Height, Human Capital and Economic Growth

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Andreas Michael Schick, M.A. Graduate Program in Economics

i i The Ohio State University 2011

Dissertation Committee: Richard H. Steckel, Advisor Trevon Logan Bruce Weinburg

i i i

Abstract

Why are some countries richer than others? Modern growth theories assign a

pivotal role to human capital accumulation, particularly in the nineteenth and early

twentieth centuries. Unfortunately, testing this conjecture is relatively impractical due to

data limitations. Standard human capital measures, such as cognitive test scores and years

of schooling, are generally unavailable, while more data-abundant alternatives, such as

literacy rates, poorly proxy human capital. To mitigate this problem, this thesis presents

evidence that adult height is a relatively suitable proxy for human capital. Using data

from the early and late twentieth century, I show that taller individuals and populations

were, and still are, substantially more productive than their shorter counterparts.

i i Furthermore, my results attribute this outcome to the significant association between

adult height and both cognitive and non-cognitive abilities.

Chapter 1 studies the contemporary pathways through which taller workers earn

more than their shorter counterparts. Earlier studies attribute this ―stature premium‖ to

non-cognitive abilities, which are associated with height and rewarded in the labor

market. While more recent research suggests that cognitive abilities cause the stature-

wage relationship. This paper reconciles the competing views by recognizing that net

nutrition, a major determinant of adult height, is integral to cognitive and non-cognitive

development. Using contemporary British data from the National Childhood

Development Study (NCDS), I show that taller children have higher average cognitive and non-cognitive test scores, and that each aptitude accounts for a substantial and roughly equal portion of the stature premium. Together, cognitive and non-cognitive abilities explain the height premium.

Chapter 2 examines the pathways contributing to the stature premium in a historical labor market, early twentieth century Canada. The literature proposes two pathways with which stature promoted productivity in the past. First, height contributed to productivity via its direct association with various physical abilities, such as strength, endurance and motor skill. Second, the contemporary pathways attribute the stature premium to cognitive and non-cognitive abilities. To test these arguments, I construct a linked Canadian data set that matches men across their 1891 and 1911 census manuscripts and World War I registration papers. I show that taller workers earned substantially more than their shorter counterparts. Furthermore, the results indicate that only cognitive and non-cognitive abilities provide a substantial contribution to the stature premium.

Chapter 3 presents evidence that national health provides an independent causal contribution to economic growth. Good health is thought to promote national productivity via its correlation with physical capacity and human capital accumulation. To test these claims, I construct a cross-country data set covering 45 countries in 1960 to 2000. I measure a nation’s health status using a standard proxy: the average adult height among its prime-age men. I show that taller populations are substantially more productive than

iii

their shorter counterparts, and that this result is due, in large part, to a strong correlation

between adult height and accumulating human capital.

v i

Acknowledgements

First, I would like to thank my advisor, Richard Steckel, whose enthusiasm motivated me even in the darkest days of winter. Research is a marathon, and sometimes the hardest part is continuing to move forward. Thank you for giving me the encouragement necessary to endure. Without it, I would have never understood what it truly means to enjoy academics.

I am very appreciative of the contributions provided by Trevon Logan. Trevon always gave me great advice, usually through the use of three pen colors. I still do not know what the red, green and blue pen marks represent. However, they played a pivotal role in improving my abilities as a researcher.

I am also grateful to Bruce Weinberg, who always made the time to talk to me about my research and empirical techniques. Bruce’s comments were very good, and provided me with excellent insight and clear intuition.

I would also like to thank Ben Baack, whose ear was always open and willing to listen. In addition, I am grateful to Kevin Pflum and Matt Anderson. Thank you for reading my drafts and suggesting comments.

Finally, last but not least, I want to thank my family—my parents, grandmother, and sister, Elizabeth. Without them, I would have never made it this far.

Vita

November 5, 1982……………………...Born in Lansing, Michigan

June 2006…………………...... B.S. Economics, Binghamton University

June 2007…………………...... M.A. Economics, The Ohio State University

2007-Present…...... Graduate Teaching Associate, Department of

Economics, The Ohio State University

Fields of Study

Major Field: Economics

i v

Table of Contents

Section Page

Abstract ...... ii

Acknowledgments...... v

Vita ...... vi

List of Tables ...... ix

List of Figures ...... xi

Chapters:

1. Understanding the Height Premium: Cognitive and Non-Cognitive Ability...... 1

1.1 Introduction ...... 1 1.2 Height and Non-Cognitive Ability...... 4 1.3 Non-Cognitive Ability and Productivity ...... 7 i i v 1.4 Empirical Framework ...... 9 1.5 Data ...... 11 1.6 Results ...... 14 Height, Cognition and Social Skills ...... 14 Height and Earnings ...... 15 Discrimination or Stature? ...... 17 1.7 Conclusion ...... 20

2. A Historical Exploration of the Height Premium: A Case Study of Canada in 1911 ...... 33 2.1 Introduction ...... 33 2.2 Height, Cognitive Ability, and Non-Cognitive Ability ...... 35 2.3 Cognitive Ability, Non-Cognitive Ability and Productivity...... 40 2.4 Data ...... 43 2.5 Empirical Framework ...... 50 2.6 Results ...... 52 Height and Productivity ...... 52 Discrimination, Nepotism, Height and Wages ...... 55 vii

2.7 Conclusion ...... 56

3. Health Makes Wealth: The Relationship between National Productivity and Health ...... 62

3.1 Introduction ...... 62 3.2 Health and Productivity ...... 64 3.3 Health Indicators ...... 66 3.4 Empirical Framework ...... 71 3.5 Data ...... 76 3.6 Results ...... 81 Ordinary Least Squares Results ...... 81 First Stage Results...... 83 Main Results ...... 84 Economic Levels versus Economic Growth ...... 86 3.7 Conclusion ...... 87

References ...... 97

Appendices

A: Appendix for Chapter 2 ...... 117

A.1 The Correlation between Occupational Aptitudes and Ability ...... 117 A.2 Constructing a Historical Data Set ...... 121 Linking Canadian Males to their Childhood and Military Service Records ...... 121 Sample Representativeness ...... 122 Linking Biases, Data Quality Issues, Limitations and Corrections ...... 124 Name of Spouse & Name of Parent Bias ...... 124 Height Data Bias ...... 125 Wage Errors ...... 125

viii

List of Tables

Table Page Page

1.1 Descriptive Statistics in Post-Industrial Britain ...... 24

1.2 Cognitive Test Scores and Height in Childhood...... 25

1.3 Non-Cognitive Test Scores and Height in Childhood ...... 26

1.4 Log Average Hourly Earnings, Cognition, Non-Cognitive Ability and the Returns to Height (Men)...... 27

1.5 Log Average Hourly Earnings, Cognition, Non-Cognitive Ability and the Returns to Height (Women) ...... 28

1.6 Male Occupational Placement and Stature ...... 29

1.7 Log Average Hourly Earnings, Cognition, Non-Cognitive Ability, Beauty and the Returns to Height (Men) ...... 30

2.1 Descriptive Statistics in Historical Canada ...... 58

2.2 Log Average Hourly Earnings, Cognitive Skill, Non-Cognitive Skill and the Returns to Height in Historical Canada ...... 59

2.3 The Relationship between Earnings and Nepotism ...... 60

3.1 The Relationship between Health and Economic Performance ...... 89

3.2 Average Adult Height among Prime-Age Men ...... 92

3.3 Descriptive Statistics across Countries ...... 93

3.4 The Relationship between Insolation and Skin Color to Adult Height ...... 94

3.5 The Relationship between Economic Growth, Health, Physical Capacity and

Human Capital ...... 95

3.6 The Relationship between Economic Levels, Health, Physical Capacity and Human

Capital (2SLS) ...... 96

A.1 Factor Analysis on Second Edition Dictionary of Occupational Title Variables ......

...... 127

A.2 The Relationship between Cognitive Ability and Cognitive Occupational Skill ......

...... 129

A.3 The Relationship between Non-Cognitive Ability and Non-Cognitive

Occupational Skill ...... 129

A.4 Hourly Earnings, Height, Actual Cognitive and Non-Cognitive Ability, and

Occupational Requirements in Post-Industrial Britain ...... 130

A.5 Relationship between Linking Success and Various Individual Characteristics

...... 131

List of Figures Table Page

1.1 The Biological Link between Stature and Ability ...... 31

1.2 Log Earnings and Stature for Men in Post-Industrial Britain ...... 32

2.1 Log Earnings and Stature for Men in Historical Canada ...... 61

i x

Chapter 1

Understanding the Height Premium: Cognitive and Non-Cognitive†

1.1 Introduction

Taller workers receive a notable wage or earnings premium in developing

economies where size, strength and stamina enhance labor productivity (Steckel, 1995;

Strauss and Thomas, 1998). The persistence of the relationship in modern industrial

economies, where most occupations are sedentary, calls for a different explanation.

Recent papers link physical growth in early childhood to the formation of cognitive skills

that are rewarded in labor markets (Case and Paxson, 2008; Heineck, 2009). A

somewhat longer tradition suggests that stature is correlated with non-cognitive ability

(Stogdill, 1948; Baker and Redding, 1962; Adams, 1980; Judge and Cable, 2004; Persico

et al. 2004). This paper measures the returns to stature in present-day Britain, and

examines the extent to which cognitive and non-cognitive abilities contribute to the

height premium.

It is well documented that stature is associated with personal and economic

success. Gaius Julius Caesar, for example, attributed Rome’s success in part to their

† This chapter was co-authored with my adviser, Professor Richard Steckel. 1

great stature. The prominent 19th century physician, Julien-Joseph Virey, argued that

taller individuals were more motivated and industrious (Hall, 2006). In western countries,

an increase in adult height from the 25th to the 75th percentile of the height distribution—

an increase of approximately 4 to 5 inches—is associated with a 9 to 15 percentage point

increase in earnings (Judge and Cable, 2004; Persico et al. 2004; Heineck, 2005; Case

and Paxson, 2008; Hubler, 2009). This return is roughly as large as completing an

additional 1 to 2 years of schooling.

Most studies attribute the height premium to a strong correlation between height

and non-cognitive ability. Persico et al. (2004) argue that this mechanism is formed in

the early teenage years, at ages when adolescents believe that taller individuals excel at

sports. Peer pressure leads taller individuals to join sports teams where they are more

likely to accumulate productive non-cognitive abilities (social skills) such as team work,

2 discipline, confidence, and leadership.

Persico et al. (2004) test the social-skills hypothesis using data from the National

Longitudinal Survey of Youth 1979, which shows that individuals who were taller in

adolescence receive a wage premium. In these data a one inch increase in adolescent

stature is associated with a 2.3 percentage point increase in adult wages. They next

consider the extent to which participation in social activities, such as sports and academic

clubs, contribute to the relationship. Controlling for social participation reduces the

height coefficient approximately 20 percent. A one inch increase in adolescent height is

now associated with a 1.8 percentage point increase in wages. These results suggest that

non-cognitive abilities play a moderate role explaining the returns to stature. However,

their analysis may understate the extent to which non-cognitive abilities contribute 2

because social participation may inadequately capture the various non-cognitive abilities

associated with height.1

Case and Paxson (2008) recently challenged the social-skill hypothesis,

suggesting that cognitive, rather than non-cognitive abilities, explain the stature premium.

They argue that stature is strongly associated with cognition via a biological pathway:

insulin-like growth factors. These channels, they argue, stimulate simultaneous neural

and physical growth, and also develop neurological regions which manage cognitive

capacity. Using data from the National Childhood Development Study (NCDS), the

authors show that taller men receive an earnings premium, and that this premium is due,

in large part, to the strong correlation between height and cognitive ability. A one inch

increase in male stature is associated with a 2.3 percentage point increase in earnings.

Including cognitive test scores into their analysis reduces the stature-earnings relationship

3 approximately 45 percent. A one inch increase in stature is now associated with a 1.3

percentage point increase in earnings. The reduction is substantial; however, the premium

remains moderately sizeable and statistically significant at the 5 percent level, suggesting

that other characteristics, such as non-cognitive abilities, may provide an important

independent contribution to the relationship.

This paper makes two contributions to the literature, first by presenting evidence

that taller children are more cognitively and non-cognitively able. Using data from the

NCDS, we find that a one standard deviation increase in childhood stature

(approximately 2 to 2.5 inches) is associated with a 0.10 standard deviation increase in

1 These characteristics are authority, communication, confidence, courtesy, discipline, ethical conduct, motivation, optimism persuasion, sociability. Persico et al. (2004) did not include these measures in their analysis due to data limitations. 3

cognitive and non-cognitive test scores. These effects are as large as growing up in

middle class family versus a lower class family.

Second, the paper measures the extent to which cognitive and non-cognitive

abilities contribute to the statue-earnings relationship. Using data from the NCDS, we

show that taller workers earn substantially more than their shorter counterparts. For men,

a one inch increase in adult stature is associated with a 2.2 percentage point increase in

adult earnings. Separately including either cognitive or non-cognitive assessments into

the analysis reduces the stature-earnings relationship roughly the same amount—from 2.2

percent to 0.9 percentage points. Including non-cognitive measures, in addition to

cognitive test scores, reduces the relationship to approximately zero—from 0.9 to 0.5

percentage points—and renders it statistically insignificant. These results indicate that

neither aptitude separately explains the entire stature premium; rather, both abilities are

4 necessary to account for the entire relationship

1.2 Height and Non-Cognitive Ability

A popular view links stature with various personality traits that promote worker

productivity, such as emotional stability and extraversion.2 This link operates via several

environmental pathways. For instance, some researchers argue that taller men are

perceived to be more attractive, and that attractiveness is associated with competence.

These perceptions encourage individuals to provide taller men with more attention,

praise, and social grooming. As a result, taller workers accumulate more extroversive

2 These characteristics are sometimes referred to as social skills. 4

characteristics, such as optimism and clear, persuasive communication skills (Ross and

Ferris 1981; Harper, 2000; Judge and Cable 2004; Mobius and Rosenblat, 2006).

Another argument holds that taller individuals were raised in more nurturing

environments, as can be provided by better educated parents. Educated parents earn more

than their less educated counterparts, and thus can devote more resources to their child’s

nutrition and medical care. These investments reduce the chances their children undergo

nutritional insults that stunt growth. Cogent parents are also more adept at creating

environments that are conducive to social development (Patterson et al. 1990; Brooks-

Gunn and Duncan, 1997; Bradley and Corwyn, 2002).

Recent neurobiological research suggests that stature is positively associated with

cognitive and non-cognitive potential via a nature-nurture interaction.3 Figure 1.1

illustrates the mechanism. Adult stature reflects an individual’s net nutrition history.

5 During the growth process, the body prioritizes nutrients to combat diseases, maintain

metabolism and as energy for physical activities. Remaining nutrients (surplus nutrition)

are then available for producing growth materials and growth stimulating components:

e.g., shared insulin-like growth factors, such as thyroid and growth hormones (Tanner,

1978; Gunnell et al. 2005; Scheepens et al. 2005). These components stimulate

simultaneous physical and neural growth, and develop the neurological regions that

manage our cognitive and non-cognitive processes (Oppenheimer and Schwartz, 1997;

Thompson and Potter, 2000; Fuster, 2001; Blair, 2004; Bechara, 2005).4

3 This channel is sometimes referred to as the biological pathway due to its associated with the biological literatures. 4 These regions are the insular, anterior cingulated, medial prefrontal and frontal cortices. 5

The social brain hypothesis suggests that neural growth promotes emotionally

stable personality characteristics. The theory maintains that behavior is influenced by our

instincts, what we innately want to do, and our experience, what common knowledge tells

us is appropriate. When individuals choose between competing behaviors their instincts

and experience assign weights to each alternative. Given this feedback, the mind chooses

the action with the highest positive weight. The experience regions promote socially

appropriate behavior, while our instincts occasionally encourage improper actions. For

example, when we are tired, our instincts may encourage us to interact with others in an

anti-social manner. However, society indicates that these actions are inappropriate, which

encourages our experience regions to provide stronger signals to counter our instincts.

Healthy neural development increases the experience regions authority in the

decision making process. Initially, the instinctual regions are more developed.5 This

6 advantage allows our instincts to send stronger signals, and thus command more control

over our actions. As the brain develops, the experience regions receive relatively more

growth, which increases their relative authority over our behavior.

Empirical research supports the neurobiological theory. Liu et al. (2003) examine

the extent to which neural growth contributes to social development. They measure a

child’s neural growth using a standard proxy: the child’s nutritional status. Using data

from the Mauritius Longitudinal Study, they show that malnourishment increases the

frequency with which a child engages in an aggressive, antisocial, dishonest, and socially

inappropriate action from occasionally to constantly. Other studies report similar results

On average, the extent to which these biological channels operate on physical and neural growth is substantial. However, these studies also indicate that responses vary at the individual level. 5 These synapses are also denser, and thus can transmit more signals. 6

using similar methods (Stoch and Smythe 1963; Chase and Martin 1970; Grantham-

McGreggor et al. 1982; Galler et al. 1983; Klein 1987). 6

1.3 Non-Cognitive Ability and Productivity

Most social scientists recognize that emotional stability and extraversion play an

important role in worker productivity (Judge et al. 1999). Emotionally stable individuals

are more adept at controlling their emotions and cultivating positive, rational personality

traits, such as composure and optimism. These traits are thought to promote more

amiable, ambitious and courteous dispositions, and are conducive to coping with stress

and managing demands. As a result, psychologists commonly link emotional stability to

productive personality traits. including authority, courtesy, discipline, ethical conduct,

7 optimism and motivation (Goldberg, 1990).7

Emotionally stable personality characteristics increase worker productivity.

Authoritative workers are more attentive, reliable, and talented at managing tasks and

stress, which promotes their ability to recognize, analyze, and solve problems. Motivated

and disciplined employees work harder and longer, and engage in more activities that

enhance and broaden their skills (Goleman, 1998). Similarly, ethical employees oppose

shirking, and thus work more. They are also more apt to follow rules, which are

conducive to carrying out instructions and engaging in teamwork (Minkler, 2008).

6 We would provide more precise empirical results. However, the estimates in these papers are not easy to interpret 7 Discipline, ethics and motivation are also associated with conscientiousness. For this paper’s purpose, this distinction is not very important as conscientiousness is non-cognitive characteristics. 7

Courteous individuals are more amiable and polite. These personality traits reduce

the tendency to engage in counterproductive activities, such as antagonizing,

intimidating, and threatening co-workers (Noland and Bakke, 1977). These actions can

present substantial costs to an employer. For example, Leymann (1990) estimated that

antagonizing an additional employee is associated with a $30,000 to $100,000 increase in

operating costs. Optimism enhances our capacity to cope with psychological states that

promote apathy and shirking, such as anxiety and depression (DuPont et al. 2006).

Extroverts are relatively adept at communicating clearly and engaging others in

social situations. These social skills play an important role in production. Clear

communication reduces the time required to tell individuals what to do and how to do it,

while more engaging individuals are more persuasive and adept at negotiation (Goleman,

1998; Betz, 2008). These traits are useful for white collar workers. For example, most

8 patients prefer physicians who are empathetic, can communicate clearly, and most

importantly, have good bedside manners (Blue, 2007). These traits are also important for

most non-professional workers. For instance, carpenters must clearly convey instructions

and progress to their peers in order to expedite construction and reduce work related

accidents.

Recent empirical evidence supports the intuition. In western countries, a one

standard deviation increase in emotional stability is associated with a 7 to 11 percent

increase in earnings. Extraversion’s contribution to earnings is unclear. Most studies

indicate that extraversion increases productivity. However, its return to earnings falls to

approximately zero once emotional stability measures are included in the earnings

equation (Gelissen and de Graaf, 2006; Mueller and Plug, 2006; Heineck, 2007). 8

1.4 Empirical Framework

Our empirical task is to estimate the extent to which cognitive and non-cognitive

abilities contribute to the stature-earnings relationship, which can be measured by

applying ordinary least squares (OLS) to the following equation:

= β + ρ + , (1)

where is the natural logarithm of individual i’s hourly earnings, is adult height, is

a vector of exogenous covariates determined before labor market entry (e.g., race,

residence, and parental investments) and is an error term (Case and Paxson, 2008).

excludes measures of worker productivity, such as occupational status and schooling. The

9 rationale is that smarter, more socially adept workers commonly choose to complete

more schooling and pursue more lucrative careers. Therefore, these characteristics are

correlated with cognition and social skills, and thus their inclusion in equation (1) could

understate the extent to which these abilities contribute to the height premium.

The two views suggest that the estimated stature coefficient, OLS, represents the

extent to which stature is correlated with earnings through its positive association with

cognitive or non-cognitive ability. We measure each aptitude’s respective contribution to

the stature premium by separately including cognitive and non-cognitive controls in

equation (1). A substantial reduction in the resulting stature coefficient would suggest

that height is strongly correlated with the respective ability. Assuming the social-skills

hypothesis is correct, including non-cognitive controls should reduce the stature estimate

the most, and vice versa.

Social skill is correlated with cognition (Heckman, 2006). As a result, including

non-cognitive controls in equation (1) may reduce the estimated stature coefficient due to

social skill’s correlation with cognitive ability, rather than due to its independent

association with stature.8 We estimate social skill’s separate contribution to the stature

premium by including non-cognitive, in addition to cognitive controls, in the equation. A

substantial reduction in the resulting stature coefficient, as compared to the estimate

obtained using only cognitive ability controls, would suggest that social skills provide a

substantial, separate contribution to the premium.

1.5 Data

0 1

The analysis requires panel data containing measures of height, cognitive ability,

non-cognitive ability, and adult labor market outcomes, which are available from the

1958 National Childhood Development Study (NCDS).9 The NCDS is a longitudinal

survey that began as a perinatal mortality study of all children born in Britain during the

week of March 3, 1958.10 Several follow up surveys (sweeps) were conducted at ages 7,

8 The average cross-correlation coefficient between the cognitive and non-cognitive measures is 0.10. Hence, the correlation is not large enough that multicollinearity is an issue. 9 The NLSY 79 Child and Young Adult surveys also contain these measures. We do not use these data because the samples are relatively small. Also, many of the children are still not adults, and thus the survey lacks information on their wages at age 30. 10 Environmental factors explain most average stature differences across populations (Malcolm, 1974; Martorelli and Habicht, 1986). Assuming environmental circumstances are significantly different between spring and the other seasons and these differences significantly affect individual characteristics, spring-born individuals may not adequately represent individuals born in other seasons. These conditions may hold 10

11, 16, 23, 33 and 42, collecting a broad range of health, socioeconomic, cognitive, and

non-cognitive measures.

The NCDS provides several measures of emotional stability and extraversion.

Individuals evaluated their motivation, optimism and authority at ages 16, 23 and 33,

respectively. The optimism assessment contains 24 questions, each asking whether the

individual experienced various pessimistic temperaments, such as inadequacy, cynicism,

anxiety, and sorrow. A higher score suggests that the individual was more pessimistic.

The motivation assessment has 8 questions, each asking the individual their opinion

regarding activities associated with ambition (e.g., it is important to work hard and

complete more education). The answers are scaled using a 5 point system ranging from 1

– not true to 5 – very true. A higher score suggests that the individual was more

motivated. Authority is attributed to management skills, such as leadership and the ability

1 1 to give instructions. Individuals rated themselves on these characteristics using a 2 point

scale ranging from 0 – not competent to 2 – very competent.

Ethical individuals comply with rules and authority figures. Teachers and parents

evaluated each individual’s integrity at ages 11 and 16, and at the latter age they also

assessed the adolescent’s honesty, truancy, vandalism record, minor crimes record,

compliance with rules, and aggression towards peers. The questions were measured using

a 2 point system ranging from 0 – the individual never expresses the characteristic to 2 –

because the spring disease environment is relatively gentle, especially compared to autumn and winter— the so-called cold and flu seasons. Also, springs-births are exposed to more sunlight during infancy because the length of day increases during spring. Sunlight is required to produce vitamin D, which is required to use calcium. Another issue is individuals with significantly different characteristics may conceive children during different seasons; however, Card (2001) indicates these differences modestly affect child characteristics and outcomes. 11

the individual constantly displays the temperament. At age 11, teachers assessed the

student’s hostility and arrogance towards peers and authorities. These questions were

measured using a 10 point scale ranging from 1 – they are not hostile / arrogant to 10 –

they are very hostile / arrogant.

Courtesy is associated with manners and an amiable, easy going attitude.

Teachers evaluated their student’s courtesy at age 16 by rating irritability, moodiness,

social flexibility and restlessness. Each question was measured using the same 2 point

scale employed to assess ethical conduct.

Extroverts are persuasive, gregarious and adept at clear communication. At age

33, individuals rated their ability to communicate and persuade individuals. Each

question was measured using the same scale employed to assess individual authority. At

age 16, teachers assessed their student’s inclination to engage in social and solitary

2 1 activities. Sociability was measured using a 5 point scale ranging from 1 – very social

and amiable to 5 – very withdrawn, and introversion was assessed using the same 2 point

scale employed to measure courtesy.11

We measure cognition using the variables employed by Case and Paxson (2008):

the individual’s math and reading test scores reported at age 11. We also include the

11 We separately include the above personality assessments in our analysis to estimate the extent to which emotional stability and extraversion contribute to the stature premium. This approach restricts our capability to report each social skill’s individual return to earnings. First, the method requires me to include over a dozen personality assessments, and thus there is not enough room to accommodate these variables in a single-page table. Second, these variables are relatively collinear, which reduces their respective precision. We try to resolve this problem using a principle components analysis. However, it is unclear which temperaments the resulting components represent. Also, the analysis is unable to reduce the available measures into a smaller number of orthogonal components. For these reasons, we do not report each social skill’s individual contribution to earnings in the main analysis. We did, however, examine the extent to which each cognitive and non-cognitive aptitude individually contributes to earnings. The results indicate that most abilities play an important role in earnings. Please e-mail Andreas Schick at [email protected] to obtain these results. 12

individual’s problem solving assessment reported at age 33, which asked individuals to

evaluate their capacity to solve problems using computers with a 2 point scale ranging

from 0 – not competent to 2 – very competent.

Height is reported at age 33. Nurses measured each individual’s stature, accurate

to one centimeter, using a tape measure or height scale. Consistent with previous studies,

we converted our stature measures from centimeters to inches (Perisco et al. 2004; Case

and Paxson, 2008). Hourly earnings were also reported at age 33. To increase the sample

size, we also included individuals who only reported earnings at age 42.

Table 1.1 presents summary statistics for two datasets: the total sample, full-time

workers with height and earnings measures; and the main sample, the previous sample

restricted to workers with measures of cognition and social skill.12 The main sample

consists almost entirely of individuals of European Caucasian descent. On average, men

3 1 stood 5 feet 10 inches tall as adults and women 5 feet 4 inches. The average logarithm of

gross hourly earnings for men and women—in terms of the value of the pound between

1999-2000—is £10.0 ($16.17) and £7.2 ($11.64), respectively. Approximately 53 percent

of the main sample was born to middle socioeconomic status (skilled labor) fathers; 16

percent to high socioeconomic (managers and professionals) fathers; and lastly, 31

percent to low socioeconomic (low skilled or semi-skilled labor) fathers.

Restricting the sample to workers with cognitive and non-cognitive measures may

introduce selection bias if the availability of these measures is correlated with unobserved

determinants of earnings. The results in table 1.1 indicate that the two samples have

12 Full-time workers are individuals who work 1000 or more hours a year. 13

similar observable characteristics, which suggests that the bias introduced by the

restriction is likely to be small.

1.6 Results

Height, Cognition and Social Skills

In the NCDS, stature is strongly correlated with cognitive and non-cognitive

abilities. To facilitate comparisons across ages and assessments, we convert our stature

measures to z-scores using the 2000 growth charts from the Centers for Disease Control

(2002). Table 1.2 reports OLS estimates of cognitive z-scores at age 11 on height z-

scores at age 7, while table 1.3 presents logistic results of social skill indicators at age 16

4 1 on z-scores at age 11.13 Column I controls for the individual’s race, region of residence,

and medical examination date. Column II includes an extensive range of parental

investment variables, such as the father’s socioeconomic group at age 7, household

income, parents’ academic achievement, parents’ stature, and parents’ involvement in

their child’s education. These extended controls represent environmental investments

contributing to physical growth and social development. If the social-skills hypothesis is

correct, then including these characteristics should substantially reduce the estimated

height coefficient.

13 The non-cognitive measures generally report whether an individual rarely, occasionally, or constantly displays a behavior. We transform these measures into dummy variables to simplify their interpretation (i.e., 0, the individual does not display the behavior and 1, the individual displays the behavior). This transformation does not significantly change the results. 14

The results indicate that taller children are more cognitively and non-cognitively

able than their shorter peers. After including the extended controls, a one point increase

in a boy’s z-score at age 7—approximately 2 inches—is associated with a 0.10 standard

deviation increase in math and reading test scores. Similarly, a one point increase in a

boy’s z-score at age 11—roughly 2.5 inches—is associated with a 2 percentage point

average increase in the measure of non-cognitive ability.14 These effects are roughly as

large as a two standard deviation increase in family income.15 Similar results are reported

for girls.

The environmental controls explain a substantial portion of the relationship

between stature and both cognitive and non-cognitive ability. Including these measures

reduces the stature estimates approximately 30 percent, on average, and in some cases—

such as motivation and optimism—explains the entire association. However, in most

5 1 cases, roughly two-thirds of the correlation between stature and ability remains

unexplained, which suggests that another pathway, such as the neurobiological channel,

may play an important role in determining this relationship.16

Height and Earnings

We estimate the extent to which cognitive and non-cognitive ability separately

contribute to the height premium. Tables 1.4 and 1.5 present regressions on the natural

14 This is equivalent to a 0.10 standard deviation increase in the respective measure of non-cognitive ability. 15 For girls, a one standard deviation increase in family income is associated with a 7 percent of a standard deviation increase in reading score at age 11 and a 1 percent increase in average cognitive ability. 16 We would examine this topic further, but it is well beyond this paper’s scope. 15

logarithm of gross hourly earnings on adult stature for men and women, respectively.

Column I includes experience, ethnicity and region of residency controls. Column II

controls for the father’s socioeconomic status, household income, parents’ education and

the parents involvement in their child’s education. Columns III-V include cognitive

controls, non-cognitive controls and both ability controls, respectively.

Taller men and women earned substantially more than their shorter counterparts.

A one inch increase in adult stature is associated with a 2.2 percentage point increase in

earnings for men and a 1.9 percentage point increase for women. These estimates are

approximately equal to those reported in Case and Paxson (2008) and Persico et al.

(2004).17

The results in column III, tables 1.4 and 1.5, show that cognition does not explain

the entire stature-earnings relationship. Including cognitive scores substantially reduces

6 1 the height estimates, from 0.015 to 0.009 for men and 0.010 to 0.003 for women.18 The

female premium is approximately equal to zero; however, the male premium remains

substantial and statistically significant at the 10 percent level. This result suggests that

another pathway, such as non-cognitive ability, may play an important role in

determining the male stature premium.

The results in column IV show that social skills contribute as much to the stature-

earnings relationship as cognition. Including non-cognitive controls reduces the stature

17 As an interesting note, the disparity in stature between men and women does not explain the gender gap in earnings. We combine the men and women samples and estimate a regression of earnings on a male indicator. The estimate male earnings premium is approximately 40 percent, which is approximately equal in value to the gender gap reported in Case and Paxson (2008). Controlling for cognition, social skills, stature and parental investment did not change the results. 18 To be clear, a stature estimate equal to 0.015 is interpreted as follows: a one inch increase in adult stature is associated with a 1.5 percent increase in adult earnings, ceteribus paribus. 16

estimates a sizeable amount: from 0.015 to 0.008 for men and 0.010 to 0.003 for women.

The results in column V suggest that non-cognitive ability accounts for a substantial,

independent portion of the height premium. Including non-cognitive controls, in addition

to cognitive controls, reduces the stature estimates an additional standard deviation—

from 0.009 to 0.005 for men and 0.003 to 0.000 for women—and renders them

statistically insignificant and approximately equal to zero. Comparing the estimates in

columns I and V, social skills individually reduce the height estimates roughly 20–35

percent. These results support the social-skills hypothesis.

The evidence suggests that neither view is entirely correct. Cognition and social

skills play an equally important role in determining the stature-earnings relationship.

However, neither aptitude individually explains the entire relationship. More importantly,

the results imply that the stature premium is mostly attributed to stature’s correlation with

7 1 cognitive and non-cognitive abilities. Controlling for both abilities reduces the male and

female stature premium approximately 75 and 100 percent, respectively.

Discrimination or Stature?

Some social scientists suggest that the male stature premium represents

discrimination. Taller men, they argue, are not smarter or more socially adept; rather,

most individuals associate stature with superiority, and thus are more inclined to employ

and promote taller men into more prestigious positions (Saul, 1971). This argument

implies that the male stature premium is associated with occupational sorting. Societies

sort taller men into relatively well-paying professions, resulting in these men earning

more on average.

We test this hypothesis by regressing occupational status on stature. The NCDS

reports three occupational groups: white collar workers, managers and professionals;

skilled workers, manual and non-manual; and blue collar workers, semi-skilled and

unskilled workers. The NCDS assigns workers into a group, in part, using the

occupation’s average earnings. In general, white (blue) collar workers receive the highest

(lowest) earnings.

Table 1.6 presents multinomial logistic regressions of occupational status on

stature (the base category is skilled workers). Column I includes race and region controls.

Column II controls for the father’s socioeconomic status at age 7, parents’ education

levels, household income, and the parents involvement in their child’s education. Column

8 1 III includes cognitive and non-cognitive controls.

The evidence indicates that taller men are relatively more likely to select into

white collar occupations than their shorter counterparts, and that this result is largely due

to stature’s association with cognitive and non-cognitive abilities. A one inch increase in

adult stature is associated with a 2.7 percentage point increase in the chances of attaining

a white collar occupation over a skilled vocation. The effect is as large as a one standard

deviation increase in household income. Including parental measures marginally reduces

the stature estimates. However, including cognitive and non-cognitive scores reduces the

estimated height coefficient to approximately zero and renders it statistically

insignificant. A one inch increase in stature is now associated with a 0.8 percentage point

increase in acquiring a white collar occupation relative to a skilled job.19

The discrimination view also suggests that taller men are perceived to be more

physically attractive, and that some societies associate attractiveness with superiority.

This view may seem to be a form of the social-skills hypothesis. The two views,

however, are distinct. The social-skills hypothesis suggests that attractive men

accumulate more non-cognitive abilities, and thus attribute the attractiveness premium to

social skills. In contrast, the discrimination view argues that attractiveness is uncorrelated

with non-cognitive ability; rather, employers tend to overestimate an attractive worker’s

productivity, and thus pay them more than they are worth (Mobius and Rosenblat, 2006).

We examine the extent to which attractiveness contributes to the stature premium

by including beauty controls in the main analysis. A substantial reduction in the resulting

9 1 stature coefficient would suggest that attractiveness plays an important role in

determining the premium. Beauty is measured at ages 11 and 33, the latter being a self

report on whether or not they are overweight. At age 11, teachers rated students as

attractive, average or unattractive.20

Table 1.7 presents evidence that more attractive men receive an earnings

premium. The results in column III indicate that attractive 11 year olds earn

approximately 6.5 percent more as adults than their unattractive peers. However,

19 We conduct the same analysis using our female sample. The results indicate that stature is uncorrelated with occupational status among women. This result is consistent with the discrimination argument, which attributes the pathway solely to men. 20 We acknowledge that these measures are not ideal. However, they are the best measures given the available data. 19

including attractiveness controls does not change the stature estimate, which suggests that

beauty’s true return to the height premium is modest.

The discrimination views argue that only distinctively tall men (those who are one

to two standard deviations taller than average) are perceived as more productive, giving

these individuals a sizeable increase in earnings relative to their shorter and substantially

taller counterparts.21 To test this view, we regress earnings on several stature categories:

very short, short, average, tall and very tall.22 Figure 1.2 plots the returns to stature

associated with each stature category (with very short individuals as the omitted

category). The results indicate that distinctively taller individuals do not receive a sizable,

discontinuous increase in earnings; rather, the returns to stature increase at a decreasing

rate, and approach zero at 72 inches (approximately one standard deviation above

average stature).23 Gains in stature only provide a substantial contribution to earnings for

0 2 shorter men. This result is consistent with the neurobiological pathway, which claims that

the correlation between physical and neural development decreases as individuals

undergo more physical growth.

1.7 Conclusion

Researchers have put forward two explanations for the height premium. The more

established view claims that stature is positively correlated with non-cognitive abilities

21 Distinctively taller men are one to two standard deviations taller than average. 22 Very short men are under 64 inches tall, short men are between 64 to 66 inches tall, average men are between 66 to 72 inches tall, tall men are 72 to 75 inches tall, and very tall men are above 75 inches tall. 23 Hubler (2009) and Case and Paxson (2008) report relatively similar results using the German Socio- Economic Panel and NCDS data, respectively. 20

that are rewarded in the labor market (Stogdill, 1948; Baker and Redding, 1962; Adams,

1980; Judge and Cable, 2004; Persico et al. 2004). Another view recently challenged this

mechanism, arguing that cognitive development accompanying vigorous physical growth

accounts for the relationship (Case and Paxson, 2008; Heineck, 2009).

This paper tests the competing hypotheses. Using data from the National

Childhood Development Study (NCDS), we show that taller children are more

cognitively able and socially adept than their shorter cohorts. A one standard deviation

increase in stature at age 7 (approximately 2 inches) is associated with a 0.10 standard

deviation increase in math and reading test scores reported at age 11. Similarly, a one

standard deviation increase in height at age 11 (approximately 2.5 inches) is associated

with a 2 percent average increase in non-cognitive ability. These effects are as large as

growing up in a middle class versus a lower class family.

1 2 We also show that each aptitude accounts for a substantial and approximately

equal portion of the stature premium. Separately including either cognitive or non-

cognitive controls in the standard earnings equation reduces the estimated stature

coefficient roughly the same amount, from 0.015 to 0.009 for men and 0.010 to 0.003 for

women. The non-cognitive controls explain a substantial, independent portion of the

stature-earnings relationship. Including non-cognitive, in addition to cognitive controls,

reduces the stature estimates to approximately zero, from 0.009 to 0.005 for men and

0.003 to 0.000 for women. These results indicate that neither pathway individually

explains the entire relationship; rather, both abilities are involved.

Our findings suggest that researchers should include adult stature in the Mincer

earnings equation, which are used to estimate the returns to schooling, when cognitive 21

and non-cognitive measures are unavailable. Schooling is positively associated with

cognitive and non-cognitive ability, and thus analyses that omit these measures will

produce upward biased estimates of the returns to schooling. Our results show that adult

height is strongly correlated with cognitive and non-cognitive ability. Hence, researchers

can use stature to mediate this bias when ability scores are unavailable.

Our paper suggests several areas for further research, one of which extends study

of the stature premium to poorer populations, as may be found in developing nations. The

neurobiological pathway suggests that the returns to productivity that operate through

gains in stature are nonlinear, i.e. they are relatively greater the larger is the baseline of

physiological deprivation. Because poor populations are shorter, their gain in physical

growth generates relatively more neural growth, and thus a greater stature premium. In

this setting it would be interesting to assess the relative degree to which cognitive and

2 2 non-cognitive abilities contribute to the stature-earnings relationship.

Policy makers will be interested in the extent to which environmental and

biological forces (nurture and nature) separately contribute to the stature premium. One

would think that failure to grow for environmental reasons would have a greater impact

on neural growth and cognitive and non-cognitive ability, and therefore economic

performance, than would physical growth constraints (e.g. genes influencing height)

imposed by biology. Pediatric research reports that a nutritionally diverse diet, especially

during pivotal growth stages, helps create more growth stimulating components, and thus

both physical growth and cognitive and non-cognitive development (Williams et al. 1978;

Richards et al. 2002; Scheepens et al. 2005; Liu and Raine, 2006; Kiddie et al. 2010). It

would be useful to test this mechanism, and study which nutrients and growth stages play

the more important roles in determining the stature-ability relationship.

3 2

My Sample Total Sample Number of Observations 2,577 6,838

Ethnicity European Caucasian 0.99 0.99

Adult Height (Inches) Men 69.8 69.7 Women 64.3 64.2

Adult Gross Hourly Earnings (£) Men 10.0 9.6 Women 7.2 7.0

Father's socioeconomic group White Collar 0.17 0.16 Skilled 0.53 0.53 Table 1.1: Descriptive Statistics in Post-Industrial Britain

Note - The Total Sample is restricted to full-time workers that report wage and height data at age 33. My sample is the Total Sample restricted to individuals with cognition, non-cognition and socioeconomic group measures.

4 2

Boys Girls

Age 7 Height for Age z-Score Limited Extended Limited Extended Dependent variables Controls Controls Controls Controls Reading at age 11 1.04*** 0.70*** 1.12*** 0.73*** (0.12) (0.12) (0.11) (0.11) Math at age 11 1.56*** 0.93*** 1.89*** 1.37*** (0.20) (0.20) (0.19) (0.19) Table 1.2: Cognitive Test Scores and Height in Childhood

5 2

Boys Girls Age 11 Height for Age z-Score Limited Extended Limited Extended Dependent variables Controls Controls Controls Controls Indicator Behavior Scores at age 16 Disobedient -0.024*** -0.022** -0.018*** -0.01 (0.008) (0.009) (0.007) (0.008) Solitary -.0.029*** -.0.031*** -0.025*** -0.030*** (0.009) (0.011) (0.009) (0.010) Confident 0.034*** 0.045*** 0.020** 0.023** (0.008) (0.009) (0.008) (0.009) Dishonest -0.028*** -0.021*** -0.021*** -0.014** (0.007) (0.008) (0.006) (0.006) Restless -0.038*** -0.023** -0.021*** -0.009** (0.008) (0.009) (0.006) (0.006) Thief -0.018*** -0.010*** -0.004** -0.003* (0.004) (0.004) (0.002) (0.002) Rude -0.01 0.00 -0.022*** -0.013* (0.007) (0.008) (0.008) (0.008) Indexed Behavioral Scores Motivation at age 16 0.313** 0.008 0.591*** 0.416*** (0.125) (0.134) (0.116) (0.124) Pessimism at age 23 -0.116** -0.008 -0.324*** -0.263*** (0.055) (0.135) (0.062) (0.067) Table 1.3: Non-Cognitive Test Scores and Height in Childhood

6 2

Note.— ***: statistically significant at the 1 percent level; **: statistically significant at the 5 percent level; *: statistically significant at the 10 percent level. Sample sizes are 2,908 for men and 2,833 for women. Limited controls inclue the individual's race, region and medical exam date. Extended controls include household income, father's socioeconomic group, mother's and father's height, education and involvement in child's education. Each indicator behavioral score is equal to one when the individual expresses the characteristics either occasionally or constantly, and zero when they do not express the trait. These characteristics are regressed using a multinomial logit model, and the height coefficients represent marginal frequencies. The indexed behavioral scores are regressed using OLS.

Men Dependent Variable: Log Gross Hourly Earnings (1) (2) (3) (4) (5) Height at age 33 0.022*** 0.016*** 0.009* 0.008 0.005 (0.005) (0.005) (0.005) (0.005) (0.005) Test for Overall Significance (F-Test) Cognitive Scores F-test (p-value) 20.04 (0.00) Non-Cognitive Scores F-test (p-value) 5.63 (0.00) Both Scores F-Test (p-value) 8.26 (0.00) N 1,383 1,383 1,383 1,383 1,383

27 Adjusted R2 0.03 0.08 0.19 0.14 0.22

Table 1.4: Log Average Hourly Earnings, Cognitive, Non-Cognitive Ability and the Returns to Height (Men)

Note.—***: statistically significant at the 1 percent level; **: statistically significant at the 5 percent level; *: statistically significant at the 10 percent level Column (1) includes experience, region of residence in 1911, and ethnicity measures Column (2) includes father's socioeconomic group and parental involvement in child's education measures Column (3) includes cognitive math and reading test scores, reported at age 11, and problem solving skills reported at age 33 Column (4) includes emotional stability and extraversion assessment scores reported at ages 11, 16, 23 and 33. Column (5) includes cognition and non-cognition controls

Women Dependent Variable: Log Gross Hourly Earnings (1) (2) (3) (4) (5) Height at age 33 0.019*** 0.010* 0.003 0.003 0.000 (0.006) (0.006) (0.006) (0.006) (0.006) Test for Overall Significance (F-Test) Cognitive Scores F-test (p-value) 16.12 (0.00) Non-Cognitive Scores F-test (p-value) 5.33 (0.00) Both Scores F-Test (p-value) 6.9 (0.00) N 1,167 1,167 1,167 1,167 1,167

28 Adjusted R2 0.01 0.09 0.18 0.16 0.21

Table 1.5: Log Average Hourly Earnings, Cognitive, Non-Cognitive Ability and the Returns to Height (Women)

Men

Dependent Variable (1) (2) (3)

Height at age 33 (Marginal Effects) White Collar 0.027*** 0.024*** 0.008 (0.006) (0.006) (0.007) Blue Collar -0.006 -0.005 0.001 (0.004) (0.004) (0.004) Table 1.6: Male Occupational Placement and Stature

Note.— ***: statistically significant at the 1 percent level. The sample size is 963 men, and skilled workers are the base category. Column I include race and region controls. Column II includes parental education measures, family income, parental involvement in child's education and father's socioeconomic group at age 7. Column III includes cognitive and non-cognitive controls.

9 2

Men Dependent Variable: Log Gross Hourly Earnings (1) (2) (3) (4) (5) (6) Height at age 33 0.023*** 0.018*** 0.018*** 0.011** 0.009 0.005 (0.006) (0.006) (0.006) (0.005) (0.006) (0.005) Attractive at age 11 0.065** (0.030) Over weight at age 33 (0.011) (0.029) Controls: Family background X X X X X Cognitive test scores X X Non-cognitive test scores X X

30 30 N 1,260 1,260 1,260 1,260 1,260 1,260

Adjusted R2 0.03 0.07 0.08 0.19 0.15 0.22

0 3 Table 1.7: Log Average Hourly Earnings, Cognition, Non-Cognitive Ability, Beauty and the Returns to Height (Men)

Note.—***: statistically significant at the 1 percent level; **: statistically significant at the 5 percent level; *: statistically significant at the 10 percent level Column (1) includes experience, region of residence in 1911, and ethnicity measures Column (2) includes father's socioeconomic group at age 7, parents' academic achievement, household income, and parental involvement in child's education Column (3) include attractiveness controls Column (4) includes math and reading test scores, reported at age 11 Column (5) includes emotional stability and extraversion scores, reported at ages 11, 16, 23 and 33 Column (6) includes cognitive and non-cognitive assessment scores

Nutrients

Devoted to

Biological Maintenance, Combating Disease and

Physical Activity

Remaining Nutrients

Devoted to

Insulin- like Growth

Factors

e.g. Thyroid Hormones

Growth Hormones

Physical Growth Neural Growth

1 3

Adult Height Cognitive &

Non-Cognitive Ability

Figure 1.1: The Biological Link between Stature and Ability

0.25

0.2

0.15

0.1

0.05 Log Earnings Log

0 64 66 68 70 72 74 75+ -0.05 Height at age 33

Figure 1.2: Log Earnings and Stature for Men in Post-Industrial Britain

Notes: Figure 1.2 reports the returns to stature associated with short, average statured,

above average statured and very tall individuals. The results include, age, race, and

2 3 matching controls.

Chapter 2

A Historical Exploration of the Height Premium: A Case Study of Canada in 1911

2.1 Introduction

Why are some countries richer than others? Modern growth theories assign a

pivotal role to human capital accumulation, particularly in the nineteenth and early

twentieth centuries (Mankiew et al., 1992; Barro, 2001; Galor, 2011). But recent

empirical evidence suggests otherwise—that human capital provided a modest

contribution to productivity (Mitch, 1999). Some researchers attribute the modest

3 3 empirical results to poor data. Standard human capital measures, they argue, are generally

unavailable before the twentieth century. As a result, most researchers use more data-

abundant alternatives, such as literacy rates and schooling enrollment ratios; however,

these measures poorly proxy individual ability, and thus produce estimates that understate

the gains in productivity which operate via increases in human capital (A’Hearn et al.,

2009). Given these issues, it is important that researchers identify new, useful proxies for

historical human capital in order to measure its contribution to economic performance.

I propose that adult height (stature) is a useful proxy for historical human capital.

Historical height data are relatively abundant starting in the mid-eighteenth century

(Steckel, 1995). During this period, photography was expensive or unavailable, and thus

most institutions identified individuals using permanent physical characteristics, such as

stature. As a result, numerous European and Neo-European institutions (e.g. militaries

and governments) collected nationally and regionally representative height estimates.

Recent studies indicate that stature is strongly associated with cognitive and non-

cognitive potential (Perisco et al., 2004; Case and Paxson, 2008). Using data from the

National Childhood Development Study (NCDS), Schick and Steckel (2010) show that

taller children score substantially higher on cognitive and non-cognitive exams. They

report that a one standard deviation increase in childhood height (approximately two

inches) is associated with a 10 percent of a standard deviation increase in cognitive and

4 3 non-cognitive test scores. This effect is approximately as large as a two standard

deviation increase in household income. Their results also indicate that stature is so

strongly associated with these abilities that the association results in taller workers

receiving a moderate wage premium (i.e., they are more productive). Excluding cognitive

and non-cognitive test scores, the authors report that a one inch increase in adult height is

associated with a 2.2 percent increase in earnings for men. Including the test scores into

their analysis reduces the returns to stature approximately 80 percent and renders it

statistically insignificant. A one inch increase in stature is now associated with a modest

0.5 percent increase in wages.

The contemporary evidence suggests that stature is a suitable proxy for human

capital. However, there is no guarantee that the extent to which stature is correlated with

human capital in the present-day accurately reflects the correlation’s magnitude in the

past. This paper provides a unique contribution to the literature: it empirically examines

the extent to which cognitive and non-cognitive abilities contribute to the relationship

between stature and productivity. To examine this topic, I construct a historical panel

data set that matches Canadian men across their World War I registration papers, which

record individual stature, and their 1891 and 1911 census manuscripts. I show that taller

workers are substantially more productive than their shorter counterparts, and that this

result is due, in large part, to stature’s strong correlation with cognitive and non-cognitive

abilities.

5 3 2.2 Height, Cognitive Ability, and Non-Cognitive Ability

Recent studies suggest that stature is strongly associated with cognitive and non-

cognitive development. The association is thought to work through various

environmental and neurobiological channels. Figure 1.1 illustrates the neurological

channel. Formally, adult stature represents an individual’s net nutrition history. As an

individual grows, their body initially uses nutrients to survive (carry out biological

maintenance, physical exertion, and combat diseases). Individuals that consume more

nutrients than is necessary to survive, convert surplus nutrients into growth materials and

growth stimulating chemical messengers (Tanner, 1978; Hoppe et al. 2004; Kerver et al.

2010). These messengers stimulate simultaneous physical and neural growth, and are

particularly conducive towards developing the neurological regions managing our

cognitive and non-cognitive processes. Hence, this pathway suggests that taller

individuals (those that are more likely to complete their physical and neurological growth

process) are more cognitively and non-cognitively developed on average (Oppenheimer

and Schwartz, 1997; Thompson and Potter, 2000; Fuster, 2001; Blair, 2004; Bechara,

2005).

Neurological development promotes our cognitive and non-cognitive capabilities.

The brain grows via producing grey and white matter, and reinforcing synapses. Grey

matter increases the rate with which we process and interpret stimuli, while white matter

and reinforced synapses improve our ability to organize and transmit neurological signals

(thus expediting our ability to coordinate thoughts). These developments are conducive to

6 3 cognitive and non-cognitive functions (Paus et al., 1999). For example, communicating

an abstract thought requires an individual to organize their thoughts as they talk, to

process the audience’s body language, and adapt their message in response to audience

reactions.

Recent psychological studies indicate that neurological growth may play a

particularly important role in developing two social skills: emotional stability and

extroversion.24 According to the social brain hypothesis, our behavior is influenced by

24 Emotional stability and extroversion are discussed in detail in the next section. Briefly, emotional stability enables us to control our emotions, and is conducive to producing positive, rationale thoughts. As a result, psychologists commonly argue that emotional stability enhances our ability to develop several social skills that generally improve individual productivity, such as authority, courtesy, discipline, ethical conduct, motivation and optimism. Extroversion is our inclination towards assertive, gregarious social

our instincts, what we innately want to do, and our experience, what society tells us is

appropriate. When individuals choose between competing behaviors their instincts and

experience assign weights to each alternative. Given this feedback, the mind chooses the

action with the highest net return. The experience regions commonly promote socially

appropriate (emotionally stable and extroversive) behavior, while our instincts

occasionally encourage the opposite. For example, when we are tired, our instincts may

encourage us to interact in an anti-social manner. However, society teaches us that these

actions are inappropriate, and thus our experiences may provide stronger signals to

overrule our instincts advice to act impolitely.

Healthy neural growth increases the experience regions’ authority in the decision

making process. Initially, the instinctual regions are more developed.25 This advantage

allows our instincts to send stronger weighted signals, and thus command more control

7 3 over our actions. As the brain develops, the experience regions receive relatively more

growth, which increases their relative authority over our behavior.

Empirical research supports the neurobiological pathway. Liu et al. (2003)

examines the extent to which neural growth contributes to cognitive and non-cognitive

development. They measure a child’s neural growth using a standard proxy: the child’s

nutritional status. Using data from the Mauritius Longitudinal Study, they show that

malnourished children score substantially lower on their cognitive and non-cognitive

assessments. Malnourishment at age 11 is associated with a 6 percent decrease in average

interaction. This characteristic is conductive to clear communication, and courteous, amicable social interaction. 25 These synapses are also denser, and thus can transmit more signals.

cognitive test scores, and a 12 percent increase in exhibiting unproductive social

characteristics, such as aggression, conduct disorder, dishonesty, delinquency, and

hyperactivity. Other studies report similar results using similar methods (Chase and

Martin, 1970; Galler et al., 1983; Klein, 1987).

The environmental pathways also suggest that stature is correlated with cognitive

and non-cognitive development. As mentioned earlier, a standard pathway is that taller

individuals are raised in more nurturing environments (e.g., growing up with more

educated parents). Educated individuals earn relatively more than their less educated

counterparts, and thus can devote more resources to their child’s nutrition and health

care. As a result, their children are less prone to nutritional insults that stunt growth, and

thus can grow taller on average. Cogent parents are also more adept at interacting with

their children in a cognitively and non-cognitively stimulating manner. For instance, they

8 3 commonly communicate with their children using correct diction and a more extensive

vocabulary. These activities play a pivotal role in promoting their child’s capability to

improve their diction and grow their vocabulary (Gibson and Peterson, 1991).

Some environmental pathways also argue that height is particularly associated

with emotional stability and extroversion. A standard channel is that moderately taller

men—males that are one to two standard deviations taller than average—are perceived to

be more attractive, and that attractiveness is associated with competence. These

perceptions encourage individuals to provide their taller counterparts with more attention,

praise and investments in their social mentoring. As a result, taller workers accumulate

more emotionally stable and extroversive characteristics, such as optimism, social grace,

and clear, persuasive communication skills (Ross and Ferris, 1981; Mobius and

Rosenblat, 2006).

Some social scientists attribute the male stature premium (i.e., that taller workers

earn more than their shorter counterparts) to height discrimination. As mentioned earlier,

moderately tall men are considered to be more attractive, and attractiveness is associated

with competence. This perception may cause employers to overestimate an attractive

(tall) employee’s productive capacity, resulting in the employer paying their taller

employees more than they are worth.

There is contemporary evidence that stature is weakly correlated with

attractiveness. Harper (2000) measures an individual’s attractiveness using peer

evaluations reported in the National Childhood Development Study. The variables

indicate whether the individual was attractive, unattractive, or unhealthy in appearance at

9 3 ages 7 and 11. He finds that attractive men earn approximately 15 percent more than their

unattractive peers. However, the results also indicate that attractive and unattractive men

achieve approximately similar average adult heights, 1.63 and 1.62 meters, respectively.

Recent research indicates that parents invested in their children’s development

during some historical periods, such as early twentieth century Canada. Cranfield et al.

(2007) use stature data from Canadian World War I soldiers to measure the extent to

which parents invested in their children’s physical development. Using a father’s

socioeconomic group to proxy for parental investment, they show that children born in

higher socioeconomic households are taller on average. Individuals raised by white collar

fathers are approximately one inch taller than individuals raised by general laborers.

Gaffield and Bouchard (1989) provide evidence that wealthier parents also invested more

in their children’s education. They show that the sons of merchants and managers were

twice as likely to attend school then the sons of general laborers.

2.3 Cognitive Ability, Non-Cognitive Ability and Productivity

It is common knowledge that cognitive abilities provide a substantial contribution

to worker productivity. Cognition enhances our capacity to process and interpret

information, which in turn, expedites our ability to accumulate and use knowledge to

solve problems. These characteristics also enhance our ability to conceptualize the net

returns attributed to ideas, which suggests that cogent individuals are more adept at

constructing, adopting, and investing in productive capital, methods and technologies.

0 4 For example, intelligent farmers may be better able to visualize, construct and implement

improved irrigation systems (Bils and Klenow, 2000).

Becker (1993) argues that work-related experience is pivotal to worker

productivity. Experienced workers are less prone to error, can carry out a wider range of

tasks, and require less time to complete their tasks. Therefore, cogent workers have the

potential to become more productive because their cognitive capacities improve their

ability to accumulate and master work-related experiences.

Empirical studies suggest that workers received a substantial intelligence

premium in historical periods. To examine this topic, researchers commonly measure the

intelligence-earnings relationship in developing countries, since these societies share

similar conditions to historical societies (i.e. poorer access to nutritional, educational,

capital, and health resources, and underdeveloped communication, transportation and

technology resources). Jolliffe (1998) examines the extent to which mathematics test

scores contribute to earnings among manual laborers in Ghana. Using data from the

Ghana Living Standards Survey, he finds that the association between math scores and

earnings is substantial and statistically significant at the 5 percent level. A one standard

deviation increase—7.58 points—in mathematics test score is associated with a 14.4

percent increase in earnings.

An emerging literature suggests that social skills, such as emotional stability and

extroversion, also play an important role in worker productivity (Judge et al., 1999).

Emotional stability promotes positive, rational emotions, and enhances our ability to cope

with stress and multiple social stimuli. Psychologists argue that these characteristics are

1 4 conducive to developing ambitious, composed, and optimistic personality traits. As a

result, psychologists commonly link emotional stability to the social skills associated

with these personality traits, such as authority, courtesy, discipline, optimism and

motivation (Goldberg, 1990).

Emotionally stable characteristics intuitively increase worker productivity. For

instance, authoritative workers are more attentive, reliable and adept at managing

resources and stress; characteristics which are argued to improve a worker’s aptitude at

recognizing, analyzing, and solving problems. Motivated and disciplined workers

commonly work harder and longer, and are also more inclined to participate in activities

that improve and expand their skills (Goleman, 1998). Courtesy reduces our inclination to

engage in various unproductive activities, such as antagonizing and arguing with co-

workers (Noland and Bakke, 1977). These actions reduce the time employees work, and

also present substantial costs to employers. For example, Leymann (1990) estimated that

antagonizing an additional employee is associated with a $30,000 to $100,000 average

increase in an employer’s operating costs. Finally, optimism helps individuals cope with

emotional states that reduce motivation and encourage shirking, such as anxiety, stress

and depression (DuPont et al., 2006).

Extroversion promotes an individual’s aptitude to interact in social settings, and

communicate clearly and courteously. These social skills promote worker productivity

because most occupations involve social activities, such as working in groups. As a

result, extroversive employees commonly excel in these situations because clear,

courteous communication expedites these social activities. For example, cargo movers

2 4 work in groups to organize and maneuver goods. Clear communication ensures that

everyone knows what to do and how to coordinate their actions to quickly transport their

merchandise. Miscommunication is costly; it can prolong transportation times, or worse,

result in a work-related accident.

Historical records indicate that motivation, discipline, and communication were

particularly conducive to worker productivity during industrial periods. Pollard (1963)

presents evidence that factory managers commonly complained about unmotivated and

undisciplined workers, which were prone to skip work, shirk and quit in a couple days.

Labor turnover presented such a substantial cost that most employers argued that ―the

stable worker was worth the most to the manufacturer,‖ (p. 255). Industrial work also

incorporated relatively complex capital and technology to the production process. Some

machines required several workers to coordinate their activities in order to operate

properly, which suggests that cooperative communication and group work enhanced the

production process.

Recent empirical evidence suggests that non-cognitive abilities provide a

substantial contribution to labor market outcomes. Mueller and Plug (2006) examine the

extent to which emotional stability contributes to earnings. Using data from the

Wisconsin Longitudinal Study, they find that a one point increase in emotional stability is

associated with a 3 percent increase in average hourly earnings for men and a 2 percent

increase for women.

2.4 Data

3 4

My analysis requires historical panel data containing measures of adult height,

earnings, cognitive ability, and non-cognitive ability. To my knowledge, no single data

set provides these measures, so the best alternative is to link individuals across several

cross-sectional studies. Several suitable data sets are the Canadian World War I

Expeditionary forces (CEF) and the 1891 and 1911 Canadian censuses.

The CEF provides every serviceman’s registration and medical records. The

medical records state each serviceman’s stature, accurate to one quarter of an inch, as

recorded using a tape measure. The registration records also report other useful individual

characteristics, such as their occupation, residence, religion, and marital status. Most

servicemen are young, ages 20 to 40, so constructing a matched sample using these data

produces a sample that, at best, only represents this particular age–demographic.

The 1891 and 1911 Canadian censuses are nationally representative surveys

which sampled the entire population in their respective years. The 1911 census reports an

individual’s characteristics in adulthood, such as their annual earnings, weeks worked per

year, hours worked per week, occupation, race, religion, and marital status. The 1891

census reports an individual’s characteristics in childhood, such as their household’s size,

father’s occupation, mother’s and father’s race, creed and ability to read and write.26

I match individuals across these surveys to construct a linked cross-sectional

sample of 1,412 Canadian servicemen, ages 25 to 35 in 1910.27 The data appendix

discusses the matching procedure in detail. Briefly, sons were linked using their given

name, surname, province of birth, ethnicity, and year of birth (give or take two years).

4 4 Furthermore, most individuals were also matched using their parents’ names, spouse’s

name, siblings’ names, month of birth and their middle name.

I measure an individual’s ability using their occupation’s cognitive and non-

cognitive requirements. The rationale is that workers generally select into occupations

that tightly match their capabilities, suggesting that their occupation’s requirements

roughly reflect their true ability (Ingram and Neumann, 2008). In the appendix, I use

contemporary data to examine the extent to which workers select into occupations that

tightly match their capabilities. The results indicate that most matches are tight (e.g.,

26 Please e-mail me to acquire the complete list of variables reported in each census. 27 I restrict the sample to this particular age group because older individuals are less likely to be observed with their parents in 1891, and younger individuals are unlikely to select into occupations that tightly match their skill set in 1911. The 1911 census data was collected in 1910.

more non-cognitively able individuals are substantially more likely to select into jobs

requiring more social skills).

I estimate an occupation’s requirements using the second edition of the Dictionary

of Occupational Titles (DOT).28 This resource reports 48 occupational characteristics, per

vocation, for over 10,000 jobs. The characteristics are commonly grouped into five

groups: intellectual aptitudes, physical aptitudes, temperaments, interests, and working

conditions.29 Briefly, the intellectual assessments measure general intelligence, verbal

aptitude and numerical capability. These aptitudes capture a worker’s ability to process

and interpret data, use and understand written and oral communication, and to

competently conduct mathematical operations quickly and accurately, respectively.30

Each assessment is scaled using a 5 point system ranging from 1 – the occupational

requirements are modest to 5 – the occupation requirements are substantial.

5 4 Temperaments and interests measure various social competences, such as a

worker’s capacity to manage employees, communicate clearly, interact in social settings,

work in groups, cope with stress, and adapt to new social situations and environments.

Each assessment is scaled using an indicator function, equal to one when the

temperament/interest is present.

28 This is the earliest edition that reports occupational characteristics. These measures were constructed using assessments conducted in the 1920’s and 1930’s. Therefore, these measures should adequately represent the average qualifications of early twentieth occupations. 29 Please e-mail me to acquire a complete list and summary of every characteristic reported in the DOT. 30 The DOT also provides the individuals specific vocational preparation (SVP) and general education development (GED). The former is defined as the amount of time required for the typical worker to acquire and master the techniques and knowledge required to adequately perform their vocational duties and tasks. The latter aptitude is more obscure. It represents the aspects of education and learning that promote worker performance. These characteristics are thought to measure cognition and non-cognition. More cogent individuals are more adept at accumulating experience; likewise, so are more motivated, industrious individuals.

The other characteristics capture physical abilities and working conditions. The

physical capacity assessment measures strength and endurance. It is scaled using a 5

point system from 1 – the occupation is sedentary to 5 – the occupation requires

substantial strength and endurance. The physical aptitude assessments capture fine-motor

requirements, such as finger dexterity, motor coordination, spatial perception, manual-

dexterity and eye-hand-foot coordination. These aptitudes are scaled using the same point

system as the cognitive aptitudes. Consistent with the literature, I omit the working

condition variables because these measures weakly capture individual skill (Usui and

Okumura, 2010).

It is unlikely that each vocational characteristic represents a unique skill (Cain and

Tremain, 1981). For instance, ―numerical aptitude‖ and ―mathematical development‖

both capture a worker’s quantitative aptitude. Similarly, ―motor coordination‖ and ―eye-

6 4 hand-foot coordination‖ both assess coordination. Consistent with previous studies, I use

a principal components analysis (PCA) to combine a large set of similar characteristics

into a smaller set of orthogonal skills (Ingram and Neumann, 2008; Usui and Okumura,

2010). The PCA is an orthogonal transformation which converts a large set of correlated

variables into a smaller set of uncorrelated variables, called principal components. Each

component is a linear combination of the original variables, where each variable is

assigned a weight between -1 to 1. Components receiving greater positive weights with

respective to certain variables are interpreted to capture the common characteristic shared

among those variables. For example, suppose a researcher conducts a PCA on the DOT

characteristics, resulting in a component that is strongly associated with numerical

aptitude and mathematical development. As a result, this component is interpreted to

capture the characteristic these variables share (for example, general quantitative

aptitude).31

Appendix table A.1 reports the weights associated with the PCA. The procedure

identified 8 components which comprise over 75 percent of the total shared variance. The

first component is strongly associated with general intelligence (.67), numerical ability

(.84) and verbal ability (.59). Therefore, it probably captures ―general cognitive ability‖.

The second factor is weighted with characteristics promoting communication (.66),

motivation (.59) and verbal aptitude (.63), and thus reflects ―general non-cognitive

ability‖. The third factor is correlated with spatial perception (.47) and working with

machines (.47), and thus represents ―general motor-skill‖. The fourth factor is associated

with abstract thought, judgment and critical thinking, so it is associated with ―abstract

7 4 thought‖. The fifth and six components are correlated with persuasion, coping with stress,

and instructing/managing individuals, and thus capture ―communication‖ and ―authority‖.

Finally, the last two skills are almost entirely weighted with social isolation and physical

capacity, respectively. Therefore, they capture ―social isolation‖ and ―physical capacity‖.

I group individuals into four socioeconomic classes using the occupational

classification system provided in the 1911 census manuscripts. The socioeconomic

categories are white collar workers, skilled workers, farmers and blue collar workers.

White collar workers include managers, government officials and professionals. Skilled

laborers comprise kindred workers, clerks, operators and other craftsmen. Agricultural

31 Please consult Ingram and Neumann, ―The Returns to Skill.‖ for an in depth discussion regarding the principal components analysis.

workers are farmers, landowners, ranchers, gardeners and certain groundskeepers. Lastly,

blue collar workers represent semi-skilled and unskilled laborers.

I estimate parental investment using a mother’s and father’s ethnicity, mother’s

and father’s ability to read and write and father’s socioeconomic group. Sociological

studies indicate that ethnicity and socioeconomic status are strongly correlated with

parental investment during late 19th century Canada. For example, Katz and Davey use

data from Hamilton, Ontario to examine the extent to which these characteristics

contribute to a child’s secondary school attendance (Katz and Davey, 1978). They show

that Scottish children were 20 and 30 percent more likely to attend secondary school then

English and Irish boys, respectively. Scottish children also attended secondary school

more regularly, and completed more years of schooling.32 Similarly, white collar sons

were 25 and 45 percent more likely to attend secondary school than skilled and unskilled

8 4 sons, respectively.

I measure environmental investments using an individual’s province of residence

in 1890, and whether the individual resided in an urban or rural township in 1890. In the

late 19th century, most provinces had approximately similar maximum age to entry and

minimum age to exit compulsory schooling laws. However, each province enforced these

laws differently. For instance, the average grade attainment, in 1910, was 9 years in

British Columbia, 7 years in Quebec, 6 years in Newfoundland and 8 years in the

remaining provinces (Oreopoulos, 2006). Several studies also indicate that schooling

opportunities were more available in rural townships. Denton and George (1974) use data

32 They do not specify how many more years of schooling Scottish children complete.

from Wentworth, Ontario to examine the extent to which schooling attendance varied

across rural and urban communities. Controlling for the father’s socioeconomic group,

mother’s and father’s ethnicity and religion, they find that urban children are 13

percentage points less likely to attend primary school than their rural counterparts.33

Table 2.1 presents summary statistics for my matched sample and the adult male

population in early twentieth century Canada. Approximately one half of the linked

sample lives in Ontario, one quarter in Quebec and the rest in Nova Scotia, New

Brunswick, Manitoba and British Columbia. The sample equally represents men of

English, Scottish, Irish and French descent. The average male stature is 5 feet and 7

inches, which is approximately 3 inches shorter than the average stature among modern

prime-aged Canadian men (Cranfield et al., 2007). Roughly 8 percent of the linked

sample was born to white collar fathers; 30 percent to skilled laborers; 38 percent to

9 4 farmers and agricultural labors; and lastly, 24 percent to unskilled workers.

The linked sample is not representative of the population of prime-aged Canadian

men. The summary statistics indicate that my sample over represents blue collar workers

and Scotts, and under represents agricultural workers and non-English/French ethnic

groups. These results are not attributed to the linkage procedure; rather, the Canadian

government primarily recruited in urban centers, and thus drafted relatively more

unskilled, semi-skilled, skilled and white collar workers (Brown and MacKenzie, 2005).

33 Urban communities contain 5,000 or more individuals.

2.5 Empirical Framework

My empirical task is to assess the extent to which cognitive and non-cognitive

abilities contribute to the correlation between height and wages. I estimate the stature

premium by applying ordinary least squares (OLS) to the following regression equation

(omitting individual subscripts):

= β + ρX + , (1)

where is the logarithm of average hourly earnings, is adult height, X is a vector of

exogenous covariates that are determined before labor market entry (e.g., their age and

age squared) and is an error term.

0 5

The OLS coefficient on height, OLS, represents stature’s direct contribution to

earnings, and the extent to which stature is correlated with earnings via its positive

association with cognitive and non-cognitive aptitude. I measure each aptitude’s

respective contribution to the height premium by separately including cognitive and non-

cognitive controls into equation one.34 A substantial reduction in the coefficient on height

indicates that taller workers are relatively more productive, due in large part, to stature’s

strong correlation with the respective ability, which suggests that height is a relatively

useful proxy for historical human capital.

34 To be clear, the cognitive controls include the general cognitive ability and abstract thought occupation requirements. The non-cognitive measures include the general non-cognitive ability, communication, authority and social isolation requirements. To examine the extent to which physical abilities contribute to the stature premium, I include the fine-motor skill and physical capacity vocational requirements.

My analysis may overstate the extent to which cognitive and non-cognitive

abilities contribute to the height-earnings relationship. For instance, some social scientists

attribute occupational placement to discrimination. Most individuals, they argue, perceive

moderately taller workers to be more attractive and associate attractiveness with

competence.35 This perception encourages employers to overestimate their taller

employees’ productivity, and thus promote them into positions requiring more ability. In

this case, a taller worker’s occupational requirements are associated with their actual

aptitudes and an attractiveness premium. Hence, including occupational requirements

into equation one removes the extent to which both ability and attractiveness contributes

to the stature premium.

To test the discrimination conjecture against my pathway, I compare the returns to

stature associated with individuals in the following stature categories: very short, short,

1 5 average, above average, and very tall.36 I estimate each group’s stature premium by

regressing the natural logarithm of wages on the above stature categories, where very

short individuals is the omitted category. The attractiveness argument indicates that the

stature premium is a step function; only moderately taller workers receive a premium,

while average, short and very tall workers receive no premium. My hypothesis, in

contrast, suggests that the height premium is non-linear. I attribute the returns to stature,

in part, to the neural growth accompanying physical growth. This biological relationship

35 This pathway is similar to the environmental argument mentioned earlier. However, this particular view argues that individuals do not provide more attractive individuals with more investments in their social skill. 36 Very tall individuals are more than two standard deviations taller than average. Above average individuals are one to two standard deviations taller than average. Short individuals are one to two standard deviations shorter than average, and very short individuals are more than two standard deviations shorter than average.

decreases as individuals grow, which implies that growth provides shorter individuals

(individuals with more potential growth) with relatively more neural development, and

thus greater returns to stature.

Nepotism may also contribute to the stature premium. For example, white collar

workers may raise taller sons, and use their connections to ensure that their sons’ receive

lucrative careers. In early twentieth century Canada, transportation and communication

were relatively slow and limited, suggesting that a parent’s connections were probably

restricted to their township. Therefore, sons residing outside their parents’ community are

less likely to acquire occupations via parental connections. Assuming nepotism strongly

contributes to the stature premium, sons residing near their parents should receive

relatively greater stature premiums.37

2 5 2.6 Results

Height and Productivity

I examine the extent to which cognitive and non-cognitive abilities contribute to

the height premium. Table 2.2 presents regression results of the natural logarithm of

37 The choice to migrate is potentially influenced by unobservable characteristics that affect earnings. For example, more intelligent individuals may migrate to areas that pay a cognition premium. I examine the extent to which numerous individual characteristics influence the migration choice by regressing the choice to migrate on individual income, stature, age, ethnicity, father’s and mother’s ability to read, and father’s socioeconomic group. The results indicate that only ethnicity plays an important role in the migration choice. English and Irish individuals are more likely to move away from their parents, while the French are more likely to stay with their parents. These results are available on request. It is possible that stature provides a direct contribution to individual productivity in more lucrative occupations. However, I do not know of any studies that address or study this topic.

gross hourly earnings on stature.38 Column I includes age and matching controls.

Columns II and III include proxies for parental and non-parental invests in education,

respectively. Columns IV to VIII include various combinations of cognitive, non-

cognitive and physical capability occupational requirements.

The results indicate that the height premium is substantial and statistically

significant at the 1 percent level. A one inch increase in adult stature is associated with a

7.0 percent increase in earnings, ceteribus paribus. This estimate is approximately nine

standard deviations greater in value than the stature-earnings estimates reported using

modern data (Case and Paxson, 2008; Schick and Steckel, 2010). This result supports the

neurobiological pathway, which suggests that shorter individuals receive relatively more

neurological development—and thus productivity gains—via increases in statures. The

Canadian servicemen are approximately three inches shorter, on average, than the

3 5 individuals examined in the respective modern studies. Therefore, an increase in stature

should provide the servicemen with relatively more neurological growth, and thus a more

sizeable stature premium.

The evidence in columns II and III suggest that education-related investments

explain a substantial portion of the stature premium. Controlling for non-parental

investments reduces the height coefficient from 0.070 to 0.063, and including parental

investments further decreases the estimate from 0.063 to 0.055. Comparing the stature

38 Clustering the standard errors at the individual levels does not change the results.

estimates in columns I and III, education-related investments explain at least 20 percent

of the stature premium.39

The results in columns IV–IX suggest that cognitive and non-cognitive abilities

provide a sizeable contribution to the stature premium. Including only cognitive

occupational requirements reduces the height coefficient approximately two standard

deviations, from 0.055 to 0.047. Controlling for cognitive and non-cognitive

requirements further reduces the stature estimate an additional two standard deviations,

from 0.047 to 0.039. Including physical abilities, in addition to cognitive and non-

cognitive aptitudes, does not change the estimated height coefficient, which suggests that

physical skills provide an unsubstantial, independent contribution to the stature premium.

Comparing the stature estimates in columns I and IX, human capital proxies explain at

least 45 percent of the height premium.

4 5 The appendix provides present-day evidence that occupational requirements

roughly capture innate ability. This result suggests that my analysis may substantially

understate the extent to which cognitive and non-cognitive abilities contribute to the

stature premium. To test this conjecture, I use a contemporary study, the British National

Childhood Development Study, to examine the extent to which innate abilities and

occupational requirements explain the stature-earnings relationship.40 The results indicate

that present-day occupational requirements understate aptitude’s contribution to the

stature premium roughly 60 percent. Assuming the contemporary evidence approximately

39 It is likely that the occupational characteristics are also correlated with environmental investments. 40 The appendix provides a detailed discussion regarding the National Childhood Development Study, the methods used to test the conjecture, and the results associated with the analysis.

applies to my study, the results indicate that innate ability may explain up to 80 percent

of the historical stature premium.41

Discrimination, Nepotism, Height and Wages

I examine the extent to which stature is associated with discrimination by

regressing earnings on several stature indicators: short, average, moderately tall, and very

tall (very short is the omitted category). Figure 2.1 graphs the results. The results indicate

that the discrimination pathway provides a modest contribution to the stature premium.

Moderately taller workers (the workers most likely to receive a discrimination premium)

receive only a modest, discontinuous increase in earnings. The results are consistent with

the neurobiological pathway. For shorter men, an increase in stature results in a greater

5 5 increase in earnings compared to their taller counterparts. Furthermore, the returns to

stature increase at a decreasing rate.42

41 Formally, including cognitive and non-cognitive scores, in addition to individual and parental background measures, may reduce the estimated height coefficient from 0.070 to 0.015. It is worth emphasizing that the remaining stature premium is substantial. A one inch increase in adult height is associated with a 1.5 percentage point increase in earnings. Considering that occupational requirements moderately capture cognitive and non-cognitive abilities, it is reasonable to assume that they also roughly measure physical ability. This rationale suggests that the remaining premium may be due to stature’s direct contribution to strength, endurance and fine-motor skill. 42 My results may overstate stature’s association with innate ability. The stature premium also captures the extent to which the market rewards aptitude (i.e., the more the market values aptitude, the greater the stature premium). Hence, in markets which particularly value ability, my results may reflect the market’s preference towards aptitude more so than that stature is strongly associated with ability. However, the evidence in the appendix suggests that this issue is a minor concern for my study. The returns to cognitive and non-cognitive occupational requirements are approximately similar in early twentieth century Canada and contemporary Britain. Furthermore, early twentieth century Canadians are shorter and receive a statistically significantly greater stature premium than present-day Britannians. Given this evidence, it is sensible to conclude that my results present evidence that stature is strongly associated with innate ability in the past, rather than reflecting a strong market preference towards aptitude.

I next examine whether nepotism contributes to the extent to which occupational

requirements explain the stature-earnings relationship. Table 2.3 reports regression

results of hourly earnings on stature for sons living ―near‖ and ―far‖ from their parents.43

The results indicate that nepotism plays an unimportant role in determining the stature

premium. Near sons receive a lower stature premium than far sons. A one inch increase

in adult stature is associated with 6.4 and 7.2 percent increase in earnings among near and

far sons, respectively. In addition, occupational requirements explain an approximately

equal portion of the stature premium among near and far sons (approximately 47

percent).

2.7 Conclusion

6 5 Human capital plays a pivotal role in economic growth. Unfortunately, standard

human capital measures are scarce (especially in developing and historical economies)

which has impeded empirical work. This paper presents microeconomic evidence that

adult height is a useful proxy for human capital, particularly in economies where human

capital measures are scarce (i.e., historical and developing economies). To examine this

topic, I construct a historical panel data set that matches Canadian men across their

World War I registration papers and their 1891 and 1911 census manuscripts. I show that

taller men earn substantially more than their shorter counterparts, and that a sizeable

43 Sons that live in the same township as their parents are considered to live near their parents and distant otherwise. This analysis omits distant sons residing with a close relative, such as their grandfather, uncle or brother.

portion of the height-earnings relationship is due to height’s association with both

cognitive and non-cognitive abilities.

The results suggest several areas for further research. One is to examine the extent

to which stature was correlated with cognitive and non-cognitive abilities in the past.

These studies will provide more conclusive evidence that stature was associated with

ability during historical periods, and thus is a suitable proxy for historical human capital.

Historical cognitive and non-cognitive measures are scarce. However, several sources

report these measures along with stature data. For example, the United States military

conducted cognitive and non-cognitive exams during WWI.

This paper suggests that taller populations are more cognitively and non-

cognitively able, and thus more productive. Researchers can examine this question using

a standard cross-region growth regression framework, where a region’s gross domestic

7 5 product (GDP) per capita is expressed as a function of height, human capital, and other

variables that contribute to economic growth (e.g. physical capital, institutions, and

technology). If height is a suitable proxy for a nation’s human capital stock then a

regression of GDP on the average stature of a region’s working-age population should be

positive, substantial and statistically significant. Furthermore, including human capital

measures should reduce the height estimate to approximately zero. 44

44 A region’s average intelligence and stature are positively associated with its GDP, and thus this analysis will overstate cognition’s contribution to the stature premium. A solution to this problem is to instrument stature with a suitable instrument.

Adult Male Population in 1911 My Sample Residing in Province Prince Edward Island 0.01 0.00 Nova Scotia 0.07 0.08 New Brunswick 0.05 0.05 Quebec 0.28 0.22 Ontario 0.35 0.47 Manitoba 0.06 0.05 Saskatchewan 0.07 0.02 Alberta 0.05 0.03 British Columbia 0.05 0.08 Yukon Territory 0.00 0.00 Northwest Territories 0.00 0.00

Ethnic Origin in 1911 English 0.26 0.29 Irish 0.15 0.20 Scottish 0.14 0.23 French 0.29 0.22 Other 0.16 0.06

Household Characteristics in 1891 Average Household Size 5.30 7.13 Father can Read 0.87 Father can Write 0.84 Mother can Read 0.90

8 5 Mother can Write 0.85 Father White Collar 0.08 Father Skilled 0.30 Father Farmer or Agriculatural Laborer 0.38

Socio-Economic Class 1911 White Collar 0.11 0.12 Skilled 0.37 0.42 Farmers & Agricultural Laborers 0.39 0.12 Laborers 0.14 0.34

Individual Characterisitics Adult Height (Inches) 67.01 Average Hourly Wage (C$) 0.23 Married during WWI 0.52 0.49 Population Living in Urban Township 0.45 0.57 N 1,413 Table 2.1: Descriptive Statistics in Historical Canada

Dependent Variable: Log Hourly Earnings (1) (2) (3) (4) (5) (6) (7) (8) Adult Height (Inches) 0.070*** 0.059*** 0.055*** 0.047*** 0.049*** 0.048*** 0.039*** 0.039*** (0.005) (0.005) (0.005) (0.004) (0.004) (0.004) (0.004) (0.004) Controls Province of Residence in 1890 and 1910 X X X X X X X Resided in an urban township in 1890 and/or 1910 X X X X X X X Parental Investment Proxies X X X X X X Cognitive Ability X X X Non-Cognitive Ability X X X Fine-motor skill and strength X X N 1,412 1,412 1,412 1,412 1,412 1,412 1,412 1,412 Adjusted R2 0.140 0.256 0.289 0.388 0.335 0.362 0.443 0.443 Table 2.2: Log Average Hourly Earnings, Cognitive Skill, Non-Cognitive Skill, and the Returns to Height in Historical Canada

9 5 squared in 1910, whether they were married, and whether they lived with their parents. Parental investment comprises the mother’s and father’s ability to read and write, household size, and the father’s socioeconomic group. Cognition separately includes the general intelligence and problem solving measures Non-cognition separately includes the general social intelligence, communication, social interaction and coping with stress components; and fine-motor skilland strength is proxied for using the motor-skill and physical capacity components.

(1) (2) (3) (4) (5) (6) Sons that live near parents Sons that live far from their Adult Height (Inches) 0.072*** 0.052*** 0.037*** 0.064*** 0.052*** 0.034*** (0.015) (0.011) (0.010) (0.005) (0.005) (0.005) Controls: Parental and non-parental investment controls X X X X Occupational requirement controls X X N 279 279 279 1,034 1,034 1,034 Adjusted R2 0.14 0.27 0.41 0.14 0.27 0.42 Table 2.3: The Relationship between Earnings and Nepotism

60 0 6 income, father's socio-economic group during childhood, mother and father's academic attainment, mother and father's

involvement in child's education. Sons that live in the same township as their parents are considered to live near their parents. This analysis omits sons that live far from their parents, but also live with a close relative.

0.600

0.500

0.400

0.300

0.200 Log Earnings Log

0.100

0.000 164-169 169-173 173-177 177-180 180+ Adult height reported in WWI medical documents(centimeters) Figure 2.1: Log Earnings and Stature for Men in Historical Canada

Notes: Figure 2.1 reports the returns to stature associated with short, average statured, above average

1 6 statured, and very tall individuals. The results reported include age, race, and matching controls.

Chapter 3

Health Makes Wealth: The Relationship between National Productivity and Health

3.1. Introduction

In the coming years, the United Nations (UN) plans to spend approximately $75

billion each year on programs which improve health outcomes in developing nations.

Some policy makers argue that these programs are too expensive; however, the UN

maintains that improving health is a practical investment. Good health, they argue, plays

an important role in economic development. Healthy individuals are stronger and more

2 6 energetic, and thus are more able to complete jobs requiring more strenuous physical

activity. Furthermore, good health promotes attentiveness and concentration, which

enhances an individual’s ability to accumulate human capital. Given this debate, I

measure the gains in national productivity associated with an increase in the respective

nation’s health status, and examine the extent to which physical capacity and

accumulating human capital contribute to the relationship between health and

productivity.

Recent research suggests that health plays an important role in national

productivity. Table 3.1 summarizes sixteen studies that estimate the extent to which

health contributes to economic performance.45 The results generally indicate that the

relationship between health and productivity is positive, substantial and statistically

significant. For instance, Bloom et al. (2004) measure the extent to which health (which

they capture using life expectancy) is associated with real gross domestic product (GDP)

per capita in 1990. Using nonlinear two stage least squares, they show that a one year

increase in life expectancy results in a 4 percent increase in output per worker.

Good health is thought to contribute to productivity via its correlation with

physical capability and accumulating human capital. Using data from the Medical

Outcomes Study Short-Form Healthy Survey, Stokes et al. (2006) show that good health

provides a substantial contribution to individual physical capability. They measure

3 6 individual health using the incidence with which the individual contracts allergic rhinitis,

a mild yet chronic allergic reaction that is prevalent in most populations. The authors

show that constant allergic rhinitis is associated with a 35 to 40 percent reduction in work

capacity (hours worked). Similarly, Mannerfeldt and Pettersson (2004) provide evidence

that health provides a moderate contribution to accumulating human capital. They

conducted a survey examining secondary students in Sweden. They show that constant

allergic rhinitis is associated with a one to two reduction in letter grades.

The microeconomic evidence suggests that physical capability and accumulating

human capital contribute to the relationship between individual health and productivity.

45 These studies do not indicate whether health directly contributes to productivity, or whether it is merely a proxy for other missing or mismeasured variables. 63

However, there is no guarantee that this microeconomic relationship translates into a

comparable macroeconomic relationship. This paper contributes to the literature by

examining the extent to which the above pathways explain the gains in national

productivity associated with an increase in national health. To examine this topic, I

construct a cross-country data set covering 45 countries in 1960 to 2000. I show that

healthier populations are substantially more productive than their unhealthy counterparts,

and that this result is largely due to a strong correlation between health and human

capital.

3.2 Health and Productivity

Health directly contributes to productivity via increases in physical capability.

4 6 Unhealthy individuals are more prone to catching diseases. To combat most diseases, the

body allocates more nutrients to immune processes and less to non-immune processes

(e.g., physical processes, such as strength, stamina and motor coordination). As a result,

unhealthy individuals are relatively weaker, slower and uncoordinated, and thus are

substantially less able to complete strenuous physical activities requiring complex

coordinated actions.

Healthy individuals also undergo more muscular and joint development (Eveleth

and Tanner, 1990). These developments improve physical strength and stamina, which

enhance an individual’s ability to work harder and longer. Muscular and joint

developments also improve various fine-motor skills, such as physical dexterity and

hand-eye-foot coordination (Malina et al., 2004). As a result, healthier individuals are

more productive in occupations requiring more complex coordinated actions.

Health also indirectly contributes to productivity via increases in human capital.

Sick individuals are more inclined to skip school. Furthermore, their malady reduces their

capacity to concentrate and pay attention, which impairs their ability to learn. Behrman et

al. (2004) suggest that poor health also delays when children enroll in school. Most

parents, they argue, allow their children to attend school once they are as physically

developed as their peers. Because unhealthy children are relatively shorter than their

healthy counterparts, the authors conclude that these children enter school substantially

later.46

The epidemiology literature suggests that an important prerequisite to good

health, consuming a nutrient diverse diet, promotes cognitive development. Figure 1.1

5 6 illustrates the epidemiology pathway. Formally, when growing individuals consume

nutrients, their bodies initially use the nutrition to survive (carry out biological

maintenance and combat diseases). Bodies that consume more nutrients than are

necessary to survive, use the surplus nutrition to simultaneous promote physical and

neurological growth (Tanner, 1978; Hoppe et al. 2004; Gunnell et al. 2005; Scheepens et

al. 2005; Kerver et al. 2010).

Neurological development promotes cognitive development. The brain grows via

strengthening synapses and producing grey and white matter. Grey matter increases the

rate with which we process and interpret stimuli, while white matter and stronger

46 The authors suggest that unhealthy children attend school approximately one to two years later than their healthy counterparts. 65

synapses improve our ability to organize and transmit neurological signals (thus

expediting our ability to coordinate thoughts). These developments increase our cognitive

potential (Paus et al., 1999). For instance, problem solving requires an individual to

process and interpret a problem, organize their thoughts in order to propose solutions, and

use both capabilities to test which solution is best.

Recent studies support the epidemiology pathway. For example, Liu et al. (2003)

examines the extent to which nutritional status contributes to cognitive ability. Using data

from the Mauritius Longitudinal Study, they show that malnourished children score

substantially lower cognitive exams. Malnourishment at age 11 is associated with a 6

percent decrease in average cognitive scores.

3.3. Health Indicators

6 6

Health is complicated to measure. As table 3.1 indicates, most studies assess a

population’s vigor using its average life expectancy. However, some researchers concede

that serious measurement problems reduce this indicator’s versatility (Pradhan et al.

2003). For instance, life expectancy is commonly estimated using data that does not

apply to every age group in the present population (e.g., prime-age adults). Furthermore,

data limitations are common, resulting in researchers constructing indicators using small

sample sizes with incomplete information. As a result, most measures are noisy and are

not comparable across countries (Stiefel et al. 2010).

Given these issues, I measure a population’s vigor using its average adult height

(stature) among prime-age men. Stature data present two particular advantages. The data

are readily available, and stature is associated with both the direct and indirect pathways

through which health contributes to productivity (Haddad and Bouis, 1991; Steckel,

1995; Strauss and Thomas, 1998; Case and Paxson, 2008; Schick and Steckel, 2010).

Height is widely recognized as a useful proxy for health (Steckel, 1995; Schultz,

2005; Weil, 2007). Intuitively, a person’s stature represents their net nutrition up to the

mid-twenties. Individuals who allocate more nutrients towards growth commonly grow

up to be taller adults. The extent to which nutrients are invested in growth depends on

nutritional inputs and claims. Prior to investing nutrients in growth, the body claims

nutrients to maintain metabolism, combat diseases and as energy for physical activities.

This “growth” process suggests that individuals who are raised in more healthy

7 6 environments (i.e., environments with less disease, less arduous physical activities, and

more available, nutritious food) should grow up to be taller adults on average, implying

that stature is a useful proxy for health (Tanner, 1978; Hoppe et al. 2004; Bozzoli et al.

2009; Kerver et al. 2010).

Contemporary studies support the intuition. For example, Smith et al. (2000)

present evidence that taller individuals live longer than their shorter counterparts. Using

British data from the Renfrew/Paisley general population study, the authors show that an

increase in stature from the shortest to tallest quintile—an increase of approximately 10

centimeters—reduces mortality rates approximately 25 percent.47 Bozzoli et al. (2009)

47 In particular, the mortality rates associated with stroke, respiratory disease and chronic heart disease. 67

find that stature is strongly correlated with another popular health indicator: postneonatal

mortality (PNM), the number of deaths occurring between 28 days to 11 months of life.

Using data from the United States and 11 European countries, the authors report that

PNM explains more than 60 percent of the variation in stature across these countries.

Furthermore, reductions in PNM across 1950 to 1980 explain approximately the entire

increase in adult heights across this time period.

Individual height is closely related to the pathway through which health

contributes to physical capability. Taller individuals undergo more muscular and joint

development, which increases their strength, endurance and overall physical resilience

(Malina et al., 2004). These characteristics provide substantial contributions to individual

physical capacity. For example, strength and endurance enable workers to work longer

and harder, while resilience substantially reduces susceptibility to pathogens (Samaras,

8 6 2009).48

Individual height is also associated with the indirect pathway through which

health contributes to human capital. Figure 1.1 indicates that the body grows by

converting surplus nutrients into growth materials and growth stimulating chemical

messengers: e.g., shared insulin-like growth components, such as thyroid and growth

hormones (Tanner, 1978; Gunnell et al. 2005; Scheepens et al. 2005). These messengers

are argued to stimulate simultaneous physical and neural growth, and to substantially

48 Most epidemiologists conclude that taller individuals are healthier on average. This conclusion is associated with studies that examined the health outcomes between populations in developed and developing countries. Most developed nations are substantially healthier and taller than developing populations, and thus the conclusion (Samaras, 2009). 68

develop the neurological regions managing our cognitive processes.49 Hence, this

pathway suggests that taller, and thus healthier, adults are more cognitively developed

(Oppenheimer and Schwartz, 1997; Thompson and Potter, 2000; Fuster, 2001; Blair,

2004; Bechara, 2005).

Recent microeconomic research indicates that stature is strongly associated with

cognitive ability. Using modern British data from the National Childhood Development

Study, Case and Paxson (2008) show that taller children score substantially higher on

cognitive assessments. A one standard deviation increase in height at age 7

(approximately two inches) is associated with a 10 percent of a standard deviation

increase in math and reading test scores at age 11. This effect is approximately as large as

a two standard deviation increase in household income.

Despite its strengths, stature is a noisy measure of health status. Individuals with

9 6 comparable adult heights occasionally achieve moderately disparate outcomes in health,

and vice versa. For example, consuming a nutritious diet may enable individuals that are

genetically predisposed towards shorter statures to reach a comparable adult height to

those that consume relatively poor diets and are predisposed to taller statures. In this

example, adult height indicates that these individuals have approximately similar physical

and cognitive capabilities, while in actuality, the former are probably more physically and

cognitive capable.

Height represents an individual’s health environment up to their mid-twenties. As

a result, stature cannot convey the environmental conditions occurring in adulthood. This

49 These regions are the insular, anterior cingulated, medial prefrontal and frontal cortices. On average, the extent to which these biological channels operate on physical and neural growth is substantial. However, these studies also indicate that responses vary at the individual level. 69

characteristic is particularly problematic when studying developing countries. For

example, children in developing countries may be raised in very oppressive

environments, and thus achieve relatively short statures. However, during adulthood their

environment may substantially improve (e.g., their nation may remove various sources of

disease, improve sanitation, and provide cleaner water) resulting in better health.

Gender inequality reduces the extent to which male stature accurately represents

population health. For instance, in more male-oriented societies, men receive more health

resources, and thus are relatively healthier. As a result, male stature overstates female

health, and thus population health, in these countries.

Finally, some researchers argue that genetic variation intuitively explains most

cross country differences in stature, and thus conclude that adult height poorly captures

health. However, contemporary studies suggest otherwise. L.A. Malcolm (1974)

0 7 concludes that variation in height across populations is almost entirely due to

environmental disparities. Martorell and Habicht (1986) provide empirical evidence

supporting this claim. They show that children residing in Europe or claiming European

descent, residing in Africa or claiming African descent, or residing in the Middle East or

claiming Middle Eastern descent exhibit approximately similar growth profiles when

growing up in similar environmental conditions.

3.4. Empirical Framework

My empirical task is to examine the extent to which health contributes to

economic prosperity. The standard way to measure this relationship is to apply ordinary

least squares (OLS) to:

= β + ρ + , (1)

where is country i’s average annual growth rate in gross domestic product (GDP) per

capita across 1960 to 2000, is country i’s average health status as represented using

the average adult stature among prime-age men, are other characteristics contributing

to nation i’s productivity (e.g., a country’s GDP per capita in 1960, institutional quality,

1 7

climate, geographic characteristics and regional characteristics) and is an error term

(Hanushek and Woessmann, 2008).

My analysis raises two endogeneity issues. First, there exists a positive

simultaneity bias between national productivity and health. Wealthier populations can

consume more nutrition and health care. Furthermore, rich nations can invest more

resources in programs promoting societal health, such as sanitation, hygiene, clean water

projects, eliminating environment related diseases and constructing health care facilities.

Second, health is measured with error, resulting in a substantial downward attenuation

bias.50 Hence, the estimated health coefficient may exhibit either a net upward or

downward bias.

The standard way to remove these biases is to use two-stage least squares (2SLS).

This procedure requires instrumental variables (IV) that contribute to economic

performance only through the productivity enhancing pathways associated with health,

and that are uncorrelated with the other unobservable characteristics determining

economic performance.51 Some researchers suggest that some convenient instruments are

a nation’s average solar insolation, skin color (pigmentation) and insolation interacted on

pigmentation (Jablonski and Chaplin, 2000; Carson, 2009).

Insolation and skin color are the primary sources of vitamin D3 (Holick, 2007).

Insolation is the extent to which an individual is exposed to sunlight, and thus ultraviolet

52 radiation. This particular radiation induces the skin to produce vitamin D3 until it

2 7 generates enough vitamin D3 to optimize individual health. Hence, individuals receiving

more sunlight probably consume more optimal amounts of vitamin D3 (Carson, 2009).

Skin pigments, such as melanin, reduce the extent to which ultraviolet radiation

penetrates the skin, and thus induces vitamin D3 production. Hence, individuals with

more melanin (darker skin colors) require more solar insolation to produce the same

amount of vitamin D3 as individuals with lighter pigments (Jablonsky and Chaplin, 2000).

50 Recent studies support this conjecture (Schultz, 2005; Weil, 2007). Using data from the Living Standard Measurement Surveys in Ghana, Schultz (2006) shows that the instrumental variables estimate on male health is 3.8 times greater than the OLS estimate. Using OLS, a one inch increase in adult male health is associated with a 1.48 percent increase in average hourly earnings. Whereas, using two-stage least squares, a similar increase in health is associated with a 5.7 percent increase in wages. 51 A particularly suitable IV should contribute to productivity through the pathways associated with the health characteristics correlated with adult stature. 52 Insolation is the incidence with which sunlight directly reaches objects on the earth’s surface (i.e., the extent to which an individual is exposed to sunlight). Formally, it is a measure for solar energy received for

a given surface area at a given time. If w equal watts and m equals meters, and i equals insolation, i = .

Vitamin D3 is conducive to physical growth (Fang et al., 2007; Carson, 2008;

Kremer et al., 2009). The nutrient promotes calcium absorption and plays a pivotal role in

skeletal growth. Recent research provides evidence supporting this conjecture. For

instance, Banegas et al. (2001) show that solar insolation provides a moderate

contribution to adult stature. Using cross-sectional data from Spain, they show that men

raised in more insolated regions are approximately three centimeters taller on average.

Recent research indicates that vitamin D3 also plays a pivotal role in immunology

(Deluca and Cantorna, 2001; Von Essen, 2010). The immune system removes pathogens

using specialized immune cells. These cells use vitamin D3 as an energy source,

suggesting that the vitamin is integral to a healthy immune system (Von Essen, 2010).

Immunology studies support this conjecture, indicating that individuals with vitamin D3

deficiencies are more prone to catching diseases and remaining sicker longer (GIMR,

3 7 2009)

Insolation and skin pigmentation are exogenous to economic performance.

However, these variables are potentially correlated with other characteristics that are

correlated with national productivity. For example, both variables are strongly associated

with latitude and climate. Regions closer to the equator receive relatively more insolation

and are more populated with individuals containing more pigments. Most economists

argue that equatorial (tropical) climates impede property right institutions, and thus

productivity (Acemoglu et al., 2002; Easterly and Levine, 2003). For example, Hall and

Jones (1999) suggest that tropical regions were particularly oppressive to Europeans,

resulting in most Europeans settling non-tropical climates. As a result, non-tropical

regions were more likely to adopt European property rights, which the authors argue were

the most conducive to economic development.

Tropical environments are also relatively oppressive to agricultural productivity.

For example, most tropical soils are quite poor. Compared to the soils in most temperate

regions, the earth in tropical regions is more acidic and nutritionally destitute. Hence,

tropical soils are relatively ill-suited towards growing most high yield crops, such as rice,

wheat and corn (Gallup, Sachs and Mellinger, 1999).

Another potentially suitable instrument is a country’s infant mortality rate (IMR)

in 1960.53 The IMR is the number of deaths in children under one year of age per 1000

live births in the same year. This measure is intuitively associated with population health.

Countries with poorer health conditions (e.g., malnutrition, disease, poor hygiene and

sanitation, and unclean water) probably report relatively higher infant mortality rates

4 7 (WHO, 2009). Empirical research substantiates the intuition. Using data on 180

countries, Reidpath and Allotey (2003) examine the extent to which IMR is correlated

with another popular health indicator: the disability adjusted life expectancy (DALE).

The authors show that the correlation between IMR and DALE is 0.91.

The epidemiological literature indicates that good neonatal health is strongly

associated with good adult health outcomes. Using data from the Demographic and

Health Surveys, Victoria et al. (2008) show that poor neonatal health (as measured using

birth weight) impedes adult health outcomes in several countries. A one kilogram

increase in birth weight is associated with a 3.3 centimeter (cm) increase in adult height.

53 A nation’s infant mortality in 1960 is due in part to the nation’s per capital income in 1960 and national health accounts. As a result, I include these variables into my analysis. 74

Furthermore, low birth weight children are also more susceptible to pathogens, and

display poorer motor skills. The authors also provide evidence that poor neonatal health

is associated with human capital accumulation. A one kilogram increase in birth weight is

associated with an additional 0.30 years of schooling.

A nation’s IMR in 1960 contributes to the population’s productivity in 1960 to

2000 via its relationship with health. All else equal, the higher a country’s IMR in 1960,

the unhealthier its resulting prime-age population in 1960 to 2000. However, infant

mortality rates are also potentially associated with other characteristics that are correlated

with economic performance. For instance, tropical climates are relatively conducive to

more oppressive disease environments, and thus are particularly ruinous to productivity

and reducing infant mortality.

Institutional quality also plays an important role in reducing infant mortality.

5 7 Using data from the World Development Report, Gupta et al. (2002) show that the

relationship between corruption and infant mortality is positive, sizeable and statistically

significant. The authors measure corruption using the institutional measures reported in

Kaufmann et al. (2008). Using two-stage least squares, they show that a one standard

deviation increase in corruption is associated with a 20 percent increase in child mortality

The resulting 2SLS health estimate, 2SLS, represents the extent to which national

health is correlated with economic performance via its positive association with physical

capacity and human capital. To measure each pathway’s respective contribution to the

above relationship, I separately include cognitive test scores and physical capacity

proxies into equation one. A substantial reduction in 2SLS suggests that health is strongly

correlated with the respective channel.

3.5. Data

My goal is to examine the extent to which health contributes to national

productivity. To study this topic, I construct a cross-country data set covering 45

countries. The data contains each country’s average institutional quality, national

productivity latitude, climate, solar insolation, cognitive ability, physical capability, and

adult height among prime-aged men.

I measure a nation’s productivity using its average annual growth rate in GDP per

capital across 1960 to 2000. GDP per capita data were adjusted for international

6 7 purchasing power parity (PPP) dollars, and were obtained from the Penn World Tables

Mark, version 6.1 (Summers, Heston and Aten, 2002).

Height data were collected from various surveys examining national health,

nutrition and anthropometry. Table 3.2 presents the average adult height (in centimeters)

among prime-aged men in each respective country. The table also reports each study

associated with the respective stature measure and each study’s observation period,

sample size, age range, and whether the stature data was measured or self-reported. Most

surveys were conducted during 1990 to 2000, and sampled prime-age men—males ages

18 to 40. Furthermore, stature measures were commonly assessed using a tape measure or

scale, and are accurate to one centimeter.54 The average height is 173 centimeters, and the

standard deviation is 5.2 centimeters. Men are shortest in Peru (161.4 centimeters) and

tallest in the Netherlands (182.2 centimeters).

I measure a population’s human capital stock using its average years of schooling

and cognitive ability test scores. Cohen and Soto (2007) report each population’s average

years of schooling in 1960. Cognitive test scores were provided by Hanushek and

Woessmann (2008). They estimate cognitive achievement using various international

math, science and reading test scores.55 The tests cover secondary school concepts that

are common across countries, and were conducted during 1964 to 2003. Using these data,

Hanushek and Woessmann (2008) construct a composite measure of cognitive

achievement that is comparable across time and countries. For a detailed discussion

regarding this measure, please consult Hanushek and Woessmann (2008).

7 7 Solar insolation measures were obtained from the NASA Total Ozone Mapping

Spectrometer (TOMS). The data represent an area’s average annual exposure to sunlight

during one complete insolation cycle, 1996 to 2005. To estimate a nation’s exposure to

sunlight, I take the weighted average insolation of every urban township in that respective

nation. Urban townships are areas containing one percent or more of a nation’s

population. TOMS recognizes areas by their latitude and longitude. I acquire each urban

township’s precise latitude and longitude coordinates (accurate to one half of a degree)

54 My analysis includes indicators capturing whether stature measures were self-reported, and whether a sample surveyed males ages 16 or 17. 55 These exams were conducted by the Organisation for Economic Co-operation and Development (OECD) and the International Association for the Evaluation of Educational Achievement (IEA). The exams cover secondary school 77

using data reported by the Geographic Names Information System. I also use these

latitude and population data to estimate each nation’s average weighted latitude.

Climate data was obtained from Gallup et al. (1999). The study reports the

percentage of the population residing in tropical climates. For a detailed discussion

regarding this measure, please consult Gallup et al. (1999). Infant mortality rates in 1960

were collected from the University of California, Santa Cruz Atlas of Global Inequality.

Skin color is most accurately measured using an E.E.L. (Evans Electric Limited)

reflectometer. Unfortunately, it is prohibitively expensive to collect aggregate

pigmentation data using this instrument, and thus these data are scarce. A more

economical approach is to evaluate a population’s average pigmentation using a standard

skin color scale, such as ranging from very light to very dark. However, most researchers

argue that this method is inaccurate. Hence, the next best alternative is to use an

8 7 estimation technique. Reletheford (1997) uses a nonlinear piecewise regression to

estimate the relationship between skin color and various environmental inputs. He shows

that latitude is the only input which determines skin color. Among males, skin color

increases 8.2 percent for every 10 degree decrease in absolute latitude in the Northern

Hemisphere, but only 3.3 percent for every 10 degree decrease in absolute latitude in the

Southern Hemisphere.

I estimate a population’s physical capacity using its average depth of hunger

(DOH). This measure represents the extent to which individuals fail to meet their

minimum dietary needs in terms of dietary energy. Formally, it captures dietary deficits

in kilocalories per person per day. DOH data were collected from the Food and

Agriculture Organization. This institution estimates each country’s average DOH in 1997 78

by comparing the average amount of dietary energy that undernourished people get from

the foods they eat with the minimum amount of dietary energy they need to maintain

their respective body weights and undertake light activity. The current child mortality

rate (CMR) is also another potential proxy for physical capability.56 The rationale is that

countries with higher child mortality rates contain relatively more oppressive health

environments which impair physical capability. I collect current CMR data from the

World Bank. This resource reports mortality rates in 2002.57

Institutional quality data was obtained from Kaufmann et al. (2008). The study

provides six measures of institutional quality:

1. Voice and Accountability – represents the extent to which a country’s citizens

can choose their government, as well as exercise and voice their political rights, civil

liberties, and views.

9 7 2. Political Instability and Violence – assesses perceptions regarding the chance

that the government may collapse via unconstitutional or violent means.

3. Government Effectiveness – measures the quality of public and civil services,

and their independence from political pressures. It also captures a government’s

competence at constructing and execute programs, and its credibility to continue to

support programs and services.

4. Regulatory Quality – appraises perceptions of a government’s ability to

construct and execute policies and regulations that allow and advocate private sector

development.

56 The child mortality rate is defined as the number of deaths of children under age 5 per 1000 live births. 57 IMR in 2002 is moderately correlated with IMR in 1960. The correlation is 0.48. 79

5. Rule of Law – the extent to which agents trust and accept a society’s rules and

laws. This measure examines the quality with which a country enforces contracts,

protects property rights, controls crime and violence, and maintains reliable and just

police and court services.

6. Control of Corruption – the degree to which public power is used to promote

private gain, especially among elites. This measure captures corruption ranging from

petty to grand offenses.

Separate measures were constructed to capture these governance characteristics in

1996, 1998 and 2000. To estimate a nation’s governance in, for example, voice and

accountability, I take the average of its voice and accountability scores in 1996, 1998,

and 2000.58

Health care quality data were collected from the WHO. The organization reports,

0 8 between 1995 and 2000, average total health expenditures as a percentage of GDP (THE)

and total government health expenditures as percent of (THE). The WHO argues that

these measures are suitable proxies for health care quality and the overall institutional

quality of health care systems. The rationale is that providing health care systems with

more money per capita allows them to provide more services and/or higher quality

services.

58 Please consult Kaufamn et al. (2008) for a detailed discussion regarding how these measures were constructed and their overall quality. I omit rule of law and voice and accountability from my analysis because they are weakly associated with economic performance once I control for the other institution variables. Furthermore, including these variables reduces my adjusted r-squared value. 80

Table 3.3 presents summary statistics for the entire sample, the poorest quartile

and the wealthiest quartile.59 The results indicate that poorer populations are substantially

worse off than their wealthy counterparts. For instance, the poorest quartile is

approximately 10 centimeters shorter than the wealthiest quartile—on average, the

poorest quartile is roughly 168.6 centimeters tall, while the wealthiest is 177.4

centimeters tall. Poorer populations also receive less education and reside in more

oppressive health environments. The wealthiest complete approximately six more years

of schooling than the poorest—8.4 years versus 2.3 years. Similarly, they are less likely

to reside in nations that are located in equatorial climates, and that provide substantially

less resources to health care.

3.6. Results

1 8

Ordinary Least Squares Results

I present evidence that good health provides a sizeable contribution to national

productivity. Table 3.5 reports OLS estimates of economic growth on health.60 Column I

includes measures of climate, latitude, institutional quality, geographic region, output per

worker in 1960, and health care quality. Columns II and III include proxies for physical

59 The poorest (wealthiest) quartile are those countries whose GDP per capita in 2000 was under (above) 6777 (23792) dollars. 60 The growth coefficients are reported in terms of basis points. As a reminder, one basis point is equal to 0.01 percentage points. 81

capacity and human capital accumulation, respectively. Finally, column IV includes

proxies for both physical capacity and human capital accumulation.

The results indicate that taller (healthier) populations are substantially more

productive than their shorter (unhealthier) counterparts. The height estimate in column I

is positive, sizeable and statistically significant at the one percent level. The

improvements in health associated with a one centimeter increase in adult height result in

a 0.12 percentage point increase in growth. This effect is quite sizeable. For example, the

upgrade in health associated with an increase in height from the 25th to the 75th percentile

of the international height distribution—an increase of approximately 10 centimeters—is

associated with an approximately 1 percentage point increase in national productivity.

This effect is as large as a two standard deviation increase in cognitive test scores.

The evidence in columns II and IV suggests that physical capacity provides a

2 8 substantial contribution to the relationship between economic growth and health.

Including only physical capacity measures reduces the stature estimate approximately 50

percent—from 0.120 to 0.058—and renders it statistically insignificant. Including

physical capacity proxies in addition to human capital measures further reduces the

stature estimate approximately 15 percent—from 0.045 to 0.025. Comparing the estimate

in column I to those in II and IV, physical capacity independently reduces the stature

estimate roughly 15 to 50 percent.

The results in columns III and IV indicate that human capital provides an even

greater contribution to the productivity-health relationship. Including only human capital

measures reduces the stature estimate approximately 60 percent—from 0.120 to 0.045.

Including human capital measures in addition to physical capacity controls further 82

reduces the stature estimate approximately 30 percent—from 0.058 to 0.025. Comparing

the estimate in column I to those in III and IV, human capital independently reduces the

stature estimate roughly 30 to 60 percent. These results suggest that human capital

contributes relatively more to the productivity-health relationship.

The evidence in column IV shows that physical capacity and human capital

accumulation explain most of the relationship between economic growth and health.

Including measures of physical capacity and human capital reduces the stature estimate

approximately 80 percent—from 0.120 to 0.025—rendering it both economically and

statistically insignificant. A one inch increase in adult health is now associated with a

0.025 percentage point increase in economic growth.

First Stage Results

3 8

This section provides evidence that adult height, and thus health, is strongly

associated with insolation, skin pigmentation, insolation-pigment interactions, and

infantile health. Table 3.4 presents the first-stage results between stature and my

instruments, condition on the above controls.61 The evidence indicates that only my

instruments are statistically significantly associated with adult height. In particular,

insolation provides a sizeable contribution to stature. Holding skin color constant, a one

unit increase in solar insolation is associated with a 2.72 centimeter increase in average

61 Scatter plots of stature on solar insolation, skin color and insolation * skin color suggest that stature is approximately linearly related to these variables. Please e-mail me at [email protected] to acquire these scatter plots. 83

adult height. This effect is as large as a four standard deviation increase in gdp per capita

in 1960.

A nation’s infantile mortality in 1960 is also strongly associated with its adult

stature in 1990 to 2000. All else equal, a one standard deviation increase in infant

mortality is associated with a 1.87 centimeter decrease in average adult height. This

effect is as large as a three standard deviation increase in gdp per capita in 1960.

Table 3.4 also presents evidence that my instruments are strong. I assess my

instruments strength using the standard Stock-Yogo test (Stock and Yogo, 2005). This

test rejects instruments when the relative bias between 2SLS and OLS is 10 percent or

more. The relative bias is approximately equal to:

≈ (2)

4 8

where is the adjusted R-squared in the first-stage regression, is the number of

instruments, and is the sample size. According to this test, the results in table 3.4

support my argument that my instruments are particularly strong. The relative bias

between 2SLS and OLS is less than 5 percent.

Main Results

I examine the extent to which a nation’s health status contributes to its economic

growth. Table 3.5 presents 2SLS regression results of economic growth on health.

Column V includes measures of a nation’s climate, latitude, institutional quality,

geographic region, output per worker in 1960, and institutional investment in health care.

Columns VI and VII include physical capacity and human capital accumulation

measures, respectfully. Finally, column VIII includes both physical capacity and human

capital accumulation controls. The corresponding OLS regression results are reported in

columns I-IV.

The results indicate that taller (healthier) countries are substantially more

productive than their shorter (unhealthier) counterparts. The health benefits associated

with a one centimeter increase in adult height results in a 0.126 percentage point increase

in growth. The contribution is very sizeable. For example, the improvement in health

associated with an increase in height from the 25th to the 75th percentile of the global

height distribution—an increase of approximately 10 centimeters—is associated with an

5 8 approximately 1 percentage point increase in national productivity. This effect is as large

as a two standard deviation increase in cognitive test scores.

The evidence in columns VI to VIII suggests that human capital may explain the

entire relationship between national productivity and health. Including only human

capital measures reduces the stature estimate to approximately zero, from 0.126 to -0.005.

A one centimeter increase in adult height is now associated with a 0.005 percent

reduction in economic growth. Including physical capacity controls, in addition to human

capital controls, reduces the stature estimate an insignificant amount, from -0.005 to -

0.014. These results support the argument that healthier populations are more productive

due to their ability to accumulate more human capital.

Economic Levels versus Economic Growth

Some economists argue that economic growth is a relatively poor measure of

economic performance. For instance, several recent growth models suggest that all

countries will grow at a common rate in the long run, and that most disparities in growth

are attributed to transitory shocks Furthermore, these models also associate a relatively

small amount of variation in economic welfare to growth. Instead, these models attribute

most disparities in performance to economic levels (for example, output per worker)

suggesting that output per worker is a more suitable proxy for economic prosperity than

economic growth (Parente and Prescott, 1994; Barro and Sala-i-Martin, 1997; Eaton and

Kortum, 1996). Given this argument, I re-examine my analysis using the natural

logarithm of real GDP per capita in 2000 as my measure of national productivity.

6 8 Table 3.6 presents 2SLS regression results of output per worker in 2000 on health.

Column I includes measures of a nation’s climate, latitude, institutional quality,

geographic region, output per worker in 1960, and institutional investment in health care.

Columns II and III include physical capacity and human capital accumulation measures,

respectfully. Finally, column IV includes both physical capacity and human capital

accumulation controls.

The results in Table 3.6, column I are relatively similar to those reported in Table

3.5, column V. Taller countries are substantially more productive than their shorter

counterparts. A one centimeter increase in adult height is associated with a 7.7 percentage

point increase in GDP per capita. This effect is as large as completing an additional year

of schooling. 86

The evidence in column II indicates that physical capability is weakly associated

with the relationship between national productivity and health. Including only physical

capacity proxies reduces the stature estimate from 0.077 to 0.076. The results in column

III suggests that human capital provides a sizeable, independent contribution to the

productivity-health relationship. Including only human capital measures reduces the

stature estimate approximately 50 percent—from 0.077 to 0.036—and renders it

statistically insignificant. Including physical capability controls in addition to human

capital controls does not significantly change the results. Overall, these results also

support the argument that healthier populations are more productive due to their ability to

accumulate more human capital.

3.7. Conclusion

7 8

There is an emerging view among growth economists that good health plays a

pivotal role to economic development through its association with accumulating human

capital and improving various physical capabilities. Using a self-constructed cross-

country data set covering 45 nations, I provide evidence that healthier populations are

substantially more productive than their unhealthy counterparts, and that this result is

largely due to the strong correlation between health and accumulating human capital.

Modern growth theories assign a pivotal role to human capital; however, its

contribution to growth (particularly, in developing and pre-twentieth century economies)

is little studied due to data limitations. Standard human capital measures, such as

cognitive test scores and accurate schooling data, are generally unavailable, while more

abundant alternatives, such as literacy rates, poorly proxy individual human capital

(A’Hearn et al., 2009; Albers, 1997; Woessmann, 2003). This paper presents evidence

indicating that average adult stature among prime-age men is so strongly associated with

human capital, that it reflects a relatively sizeable portion of the relationship between

human capital and national productivity. This is an important contribution to this

literature, as it suggests that the relatively data-abundant measure, stature, is a relatively

suitable proxy for human capital.

My results suggest other areas to study. For instance, a natural extension to this

paper is to examine the pathways through which other population health measures (for

example, morbidity rates, body mass index, and life expectancy) contribute to economic

performance. No indicator perfectly proxies health; instead each measure captures certain

8 8 health characteristics better than others. For example, life expectancy is relatively

conducive to capturing physical capacity; individuals that live longer probably work

longer. As a result, using life expectancy may provide a more accurate measure as to the

extent to which health contributes to productivity via its association with physical

capacity.

Health measure Coefficient Other covariates (all paper have the log of initial Study Data Estimator (in logs) (standard error) income per capita or per worker) Barro and Lee Life expectancy 0.073 (0.013) n = 85 in 1965 - 75, SUR with Male and female secondary schooling, I/GDP, (1994) n = 95 in 1975 - 85 country G/GDP, log(1 + black market premium), random effects revolutions

Barro and Sala-i- Life expectancy 0.058 (0.013) n = 87 in 1965 - 75 SUR with Secondary higher education, log(GDP)x human Martin (1995) n = 97 in 1975 - 85 country capital, public spending on education/GDP, I/GDP, random effects G/GDP, log(1 + black market premium), political instability, growth rate in terms of trade

Barro (1996) Life expectancy 0.042 (0.014) n = 80 in 1965 - 75 3SLS Male secondary and higher education, log(GDP) x n = 87 in 1975 - 85 male education, log fertility rate, government n = 84 in 1985 - 90 consumption ratio, rule of law index, terms of trade change, democracy index, democracy index squared,

87 inflation rate, continent dummies

Caselli Esquivel Life expectancy -0.001 (0.032) 25 year panel at GMM (Arellano- Male and female schooling, I/GDP, G/GDP, black

9 8 and Lefort 5 year intervals, Bond method) market premium, revolutions (1996) 1960 - 85, n = 91

Sachs and Warner Life expectancy 45.48 (17.48) 25 year cross section, OLS Openness, openness x log(GDP), land locked, (1997) Life expectancy -5.40 (2.24) n = 79 government savings, climate, instutitions, natural squared resource exports, growth in economically active population net population growth

Bloom and Life expectancy 0.027 (0.107) 25 year cross section, OLS Log(years secondary schooling), population Malaney (1998) 1965 - 90, n = 77 growth, growth of economically active populations, natural resources, openness, institutions, access to ports, government savings, tropics, ratio of coastline distance to land area Table 3.1 (Cont): The Relationship between Health and Economic Performance

Table 3.1 (Cont) Bloom and Sachs Life expectancy 0.037 (0.011) 25 year cross section, OLS Log(secondary schooling), openness, institutional (1998) 1965 - 90, n = 65 quality, central government deficit, percentage area in tropics, log coastal population density, log inland population density, total population growth rate, working-age population growth rate, Africa dummy

Bloom and Life expectancy -0.001 (0.032) 25 year panel at GMM (Arellano- Log(years of secondary schooling) openness, Williamson 5 year intervals Bond method) institutional quality, average government savings, (1998) 1960 - 85, n - 91 savings rate, population growth rate, working-age population growth rate, natural resource abundance, access to port, tropics dummy, ratio of coastline to land area

Bloom et al. (1999) Life expectancy 0.019 (0.012) 25 year cross section, 2SLS Log(years of secondary schooling) openness,

90 1965 - 90, n = 80 institutional quality, population growth rate,

working-age population growht rate, log of ratio of total population to working-age population, tropics

0 9

Hamoudi and Life expectancy 0.072 (0.020) 15 year cross section, OLS Openness, institutional quality, net government Sachs (1999) 1980 - 95, n = 78 savings, tropics area, log(coastal population density), population growth rate, working-age population growth rate, Afrlica dummy

Bloom, Canning Life expectancy 0.063 (0.016) 25 year panel at Pooled OLS Log(years of secondary schooling) openness, and Malaney 5 year intervals institutional quality, population growth, working-age (2000) 1965 - 90, n = 391 population growth, log(ratio of working-age to total population), population density, period dummies, GDP per worker, tropics, land locked

Table 3.1 (Cont) Gallup and Sachs Life expectancy 0.030 (0.009) 25 year cross section, OLS Years of secondary schooling, openness, public (2000) 1965 - 90, n = 75 institutional quality, population within 100 kilometers of the coast, malaria index in 1966, change in malaria index from 1966 to 1994, tropical area, land locked

Bhargava, Jamison Adult survival 0.358 (0.114) 25 year panel at Dynamic random Openness, log(fertility) tropics, log(I/GDP) Lau, and Murray rate (ASR) -0.048 (0.016) 5 year intervals, effects (2001) ASR x log( GDP - 1965 - 90, n = 92 per capita)

Jamison, Lau and log(ASR) 0.50 (0.23) 25 year panel at Maximum Average years of schooling for men, log(average Wang (2005) 5 year intervals, Liklihood capital stock per worker), openness, years elapsed, 1965 - 90, n = 53 tropical area, coastal area, fertility rate

Bloom, Cannings Life expectancy 0.040 (0.019) 30 year panel at Nonlinear 2SLS Average years schooling, size of economically

1 9 and Sevilla 10 year intervals active population, governance, capital stock, (2004) 1960 - 90, n = 104 percentage of land area in tropics, Technological catch up, average population work experience, work experience squared

Bloom and ASR 0.024 (0.007) 35 year panel at Nonlinear least Average years schooling, openness, institutional Canning (2005) squares quality, percentage of land within 100 kilometers 5 year intervals, of the coast, percentage of land area in the tropics, 1960 - 95, n = 416 technological catch up, size of labor force, and capital stock

Note—.ASR: adult survival rate; GDP: gross domestic product; GMM: generalized method of moments; OLS: ordinary least squares: 2SLS: two-stage least squares; 3SLS: three stage least squares; SUR: seemingly unrelated regression

Country Year N Height Ages Reference Argentina 1992-1993 68212 170.90 18 MSAS Australia 1995 2048 176.30 25-44 Australian Bureau of Statistics Austria* 1994-2001 3578 178.96 23-40 Garcia et al., Belgium* 1994-2001 2502 178.56 23-40 Garcia et al., Brazil 1989 1,104 169.60 22 Monteiro et al. Canada 1995 Hundreds 178.00 25-44 Progress in Prevention Chile 2001-2003 558 168.20 22-28 Rona et al. China 2002 Thousands 170.20 17 Yang Colombia 2002 10,000+ 170.64 18-22 Meisel et al., Cyprus 1999-2000 200 174.55 16-17 Savva et al., Denmark* 1994-2001 2530 181.58 23-40 Garcia et al., Egypt# 2004 500 171.43 19-21 El-Meligy et al. Finland* 1994-2001 2625 178.73 23-40 Garcia et al., France* 1996 Thousands 176.60 25-34 Reif et al,. Ghana 1987-89 3414 169.46 25-29 Schultz Greece 1994-2001 5089 177.53 23-40 Garcia et al., Iceland 1983-1987 303 180.24 18-20 Dagbjartsson et al., India 2005-2006 10,000+ 164.50 20 Deaton Indonesia 2000 3924 162.00 26-47 Maccini et al., Iran 2005 Thousands 172.52 24-35 Haghdoost et al., Ireland* 1994-2001 2837 176.73 23-40 Garcia et al., Israel 1980-2000 19747 175.60 19-24 Mandel et al., Italy* 1994-2001 10290 175.20 23-40 Garcia et al., Japan 2002 6805 171.97 20-40 Physical Fitness Study Jordan 1997 Hundreds 169.60 17.5 Hasan et al Korea 2005 Thousands 174.40 18-20 Kim et al., Malaysia 1996 6626 166.08 20-39 Lim et al., Mexico 2000 1444 168.05 17-18 Navarro et al., Netherlands* 1996 Thousands 182.20 25-34 Reif et al,.

2 9 New Zealand 1996/97 Hundreds 177.40 19-24 Russel et al., Norway 1992-94 199 179.60 20-39 Bolstad et al., Peru 1975 119 161.40 20-40 Peru National Insitution of Health Philippines 2003 Hundreds 163.49 20-39 Philippine Facts and Figures 2003 Portugal* 1994-2001 6610 171.13 23-40 Garcia et al., Singapore 1995 832 168.50 ??? South Africa DemographicSingh et al. and Health South Africa 1998 5566 169.00 25-34 Survey Spain* 1994-2001 8173 174.79 23-40 Garcia et al., Sweden* 1994-2001 2631 180.69 23-40 Garcia et al., Switzerland 1996 1097 178.00 18-27 Miedinger et al. Thailand*# 1991-1995 31320 167.50 20-50 Boonchai Somboonsook Tunisia*# 2000's 451 174.30 16-17 Ghannem et al., Turkey 2004-2006 5128 175.05 18-39 Ozer UK 1996 5579 176.14 25-44 National Health Survey England USA* 1999-2002 1441 176.55 20-40 Ogden et al., Zimbabwe 1998-1999 459 173.39 18-30 Siffert et al. Table 3.2: Average Adult Height among Prime-Age Men

Note—.*: Self-Reported; #: Student

By quartiles of GDP per capita in 2000 Complete Poorest Wealthiest Sample Quartile Quartile Log average annual growth rate of GDP per capita (PPP) 0.027 0.022 0.030 from 1960 to 2000 (0.012) (0.01) (0.013)

GDP per capita in 1960 (1000's) 5.270 1.73 8.700 (3.740) (0.84) (3.560) Adult height of prime-age men (centimeters) 173.30 168.580 177.36 (5.28) (4.52) (3.91) Instruments Insolation 4.22 20.51 3.34 (1.02) (12.24) (0.80) Skin Color 0.29 0.16 0.41 (0.18) (0.12) (0.19) Insolation * Skin Color 1.07 0.80 1.27 (0.56) (0.59) (0.51) Infant Mortality in 1960 71.47 133.25 23.42 (52.53) (30.81) (6.11) Geography Latitude 34.20 20.51 45.55 (16.72) (12.24) (17.27) Percentage of population residing in tropical climates 0.11 0.29 0.00 (0.24) (0.35) (0.00) Institutional Quality Voice and Accountability 0.56 -0.57 1.30 (0.89) (0.51) (0.42) Political Instability and Violence 0.31 -0.63 1.17 (0.87) (0.55) (0.15) Government Effectiveness 0.92 -0.16 1.83 (0.90) (0.28) (0.29) Regulatory Quality 0.76 -0.10 1.39 (0.72) (0.61) (0.32) Rule of Law 0.78 -0.32 1.68 3 9 (0.92) (0.43) (0.19) Control of Corruption 0.86 -0.37 1.92 (1.05) (0.27) (0.38) Total Health Expenditures as a % of GDP per capita (THE) 6.93 5.59 8.25 (2.32) (2.21) (2.27) Government Health Expenditures as a percentage of THE 59.16 46.45 68.03 (16.15) (11.53) (14.65) Physical Capability Depth of Hunger 156.67 216.67 121.67 (58.89) (56.14) (13.37) Child Mortality Rate in 2002 16.05 30.44 7.24 (17.13) (22.07) (2.83) Human Capital Cognitive ability 4.54 4.00 5.07 (0.59) (0.45) (0.15) Years of Schooling in 1960 5.53 2.34 8.40 (2.93) (1.36) (1.89) Number of Observations 45 12 12 Table 3.3: Descriptive Statistics across Countries

Note—.Standard deviations are in parentheses. Higher skin color values indicate lighter complexions. The poorest quartile contains countries whose real gdp per capita in 2000 was under 6777 dollars, whereas the wealthiest quartile contains countries whose real gdp per capita in 2000 exceeded 23792 dollars.

Dependent Variable : Adult Height (cm) (1) Insolation -4.37*** (1.54) Skin Color -29.35* (15.63) Skin Color * Insolation 7.09** (3.43) Infant Mortality Per 1000 Births (1960) -0.036* (0.019) Latitude 0.087 (0.09) Tropical Population (%) 5.11 (3.21) GDP per capita in 1960 0.19 (0.24) N 45 F-Test 4.34 Adjusted R2 0.80 Table 3.4: The Relationship between Insolation and Skin Color to Adult Height

Note—.***: 3tatistically 3ignificant at the 1 percent level; **: 3tatistically significant at the 5

4 9 percent level; *: 3tatistically 3ignificant at the 10 percent level. Standard deviations are presented in parentheses. Insolation equals kilowatt-hours per square meter per day. Skin color estimates represent the extent to which skin reflects light, and thus higher values indicate lighter complexions. The regression results control for gdp per capita in 1960, institutional quality, geographic region, and whether the height measures were self-reported, collected from students, and males ages 16 or 17.

OLS 2SLS Dependent Variable : Log economic growth in 1960 to 2000 (%) (1) (2) (3) (4) (5) (6) (7) (8) Adult Height among prime-age men (Centimeters) 0.120*** 0.058 0.045 0.025 0.126** 0.048 -0.005 -0.014 (0.043) (0.056) (0.050) (0.057) (0.060) (0.065) (0.076) (0.068) Physical Capacity Proxies: Depth of Hunger 2000 X X X X Incidence of Communicable Diseases (%) X X X X

Human Capital Proxies: Cognitive Test Scores X X X X Years of Schooling 1960 X X X X N 45 45 45 45 45 45 45 45 Adjusted R2 0.76 0.77 0.79 0.79 0.86 0.88 0.88 0.89

95 Table 3.5: The Relationship between Economic Growth, Health, Physical Capacity and Human Capital

5 9 Note—.***: statistically significant at the 1 percent level; **: statistically significant at the 5 percent level; *: statistically significant at the 10 percent level. Standard deviations are presented in parentheses. All regression results include measures of institutional quality with respect to governance and health care, real GDP per capita in 1960, climate, latitude, geographic region/continent, whether the height measures were self-reported, taken from students or taken from young men. The coefficients represent basis points.

Dependent Variable : Real GDP per capita in 2000 (1) (2) (3) (4) Adult Height among prime-age men (Centimeters) 0.077** 0.076** 0.036 0.038 (0.033) (0.036) (0.043) (0.039) Physical Capacity Proxies: Depth of Hunger 2000 X X Incidence of Communicable Diseases (%) X X

Human Capital Proxies: Cognitive Test Scores X X Years of Schooling 1960 X X N 45 45 45 45 Adjusted R2 0.91 0.92 0.92 0.93

96 Table 3.6: The Relationship between Economic Levels, Health, Physical Capacity, and Human Capital (2SLS)

6 9

References

Health Statistics: Key Data on Health 2002. European Commission, 2002.

Health Survey for England 2006 Latest Trends. The NHS Information Centre, 2008.

How Australians Measure Up. Australian Bureau of Statistics, 1995.

National Survey of Cultural, Socioeconomic, Biochemical, and Nutritious Indicators Related to the Degenerative Chronic Illnesses. Instituto Nacional de Salud, 2006.

Philippine Facts and Figures 2003. Food and Nutrition Research Institute, 2003.

"South Africa Demographic and Health Survey." Department of Health National Health Information Systems of South Africa, 1998.

"Statistical Yearbook 2007." Statistics Norway, 2007.

7 9 (GIMR), Garvan Institute of Medical Research. "Acutely Ill Patients have Low Vitamin D Levels." Science Daily 2009.

Abounaja, S., and W.H. Gilmour. "Standards for Height and Weight of Libyan Children Aged 6 to 17 years." Arab Journal of Medicine 4 (1985): 13-18.

Acemoglu, Daron, Simon Johnson, and James A. Robinson. "Reversal of Fortune: Geography and Institutions in the Making of the Modern World Income Distribution." Quarterly Journal of Economics 117 (2002): 1231-94.

Adams, R. "Social Psychology and Beauty: Effects of Age, Height, and Weight on Self-Reported Personality Traits and Social Behavior." Journal of Social Psychology 112 (1980): 287-93.

A'Hearn, Brian, Jörg Baten, and Dorothee Crayen. "Quantifying Quantitative Literacy: Age Heaping and the History of Human Capital." The Journal of Economic History 69 (2009): 783-808.

Albers, Ronald. "Human Capital and Economic Growth: Operationalising Growth Theory, with Special Reference to the Netherlands in the 19th Century." Research Memorandum: University of Groningen, 1997.

Albrecht, James, and Susan Vroman. "A Matching Model with Endogenous Skill Requirements." International Economic Review 43.1 (2002): 283-305.

Al-Hazzaa, H.M. "Anthropometric Measurements of Saudi Boys Aged 6-14 years." Annals of Human Biology 17 (1990): 33-40.

Allen, Robert C. "Progress and Poverty in Early Modern Europe." Explorations in Economic History 56.3 (2003): 403-43.

Alnwick, D. "The Weight, Length and Mid-Upper Arm Circumference of Kenyan Children in Nairobi Nursery Schools (UNICEF)." Social Statistics Bulletin 3.1 (1980): 1-19.

Anderson, N.E., et al. "Hojde og Vaget hos Danske Born." Ugeskrift for Laeger 144 (1982): 1760-65.

Arechiga, J., and C. Serrano. "Parametros Antropometricos de Crecimiento en un Grupo Indigena Mayance (Tojolabal)." Estudious de Cultura Maya 13 (1981): 307-37.

8 9 Ashcroft, M.T. , and H.A. Lovell. "Heights and Weights of Jamaican Children of Various Racial Origins." Tropical and Geographical Medicine 4 (1964): 346-53.

Attalah, N.L. "Adult Medical Students at Alexandria." South Asian Anthropologist.470-472 (1987).

Bagenholm, G., B. Kristiansson, and A.A.A. Nasher "Growth and Malnutrition Among Preschool Children in Democratic Yemen." Bulletin of the World Health Organization 66 (1988): 491-98.

Baker, Eldon, and Charles Redding. "The Effects of Perceived Tallness in Persuasive Speaking: An Experiment." Journal of Communication 12 (1962): 51-53.

Banegas, J.R., et al. "Month of Birth and Height of Spanish Middle-Aged Men." Annals of Human Biology 28.1 (2001): 15-20.

Barro, R. Health and Economic Growth, Mimeo. Cambridge: Harvard University, 1996.

Barro, R., and J. Lee. "Sources of Economic Growth." Carnegie-Rochester Conference Series on Public Policy 40, 1994. 1-46.

Barro, R., and X. Sala-I-Martin. Economic Growth. New York: McGraw-Hill, 1995.

Barro, Robert J. "Human Capital: Growth, History, and Policy-A Session to Honor Stanley Engerman." The American Economic Review 91.2 (2001): 12-17.

Barro, Robert J., and Xavier Sala-i-Martin. "Technological Diffusion, Convergence, and Growth " Journal of Economic Growth 2.1 (1997): 1- 26.

Bayoumi, A., and M.A.A. Moussa. "Kuwait Nutritional Survey: Comparisons of the Nutritional Status of Kuwaiti Children Aged 0-5 years with the NCHS/CDC Reference Population." Bulletin of the World Health Organization 63 (1985): 521-26.

Bechara, Antoine. "Decision Making, Impulse Control and Loss of Willpower to Resist Drugs: A Neurocognitive Perspective." Nature Neuroscience 9.11 (2005): 1458-63.

Becker, Gary Stanley. Human Capital: A Theoretical and Empirical Analysis with

9 9 Special Reference to Education. Chicago: University of Chicago Press, 1993.

Behrman, Jere R., Harold Alderman, and John Hoddinott. "Hunger and Malnutrition." Copenhagen Consensus – Challenges and Opportunities. Copenhagen, 2004.

Betz, Frank. "Returns to Sociability. Unpublished Manuscript." 2008.

Bhargava, A., et al. "Modeling the Effects of Health on Economic Growth." Journal of Health Economics 20.3 (2001): 423-40.

Bilicki, T., et al. "Changing Patterns of Social Variation in Stature in Poland: Effects of Transition from a Command Economy to the Free-Market System?" J. Biosoc. Sci. 37 (2005): 427-34.

Billewicz, W.Z., and I.A. McGregor. "A Birth-to-Maturity Longitudinal Study of Heights and Weights in Two African (Gambian) Villages, 1951-1975." Annals of HUman Biology 9 (1982): 309-20.

Bils, Mark, and Peter J. Klenow. "Does Schooling Cause Growth or the Other Way Around?" American Economic Review 90 (2000): 1169-83.

Bindon, J.R., and P.T. Baker. "Modernization, Migration, and Obesity among Samoan Adults." Annals of Human Biology 12 (1985): 67-76.

Bindon, J.R., and S.M. Zansky. "Growth Patterns of Height and Weight Among the Three Groups of Samoan Preadolescents." Annals of Human Biology 13 (1986): 171-78.

Birkbeck, J.A., and A.M. Guthrie. "Anthropometric Studies on Dunedin Five and Six Year Old Children." New Zealand Medical Journal 659 (1980): 331- 34.

Blaha, P. "Anthropometric Studies of the Czechoslovak Population: from 6 to 55 years." Czechoslovak Spartakiade 2 (1986).

Blair, R. "The Roles of Orbital Frontal Cortex in the Modulation of Antisocial Behavior." Brain and Cognition 55 (2004): 198-208.

Bloom, D., et al. "Population Chance and Human Development in Latin America." Inter-American Development Bank 1999.

Bloom, D., and J. Sachs. "Geography, Demography, and Economic Growth in

0 0 1 Africa." Brookings Papers on Economic Activity 2 (1998): 207-73.

Bloom, D. E., and D. Canning. "Health and Economic Growth: Reconciling the Micro and Macro Evidence." 2005.

Bloom, D. E., D. Canning, and P. N. Malaney. "Demographic Change and Economic Growth in Asia." Population and Development Review 26 (2000): 257-90.

Bloom, David E., David Canning, and Jaypee Sevilla. "The Effect of Health on Economic Growth: A Production Function Approach." World Development 32.1 (2004): 1-13.

Bloom, D. E., and P. Malaney. "Macroeconomic Consequences of the Russian Mortality Crisis." World Development 26 (1998): 2073-85.

Bloom, D. E., and J. G. Williamson. "Demographic Transitions and Economic Miracles in Emerging Asia." World Bank Economic Review 12.3 (1998): 419-55.

Blue, Laura. "Better Bedside Manners." Time 2007. 100

Bogin, B.A., and R.B. MacVean. "Growth in Height and Weight of Urban Guatemalan Primary School Children of Low and High Socioeconomic Status." Human Biology 50 (1978): 477-88.

Bolstad, Greta, Bibba Benum, and Aase Rokne. "Anthropometry of Norwegian Light Industry and Office Workers." Applied Ergonomics 32 (2001): 239- 46.

Boulton, T.J.C., and S. Weinstein. "A Comparative Anthropometric Study of Cook Islands and Australian Preschool Children." New Zealand Medical Journal 88 (1978): 489-90.

Bozzoli, Carlos, Angus Deaton, and Climent Quintana-Domeque. "Adult Height and Childhood Disease." Demography 46.4 (2009): 647-69.

Bradley, Robert, and Robert Corwyn. "Socioeconomic Status and Child Development." Annual Review of Psychology 53 (2002): 371-99.

Brooks-Gunn, Jeanne, and Greg Duncan. "The Effects of Poverty on Children." Children and Poverty 2 (1997): 55-71.

Brown, Robert Craig, and David Clark MacKenzie. Canada and the First World War: Essays in Honour of Robert Craig Brown. University of Toronto

1 0 1 Press, 2005.

Brown, T., and G.C. Townsend. "Adolescent Growth in Height of Australian Aboriginals Analyzed by the Preece-Baines Function: a Longitudinal Study, 1961-1971." Annals of Human Biology 9 (1982): 495-506.

Brundtland, G.H., K. Liestol, and L.W. Walloe. "Height, Weight, and Menarcheal Age of Oslo School Children during the last 60 years." Annals of Human Biology 7.307-322 (1980).

Cain, Pamela S., and Donald J. Tremain. "The Dictionary of Occupational Titles as a Source of Occupational Data." American Sociological Review 46.3 (1981): 253-78.

Capucci, E. "II. Antropometria dei Soggetti in Accresimento." Ricerche Antropologiche in una Comunita del Nicaragua. Ed. G.F. De Stefano. Vol. 64, 1986. 185-98.

Capucci, E., et al. "Statura e Peso in un Campione Della Popolazione Italiana in Accrescimento (4-15 Anni)." Rivista di Antropologia 62 (1982-83): 255- 71. 101

Card, David. "Estimating the Return to Schooling: Progress on Some Persistent Econometric Problems." Econometrica 69.5 (2001): 1127-60.

Carson, Scott Alan. "Geography, Insolation, and Vitamin D in Nineteenth Century US African-American and White Statures." Explorations in Economic Historu 46.1 (2009): 149-59.

Case, Anne, and Christina Paxson. "Stature and Status: Height, Ability, and Labor Market Outcomes." Journal of Political Economy 116.3 (2008): 499-532.

Caselli, F., G. Esquivel, and F. Lefort. "Reopening the Convergence Debate: A New Look at Cross Country Growth Empirics." Journal of Economic Growth 1 (1996): 363-89.

Cavalli-Sforza, L.L. African Pygmies. New York: Academic Press, 1986.

Centers for Disease Control. "2000 CDC Growth Charts for the United States: Methods and Development." Vital and Health Statistics 11.246 (2002).

Chase, H. P., and H. P. Martin. "Undernutrition and Child Development." New England Journal of Medicine 282 (1970): 933-39.

Clegg, E.J. "The Growth of Melanesian and Indian Children in Fiji." Annals of

2 0 1 Human Biology 16 (1989): 507-28.

Cohen, Daniel, and Marcelo Soto. "Growth and Human Capital: Good Data, Good Results." Journal of Economic Growth 12 (2007): 51-76.

Coodin, F.J., et al. "Growth and Nutrition of Manitoba Preschool Indian Children. III. Anthropometry." Human Biology 52 (1980): 563-78.

Corlett, J.T. "Growth of Urban School Children in Botswana." Annals of Human Biology 13 (1986): 73-82.

Cranfield, John, Kris Inwood, and Asher Kirk-Elleker. "A Longitudinal Perspective on the Physical Well-Being of Canadians born before 1901." 2007.

Crayen, Dorothee, and Joerg Baten. "New Evidence and New Methods to Measure Human Capital Inequality before and during the Industrial Revolution: France and the US in the Seventeenth to Nineteenth Centuries." The Economic History Review 63.2 (2010): 452-78.

102

Cresta, M., and N. Avoundogra. "Risultati Della Studio Longitudinale Dalla Nascita a 5 Anni in un Gruppo di Bambini di Porto Novo (Repubblica Populare del Benin)." Rivista di Antropologia 61 (1980-81): 41-132.

Dacou-Voutetakis, C., et al. "Growth Charts."

Dagbjartsson A, Þórsson , and Arnórsson Pálsson GI. "Height and Weight of Icelandic Children 6-20 Years of Age." LÆKNABLAÐIÐ 86 (2000): 509- 14.

Danker, H., P. Thiel, and G. Tsiakalos. "Untersuchungen zur Bevolkerungsbiologie Norddeutschlands." Anthropologischer Anzeiger 39 (1981): 82-88.

De Stefano, G.F., and J.M. Jenkins. "Richerche de Antropologia Biologia su Popolazioni Nicaraguensi." Rivista di Antropologia 58 (1972): 289-302.

Deaton, Angus. "Height, Health, and Inequality: The Distribution of Adult Heights in India." American Economic Review 98.2 (2008): 468-74.

Del-Rio-Navarro, BE, et al. "Mexican Anthropometric Percentiles for Ages 10– 18." European Journal of Clinical Nutrition 61 (2007): 12.

Deluca, Hector F., and Margherita T. Cantorna. "Vitamin D: its Role and Uses in

3 0 1 Immunology." The FASEB Journal 15 (2001): 2579-85.

Demirjian, A., and M. Brault-Dubuc. Croissance et Developpement de l'Enfant Quebecois de la Naissance a Six Ans. Montreal: Les Presses de l'Universite de Montreal, 1985.

Denton, Frank T., and Peter J. George. "Socio-Economic Influences on School Attendance: A Study of a Canadian County in 1871." History of Education Quarterly 14.2 (1974): 223-32.

Deurenberg, Paul, et al. "Prediction of Percentage Body Fat from Anthropometry and Bioelectrical Impedance in Singaporean and Beijing Chinese." Asia Pacific Journal of Clinical Nutrition 9.2 (2000): 5.

Dietz, W.H., et al. "Nutritional Status of Efe Pygmies and Lese Horticulturists." American Journal of Physical Anthropology 78 (1989): 509-18.

DuPont, Robert, et al. "Economic Costs of Anxiety Disorders." Anxiety 2 (1996): 167-72.

103

Easterly, William, and Ross Levine. "Tropics, Germs, and Crops: How Endowments Influence Economic Development " Journal of Monetary Economics 50.1 (2003): 3-39.

Eaton, Jonathan, and Samuel Kortum. "Trade in Ideas Patenting and Productivity in the OECD." Journal of International Economics 40.3-4 (1996): 251-78.

Eiben, O.G. "The Kormend Growth Study: Body Measurements." Anthropologai Kozlemenyek 26 (1982): 181-210.

Eiben, O.G., and E. Panto. "The Hungarian National Growth Standards." Anthropologai Kozlemenyek 30 (1986): 1-40.

El-Meligy, Manal M.S., et al. "Estimation of Human Body Built in Egyptians." Forensic Science International 159 (2006): 27-31.

El-Nofely, A.A. "Anthropometric Study of Growth of Egyptian Nubian Children." Human Biology 50 (1978): 183-208.

Essen, Marina Rode von, et al. "Vitamin D Controls T Cell Antigen Receptor Signaling and Activation of Human T Cells." Nature Immunology 2010.11 (2010).

Eveleth, P. Bowers, E.J., and J.I. Schall. "Secular Change in Growth in

4 0 1 Philadelphia Black Adolescents." Human Biology 51 (1979): 213-28.

Eveleth, Phyllis B., and James Mourilyan Tanner. Worldwide Variation in Human Growth. Cambridge University Press, 1990.

Fang, Yue, et al. "Vitamin D Receptor Gene Haplotype Is Associated with Body Height and Bone Size." THe Journal of Clinical Endocrinology & Metabolism 92.4 (2007): 1491-501.

Farquharson, S.M. "Growth Patterns and Nutrition in Nepali Children." Archives of Diseases in Childhood 51 (1976): 3-12.

Faulhaber, J. Crecimento: Somatometria de la Adolescencia. Mexico City: Universidad Nacional Autonoma de Mexico, 1989.

---. Investigacion Longitudinal del Crecimiento 26. Nacional de Antropologia e Historia, Collecion Cientifica. Cordoba: Antropologia Fisca, 1976.

104

Ferrie, Joseph. "A New Sample of Males Linked from the Public Use Micro Sample of the 1850 U.S. Federal Census of Population to the 1860 U.S. Federal Census Manuscript Schedules." Historical Methods 29 (1996): 141-56.

Fogel, Robert W., Stanley L. Engerman, and James Trussell. "Exploring the Uses of Data on Height: The Analysis of Long-Term Trends in Nutrition, Labor Welfare, and Labor Productivity." Social Science History 6.4 (1982): 401- 21.

Foster, T.A., et al. "Anthropometric and Maturation Measurements of Children Ages 5 to 14 years in Biracial Community - Bogalusa Heart Study." American Journal of Clinical Nutrition 30 (1977): 582-91.

Fox, P.T. 1981/82 Survey of growth in [re-School Children.: London School of Hygiene and Tropical Medicine, 1984.

Fox, P.T. , M.D. Elston, and Waterlow J.C. Pre-school Child Survey. London, Her Majesty's Staionery Office: Dept of Health and Social Security 1981.

Friendlander, J.S. The Solomon a Long-term study of Health, Human Biology, and Culture Change Project. The Solomon Islands Project. Oxford: Clarendon Press, 1987.

5 0 1 Frisancho, A.R., G.A. Borkan, and E.K. Layman. "Pattern of Growth of Lowland and Highland Peruvian Quechua of Similar Genetic Composition." Human Biology 47 (1975): 233-43.

Frisancho, A.R., et al. "Nutritional Infleunce on Childhood Development and genetic Controls of Adolscent Growth of Quechuas and Mestizas from the Peruvian Lowlands." Americans Journal of Physical Anthropology 52 (1980): 367-75.

Fung, K.P., et al. "A Survey of Growth of Honk Kong Children " Hong Kong Journal of Paediatrics 2 (1985): 105-16.

Fuster, Joaquin. "The Prefrontal Cortex – An Update: Time Is of the Essence." Neuron 20 (2001): 319-33.

Gaffield, Chad, and Gerard Bouchard. "Literacy, Schooling, and Family Reproduction in Rural Ontario and Quebec." Historical Studies in Education 1 (1989): 201-18.

105

Galler, J. R., et al. "The Influence of Early Malnutrition and Subsequent Behavioral Development." Journal of the American Academy of Child Psychiatry 22 (1983): 16-22.

Gallery, J. R., et al. "The Influence of Early Malnutrition and Subsequent Behavioral Development." Journal of the American Academy of Child Psychiatry 22 (1983): 16-22.

Gallo, P.G., and M.F. Mestriner. "Growth of Children in Somalia." Human Biology 52 (1980): 547-62.

Gallup, J., and J. Sachs. "The Economic Burden of Malaria." Center for International Development. Cambridge, MA: Harvard University, 2000.

Gallup, John Luke, Jeffrey D. Sachs, and Andrew D. Mellinger. "Geography and Economic Development." International Regional Science Review 22.2 (1999): 179-232.

Galor, Oded. Unified Growth Theory. Princeton University Press, 2011.

Garcia, Jaume, and Climent Quintana-Domeque. The Evolution of Adult Height in Europe: A Brief Note: Department of Economics and Business, Universitat Pompeu Fabra, 2006.

6 0 1 Gelissen, John, and Paul M. de Graaf. "Personality, Social Background, and Occupational Career Success." Social Science Research 35 (2006): 702- 26.

Gerverm, W.J.M. Measurements of the Body Proportions in Children. Lordgraf: Drukkerrij Groenavelt B.V., 1988.

Ghannem, H., et al. "Clustering of Cardiovascular Risk Factors Among Obese Urban Schoolchildren in Sousse, Tunisia." Eastern Mediterranean Health Journal 9.1/2 (2003).

Gibson, Kathleen Rita, and Anne C. Petersen. Brain maturation and Cognitive Development: Comparative and Cross-Cultural Perspectives Foundations of Human Behavior. Transaction Publishers, 1991.

Goldberg, Lewis. "An Alternative ―Description of Personality‖: The Big-Five Factor Structure." Journal of Personalty and Social Psychology 56.9 (1990): 1216-29.

106

Goleman, Daniel. Working with Emotional Cognition. New York: Bantam, 1998.

Grantham-McGregor, Sally, et al. "Longitudinal Study of Growth and Development of Young Jamaican Children Recovering from Severe Protein-Energy Malnutrition." Development Medicine and Child Neurology 24 (1982): 321-31

Gunnell, David, et al. "Association of Insulin-like Growth Factor I and Insulin- like Growth Factor Binding Protein-3 with Intelligence Quotient among 8 to 9 year old Children in the Avon Longitudinal Study of Parents and Children." Pediatrics 116.5 (2005): e681-86.

Gyenis, Gyula, and Kálmán Joubert. "Socioeconomic Determinants of Anthropometric Trends among Hungarian Youth." Economics and Human Biology 2 (2004): 321–33.

Haddad, Lawrence J., and Howarth E. Bouis. "The Impact of Nutritional Status on Agricultural Productivity: Wage Evidence from the Philippines." Oxford Bulletin of Economics and Statistics 53.1 (1991): 45-68.

Haghdoost, A. A., A. Mirzazadeh, and S. Alikhani. "Secular Trend of Height Variations in Iranian Population Born between 1940 and 1984." Iranian Journal of Public Health 37.1 (2008): 1-7.

7 0 1 Hall, Robert E., and Charles I. Jones. "Why Do Some Countries Produce So Much More Output Per Worker Than Others?" The Quarterly Journal of Economics 114.1 (1999): 83-116.

Hall, Stephen. Size Matters: How Height Affects the Health, Happiness, and Success of Boys—And the Men they Become. New York: Houghton Mifflin Company, 2006.

Hamoudi, A., and J. Sachs. "Economic Consequences of Health Status: A Review of the Evidence." Harvard Center for International Development. Cambridge, MA: Harvard University, 1999.

Hanushek, Eric A., and Ludger Woessmann. "The Role of Cognitive Skills in Economic Development " Journal of Economic Literature 46.3 (2008): 607-68.

Harper, Barry. "Beauty, Stature and the Labour Market: A British Cohort Study." Oxford Bulletin of Economics and Statistics 62.s1 (2000): 771-800.

Hasan, MA, et al. "Growth Status of Jordanian schoolchildren in Military-Funded Schools." European Journal of Clinical Nutrition 55 (2001): 380-86. 107

Heckman, James, Jora Stixrud, and Sergio Urzua. "The Effects of Cognitive and Non-Cognitive Abilities on Labor Market Outcomes and Social Behavior." NBER Working Paper Series, 2006.

Heineck, Guido. "Does it Pay to be Nice? Personality and Earnings in the UK." LASER Discussion Paper Series, 2007.

---. "Up in the Skies? The Relationship between Body Height and Earnings in Germany." Labour 19.3 (2005): 469-89.

Holick, Michael F. "Vitamin D Deficiency." The New England Journal of Medicine 357 (2007): 266-81.

Hoppe, Camilla, et al. "Animal Protein Intake, Serum Insulin-like Growth Factor 1, and Growth in Healthy 2.5-y-old Danish Children." American Journal of Clinical Nutrition 80.2 (2004): 447-52.

Hübler, Olaf. "The Nonlinear Link Between Height and Wages in Germany, 1985–2004." Economics and Human Biology 7.2 (2009): 191-99.

Ingram, Beth F., and George R. Neumann. "The Returns to Skill." Labour Economics 13.1 (2006): 35-59.

8 0 1 Institute, Canadian Fitness and Lifestyle Research. "Body Mass Index." Progress in Prevention 15 (1997).

Inwood, Kris, and John Cranfield. "The Great Transformation: A Long-Run Perspective on Physical Well-Being in Canada." Economics and Human Biology 5.2 (2007): 204-28.

Jablonski, Nina G., and George Chaplin. "The Evolution of Human Skin Coloration." Journal of Human Evolution 39.1 (2000): 57-106.

Jamison, D. T., L. J. Lau, and J. Wang. "Health's Contribution to Economic Growth in an Environment of Partially Endogenous Technical Progress." Health and Economic Growth: Findings and Policy. Eds. Guillem Lopez Casasnovas, Berta Rivera and Luis Currais, 2005.

Jolliffe, Dean. "Skills, Schooling, and Household Income in Ghana." World Bank Economic Review 12.1 (1998): 81-104.

Judge, Timothy A., and Daniel M. Cable. "The Effect of Physical Height on Workplace Success and Income: Preliminary Test of a Theoretical Model." Journal of Applied Psychology 89.3 (2004): 428-41. 108

Judge, Timothy A., et al. "The Big Five Personality Traits, General Mental Ability, and Career Success Across the Life Span." Personnel Psychology 52.3 (1999): 621-52.

---. "The Big Five Personality Traits, General Mental Ability, and Career Success Across the Life Span." Personnel Psychology 52.3 (1999): 621-52.

Katz, Michael B., and Ian E. Davey. "Youth and Early Industrialization in a Canadian City." The American Journal of Sociology 84 Supplement: Turning Points: Historical and Sociological Essays on the Family (1978): S81-S119.

Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi. "Governance Matters VII: Aggregate and Individual Governance Indicators 1996-2007." Policy Research Working Paper Series: The World Bank, 2008.

Kerver, Jean, et al. "Dietary Predictors of the Insulin-like Growth Factor System in Adolescent Females: Results from the Dietary Intervention Study in Children (DISC)." The American Journal of Clinical Nutrition 91.3 (2010): 643-50.

Khadera, Yousef, et al. "High Prevalence of the Metabolic Syndrome among Northern Jordanians." Journal of Diabetes and Its Complications 21

9 0 1 (2007): 214-19.

Kiddie, Joy, et al. "Nutritional Status of Children with Attention Deficit Hyperactivity Disorder: A Pilot Study." International Journal of Pediatrics 2010 (2010): 1-7.

Kim, Ji-Yeong, et al. "Anthropometric Changes in Children and Adolescents From 1965 to 2005 in Korea." American Journal of Physical Anthropology 136 (2008): 230-36.

Klein, R. G. "Prognosis of Attention Deficit Disorder and its Management of Adolescence." Pediatrics Review 8 (1987): 216-22.

Krassa, Gerasimos E, et al. "Self-Reported Prevalence of Obesity Among 20,329 Adults from Large Territories of Greece, Serbia and Turkey." Hormones (2003): 49-54.

Kremer, Richard, et al. "Vitamin D Status and Its Relationship to Body Fat, Final Height, and Peak Bone Mass in Young Women." The Journal of Clinical Endocrinology & Metabolism 94.1 (2009): 67-73.

109

Leung, S. S. F., et al. "Secular Changes in Standing Height, Sitting Height and Sexual Maturation of Chinese—The Hong Kong Growth Study, 1993." Annals of Human Biology 23 (1996): 297-306.

Leymann, Heinz. "Mobbing and Psychological Terror at Workplaces." Violence and Victims 5.2 (1990): 119-26.

Lim, T., et al. "Distribution of Body Weight, Height and Body Mass Index in a National Sample of Malaysian Adults." Medical Journal Malaysia (2000): 108-28.

Liu, Jianghong, and Adrian Raine. "The Effect of Childhood Malnutrition on Externalizing Behavior." Current Opinion in Pediatrics 18.5 (2006): 565- 70.

Liu, Jianghong, et al. "Malnutrition at Age 3 Years and Lower Cognitive Ability at Age 11 Years." Archives of Pediatrics & Adolescent Medicine 157 (2003): 593-600.

Maccini, Sharon, and Dean Yang. "Under the Weather: Health, Schooling, and Economic Consequences of Early-Life Rainfall." NBER Working Paper: National Bureau of Economic Research, 2008.

Malcolm, L. A. "Ecological Factors Relating to Child Growth and Nutritional

0 1 1 Status." Nutrition and Malnutrition: Identification and measurement. Eds. A. F. Roche and F. Falkner. New York: Plenum Press, 1974.

Malina, Robert M., Claude Bouchard, and Oded Bar-Or. Growth, Maturation, and Physical Activity. Human Kinetics, 2004.

Mandel, Dror, et al. "Height-Related Changes in Body Mass Index: A Reappraisal." Journal of American College Nutrition 23.1 (2004): 51-54.

Mankiw, Gregory, David Romer, and David Weil. "A Contribution to the Empirics of Economic Growth." Quarterly Journal of Economics 107.2 (1992): 407-37.

Mannerfeldt, R., and A. Pettersson. Fatta Att Man kan Bli Trott. Skolerfarenheter och Skolprestationer Bland Elever Med Bjorkpollenallergi i Skolar 8. PRIM, 2004.

Martorell, Reynaldo, and Jean-Pierre Habicht. "Growth in Early Childhood in Developing Countries." Human Growth: A Comprehensive Treatise. Eds. F. Falkner and J. M. Tanner. New York: Plenum Press, 1986.

110

Meisel, Adolfo, and Margarita Vega. A Tropical Sucess Story: A Century of Improvements in the Biological Standard of Living, Colombia 1910-2002: Banco de la República, 2004.

Miedingera, David, et al. "Respiratory Symptoms and Smoking Behaviour in Swiss Conscripts." Swiss Medical Weekly 136 (2006): 4.

Ministry of Education, Culture, Sports, Science and Technology-Japan. "Physical Fitness Study". 2002. . Minkler, Lanse. Integrity and Agreement. University of Michigan Press, 2008.

Mitch, David. "The Role of Education and Skill in the British Industrial Revolution." The British Industrial Revolution: An Economic Perspective. Ed. Joel Mokyr: Westview Press, 1999.

Mobius, Markus M., and Tanya S. Rosenblat. "Why Beauty Matters." The American Economic Review 96.1 (2006): 222-35.

Monteiro, C.A., M.H. D'Aquino Benicio, and N. da Cruz Goveia. "Secular growth trends in Brazil over three decades." Annals of Human Biology 21.4 (1994): 381-90.

Morton, Desmond. When Your Number's Up: The Canadian Soldier in the First

1 1 1 World War. Vintage Canada, 1994.

MSAS, (Ministerio de Salud y Acción Social de la Nación). Evaluación del Estado de Nutrición de la Población de Varones de 18 Años. Argentina 1992-1993, 1999.

Mueller, Gerrit, and Erik Plug. "Estimating the Effect of Personality on Male and Female Earnings." Industrial & Labor Relations Review 60.1 (2006): 1- 22.

Nogueról, Luiz Paulo, Cláudio D. Shikida, and Leonardo M. Monasterio. "Seis Centímetros: Uma Analíse Antropométrica da Pof 2002-2003." Associação Nacional dos Centros de Pósgraduação em Economia, 2005.

Noland, Edward, and Edward Bakke. Workers Wanted. New York: Harper, 1977.

Ogden, Cynthia, et al. Mean Body Weight, Height, and Body Mass Index, United States 1960–2002: Center for Disease Control and Prevention, 2004.

111

Okumura, Tsunao, and Emiko Usui. "Do Parents' Social Skills Influence Their Children's Sociability?" IZA Discussion Papers: Institute for the Study of Labor (IZA), 2010.

Oppenheimer, Jack, and Harold L. Schwartz. "Molecular Basis of Thyroid Hormone-Dependent Brain Development." Endocrine Reviews 18 (1997): 462-75.

---. "Molecular Basis of Thyroid Hormone-Dependent Brain Development. Endocrine Reviews 18:462-475." Endocrine Reviews 18 (1997): 462-75.

Oreopoulos, Philip. "The Compelling Effects of Compulsory Schooling: Evidence from Canada." Canadian Journal of Economics 39.1 (2006): 22-52.

Organization, The World Health. Children: Reducing Mortality, 2009.

Özer, Basak Koca. "Secular Trend in Body Height and Weight of Turkish Adults." Anthropological Science 116.3 (2008): 191-99.

Parente, Stephen L., and Edward C. Prescott. "Monopoly Rights: A Barrier to Riches". 1996. Mimeo. University of Minnesota.

Patterson, Gerald, Barbara Debaryshe, and Elizabeth Ramsey. "A Developmental Perspective on Antisocial Behavior." American Psychologist 44 (1990):

2 1 1 329-35.

Paus, Tomáš, et al. "Structural Maturation of Neural Pathways in Children and Adolescents: In Vivo Study." American Association for the Advancement of Science 283.5409 (1999): 1908-11.

Persico, Nicola, Andrew Postlewaite, and Daniel Silverman. "The Effect of Adolescent Experience on Labor Market Outcomes: The Case of Height." Journal of Political Economy 112 (2004): 1019-53.

Pollard, Sidney. "Discipline in the Industrial Revolution." The Economic History Review 16.2 (1963): 254-71.

Pradhan, Menno, David Sahn, and Stephen Younger. "Decomposing World Health Inequality." Journal of Health Economics 22.2 (2003): 271-93.

Reidpath, D. D., and P. Allotey. "Infant Mortality Rate as an Indicator of Population Health." Journal of Epidemiology & Community Health 57 (2003): 344-46.

112

Reif, K., and E. Marlier. "Health Care Issues and Public Security." Eurobarameter 44.3 (1996).

Relethford, John H. "Hemispheric Difference in Human Skin Color." American Journal of Physical Anthropology 104 (1997): 449-57.

Richards, Marcus, et al. "Birthweight, Postnatal Growth and Cognitive Function in a National U.K. Birth Cohort." International Journal of Epidemiology 31.342-348 (2002).

Rona, Roberto J., et al. "Untreated Asthma, Final Height and Sitting Height/Leg Length Ratio in Chile." Respiratory Medicine 100 (2006): 911-17.

Ross, Jerry, and Kenneth Ferris. "Interpersonal Attraction and Organizational Outcomes: A Field Examination." Administrative Science Quarterly 26 (1981): 617-32.

Russell, David, Winsome Parnell, and Noela Wilson. NZ Nutrition: NZ People Key Results of the 1997 National Nutrition Survey: New Zealand Ministry of Health, 1999.

Sachs, J., and A. Warner. "Sources of Slow Growth in African Economies." Journal of African Economics 6 (1997): 335-37.

3 1 1 Samaras, Thomas T. "Should we be Concerned over Increasing Body Height and Weight?" Experimental Gerontology 44.1-2 (2009): 83-92.

Saul, Feldman. The Presentation of Shortness in Everyday Life—Height and Heightism in American Society: Toward a Sociology of Stature Paper. Chicago: American Sociological Association, 1971.

Savva, Savvas C., et al. "Reference Growth Curves for Cypriot Children 6 to 17 Years of Age." Obesity Research 9.12 (2001): 8.

Scheepens, Arjan, Tanja A. E. Möderscheim, and Peter D. Gluckman. "The Role of Growth Hormone in Neural Development." Hormone Research 64.s3 (2005): 66-72.

Schick, Andreas M., and Richard H. Steckel. "Height as a Proxy for Cognitive and Non-Cognitive Ability." NBER Working Paper Series, 2010.

Schultz, Paul. "Productive Beneﬁts of Health: Evidence from Low-Income Countries." Health and Economic Growth: Findings and Policy Implications. Eds. Guillem López i Casasnovas, Berta Rivera and Luis Currais: MIT Press, 2005. 113

Schultz, Paul T. "Productive Benefits of Improving Health: Evidence from Low- Income Countries." Working Papers: Economic Growth Center, Yale University, 2005.

Siffert, Winfried, et al. "Worldwide Ethnic Distribution of the G Protein b3 Subunit 825T Allele and Its Association with Obesity in Caucasian, Chinese, and Black African Individuals." J Am Soc Nephrol 10 (1999): 1921-30.

Silventoinen, Karri. "Determinants of Variation in Adult Body Height." J. Biosocial Sci 35 (2003): 263-85.

Singh, Jarnail, et al. "An Anthropometric Study of Singapore Candidate Aviators." Ergonomics 38.4 (1995): 651-58.

Smith, George Davey, et al. "Group Height and Risk of Death among Men and Women: Aetiological Implications of Associationswith Cardiorespiratory Disease and Cancer Mortality." Journal of Epidemiology and Community Health 54.2 (2000): 97-103.

Somboonsook, Boonchai. Obesity Among STOU Students, 1995.

Song, Yun-Mi, George Davey Smith, and Joohon Sung. "Adult Height and Cause-

4 1 1 specific Mortality: A Large Prospective Study of South Korean Men." American Journal of Epidemiology 158.5 (2003): 6.

Steckel, Richard H. "Census Matching and Migration: A Research Strategy." Historical Methods 21 (1988): 52-60.

---. "Stature and the Standard of Living." Journal of Economic Literature 33 (1995): 1903-40.

Stiefel, Matthew, Rocco Perla, and Bonnie Zell. "A Healthy Bottom Line: Healthy Life Expectancy as an Outcome Measure for Health Improvement Efforts." The Milbank Quarterly 88.1 (2010): 30-53.

Stoch, M. B., and P. M. Smythe. "Does Undernutrition during Infancy Inhibit Brain Growth and Subsequent Intellectual Development." New England Journal of Medicine 282 (1963): 933-39.

Stock, James H., and Motohiro Yogo. Testing for Weak Instruments in IV Regression. Identification and Inference for Econometric Models: A Festschrift in Honor of Thomas Rothenberg. Ed. Donald W. K. Andrews and James H. Stock: Cambridge University Press, 2005. 114

Stogdill, R. M. "Personal Factors Associated with Leadership: A Survey of the Literature." Journal of Psychology 23.3 (1948): 35-71.

Stokes, J., E. Fenstad, and T.B. Casale. "Managing Impairment in Patients with Allergic Rhinitis." Allergy Asthma Proc 27 (2006): 12-16.

Strauss, John, and Duncan Thomas. "Health, Nutrition, and Economic Development." Journal of Economic Literature 36.2 (1998): 766-817.

Su, Pen-hua, Shu-li Wang, and Jia-yuh Chen. "Estimating Final Height from Parental Heights and Sex in Taiwanese." Human Biology 79.3 (2007): 283-92.

Summers, Robert, Alan Heston, and Bettina Aten. Penn World Table Version 6.1. Center for International Comparisons at the University of Pennsylvania.

Tanner, James. Foetus into Man: Physical Growth from Conception to Maturity. Cambridge: Harvard U. Press, 1978.

---. "Foetus into Man: Physical Growth from Conception to Maturity." Cambridge Harvard University Press, 1978.

Thompson, Catherine, and Gregory B. Potter. "Thyroid Hormone Action in

5 1 1 Neural Development." Cerebral Cortex 10 (2000): 939-45.

Victora, Cesar G., et al. "Maternal and Child Undernutrition: Consequences for Adult Health and Human Capital." The Lancet 371.9609 (2008): 340-57.

Wada, K, et al. "Validity of Self-Reported Height and Weight in a Japanese Workplace Population." International Journal of Obesity 29 (2005): 1093- 99.

Wahab, Abdul, et al. National Nutrition Survey for Adult Bahrainis Aged 19 Years and Above: Ministry of Health for Kingdom of Bahrain, 2002.

Watson, R. S., and K. M. Etta. "Anthropometric Measurements and Amounts of Body Fat and Serum Cholesterol in Samples of Three Nigerian Populations." British Journal of Nutrition 33 (1975): 141-48.

Weil, David N. "Accounting for the Effect Of Health on Economic Growth." The Quarterly Journal of Economics 122.3 (2007): 1265-306.

Williams, Ivan, et al. "Relative Effects of Drugs and Diet on Hyperactive Behaviors: An Experimental Study." Pediatrics 61.6 (1978): 811-17. 115

Woessmann, Ludger. "Specifying Human Capital." Journal of Economic Surveys 17.3 (2003): 239-70.

Wong, Linda Y. "Can the Mortensen-Pissarides Model with Productivity Change Explain U.S. Wage Inequality." Journal of Labor Economics 21.1 (2003): 70-105.

Yang, X. G., et al. "Study on Weight and Height of the Chinese People and the Differences Between 1992 and 2002." Zhonghua Liu Xing Bing Xue Za Zhi 26.7 (2005): 489-93.

Zemel, B., and C. Jenkins. "Dietary Change and Adolescent Growth Among the Bundi (Gende-speaking) People of Papua New Guinea." American Journal of Human Biology 1 (1989): 709-18.

6 1 1

116

Appendix A

Appendix for Chapter 2

A.1 The Correlation between Occupational Aptitudes and Ability

I capture innate cognitive and non-cognitive ability using occupational

requirements. To substantiate their use as ability proxies, I must present evidence that

occupational requirements are strongly correlated with ability in the past. This analysis

requires historical cognitive and non-cognitive measures. Unfortunately, these measures

are very scarce, and thus my best alternative is to examine this relationship using

7 1 1 contemporary data. A useful data set is the National Childhood Development Study

(NCDS). The NCDS is a panel study that sampled every baby born in England, Scotland

and Wales during the week of March 3, 1958. The study originated as a perinatal survey,

but was eventually expanded to study other topics, such as cognitive and non-cognitive

development.

The NCDS provides cognitive and non-cognitive measures at ages 11, 16 and 33.

Individuals completed math and reading cognitive exams at age 11, and a problem

solving assessment at age 33. Parents and teachers evaluated the individual’s non-

cognitive abilities at ages 11 and 16. At age 11, teachers assessed their students’ non-

117

cognitive abilities using the Bristol Social Adjustment Guide (BSAG). The BSAG is an

emotional intelligence assessment that contains approximately 150 attitude/temperament

questions. Each question asks the student’s teacher to rate their student’s attitude towards

a particular social situation, such as their interest in interacting with classmates. Together,

the questions are summed together to quantitatively measure a student’s social

withdrawal, depression, anxiety, hostility, restlessness, and other unproductive social

characteristics. The individual’s parents and teacher used similar assessments to evaluate

the individual’s non-cognitive abilities at age 16.

The NCDS reports an individual’s occupational outcome at age 33 (in 1991). I

measure an occupation’s cognitive and non-cognitive requirements using the revised 4th

edition of the DOT (which was printed in 1991). This edition reports 53 occupational

characteristics, which are grouped into 5 categories: intellectual aptitudes, physical

8 1 1 aptitudes, temperaments, interests, and working conditions. I use a principal components

analysis to capture the cognitive and non-cognitive aptitudes common across

requirements. The analysis identified two cognitive abilities (general cognitive ability and

critical thinking/judgment) and non-cognitive abilities (general non-cognitive ability and

sociability).

Appendix table 2 presents evidence that more cogent individuals are more apt to

select into more cognitive-intensive occupations. Columns I and III regress an

individual’s general cognitive requirement z-score on their mathematics and reading z-

scores, respectively, while columns II and IV regress an individual’s critical thinking z-

score on their mathematics and reading z-scores, respectively. Every column controls for

the individual’s race, residence, household income in childhood, mother’s and father’s 118

academic achievement, socioeconomic status, and involvement in their child’s education.

A one standard deviation increase in an individual’s mathematics score is associated with

a 23 and 37 percent of a standard deviation increase in their critical thinking and

cognitive requirements, respectively. The effects are larger than a 5 standard deviation

increase in household income. Similar results are reported with reading scores.

Appendix table 3 presents evidence that more socially adept individuals are

substantially more apt to select into occupations requiring more non-cognitive ability.

Columns I and II regress an individual’s general non-cognitive requirement z-score and

sociability z-score on their BSAG z-score, respectively. Every column includes the same

controls used in appendix table 2. A one standard deviation increase in an individual’s

BSAG z-score is associated with a 9 and 12 percent of a standard deviation increase in

their non-cognitive and sociability requirements, respectively. The effects are larger than

9 1 1 a 3 standard deviation increase in household income.

The evidence indicates that occupational requirements are roughly correlated with

ability, which suggests that my analysis may substantially understate the extent to which

innate aptitude contributes to the stature premium. Appendix table 4 presents evidence

supporting this conjecture. The table reports the estimated stature coefficients obtained

via regressing adult hourly earnings on adult height.62 Column I includes age, race and

residence controls. Column II includes parental investment controls, while columns III

62 The NCDS collects earnings data at ages 33 and 42. I measure most individuals earnings using the measure reported at age 33. When this measure is missing, I use their earnings reported at age 42. The NCDS reports adult stature at age 33. Nurses measured each individual’s stature, accurate to one centimeter, using a tape measure or height scale. Consistent with previous studies, we converted our stature measures from centimeters to inches (Perisco et al. 2004; Case and Paxson, 2008). Hourly earnings were also reported at age 33. To increase the sample size, we also included individuals who only reported earnings at age 42. 119

and IV include cognitive and non-cognitive occupational requirement and innate ability

measures, respectively.

The results support the conjecture. Column I indicates that taller workers earn

substantially more than their shorter counterparts. A one inch increase in adult stature is

associated with a 2.2 percentage point increase in adult hourly earnings. Including only

occupational requirements, in addition to parental investment controls, reduces the stature

premium approximately 0.60 standard deviations, from 0.015 to 0.011. On the other

hand, including only innate ability scores, in addition to parental investment, reduces the

stature premium approximately two standard deviations, from 0.015 to 0.005, and renders

it statistically insignificant. Comparing the results in columns III and IV, occupational

requirements understate the extent to which ability contributes to the stature premium by

approximately 60 percent.

0 2 1 As an interesting note, occupational requirements provide an approximately

similar return to earnings in the past and present.63 A one standard deviation increase in

cognitive requirements is associated with a 0.39 and 0.40 percent of a standard deviation

increase in earnings in historical Canada and modern Britain, respectively. The results are

also approximately similar across non-cognitive requirements.

63 To make the results comparable, I converted the earnings and occupational requirements measures, across samples, into similar unit values.

120

A.2 Constructing a Historical Data Set

Linking Canadian Males to their Childhood and Military Service Records

I matched individuals using a variant of Ferrie’s (1996) iterative matching

procedure. The procedure steps are as follows:

1. Names were converted to a phonetic code using the Jaro-Winkler technique.64

This procedure assigns an identical phonetic code to similar sounding names, such as

Alexander and Alexandar, and thus accounts for minor spelling variations in names.

2. Individuals were matched using their available time constant information, such

as their country of birth, province of birth, given name, surname, ethnicity, and year of

birth (give or take two years). This step identified 16,291 individuals, of which 10,000

1 2 1 were randomly sampled.

3. I discarded approximately 6,654 records because they were blank, illegible,

missing, or misreported.

4. Multiple matches were manually compared using additional non-digitized time

constant information provided in each survey, such as the individual’s middle name,

spouse’s name, parent’s name and the date the individual migrated to Canada. Individuals

were discarded if:

i. Their wife’s name differed between the CEF and 1911 census.

ii. Their parent’s name differed between surveys.

64 This is the phonetic algorithm used by IPUMS. An advantage of this algorithm is that it calculates the distance between strings. 121

iii. Their middle name differed between surveys.

iv. The individual migrated to Canada after 1891.

5. Remaining multiple matches were distinguished using their month of birth.

Matches were rejected when an individual’s month of birth differed between the CEF and

1911 census. I could not compare several month clusters—March and May, April and

August, and June, July and January—because most pen scripts made these months

indiscernible. For example, March was commonly reported as Mar, which looks like

May.

6. Remaining ambiguous matches were identified using their day of birth. I

matched each individual to their 1901 census record using steps 3 to 5. Matches were

rejected when an individual’s day of birth differed between the CEF and 1901 census

(give or take two days). No multiple matches remained after this step.

2 2 1 7. Any remaining unidentified individuals were matched with individuals born

within five years of their birth year using steps 3 to 6. No unidentified individuals

remained after this step.

The procedure linked 1,570 of the remaining 3,358 individuals, resulting in an

overall linkage success rate—between the three surveys— of 47 percent. 158 of the 1,570

records were omitted because they contained missing parental investment measures.

Sample Representativeness

My method matches individuals using certain individual characteristics, which

implies that the resulting sample is unrepresentative. I examine the extent to which these 122

characteristics impair my sample’s representativeness by comparing the statistical

variation between these variables across the linked and unlinked samples. Appendix table

4 presents marginal frequency logistic regression results of matching success on each

matching characteristic; the dependent variable is equal to 1 if a successful match was

made and equal to zero otherwise.

The results indicate that occupation is weakly correlated with matching success.

Blue collar workers were matched more often than craftsmen and professionals, but the

variation is modest and statistically insignificant. Blue collar workers were matched

substantially more than agricultural workers. However, the low linkage rate among

agricultural workers is mostly attributed to their tendency to not report earnings data.65

Linkage was strongly associated with an individual’s province of birth.

Individuals born in Nova Scotia, New Brunswick and Manitoba were substantially more

3 2 1 likely to be matched than individuals born in Ontario. For example, individuals born in

New Brunswick have a 19.4 percent higher linkage success rate than those born in

Ontario. These results indicate that it is substantially easier to match individuals in more

rural provinces, where there are fewer individuals with similar characteristics.

The linkage success rate was also associated with marital status and whether the

individual’s name was common. Married individuals were 11.9 percent more likely to be

matched, on average. Similarly, uncommon named individuals had a 9 percent higher

linkage success rate.66 To test whether marital status and name frequency impact my

65 In words, I could match agricultural workers, but commonly excluded them from the sample because they did not report earnings data. 66 Consistent with the literature, a common name is a surname/given name combination that appeared more than 10 times in the 1911 census (Steckel, 1988; Ferrie, 1996). 123

analysis, I measure the extent to which these characteristics are associated with stature,

earnings, parental investment and occupational skill. The results indicate that marital

status and name frequency are weakly associated with these variables. Hence, the sample

selection biases associated with over matching married and uncommon named

individuals should not substantially affect this paper’s results.67

Linking Biases, Data Quality Issues, Limitations and Corrections

This section discusses the representativeness of the data. It describes the

following topics: one, the biases introduced by the linking procedure; two, how to correct

for the biases; three, the quality of the wage and height measures; finally, the general

limitations of the data set.

4 2 1

Name of Spouse & Name of Parent Bias

The sample may over represent married men or men that live with their parents in

1910 (independents) because I used parent and spouse information to evaluate most

matches. An OLS regression may estimate inconsistent coefficients if the characteristics

of either married men or dependents significantly differ from those of singles or

independents. A researcher can include marriage and independent indicators into their

analysis to control for each group’s respective unobserved differences, and thus correct

for these biases.

67 These results are available upon request. 124

Height Data Bias

Soldier stature data commonly suffers from truncation bias because most

militaries enforce minimum height requirements. According to Morton (1994), the CEF

did not have a minimum height requirement, but some officers preferred their soldiers to

be 63 inches or taller. This preference induces ―fuzzy‖ truncation, which can be corrected

using a maximum likelihood truncated regression (Fogel et al., 1982). However, the

height data do not appear to be significantly affected by ―fuzzy‖ truncation. Sixty-two

individuals are shorter than 63 inches, and a frequency distribution indicates that the

height data is approximately normal.

5 2 1 Wage Errors

Approximately 20 individuals report missing values for hours worked per week.

This problem is solved as follows

1. Most of these individuals lived in a district with a single employer, such as a

mine, mill or factory. In general, workers with the same occupation had roughly similar

annual earnings, weeks worked per year and hours worked per week. Therefore,

individuals with missing hours were assigned the hours reported by workers in the same

occupation.

2. Remaining workers were compared to workers employed in the same

occupation. If the worker reported the same annual earnings and weeks worked per year 125

as his peers, then they were assigned an average hours worked per week equivalent to his

peers. In all cases, this value was 60 hours.

6 2 1

126

Cognitive Skill Social Skill Fine Motor Skill Factor Factor Factor Variables Loadings Variables Loadings Variables Loadings Numerical 0.837 Social Interaction 0.664 Emotional Intelligence 0.488 Form Perception 0.759 Verbal 0.633 Machine Aptitude 0.471 Manual Dexterity 0.734 Motivation 0.591 Spatial 0.471 Instructing and Finger Dexterity 0.733 Management 0.535 Social Interaction 0.360 Motor Coordination 0.693 Intelligence 0.481 Dealing with Change 0.356 Preference towards Preference towards Social Tangible Outcomes 0.672 Interaction 0.478 Intelligence 0.350 Intelligence 0.669 Numerical 0.354 Motivation 0.339 Preference to deal with Clerical 0.628 things 0.340 Critical Thinking I 0.302 Preference towards Social Verbal 0.589 Persuasion 0.332 Interaction 0.258 Preference towards Color 0.558 Clerical 0.307 Science 0.257 Eye-Hand-Foot Preference to Improve Preference towards Coordination 0.533 Social Welfare 0.195 Communication 0.251 Spatial 0.485 Solitary 0.186 Verbal 0.238 Instructing and Preference towards Management 0.456 Communication 0.184 Critical Thinking II 0.232 Preference to Improve Solitary 0.424 Critical Thinking I 0.084 Social Welfare 0.228 Dealing with Change 0.387 Preference towards Science 0.068 Clerical 0.180 Emotional Intelligence 0.286 Critical Thinking II 0.039 Persuasion 0.177 Preference toward Machine Aptitude 0.279 Color -0.011 Abstract Thought 0.123 Critical Thinking II 0.180 Taking Instructions -0.037 Numerical 0.108 Preference toward Abstract Thought 0.164 Coping with Stress -0.065 Judgement 0.095 7 2 1 Preference towards Eye-Hand-Foot Science 0.140 Coordination -0.073 Form Perception 0.060 Preference toward Abstract Coping with Stress 0.115 Thought -0.079 Coping with Stress -0.012 Preference towards a Motivation 0.104 Judgement -0.088 routine -0.016 Preference towards a Judgement 0.094 routine -0.156 Dealing with Repitition -0.036 Preference to Improve Social Welfare 0.068 Dealing with Repitition -0.250 Manual Dexterity -0.038 Social Interaction 0.052 Form Perception -0.286 Taking Instructions -0.125 Critical Thinking I 0.035 Spatial -0.304 Finger Dexterity -0.228 Persuasion 0.016 Finger Dexterity -0.323 Motor Coordination -0.233 Preference towards Social Interaction 0.011 Motor Coordination -0.325 Color -0.397 Preference towards Preference towards Communication -0.067 Dealing with Change -0.356 Tangible Outcomes -0.421 Preference to deal with Preference towards Eye-Hand-Foot things -0.244 Tangible Outcomes -0.360 Coordination -0.515 Instructing and Taking Instructions -0.808 Manual Dexterity -0.403 Management -0.541 Preference to deal with Dealing with Repitition -0.847 Machine Aptitude -0.590 things -0.624 Preference towards a routine -0.852 Emotional Intelligence -0.640 Solitary -0.824 Table A.1: Factor Analysis on Second Edition Dictionary of Occupational Title Variables

127

Table A.1 (Cont)

8 2 1

Motivation 0.104 Judgement -0.088 Preference towards a routine -0.016 Judgement 0.094 Preference towards a routine -0.156 Dealing with Repitition -0.036 Preference to Improve Social Welfare 0.068 Dealing with Repitition -0.250 Manual Dexterity -0.038 Social Interaction 0.052 Form Perception -0.286 Taking Instructions -0.125 Critical Thinking I 0.035 Spatial -0.304 Finger Dexterity -0.228 Persuasion 0.016 Finger Dexterity -0.323 Motor Coordination -0.233 Preference towards Social Interaction 0.011 Motor Coordination -0.325 Color -0.397 Preference towards Preference towards Tangible Communication -0.067 Dealing with Change -0.356 Outcomes -0.421 Preference towards Tangible Eye-Hand-Foot Preference to deal with things -0.244 Outcomes -0.360 Coordination -0.515

Taking Instructions -0.808 Manual Dexterity -0.403 Instructing and Management -0.541 Preference to deal with Dealing with Repitition -0.847 Machine Aptitude -0.590 things -0.624 Preference towards a routine -0.852 Emotional Intelligence -0.640 Solitary -0.824

128

(1) (2) (3) (4) General Cognitive Requirement z-score Critical Thinking Requirement z-score Math Test z-score at age 11 0.37*** 0.23***

Reading Test z-score at age 11 0.35*** 0.22***

N 2,127 2,127 2,127 2,127 Adjusted R2 0.22 0.2 0.08 0.08 Table A.2: The Relationship between Cognitive Ability and Cognitive Occupational Skill

Note—.***: statistically significant at the 1 percent level; **: statistically significant at the 5 percent level; *: statistically significant at the 10 percent level. The above regression results control for individual age, race, region of residence, household income, father's socio-economic group during childhood, mother and father's academic attainment, mother and father's involvement in child's education.

129

9 2 1 (1) (2) Non-Cognitive Requirement z-score Communication Requirement z-score Bristol Social Adjustment z-score at age 11 0.09*** 0.12***

N 2,127 2,127 Adjusted R2 0.02 0.04 Table A.3: The Relationship between Non-Cognitive Ability and Non-Cognitive Occupational Skill

Dependent Variable: Log Hourly Earnings (1) (2) (3) (4) Adult Height (Inches) 0.022*** 0.015*** 0.011** 0.005 (0.006) (0.006) (0.005) (0.005) Controls Parental Investment Measures X X X Cognitive Test Scores at ages 11 and 33 X Non-Cognitive Test Scores at ages 11 and 16 X Cognitive Occupational Requirements X Non-Cognitive Occupational Requirements X N 1,233 1,233 1,233 1,233 Adjusted R2 0.030 0.07 0.20 0.19 Table A.4: Hourly Earnings, Height, Actual Cognitive and Non-Cognitive Ability, and Occupational Requirements in 130 Post-Industrial Britain

0 3 1

Note—.***: statistically significant at the 1 percent level; **: statistically significant at the 5 percent level; *: statistically significant at the 10 percent level. OLS regression coefficients presented with standard errors in parentheses. All regression results include the individual's experience, race and region of residence. Parental investment controls include separate measures of the father's socio-economic group at ages 7 and 16, household income at age 11, father's and mother's education attainment, and the parents involvement in the individual's adolescent education. Cognitive test scores include separate scores of the individuals math, reading and problem solving exam scores. Non-cognitive test scores include separate scores of the individual's neurotocism and extraversion assessments. Finally, the cognitive and non-cognitive occupational requirements include separate measures of the individual's cognitive, critical thinking, non-cognitive and sociability requirements.

Logistic Marginal Frequencies Dependent Variable : Individual was linked (1) Adult Height (Inches) -0.002 (0.004) Married 0.119*** (0.018) Common Name -0.092*** (0.024) Individual's Socio-Economic Group White Collar -0.045 (0.030) Skilled -0.011 (0.021) Farmer -0.118*** (0.030) Individual's Birth Residence Quebec 0.057** (0.022) Nova Scotia 0.261*** (0.040) New Brunswick 0.194*** (0.046) Manitoba 0.183*** (0.063) British Columbia 0.080

1 3 1 (0.111) Territories -0.159 (0.142) British -0.121*** (0.330) Other Foreign Nations -0.265 (0.121) N 3,358 Pseudo R2 0.04 Table A.5: The Relationship between Linking Success and Various Individual Characteristics

Note—.***: statistically significant at the 1 percent level; **: statistically significant at the 5 percent level; *: statistically significant at the 10 percent level. Multinomial logit coefficients are presented with standard deviation in parentheses. The omitted birth residence is Ontario.

131