The College Enrollment Behavior of Class of 1999 Graduates from the : Findings of the Winter 2000 Follow-Up Survey and Comparisons with those for Previous Years

Prepared by: lshwar Khatiwada Andrew M. Sum With Sheila Palma

Center for Labor Market Studies Northeastern University

Prepared for: Boston Private Industry Council April 2001 Table of Contents

College Enrollment Rates ofClass of I 999 Boston Public High School Graduates and Historical Trends in College Enrollment ______2

The Logistic Regression Models ofthe College Enrollment Behavior ofClass of 1999 Boston Public High school Graduates ______8

Estimating the Likelihood ofCollege Attendance Among the Entire Sample ofGraduates from the ofClass of I 999 IO

Predicting the Probability ofCollege Enrollment for Selected High School Graduates From the Class of I 999 I 7

Summary ofFindings and Future Research Recommendations I 9

Appendix A: 2 5

Appendix B: 29

Appendix C: 3 2

(.' ' . ) Introduction

The Boston Private Industry Council (PIC) has been actively engaged in the - formulation and implementation of programs in the education and workforce development arenas in the city of Boston since its establishment in the early 1980s. Every year, since the mid 1980s, the PIC has conducted a follow-up survey of all of the graduates of Boston public high schools from the preceding calendar year. The main objective of the follow-up survey is to obtain information on Boston public high school graduates' transition from school to college and work, including information on their college enrollment and employment status. the types of colleges and post secondary training schools attended, their financial aid status, and key characteristics of their jobs, such as hours of work, hourly wages, occupations and industries of their employers. and I their job-related training activities. During the late winter and early spring of 2000, the Boston PIC made an effort to interview each of the nearly 3,000 graduates of Boston public high schools from the Class of 1999. Follow-up interviews were completed with nearly 80% of the graduates.

This research paper is devoted to an analysis of the findings of the winter 2000 follow-up survey with respect to the college and post-secondary training school enrollment behavior of Class of 1999 graduates at the time of the survey. Findings on these college enrollment rates will be presented for all high school graduates, for men and women, for graduates in each of from major race-ethnic groups, and for graduates from the exam and non-exam high schools. The college/post-secondary training program enrollment rates for graduates from the Class of 1999 also will be compared to those of graduates from preceding years going back to the late l 980's. The paper also will present key findings of a multivariate statistical analyses of the college enrollment behavior of graduates from the Class of 1999, employing a technique known as logistic regression. The model will be used to identify how the college enrollment behavior of Class of 1999 graduates of Boston public high schools associated with their gender, race-ethnicity, types of high school attended (exam and non-exam schools), their participation in school- to-career programs. and their senior year and summer job experiences. We have designed a set of logistic probability regression models to estimate the probability of college enrollment among Class of 1999 Boston public high school graduates from all high schools combined and from the exam and non-exam schools separately.

The findings of this research paper are organized into three main parts. The first part describes the college enrollment status of all graduates and key demographic sub- groups of Class of 1999 Boston public high school graduates, the second part provides findings of the logistic regression models of the probability of college enrollment, and the third part presents a summary of key findings and recommendations for future research.

College Enrollment Rates of Class of 1999 Boston Public High School Graduates and Historical Trends in College Enrollment I

Findings in Chart 1 display historical trends in college enrollment rates of Boston public high school graduates from 1985 to 1999. Over the past fifteen years (1985-1999), the college enrollment rate of Boston public high school graduates has increased by 17 percentage points from 50 percent in 1985 to 67 percent in 1999, a relative increase of 34% over this fourteen year period. This enrollment growth rate of 34 percent between 1985 and 1999 should be considered a very favorable trend in enrollment. The 67% enrollment rate for graduates from the Class of 1999 matches the U.S. national college enrollment rate; however, the graduates from the Boston public high schools college enrollment rates for Black and Hispanic have clearly outperformed the U.S. national rate of college enrollment for the same race-ethnic groups. The overall college enrollment rate for Boston's public schools was at an all time high of 68% in 1997, but the 67.4% rate in 1999 includes findings for several alternative high schools that were not surveyed in prior years. Adjusting for the expanded universe of high schools in the 2000 follow-up survey, the college enrollment rates for the classes of 1997 and 1999 were basically identical.

2 Chart 1: College Enrollment Rates for Boston Public High School Graduates, Classes of 1985 to 1999

66.0% 65% 64.0% 65.8% 64.0'1, 83.0% 62.0'/, 60% j• 0 u c ~%

ec w ~ 50% 50.0'/,

45'/i

40'/, 1985 1986 1987 1988 1989 1990 1991 1992 1993 1995 1996 1997 1998 1999

How have male and female high school graduates fared in transitioning from high school to college in the first year after graduation from high school? Findings of the Winter 2000 follow-up survey revealed that nearly 72% of female graduates from the Class of 1999 were attending a college or post-secondary training institution versus only 62% of the men, a l O percentage point gap in favor of women. Large gender gaps in college enrollment rates have existed since the mid-1990s. (See Chart 3). Chart 2 depicts college enrollment rates of male and female graduates of Boston public high school for selected years over the 1987-99 period. Throughout the 1990s, female graduates' college enrollment rates have outpaced those of men's. In 1987, there was a modest one percentage point difference between the college enrollment rates of men and women in favor of men, but the enrollment gap has reversed direction and widened considerably in the 1990s, reaching an all time high of 12 percentage points in 1998 before narrowing slightly in 1999 to 9.3 percentage points.

3 .... _

Chart 2: Trends in College Enrollment Rates of Boston Public High School Graduates, by Gender, Selected Time Periods, 1987-1999

r- 80% --Men I • Women 1

70% 71 .5%

"C ..! 62.2% 60.8% ec 60% w

0~ 55.2%

50%

1987 1991 1996 1997 1998 1999

Chart 3: Percentage Points Gap Between the College Enrollment Rates of female and Male Graduates, Selected Years: 198 7-1999

14% 12.5% 12%

g. 10% C> !? c: ·5 8% 0.. Cl C) !! c: Cl ~ l 4%

2%

0% 1987 1991 1996 1998 1999

4 Chart 4 displays trends in the college enrollment rates of Boston public high school graduates across four race-ethnic groups for selected years over the I 987-99 time period. In each time period, Asian graduates from Bosto.n public high schools had the highest college enrollment rates, and their college enrollment rate was 82 percent or higher in each year. Following Asian youth were White non-Hispanic youth, whose college enrollment rate varied over a range of 58% to 71% over the past decade. The enrollment rate of Black graduates has increased substantially since the late 1980s rising from 49.0 percent in 1987 to 64.4 percent in 1999, a gain of 15 percentage points. representing a relative increase of 31 percent over this time period. fn 1999, the Black college enrollment rate was just under 65%, or 6 percentage points below that of Whites. Another positive trend in college enrollment rates among Boston public high school graduates is that of Hispanic graduates whose college enrollment rate increased from 53 percent in 1987 to 60.5 percent in 1999, a relative increase of 13 percent over this 12 year period. Their college enrollment rate in 1999 was well above that of their Hispanic counterparts across the country. Despite these impressive gains, their college enrollment rate has remained the lowest among the four race-ethnic groups in the city of Boston.

Chart 4: Trends in College Enrollment Rates of Boston Public High School Graduates, Selected Time Periods, by Race-Ethnic Groups 1987-1999

100%

92.0% 90% 88.3% 88.1% 82.6% 82.9% 84.1% 80%

"O ~ 70% ec: w 60% 0~

50% ~ Black ---White 40% Asian --- Hispanic 30% 1987 1991 1996 1997 1998 1999

5 Do the female-male differences in college enrollment rates prevail in each race- ethnic group and type of high school (exam or non-exam) or are they confined to only a few groups? To answer this question, we analyzed the PIC follow-up survey data on the college attendance behavior of male and female graduates from the Classes of 1998 and 1999 in each race-ethnic group and among both exam and non-exam school graduates. Findings in Table 1 display the differences in college enrollment rates of Class of 1998 and 1999 Boston public high school graduates by gender in each of four race-ethnic groups. In both years, the college enrollment rates of female graduates in each of the race/ethnic groups were greater than those of male graduates. The sizes of the gender differences have changed somewhat over the past two years. The gender gaps in college enrollment rates for men and women within the Black and White race groups have declined in 1999 as compared to 1998 by 5 to 6 percentage points. In contrast, the college enrollment rate differences between male and female graduates of Hispanic and Asian origin have increased between 1998 and 1999. The gender gap for Hispanic high school graduates was quite considerable with female graduates enjoying a near l 8-percentage point higher college enrollment rate. (Chart 5). There was a large difference in the college enrollment gaps between men and women in the exam and non-exam high schools. The gender gap for graduates from the exam schools was negligible with a small advantage for men; however, the college enrollment rate gap between women and men in the non- exam schools during 1999 was close to 10 percentage points.

Table l: Differences in Female and Male College Enrollment Rates Among Class of 1998 and 1999 Boston Public High School Graduates, Total and by Gender, Race & Ethnicity 1998 1999 Female Male Difference Female Male Difference All 72.5% 60.5% 12.0% 71 ,5% 62.2% 9.3% Black 68.6% 54.6% 14.0% 68.3% 59.6% 8.6% White 76.0% 64.3% 11 .7% 73.8% 68.1% 5.7% Asian 89.9% 86.2% 3.7% 872% 80.4% 6.8% Hispanic 61.9% 48.2% 13.7% 68.5% 50.8% 17.6% Exam School 92.1% 92.5% -0.4% Non-exam School 63.7% 53.9% 9.8%

6 ..

Chart: 5 Percentage Points Gap Between the College Enrollment Rates of Female and Male Graduates of the Class of 1999, Total and by Race-Ethnic Origin

20"/o 18% · 17.&°/o c. 1&°/o (!)"' 14% · J!l .. .!: 12% - 0 CL a, 10% C> 8% . c"' -Q) CJ &°/o ...Q) CL 4% 2% 0% All Asian Black, non- Hispanic Wlite,non- Hispanic Hispanic

College enrollment rates of Class of I 999 graduates varied quite considerably by high school. Table 2 displays the percent of Class of 1999 Boston public high school graduates who were enrolled in college or a post-secondary training program by high school. Enrollment rates varied from lows of 48% and 49% for and Egleston High School to highs of 93% and 94% for graduates from and Boston Latin High School, a relative difference of approximately 2 to 1 between the top and bottom performers on this measure.

7 r

Table 2: Percent of 1999 Graduates of Boston Public High Schools Who Were Enrolled in a College or Post-Secondary Training School, by High School (ranked from Highest to Lowest) Rank School % Rank School % Enrolled Enrolled 1 Boston Latin H.S. 94.6% 11 Burke H.S. 6 1.7% 2 Boston Latin Academy H.S. 93.1% 12 English H.S. 60.7% 3 O'Bryant H.S. 86.9% 13 West Roxbury H.S. 58.9% 4 ACC 86.9% 14 Hyde Park H.S. 52.9% 5 Fenway 78.6% 15 Dorchester H.S. 51 .5% 6 Health Careers 72.2% 16 Boston H.S. 51.3% 7 Brighton H.S. 70.5% 17 Mckinley Tech 50.0% 8 H.S. 69.1% 18 Madison H.S. 49.4% 9 Snowden International H.S. 68.3% 19 Egleston 49.2% 10 Charlestown H.S. 65.6% 20 East Boston H.S. 48.3%

The Logistic Regression Models of the College Enrollment Behavior of Class of 1999 Boston Public High school Graduates

To more fully understand the factors influencing the probability of college enrollment among high school graduates from the Class of 1999, we have estimated a set of logistic regression models of college attendance. The logistic regression model is a statistical technique used for estimating models in which the dependent variable is a dichotomous variable such as the college enrollment status of a graduate, which takes on only two val ues, 0 and I, where l represents a graduate who enrolled in college and O represents a graduate who was not enrolled in college. The logistic model is used to help predict the probability whether or not some event will occur. For our analysis, we have estimated several logistic models of college enrollment to get a better understanding of the college enrollment behavior of graduates in both exam and non-exam schools. Our analysis examines the degree to which the probability of college enrollment is influenced by the across gender and race-ethnic origin of graduates, the type of high school they attended, whether they participated in various types of various school-to-career programs and whether they worked in the summers and senior year during their high school years. I The dependent variable in all of the logistic models in our analysis is the college enrollment status of a graduate, which takes on the values O (not-enrolled) and 1 (enrolled). The predictor variables used in our models of college attendance are a set of dummy variables representing the gender and race/ethnicity of graduates, the high schools attended, their participation in school-to-career programs, their summer job experiences, and their senior year work experience. (Please see Appendix A for definitions of all dependent and predictor variables in our models).

The coefficients (Bs) derived from the logistic probability model represent the change in the estimated odds ratio for college enrollment among Class of 1999 Boston public high school graduates from a one unit change in the predictor variable. For example, the coefficient of .867 for Asian youth in Table 3 indicates that the odds ratio of enrolling in college would, ceteris paribus, increase by .87 if the graduate were Asian. The Beta coefficient obtained from the logistic regression model tells us the change in the odds ratio of a graduate's enrollment in college; however, we may be interested in seeing the marginal change in the probability of college enrollment from a change in the characteristic of a graduate. The estimated coefficients in the logistic regression also can be converted into a set of marginal effects on the probability of being enrolled in college. The estimated values of these marginal effects, however, will depend on the values of the other right hand side variables in the model. A standard practice in the literature is to calculate these marginal effects at the means of all right hand side variables. We can, thus, convert the logistic regression coefficients (Bs) into a set of marginal effects by multiplying the value of each B by (P) and (1-P), where P is the percent of high school graduates in the sample who were attending either college or a post-secondary training institution at the time of the Winter 2000 follow-up survey. The estimated marginal effects for each predictor variable are displayed in the last column for each set of regression results. For example, in the model for all high school graduates, the estimated marginal effect for attending Boston Latin High School (BLATINHS) is 0.53, indicating

9 that graduates of this school were 53 percentage points more likely to enroll in college than their counterparts from the city's non-exam schools.

Estimating the Likelihood of College Attendance Among the Entire Sample of Graduates from the of Class of 1999

Our first logistic regression model analyzes the probability of college enrollment for the entire sample of Class of 1999 graduates. Findings are displayed in Table 3. The predictor variables in the model include the gender and race-ethnic origin of the graduate, three exam high school variables, a variable representing the summer job experiences of the graduate, and a variable representing his/her employment experience during the senior year of high school. At this time, we do not have any data on the academic performance and course-taking behavior of graduates during their high school years. When these data become available, they will be incorporated into these logistic regression models as additional explanatory variables. In our present model, the three exam schools are serving as proxies for the academic achievement (reading, math, writing proficiencies) of graduates and their course-taking behavior. The base group for our model is a White, non-Hispanic male who attended one of the non-exam high schools and who did not work either during the summers of his high school years or the senior year of high school. The estimated marginal effects of the right hand side variables on the probability of college attendance are presented in the last column of the table.

The main findings of the analysis can be summarized in the following manner using the marginal effects of each variable. Female graduates were 8 percentage points more likely to enroll in college than their male counterparts. Asian graduates were signifi cantly more likely from White non-Hispanic graduates to attend college while Black and Hispanic graduates were no more or less likely to attend college than Whites. Exam school graduates were considerably more likely to attend college from their non- exam school peers. Graduates from Boston Latin Academy, Boston Latin High School,

10 and O'Bryant Technical High School were 44%, 53%, and 28%, respectively, more likely to enroll in college than their counterparts from the non-exam schools. Graduates who reported that they worked during the summers in high school were 12 percent more likely to enroll in college than their counterparts who did not work during the summer, but work during the senior year had no signifi_cant effect on the probability of college attendance.

Table 3: Findings of the Logistic Regression Analysis of the Probability of College Enrollment of Class of 1999 Boston Public High School Graduates, All High School

1 B Wald Stat p 1-P B*P*(1-P) (Means of Enrollment) (Marginal Effect) FEMALE 0.372 15.4*** 0.675 0.325 0.082 BLACK 0.132 1.1 0.675 0.325 0.029 ASIAN 0.788 14.1*** 0.675 0.325 0.173 HISPANIC 0.003 0.0 0.675 0.325 0.001 BLATINAC 2.007 42.6*** 0.675 0.325 0.440 BLATINHS 2.432 72.1 *** 0.675 0.325 0.534 OBRYANT 1.293 24.4*** 0.675 0.325 0.284 SNYRJOB -0.482 1.2 0.675 0.325 -0.106 SUMMJOB 0.548 16.7*** 0.675 0.325 0.120 MISSRWK -0.27 0.4 0.675 0.325 -0.059 Constant 0.041 0.0 0.675 0.325 0.009 Log Likelihood 2612.2 Degrees of Freedom 10, 2299 Cox & Snell R-Sq 0.121 Nagelker R-Sq 0.169 Chi-Square 297.08*** Base Group is Non-Exam School Graduates. *implies signi fi cant at the . IO level. **implies significant at .05 level. ***implies significant at .oI level

Our second logistic regression model is used to analyze the probability of college enrollment among Class of 1999 graduates from the city's three exam schools only, treating O'Bryant Technical High School as the base high school. Findings of the analysis are displayed in Table 4. None of the gender, race-ethnic, or in-school employment variables had any significant effect on the probability of college anendance

I The Wald statistic is used to test the significance of the individual coefficients in the logistic regression model. The value of the Chi-Square statisricsis used to identify the signifi cance of the explanatory power of the enti re model.

II among graduates of the city's three exam high schools. Unlike the case for all high schools, female graduates from the city's three exam high schools. ceteris paribus, were no more likely to attend college than male counterparts from the same high schools, and Asians, Blacks, and Hispanics were just as likely to attend college as White, non- Hispanic graduates from the city's three exam high schools. The only significant variable in the model was attendance at Boston Latin High School. Graduates from Boston Latin High School were 6 percentage points more likely to enroll in college than their counterparts from O'Bryant High School. The coefficient on this high school variable was significant at the 5 percent level (two-tailed test). The entire model of college attendance for exam school graduates has very limited explanatory power.

Table 4: Findings of the Logistic Regression Analysis of Probability of College Enrollment Among Class of l 999 Boston Public High School Graduates, in the Exam Schools B Wald p 1-P B*P*(1-P) Stat (Means of Enrollment) (Marginal Effect) FEMALE 0.037 0.01 0.921 0.079 0.003 BLACK -0.583 1.87 0.921 0.079 -0.042 ASIAN 0.322 0.36 0.921 0.079 0.023 HISPANIC -0.488 0.81 0.921 0.079 -0.036 BLATINAC 0.533 1.65 0.921 0.079 0.039 BLATINHS 0.86 4.21** 0.921 0.079 0.063 SNYRJOB -4.769 0.09 0.921 0.079 -0.347 SUMMJOB 0.629 1.58 0.921 0.079 0.046 MISSRWK -4.673 0.09 0.921 0.079 -0.340 Constant 6.33 0.16 0.921 0.079 0.461 Log Likelihood 295.4 Degrees of 9, 548 Freedom Cox & Snell R-Sq 0.022 Nagelker R-Sq 0.053 Chi-Square 12.61 Base Group is O 'Bryant High School *implies significant at the . IO level. **implies significant at .05 level. ***implies significant at .0 I level

12 Our third model of college enrollment behavior is based solely on the experiences of the graduates of the city's non-exam schools. Findings of the multivariate stati stical analysis of the probability of college enrollment among Class of 1999 graduates from the non-exam schools are displayed in Table 5. The base high schools for the analysis are East Boston and Madison Park High School. All other non-exam schools are entered into the model as additional predictors. The estimated marginal effects for the right hand side variables reveal that female graduates from the non-exam schools were 10 percentage points more likely to enroll in college than their male counterparts. Each of the three race-ethnic variables had positive signs with statistically significant coefficients except for Hispanic youth, indicating that Asian and Black youth were more likely to attend college than their White non-Hispanic counterparts in the non-exam schools. The probability of college enrollment was 20 percentage points higher among Asian youth than among White youth. A Black youth's probability of college enrollment was 9 percentage points higher than that of White graduates, and the coefficient on this variable was statistically significant at the 5 percent level. Graduates from ACC, Brighton, Burke, Charlestown, English, Snowden, South Boston, and West Roxbury High Schools were significantly more likely to enroll in college than their counterparts from Madison and East Boston High School. All of the coefficients were statistically significant at the 5 percent level or better, and several were quite large. For example, the marginal effects on the probability of college attendance were equal to 47 percentage points for graduates of ACC, 22 percentage points for graduates of Brighton High School, and 19 percentage points for graduates of . Those graduates who worked during the summers in high school were found to be significantly more likely to attend college than their counterparts who did not work during the summers. The marginal effect of the summers job variable was 14.2 percentage points and was significant at the .0 I level. The senior year employment variable had no significant effect.

13 Table 5: Findings of the Logistic Regression Analysis on the School Effects on Probability of College Enrollment of Class of 1999 Boston Public High School Graduates, by Non-Exam School B Wald p 1-P B*P*(1-P) Stat (Means of (Marginal Effect) Enrollment) FEMALE 0.422 17.0*** 0.597 0.403 0.102 BLACK 0.37 5.9** 0.597 0.403 0.089 ASIAN 0.821 11 .5*** 0.597 0.403 0.198 HISPANIC 0.197 1.3 0.597 0.403 0.047 BOSTON HS 0.024 0.01 0.597 0.403 0.006 BRIGHTON 0.898 16.8*** 0.597 0.403 0.216 BURKE 0.487 3. 8** 0.597 0.403 0.117 CHARLEST 0.787 13.6"** 0.597 0.403 0.189 DORCHEST -0.003 0.00 0.597 0.403 -0.001 ENGLISH 0.568 8.1** 0.597 0.403 0.137 HYDE PARK 0.145 0.47 0.597 0.403 0.035 SNOWDEN I 0.657 4.4** 0.597 0.403 0.158 SOUTH BOS 0.681 8.7** 0.597 0.403 0.164 ACC 1.969 21 .3*** 0.597 0.403 0.474 WROXBURY 0.382 4.3**' 0.597 0.403 0.092 FENWAY 1.243 8.7** 0.597 0.403 0.299 NEWMISSI -1 .359 1.36 0.597 0.403 -0.327 MCKENLVO -4.779 0.38 0.597 0.403 -1 .150 MCKENLTE 0.161 0.03 0.597 0.403 0. 039 HEALTHCA 0.745 2. 8* 0.597 0.403 0.179 EGLESTON -1 .149 1.84 0.597 0.403 -0.276 BOSTON EV -5.688 0.18 0.597 0.403 -1.368 HORRMANN -5.489 0.33 0.597 0.403 -1.321 SNYRJOB -0.389 0.68 0.597 0.403 -0.094 SUMMJOB 0.589 16.3*** 0.597 0.403 0.142 MISSRWK -0.201 0.18 0.597 0.403 -0.048 Constant -0.687 2.02 0.597 0.403 -0.165 Log Likelihood 2230.2 Degrees of 26, 1725 Freedom Cox & Snell R-Sq 0.072 Nagelker R-Sq 0.097 Chi-Square 130.6*** Base Group is East Boston High School and lvfadison Hif{h School. •implies significanl at the . IO level. Himplies significant at .05 level. 0 •implics significant at .0 I level

14 Our final model of the college enrollment behavior of graduates from the city's non-exam schools is similar to that of model three, except that it includes fi ve additional variables representing participation in five different school-to-career programs during the high school years. Those five programs were the Academy of Finance, the Academy of Public Service, the Academy of Travel and Tourism, Career Pathways, and Pro Tech. Each of these variables is a dichotomous variable that takes on the value of l if the graduate participated in such a program during the hjgh school years and the value of zero if they did not participate in such a program.

Results in Table 6 display the findings of this last regression model on the probability of college enrollment among graduates of the non-exam schools. Of the five school-to-career programs, the coefficients of two programs (the Academy of Finance and Pro Tech) had statistically significant positive coefficients. Graduates who were participants in the Academy of Finance and the Pro Tech program were 34 and 24 percentage points more likely to enroll in college than their counterparts who did not participate in a school-to-career program.

15 Table 6: Findings of the Logistic Regression Analysis of the Probability of College Enrollment Among Class of 1999 Boston Public High School Graduates. by Non-Exam Schools B Wald p 1-P B*P*(1-P) Stat (Means of Enrollment) (Marginal Effect) FEMALE 0.401 15.0*** 0.597 0.403 0.096 BLACK 0.309 3.9** 0.597 0.403 0.074 ASIAN 0.774 10.1 *** 0.597 0.403 0.186 HISPANIC 0.108 0.391 0.597 0.403 0.026 BOSTON HS -0.084 0.11 0.597 0.403 -0.020 BRIGHTON 0.774 11 .5**1t 0.597 0.403 0.186 BURKE 0.655 6.01** 0.597 0.403 0.158 CHARLE ST 0.767 12.0°1\" 0.597 0.403 0.185 DORCHEST -0.082 0.083 0.597 0.403 -0.020 ENGLISH 0.581 7.25** 0.597 0.403 0.140 HYDE PARK 0.071 0.08* 0.597 0.403 0.017 SNOWDEN I 0.823 6.39** 0.597 0.403 0.198 SOUTH BOS 0.833 11 .2*** 0.597 0.403 0.200 ACC 2.09 23.0*** 0.597 0.403 0.503 WROXBURY 0.518 6.8** 0.597 0.403 0.125 FENWAY 1.317 9.5° 0.597 0.403 0.317 NEWMISSI -1.174 1.003 0.597 0.403 -0.282 MCKENLVO -4.648 0.359 0.597 0.403 -1 .118 MCKENLTE 0.316 0.096 0.597 0.403 0.076 HEALTHCA 0.909 4.0** 0.597 0.403 0.219 EGLESTON -0.952 1.243 0.597 0.403 -0.229 BOSTON EV -5.526 0.168 0.597 0.403 -1 .330 HORRMANN -5.337 0.315 0.597 0.403 -1 .284 SNYRJOB -0.508 1.136 0.597 0.403 -0.122 SUMMJOB 0.502 11.5°* 0.597 0.403 0.121 MISSRWK -0.282 0.352 0.597 0.403 -0.068 FIN1 1.434 7.0* 0.597 0.403 0.345 PBSERV1 0.715 2.476 ' 0.597 0.403 0.172 PROTECH1 1.003 19.3*** 0.597 0.403 0.241 TRAVTOR1 0.19 0.267 0.597 0.403 0.046 CARPATH1 0.099 0.403 0.597 0.403 0.024 Constant -0.597 1.52 0.597 0.403 -0.144 Log Likelihood 2198.0 Degrees of 31 , 1720 Freedom Cox & Snell R-Sq 0.089 Nagelker R-Sq 0.12 Chi-Square 162.81 Base Group is East Boston Hiqh School and Madison High School. *1mp1J~s significant at . 10 level ..implies s1gn 1ficant at .05 level ***implies significant at .01 level

16 Predicting the Probability of College Enrollment for Selected High School Graduates From the Class of 1999

The findings of the logistic regression models also can be used to predict the probability of college enrollment for high school graduates with given demographic, schooling, and work experience characteristics. The specific formulas used to generate these probability estimates are explained more fully in Appendix C. We have selected seven hypothetical high school graduates from the Class of 1999 to illustrate the range in the expected probabilities of being enrolled in college at the time of the winter 2000 follow-up survey. These seven hypothetical graduates have varying combinations of gender and race/ethnic characteristics, schooling backgrounds, and work experiences. The traits of each of these seven graduates and their predicted probabilities of college attendance are displayed in Table 7.2

The predicted probabilities of college attendance for these seven hypothetical graduates varied considerably, ranging from a low of 35% to a high of nearly 99%. The highest predicted probabilities of college enrollment were for the three hypothetical graduates from the three exam schools. A female, Asian graduate from Boston Latin High School who worked during the surnrners of her high school years had a predicted probability of college attendance of 98.5%. Tne two other hypothetical graduates from O' Bryant Technical High School and Boston Latin Academy also had very high predicted probabilities of college attendance, 92% and 89%. respectively (Table 7).

The four hypothetical graduates from the non-exam schools had quite variable predicted probabilities of college attendance, ranging from 35% to nearly 82%. The probability of college attendance was considerably affected by the high school they attended, their race/ethnic origin, their work experiences in high school, and their

2 The estimated probabilities of college attendance for the first three graduates are based on the findings of the logistic regression model for all high school graduates. The predicted probabilities of college enrollment for the other four graduates are based on the findings of the logistic regression model for the non-exam school graduates (model # 4).

17 participation in school to career programs. The hypothetical Black male, who graduated from Brighton High School, had substantial work experiences in high school, and participated in a Pro-Tech program had an 82% predicted probability of attending college. The hypothetical Asian male who attended Snowden International High School had a similar high probability of college attendance (82%). In contrast, the two hypothetical males who attended East Boston and Madison Park High School, had no substantial work experience during high school, and clid not participate in any school-to- career program had a predicted probability of college attendance of only 36 to 43 per cent (Table 7).

The academic backgrounds, academic performance, and course-taking behavior of these hypothetical students likely varied to a considerable degree across these seven high schools. The estimated, substantial independent effects of the high school attended on college attendance, thus, likely represent differences in the academic proficiencies of these graduates at the beginning and end of their high school years and their course taking behavior. The availability of the BPS transcript data for individual students should help us identify the independent influence of academic achievement, grade point averages, and high school courses on the college enrollment behavior of recent graduates from the city's public high schools.

I& Table 7: Predicted Probabilities of College Attendance Among a Selected Hypothetical Set of Boston Public High School Graduates from the Class of 1999

Student Characteristics Predicted Probability

Asian, female, attended Boston Latin High 98.5% School, worked during the summers, but not during senior year of high school. Hispanic, male, attended Boston Latin 89.2% Academy, worked during the summer and senior year of high school Black, female, attended O' Bryant 91.6% Technical High School, worked during the summer, but not the senior year of high school. Black, male, attended Brighton High 81.5% School, worked during the summer and senior year, participated in a Pro-Tech program. Asian, male, attended Snowden 81.8% International High School, worked during the summer, but not during senior year. White, male, attended East Boston High 35.5% School, did not work in the summer or senior year, and did not participate in a school-to-career program. Black, male, attended Madison Park, did 42.8% not work in summer or senior year, did not participate in a school-to-career program.

19 Summary of Findings and Future Research Recommendations

The findings appearing in this research paper have pro~ided both a description and an analysis of the college enrollment behavior of Class of 1999 Boston public high school graduates and a statistical analysis of the factors associated with higher and lower rates of college attendance among these graduates. Slightly more than two-thirds of the graduates from tbe Class of 1999 were attending a post-secondary education or training program at the time of the Winter 2000 follow-up survey. This college enrollment rate essentially matched the previous historical high of 68.0% that was achieved by graduates from the Class of 1997 and was nearly 18 percentage points higher than the college enrollment rate of graduates from the city's public high schools in 1985. Women's college enrollment rates both in the aggregate and across all four of our race-ethnic groups outpaced the college enrollment rate of men, continuing a trend since the early 1990s. Approximately 72% of female graduates from the Class of 1999 were attending a college or post-secondary training institution at the time of the winter follow-up survey versus only 62% of the men, a gender gap of nearly 10 percentage points. The gender gap has been in the 7 to 12 percentage point range since 1996, radically reversing the modest male college enrollment rate advantage that existed in the mid to late 1980s. Women in each of the four major race-ethnic groups were more likely than their male counterparts to be attending college, and the size of these college enrollment rate gaps varied from 6% among White, non-Hispanics to a high of nearly 18% among Hispanics.

Among graduates of the city's three exam schools, college enrollment rates of men and women were essentially identical; however. among graduates of the district and other non-exam schools, the college enrollment rate among women was l O percentage points higher than that of men. Women's overall college enrollment rate advantage is al_so partly attributable to the fact that they accounted for a clear majority 61 % of the graduates from the exam schools. College enrollment rates of Asian youth have exceed those of graduates in the other three race-ethnic groups. While 84% of Asian graduates from the Class of 1999 were attending college or a post-secondary training institutjon, only 65% of Black graduates and 61 % of Hispanic graduates were doing so. Since the

20 late I 980s, however, there has been a narrowing of the gaps between the college enrollment rates of Asian graduates and those of Black, Hispanic, and White graduates. Over the 1987-99 period, college enrollment rates for Black and Hispanic graduates rose by 15 and 11 percentage points, respectively.

The college enrollment rate of graduates from the city's exam schools was slightly over 92% while only 58% of the graduates from the city's district and other non- exam schools were enrolled in a college or post-secondary training institution. Across individual high schools, college enrollment rates ranged from lows of 48 to 49 percent in East Boston High School, Egleston, and Madison Park to highs of 93% in Boston Latin Academy and nearly 95% in Boston Lati n High School.

Findings based on our multivariate statistical analysis of the college enrollment decisions of Class of 1999 graduates revealed that the probability of attending college was significantly influenced by the high school that they attended, their gender, their race/ethnic origin, their summer job experiences, and their participation in selected types of school-to-career programs while in high school. High school graduates from each of the city's three exam schools (Boston Latin, Latin Academy, and the O' Bryant Technical School) were significantly more likely to attend college than their counterparts attending other high schools. Among graduates from the district and other non-exam high schools, those who graduated from ACC, Brighton, Charlestown, English, Fenway, Snowden International, South Boston, and West Roxbury High Schools were significantly more likely to attend college than their counterparts from the other non-exam schools.

Women were significantly more likely than male graduates from the Class of I 999 to attend college. This gender difference prevailed in the model for all high school graduates and in the model for graduates from the non-exam high schools, but there was no significant gender difference in college attendance between female and male graduates from the city's exam schools. Asian graduates, holding all other variables constant, were found to be more likely than Whites to be enrolled in college.

1 1 Those high school graduates who worked during the summers of their high school years were found to be significantly more likely to attend college than their peers with no summer work experience while work during the senior year had no significant effect on the probability of college enrollment. Among non-exam school graduates, those youth who participated in the Academy of Finance program or a Pro Tech program were significantly more likely to attend college or a post-secondary training institution than their peers who did not participate in a school-to-career program while in high school.

There are a number of future research activities that should be undertaken to improve our knowledge base on the college enrollment behavior and academic experiences of Boston public high school graduates in recent years. First, data from the school transcripts of Boston public high school graduates should be integrated into the date base from the exit and follow-up surveys to allow a more thorough analysis of the influence of graduates' high school courses, grade point averages, and achievement test scores on their post-secondary educational and training activities. National research has repeatedly revealed a key role played by high school curriculum, educational aspirations and expectations, grade point averages, and achievement test scores on the college enrollment and retention behavior of high school graduates.3 How do the high school academic behaviors and experiences of Boston high school graduates influence their college enrollment behavior in the early years after graduation?

Second, there is an immediate need to more fully understand the large gender gaps in college enrollment rates that have prevailed between men and women since the mid- l 990s, especially among Black and Hispanic youth. There are several separate, but related sets of research questions that should be addressed. What accounts for the disproportionate number of female graduates from the city's three exam schools? Of the

3 See: {i) Andrew M. Sum and W. Neal Fogg, "The Adolescent Poor and the Transition to Early Adulthood,·· in Adolescence and Poverty: Challenge for the 1990s, (Editors: Peter Edelman and Joyce Ladner), Center for National Policy Press, Washington, D.C., 1991 ; (ii) National Center for Education Statistics, High School and Beyond, A National Longitudinal Survey for the 1980s, Two Years in High School: The Status of 1980 Sophomores in 1982, U.S. Government Printing Office, Washington, D.C., 1984; (iii) Andrew M. Sum, Neeta Fogg, and Garth Mangum, Confronting the Youth Demographic Challene.e: Labor Market Prospects of At-Risk Youth, Sar Levitan Center for Social Policy Studies, Johns Hopkins University. Baltimore, 2000.

22 740 graduates from the city's three exam high schools combined, women accounted for 453 or 61 %. In each of the three exam schools, women represented a solid majority of the Class of 1999 graduates, with their shares ranging from just under 59% in Boston Latin High School to 64% in Latin Academy. Are young women more successful than men in passing the entry exams for admission into these high schools? rs the retention rate among exam school students significantly higher among women than men? Do women in the non-exam high schools achieve higher course grades and academic achievement than men? For a given level of academic performance are women from the non-exam schools significantly more likely to attend college than men? The considerably higher dropout rate of male high school students also needs to be more carefully investigated. Of the 2,660 male freshmen in the 1995-96 school year, only 1,303 or 49% graduated with the Class of 1999 versus 68% of the women. Do these young men transfer to other high schools inside and outside the city at higher rates than women or are they simply dropping out of high school at much higher rates than women? These critical research questions need to be more fully investigated, and strategies for reducing the low high school retention rates and below average college attendance rates of men need to be developed.

While the college enrollment rates of Black and Hispanic high school graduates have improved markedly over the past decade, there remain important gaps in college enrollment rates among graduates in the four race-ethnic groups. Among Class of 1999 graduates, the college enrollment rates of Hispanic and Black graduates were 24 and 20 percentage points, respectively, below those of Asian graduates, whose college enrollment rates were the highest at 84%. What types of strategies, including academic enrichment, after school tutoring, mentoring, and college counseling might the Boston public schools pursue to improve the college going rates of Black and Hispanic graduates, especially men?

Finally, there is a need to extend the length of existing follow-up surveys to track the college retention and academic performance of recent graduates over longer time periods. The existing follow-up survey currently tracks the college enrollment status of

23 graduates approximately 9 to 10 months after their initial date of graduation from high school. The follow-up survey, thus, captures information on their enrollment status and fmancial aid status about halfway through the spring semester of their freshman year. No information is currently collected on their academic performance or course credits. The Boston Private Industry Council together with research staff from the Boston Public Schools and the Center for Labor Market Studies should work with all - based colleges and universities to obtain administrative record-based information on the college retention and academic performance of all BPS graduates in recent years, tracking progress initially through their first two full years of college and then extending the follow-up through four full years. Personal follow-up interviews with a representative sample of retained college students and dropouts should be conducted to gain more personal insights into their college experiences and performance.

24 Appendix A:

Definitions of Variables Appearing in the Logistic Probability Models of College Enrollment for Class of 1999 Graduates of Boston Public High Schools Variable Name Definition COLENR A dummy variable representing the college or post- secondary training enrollment status of a h igh school graduate from the Class of 1999 =1 if enrolled in a post-sec011dary education or training program =O if not enrolled

FEMALE A dummy variable representing the gender of a high school graduate =I if female =O if male

BLACK A dummy variable representing the race-ethnicity of a graduate from the Class of 1999 = l if graduate is Black, not Hispanic =O if other

WHITE A dummy variable representing the race-ethnicity of a graduate from the Class of 1999 =l if graduate is White, not Hispanic =O if other

ASlAN A dummy variable representing the race-ethnicity of a graduate from the Class of 1999 =1 if graduate is Asian =O if other

HISPANIC A dummy variable representing the race-ethnicity of a graduate from the Class of 1999 = l if graduate is Hispanic =O if other

BLATINAC A dichotomous variable representing the Boston Latin Academy = 1 if the school was the Boston Latin Academy =O if else

BLATIHS A dichotomous variable representing the Boston Latin High School = l if the school was Boston Latin High School =O if else

25 Variable Name Definition

OBRYANT A dichotomous variable representing the O'Bryant Technical High School =l if the school was O'Bryant Technical High School =O if else

BRIGHTON A dichotomous variable representing Brighton High School =l if the school was Brighton High School =O if else

BOSTON HS A dichotomous variable representing Boston High School =1 if the school was Boston High School =O if else

BURKE A dichotomous variable representing the Burke High School = l if the school was Burke High School =O if else

CHARLEST A dichotomous variable representing Charlestown High School =I if the school was Charlestown High School =O if else

DORCHEST A dichotomous variable representing Dorchester High School =I if the school was Dorchester High School =O if else

EASTBOST A dichotomous variable representing East Boston High School =I if the school was East Boston High School =O if else

ENGLISH A dichotomous variable representing English High School =I if the school was English High School =O if else

HYDEPARK A dichotomous variable representing Hyde Park High School =l if the school was Hyde Park High School =O if else

26 Variable Name Definition

MADISON A dichotomous variable representing Madison Park High School =I if the school was Madison Park High School =Oif else

SNOWDEN A dichotomous variable representing Snowden International High School =1 if the school was Snowden International High School =Oif else

SOUTHBOS A dichotomous variable representing South Boston High School =l if the school was South Boston High School =O if else

ACC A dichotomous variable representing ACC High School =I if the school was ACC High School =O if else

WROXBURY A dichotomous variable representing West Roxbury High School =1 if the school was West Roxbury High School =O if else

FENWAY A dichotomous variable representing Fenway High School =l if the school was Fenway High School =O if else

HEALTHCA A dichotomous variable representing the Health Care Academy =l if the school was Healthcare Academy =O if else

EGLESTON A dichotomous variable representing Egleston High School =I if the school was Egleston High Schoo] =O if else

SUMMJOB A dichotomous summer job variable =l if graduate held a summer job during the high school years. =O if no summer job held.

27 Variable Name Defi11itio11

SNYRJOB A dichotomous senior year job variable =1 if graduate h_eld a job during the senior year in high. school. =O if no job was held in the senior year of high school.

FINl A dichotomous variable representing participation in a School-to-career program in high school. =l if graduate participated in an Academy of Finance program =O if else

PBSERVI A dichotomous variable representing participation in a School-to-career program in high school. = l if graduate participated in an Academy of Public Service program. =O if else

TRAVTOR A dichotomous variable representing participation in a school-to-career program in high school. = I if graduate participated in an Academy of Travel and Tourism program =O if else

CARPATH A dichotomous variable representing participation in a School-to-career program in high school. =l if graduate participated in a Career Pathway program =O if else

PROTECHl A dichotomous variable representing participation in a. School-to-career program in high school =l if graduate participated in a Pro Tech program =O if else

MISSRWK A dichotomous variable depicting whether weeks worked by a graduate during the senior year was missing =1 if weeks worked was missing =O if else

18 Appendix 8:

Means and Standard Deviation of variables Used in the First Logistic Regression Model Variables Mean Standard Deviation N COLENR 0.676 0.468 2310 FEMALE 0.560 0.497 2310 BLACK 0.470 0.499 2310 ASIAN 0.111 0.314 2310 HISPANIC 0.188 0.391 2310 OBRYANT 0.060 0.237 2310 BLATINAC 0.071 0.256 2310 BLATINHS 0.111 0.315 2310 SNYRJOB 0.701 0.458 2310 SUMMJOB 0.842 0.365 2310 MISSRWK 0.289 0.454 2310

Means and Standard Deviation of variables Used in the Second Logistic Regression Model Variables Mean Standard Deviation N COLENR 0.921 0.270 558 FEMALE 0.620 0.486 558 BLACK 0.285 0.452 558 ASIAN 0.215 0.411 558 HISPANIC 0.106 0.308 558 OBRYANT 0.247 0.432 558 BLATINAC 0.292 0.455 558 BLATINHS 0.461 0.499 558 SNYRJOB 0.638 0.481 558 SUMMJOB 0.875 0.332 558 MISSRWK 0.358 0.480 558

29 Means and Standard Deviation of variables Used in the Third Logistic Regression Model Variables Mean Standard Deviation N COLENR 0.598 0.490 1752 FEMALE 0.541 0.499 1752 BLACK 0.529 0.499 1752 ASIAN 0.078 0.268 1752 HISPANIC 0.214 0.410 1752 BOSTON HS 0.057 0.232 1752 BRIGHTON 0.080 0.272 1752 BURKE 0.055 0.228 1752 CHAR LEST 0.091 0.288 1752 DORCH EST 0.055 0.228 1752 ENGLI SH 0.094 0.292 1752 HYDEPARK 0.075 0.264 1752 SNOWDEN I 0.033 0.178 1752 SOUTH BOS 0.072 0.259 1752 ACC 0.030 0.171 1752 WROXBURY 0.116 0.320 1752 FENWAY 0.021 0.144 1752 NEWMISSI 0.002 0.048 1752 MCKENLVO 0.002 0.041 1752 MCKENLTE 0.002 0.048 1752 HEALTHCA 0.016 0.125 1752 EGLESTON 0.004 0.063 1752 BOSTON EV 0.001 0.024 1752 HORRMANN 0.001 0.034 1752 SNYRJOB 0.722 0.448 1752 SUMMJOB 0.832 0.374 1752 MISSRWK 0.267 0.443 1752

30 Means and Standard Deviation of variables Used in the Fourth Logistic Regression Model Variables Mean Standard Deviation N ENROLLED 0.598 0.490 1752 FEMALE 0.541 0.499 1752 BLACK 0.529 0.499 1752 ASIAN 0.078 0.268 1752 HISPANIC 0.214 0.410 1752 BOSTON HS 0.057 0.232 1752 BRIGHTON 0.080 0.272 1752 BURKE 0.055 0.228 1752 CHARLE ST 0.091 0.288 1752 DORCH EST 0.055 0.228 1752 ENGLISH 0.094 0.292 1752 HYDE PARK 0.075 0.264 1752 SNOWDEN I 0.033 0.178 1752 SOUTH BOS 0.072 0.259 1752 ACC 0.030 0.171 1752 WROXBURY 0.116 0.320 1752 FENWAY 0.021 0.144 1752 NEWMISSI 0.002 0.048 1752 MCKENLVO 0.002 0.041 1752 MCKENLTE 0.002 0.048 1752 HEALTHCA 0.016 0.125 1752 EGLESTON 0 ..004 0.063 1752 BOSTON EV 0.001 0.024 1752 HORRMANN 0.001 0.034 1752 SNYRJOB 0.722 0.448 1752 SUMMJOB 0.832 0.374 1752 MISSRWK 0.267 0.443 1752 FIN1 0.014 0.119 1752 PBSERV1 0.018 0.134 1752 PROTECH1 0.084 0.278 1752 TRAVTOR1 0.021 0.142 1752 CARPATH1 0.221 0.415 1752

31 Appendix C:

Estimating the Probability of a Graduate with Given Background Traits Being Enrolled in College at the Time of the Winter Follow-up Survey

The logistic regression models that were used to analyze the college enrollment status of Class of 1999 graduates provided estimates of the sign and size of the influence of individual variables on the log of the odds of enrolling in college. The coefficients on the independent variables also were converted into estimated marginal effects evaluated at the means of all of the predictor variables in the model. One can also use the findings of the logistic regression model to estimate the probability of a given high school graduates being enrolled in college at the time of the follow-up survey.

The procedure for estimating the probability of college enrollment for a graduate with given traits is relatively straightforward. The probability that a given high school graduate will be attending college is equal to the following:

p , l I + e a + /3 x

To solve for Pi , we begin by calculating the value of a + f)x for an individual with given traits, Xi (i.e., gender, race-ethnic origin, high school attended). The values of a and P's are those generated by the logistic regression model. We then calculate the value of ea+flxi. The value of the denominator is simply equal to l + ea+flxi. The ratio of these two values would then yield the estimated probability of college attendance for this individual.

To illustrate how this procedure id used, let us use the findings from our first model of college enrollment decision for all high school graduates in Table 3. Let us take a White, non-Hispanic female who graduated from the Boston Latin Academy High

32 School, who worked during the summers while in high school, but did not work during the senior years of high school. The value of a + /Jx; for this given individual was 2.968. Raising e to the power 2.968 yields a value of 19.45. The denominator for the probability formula is 20.45. The ratio of these two values is equal to .951 , implying that the probability of college attendance on this female graduate from the Boston Latin Academy High School was equal to 95.5%. These probability estimates can be derived for any given individual with a known set of demographic, schooling, and work experience traits.

33