Ethnic-Identity Intensity As a Moderator of the Technology Acceptance Model and Its Antecedents Marcus Anthony Alexander Iowa State University

Iowa State University Capstones, Theses and Retrospective Theses and Dissertations Dissertations

2008 Ethnic-identity intensity as a moderator of the Technology Acceptance Model and its antecedents Marcus Anthony Alexander Iowa State University

Follow this and additional works at: https://lib.dr.iastate.edu/rtd Part of the Computer Sciences Commons, Industrial Engineering Commons, Race and Ethnicity Commons, and the Race, Ethnicity and Post-Colonial Studies Commons

Recommended Citation Alexander, Marcus Anthony, "Ethnic-identity intensity as a moderator of the Technology Acceptance Model and its antecedents" (2008). Retrospective Theses and Dissertations. 15676. https://lib.dr.iastate.edu/rtd/15676

This Dissertation is brought to you for free and open access by the Iowa State University Capstones, Theses and Dissertations at Iowa State University Digital Repository. It has been accepted for inclusion in Retrospective Theses and Dissertations by an authorized administrator of Iowa State University Digital Repository. For more information, please contact [email protected]. Ethnic-identity intensity as a moderator of the Technology Acceptance Model and its antecedents

Marcus Anthony Alexander

A dissertation submitted to the graduate faculty

in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

Major: Human-Computer Interaction

Program of Study Committee: Anthony M. Townsend, Major Professor George A. Jackson Russell N. Laczniak Patricia Leigh Brian Mennecke Kevin P. Scheibe

Iowa State University

Ames, Iowa

2008

3307103 2008 ii

TABLE OF CONTENTS

LIST OF TABLES...... vii

LIST OF FIGURES...... xv

ABSTRACT...... xx

CHAPTER 1: INTRODUCTION...... 1

OVERVIEW ...... 1

PURPOSE ...... 1

THE ORIGIN OF THE CONCEPT OF DIGITAL DIVIDE ...... 2

Digital Exclusion...... 3

The Digital Divide as an Issue of Socio-Economic Division...... 12

The Digital Divide is Not an Issue of Access, but Use...... 13

The Digital Divide Begins in the home Environment, Highly Influenced by Parents ...... 15

WHY STUDY THE DIGITAL DIVIDE...... 17

The Digital Divide is Growing Larger...... 19

There is a Need for a Different Research Approach to Studying the Digital Divide ...... 20

Must Include Ethnicity in Digital Divide Research ...... 22

Need for Digital Diversification...... 24

Must Eradicate the Digital Divide...... 25

Bridging the Digital Divide Requires More...... 26

Digital Divide Deniers...... 27

SUMMARY TABLE...... 28

SUMMARY...... 30

CHAPTER 2: LITERATURE REVIEW...... 32

OVERVIEW ...... 32

iii

RACE AND TECHNOLOGY PARTICIPATION IN THE DIGITAL DIVISION - GENERAL MOTIVATIONAL RACE

DIFFERENCES ...... 32

RACE VS. ETHNICITY / RACIAL-IDENTITY VERSUS ETHNIC-IDENTITY...... 39

ETHNIC-IDENTITY AS A FUNCTION OF SOCIAL IDENTITY AND ITS CORRELATION TO SELF- EFFICACY ...... 42

ETHNIC-IDENTITY AS A FUNCTION OF SOCIAL IDENTITY AND ITS CORRELATION TO TECHNOLOGY ...... 45

TECHNOLOGY ACCEPTANCE MODEL ...... 55

TAM, Antecedents and Extensions...... 57

EXTENDED TAM MODEL ...... 64

RESEARCH QUESTIONS AND HYPOTHESES ...... 65

Classic TAM...... 67

Symbolic and Functional Utilities ...... 67

Self-Efficacy...... 70

Ethnic-Identity...... 72

SUMMARY...... 74

CHAPTER 3: METHODOLOGY...... 75

OVERVIEW ...... 75

MEASURES ...... 75

Ethnic-Identity...... 75

Measures Considered for Ethnic Identity...... 75

Multigroup Ethnic-Identity Measure ...... 77

Ease of Use Construct ...... 78

Technical Self-Efficacy Measures ...... 79

Trait Efficacy ...... 79

State Efficacy ...... 79

Usefulness Construct...... 80

Symbolic Utility...... 81

Functional Utility...... 82

SAMPLE SIZE AND CHARACTERISTICS OF RESEARCH ...... 83

PROCESS ...... 85

ANALYSIS ...... 86

PROBLEMS AND LIMITATIONS ...... 87

PROPOSED SURVEY INSTRUMENT ...... 88

SUMMARY...... 95

CHAPTER 4: ANALYSIS OF DATA...... 96

INTRODUCTION...... 96

ORGANIZATION OF DATA ANALYSIS...... 96

PRESENTATION OF DESCRIPTIVE CHARACTERISTICS OF RESPONDENTS...... 98

RESEARCH QUESTIONS AND ASSOCIATED HYPOTHESES ...... 101

Part 1: Single-Group Analysis ...... 101

Part 2: Multi-group Analysis ...... 103

ANALYSIS OF DATA ...... 104

Overview...... 104

Construct Reliability...... 106

Part 1: Single Group Analysis ...... 107

Measurement Model Fit...... 112

Causal / Structural Model ...... 113

Hypotheses 1 Results...... 114

Hypotheses 2 Results...... 115

Model Fit ...... 118

Post Hoc Analysis ...... 118

Model Fit Summary...... 121

Part 2: Multi-group Analysis ...... 122

Multi-group Measurement Models ...... 123

Measurement Model Fit – Lower-Identifiers...... 129

Measurement Model Fit – Higher-Identifiers...... 134

Causal Models...... 135

Unconstrained...... 135

Constrained...... 141

Model Fit and Comparison...... 147

Hypothesis 3 Results ...... 149

Post Hoc Analysis 1: Unique Identifier Models ...... 149

Lower-Identifiers...... 149

Higher-Identifiers...... 150

Model Fit and Comparison...... 159

Post Hoc Analysis 2: Parsimonious Comparison...... 161

Unconstrained...... 161

Constrained...... 165

Model Fit and Comparison...... 168

SUMMARY...... 170

CHAPTER 5: DISCUSSION AND CONCLUSIONS...... 171

INTRODUCTION...... 171

SUMMARY OF STUDY ...... 171

FINDINGS ...... 173

Part 1: Single Group Analysis ...... 173

Hypotheses 1...... 173

Hypotheses 2...... 175

Post Hoc Analysis ...... 175

Part 2: Multi-group Analysis ...... 176

Hypothesis 3 ...... 176

Post Hoc Analyses...... 177

CONCLUSIONS ...... 178

Part 1: Single Group Analysis ...... 178

Hypothesis 1 ...... 178

Hypothesis 2 ...... 182

Post Hoc Analysis ...... 182

Part 2: Multi-group Analysis ...... 183

Hypothesis 3 ...... 183

Post Hoc Analyses...... 184

IMPLICATIONS ...... 186

LIMITATIONS ...... 188

FUTURE RESEARCH...... 190

SUMMARY...... 192

REFERENCES...... 194

APPENDIX A: CHAPTER 1 DIGITAL DIVIDE GRAPHS...... 236

APPENDIX B: ADDITIONAL CHAPTER 4 ANALYSIS OF DATA RESULTS...... 247

vii

LIST OF TABLES

TABLE 1: ACCESS STATISTICS FROM PARTICIPANTS OF BROOKLYN COMMUNITY TECHNOLOGY CENTER ...... 13

TABLE 2: DETAILED SUMMARY OF CHAPTER 1 ...... 28

TABLE 3: AGE FREQUENCIES OF RESEARCH PARTICIPANTS ...... 98

TABLE 4: GENDER FREQUENCIES OF RESEARCH PARTICIPANTS ...... 99

TABLE 5: PARENTAL EDUCATION FREQUENCIES OF RESEARCH PARTICIPANTS...... 99

TABLE 6: PARENTAL EDUCATION FREQUENCIES OF RESEARCH PARTICIPANTS...... 100

TABLE 7: REGION OF ORIGIN FREQUENCIES OF RESEARCH PARTICIPANTS ...... 100

TABLE 8: COLLEGE ATTENDING FREQUENCIES OF RESEARCH PARTICIPANTS...... 101

TABLE 9: CONSTRUCT RELIABILITY CALCULATIONS...... 106

TABLE 10: UNSTANDARDIZED REGRESSION WEIGHTS (FACTOR LOADINGS) OF INDICATORS ON THEIR

RESPECTIVE LATENT CONSTRUCT OF MEASUREMENT MODEL – ENTIRE SAMPLE ...... 111

TABLE 11: STANDARDIZED REGRESSION WEIGHTS (FACTOR LOADINGS) OF INDICATORS ON THEIR

RESPECTIVE LATENT CONSTRUCT OF MEASUREMENT MODEL – ENTIRE SAMPLE ...... 111

TABLE 12: CHI-SQUARE SIGNIFICANCE TEST OF MEASUREMENT MODEL - ENTIRE SAMPLE ...... 113

TABLE 13: COMPARATIVE FIT INDEX GOODNESS-OF-FIT COMPARISON OF MEASUREMENT MODEL –

ENTIRE SAMPLE ...... 113

TABLE 14: ROOT MEAN SQUARE ERROR OF APPROXIMATION GOODNESS-OF-FIT COMPARISON OF

MEASUREMENT MODEL– ENTIRE SAMPLE ...... 113

TABLE 15: UNSTANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL – ENTIRE SAMPLE ...... 117

TABLE 16: STANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL– ENTIRE SAMPLE...... 117

TABLE 17: CHI-SQUARE SIGNIFICANCE TEST OF CAUSAL MODEL – ENTIRE SAMPLE ...... 118

TABLE 18: COMPARATIVE FIT INDEX GOODNESS-OF-FIT COMPARISON OF CAUSAL MODEL –

ENTIRE SAMPLE ...... 118

TABLE 19: ROOT MEAN SQUARE ERROR OF APPROXIMATION GOODNESS-OF-FIT COMPARISON OF CAUSAL

MODEL – ENTIRE SAMPLE ...... 118 viii

TABLE 20: POST HOC UNSTANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL – ENTIRE SAMPLE ...... 121

TABLE 21: POST HOC STANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL– ENTIRE SAMPLE...... 121

TABLE 22: POST HOC CHI-SQUARE SIGNIFICANCE TEST OF CAUSAL MODEL – ENTIRE SAMPLE...... 121

TABLE 23: POST HOC COMPARATIVE FIT INDEX GOODNESS-OF-FIT COMPARISON OF CAUSAL MODEL –

ENTIRE SAMPLE ...... 122

TABLE 24: POST HOC ROOT MEAN SQUARE ERROR OF APPROXIMATION GOODNESS-OF-FIT COMPARISON

OF CAUSAL MODEL – ENTIRE SAMPLE ...... 122

TABLE 25: UNSTANDARDIZED REGRESSION WEIGHTS (FACTOR LOADINGS) OF INDICATORS ON THEIR

RESPECTIVE LATENT CONSTRUCT OF MEASUREMENT MODEL – LOWER-IDENTIFIERS ...... 128

TABLE 26: STANDARDIZED REGRESSION WEIGHTS (FACTOR LOADINGS) OF INDICATORS ON THEIR

RESPECTIVE LATENT CONSTRUCT OF MEASUREMENT MODEL – LOWER-IDENTIFIERS ...... 128

TABLE 27: CHI-SQUARE SIGNIFICANCE TEST OF MEASUREMENT MODEL – LOWER-IDENTIFIERS ...... 129

TABLE 28: COMPARATIVE FIT INDEX GOODNESS-OF-FIT COMPARISON OF MEASUREMENT MODEL –

LOWER-IDENTIFIERS ...... 129

TABLE 29: ROOT MEAN SQUARE ERROR OF APPROXIMATION GOODNESS-OF-FIT COMPARISON OF

MEASUREMENT MODEL – LOWER-IDENTIFIERS ...... 129

TABLE 30: UNSTANDARDIZED REGRESSION WEIGHTS (FACTOR LOADINGS) OF INDICATORS ON THEIR

RESPECTIVE LATENT CONSTRUCT OF MEASUREMENT MODEL – HIGHER-IDENTIFIERS ...... 133

TABLE 31: UNSTANDARDIZED REGRESSION WEIGHTS (FACTOR LOADINGS) OF INDICATORS ON THEIR

RESPECTIVE LATENT CONSTRUCT OF MEASUREMENT MODEL – HIGHER-IDENTIFIERS ...... 133

TABLE 32: CHI-SQUARE SIGNIFICANCE TEST OF MULTI-GROUP MEASUREMENT MODELS ...... 134

TABLE 33: COMPARATIVE FIT INDEX GOODNESS-OF-FIT COMPARISON OF MULTI-GROUP MEASUREMENT

MODELS ...... 134

TABLE 34: ROOT MEAN SQUARE ERROR OF APPROXIMATION GOODNESS-OF-FIT COMPARISON OF

MULTI-GROUP MEASUREMENT MODELS ...... 134

TABLE 35: UNCONSTRAINED UNSTANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL – LOWER-

IDENTIFIERS...... 137

TABLE 36: UNCONSTRAINED STANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

LOWER-IDENTIFIERS ...... 138

TABLE 37: UNCONSTRAINED UNSTANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL – HIGHER-

IDENTIFIERS...... 141

TABLE 38: UNCONSTRAINED STANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

HIGHER-IDENTIFIERS...... 141

TABLE 39: CONSTRAINED UNSTANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

LOWER-IDENTIFIERS ...... 143

TABLE 40: CONSTRAINED STANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

LOWER-IDENTIFIERS ...... 144

TABLE 41: CONSTRAINED UNSTANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

HIGHER-IDENTIFIERS...... 146

TABLE 42: CONSTRAINED STANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

HIGHER-IDENTIFIERS...... 147

TABLE 43: CHI-SQUARE SIGNIFICANCE TEST OF MULTI-GROUP CAUSAL MODELS...... 148

TABLE 44: COMPARATIVE FIT INDEX GOODNESS-OF-FIT COMPARISON OF MULTI-GROUP CAUSAL MODELS .... 148

TABLE 45: ROOT MEAN SQUARE ERROR OF APPROXIMATION GOODNESS-OF-FIT COMPARISON OF

MULTI-GROUP CAUSAL MODELS – ENTIRE SAMPLE...... 148

TABLE 46: DIFFERENCE: MODEL COMPARISON UNCONSTRAINED VERSUS CONSTRAINED ...... 148

TABLE 47: POST HOC UNCONSTRAINED UNSTANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

LOWER-IDENTIFIERS ...... 153

TABLE 48: POST HOC UNCONSTRAINED STANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

LOWER-IDENTIFIERS ...... 153

TABLE 49: POST HOC UNCONSTRAINED UNSTANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

HIGHER-IDENTIFIERS...... 155

TABLE 50: POST HOC UNCONSTRAINED STANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

HIGHER-IDENTIFIERS...... 155

TABLE 51: POST HOC CONSTRAINED UNSTANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

LOWER-IDENTIFIERS ...... 157

TABLE 52: POST HOC CONSTRAINED STANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

LOWER-IDENTIFIERS ...... 157

TABLE 53: POST HOC CONSTRAINED UNSTANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

HIGHER-IDENTIFIERS...... 159

TABLE 54: POST HOC CONSTRAINED STANDARDIZED REGRESSION WEIGHTS OF CAUSAL MODEL –

HIGHER-IDENTIFIERS...... 159

TABLE 55: CHI-SQUARE SIGNIFICANCE TEST OF MULTI-GROUP CAUSAL MODELS...... 160

TABLE 56: COMPARATIVE FIT INDEX GOODNESS-OF-FIT COMPARISON OF MULTI-GROUP CAUSAL MODELS .... 160

TABLE 57: ROOT MEAN SQUARE ERROR OF APPROXIMATION GOODNESS-OF-FIT COMPARISON OF

MULTI-GROUP CAUSAL MODELS – ENTIRE SAMPLE...... 160

TABLE 58: DIFFERENCE: MODEL COMPARISON UNCONSTRAINED VERSUS CONSTRAINED ...... 160

TABLE 59: PARSIMONIOUS POST HOC UNSTANDARDIZED UNCONSTRAINED RESULTS –

LOWER-IDENTIFIERS ...... 164

TABLE 60: PARSIMONIOUS POST HOC STANDARDIZED UNCONSTRAINED RESULTS – LOWER-IDENTIFIERS ...... 164

TABLE 61: PARSIMONIOUS POST HOC UNSTANDARDIZED UNCONSTRAINED RESULTS –

HIGHER-IDENTIFIERS...... 164

TABLE 62: PARSIMONIOUS POST HOC STANDARDIZED UNCONSTRAINED RESULTS – HIGHER-IDENTIFIERS ...... 164

TABLE 63: PARSIMONIOUS POST HOC UNSTANDARDIZED CONSTRAINED RESULTS – LOWER-IDENTIFIERS...... 167

TABLE 64: PARSIMONIOUS POST HOC STANDARDIZED CONSTRAINED RESULTS – LOWER-IDENTIFIERS ...... 167

TABLE 65: PARSIMONIOUS POST HOC UNSTANDARDIZED CONSTRAINED RESULTS – HIGHER-IDENTIFIERS ...... 168

TABLE 66: PARSIMONIOUS POST HOC STANDARDIZED CONSTRAINED RESULTS – HIGHER-IDENTIFIERS...... 168

TABLE 67: CHI-SQUARE SIGNIFICANCE TEST OF PARSIMONIOUS CAUSAL MODELS ...... 169

TABLE 68: COMPARATIVE FIT INDEX GOODNESS-OF-FIT COMPARISON OF PARSIMONIOUS CAUSAL MODELS ... 169

TABLE 69: ROOT MEAN SQUARE ERROR OF APPROXIMATION GOODNESS-OF-FIT COMPARISON OF

PARSIMONIOUS CAUSAL MODELS – ENTIRE SAMPLE ...... 169

TABLE 70: DIFFERENCE: MODEL COMPARISON UNCONSTRAINED VERSUS CONSTRAINED ...... 170

TABLE 71: DESCRIPTIVE STATISTICS FOR EASE OF USE QUESTIONS...... 179

TABLE 72: FREQUENCIES OF RESPONSES FOR EASE OF USE QUESTION 8...... 179

TABLE 73: FREQUENCIES OF RESPONSES FOR EASE OF USE QUESTION 9...... 180

TABLE 74: FREQUENCIES OF RESPONSES FOR EASE OF USE QUESTION 8...... 180

TABLE 75: COVARIANCES FOR ORIGINAL MEASUREMENT MODEL...... 249

TABLE 76: CORRELATIONS FOR ORIGINAL MEASUREMENT MODEL ...... 250

TABLE 77: VARIANCES FOR ORIGINAL MEASUREMENT MODEL...... 250

TABLE 78: SQUARED MULTIPLE CORRELATIONS FOR ORIGINAL MEASUREMENT MODEL ...... 251

TABLE 79: REGRESSION WEIGHTS FOR SINGLE-GROUP CAUSAL MODEL ...... 254

TABLE 80: STANDARDIZED REGRESSION WEIGHTS FOR SINGLE-GROUP CAUSAL MODEL...... 255

TABLE 81:COVARIANCES FOR SINGLE-GROUP CAUSAL MODEL ...... 255

TABLE 82:CORRELATIONS FOR SINGLE-GROUP CAUSAL MODEL...... 256

TABLE 83: VARIANCES FOR SINGLE-GROUP CAUSAL MODEL...... 256

TABLE 84: SQUARED MULTIPLE CORRELATIONS FOR SINGLE-GROUP CAUSAL MODEL ...... 256

TABLE 85:CMIN FOR SINGLE-GROUP CAUSAL MODEL ...... 257

TABLE 86: RMR, GFI FOR SINGLE-GROUP CAUSAL MODEL ...... 257

TABLE 87: BASELINE COMPARISONS FOR SINGLE-GROUP CAUSAL MODEL...... 257

TABLE 88: PARSIMONY-ADJUSTED MEASURES FOR SINGLE-GROUP CAUSAL MODEL...... 257

TABLE 89: NCP FOR SINGLE-GROUP CAUSAL MODEL ...... 258

TABLE 90: FMIN FOR SINGLE-GROUP CAUSAL MODEL ...... 258

TABLE 91: RMSEA FOR SINGLE-GROUP CAUSAL MODEL...... 258

TABLE 92: AIC FOR SINGLE-GROUP CAUSAL MODEL ...... 258

TABLE 93: ECVI FOR SINGLE-GROUP CAUSAL MODEL...... 258

TABLE 94: HOELTER FOR SINGLE-GROUP CAUSAL MODEL ...... 258

TABLE 95: REGRESSION WEIGHTS FOR POST HOC SINGLE-GROUP CAUSAL MODEL...... 261

TABLE 96: STANDARDIZED REGRESSION WEIGHTS FOR POST HOC SINGLE-GROUP CAUSAL MODEL ...... 262

xii

TABLE 97: COVARIANCES FOR POST HOC SINGLE-GROUP CAUSAL MODEL...... 262

TABLE 98: CORRELATIONS FOR POST HOC SINGLE-GROUP CAUSAL MODEL ...... 263

TABLE 99: VARIANCES FOR POST HOC SINGLE-GROUP CAUSAL MODEL ...... 263

TABLE 100: SQUARED MULTIPLE CORRELATIONS FOR POST HOC SINGLE-GROUP CAUSAL MODEL...... 263

TABLE 102: TOTAL EFFECTS FOR POST HOC SINGLE-GROUP CAUSAL MODEL...... 264

TABLE 103: STANDARDIZED TOTAL EFFECTS FOR POST HOC SINGLE-GROUP CAUSAL MODEL ...... 265

TABLE 104: DIRECT EFFECTS FOR POST HOC SINGLE-GROUP CAUSAL MODEL...... 265

TABLE 105: STANDARDIZED DIRECT EFFECTS FOR POST HOC SINGLE-GROUP CAUSAL MODEL ...... 266

TABLE 106: INDIRECT EFFECTS FOR POST HOC SINGLE-GROUP CAUSAL MODEL...... 266

TABLE 107: STANDARDIZED INDIRECT EFFECTS FOR POST HOC SINGLE-GROUP CAUSAL MODEL ...... 267

TABLE 108: REGRESSION WEIGHTS FOR MEASUREMENT MODEL OF LOWER-IDENTIFIERS...... 267

TABLE 109: STANDARDIZED REGRESSION WEIGHTS FOR MEASUREMENT MODEL OF LOWER-IDENTIFIERS ...... 268

TABLE 110: COVARIANCES FOR MEASUREMENT MODEL OF LOWER-IDENTIFIERS ...... 268

TABLE 111: CORRELATIONS FOR MEASUREMENT MODEL OF LOWER-IDENTIFIERS...... 269

TABLE 112: VARIANCES FOR MEASUREMENT MODEL OF LOWER-IDENTIFIERS ...... 269

TABLE 113: SQUARED MULTIPLE CORRELATIONS FOR MEASUREMENT MODEL OF LOWER-IDENTIFIERS...... 270

TABLE 115: REGRESSION WEIGHTS FOR MEASUREMENT MODEL OF HIGHER-IDENTIFIERS ...... 270

TABLE 116: STANDARDIZED REGRESSION WEIGHTS FOR MEASUREMENT MODEL OF HIGHER-IDENTIFIERS...... 271

TABLE 117: COVARIANCES FOR MEASUREMENT MODEL OF HIGHER-IDENTIFIERS ...... 271

TABLE 118: CORRELATIONS FOR MEASUREMENT MODEL OF HIGHER-IDENTIFIERS ...... 272

TABLE 119: VARIANCES FOR MEASUREMENT MODEL OF HIGHER-IDENTIFIERS...... 272

TABLE 120: SQUARED MULTIPLE CORRELATIONS FOR MEASUREMENT MODEL OF HIGHER-IDENTIFIERS ...... 273

TABLE 122: CMIN FOR MULTI-GROUP MEASUREMENT MODEL ...... 273

TABLE 123: RMR, GFI FOR MULTI-GROUP MEASUREMENT MODEL...... 273

TABLE 124: BASELINE COMPARISONS FOR MULTI-GROUP MEASUREMENT MODEL ...... 273

TABLE 125: PARSIMONY-ADJUSTED MEASURES FOR MULTI-GROUP MEASUREMENT MODEL ...... 274

TABLE 126: NCP FOR MULTI-GROUP MEASUREMENT MODEL...... 274

xiii

TABLE 127: FMIN FOR MULTI-GROUP MEASUREMENT MODEL...... 274

TABLE 128: RMSEA FOR MULTI-GROUP MEASUREMENT MODEL ...... 274

TABLE 129: AIC FOR MULTI-GROUP MEASUREMENT MODEL...... 274

TABLE 130: ECVI FOR MULTI-GROUP MEASUREMENT MODEL ...... 274

TABLE 131: HOELTER FOR MULTI-GROUP MEASUREMENT MODEL...... 274

TABLE 132: UNSTANDARDIZED REGRESSION WEIGHTS FOR UNCONSTRAINED MULTI-GROUP CAUSAL

MODEL – LOWER-IDENTIFIERS ...... 279

TABLE 133: POST HOC UNSTANDARDIZED REGRESSION WEIGHTS FOR UNCONSTRAINED -

LOWER-IDENTIFIERS ...... 287

TABLE 134: POST HOC STANDARDIZED REGRESSION WEIGHTS FOR UNCONSTRAINED -

LOWER-IDENTIFIERS ...... 288

TABLE 135: POST HOC COVARIANCES FOR UNCONSTRAINED LOWER-IDENTIFIERS ...... 288

TABLE 136: POST HOC CORRELATIONS FOR UNCONSTRAINED LOWER-IDENTIFIERS...... 288

TABLE 137: POST HOC VARIANCES FOR UNCONSTRAINED LOWER-IDENTIFIERS ...... 289

TABLE 138: POST HOC SQUARED MULTIPLE CORRELATIONS FOR UNCONSTRAINED LOWER-IDENTIFIERS...... 289

TABLE 139: POST HOC UNSTANDARDIZED REGRESSION WEIGHTS FOR UNCONSTRAINED

HIGHER-IDENTIFIERS...... 290

TABLE 140: POST HOC STANDARDIZED REGRESSION WEIGHTS FOR UNCONSTRAINED HIGHER-IDENTIFIERS.... 290

TABLE 141: POST HOC COVARIANCES FOR UNCONSTRAINED HIGHER-IDENTIFIERS ...... 291

TABLE 142: POST HOC CORRELATIONS FOR UNCONSTRAINED HIGHER-IDENTIFIERS...... 291

TABLE 143: POST HOC VARIANCES FOR UNCONSTRAINED HIGHER-IDENTIFIERS...... 291

TABLE 144: POST HOC SQUARED MULTIPLE CORRELATIONS FOR UNCONSTRAINED HIGHER-IDENTIFIERS ...... 292

TABLE 145: POST HOC REGRESSION WEIGHTS FOR CONSTRAINED LOWER-IDENTIFIERS ...... 292

TABLE 146: POST HOC STANDARDIZED REGRESSION WEIGHTS FOR CONSTRAINED LOWER-IDENTIFIERS...... 293

TABLE 147: POST HOC COVARIANCES FOR CONSTRAINED LOWER-IDENTIFIERS...... 293

TABLE 148: POST HOC CORRELATIONS FOR CONSTRAINED LOWER-IDENTIFIERS ...... 293

TABLE 149: POST HOC VARIANCES FOR CONSTRAINED LOWER-IDENTIFIERS...... 294

xiv

TABLE 150: POST HOC SQUARED MULTIPLE CORRELATIONS FOR CONSTRAINED LOWER-IDENTIFIERS ...... 294

TABLE 151: POST HOC REGRESSION WEIGHTS FOR CONSTRAINED HIGHER-IDENTIFIERS...... 295

TABLE 152: POST HOC STANDARDIZED REGRESSION WEIGHTS FOR CONSTRAINED HIGHER-IDENTIFIERS ...... 295

TABLE 153: POST HOC COVARIANCES FOR CONSTRAINED HIGHER-IDENTIFIERS...... 296

TABLE 154: POST HOC CORRELATIONS FOR CONSTRAINED HIGHER-IDENTIFIERS ...... 296

TABLE 155: POST HOC VARIANCES FOR CONSTRAINED HIGHER-IDENTIFIERS ...... 296

TABLE 156: POST HOC SQUARED MULTIPLE CORRELATIONS FOR CONSTRAINED HIGHER-IDENTIFIERS ...... 297

TABLE 157: POST HOC MULTI- GROUP CMIN RESULTS ...... 297

TABLE 158: POST HOC MULTI- GROUP RMR, GFI RESULTS...... 297

TABLE 159: POST HOC MULTI- GROUP BASELINE COMPARISONS ...... 298

TABLE 160: POST HOC MULTI- GROUP PARSIMONY-ADJUSTED MEASURES ...... 298

TABLE 161: POST HOC MULTI- GROUP NCP RESULTS...... 298

TABLE 162: POST HOC MULTI- GROUP FMIN RESULTS...... 298

TABLE 163: POST HOC MULTI- GROUP RMSEA RESULTS ...... 299

TABLE 164: POST HOC MULTI- GROUP AIC RESULTS ...... 299

TABLE 165: POST HOC MULTI- GROUP ECVI RESULTS ...... 299

TABLE 166: POST HOC MULTI- GROUP HOELTER RESULTS...... 299

TABLE 167: POST HOC MULTI- GROUP NESTED MODEL COMPARISONS ASSUMING MODEL

UNCONSTRAINED TO BE CORRECT ...... 300

TABLE 168: POST HOC MULTI- GROUP NESTED MODEL COMPARISONS ASSUMING MODEL MEASUREMENT

WEIGHTS TO BE CORRECT...... 300

TABLE 169: POST HOC MULTI- GROUP NESTED MODEL COMPARISONS ASSUMING MODEL

STRUCTURAL WEIGHTS TO BE CORRECT: ...... 300

TABLE 170: POST HOC MULTI- GROUP NESTED MODEL COMPARISONS ASSUMING MODEL STRUCTURAL

COVARIANCES TO BE CORRECT...... 300

TABLE 171: POST HOC MULTI- GROUP NESTED MODEL COMPARISONS ASSUMING MODEL STRUCTURAL

RESIDUALS TO BE CORRECT...... 300

LIST OF FIGURES

FIGURE 1: PERSONS USING THE INTERNET AT HOME (BY RACE) ...... 17

FIGURE 2: PROPOSED TAM EXTENSION ...... 65

FIGURE 3: COMPUTER SELF-EFFICACY RESEARCH MODEL BY COMPEAU AND HIGGINS (1995)...... 80

FIGURE 4: CAUSAL MODEL WITH HYPOTHESES 1 AND 2 LABELED ...... 102

FIGURE 5: CAUSAL MODEL WITH HYPOTHESIS 3 LABELED ...... 104

FIGURE 6: MEASUREMENT MODEL – ENTIRE SAMPLE ...... 108

FIGURE 7: UNSTANDARDIZED RESULTS OF MEASUREMENT MODEL – ENTIRE SAMPLE...... 109

FIGURE 8: STANDARDIZED RESULTS OF MEASUREMENT MODEL – ENTIRE SAMPLE ...... 110

FIGURE 9: CAUSAL MODEL WITH HYPOTHESES 1 AND 2 LABELED ...... 114

FIGURE 10: UNSTANDARDIZED RESULTS OF CAUSAL MODEL– ENTIRE SAMPLE...... 116

FIGURE 11: STANDARDIZED RESULTS OF CAUSAL MODEL – ENTIRE SAMPLE ...... 117

FIGURE 12: POST HOC CAUSAL MODEL – ENTIRE MODEL...... 119

FIGURE 13: POST HOC UNSTANDARDIZED RESULTS OF CAUSAL MODEL – ENTIRE MODEL...... 120

FIGURE 14: POST HOC STANDARDIZED RESULTS OF CAUSAL MODEL – ENTIRE MODEL ...... 120

FIGURE 15: CAUSAL MODEL WITH HYPOTHESIS 3 LABELED ...... 123

FIGURE 16: MULTI-GROUP MEASUREMENT MODEL WITH MEASUREMENT LOADINGS CONSTRAINED –

LOWER-IDENTIFIERS ...... 125

FIGURE 17: UNSTANDARDIZED RESULTS OF MULTI-GROUP MEASUREMENT MODEL WITH MEASUREMENT

LOADINGS CONSTRAINED – LOWER IDENTIFIERS...... 126

FIGURE 18: STANDARDIZED RESULTS OF MULTI-GROUP MEASUREMENT MODEL WITH MEASUREMENT

LOADINGS CONSTRAINED – LOWER IDENTIFIERS...... 127

FIGURE 19: MULTI-GROUP MEASUREMENT MODEL WITH MEASUREMENT LOADINGS CONSTRAINED –

HIGHER-IDENTIFIERS...... 130

FIGURE 20: UNSTANDARDIZED RESULTS OF MULTI-GROUP MEASUREMENT MODEL WITH MEASUREMENT

LOADINGS CONSTRAINED – HIGHER-IDENTIFIERS...... 131 xvi

FIGURE 21: STANDARDIZED RESULTS OF MULTI-GROUP MEASUREMENT MODEL WITH MEASUREMENT

LOADINGS CONSTRAINED – HIGHER-IDENTIFIERS...... 132

FIGURE 22: UNCONSTRAINED UNSTANDARDIZED RESULTS OF CAUSAL MODEL – LOWER-IDENTIFIERS ...... 136

FIGURE 23: UNCONSTRAINED STANDARDIZED RESULTS OF CAUSAL MODEL – LOWER-IDENTIFIERS...... 136

FIGURE 24: UNCONSTRAINED UNSTANDARDIZED RESULTS OF CAUSAL MODEL – HIGHER-IDENTIFIERS ...... 139

FIGURE 25: UNCONSTRAINED STANDARDIZED RESULTS OF CAUSAL MODEL – HIGHER-IDENTIFIERS ...... 140

FIGURE 26: CONSTRAINED UNSTANDARDIZED RESULTS OF CAUSAL MODEL – LOWER-IDENTIFIERS...... 142

FIGURE 27: CONSTRAINED STANDARDIZED RESULTS OF CAUSAL MODEL – LOWER-IDENTIFIERS ...... 143

FIGURE 28: CONSTRAINED UNSTANDARDIZED RESULTS OF CAUSAL MODEL – HIGHER-IDENTIFIERS ...... 145

FIGURE 29: CONSTRAINED STANDARDIZED RESULTS OF CAUSAL MODEL – LOWER-IDENTIFIERS ...... 145

FIGURE 30: POST HOC LOWER-IDENTIFIER CAUSAL MODEL ...... 150

FIGURE 31: POST HOC HIGHER-IDENTIFIER CAUSAL MODEL...... 151

FIGURE 32: POST HOC UNCONSTRAINED UNSTANDARDIZED RESULTS OF CAUSAL MODEL –

LOWER-IDENTIFIERS ...... 152

FIGURE 33: POST HOC UNCONSTRAINED STANDARDIZED RESULTS OF CAUSAL MODEL –

LOWER-IDENTIFIERS ...... 152

FIGURE 34: POST HOC UNCONSTRAINED UNSTANDARDIZED RESULTS OF CAUSAL MODEL –

HIGHER-IDENTIFIERS...... 154

FIGURE 35: POST HOC UNCONSTRAINED STANDARDIZED RESULTS OF CAUSAL MODEL –

HIGHER-IDENTIFIERS...... 154

FIGURE 36: POST HOC CONSTRAINED UNSTANDARDIZED RESULTS OF CAUSAL MODEL –

LOWER-IDENTIFIERS ...... 156

FIGURE 37: POST HOC CONSTRAINED STANDARDIZED RESULTS OF CAUSAL MODEL – LOWER-IDENTIFIERS.... 156

FIGURE 38: POST HOC CONSTRAINED UNSTANDARDIZED RESULTS OF CAUSAL MODEL –

HIGHER-IDENTIFIERS...... 158

FIGURE 39: POST HOC CONSTRAINED STANDARDIZED RESULTS OF CAUSAL MODEL – HIGHER-IDENTIFIERS ... 158

xvii

FIGURE 40: PARSIMONIOUS POST HOC UNSTANDARDIZED UNCONSTRAINED RESULTS –

LOWER-IDENTIFIERS ...... 162

FIGURE 41: PARSIMONIOUS POST HOC STANDARDIZED UNCONSTRAINED RESULTS – LOWER-IDENTIFIERS...... 162

FIGURE 42: PARSIMONIOUS POST HOC UNSTANDARDIZED UNCONSTRAINED RESULTS –

HIGHER-IDENTIFIERS...... 163

FIGURE 43: PARSIMONIOUS POST HOC STANDARDIZED UNCONSTRAINED RESULTS – HIGHER-IDENTIFIERS ..... 163

FIGURE 44: PARSIMONIOUS POST HOC UNSTANDARDIZED CONSTRAINED RESULTS – LOWER-IDENTIFIERS...... 165

FIGURE 45: PARSIMONIOUS POST HOC STANDARDIZED CONSTRAINED RESULTS – LOWER-IDENTIFIERS ...... 166

FIGURE 46: PARSIMONIOUS POST HOC UNSTANDARDIZED CONSTRAINED RESULTS – HIGHER-IDENTIFIERS ..... 166

FIGURE 47: PARSIMONIOUS POST HOC STANDARDIZED CONSTRAINED RESULTS – HIGHER-IDENTIFIERS ...... 167

FIGURE 48: HOUSEHOLDS WITH INTERNET ACCESS BY RACE/HISPANIC ORIGIN ...... 236

FIGURE 49: TARE OF GROWTH OF INTERNET PENETRATION BY RACE/HISPANIC ORIGIN...... 236

FIGURE 50: INCOME AND EDUCATION DIFFERENCES ACCOUNT FOR HALF OF THE GAP BETWEEN

BLACKS AND HISPANICS AND NATIONAL AVERAGE...... 237

FIGURE 51: HOUSEHOLDS WITH A COMPUTER BY RACE/HISPANIC ORIGIN ...... 237

FIGURE 52: INTERNET USE BY RACE/ETHNICITY...... 238

FIGURE 53: HOUSEHOLD ACCESS RATES BY RACE/ETHNICITY DO NOT CLOSELY TRACK INTERNET

USE BY PERSONS...... 238

FIGURE 54: INTERNET USE BY LOCATION AND RACE/ETHNICITY...... 239

FIGURE 55: PERCENT OF US HOUSEHOLDS WITH A COMPUTER BY RACE/HISPANIC ORIGIN ...... 239

FIGURE 56: HOUSEHOLDS WITH A COMPUTER BY INCOME BY RACE/HISPANIC ORIGIN ...... 240

FIGURE 57: HOUSEHOLDS WITH INTERNET ACCESS BY RACE/HISPANIC ORIGIN ...... 240

FIGURE 58: HOUSEHOLDS WITH INTERNET ACCESS...... 241

FIGURE 59: PERSONS USING THE INTERNET BY RACE/HISPANIC ORIGIN ...... 241

FIGURE 60: PERSONS USING THE INTERNET AT HOME BY RACE/HISPANIC ORIGIN ...... 242

FIGURE 61: PERSONS USING THE INTERNET OUTSIDE THE HOME BY RACE/HISPANIC ORIGIN ...... 242

FIGURE 62: PERSONS USING THE INTERNET OUTSIDE THE HOME BY RACE/HISPANIC ORIGIN ...... 243

xviii

FIGURE 63: PERSONS USING THE INTERNET OUTSIDE THE HOME AT SCHOOLS ...... 243

FIGURE 64:PERSONS USING THE INTERNET OUTSIDE THE HOME AT WORK ...... 244

FIGURE 65: REASONS FOR HOUSEHOLD WITH A COMPUTER/WEBTV NOT USING THE INTERNET AT

HOME BY RACE/HISPANIC ORIGIN ...... 244

FIGURE 66: HOUSEHOLDS WITH HOME INTERNET ACCESS BY RACE/HISPANIC ORIGIN...... 245

FIGURE 67: RATE OF GROWTH OF INTERNET PENETRATION BY RACE/HISPANIC ORIGIN...... 245

FIGURE 68: INTERNET ACCESS AT HOME BY RACE/ETHNICITY AND DISABILITY STATUS...... 246

FIGURE 69: REGULARLY USES A PC BY RACE/ETHNICITY AND DISABILITY STATUS...... 246

FIGURE 70: SINGLE-GROUP MEASUREMENT MODEL...... 247

FIGURE 71: UNSTANDARDIZED RESULTS OF ORIGINAL MEASUREMENT MODEL SHOWING NEGATIVE

ERROR VARIANCE...... 248

FIGURE 72: UNSTANDARDIZED RESULTS OF ORIGINAL MEASUREMENT MODEL SHOWING FU.I IS

TOO-STRONGLY CORRELATED WITH THE OTHER INDICATORS...... 249

FIGURE 73: SINGLE-GROUP CAUSAL MODEL ...... 252

FIGURE 74: SINGLE-GROUP UNSTANDARDIZED RESULTS ...... 253

FIGURE 75: SINGLE-GROUP STANDARDIZED RESULTS ...... 254

FIGURE 76: POST HOC SINGLE-GROUP CAUSAL MODEL ...... 259

FIGURE 77: UNSTANDARDIZED RESULTS FOR POST HOC SINGLE-GROUP CAUSAL MODEL ...... 260

FIGURE 78: STANDARDIZED RESULTS FOR POST HOC SINGLE-GROUP CAUSAL MODEL...... 261

FIGURE 79: MULTI-GROUP CAUSAL MODEL UNCONSTRAINED ...... 275

FIGURE 80: UNSTANDARDIZED RESULTS FOR MULTI-GROUP CAUSAL MODEL UNCONSTRAINED ...... 276

FIGURE 81: STANDARDIZED RESULTS FOR MULTI-GROUP CAUSAL MODEL UNCONSTRAINED...... 277

FIGURE 82: UNSTANDARDIZED RESULTS FOR MULTI-GROUP CAUSAL MODEL CONSTRAINED...... 278

FIGURE 83: STANDARDIZED RESULTS FOR MULTI-GROUP CAUSAL MODEL CONSTRAINED ...... 279

FIGURE 84: POST HOC CAUSAL MODEL – LOWER-IDENTIFIERS ...... 280

FIGURE 85: POST HOC STANDARDIZED RESULTS – LOWER-IDENTIFIERS...... 281

xix

FIGURE 86: UNSTANDARDIZED RESULTS MULTI-GROUP CAUSAL MODEL UNCONSTRAINED –

HIGHER IDENTIFIERS ...... 282

FIGURE 87: STANDARDIZED RESULTS MULTI-GROUP CAUSAL MODEL UNCONSTRAINED –

HIGHER IDENTIFIERS ...... 283

FIGURE 88: UNSTANDARDIZED RESULTS MULTI-GROUP CAUSAL MODEL CONSTRAINED –

LOWER-IDENTIFIERS ...... 284

FIGURE 89: STANDARDIZED RESULTS MULTI-GROUP CAUSAL MODEL UNCONSTRAINED –

LOWER-IDENTIFIERS ...... 285

FIGURE 90: UNSTANDARDIZED RESULTS MULTI-GROUP CAUSAL MODEL CONSTRAINED –

HIGHER-IDENTIFIERS...... 286

FIGURE 91: STANDARDIZED RESULTS MULTI-GROUP CAUSAL MODEL CONSTRAINED –

HIGHER-IDENTIFIERS...... 287

ABSTRACT

The goal of this dissertation is oriented to study the motivation of minorities to adopt and learn new, innovative technologies. The primary research question: is there some reason the

African-American experience is driving certain sub-groups of the population to the wrong side of the digital divide? To examine this, I introduce ethnic identity as a moderating variable to a leftward extended technology acceptance model (TAM). Additionally, both components of the classic TAM (ease of use and usefulness) are divided into two antecedents: 1) ease of use- a) trait efficacy and b) state efficacy and 2) usefulness

(completely replaced by) - a) symbolic utility and b) functional utility. State efficacy has a significant relationship to both ease of use as well as intent to use. Functional utility has a significant relationship to intent to use. Ethnic-Identity and its intensity does have a moderating effect to a more parsimonious model (were significant paths are compared for invariance.

CHAPTER 1: INTRODUCTION

Overview The goal of this dissertation is oriented to study the motivation of minorities to adopt and learn new, innovative technologies. Towards this end, I will extensively review the relevant descriptive literature that explains the digital divide, with specific attention to how it is characterized. I will demonstrate that while the digital divide is characterized by numerous separators (e.g. income, geography), race (and as I will show later, more importantly ethnicity) is still a proxy of the majority of those negatively affected by this division. As this introduction is extremely detailed and thorough, a detailed summary table of this entire first chapter is included at the end of this chapter.

In chapter 2, as many indicators are proxies of race, it is useful to review what is known concerning racial differences with regards to technology and information technology.

I will explore the theoretical bases of differences between races transitioning into the difference between race and ethnicity (and thusly, racial identity and ethnic identity). I will then describe a new proposed model of ethnic identity and technology adoption incorporating technical self efficacy as well as hedonic and utilitarian characteristics.

Purpose The goal of this dissertation is oriented to study the motivation of minorities to adopt and learn new, innovative technologies. The primary research question: is there some reason the African-American experience is driving certain sub-groups of the population to the wrong side of the digital divide? Other questions as part of this research include:

1. Are there differences in technical consumption that differs by ethnicity and the

intensity of ethnic affiliation?

2. Does ethnic identity moderate the traditional (and the proposed extended) technology

acceptance model (TAM).

The motivation for this study is rooted in the issue of the digital divide. As a result, extensive detailed digital divide and technology adoption follows.

The Origin of the Concept of Digital Divide

The digital divide is a classification of those who have access to innovation technology (haves) and those who do not (have-nots). This also encompasses physical access to technology hardware as well as the skills and resources needed for effective and efficient use. As alluded to earlier, the context of a digital divide is often segmented by socioeconomic status (rich/poor), racial (white/minority), or geographical (urban/rural), just to name a few.

According to authors Tichenor, Olien, and Donohue, the concept of information

“have-nots” first arose in the context of imbalances in newspaper readership which reportedly favored the well-off middle class. As a result, they proposed the theory of the

"knowledge gap" where they argued that the informational deprivation of lower classes directly contributes to their alienation from social (and political) life, reinforcing their powerlessness. According to the original formulation of the theory, with the increase of the volume of information transmitted by mass media, social segments with higher socio- economic status are more likely to adopt this information faster than lower socio-economic social segments, thereby increasing the knowledge gap. Katzman (1974) applies Tichenor, et

3 al to the diffusion of new information and communication technologies (ICT) claiming that

“information rich” or haves’ social groups derive more benefits from ICTs than the

"information poor" or have-nots; consequently, this inequality between these two groups grows even wider. With the later proliferation of new information and communication technologies and the evolution of new media, the contrast between the haves and have nots acquired new significance. As these new media grew into powerful conduits of information and knowledge in society, access to them was recognized as a critical dimension of social inclusion: thus the term "digital divide" was created (Bakardjieva, 2003), representing the inequitable distribution of access to innovative technology and its complex and far-reaching consequences.

Bakardjieva (2003) goes into further detail by describing a type of politic that can be distinguished in relation to projects of self-identity that he borrows from Giddens. He refers to this as "emancipatory politics,” which its main points of interests lay in the idea of

"liberating individuals and groups from constraints which adversely affect their life chances."

Giddens notes that because emancipatory politics are essentially preoccupied with the elimination of exploitative, unequal, or oppressive social relations, it is of a kind of politics whose main orientation is "away from" rather than "towards." He concludes that it gives little to no indication as to what the substance of the free and autonomous choices made by emancipated individuals would be (Bakardjieva, 2003).

Digital Exclusion

Numerous authors have identified the existence of such a digital divide: What usually differs is the means by which the divide is categorized.

Also referred to as digital exclusion, the primary issues with the digital divide are the negative consequences for both the population affected and economy with the creation of the

“digital underclass,” (the significant amount of individuals who, for whatever reason are being left out of the digital revolution)("DIGITAL INCLUSION: All inclusive," 2006). One may argue that there are people who do not want to “get connected,” and that the consequences of their actions are those of personal choice. Unfortunately, the choice to be connected may not be wholly willful (Ilie, 2005), and those who appear to have chosen digital ignorance may not have done so from a fully-informed position (Riggins, 2004).

Further, as access to fundamental services increasingly shifts online and the economy begins to demand digital capacity from its citizenry, those without the Internet will not only find themselves severely disadvantaged but society also loses the impact and influence of their participation ("DIGITAL INCLUSION: All inclusive," 2006) Consequently, society can no longer “sit aside” and do nothing while the digital divide continues to marginalize a section of population.

The digitally excluded are not particularly easy to categorize; indeed, the literature not only supports the existence of a digital divide, but demonstrates that there are many types of digital divides, among a number of demographic variables such as race, income, gender, etc., as well as a number of differing ways to classify the digitally divided. Rice and Katz

(2003) identify and analyze three kinds of digital divides for both the Internet and mobile phones technologies alone:

1. user versus nonuser

2. veteran versus recent adopter

3. continuing user versus dropout

They also describe the depth of similarities and differences among these digital divide categories based on demographic variables. For example, the gap between Internet users and nonusers is associated with income and age; the gap between mobile phone users and nonusers is associated with income, work status, and marital status. The veteran/novice

Internet gap is predicted by income, age, education, phone user, membership in community religious organizations, having children, and gender; for mobile phones, age, work status and marital status are the primary predictors. The gap between continuing and dropout users is predicted by education for Internet usage and by income for mobile phone usage. Marginally significant cross-categorization of Internet and mobile phone usage/non-usage is distinguished (significantly although weakly) primarily by income and education. Thus, there are several digital divides, each predicted by somewhat different variables (Rice & Katz,

2003).

Sarkar (2005) goes a step further in his research where he studies the diffusion of the personal computer and the Internet across households in the United States. He asserts that existing studies in this context are of a descriptive nature and suffer from serious methodological problems1. He was therefore unable to investigate relevant policy questions, such as the predicted future dimension of the digital divide and the dynamic impact of government programs initiated to bridge this divide. His primary contribution is a set of household-level findings that indicate that the digital divide exists across different groups of the U.S. population. He also concludes, everything else remaining constant, this divide will not, as some economists have argued, close in the near future. He forecasts the exact

1 An entire sub-section of this chapter will discuss numerous authors’ support of the current, serious methodological issues in studying the digital divide as referenced in this statement.

6 magnitude of the divide across various dimensions such as income, education, race, etc.

Secondly, his research yields strong evidence in favor of social learning or network effects that are commonly cited as justification for government interventions. This suggests a different set of policy prescriptions, such as selective tutoring, compared to existing price- based policies such as subsidies to encourage the adoption of the Internet among diverse groups (Sarkar, 2005).

Despite the new economy that has brought a plethora of prosperity to America, many authors identity the seriousness of the digital divide which is emerging, creating continually widening gaps between technological haves and have-nots by region, race, socioeconomic class, and gender (DLC, 2001; Richard, 1999; Gaillard, 2001). A report by the Progressive

Policy Institute, the think tank of the centrist Democratic Leadership Council, identifies further the existence of this divide within the United States. Their reports rank the 50 states on how well they are adapting to the new economy. The report identifies a clear geographical pattern: the West Coast and Eastern Seaboard from New Hampshire to Virginia are at the forefront of the 21st Century Economy (Dunham, 1999).

Gaillard (2001) also identifies the existence of an electronic digital divide within the

United States. He identifies the digital divide as it concerns access to the Internet and its corresponding technologies. “The U.S. government is concerned about the digital divide because it appears that certain ethnic groups and income levels are being excluded from computer technologies and the Internet. These groups include African Americans and

Hispanics, who are lagging Whites significantly in gaining access to the Internet” (Gaillard,

2001). The gap between majority and minority groups appear to be widening. Gaillard’s research also has industry implications as well as providing evidence as to why the digital

7 divide must be eradicated. Since Internet access is a prerequisite to electronic commerce, an understanding of the relationship between the digital divide (especially as it related to technology access as well as usage when the access is abundant) and marketing (in particular as it is related to target marketing or marketing segmentation2) is essential. Federal, state, and local governments are trying to reduce or eliminate the digital divide to ensure equal access to all citizens. Marketing would benefit if equal access also meant increased electronic commerce. As a result, business leaders are also concerned about the digital divide because, as previously stated, it affects access to the Internet and corresponding technologies. If the consumers are denied access to the Internet, it will be difficult for them to participate in business through consumer level electronic commerce. However, his research has shown statistically that solving the problems of the digital divide will not necessarily aid business to consumer (B2C) level electronic commerce. The research has further found that the apparent reasons for the digital divide, currently thought to be income, education, and ethnic orientation, may be less important than initial government surveys indicate. The research demonstrates that between Internet access and consumer intent to purchase goods and services in business to consumer electronic commerce lies at least three other considerations that need to be addressed by business leaders: consumer trust, consumer commitment, and consumer involvement with Internet technologies. These issues vary largely by an individual’s intensity of social-identification (discussed later). All of the aforementioned considerations are important links between using the technology in general as well as using the technology for business to consumer (B2C) e-commerce. Gaillard's (2001) research also

2 Target marketing / market segmentation - dividing a consumer market into distinct subsets or segments that behave in the same way or have similar needs.

8 shows that these three areas have a combined relationship to the magnitude of the digital divide. Conclusively, any factors / actions that affect these three constructs will also affect the digital divide and vice versa. Business leaders seeking to engage in B2C e-commerce must pay attention to these three characteristics, consumer trust, consumer commitment, and optimizing the consumer experience (involvement), when using the Internet. Not addressing these issues proactively will increase the likelihood of failure while engaging in electronic commerce- or stagnating the digital divide (Gaillard, 2001).

Kvasny & Keil's (2006) research examines efforts undertaken by two U.S. cities in the state of Georgia (Atlanta and LaGrange) to address the digital divide. Her research observes that Atlanta's initiative has taken the form of community technology centers; citizens can go to community technology centers for exposure to Internet technologies, and learn about computers and various applications. On the other hand, LaGrange has taken a very different approach, providing free internet access to the home via a digital cable set-top box. Using theoretical constructs from Bourdieu3, this research analyzed 1) how the target populations and service providers reacted to the two initiatives, 2) how these reactions served to reproduce the digital divide, and 3) the lessons for future digital divide initiatives – the latter includes the reasoning for this research’s incorporation into this dissertation. Within her findings and analysis, there was no question that learning does indeed occur and that the programs are beneficial; however, there was no mechanism for people to go to the next step

(e.g. technical certification, college, purchasing a personal computer or “escaping the poverty that put them on the losing end of the digital divide in the first place” (Kvasny & Keil, 2006).

3 Theoretical Constructs from Bourdieu – this included but is not limited to social capital, symbolic capital, cultural capital, field, habitus, illusio, and reflexivity principles.

Findings of the researchers conclude that the Atlanta and LaGrange programs could be classified as successes in the sense that they provided access and basic computer literacy to people lacking these resources. However, both programs were, at least initially, conceived rather narrowly and represent short-term, technology-centric fixes to a problem that is deeply rooted in long-standing and systemic patterns of spatial, political, economic and most importantly for the purposes of this dissertation, social disadvantages. A persistent divide exists even when cities are giving away (theoretically) free goods and services (Kvasny &

Keil, 2006). In conclusion, the digital divide is a serious problem, not something resolved with a quick fix.

Therefore, part of the concern with the digital divide comes the issue of the diffusion

(and use) of technology (as opposed to access to technology). Rivas (2004) investigates differences in computer ownership levels between the four major racial/ethnic groups in the

United States: African-Americans, Hispanics, Asians and Whites. Rivas's (2004) study improves on previous studies in three major ways: First, he conceptually frames the digital divide within social stratification theory4 (which provides an appropriate starting point from which to understand diffusion of innovation theory5). Diffusion of innovation theory serves to explain the process of how an innovation is adopted, but does not explain how certain groups become privileged enough to be early adopters (as mentioned in the Crossing the

Chasm covered earlier in Introduction). Rivas's (2004) technique, via framing diffusion of innovation theory within an understanding of social inequality and social stratification

4 Social Stratification Theory – a theory developed by Weber (1957); its basic principle are based on 1) how societies are organized in hierarchical systems of domination and subordination and 2) the significance of power in the determination of social relationships based on domination and subordination. 5 Diffusion of Innovation Theory – developed by Rogers (1995); individuals are seen as possessing different degrees of willingness to adopt innovations (segmented into the groups (from most willing to adopt to least willing to adopt): innovators, early adaptors, early majority, late majority, laggards).

10 processes, his study advances a better understanding of why the digital divide exists between and even within the four major racial groups in the United States: African-Americans,

Hispanics, Asians and Whites. He also argues for and expands on the idea of social context as it may affect computer ownership. Additionally, he makes use of the four most recent cross-sections of Current Population Survey6 data which contains information on computer ownership and finds that even after controlling for household composition, demographic and socioeconomic factors, the gap between African-American and Hispanic households in relation to White households remains significant. As a side note, for Asian households, the racial effect was explained by the variables in the models7, but this is not the case for

African-American or Hispanic households. Additionally, there exists significant interaction between the presence of immigrants and race of household, demonstrating that African-

American immigrant and Asian immigrant households are much more likely to own computers than their similarly situated non-immigrant counterparts (African-Americans and

Asian-Americans, respectfully). No evidence suggested that computer adoption is occurring at a faster or slower rate for African-American or Hispanic households in relation to White or

Asian households- the significance seemed more within the latter mentioned races. He concludes with evidence of a social context effect as well (Rivas, 2004).

Furthermore, this dissertation investigates the ethnic differences in technology use, adoption and access (as well as the motivations for such) as it relates to African-Americans and Whites. While most research supports the theory that the Digital Divide (in this research the terms digital divide and digital division will be used interchangeably) is a separation of

6 Current Population Survey (CPS) - a statistical survey conducted by the United States Census Bureau for the Bureau of Labor Statistics (BLS) who use the data to provide a monthly report on the U.S. employment statistics. 7 This dissertation will not cover such effects as it pertains to the Asian-American population.

11 haves versus have-nots on the bases of income and access, the aforementioned and upcoming sections aimed to explicitly demonstrate that while this is true, African-Americans have the lower income levels and the limited access to / availability of innovative8 technology and even more importantly, if access is available, less motivation to use than their white counterparts (the same holds true for females as compared to males). In their study on racial differences in the magnitude and composition of wealth, Blau and Graham (1990) analyzed data from the 1976 and 1978 National Longitudinal Surveys of young men and young women. Their findings revealed that young African-American families held 18% of the wealth of young white families. Additionally, African-Americans shared several characteristics that could lower their net worth (relative to Whites), including lower income and demographic factors such as a higher incidence of central city residence and families headed by single women. Although the income difference between African-Americans and

Whites was the largest single factor explaining racial differences in wealth, 75% of the wealth gap remained after controlling for income and other demographic factors. The researchers concluded that racial differences in intergenerational transfers of wealth and, to a lesser extent, barriers to the accumulation of business and home equity most likely play a role in the economic divide between African-Americans and Whites (Blau & Graham, 1990).

Bartel goes further to substantiate that African-Americans earn less than Whites in America

(1981).

8 Innovative technology – a technology that is perceived as new by the consumer

The Digital Divide as an Issue of Socio-Economic Division

Numerous researchers support the digital divide as an issue of socio-economic division (Wareham, Levy, & Shi, 2004; Anindya, Kenneth, & John, 2005; Campbell, 2001;

Yap, Das, Burbridge, & Cort, 2006).

Anindya et al. (2005) analyzes the impact of a variety of socio-economic influences on households' decision to pay for basic Internet access. Traditional socio-demographic variables are apparent; according to Anindya’s research, the influences of income and education seem to be particularly strong. The analysis also suggests that the simple subscription decision is only modestly sensitive to price implying access subsidies for basic access are unlikely to be a highly effective tools in bridging the digital divide(Anindya et al.,

2005). In other words, something other than price and access is influencing whether or not

American households will pay for basic internet access.

Campbell (2001) asserts that the use of new information and communication technologies (ICT) is advancing rapidly, but diffusion patterns are less clear and quickly change. There is grave concern that such rapid, uneven change is further widening the digital divide and may exacerbate the existing socio-economic divide between them the two sides

(Campbell, 2001).

Yap et al. (2006) poses the research question: Why are some successful with e- commerce while others flounder? The purpose of his article was to study the impact of technology, cultural, and socio-economic factors on the global diffusion of e-commerce.

While past studies have focused on technology reasons alone, his research includes cultural and socio-economic factors as well. Having access to the Internet does not necessarily translate to e-commerce usage. He concludes that, fundamentally, culture and socio-

13 economic factors are pivotal in bridging the gap between Internet usage and e-commerce diffusion (Yap et al., 2006).

The Digital Divide is Not an Issue of Access, but Use

The emergence of information technology has created a need for newly skilled workers able to use technology to advance society. Acquiring proficiency in working with technology, however, initially requires access to the technology. This increase in technology has brought attention to those people in various communities who have been digitally divided. The digital divide represents a divide (with respect to technology) between two distinct groups of people--the information rich and the information poor. A more detailed definition of the digital divide incorporates access to technology, quality electronic resources, and training on the use of the technology (President, 2005). This particular study examined one of Brooklyn’s community technology centers (whose primary purpose is to help close the digital divide) and also explored the perceptions and attitudes about technology of those who frequent this community technology center.

Access At Home At School At Work To Computer 31% 22% 18%

To Internet 22% 25% 9%

Table 1: Access Statistics from Participants of Brooklyn Community Technology Center

The top 3 reasons participants came to the CTC include:

1. check email (59%)

2. surf the Internet (53%)

3. improve computer skills (51%)

Other statistics from this study include:

 54% felt the Internet was too expensive

 75% felt they were missing out by not using the Internet

 88% earned $30,000 or less annually

Overall participants in this study were dissatisfied with their level of access. Since they valued the importance of computers and the Internet, they attended this center to better themselves and acquire new skills. Additionally, by coming to Brooklyn’s CTC, they were able to compensate for the lack of access to computers and Internet access at home, work and school. Some of the participants felt they had enough access because of this center while others still felt they did not have enough access, despite the center (President, 2005).

Brooklyn’s community technology centers provided Internet access to people in the community; however, the centers by themselves were unable to meet the access needs of all participants (Silva, 2002).

Terrell-Simmons (2002) conducted a study that analyzed the African American digital divide in New Jersey from the perspective of graduates of digital divide programs.

Seventy-five African American respondents were surveyed to collect pre and post-graduate data concerning their computer and Internet access and usage trends. The digital divide has been identified as the separation between those with and those without technology. Terrell-

Simmons (2002) attests that Northeastern states such as New Jersey have experienced great challenges in bringing technology to those who live in the inner city. The challenge includes not only in bringing technology to the have-nots, but also getting them to use it. Results from the data he collected suggest that computer access and usage as well as Internet access

15 and usage increased for those respondents graduating from digital divide programs. The data also suggests applications of such programs have the potential for effectively improving computer and Internet skills of those who utilize them. While digital divide programs show promise for helping those who use it, others will still have difficulty becoming successful in society due to their lack of access and usage of such technology due to economic reasons. He concludes that further research is necessary to assess the persistent digital divide disparity among African Americans (Terrell-Simmons, 2002).

Goldsborough (2000) asserts that many people lack access to the Internet, particularly minorities, the poor, the less educated, and those living in rural and inner city areas

(Goldsborough, 2000). The chasm is between people living in metropolitan areas and those in “micropolitan areas” -- the federal government's name for what used to be called "small towns" is continually growing (Bulik, 2006). In an attempt to reach the most people with the least dollars, technology companies concentrate their marketing spending in consumer-dense metropolitan ones, giving city dwellers an information heads-up. Many agree that the metro- micro tech gap is burgeoning (Bulik, 2006).

Thusly, this section demonstrates the digital divide is not necessarily a function of access, but use.

The Digital Divide Begins in the home Environment, Highly Influenced by Parents

Numerous authors suggest that the digital divide is highly rooted in African-

American homes. Furthermore, while public schools have made significant gains in providing computer and Internet access, minority and poor students lack computer access outside of regular school hours, according to two reports by the National Center for

Education Statistics at the US Department of Education (Roach, 2003). This section of the dissertation helps demonstrate where the social context and influence of the digital divide may originate.

Dalton & Yeung (2005) attributes possible detriments to the African-American community as far as the digital divide is concerned to the parents. (Figure 1 gives a graphical representation of this divide in the home.) He examines whether differences in parental occupational prestige mediate or moderate race differences in four indicators of child development - reading scores, math scores, Behavior Problems Index, and health status - using data from the Panel Study of Income Dynamics Child Development Supplement.

Dalton & Yeung's (2005) findings reveal that although for behavioral problems there is no impact of parental occupational prestige, for reading, math, and health, there are significant academic returns to parental occupational prestige, but only for White families. The author hypothesizes that this racially distinct dynamic may be a result of ongoing discrimination in the labor market, thereby reducing the association between ability (job and parenting) and prestige; or it may be a result of the difficulty of African-Americans to translate occupational prestige gains into other benefits as a result of discrimination outside the labor market; or finally, it may be the result of a generational lag between occupational status and parenting practices (Dalton & Yeung, 2005).

Similarly, Allan suggests that the influence of the role model unit is essential in shaping the child’s motivation as mentioned earlier. African-Americans were found to have a greater motivation to manage than Whites. It is suspected that differences in motivation between the two groups may have been due social factors relating to work experience, influence of family members, the existence of role models in high school,

Figure 1: Persons Using the Internet at Home (by Race)

college, the family, or the community, and other factors. The results suggest that African-

American male college business students may be an important but relatively underutilized source of management talent for US business (Allan & Nellen, 1992).

Why Study the Digital Divide

The digital divide is not only about offering Internet access to every citizen, nor is it only about social policy or computer penetration. Brotman (2002) states that in fact, the stakes are much greater. Bridging the digital divide will enable businesses to thrive at a new level of post-industrial innovation. As he coins the term “digital dividend” this represents the set of outcomes that the private sector can achieve by promoting widespread penetration and more importantly use of digital technologies. Also, from an industry perspective, this can

18 translate into better-trained, more productive employees, expanded sales and marketing opportunities at home and abroad, as well as a more diverse supply chain (Brotman, 2002).

In addition to the aforementioned economic prosperity that bridging the digital divide will bring, Mistry (2005) asserts that studying the digital divide requires an interdisciplinary approach (as this dissertation will explore in further depth later). In the current era of technological growth and innovation, the role of information and communication technology

(ICT) as a catalyst to enhance economic development and the quality of life has become an increasingly important issue. The term digital divide has been used to refer to the gap between those who have access to and again, more importantly utilize technology versus those who do not. Based on the literature on the digital divide, the phenomenon calls for broad-spectrum interdisciplinary frameworks to guide future research (Mistry, 2005).

In addition to the need for an interdisciplinary approach, the digital divide itself has been poorly studied. The digital divide has traditionally described inequalities in access to computers and the Internet between groups of people based on one or more social or cultural identifiers (Gorski, 2002). Accordingly, researchers tend to compare rates of physical access to, or rates of actual use of, these technologies across individuals or schools based on race, gender, socioeconomic status, education level, disability status, and first or primary language.

The divide refers to the difference in access rates among one or more groups; for example, the racial digital divide is the difference in computer and Internet access and usage rates (at home, school, work, or other locations) between those groups with higher rates of access and usage (white people and Asian Pacific Islander people) and those with relatively lower rates of access and usage (Native American, African American, and Hispanics). However, this traditional understanding and corresponding research of the digital divide fails to capture the

19 full picture of inequity and privilege recycled by these gaps and the resulting educational, social, cultural, and economic ramifications, primarily for groups of people already educationally, socially, culturally, and economically oppressed. Meanwhile, such a limited view of the digital divide serves the interests of privileged groups and individuals, who can continue critiquing and working to dissolve gaps in physical access and use rates while failing to think critically and reflectively about their personal and collective roles in recycling old inequities in a new era of ITCs. A full understanding of these disparities must begin with an examination of educational and socialization processes as early as elementary school

(Gorski, 2003).

The Digital Divide is Growing Larger

Left unchecked, the digital divide will continue to grow and began to more adversely affect the U.S. and the global economy (Kallol, 2005).

Although home computing and broadband have become more affordable and more widespread, the digital divide is still expected to widen. Many poorer households are likely to remain unable to afford the cost of a computer and connectivity and as emphasized in this dissertation. However, wealthier households with Internet access will pay less to connect to a growing range of products and services, from innovative technology to voice communication. The overall fundamental problem remains the economic gap between rich and poor, between working class and middle class. The digital divide is just one of many causes that must be addressed. For example, providing free Wi-Fi coverage to everyone is laudable, but it is unlikely to have much impact unless combined with subsidized equipment and training (Tansley, 2006).

In general, as technology continues to advance, the digital divide grows wider. The paradox of the digital age is that while technology expands opportunities for a significant number of people, it also can divide them further by highlighting the differences between the haves and the have-nots. Some refer to this problem as the digital divide, but it is really another manifestation of the issue of poverty. At its core, the economic gap in America is more than just the difference in income, it is an asset gap - a gap in skills, information, and resources (Ramsey, 2003).

There is a Need for a Different Research Approach to Studying the Digital Divide

Numerous authors identify the need for a different research approach to the digital divide as well as information technology in general. This is coupled with the need to include ethics in digital divide research (mentioned in the previous section).

One such author introduces the notion of “diffusionism9.” It is pervasive within IS research and practice. Generically, diffusionism denotes an asymmetrical view of innovation as originating exclusively in "progressive" centers, from which it spreads through an essentially passive recipient community. This malicious model privileges an élite few over the majority, with the deception of the “innovator/imitator dichotomy” presented as natural, moral and inevitable. Research failed to find any empirical support for diffusionism; however, concludes empirical examples demonstrate both the ubiquity of the diffusionist mindset in IS research and practice (McMaster & Wastell, 2005). In other words, his research concludes by arguing for a more critical approach within IS research on innovation, the use

9 Diffusionism- a notion that provides a general way of understanding innovation and human progress

21 of richer, process-based theories, and greater partnership with practitioners in order to close the research/practice gap – as this research is aimed to do (McMaster & Wastell, 2005).

The digital divide has been identified as a primary concern for those interested in both domestic and global IT development in view of its impact on economic development and quality of life. However, conducting research on the subject requires a broader perspective than found in traditional IS research. A broad-spectrum interdisciplinary approach is needed.

Prior research in technology adoption can be a source of research models; however, these models must be broadened to deal with the range of economic and cultural factors on societal levels. The current models should also be expanded to address IT adoption both within and outside of their social constructs. In other words, such social determinants can no longer be taken as a given. Hence, this dissertation research is needed to help explain IT adoption within and outside of varying social constructs. In addition, the effects of technological changes over time need to be studied since they are integral to the Internet revolution.

Findings from such research should benefit economic and education policy making, corporate business strategy formulation and would improve disadvantaged individuals' quality of life (Lu, 2001; Kauffman & Techatassanasoontorn, 2005).

An important distinction of this research is the realization that we must refocus research on people not technology. The future role of new technologies and their poverty reduction potential should be measured against a thorough understanding of the social subjects that they generate rather than upon examination of the technologies themselves.

Optimism must be tempered with realism if a suitable effective framework is to be developed

(Cawkell, 2002). This can began by incorporating youth in digital divide research. Trudeau outlined his vision for closing the digital divide and giving youth a voice in shaping the

22 technology agenda. If electronic information is presented in a native language, then that culture begins to understand it does have power over the information world. By disenfranchising the youth from the technology equation, an incredible split will occur when youth reach their 30s (MacInnis, 2002). This is one reason for this research utilizing college undergraduate students as a sample.

Must Include Ethnicity in Digital Divide Research

As there are many reports about the digital divide, there are many discrepant interpretations of what the reports indicate. This pattern of inconsistent analyses, often in relation to identical data sets, has existed at least since the last decade. Hacker & Mason

(2003) argues that a major problem with much of the digital divide research is the failure to include ethnicity concerns as an explicit part of analyzing and interpreting digital divide gaps

(Hacker & Mason, 2003). If researchers include more recognition of ethnicity with their findings about divide gaps, it is likely that they will produce better research and findings as well as more defensible linkages between study reports and policy deliberations (Hacker &

Mason, 2003).

As a follow-up article, Phillip Brey calls for more analysis of ethical issues when it comes to technology, particularly of virtual reality (Ford, 2001).

It is important for African Americans and other minorities to understand the path of technology evolution and the implications it will have on their lives. Bill Gates, in his book

“Business @ the Speed of Thought” (Warner Books, 1999), talks about the adoption of a

"Web Lifestyle" in which people constantly look to the Internet to retrieve information, conduct fundamental household transactions, communicate with their friends and colleagues

23 and do a whole host of other things. He suggests that this ever-widening connection will evolve into a "Digital Nervous System" from which the world will operate. Those most outspoken about the digital divide would have you believe that because African Americans and Hispanics are significantly behind the curve in terms of having personal computers in their homes that they will be woefully and forever left behind. The facts provide a slightly different picture. While naming the initial disparate distribution of Internet access the digital divide has certainly been valuable in bringing attention to a trend that needs to be addressed; however, Morse argues that that same name does a disservice to the very communities it seeks to help because it causes African Americans and other minorities to be prematurely placed in a collective socio-economic and psychological ethos of inferiority before the technology game has even been played (Morse, 2000).

Johnson (2002) also asserts that race cannot be ignored as an aspect of the digital divide. The digital divide is becoming a huge gaping wound for many people, including

Hispanics and African-Americans, in the landscape of America (Johnson, 2002). According to a recent study by Gardyn (2001), African-American and Hispanic Americans use and perceive the Internet differently than general market consumers as a result of their distinct cultural backgrounds and values. As Internet participation rates of multicultural segments grow, shrinking the so-called digital divide, marketers who pay attention to these differences will be better able to serve the needs and capture the loyalty of these online consumer segments. All Internet users, regardless of race or ethnicity, overwhelmingly find the medium to be a positive addition to their lives and one that makes their lives easier, according to the study (Gardyn, 2001).

Despite the increasing availability and affordability of information technology, minorities are likely to remain confined to being technology "have nots," according to a

Creighton University law professor. “In The Digital Divide: Standing in the Intersection of

Race and Technology," Raneta Lawson Mack examines the racial component of the digital divide and discusses historical reasons why minority communities might fail to embrace computer technology even as it becomes easier to acquire (Roach, 2002).

The digital divide is not just a pressing issue of social equity and race. It is fundamental to Americans' ability to innovate, stay competitive, and generate economic wealth. The US is truly multinational, multicultural, and multiracial. In an age where talent is the key factor of production, we must utilize America's vast and diverse human resources.

But many people are being left behind as the move to a high-tech knowledge economy reinforces long-standing social and economic cleavages. For example, only one-third of

African-American and Hispanic households own computers, compared to more than half of all American households (Florida, 2001).

Need for Digital Diversification

Studying the digital divide will make tracks into moving toward digital diversification.

According to Kang (2000), most inquires to date into race and cyberspace have focused on the digital divide asking whether racial minorities have access to advanced computing-communication technologies. A more fundamental question is: Can cyberspace change the way that race functions in American society? Kang’s analysis starts with a social- cognitive account of American racial mechanics that centers the role of racial schemas. It

25 argues that cyberspace can disrupt racial schemas because it alters the architecture of both identity presentation and social interaction. It concludes that society need not adopt a single uniform design strategy for all of cyberspace. Instead, society can embrace a policy of digital diversification (Kang, 2000).

Similarly, according to a new report by Insight Research, Internet penetration rates for Hispanics and Asians in America will grow several times faster than that of the rest of the population over the next 5 years. The report advises Web portal, Internet, and IP telephony providers to develop affinity portals based on demographics such a gender and race to differentiate themselves from all-purpose portals (Sherwood, 2001).

Must Eradicate the Digital Divide

Payton (2003) reveals the perspectives on the digital divide of 41 African-American teenagers attending 7 high schools and one middle school in the Wake County public school system in Raleigh, NC. The students were asked about their views regarding college, intended major, family education, and role models. Most of the teenagers specified that they planned to pursue higher education. Nearly all of the students said their parents had completed high school, 61% of their parents had completed college, and 14% said their parents had attended college without earning a degree. With respect to the digital divide, 90% of the participants had never heard of the phenomenon, but the 4 students who were aware of it had strong views on the resources needed to reduce it. While efforts to eliminate the digital divide are valuable, public policymakers must seek to improve the social network that will help the digital divide resolve itself (Payton, 2003; Chaker, 2005).

Bridging the Digital Divide Requires More

"For companies that are increasingly focused on global emerging markets, understanding socio-economic factors impacting technology adoption in the various regions is absolutely crucial," said Patrick Moorhead, chairman of the GCAB10 and vice president of corporate marketing at AMD. "Bridging the digital divide requires more than simply offering computers and Internet access" (Hachman, 2003).

Similarly, when discussing the digital divide between developed countries and smaller, poorer countries, Adam Peake stated, “You cannot expect installing a high-end network infrastructure in a small country can help them bridge the divide" (Ridzuan, 2001).

Ownership issue should always be considered in tackling digital divide. The idea is not just to get people to have things or technologies but to ensure that continuous education is carried along the process (Ridzuan, 2001).

Roach says the lack of community-based programs and resources to equip and sustain underserved populations with significant information technology skills can override the effect of basic efforts to bring information technology to distressed communities. Dr. Lynette

Kvasny's conclusions come from her research of two Georgia cities that implemented technology initiatives to address the digital divide (Roach, 2003).

As the digital divide refers to the separation between those who have access to digital information and communications technology (ICT) and those who do not, many believe that universal access to ICT would bring about a global community of interaction, commerce, and learning resulting in higher standards of living and improved social welfare. However, the digital divide threatens this outcome, leading many public policy makers to debate the best

10 GCAB – Global consumer Advisory Board – sponsors the technical terminology and complexity test.

27 way to bridge the divide. Much of the research on the digital divide focuses on first order effects regarding who has access to the technology, but some work addresses the second order effects of inequality in the ability to use the technology among those who do have access. In Sanjeev’s research, both first and second order effects of the digital divide at three levels of analysis: the individual level, the organizational level, and the global level. He also suggests a series of research questions at each level of analysis to guide researchers seeking to further examine the digital divide and how it impacts citizens, managers, and economies

(Dewan & Riggins, 2005).

Mack focuses on corporate and community-based solutions giving special attention to the individual needs of specific communities. Mack contends that resolving the digital divide doesn't merely depend on placing high-speed computers in homes. Instead, she says, it means educating those who are not aware of technology's benefits, designing content that is relevant to those communities, and increasing access. "As the saying goes, it's the difference between giving a person a fish and teaching him to fish," Mack notes (Roach, 2002).

Digital Divide Deniers

There are numerous researchers who believe the digital divide is a “hype” “myth” or manipulation of statistics. Lee Hubbard is one such researcher who refutes much of the theory of the digital divide between minorities and whites. Lee suggests that the level of computer and Internet usage among African-Americans might be much higher than the statistics would indicate. He goes further to acknowledge the possibility of a digital divide but asserts that is a function of socio-economic class rather than one of race (Hubbard, 2000).

It is for this reason I take time to assert that indeed the digital divide exist and thus reasoning for our research.

Summary Table

Table 2: Detailed Summary of Chapter 1 Digital Divide (DD) Authors Key Notes Origin of DD Tichenor, Olien, & Authors proposed the theory of Donohue the "knowledge gap" Katzman, 1974 “information rich” or haves’ derive more benefits from ICTs than the "information poor" or have-nots’. Bakardjieva, 2003 Term “digital divide” was created as new media grew into powerful conduits of information. Bakardjieva, Giddens emancipatory politics - 2003 liberation constraints which adversely affect life chances life-politics - remoralising" of social life Consequences of a DD Riggins, 2004 Fundamental services are shifting online and economy demands digital capacity. People w/out access miss numerous opportunities and society misses their input. Rice and Katz 2003 There are many types of digital divides. The gap is predicted by Ronald, 2003 different variables. Sarkar, 2005 Research yields strong evidence in favor of social learning or network effects DLC, 2001; Richard, There is a continually widening 1999; Gaillard, 2001 gaps between technological haves and have-nots by region, race, socioeconomic class, and gender

Table 2: (continued) Richard, 1999; Kvasny Geographical differences in 2006; Roach, 2003 digital divide Rivas, 2004 Racial divides Blau and Graham Digital Divide as an Issue Jonathan, Anindya, 2005; of Socio-Economic Duncan, 2001; Division Alexander, 2006 Not an issue of access, but President, 2005; Ronald, DD programs can help increase use 2003; Terrell-Simmons; computer access, usage, & Beth Snyder Internet access and usage Digital Divide Begins in Dalton, 2005; Allan Parents/ role model influence in the home child tech. access and use; African-American males had a higher motivation to manage than white males, white females, and black females Why Study the Digital Stuart, 2002; Jamshed, Decreasing DD yields corporate Divide 2005; Gorski, 2002; advantages including better- trained, productive employees, diverse supply chain; future research must address broad- spectrum interdisciplinary frameworks and inequity and privilege recycled by DD gaps The Digital Divide is Kallol, 2005; Rey, 2003 The overall fundamental Growing Larger economic gap exists between rich and poor, between working class and middle class; Additionally, there is an asset gap - a gap in skills, information, and resources Different approach needed Tom, 2005; Ming-te, More critical approach needed to study the DD 2001; Robert, 2005; within IS research and to Tony, 2002; Patricia, explain IT adoption within and 2002 outside of varying social constructs. Focus on people (vs. technology) Ethics in DD research Kenneth, 2003; Paul, Need for sound analyses and 2001; Morse, 2000; interpretation of DD; ethics in Lance, 2002; Rebecca, virtual reality and multi-user 2001; Ronald, 2002; environments; Adverse effects Richard, 2001 of the term DD; distinctions between minority groups. Need for Digital Jerry, 2000; Sonja, 2001 Diversification

Table 2: (continued) We Must Eradicate the Fay Cobb, 2003 Society must develop a social Digital Divide network that will help the digital divide resolve itself Bridging the Digital Mark, 2003; Ridzuan, Proper sustained support, Divide Requires More 2001; Ronald, 2003; community based programs and Sanjeev, 2005; Roach, resources needed to bridge DD 2002 DD Deniers Lee, 2000 Suggest DD between race is a manipulation of statistics there is a DD in socio-economic class

Summary

As a general review of this section, research that substantiates the fact that race matters and must be included in research was presented. This chapter provides extensive evidence that indeed the digital divide does exist. This evidence is taken a step further by showing evidence that there is a digital divide present in the United States.

Following this evidence, I identify the significance of culture’s socio-economic status concerning its impact on the digital divide. This was transitioned into the issue of access, identifying that more important than access, is actual usage (i.e. the problem of access is decreasing, yet the problem of usage of innovative technology is increasing, although it is available). After supporting this existence, I began to explore its social contexts which are found in its origin. Additionally, I continue its social implications via demonstration of digital divides social influence beginning in the home environment.

After this stage is set, this chapter goes into depth of “why,” why study the digital divide and more importantly, why conduct this research. As a brief review, this research is conducted for the following five main reasons:

1. the digital divide is getting larger

2. numerous researchers have identified the need for different and

interdisciplinary approaches to study the digital divide

 including the incorporation of ethics in digital divide research

3. the need for digital diversification

 benefits for the next generation, especially

4. the fact that we must bridge the digital divide

 essentially realizing that bridging the digital divide will require more than

a “quick fix.”

After setting this important background, reasoning and support, the next chapter

(literature review) will go into depth of the issue of technology self-efficacy as well as tie in a new approach to digital divide research: the incorporation of social identity (i.e. primarily ethnic identity measure).

CHAPTER 2: LITERATURE REVIEW

Overview The last chapter discusses in extensive detail, the digital divide as it exists as a function mainly of use, as opposed to a function of access, income and demographic differences (although these variables may serve as moderators to innovative technology use as well as its predictors/antecedents).

Additionally, as stated in the previous chapter, the goal of this dissertation is oriented to study the motivation of minorities to adopt, learn and ultimately use new, innovative technologies. In this chapter, I will extensively review the literature of African-American /

White differences as it pertains to general motivation as well as motivation with IT and innovative technology. I will then explicitly differentiate between race and ethnicity as well as racial identity and ethnic identity and why ethnic identity was chosen as a moderating variable for the proposed model. The literature of ethnic identity (as a function of self and social identity) as it relates to both self-efficacy and technology will then be reviewed.

Towards this I will then proposed a leftward-extended technology acceptance model (TAM), incorporating the moderating variable of ethnic identity.

Race and Technology Participation in the Digital Division - General Motivational Race Differences

In order to solidify support for this research, it is essential to explain why “Race11

Matters” (Miller, 1998; West, 1993), particularly as it pertains to this research in explaining and supporting why the digital divide exists. Additionally, examining how and why the

11 Race versus Ethnicity- Later in this dissertation, I will further distinguish between race and ethnicity and then why ethnic identity was chosen over racial identity as a moderator in this dissertation research study.

33 digital divide exists as well as how and why it is expanding will better prepare society to combat this digital division in an effort to reach digital inclusion (West, 1993). In today’s contemporary world, access to information and knowledge play a key role in advancing economic and social well-being. The improvement of information and communication technologies (ICTs) has enabled large amounts of information to circulate globally while simultaneously being stored at increasingly higher speeds complimented by much lower costs. As a result, the relevance of physical distance has been diminished (Fors & Moreno,

2002). Thomas Freidmann (2005) also describes the reduction of physical distance as a result of technology. This demonstrates one of the key benefits of the information age. This also provides implications into why the Digital Divide must not only be acknowledged and studied, but seen as an issue that must be immediately addressed.

As demonstrated in Chapter 1, while many researchers posit that the digital divide is a function of income not race/ethnicity, this literature supports that African-Americans, statistically, have the lower incomes, lower level of access, etc. Proof of general racial differences is solidified even further by Blau and Graham (1990)whose study found racial differences in the magnitude and composition of wealth . Not only did young African-

American families hold only eighteen percent (18%) of the wealth of young white families, but African-American families also were shown to hold their wealth in proportionally different forms. Characteristics that lowered African-American’s net worth as compared to whites included (but was not limited to) lower income and a higher frequency of central city residence: income was still the largest single factor explaining the difference in wealth among the races. However, even after controlling for income and other demographic factors, as much as 75% of the wealth gap remained between the races (Blau & Graham, 1990).

There are numerous documented differences between races and various aspects of our society. Simply put, African-Americans and Whites are different as are males and females.

These differences range from differences in biological motivations (Guay, 2001) to expectancy outcomes (Expectancy Theory12) that may associate with one’s motivations.

Peter Allan and Eugene Nellen identify motivational differences among African-

American college business students, especially as it pertains to their motivation to manage.

According to their study, African-American males were less interested than Whites in standing out from the group as well as imposing their will on others, but did have a greater desire to assert themselves and take charge. The authors also note that the motivation to manage among African-American males differs depending on their collegiate environment.

For example, African-American students attending a Historically Black College and

University (HBCU)13 demonstrated higher levels of self-esteem, confidence and aspiration than those of African-American students attending predominately white colleges and universities (Allan & Nellen, 1992)- indicating the influence of one’s environment on their motivations (this importance will resurface particularly in the symbolic value of technology section of this paper).

Racial differences have also been identified in not only job satisfaction but also job expectations (Brenner & Tomkiewicz, 1982). As far as managerial positions in industry, mainly occupied by whites, one of their primary research goals was to determine whether

12 Expectancy theory of motivation was proposed by Victor Vroom. This theory is basically an attempt to come up with a model of how people would rationally decide whether or not to be motivated to pursue a particular course of action: In the case of this research study, the course of action is to adopt and utilize innovative technology. 13 Abbreviated (HBCU) – this heading refers to the 100’s of U.S. institutions, accredited by a by a nationally recognized accrediting agency or association determined by the Secretary of Education, established before 1964 who were founded on the continuing premise of the education of African-Americans

African-Americans and Whites expect the same things from these jobs. Brenner’s research shows strong evidence that African-Americans desire different characteristics in a job than whites. His research looked at twenty-five different job characteristics, of which, 40% revealed significant differences in job oriented preferences according to race. African-

Americans rated nine out of the ten preferences more important than whites. Their factor analysis revealed that African-Americans value long-range career goals and structure considerably more than their white counterparts, while their (African-Americans’) preference for intrinsic and extrinsic factors14 are not as strong. Examining the greater importance

African-Americans place on all job characteristics, the research suggests reasons for this outcome may include the possibility that African-Americans feel that by insisting on more of everything, they will be placed in a better bargaining position to obtain at least some of their own goals (Brenner & Tomkiewicz, 1982). Similarly, research by Thomas and Wise (1999) suggest that similar marked differences in employment preferences, especially among races

(also among gender which is not covered by this dissertation research) (Thomas & Wise,

1999). This is particularly important as it allows us to relate these differences in motivation to differences in motivations to adopt and utilize innovative technology.

Motivational differences are also observed in the racial/ethnic level early in child development. To identify the consequences of TV advertising on different types of children, a study was conducted to discover the cognitive responses and extra-product expectations

14 Intrinsic factors are those associated with human well-being through the satisfaction of three universal psychological needs: 1) autonomy, 2) competence, and 3) social relatedness. Intrinsically motivated behavior is considered as behavior which is freely engaged in; for example, enjoyment-based. Extrinsic factors are those most often associated with the engagement of activities only because they lead to desirable outcomes which are separate from the activity itself (such as tangible rewards). In other words, this type of behavior is a means to an end and not engaged in for its own sake. Furthermore, intrinsic motivation is most often associated with the involvement of complex tasks (for example those leading to highly valued outcomes such as creativity, vitality, spontaneity and quality); whereas extrinsic motivation’s importance is coupled in relation to those simple and unattractive tasks.

36 fostered by TV commercials for 2 specific groups: white and African-American children.

Study findings serve to reinforce the black/white (race) disparity in awareness of commercials' motives that was consistently found in previous investigations. The primary significance of this study to our research is the finding that African-American children generally perceived the commercials' social relationships and goods as superior to their own situations (Donohue, Meyer, & Henke, 1978). This importance will again resurface in the symbolic/social value of technology later in this chapter.

As this racial/ethnic disparity begins early in life, it is seen to progressively get worse as the child grows older and as the use of new technologies has moved from the periphery of the American educational experience to its very center. However, without any formal technology prerequisites, undergraduates start their college careers with differing technological knowledge, stratified by gender, socio-economic status, and racial backgrounds

(Goode, 2004). Much of this division can be traced to high school computing experiences, or lack thereof, which varies between high schools serving different student populations.

Findings from a case study of course offerings in California, reveal that high schools serving predominantly low-income, students of color offer low-level and remedial level computing curriculum, while schools attended by White and Asian students provide a more academically challenging set of computer courses for students (Goode, 2004). As this dissertation does not discuss gender, just ethnicity (or ethnic identification, for our purposes)

Goode’s research also reveals that even when academic computing experiences are available, females find the content uninteresting and do not participate at the same level as males. As a result, without authentic learning experiences that involve computers, students are less prone

37 to develop a technology identity, and a sense of competency which influences their university work as well as the technology used to complete their work (and play)(Goode, 2004).

Further solidifying the reasoning for and connection to this research, differences among African-American and white management information systems (MIS) professionals and managers are acknowledged by (Igbaria & Wormley, 1992), In their study, sixty-nine black-white pairs from a telecommunications company were matched on age, job function, tenure with the organization, and organizational level. His research yields that African-

Americans saw less discretion and less support from supervisors in their jobs than their white counterparts; additionally, African-Americans also felt that their expectations had not been met as well as their white counterparts. However, African-Americans did not, feel less acceptance by their organization or that they were participating less in training programs. Job discretion, career support, and participation in training programs all proved to be related to supervisor ratings of job performance while feelings of acceptance and met expectations were not. Whites received higher performance ratings than African-Americans, even after controlling statistically for the effects of organizational experience. Job performance was positively related to supervisor predictions of advancement prospects. White employees with higher advancement prospects were more satisfied with their careers (Igbaria & Wormley,

1992). As evidenced here, there is even a divide among technically inclined African-

Americans and whites, further solidifying our reasoning for this study.

The aforementioned paragraph included a short list of the many research-supported differences between African-Americans and Whites which further supports the basis for this research. African-Americans and Whites are different (as are males and females- however, more evidence is supported here concerning racial differences due to its more controversial

38 nature). The above examples are provided to begin to demonstrate our basis for this research, African-Americans and whites, are different and those differences must be understood in order to utilize the benefits of diversity.

It is essential to note that such differences are not included to create or perpetuate the division between African-Americans and Whites but to acknowledge those differences, learn from them and identify better ways to study these differences to contribute knowledge toward reaching digital inclusion or technical Empowerment. To visualize this, picture individuals with different personalities. Psychologists may use a Myers-Briggs15 assessment to determine the personality-identity of individuals within a group. The research conducted by the aforementioned was designed to not only acknowledge the different personality- identity people have, but also to study and recommend optimal ways the diverse personality identities can interact with each other to utilize the full potential that each has to offer. For example, an ENFJ16 personality- identity can best react to an ISTP17 personality-identity to achieve maximum productivity if they have the right tools and knowledge and ultimately use that knowledge to most effectively communicate with the differing personality. Similarly this research acknowledged those differences between African-Americans and Whites (which is analogous to ethnic identification and its intensity) and aims to contribute to ways help shrink the digital divide. Understanding that the intent to adopt technology is the key to the digital divide, this dissertation aims to study ethnic identity and its intensity as a moderator to adoption and use of technology.

15 Myers-Briggs - Myers-Briggs Type Indicator (MBTI) is a personality assestment designed to identify certain psychological differences according to the typological theories developed by Carl Gustav Jung in 1921. 16 ENFJ – Based on the MBTI, represents an individual with Extroverted Intuitive Feeling Judging personality type. 17 ISTP – Based on the MBTI, represents and individual with Intoverted Sensing Thinking Perceiving personality type.

Race vs. Ethnicity / Racial-Identity versus Ethnic-Identity

The previous section briefly surveys general motivational difference between races.

Now, I will briefly distinguish between race and ethnicity and why ethnic-identity was chosen over racial-identity as a moderator in this dissertation research study.

Race is defined as a phenotypical concept that refers to the gene frequencies in a population. Races are differentiated by inherited characteristics; as a result, racial categories are often defined based on physical appearance (Cross, 1991). Race itself is not an empirically derived concept; on the contrary, it is socially constructed. Furthermore, there is no genetic basis for mutually exclusive racial categories, simply because features and characteristics are distributed along a continuous distribution (1991). Moreover, there is actually more variation within a group than between groups. Race is essentially a Post Hoc grouping criterion where groups are defined based on sociocultural conventions and classification criteria and selected based on their usefulness in discriminating between the socially constructed categories. Additionally, an individual's racial identity can change over time, and self-ascribed race can differ from assigned race (Kressin et al. 2003).

According to Gardym (2001), race has always been and will always be a tool to justify social and political inequity and has little scientific value in research where the goal is to predict variability. If there is as much variability within racial groups as between groups, then there is little predictive value in measuring race itself. Therefore a tool is needed which actually predicts behavior between groups: Ethnicity.

Ethnicity is defined as a sense of being different from other groups because of cultural tradition, ancestry, national origin, history, or religion (Cross, 1991). In other words, ethnicity is defined by a sense of shared experience and “peoplehood”, defined by a unique socio-cultural heritage that is transmitted across generations. While it is similar to culture, it is more context specific (e.g. there are many ethnicities contained within a

Culture) (Gardyn, 2001). As described, ethnicity has the potential to be a better predictor of human behavior (than race).

Ethnicity itself as a measure or predictor, however, does posses it problems:

 viewing ethnicity as mutually exclusive categories suffers the same problem as race

 it may not be very descriptive of all group members (such as Whites)

 it may not be as useful for prediction as other considerations

 need to find aspects that can be measured on a continuum rather than in categories

(Cross, 1991).

These issues segue to the reason ethnic-identity was selected as opposed to the other considerations mentioned above. Ethnic Identity is defined as a subjective sense of ethnic group membership held by group members. It is to what degree (intensity) ethnic identity is part of self in that it incorporates:

 Sense of belonging

 Positive evaluation of the group

 Preference for group membership

 Ethnic interest and knowledge

 Involvement with group activities (Cross, 1991)

The process one undergoes toward ethnic-identity development:

1. ethnicity taken for granted (based on what others/society thinks)

2. Exploration: Investigate meaning and implications of group membership

3. Achieved ethnic identity: Fully integrated view of ethnicity with self view. (Note:

This is not a static point of development as ethnic identity may always remain in

flux)

Williams and Cross (1991) describes the ethnic-identity for African-Americans as a four-step process:

1. Pre-Encounter – African-American identity is devalued; belief in White Superiority

2. Encounter - experience discrimination and begin to value African-American identity

3. Immersion/Emersion-

a. Immersion - investigate African-American identity; complete rejection of

White culture (Radicalization / Militancy)

b. Emersion- begin to accept white culture

4. Internalization- achieve a self confident and secure African-American identity and

can learn from other cultures.

Another 2005 study’s results indicate that for African-Americans racial-ethnic identity is a more important component of self-concept than it is for multiracial individuals and whites (both of whom say they place little importance on it) (Rivas, 2004). Other findings show that the importance of racial-ethnic identity varies across settings (e.g., it is most important for African-Americans at work and least important at home) (Rivas, 2004).

Consequently, for this research, it is hypothesized that ethnic-identity will yield the most relevant and transferable data, results and conclusions and thus serve as a moderator.

Ethnic-identity is an essential component of the self-concept; similar to other aspects of identity, it can be particularly salient during adolescence and thus the reasoning for utilizing college students in this dissertation study. Most previous research on ethnic-identity focused on the unique elements that distinguish particular ethnic groups; however, it is important to study and compare ethnic identity and its correlates across groups (Phinney,

1992). The relationship of ethnic identity to various demographic variables and to self- esteem was examined in his study as well, and thus this current dissertation will further extend this concept to attempt to identify ethnic identity toward a technology adoption correlation.

Ethnic-Identity as a function of Social Identity and its Correlation to Self- Efficacy

The last section demonstrates the differences between race and ethnicity, as well as racial-identity versus ethnic-identity. As this dissertation research approach utilizing ethnic- identity as a moderator of technology acceptance is the first of its kind, there is currently extremely little literature on ethnic-identity and technology and its acceptance. As a result, as ethnic-identity is a function of social-identity, social-identity was then utilized as a predictor of ethnic identity to assist in drawing the initial hypothesis for this research. Even using social-identity as a proxy of ethnic-identity, the literature is still very scarce, indicating the new, innovative nature of this research.

This section, through the reviewed literature, further solidifies ethnic identity as a component of social-identity as a necessary component of this dissertation research study.

The relationship between social identity theory and goal setting has been well- explored (Britt, 1991). In his study, two hypotheses were being tested:

1. Subjects categorized into a group would rate their particular group favorably on a

number of dimensions, including their ability to perform a task.

2. Subjects would then attribute these characteristics to themselves and have higher

expectations, goals, and performance on the task than subjects not categorized into a

group

Two experiments were conducted:

1. Experiment One, in which subjects were either categorized or not categorized into a

group, then rated their self-efficacy regarding a particular task, set a goal for

themselves, and performed the task.

2. Experiment Two, identical to experiment one, but differed only in that subjects were

divided into naturally occurring groups instead of categorizing subjects into groups.

Results of the study yielded that although subjects wanted their group to do well on the task in either experience, they did not show in-group favoritism. Further analysis of the data yielded subjects who strongly identified with their group felt more efficacious when they listened to the in-group, than when they listened to the out-group (Britt, 1991).

Harris (2003) conducted a study which examined the relationship between public employee racial identity status and contextual performance indicators defined as altruism, conscientiousness, organizational commitment, and self-efficacy. U.S. Census data indicates that African-Americans tend to be over-represented in jobs with low to moderate levels of power and authority (Harris, 2003). As a result, his particular study contends that race membership may lack measurement sensitivity to pick up contextual performance group

44 differences; therefore, use of ethnic-identity status is a better, culturally competent measure.

Previous literature and research suggests that organizational advancement and rewards are influenced by contextual performance; additionally, supervisory performance ratings of subordinates are more reliant on contextual rather than objective factors. Non-task performance is difficult to measure and rely on single rater assessment opportunities.

Consequently, there is a bias to affect worker ratings. Some performance studies have associated achievement/performance gaps or perceived differences between African-

Americans and Whites with racial bias; therefore, given the limited amount of research in this area, the relationship cannot be supported or refuted. In examining its role, racial self- identity, was measured using Helms' (1990) Black Racial Identity Attitude Scale (BRIAS) and White Racial Identity Attitude Scale (WRIAS). Contextual performance indicators were self-efficacy, altruism, conscientiousness and organizational commitment. Black Racial

Identity Attitude Scale Conformity/Immersion-Emersion, White Racial Identity Attitude

Scale Disintegration/Reintegration and Combined Racial Identity Measures’ scores were found to explain more variation in contextual performance than race membership alone

(Harris, 2003). Furthermore, conformity, reintegration and Combined Racial Identity

Measures were negative predictors of self-efficacy. Other results of the study included:

1. Immersion-Emersion and Disintegration were positive predictors of self-efficacy;

2. Contact were positive predictor of conscientiousness

3. Combined Racial Identity Scores were positive predictors of altruism (Harris, 2003).

Ethnic-Identity as a function of Social Identity and its Correlation to Technology

The 2006 study by Jaeki & Yong Jin on social influences in the acceptance of technological services, investigates the effect of subjective norms, tendency to social comparison, and social identity on behavioral intention to use technological services. The use of various technological services was regarded as social behavior or a behavior affected by social factors. Jaeki & Yong Jin (2006) identifies a link between subjective norms and behavioral intention in the theory of reasoned action, social identity theory, and social comparison literature and unveils how social factors including subjective norms, social identity, and tendency to social comparison affect behavioral intention to use a specific technology. Additionally, it is essential for managers and academics to gain further understanding of the social nature of customer behavior, especially with regard to utilizing technological services, thus providing further insight for the development of technology driven e-commerce for various populations (Jaeki & Yong Jin, 2006).

Social influence on technology acceptance behavior has been acknowledged but needs to be further articulated and researched (Younghwa, Jintae, & Zoonky, 2006). While

Subjective Norm, as presented in the above article, has been dominantly used to capture the essence of social influence on technology adoption, to date, the findings have led some researchers to question its holistic validity as to whether or not subjective norm fully captures the extent of social influence on technology adoption. Social psychologists have recently examined self-identity as a construct reflecting social influence on behavior. It (self-identity) has been shown to have significant influence on voluntary behavior and have enduring effects particularly in situations where the subjective norm had little effect. In their recent

46 study, the effect of self-identity on technology acceptance decision in the context of a web- based class support system under the Technology Acceptance Model is examined (2006).

The result of the study demonstrates that there exists significant direct and indirect effect of self-identity (or ethnic-identity for our purposes) on technology acceptance (Younghwa et al.,

2006). The results also yield that self-identity (social-identity) has significant and direct effects on the technology acceptance in both voluntary (hedonic) and experienced

(utilitarian) situations, while subjective norm showed no significant effect in both situations

(Younghwa et al., 2006).

As stated earlier, social psychological theories of social identification and social representation are abundantly applied in a number of arenas, including within the organizational context. This includes, but is not limited to the shared values and beliefs identified as underpinning organizational cultures as well as the fact that social identity theory suggests people will more readily accept as valid ideas from members of their own in- group. Additionally, social representation theory itself addresses how beliefs are negotiated through conversation and interaction, and how they are used as explanation in social life. As a result, these two theories may clarify general understanding of how underlying beliefs and assumptions come to be shared within various environments (Hayes, 1991). Of great importance for our research, theories of organizational culture also identify subcultures and counter-cultures within the organization. Moreover, social identity theory may be useful in explaining how individuals, through membership of smaller groups, come to identify with, or distance themselves from, the larger culture; further implying that the working group (sub- group similar to this dissertation study of sub-culture) becomes an appropriate unit of analysis (Hayes, 1991).

Some research and attention has been given to social-identity issues and technology within an organizational setting. Recently, a great amount of information technology research suggests that increased user involvement and influence on information technology development can overcome user versus developer differences (for the purposes of this research, late adopters versus early adopters and innovators from the Technology Adoption

Life Cycle reviewed in Chapter 1) and thereby increase user acceptance (Gefen & Ridings,

2003). However, realization of such an objective liaisons possible problematic outcomes when implementing new information technology within a large cultural setting such as within organizations wherein such user involvement is impractical. As such Gefen & Ridings

(2003) examined information technology in view of Social Identity Theory, which again, focuses on how identification with groups affects individuals' beliefs and behavior, and for the purposes of our research, technology adoption. The results of the study yielded users' acceptance of information technology increased when users believed that the inter-group boundary between them and the developers was reduced as well as when users believed that they shared values with the developers group. Consequently, both of these beliefs increased when developers were perceived to be responsive to user requests. The results further suggest that, even in large settings such as organizations where user influence on information technology development is limited, developers’ (e.g. innovators and early adopters) responsiveness can increase user (later adopters’ / laggards’) acceptance of new information technology (Gefen & Ridings, 2003).

In Accommodating Technological Innovation, Bernstein (2004), explains how social tensions created by recent technological innovations are investigated. More specifically, one of the innovations exerting significant influence on peoples' lives, the Internet, and its impact

48 on the normative conception of identity is examined. Similar to this dissertation study,

(Bernstein, 2004) utilizes a socially-oriented approach; as a result, his study suggests direct incorporation of an independent (social) identity interest as well as indirect incorporation through the readjustment of currently existing and future literature (Bernstein, 2004).

The growth and increasing sophistication of information technology has led to the increasing importance of information accessibility in various social environments. Following this, within the workplace environment, pervasiveness of the resultant knowledge-based economy has centered attention on issues of employee group (social) identity. As a result, employee perceptions of group membership and how they guide the change outcomes of organizations implementing new information technology is explored by Gavin (2005). Using social identity theory as a framework, the salient inter-group relationships of two groups of employees, management and information technology implementation teams, and how these employees use their different group memberships to reframe positions of authority or knowledge around technology change, is investigated as well as the extent to which perceptions of social identity legitimate institutional structures already in place despite the potential of new technology (Gavin, 2005).

For our purposes, not only is social-identification a factor, but the strength or intensity (as mentioned earlier) is a major factor and a large part of this dissertation study

(Deshponde, Hoyer, & Donthu, 1986). In 1995, Houston (1995) conducted an experiment on this very aspect as well. One perspective of business and marketing strategy formulation and implementation views the organization as an open social system wherein multiple coalitions engage in political action to guide an organization’s direction. As such, group and individual goals are tempered by recognition of a mutual dependence on the achievement of broader

49 organization (system) goals. Houston's (1995) study conceptualized strategy development and implementation processes as emerging from a context of interaction patterns within a network of system actors who possess unique, potentially incongruent mental models regarding strategic goals. The study, itself, drew upon social identity theory and a cognitive perspective of organizational culture to examine how an individual's strength of identification with a sub-group affects organizational social interaction patterns and members' perspectives regarding strategy. “The strength of identification with an organizational sub-group can directly influence the convergence of beliefs within and the divergence of beliefs between interacting sub-groups” (Houston, 1995). The study examined the beliefs of senior executives connected by a web of cross-group, cross-level internal working relationships activated by a major strategic initiative an organization. Multiple methods were used to compare the beliefs of the twenty-nine "key players" in a particular initiative regarding competitive advantage, the strategic significance of such initiative, and critical issues surrounding the launch of this new initiative. The beliefs were linked to a social network analysis and revealed information rich and compelling portraits of the diverse cognitive processes that created inter-group barriers to strategy processes (Houston, 1995).

Participants who identified more strongly (versus weakly) with the new technology initiative:

 mentioned more positive beliefs

 were less critical toward the initiative team when articulating negative beliefs,

 were highly integrated into the social network surrounding the technology

initiative

 were more likely to share positive beliefs concerning the venture.

Respondents with a weaker level of identification:

 were more likely to be negative and critical toward the new technology initiative

 were not highly integrated into the social network,

 were more likely to share negative beliefs concerning the venture.

The technology initiative was not effectively linked to existing organizational systems and no effective champion emerged to build a shared sense of vision for the new technology initiative between the groups within the organization as a whole (Houston, 1995).

In addition to examining the historical development of e-mail, the research by Kilker

(1999) contributes to the emerging and continuing social history of information technology.

His case study examined the role of social identifications in discussions about early networked electronic mail. Furthermore, the overall goal was to extend the social construction of technology model using concepts from social identity theory. Social identifications were found to be present in the heterogeneous group's discussions at all levels including the institutional, group, and individual levels, as well as in members' constructions of e-mail users through identity bootstrapping (which attempts, to distinguish between expert and user identities). Additionally, identifications also appeared to influence evaluations of e- mail systems and of the discussions (Kilker, 1999). Interestingly, despite stereotypes of computer developers as insensitive to social issues, members of Kilker ’s study group repeatedly reflected on the broader social implications of e-mail and on specific impacts of it on the group's communication (1999). Social and technical closure appeared to be inhibited due, largely, to the difficulty of creating a super ordinate group identity. Overall, however, both groups (users and developers) appeared to indirectly influence both e-mail technology and the social process of Internet standards development. As a result, using the social construction of technology model with concepts from social identity theory highlights the

51 potential influence of individual and group identities in technology design, and goes further to suggest even more realistic expectations for collaboration within heterogeneous technical groups (Kilker, 1999).

Further supporting the social-identity influences on technology adoption and diffusion, the uses and gratifications theory was used to explain why young people use cellular telephones to satisfy their social and psychological needs in a study by Kondo

(2003). Analysis of the data identified dimensions of connectedness and security, bonding, fashion/status and, most importantly for this dissertation research, social identity as information sought as significant predictors of cellular phone expectations. Gender and students' academic status also proved to be significant factors. From a social standpoint, cellular phones were seen as an important medium for maintaining relationships with peers.

The implications of the study suggest that cellular phones provide for more than just communication, they fulfill socialization desires and help set the social status of young people (Kondo, 2003).

Although culture has recently been recognized as one factor in interface design, computer science and engineering practices and doctrines are generally thought to be culturally neutral. Matti, Erkki, Esko, & Piet (2006) recommended and presented an approach which recognizes society and culture in computational concepts and applications and as a necessary component.

The effects of social identity can even be seen within a top-down structure. Porter

(1997) explored the effect of managers' race and gender on their attributions and conceptions of subordinate commitment to their organization. Results indicated that:

1. managers experience in-group favoritism based on race, but not on sex

2. all managers hold stereotypes which influence the level of commitment they

attribute to an actor based on the actor's behavior, the actor's race and the actor's

sex (Porter, 1997).

Social identity theory predicts that in most cases self-enhancement will have a significant effect on identity salience. However, less consistent results were obtained for the explanations of identity salience via a demographic perspective (Randel, 1999). According to

Randel (1999) more research is needed in this area.

Little research has been directed towards exploring initial trust formation in a computer-mediated social action network. As a result, further research is required to examine the impact of information and communication technology on social identity, more specifically, trust formation in contextually diverse online communities (Ryan, 2003).

In his study, Scott (1997) integrated theories of group effectiveness with social identity and social categorization theory and found that team social identification was an important predictor of team performance. This outcome also explained the variance by cohesiveness and external communication. Additionally, team social identification itself was influenced by support and recognition from top management, project leader organizational influence, the extent to which members' time was dedicated to the team, and team functional diversity (Scott, 1997).

Similar patterns are also noted across age groups. In 2002, a case study of the adoption and use of video mail was conducted by Seede (2002). As a computer-mediated communications technology, video-mail was selected due to its unique characteristics

(recorded synchronously, distributed asynchronously, conveys multiple simultaneous cues with persistent messages, and requires special equipment to compose but not to view) that

53 affect its adoption, diffusion and use. As a result, a number of disciplines have offered theories that may be used to explain video-mail use and adoption. These theories include:

 Communications Theories:

o cues-filtered-out

o Information Richness,

o Social Presence

o Social Influence

o Social Information Processing

 Mass Media Theories (limited effects approaches):

o Uses, Gratifications, and Dependency Theory (UGD)

o Expectancy Value Theory

o Dependency Theory

 Social Psychology Theories:

o Social-Identity De-Individuation Theory

o Flow Theory (Seede, 2002)

Many of the results were consistent with theoretically expected behavior suggesting that users respond in-kind to video-mail they receive, which is also consistent with social influence theory. Additionally, flow theory successfully predicted significant early use of video-mail due to its novelty and high degree of message control while cues-filtered-out correctly predicted a high degree of social presentation within the video-mail; furthermore, dependency theory proved consistent with video-mail viewing behavior. On the other hand, social presence and information richness theories did indeed accurately predict a high degree of socio-emotional activity. While a modest level of hyper-personal communications

54 occurred as anticipated by social psychological theories, there was no evidence to support the contention that higher status individuals would be more likely to adopt video-mail than lower status ones. Furthermore, the relatively high level of personal presence of video-mail was perceived as having greater social risk according to the study. Video-mail composers used affinity, familiarity and proximity as a selection criterion for recipients to avoid being judged for less than perfect communications and required uninterrupted private time to compose messages, as they were concerned about their recipient's ability to receive messages and respond in-kind. The final finding of the study yielded that technical proficiency and training were correlated with high levels of use as well as low failure rates (Seede, 2002).

The implications of identity fit (defined as the degree to which individuals know, and are able to identify, those identities that are most important to the focal individual) for diverse work teams is explored in view of how increased levels of diversity and more frequent use of teams in the workplace affects individuals and their interactions within groups. Previous research on diversity and teams has focused on linking demographic differences to process and outcome variables but has neglected to measure the psychological influences that are presumed to underlie the effects of diversity and group processes.

Therefore the identity fit construct was developed as a means of understanding how the psychological mechanisms inherent in group work and the situational effects defined by the team and organizational context interact to determine individual outcomes as well as examine the implications of identity fit and its antecedents (e.g., diversity context) on group process and individual performance (Thatcher, 2000). Building on a plethora of theories including social identity, categorization, self-verification, and person environment fit, results of the study yield that high levels of identity fit have significant and positive effects on creative

55 performance whereas low levels of identity fit (what the authors referred to as “identity misfit”) lead to high levels of absenteeism and team-related stress. As was expected, individuals with many demographic differences relative to team members had low levels of identity fit (Thatcher, 2000). Contradictory to expectations derived from previous research, individuals who communicated with team members on a face-to-face basis had lower levels of identity fit than individuals who communicated with team members using e-mail or the telephone. However, the results support previous research on diversity in that it often presumed demographic differences cause negative outcomes through a mechanism of social identity. An important contribution of Thatcher's (2000) study suggests that if individuals in teams focus on identity knowledge and acceptance some of the negative effects attributed to demographic differences may be reduced. As a side note, industry implications suggest that managers must understand the consequences of using various forms of technology in their organizations since communication medium may influence identity fit which, in turn, affects creativity, absenteeism, and team stress (Thatcher, 2000).

Technology Acceptance Model

As its base resides within psychological research, the Technology Acceptance Model

(TAM) (Davis et. al. 1989) is an adaptation of the Theory of Reasoned Action (TRA)

(Fishbein et. al. 1975) to the information technology realm. TAM posits that its two main independent constructs (1. perceived usefulness and 2. perceived ease of use) both determine the primary dependent construct of an individual's intention to use a system with intention to use serving as a mediator of actual system use (York University, 2006). Research has also

56 demonstrated that perceived usefulness is directly impacted by perceived ease of use in this model (Wang, 1998) (however, for the purposes of this dissertation, the two independent constructs will not be tested with this relationship in consideration).

Some research has simplified the TAM by removing the attitude construct found in the TRA (Venkatesh et. al., 2003). Additionally, other research has attempted to extend the

TAM, generally by:

1. introducing factors from related models,

2. introducing additional or alternative belief factors, or

3. examining antecedents and moderators of perceived usefulness and perceived ease of

use (Wixom and Todd, 2005) as this dissertation research aims to do.

Both the TRA and the TAM, have strong behavioral elements and assume that when someone forms an intention to act or use technology, that they will be free to act or use without limitation. However, in practice, constraints such as limited ability, time, environmental or organizational limits, as well as unconscious habits also limit the freedom to act or use (York University, 2006). One limitation of TAM is that it assumes usage is volitional, that is, there are no barriers that would prevent an individual from using a particular technology if he or she chose to do so. As such, one study extends TAM by adding perceived user resources to the model, with careful attention to placing the construct in

TAM's existing nomological18 structure (Mathieson, Peacock, & Chin, 2001).

18 Nomological- relating to or expressing basic physical laws or rules of reasoning (such as logical laws).

Figure 2: The Technology Acceptance Model

TAM, Antecedents and Extensions

Numerous authors acknowledge the importance of understanding what motivates individuals toward acceptance of technology and that unless there is a clearer understanding of what motivates individuals to use technology, there is a greater likelihood that technology will continue to be under-used (i.e. the Digital Divide as a function of Use as opposed to access, socio economic status, and other demographic variables) (Griffin, 2006; Guus &

Kees van, 2005; Wang, 1998). In order for individuals to successfully and effectively adopt and use technology, understanding the role, antecedents and intentions of its acceptance is essential (Guus & Kees van, 2005).

In their study over computer self efficacy, Compeau et al (1999) used longitudinal data to develop a model based on Bandura's (1977, 1986) Social Cognitive Theory.

Specifically, the study tested the influence of computer self-efficacy, outcome expectations, affect and anxiety on computer usage. Compeau et al found significant relationships between computer self-efficacy and outcome expectations as well as between self-efficacy and affect and anxiety and use. Overall, the findings provided strong confirmation that both self-

58 efficacy and outcome expectations impact on an individual's affective and behavioral reactions to information technology (Compeau, Higgins, & Huff, 1999).

In a study using the TAM’s behavioral constructs of perceived usefulness and perceived ease of use in testing for predicting user acceptance of the World Wide Web, the results yielded that the model was not only strengthened but validated by the inclusion of an additional construct, computer self-efficacy (Fenech, 1998).

Similarly, more individuals than ever are using technology, in this particular study, surfing the Internet, to fulfill a variety of social, personal and business needs. As a result,

Glassberg (2000) investigates if:

1. specific factors be identified which motivate individuals to use technology (in their

case the Web) in the multiple domains of their life

2. a theoretically rich model be developed to predict acceptance behavior of new multi-

functional technologies with greater accuracy than existing models (e.g. TAM)

Consequently, his research yielded the development of the Technology Acceptance Model for the World Wide Web (TAMWWW) (Glassberg, 2000).

Igbaria and Iivari (1995) introduces an extended TAM is introduced which explicitly incorporates self-efficacy and its determinants (experience and organizational support) as factors affecting computer anxiety, perceived ease of use, perceived usefulness and the use of computer technology. Consistent with the original TAM, perceived usefulness has a strong direct effect on usage, while perceived ease of use had indirect effect on usage through perceived usefulness. The results of the study yielded that self-efficacy had both direct and indirect effects on usage, further demonstrating its importance in the decision to use computer technology. Additionally, self-efficacy had a strong, direct effect on perceived ease

59 of use, but only an indirect effect on perceived usefulness via perceived ease of use (Igbaria

& Iivari, 1995).

Of particular importance to this dissertation research study is the particularization of computer-efficacy and its influence on the technology acceptance model. As previously discussed, there is resistance in using information technology despite potentially tremendous benefits. As investments in IT continue to grow exponentially and have increasingly become more of a competitive necessity, practitioners and researchers alike are increasingly interested in understanding the circumstances surrounding such resistance. Researchers generally cite the technology acceptance model (TAM) as one of the most influential computer-usage models and while the TAM has empirically demonstrated its validity, the model accounts for only a fraction of system usage variance. Other researchers suggest that the TAM constructs may be too general (McFarland, 1999). By simultaneously considering diverse attributes, individuals become confused regarding what is being measured and how much weight should be given to a particular attribute. Other researchers suggest that by not explicitly considering the influences of external variables, the results of the TAM may be spurious or misleading. As a result, research suggests that additional IT-specific variables should be incorporated into the TAM (McFarland, 1999). Among those additional variables that may provide added insight into human computing behaviors include self-efficacy and the characteristics of characteristics of the technical task at hand. Although prior studies have defined computer-efficacy measures, they do not specifically consider the technical task nor the particular situational circumstances; therefore, based on self-efficacy theory, these general measures of efficacy are fundamentally inaccurate and vague (McFarland, 1999).

Empirical studies investigating computer-efficacy have been able to explain very little of the

60 computer-efficacy variance. McFarland's (1999) study not only developed a new particularized computer-efficacy measure which incorporated IT but also those IT-specific antecedents believed to influence IT into the TAM; therefore, the findings of his study provide strong support for the particularization of IT behavioral research as well as social cognitive theory, in that the contextual variables directly affect system usage (McFarland,

1999).

Although information technology has played an increasingly important role in contemporary education, resistance to IT remains significant. Min, Yan, & Yuecheng's

(2004) study aimed to identify additional key determinants of IT acceptance. The current technology acceptance model (TAM) and the social cognitive theory (SCT) are combined to provide a new framework for this analysis. Results of the study were consistent with the

TAM factors for explaining behavioral intention (reviewed earlier). The study also indicated that computer self-efficacy (CSE) has substantial influence on the technology acceptance

(Min et al., 2004).

Similarly, based on an augmented TAM, Chau (2001) examined the influence of computer attitude and self-efficacy on IT usage behavior. Computer attitude and self- efficacy were explicitly incorporated in the research model as external variables affecting perceived usefulness and perceived ease of use, the two key factors influencing the IT usage behavior according to the original TAM. The results yield that computer attitude has a significant and positive effect on perceived usefulness and perceived ease of use. Computer self-efficacy, on the other hand, has a relatively small, but negative, effect on perceived usefulness and no significant effect on perceived ease of use (Chau, 2001).

An integrative model explaining intentions to use an information technology yielded results revealing strong influences of both personal innovativeness and computer self- efficacy (Thompson, Compeau, & Higgins, 2006).

Several research studies have developed technology acceptance models to determine the influences on the adoption of various technologies. However, and most importantly for our research purposes, limited research has explored the influence of electronic business technologies in a business-to-business environment (Sousa, 2003). His study, like this dissertation, expands current research to develop a model which identifies the factors having a significant influence on the users' perception to adopt technology, within a business purchasing environment. The model developed by his study, Electronic Business

Technology Acceptance Model (EbTAM), identifies and explains the influences on the use of electronic business technology (Sousa, 2003). The additional constructs, which expand the original technology acceptance models, provide substantial evidence on the attitudes supporting electronic business technology adoption. The results of the study support the findings of previous studies which assert the influence of subjective norms, self-efficacy, web skills, perceived ease of use and perceived usefulness on the behavioral intent to use technology. The external factors associated with the competitive and customer environment illustrated a significant influence on the user's behavior. Sousa's (2003) study concludes that as the customer orientation increases within the organization, the user's reliance on electronic business systems decreases suggesting that these systems do not replace the need for traditional marketing tasks. The results of this study’s EbTAM model do not support the influence of the system's ease of use as a predictor on the intention to adopt technology. This

62 suggests, and most significant for our research, that the perception of whether these systems are useful is the dominant factor leading to adoption (Sousa, 2003).

Research such as that of Tremblay (2000) examined the extent to which the level of acceptance or rejection of an IT project can be predicted on the basis of the psychosocial development profile of individuals whose perceived computer self-efficacy, attitude toward technology in general, and certain biographical variables have been statistically controlled.

The research study yielded that the level of acceptance or rejection of the technology project was statistically associated with specific characteristics of the individual’s psychosocial development profile (coefficient of correlation varying between 0.157 to 0.311), their perceived computer self-efficacy (0.230), their attitude toward technology in general (.350), as well as demographic independent variables such as their gender. The study’s multiple regression model which contained the third positive psychosocial development scale based on technical use initiative, the respondent’s gender, general demonstrated attitude, and perceived computer self-efficacy of the respondent showed the strongest results. However, these variables could explain only 17.8% of the variance in the response to the Internet project which, for the purposes of our research indicates that additional variables need to be included in this model and other technology acceptance models to better explain the level of acceptance or rejection of the technical projects (Tremblay, 2000). Thus, partial support for the positive correlation between ethnic-identity (EI) and the mediators.

User acceptance of information technologies is an important issue since productivity gains associated with technology use can be achieved only if technology is not only accepted but used by target, and possibly other, users (Venkatesh, 1998). Going further to identify and

63 address five important issues related to user acceptance, he utilized the following key research inquires:

1. How do the different models of user acceptance (e.g., Technology Acceptance

Model, Decomposed Theory of Planned Behavior, Innovation Diffusion Theory

and a Model of PC Utilization) compare with each other in terms of predictive

validity? He expected that the TAM, which comprises of perceived usefulness

and perceived ease of use, to be the most powerful model predicting acceptance.

2. What are the determinants of perceived usefulness? He hypothesized that

subjective norm and innovation characteristics to be the key determinants of

perceived usefulness.

3. What are the determinants of perceived ease of use? He expected perceived

behavioral control, computer anxiety, enjoyment, computer self-efficacy and

objective usability to be the critical determinants of perceived ease of use.

4. Does behavioral intention indeed predict actual usage behavior or does it only

predict self-reported usage behavior? He expected usage behavior (measured in

terms of duration of use) to predict behavioral intention most effectively.

5. Are there any additional predictors of behavior? Typically, behavioral intention to

use is the key predictor of usage behavior. The role of habit as an additional

predictor of behavior was examined.

Results yielded strong support for the models proposed in his research (Venkatesh, 1998).

The TAM has been widely used to predict user acceptance and use based on perceived ease of use and usefulness. However, in order to design effective training interventions to improve user acceptance, it is necessary to better understand the antecedents

64 and determinants of key acceptance constructs. Another study by Venkatesh & Davis (1996) focused on understanding the determinants of perceived ease of use. The results supported the hypothesis that an individual's perception of a particular system's ease of use is anchored to her general computer self-efficacy at all times, and objective usability has an impact on ease of use perceptions about a specific system only after direct experience with the system

(Venkatesh & Davis, 1996).

Extended TAM Model

Researchers suggest that the TAM constructs may be too general (McFarland, 1999).

It is essential to understand what influences individuals’ acceptance or rejection of new information technology (Wang, 1998). Recently, , advances in wireless technology have increased the number of people using mobile devices and accelerated the rapid development of mobile service conducted with these devices. However, although many companies today are making considerable investments to take advantage of the new business possibilities offered by wireless technology, research on mobile commerce suggests potential consumers may not adopt these mobile services and technology despite their availability. As a result, there is a need for research to identify the factors that affect consumer intention to use mobile technology (Yi-Shun, Hsin-Hui, & Pin, 2006) as the aim of this dissertation research.

Towards this, this dissertation extends the original TAM leftward to include ‘trait’ and ‘state’ efficacy as antecedents of ‘perceived ease of use.’ For the usefulness construct

‘symbolic’ and ‘functional’ utilities will replace ‘perceived usefulness’ as I am proposing that these two components combine to form usefulness. Furthermore, this dissertation research introduces the variable of ‘ethnic identity’ and its intensity as a moderator to all the

65 antecedent variables of ‘trait’ and ‘state’ efficacies as well as ‘symbolic’ and ‘functional’ utilities to their respective exogenous variables. This model is diagramed below. Following the diagram, the hypotheses for this dissertation research will be explored.

Trait Efficacy

Ease of Use

State Efficacy Intent to Use/Master Ethnic Identity

Symbolic Utility

Usefulness Usefulness

Functional Utility

Figure 2: Proposed TAM Extension

Research Questions and Hypotheses

Research Question 1 – Is there a relationship between one’s perceived ease of use as well as their combined functional and symbolic utilities to their intention to adopt and use new, innovative technology?

H1(a): There exists a positive relationship between one’s perceived ease of use to

technology and their likelihood to adopt new innovative technology.

H1(b): There exists a positive relationship between one’s symbolic utility to their

likelihood to adopt.

H1(c): There exists a positive relationship between one’s functional utility to their

likelihood to adopt.

Research question 2: Is there a relationship between one’s combined trait and state efficacies to their perceived ease of use of new innovative technology.

H2(a): There exists a positive relationship between one’s trait and state efficacy to

their perceived ease of use.

H2(b): There exists a positive relationship between one’s state efficacy to their

perceived ease of use.

Research Question 3: To what degree does one’s ethnic identity intensity affect their perceived ease of use and intention to adopt and use new, innovative technology?

H3(a): The greater the intensity of ethnic identity, the stronger relationship between

trait efficacy and perceived ease of use.

H3(b): The greater the intensity of ethnic identity, the stronger relationship between

state efficacy and perceived ease of use

H3(c): The greater the intensity of ethnic identity, the stronger relationship between

symbolic utility and intent to adopt / use.

H3(d): The greater the intensity of ethnic identity, the stronger relationship between

functional utility and intent to adopt / use.

Classic TAM

The first sets of hypotheses are supported by the classic TAM literature as well as validating research experiments. As a result, for this dissertation research, it is hypothesized:

H1(a): There exists a positive relationship between one’s perceived ease of use to

technology and their likelihood to adopt new innovative technology.

Symbolic and Functional Utilities

The symbolic / social value of technology is largely under-studied in the IT realm as numerous authors identify the lack of existing research and literature on the symbolic / social value of technology (Baskerville, DeGross & Stage 2000; Gumm 2001; Barrantes 2006).

Researchers also identify that social influence from peers strongly affects individual's attitude and behavioral intention to adopt a new information technology (Wang, 1998).

Several theoretical models from technology and social (or socio-technical) psychology are available to support the implementation of innovations. Research in this realm suggests that the most significant finding among such research is the key roles perceived fun and perceived enjoyment play as external variables in influencing technology beliefs, attitude, efficacies and usage. Thereby, emphasizing the entertainment (symbolic) value of technology, computer anxiety of individuals can be reduced and diminished while improving computer self-efficacy (Guus & Kees van, 2005).

For increased predictive power, the TAM has often been augmented to include additionally antecedents and predictors of technology acceptance. Leo (2004) extends the model to include social influence (attitude) of consumer intention to use (on-line) technology. Besides ease of use and usefulness, compatibility, privacy, security, normative beliefs, and self- efficacy are included in an augmented TAM. The intention to use (on-line) technology was strongly influenced by attitude toward (on-line) technology, normative beliefs, and self-efficacy (Leo, 2004).

Sanna (2005) introduced a social constructionist approach to information technology literacy. It is introduced via the concept of “IT self” as a description of the momentary, context-dependent, and multilayered nature of interpretations of IT competencies (Sanna,

2005). This description asserts descriptions of IT competencies and computer-related attitudes are dialogic social constructs and closely tied with more general implicit understandings of the nature of technical artifacts and technical knowledge. These and like implicit theories, applications and assumptions are rarely taken under scrutiny in discussions of IT literacy yet they have profound implications for the aims and methods in technology adoption (Sanna, 2005).

As the TAM demonstrates, individuals’ perceptions are what drive technology adoption. To further support this, Wang (1998) conducts a study where four protocols were developed to enhance potential users' perception of new IT in user pre-training technology introduction. Towards this a 2 x 2 framework was developed to classify these protocols as

Technology Introduction Protocols (TIP) along two dimensions:

1. the 'focus of introduction' dimension – which centers on "what approaches do we use

to deliver the information regarding the new technology" utilizing featured-focused

protocols and application-focused protocols

2. the 'source of influence' dimension – which concentrates on "who delivers the

information regarding the new technology" utilizing expert-centered protocols and

peer-supported protocols (Wang, 1998).

Results of the study indicated that the feature-focused approach (for the ‘focus of introduction’ dimension) and peer-supported approach (for the ‘source of influence’ dimension) generated the best results in subjects' computer self-efficacy, perceived usefulness, and perceived ease of use in most cases (Wang, 1998).

Conclusively, the perception of whether these systems are useful is the dominant factor leading to adoption (Sousa, 2003). Therefore, based on the above literature as well as the literature presented in the previous section and using subjective norm as a partial predictor of symbolic/social value of technology, combined with the utilitarian view of technology needed to complete primarily work, I derive the hypotheses:

H1(b): There exists a positive relationship between one’s symbolic utility to their

likelihood to adopt.

H1(c): There exists a positive relationship between one’s functional utility to their

likelihood to adopt.

Self-Efficacy

Self-efficacy is defined as one’s own belief in their capabilities of using a computer in the accomplishment of specific tasks, on computer usage (Igbaria & Iivari, 1995). Self- efficacy has a strong, direct effect on perceived ease of use, but only an indirect effect on perceived usefulness via perceived ease of use (1995). Computer self-efficacy has substantial influence on the technology acceptance (Min et al., 2004).

Those who have a higher level of computer self-efficacy may be more accepting of technology and may be more likely to use all its available features (Griffin, 2006).

Numerous authors suggest that measures of computer self-efficacy measures often based on self-assessments that measure interpretations of skills as opposed to performance in practice.

Additionally, numerous authors have attempted to enhance the TAM with constructs such as self-efficacy. Yi-Shun et al. (2006) re-specifies and validates an integrated model for predicting consumer intention to use technology by adding a trust-related construct of

'perceived credibility' (this trust element segues into the Symbolic/social utility of this augmented TAM presented by this research) and two resource-related constructs ('self- efficacy' and 'perceived financial resources') to the TAM's nomological structure and re- examining the relationships between the proposed constructs (Yi-Shun et al., 2006).

As mentioned Earlier, in his 1991 study, Britt’s analysis of research data yielded subjects who strongly identified with their group felt more efficacious when they listened to the in-group, than when they listened to the out-group (Britt, 1991).

Furthermore, overall, findings of another study provided strong confirmation that both self-efficacy and outcome expectations impact on an individual's affective and behavioral reactions to information technology (Compeau et al., 1999). Moreover, in a study

71 using the TAM’s behavioral constructs of perceived usefulness and perceived ease of use in testing for predicting user acceptance, the results yielded that the model was not only strengthened but validated by the inclusion of an additional construct, computer self-efficacy

(Fenech, 1998).

As evidenced in the previous section, self-efficacy and technology acceptance has been well studied. This proposed model extends upon other uses of self-efficacy or computer self-efficacies based on Eden’s (1982) breakdown of self-efficacy: trait (TE) and state (SE) efficacy19. As a result, it is hypothesized based on the reviewed literature that:

H2(a): There exists a positive relationship between one’s trait and state efficacy to

their perceived ease of use.

H2(b): There exists a positive relationship between one’s state efficacy to their

perceived ease of use.

This hypothesis is formed based on previous research studies demonstrating the positive relationship between CSE and perceived Ease of Use.

19 Trait and State Efficacy - based on the work of Dov Eden (1988), Self-efficacy is comprised of two components: trait efficacy (or generalized efficacy) and state efficacy (task-specific self-efficacy).

Ethnic-Identity

The primary question of this research: is there some reason the African-American experience is driving certain sub-groups of the population to the wrong side of the digital divide?

Building on a plethora of theories including social identity, categorization, self- verification, and person environment fit, results of the study yield that high levels of identity fit have significant and positive effects on creative performance (Thatcher, 2000).

In a 2005 study investigating the effects of the TAM with ethnicity as a mediator, computer self-efficacy and subjective norms were determinants of perceived usefulness and perceived ease of use, respectfully, which in turn determined the attitude of students using computer. Their analysis indicated that this particular research did not support TAM as is and suggested that perceived usefulness is the most significant predictor of perceived ease of use and thus the most important predictor of technology adoption (Sen, 2005). Their research goes further to suggest that additional research endeavors should be devoted to the measurement of technology use in various environments with verifying ethnic backgrounds to further analyze students' acceptance or rejection of technology. (Sen, 2005). Thus, the incorporation of ethnic identity in this model.

Using social identity as a proxy, social identity theory predicts that in most cases self- enhancement will have a significant effect on identity salience (Randel, 1999). Furthermore, in his study, Scott (1997) integrated theories of group effectiveness with social identity and social categorization theory and found that team social identification was an important predictor of team performance.

As mentioned earlier, results of a study yielded that self-identity (social-identity) has significant and direct effects on the technology acceptance in both voluntary (hedonic or symbolic for the purpose of this dissertation) and experienced (utilitarian our functional) situations, while subjective norm showed no significant effect in both situations (Younghwa et al., 2006).

Furthermore, shared values and beliefs are identified as underpinning characteristics of various cultures as well as the fact that social identity theory suggests people will more readily accept as valid ideas from members of their own in-group (Hayes, 1991).

Drawing from the literature and largely using social-identity as a predictor of the role of ethnic-identity, the following it hypothesized concerning the moderating relationship of ethnic-identity:

H3(a): The greater the intensity of ethnic identity, the stronger relationship between

trait efficacy and perceived ease of use.

H3(b): The greater the intensity of ethnic identity, the stronger relationship between

state efficacy and perceived ease of use

H3(c): The greater the intensity of ethnic identity, the stronger relationship between

symbolic utility and intent to adopt / use.

H3(d): The greater the intensity of ethnic identity, the stronger relationship between

functional utility and intent to adopt / use.

Further supporting these innovative hypotheses concerning ethnic-identity is reasoning that as one spends more time and has more experiences and devotes more mental capacity to not

74 only form but solidify their ethnic identity, the more valuable their info and symbolic utility as well as the higher their self-efficacies.

This statement serves as a research proposition that acknowledges the lack of directional guidance in the literature and thus this literature review; however, be reminded that TAM and its antecedents differ by ethnicity/race which for the purposes of this research is analogous to strong identifiers of ethnic affiliation versus weak identifies of ethnic affiliation; consequently, it is hypothesized that the extended TAM presented here will differ by strong identifiers to weak identifiers as well. In other words, it is expected that the intensity of ethnic identity affiliation will have a moderating effect on the rest of the proposed model.

Summary

This section discusses the racial differences in general motivation and then explains the differences between race and ethnicity and racial identity and ethnic identity. Following, extensive literature is provided to support the moderating affects of ethnic identity (as a component of social and self identity) hypotheses towards the mediating variables (trait and state efficacy as well as symbolic and informational utilities). The TAM is then briefly surveyed and the connection of the aforementioned mediating variables is drawn from the supporting literature. Finally, a new, extended TAM model is presented.

CHAPTER 3: METHODOLOGY

Overview

The goal of this dissertation is to study the motivation of minorities to adopt, diffuse and learn new, innovative technologies. Specifically, this study will examine the effect of ethnic identification as a moderator variable in this motivation. Toward this end, a web survey will be administered as a means to collect data from participants after they agree to participate in the study via an electronic consent form preceding the online survey.

Measures

Ethnic-Identity

Measures Considered for Ethnic Identity

To date, various methods have been developed and introduced to measure ethnic identity. Some of the measures that were reviewed to measure ethnic identity for this dissertation study include:

 Racial identity scale [RIAS] (1985)

Helms JE; Parham TA

 Racial identity scale short form [RIAS] social attitudes inventory (1985)

Helms JE; Parham TA

 Optimal theory applied to identity development [OTAID] instrument [1994]

Haggins KL; Myers LJ; Speight S; Highlen P; Cox C; Reynolds A

 Milliones’ developmental inventory of Black consciousness [DIBC] revision 2 (1985)

Taylor J; Brown AB; Denton SE

 African self consciousness scale [ASCS] (1984)

Kambon K; Baldwin JA

 Black nationalistic ideology scale [BIS] revised (1976)

Terrell F; Taylor J

 Perceived racism scale [PRS] (1996)

McNeilly MD; Anderson NB; Robinson EL; McManus CB; Armstead CA; Clark R;

Pieper CF; Simons CE; Saulter TD

 National survey of Black Americans self esteem index (1981)

Clark ML

 National survey of Black Americans racial stereotypes index (1981)

Clark ML

 National survey of Black Americans satisfaction index (1981)

Clark ML

 National survey of Black Americans family provider role expectancies (1995)

Bowman PJ

 Black male experiences measure [BMEM] (1985)

Cunningham M; Spencer MB.

 Scale of racial socialization [SORS A] adolescent version (1993)

Stevenson HC Jr.

 National study of Black college students [NSBCS] (1981)

Allen WR

In previous attempts to measure ethnic identity, many researchers have generally focused on specific groups and tried to identify and assess key components of ethnic identity within those ethnic groups; as a result such studies have included many aspects of ethnic identity (Phinney, 1992). The issue arises in that these components have varying importance for different groups. For example, political attitudes are important in measure of African-

American Identity, language is salient in Mexican-American measure, and cultural attitudes play a major role in Asian-American identity (Phinney, 1992). The focus of the measure chosen for this dissertation study is on ethnic identity as a general phenomenon because it is relevant across groups. While it is clear that each group has its unique history, traditions, and values, the concept of a group identity (a sense of identification with, or belonging to, one’s own group) is common to all human beings. Therefore, general aspects of ethnic identity can be examined by focusing on those components that are common across groups. To study and compare the role of ethnic identity in development, including its precursors, correlates and influences on behavior and attitudes towards technology, a measure of ethnic identity is needed that can be used with diverse populations (Phinney, 1992).

Multigroup Ethnic-Identity Measure

Although numerous methods were surveyed to measure ethnic identity within this study, the Multigroup Ethnic Identity Measure by Jean S. Phinney (1992) was chosen to measure the moderator variable of ethnic identity in this dissertation study. This measure was chosen because of its validity and ability to measure across different ethnicities while using the same measurement instrument survey. Phinney presents a questionnaire measure

78 of ethnic identity based on the elements of ethnic identity that are common across groups; as a result, it can be used with all ethnic groups. In his experiment, the questionnaire was administered to 417 high school students and 136 college students from ethnically diverse schools. The reliability of his measure was assessed by Cronbach’s alpha20 at .81 for the high school sample and at .90 for the college sample (Phinney, 1992). Phinney’s measure can also be used to examine similarities and differences in ethnic identity and its correlates among youths from different ethnic groups.

Ease of Use Construct

‘Ease of Use’ refers to the degree to which an individual believes that using a particular system would be free from physical and mental effort (Davis, 1989).

‘Ease of use’ will be measured via a three-item, six-point Likert21 scale. ‘ease of use’ questions were adapted from the original TAM perceived usefulness measure by Davis

(Davis, 1989). An even-point-Likert scale continuum is used in order to force precedence of the respondents; in other words, a respondent is not able to ‘sit on the fence’ of a given answer by selecting ‘neutral’ or ‘not sure.’ Ceteris paribus22, a technology high in ‘ease of use’ is more likely to be adopted.

20 Cronbach’s alpha- a test for a model or survey's internal consistency. Sometimes called a "scale reliability coefficient" Cronbach’s alpha has an important use as a measure of the reliability of psychometric instruments and has a value ranging from 0 to 1, 1 indicating the highest reliability rating. 21 Likert Scale - a type of attitudinal rating scale developed by Rensis Likert which asks participants to indicate the extent to which they agree or disagree with statements. Responses are generally on a continuum including but not limited to "strongly agree," "agree," "disagree," and "strongly disagree." 22 Ceteris paribus - a Latin phrase, literally translated as "with other things [being] the same," and usually rendered in English as "all other things being equal." A prediction, or a statement about causal or logical connections between two states of affairs, is qualified by ceteris paribus in order to acknowledge, and to rule out, the possibility of other factors which could override the relationship between the antecedent and the consequent.

Further extending the ‘ease of use’ construct, based on the literature, this research posits the antecedents of ‘ease of use’ will be ‘trait efficacy’ and ‘state efficacy.’

Technical Self-Efficacy Measures

Computer Self-Efficacy (CSE) has been studied heavily in the last two decades. A number of computer self-efficacy measures were considered for this study as well. Since computer self-efficacy is such an explored aspect of technology use and adoption, this dissertation does not go into as much detail concerning its measurement selection as it does with ethnic identity.

Trait Efficacy

For this research study, CSE is segmented based on the work of Dov Eden (1988) into trait and state efficacies. Furthermore, to measure trait efficacy (or generalized self- efficacy), a twelve-item, six-point Likert scale developed by Sherer et al (1982) will be utilized. This measurement by Sherer et al is recommended by Eden (1988) as an accurate and effective measurement of trait efficacy (Townsend, 1993). Sherer (1982) reposts that this twelve-item measure for trait efficacy has reliability coefficient (Cronbach’s alpha) of

.89.

State Efficacy

State efficacy (or technological task-specific self efficacy) will be measured via a combination of refocused CSE measures. As a result, for this study I have adopted a hybrid compilation of a couple of different, efficient, computer self –efficacy measures. Those utilized are adapted measures from Compeau and Higgins (1995), Murphy (1989), and

Khorrami-Arani (2001). Compeau and Higgins’ model is diagramed in the following figure.

Also, as recommended by Eden (1988) this hybrid of scales will consist of both “magnitude” and “strength” components to more accurately capture state efficacy (Townsend 1993).

Furthermore, for the state efficacy measures, the sources above were adapted utilizing the state Internet self-efficacy and Web-specific self-efficacy based on existing research on

Internet self-efficacy and social cognitive theory (Hsu & Chiu, 2004). State efficacy will be measured with a nineteen-item, six-point Likert scale. The measure used has a reliability rating of .97. Furthermore, as the Internet is, by far, the most widely used / common, application on the largest spectrum of technologies, this dissertation research uses the measure for the Internet’s state efficacy as a proxy for technological state efficacy.

Figure 3: Computer Self-Efficacy Research Model by Compeau and Higgins (1995)

Usefulness Construct

Perceived usefulness is defined as the “degree to which a person believes that using a particular system would enhance his or her job performance" (Davis, 1989). ‘Usefulness’

81 will be replaced with the ‘symbolic utility’ and ‘functional utility’ constructs. Ceteris paribus, a technology high in ‘usefulness’ is one where a user believes in the existence of a positive use-performance relationship and is thus more likely to be adopted.

Further extending the ‘usefulness’ construct, based on the literature, this research posits the antecedents of ‘usefulness’ will be the users’ ‘symbolic utility’ and ‘functional utility.’

Symbolic Utility

Based on evidence from the literature review, the symbolic utility as this dissertation study examines is a new concept applied to technology acceptance. As stated in the previous chapter, the symbolic / social value of technology is largely under-studied in the IT realm as numerous authors identify the lack of existing research and literature on the symbolic / social value of technology (Baskerville, DeGross & Stage 2000; Gumm, 2001; Barrantes, 2006).

Researchers also identify that social influence from peers strongly affects individual's attitude and behavioral intention to adopt a new information technology (Wang, 1998).

Consequently, as a proxy for the symbolic utility of this research, the value expressive component of the Consumer Susceptibility to Reference Group Influence (Park &

Lessig, 1977). Reference group influence itself is defined as the “influence from an actual or imaginary individual or group conceived of having significant relevance upon an individual’s evaluations, aspirations or behavior.” Reference group influence has three motivational components, one of which being the value expressive component (the others being

82 informational and utilitarian components) (Park & Lessig, 1977). The value expressive construct of reference group influence is defined more specifically as the influence relating to an individual’s desire to enhance his/her self concept in the eyes of others (i.e. the individual identifies with positive referents and dissociates him/herself from negative referents) (Park &

Lessig, 1977).

Furthermore, with the aforementioned as a basis, the symbolic utility measure for this research is adapted from the value expressive component of reference group influence measure. This component has a Cronbach’s alpha of .97 among student populations.

Moreover, this construct has been adapted to the new Apple iPhone. The Apple iPhone was chosen due to its combined hedonic (enjoyment) and utilitarian (functional) features in addition to its perceived innovativeness. Akin to the iPod, the iPhone was also selected due to the stereotypical symbolic value of owning Apple technology (particularly, Apple mobile technology).

Functional Utility

The functional utility of this research represents the users’ perception of how well the technology will have a functional purpose to their lives in the light of, predominately, their productivity. For this study, the classic TAM perceived usefulness measure will be adapted to measure the functional utility as hypothesized by this research. This measure has a

Cronbach’s alpha of .89. Moreover, this construct has also been adapted to the new Apple iPhone, again, chosen due to its combined hedonic (enjoyment) and utilitarian (functional) in addition to its perceived innovativeness. The Apple iPhone combines numerous hedonic and utilitarian features; however the primary utilitarian features of the iPhone focused upon via

83 this research will be the actual phone feature, the MP323 feature, and the PDA functionality

(which includes an organizer, calendar, contacts, and to-do-lists among other productivity- focused features).

Sample Size and Characteristics of Research

Although the calculated sample size for this study yielded 487 participants24, will collected survey data from approximately 257 individuals. Individuals will be college students between the ages of 18 and 26 from predominantly African-American backgrounds and will be within an engineering discipline.

This sample size is based on a 99% confidence interval.

Minority participants were surveyed primarily through the National Society of Black

Engineers (NSBE) Online Survey utility as Iowa State University’s population does not have sufficient minority population for this study. Again, a sample of approximately 500 will be collected from the NSBE participants. An online survey utility was selected due to its simplicity in administration to the population and ease of data entry.

College students are selected because the traditional college student age range is ideal for capturing data. Additionally, students are still impressionable in external motivation

23 MP3 – MPEG-1 Audio Layer-3 is a compression scheme used to transfer audio files via the Internet and store in portable players as well as digital audio servers. 24 Sample size was calculated via the formula: 2  z    / 2 (based on Phinney’s multigroup n    ,where z / 2  2.576,   .01, E  .003,and   .59  E  ethnic study)

84 factors that would cause them to adopt technology. Furthermore, identity formation is identity formation is widely acknowledged as one of the central tasks of adolescence

(Erikson 1968; Marcia 1980; Waterman 1985). High school students were considered for this survey; however, they were not chosen due to time constraints and the scope of this dissertation study. Professional participants were also considered for this study. However, they were not chosen as a sample due largely to limited access to participants and the expectation of the researcher to participate in additional activities outside of the research project. Participation in extra activities may have conflicted with the goals of the study and thus diminished the quality of the research. Students of an engineering discipline were chosen so the longitudinal motivations of technology adoption via their pursuance of a technical degree could be assessed, measured and analyzed.

Furthermore, Phinney’s measure was adopted for this study not only because of the versatility of the measure as it is able to be utilized across various ethnicities, but also because of its breadth of and holistic approach to the components that are measured.

Phinney’s multicultural ethnic identity study measures the following aspects of ethnic identity (these components were explored in greater detail in Chapters 1 and 2 of this dissertation):

1. Self-Identification and Ethnicity – the ethnic label one uses for oneself

2. Ethnic Behaviors and Practices – involvement in social activities with members of

one’s group and participation in cultural traditions

3. Affirmation and Belonging – ethnic pride, feeling good about one’s background,

being happy with one’s group membership as well as feelings of belonging and

attachment to the group

4. Ethnic Identity Achievement – a continuous variable, ranging from the lack of

exploration and commitment, reflected in efforts to learn more about one’s

background and a clear understanding of the role of ethnicity for oneself

5. Attitudes towards other groups – this component is not directly a part of ethnic

identity, buy may interact with it as a factor of one’s social identity in the larger

society.

Process

The method of data collection for this dissertation will be via an ISU IRB approved online survey instrument. Participants from the National Society of Black Engineers (NSBE) will be directed to the NSBE Online (NOL) survey feature. There, if they choose to participate, they will agree to the electronic consent form by clicking the “I Agree” button and will proceed to complete the 82-item survey.

The website automatically connects the entered data to a pre-formatted database.

Once the desired number of participants have consented and completed the survey, the database information will be transferred to statistical software. Non-qualitative data will be properly coded for analysis and the data will be analyzed by the statistical software through an Amos Model (explained below) and verified through Iowa State University’s Statistical

Consulting resource. Following this statistical verification, a thorough analysis of the results will be conducted with their discussion. Conclusions, implications as well as future directions for future research will follow.

Analysis

An Amos Model analysis will be applied to the collected data. This model is a type of structured equation modeling (SEM) which itself is a hybrid technique that encompasses aspects of confirmatory factor analysis25, path analysis and regression. Additionally, SEM encourages confirmatory, rather than exploratory, modeling; thus, it is suited to theory testing, rather than theory development. SEM allows researchers in the social sciences, management sciences, behavioral sciences, biological sciences, educational sciences and other fields to empirically assess their theories or in other words confirms relationships in

Attitudinal and Behavioral models. These theories are usually formulated as theoretical models for observed and latent26 variables.

Amos itself is powerful SEM statistical analysis software which enables researchers to support research and theories by extending standard multivariate analysis methods, including regression, factor analysis, correlation, and analysis of variance. Amos will allow the data to be specified, estimated, assessed, and presented via my proposed TAM extension model in an intuitive path diagram to show the hypothesized relationships among variables.

25 Confirmatory Factor Analyses - In theory, exploratory and confirmatory factor analyses can be thought of as two ends of a spectrum. In exploratory analysis, one is trying to make sense of the data; for example “How many factors are there? What are they?” In other words, exploratory analysis can be thought of as a technique for data reduction. On the contrary, in confirmatory analysis, one tests hypotheses corresponding to prior theoretical notions, which can include the number and nature of factors, but can include much more complex hypotheses, such as the equality of factor pattern matrices across populations (<.001NERN<.001). Bother methodologies have their advantages and disadvantages (<.001NERN<.001). 26 Latent variable- unobservable variable

Furthermore, with Amos, one can build attitudinal and behavioral models that more realistically reflect complex relationships, because any numeric variable, whether observed

(such as data from a survey) or latent (such as satisfaction and loyalty), can be used to predict any other numeric variable.

Problems and Limitations

Limitations for this study will include but are not limited to:

 The effort of participants- as there is no incentive to complete the survey, students

may not give the online survey a full effort

 The model presented by this research has been simplified for the purpose and focus of

this dissertation

 Most importantly, however, is the lack of existing research and literature on the

symbolic (social) value of technology (Baskerville, DeGross & Stage 2000; Gumm

2001; Barrantes 2006) and as the focus of this research is the antecedent ethnic

identity, little attention to the construction, measurement and validation of the social

value of technology could be given in this research study.

Proposed Survey Instrument

1. Age

2. Gender M F

3. Parental Education

a. Some High School

b. High School / GED

c. Associate Degree

d. Some College

e. Bachelor’s Degree

f. Some Graduate School

g. Master’s Degree

h. Doctorate (includes M.D., Ph.D. J.D. E.D.)

i. Post Doctorate

4. Adolescent Support

a. Choices TBD

5. Economic Status

a. Choices TBD

6. (US) Region of Origin

a. South East

b. North East

c. Mid-West

d. South West

e. West

7. Do (did) you attend college at an HBCU?

(Classic TAM: Ease of Use)  this and next line of text not included in actual survey

Utilizing a 6-point Likert scale ranging from Strongly disagree to Strongly agree:

8. Getting the information I want from a website is easy

9. Learning to use a website is easy

10. Becoming skillful at using a website is easy

(Symbolic Utility)  this text not included in actual survey

Use the number below to indicate how relevant to you feel each statement.

4 – Highly relevant | 3- Medium Relevance | 2 – Low Relevance | 1 – Not Relevant

11. I feel that the purchase or use of an Apple iPhone will enhance the image which

others will have of me.

12. I feel that those who purchase of use as Apple iPhone possess the characteristics

which I would like to have.

13. I sometimes feel that it would be nice to be like the type of person which

advertisements show using an Apple iPhone.

14. I feel that the people who purchase an Apple iPhone are admired or respected by

others.

15. I feel that the purchase of an Apple iPhone helps me show others who I am, or would

like to be (such as technically savvy, technically up-to-date, or technically ahead-of-

the-rest).

(Functional Utility)  this and next line of text not included in actual survey

Utilizing a 6-point Likert scale ranging from Strongly disagree to Strongly agree:

16. Using the Apple iPhone’s phone feature can enhance my effectiveness at my job

17. Using the Apple iPhone’s phone feature can increase my overall productivity

18. Using the Apple iPhone’s phone feature can improve my job performance

19. Using the Apple iPhone’s PDA features (including email, calendar, to-do lists,

organizer, etc) can enhance my effectiveness at my job

20. Using the Apple iPhone’s PDA features (including email, calendar, to-do lists,

organizer, etc) can increase my overall productivity

21. Using the Apple iPhone’s PDA features (including email, calendar, to-do lists,

organizer, etc) can improve my job performance

22. Using the Apple iPhone’s mp3 feature can enhance my effectiveness at my job

23. Using the Apple iPhone’s mp3 feature can increase my overall productivity

24. Using the Apple iPhone’s mp3 feature can improve my job performance

(Trait Efficacy)  this and next line of text not included in actual survey

Utilizing a 6-point Likert scale ranging from Strongly disagree to Strongly agree:

25. If I cannot to a job the first time, I keep trying until I can.

26. When I set important goals for myself, I rarely achieve them. (r)

27. I give up on things before completing them. (r)

28. When I have something unpleasant to do, I stick to it until I finish it.

29. When I decide to do something, I get right to work on it.

30. When trying to learn something new, I soon give up if I am not initially successful. (r)

31. When unexpected problems occur, I don’t handle them well. (r)

32. I avoid trying to learn new things when they look too difficult for me. (r)

33. Failure just makes me try harder.

34. I feel insecure about my ability to do things. (r)

35. I give up easily. (r)

36. I do not seem capable of dealing with most problems that come up in life. (r)

{Insert iPhone picture and list of features, applications and technical specs here}

(State Efficacy)  this and next line of text not included in actual survey

Utilizing a 6-point Likert scale ranging from Strongly disagree to Strongly agree:

I Feel Confident:

37. navigating the World Wide Web by following hyperlinks.

38. visiting a Web site by entering its address (URL) in the browser.

39. going backward and forward to previously visited Web pages without being lost in

the hyperspace (cyberspace).

40. finding information by using a search engine.

41. finding information in a Web directory or portal.

42. looking for information by querying a Web database.

43. receiving e-mail messages.

44. sending e-mail messages.

45. saving the files attached to e-mail.

46. attaching files to e-mail.

47. posting messages in a Web bulletin board.

48. exchanging messages with other users in discussing forums.

49. chatting on the WWW.

50. downloading files and software.

51. uploading files to a Web site or FTP site.

52. connecting to the Internet through a modem, ADSL, etc.

53. creating a Web page for the World Wide Web.

54. filling out and submitting Web forms.

55. installing an application or software.

(Intention to Use/Adopt)  this text not included in actual survey

56. Assuming I had access to an Apple iPhone, I intend to use it.

57. Given that I had access to an Apple iPhone, I predict that I would use it.

(Multigroup Ethnic Identity Measure)  this text not included in actual survey

58. In terms of ethic group, I consider myself to be: (answer will be coded)

59. How strongly do you feel you identify with the ethnic group you specified in the

previous question?

4 – Very Strongly | 3- Strongly | 2 - Weakly | 1 – Very Weakly

Use the number below to indicate how much you agree or disagree with each statement.

4 - Strongly agree | 3- Somewhat agree | 2 - Somewhat disagree | 1 - Strongly disagree

60. I have spent time trying to find out more about my own ethnic group, such as history,

tradition, and customs,

61. I am active in organization or social groups that include mostly members of my own

ethnic group.

62. I have a clear sense of my ethnic background and what it means for me.

63. I like meeting and getting to know people from ethnic groups other than my own.

64. I think a lot about how my life will be affected by my ethnic group membership.

65. I am happy that I am a member of the group I belong to.

66. I sometimes feel it would be better if different ethnic groups didn’t try to mix

together.

67. I am not very clear about the role of my ethnicity in my life.

68. I often spend time with people from ethnic groups other than my own.

69. I really have not spent much time trying to learn more about the culture and history of

my ethnic group.

70. I have a strong sense of belonging to my own ethnic group.

71. I understand pretty well what my ethnic group membership means to me, in terms of

how to relate to my own group and other groups.

72. In order to learn more about my ethnic background, I have often talked to other

people about my ethnic group.

73. I have a lot of pride in my ethnic group and it accomplishments.

74. I don’t try to become friends with people from other ethnic groups.

75. I participate in cultural practices of my own group, such as special food, music, or

customs.

76. I am involved with people from other ethnic groups.

77. I feel a strong attachment towards my own ethnic group.

78. I enjoy being around people from ethnic groups other than my own.

79. I feel good about my cultural or ethnic background.

Select the number that gives the best answer to each question:

80. My ethnicity is:

(1) Asian, Asian American, or Oriental

(2) Black or African American

(3) Hispanic or Latino

(4) White, Caucasian, European, not Hispanic

(5) American Indian

(6) Mixed; parents are from two different groups

(7) Other (write in):

81. My father’s ethnicity is (use numbers above):

82. My mother’s ethnicity is (use numbers above):

Scoring for the Multiethnic Identity Measure section of Survey Instrument

This measure will be scored by reversing the items (indicated by “R”), summing across item, and obtaining the mean (Items 1, 2, 3, 5, 6, 8R, 10R, 11, 12, 13, 14, 16, 18, and 20).

Subscales are as follow: Affirmation and Belonging (Items 6, 11, 14, 18, and 20); Ethnic

Identity Achievement (Items 1, 3, 5, 8R, 10R, 12 and 13); and Ethnic Behaviors (items 2 and

16). Ethnic self-identification (open-ended response), ethnicity (Item 21), and parents’ ethnicity (Items 22 and 23) are not scored but are used as background information.

Summary The goal of this dissertation is oriented to study the motivation of minorities to adopt, diffuse and learn new, innovative technologies and particularly, the effect of ethnic identification in this motivation. Toward this end, this dissertation extends the original TAM leftward to include ‘trait’ and ‘state’ efficacy as moderators of ‘perceived ease of use’ and

‘symbolic’ and ‘functional’ utility as moderators of ‘perceived usefulness.’ Furthermore, this dissertation research introduces the variable of ‘ethnic identity’ and its intensity as a moderator to all for mediator variables of ‘trait’ and ‘state’ efficacies as well as ‘symbolic’ and ‘functional’ utilities.

My method of data collection consists of a web survey administered to willing participants after they agree to participate in the study via an electronic consent form preceding the online survey. The focus of our ISU IRB approved 82-item survey is that of

Phinney’s Multigroup Ethnic Identity Measure (1992). The resulting data was analyzed using an Amos Model. I predict that ethnic identity will have a positive correlation to aforementioned mediator variables, and ultimately the dependent variable, Likelihood to

Adopt.

CHAPTER 4: ANALYSIS OF DATA

Introduction

The goal of this dissertation is to study the motivation of underrepresented minorities to adopt, diffuse and use new, innovative technologies. Specifically, this study examined the effect of ethnic identification and its intensity as a moderator variable in this motivation.

Chapter 3 described the methodology used to test the hypotheses detailed after the literature review in Chapter 2 of this dissertation. This chapter will present the collected demographic characteristics of the sample as well as the results of the SEM analysis described in Chapter 3.

Organization of Data Analysis

The data in this chapter will be presented two parts. The first part incorporates the components of hypothesis one and two, where the entire sample was used in the SEM analysis. The second part is comprised of the components of hypothesis 3 where the sample was divided into two groups in order to perform a multiple group analysis.

I will first present the descriptive statistics for the demographic characteristics of the survey sample. Following, I will then introduce the data in the two parts mentioned above.

In Part 1, first the hypothesis will again be presented, followed by the validation of the measurement model. Unstandardized and standardized estimates as well as goodness-of-fit statistics will also be presented for this entire-sample measurement model. Reports of each construct’s reliability will be presented following the validation of the measurement model.

Following the report of construct reliability, the results for the causal model will be presented as well as the unstandardized and standardized estimates of the casual paths and their significance. Next, goodness-of-fit statistics will be presented for the causal model. I will then present result of post hoc analyses as well as the unstandardized and standardized estimates of the casual paths and their significance. The goodness-of-fit statistics for the post hoc causal model will then be presented.

In Part 2, first the hypothesis will again be presented, followed by the validation of the multi-group measurement model. The two groups used in this multi-group model will be those that more highly identify with their ethnic affiliation (higher-identifiers) and those that identify with their ethnic affiliation to a lower extent (lower-identifiers). Unstandardized and standardized estimates as well as goodness-of-fit statistics will also be presented for this multi-group measurement model. The results then for the multi-group causal model will be presented as well as the unstandardized and standardized estimates of the casual paths and their significance. Goodness-of-fit statistics will then be presented for the multi-group causal model. Following the multi-group casual model presentation, nested model comparisons of both the multi-group models’ unconstrained and constrained versions will be conducted followed by a chi-square difference test for significance of group invariance. I will then present result of post hoc analyses as well as the unstandardized and standardized estimates of the casual paths and their significance. The goodness-of-fit statistics for the post hoc causal model will then be presented. Following the post hoc multi-group model analysis, nested model comparisons of both the post hoc multi-group models’ unconstrained and constrained versions will be conducted followed by a chi-square difference test for significance of group invariance.

Presentation of Descriptive Characteristics of Respondents

At total of two-hundred fifty-seven (257) respondents completed the online survey.

Although demographic variables were not included in the analysis of the data (i.e. are not part of the proposed models and were not used to determine any endogenous variables), their descriptive frequencies are presented here so illustrate the spread of survey participants completing the online survey.

Table 3 shows the number of responses available for the age demographic variable. Of the

257 completed and valid survey participants, one (.4%) was 17 and thirty-six (14%) chose not to answer these questions. Seventy-eight (30.4%) were between ages 18 – 20, fifty-six

(21.8%) were between the ages of 21 – 23 and thirty (11.7%) were between the ages of 24 –

26 yielding a total of 63.9% of the survey population between the ages of 18 – 26. This age group represents the U.S. average of the undergraduate student.

Age Frequency Percent 36 14.0 <= 17 1 .4 >30 30 11.7 18-20 78 30.4 21-23 56 21.8 24-26 30 11.7 27-30 26 10.1 Total 257 100.0 Table 3: Age Frequencies of Research Participants

Table 4 shows the available responses available across gender for this study. Of the 257 valid and completed surveys, 108 (42%) are female and the remaining 149 (58%) are male.

Gender Frequency Percent Female 108 42.0 Male 149 58.0 Total 257 100.0 Table 4: Gender Frequencies of Research Participants

Tables 5 and 6 show the available responses of participants as it relates to their parents’ education and income levels, respectively. Slightly over half of the survey participants’ parents have completed some college (26.1%), received their Associate’s degree (4.7%) or their Bachelor’s degree (26.1%) as Table 5 demonstrates.

Parental Education Frequency Percent Associate Degree 12 4.7 Bachelor's Degree 67 26.1 Doctorate (includes M.D., Ph.D., Ed.D., 6 2.3 J.D. etc.) High School / GED 24 9.3 Master's Degree 44 17.1 Some College 67 26.1 Some Graduate School 15 5.8 Some High School 13 5.1 Total 257 100.0 Table 5: Parental Education Frequencies of Research Participants

Concerning parental income, Table 6 shows that 34.6% of the participants’ parents earn less than $30,000 per year. Over half of the survey populations’ parents (50.2%) earn less than

$48,000 per year.

100

Parent Income Frequency Percent < $30,000 89 34.6 > $100,000 29 11.3 $30,000 - $48,000 40 15.6 $48,000 - $78,000 62 24.1 $78,000 - $100,000 37 14.4 Total 257 100.0 Table 6: Parental Education Frequencies of Research Participants

Finally, Tables 7 and 8 show the available responses of participants’ geographic origin and geographic region of the college they are attending, respectively. The greatest number of participants (31.9%) are from the South Eastern region of the U.S. as denoted in Table 7.

Region of Origin Frequency Percent International 46 17.9 Mid-West 70 27.2 North East 14 5.4 South East 82 31.9 South West 40 15.6 West 5 1.9 Total 257 100.0 Table 7: Region of Origin Frequencies of Research Participants

Table 8 indicates that the greatest representation of regions from which participants are attending college is the Mid-West with 40.5%.

101

Region - College Frequency Percent International 6 2.3 Mid-West 104 40.5 North East 13 5.1 South East 80 31.1 South West 49 19.1 West 5 1.9 Total 257 100.0 Table 8: College Attending Frequencies of Research Participants

Research Questions and Associated Hypotheses

The purpose of this study is to examine the motivation of underrepresented minorities to adopt, diffuse and use new, innovative technologies. Specifically, this study examined the effect of ethnic identification and its intensity as a moderator variable in this motivation.

Part 1: Single-Group Analysis

Part 1 of this analysis will address research questions 1 and 2 as well as their associated hypotheses 1 and 2 which are stated below.

H1(a): There exists a positive relationship between one’s perceived ease of use to

technology and their likelihood to adopt new innovative technology.

H1(b): There exists a positive correlation between one’s symbolic utility to their

likelihood to adopt.

102

H1(c): There exists a positive correlation between one’s functional utility to their

likelihood to adopt.

Research question 2: Is there a relationship between one’s combined trait and state efficacies to their perceived ease of use of new innovative technology.

H2(a): There exists a positive correlation between one’s trait and state efficacy to

their perceived ease of use.

H2(b): There exists a positive correlation between one’s state efficacy to their

perceived ease of use.

H2(a) Trait Efficacy

H2(b) Ease of Use H1(a) State Efficacy

Ethnic Identity Intent to Use H1(b)

H1(c) Symbolic Utility

Functional Utility

Figure 4: Causal Model with Hypotheses 1 and 2 labeled

103

Part 2: Multi-group Analysis

Part 2 of this analysis will include research question 3 and its associated hypothesis 3 which are stated below.

Research Question 3: To what degree does one’s ethnic identity intensity affect their perceived ease of use and intention to adopt and use new, innovative technology?

H3(a): The greater the intensity of ethnic identity, the stronger relationship between

trait efficacy and perceived ease of use.

H3(b): The greater the intensity of ethnic identity, the stronger relationship between

state efficacy and perceived ease of use

H3(c): The greater the intensity of ethnic identity, the stronger relationship between

symbolic utility and intent to adopt / use.

H3(d): The greater the intensity of ethnic identity, the stronger relationship between

functional utility and intent to adopt / use.

104

Trait Efficacy

H3(a) Ease of Use State Efficacy H3(b)

Ethnic Identity Intent to Use

H3(c)

Symbolic Utility H3(d)

Functional Utility

Figure 5: Causal Model with Hypothesis 3 labeled

Analysis of Data

Overview

This section presents a test of the extended TAM described and diagramed in Chapter

2. The model consists of four exogenous variables, state efficacy, trait efficacy, symbolic utility and functional utility. The model consists of two endogenous variables, ‘ease of use’ and ‘intent to use/adopt.’ There are causal paths from state and trait efficacy to ‘ease of use.’

There is a causal path from ‘ease of use’ to ‘intent to use/adopt’ as well as causal paths from symbolic and functional utilities to ‘intent to use/adopt.’ Ethnic Identity serves as a moderator to all the paths originating from the exogenous variables (state efficacy, trait efficacy, symbolic utility, and functional utility). As diagramed above, SEM analysis

105 assumes correlation between all of the exogenous variables. This is due to the fact that there are indeed antecedents to each of the exogenous variables as well as other factors that are not included in the model or its analysis.

A latent variable27 model was used for this SEM analysis. This approach allows for the compensation of random error of the indictor variables28 on their respective latent variable.

To create this latent variable model, individual items from the scale were divided into three groups, or parcels, and summed together to form three measured variables for each latent variable in order to operationalize each latent construct (Russsell et al 1998). To develop these item parcels, a one-factor model was fit to the total number of items assessing each respective measure. For example, for the trait efficacy measure, there were 12 items used to assess trait efficacy in the survey. A one factor model was fit to these 12 items assessing trait efficacy. The items are then rank ordered based on their factor loadings and assigned to groups so as to equate the average loadings of each group of items on the factor.

More specifically, with the trait efficacy measure, items 1,4, 9 and 12 were assigned to group

1 (TE.I in Figure 7); items 2, 5, 8 and 11 assigned to group 2 (TE.II in Figure 7); and items 3,

6, 7 and 10 were assigned to group 3 (TE.III in Figure 7). When this type of procedure is used, the resulting item parcels should reflect the underlying construct of trait efficacy to an equal degree (1998). Russell et al also notes that that analyzing parcels as opposed to individual items improves the overall fit of the model. This effect is due to improvements in

27 Latent variable - the unobserved variables or constructs or factors which are measured by their respective indicators 28 Indicator variable - observed variables, also called manifest variables or reference variables, such as items in a survey instrument. By convention, indicators should have factor loadings of .7 or higher on their latent factors.

106 the distribution of the measured variables as well as to fewer parameters being estimated as a consequence of a simpler measurement model that is more parsimonious without being exclusive (1998).

Construct Reliability Construct reliability29 was calculated for each latent construct based on the sample data and the resulting factor loadings of the indicator variables on their respective latent variables. The reliability of the state efficacy construct is .960 and the reliability of the trait efficacy construct was found to be .916. The reliability of the symbolic utility construct is

.953 and the reliability of the functional utility construct was found to be .948. And finally, the reliability of the ‘ease of use’ construct is .885 and the reliability of the ‘intent to use’ construct was found to be .979. The construct reliability calculations are summarized in

Table 9.

Construct Reliability State Efficacy .960 Trait Efficacy .916 Symbolic Utility .953 Functional Utility .948 Ease of Use .885 Intent to Use / Adopt .979 Table 9: Construct Reliability Calculations

29 2 2 Construct reliability = [(SUM(sli)) ]/[(SUM(sli)) + SUM(ei))], where sli = standardized loadings for the indicators and ei = the corresponding error terms, where error is 1 minus the reliability of the indicator, which is the square of the indicator's standardized loading

107

Part 1: Single Group Analysis

In part 1 of the analysis, the entire sample was analyzed as a single group. The first phase of the analysis incorporates hypotheses one and two, where the entire sample was used in the SEM analysis.

SEM analysis centers around two basic steps; validating the measurement model30 and then fitting the structural (or causal) model31 (Kline 1998). As hypotheses 1 and 2 require the entire sample to be analyzed together, I will first validate the measurement model for the entire sample.

In validating the original measurement model (see Appendix B), a negative error variance yielded from the measured variable FU.I which is one of the indicators for the functional utility construct. This negative error variance is due to an extremely high correlation between FU.I and the other two indicators of functional utility, FU.II and FU.III.

Since this indicator was correlating too highly with the others, this indicator was deleted from the model (National Statistics, 2001) as diagramed in Figure 7 (detailed estimates of the original model are included in Appendix B). Figure 8 shows the unstandardized results for the model and Figure 9 shows the standardized results for the model.

30 Measurement model - part (or possibly all) of a SEM model which deals with the latent variables and their indicators 31 Structural (causal) model - may be contrasted with the measurement model. It consists of the set of exogenous and endogenous variables in the model, together with the direct effects connecting them, any correlations among the exogenous variable or indicators, and the disturbance terms for these variables (reflecting the effects of unmeasured variables not in the model)

108

e13 e14 e15 e1 e2 e3 1 1 1 e16 e17 1 1 1 1 1 EoU.8 EoU.9 EoU.10 TE.I TE.II TE.III 1 Int.51 Int.52 1 1 EaseOfUse TraitEff Intent to Use

StateEff SymbUtil FuncUtil 1 1 1

SE.I SE.II SE.III SU.III SU.II SU.I FU.II FU.III 1 1 1 1 1 1 1 1

e9 e8 e7 e6 e5 e4 e11 e10

Figure 6: Measurement Model – Entire Sample

109

.37 .17 .16 3.34 2.98 2.43 e13 e14 e15 .06 .25 e1 e2 e3 1 1 1 e16 e17 1 1 1 1 1 EoU.8 EoU.9 EoU.10 TE.I TE.II TE.III 1.00 1.281.41 Int.51 Int.52 1.00 1.271.45 .23 4.06 1.00 .88 1.94 EaseOfUse TraitEff Intent to Use .12 .21 -.13

.64 1.37 -1.11 .33 .78 .04 3.45 -1.07 .56

-.25 2.77 3.82 8.73 2.92 18.37

StateEff SymbUtil FuncUtil 1.56 1.00 .50 1.00 1.00 .99 1.00 .86

SE.I SE.II SE.III SU.III SU.II SU.I FU.II FU.III 1 1 1 1 1 1 1 1 6.47 1.74 1.88 .40 .46 .52 5.04 1.64 e9 e8 e7 e6 e5 e4 e11 e10

Figure 7: Unstandardized Results of Measurement Model – Entire Sample

110

e13 e14 e15 e1 e2 e3 .39 .69 .74 e16 e17 .55 .69 .78 EoU.8 EoU.9 EoU.10 .97 .86 TE.I TE.II TE.III .62 .83 .86 Int.51 Int.52 .74 .83 .88 .99 .93 EaseOfUse TraitEff Intent to Use .18 .22 -.05

.45 .33 -.13 .16 .13 .05 .58 -.31 .23

-.05 .22 .52

StateEff SymbUtil FuncUtil .88 .91 .80 .92 .96 .91 .93 .86 .77 .83 .82 .65 .86 .85 .73 .92 SE.I SE.II SE.III SU.III SU.II SU.I FU.II FU.III

e9 e8 e7 e6 e5 e4 e11 e10

Figure 8: Standardized Results of Measurement Model – Entire Sample

The unstandardized factor loadings of each indictor on its respective latent variable are presented in Table 10. The standardized factor loadings of each indicator on its respective latent variable are presented in Table 11. Table 10 demonstrates that all of the factor loadings for this measurement model are significant at the .001 (two-tailed) confidence interval level.

111

Indicator Construct Estimate S.E. C.R. P TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.273 .101 12.571 <.001 TE.III <--- TraitEff 1.451 .113 12.796 <.001 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil 1.000 .043 23.109 <.001 SU.III <--- SymbUtil .505 .029 17.636 <.001 SE.III <--- StateEff 1.000 SE.II <--- StateEff .993 .045 21.944 <.001 SE.I <--- StateEff 1.560 .077 20.361 <.001 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .864 .053 16.452 <.001 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.282 .130 9.839 <.001 EoU.10 <--- EaseOfUse 1.406 .143 9.852 <.001 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .882 .040 21.784 <.001 Table 10: Unstandardized Regression Weights (Factor Loadings) of indicators on their respective latent construct of Measurement Model – Entire Sample

Indicator Construct Estimate TE.I <--- TraitEff .741 TE.II <--- TraitEff .830 TE.III <--- TraitEff .882 SU.I <--- SymbUtil .921 SU.II <--- SymbUtil .930 SU.III <--- SymbUtil .805 SE.III <--- StateEff .907 SE.II <--- StateEff .912 SE.I <--- StateEff .876 FU.III <--- FuncUtil .958 FU.II <--- FuncUtil .855 EoU.8 <--- EaseOfUse .623 EoU.9 <--- EaseOfUse .830 EoU.10 <--- EaseOfUse .858 Int.51 <--- Intent to Use .986 Int.52 <--- Intent to Use .927 Table 11: Standardized Regression Weights (Factor Loadings) of indicators on their respective latent construct of Measurement Model – Entire Sample

112

Measurement Model Fit

Three primary goodness-of-fit measures will be used to ascertain the fit of the models tested in this research. These include the Chi-square significance test, the Comparative Fir

Index (CFI) and the Root Mean Square Error of Approximation (RMSEA). The chi-square statistic is determined by its p-value: a p-value of .05 or less is considered significant. The

CFI is essentially a normed-fit index (NFI32) designed to take sample size into consideration.

CFI values of .95 or greater are considered representative in a well-fitting model (Bentler

1990) or even superior fitting model (Hu and Bentler 1999). RMSEA values less than .06 indicate good fit (Brown and Cudeck 1993).

The measurement model presented in Figures 7 – 9 yields a chi-square (CMIN) value of 151.309 with 89 degrees of freedom. This indicates that the chi-square is significant at the

.05 level. Table 12 presents the chi-square for the measurement model presented in Figures 7

– 9 (listed as the default model33). The model fit tables also include the values for the saturated models34 and independence models35.

32 Normed fit Index (NFI) – value of .95 indicates good fit; however, this index is nt used much anymore due its revision /upgrade to the CFI. 33 Default model- the researcher's structural model which is always more parsimonious than the saturated model and almost always fitting better than the independence model with which it is compared using goodness of fit measures. In other words, the default model will have a goodness of fit between the perfect explanation of the trivial saturated model and terrible explanatory power of the independence model, which assumes no relationships. 34 Saturated models- the fully explanatory models in which there are as many parameter estimates as degrees of freedom. This is the most un-parsimonious models possible 35 Independence models- assume all relationships among measured variables are 0. This implies the correlations among the latent variables are also 0.

113

Model NPAR CMIN DF P CMIN/DF Default model 47 151.309 89 .000 1.700 Saturated model 136 .000 0 Independence model 16 2982.847 120 .000 24.857 Table 12: Chi-Square significance test of Measurement Model - Entire Sample

Tables 13 and 14 show the CFI and RMSEA, respectively, values for this measurement model. The CFI is calculated to be .978 indicating superior fit of the model.

The RMSEA is calculated at .052 also indicating good-fit for the measurement model.

NFI RFI IFI TLI Model CFI Delta1 rho1 Delta2 rho2 Default model .949 .932 .978 .971 .978 Saturated model 1.000 1.000 1.000 Independence model .000 .000 .000 .000 .000 Table 13: Comparative Fit Index Goodness-of-fit Comparison of Measurement Model – Entire Sample

Model RMSEA LO 90 HI 90 PCLOSE Default model .052 .038 .066 .380 Independence model .305 .296 .315 .000 Table 14: Root Mean Square Error of Approximation Goodness-of-fit Comparison of Measurement Model– Entire Sample

Causal / Structural Model

The previous section demonstrated the validity of the measurement model. As it is shown to be valid for the data, this section will present the result for the causal model used to examine hypotheses 1 and 2 (which uses the entire, un-moderated, sample population).

Figure 10 shows a graphical depiction of the hypotheses transposed onto the diagram of the model.

Figure 11 shows the associated unstandardized path coefficients transposed onto the model while Figure 12 shows the associated standardized path coefficients transposed onto

114 the model. Table 15 presents the unstandardized path coefficients, their standard error and p- values while Table 16 presents the unstandardized path coefficients.

Trait Efficacy H2(a)

H2(b) Ease of Use H1(a) State Efficacy

Intent to Use Ethnic Identity H1(b)

Symbolic Utility H1(c)

Functional Utility

Figure 9: Causal Model with Hypotheses 1 and 2 labeled

Hypotheses 1 Results

Hypotheses one addressed those functions where ‘intent to use’ is the endogenous variable. Intent to use is presented as a function of ease of use as well as symbolic and functional utilities.

H1(a): There exists a positive relationship between one’s perceived ease of use to

technology and their likelihood to adopt new innovative technology.

The regression weight from Ease of use to intent to use is .303 which is not

significant at the .05 (two-tailed) level. Thus, H1(a) is rejected.

H1(b): There exists a positive correlation between one’s symbolic utility to their

likelihood to adopt.

115

The regression weight of symbolic utility to intent to use is -.076 which is not

significant at the .05 (two-tailed) level. Thus H1(b) is rejected.

H1(c): There exists a positive correlation between one’s functional utility to their

likelihood to adopt.

The regression weight from functional utility to intent to use is .201 which is

significant at the .001 (two-tailed) level. Thus the analysis of the data fails to

reject H1(c).

Hypotheses 2 Results

Hypotheses two addresses those functions where ‘ease of use’ is the endogenous variable.

Ease of use is presented as a function of trait and state efficacies.

H2(a): There exists a positive correlation between one’s trait to their perceived ease

of use.

The regression weight from trait efficacy to ease of use is .038 which is

significant at the .05 (two-tailed) level. Thus the analysis of the data fails to

reject H2(a).

H2(b): There exists a positive correlation between one’s state efficacy to their

perceived ease of use.

116

The regression weight from state efficacy to ease of use is .07 which is significant at the .001 (two-tailed) level. Thus the analysis of the data fails to reject H2(b).

Trait Efficacy .04

.07 Ease of Use State Efficacy .30

Intent to Use Ethnic Identity -.08

Symbolic Utility .20

Functional Utility

Figure 10: Unstandardized Results of Causal Model– Entire Sample

117

Trait Efficacy .16

.43 Ease of Use State Efficacy .11

Intent to Use Ethnic Identity -.09

Symbolic Utility .62

Functional Utility

Figure 11: Standardized Results of Causal Model – Entire Sample

Endogenous Exogenous Estimate S.E. C.R. P EaseOfUse <--- TraitEff .038 .016 2.296 .022 EaseOfUse <--- StateEff .070 .012 5.781 <.001 Intent to Use <--- EaseOfUse .303 .168 1.808 .071 Intent to Use <--- FuncUtil .201 .023 8.560 <.001 Intent to Use <--- SymbUtil -.076 .055 -1.389 .165 Table 15: Unstandardized Regression Weights of Causal Model – Entire Sample

Endogenous Exogenous Estimate EaseOfUse <--- TraitEff .157 EaseOfUse <--- StateEff .431 Intent to Use <--- EaseOfUse .105 Intent to Use <--- FuncUtil .617 Intent to Use <--- SymbUtil -.093 Table 16: Standardized Regression Weights of Causal Model– Entire Sample

118

Model Fit

Based on the goodness-of-fit measures of the chi-square, CFI and RMSEA, this causal model fits very well. It fits close to that of the measurement model. The chi-square for this causal model is 165.82 with 93 degrees of freedom. This is significant at the .05 level. The CFI for this causal model is .975 which indicates superior fit. The RMSEA for this causal model is

.055 which also indicates that it is a good fit to the data. Tables 17, 18 and 19 present the chi-square, CFI and RMEA fit statistics, respectively, compared to the saturated and independence models.

Model NPAR CMIN DF P CMIN/DF Default model 43 165.822 93 .000 1.783 Saturated model 136 .000 0 Independence model 16 2982.847 120 .000 24.857 Table 17: Chi-square Significance Test of Causal Model – Entire Sample

NFI RFI IFI TLI Model CFI Delta1 rho1 Delta2 rho2 Default model .944 .928 .975 .967 .975 Saturated model 1.000 1.000 1.000 Independence model .000 .000 .000 .000 .000 Table 18: Comparative Fit Index Goodness-of-fit Comparison of Causal Model – Entire Sample

Model RMSEA LO 90 HI 90 PCLOSE Default model .055 .041 .069 .251 Independence model .305 .296 .315 .000 Table 19: Root Mean Square Error of Approximation Goodness-of-fit Comparison of Causal Model – Entire Sample

Post Hoc Analysis

An additional, theory-supported post hoc modification of the model suggests a better fit of the model if a causal path is added from state efficacy to ease of use. This theory is

119 discussed in Chapter 5. The path coefficient for the newly added causal path from state efficacy to intent to use is .097 which is significant at the .001 level. The (non)significance of the of the original causal paths do not change with this post-hoc modification. The regression weight from ease of use to intent to use is .017 which is not significant at the .05

(two-tailed) level. The regression weight of symbolic utility to intent to use is -.037 which is not significant at the .05 (two-tailed) level. The regression weight from functional utility to intent to use is .181 which is significant at the .001 (two-tailed) level. The path coefficient from trait efficacy to ease of use is .037 which is significant at the .05 (two-tailed) level. The path coefficient from state efficacy to ease of use is .07 which is significant at the .001 (two- tailed) level. Tables 20 and 21 show these statistics in tabular form.

Trait Efficacy

Ease of Use State Efficacy

Intent to Use Ethnic Identity

Symbolic Utility

Functional Utility

Figure 12: Post Hoc Causal Model – Entire Model

120

Trait Efficacy .04

.07 Ease of Use State Efficacy .02

.10 Intent to Use Ethnic Identity -.04

Symbolic Utility .18

Functional Utility

Figure 13: Post Hoc Unstandardized Results of Causal Model – Entire Model

Trait Efficacy .16

.43 Ease of Use State Efficacy .01

.21 Intent to Use Ethnic Identity -.05

Symbolic Utility .56

Functional Utility

Figure 14: Post Hoc Standardized Results of Causal Model – Entire Model

121

Endogenous Exogenous Estimate S.E. C.R. P EaseOfUse <--- TraitEff .037 .016 2.286 .022 EaseOfUse <--- StateEff .070 .012 5.732 <.001 Intent to Use <--- EaseOfUse .017 .185 .093 .926 Intent to Use <--- FuncUtil .181 .024 7.613 <.001 Intent to Use <--- SymbUtil -.037 .054 -.691 .490 Intent to Use <--- StateEff .097 .030 3.213 .001 Table 20: Post Hoc Unstandardized Regression Weights of Causal Model – Entire Sample

Endogenous Exogenous Estimate EaseOfUse <--- TraitEff .156 EaseOfUse <--- StateEff .428 Intent to Use <--- EaseOfUse .006 Intent to Use <--- FuncUtil .556 Intent to Use <--- SymbUtil -.046 Intent to Use <--- StateEff .206 Table 21: Post Hoc Standardized Regression Weights of Causal Model– Entire Sample

Model Fit Summary

The goodness of fit statistics for the model with the post hoc addition of the causal path from State Efficacy to ‘Ease of Use’ does fit slightly better. The chi-square for this causal model is 155.59 with 92 degrees of freedom. This is significant at the .05 level. The

CFI for this causal model is .978 which indicates superior fit. The RMSEA for this causal model is .052 which also indicates that it is a good fit to the data. Tables 22, 23 and 24 present the chi-square, CFI and RMEA fit statistics, respectively, compared to the saturated and independence models.

Model NPAR CMIN DF P CMIN/DF Default model 44 155.592 92 .000 1.691 Saturated model 136 .000 0 Independence model 16 2982.847 120 .000 24.857 Table 22: Post Hoc Chi-square Significance Test of Causal Model – Entire Sample

122

NFI RFI IFI TLI Model CFI Delta1 rho1 Delta2 rho2 Default model .948 .932 .978 .971 .978 Saturated model 1.000 1.000 1.000 Independence model .000 .000 .000 .000 .000 Table 23: Post Hoc Comparative Fit Index Goodness-of-fit Comparison of Causal Model – Entire Sample

Model RMSEA LO 90 HI 90 PCLOSE Default model .052 .037 .066 .394 Independence model .305 .296 .315 .000 Table 24: Post Hoc Root Mean Square Error of Approximation Goodness-of-fit Comparison of Causal Model – Entire Sample

Part 2: Multi-group Analysis This second phase of the analysis is comprised of the components of hypothesis 3 where the sample was divided into two groups in order to perform a multiple group analysis. The two groups were divided based on a median split of the ethnic-identity score using the formula presented in Chapter 3. The ethnic-identity score ranges from one 1 to 4. The median value for this population was 3.571. This yielded in 129 participants being classified in the Lower-

Identifier group and 128 participants classified in the Higher-Identifier group. Lower and higher were used as the descriptors for the median split values due to the lack of polarity within the sample36.

Hypotheses 3 include:

H3(a): The greater the intensity of ethnic identity, the higher the trait efficacy towards

technology.

36 Lack of polarity – originally, this study aimed to examine low and high identifiers; however, with a median split at 3.571 many of the sample placed in the lower-identifiers groups actually had an ethnic identity score above 3 (on a scale of 1 – 4).

123

H3(b): The greater the intensity of ethnic identity, the higher the state efficacy

towards technology.

H3(c): The greater the intensity of ethnic identity, the higher the symbolic utility

towards technology.

H3(d): The greater the intensity of ethnic identity, the higher the functional utility

towards technology.

Trait Efficacy

H3(a) Ease of Use State Efficacy H3(b)

Ethnic Identity Intent to Use

H3(c)

Symbolic Utility H3(d)

Functional Utility

Figure 15: Causal Model with Hypothesis 3 labeled

Multi-group Measurement Models

124 group measurement models37 the factor weights of each of the indicator variables on its respective latent variable are fixed; all else are allowed to vary across groups.

In validating each original multi-group measurement model (see Appendix B), pertaining to the Lower-Identifier group, a negative error variance yielded from the measured variable

Int.51 which is one of the indicators for the ‘Intent to Use’ construct. This negative error variance is due to an extremely high correlation between Int.51 and the indicator of ‘Intent to

Use’, Int.52. As a result, since this indicator was correlating too highly with the other, this measured indicator’s error variance was fixed at 0. Figure 17 shows the lower identifiers measurement model and detailed estimates for this measurement model are presented in

Tables 25 - 26. Figure 18 shows the unstandardized results for the model and Figure 19 shows the standardized results for the model.

37 Multi-group measurement models- measurement models across groups (such as across lower and higher identifiers)

125

e13 e14 e15 0 e1 e2 e3 1 1 1 e16 e17 1 1 1 1 1 EoU.8 EoU.9 EoU.10 TE.I TE.II TE.III 1 FL8 FL9 Int.51 Int.52 1 FL1 FL2 1 FL10 EaseOfUse TraitEff Intent to Use

StateEff SymbUtil FuncUtil FL6 1 FL4 1 1 FL5 FL3 FL7

SE.I SE.II SE.III SU.III SU.II SU.I FU.II FU.III 1 1 1 1 1 1 1 1

e9 e8 e7 e6 e5 e4 e11 e10

Figure 16: Multi-group Measurement Model with Measurement Loadings Constrained – Lower-Identifiers

126

.38 .18 .12 4.01 3.02 3.37 e13 e14 e15 .00 .28 e1 e2 e3 1 1 1 e16 e17 1 1 1 1 1 EoU.8 EoU.9 EoU.10 TE.I TE.II TE.III 1.00 1.271.42 Int.51 Int.52 1.00 1.281.45 .24 3.34 1.00 .88 2.18 EaseOfUse TraitEff Intent to Use .09 .14 -.23

.42 1.59 -1.52 .06

.29 .00 3.71 -1.15 .63

-.06 3.74 2.93 9.18 3.14 18.65

StateEff SymbUtil FuncUtil 1.56 1.00 .51 1.00 1.00 .99 1.00 .88

SE.I SE.II SE.III SU.III SU.II SU.I FU.II FU.III 1 1 1 1 1 1 1 1 6.20 1.58 1.86 .35 .41 .51 4.43 1.27 e9 e8 e7 e6 e5 e4 e11 e10

Figure 17: Unstandardized Results of Multi-group Measurement Model with Measurement Loadings Constrained – Lower Identifiers

127

e13 e14 e15 e1 e2 e3 .39 .69 .80 e16 e17 .45 .64 .67 EoU.8 EoU.9 EoU.10 1.00 .86 TE.I TE.II TE.III .62 .83 .89 Int.51 Int.52 .67 .80 .82 1.00 .93 EaseOfUse TraitEff Intent to Use .13 .16 -.08

.28 .36 -.19 .03

.05 .00 .58 -.35 .24

-.01 .49 .22

StateEff SymbUtil FuncUtil .88 .91 .84 .93 .97 .92 .94 .88 .78 .85 .83 .70 .88 .86 .77 .94 SE.I SE.II SE.III SU.III SU.II SU.I FU.II FU.III

e9 e8 e7 e6 e5 e4 e11 e10

Figure 18: Standardized Results of Multi-group Measurement Model with Measurement Loadings Constrained – Lower Identifiers

128

Indicator Construct Estimate S.E. C.R. P Label TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.279 .099 12.864 <.001 FL1 TE.III <--- TraitEff 1.445 .108 13.405 <.001 FL2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil 1.002 .043 23.277 <.001 FL3 SU.III <--- SymbUtil .510 .028 17.977 <.001 FL4 SE.III <--- StateEff 1.000 SE.II <--- StateEff .993 .045 21.842 <.001 FL5 SE.I <--- StateEff 1.560 .077 20.288 <.001 FL6 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .882 .052 16.938 <.001 FL7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.274 .130 9.777 <.001 FL8 EoU.10 <--- EaseOfUse 1.423 .146 9.773 <.001 FL9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .879 .028 30.855 <.001 FL10 Table 25: Unstandardized Regression Weights (Factor Loadings) of indicators on their respective latent construct of Measurement Model – Lower-Identifiers

Indicator Construct Estimate TE.I <--- TraitEff .674 TE.II <--- TraitEff .802 TE.III <--- TraitEff .821 SU.I <--- SymbUtil .927 SU.II <--- SymbUtil .940 SU.III <--- SymbUtil .836 SE.III <--- StateEff .912 SE.II <--- StateEff .923 SE.I <--- StateEff .885 FU.III <--- FuncUtil .968 FU.II <--- FuncUtil .875 EoU.8 <--- EaseOfUse .621 EoU.9 <--- EaseOfUse .829 EoU.10 <--- EaseOfUse .892 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .925 Table 26: Standardized Regression Weights (Factor Loadings) of indicators on their respective latent construct of Measurement Model – Lower-Identifiers

129

Measurement Model Fit – Lower-Identifiers

The measurement model presented in Figures 17 -19 yields a chi-square (CMIN) value of 288.34 with 189 degrees of freedom. This indicates that the chi-square is significant at the .05 level. Table 27 presents the chi-square for the measurement model presented in

Figures 17 – 19 (listed as the default model). The model fit tables also include the values for the saturated models and independence models.

Tables 28 and 29 show the CFI and RMSEA, respectively, values for this measurement model. The CFI is calculated to be .966 indicating superior fit of the model.

The RMSEA is calculated at .045 also indicating good-fit for the measurement model.

Model NPAR CMIN DF P CMIN/DF Default model 83 288.343 189 .000 1.526 Saturated model 272 .000 0 Independence model 32 3126.702 240 .000 13.028 Table 27: Chi-square Significance Test of Measurement Model – Lower-Identifiers

NFI RFI IFI TLI Model CFI Delta1 rho1 Delta2 rho2 Default model .908 .883 .966 .956 .966 Saturated model 1.000 1.000 1.000 Independence model .000 .000 .000 .000 .000 Table 28: Comparative Fit Index Goodness-of-fit Comparison of Measurement Model – Lower-Identifiers

Model RMSEA LO 90 HI 90 PCLOSE Default model .045 .035 .056 .760 Independence model .217 .210 .224 .000 Table 29: Root Mean Square Error of Approximation Goodness-of-fit Comparison of Measurement Model – Lower-Identifiers

130

As SEM analysis centers around two basic steps, validating the measurement model and then fitting the structural (or causal) model (Kline 1998), measurement models were validated for each group (higher and lower identifiers) individually. In evaluating multi- group measurement models the factor weights of each of the indicator variables on its respective latent variable are fixed; all else are allowed to vary across groups.

Figure 20 shows the higher-identifiers measurement model and detailed estimates for this measurement model are presented in Tables 30 - 32. Figure 21 shows the unstandardized results for the model and Figure 22 shows the standardized results for the model.

e13 e14 e15 e1 e2 e3 1 1 1 e16 e17 1 1 1 1 1 EoU.8 EoU.9 EoU.10 TE.I TE.II TE.III 1 FL8 FL9 Int.51 Int.52 1 FL1 FL2 1 FL10 EaseOfUse TraitEff Intent to Use

StateEff SymbUtil FuncUtil FL6 1 FL4 1 1 FL5 FL3 FL7

SE.I SE.II SE.III SU.III SU.II SU.I FU.II FU.III 1 1 1 1 1 1 1 1

e9 e8 e7 e6 e5 e4 e11 e10

Figure 19: Multi-group Measurement Model with Measurement Loadings Constrained – Higher-Identifiers

131

.35 .19 .18 2.67 2.93 1.44 e13 e14 e15 .11 .22 e1 e2 e3 1 1 1 e16 e17 1 1 1 1 1 EoU.8 EoU.9 EoU.10 TE.I TE.II TE.III 1.00 1.271.42 Int.51 Int.52 1.00 1.281.45 .22 4.62 1.00 .88 1.70 EaseOfUse TraitEff Intent to Use .15 .23 -.14

.81 1.15 -1.02 .56 1.03 .08 3.08 -.97 .45

-.47 2.32 3.85 8.08 2.67 17.29

StateEff SymbUtil FuncUtil 1.56 1.00 .51 1.00 1.00 .99 1.00 .88

SE.I SE.II SE.III SU.III SU.II SU.I FU.II FU.III 1 1 1 1 1 1 1 1 6.51 1.87 1.97 .44 .50 .55 5.18 2.30 e9 e8 e7 e6 e5 e4 e11 e10

Figure 20: Unstandardized Results of Multi-group Measurement Model with Measurement Loadings Constrained – Higher-Identifiers

132

e13 e14 e15 e1 e2 e3 .38 .65 .70 e16 e17 .63 .72 .87 EoU.8 EoU.9 EoU.10 .94 .86 TE.I TE.II TE.III .62 .81 .84 Int.51 Int.52 .80 .85 .93 .97 .93 EaseOfUse TraitEff Intent to Use .24 .23 -.05

.61 .31 -.11 .29 .17 .11 .57 -.28 .21

-.10 .20 .57

StateEff SymbUtil FuncUtil .87 .90 .78 .91 .94 .90 .92 .85 .75 .81 .80 .61 .84 .83 .72 .88 SE.I SE.II SE.III SU.III SU.II SU.I FU.II FU.III

e9 e8 e7 e6 e5 e4 e11 e10

Figure 21: Standardized Results of Multi-group Measurement Model with Measurement Loadings Constrained – Higher-Identifiers

133

Indicators Constructs Estimate S.E. C.R. P Label TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.279 .099 12.864 <.001 FL1 TE.III <--- TraitEff 1.445 .108 13.405 <.001 FL2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil 1.002 .043 23.277 <.001 FL3 SU.III <--- SymbUtil .510 .028 17.977 <.001 FL4 SE.III <--- StateEff 1.000 SE.II <--- StateEff .993 .045 21.842 <.001 FL5 SE.I <--- StateEff 1.560 .077 20.288 <.001 FL6 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .882 .052 16.938 <.001 FL7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.274 .130 9.777 <.001 FL8 EoU.10 <--- EaseOfUse 1.423 .146 9.773 <.001 FL9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .879 .028 30.855 <.001 FL10 Table 30: Unstandardized Regression Weights (Factor Loadings) of indicators on their respective latent construct of Measurement Model – Higher-Identifiers

Indicators Constructs Estimate TE.I <--- TraitEff .796 TE.II <--- TraitEff .849 TE.III <--- TraitEff .933 SU.I <--- SymbUtil .911 SU.II <--- SymbUtil .918 SU.III <--- SymbUtil .783 SE.III <--- StateEff .897 SE.II <--- StateEff .900 SE.I <--- StateEff .867 FU.III <--- FuncUtil .939 FU.II <--- FuncUtil .850 EoU.8 <--- EaseOfUse .617 EoU.9 <--- EaseOfUse .809 EoU.10 <--- EaseOfUse .839 Int.51 <--- Intent to Use .970 Int.52 <--- Intent to Use .926 Table 31: Unstandardized Regression Weights (Factor Loadings) of indicators on their respective latent construct of Measurement Model – Higher-Identifiers

134

Measurement Model Fit – Higher-Identifiers

The measurement model presented in Figures 20 -22 yields a chi-square (CMIN) value of 288.34 with 189 degrees of freedom. This indicates that the chi-square is significant at the .05 level. Table 32 presents the chi-square for the measurement model presented in

Figures 20 – 22 (listed as the default model). The model fit tables also include the values for the saturated models and independence models.

Tables 33 and 34 show the CFI and RMSEA values, respectively, for this measurement model. The CFI is calculated to be .966 indicating superior fit of the model.

The RMSEA is calculated at .045 also indicating good-fit for the measurement model.

Model NPAR CMIN DF P CMIN/DF Default model 83 288.343 189 .000 1.526 Saturated model 272 .000 0 Independence model 32 3126.702 240 .000 13.028 Table 32: Chi-Square significance test of Multi-group Measurement Models

NFI RFI IFI TLI Model CFI Delta1 rho1 Delta2 rho2 Default model .908 .883 .966 .956 .966 Saturated model 1.000 1.000 1.000 Independence model .000 .000 .000 .000 .000 Table 33: Comparative Fit Index Goodness-of-fit Comparison of Multi-group Measurement Models

Model RMSEA LO 90 HI 90 PCLOSE Default model .045 .035 .056 .760 Independence model .217 .210 .224 .000 Table 34: Root Mean Square Error of Approximation Goodness-of-fit Comparison of Multi-group Measurement Models

135

Causal Models

As when testing multi-group models, the primary concern is testing for invariance across groups. Thusly, both lower identifiers and higher identifiers causal models are evaluated by first allowing their causal paths to vary across groups (unconstrained) and then with those casual paths of interest fixed across groups (constrained). Following, a chi-square difference of comparison is conducted to see if the two tests significantly differ from each other.

Unconstrained

The regression weight from ease of use to intent to use is .34 which is not significant at the .05 (two-tailed) level. The regression weight of symbolic utility to intent to use is -.01 which is not significant at the .05 (two-tailed) level. The regression weight from functional utility to intent to use is .19 which is significant at the .001 (two-tailed) level. The regression weight from trait efficacy to ease of use is .04 which is not significant at the .05 (two-tailed) level. The regression weight from state efficacy to ease of use is .05 which is significant at the .01 (two-tailed) level. Figures 23 and 24 show the unstandardized and standardized path regressions, respectively. Tables 35 and 36 detail the unstandardized and standardized estimates, respectively as well.

136

Trait Efficacy .04

.05 Ease of Use State Efficacy .34

Intent to Use Ethnic Identity -.01

Symbolic Utility .19

Functional Utility

Figure 22: Unconstrained Unstandardized Results of Causal Model – Lower-Identifiers

Trait Efficacy .14

.28 Ease of Use State Efficacy .11

Intent to Use Ethnic Identity -.02

Symbolic Utility .58

Functional Utility

Figure 23: Unconstrained Standardized Results of Causal Model – Lower-Identifiers

137

Endogenous Exogenous Estimate S.E. C.R. P Label EaseOfUse <--- TraitEff .039 .027 1.431 .152 R1 EaseOfUse <--- StateEff .046 .016 2.903 .004 R2 Intent to Use <--- EaseOfUse .336 .233 1.440 .150 R3 Intent to Use <--- FuncUtil .192 .027 7.159 <.001 R4 Intent to Use <--- SymbUtil -.014 .069 -.202 .840 R5 TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.280 .100 12.842 <.001 FL1 TE.III <--- TraitEff 1.449 .108 13.366 <.001 FL2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil .999 .043 23.150 <.001 FL3 SU.III <--- SymbUtil .510 .028 17.988 <.001 FL4 SE.III <--- StateEff 1.000 SE.II <--- StateEff .989 .045 21.737 <.001 FL5 SE.I <--- StateEff 1.561 .077 20.363 <.001 FL6 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .805 .034 23.374 <.001 FL7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.284 .131 9.819 <.001 FL8 EoU.10 <--- EaseOfUse 1.393 .142 9.783 <.001 FL9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .873 .029 30.589 <.001 FL10 Table 35: Unconstrained Unstandardized Regression Weights of Causal Model – Lower-Identifiers

138

Endogenous Exogenous Estimate EaseOfUse <--- TraitEff .145 EaseOfUse <--- StateEff .283 Intent to Use <--- EaseOfUse .112 Intent to Use <--- FuncUtil .580 Intent to Use <--- SymbUtil -.017 TE.I <--- TraitEff .673 TE.II <--- TraitEff .802 TE.III <--- TraitEff .822 SU.I <--- SymbUtil .927 SU.II <--- SymbUtil .940 SU.III <--- SymbUtil .836 SE.III <--- StateEff .912 SE.II <--- StateEff .922 SE.I <--- StateEff .886 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .839 EoU.8 <--- EaseOfUse .624 EoU.9 <--- EaseOfUse .838 EoU.10 <--- EaseOfUse .883 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .925 Table 36: Unconstrained Standardized Regression Weights of Causal Model – Lower-Identifiers

The regression weight from ease of use to intent to use is .30 which is not significant at the .05 (two-tailed) level. The regression weight of symbolic utility to intent to use is -.07 which is not significant at the .05 (two-tailed) level. The regression weight from functional utility to intent to use is .17 which is significant at the .001 (two-tailed) level. The regression weight from trait efficacy to ease of use is .03 which is not significant at the .05 (two-tailed) level. The regression weight from state efficacy to ease of use is .10 which is significant at the .0o1 (two-tailed) level. Figures 25 and 26 show the unstandardized and standardized path

139 regressions, respectively. Tables 37 and 38 detail the unstandardized and standardized estimates, respectively as well.

Trait Efficacy .03

.10 Ease of Use State Efficacy .30

Intent to Use Ethnic Identity -.07

Symbolic Utility .17

Functional Utility

Figure 24: Unconstrained Unstandardized Results of Causal Model – Higher-Identifiers

140

Trait Efficacy .13

.59 Ease of Use State Efficacy .11

Intent to Use Ethnic Identity -.09

Symbolic Utility .59

Functional Utility

Figure 25: Unconstrained Standardized Results of Causal Model – Higher-Identifiers

141

Endogenous Exogenous Estimate S.E. C.R. P Label EaseOfUse <--- TraitEff .028 .019 1.488 .137 R-1 EaseOfUse <--- StateEff .098 .016 5.920 <.001 R-2 Intent to Use <--- EaseOfUse .301 .235 1.277 .201 R-3 Intent to Use <--- FuncUtil .172 .026 6.558 <.001 R-4 Intent to Use <--- SymbUtil -.072 .074 -.968 .333 R-5 TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.280 .100 12.842 <.001 FL1 TE.III <--- TraitEff 1.449 .108 13.366 <.001 FL2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil .999 .043 23.150 <.001 FL3 SU.III <--- SymbUtil .510 .028 17.988 <.001 FL4 SE.III <--- StateEff 1.000 SE.II <--- StateEff .989 .045 21.737 <.001 FL5 SE.I <--- StateEff 1.561 .077 20.363 <.001 FL6 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .805 .034 23.374 <.001 FL7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.284 .131 9.819 <.001 FL8 EoU.10 <--- EaseOfUse 1.393 .142 9.783 <.001 FL9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .873 .029 30.589 <.001 FL10 Table 37: Unconstrained Unstandardized Regression Weights of Causal Model – Higher-Identifiers

Endogenous Exogenous Estimate EaseOfUse <--- TraitEff .130 EaseOfUse <--- StateEff .595 Intent to Use <--- EaseOfUse .108 Intent to Use <--- FuncUtil .589 Intent to Use <--- SymbUtil -.090 Table 38: Unconstrained Standardized Regression Weights of Causal Model – Higher-Identifiers

Constrained

The regression weight from ease of use to intent to use is .30 which is not significant at the .05 (two-tailed) level. The regression weight of symbolic utility to intent to use is -.07

142 which is not significant at the .05 (two-tailed) level. The regression weight from functional utility to intent to use is .17 which is significant at the .001 (two-tailed) level. The regression weight from trait efficacy to ease of use is .03 which is not significant at the .05 (two-tailed) level. The regression weight from state efficacy to ease of use is .10 which is significant at the .001 (two-tailed) level. Figures 27 and 28 show the unstandardized and standardized path regressions, respectively. Tables 39 and 40 detail the unstandardized and standardized estimates, respectively as well.

Trait Efficacy .04

.07 Ease of Use State Efficacy .31

Intent to Use Ethnic Identity -.04

Symbolic Utility .18

Functional Utility

Figure 26: Constrained Unstandardized Results of Causal Model – Lower-Identifiers

143

Trait Efficacy .12

.42 Ease of Use State Efficacy .11

Intent to Use Ethnic Identity -.05

Symbolic Utility .56

Functional Utility

Figure 27: Constrained Standardized Results of Causal Model – Lower-Identifiers

Endogenous Exogenous Estimate S.E. C.R. P Label EaseOfUse <--- TraitEff .035 .016 2.220 .026R1 EaseOfUse <--- StateEff .073 .012 5.984 <.001R2 Intent to Use <--- EaseOfUse .311 .166 1.875 .061R3 Intent to Use <--- FuncUtil .181 .019 9.656 <.001R4 Intent to Use <--- SymbUtil -.042 .051 -.828 .408R5 TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.281 .100 12.842 <.001FL1 TE.III <--- TraitEff 1.450 .108 13.367 <.001FL2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil 1.000 .043 23.150 <.001FL3 SU.III <--- SymbUtil .510 .028 17.980 <.001FL4 SE.III <--- StateEff 1.000 SE.II <--- StateEff .987 .045 21.758 <.001FL5 SE.I <--- StateEff 1.557 .076 20.361 <.001FL6 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .805 .034 23.374 <.001FL7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.300 .131 9.892 <.001FL8 EoU.10 <--- EaseOfUse 1.388 .140 9.896 <.001FL9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .874 .029 30.473 <.001FL10 Table 39: Constrained Unstandardized Regression Weights of Causal Model – Lower-Identifiers

144

Endogenous Exogenous Estimate EaseOfUse <--- TraitEff .123 EaseOfUse <--- StateEff .421 Intent to Use <--- EaseOfUse .113 Intent to Use <--- FuncUtil .558 Intent to Use <--- SymbUtil -.052 TE.I <--- TraitEff .673 TE.II <--- TraitEff .802 TE.III <--- TraitEff .823 SU.I <--- SymbUtil .927 SU.II <--- SymbUtil .940 SU.III <--- SymbUtil .836 SE.III <--- StateEff .911 SE.II <--- StateEff .921 SE.I <--- StateEff .884 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .839 EoU.8 <--- EaseOfUse .648 EoU.9 <--- EaseOfUse .858 EoU.10 <--- EaseOfUse .890 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .922 Table 40: Constrained Standardized Regression Weights of Causal Model – Lower-Identifiers

The regression weight from Ease of use to intent to use is .31 which is not significant at the .05 (two-tailed) level. The regression weight of symbolic utility to intent to use is -.04 which is not significant at the .05 (two-tailed) level. The regression weight from functional utility to intent to use is .18 which is significant at the .001 (two-tailed) level. The regression weight from trait efficacy to ease of use is .04 which is significant at the .05 (two-tailed) level. The regression weight from state efficacy to ease of use is .07 which is significant at the .001 (two-tailed) level. Figures 29 and 30 show the unstandardized and standardized path regressions, respectively. Tables 41 and 42 detail the unstandardized and standardized estimates, respectively as well.

145

Trait Efficacy .04

.07 Ease of Use State Efficacy .31

Intent to Use Ethnic Identity -.04

Symbolic Utility .18

Functional Utility

Figure 28: Constrained Unstandardized Results of Causal Model – Higher-Identifiers

Trait Efficacy .17

.48 Ease of Use State Efficacy .10

Intent to Use Ethnic Identity -.05

Symbolic Utility .60

Functional Utility

Figure 29: Constrained Standardized Results of Causal Model – Lower-Identifiers

146

Endogenous Exogenous Estimate S.E. C.R. P Label EaseOfUse <--- TraitEff .035 .016 2.220 .026R1 EaseOfUse <--- StateEff .073 .012 5.984 <.001R2 Intent to Use <--- EaseOfUse .311 .166 1.875 .061R3 Intent to Use <--- FuncUtil .181 .019 9.656 <.001R4 Intent to Use <--- SymbUtil -.042 .051 -.828 .408R5 TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.281 .100 12.842 <.001FL1 TE.III <--- TraitEff 1.450 .108 13.367 <.001FL2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil 1.000 .043 23.150 <.001FL3 SU.III <--- SymbUtil .510 .028 17.980 <.001FL4 SE.III <--- StateEff 1.000 SE.II <--- StateEff .987 .045 21.758 <.001FL5 SE.I <--- StateEff 1.557 .076 20.361 <.001FL6 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .805 .034 23.374 <.001FL7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.300 .131 9.892 <.001FL8 EoU.10 <--- EaseOfUse 1.388 .140 9.896 <.001FL9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .874 .029 30.473 <.001FL10 Table 41: Constrained Unstandardized Regression Weights of Causal Model – Higher-Identifiers

147

Endogenous Exogenous Estimate EaseOfUse <--- TraitEff .173 EaseOfUse <--- StateEff .478 Intent to Use <--- EaseOfUse .102 Intent to Use <--- FuncUtil .604 Intent to Use <--- SymbUtil -.051 TE.I <--- TraitEff .795 TE.II <--- TraitEff .849 TE.III <--- TraitEff .934 SU.I <--- SymbUtil .913 SU.II <--- SymbUtil .916 SU.III <--- SymbUtil .784 SE.III <--- StateEff .903 SE.II <--- StateEff .896 SE.I <--- StateEff .870 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .810 EoU.8 <--- EaseOfUse .595 EoU.9 <--- EaseOfUse .816 EoU.10 <--- EaseOfUse .798 Int.51 <--- Intent to Use .976 Int.52 <--- Intent to Use .924 Table 42: Constrained Standardized Regression Weights of Causal Model – Higher-Identifiers

Model Fit and Comparison

Based on the goodness-of-fit measures of the chi-square, CFI and RMSEA, each of the causal models fit very well, close to that of the measurement model. The chi-square for the unconstrained causal model is 311.75 with 199 degrees of freedom while the chi-square for the constrained causal model is 318.94 with 204 degrees of freedom. Both are significant at the .05 level. The CFI for the unconstrained causal model is .961 while the CFI for the unconstrained causal model is .960 which both indicate superior fits. The RMSEA for the unconstrained causal model is .047 while the RMSEA for the constrained causal model is also .047 which both also indicate good fits to the data. Tables 43, 44 and 45 present the chi-

148 square, CFI and RMEA fit statistics, respectively, compared to the saturated and independence models.

Model NPAR CMIN DF P CMIN/DF Unconstrained 73 311.751 199 .000 1.567 Constrained 68 318.935 204 .000 1.563 Saturated model 272 .000 0 Independence model 32 3126.702 240 .000 13.028 Table 43: Chi-square Significance Test of Multi-group Causal Models

NFI RFI IFI TLI Model Delta1 rho1 Delta2 rho2 CFI

Unconstrained .900 .880 .961 .953 .961 Constrained .898 .880 .961 .953 .960 Saturated model 1.000 1.000 1.000 Independence model .000 .000 .000 .000 .000 Table 44: Comparative Fit Index Goodness-of-fit Comparison of Multi-group Causal Models

Model RMSEA LO 90 HI 90 PCLOSE Unconstrained .047 .037 .057 .672 Constrained .047 .037 .057 .683 Independence model .217 .210 .224 .000 Table 45: Root Mean Square Error of Approximation Goodness-of-fit Comparison of Multi-group Causal Models – Entire Sample

In comparing the two models for invariance, the difference in chi-square values between the constrained and unconstrained models is 6.934 with 5 degrees of freedom. This is not significant at the .05 level. Thus the two groups for this data and model are invariant.

Model DF CMIN P Difference 5 6.934 .226 Table 46: Difference: Model Comparison Unconstrained versus Constrained

149

Hypothesis 3 Results H3(a): The greater the intensity of ethnic identity, the higher the trait efficacy towards

technology.

H3(b): The greater the intensity of ethnic identity, the higher the state efficacy

towards technology.

H3(c): The greater the intensity of ethnic identity, the higher the symbolic utility

towards technology.

H3(d): The greater the intensity of ethnic identity, the higher the functional utility

towards technology.

In comparing the two models for invariance, the difference in chi-square

values between the constrained and unconstrained models is 6.934 with 5

degrees of freedom. This is not significant at the .05 level. Thus the two

groups for this data and model are invariant; therefore, H3 is not supported by

the data.

Post Hoc Analysis 1: Unique Identifier Models

Lower-Identifiers

An additional, theory-supported post hoc modification of the model suggests a better fit of the model for lower identifiers if a causal path is added from state efficacy to intent of use. This theory is discussed in Chapter 5.

The path coefficient for the newly added causal path from state efficacy to intent to use is .097 which is significant at the .001 level.

150

Trait Efficacy

Ease of Use State Efficacy

Intent to Use Ethnic Identity

Symbolic Utility

Functional Utility

Figure 30: Post Hoc Lower-Identifier Causal Model

Higher-Identifiers

Two additional, theory-supported post hoc modifications of the higher-identifier model suggest a better fit of the model if a causal path is added from symbolic utility to ease of use as well as a causal path added from functional utility to ease of use. This theory is discussed in Chapter 5.

151

Trait Efficacy

Ease of Use State Efficacy

Intent to Use Ethnic Identity

Symbolic Utility

Functional Utility

Figure 31: Post Hoc Higher-Identifier Causal Model

The newly added regression weight from state efficacy to intent to use is .11 which is significant at the .01 level. The regression weight from ease of use to intent to use is .13 which is not significant at the .05 (two-tailed) level. The regression weight of symbolic utility to intent to use is .01 which is not significant at the .05 (two-tailed) level. The regression weight from functional utility to intent to use is .17 which is significant at the .001

(two-tailed) level. The regression weight from trait efficacy to ease of use is .04 which is not significant at the .05 (two-tailed) level. The regression weight from state efficacy to ease of use is .04 which is significant at the .01 (two-tailed) level. Figures 33 and 34 show the unstandardized and standardized path regressions, respectively. Tables 47 and 48 detail the unstandardized and standardized estimates, respectively as well.

152

Trait Efficacy .04

.04 Ease of Use State Efficacy .13

.11 Intent to Use Ethnic Identity .01

Symbolic Utility .17

Functional Utility

Figure 32: Post Hoc Unconstrained Unstandardized Results of Causal Model – Lower-Identifiers

Trait Efficacy .14

.28 Ease of Use State Efficacy .04

.22 Intent to Use Ethnic Identity .01

Symbolic Utility .52

Functional Utility

Figure 33: Post Hoc Unconstrained Standardized Results of Causal Model – Lower-Identifiers

153

Endogenous Exogenous Estimate S.E. C.R. P Label EaseOfUse <--- TraitEff .038 .027 1.405 .160R1 EaseOfUse <--- StateEff .045 .016 2.845 .004R2 Intent to Use <--- EaseOfUse .128 .236 .542 .588R3 Intent to Use <--- FuncUtil .171 .027 6.328 <.001R4 Intent to Use <--- SymbUtil .012 .068 .178 .859R5 Intent to Use <--- StateEff .108 .038 2.824 .005R6 Table 47: Post Hoc Unconstrained Unstandardized Regression Weights of Causal Model – Lower-Identifiers

Endogenous Exogenous Estimate EaseOfUse <--- TraitEff .142 EaseOfUse <--- StateEff .277 Intent to Use <--- EaseOfUse .043 Intent to Use <--- FuncUtil .516 Intent to Use <--- SymbUtil .015 Intent to Use <--- StateEff .223 Table 48: Post Hoc Unconstrained Standardized Regression Weights of Causal Model – Lower-Identifiers

The newly added regression weight from symbolic utility to ‘ease of use’ is .01 which is not significant at the .1 level. The newly added regression weight from symbolic utility to

‘ease of use’ is .05 which is not significant at the .1 level. The regression weight from ease of use to intent to use is .29 which is not significant at the .05 (two-tailed) level. The regression weight of symbolic utility to intent to use is -.08 which is not significant at the .05

(two-tailed) level. The regression weight from functional utility to intent to use is .17 which is significant at the .001 (two-tailed) level. The regression weight from trait efficacy to ease of use is .04 which is not significant at the .05 (two-tailed) level. The regression weight from state efficacy to ease of use is .10 which is significant at the .01 (two-tailed) level. Figures

35 and 36 show the unstandardized and standardized path regressions, respectively. Tables

49 and 50 detail the unstandardized and standardized estimates, respectively as well.

154

Trait Efficacy .04

.10 Ease of Use State Efficacy .29

.05 .01 Intent to Use Ethnic Identity -.08

Symbolic Utility .17

Functional Utility

Figure 34: Post Hoc Unconstrained Unstandardized Results of Causal Model – Higher-Identifiers

Trait Efficacy .19

.58 Ease of Use .10 State Efficacy

.17 .09 Intent to Use Ethnic Identity -.10

Symbolic Utility .58

Functional Utility

Figure 35: Post Hoc Unconstrained Standardized Results of Causal Model – Higher-Identifiers

155

Endogenous Exogenous Estimate S.E. C.R. P Label EaseOfUse <--- TraitEff .041 .019 2.103 .035R-1 EaseOfUse <--- StateEff .095 .017 5.722 <.001R-2 EaseOfUse <--- FuncUtil .009 .010 .908 .364R-6 EaseOfUse <--- SymbUtil .049 .030 1.642 .101R-7 Intent to Use <--- EaseOfUse .294 .240 1.223 .221R-3 Intent to Use <--- FuncUtil .172 .027 6.417 <.001R-4 Intent to Use <--- SymbUtil -.077 .073 -1.055 .291R-5 Table 49: Post Hoc Unconstrained Unstandardized Regression Weights of Causal Model – Higher-Identifiers

Endogenous Exogenous Estimate EaseOfUse <--- TraitEff .189 EaseOfUse <--- StateEff .584 EaseOfUse <--- FuncUtil .089 EaseOfUse <--- SymbUtil .172 Intent to Use <--- EaseOfUse .105 Intent to Use <--- FuncUtil .584 Intent to Use <--- SymbUtil -.096 Table 50: Post Hoc Unconstrained Standardized Regression Weights of Causal Model – Higher-Identifiers

The newly added regression weight from state efficacy to intent to use is .11 which is significant at the .01 level. The regression weight from Ease of use to intent to use is .13 which is not significant at the .05 (two-tailed) level. The regression weight of symbolic utility to intent to use is .01 which is not significant at the .05 (two-tailed) level. The regression weight from functional utility to intent to use is .17 which is significant at the .001

(two-tailed) level. The regression weight from trait efficacy to ease of use is .04 which is not significant at the .05 (two-tailed) level. The regression weight from state efficacy to ease of use is .04 which is significant at the .01 (two-tailed) level. Figures 37 and 38 show the unstandardized and standardized path regressions, respectively. Tables 51 and 52 detail the unstandardized and standardized estimates, respectively as well.

156

Trait Efficacy .04

.07 Ease of Use State Efficacy .19

.10 Intent to Use Ethnic Identity -.03

Symbolic Utility .17

Functional Utility

Figure 36: Post Hoc Constrained Unstandardized Results of Causal Model – Lower-Identifiers

Trait Efficacy .15

.41 Ease of Use .07 State Efficacy

.21 Intent to Use Ethnic Identity -.04

Symbolic Utility .52

Functional Utility

Figure 37: Post Hoc Constrained Standardized Results of Causal Model – Lower-Identifiers

157

Endogenous Exogenous Estimate S.E. C.R. P Label EaseOfUse <--- TraitEff .043 .016 2.624 .009R1 EaseOfUse <--- StateEff .071 .012 5.807 <.001R2 Intent to Use <--- EaseOfUse .193 .172 1.126 .260R3 Intent to Use <--- FuncUtil .172 .019 9.144 <.001R4 Intent to Use <--- SymbUtil -.030 .050 -.596 .551R5 Intent to Use <--- StateEff .104 .038 2.742 .006R6 Table 51: Post Hoc Constrained Unstandardized Regression Weights of Causal Model – Lower-Identifiers

Endogenous Exogenous Estimate EaseOfUse <--- TraitEff .149 EaseOfUse <--- StateEff .411 Intent to Use <--- EaseOfUse .069 Intent to Use <--- FuncUtil .522 Intent to Use <--- SymbUtil -.036 Intent to Use <--- StateEff .214 Table 52: Post Hoc Constrained Standardized Regression Weights of Causal Model – Lower-Identifiers

The newly added regression weight from symbolic utility to ‘ease of use’ is .01 which is not significant at the .1 level. The newly added regression weight from symbolic utility to

‘ease of use’ is .05 which is not significant at the .1 level. The regression weight from Ease of use to intent to use is .29 which is not significant at the .05 (two-tailed) level. The regression weight of symbolic utility to intent to use is -.08 which is not significant at the .05

39 and 40 show the unstandardized and standardized path regressions, respectively. Tables

53 and 54 detail the unstandardized and standardized estimates, respectively as well.

158

Trait Efficacy .04

.07 Ease of Use State Efficacy .19

.04 .01 Intent to Use Ethnic Identity -.03

Symbolic Utility .17

Functional Utility

Figure 38: Post Hoc Constrained Unstandardized Results of Causal Model – Higher-Identifiers

Trait Efficacy .21

.46 Ease of Use .06 State Efficacy

.13 .15 Intent to Use Ethnic Identity -.04

Symbolic Utility .58

Functional Utility

Figure 39: Post Hoc Constrained Standardized Results of Causal Model – Higher-Identifiers

159

Endogenous Exogenous Estimate S.E. C.R. P Label EaseOfUse <--- TraitEff .043 .016 2.624 .009R1 EaseOfUse <--- StateEff .071 .012 5.807 <.001R2 EaseOfUse <--- FuncUtil .015 .010 1.452 .147R-6 EaseOfUse <--- SymbUtil .036 .029 1.234 .217R-7 Intent to Use <--- EaseOfUse .193 .172 1.126 .260R3 Intent to Use <--- FuncUtil .172 .019 9.144 <.001R4 Intent to Use <--- SymbUtil -.030 .050 -.596 .551R5 Table 53: Post Hoc Constrained Unstandardized Regression Weights of Causal Model – Higher-Identifiers

Endogenous Exogenous Estimate EaseOfUse <--- TraitEff .208 EaseOfUse <--- StateEff .463 EaseOfUse <--- FuncUtil .149 EaseOfUse <--- SymbUtil .134 Intent to Use <--- EaseOfUse .065 Intent to Use <--- FuncUtil .581 Intent to Use <--- SymbUtil -.037 Table 54: Post Hoc Constrained Standardized Regression Weights of Causal Model – Higher-Identifiers

Model Fit and Comparison

Based on the goodness-of-fit measures of the chi-square, CFI and RMSEA, each of the causal models fit very well, close to that of the measurement model. The chi-square for the unconstrained causal model is 296.88 with 196 degrees of freedom while the chi-square for the constrained causal model is 303.82 with 201 degrees of freedom. Both are significant at the .05 level. The CFI for the unconstrained causal model is .965 while the CFI for the constrained causal model is .964 which both indicate superior fits. The RMSEA for the unconstrained causal model is .045 while the RMSEA for the constrained causal model is also .047 which both also indicate good fits to the data. Tables 55, 56 and 57 present the chi-

160 square, CFI and RMEA fit statistics, respectively, compared to the saturated and independence models.

Model NPAR CMIN DF P CMIN/DF Unconstrained 76 296.881 196 .000 1.515 Measurement weights 76 296.881 196 .000 1.515 Structural weights 71 303.816 201 .000 1.512 Table 55: Chi-square Significance Test of Multi-group Causal Models

NFI RFI IFI TLI Model Delta1 rho1 Delta2 rho2 CFI

Unconstrained .905 .884 .966 .957 .965 Measurement weights .905 .884 .966 .957 .965 Structural weights .903 .884 .965 .957 .964 Table 56: Comparative Fit Index Goodness-of-fit Comparison of Multi-group Causal Models

Model RMSEA LO 90 HI 90 PCLOSE Unconstrained .045 .034 .055 .787 Measurement weights .045 .034 .055 .787 Structural weights .045 .034 .055 .797 Table 57: Root Mean Square Error of Approximation Goodness-of-fit Comparison of Multi-group Causal Models – Entire Sample

Model DF CMIN P Difference 5 6.934 .226 Table 58: Difference: Model Comparison Unconstrained versus Constrained

161

Post Hoc Analysis 2: Parsimonious Comparison

An additional post hoc analysis of the most significant parsimonious model does yield a significant difference between the two groups. In this more-parsimonious model comparison, only the paths significant to both groups (i.e. paths significant to both lower- identifiers as well as higher-identifiers) are constrained in order to test for invariance (Medlin et al, 2002). Thus, in this more-parsimonious model comparison, paths from state efficacy to intent to use as well as the path from functional utility to intent to use are constrained contrast while other paths, residuals, etc are allow to vary across groups. (As in the earlier comparisons this model is compared to a model where all paths are freely allowed to vary across groups.) For this comparison, both lower- and higher- identifiers were compared using the same original model (as this model contains the two paths which are common to both groups).

Unconstrained

Figures 41 and 42 show the unstandardized and standardized constrained results for the lower-identifiers and Figures 43 and 44 show the unstandardized and standardized results for the higher-identifiers.

162

Trait Efficacy .04

.05 Ease of Use State Efficacy .34

Ethnic Identity Intent to Use -.01

Symbolic Utility .19

Functional Utility

Figure 40: Parsimonious Post Hoc Unstandardized Unconstrained Results – Lower-Identifiers

Trait Efficacy .14

.28 Ease of Use State Efficacy .11

Ethnic Identity Intent to Use

-.02

Symbolic Utility .55

Functional Utility

Figure 41: Parsimonious Post Hoc Standardized Unconstrained Results – Lower-Identifiers

163

Trait Efficacy .03

Ease of Use .10 State Efficacy .30

Ethnic Identity Intent to Use

-.07

Symbolic Utility .17

Functional Utility

Figure 42: Parsimonious Post Hoc Unstandardized Unconstrained Results – Higher-Identifiers

Trait Efficacy .13

.59 Ease of Use .11 State Efficacy

Ethnic Identity Intent to Use -.09

Symbolic Utility .59

Functional Utility

Figure 43: Parsimonious Post Hoc Standardized Unconstrained Results – Higher-Identifiers

164

Tables 59 and 60 show the unstandardized and standardized unconstrained results for the lower-identifiers and Tables 61 and 62 show the unstandardized and standardized constrained results for the higher-identifiers.

Endogenous Exogenous Estimate EaseOfUse <--- TraitEff .130 EaseOfUse <--- StateEff .595 Intent to Use <--- EaseOfUse .108 Intent to Use <--- FuncUtil .589 Intent to Use <--- SymbUtil -.090 Table 62: Parsimonious Post Hoc Standardized Unconstrained Results – Higher-Identifiers

165

Constrained

Figures 45 and 46 show the unstandardized and standardized constrained results for the lower-identifiers and Figures 47 and 48 show the unstandardized and standardized results for the higher-identifiers.

Trait Efficacy .04

.07 Ease of Use State Efficacy .35

Ethnic Identity Intent to Use .00

Symbolic Utility .18

Functional Utility

Figure 44: Parsimonious Post Hoc Unstandardized Constrained Results – Lower-Identifiers

166

Trait Efficacy .13

.42 Ease of Use State Efficacy .13

Ethnic Identity Intent to Use

.00

Symbolic Utility .55

Functional Utility

Figure 45: Parsimonious Post Hoc Standardized Constrained Results – Lower-Identifiers

Trait Efficacy .03

Ease of Use .07 State Efficacy .27

Ethnic Identity Intent to Use

-.09

Symbolic Utility .18

Functional Utility

Figure 46: Parsimonious Post Hoc Unstandardized Constrained Results – Higher-Identifiers

167

Trait Efficacy .17

.48 Ease of Use .48 State Efficacy

Ethnic Identity Intent to Use -.11

Symbolic Utility .62

Functional Utility

Figure 47: Parsimonious Post Hoc Standardized Constrained Results – Higher-Identifiers

Tables 63 and 64 show the unstandardized and standardized constrained results for the lower-identifiers and Tables 65 and 66 show the unstandardized and standardized results for the higher-identifiers.

Endogenous Exogenous Estimate S.E. C.R. P Label EaseOfUse <--- TraitEff .037 .027 1.340 .180 R1 EaseOfUse <--- StateEff .073 .012 5.976 <.001 R2 Intent to Use <--- EaseOfUse .353 .218 1.619 .105 R3 Intent to Use <--- FuncUtil .182 .019 9.722 <.001 R4 Intent to Use <--- SymbUtil -.002 .065 -.027 .979 R5 Table 63: Parsimonious Post Hoc Unstandardized Constrained Results – Lower-Identifiers

Endogenous Exogenous Estimate EaseOfUse <--- TraitEff .128 EaseOfUse <--- StateEff .422 Intent to Use <--- EaseOfUse .126 Intent to Use <--- FuncUtil .554 Intent to Use <--- SymbUtil -.002 Table 64: Parsimonious Post Hoc Standardized Constrained Results – Lower-Identifiers

168

Endogenous Exogenous Estimate S.E. C.R. P Label EaseOfUse <--- TraitEff .035 .019 1.807 .071 R-1 EaseOfUse <--- StateEff .073 .012 5.976 <.001 R2 Intent to Use <--- EaseOfUse .272 .249 1.092 .275 R-3 Intent to Use <--- FuncUtil .182 .019 9.722 <.001 R4 Intent to Use <--- SymbUtil -.086 .068 -1.262 .207 R-5 Table 65: Parsimonious Post Hoc Unstandardized Constrained Results – Higher-Identifiers

Endogenous Exogenous Estimate EaseOfUse <--- TraitEff .169 EaseOfUse <--- StateEff .478 Intent to Use <--- EaseOfUse .091 Intent to Use <--- FuncUtil .616 Intent to Use <--- SymbUtil -.107 Table 66: Parsimonious Post Hoc Standardized Constrained Results – Higher-Identifiers

Model Fit and Comparison

Both are significant at the .05 level. The CFI for the unconstrained parsimonious causal model is .961 while the CFI for the constrained parsimonious causal model is .959 which both indicate superior fits. The RMSEA for the unconstrained parsimonious causal model is

.047 while the RMSEA for the constrained parsimonious causal model is also .048 which

169 both also indicate good fits to the data. Tables 67, 68 and 69 present the chi-square, CFI and

RMEA fit statistics, respectively, compared to the saturated and independence models.

Model NPAR CMIN DF P CMIN/DF Unconstrained 73 311.751 199 .000 1.567 Constrained 71 317.939 201 .000 1.582 Saturated model 272 .000 0 Independence model 32 3126.702 240 .000 13.028 Table 67: Chi-square Significance Test of Parsimonious Causal Models

NFI RFI IFI TLI Model CFI Delta1 rho1 Delta2 rho2 Unconstrained .900 .880 .961 .953 .961 Constrained .898 .879 .960 .952 .959 Saturated model 1.000 1.000 1.000 Independence model .000 .000 .000 .000 .000 Table 68: Comparative Fit Index Goodness-of-fit Comparison of Parsimonious Causal Models

Model RMSEA LO 90 HI 90 PCLOSE Unconstrained .047 .037 .057 .672 Constrained .048 .038 .057 .635 Independence model .217 .210 .224 .000 Table 69: Root Mean Square Error of Approximation Goodness-of-fit Comparison of Parsimonious Causal Models – Entire Sample

In comparing the two models for invariance, the difference in chi-square values between the constrained and unconstrained models is 6.188 with 2 degrees of freedom. This is significant at the .05 level. Thus there is a significant difference between parsimonious model for the two groups (lower- and higher- identifiers).

170

Model DF CMIN P Difference 2 6.188 .045 Table 70: Difference: Model Comparison Unconstrained versus Constrained

Summary

Only partial support was found for the relationship proposed by the extended model at the entire-sample level. In part 1, statistically significant, positive correlations were found between state efficacy and ease of use, trait efficacy and ease of use as well as functional utility and intent to use. Post hoc analysis reveals possible additional causal paths between state efficacy and intent to use.

In part 2, the original constrained and unconstrained multi-group models for the current sample were shown to be invariant. Post hoc analyses reveal possible additional causal paths from functional and symbolic utility to ease of use for the higher identifier group as well as an additional causal path between state efficacy and intent to use for the lower identifier group. Additional post hoc analyses reveal significant difference between the two groups when only common statistically significant paths are compared.

This chapter has presented the results of the analysis detailed in Chapter 3 and has test the hypotheses discussed in Chapter 2. The next chapter will present a discussion of the result o the analysis and draw conclusions. It will also discuss the practical and theoretical implications of the current research, the limitations of the present research and it will discuss directions for addition research that are suggested by this study.

171

CHAPTER 5: DISCUSSION AND CONCLUSIONS

Introduction

Chapter 4 presented the results of the analysis of data described in Chapter 3. This final chapter of the dissertation will build on those results via the presentation of pertinent conclusions, implications and areas for future research.

This chapter consists of six sections. The first section will provide a summary of the entire study. The second section provides an overview of the results presented in Chapter 4.

The third section of this chapter will present conclusion based on those findings. Practical and theoretical implications will be presented in the fourth section of this chapter. The fifth section will provide a description of the limitations of this study leading into the sixth section which will provide recommendations for future research. This chapter will conclude with a brief summary of the chapter.

Summary of Study

The goal of this dissertation is oriented to study the motivation of underrepresented minorities to adopt and learn new, innovative technologies. Chapter 1 extensively reviewed the relevant descriptive literature that explains the digital divide, with specific attention to how it is characterized. Chapter 1 demonstrated that while the digital divide is characterized by numerous separators (e.g. income, geography), race (and as I will show later, more importantly ethnicity) is still a proxy of the majority of those negatively affected by this division.

172

The primary research question: Is there a reason the African-American experience is driving certain subgroups of the population to the wrong side of the digital divide? Other questions as part of this research include:

1. Are there differences in technical consumption that differs by ethnicity and the

intensity of ethnic affiliation?

2. Does ethnic identity moderate the traditional (and the proposed extended)

technology acceptance model (TAM).

The motivation for this study is rooted in the issue of the digital divide and possible ways to lessen the divide within the United States.

In Chapter 2, as many indicators are proxies of race, I reviewed literature concerning racial differences with regards to technology and information technology. The theoretical bases of differences between races were explored transitioning into the difference between race and ethnicity (and thusly, racial identity and ethnic identity). I described a new proposed model of ethnic identity and technology adoption incorporating technical self efficacy as well as hedonic and utilitarian characteristics.

More specifically, I presented a survey of the literature of African-American / White differences as it pertains to general motivation as well as motivation with Information

Technology (IT) and innovative technology. I explicitly differentiate between race and ethnicity as well as racial identity and ethnic identity. Moreover, I describe why ethnic identity was chosen as a moderating variable for the proposed model. The literature of ethnic identity (as a function of self and social identity) as it relates to both self-efficacy and technology was also reviewed. Finally, I presented my proposed leftward-extended

173 technology acceptance model (TAM), incorporating the moderating variable of ethnic identity.

To test the model, research questions and hypotheses, an online 82-item survey was completed by 257 National Society of Black Engineers (NSBE) members. This survey consisted on of 6 major constructs – State Efficacy, Trait Efficacy, Ease of Use, Symbolic

Utility, Functional Utility and Intention to Use – designed to capture each of these components for the proposed model.

Findings

This section summarizes the results from chapter 4, organized in the same order as presented in Chapter 4.

Part 1: Single Group Analysis

Part 1 of this analysis addressed research questions 1 and 2 as well as their associated hypotheses 1 and 2 which are stated below along with their conclusions.

Hypotheses 1

Hypotheses one addressed those functions were ‘intent to use’ is the endogenous variable. Intent to use is presented as a function of ease of use as well as symbolic and functional utilities.

174

H1(a): There exists a positive relationship between one’s perceived ease of use of

technology and their likelihood to adopt new innovative technology.

The regression weight from Ease of use to intent to use is .303 which is not

significant at the .05 (two-tailed) level. Thus, H1(a) is rejected and is

therefore not supported by the data.

H1(b): There exists a positive correlation between one’s symbolic utility to their

likelihood to adopt.

The regression weight of symbolic utility to intent to use is -.076 which is not

significant at the .05 (two-tailed) level. Thus H1(b) is rejected and is

therefore not supported by the data.

H1(c): There exists a positive correlation between one’s functional utility to their

likelihood to adopt.

The regression weight from functional utility to intent to use is .201 which is

significant at the .001 (two-tailed) level. Thus the analysis of the data fails to

reject H1(c) and is therefore supported by the data.

175

Hypotheses 2

Hypotheses two addresses those functions where ‘ease of use’ is the endogenous variable. Ease of use is presented as a function of trait and state efficacies.

Research question 2: Is there a relationship between one’s combined trait and state efficacies to their perceived ease of use of new innovative technology?

H2(a): There exists a positive correlation between one’s trait to their perceived ease

of use.

The regression weight from trait efficacy to ease of use is .038 which is

significant at the .05 (two-tailed) level. Thus the analysis of the data fails to

reject H2(a) and is therefore supported by the data.

H2(b): There exists a positive correlation between one’s state efficacy to their

perceived ease of use.

The regression weight from state efficacy to ease of use is .07 which is

significant at the .001 (two-tailed) level. Thus the analysis of the data fails to

reject H2(b) and is therefore supported by the data.

Post Hoc Analysis

An additional, theory-supported post hoc modification of the model suggests a better fit of the model if a causal path is added from state efficacy to ease of use. This theory

176 is discussed later in this chapter. The path coefficient for the newly added causal path from state efficacy to intent to use is .097 which is significant at the .001 level.

Part 2: Multi-group Analysis

Part 2 of this analysis included research question 3 and its associated hypothesis 3 which are stated below.

Hypothesis 3

Research Question 3: To what degree does one’s ethnic identity intensity affect their perceived ease of use and intention to adopt and use new, innovative technology?

H3(a): The greater the intensity of ethnic identity, the higher the trait efficacy towards

technology.

H3(b): The greater the intensity of ethnic identity, the higher the state efficacy

towards technology.

H3(c): The greater the intensity of ethnic identity, the higher the symbolic utility

towards technology.

H3(d): The greater the intensity of ethnic identity, the higher the functional utility

towards technology.

In comparing the two models for invariance, the difference in chi-square

values between the constrained and unconstrained models is 6.934 with 5

177

degrees of freedom. This is not significant at the .05 level. Thus the two

groups for this data and model are invariant; therefore, H3 is not supported by

the data.

Post Hoc Analyses

The newly added regression weight from state efficacy to intent to use is .11 which is significant at the .01 level. The newly added regression weight from symbolic utility to ‘ease of use’ is .01 which is not significant at the .1 level. The newly added regression weight from symbolic utility to ‘ease of use’ is .05 which is not significant at the .1 level.

An additional post hoc analysis does reveal a significant difference between the two groups when only those paths commonly significant to both groups are compared, referred to the parsimonious post hoc analysis.

This chapter will examine these findings and will raise critical issues for future research.

178

Conclusions

Part 1: Single Group Analysis

Hypothesis 1

For Hypothesis 1, only one of the three predicted relationships, functional utility, proved to be significant. Ease of Use and symbolic utilities regression on intent to use was found not to be significant.

The significance of the functional utility path indicates such importance to one’s decision to adopt innovative technology. This suggests that given an innovation, its acceptance holds with the ability to improve one’s effectiveness in completing desired task(s). This conclusion comes to no surprise.

What is surprising concerning this first set of results is the non-significant relationship between ‘ease of use’ and ‘intent to use.’ As this path is part of the traditional

TAM, a significant relationship here is almost automatically assumed.

This is most likely explained by the lack of variance of each of the ‘ease of use’ indicators.

The ‘ease of use’ construct consisted of questions 8 – 10 presented below:

8. Getting the information I want from a website is easy

9. Learning to use a website is easy

10. Becoming skillful at using a website is easy

The descriptive statistics of these indicators are presented below. Table 71 includes the mean, mode, median and variance of each of these indicators. To further solidify this point of non-variance for these indicators, Tables 71 - 74 present the frequencies of responses for each of these questions.

179

Question 8- Getting the Question 10- information I Question 9- Becoming skillful want from a Learning to use a at using a website website is easy. website is easy. is easy.

Valid 257 257 257

Mean 5.14 5.35 5.29

Std. Error of Mean .048 .047 .049

Median 5.00 5.00 5.00

Mode 5 6 6

Std. Deviation .776 .747 .792

Variance .602 .558 .628

Skewness -.696 -1.078 -.799

Std. Error of Skewness .152 .152 .152 Table 71: Descriptive Statistics for Ease of Use Questions

Frequency Percent

2 1 .4

3 5 1.9

4 41 16.0

5 121 47.1

6 89 34.6

Total 257 100.0 Table 72: Frequencies of Responses for Ease of Use Question 8

180

Frequency Percent

2 1 .4

3 3 1.2

4 27 10.5

5 99 38.5

6 127 49.4

Total 257 100.0 Table 73: Frequencies of Responses for Ease of Use Question 9

Frequency Percent

3 5 1.9

4 39 15.2

5 90 35.0

6 123 47.9

Total 257 100.0 Table 74: Frequencies of Responses for Ease of Use Question 8

What these statistics imply is that for this proposed model and this data set, ‘ease of use’ is not a predictor of ‘intent to use’ as there is too little variance in the response for this construct to accurately predict the endogenous variable. This may mean two possible implications for the ease of use constructs, as I will mention in the Implications and Future

Research sections of this chapter.

Similarly, the symbolic utility construct did not yield the expected effects. The effect on intent to use was not only non-significant, but was also a negative relationship. This suggests that for this data, symbolic utility, as measured, did not play a major part in one’s intent to use/adopt. Of important mention here is that ‘symbolic utility’ was actually not measured, but a proxy for the symbolic utility of this research, represented via the value expressive component of the Consumer Susceptibility to Reference Group Influence (Park &

Lessig, 1977). This measure was used to approximate the symbolic utility measure as there

181 currently is no measure for the symbolic implication and/or relevance of technology.

Reference group influence itself is defined as the “influence from an actual or imaginary individual or group conceived of having significant relevance upon an individual’s evaluations, aspirations or behavior” (1977). Reference group influence has three motivational components, including the value expressive component (the others being informational and utilitarian components) (Park & Lessig, 1977). The value expressive construct of reference group influence is defined more specifically as the influence relating to an individual’s desire to enhance his/her self concept in the eyes of others. Additionally, it is possible that symbolic implications of adopting the technology is more intrinsic and for internal symbolic purposes as opposed to extrinsic purposes for outside symbolic importance as predicted.

Another possibility for the insignificant relationship between symbolic utility and intent to use (other than that the value expressive component of the reference group influence measure was an inaccurate approximation) is that this particular measure does give a reasonable approximation, suggesting that symbolic utility has no effect on one’s intent to use technology. However, I do not consider this perspective to be the case, namely due to the number of other authors and researchers that support the notion and importance of the symbolic implications of technology. The lack of existing research and literature on the symbolic (social) value of technology demonstrates a serious gap in current technology adoption literature, models and studies (Baskerville, DeGross & Stage 2000; Gumm 2001;

Barrantes 2006).

182

Hypothesis 2

For Hypothesis 2, both predicted relationships proved to be significant. Trait and state efficacy both have a positive causal relationship to Ease of Use. This suggests that both trait and state efficacy have predictive power when it comes to one’s perceived ease of use.

State efficacy’s impact is more powerful than that of trait efficacy in predicting one’s perceived ease of use. This demonstrates the greater impact one’s task-specific efficacies have on their perceived ease to use and ultimately intent to use / adopt (explained in the next section).

By far, functional utility and state efficacy are the strongest antecedents of the original model for the entire sample.

Post Hoc Analysis

The additional post hoc modification of the model suggested a better fit of the model if a causal path from added from state efficacy to ease of use. Many authors have also identified this relationship – a causal relationship from application-specific self-efficacy to one’s use of a technology (Hwang et al, 2002). This further solidifies state (task-specific) efficacy’s direct and indirect roles in the adoption and use of new innovative technology38.

38 Innovative technology – since an “innovation” is relative based on one’s perspective, innovative technology is defined as new or innovative by the user

183

Part 2: Multi-group Analysis

Hypothesis 3

For hypothesis 3, the initial test for invariance across groups yielded no significant difference between the two groups.

This is most likely explained due to the lack of polarity between the survey participants. More specifically, the ethnic-identity scale has a range of 1 to 4. As I divided the population using a median split, a normal distribution would have allowed for the testing of the extremes; for example, testing those participants who scored 1 or 2 versus those participants who scored 3 or 4. However, the median value was 3.571. This is an extremely high median value for the given range. Therefore, within the data there were over a dozen participants who scored 3.571 and approximately 103 participants forced in the ‘lower- identifier’ category but who indeed scored what normally would have been considered a high identifier score (between 2.5 and 3.571). The lowest value was 1.68 which was placed in the two category; as a result, there really were no ‘”low” identifiers to which to compare the high identifiers. Consequently, the highest of the high identifiers (ethnic-identity score > 3.571; thus labeled ‘higher-identifiers’) were contrasted with the lower of the high identifiers combined with the moderate identifiers (ethnic-identity score between 2 and 3.571; thus labeled ‘lower-identifiers’). As a result, not finding significant differences between the two compared groups is not as surprising as it would have been had the two extremes (low and high identifiers) been compared and still yielded no significant differences.

184

Post Hoc Analyses

Although the initial test for group invariance supported that the two groups are invariant, several post hoc analysis still point to key differences between the two groups.

The first key difference between the two groups is that each of the groups suggests a different model. As presented in the previous chapter, the lower-identifier model has a post hoc addition identical to that of the post hoc addition for the entire-sample: the addition of a causal path from state efficacy to ‘intent to use.’ On the other hand, the higher identifier model has two post hoc additions; one additional causal path from functional utility to ‘ease of use’ and another additional causal path from symbolic utility to ‘ease of use.’ However, this model does not share the same causal path from state efficacy to ‘ease of use’ as does the lower-identifier model nor the model for the entire sample. This not only suggests that there may be differences in motivations and decisions to use/ adopt innovative technology between the two groups but also, as mentioned in Chapter 2, that the same backgrounds and experiences that helped shaped an individuals identity intensity play a part in their considerations when adopting the innovative technology. As the model for the entire sample is the same as that of the lower-identifier, it is extremely interesting that this path from state efficacy to intent to use is non-existent in the higher identifier model. It is also very intriguing how the combined affects of the two groups cloaks those effects unique to the higher-identifiers when isolated. This effect is more than likely caused by the non-significant nature between the two groups. As the initial test for group invariance supported no significance differences between the two groups, this may help explain why the unique effect of the higher identifiers is masked when analyzed with the entire sample.

185

Before addressing the second of the key differences between the two groups, I want to bring special attention to the post hoc model for the higher-identifiers. As mentioned above, there are causal paths from both, functional utility as well as symbolic utility to ‘ease of use.’ Recall in Chapter 2, that the traditional usefulness construct of TAM was replaced by the functional and symbolic utilities construct. As the post hoc modification suggests paths from both functional and symbolic utilities to ‘ease of use,’ this model and data suggest, essentially, a causal path from usefulness to ease of use. This is intriguing as the vast majority of the current IT literature insists that there is only a unidirectional causal path from ease of use to usefulness. This could possibly represent evolving diffusions of innovations as innovations are exponentially evolving. More specifically, I am referring to a type of affective (influential) bias. Amitai Etzioni developed a type of decision-making continuum with normative-affective factors on one end and logical empirical factors on the other (1987). Normative-affective factors can and do influence the selection of means by excluding the role of logical-empirical considerations in many areas (i.e., choice is made exclusively on normative-affective grounds); in other areas - by infusing the deliberations in such a way that logical empirical considerations play a relatively minor or secondary role to normative-affective factors. Etzioni also states how normative-affective factors shape (to a significant extent) decision-making, to the extent it takes place, the information gathered, the ways it is processed, the inferences that are drawn, the options that are being considered, and those that are finally chosen. Therefore, to a significant extent, cognition, inference, and judgment are not logical-empirical endeavors but governed by normative-affective (non- cognitive) factors, reflecting individual, psychodynamic as well as collective processes

(Etzioni, 1987).

186

The second key difference between the two groups is that each of the groups suggests that the two groups are significantly different when a more parsimonious model is compared.

More specifically, when the causal paths significant to both groups (the path from state efficacy to intent to use and the path from functional utility to intent to use) are tested for invariance with the original proposed multi-group model, the two groups are significantly different from each other. This is an important finding for numerous reasons. Firstly, it supports the continuation of similar research as it relates to technology adoption moderated by ethnic identity and its intensity. Secondly, as a significant difference was found between the two groups when they are categorized into the highest of the high (higher) identifiers and the lower of the high as well as moderate (lower) identifiers, this could mean that an even greater significance could be found if the extremes (low scores of 1 and high scores of 4) are compared. Such studies would provide a good basis for future research.

Implications

The last section discussed conclusions based on the findings of this study. This section provides possible implications based on these conclusions and findings. There are various types of implications gathered from this study, including industry, educational and development implications.

As this purpose of this study is to understand how one’s background and experiences shape their motivations to adopt and utilize new, innovative technology, other authors have identified this same need to understand these background experiences and their relation to new technology development and adoption. Rogers concluded from his study that by

187 determining the types of experiences desired by different groups of Internet consumers, as well as the preferences associated with those types, innovators can gain knowledge of how best to accommodate individual differences in website design and content (Rogers, 2004).

For example, various experiences (such as those which shape one’s ethnic-identity) may lead to varying preferences in features and/or the navigation of web-sites.

Educational implications are also concluded from this study. One may automatically look to technical training to improve computer self efficacy. However, group comparisons conducted between treatment and control groups, indicated that brief training interventions improve Internet and computer attitudes but not technical self-efficacy (Lin, 2004). This finding supports the notion of starting technical training early in children’s education. In

2004, the findings of Goode’s research emphasized the importance of developing K-16 policies that explicitly state the academic technology knowledge expected of students, the role of high schools in preparing these scholars, and the responsibility of the university for supporting the ongoing digital needs of its student population (Goode, 2004). Helping to strengthen one’s task-specific (state) efficacies (in this case, computer or technical efficacies) early in life, should increase their motivations toward stronger state efficacies toward the adoption of new, innovative technologies.

For individuals currently past the K – 16 ages the aforementioned may have little impact these individuals; however, for the optimum positive effects of IT skills employees and managers should consider the alignment among level of computer skills, level of IS functionality, and level of task difficulty. Employers should take the appropriate measures to enhance the level of their employees' computer self-efficacy via the appropriate information technology training, education and support (Bani Ali, 2005).

188

Additionally, in both of the aforementioned cases, managers and educators (for example) can intervene with individuals who perceive their performance with a technical system to be low in order to improve their performance (John, 1999). Additional industry implications suggest that managers must understand the implications of using various forms of technology in their organizations, including those which are more readily adopted than others, since communication mediums’ adoption may influence identity fit which, in turn, affects creativity, absenteeism, and team stress (Thatcher, 2000).

Finally, there are possible technical marketing implications as well. Marketers should be cognizant of how one’s background and experiences shape their perception of technologies. Implications for retailers attempting to implement self-service technology, for example, should not only be mindful of the benefits or utility of the technology, but also how the consumer perceives the process or enjoyment of using the technology as self-service consumers will use both types of these value judgments in forming an attitude about the technology which ultimately influences their future behavioral intentions of adopting and using the technology (Collier, 2006).

This section describes some of the implications of this study. The next section will discuss the limitations of the study and lead into future research suggested by this study.

Limitations

There are several limitations of this study. The effort of participants is one such limitation. As there was no incentive to complete the survey, students may not give the online survey their full effort.

189

Additionally, the models presented by this research study have been simplified for the scope and focus of this dissertation. Areas for future and continuing research are discussed in the follow section.

Another possible limitation includes the concentration of the geographic locations of the participants. The majority of the NSBE participants are from the Midwest region and the second largest concentration of participants are from the Southeast region. As most regions within the US have their own distinctive cultures and experiences, a more equal distribution of geographic location of participants may have had an impact on the findings and results.

The fact that an underrepresented minority network was chosen for this study may also be a limitation. It has been suggested that this could also be a reason for the non- polarity of the data; suggesting that because the participants already belonged to such an organization (one dedicated to African-American issues) most of the individuals completing the survey from the organization could have a higher ethnic identity than if African-

American engineers, not apart of such an organization were surveyed.

Most importantly, I persist, is the lack of existing research and literature on the symbolic (social) value of technology (Baskerville, DeGross & Stage 2000; Gumm 2001;

Barrantes 2006). As the focus of this research is the moderator ethnic identity, little attention to the construction, measurement and validation of the social value of technology could be given in this research study. Additionally, the measure used in this study to approximate symbolic utility may not have been as accurate of a proxy of this missing construct as I had predicted.

190

Future Research

As demonstrated throughout this chapter (and the entire study) there are numerous areas for future research.

As the post hoc analysis of this study demonstrated, future research concerning ethnic identity intensity and its impact on technology adoption and use should be further examined.

As the split between the two groups examined in the study yielded different causal models and as the most parsimonious version of the model demonstrated a significant difference between the two groups, ethnic identity intensity and its relationship toward technology adoption and use should no longer be ignored in research and technology development.

Also, as suggested numerous times throughout the study, a symbolic (social) implication of technology scale should be developed (Baskerville, DeGross & Stage 2000;

Gumm 2001; Barrantes 2006). This is a surprisingly overlooked aspect within the diffusion and adoption of innovation literature, research and practice. Once a reliable scale is developed and validated, this research and model should be re-examined utilizing the new measure.

The ease of use construct for this research did not hold true with literature on the classic TAM. One possible reason, in addition to those mentioned earlier this chapter, is the archaic nature of the classic TAM. While the classic TAM serves as an excellent basis for research, the constructs therein are now outdated as they are now approaching two decades since their initial development (Davis et. al., 1989). Furthermore, most TAM applications, revisions and extensions are applied specifically to software and IT. Refining TAM to be more applicable to the innovations of today and tomorrow may provide a more accurate depiction of many of the underlying motivations for adoption and use of current and future

191 innovative technology. Also, such a refinement would help in the development of such technologies as a better understanding of motivations of individuals to adopt those technologies could be more accurately applied.

Similarity to ethnic-identity is genderial identity; more specifically, how one identifies with their sex and to what extent they identify. A similar study to my current study should be conducted with an aim to measure genderial identity and its impact on motivations to adopt and use new innovative technology.

This study should also be re-examined studying different technologies, such as other brands of mobile technology as well as other types of mobile technology (and technology in general). Ubiquitous and always-on computing, for example, are becoming ever more prevalent and including such types of technologies into a study as this could lead to better development of ubiquitous and always-on computing due to a greater understanding and forecasting of its diffusion through adopters.

Additionally, this model and research should be re-examined using polar data with respect to ethnic identity (i.e. with true low and high identifiers). With the current evidence of possible differences between the two non-polar groups (lower and higher identifiers), the current research re-examined with more polar data, differences may be more salient.

This model should also be re-examined with the underrepresented minority group that propelled the study of ethnic identity intensity, the Hispanic/Latino community. As

Deshponde et. al. (1986) noted, Latino ethnic identity intensity have very salient effects when it came to such decisions relating to brand loyalty and even politics. The study re-examined with Hispanic participants may demonstrate that ethnic-identity intensity plays a crucial role in technology brand loyalty and adoption.

192

In conclusion, in most studies, there is always room to refine the proposed model itself – this study is no different. This model could be refined utilizing, especially, some of the significant relationships as well as relationships not significant. Additional constructs, such as enjoyment could also be added to the model as well as additional antecedents. This could even be further enhanced via cross referencing data and patterns across various demographic variables, such as income or parental education for example.

Summary

In closing, this study aims to examine how underrepresented minorities’ background and experiences shape and influence their motivation to adopt and use new, innovative technology.

This study found partial support for the proposed model with the antecedents of functional utility and state efficacy being the most significant components in all models; this includes analysis were the entire sample was used as well as both models specific to each group within the multi-group analysis. Also, (marginally) significant differences were detected between the two tested groups within post hoc analyses of the multi-group models.

This final chapter has described a brief summary of the entire study, including introductory material as well as review of the literature and methods for this research. An overview of the results was then presented followed by conclusions based on those findings.

Practical and theoretical implications were then presented in the fourth section of this chapter. Finally, the fifth section provided a description of the limitations of this study leading into the sixth section which suggested recommendations for future research.

193

It is my hope that this dissertation research study as well as the associated findings, conclusions and suggestions for future research make a significant impact to the field of adoption, diffusion, development and use of innovative (mobile) technology.

194

REFERENCES

Adair, D. (2003). Bridging the digital divide: Providing essential communication. Design

Engineering, 49, 14.

Addae, H. M. (2003). Dimensions, antecedents and consequences of absence legitimacy:

From theory to empirical evidence in a nine-nation study. Unpublished Ph.D.,

Concordia University (Canada), Canada.

African Women Take On Gender Digital Divide. (2002). African Times, 15(13), 17.

Agosta, A., Jr. (2003). The relationship between the social network size and self-efficacy in a

study of housed and homeless adolescents. Unpublished M.A., Roosevelt University,

United States -- Illinois.

Ahlgren, R. L. (2001). An investigation of demographic, psychosocial and self-reported

behavioral influences on career maturity levels of college student-athletes.

Unpublished Ed.D., West Virginia University, United States -- West Virginia.

Al-Gahtani, S. S., & King, M. (1999). Attitudes, satisfaction and usage: factors contributing

to each in the acceptance of information technology. Behaviour & Information

Technology, 18(4), 277-297.

Al-Jalahma, D. R. (2003). Information technology: An assessment of the unique factors

leading to IT adoption and use in a developing country. Unpublished Ph.D.,

Vanderbilt University, United States -- Tennessee.

Alex Van Der, M. (2006). Identifying Some Constraints In First Year Economics Teaching

And Learning At A Typical South African University Of Technology. South African

Journal of Economics, 74(1), 150.

195

Alfeld-Liro, C. J. (1999). Gender, achievement, motivation, and mental health among

adolescents in the 1990s. Unpublished Ph.D., University of Michigan, United States -

- Michigan.

Ali, M., & Alamin, M. (2001). The digital revolution and the new reformation. Harvard

International Review, 23(1), 52.

Allan, P., & Nellen, E. H. (1992). The Motivation to Manage: Differences Between College

Students of Business According to Race. The Mid - Atlantic Journal of Business,

28(3), 273.

Almahboub, S. F. (2000). Attitudes toward computer use and gender differences among

Kuwaiti sixth-grade students. Unpublished Ph.D., University of North Texas, United

States -- Texas.

Amoth, J. E. (2006). Single-gender classes and computer-assisted instruction: A study of the

computer attitudes of middle school girls. Unpublished Ph.D., Capella University,

United States -- Minnesota.

Ancis, J. R., & Phillips, S. D. (1996). Academic gender bias and women's behavioral agency

self-efficacy. Journal of Counseling and Development : JCD, 75(2), 131.

Anderson, D. M. C. (2001). Assessing attitudes toward computers and the use of Internet

resources among undergraduate microbiology students. Unpublished Ph.D., The

University of Southern Mississippi, United States -- Mississippi.

Anindya, C., Kenneth, S. F., & John, H. (2005). An analysis of the determinants of internet

access. Telecommunications Policy, 29(9,10), 731.

Anonymous. (1998). Bridging the digital divide: Simplification. Growth Strategies(897), 1.

Anonymous. (2000). The digital divide. EDUCAUSE Review, 35(6), 29.

196

Anonymous. (2001a). The digital divide: Black-white or rich-poor. Campus - Wide

Information Systems, 18(2), 56.

Anonymous. (2001b). Learning to Bridge the Digital Divide. Education & Training, 43(2/3),

125.

Anonymous. (2003a). 'Digital Divide' Shrinking, Study Says. Communications Today, 9(50),

Anonymous. (2003b). Gender and Computers: Understanding the Digital Divide. Equal

Opportunities International, 22(3), 72.

Anonymous. (2004). News and Views; The Racial Digital Divide: A Narrowing Rift in the

Classroom But Still a Chasm in the American Home. The Journal of African-

Americans in Higher Education(44), 34.

Anonymous. (2006a). Articles in French - II: E-Learning Adoption Determinants: A

Modified Technology Acceptance Model. Communications of the Association for

Information Systems, 18, 3.

Anonymous. (2006b). The Geek Gap: Why Business and Technology Professionals Don't

Understand Each Other and Why They Need Each Other to Survive. Publishers

Weekly, 253(16), 179.

Antonelli, C. (2003). The digital divide: Understanding the economics of new information

and communication technology in the global economy. Information Economics and

Policy, 15(2), 173.

Asfaw, M. (2006). Information and communication technologies within Ethiopia: Socio-

personal factors affecting adaptation and use. Unpublished Ph.D., Walden

University, United States -- Minnesota.

197

Auerbach, S. (2001). Under co-construction: Parent roles in promoting college access for

students of color. Unpublished Ph.D., University of California, Los Angeles, United

States -- California.

Awang-Hashim, R., Harold F. O'Neil, J., & Hocevar, D. (2002). Ethnicity, Effort, Self-

Efficacy, Worry, and Statistics Achievement in Malaysia: A Construct Validation of

the State-Trait Motivation Model. Educational Assessment, 8(4), 341-364.

Babbitt, T. G. (2001). Understanding the Digital Economy: Data Tools, and Research.

Academy of Management. The Academy of Management Review, 26(3), 463.

Bagchi, K. (2005). Factors Contributing to Global Digital Divide: Some Empirical Results.

Journal of Global Information Technology Management, 8(3), 47.

Bakardjieva, M. (2003). What knowledge? Whose fingertips? Negotiating and serving

diverse identities through information technology. Canadian Ethnic Studies, 35(3),

133.

Bani Ali, A. S. (2005). An assessment of the impact of the fit among computer self-efficacy,

task characteristics and systems characteristics on performance and information

systems utilization. Unpublished Ph.D., The George Washington University, United

States -- District of Columbia.

Bansavich, J. C. (2005). Factors influencing preservice teachers' readiness to integrate

technology into their instruction. Unpublished Ed.D., University of San Francisco,

United States -- California.

Bartel, A. P. (1981). Race Differences in Job Satisfaction: A Reappraisal. The Journal of

Human Resources, 16(2), 294.

198

Bassin, R. (2000). The Digital Divide: A Socio-Eco-Techno Phenomenon or Informational

Vacuum? African American Magazine, 3(3), 26.

Beazley, J. A. (2000). Understanding the basis for gender differences in course Web site

usage at a university. Unpublished Ph.D., University of California, Los Angeles,

United States -- California.

Becchetti, L., & Adriani, F. (2005). Does the digital divide matter? The role of information

and communication technology in cross-country level and growth estimates.

Economics of Innovation and New Technology, 14(6), 435.

Belloli, J. R. (1999). Career self-efficacy in the prediction of expected salary. Unpublished

M.A., California State University, Long Beach, United States -- California.

Bernstein, G. (2004). Accommodating Technological Innovation: Identity, Genetic Testing

and the Internet. Vanderbilt Law Review, 57(3), 963.

Beyond the digital divide. (2004). The Economist, 370(8366), 8.

Birgit, S., & Karin, S. (2005). Exploring gender differences in leaders' occupational self-

efficacy. Women in Management Review, 20(7/8), 513.

Blau, F. D., & Graham, J. W. (1990). Black-White Differences in Wealth and Asset

Composition. The Quarterly Journal of Economics, 105(2), 321.

Braundy, M. A. (2005). Men and women and tools: Reflections on male resistance to women

in trades and technology. Unpublished Ph.D., The University of British Columbia

(Canada), Canada.

Brenner, O. C., & Tomkiewicz, J. (1982). Job Orientation of Black and White College

Graduates in Business. Personnel Psychology, 35(1), 89.

199

Brewster, K. L. (1994). Race differences in sexual activity among adolescent women: The

role of neighborhood characteristics. American Sociological Review, 59(3), 408.

Bridging the digital divide. (2003). Malaysian Business, 12.

Britt, M. A. (1991). The role of social identity theory in self-efficacy, goal-setting, and

performance. Unpublished Ph.D., State University of New York at Albany, United

States -- New York.

Brotman, S. N. (2002). The digital dividend. Technology Review, 105(2), 92.

Brown, I., & Licker, P. (2003). Exploring Differences in Internet Adoption and Usage

between Historically Advantaged and Disadvantaged Groups in South Africa. Journal

of Global Information Technology Management, 6(4), 6.

Buelna, G. (2003). Information technology and the Latina/o community: The vision of

California Latino legislators toward information technology. Unpublished Ph.D., The

Claremont Graduate University, United States -- California.

Bulik, B. S. (2006). Staring Into A Digital Divide. Advertising Age, 77(24), 3.

Burnham, J. D. (2004). An exploration of the deterrents to demonstrate and seek proficiency

in adult basic education among blue collar workers. Unpublished Ph.D., The

University of Mississippi, United States -- Mississippi.

Bush Misses Black Americans: Administration's 'Blueprint For New Beginnings' Fails In Its

Attempt To Impact African-Americans and Latinos; Part I of Three Parts. (2001).

Sacramento Observer, p. C6.

Campbell, D. (2001). Can the digital divide be contained? International Labour Review,

140(2), 119.

200

Cawkell, T. (2002). The information society: History and futures. Journal of Information

Science, 28(5), 433.

Chaker, M. N. (2005). Managing Disequilibria in the 'New Economy'. Journal of American

Academy of Business, Cambridge, 6(1), 58.

Chau, P. Y. K. (2001). Influence of computer attitude and self-efficacy on IT usage behavior.

Journal of End User Computing, 13(1), 26.

Chaudhuri, A., Flamm, K. S., & Horrigan, J. (2005). An analysis of the determinants of

internet access. Telecommunications Policy, 29(9,10), 731.

Childers, T. L., Carr, C. L., Peck, J., & Carson, S. (2001). Hedonic and utilitarian

motivations for online retail shopping behavior. Journal of Retailing, 77(4), 511.

Chung, Y. B. (2002). Career Decision-Making Self-Efficacy and Career Commitment:

Gender and Ethnic Differences Among College Students. Journal of Career

Development, 28(4), 277.

Clark, D. (2006). Intel Aims to Bridge Digital Divide; Developing Nations to Get $1 Billion

Over Five Years To Broaden PC, Web Access. Wall Street Journal, p. B.2.

Cobham, B. A. (2003). As I walk through the valley: The self-efficacy of tenured Black

women faculty employed at predominately White institutions. Unpublished Ph.D.,

Indiana University, United States -- Indiana.

Collier, J. E. (2006). Examining customers' intentions to use self-service technology through

utilitarian and hedonic value judgments. Unpublished Ph.D., The University of

Memphis, United States -- Tennessee.

201

Combs, G. M. (2000). A field experimental analysis of the impact of self-efficacy on diversity

training effectiveness. Unpublished Ph.D., The University of Nebraska - Lincoln,

United States -- Nebraska.

Compeau, D., Higgins, C. A., & Huff, S. (1999). Social cognitive theory and individual

reactions to computing technology: A longitudinal study. MIS Quarterly, 23(2), 145.

Cook, P. J., & Ludwig, J. (1997). Weighing the "burden of 'acting white'": Are there race

differences in attitudes toward education? Journal of Policy Analysis and

Management, 16(2), 256.

Cornelia, B. (2005). A Review of: "Gender and Computers: Understanding the Digital

Divide'. Information Society, 21(3), 221.

Crenshaw, S. J. N. (2005). The role of systemic public education in the retention and attrition

of students placed at-risk: Closing or creating the achievement gap. Unpublished

Ed.D., Auburn University, United States -- Alabama.

Cubba, S. (2004). Uncovering hidden leaders: The effects of gender and ethnicity on leader

attributes, leadership ability, and citizen leader potential. Unpublished D.P.A.,

University of La Verne, United States -- California.

Culleeney, M. A. (1996). Computer-assisted instruction in a college setting: Survey of

student demographic and motivational characteristics. Unpublished Ph.D., Loyola

University of Chicago, United States -- Illinois.

Cullen, R. (2001). Addressing the digital divide. Online Information Review, 25(5), 311.

Dalton, C., & Yeung, W. J. (2005). Black-White Differences in Occupational Prestige: Their

Impact on Child Development. The American Behavioral Scientist, 48(9), 1229.

202

Daniels, G. L. (2002). Uncovering factors in urban high school students' media career

considerations: A Media Career Interest Model. Unpublished Ph.D., University of

Georgia, United States -- Georgia.

Dario, B., & Fabrizia, M. (2003). Self-image and perceived self-efficacy during adolescence.

Journal of Youth and Adolescence, 32(5), 337.

Davenport, E. (2003). Mobilizing the Information Society: Strategies for Growth and

Opportunity. Journal of Documentation, 59(3), 359.

Davis, F. D. (1989). Perceived Usefulness, Perceived Ease Of Use, And User Accep. MIS

Quarterly, 13(3), 319.

Decker, L. A. (2003). Brain-compatible classroom effects on college students in a public

speaking class. Unpublished Ed.D., Regent University, United States -- Virginia.

Deng, X. (2000). Developing an information technology learning model in a computer-

integrated manufacturing (CIM) context. Unpublished Ph.D., The University of

Toledo, United States -- Ohio.

Deshponde, R., Hoyer, W. D., & Donthu, N. (1986). The Intensity of Ethnic Affiliation: A

Study of the Socialogy of Hispanic Consumption. Journal of Consumer Research, 13,

214 - 220.

Devi, C. (2002). Providing students tech choices. Computimes Malaysia, 1.

Dewan, S., Ganley, D., & Kraemer, K. L. (2005). Across the Digital Divide: A Cross-

Country Multi-Technology Analysis of the Determinants of IT Penetration. Journal

of the Association for Information Systems, 6(12), 1.

Dewan, S., & Riggins, F. J. (2005). The Digital Divide: Current and Future Research

Directions. Journal of the Association for Information Systems, 6(12), 1.

203

Diane, P. (2004). Understanding Social and Spatial Divisions in the New Economy: New

Media Clusters and the Digital Divide. Economic Geography, 80(1), 45.

Digital divide widening, UN secretary general says. (2002). Europemedia, 1.

Digital divide widens. (2002). Europemedia, 1.

Digital Inclusion: All inclusive. (2006). New Media Age, 23.

DiMaggio, P., Hargittai, E., W. Russell, N., & Robinson, J. P. (2001). Social implications of

the Internet. Annual Review of Sociology, 27, 307.

Doherty, N. F., & Ellis-Chadwick, F. E. (2003). The relationship between retailers' targeting

and e-commerce strategies: An empirical analysis. Internet Research, 13(3), 170.

Dolinsky, A. L., Caputo, R. K., & Pasumarty, K. (1994). Long-term entrepreneurship

patterns: A national study of black and white female entry and stayer status

differences. Journal of Small Business Management, 32(1), 18.

Donohue, T. R., Meyer, T. P., & Henke, L. L. (1978). Black and White Children: Perceptions

of TV Commercials. Journal of Marketing, 42(4), 34.

Dugdale, A. M. (1995). Devices and desires: Constructing the intrauterine device, 1908-

1988. Unpublished Ph.D., University of Wollongong (Australia), Australia.

Dunham, R. S. (1999). Across America, a Troubling `Digital Divide'. Business Week(3640),

40.

Dvir, T., Eden, D., & Banjo, M. L. (1995). Self-fulfilling prophecy and gender: Can women

be Pygmalion and Galatea? Journal of Applied Psychology, 80(2), 253.

Eden, D. (1986). OD and Self-Fulfilling Prophecy: Boosting Productivity by Raising

Expectations. The Journal of Applied Behavioral Science, 22(1), 1.

204

Eden, D. (1988). Pygmalion, Goal Sitting, And Expectancy: Compatible Ways To. Academy

of Management. The Academy of Management Review, 13(4), 639.

Eden, D., Geller, D., Gewirtz, A., Gordon-Terner, R., & et al. (2000). Implanting Pygmalion

leadership style through workshop training: Seven field experiments. Leadership

Quarterly, 11(2), 171.

Eden, D., & Kinnar, J. (1991). Modeling Galatea: Boosting Self-Efficacy to Increase

Volunteering. Journal of Applied Psychology, 76(6), 770.

Eden, D., & Zuk, Y. (1995). Seasickness as a self-fulfilling prophecy: Raising self-efficacy

to boost performance at sea. Journal of Applied Psychology, 80(5), 628.

Edwards, Y. D. (2005). Looking Beyond the Digital Divide. Federal Communications Law

Journal, 57(3), 585.

Elsass, P. M. (1998). Sometimes a good excuse is just not enough. The Academy of

Management Executive, 12(4), 131.

Fenech, T. (1998). Using perceived ease of use and perceived usefulness to predict

acceptance of the World Wide Web. Computer Networks & ISDN Systems, 30(1-7),

629.

Finnigan, K. S. (2003). Principal leadership and teacher expectancy in a high-stakes

accountability policy context. Unpublished Ph.D., The University of Wisconsin -

Madison, United States -- Wisconsin.

Fitzgerald, M. (1993). Minorities managing minorities. Editor & Publisher, 126(35), 28.

Flanagan, L., & Jacobsen, M. (2003). Technology leadership for the twenty-first century

principal. Journal of Educational Administration, 41(2), 124.

Florida, R. (2001). E-inclusion: It's not a choice. InformationWeek(830), 136.

205

Ford, P. J. (2001). A further analysis of the ethics of representation in virtual reality: Multi-

user environments. Ethics and Information Technology, 3(2), 113.

Fors, M., & Moreno, A. (2002). The benefits and obstacles of implementing ICTs strategies

for development from a bottom-up approach. Aslib Proceedings, 54(3), 198.

Fossett, M. A., Galle, O. R., & Kelly, W. R. (1986). Racial Occupational Inequality, 1940-

1980: National and Regional Trends. American Sociological Review, 51(3), 421.

Foster, J. J., II. (2001). Self-efficacy, African-American youth, and technology: Why the belief

that "I cannot do that" is driving the digital divide. Unpublished Ph.D., The

University of Alabama, United States -- Alabama.

Fraser, O. A. (2002). Accessing the Gift: African-Americans and the Digital Divide. Hyde

Park Citizen, p. 14.

Friedman, T. L. (2005). The World is Flat: A brief History of the Twenty-first centruy.

Gaillard, F. D. (2001). Understanding the digital divide as it relates to electronic commerce.

Unpublished Ph.D., Old Dominion University, United States -- Virginia.

Gangseog, R., Jongchul, P., & Lawrence, F. (2006). The Role of Product Type and Country-

of-Origin in Decisions about Choice of Endorser Ethnicity in Advertising.

Psychology & Marketing, 23(6), 487.

Gardyn, R. (2001). True colors. American Demographics, 23(4), 14.

Garnes, D. M. B. (2005). Hope and self-efficacy as motivational influences in technology

adoption. Unpublished Ph.D., Regent University, United States -- Virginia.

Gavin, M. S., & Bernadette, M. W. (2005). The Influence of Perceptions of Social Identity

on Information Technology-Enabled Change. Group & Organization Management,

30(3), 289.

206

Gefen, D., & Ridings, C. M. (2003). IT acceptance: Managing user - IT group boundaries.

Database for Advances in Information Systems, 34(3), 25.

Gilad, C., Stanley, M. G., & Dov, E. (2001). Validation of a new general self-efficacy scale.

Organizational Research Methods, 4(1), 62.

Gilad, C., Stanley, M. G., & Dov, E. (2004). General self-efficacy and self-esteem: toward

theoretical and empirical distinction between correlated self-evaluations. Journal of

Organizational Behavior, 25(3), 375.

Gittleman, M., & Wolff, E. N. (2004). Racial Differences in Patterns of Wealth

Accumulation. The Journal of Human Resources, 39(1), 193.

Glassberg, B. C. (2000). Individual use of Web techonology: A reconceptualization and

empirical test of the technology acceptance model. Unpublished Ph.D., University of

South Carolina, United States -- South Carolina.

Golden, D. (2000). Unequal Signs: For Inner-City Schools, Calculators May Be The Wrong

Answer --- More Poor, Black Kids Use Machines to Do Math, But Scores Remain

Low --- An Inverse `Digital Divide'. Wall Street Journal, p. A.1.

Goldsborough, R. (2000). Bridging the digital divide. Office Solutions, 17(4), 11.

Goldsmith, R. E., White, D. J., & Stith, M. T. (1987). Values of Middle-Class African-

Americans and Whites: A Replication and Extension. Psychology & Marketing

(1986-1998), 4(2), 135.

Goode, J. (2004). Mind the gap: The digital dimension of college access. Unpublished Ph.D.,

University of California, Los Angeles, United States -- California.

Gordon, C. B., II, & Anand, K. (2005). Explaining consumer acceptance of handheld Internet

devices. Journal of Business Research, 58(5), 553.

207

Gorski, P. C. (2003). Privilege And Repression In The Digital Era: Rethinking The

Sociopolitics Of The Digital Divide. Race, Gender & Class, 10(4), 145.

Graczyk, A. J. (2005). An examination of the influence of formal characteristics of paper-

and-pencil selection test procedures on outcomes: A policy-capturing approach.

Unpublished Ph.D., University of Houston, United States -- Texas.

Greenhaus, J. H., & Parasuraman, S. (1993). Job performance attributions and career

advancement prospects: An examination of gender and race effects. Organizational

Behavior and Human Decision Processes, 55(2), 273.

Greenwood, A. (2005). Sailing into the future. New Statesman, 18(840), 34.

Griffin, C. (2006). Investigating the effects of stable personality traits on computer self-

efficacy with repeated training. Unpublished Ph.D., Southern Illinois University at

Carbondale, United States -- Illinois.

Grubesic, T. H. (2001). A spatial analysis of Internet accessibility. Unpublished Ph.D., The

Ohio State University, United States -- Ohio.

Gupta, S. (2006). Longitudinal investigation of collaborative e-learning in an end-user

training context. Unpublished Ph.D., Georgia State University, United States --

Georgia.

Guus, G. M. P., & Kees van, M. (2005). An Investigation of Factors that Influence Senior

Executives to Accept Innovations in Information Technology. International Journal

of Management, 22(4), 542.

Hachman, M. (2003). New Study Attempts To Track "Digital Divide" ; A study funded by

Advanced Micro Devices released Wednesday found that the gap between the wired

208

and unwired is expanding across economic and social lines that differ throughout the

world. ExtremeTech.com, 1.

Hacker, K. L., & Mason, S. M. (2003). Ethical gaps in studies of the digital divide. Ethics

and Information Technology, 5(2), 99.

Hadas, A. (2001). Predictors of perceived self-efficacy among White and African American

psychologically distressed older adults. Unpublished Ph.D., Columbia University,

United States -- New York.

Hadley, J. G. (2003). A test of Bandura's theory: Generalized self-efficacy and the

personality traits of introversion and extroversion as measures of job performance.

Unpublished Ph.D., Saybrook Graduate School and Research Center, United States --

California.

Hall, J. M. (2002). A comparative analysis of mathematics self-efficacy of developmental and

non-developmental college freshman mathematics students. Unpublished Ph.D., The

University of Mississippi, United States -- Mississippi.

Hans van der, H. (2004). User Acceptance Of Hedonic Information Systems. MIS Quarterly,

28(4), 695.

Harper, C. E. (2004). The conflict management styles, strength of conflict management self-

efficacy, and moral development levels of school counselors. Unpublished Ph.D.,

North Carolina State University, United States -- North Carolina.

Harris, A. F. (2003). Towards cultural competence: An exploratory study of the relationship

between racial identity and contextual performance indicators of public employees.

Unpublished Ph.D., North Carolina State University, United States -- North Carolina.

209

Hartman, J. B. (2004). Hierarchical influence of personal values and innovativeness on

adolescent Web consumption. Unpublished Ph.D., The University of Arizona, United

States -- Arizona.

Havelka, D. (2003). Predicting software self efficacy among business students: A preliminary

assessment. Journal of Information Systems Education, 14(2), 145.

Hayes, N. (1991). Social Identity, Social Representations And Organizational Culture.

Unpublished Ph.D., Council for National Academic Awards (United Kingdom),

England.

Heafner, T. L. (2002). Powerful methods: A framework for effective integration of technology

in secondary social studies. Unpublished Ph.D., The University of North Carolina at

Greensboro, United States -- North Carolina.

Henry, J. W., & Stone, R. W. (1999). The impacts of end-user gender, education,

performance, and system use on computer self-efficacy and outcome expectancy.

Southern Business Review, 25(1), 10.

Hewes, G. M. (2003). Black-white differences in the gender wage ratio. Unpublished Ph.D.,

Harvard University, United States -- Massachusetts.

Hinkelman, M. (2005). Race, income, education found to affect high-speed Net usage.

Knight Ridder Tribune Business News, 1.

Hodge, S. M. (2001). Social indicators, dysfunctional cognitions, and career decision-

making self-efficacy in work role participation of TANF recipients. Unpublished

Ph.D., University of Georgia, United States -- Georgia.

210

Holden, C. (2003). When racial discrimination can help cure: DRUGS: Research suggests

that race affects many diseases, but scientists vary on the usefulness of such findings.

Constance Holden reports. Financial Times, p. 15.

Holladay, C. L. (2004). An evaluation of diversity training: Effects of trainer characteristics

and training focus. Unpublished Ph.D., Rice University, United States -- Texas.

Homer-Dixon, T. (2003). Winners and losers in the information age. IEEE Spectrum, 40(5),

11.

Houston, M. B. (1995). Managerial representations of a strategic market initiative: The

influence of identity. Unpublished Ph.D., Arizona State University, United States --

Arizona.

Hsu, M.-H., & Chiu, C.-M. (2004). Internet Self-Efficacy and Electronic Service Acceptance.

Decision Support Systems, 38(2004), 369– 381.

Hu, P. J.-H. (1998). Management of telemedicine technology in healthcare organizations:

Technology acceptance, adoption, evaluation, and their implications. Unpublished

Ph.D., The University of Arizona, United States -- Arizona.

Hubbard, L. (2000a). Don't believe the hype: Is there really a digital divide? Call & Post, p.

B5.

Hubbard, L. (2000b). Income, not race, defines digital divide, says study. Recorder, p. A6.

Hubbard, L. (2000c). Study looks at digital divide. New York Amsterdam News, p. 4.

Hubbard, L. (2000d). Study looks at digital divide. Afro - American Red Star, p. B9.

Hubbard, T. M. (2005). It's about more than "just be consistent" or "out-tough them":

Culturally responsive classroom management. Unpublished Ph.D., The Ohio State

University, United States -- Ohio.

211

Hudson, C. K. (2000). Poverty and the secondary labor market. Unpublished Ph.D., The

University of North Carolina at Chapel Hill, United States -- North Carolina.

Hunter, S. L. (2005). The role of gender, race and self-efficacy in career choice for primarily

male-oriented, primarily female-oriented and minority under-represented

professions. Unpublished Ph.D., University of Kentucky, United States -- Kentucky.

Igbaria, M., & Iivari, J. (1995). The effects of self-efficacy on computer usage. Omega,

23(6), 587.

Igbaria, M., & Wormley, W. M. (1992). Organizational experiences and career success of

MIS professionals and managers: An examination of race differences. MIS Quarterly,

16(4), 507.

Igbaria, M., & Wormley, W. M. (1995). Race differences in job performance and career

success. Association for Computing Machinery. Communications of the ACM, 38(3),

82.

Ilie, V. (2005). Opening Up The Adoption Shell. “Shallow” And “Deep Level” Usage

Patterns In Healthcare Organizations. Proceedings of the 2005 Southern Association

of Information Systems Conference, 5.

Jacobsen, D. M. (1998). Adoption patterns and characteristics of faculty who integrate

computer technology for teaching and learning in higher education. Unpublished

Ph.D., University of Calgary (Canada), Canada.

Jaeki, S., & Yong Jin, K. (2006). Social influence process in the acceptance of a virtual

community service. Information Systems Frontiers, 8(3), 241.

Jan van, D., & Kenneth, H. (2003). The digital divide as a complex and dynamic

phenomenon. Information Society, 19(4), 315.

212

Jawahar, I. M., & Elango, B. (2001). The effect of attitudes, goal setting and self-efficacy on

end user performance. Journal of End User Computing, 13(2), 40.

Jayaratne, T. E., Ybarra, O., Sheldon, J. P., Brown, T. N., & et al. (2006). White Americans'

Genetic Lay Theories of Race Differences and Sexual Orientation: Their Relationship

with Prejudice toward African-Americans, and Gay Men and Lesbians. Group

Processes & Intergroup Relations, 9(1), 77.

Jeffreys, J. W. (2001). The impact of supervision on social worker perception of self-efficacy.

Unpublished Ph.D., State University of New York at Albany, United States -- New

York.

Jenkins, H. (2002). Cyberspace and race. Technology Review, 105(3), 89.

John, W. H., & Robert, W. S. (1999). The impacts of end-user gender, education,

performance, and system use on computer self-efficacy and outcome expectancy.

Southern Business Review, 25(1), 10.

Johnson, G. (2004). Sisters united in techno-empowerment; International organizations

sponsor a computer skills class for African women. Philadelphia Tribune, p. 1B.

Johnson, L. (2002). Closing the technology gap by adding color. New York Amsterdam News,

p. 33.

Johnson, R. D., & Marakas, G. M. (2000). Research report: The role of behavioral modeling

in computer skills acquisition - Toward refinement of the model. Information Systems

Research, 11(4), 403.

Jones, W. J., Childers, T. L., & Kaufman-Scarborough, C. (2006). Hedonic and Utilitarian

Dimensions of Online Retail Shopping Behavior: Does Disability Matter? American

Marketing Association. Conference Proceedings, 17, 168.

213

Joo-Young, J., Jack Linchuan, Q., & Yong-Chan, K. (2001). Internet connectedness and

inequality: Beyond the "divide". Communication Research, 28(4), 507.

Joyce, S., & Paul, R. S. (2006). Gender Inequities of Self-Efficacy on Task-Specific

Computer Applications in Business. Journal of Education for Business, 81(5), 269.

Kallol, B. (2005). Factors Contributing to Global Digital Divide: Some Empirical Results.

Journal of Global Information Technology Management, 8(3), 47.

Kang, J. (2000). Cyber-race. Harvard Law Review, 113(5), 1130.

Kathleen, H. (2003). How self-efficacy and gender issues affect software adoption and use.

Association for Computing Machinery. Communications of the ACM, 46(9), 167.

Kauffman, R. J., & Techatassanasoontorn, A. A. (2005). I There a Global Digital Divide for

Digital Wireless Phone Technologies? Journal of the Association for Information

Systems, 6(12), 1.

Kelly, L. Z., Wayne, A. H., Pamela, L. P., Angela, K. M., & Christian, K. (2001). Beyond

self-efficacy: Interactive effects of role conflict and perceived collective efficacy.

Journal of Managerial Issues, 13(4), 483.

Kilker, J. A. (1999). Networking identity: A case study examining social interactions and

identity in the early development of e-mail technology. Unpublished Ph.D., Cornell

University, United States -- New York.

Kondo, H. (2003). Gratifications sought from new technology: Cellular phones in the lives of

Japanese young people. Unpublished M.A., California State University, Fullerton,

United States -- California.

Kouzmin, A. (2000). Information technology and development: is the 'digital divide' an

inevitability? Public Administration & Development, 20(3), 167.

214

Kulviwat, S. (2004). An investigation of factors predicting consumer adoption of high-

technology innovations: The integrative roles of cognition and affect. Unpublished

Ph.D., Southern Illinois University at Carbondale, United States -- Illinois.

Kushairi, A. (2002). Digital divide among computer users. Computimes Malaysia, 1.

Kvasny, L., & Keil, M. (2006). The challenges of redressing the digital divide: a tale of two

US cities. Information Systems Journal, 16(1), 23.

Kwok, S., & Uncles, M. (2005). Sales promotion effectiveness: the impact of consumer

differences at an ethnic-group level. The Journal of Product and Brand Management,

14(2/3), 170.

Kwon, H. (1996). Adoption of cellular telephone technologies and services: User perceptions

and motivations in the United States (Hawaii) and South Korea. Unpublished Ph.D.,

University of Hawai'i, United States -- Hawaii.

Laurel, A. H., Patricia, M., & Carol, L. S. (1998). Procrastination in college students: The

role of self-efficacy and anxiety. Journal of Counseling and Development : JCD,

76(3), 317.

Laurentis, C. D. (2006). Digital Knowledge Exploitation: ICT, Memory Institutions and

Innovation from Cultural Assets. Journal of Technology Transfer, 31(1), 77.

Layne, C. M., & Hohenshil, T. H. (2005). High Tech Counseling: Revisited. Journal of

Counseling and Development : JCD, 83(2), 222.

Lederer, A. L., Maupin, D. J., Sena, M. P., & Zhuang, Y. (2000). The technology acceptance

model and the World Wide Web. Decision Support Systems, 29, 269 - 282.

215

Lee, H.-H., Fiore, A. M., & Kim, J. (2006). The role of the technology acceptance model in

explaining effects of image interactivity technology on consumer responses.

International Journal of Retail & Distribution Management, 34(8), 621.

Lee, S.-G. (2003). An integrative study of mobile technology adoption based on the

technology acceptance model, theory of planned behavior and diffusion of innovation

theory. Unpublished Ph.D., The University of Nebraska - Lincoln, United States --

Nebraska.

Leo, R. V. (2004). Predicting consumer intentions to use on-line shopping: the case for an

augmented technology acceptance model. Information & Management, 41(6), 747.

Levin, M. (1990). Implications of race and sex differences for compensatory affimative

action and the concept of discrimination. The Journal of Social, Political, and

Economic Studies, 15(2), 175.

Levin, M. (1991). Race Differences: An Overview. The Journal of Social, Political, and

Economic Studies, 16(2), 195.

Lewis, W., Agarwal, R., & Sambamurthy, V. (2003). Sources Of Influence On Beliefs About

Information Technology Use: An Empirical Study Of Knowledge Workers. MIS

Quarterly, 27(4), 657.

Lewis, W., Jr. (1999). Technology beliefs, belief antecedents, and IT usage: A study of the

World Wide Web. Unpublished Ph.D., The Florida State University, United States --

Florida.

Liang, X. (2005). Development and validation of a new computer self-efficacy scale for use

in complex technology contexts. Unpublished Ph.D., State University of New York at

Albany, United States -- New York.

216

Lin, C.-H. (2003). Intergenerational parallelism of self-efficacy: Moderating variables,

mediating variables, and common antecedents. Unpublished Ph.D., Texas A&M

University, United States -- Texas.

Lin, L. F. (2004). An investigation of applicant reactions to Internet-based selection

procedures. Unpublished Ph.D., Bowling Green State University, United States --

Ohio.

Lin, M. H. (2002). The relationship between sports employee characteristics and the factors

of computer technology self-efficacy, attitudes toward computer technology, and use

of computer technology. Unpublished Ph.D., University of Kansas, United States --

Kansas.

Littrell, R. F., & Nkomo, S. M. (2005). Gender and race differences in leader behaviour

preferences in South Africa. Women in Management Review, 20(7/8), 562.

Livingston, V. (2000). An analysis of the personal and social characteristics that impact

black female career mobility. Unpublished D.S.W., Norfolk State University, United

States -- Virginia.

Lonergan, J. M. (2002). Development of a model of applicant reactions to cognitive ability

tests: Shedding light on racial subgroup test score differences. Unpublished Ph.D.,

Northern Illinois University, United States -- Illinois.

Loney, N. (2003). Digital divide holds back rural areas, says report. Regeneration &

Renewal, 2.

Lu, M.-t. (2001). Digital divide in developing countries. Journal of Global Information

Technology Management, 4(3), 1.

217

Lynch, L. L. S. (2001). Technology value-beliefs and technology skill self-efficacy of

preservice teachers: A measurement and structural model. Unpublished Ph.D.,

University of Missouri - Columbia, United States -- Missouri.

Lynch, M. T. (2003). The effects of research training and research efficacy on empirical

research productivity by practicing social workers. Unpublished Ph.D., University of

Pittsburgh, United States -- Pennsylvania.

Lynn, R. (2000). Race differences in sexual behavior and their demographic implications.

Population and Environment, 22(1), 73.

MacInnis, P. (2002). Youth ignored on IT agenda. Computing Canada, 28(18), 5.

Maertz, C. P., Jr., Bauer, T. N., Mosley, D. C., Jr., Posthuma, R. A., & Campion, M. A.

(2005). Predictors of self-efficacy for cognitive ability employment testing. Journal

of Business Research, 58(2), 160.

Mahmoud, A. R. (1999). Effectiveness of Web-based virtual learning environments in

business education: Focusing on basic skills training for information technology.

Unpublished Ph.D., Louisiana State University and Agricultural & Mechanical

College, United States -- Louisiana.

Mallory, N. (2000). Dot Coms: Diversity & Profits; Can it Be? How Diversity is Quietly

Influencing the New Market. Next Step Magazine, 42.

Mapp-Embry, D. L. T. (2005). Teachers' sense of efficacy and African American student

engagement: A case study of classes in a selected Kentucky middle school.

Unpublished Ed.D., University of Kentucky, United States -- Kentucky.

Margo, R. A., & Smith, J. P. (1986). Race and Human Capital: Comment/Reply. The

American Economic Review, 76(5), 1221.

218

Mason, L. K. (2004). Attachment classification, parental bonding and parental authority as

they relate to anxiety, depression, and self-efficacy from a multicultural perspective.

Unpublished Ph.D., Fordham University, United States -- New York.

Mathieson, K., Peacock, E., & Chin, W. W. (2001). Extending the technology acceptance

model: The influence of perceived user resources. Database for Advances in

Information Systems, 32(3), 86.

Matthews, D., & Victoria, A. (2002). Cutting a figure: Tailoring, technology and social

identity in nineteenth-century Paris. Unpublished Ph.D., Stanford University, United

States -- California.

Matthews, S. (2001). Making it: A qualitative study of the experiences of black women who

have held middle management positions in United States corporations. Unpublished

Ph.D., New York University, United States -- New York.

Matti, T., Erkki, S., Esko, K., & Piet, K. (2006). Ethnocomputing: ICT in cultural and social

context. Association for Computing Machinery. Communications of the ACM, 49(1),

126.

Matwyshyn, A. M. (2004). Silicon ceilings: The gendered digital production divide and

single-sex technology education. A study of computer attitudes of girls in single-sex

and coeducational high schools. Unpublished Ph.D., Northwestern University, United

States -- Illinois.

Maupin, D. L. (2003). A comparison of demographics, motivation, and learning strategies of

college students taking traditional campus-based courses and Internet-based distance

learning courses. Unpublished Ed.D., Florida Atlantic University, United States --

Florida.

219

Mayall, H. J. (2002). An exploratory/descriptive look at gender differences in technology

self-efficacy and academic self-efficacy in the GlobalEd Project. Unpublished Ph.D.,

The University of Connecticut, United States -- Connecticut.

McFarland, D. J. (1999). The particularization of computer-efficacy and its influence on the

technology acceptance model: A field study. Unpublished Ph.D., Drexel University,

United States -- Pennsylvania.

McKay, P. F., Doverspike, D., Bowen-Hilton, D., & McKay, Q. D. (2003). The Effects of

Demographic Variables and Stereotype Threat on Black/White Differences in

Cognitive Ability Test Performance. Journal of Business and Psychology, 18(1), 1.

McMaster, T., & Wastell, D. (2005). Diffusion - or delusion? Challenging an IS research

tradition. Information Technology & People, 18(4), 383.

Meng-chun, L., & Gee, S. (2006). Social Learning and Digital Divides: A Case Study of

Internet Technology Diffusion. Kyklos, 59(2), 307.

Michie, S., & Nelson, D. L. (2006). Barriers women face in information technology careers:

Self-efficacy, passion and gender biases. Women in Management Review, 21(1), 10.

Miller, E. M. (1998). Why Race Matters: Race Differences and What They Mean. The

Journal of Social, Political, and Economic Studies, 23(3), 360.

Millet, M. J. (2005). An examination of factors related to the retention of new teachers in the

Metropolitan Nashville Tennessee School System. Unpublished Ed.D., Tennessee

State University, United States -- Tennessee.

Min, G., Yan, X., & Yuecheng, Y. (2004). An Enhanced Technology Acceptance Model for

Web-Based Learning. Journal of Information Systems Education, 15(4), 365.

220

Mistry, J. J. (2005). A Conceptual Framework for the Role of Government in Bridging the

Digital Divide. Journal of Global Information Technology Management, 8(3), 28.

Moore, G. A. (1999). Crossing the Chasm: Marketing and selling high tech products to

mainstream customers (2 ed.): Harper Business.

Moore, P. B. (2001). Access and success in Web courses at an urban multicultural

community college: The student's perspective. Unpublished Ed.D., Northern Arizona

University, United States -- Arizona.

Morrison, J. A., James, F. W., Sprecher, D. L., Khoury, P. R., & Daniels, S. R. (1999). Sex

and race differences in cardiovascular disease risk factor changes in schoolchildren,

1975-1990: The Princeton School study. American Journal of Public Health, 89(11),

1708.

Morse, S. (2000). How Long Will the 'Digital Divide' Last? Sentinel, p. A6.

Morse, S. (2001). "The New President's Urban Technology Efforts Show Little Promise So

Far". Take Pride! Community Magazine, 3.

Newberry, B. (2003). Effects of social motivation for learning and student social presence on

engagement and satisfaction in online classes. Unpublished Ph.D., University of

Kansas, United States -- Kansas.

Newhagen, J. E. (1994). Media Use and Political Efficacy: The Suburbanization of Race and

Class. Journal of the American Society for Information Science (1986-1998), 45(6),

386.

Ng-Mak, D. S., Dohrenwend, B. P., Abraido-Lanza, A. P., & Turner, J. B. (1999). A further

analysis of race differences in the National Longitudinal Mortality Study. American

Journal of Public Health, 89(11), 1748.

221

Nievar, M. A. (2004). Parental efficacy, attachment, and child self-control: A comparison of

processes among demographic groups. Unpublished Ph.D., Michigan State

University, United States -- Michigan.

Nigel, F., David, M., & Nicola, M. (2001). The role of individual differences in Internet

searching: An empirical study. Journal of the American Society for Information

Science and Technology, 52(12), 1049.

Noureddine, S. N. (2001). Self cognitions as motivational predictors of healthy eating.

Unpublished Ph.D., University of Michigan, United States -- Michigan.

O'Neil, D. V., & Baker, P. M. A. (2003). The role of institutional motivations in

technological adoption: Implementation of DeKalb County's Family Technology

Resource Centers. Information Society, 19(4), 305.

Oberman, P. S. (2002). Academic help-seeking in the high school computer science

classroom: Relationship to motivation, achievement, gender, and ethnicity.

Unpublished Ph.D., Emory University, United States -- Georgia.

Ohayia, C. A. (2005). The underrepresentation of African Americans in the information

technology professions: An examination of contributing factors. Unpublished Ph.D.,

Capella University, United States -- Minnesota.

Okech, A. P. (2004). An exploratory examination of the relationships among emotional

intelligence, elementary school science teacher self-efficacy, length of teaching

experience, race/ethnicity, gender, and age. Unpublished Ed.D., Texas A&M

University - Kingsville, United States -- Texas.

Okoli, C., & Mbarika, V. A. W. (2003). A framework for assessing e-commerce in Sub-

Saharan Africa. Journal of Global Information Technology Management, 6(3), 44.

222

Ones, D. S., & Viswesvaran, C. (1998). Gender, age, and race differences on overt integrity

tests: Results across four large-scale job applicant data sets. Journal of Applied

Psychology, 83(1), 35.

Ooi, J. (2005). The next warfront. Malaysian Business, 11.

Oranit, B. D., & Dov, E. (2000). Remedial self-fulfilling prophecy: Two field experiments to

prevent Golem effects among disadvantaged women. Journal of Applied Psychology,

85(3), 386.

Park, B. (2003). Faculty adoption and utilization of Web-assisted instruction (WAI) in higher

education: Structural equation modeling (SEM). Unpublished Ph.D., The Florida

State University, United States -- Florida.

Park, C. W., & Lessig, V. P. (1977). Students and Housewives- Differences in Susceptibility

to Reference Group Influence. Journal of Consumer Research, 4(2), 102 - 110.

Park, H. (2004). Factors that affect information technology adoption by teachers.

Unpublished Ph.D., The University of Nebraska - Lincoln, United States -- Nebraska.

Paulsen, A. M., & Betz, N. E. (2004). Basic Confidence Predictors of Career Decision-

Making Self-Efficacy. The Career Development Quarterly, 52(4), 354.

Pavlou, P. A., & Fygenson, M. (2006). Understanding and Predicting Electronic Commerce

Adoption: An Extension of the Theory of Planned Behavior. MIS Quarterly, 30(1),

115.

Payton, F. C. (2003). Rethinking the digital divide. Association for Computing Machinery.

Communications of the ACM, 46(6), 89.

223

Pearson, T. R. (2003). Selfish girls: Perceived understanding of gender, race, and class in

discussions about technology with pre-adolescent, low-income, African American

girls. Unpublished Ph.D., University of Florida, United States -- Florida.

Petterson, S. M. (1997). Are young black men really less willing to work? American

Sociological Review, 62(4), 605.

Phelps, D. C. (2005). Information system security: Self-efficacy and security effectiveness in

Florida libraries. Unpublished Ph.D., The Florida State University, United States --

Florida.

Phinney, J. S. (1992). The Multigroup Ethnic Identity Measure: A New Scale for Use With

Diverse Groups. Journal of Adolescent Research, 7(2), 156-176.

Piccoli, G., Rami, A., & Ives, B. (2001). Web-based virtual learning environments: A

research framework and a preliminary assessment of effectiveness in basic IT skills

training. MIS Quarterly, 25(4), 401.

Ponders, V. E. (2001). The influence of teacher characteristics and preparation on student

achievement in publicly funded Detroit schools. Unpublished Ph.D., University of

Michigan, United States -- Michigan.

Porter, D. M., Jr. (1997). The eye of the beholder: The impact of race and gender on

manager's attributions and conceptions of commitment to the organization.

Unpublished Ph.D., Harvard University, United States -- Massachusetts.

President, T. (2005). A study of a community technology center in Brooklyn: An exploration

of the contemporary digital divide. Unpublished Ed.D., Columbia University

Teachers College, United States -- New York.

224

Purnendu, M. (2004). Inter-country Analysis of E-Business. Journal of Global Information

Technology Management, 7(2), 1.

Putrevu, S. (2001). Exploring the Origins and Information Processing Differences Between

Men and Women: Implications for Advertisers. Academy of Marketing Science

Review, 2001, 1.

"Race difference in attitudinal and motivational factors in. (1993). Social Security Bulletin,

56(3), 108.

Rachel Gali, C. (2006). Anticipated Work-Family Conflict: Effects of Gender, Self-Efficacy,

and Family Background. The Career Development Quarterly, 54(3), 202.

Ragnar, A. (2006). Technology and Social Inclusion. Rethinking the Digital Divide. Journal

of Documentation, 62(2), 293.

Ramsey, R. (2003). Digital solutions to the economic divide. Journal of Housing and

Community Development, 60(2), 7.

Randel, A. E. (1999). Contextual antecedents and performance-based consequences of

identity salience in organizational project groups. Unpublished Ph.D., University of

California, Irvine, United States -- California.

Reed, K. L. (1998). New age technology and new "aged" workers: The impact of age on

computer technology skill acquisition and the influence of computer self-efficacy,

age-related beliefs, and change attitudes. Unpublished Ph.D., The University of

Nebraska - Lincoln, United States -- Nebraska.

Reynolds, N. A. (2003). New media in old buildings: A critical ethnography of urban

secondary students' interaction with computers and Internet access. Unpublished

Ed.D., The George Washington University, United States -- District of Columbia.

225

Rice, R. E., & Katz, J. E. (2003). Comparing internet and mobile phone usage: Digital

divides of usage, adoption and dropouts. Telecommunications Policy, 27(8,9), 597.

Ridzuan, A. (2001). Computimes (Malaysia): Eliminating global digital divide. Computimes

Malaysia, 1.

Riggins, F. J. (2004). A Multichannel Model of Separating Equilibrium in the Face of the

Digital Divide. Journal of Management Information Systems, 21(2), 161 - 179.

Rivas, S. (2004). The United States digital divide: Race, SES, and social context as

predictors of computer ownership, 1997--2001. Unpublished Ph.D., University of

Michigan, United States -- Michigan.

Roach, R. (2002a). Information Technology Group to address issues of equity. Black Issues

in Higher Education, 19(6), 47.

Roach, R. (2002b). Law professor explores digital divide, race in new book. Black Issues in

Higher Education, 18(23), 35.

Roach, R. (2002c). Tech Talk: Law Professor Explores Digital Divide, Race in New Book.

Black Issues in Higher Education, 18(23), 35.

Roach, R. (2003a). Report: Digital Divide Rooted in Home Computer Ownership. Black

Issues in Higher Education, 20(20), 50.

Roach, R. (2003b). Study analyzes efforts to bring technology to distressed areas. Black

Issues in Higher Education, 19(24), 31.

Robertson, R. L. (2001). An urban perspective to technology. Michigan Chronicle, p. A6.

Roblin, C. (2005). Crossfire: Gender, cyberfeminism and World Wide Web visual culture.

Unpublished M.A., Concordia University (Canada), Canada.

226

Rodriguez, H. R. (2005). Social, cognitive, and attitudinal influences on birth control use

among adolescents. Unpublished Ph.D., Purdue University, United States -- Indiana.

Rogers, B. L. (2004). Development of the Internet Experience Scale: Investigating the

relationships between cognitive, affective, and conative components of Internet

experience. Unpublished Ph.D., Wichita State University, United States -- Kansas.

Rogers, M., & Oder, N. (1999). Feds focus on "Digital divide" between whites and

minorities. Library Journal, 124(13), 14.

Ross, C. (2005). Experiences and expectations: What prompts an educator to use computers

in the classroom? Unpublished M.A., Wilfrid Laurier University (Canada), Canada.

Ruggles, S. (1994). The origins of African-American family structure. American Sociological

Review, 59(1), 136.

Ryan, S. E. (2003). Don't trust anyone outside your pack: Initial trust formation in an online

social activist network. Unpublished M.A., University of Toronto (Canada), Canada.

Saleem, H. (2005). Antecedents of computer self-efficacy and behavioral intention to use new

IT: An investigation of the role of personality traits. Unpublished M.Sc., Concordia

University (Canada), Canada.

Sampson, E. E., & Kardush, M. (1965). Age, sex, class, and race differences in response to a

two-person non-zero-sum game1. The Journal of Conflict Resolution (pre-1986),

9(2), 212.

Sangaran, S. (2002). Training to use technology well. Computimes Malaysia, 1.

Sanjeev, D., Dale, G., & Kenneth, L. K. (2005). Across the Digital Divide: A Cross-Country

Multi-Technology Analysis of the Determinants of IT Penetration. Journal of the

Association for Information Systems, 6(12), 1.

227

Sanna, T. (2005). The social and discursive construction of computing skills. Journal of the

American Society for Information Science and Technology, 56(1), 13.

Sarkar, M. (2005). Empirical studies on the diffusion and valuation of the Internet.

Unpublished Ph.D., Yale University, United States -- Connecticut.

Saunders, J., Davis, L., Williams, T., & Williams, J. H. (2004). Gender differences in self-

perceptions and academic outcomes: A study of African American high school

students. Journal of Youth and Adolescence, 33(1), 81.

Schaub, M. (2003). Social cognitive career theory: Examining the mediating role of

sociocognitive variables in the relation of personality to vocational interests.

Unpublished Ph.D., The University of Akron, United States -- Ohio.

Scherer, R. F., Brodzinski, J. D., & Wiebe, F. A. (1990). Entrepreneur Career Selection and

Gender: A Socialization Approach. Journal of Small Business Management, 28(2),

37.

Schreckenberg, S. (2004). The digital divide: A study of the intra-ethnic divide within the

African American population in Johnson City, Tennessee. Unpublished M.S., East

Tennessee State University, United States -- Tennessee.

Schroeder, J. C. (2001). The effects of self-efficacy, outcome expectancy, ease of use, and

usefulness on job performance: Evaluating the role of information technology.

Unpublished M.A., Stephen F. Austin State University, United States -- Texas.

Schwoerer, C. E., & May, D. R. (1996). Age and work outcomes: the moderating effects of

self-efficacy and tool design effectiveness. Journal of Organizational Behavior

(1986-1998), 7(5), 469.

228

Scott, S. G. (1997). Social identification effects in products and process development teams.

Journal of Engineering and Technology Management, 14(2), 97.

Seede, G. K. (2002). A case study of the adoption and use of video-mail in an elementary

school setting. Unpublished Ph.D., The Claremont Graduate University, United States

-- California.

Sen, S. (2005). The effect of technology acceptance on post-secondary African-American

students' achievement in mathematics: A path analytic inquiry. Unpublished Ed.D.,

University of Central Florida, United States -- Florida.

Sharifah, K. (2002). Increasing use of ICT in rural areas. Computimes Malaysia, 1.

Sheng, Y. P., Pearson, J. M., & Crosby, L. (2003). Organizational culture and employees'

computer self-efficacy: An emperical study. Information Resources Management

Journal, 16(3), 42.

Sherri, L. T., Jason, C. S., & Richard, T. L. (2004). Family Factors Associated With Sixth-

Grade Adolescents' Math and Science Career Interests. The Career Development

Quarterly, 53(1), 41.

Sherry, L. C. (1998). Diffusion of the Internet within a graduate school of education.

Unpublished Ph.D., University of Colorado at Denver, United States -- Colorado.

Sherwood, S. (2001). Digital divide & conquer. Chief Executive(167), 11.

Silva, R. D. (2002). Tackling digital divide (HL). Computimes Malaysia, 1.

Smith, G. (2001). Closing the digital divide. Information World Review(172), 10.

Smith, S. M. (2002). Using the social cognitive model to explain vocational interest in

information technology. Information Technology, Learning, and Performance

Journal, 20(1), 1.

229

Smith, S. M. (2004). Software Skills Acquisition: Confidence Vs. Competence. Information

Technology, Learning, and Performance Journal, 22(2), 33.

Smith, S. M. (2005). The Digital Divide: Gender And Racial Differences In Information

Technology Education. Information Technology, Learning, and Performance

Journal, 23(1), 13.

Sockel, H. M. (2000). The impact of technological innovation on the information systems'

software knowledge workers. Unpublished D.B.A., Cleveland State University,

United States -- Ohio.

Solomon, B. S. (2000). Computer assisted technology transfer: Understanding technology

diffusion in Oklahoma small-to-medium sized manufacturers. Unpublished Ed.D.,

Oklahoma State University, United States -- Oklahoma.

Somerson, P. (2000). Great divide. Ziff Davis Smart Business for the New Economy, 13(5),

35.

Songpol, K. (2004). An investigation of factors predicting consumer adoption of high-

technology innovations: The integrative roles of cognition and affect. Unpublished

Ph.D., Southern Illinois University at Carbondale, United States -- Illinois.

Sousa, K. J. (2003). Factors influencing the adoption of electronic business in the purchasing

activities within a business organization using an extended technology acceptance

model. Unpublished Ph.D., University of Rhode Island, United States -- Rhode

Island.

Stanley, L. D. (2003). Beyond Access: Psychosocial Barriers to Computer Literacy.

Information Society, 19(5), 407.

230

Stewart, T. (2002). The relationship between Myers-Briggs Type profiles and key factors that

affect academic program completion. Unpublished Ed.D., University of

Massachusetts Lowell, United States -- Massachusetts.

Strong, D. M., Dishaw, M. T., & Bandy, B. D. (2006). Extending Task Technology Fit with

Computer Self-Efficacy. Database for Advances in Information Systems, 37(2/3), 96.

Study looks at digital divide. (2000). Take Pride! Community Magazine, 1.

Subbiah, A. (2003). Information for research in developing countries: Information

technology - friend or foe? Bulletin of the American Society for Information Science

and Technology, 29(5), 16.

Sushil, K. S., & Jatinder, N. D. G. (2003). Socio-economic influences of e-commerce

adoption. Journal of Global Information Technology Management, 6(3), 3.

Syed, H. A. (2003). Digital divide and purchase intention: Why demographic psychology

matters. Journal of Economic Psychology, 24(3), 321.

Tansley, D. (2006). Mind the Gap: 2006 will witness the deepening of the digital divide.

Financial Times, p. 2.

Taylor, B. G. (1997). Design, development and evaluation of Web-based motivational

strategies training: Individual and situational aspects of academic motivation.

Unpublished Ph.D., Georgia State University, United States -- Georgia.

Techatassanasoontorn, A. A. (2006). The state-based and regional contagion theories of

technology diffusion. Unpublished Ph.D., University of Minnesota, United States --

Minnesota.

Technology gap may add to instability. (2002). EE Times, 18.

231

Terrell-Simmons, V. S. (2002). Bridging the African American digital divide: An analysis of

computer and Internet access and usage trends among African American graduates

of digital divide programs in New Jersey. Unpublished Ph.D., Capella University,

United States -- Minnesota.

Tharp, A. M. (2002). Career development of women engineers: The role of self-efficacy and

supports-barriers. Unpublished Ph.D., University of Minnesota, United States --

Minnesota.

Thatcher, S. M. B. (2000). Do you really know me? The implications of identity fit for diverse

workteams. Unpublished Ph.D., University of Pennsylvania, United States --

Pennsylvania.

Thomas, K. M., & Wise, G. P. (1999). Organizational Attractiveness and Individual

Differences: Are Diverse Applicants Attracted by Different Factors? Journal of

Business and Psychology, 13(3), 375.

Thompson, D. E., & Thompson, T. A. (1985). Task-Based Performance Appraisal for Blue-

Collar Jobs: Evaluation of Race and Sex Effects. Journal of Applied Psychology,

70(4), 747.

Thompson, R., Compeau, D., & Higgins, C. (2006). Intentions to Use Information

Technologies: An Integrative Model. Journal of Organizational and End User

Computing, 18(3), 25.

Townsend, A. M. (1993). Determinants of Goal Commitment in an Incentive Paid

Workforce.

Tremblay, D. R. (2000). Resistance to information technology: A psychosocial explanation.

Unpublished Ph.D., Walden University, United States -- Minnesota.

232

Tresemer, D. (1974). FEAR OF SUCCESS - POPULAR, BUT UNPROVEN. Psychology

Today, 7(10), 82.

Twenge, J. M., & Crocker, J. (2002). Race and self-esteem: Meta-analyses comparing

Whites, African-Americans, Hispanics, Asians, and American Indians and comment

on Gray-Little and Hafdahl (2000). Psychological Bulletin, 128(3), 371.

Venkatesh, V. (1998). User acceptance of information technology: A unified view.

Unpublished Ph.D., University of Minnesota, United States -- Minnesota.

Venkatesh, V., & Brown, S. A. (2001). A longitudinal investigation of personal computers in

homes: Adoption determinants and emerging challenges. MIS Quarterly, 25(1), 71.

Venkatesh, V., & Davis, F. D. (1996). A model of the antecedents of perceived ease of use:

Development and test. Decision Sciences, 27(3), 451.

Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User Acceptance Of

Information Technology: Toward A Unified View. MIS Quarterly, 27(3), 425-478.

Veum, J. R. (1996). Gender and race differences in company training. Industrial Relations,

35(1), 32.

VilaCruz, G. G. (2003). Barrier perceptions, coping efficacy, and other career-related

factors within a community college population: A social cognitive career theory

perspective. Unpublished Ph.D., The University of Wisconsin - Madison, United

States -- Wisconsin.

Viswanath, V., Michael, G. M., & Phillip, L. A. (2000). A longitudinal field investigation of

gender differences in individual technology adoption decision-making processes.

Organizational Behavior and Human Decision Processes, 83(1), 33.

233

Wainer, H., & Brown, L. M. (2004). Two Statistical Paradoxes in the Interpretation of Group

Differences: Illustrated with Medical School Admission and Licensing Data. The

American Statistician, 58(2), 117.

Wakefield, R. L., & Whitten, D. (2006). Mobile computing: a user study on

hedonic/utilitarian mobile device usage. European Journal of Information Systems,

15(3), 292.

Wang, S.-Y. M. (1998). Facilitating the application of the technology acceptance model: An

experimental evaluation of alternative protocols for technology introduction.

Unpublished Ph.D., University of South Carolina, United States -- South Carolina.

Wareham, J., Levy, A., & Shi, W. (2004). Wireless diffusion and mobile computing:

Implications for the digital divide. Telecommunications Policy, 28(5,6), 439.

Waskiewicz, M. V. K. (2002). The mediating role of sense of community in the relationship

among principals' characteristics, district characteristics, and self-efficacy beliefs of

public school principals in Florida. Unpublished Ph.D., University of Florida, United

States -- Florida.

Wasonga, T. A. (2002). An analysis of the relationship between external assets, internal

assets and academic achievement among urban students. Unpublished Ed.D.,

University of Missouri - Columbia, United States -- Missouri.

Weaver, N. L. (2004). Injury as a barrier to physical activity and the moderating effects of

intrinsic motivation, self-efficacy, race and gender. Unpublished Ph.D., The

University of North Carolina at Chapel Hill, United States -- North Carolina.

West, C. (1993). Race Matters.

234

Wheeler, D. L. (2006). Gender sensitivity and the drive for IT: Lessons from the NetCorps

Jordan Project. Ethics and Information Technology, 8(3), 131.

Woolford-Singh, A. (2004). An examination of relationships of reading and writing self-

efficacy beliefs, standardized placement test scores, and diverse community college

students' perceptions of those relationships. Unpublished Ph.D., Old Dominion

University, United States -- Virginia.

Yap, A., Das, J., Burbridge, J., & Cort, K. (2006). A Composite-Model for E-Commerce

Diffusion: Integrating Cultural and Socio-Economic Dimensions to the Dynamics of

Diffusion. Journal of Global Information Management, 14(3), 17.

Yi-Shun, W., Hsin-Hui, L., & Pin, L. (2006). Predicting consumer intention to use mobile

service. Information Systems Journal, 16(2), 157.

Yosso, T. J. (2000). A critical race and LatCrit approach to media literacy: Chicana/o

resistance to visual microaggressions. Unpublished Ph.D., University of California,

Los Angeles, United States -- California.

Younghwa, L., Jintae, L., & Zoonky, L. (2006). Social Influence on Technology Acceptance

Behavior: Self-Identity Theory Perspective. Database for Advances in Information

Systems, 37(2/3), 60.

Zealand, R. A. (2004). Relationships among achievement, perceptions of control, self-

regulation, and self-determination of students with and without the classification of

learning disabilities. Unpublished Ph.D., Columbia University, United States -- New

York.

235

Zhu, J. (2003). Application of computer technology in public school classrooms: Usage

dimensions and influencing factors. Unpublished Ph.D., The Pennsylvania State

University, United States -- Pennsylvania.

Zollinhofer, M. H. (1998). Boon or bane: The role of computing technologies on the

professional practice of higher education faculty. Unpublished Ph.D., University of

Virginia, United States -- Virginia.

236

APPENDIX A: CHAPTER 1 DIGITAL DIVIDE GRAPHS

Figure 48: Households with Internet Access by Race/Hispanic Origin

Figure 49: Tare of Growth of Internet Penetration by Race/Hispanic Origin

237

Figure 50: Income and Education Differences Account for Half of the Gap between Blacks and Hispanics and National Average

Figure 51: Households with a Computer by Race/Hispanic Origin

238

Figure 52: Internet Use by Race/Ethnicity

Figure 53: Household Access Rates by Race/Ethnicity Do not Closely Track Internet Use by Persons

239

Figure 54: Internet Use by Location and Race/Ethnicity

Figure 55: Percent of US Households with a Computer by Race/Hispanic Origin

240

Figure 56: Households with a Computer by Income by Race/Hispanic Origin

Figure 57: Households with Internet Access by Race/Hispanic Origin

241

Figure 58: Households with Internet Access

Figure 59: Persons Using the Internet by Race/Hispanic Origin

242

Figure 60: Persons Using the Internet at Home by Race/Hispanic Origin

Figure 61: Persons Using the Internet Outside the Home by Race/Hispanic Origin

243

Figure 62: Persons Using the Internet Outside the Home by Race/Hispanic Origin

Figure 63: Persons Using the Internet Outside the Home at Schools

244

Figure 64:Persons Using the Internet Outside the Home at Work

Figure 65: Reasons for Household with a Computer/WebTV Not Using the Internet at Home by Race/Hispanic Origin

245

Figure 66: Households with Home Internet Access by Race/Hispanic Origin

Figure 67: Rate of Growth of Internet Penetration by Race/Hispanic Origin

246

Figure 68: Internet Access at Home by Race/Ethnicity and Disability Status

Figure 69: Regularly Uses a PC by Race/Ethnicity and Disability Status

247

APPENDIX B: ADDITIONAL CHAPTER 4 ANALYSIS OF DATA RESULTS

e13 e14 e15 e1 e2 e3 1 1 1 e16 e17 1 1 1 1 1 EoU.8 EoU.9 EoU.10 TE.I TE.II TE.III 1 Int.51 Int.52 1 1 EaseOfUse TraitEff Intent to Use

StateEff SymbUtil FuncUtil 1 1 1

SE.I SE.II SE.III SU.III SU.II SU.I FU.II FU.III FU.I 1 1 1 1 1 1 1 1 1

e9 e8 e7 e6 e5 e4 e11 e10 e12

Figure 70: Single-Group Measurement Model

248

.37 .17 .17 3.34 2.97 2.45 e13 e14 e15 .08 .23 e1 e2 e3 1 1 1 e16 e17 1 1 1 1 1 EoU.8 EoU.9 EoU.10 TE.I TE.II TE.III 1.00 1.281.40 Int.51 Int.52 1.00 1.271.45 .23 4.06 1.00 .89 1.92 EaseOfUse TraitEff Intent to Use

.12 -.14 .21 1.37 -1.30 .64 .26

.78 .04 2.92

-1.07 .56 -.25 3.48 2.41

8.73 2.90 16.60

StateEff SymbUtil FuncUtil 1.56 1.00 .51 1.00 .98 1.07 .99 1.01 1.00

SE.I SE.II SE.III SU.III SU.II SU.I FU.II FU.III FU.I 1 1 1 1 1 1 1 1 1 6.48 1.73 1.89 .40 .44 .54 2.96 3.42 -.53 e9 e8 e7 e6 e5 e4 e11 e10 e12

Figure 71: Unstandardized Results of Original Measurement Model showing negative error variance

249

e13 e14 e15 e1 e2 e3 .39 .69 .74 e16 e17 .55 .69 .78 EoU.8 EoU.9 EoU.10 .96 .87 TE.I TE.II TE.III .62 .83 .86 Int.51 Int.52 .74 .83 .88 .98 .93 EaseOfUse TraitEff Intent to Use

.18 .22 -.05 .45 .33 -.16 .13

.13 .05 .52

-.31 .24 -.05 .50 .20

StateEff SymbUtil FuncUtil .88 .91 .80 .92 .92 1.01 .91 .93 .91 .77 .83 .82 .65 .87 .84 .84 .83 1.03 SE.I SE.II SE.III SU.III SU.II SU.I FU.II FU.III FU.I

e9 e8 e7 e6 e5 e4 e11 e10 e12

Figure 72: Unstandardized results of Original Measurement Model showing FU.I is too-strongly correlated with the other indicators

Table 75: Covariances for Original Measurement Model Estimate S.E. C.R. P Label TraitEff <--> StateEff .782 .419 1.867 .062 par_11 SymbUtil <--> StateEff -.252 .340 -.740 .459 par_12 StateEff <--> FuncUtil 2.774 .872 3.182 .001 par_13 SymbUtil <--> FuncUtil 3.820 .555 6.879 <.001 par_14 EaseOfUse <--> Intent to Use .123 .048 2.569 .010 par_15 FuncUtil <--> Intent to Use 3.453 .449 7.688 <.001 par_16 SymbUtil <--> Intent to Use .558 .160 3.479 <.001 par_17 TraitEff <--> EaseOfUse .209 .075 2.800 .005 par_18 StateEff <--> EaseOfUse .639 .121 5.262 <.001 par_19 SymbUtil <--> EaseOfUse .042 .058 .722 .470 par_20 FuncUtil <--> EaseOfUse .330 .149 2.211 .027 par_21 TraitEff <--> FuncUtil -1.114 .605 -1.839 .066 par_22 TraitEff <--> SymbUtil -1.070 .257 -4.169 <.001 par_23 StateEff <--> Intent to Use 1.370 .286 4.793 <.001 par_24 TraitEff <--> Intent to Use -.135 .190 -.708 .479 par_25

250

Table 76: Correlations for Original Measurement Model Estimate TraitEff <--> StateEff .131 SymbUtil <--> StateEff -.050 StateEff <--> FuncUtil .219 SymbUtil <--> FuncUtil .522 EaseOfUse <--> Intent to Use .183 FuncUtil <--> Intent to Use .579 SymbUtil <--> Intent to Use .234 TraitEff <--> EaseOfUse .215 StateEff <--> EaseOfUse .448 SymbUtil <--> EaseOfUse .051 FuncUtil <--> EaseOfUse .159 TraitEff <--> FuncUtil -.129 TraitEff <--> SymbUtil -.311 StateEff <--> Intent to Use .333 TraitEff <--> Intent to Use -.048

Table 77: Variances for Original Measurement Model Estimate S.E. C.R. P Label TraitEff 4.059 .621 6.538 <.001 par_26 SymbUtil 2.921 .310 9.419 <.001 par_27 StateEff 8.732 .948 9.209 <.001 par_28 FuncUtil 18.373 1.949 9.428 <.001 par_29 EaseOfUse .233 .045 5.154 <.001 par_30 Intent to Use 1.938 .190 10.182 <.001 par_31 E1 3.336 .363 9.178 <.001 par_32 E2 2.983 .427 6.988 <.001 par_33 E3 2.435 .483 5.037 <.001 par_34 E4 .521 .089 5.885 <.001 par_35 E5 .458 .086 5.353 <.001 par_36 E6 .404 .041 9.780 <.001 par_37 E7 1.882 .273 6.902 <.001 par_38 E8 1.737 .262 6.620 <.001 par_39 E9 6.472 .782 8.277 <.001 par_40 E10 1.642 .843 1.949 .051 par_41 E11 5.035 .763 6.596 <.001 par_42 E13 .367 .037 10.047 <.001 par_43 E14 .173 .029 6.006 <.001 par_44 E15 .165 .033 5.050 <.001 par_45

251

Estimate S.E. C.R. P Label E16 .057 .072 .793 .428 par_46 E17 .247 .060 4.111 <.001 par_47

Table 78: Squared Multiple Correlations for Original Measurement Model Estimate Int.52 .859 Int.51 .971 EoU.10 .737 EoU.9 .689 EoU.8 .388 FU.II .732 FU.III .918 SE.I .767 SE.II .832 SE.III .823 SU.III .648 SU.II .864 SU.I .849 TE.III .778 TE.II .688 TE.I .549

252

e1 e2 e3 e13 e14 e15 1 1 1 1 1 1

TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 1 1 e16 e17 TraitEff EaseOfUse 1 1 1 Int.51 Int.52 res1 1 e10 e11 StateEff 1 1 Intent to Use 1 FU.III FU.II 1 1 SE.I SE.II SE.III 1 1 1 res2 FuncUtil e9 e8 e7

SymbUtil 1

SU.III SU.II SU.I 1 1 1

e6 e5 e4

Figure 73: Single-Group Causal Model

253

3.34 2.99 2.40 .37 .16 .18 e1 e2 e3 e13 e14 e15 1 1 1 1 1 1

TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 1.30 1.00 1.271.46 1.00 1.39 .05 .25 4.05 e16 e17 .04 TraitEff EaseOfUse 1 1 1 .18 .07 .30 Int.51 Int.52 1.71 5.03 1.00 .88 .788.76 res1 e10 e11 StateEff -1.111 1 Intent to Use 1.56 1.00 -1.06 .99 2.93 FU.III FU.II .20 1 1.00 .87 1.25 SE.I SE.II SE.III -.24 18.31 -.08 1 1 1 res2 6.43 1.77 1.86 FuncUtil e9 e8 e7 3.83 2.92

SymbUtil .50 1.00 1.00

SU.III SU.II SU.I 1 1 1 .40 .46 .52 e6 e5 e4

Figure 74: Single-Group Unstandardized Results

254

e1 e2 e3 e13 e14 e15

.55 .69 .78 .39 .71 .72 TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 .62 .84 .85 .74 .83 .88 .23 e16 e17 .16 EaseOfUse TraitEff .97 .86

.43 .11 Int.51 Int.52 .13 .99 .93 res1 .35 e10 e11

StateEff -.13 .91 .73 Intent to Use .88 .91 -.31 .91 .23 FU.III FU.II .77 .83 .82 .62 .96 .86 SE.I SE.II SE.III -.05 -.09 res2 FuncUtil e9 e8 e7 .52

SymbUtil .80 .92 .93 .65 .86 .85 SU.III SU.II SU.I

e6 e5 e4

Figure 75: Single-Group Standardized Results

Table 79: Regression Weights for Single-Group Causal Model Estimate S.E. C.R. P Label EaseOfUse <--- TraitEff .038 .016 2.296 .022 par_11 EaseOfUse <--- StateEff .070 .012 5.781 <.001 par_12 Intent to Use <--- EaseOfUse .303 .168 1.808 .071 par_13 Intent to Use <--- FuncUtil .201 .023 8.560 <.001 par_14 Intent to Use <--- SymbUtil -.076 .055 -1.389 .165 par_15 TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.273 .101 12.557 <.001 par_1 TE.III <--- TraitEff 1.455 .114 12.778 <.001 par_2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil .999 .043 23.097 <.001 par_3 SU.III <--- SymbUtil .504 .029 17.643 <.001 par_4 SE.III <--- StateEff 1.000 SE.II <--- StateEff .990 .045 21.875 <.001 par_5 SE.I <--- StateEff 1.560 .076 20.420 <.001 par_6

255

Estimate S.E. C.R. P Label FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .866 .052 16.806 <.001 par_7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.298 .132 9.821 <.001 par_8 EoU.10 <--- EaseOfUse 1.391 .142 9.825 <.001 par_9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .880 .043 20.684 <.001 par_10

Table 80: Standardized Regression Weights for Single-Group Causal Model Estimate EaseOfUse <--- TraitEff .157 EaseOfUse <--- StateEff .431 Intent to Use <--- EaseOfUse .105 Intent to Use <--- FuncUtil .617 Intent to Use <--- SymbUtil -.093 TE.I <--- TraitEff .740 TE.II <--- TraitEff .829 TE.III <--- TraitEff .884 SU.I <--- SymbUtil .922 SU.II <--- SymbUtil .929 SU.III <--- SymbUtil .805 SE.III <--- StateEff .908 SE.II <--- StateEff .910 SE.I <--- StateEff .876 FU.III <--- FuncUtil .956 FU.II <--- FuncUtil .856 EoU.8 <--- EaseOfUse .623 EoU.9 <--- EaseOfUse .840 EoU.10 <--- EaseOfUse .848 Int.51 <--- Intent to Use .987 Int.52 <--- Intent to Use .925

Table 81:Covariances for Single-Group Causal Model Estimate S.E. C.R. P Label SymbUtil <--> FuncUtil 3.830 .556 6.894 <.001 par_16 TraitEff <--> StateEff .778 .419 1.857 .063 par_17 TraitEff <--> FuncUtil -1.109 .603 -1.838 .066 par_18 SymbUtil <--> StateEff -.237 .340 -.696 .486 par_19 StateEff <--> FuncUtil 2.926 .874 3.349 <.001 par_20

256

Estimate S.E. C.R. P Label TraitEff <--> SymbUtil -1.057 .256 -4.128 <.001 par_21

Table 82:Correlations for Single-Group Causal Model Estimate SymbUtil <--> FuncUtil .524 TraitEff <--> StateEff .131 TraitEff <--> FuncUtil -.129 SymbUtil <--> StateEff -.047 StateEff <--> FuncUtil .231 TraitEff <--> SymbUtil -.307

Table 83: Variances for Single-Group Causal Model Estimate S.E. C.R. P Label TraitEff 4.052 .621 6.530 <.001 par_22 SymbUtil 2.924 .310 9.428 <.001 par_23 StateEff 8.755 .949 9.225 <.001 par_24 FuncUtil 18.307 1.933 9.472 <.001 par_25 res1 .179 .036 5.055 <.001 par_26 res2 1.251 .140 8.928 <.001 par_27 e1 3.343 .364 9.184 <.001 par_28 e2 2.995 .428 6.992 <.001 par_29 e3 2.403 .486 4.947 <.001 par_30 e4 .518 .089 5.844 <.001 par_31 e5 .462 .086 5.388 <.001 par_32 e6 .404 .041 9.779 <.001 par_33 e7 1.859 .273 6.812 <.001 par_34 e8 1.774 .265 6.697 <.001 par_35 e9 6.428 .781 8.228 <.001 par_36 e10 1.709 .807 2.117 .034 par_37 e11 5.028 .744 6.762 <.001 par_38 e13 .368 .037 10.042 <.001 par_39 e14 .164 .029 5.673 <.001 par_40 e15 .176 .033 5.387 <.001 par_41 e16 .052 .077 .672 .502 par_42 e17 .251 .064 3.934 <.001 par_43

Table 84: Squared Multiple Correlations for Single-Group Causal Model Estimate

257

Estimate EaseOfUse .228 Intent to Use .352 Int.52 .856 Int.51 .974 EoU.10 .719 EoU.9 .705 EoU.8 .388 FU.II .732 FU.III .915 SE.I .768 SE.II .829 SE.III .825 SU.III .648 SU.II .863 SU.I .850 TE.III .781 TE.II .687 TE.I .548

Table 85:CMIN for Single-Group Causal Model Model NPAR CMIN DF P CMIN/DF Default model 43 165.822 93 .000 1.783 Saturated model 136 .000 0 Independence model 16 2982.847 120 .000 24.857

Table 86: RMR, GFI for Single-Group Causal Model Model RMR GFI AGFI PGFI Default model .423 .926 .891 .633 Saturated model .000 1.000 Independence model 2.822 .377 .294 .333

Table 87: Baseline Comparisons for Single-Group Causal Model NFI RFI IFI TLI Model CFI Delta1 rho1 Delta2 rho2 Default model .944 .928 .975 .967 .975 Saturated model 1.000 1.000 1.000 Independence model .000 .000 .000 .000 .000

Table 88: Parsimony-Adjusted Measures for Single-Group Causal Model Model PRATIO PNFI PCFI Default model .775 .732 .755

258

Model PRATIO PNFI PCFI Saturated model .000 .000 .000 Independence model 1.000 .000 .000

Table 89: NCP for Single-Group Causal Model Model NCP LO 90 HI 90 Default model 72.822 40.752 112.740 Saturated model .000 .000 .000 Independence model 2862.847 2688.591 3044.424

Table 90: FMIN for Single-Group Causal Model Model FMIN F0 LO 90 HI 90 Default model .648 .284 .159 .440 Saturated model .000 .000 .000 .000 Independence model 11.652 11.183 10.502 11.892

Table 91: RMSEA for Single-Group Causal Model Model RMSEA LO 90 HI 90 PCLOSE Default model .055 .041 .069 .251 Independence model .305 .296 .315 .000

Table 92: AIC for Single-Group Causal Model Model AIC BCC BIC CAIC Default model 251.822 257.939 404.432 447.432 Saturated model 272.000 291.347 754.674 890.674 Independence model 3014.847 3017.123 3071.632 3087.632

Table 93: ECVI for Single-Group Causal Model Model ECVI LO 90 HI 90 MECVI Default model .984 .858 1.140 1.008 Saturated model 1.063 1.063 1.063 1.138 Independence model 11.777 11.096 12.486 11.786

Table 94: HOELTER for Single-Group Causal Model HOELTER HOELTER Model .05 .01 Default model 180 198 Independence model 13 14

259

e1 e2 e3 e13 e14 e15 1 1 1 1 1 1

TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 1 1 e16 e17 TraitEff EaseOfUse 1 1 1 Int.51 Int.52 res1 1 e10 e11 StateEff 1 1 Intent to Use 1 FU.III FU.II 1 1 SE.I SE.II SE.III 1 1 1 res2 FuncUtil e9 e8 e7

SymbUtil 1

SU.III SU.II SU.I 1 1 1

e6 e5 e4

Figure 76: Post Hoc Single-group Causal Model

260

3.34 3.00 2.40 .37 .17 .17 e1 e2 e3 e13 e14 e15 1 1 1 1 1 1

TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 1.30 1.00 1.271.46 1.00 1.40 .06 .25 4.05 e16 e17 .04 TraitEff EaseOfUse 1 1 1 .18 .07 .02 Int.51 Int.52 1.67 5.01 1.00 .88 .788.73 res1 e10.10 e11 StateEff -1.101 1 Intent to Use 1.56 1.00 -1.06 .99 2.81 FU.III FU.II .18 1 1.00 .87 1.20 SE.I SE.II SE.III -.23 18.35 -.04 1 1 1 res2 6.45 1.75 1.88 FuncUtil e9 e8 e7 3.83 2.92

SymbUtil .50 1.00 1.00

SU.III SU.II SU.I 1 1 1 .40 .46 .52 e6 e5 e4

Figure 77: Unstandardized Results for Post Hoc Single-group Causal Model

261

e1 e2 e3 e13 e14 e15

.55 .69 .78 .39 .70 .73 TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 .62 .84 .85 .74 .83 .88 .22 e16 e17 .16 EaseOfUse TraitEff .97 .86

.43 .01 Int.51 Int.52 .13 .99 .93 res1 .38 e10.21 e11

StateEff -.13 .92 .73 Intent to Use .88 .91 -.31 .91 .22 FU.III FU.II .77 .83 .82 .56 .96 .86 SE.I SE.II SE.III -.05 -.05 res2 FuncUtil e9 e8 e7 .52

SymbUtil .81 .92 .93 .65 .86 .85 SU.III SU.II SU.I

e6 e5 e4

Figure 78: Standardized Results for Post Hoc Single-group Causal Model

Table 95: Regression Weights for Post Hoc Single-group Causal Model Estimate S.E. C.R. P Label EaseOfUse <--- TraitEff .037 .016 2.286 .022 par_11 EaseOfUse <--- StateEff .070 .012 5.732 <.001 par_12 Intent to Use <--- EaseOfUse .017 .185 .093 .926 par_13 Intent to Use <--- FuncUtil .181 .024 7.613 <.001 par_14 Intent to Use <--- SymbUtil -.037 .054 -.691 .490 par_15 Intent to Use <--- StateEff .097 .030 3.213 .001 par_22 TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.273 .101 12.557 <.001 par_1 TE.III <--- TraitEff 1.455 .114 12.778 <.001 par_2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil .998 .043 23.075 <.001 par_3 SU.III <--- SymbUtil .505 .029 17.661 <.001 par_4

262

Estimate S.E. C.R. P Label SE.III <--- StateEff 1.000 SE.II <--- StateEff .993 .045 21.924 <.001 par_5 SE.I <--- StateEff 1.561 .077 20.378 <.001 par_6 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .866 .053 16.459 <.001 par_7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.297 .132 9.793 <.001 par_8 EoU.10 <--- EaseOfUse 1.403 .143 9.796 <.001 par_9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .882 .041 21.759 <.001 par_10

Table 96: Standardized Regression Weights for Post Hoc Single-group Causal Model Estimate EaseOfUse <--- TraitEff .156 EaseOfUse <--- StateEff .428 Intent to Use <--- EaseOfUse .006 Intent to Use <--- FuncUtil .556 Intent to Use <--- SymbUtil -.046 Intent to Use <--- StateEff .206 TE.I <--- TraitEff .740 TE.II <--- TraitEff .829 TE.III <--- TraitEff .884 SU.I <--- SymbUtil .922 SU.II <--- SymbUtil .929 SU.III <--- SymbUtil .805 SE.III <--- StateEff .907 SE.II <--- StateEff .912 SE.I <--- StateEff .876 FU.III <--- FuncUtil .957 FU.II <--- FuncUtil .856 EoU.8 <--- EaseOfUse .621 EoU.9 <--- EaseOfUse .837 EoU.10 <--- EaseOfUse .853 Int.51 <--- Intent to Use .985 Int.52 <--- Intent to Use .927

Table 97: Covariances for Post Hoc Single-group Causal Model Estimate S.E. C.R. P Label SymbUtil <--> FuncUtil 3.827 .556 6.887 <.001 par_16 TraitEff <--> StateEff .776 .418 1.856 .063 par_17

263

Estimate S.E. C.R. P Label TraitEff <--> FuncUtil -1.105 .604 -1.830 .067 par_18 SymbUtil <--> StateEff -.230 .340 -.678 .498 par_19 StateEff <--> FuncUtil 2.807 .872 3.220 .001 par_20 TraitEff <--> SymbUtil -1.057 .256 -4.129 <.001 par_21

Table 98: Correlations for Post Hoc Single-group Causal Model Estimate SymbUtil <--> FuncUtil .522 TraitEff <--> StateEff .131 TraitEff <--> FuncUtil -.128 SymbUtil <--> StateEff -.046 StateEff <--> FuncUtil .222 TraitEff <--> SymbUtil -.307

Table 99: Variances for Post Hoc Single-group Causal Model Estimate S.E. C.R. P Label TraitEff 4.052 .621 6.530 <.001 par_23 SymbUtil 2.924 .310 9.427 <.001 par_24 StateEff 8.731 .948 9.208 <.001 par_25 FuncUtil 18.346 1.948 9.417 <.001 par_26 res1 .179 .036 5.042 <.001 par_27 res2 1.199 .133 9.047 <.001 par_28 e1 3.343 .364 9.183 <.001 par_29 e2 2.995 .428 6.993 <.001 par_30 e3 2.403 .486 4.946 <.001 par_31 e4 .518 .089 5.838 <.001 par_32 e5 .464 .086 5.397 <.001 par_33 e6 .403 .041 9.770 <.001 par_34 e7 1.883 .273 6.905 <.001 par_35 e8 1.745 .263 6.643 <.001 par_36 e9 6.450 .781 8.261 <.001 par_37 e10 1.669 .842 1.983 .047 par_38 e11 5.015 .763 6.571 <.001 par_39 e13 .369 .037 10.059 <.001 par_40 e14 .166 .029 5.736 <.001 par_41 e15 .170 .033 5.179 <.001 par_42 e16 .058 .072 .802 .422 par_43 e17 .246 .060 4.097 <.001 par_44

Table 100: Squared Multiple Correlations for Post Hoc Single-group Causal Model Estimate

264

Estimate EaseOfUse .225 Intent to Use .381 Int.52 .860 Int.51 .971 EoU.10 .728 EoU.9 .700 EoU.8 .385 FU.II .733 FU.III .917 SE.I .767 SE.II .831 SE.III .823 SU.III .649 SU.II .863 SU.I .850 TE.III .781 TE.II .687 TE.I .548

Table 101: Total Effects for Post Hoc Single-group Causal Model FuncUtil StateEff SymbUtil TraitEff EaseOfUse Intent to Use EaseOfUse .000 .070 .000 .037 .000 .000 Intent to Use .181 .098 -.037 .001 .017 .000 Int.52 .159 .087 -.033 .001 .015 .882 Int.51 .181 .098 -.037 .001 .017 1.000 EoU.10 .000 .098 .000 .052 1.403 .000 EoU.9 .000 .090 .000 .048 1.297 .000 EoU.8 .000 .070 .000 .037 1.000 .000 FU.II .866 .000 .000 .000 .000 .000 FU.III 1.000 .000 .000 .000 .000 .000 SE.I .000 1.561 .000 .000 .000 .000 SE.II .000 .993 .000 .000 .000 .000 SE.III .000 1.000 .000 .000 .000 .000 SU.III .000 .000 .505 .000 .000 .000 SU.II .000 .000 .998 .000 .000 .000 SU.I .000 .000 1.000 .000 .000 .000 TE.III .000 .000 .000 1.455 .000 .000 TE.II .000 .000 .000 1.273 .000 .000 TE.I .000 .000 .000 1.000 .000 .000

265

Table 102: Standardized Total Effects for Post Hoc Single-group Causal Model Intent to FuncUtil StateEff SymbUtil TraitEff EaseOfUse Use EaseOfUse .000 .428 .000 .156 .000 .000 Intent to .556 .209 -.046 .001 .006 .000 Use Int.52 .516 .194 -.043 .001 .006 .927 Int.51 .548 .206 -.045 .001 .006 .985 EoU.10 .000 .365 .000 .133 .853 .000 EoU.9 .000 .358 .000 .131 .837 .000 EoU.8 .000 .265 .000 .097 .621 .000 FU.II .856 .000 .000 .000 .000 .000 FU.III .957 .000 .000 .000 .000 .000 SE.I .000 .876 .000 .000 .000 .000 SE.II .000 .912 .000 .000 .000 .000 SE.III .000 .907 .000 .000 .000 .000 SU.III .000 .000 .805 .000 .000 .000 SU.II .000 .000 .929 .000 .000 .000 SU.I .000 .000 .922 .000 .000 .000 TE.III .000 .000 .000 .884 .000 .000 TE.II .000 .000 .000 .829 .000 .000 TE.I .000 .000 .000 .740 .000 .000

Table 103: Direct Effects for Post Hoc Single-group Causal Model FuncUtil StateEff SymbUtil TraitEff EaseOfUse Intent to Use EaseOfUse .000 .070 .000 .037 .000 .000 Intent to Use .181 .097 -.037 .000 .017 .000 Int.52 .000 .000 .000 .000 .000 .882 Int.51 .000 .000 .000 .000 .000 1.000 EoU.10 .000 .000 .000 .000 1.403 .000 EoU.9 .000 .000 .000 .000 1.297 .000 EoU.8 .000 .000 .000 .000 1.000 .000 FU.II .866 .000 .000 .000 .000 .000 FU.III 1.000 .000 .000 .000 .000 .000 SE.I .000 1.561 .000 .000 .000 .000 SE.II .000 .993 .000 .000 .000 .000 SE.III .000 1.000 .000 .000 .000 .000 SU.III .000 .000 .505 .000 .000 .000 SU.II .000 .000 .998 .000 .000 .000 SU.I .000 .000 1.000 .000 .000 .000 TE.III .000 .000 .000 1.455 .000 .000

266

FuncUtil StateEff SymbUtil TraitEff EaseOfUse Intent to Use TE.II .000 .000 .000 1.273 .000 .000 TE.I .000 .000 .000 1.000 .000 .000

Table 104: Standardized Direct Effects for Post Hoc Single-group Causal Model FuncUtil StateEff SymbUtil TraitEff EaseOfUse Intent to Use EaseOfUse .000 .428 .000 .156 .000 .000 Intent to Use .556 .206 -.046 .000 .006 .000 Int.52 .000 .000 .000 .000 .000 .927 Int.51 .000 .000 .000 .000 .000 .985 EoU.10 .000 .000 .000 .000 .853 .000 EoU.9 .000 .000 .000 .000 .837 .000 EoU.8 .000 .000 .000 .000 .621 .000 FU.II .856 .000 .000 .000 .000 .000 FU.III .957 .000 .000 .000 .000 .000 SE.I .000 .876 .000 .000 .000 .000 SE.II .000 .912 .000 .000 .000 .000 SE.III .000 .907 .000 .000 .000 .000 SU.III .000 .000 .805 .000 .000 .000 SU.II .000 .000 .929 .000 .000 .000 SU.I .000 .000 .922 .000 .000 .000 TE.III .000 .000 .000 .884 .000 .000 TE.II .000 .000 .000 .829 .000 .000 TE.I .000 .000 .000 .740 .000 .000

Table 105: Indirect Effects for Post Hoc Single-group Causal Model FuncUtil StateEff SymbUtil TraitEff EaseOfUse Intent to Use EaseOfUse .000 .000 .000 .000 .000 .000 Intent to Use .000 .001 .000 .001 .000 .000 Int.52 .159 .087 -.033 .001 .015 .000 Int.51 .181 .098 -.037 .001 .017 .000 EoU.10 .000 .098 .000 .052 .000 .000 EoU.9 .000 .090 .000 .048 .000 .000 EoU.8 .000 .070 .000 .037 .000 .000 FU.II .000 .000 .000 .000 .000 .000 FU.III .000 .000 .000 .000 .000 .000 SE.I .000 .000 .000 .000 .000 .000 SE.II .000 .000 .000 .000 .000 .000 SE.III .000 .000 .000 .000 .000 .000 SU.III .000 .000 .000 .000 .000 .000

267

FuncUtil StateEff SymbUtil TraitEff EaseOfUse Intent to Use SU.II .000 .000 .000 .000 .000 .000 SU.I .000 .000 .000 .000 .000 .000 TE.III .000 .000 .000 .000 .000 .000 TE.II .000 .000 .000 .000 .000 .000 TE.I .000 .000 .000 .000 .000 .000

Table 106: Standardized Indirect Effects for Post Hoc Single-group Causal Model FuncUtil StateEff SymbUtil TraitEff EaseOfUse Intent to Use EaseOfUse .000 .000 .000 .000 .000 .000 Intent to Use .000 .003 .000 .001 .000 .000 Int.52 .516 .194 -.043 .001 .006 .000 Int.51 .548 .206 -.045 .001 .006 .000 EoU.10 .000 .365 .000 .133 .000 .000 EoU.9 .000 .358 .000 .131 .000 .000 EoU.8 .000 .265 .000 .097 .000 .000 FU.II .000 .000 .000 .000 .000 .000 FU.III .000 .000 .000 .000 .000 .000 SE.I .000 .000 .000 .000 .000 .000 SE.II .000 .000 .000 .000 .000 .000 SE.III .000 .000 .000 .000 .000 .000 SU.III .000 .000 .000 .000 .000 .000 SU.II .000 .000 .000 .000 .000 .000 SU.I .000 .000 .000 .000 .000 .000 TE.III .000 .000 .000 .000 .000 .000 TE.II .000 .000 .000 .000 .000 .000 TE.I .000 .000 .000 .000 .000 .000

Table 107: Regression Weights for Measurement Model of Lower-Identifiers Estimate S.E. C.R. P Label TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.279 .099 12.864 <.001 FL1 TE.III <--- TraitEff 1.445 .108 13.405 <.001 FL2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil 1.002 .043 23.277 <.001 FL3 SU.III <--- SymbUtil .510 .028 17.977 <.001 FL4 SE.III <--- StateEff 1.000 SE.II <--- StateEff .993 .045 21.842 <.001 FL5 SE.I <--- StateEff 1.560 .077 20.288 <.001 FL6 FU.III <--- FuncUtil 1.000

268

Estimate S.E. C.R. P Label FU.II <--- FuncUtil .882 .052 16.938 <.001 FL7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.274 .130 9.777 <.001 FL8 EoU.10 <--- EaseOfUse 1.423 .146 9.773 <.001 FL9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .879 .028 30.855 <.001 FL10

Table 108: Standardized Regression Weights for Measurement Model of Lower-Identifiers Estimate TE.I <--- TraitEff .674 TE.II <--- TraitEff .802 TE.III <--- TraitEff .821 SU.I <--- SymbUtil .927 SU.II <--- SymbUtil .940 SU.III <--- SymbUtil .836 SE.III <--- StateEff .912 SE.II <--- StateEff .923 SE.I <--- StateEff .885 FU.III <--- FuncUtil .968 FU.II <--- FuncUtil .875 EoU.8 <--- EaseOfUse .621 EoU.9 <--- EaseOfUse .829 EoU.10 <--- EaseOfUse .892 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .925

Table 109: Covariances for Measurement Model of Lower-Identifiers Estimate S.E. C.R. P Label TraitEff <--> StateEff .287 .558 .514 .607 par_11 SymbUtil <--> StateEff -.063 .506 -.125 .901 par_12 SymbUtil <--> FuncUtil 3.735 .795 4.697 <.001 par_13 EaseOfUse <--> Intent to Use .090 .069 1.305 .192 par_14 FuncUtil <--> Intent to Use 3.714 .667 5.571 <.001 par_15 SymbUtil <--> Intent to Use .635 .245 2.587 .010 par_16 TraitEff <--> EaseOfUse .142 .095 1.492 .136 par_17 StateEff <--> EaseOfUse .422 .155 2.726 .006 par_18 SymbUtil <--> EaseOfUse -.003 .085 -.037 .970 par_19 FuncUtil <--> EaseOfUse .056 .206 .273 .785 par_20 TraitEff <--> FuncUtil -1.524 .805 -1.892 .058 par_21 TraitEff <--> SymbUtil -1.147 .347 -3.300 <.001 par_22

269

Estimate S.E. C.R. P Label StateEff <--> Intent to Use 1.588 .434 3.656 <.001 par_23 TraitEff <--> Intent to Use -.228 .264 -.863 .388 par_24 StateEff <--> FuncUtil 2.929 1.259 2.327 .020 par_25

Table 110: Correlations for Measurement Model of Lower-Identifiers Estimate TraitEff <--> StateEff .052 SymbUtil <--> StateEff -.012 SymbUtil <--> FuncUtil .488 EaseOfUse <--> Intent to Use .125 FuncUtil <--> Intent to Use .583 SymbUtil <--> Intent to Use .243 TraitEff <--> EaseOfUse .159 StateEff <--> EaseOfUse .285 SymbUtil <--> EaseOfUse -.004 FuncUtil <--> EaseOfUse .027 TraitEff <--> FuncUtil -.193 TraitEff <--> SymbUtil -.354 StateEff <--> Intent to Use .355 TraitEff <--> Intent to Use -.084 StateEff <--> FuncUtil .224

Table 111: Variances for Measurement Model of Lower-Identifiers Estimate S.E. C.R. P Label TraitEff 3.338 .629 5.303 <.001 par_41 SymbUtil 3.141 .445 7.060 <.001 par_42 StateEff 9.178 1.323 6.939 <.001 par_43 FuncUtil 18.649 2.630 7.091 <.001 par_44 EaseOfUse .239 .053 4.509 <.001 par_45 Intent to Use 2.179 .272 8.000 <.001 par_46 e16 .000 e1 4.008 .595 6.738 <.001 par_47 e2 3.022 .597 5.065 <.001 par_48 e3 3.369 .718 4.691 <.001 par_49 e4 .514 .115 4.487 <.001 par_50 e5 .412 .108 3.821 <.001 par_51 e6 .352 .051 6.835 <.001 par_52 e7 1.860 .368 5.056 <.001 par_53 e8 1.582 .343 4.620 <.001 par_54 e9 6.201 1.049 5.912 <.001 par_55

270

Estimate S.E. C.R. P Label e10 1.270 .986 1.288 .198 par_56 e11 4.433 .939 4.722 <.001 par_57 e13 .381 .053 7.257 <.001 par_58 e14 .177 .038 4.662 <.001 par_59 e15 .125 .041 3.039 .002 par_60 e17 .282 .035 8.000 <.001 par_61

Table 112: Squared Multiple Correlations for Measurement Model of Lower-Identifiers Estimate Int.52 .856 Int.51 1.000 EoU.10 .795 EoU.9 .687 EoU.8 .386 FU.II .766 FU.III .936 SE.I .783 SE.II .851 SE.III .831 SU.III .699 SU.II .884 SU.I .859 TE.III .674 TE.II .644 TE.I .454

Table 113: Regression Weights for Measurement Model of Higher-Identifiers Estimate S.E. C.R. P Label TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.279 .099 12.864 <.001 FL1 TE.III <--- TraitEff 1.445 .108 13.405 <.001 FL2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil 1.002 .043 23.277 <.001 FL3 SU.III <--- SymbUtil .510 .028 17.977 <.001 FL4 SE.III <--- StateEff 1.000 SE.II <--- StateEff .993 .045 21.842 <.001 FL5 SE.I <--- StateEff 1.560 .077 20.288 <.001 FL6

271

Estimate S.E. C.R. P Label FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .882 .052 16.938 <.001 FL7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.274 .130 9.777 <.001 FL8 EoU.10 <--- EaseOfUse 1.423 .146 9.773 <.001 FL9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .879 .028 30.855 <.001 FL10

Table 114: Standardized Regression Weights for Measurement Model of Higher-Identifiers Estimate TE.I <--- TraitEff .796 TE.II <--- TraitEff .849 TE.III <--- TraitEff .933 SU.I <--- SymbUtil .911 SU.II <--- SymbUtil .918 SU.III <--- SymbUtil .783 SE.III <--- StateEff .897 SE.II <--- StateEff .900 SE.I <--- StateEff .867 FU.III <--- FuncUtil .939 FU.II <--- FuncUtil .850 EoU.8 <--- EaseOfUse .617 EoU.9 <--- EaseOfUse .809 EoU.10 <--- EaseOfUse .839 Int.51 <--- Intent to Use .970 Int.52 <--- Intent to Use .926

Table 115: Covariances for Measurement Model of Higher-Identifiers Estimate S.E. C.R. P Label TraitEff <--> StateEff 1.031 .601 1.715 .086 par_26 SymbUtil <--> StateEff -.470 .450 -1.044 .297 par_27 StateEff <--> FuncUtil 2.324 1.167 1.991 .046 par_28 SymbUtil <--> FuncUtil 3.849 .750 5.133 <.001 par_29 EaseOfUse <--> Intent to Use .147 .063 2.328 .020 par_30 FuncUtil <--> Intent to Use 3.080 .586 5.252 <.001 par_31 SymbUtil <--> Intent to Use .454 .206 2.206 .027 par_32 TraitEff <--> EaseOfUse .231 .106 2.174 .030 par_33 StateEff <--> EaseOfUse .811 .168 4.839 <.001 par_34 SymbUtil <--> EaseOfUse .081 .078 1.044 .297 par_35 FuncUtil <--> EaseOfUse .556 .209 2.666 .008 par_36

272

Estimate S.E. C.R. P Label TraitEff <--> FuncUtil -1.016 .875 -1.161 .246 par_37 TraitEff <--> SymbUtil -.970 .355 -2.730 .006 par_38 StateEff <--> Intent to Use 1.155 .367 3.146 .002 par_39 TraitEff <--> Intent to Use -.137 .266 -.515 .606 par_40

Table 116: Correlations for Measurement Model of Higher-Identifiers Estimate TraitEff <--> StateEff .169 SymbUtil <--> StateEff -.101 StateEff <--> FuncUtil .197 SymbUtil <--> FuncUtil .566 EaseOfUse <--> Intent to Use .242 FuncUtil <--> Intent to Use .568 SymbUtil <--> Intent to Use .213 TraitEff <--> EaseOfUse .231 StateEff <--> EaseOfUse .614 SymbUtil <--> EaseOfUse .107 FuncUtil <--> EaseOfUse .288 TraitEff <--> FuncUtil -.114 TraitEff <--> SymbUtil -.276 StateEff <--> Intent to Use .311 TraitEff <--> Intent to Use -.049

Table 117: Variances for Measurement Model of Higher-Identifiers Estimate S.E. C.R. P Label TraitEff 4.620 .820 5.633 <.001 par_62 SymbUtil 2.675 .386 6.937 <.001 par_63 StateEff 8.080 1.183 6.831 <.001 par_64 FuncUtil 17.291 2.546 6.792 <.001 par_65 EaseOfUse .216 .048 4.492 <.001 par_66 Intent to Use 1.703 .230 7.403 <.001 par_67 e1 2.672 .410 6.517 <.001 par_68 e2 2.928 .526 5.569 <.001 par_69 e3 1.437 .505 2.845 .004 par_70 e4 .546 .122 4.460 <.001 par_71 e5 .501 .119 4.200 <.001 par_72 e6 .439 .063 6.959 <.001 par_73 e7 1.970 .383 5.143 <.001 par_74 e8 1.871 .371 5.043 <.001 par_75 e9 6.514 1.100 5.922 <.001 par_76

273

Estimate S.E. C.R. P Label e10 2.304 1.015 2.270 .023 par_77 e11 5.180 .998 5.190 <.001 par_78 e13 .351 .049 7.133 <.001 par_79 e14 .186 .036 5.164 <.001 par_80 e15 .184 .041 4.509 <.001 par_81 e16 .106 .072 1.466 .143 par_82 e17 .218 .061 3.564 <.001 par_83

Table 118: Squared Multiple Correlations for Measurement Model of Higher-Identifiers Estimate Int.52 .858 Int.51 .942 EoU.10 .704 EoU.9 .654 EoU.8 .381 FU.II .722 FU.III .882 SE.I .751 SE.II .810 SE.III .804 SU.III .613 SU.II .843 SU.I .831 TE.III .870 TE.II .721 TE.I .634

Table 119: CMIN for Multi-group Measurement Model Model NPAR CMIN DF P CMIN/DF Default model 83 288.343 189 .000 1.526 Saturated model 272 .000 0 Independence model 32 3126.702 240 .000 13.028

Table 120: RMR, GFI for Multi-group Measurement Model Model RMR GFI AGFI PGFI Default model .499 .882 .830 .613 Saturated model .000 1.000 Independence model 2.815 .373 .289 .329

Table 121: Baseline Comparisons for Multi-group Measurement Model

274

NFI RFI IFI TLI Model CFI Delta1 rho1 Delta2 rho2 Default model .908 .883 .966 .956 .966 Saturated model 1.000 1.000 1.000 Independence model .000 .000 .000 .000 .000

Table 122: Parsimony-Adjusted Measures for Multi-group Measurement Model Model PRATIO PNFI PCFI Default model .788 .715 .760 Saturated model .000 .000 .000 Independence model 1.000 .000 .000

Table 123: NCP for Multi-group Measurement Model Model NCP LO 90 HI 90 Default model 99.343 57.478 149.168 Saturated model .000 .000 .000 Independence model 2886.702 2709.927 3070.824

Table 124: FMIN for Multi-group Measurement Model Model FMIN F0 LO 90 HI 90 Default model 1.131 .390 .225 .585 Saturated model .000 .000 .000 .000 Independence model 12.262 11.320 10.627 12.042

Table 125: RMSEA for Multi-group Measurement Model Model RMSEA LO 90 HI 90 PCLOSE Default model .045 .035 .056 .760 Independence model .217 .210 .224 .000

Table 126: AIC for Multi-group Measurement Model Model AIC BCC BIC CAIC Default model 454.343 479.882 Saturated model 544.000 627.694 Independence model 3190.702 3200.549

Table 127: ECVI for Multi-group Measurement Model Model ECVI LO 90 HI 90 MECVI Default model 1.782 1.618 1.977 1.882 Saturated model 2.133 2.133 2.133 2.462 Independence model 12.513 11.819 13.235 12.551

Table 128: HOELTER for Multi-group Measurement Model

275

HOELTER HOELTER Model .05 .01 Default model 198 211 Independence model 24 25

V2 V3 V4 V15 V16 V17 e1 e2 e3 e13 e14 e15 1 1 1 1 1 1

TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 FL8 1 FL1 FL2 1 FL9 0 V18 V1 e16 e17 R1 TraitEff EaseOfUse 1 1 1 V19 R2 R3 Int.51 Int.52 0 V14 1 FL10 ccc1_1V9 res1 e10 e11 StateEff 1 1 Intent to Use FL6 1 ccc2_1 FL5 FU.III FU.II R4 1 ccc5_1 1 FL7 V20 SE.I SE.II SE.III ccc4_1 V13 R5 1 1 1 res2 V12 V11 V10 FuncUtil

e9 e8 e7 ccc3_1 Cv1 V5

SymbUtil FL4 1 FL3

SU.III SU.II SU.I 1 1 1 V8 V7 V6 e6 e5 e4

Figure 79: Multi-group Causal Model Unconstrained

276

4.01 3.03 3.35 .38 .17 .13 e1 e2 e3 e13 e14 e15 1 1 1 1 1 1

TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 1.28 1.00 1.281.45 1.00 1.39 .00 .28 3.33 e16 e17 .04 TraitEff EaseOfUse 1 1 1 .22 .05 .34 Int.51 Int.52 .00 5.42 1.00 .87 .28 9.21 res1 e10 e11 StateEff 1 1 Intent to Use 1.56 1.00 -1.47 .99 FU.III FU.II .19 1 3.16 .81 1.43 -1.14 1.00 SE.I SE.II SE.III 19.79 -.01 1 1 1 res2 6.12 1.59 1.87 FuncUtil e9 e8 e7 -.06 3.533.15

SymbUtil .51 1.00 1.00

SU.III SU.II SU.I 1 1 1 .35 .41 .51 e6 e5 e4

Figure 80: Unstandardized Results for Multi-group Causal Model Unconstrained

277

e1 e2 e3 e13 e14 e15

.45 .64 .68 .39 .70 .78 TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 .62 .84 .88 .67 .80 .82 .11 e16 e17 .14 EaseOfUse TraitEff 1.00 .86

.28 .11 Int.51 Int.52 1.00 .92 .05 res1 .35 e10 e11

StateEff 1.00 .70 Intent to Use .89 .91 -.18 .92 FU.III FU.II .79 .85 .83 .58 .84 -.35 .23 1.00 SE.I SE.II SE.III -.02 res2 FuncUtil e9 e8 e7 -.01 .45

SymbUtil .84 .93 .94 .70 .88 .86 SU.III SU.II SU.I

e6 e5 e4

Figure 81: Standardized Results for Multi-group Causal Model Unconstrained

278

4.02 3.02 3.34 .38 .17 .14 e1 e2 e3 e13 e14 e15 1 1 1 1 1 1

TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 1.30 1.00 1.281.45 1.00 1.39 .00 .28 3.33 e16 e17 .04 TraitEff EaseOfUse 1 1 1 .22 .07 .31 Int.51 Int.52 .00 5.42 1.00 .87 .26 9.14 res1 e10 e11 StateEff 1 1 Intent to Use 1.56 1.00 -1.48 .99 FU.III FU.II .18 1 3.14 .81 1.43 -1.14 1.00 SE.I SE.II SE.III 19.79 -.04 1 1 1 res2 6.19 1.60 1.87 FuncUtil e9 e8 e7 -.06 3.533.15

SymbUtil .51 1.00 1.00

SU.III SU.II SU.I 1 1 1 .35 .41 .51 e6 e5 e4

Figure 82: Unstandardized Results for Multi-group Causal Model Constrained

279

e1 e2 e3 e13 e14 e15

.45 .64 .68 .42 .74 .79 TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 .65 .86 .89 .67 .80 .82 .20 e16 e17 .12 EaseOfUse TraitEff 1.00 .85

.42 .11 Int.51 Int.52 1.00 .92 .05 res1 .31 e10 e11

StateEff 1.00 .70 Intent to Use .88 .91 -.18 .92 FU.III FU.II .78 .85 .83 .56 .84 -.35 .23 1.00 SE.I SE.II SE.III -.05 res2 FuncUtil e9 e8 e7 -.01 .45

SymbUtil .84 .93 .94 .70 .88 .86 SU.III SU.II SU.I

e6 e5 e4

Figure 83: Standardized Results for Multi-group Causal Model Constrained

Table 129: Unstandardized Regression Weights for Unconstrained Multi-group Causal Model – Lower- Identifiers Estimate S.E. C.R. P Label EaseOfUse <--- TraitEff .039 .027 1.431 .152 R1 EaseOfUse <--- StateEff .046 .016 2.903 .004 R2 Intent to Use <--- EaseOfUse .336 .233 1.440 .150 R3 Intent to Use <--- FuncUtil .192 .027 7.159 <.001 R4 Intent to Use <--- SymbUtil -.014 .069 -.202 .840 R5 TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.280 .100 12.842 <.001 FL1 TE.III <--- TraitEff 1.449 .108 13.366 <.001 FL2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil .999 .043 23.150 <.001 FL3 SU.III <--- SymbUtil .510 .028 17.988 <.001 FL4 SE.III <--- StateEff 1.000

280

Estimate S.E. C.R. P Label SE.II <--- StateEff .989 .045 21.737 <.001 FL5 SE.I <--- StateEff 1.561 .077 20.363 <.001 FL6 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .805 .034 23.374 <.001 FL7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.284 .131 9.819 <.001 FL8 EoU.10 <--- EaseOfUse 1.393 .142 9.783 <.001 FL9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .873 .029 30.589 <.001 FL10

V2 V3 V4 V15 V16 V17 e1 e2 e3 e13 e14 e15 1 1 1 1 1 1

TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 FL8 1 FL1 FL2 1 FL9 0 V18 V1 e16 e17 R1 TraitEff EaseOfUse 1 1 1 V19 R2 R3 Int.51 Int.52 0 V14 1 FL10 ccc1_1V9 res1 e10 e11 StateEff 1 R6 1 Intent to Use FL6 1 ccc2_1 FL5 FU.III FU.II R4 1 ccc5_1 1 FL7 V20 SE.I SE.II SE.III ccc4_1 V13 R5 1 1 1 res2 V12 V11 V10 FuncUtil

e9 e8 e7 ccc3_1 Cv1 V5

SymbUtil FL4 1 FL3

SU.III SU.II SU.I 1 1 1 V8 V7 V6 e6 e5 e4

Figure 84: Post Hoc Causal Model – Lower-Identifiers

281

e1 e2 e3 e13 e14 e15

.45 .64 .68 .39 .69 .79 TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 .62 .83 .89 .67 .80 .82 .10 e16 e17 .14 EaseOfUse TraitEff 1.00 .86

.28 .04 Int.51 Int.52 1.00 .92 .05 res1 .39 e10 e11 .22 StateEff 1.00 .70 Intent to Use .88 .91 -.18 .92 FU.III FU.II .78 .85 .83 .52 .84 -.35 .23 1.00 SE.I SE.II SE.III .01 res2 FuncUtil e9 e8 e7 -.01 .45

SymbUtil .84 .93 .94 .70 .88 .86 SU.III SU.II SU.I

e6 e5 e4

Figure 85: Post Hoc Standardized Results – Lower-Identifiers

282

2.68 2.92 1.43 .35 .18 .19 e1 e2 e3 e13 e14 e15 1 1 1 1 1 1

TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 1.28 1.00 1.281.45 1.00 1.41 .08 .24 4.61 e16 e17 .04 TraitEff EaseOfUse 1 1 1 .12 .10 .29 Int.51 Int.52 .00 6.71 1.00 .87 1.038.12 res1 e10 e11 1 1 .01 StateEff -.86 Intent to Use 1.56 1.00 2.65 .99 FU.III FU.II -.97 .17 1 1.00 .81.05 1.13 19.83 SE.I SE.II SE.III -.48 -.08 1 1 1 res2 6.51 1.92 1.91 FuncUtil e9 e8 e7 3.64 2.68

SymbUtil .51 1.00 1.00

SU.III SU.II SU.I 1 1 1 .44 .51 .54 e6 e5 e4

Figure 86: Unstandardized Results Multi-group Causal Model Unconstrained – Higher Identifiers

283

e1 e2 e3 e13 e14 e15

.63 .72 .87 .38 .66 .69 TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 .62 .81 .83 .80 .85 .93 .45 e16 e17 .19 EaseOfUse TraitEff .95 .85

.58 .10 Int.51 Int.52 .17 .98 .92 res1 .34 e10 e11 .09 StateEff -.09 1.00 .66 Intent to Use .87 .90 .21 .90 FU.III FU.II .75 .80 .81 -.28 .58 1.00 .81.17 SE.I SE.II SE.III -.10 -.10 res2 FuncUtil e9 e8 e7 .50

SymbUtil .78 .91 .92 .61 .84 .83 SU.III SU.II SU.I

e6 e5 e4

Figure 87: Standardized Results Multi-group Causal Model Unconstrained – Higher Identifiers

284

4.02 3.01 3.36 .38 .17 .13 e1 e2 e3 e13 e14 e15 1 1 1 1 1 1

TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 1.29 1.00 1.281.45 1.00 1.41 .00 .28 3.33 e16 e17 .04 TraitEff EaseOfUse 1 1 1 .22 .07 .19 Int.51 Int.52 .00 5.42 1.00 .87 .25 9.16 res1 e10 e11 StateEff 1 .10 1 Intent to Use 1.55 1.00 -1.48 .99 FU.III FU.II .17 1 3.15 .81 1.34 -1.14 1.00 SE.I SE.II SE.III 19.79 -.03 1 1 1 res2 6.24 1.60 1.85 FuncUtil e9 e8 e7 -.05 3.533.15

SymbUtil .51 1.00 1.00

SU.III SU.II SU.I 1 1 1 .35 .41 .51 e6 e5 e4

Figure 88: Unstandardized Results Multi-group Causal Model Constrained – Lower-Identifiers

285

e1 e2 e3 e13 e14 e15

.45 .64 .67 .42 .73 .80 TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 .65 .85 .90 .67 .80 .82 .20 e16 e17 .15 EaseOfUse TraitEff 1.00 .85

.41 .07 Int.51 Int.52 1.00 .92 .05 res1 .38 e10 e11 .21 StateEff 1.00 .70 Intent to Use .88 .91 -.18 .92 FU.III FU.II .78 .85 .83 .52 .84 -.35 .23 1.00 SE.I SE.II SE.III -.04 res2 FuncUtil e9 e8 e7 -.01 .45

SymbUtil .84 .93 .94 .70 .88 .86 SU.III SU.II SU.I

e6 e5 e4

Figure 89: Standardized Results Multi-group Causal Model Unconstrained – Lower-Identifiers

286

2.68 2.91 1.45 .35 .18 .19 e1 e2 e3 e13 e14 e15 1 1 1 1 1 1

TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 1.29 1.00 1.281.45 1.00 1.41 .08 .24 4.61 e16 e17 .04 TraitEff EaseOfUse 1 1 1 .13 .07 .19 Int.51 Int.52 .00 6.71 1.00 .87 1.058.25 res1 e10 e11 1 1 .01 StateEff -.86 Intent to Use 1.55 1.00 2.65 .99 FU.III FU.II -.97 .17 1 1.00 .81.04 1.14 19.83 SE.I SE.II SE.III -.47 -.03 1 1 1 res2 6.50 1.94 1.85 FuncUtil e9 e8 e7 3.65 2.68

SymbUtil .51 1.00 1.00

SU.III SU.II SU.I 1 1 1 .44 .51 .54 e6 e5 e4

Figure 90: Unstandardized Results Multi-group Causal Model Constrained – Higher-Identifiers

287

e1 e2 e3 e13 e14 e15

.63 .72 .87 .36 .65 .66 TE.I TE.II TE.III EoU.8 EoU.9 EoU.10 .60 .80 .81 .80 .85 .93 .35 e16 e17 .21 EaseOfUse TraitEff .95 .85

.46 .06 Int.51 Int.52 .17 .98 .92 res1 .34 e10 e11 .15 StateEff -.09 1.00 .66 Intent to Use .87 .90 .21 .90 FU.III FU.II .75 .80 .82 -.28 .58 1.00 .81.13 SE.I SE.II SE.III -.10 -.04 res2 FuncUtil e9 e8 e7 .50

SymbUtil .78 .91 .92 .61 .84 .83 SU.III SU.II SU.I

e6 e5 e4

Figure 91: Standardized Results Multi-group Causal Model Constrained – Higher-Identifiers

Table 130: Post Hoc Unstandardized Regression Weights for Unconstrained - Lower-Identifiers Estimate S.E. C.R. PLabel EaseOfUse <--- TraitEff .038 .027 1.405 .160R1 EaseOfUse <--- StateEff .045 .016 2.845 .004R2 Intent to Use <--- EaseOfUse .128 .236 .542 .588R3 Intent to Use <--- FuncUtil .171 .027 6.328 <.001R4 Intent to Use <--- SymbUtil .012 .068 .178 .859R5 Intent to Use <--- StateEff .108 .038 2.824 .005R6 TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.280 .100 12.854 <.001FL1 TE.III <--- TraitEff 1.447 .108 13.386 <.001FL2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil .999 .043 23.183 <.001FL3 SU.III <--- SymbUtil .509 .028 17.991 <.001FL4 SE.III <--- StateEff 1.000 SE.II <--- StateEff .989 .045 21.837 <.001FL5 SE.I <--- StateEff 1.557 .077 20.345 <.001FL6 FU.III <--- FuncUtil 1.000

288

FU.II <--- FuncUtil .805 .034 23.374 <.001FL7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.276 .130 9.803 <.001FL8 EoU.10 <--- EaseOfUse 1.412 .144 9.786 <.001FL9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .874 .028 30.690 <.001FL10

Table 131: Post Hoc Standardized Regression Weights for Unconstrained - Lower-Identifiers Estimate EaseOfUse <--- TraitEff .142 EaseOfUse <--- StateEff .277 Intent to Use <--- EaseOfUse .043 Intent to Use <--- FuncUtil .516 Intent to Use <--- SymbUtil .015 Intent to Use <--- StateEff .223 TE.I <--- TraitEff .674 TE.II <--- TraitEff .802 TE.III <--- TraitEff .822 SU.I <--- SymbUtil .927 SU.II <--- SymbUtil .940 SU.III <--- SymbUtil .836 SE.III <--- StateEff .912 SE.II <--- StateEff .922 SE.I <--- StateEff .885 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .839 EoU.8 <--- EaseOfUse .622 EoU.9 <--- EaseOfUse .832 EoU.10 <--- EaseOfUse .889 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .925

Table 132: Post Hoc Covariances for Unconstrained Lower-Identifiers Estimate S.E. C.R. PLabel SymbUtil <--> FuncUtil 3.536 .789 4.482 <.001Cv1 TraitEff <--> StateEff .285 .559 .511 .609ccc1_1 TraitEff <--> FuncUtil -1.472 .805 -1.829 .067ccc2_1 SymbUtil <--> StateEff -.057 .508 -.112 .911ccc3_1 TraitEff <--> SymbUtil -1.142 .347 -3.288 .001ccc4_1 StateEff <--> FuncUtil 3.158 1.269 2.489 .013ccc5_1

Table 133: Post Hoc Correlations for Unconstrained Lower-Identifiers Estimate SymbUtil <--> FuncUtil .448 TraitEff <--> StateEff .052 TraitEff <--> FuncUtil -.181

289

SymbUtil <--> StateEff -.011 TraitEff <--> SymbUtil -.353 StateEff <--> FuncUtil .234

Table 134: Post Hoc Variances for Unconstrained Lower-Identifiers Estimate S.E. C.R. PLabel TraitEff 3.333 .629 5.301 <.001V1 SymbUtil 3.149 .446 7.061 <.001V5 StateEff 9.218 1.327 6.945 <.001V9 FuncUtil 19.789 2.474 8.000 <.001V13 res1 .217 .048 4.471 <.001V19 res2 1.339 .168 7.951 <.001V20 e1 4.010 .595 6.741 <.001V2 e2 3.026 .598 5.064 <.001V3 e3 3.355 .718 4.673 <.001V4 e4 .512 .115 4.445 <.001V6 e5 .414 .108 3.827 <.001V7 e6 .351 .051 6.829 <.001V8 e7 1.854 .368 5.033 <.001V10 e8 1.591 .342 4.648 <.001V11 e9 6.190 1.048 5.907 <.001V12 e10 .000 e11 5.415 .677 8.000 <.001V14 e13 .382 .053 7.251 <.001V15 e14 .175 .038 4.591 <.001V16 e15 .127 .041 3.108 .002V17 e16 .000 e17 .282 .035 8.000 <.001V18

Table 135: Post Hoc Squared Multiple Correlations for Unconstrained Lower-Identifiers Estimate EaseOfUse .101 Intent to Use .385 Int.52 .855 Int.51 1.000 EoU.10 .791 EoU.9 .692 EoU.8 .387 FU.II .703 FU.III 1.000 SE.I .783 SE.II .850 SE.III .833 SU.III .699 SU.II .883

290

SU.I .860 TE.III .675 TE.II .644 TE.I .454

Table 136: Post Hoc Unstandardized Regression Weights for Unconstrained Higher-Identifiers Estimate S.E. C.R. PLabel EaseOfUse <--- TraitEff .041 .019 2.103 .035R-1 EaseOfUse <--- StateEff .095 .017 5.722 <.001R-2 EaseOfUse <--- FuncUtil .009 .010 .908 .364R-6 EaseOfUse <--- SymbUtil .049 .030 1.642 .101R-7 Intent to Use <--- EaseOfUse .294 .240 1.223 .221R-3 Intent to Use <--- FuncUtil .172 .027 6.417 <.001R-4 Intent to Use <--- SymbUtil -.077 .073 -1.055 .291R-5 TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.280 .100 12.854 <.001FL1 TE.III <--- TraitEff 1.447 .108 13.386 <.001FL2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil .999 .043 23.183 <.001FL3 SU.III <--- SymbUtil .509 .028 17.991 <.001FL4 SE.III <--- StateEff 1.000 SE.II <--- StateEff .989 .045 21.837 <.001FL5 SE.I <--- StateEff 1.557 .077 20.345 <.001FL6 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .805 .034 23.374 <.001FL7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.276 .130 9.803 <.001FL8 EoU.10 <--- EaseOfUse 1.412 .144 9.786 <.001FL9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .874 .028 30.690 <.001FL10

Table 137: Post Hoc Standardized Regression Weights for Unconstrained Higher-Identifiers Estimate EaseOfUse <--- TraitEff .189 EaseOfUse <--- StateEff .584 EaseOfUse <--- FuncUtil .089 EaseOfUse <--- SymbUtil .172 Intent to Use <--- EaseOfUse .105 Intent to Use <--- FuncUtil .584 Intent to Use <--- SymbUtil -.096 TE.I <--- TraitEff .795 TE.II <--- TraitEff .849 TE.III <--- TraitEff .933 SU.I <--- SymbUtil .913 SU.II <--- SymbUtil .916

291

SU.III <--- SymbUtil .783 SE.III <--- StateEff .900 SE.II <--- StateEff .897 SE.I <--- StateEff .867 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .810 EoU.8 <--- EaseOfUse .619 EoU.9 <--- EaseOfUse .812 EoU.10 <--- EaseOfUse .832 Int.51 <--- Intent to Use .977 Int.52 <--- Intent to Use .920

Table 138: Post Hoc Covariances for Unconstrained Higher-Identifiers Estimate S.E. C.R. PLabel SymbUtil <--> FuncUtil 3.645 .752 4.846 <.001Cv-1 TraitEff <--> StateEff 1.028 .602 1.708 .088ccc1_2 TraitEff <--> FuncUtil -.862 .891 -.968 .333ccc2_2 TraitEff <--> SymbUtil -.967 .355 -2.723 .006ccc3_2 StateEff <--> FuncUtil 2.650 1.199 2.209 .027ccc4_2 SymbUtil <--> StateEff -.479 .452 -1.060 .289par_76

Table 139: Post Hoc Correlations for Unconstrained Higher-Identifiers Estimate SymbUtil <--> FuncUtil .500 TraitEff <--> StateEff .168 TraitEff <--> FuncUtil -.090 TraitEff <--> SymbUtil -.275 StateEff <--> FuncUtil .209 SymbUtil <--> StateEff -.103

Table 140: Post Hoc Variances for Unconstrained Higher-Identifiers Estimate S.E. C.R. PLabel TraitEff 4.612 .820 5.627 <.001V-1 SymbUtil 2.680 .386 6.944 <.001V-5 StateEff 8.116 1.185 6.849 <.001V-9 FuncUtil 19.834 2.489 7.969 <.001V-13 res1 .120 .029 4.096 <.001V-20 res2 1.134 .160 7.102 <.001V-21 e1 2.681 .411 6.526 <.001V-2 e2 2.920 .525 5.558 <.001V-3 e3 1.434 .506 2.835 .005V-4 e4 .535 .123 4.357 <.001V-6 e5 .514 .121 4.248 <.001V-7 e6 .438 .063 6.945 <.001V-8 e7 1.909 .380 5.029 <.001V-10 e8 1.925 .376 5.118 <.001V-11

292

e9 6.512 1.102 5.912 <.001V-12 e10 .000 e11 6.715 .843 7.969 <.001V-14 e13 .348 .049 7.118 <.001V-15 e14 .182 .036 5.090 <.001V-16 e15 .192 .041 4.671 <.001V-17 e16 .082 .072 1.141 .254V-18 e17 .236 .062 3.810 <.001V-19

Table 141: Post Hoc Squared Multiple Correlations for Unconstrained Higher-Identifiers Estimate EaseOfUse .447 Intent to Use .338 Int.52 .847 Int.51 .954 EoU.10 .692 EoU.9 .659 EoU.8 .383 FU.II .657 FU.III 1.000 SE.I .751 SE.II .805 SE.III .810 SU.III .614 SU.II .839 SU.I .834 TE.III .871 TE.II .721 TE.I .632

Table 142: Post Hoc Regression Weights for Constrained Lower-Identifiers Estimate S.E. C.R. PLabel EaseOfUse <--- TraitEff .043 .016 2.624 .009R1 EaseOfUse <--- StateEff .071 .012 5.807 <.001R2 Intent to Use <--- EaseOfUse .193 .172 1.126 .260R3 Intent to Use <--- FuncUtil .172 .019 9.144 <.001R4 Intent to Use <--- SymbUtil -.030 .050 -.596 .551R5 Intent to Use <--- StateEff .104 .038 2.742 .006R6 TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.282 .100 12.857 <.001FL1 TE.III <--- TraitEff 1.446 .108 13.384 <.001FL2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil 1.000 .043 23.165 <.001FL3 SU.III <--- SymbUtil .510 .028 17.986 <.001FL4 SE.III <--- StateEff 1.000 SE.II <--- StateEff .986 .045 21.844 <.001FL5

293

SE.I <--- StateEff 1.552 .076 20.356 <.001FL6 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .805 .034 23.374 <.001FL7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.287 .130 9.921 <.001FL8 EoU.10 <--- EaseOfUse 1.406 .141 9.939 <.001FL9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .874 .028 30.717 <.001FL10

Table 143: Post Hoc Standardized Regression Weights for Constrained Lower-Identifiers Estimate EaseOfUse <--- TraitEff .149 EaseOfUse <--- StateEff .411 Intent to Use <--- EaseOfUse .069 Intent to Use <--- FuncUtil .522 Intent to Use <--- SymbUtil -.036 Intent to Use <--- StateEff .214 TE.I <--- TraitEff .673 TE.II <--- TraitEff .803 TE.III <--- TraitEff .821 SU.I <--- SymbUtil .927 SU.II <--- SymbUtil .940 SU.III <--- SymbUtil .836 SE.III <--- StateEff .912 SE.II <--- StateEff .921 SE.I <--- StateEff .883 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .839 EoU.8 <--- EaseOfUse .647 EoU.9 <--- EaseOfUse .852 EoU.10 <--- EaseOfUse .896 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .924

Table 144: Post Hoc Covariances for Constrained Lower-Identifiers Estimate S.E. C.R. PLabel SymbUtil <--> FuncUtil 3.533 .789 4.481 <.001Cv1 TraitEff <--> StateEff .251 .556 .450 .652ccc1_1 TraitEff <--> FuncUtil -1.483 .804 -1.843 .065ccc2_1 SymbUtil <--> StateEff -.047 .506 -.094 .925ccc3_1 TraitEff <--> SymbUtil -1.140 .347 -3.288 .001ccc4_1 StateEff <--> FuncUtil 3.154 1.265 2.494 .013ccc5_1

Table 145: Post Hoc Correlations for Constrained Lower-Identifiers Estimate SymbUtil <--> FuncUtil .448

294

TraitEff <--> StateEff .045 TraitEff <--> FuncUtil -.183 SymbUtil <--> StateEff -.009 TraitEff <--> SymbUtil -.352 StateEff <--> FuncUtil .234

Table 146: Post Hoc Variances for Constrained Lower-Identifiers Estimate S.E. C.R. PLabel TraitEff 3.327 .628 5.301 <.001V1 SymbUtil 3.146 .446 7.059 <.001V5 StateEff 9.158 1.318 6.948 <.001V9 FuncUtil 19.789 2.474 8.000 <.001V13 res1 .220 .049 4.486 <.001V19 res2 1.344 .169 7.949 <.001V20 e1 4.022 .596 6.748 <.001V2 e2 3.015 .597 5.054 <.001V3 e3 3.358 .717 4.684 <.001V4 e4 .513 .115 4.453 <.001V6 e5 .412 .108 3.813 <.001V7 e6 .352 .052 6.832 <.001V8 e7 1.850 .367 5.038 <.001V10 e8 1.603 .341 4.696 <.001V11 e9 6.241 1.050 5.943 <.001V12 e10 .000 e11 5.415 .677 8.000 <.001V14 e13 .381 .053 7.245 <.001V15 e14 .171 .038 4.487 <.001V16 e15 .132 .041 3.247 .001V17 e16 .000 e17 .282 .035 8.000 <.001V18

Table 147: Post Hoc Squared Multiple Correlations for Constrained Lower-Identifiers Estimate EaseOfUse .196 Intent to Use .377 Int.52 .854 Int.51 1.000 EoU.10 .803 EoU.9 .726 EoU.8 .418 FU.II .703 FU.III 1.000 SE.I .780 SE.II .848 SE.III .832 SU.III .699

295

SU.II .884 SU.I .860 TE.III .675 TE.II .645 TE.I .453

Table 148: Post Hoc Regression Weights for Constrained Higher-Identifiers Estimate S.E. C.R. PLabel EaseOfUse <--- TraitEff .043 .016 2.624 .009R1 EaseOfUse <--- StateEff .071 .012 5.807 <.001R2 EaseOfUse <--- FuncUtil .015 .010 1.452 .147R-6 EaseOfUse <--- SymbUtil .036 .029 1.234 .217R-7 Intent to Use <--- EaseOfUse .193 .172 1.126 .260R3 Intent to Use <--- FuncUtil .172 .019 9.144 <.001R4 Intent to Use <--- SymbUtil -.030 .050 -.596 .551R5 TE.I <--- TraitEff 1.000 TE.II <--- TraitEff 1.282 .100 12.857 <.001FL1 TE.III <--- TraitEff 1.446 .108 13.384 <.001FL2 SU.I <--- SymbUtil 1.000 SU.II <--- SymbUtil 1.000 .043 23.165 <.001FL3 SU.III <--- SymbUtil .510 .028 17.986 <.001FL4 SE.III <--- StateEff 1.000 SE.II <--- StateEff .986 .045 21.844 <.001FL5 SE.I <--- StateEff 1.552 .076 20.356 <.001FL6 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .805 .034 23.374 <.001FL7 EoU.8 <--- EaseOfUse 1.000 EoU.9 <--- EaseOfUse 1.287 .130 9.921 <.001FL8 EoU.10 <--- EaseOfUse 1.406 .141 9.939 <.001FL9 Int.51 <--- Intent to Use 1.000 Int.52 <--- Intent to Use .874 .028 30.717 <.001FL10

Table 149: Post Hoc Standardized Regression Weights for Constrained Higher-Identifiers Estimate EaseOfUse <--- TraitEff .208 EaseOfUse <--- StateEff .463 EaseOfUse <--- FuncUtil .149 EaseOfUse <--- SymbUtil .134 Intent to Use <--- EaseOfUse .065 Intent to Use <--- FuncUtil .581 Intent to Use <--- SymbUtil -.037 TE.I <--- TraitEff .795 TE.II <--- TraitEff .850 TE.III <--- TraitEff .932 SU.I <--- SymbUtil .913 SU.II <--- SymbUtil .916

296

SU.III <--- SymbUtil .784 SE.III <--- StateEff .904 SE.II <--- StateEff .897 SE.I <--- StateEff .868 FU.III <--- FuncUtil 1.000 FU.II <--- FuncUtil .810 EoU.8 <--- EaseOfUse .596 EoU.9 <--- EaseOfUse .803 EoU.10 <--- EaseOfUse .815 Int.51 <--- Intent to Use .977 Int.52 <--- Intent to Use .922

Table 150: Post Hoc Covariances for Constrained Higher-Identifiers Estimate S.E. C.R. P Label SymbUtil <--> FuncUtil 3.645 .752 4.847 <.001 Cv-1 TraitEff <--> StateEff 1.049 .606 1.730 .084 ccc1_2 TraitEff <--> FuncUtil -.862 .891 -.967 .333 ccc2_2 TraitEff <--> SymbUtil -.969 .355 -2.727 .006 ccc3_2 StateEff <--> FuncUtil 2.649 1.208 2.193 .028 ccc4_2 SymbUtil <--> StateEff -.466 .455 -1.025 .305 par_71

Table 151: Post Hoc Correlations for Constrained Higher-Identifiers Estimate SymbUtil <--> FuncUtil .500 TraitEff <--> StateEff .170 TraitEff <--> FuncUtil -.090 TraitEff <--> SymbUtil -.276 StateEff <--> FuncUtil .207 SymbUtil <--> StateEff -.099

Table 152: Post Hoc Variances for Constrained Higher-Identifiers Estimate S.E. C.R. PLabel TraitEff 4.611 .819 5.628 <.001V-1 SymbUtil 2.680 .386 6.941 <.001V-5 StateEff 8.249 1.200 6.876 <.001V-9 FuncUtil 19.834 2.489 7.969 <.001V-13 res1 .127 .030 4.203 <.001V-20 res2 1.144 .161 7.124 <.001V-21 e1 2.678 .410 6.524 <.001V-2 e2 2.912 .525 5.546 <.001V-3 e3 1.448 .506 2.863 .004V-4 e4 .538 .123 4.360 <.001V-6 e5 .513 .121 4.228 <.001V-7 e6 .436 .063 6.938 <.001V-8 e7 1.850 .379 4.877 <.001V-10 e8 1.944 .381 5.101 <.001V-11

297

e9 6.504 1.105 5.884 <.001V-12 e10 .000 e11 6.715 .843 7.969 <.001V-14 e13 .351 .049 7.126 <.001V-15 e14 .177 .036 4.951 <.001V-16 e15 .194 .041 4.711 <.001V-17 e16 .083 .072 1.152 .249V-18 e17 .235 .062 3.788 <.001V-19

Table 153: Post Hoc Squared Multiple Correlations for Constrained Higher-Identifiers Estimate EaseOfUse .346 Intent to Use .343 Int.52 .850 Int.51 .954 EoU.10 .664 EoU.9 .645 EoU.8 .356 FU.II .657 FU.III 1.000 SE.I .754 SE.II .805 SE.III .817 SU.III .615 SU.II .839 SU.I .833 TE.III .870 TE.II .722 TE.I .633

Table 154: Post Hoc Multi- group CMIN Results Model NPAR CMIN DF P CMIN/DF Unconstrained 76 296.881 196 .000 1.515 Measurement weights 76 296.881 196 .000 1.515 Structural weights 71 303.816 201 .000 1.512 Structural covariances 60 322.301 212 .000 1.520 Structural residuals 58 328.728 214 .000 1.536 Measurement residuals 43 345.252 229 .000 1.508 Saturated model 272 .000 0 Independence model 32 3126.702 240 .000 13.028

Table 155: Post Hoc Multi- group RMR, GFI Results Model RMR GFI AGFI PGFI Unconstrained .527 .879 .832 .633 Measurement weights .527 .879 .832 .633 Structural weights .547 .875 .831 .647

298

Structural covariances .794 .869 .831 .677 Structural residuals .794 .867 .831 .682 Measurement residuals .774 .859 .833 .723 Saturated model .000 1.000 Independence model 2.815 .373 .289 .329

Table 156: Post Hoc Multi- group Baseline Comparisons NFI RFI IFI TLI Model Delta1 rho1 Delta2 rho2 CFI

Unconstrained .905 .884 .966 .957 .965 Measurement weights .905 .884 .966 .957 .965 Structural weights .903 .884 .965 .957 .964 Structural covariances .897 .883 .962 .957 .962 Structural residuals .895 .882 .961 .955 .960 Measurement residuals .890 .884 .960 .958 .960 Saturated model 1.000 1.000 1.000 Independence model .000 .000 .000 .000 .000

Table 157: Post Hoc Multi- group Parsimony-Adjusted Measures Model PRATIO PNFI PCFI Unconstrained .817 .739 .788 Measurement weights .817 .739 .788 Structural weights .838 .756 .808 Structural covariances .883 .792 .850 Structural residuals .892 .798 .856 Measurement residuals .954 .849 .916 Saturated model .000 .000 .000 Independence model 1.000 .000 .000

Table 158: Post Hoc Multi- group NCP Results Model NCP LO 90 HI 90 Unconstrained 100.881 58.429 151.300 Measurement weights 100.881 58.429 151.300 Structural weights 102.816 59.849 153.749 Structural covariances 110.301 65.858 162.705 Structural residuals 114.728 69.679 167.731 Measurement residuals 116.252 70.224 170.249 Saturated model .000 .000 .000 Independence model 2886.702 2709.927 3070.824

Table 159: Post Hoc Multi- group FMIN Results Model FMIN F0 LO 90 HI 90 Unconstrained 1.164 .396 .229 .593 Measurement weights 1.164 .396 .229 .593 Structural weights 1.191 .403 .235 .603

299

Structural covariances 1.264 .433 .258 .638 Structural residuals 1.289 .450 .273 .658 Measurement residuals 1.354 .456 .275 .668 Saturated model .000 .000 .000 .000 Independence model 12.262 11.320 10.627 12.042

Table 160: Post Hoc Multi- group RMSEA Results Model RMSEA LO 90 HI 90 PCLOSE Unconstrained .045 .034 .055 .787 Measurement weights .045 .034 .055 .787 Structural weights .045 .034 .055 .797 Structural covariances .045 .035 .055 .785 Structural residuals .046 .036 .055 .752 Measurement residuals .045 .035 .054 .821 Independence model .217 .210 .224 .000

Table 161: Post Hoc Multi- group AIC Results Model AIC BCC BIC CAIC Unconstrained 448.881 472.267 Measurement weights 448.881 472.267 Structural weights 445.816 467.662 Structural covariances 442.301 460.763 Structural residuals 444.728 462.575 Measurement residuals 431.252 444.483 Saturated model 544.000 627.694 Independence model 3190.702 3200.549

Table 162: Post Hoc Multi- group ECVI Results Model ECVI LO 90 HI 90 MECVI Unconstrained 1.760 1.594 1.958 1.852 Measurement weights 1.760 1.594 1.958 1.852 Structural weights 1.748 1.580 1.948 1.834 Structural covariances 1.735 1.560 1.940 1.807 Structural residuals 1.744 1.567 1.952 1.814 Measurement residuals 1.691 1.511 1.903 1.743 Saturated model 2.133 2.133 2.133 2.462 Independence model 12.513 11.819 13.235 12.551

Table 163: Post Hoc Multi- group HOELTER Results HOELTER HOELTER Model .05 .01

Unconstrained 199 212 Measurement weights 199 212 Structural weights 199 212 Structural covariances 197 209

300

Structural residuals 195 207 Measurement residuals 197 210 Independence model 24 25

Table 164: Post Hoc Multi- group Nested Model Comparisons assuming model Unconstrained to be correct NFI IFI RFI TLI Model DF CMIN P Delta-1 Delta-2 rho-1 rho2

Structural weights 5 6.934 .226 .002 .002 .000 .000 Structural covariances 16 25.419 .063 .008 .009 .000 .000 Structural residuals 18 31.846 .023 .010 .011 .002 .002 Measurement residuals 33 48.370 .041 .015 .017 -.001 -.001

Table 165: Post Hoc Multi- group Nested Model Comparisons assuming model Measurement weights to be correct NFI IFI RFI TLI Model DF CMIN P Delta-1 Delta-2 rho-1 rho2

Table 166: Post Hoc Multi- group Nested Model Comparisons assuming model Structural weights to be correct: NFI IFI RFI TLI Model DF CMIN P Delta-1 Delta-2 rho-1 rho2

Structural covariances 11 18.485 .071 .006 .006 .001 .001 Structural residuals 13 24.912 .024 .008 .009 .002 .002 Measurement residuals 28 41.436 .049 .013 .014 .000 .000

Table 167: Post Hoc Multi- group Nested Model Comparisons assuming model Structural covariances to be correct NFI IFI RFI TLI Model DF CMIN P Delta-1 Delta-2 rho-1 rho2

Structural residuals 2 6.427 .040 .002 .002 .001 .001 Measurement residuals 17 22.951 .151 .007 .008 -.001 -.001

Table 168: Post Hoc Multi- group Nested Model Comparisons assuming model Structural residuals to be correct NFI IFI RFI TLI Model DF CMIN P Delta-1 Delta-2 rho-1 rho2

Measurement residuals 15 16.524 .348 .005 .006 -.002 -.002

301