School spirit: exploring the long-term effects of the U.S. temperance movement on educational attainment Andrew Francis-Tana Emory University National University of Singapore Cheryl Tanb Emory University Ruhan Zhangc University of Chicago This study examines the long-term consequences of the U.S. temperance movement, one of the largest-scale policy changes impacting alcohol consumption in history. Using a sample of persons born between 1900 and 1925 drawn from nationally representative census microdata, the study investigates the effect of prenatal temperance environment on adult educational outcomes. The quantitative analysis uses proxies for temperance environment: the strength of temperance laws and the density of liquor retailers. In sum, the evidence suggests that men and women exposed to the temperance movement in utero had modestly higher education. The effect on eighth grade completion was weakly significant at best, while the effect on high school completion was robustly significant, though relatively small in magnitude. Estimates imply that the adoption of temperance laws increased the odds of high school completion by about 3-8%, and analogously that the reduction of retail liquor density, to the extent it changed from 1913 to 1921, increased the odds of high school completion by about 4-8%. The findings are consistent with several causal mechanisms including a reduction in fetal alcohol exposure and the adverse developmental outcomes with which it is associated. Keywords: temperance movement, educational attainment, prenatal environment, alcohol policy, early twentieth century history, fetal alcohol spectrum disorders (FASD). JEL classifications: I12, I21, J18, K32.

a

Andrew Francis-Tan, PhD, Associate Professor of Economics at Emory University and Visiting Associate Professor at the Lee Kuan Yew School of Public Policy, National University of Singapore, 469B Bukit Timah Road, #02-02, Singapore 259771; [email protected]; +65 6601 1179 (corresponding author). b Cheryl Tan, MPH, Rollins School of Public Health, Emory University, Atlanta, GA. c Ruhan Zhang, MS, Department of Statistics, University of Chicago, Chicago, IL.

1

1. Introduction Temperance was a popular movement to reduce alcohol consumption in the U.S. It consisted of several waves spanning from the early 1800s to the 1930s (Kobler 1974, Rumbarger 1989, Murdock 1998, Okrent 2010). Proponents argued that alcohol had harmful effects on the health and morality of drinkers and society as a whole (Cherrington 1908, 1915, 1920). During the nineteenth century, temperance reformers achieved several legislative victories to regulate alcohol. However, by the 1850s, most of their gains quickly withered under popular and industry pressure (Rumbarger 1989, Murdock 1998). The last wave of temperance proved to be the most successful, by far. Founded in 1893, the Anti-Saloon League guided the dry movement to many victories during the early twentieth century (Kobler 1974, Murdock 1998, Okrent 2010). A number of states enacted either prohibitory laws, which banned the commercial sale of alcohol state-wide, or local option laws, which allowed localities (usually counties) to vote to restrict alcohol sale (Cherrington 1908, 1915, 1920). The movement's efforts culminated in the 18th amendment which prohibited "the manufacture, sale, or transportation of intoxicating liquors within, the importation thereof into, or the exportation thereof from the United States" (Kobler 1974, Rumbarger 1989, Murdock 1998, Okrent 2010). In January 1920, federal prohibition commenced with strong public approval and strict enforcement. Nevertheless, support for temperance faded substantially during the mid to late 1920s. America's prohibition experiment ended in 1933 with the repeal of the 18th amendment (Kobler 1974, Rumbarger 1989, Murdock 1998, Okrent 2010). Despite the common belief that prohibition was a failure, it dramatically reduced alcohol consumption for several years (Warburton 1932, Miron and Zwiebel 1991, Dills, Jacobson, and

2

Miron 2004, Blocker 2006). Figure 1 illustrates the trends in alcohol consumption per capita and alcoholism deaths per 100,000 population from 1900 to 1930 (Warburton 1932, Bureau of the Census, 1900-1931). As the figure shows, both measures decreased significantly in the late 1910s, reached their lowest level in 1920, and increased to roughly 70% of their 1900-1915 average by the mid-1920s. Thus, considerably less alcohol was consumed during the height of the temperance movement. In this study, we examine the long-term consequences of the U.S. temperance movement, one of the largest-scale policy changes impacting alcohol consumption in history. Using a sample of persons born between 1900 and 1925 drawn from nationally representative U.S. census microdata, the study investigates the effect of prenatal temperance environment on adult educational outcomes. The quantitative analysis uses proxies – the strength of temperance laws and the density of liquor retailers – to measure the degree to which the temperance movement shaped not only the physical accessibility of alcohol but also the social norms governing its consumption. To summarize, the evidence suggests that men and women exposed to the temperance movement in utero had modestly higher education. The effect on eighth grade completion was weakly significant at best, while the effect on high school completion was robustly significant, though relatively small in magnitude. Our preferred estimates from logistic models imply that the adoption of temperance laws increased the odds of high school completion by about 3-8%, and analogously that the reduction of retail liquor density, to the extent it changed from 1913 to 1921, increased the odds of high school completion by about 4-8%. Comparable estimates from linear probability models reveal an increase in high school completion of roughly 1.5 percentage points. We argue the findings are consistent with several causal mechanisms including a

3

reduction in fetal alcohol exposure and the adverse developmental outcomes with which it is associated. All in all, this study contributes to the literature. Social scientists have been increasingly interested in the link between prenatal environment and educational outcomes (e.g., Fuller 2014, Lee 2014, Neelsen and Stratmann 2011, Sanders 2012). Previous research on the effects of the temperance movement has focused almost entirely on contemporaneous outcomes, e.g., alcohol consumption and violence (Owens 2011, 2014, Livingston 2016, Warburton 1932, Miron and Zwiebel 1991, Dills, Jacobson, and Miron 2004, Blocker 2006, Miron 1999, Asbridge and Weerasinghe 2009). This is the first study to link prenatal temperance environment with educational attainment. Additionally, previous epidemiological research on the effects of maternal drinking on children's outcomes may be subject to some degree of reporting bias, which can arise from reliance on maternal self-reported drinking behaviors, as well as omitted variable bias, which can arise when persons prenatally exposed to alcohol and those unexposed differ in unobserved factors. By using a policy experiment, this study aims to minimize the influence of such biases that may be present in the literature. The remainder of the article is organized as follows. Section 2 describes the data and methods, Section 3 presents the results, and Section 4 discusses the conclusions.

2. Data and methods 2.1 Sample In a multivariate logistic regression framework, a person's educational attainment is regressed on the state-level temperance environment while he or she was in utero. To create the dataset, a sample of persons born in the U.S. between 1900 and 1925 is drawn from nationally

4

representative U.S. census microdata available from IPUMS (Ruggles et al. 2010). Samples from census years 1950, 1960, and 1970 are selected. Each of these are 1-in-100 national random samples of the population. In those census years, nearly all persons born between 1900 and 1925 had completed their schooling and were still alive. Advantageously, the 1960 and 1970 censuses identify year and quarter of birth and provide information on educational attainment for all persons, whereas the 1950 census identifies only year of birth and provides information on educational attainment for only "sample-line" (randomly selected) persons. For this reason, only "sample-line" persons from census year 1950 are retained. Each person in the sample is matched with measures of temperance environment corresponding with place (state) and time (first trimester) in utero. It is assumed that state in utero was identical to state of birth. For census years 1960 and 1970, it is assumed that persons were in utero three quarters (i.e., nine months) prior to their reported quarter of birth. For example, someone born in the first quarter of 1910 is assumed to be in utero in the second quarter of 1909. For census year 1950, it is assumed that persons were in utero during the year prior to their reported birth year, given that only birth year is available.

2.2 Variables According to Goldin (1998), educational attainment steadily increased during the first half of the twentieth century. Throughout the period, females were more likely than males to graduate from high school, which, in Goldin's view, suggests that the return to high school completion was larger for women. Indeed, tabulations using our census data confirm that females had higher eighth grade and high school completion for every birth cohort in the sample population. In the empirical analysis, the dependent variables consist of two measures of

5

educational attainment: whether a person completed eighth grade or higher (binary) and whether a person completed high school or more (binary). Both variables are from IPUMS. The main exposure variable is temperance environment, which is measured by either the strength of temperance laws or the density of liquor retailers in separate regressions. These measures reflect the extent to which economic and social barriers (e.g., prices, legal penalties, social acceptability, and availability) constrained the alcohol consumption of men and women and, therefore, serve as proxies for prenatal alcohol exposure. The strength of temperance laws is coded from Anti-Saloon League reports (Cherrington 1908, 1915, 1920). Early twentieth century state temperance laws are described in Appendix Table 1. "Weak" is a binary indicator that equals one if there was a local option or other substantive anti-liquor law in effect when and where a person was in utero. "Strong" is a binary indicator that equals one if the commercial sale of alcohol was effectively prohibited when and where a person was in utero. This variable takes a value of one for all states from the third quarter of 1919 to the first quarter of 1925, a period which commenced with the Wartime Prohibition Act and continued with the 18th amendment. Note that it is problematic to consider cohorts born after 1925, as drinking patterns changed drastically in the mid to late 1920s, even though prohibitory laws were still on the books. Figure 2 depicts the number of states with temperance laws during 1899-1925. Until 1909, most U.S. states had no substantive temperance laws. The number of states with strong laws increased modestly between 1908 and 1909 and then dramatically between 1914 and 1920. The density of liquor retailers is constructed from Bureau of Internal Revenue reports (1899-1925) and Bureau of the Census (1996) population estimates. The Bureau of Internal Revenue (later renamed the Internal Revenue Service) published data on liquor retailers during

6

the entire period of study, except for 1919. "Liquor retailer density" equals the number of liquor retailers per 1,000 population when and where a person was in utero. Figure 3 depicts the number of liquor retailers per 1,000 population during 1899-1925. According to the figure, the density of liquor retailers was relatively constant until about 1907 when it started to trend downward. The decline was moderate between 1907 and 1917 but severe between 1917 and 1921. It is critical to account for potential omitted variables which may be correlated with both educational attainment and temperance environment in utero. Individual-level controls are included in regressions. They are sex, race/ethnicity, census year, birth year fixed effects, and birth state fixed effects. All of these are from census years 1950, 1960, and 1970. State-level prenatal demographic controls, which vary by state and year in utero, are also included in regressions. They are constructed from census years 1900, 1910, 1920, and 1930. Values are linearly interpolated between years. State-level demographic controls include male labor force participation (the percentage of men age 16 and above who reported any gainful occupation), foreign-born population (the percentage of persons born outside of the U.S.), urban population (the percentage of persons in cities and incorporated places with 2,500+ inhabitants), adult literacy (the percentage of persons age 18 and above who could read and write), population age 18+ (the percentage of persons age 18 and above), white race (the percentage of persons identified as non-Hispanic whites), and high status occupation (the percentage of employed persons age 16 and above who were classified as professional, technical, managers, officials, and proprietors). Evidence on early twentieth century England and Wales suggests that male cirrhosis mortality was highest among the professional and managerial classes (Crombie and Precious 2011).

7

In addition to state-level prenatal demographic controls, some regressions also include region-specific trends in birth year to capture the influence of other background variables. To do so, we use the standard nine U.S. census regions.

2.3 Statistical models All analyses are performed with Stata 14. Multivariate logistic regressions are employed to investigate the association between educational attainment and temperance environment in utero. In the first set of regressions, the key independent variables are indicators for weak law and strong law, and the reference category is no substantive law. In the second set, the key independent variable is liquor retailer density, i.e., the number of liquor retailers per 1,000 population. The basic empirical model can be expressed as follows: y   t    x       w   ,

where y is educational attainment, t is prenatal temperance environment, x is a vector of individual-level controls,  is a birth year fixed effect,  is a birth state fixed effect, and w is a vector of state-level prenatal demographic controls. In the regressions, adjusted odds ratios (aOR) and 95% confidence intervals (CI) are reported. Robust standard errors are adjusted for clustering on birth state. To explore robustness, several extensions of the basic model are implemented. Multivariate logistic regressions are run with interactions between prenatal temperance environment and state pre-regulatory liquor intensity. The aim of this exercise is to confirm that persons born in states with historically elevated alcohol consumption are the ones that experience larger educational effects post-regulation. Furthermore, regressions are run separately for persons born in 1900-1919 and 1920-1925. Given that persons who belong to the younger cohort (1920-

8

1925) were exposed to alcohol regulations in utero and early childhood but not in later childhood or adulthood, it is informative to examine the size and significance of educational effects for these persons specifically. Additionally, regressions are run separately on various restricted samples including samples composed of only persons from the 1950 census, only persons from the 1960 census, and only persons from the 1970 census. Regressions are also run on a restricted sample excluding persons in utero when the influenza pandemic struck the U.S. during the fourth quarter of 1918 and the first quarter of 1919 (Almond 2006). Finally, linear probability models are run to confirm the robustness of the logistic models.

3. Results Table 1 displays means of dependent variables, prenatal temperance environment, and controls for persons born in the U.S. between 1900 and 1925. As the table shows, 82% of men and women had completed eighth grade or higher and 45% completed high school or more. It is also worth noting that almost 90% of the sample is identified as white, roughly half is female, and about 85% is drawn from census years 1960 and 1970. Table 2 explores the association between educational attainment and the strength of temperance laws while in utero. As the table reveals, individuals prenatally exposed to temperance laws had higher education. In the baseline model, the adjusted odds of completing eighth grade or higher are 1.06 (95% CI: 1.00, 1.12) times higher among people in utero under weak laws, relative to those in utero under no law. With the inclusion of region-specific trends, the adjusted odds of completing eighth grade or higher are 1.04 (95% CI: 1.00, 1.09) times higher among those in utero under weak laws. The estimated effect of strong law is not significant in either specification. Nevertheless, the effects on high school completion appear to

9

be larger and more significant than the effects on eighth grade completion. In the baseline model, the adjusted odds of completing high school or more are 1.05 (95% CI: 1.01, 1.10) times higher among people in utero under weak laws and 1.08 (95% CI: 1.02, 1.14) times higher among those in utero under strong laws, relative to those in utero under no law. With region-specific trends, the adjusted odds of completing high school or more are 1.03 (95% CI: 1.00, 1.06) times higher among people in utero under weak laws and 1.05 (95% CI: 1.01, 1.09) times higher among those in utero under strong laws. Regarding the individual-level covariates, it is worth noting that being female is associated with higher educational attainment, while being non-white is associated with lower attainment. Table 3 explores the association between educational attainment and the density of liquor retailers while in utero. As the table shows, educational attainment varied inversely with liquor retailer density. In the baseline model, the adjusted odds of completing eighth grade or higher increase by 4% (aOR: 0.96, 95% CI: 0.92, 1.01) with a one unit decrease in liquor retailer density. This estimated effect is significant at the 10% level. However, the effect becomes insignificant when region-specific trends are included in the model. Contrastingly, the effects on high school completion are consistently significant. In the baseline model, the adjusted odds of completing high school or more increase by 4% (aOR: 0.96, 95% CI: 0.94, 0.98) with a one unit decrease in liquor retailer density. With region-specific trends, the adjusted odds of completing high school or more increase by 2% (aOR: 0.98, 95% CI: 0.97, 0.99) with a one unit decrease in liquor retailer density. Like previous results, being female is associated with higher odds of educational attainment, and being non-white is associated with lower odds. Table 4 displays the results from several robustness exercises. First, regressions are run with interactions between prenatal temperance environment and pre-regulatory liquor intensity.

10

The objective is to investigate whether persons born in states with historically elevated alcohol consumption are the ones who experience larger educational effects post-regulation. "Low intensity" states are defined as those whose liquor retailer density is less than 2.33 in 1900 (e.g., Florida, Iowa, Massachusetts, Texas), and "high intensity" states are those whose liquor retailer density is more than 2.33 in 1900 (e.g., California, Maryland, New York, Wisconsin). Roughly half of the sample was born in a low intensity state and half in a high intensity state. The top panel shows the results. In sum, the only statistically significant interactions are those between temperance environment and high intensity. For example, the adjusted odds of completing high school or more are 1.06 (95% CI: 1.01, 1.11) times higher among people in utero in high intensity states with weak laws and 1.12 (95% CI: 1.04, 1.20) times higher among those in utero in high intensity states with strong laws. These findings make sense as states with historically elevated alcohol consumption are impacted most by changes in alcohol regulations. Next, regressions are run separately for persons born in 1900-1919 and 1920-1925. The objective is to explore whether the mechanisms driving the association between educational attainment and temperance environment involve exposure early in life or exposure later in life. Persons born between 1900 and 1919 were not only exposed to alcohol regulations in utero but also in childhood and adulthood. Persons born between 1920 and 1925, however, were exposed to alcohol regulations in utero and early childhood but were not exposed in later childhood or adulthood, since norms and enforcement of laws began to relax in the mid-1920s. Therefore, it is useful to see if the estimated effects are as strong for this younger cohort as they are for the older cohort. To conduct the exercise, it is only possible to measure temperance environment with liquor retailer density, because state temperance laws were invariant during 1920-1925. The middle panel shows the results. For cohorts born in 1900-1919, results are similar to previous

11

ones. Liquor retailer density is inversely associated with both measures of educational attainment, significantly so for high school completion. For cohorts born in 1920-1925, liquor retailer density is significantly associated with high school completion (aOR: 0.92, 95% CI: 0.90, 0.95) but not eighth grade completion. Strikingly, the estimated effects are larger, not smaller, for the younger cohort. This exercise suggests that the association between educational attainment and temperance environment is not just a phenomenon of later childhood or adulthood. Exposure occurring in utero or early childhood plays an important role. Lastly, regressions are run separately on various restricted samples. The analysis is conducted with samples composed of only persons from the 1950 census, only persons from the 1960 census, and only persons from the 1970 census. It is useful to confirm that the results are not an artifact of just one census year, which would raise concerns of sample selection. The analysis is also conducted with a sample excluding persons in utero when the influenza pandemic struck the U.S. during the fourth quarter of 1918 and the first quarter of 1919 (Almond 2006). In light of the well-established finding that prenatal exposure to flu reduced educational attainment, it is useful to confirm that the mechanism linking temperance environment and schooling is independent of the influenza pandemic. The bottom panel of Table 4 shows the results. Empirical patterns are similar across the census samples and the sample excluding flu quarters. Consistent with previous tables, both measures of prenatal temperance environment tend to have a more significant association with high school completion than with eighth grade completion, and for high school completion, the effect of strong law tends to be larger than that of weak law.

4. Discussion

12

Using a sample of persons born between 1900 and 1925, this study estimated the effect of prenatal temperance environment on adult educational outcomes. The quantitative analysis used proxies – the strength of temperance laws and the density of liquor retailers – to measure the degree to which the temperance movement shaped not only the physical accessibility of alcohol but also the social norms governing its consumption. In sum, the evidence suggests that men and women exposed to the temperance movement in utero had modestly higher education. The effect on eighth grade completion was weakly significant at best, while the effect on high school completion was consistently significant, though relatively small in magnitude. Estimates from logistic models imply that the adoption of temperance laws increased the odds of high school completion by about 3-8%, and analogously that the reduction of retail liquor density, to the extent it changed from 1913 to 1921, increased the odds of high school completion by about 48%. Comparable estimates from linear probability models reveal an increase in high school completion of roughly 1.5 percentage points. In light of the decline in alcohol consumption during the temperance movement (Warburton 1932, Miron and Zwiebel 1991, Dills, Jacobson, and Miron 2004, Blocker 2006), several possible causal mechanisms are consistent with the study's findings. On one hand, the mechanism may be behavioral. It could be that households decreased their expenditures on alcohol and, in turn, increased their expenditures on children's education. It also could be that parents consumed less alcohol, which improved the home environment in ways that stimulated children's education. On the other hand, the mechanism may be biological. That is, the results may reflect changes in the incidence of fetal alcohol spectrum disorders (FASD). FASD encompasses the entire range of disorders associated with prenatal alcohol exposure (Hoyme et al. 2005, Mattson et al. 2011).

13

Given the growing interest in FASD, it is worth discussing this possible mechanism further. Persons diagnosed with FASD may have physical impairments, learning disabilities, and behavioral problems (Kodituwakku 2007, Streissguth et al. 1991, Spohr et al. 1993, Mattson et al. 2011). They are less likely to complete secondary education and more likely to experience negative outcomes including alcohol or drug abuse (Yates et al. 1998, Griesler and Kandel 1998, Baer et al. 2003, Streissguth et al. 2004, Rangmar et al. 2015). Then, to assess the hypothesis that FASD might explain the results, it is necessary to consider the extent to which pregnant women consumed alcohol during the early twentieth century. While the alcohol problem was often portrayed as a masculine one, women did drink. Lender (1981) compiles historical surveys of persons institutionalized for alcohol abuse and finds that 10-25 percent of alcoholics were female at the turn of the century. Based on a variety of historical evidence, Murdock (1998) argues that "significant numbers of respectable American women consumed alcohol," and that most female drinking was moderate and happened at home. Therefore, FASD is a potential causal mechanism. This study makes several contributions to the literature. It yields historical insights, as it is the first to examine the long-term consequences of the U.S. temperance movement on educational attainment. Thus, it contributes to research on the effects of temperance (e.g., Owens 2011, 2014, Livingston 2016, Blocker 2006, Miron and Zwiebel 1991). The most comparable study is one that takes advantage of a 1960s policy experiment which allowed the sale of strong beer in Swedish grocery stores (Nilsson 2017). It finds that those exposed to the Swedish policy in utero had lower earnings, educational attainment, and cognitive ability. Furthermore, this study bolsters the wider social science literature on prenatal environment, which demonstrates that prenatal exposure to maternal illness, natural disaster, war, famine, and pollution can reduce

14

educational attainment later in life (Almond 2006, Fuller 2014, Lee 2014, Neelsen and Stratmann 2011, Sanders 2012). To the degree they reflect changes in the incidence of FASD, the results are congruent with evidence that persons diagnosed with FASD have deficits in intellectual performance and executive functioning which persist beyond childhood (Streissguth et al. 1991, Spohr et al. 1993, Mattson et al. 2011, Rangmar et al. 2015). However, relative to most epidemiological research, this study is uncommon in its approach to estimating the effects of prenatal alcohol exposure. By investigating an unprecedented policy change, it sought to minimize the distortionary influence of reporting bias, which can arise from reliance on maternal self-reported drinking behaviors, and omitted variable bias, which can arise when persons prenatally exposed to alcohol and those unexposed differ in unobserved factors. In this sense, the study is comparable to recent investigations which make use of plausibly exogenous variation in prenatal alcohol exposure (von Hinke Kessler Scholder et al. 2014, Nilsson 2017). Although the study is historical in focus, it holds significance today. The findings demonstrate that the consequences of prenatal exposure to alcohol policies extend beyond childhood into adulthood and beyond health outcomes into other life outcomes. This does not imply that society should return to prohibition-era policies, but it does underscore the urgent need to reduce alcohol use among pregnant women. Indeed, fetal alcohol exposure still occurs at an unacceptably high level, since research by the U.S. Centers for Disease Control and Prevention (Tan et al. 2015) finds that one in 10 pregnant women reported consuming alcohol in the past 30 days and one in 33 reported binge drinking.

15

References Almond D. Is the 1918 Influenza Pandemic Over? Long-Term Effects of In Utero Influenza Exposure in the Post-1940 U.S. Population. Journal of Political Economy. 2006;114(4):672– 712. Asbridge M, Weerasinghe S. Homicide in Chicago from 1890 to 1930: prohibition and its impact on alcohol- and non-alcohol-related homicides. Addiction. 2009;104(3):355-364. Baer JS, Sampson PD, Barr HM, Connor PD, Streissguth AP. A 21-year longitudinal analysis of the effects of prenatal alcohol exposure on young adult drinking. Archives of General Psychiatry. 2003;60:377–385. Blocker JS. Did Prohibition Really Work? Alcohol Prohibition as a Public Health Innovation. Am J Public Health. 2006;96(2):233–243. Bureau of Internal Revenue. Annual report of the Commissioner of Internal Revenue. Washington, DC: US Government Printing Office; 1899-1925. Available online, http://www.irs.gov/uac/SOI-Tax-Stats-Archive-1863-to-1999-Annual-Reports-and-IRSData-Books, accessed on June 17, 2015. Bureau of the Census. Intercensal Estimates of the Total Resident Population of States. 1996. Available online, https://www.census.gov/popest/data/state/asrh/1980s/80s_st_totals.html, accessed on June 17, 2015. Bureau of the Census. Mortality Statistics. Washington, DC: US Government Printing Office; 1900-1931. Available online, http://www.cdc.gov/nchs/products/vsus/vsus_1890_1938.htm, accessed on June 17, 2015. Cherrington EH. The Anti-Saloon League Year Book: An Encyclopedia of Facts and Figures Dealing with the Liquor Traffic and the Temperance Reform. Westerville, OH: The AntiSaloon League of America;1908, 1915, 1920. Crombie, Iain K. and Elaine Precious. 2011. "Changes in the Social Class Gradient of Cirrhosis Mortality in England and Wales across the 20th Century." Alcohol and Alcoholism, 46(1): 80-82. Dills AK, Jacobson M, Miron JA. The effect of alcohol prohibition on alcohol consumption: evidence from drunkenness arrests. Economics Letters. 2004;86:279-284. Fuller, Sarah C. 2014. "The Effect of Prenatal Natural Disaster Exposure on School Outcomes." Demography, 51(4): 1501-1525. Goldin, Claudia. 1998. "America's Graduation from High School: The Evolution and Spread of Secondary Schooling in the Twentieth Century." The Journal of Economic History, 58(2): 345-374. Griesler PC, Kandel DB. The impact of maternal drinking during and after pregnancy on the drinking of adolescent offspring. Journal of Studies on Alcohol. 1998;59(3):292–304. Hoyme HE, May PA, Kalberg WO, et al. A practical clinical approach to diagnosis of fetal alcohol spectrum disorders: clarification of the 1996 institute of medicine criteria. Pediatrics. 2005;115(1):39–47. Kobler, J. Ardent Spirits: The Rise and Fall of Prohibition. London: Michael Joseph; 1974. Kodituwakku PW. Defining the behavioral phenotype in children with fetal alcohol spectrum disorders: A review. Neuroscience and Biobehavioral Review. 2007;31(2):192-201. Lee, Chulhee. 2014. "In utero exposure to the Korean War and its long-term effects on socioeconomic and health outcomes." Journal of Health Economics, 33: 76-93.

16

Lender ME. Women Alcoholics: Prevalence Estimates and Their Problems as Reflected in Turnof-the-Century Institutional Data. International Journal of the Addictions. 1981;16(3):443448. Livingston, B. "Murder and the Black Market: Prohibition's Impact on Homicide Rates in American Cities." International Review of Law and Economics, 2016;45(March): 33-44. Mattson SN, Crocker N, Nguyen TT. Fetal Alcohol Spectrum Disorders: Neuropsychological and Behavioral Features. Neuropsychology. 2011;21(2):81-101. Miron JA, Zwiebel J. Alcohol Consumption During Prohibition. American Economic Review. 1991;81(2):242-247. Miron JA. Violence and U.S. Prohibitions of Drugs and Alcohol. American Law and Economics Review. 1999;1(1-2):78-114. Murdock CG. Domesticating Drink: Women, Men, and Alcohol in America, 1870-1940. Baltimore: The Johns Hopkins University Press;1998. Neelsen, Sven and Thomas Stratmann. 2011. "Effects of prenatal and early life malnutrition: Evidence from the Greek famine." Journal of Health Economics, 30(3): 479-488. Nilsson JP. Alcohol Availability, Prenatal Conditions, and Long-term Economic Outcomes. Journal of Political Economy, 2017; 125(4): 1149-1207. Okrent, D. Last Call: The Rise and Fall of Prohibition. New York: Scribner;2010. Owens EG. Are Underground Markets Really More Violent? Evidence from Early 20th Century America. American Law and Economics Review. 2011;13(1):1-44. Owens EG. The American Temperance Movement and Market-Based Violence. American Law and Economics Review. 2014;16(2):433-472. Rangmar J, Hjern A, Vinnerljung B, et al. Psychosocial Outcomes of Fetal Alcohol Syndrome in Adulthood. Pediatrics. 2015;135(1):E52-E58. Ruggles SJ, Alexander T, Genadek K, Goeken R, Schroeder MB, Sobek M. Integrated Public Use Microdata Series: Version 5.0. Minneapolis: University of Minnesota;2010. Rumbarger, JJ. Profits, Power, and Prohibition: Alcohol Reform and the Industrializing of America, 1800-1930. Albany: State University of New York Press;1989. Sanders, Nicholas J. 2012. "What Doesn't Kill You Makes You Weaker Prenatal Pollution Exposure and Educational Outcomes." Journal of Human Resources, 47(3): 826-850. Spohr HL, Willms J, Steinhausen HC. Prenatal alcohol exposure and long-term developmental consequences. The Lancet. 1993;341:907–910. Streissguth AP, Asse JM, Clarren SK, et al. Fetal Alcohol Syndrome in Adolescents and Adults. JAMA. 1991;265(15):1961-1967. Streissguth AP, Bookstein FL, Barr HM, Sampson PD, O’Malley K, Young JK. Risk factors for adverse life outcomes in fetal alcohol syndrome and fetal alcohol effects. Developmental and Behavioral Pediatrics. 2004;5(4):228-238. Tan, CH, Denny CH, Cheal NE, et al. Alcohol use and binge drinking among women of childbearing age–United States, 2011-2013. Morbidity and Mortality Weekly Report (MMWR). 2015;64(37):1042-1046. von Hinke Kessler Scholder S, Wehby GL, Lewis S, Zuccolo L. Alcohol Exposure In Utero and Child Academic Achievement. Economic Journal. 2014;124(576):634-667. Warburton C. The Economic Results of Prohibition. New York: Columbia University Press;1932.

17

Yates WR, Cadoret RJ, Troughton EP, et al. Effect of fetal alcohol exposure on adult symptoms of nicotine, alcohol, and drug dependence. Alcoholism: Clinical and Experimental Research. 1998;22(4):914-920.

18

8.0 7.0 6.0 5.0 4.0 3.0 2.0

1.0

Deaths per 100,000 population

2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0

0.0

1900 1902 1904 1906 1908 1910 1912 1914 1916 1918 1920 1922 1924 1926 1928 1930

Gallons per capita

Figure 1. Alcohol consumption per capita and alcoholism death rate in the United States, 19001930

Alcohol consumption

Alcoholism death rate

NOTE. Estimates of alcohol consumption per capita are from Warburton (1932). Estimates of deaths attributed to alcoholism are from the Bureau of the Census (1900-1931).

19

50 45 40 35 30 25 20 15 10 5 0

1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925

Number of U.S. states

Figure 2. Number of U.S. states with temperance laws, 1899-1925

Strong law

Weak law

NOTE. The figure displays the number of states with temperance laws in effect at the beginning of the year. "Strong" refers to laws effectively prohibiting the commercial sale of alcohol statewide, and "weak" refers to local option or other substantive anti-liquor laws short of prohibition. Variables were coded from Anti-Saloon League reports (Cherrington 1908, 1915, 1920).

20

Figure 3. Number of liquor retailers per 1,000 population, 1899-1925

Liquor retailers per 1,000 population

3 2.5 2 1.5 1 0.5

1899 1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925

0

NOTE. Variable was constructed from Bureau of Internal Revenue reports (1899-1925) and Bureau of the Census (1996) population estimates. The value for 1919 was linearly interpolated because data was unavailable.

21

Table 1 Descriptive statistics for persons born between 1900 and 1925

Dependent variables Educational attainment Eighth grade or higher High school completion or more Prenatal temperance environment Liquor retailers per 1,000 population Temperance laws No law Weak law Strong law Individual-level controls Sex Male Female Birth year cohort 1900-1904 1905-1909 1910-1914 1915-1919 1920-1925 Race/ethnicity White Black Hispanic Asian/American Indian/Other Census region of birth New England Middle Atlantic East North Central West North Central South Atlantic East South Central West South Central Mountain Pacific Census year 1950 1960 1970 State-level demographic controls Male labor force participation (%) Foreign-born population (%) Urban population (%) Adult literacy (%) Population age 18+ (%) White race (%) High status occupation (%)

Mean

SD

Min

Max

0.82 0.45

0.38 0.50

0 0

1 1

1.69

1.61

0

25

0.39 0.24 0.37

0.49 0.43 0.48

0 0 0

1 1 1

0.48 0.52

0.50 0.50

0 0

1 1

0.14 0.17 0.20 0.22 0.28

0.35 0.37 0.40 0.41 0.45

0 0 0 0 0

1 1 1 1 1

0.89 0.10 0.01 0.01

0.32 0.29 0.12 0.07

0 0 0 0

1 1 1 1

0.06 0.20 0.19 0.12 0.15 0.10 0.11 0.03 0.04

0.24 0.40 0.39 0.33 0.35 0.30 0.31 0.17 0.20

0 0 0 0 0 0 0 0 0

1 1 1 1 1 1 1 1 1

0.14 0.45 0.40

0.35 0.50 0.49

0 0 0

1 1 1

90.1 13.3 44.7 90.4 61.5 86.9 11.2

1.9 10.1 21.8 8.0 5.3 16.4 2.8

83.4 0.2 3.0 57.7 50.1 17.6 0.0

100.0 64.9 100.0 98.9 90.6 100.0 17.9

NOTE. Sample size is 1,110,643.

22

Table 2 Multivariate logistic regression analysis of the association between educational attainment and prenatal temperance environment (temperance laws) for persons born between 1900 and 1925 Eighth grade or higher aOR (95% CI) Temperance laws No law ref. Weak law 1.06 (1.00 - 1.12)* Strong law 1.02 (0.96 - 1.08) Sex Male ref. Female 1.37 (1.31 - 1.44)* Race/ethnicity White non-Hispanic ref. Black non-Hispanic 0.33 (0.29 - 0.39)* Hispanic 0.15 (0.09 - 0.28)* Other non-Hispanic 0.28 (0.18 - 0.44)* Census year 1950 ref. 1960 1.08 (1.05 - 1.11)* 1970 1.17 (1.14 - 1.21)* State demographic controls Male labor force participation 1.00 (0.96 - 1.03) Foreign-born population 0.96 (0.94 - 0.98)* Urban population 1.02 (1.00 - 1.04)* Adult literacy 0.97 (0.96 - 0.99)* Population age 18+ 0.98 (0.94 - 1.01) White race 1.01 (0.98 - 1.03) High status occupation 1.00 (0.90 - 1.10) Birth year fixed effects Birth state fixed effects Region-specific time trends N

High school completion or more

aOR (95% CI)

aOR (95% CI)

aOR (95% CI)

ref. ref. ref. 1.04 (1.00 - 1.09)+ 1.05 (1.01 - 1.10)* 1.03 (1.00 - 1.06)+ 1.00 (0.96 - 1.05) 1.08 (1.02 - 1.14)* 1.05 (1.01 - 1.09)* ref. ref. ref. 1.37 (1.31 - 1.44)* 1.21 (1.15 - 1.26)* 1.21 (1.15 - 1.26)* ref. ref. ref. 0.33 (0.29 - 0.39)* 0.35 (0.30 - 0.41)* 0.35 (0.30 - 0.41)* 0.16 (0.09 - 0.28)* 0.27 (0.17 - 0.41)* 0.26 (0.17 - 0.41)* 0.28 (0.18 - 0.44)* 0.55 (0.38 - 0.80)* 0.55 (0.38 - 0.79)* ref. ref. ref. 1.07 (1.05 - 1.10)* 1.05 (1.01 - 1.08)* 1.05 (1.01 - 1.08)* 1.17 (1.14 - 1.20)* 1.14 (1.12 - 1.17)* 1.14 (1.12 - 1.17)* 0.97 (0.95 - 1.00)* 0.95 (0.93 - 0.97)* 1.02 (1.01 - 1.03)* 1.01 (1.00 - 1.03)* 1.03 (1.00 - 1.06)+ 1.00 (0.99 - 1.02) 0.99 (0.95 - 1.03)

1.03 (1.00 - 1.05)* 0.98 (0.97 - 0.99)* 1.00 (0.99 - 1.01) 0.99 (0.98 - 1.00)+ 1.02 (1.00 - 1.04)+ 0.99 (0.97 - 1.01) 1.04 (0.99 - 1.09)

1.01 (1.00 - 1.03)+ 0.97 (0.96 - 0.99)* 1.00 (1.00 - 1.01) 1.00 (0.99 - 1.02) 1.04 (1.02 - 1.05)* 1.00 (0.99 - 1.02) 1.04 (1.01 - 1.06)*

Yes Yes No

Yes Yes Yes

Yes Yes No

Yes Yes Yes

1,110,643

1,110,643

1,110,643

1,110,643

NOTE. "aOR" indicates adjusted odds ratio from logistic regression. Robust standard errors are adjusted for clustering on birth state. An asterisk indicates significance at the 5% level, and a plus indicates significance at the 10% level.

23

Table 3 Multivariate logistic regression analysis of the association between educational attainment and prenatal temperance environment (liquor retailer density) for persons born between 1900 and 1925 Eighth grade or higher aOR (95% CI) Liquor retailer density 0.96 (0.92 - 1.01)+ Sex Male ref. Female 1.37 (1.31 - 1.44)* Race/ethnicity White non-Hispanic ref. Black non-Hispanic 0.33 (0.28 - 0.38)* Hispanic 0.15 (0.09 - 0.28)* Other non-Hispanic 0.28 (0.18 - 0.44)* Census year 1950 ref. 1960 1.07 (1.05 - 1.10)* 1970 1.17 (1.14 - 1.20)* State demographic controls Male labor force participation 0.99 (0.95 - 1.03) Foreign-born population 0.96 (0.95 - 0.98)* Urban population 1.02 (1.00 - 1.04)* Adult literacy 0.98 (0.96 - 0.99)* Population age 18+ 0.98 (0.95 - 1.02) White race 1.01 (0.99 - 1.03) High status occupation 1.00 (0.90 - 1.10) Birth year fixed effects Birth state fixed effects Region-specific time trends N

High school completion or more

aOR (95% CI) 0.99 (0.97 - 1.01)

aOR (95% CI)

aOR (95% CI)

0.96 (0.94 - 0.98)* 0.98 (0.97 - 0.99)*

ref. ref. ref. 1.37 (1.31 - 1.44)* 1.21 (1.16 - 1.27)* 1.21 (1.16 - 1.27)* ref. ref. ref. 0.33 (0.28 - 0.38)* 0.35 (0.30 - 0.41)* 0.35 (0.30 - 0.41)* 0.16 (0.09 - 0.28)* 0.26 (0.17 - 0.41)* 0.27 (0.17 - 0.42)* 0.28 (0.18 - 0.44)* 0.55 (0.38 - 0.78)* 0.55 (0.38 - 0.78)* ref. ref. ref. 1.07 (1.04 - 1.10)* 1.04 (1.01 - 1.08)* 1.04 (1.00 - 1.07)* 1.16 (1.13 - 1.19)* 1.13 (1.11 - 1.16)* 1.13 (1.11 - 1.16)* 0.97 (0.95 - 1.00)* 0.95 (0.93 - 0.97)* 1.02 (1.01 - 1.03)* 1.01 (1.00 - 1.03)* 1.04 (1.00 - 1.07)* 1.00 (0.99 - 1.02) 0.99 (0.95 - 1.03)

1.02 (1.00 - 1.04)* 0.99 (0.98 - 1.00)* 1.00 (0.99 - 1.01) 1.00 (0.98 - 1.01) 1.02 (1.01 - 1.04)* 0.99 (0.97 - 1.01) 1.04 (0.99 - 1.09)

1.01 (1.00 - 1.02) 0.97 (0.96 - 0.99)* 1.00 (1.00 - 1.01) 1.01 (1.00 - 1.02) 1.04 (1.03 - 1.06)* 1.00 (0.98 - 1.02) 1.03 (1.01 - 1.06)*

Yes Yes No

Yes Yes Yes

Yes Yes No

Yes Yes Yes

1,110,643

1,110,643

1,110,643

1,110,643

NOTE. "aOR" indicates adjusted odds ratio from logistic regression. Robust standard errors are adjusted for clustering on birth state. An asterisk indicates significance at the 5% level, and a plus indicates significance at the 10% level.

24

Table 4 Estimates from various models estimating the association between educational attainment and prenatal temperance environment Eighth grade or higher aOR (95% CI)

Logistic model Interactions with state liquor intensity in 1900

Weak law x low intensity Weak law x high intensity Strong law x low intensity Strong law x high intensity

1.00 (0.92 - 1.09) 1.15 (1.07 - 1.23)* 0.97 (0.89 - 1.07) 1.13 (0.98 - 1.31)+

High school completion aOR (95% CI) 1.05 (0.97 - 1.14) 1.06 (1.01 - 1.11)* 1.05 (0.95 - 1.17) 1.12 (1.04 - 1.20)*

Liquor retailer density x low intensity 1.00 (0.86 - 1.15) 0.94 (0.87 - 1.02) Liquor retailer density x high intensity 0.97 (0.93 - 1.00)+ 0.96 (0.94 - 0.98)*

Cohorts born 1920-25

Liquor retailer density

1.00 (0.95 - 1.04)

0.92 (0.90 - 0.95)*

Cohorts born 1900-19

Liquor retailer density

0.97 (0.93 - 1.02)

0.97 (0.94 - 1.00)*

Only 1950 census

Weak law Strong law

1.06 (0.98 - 1.14) 1.01 (0.91 - 1.12)

1.08 (1.02 - 1.15)* 1.08 (1.01 - 1.16)*

Liquor retailer density

0.96 (0.92 - 1.01)+ 0.97 (0.95 - 0.99)*

Weak law Strong law

1.06 (1.00 - 1.13)+ 1.04 (0.98 - 1.09) 1.01 (0.93 - 1.09) 1.05 (0.98 - 1.12)

Liquor retailer density

0.95 (0.91 - 0.99)* 0.96 (0.94 - 0.98)*

Weak law Strong law

1.06 (1.00 - 1.12)* 1.06 (1.01 - 1.11)* 1.04 (1.00 - 1.09)+ 1.11 (1.06 - 1.18)*

Liquor retailer density

0.98 (0.93 - 1.03)

Weak law Strong law

1.06 (1.00 - 1.12)* 1.05 (1.01 - 1.10)* 1.03 (0.97 - 1.09) 1.08 (1.02 - 1.15)*

Liquor retailer density

0.96 (0.92 - 1.01)+ 0.96 (0.94 - 0.98)*

Only 1960 census

Only 1970 census

Without flu quarters

0.96 (0.94 - 0.99)*

Eighth grade or higher b (SE)

High school completion b (SE)

Weak law Strong law

0.008 (0.004)* -0.001 (0.005)

0.012 (0.006)* 0.016 (0.007)*

Liquor retailer density

0.002 (0.003)

-0.007 (0.003)*

Linear probability model

NOTE. "aOR" indicates adjusted odds ratio from logistic regression, and "b" indicates coefficient from linear probability regression. All regressions include individual-level controls, state-level demographic controls, birth year fixed effects, and birth state fixed effects. The model "without flu quarters" omits observations associated with the fourth quarter of 1918 and the first quarter of 1919 when the influenza pandemic struck the U.S. Robust standard errors are adjusted for clustering on birth state. An asterisk indicates significance at the 5% level, and a plus indicates significance at the 10% level.

25

Appendix Table 1 U.S. state temperance laws from late 1800s to 1920 Weak Law

Alabama Alaska Arizona Arkansas California Colorado Connecticut Delaware D.C. Florida Georgia Hawaii Idaho Illinois Indiana Iowa Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota Mississippi Missouri

Start

End

Strong Law Start

1911.Q2 None 1909.Q1 1913.Q1 1911.Q1 1907.Q1 1909.Q1 1907.Q1 None 1913.Q1 1908.Q1 None 1909.Q3 1907.Q1 1905.Q1 1907.Q1 None 1906.Q1 1909.Q1 1884.Q1 1914.Q1 1916.Q1 1908.Q3 1915.Q1 1886.Q1 1913.Q1

1915.Q2 None 1914.Q4 1915.Q4 1919.Q2 1915.Q4 1919.Q2 1919.Q2 None 1915.Q2 1916.Q1 None 1915.Q4 1919.Q2 1918.Q1 1915.Q4 None 1919.Q2 1919.Q2 1905.Q1 1919.Q2 1919.Q2 1918.Q1 1919.Q2 1908.Q4 1919.Q2

1915.Q3 1918.Q1 1915.Q1 1916.Q1 1919.Q3 1916.Q1 1919.Q3 1919.Q3 1917.Q4 1915.Q3 1916.Q2 1918.Q2 1916.Q1 1919.Q3 1918.Q2 1916.Q1 1880.Q1 1919.Q3 1919.Q3 1905.Q2 1919.Q3 1919.Q3 1918.Q2 1919.Q3 1909.Q1 1919.Q3

Weak Law

Montana Nebraska Nevada New Hampshire New Jersey New Mexico New York North Carolina North Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island South Carolina South Dakota Tennessee Texas Utah Vermont Virginia Washington West Virginia Wisconsin Wyoming

Start

End

Strong Law Start

1917.Q1 1909.Q1 None 1903.Q1 1918.Q1 1913.Q1 1917.Q2 1903.Q1 None 1908.Q1 None 1913.Q1 None None 1892.Q1 1889.Q1 1909.Q3 None 1911.Q3 1903.Q3 1904.Q1 1909.Q1 1863.Q1 1914.Q3 None

1918.Q4 1917.Q1 None 1918.Q1 1919.Q2 1918.Q3 1919.Q2 1908.Q4 None 1912.Q4 None 1915.Q4 None None 1915.Q4 1917.Q2 1914.Q4 None 1917.Q2 1919.Q2 1916.Q3 1915.Q4 1914.Q2 1919.Q2 None

1919.Q1 1917.Q2 1919.Q1 1918.Q2 1919.Q3 1918.Q4 1919.Q3 1909.Q1 1889.Q4 1919.Q2 1907.Q4 1916.Q1 1919.Q3 1919.Q3 1916.Q1 1917.Q3 1915.Q1 1918.Q3 1917.Q3 1919.Q3 1916.Q4 1916.Q1 1914.Q3 1919.Q3 1919.Q3

NOTE. Start and end dates indicate the approximate quarters during which a state temperance law was effective. Q1 refers to the first quarter of the year (January-March), Q2 to the second quarter (April-June), Q3 to the third quarter (July-September), and Q4 to the fourth quarter (October-December). There were other temperance laws in addition to those listed in the table. Alabama also had a weak law from 1907.Q1 to 1908.Q4 and a strong law from 1909.Q1 to 1911.Q1. New Hampshire also had a strong law from 1855.Q1 to 1902.Q4. Oregon also had a weak law from 1904.Q3 to 1910.Q4. Vermont also had a strong law from 1852.Q1 to 1903.Q2. The primary reference for this table is Anti-Saloon League Year Book, 1920.

26