Municipal Housekeeping: The Impact of Women’s Suffrage on the Provision of Public Education Celeste K. Carruthers and Marianne H. Wanamaker∗ October 2012 Abstract Models of intrahousehold bargaining and electoral competition predict that (1) women place a higher value on child welfare than men and (2) public expenditures change to better reflect these preferences in the wake of women’s suffrage. We use newly digitized data on local educational expenditures and schooling resources in four segregated Southern states to test whether the Nineteenth Amendment to the U.S. Constitution, which extended the voting franchise to all adult women in the country, explains part of the momentous growth in educational expenditure and school quality in the interwar years. Suffrage significantly increased local educational spending – by as much as 50.7 percent in the short-term – predominantly by employing more teachers and raising teacher salaries. But because effects on white schools outpaced those for black, suffrage expansion exacerbated racial inequality in education expenditures, setting the relative ascent of black school spending back by seventeen years. JEL Codes: H75, I22, P16 ∗ Carruthers: University of Tennessee, 702 Stokely Management Center, Knoxville, TN 37996-0570, [email protected]. Wanamaker: University of Tennessee, 524 Stokely Management Center, Knoxville, TN 37996-0550, [email protected]. Ye Gu, Ahiteme Houndonougbo, Andrew Moore, James England, III, and Nicholas Busko provided outstanding research assistance. We thank Robert Whaples and Michael Price for thoughtful comments and suggestions, as well as participants at the 2012 Appalachian Spring Conference in World History and Economics. We are grateful to Larry Kenny for sharing data. Seed funding for this project was provided by the University of Tennessee Office of Research. Additional support includes a grant from the Spencer Foundation, grant number 201200064, and a grant from the University of Kentucky Center for Poverty Research via the U.S. Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation, grant number 5 U01 PE000002-06. The opinions and conclusions expressed herein are solely those of the authors and should not be construed as representing the opinion or policy of the Spencer Foundation, the U.S. Department of Health and Human Services, or any agency of the Federal Government. All errors are our own. 1 1 Introduction “The men have been carelessly indifferent to much of this civic housekeeping, as they have always been indifferent to details of the household... The very multifariousness and complexity of a city government demand the help of minds accustomed to detail and variety of work, to a sense of obligation for the health and welfare of young children and to a responsibility for the cleanliness and comfort of others.” - Jane Addams1 At the dawn of the twentieth century, the United States held a position of distinction in the provision of public education. Among its Western Hemisphere peers, the U.S. exhibited the highest “common school” (grade 1-8) enrollment rates. In addition, the U.S. was showing early leadership in the race towards mass secondary education. Indeed, the country had enjoyed a substantial and persistent mass education advantage since the middle of the previous century aided by the country’s commitment to a set of “egalitarian principles ” (Goldin, 2001). These principles included “public funding, openness, gender neutrality, local (and also state) control, separation of church and state, and an academic curriculum,” and they drove the United States to world leadership in education provision by 1900. As the twentieth century progressed, the United States strengthened its leadership position in the provision of public education and brought secondary education to the masses. By the middle of the nineteenth century, and by the dawn of the second world war, the median 19-year-old was a secondary school graduate. The available literature cites the continued application of American egalitarian principles and strong labor market demand for an educated workforce to explain the steady advance of public education provision (Goldin & Katz, 2009). Yet, as we show in Figures 1 - 3, the growth of education provision was not constant over the course of the century; the commitment of states and local school districts to the funding of public education exhibits a marked uptick around 1920, the same year that the Nineteenth Amendment to the U.S. Constitution guaranteed full voting rights to all adult females. Given the timing of the change, is there a role for universal suffrage in explaining the renewed commitment of school districts to public education finance? An extensive empirical literature documents a greater propensity of women to support the pro1 Quoted in Harper (1922), p. 178. 2 vision of public goods, to hold other-regarding preferences, to foster the expansion of government to benefit child welfare and, in some ways, to hold Goldin’s “egalitarian principles” closer to heart.2 Standard models of electoral competition indicate that policy makers will respond to shifting preferences of their electoral base by altering their own voting behavior. The testable implication is that the enfranchisement of women should have resulted in greater education expenditures following the ratification of the Nineteenth Amendment. Indeed, researchers in this literature have identified a measurable impact of expanded suffrage on government provision of other services.3 An important caveat to this expectation, however, is the idea that the impact of suffrage on public spending was fleeting as policymakers became less wary of the female vote (Moehling & Thomasson, 2012). We use a new, county-level panel dataset of education spending for four Southern U.S. states – Alabama, Georgia, Louisiana, and South Caolina – to test this hypothesis. These states are not among the twenty-nine that had already granted full voting rights to adult females by 1920. Rather, the counties in this panel were coerced into extending the voting franchise by the passage of the Nineteenth Amendment, which none of the four states even ratified until the 1950s and 1970s. This limits the scope of identifying variation at hand, since there is just one point of intertemporal variation in suffrage, and it is shared by all counties in our panel. One benefit from this limitation, however, is that policy endogeneity is definitively ruled out as a driving factor behind results. Comparing annual education finance statistics to the timing of women’s suffrage, we examine the role of suffrage in generating meaningful changes in education expenditures. Even though all states with local data were effectively treated by suffrage in 1920, cross-county variation in gender shares and voter turnout means that some counties realized a larger dosage of suffrage. Both the relative size of the female voting-age population and the turnout rates of female voters differed substantially by county. We exploit this demographic variation as a measure of the relative power of females in the democratic process to generate testable implications for education spending. Because the data spans the Jim Crow-era South, the electoral model has further implications for variation in the impact of suffrage across races. Although the Nineteenth Amendment theoretically applied to all females of voting age, the de jure outcome was an expansion of voting rights to white females more so than to black females. Severe disenfranchisement through poll 2 3 See, inter alia, Eckel & Grossman (1996); Croson & Gneezy (2009); Li et al. (2011); Doepke & Tertilt (2012). Husted & Kenny (1997); Miller (2008); Lott & Kenny (1999). 3 taxes, literacy tests, and voter intimidation meant that although the median Southern voter became more female after 1920, it did not become any less white until later decades. But because schools were segregated and education expenditures were reported by race, as a further test of our predictions, we examine whether the expansion of voting rights to females affected expenditures on white schools differently from black schools. Findings show that expanded suffrage had a significant, positive impact on both black and white school expenditures, but that white school spending gains far outpaced those for black schools. As a result we conclude that, by 1940, the level of both black and white school expenditure would have been measurably lower in the absence of women’s suffrage, but racial inequality in the same would have been substantially smaller as well. Following a short period of gains, the ratio of black to white per-capita spending fell by 6.6 percent immediately following suffrage and did not return to 1920 levels for seventeen years. 2 Theoretical Foundation This study is motivated by the observation that per capita education expenditures in the United States demonstrate a marked uptick around 1920. The accelerated growth of education funding may be the result of numerous causes: the impact of World War I and the ensuing recession, rising living standards and incomes, a modernizing work force and a rising demand by employers for formal human capital, changes in compulsory schooling requirements, or expanded “free tuition” legislation.4 Our study examines the role of an expanded electorate in the expenditure increase. Two distinct lines of economic research motivate the analysis. First, a classical model of electoral competition indicates that policymakers will vote in accordance with the view of the median voter in the electorate.5 Even if the decisive voter is not strictly at the median of income or preference indices (Brunner & Ross, 2010), expanding the electorate at the scale realized by the Nineteenth Amendment has dramatic implications for vote-maximizing behavior if new voters have systematically different preferences for the provision of public services. Changes in the size of the electorate matter inasmuch as new participants in the political process exhibit different preferences 4 See Goldin & Katz (2009) for a more thorough discussion of the social, economic, and political landscape that contributed to the growth of public education in the early 20th century. 5 See Bowen (1943), and Black (1948), and Baumgardner (1993). 4 from incumbents. The Nineteenth Amendment doubled the size of the electorate; if the new decisive voter also exhibited a greater preference for spending on public education, we should observe an acceleration in expenditures as a result. Second, a series of empirical results document a greater preference of females for goods that enhance child welfare and for the provision of public goods, in general. In the intra-household context, a number of studies have shown an increased propensity of females to invest in the health and welfare of their own children, relative to their male counterparts.6 An increase in the financial resources of women relative to men consistently results in higher expenditures on goods benefitting the household’s children (food, clothing, and child care) at the expense of goods such as alcohol and tobacco. Welfare outcomes for children in the household also tend to rise with the mother’s financial resources. Anthropometric status, nutrition, and child survival rights have all been shown to increase with the mother’s income share.7 In addition to an increased propensity to invest financial resources in their own children, women also appear to prefer a larger presence of public goods in general and goods benefitting children (not their own) in particular. Focusing on women’s voting rights, rather than their financial resources, Lott & Kenny (1999) credit the enfranchisement of women with an increase in overall government expenditures and revenue in the early twentieth century United States. At the state level, however, they find no significant impact on specific components of government expenditure including social services and education. Miller (2008) demonstrates that suffrage and the increased voting power of women resulted in a sizable increase in local public health spending and a decrease in child mortality rates. This paper focuses specifically on the impact of female enfranchisement on the total provision of public education, including state and local spending. Intra-household, Pitt & Khandker (1998) document an increased propensity of both male and female children to attend school with a rise in female income among Bangladeshi households. But we know of no specific evidence that women exhibit a higher preference for the provision of public education than do men. If policymakers respond to the decisive voter as hypothesized, and if females are more willing 6 See Doepke & Tertilt (2012) for a summary of empirical findings. See Atkin (2009) and Duflo (2003) for the anthropometric results, Rubalcava et al. (2009) for nutritional status, and Thomas (1990) for child survival results. See Wolpin (1993) for a more complete summary of the literature on health outcomes of children. 7 5 to fund publicly provided goods than males, the enfranchisement of women should have increased government expenditures on public goods, just as Lott & Kenny (1999) have found. If the higher preference of women for public goods extends to education, we should observe an increase in public resources devoted to education at the same juncture. The following testable implications flow from the idea that females have higher preferences for public education, or at least that the public sector believes this to be so: 1. The enfranchisement of women in 1920 should have increased the size of local school districts’ budgets. 2. Gains in spending should have been higher where female voter power was higher. 3. The U.S. South in 1920 was characterized by rampant, informal disenfranchisement of black voters, both male and female. Thus, the Nineteenth Amendment granted access to the political process for white women much more so than for black. If female preferences for education provision were race-discriminating, spending on white schools should have risen more than spending on black schools.8 3 Data To measure the impact of suffrage on the local provision of education, we utilize a newly transcribed dataset of county-level black and white public schooling statistics between 1910 and 1940 for four Southern states: Alabama, Georgia, Louisiana, and South Carolina. Each state’s Department of Education or equivalent office published an annual or biennial report containing statistics on revenues and expenditures, disaggregated by county and often by race.9 These data and the data collection process are described in detail in the Data Appendix and in Carruthers & Wanamaker (2012). Transcribed data are assembled into a county-by-race panel describing the finances of each county in these states. Note that local educational expenditures include all spending from state and federal transfers, since local school districts channeled intergovernmental support for 8 By “race-discriminating” preference for education, we mean whether the preference for education provision among females is relatively higher for spending on their own race. This includes the possibility that women cared only for the welfare of their own children, who would almost certainly have been of the same Census-recorded race. 9 These reports also contain enrollment, attendance, teacher salaries, term lengths, and school counts. These additional measures of education investments are examined in Section 5. 6 public education. Although related work has found no significant impact of suffrage on state-level educational spending,10 school spending reported at the county level better captures the contribution of federal, state, and local spending combined.11 Section 6 shows that the state-level spending response to suffrage was small or negligible for these four states, implying that steep changes in total educational spending were predominantly driven by local decisions. The conceptual framework outlined above generates testable implications regarding the dosage of suffrage treatment across local areas. More granular population statistics than those available in published census volumes are needed in order to exploit inter-county variation in the size and composition of the female population. We populate decennial 1910-1940 age-by-race-by-gender cells for each county in the four-state sample using data from the genealogy website Ancestry.com, which contains a 100 percent count of each 1910-1940 U.S. Census. Intercensal population estimates are linearly interpolated. The education panel and decennial demographic data are matched to additional county-level variables from population and agricultural censuses taken in 1910, 1920, 1930, and 1940. Relevant statistics from these census reports include black-white population ratios, crop value per capita, and the percent of land devoted to agriculture. Annual measures of census variables are interpolated between census years to fully populate the panel. Lastly, demographic data for each county are matched to total voter counts for the 1920 general election (Clubb et al., 2006), the first where all U.S. females could participate. We use Bayesian methods with informative priors to infer race-bygender vote shares for each county. See the Data Appendix for details on this procedure. Together, these data provide a rich description of annual school expenditures in Southern segregated states before and after the Nineteenth Amendment, as well as a unique profile of genderby-race population and voter distributions in 1920. Table 1 lists descriptive statistics for school spending outcomes and other measures of public school resources, suffrage treatment measures, and Census controls. 10 Total educational spending was one of multiple outcomes analyzed by Lott & Kenny (1999) and Miller (2008). Miller (2008) and Lott & Kenny (1999) have state expenditures in all years surrounding 1920, but most education expenditures were locally sourced. They have local spending data for only four years between 1902 and 1927, aggregated up to the state level. 11 The vast majority of school districts in these data are county districts. Data reported at sub-county level (e.g., for city or township districts) are aggregated to the county level. 7 4 Empirical Strategy and Results The stylized facts relevant to this application are presented in Figures 1 - 3. Figure 1 plots percapita school spending by year for all of the United States, where a substantial increase in expenditures is evident in the years immediately following 1920. The same metric for the transcribed four-state sample of local school data is located in Figures 2 and 3. We consider 1921 (that is, the 1920-1921 school year) to be the first post-suffrage reporting year for these states. Most of the financial reports that form the basis of these data referred to academic years (e.g., 1919-1920). The Nineteenth Amendment was officially in place as of August 1920, well after funds were spent for the 1919-1920 school year. Though there may have been anticipatory effects on spending immediately prior to the Nineteenth Amendment, we are most interested in the discrete change in school spending and resource trends after policymakers faced a new electorate. Both trends were rising prior to the Nineteenth Amendment, but exhibit a much steeper climb after 1921. Note that these gains are not uniform across all schools. For each of the states in the sample, expenditures are reported separately for segregated black and white schools. Figure 3 plots per-capita school expenditures by race and shows that post-suffrage spending gains were largely limited to white schools. Spending on black schools climbed steadily throughout this period, but in contrast to spending on white schools, exhibited no observable obvious break in trend around 1921.12 We use an estimation strategy that detects significant changes in the levels as well as growth rates of school spending after 1920. Let Zt (t ∈ [1910, 1940]) represent suffrage “treatment,” a dummy variable equal to one in 1921 and later years. Tt is a time trend equal to the current year minus 1921(so that 1921 is year “0”). The estimating equation is: ln(Yct ) = aZt + bTt Zt + cTt2 Zt + dTt + eTt2 + θc + βXct + εct , (1) where Yct is a measure of public educational spending for county c at time t (in 1925 dollars) and εct is an error term. Xct is a matrix of county observable features updated decennially and θc is a county fixed effect. Variables in Xct include the value of crops per capita, the percent 12 Here, and throughout the text, per capita statistics are measured against the voting-age population of adults (i.e., the population age 21 and older) and race-specific per capita statistics carry a race-specific denominator rather than the total adult county population. 8 of land devoted to agriculture, the black-white population ratio, a cubic function of total county population, the adult female population share, and the white and black adult female population shares.13,14 Xct variables will help to control for non-linear agricultural and demographic trends that may have affected educational resources. For this and all estimating equations, robust standard errors are clustered within counties. In essence, Equation 1 is a deviation-from-trend analysis centered around 1921. The parameter a returns the immediate and discrete shift in educational spending and other outcomes following 1921, conditional on quadratic underlying trends in Yct leading up to suffrage and then following suffrage. The coefficients b and c identify changes in the growth rate of Yct values after 1921, relative to pre-1921 growth identified by coefficients d and e. That is, a, b, and c are not so much “suffrage” effects as “post-1921” effects, including suffrage. Accordingly, Equation 1 is a starting point and other testable implications of the conceptual framework will be used to further evaluate the hypothesis that public support for schools increased in the wake of women’s political enfranchisement. Table 2 Column (1) lists estimates for coefficients a − c in Equation 1, where total per capita school spending is the outcome of interest. Results indicate that immediately following suffrage, per capita spending on education experienced a level shift of 50.7 percent relative to the level of spending just prior to 1921. Although we might expected to see a dwindling effect of suffrage over time, as analysis by Moehling & Thomasson (2012) finds in the context of public health spending, the estimated coefficients on the post-1920 linear and quadratic terms are both positive in Column (1) of Table 2. That is, the new spending trajectory was significantly higher following 1921, by 0.55Tt + 0.006Tt2 percent for each Tt beyond 1921. Together, these point estimates indicate that 1921 marked a sharp break in school spending trends, coinciding with precisely the year that women’s suffrage was first enforced. Because suffrage was granted late in the 1920 calendar year, the regression specification above uses 1921 as the first year of suffrage “treatment.” Indeed the expenditure paths charted in Figures 1 - 3 seem to implicate 1921 as the break between two expenditure regimes. Still, we validate this assumption by re-estimating Equation 1 using other years as the breakpoint. For potential 13 These variables are interpolated between census years 1910, 1920, 1930 and 1940. Agricultural economic activity was a close substitute to schooling in rural areas, but Southern economies were shifting to a more industrial emphasis over this time period. 14 9 breakpoints k ∈ [1915 − 1925], the binary indicator Zt is replaced with a post-k dummy variable, and Tt is recalculated as the current year minus year k. Equation 1 is re-estimated for each k. Figure 4 plots the coefficient estimates for a, which represents the level break at each year k. Confidence intervals are drawn from robust standard errors. The coefficient a is statistically significant for all potential breakpoints, but negative until 1920. This means that between 1915 and 1919, deviations from underlying trends were predominantly downward. By contrast, the years 1920-1925 are all associated with positive and significant level shifts, with the peak coinciding with 1921. We emphasize that Figure 4 is not by itself direct evidence that suffrage was responsible for spending shifts in 1921 and later, but it supports the assertion that 1921 divided two school spending regimes, one declining and one rising. This is a necessary condition for the idea that the passage of the Nineteenth Amendment marked a sharp departure from earlier spending patterns. We conclude that a sizable and statistically significant shift in education expenditures occurred between 1920 and 1921. But it remains possible that the post-1920 jump in education expenditures identified by our empirical model is unrelated to women’s suffrage. Next, additional implications of the conceptual framework are utilized to discern whether a post-suffrage is part of the overall post-1920 effect. Specifically, two measures of female voter power are incorporated into the model to proxy for the dosage of suffrage that counties received. First one implication of the conceptual framework of electoral competition and updated constituent preferences is that there should be a steeper climb in education spending in counties where women represented a larger share of the electorate in that year if – as hypothesized – pre-suffrage spending was below the typical female voter’s preferred level of spending. We test whether the discrete shift in spending identified by coefficient a in Equation 1 is larger in counties where females constituted a larger share of the adult population. At the same time, we test whether post-1921 trends in spending identified by b and c varied by female population share. The proxy is valid as long as the female population share is not correlated with unobservables that would otherwise have affected the growth of expenditures after 1921. While the share of females may have been correlated with the size of the school-age population and, therefore, correlated with the growth in education expenditures over the entire period, there is no particular reason that school expenditures should have risen more after 1921 than before in heavily female counties. In other words, while the percent female may have been correlated with the level 10 and even the growth of education expenditures over the entire period, an increase in expenditures relative to trend after 1921 in counties with a higher female share of the electorate is more difficult to attribute to causes other than suffrage itself.15 A second proxy for the strength of suffrage dosage is motivated by the idea that spending gains will be steeper in counties with relatively higher female turnout rates in the 1920 election, although an important caution is the possibility that there may be a common, unobservable, cause of both female turnout and education expenditures. We test whether 1921 intercept shifts and post-1921 spending trends vary by Bayesian estimates of female turnout (i.e., the share of 1920 voters who were female). Since turnout statistics are estimates, we compute bootstrapped standard errors for regressions using turnout as suffrage proxies. Mechanically, these two expectations are tested by introducing an additional variable, VctF , to Equation 1 and fully interacting VcF with pre-1921 and post-1921 trend terms. VctF is either the percentage of the electorate (age 21 and older) in year t that was female or the estimated percentage of 1920 election voters who were female. Vcf percentages are interacted with Zt and pre-1920 and post-1920 quadratic growth terms in Equation 1. This modification of Equation 1 allows the discrete change in outcomes at the 1921 breakpoint as well as underlying trends in each outcome to vary by the dosage of suffrage, as proxied by variation in demographics or female voter turnout. The new estimating equation is as follows: ln(Yct ) = aZt + ãZt VctF + bTt Zt + b̃Tt Zt VctF + cZt Tt2 + c̃Zt Tt2 VctF ˜ t V F + eT 2 + ẽT 2 V F + θc + βXct + εct +dTt + dT ct t t ct (2) The results of these interactions are located in Columns (2) and (3) of Table 2. The coefficient on the post-1920 variable interacted with female population share (Y ear ≥ 1921 ∗ VcF ) is statistically significant and carries quantitative import. Recall that the effect of Y ear ≥ 1921 measures the baseline 1921 local deviation from trend and that the effect of Y ear ≥ 1921 ∗ VcF measures the additional deviation, due to marginally higher female population shares. Consistent with a causal role for suffrage in explaining post-1921 expenditure increases, this coefficient indi15 It is still possible that females sorted into high-expenditure districts so that female shares and the growth of expenditures are positively related due to migration. But this would have been true prior to 1921 as well, and continuous sorting patterns do not not explain the discrete break in spending trends at 1921. In addition, fixing the electorate share variable at its 1920 value in Equation 1 gives qualitatively similar results. 11 cates that each percentage point gain in the share of adults who were female in a county increased 1921 expenditures by 2.6 percent. Put in other words, a one-standard deviation increase in female population shares (2.33 percentage points) is associated with a 6.1 percent increase in per-capita school spending. At the mean female population share, Column (2) indicates that spending trends shifted by 51.7 percent, close to the baseline 50.7 percent estimate in Column (1).16 Equation 2 permits inference about the short-term changes in school spending at 1921 as well as the change in post-1921 trends. Although Column (1) points to a steady increase in spending trajectories after 1921, Column (2) does not attribute any of these longer-term gains to a stronger dosage of suffrage in the form of higher female population shares. Column (3) of Table 2 indicates that female voter turnout is less powerful in predicting post1921 spending patterns in these four states. There are a number of potential explanations for this difference. First, politicians may have responded to the threat of female voters rather than the voters themselves (Moehling & Thomasson, 2012). In the present context, it may have been that politicians were cognizant of the fact that women could vote if they liked, and acted to appease these voters just as they would if they had actually appeared at the polls. The Column (2) finding that post-1921 trends did not vary by female population shares supports this possibility, in that the threat of a female voting bloc would have been more powerful in the years immediately following suffrage. Second, note that our estimates of female voter turnout are somewhat noisy, and standard errors in Column (3) are bootstrapped. We are, therefore, not surprised to observe a less tangible impact relative to the more precisely calculated female population shares in Column (2). As a final test of the causal role played by suffrage in increasing school expenditures, we measure the change in spending after 1920 for black schools and white schools separately. If changes in school spending overall were in fact attributable to changes in the electorate around the Nineteenth Amendment, white schools should have benefitted from women’s suffrage by more than black schools, given the obstacles black voters faced in exercising a meaningful voice in the political process. That is to say, the decisive voter better represented women’s preferences following the Nineteenth Amendment, but black men and women were no more enfranchised than they were before 1920.17 We replace per capita total expenditures as the dependent variable in Mean female population shares were 49.8 percent, and so the calculation is −0.778 + 49.8 ∗ 0.026 = 0.517. Official turnout statistics by race and sex are not available for this period, but we estimate both black male and black female turnout in the 1920 presidential election at 1% while white female turnout was 22 percent and white male 16 17 12 Equations 1 and 2 with separate measures for per capita white and black education expenditures separately.18 Indeed, estimated coefficients in Columns (1) and (3) of Table 3 indicate that white school spending gained 47.5 percent in 1921 relative to its previous trend, while black spending increased by 32.4 percent. If none of the post-1920 increase in black spending is attributable to suffrage, and if the trajectory of black spending is an adequate counterfactual to the trajectory of white spending, the implication is that women’s suffrage was nevertheless responsible for a large share of the overall rise in white education spending after and including 1921. Just as with the overall expenditure estimates, changes in white spending levels (but not black) were in part determined by the percentage of the electorate who were female, per Column (2) and (4) of Table 3. The striking finding that white schools benefitted far more from suffrage than did black schools suggests that the ratio of black to white school expenditures, the so-called education “gap,” suffered at the hands of the 19th Amendment. A direct test of the post-1921 change in the ratio of black to white per pupil spending (Column (5) of Table 3) shows a 6.6-point decline in the black-white spending ratio as of 1921. Coefficients on post-1921 quadratic terms suggest that the black-white spending ratio continued declining for five years following 1921 before turning upward again. Indeed, the average black-white expenditure gap did not return to its 1920 level (0.248) until 1937. 5 The White-Black Gap in School Resources The assembled data on school resources includes more than expenditures, allowing for the identification of likely mechanisms for higher post-1921 spending. Equation 1 is applied to counties’ average term length (in days), the number of teachers per 1,000 adults, and average teacher salaries (in real 1925 dollars). Each school resource is reported separately for black and white schools. Results are listed in Table 4 and depicted in Figures 5 - 7. Columns (1) and (2) describe trends in white and black term lengths, respectively. White terms did not increase significantly in 1921, and conditional on other covariates in Equation 1, white terms followed a convex path after 1921. Black term lengths also appear to have followed a convex path thereafter, following a 3.2-day turnout was 33 percent. 18 Black expenditures are per over-21 black population and white expenditures are per over-21 white population. We do not include per pupil calculations as enrollment, itself, may have been affected by suffrage. 13 upward shift that coincides with women’s suffrage. Turning to Figure 5, however, it is not clear that 1921 marks a break from underlying upward trends in white or black term lengths. Column (3) indicates that the ranks of white teachers grew by 0.932 per 1,000 white adults in 1921 (6 percent of the mean), and Figure 6 depicts a clear break from pre-suffrage trends. And though Column (4) does not point to a similarly dramatic 1921 shift in the number of black teachers, post-1921 trends in black teachers were significantly upward. Indeed, Figure 6 indicates that 1921 divided periods of declining and rising employment for black teachers. Column (5) of Table 4 indicates that white teacher salaries increased discretely by 35.2 percent in 1921, relative to years just before 1921, and furthermore, that white teacher salaries continued to climb relative to pre1921 trends. In much the same way that black teachers per capita exhibited a U-shape trend around 1921, Column (6) indicates that black teacher salaries reversed several years of downward trends and rose significantly after 1921. Specifically, Column (6) and Figure 7 suggest that black salaries did not rise discontinuously at the 1921 threshold (rather, they fell 24.5 percent), but that black salaries increased at nearly the same pace as white salaries after 1921. Collectively, these findings illustrate some of the mechanisms of discrete post-suffrage spending gains. School systems did not lengthen the school year more than they would have in the absence of women’s suffrage, but they did hire more teachers and raise teacher salaries.19 Comparing the counterfactual level of resource inequality after 1910 to what is actually observed, we estimate that, although absolute spending levels would have been substantially lower after 1920, the resource gap would have also been less stark. 6 Extension: State Spending After Suffrage The analysis in this paper is limited to the provision of local education resources, as reported to state Departments of Education, in four Southern states. The sample is thus limited in its ability to address (1) the allocation of state and local spending gains and (2) the response of states other than the four included in our panel. An inability to identify state responsiveness to suffrage is a mild limitation given the dominant 19 Unreported analyses estimate Equation 2 for these school resources and find that variation in adult female population shares is significantly associated with even larger 1921 shifts in term length (white), teachers per capita (black), and average teacher salaries (white and black). 14 role that local leaders and school districts played in collecting revenues for schools and allocating resources. Recall that school spending data in states’ annual reports includes any spending out of state appropriations, so the previous section identifies the response of total education spending to an expanded electorate. We might expect that local spending decisions are driving any post-suffrage trends because of the decentralization of both education decision-making bodies and revenue collections. Lott & Kenny (1999) and Miller (2008) find no significant change in state education expenditures in the wake of female suffrage despite large increases in overall public spending by states. If the same is true for the four Southern states that form the basis of our analysis, then the spike in educational spending following the Nineteenth Amendment was a predominantly local response, in line with conceptual expectations. We employ the same state-level finance data and socioeconomic control variables used by Lott & Kenny (1999) and Miller (2008) to replicate previous work estimating the effect of suffrage on state-level educational spending.20 We then go on to test whether the four sampled states reacted fundamentally different than the nation at large. This exercise serves two purposes: first, to infer whether principle results are driven by a state-level response rather than a local response, and second, to examine the subsample’s representativeness relative to the rest of the country. The four Southern states included in the analysis may have been systematically different in a way that limits the ability to extrapolate results to the nation. To situate results with those of Lott & Kenny (1999) and Miller (2008), we estimate a fixed effects specification informed by both studies: ln(Yst ) = α + SU F F RAGEst β1 + SU F F RAGEst ∗ F OU RST AT Es β2 + Xst ψ +θs + α(t) + θs ∗ α(t) + θt + εst , (3) where ln(Yst ) is the natural log of states’ per-capita education expenditures, SU F F RAGEst is a measure of suffrage treatment, F OU RST AT Es is an indicator equal to one for Alabama, Georgia, Louisiana, and South Carolina, and Xst is a matrix of socioeconomic controls.21 The parameter 20 We thank Larry Kenny for graciously providing these data and associated documentation. State expenditure and revenue data from 1870-1915 were originally provided to Lott and Kenny by John Wallis. 21 Controls include the presence of a literacy test, secret ballot indicator, number of motor vehicle registrations, log population density, fraction of the population rural, fraction of the population black, fraction of the population older than 65, fraction of women working, fraction of the population illiterate, fraction of the labor force in manufacturing, fraction of the population foreign-born, and the average real manufacturing wage. 15 α(t) controls for linear trends common to all states, θs is a state fixed effect, and θs ∗ α(t) controls for state-specific trends. Specifications with and without time trend controls (i.e., (α(t)+θs ∗α(t))) are estimated. Equation 3 additionally controls for year fixed effects (θt ) since cross-state variation in the timing of women’s suffrage is not collinear with any one year fixed effect. There are two measures of suffrage treatment. The first follows Lott & Kenny (1999) and defines SU F F RAGEst , that is, the “dosage” of suffrage, to be the product of the number of years since suffrage was implemented and the share of adults who are female in a given year. Second, following Miller (2008), SU F F RAGEst is defined to be a simple binary indicator equal to one in states and years with women’s suffrage. Table 5 summarizes suffrage and summary statistics for the nation and the four-state sample. Clearly, the four Southern states exhibited lower spending than the rest of the country, and they had less exposure to suffrage rights between 1870 and 1940. Although the South may not be representative in terms of the levels of spending outcomes, Equation 3 tests whether they were fundamentally different in terms of the trajectory of spending after suffrage. Equation 3 is estimated with and without an interaction between the SU F F RAGEst variable and the F OU RST AT ES indicator. Results are reported in Table 6. Columns (1) and (3) report the estimated coefficients for two different proxies of SU F F RAGEst without the interaction for our sample states. The Lott & Kenny (1999) replication in Column (1) measures no statistically significant increase in expenditures while the Column (3) specification attributes to female suffrage a marginally significant 14.3 percent gain in educational spending, relative to linear statespecific time trends. The second coefficients of Columns (2) and (4) measure the difference in post-suffrage spending in the four-state subsample, relative to the baseline impact in other states. Neither interaction is statistically significant, indicating that state educational spending in these four states responded no differently to suffrage than the rest of the nation. In 1919, on the eve of the Nineteenth Amendment, state spending in the four-state subsample averaged $2.22 per capita, and Table 6 Column (4) indicates an increase of no more than 15.1 percent, or $0.34, following the extension of female voting rights.22 At the same time, total spending as recorded by local governments averaged $5.34 per capita, and per Table 2 Column (1), increased 50.7 percent, or $2.71, following the Nineteenth Amendment. Thus, the local response to suffrage 22 The confidence interval for the Column (4) interaction is low enough to render the coefficient combination β̂1 + β̂2 statistically insignificant and close to zero. Thus, a conservative interpretation of Column (4) is that suffrage had no impact on state education spending in the four states of interest, and that 15.1 percent is in the upper range of estimates. 16 dominated any state response, as anticipated. 7 Conclusion In the four Southeastern states we examine, county-level school spending averaged $331,418 (in 1983 dollars) over the years 1920-1940, 3.8 times higher than average spending in 19101919. Controlling for underlying trends in educational spending and decennial shifts in agriculture economies and demographics, our first-pass estimates of the effect of suffrage suggest that extending the voting franchise increased education spending by 50.7 log points in 1921 and that the effect persisted, and likely grew, over time. This increase represents 37 percent of the unconditional gain in spending after 1920. Our estimates of the impact of women’s suffrage on the local provision of education does not appear to be entirely related to unobserved shocks unique to 1920. Spending gains were predominantly local, higher in white schools, and higher in counties with higher female representation in the electorate around 1920. These observations are consistent with conceptual expectations about the localized political economy response to asymmetric voter enfranchisement. Findings confirm expectations from economic models of intra-household bargaining and political economy. The reaction of public finance allocations to the extension of women’s voting rights provides strong support for the idea that suffrage shifted and increased the pivotal voter’s preferences for public education. The adoption of women’s suffrage in the United States and the subsequent impact of suffrage on public education represents a historic episode that should shape our expectations for the relationship between women’s rights and human capital accumulation in modern developing countries. As women gain electoral power, resources for public schools improve. Card & Kreuger (1992a) show that for men born after 1920, gains in school quality led to higher adult earnings. Our results suggest that part of these human capital gains should be attributed to newly enfranchised women voters. References Atkin, D. (2009). Working for the future: Female factory work and child health in mexico. Working Paper. 17 Baumgardner, J. R. (1993). Tests of median voter and political support maximization models: the case of federal/state welfare programs. Public Finance Review, 21(1), 48–83. Black, D. (1948). On the rationale of group decision making. Journal of Political Economy, 56(1), 23–34. Bowen, H. R. (1943). The interpretation of voting in the allocation of economic resources. Quarterly Journal of Economics, 58(1), 27–48. Brunner, E. J., & Ross, S. L. (2010). Is the median voter decisive? evidence from referenda voting patterns. Journal of Public Economics, 94(11-12), 898 – 910. Card, D., & Kreuger, A. B. (1992a). Does School Quality Matter? Returns to Education and the Characteristics of Public Schools in the United States. The Journal of Political Economy, 100, 1–40. Card, D., & Kreuger, A. B. (1992b). School Quality and Black-White Relative Earnings: A Direct Assessment. The Quarterly Journal of Economics, 107. Carruthers, C. K., & Wanamaker, M. H. (2012). Closing the gap? The effect of private philanthropy on the provision of African-American schooling in the U.S. South. Working Paper. Carter, S. B., Gartner, S. S., Haines, M. R., Olmstead, A. L., Sutch, R., & Wright, G. (Eds.) (2006). Historical Statistics of the United States: Earliest Times to the Present. Cambridge University Press. Clubb, J. M., Flanigan, W. H., & Zingale, N. H. (2006). Electoral Data for Counties in the United States: Presidential and Congressional Races, 1840-1972. ICPSR08611-v1. Ann Arbor, MI: Inter-university Consortium for Political and Social Research. Corder, J. K., & Wolbrecht, C. (2004). Using Prior Information to Aid Ecological Inference: A Bayesian Approach. In G. King, O. Rosen, & M. A. Tanner (Eds.) Ecological Inference: New Methodological Strategies, (pp. 144–161). Cambridge University Press. Croson, R., & Gneezy, U. (2009). Gender differences in preferences. Journal of Economic Literature, 47(2), 1–27. 18 Doepke, M., & Tertilt, M. (2012). Does female empowerment promote economic development? Working Paper. Donohue, J. J., Heckman, J. J., & Todd, P. E. (2002). The Schooling of Southern Blacks: The Roles of Legal Activism and Private Philanthropy, 1910-1960. Quarterly Journal of Economics, 117, 225–268. Duflo, E. (2003). Grandmothers and granddaughters: Old-age pensions and intrahousehold allocation in south africa. World Bank Economic Review, 17(1), 1–25. Eckel, C. C., & Grossman, P. J. (1996). Altruism in anonymous dictator games. Games and Economic Behavior, 16, 181–191. Goldin, C. (2001). The human capital century and american leadership: Virtues of the past. Journal of Economic History, 61(2), 263–92. Goldin, C., & Katz, L. F. (2009). Why the united stated led in education: Lessons from secondary school expansion, 1910 to 1940. In D. Eltis, F. D. Lewis, & K. L. Sokoloff (Eds.) Human Capital and Institutions: A Long-Run View. Cambridge University Press. Harper, I. H. (Ed.) (1922). The History of Woman Suffrage: 1900-1920, vol. 5. New York, NY: National American Woman Suffrage Association. Husted, T. A., & Kenny, L. W. (1997). The effect of the expansion of the voting franchise on the size of government. Journal of Political Economy, 105(1), 54–82. Li, S. X., Eckel, C. C., Grossman, P. J., & Brown, T. L. (2011). Giving to government: Voluntary taxation in the lab. Journal of Public Economics, 95(11-12), 1190–1201. Lott, J. R., & Kenny, L. W. (1999). Did women’s suffrage change the size and scope of government? Journal of Political Economy, 107, 1163–1198. Miller, G. (2008). Women’s suffrage, political responsiveness, and child survival in american history. Quarterly Journal of Economics, 123(3), 1287 – 1327. 19 Moehling, C. M., & Thomasson, M. A. (2012). The political economy of saving mothers and babies: The politics of state participation in the sheppard-towner program. Journal of Economic History, 72(1), 75–103. Pitt, M. M., & Khandker, S. R. (1998). The impact of group-based credit programs on poor households in bangladesh: Does the gender of participants matter? Journal of Political Economy, 106(5), 958–996. Rubalcava, L., Turuel, G., & Thomas, D. (2009). Investments, time preferences, and public transfers paid to women. Economic Development and Cultural Change, 57(3), 507–538. Thomas, D. (1990). Intra-household resource allocation: An inferential approach. The Journal of Human Resources, 25(4), 635–664. Wolpin, K. I. (1993). Determinants and consequences of the mortality and health of infants and children. In Handbook of Population and Family Economics, Volume 1, (pp. 483–557). Elsevier. 20 A Data Appendix A.1 School Finance and Resource Data This study makes use of data on Southern public school districts between 1910 and 1940 in four states (Alabama, Georgia, Louisiana, and South Carolina), including statistics on schools, teachers, students, and expenditures. These states were selected for their consistency in reporting the educational resources of interest. While several researchers have used portions of these data for specific projects, to our knowledge, our assembled dataset is unprecedented in its size, scope, and depth. We have already used these data to estimate the impact of philanthropically-funded Rosenwald schools on public support for segregated schools (Carruthers & Wanamaker, 2012), and we expect these data to serve as a valuable foundation for future research. This section describes the data in more detail and illustrates trends in white and African-American schooling resources in the early 20th century. Our primary sources of education input data are annual reports from state Superintendent Offices, Departments of Education, or equivalent governmental units. Measures of schooling resources reported separately for white and African-American schools typically include: 1. Enrollment, average daily attendance 2. Number of teachers overall 3. Expenditures 4. Teacher salaries 5. Number of schools 6. Average term lengths 7. School revenues drawn from local taxes We outsourced transcription of available statistics for these four states and assembled countyby-race panels for the years 1910-1940. Data availability is remarkably consistent across states and years. We conducted an informed 0.5 percent audit of each transcribed variable. Specifically, for each school statistic and each state, we regressed transcribed data against county fixed effects and a quadratic function of time, generating predicted values and residuals. We flag cells in the top 99.5 percent of residuals, in absolute value. This resulted in 814 flagged cells, for 16 variables and just 21 over 11,000 county-by-race rows of raw data. Then, our research assistant verified the accuracy of each flagged cell by consulting the original scanned reports and fixed any discovered errors. The realized error rate from these flags was 14.9 percent. We believe this to be an encouraging signal of the underlying fidelity of these data, considering that our audit focused on the top 0.5 percent of outliers within counties’ time series. Other researchers have used portions of these data or statewide aggregations of historic education data to characterize pre-war school resources. We construct summary statistics comparable to those that have been reported in earlier work. For instance, observe that Card & Kreuger (1992b) assemble data on pupil-teacher ratios, annual teacher pay, and term length for black and white schools from eighteen state-level reports, including the five states in our sample. Statistics in our data closely track theirs. Card & Kreuger (1992b) document a convergence in white-black term length ratio of 1.31 (1915) to 1.10 (1940) in eighteen segregated states, weighted by enrollment. We find a convergence of 1.40 to 1.09 over the same period in our five Southern states, weighted by total black and white enrollment. In related work, Card & Kreuger (1992a) document statewide aggregates of pupil-teacher ratios and term lengths using the U.S. Department of Education series Biennial Survey of Education. The earliest cohort they consider attended school between 1926 and 1947. We use our transcribed data on enrollment, teachers, and term lengths to construct statewide average values of pupil-teacher ratios and term lengths over the years 1926-1940. Summary statistics are similar to those reported by Card & Kreuger (1992a). For Alabama, Georgia, Louisiana, and the Carolinas, Card & Kreuger (1992a) report pupil-teacher ratios in the range of 33.7 - 35.9 and term lengths in the range of 154 - 161 (see their Table 1 on p. 12, first and fourth columns). For these same states and a somewhat earlier window of time, we find pupil-teacher ratios of 32.2 - 37.7 and term lengths of 148.9 - 158.3. In addition, Donohue et al. (2002) utilize the same series of Georgia Department of Education statistics that we use here, and their panel of county statistics spans 1911 to 1960. Our transcribed data from 1910 - 1940 match theirs exactly.23 A.2 Voter Turnout Data This study makes use of voter turnout estimates by race and gender to estimate the impact of females’ electoral participation on school spending and resource trends. Precise data on the gender 23 We thank the authors for providing a copy of these data. 22 and race of 1920 voters are not available at the county level. Instead, we have county-level counts of total votes by political party for every general election during this period. We use Bayesian methods with informative priors to infer race-by-gender vote shares for each county, relying on total turnout statistics as well as demographic data from the U.S. Census. For each county i, we observe the number of voters Vi , the adult population Ni , and the gender and race composition xji , where j = 1, ..., 6 indicates female, male, black-female, black-male, white-female and white-male groups, respectively. We start by modeling the number of voters Vi as a draw from a binomial distribution: Vi ∼ bin (zi , Ni ) , (4) where zi is the probability that an individual votes. That probability in turn is the weighted sum of the group-specific probabilities. It becomes: zi = x3i p3i + x4i p4i + x5i p5i + x6i p6i, (5) where p3i , p4i , p5i and p6i represent the probability that a black-female, black-male, white-female and white-male votes, respectively. Our goal is to find the posterior distributions that characterize pji for j = 3, ..., 6 for the 1920 election. We follow Corder & Wolbrecht (2004) and approximate the logistic transformations θji of the group specific probabilities pji with normal distributions. That is, pji = (1+exp(θji )) exp(θji ) and θji ∼ norm (µj , τj ), where µj is the mean of the distribution and τj is its precision. We use uninformative uniform distributions as priors on µ3 and µ5 (female groups). For µ4 and µ6 (male groups), we choose normal prior distributions that reflect the turnouts observed for the 1916 presidential election, for which women did not vote. We assume gamma distributions for the priors on the precisions τj for all groups. Equations 4 and 5, along with the specified prior distributions complete the model. Figure 1 shows a graphical representation of the model. The model is solved through numerical simulations. Specifically, we use Metropolis-Hasting algorithms and Markov Chain Monte Carlo (MCMC) techniques. The posterior distributions of p3i , p4i , p5i and p6i are obtained based on the data (Vi , Ni and xji ) and MCMC draws from the prior distributions of µj and τj . The point estimates of the group specific turnouts are obtained by sampling from their respective posterior 23 distributions. For each MCMC draw, we also compute the mean of pji across counties. The results yield the aggregate posterior distributions of pj . The aggregate point estimates of pj are obtained by sampling from the aggregate posterior distributions. 24 Tables and Figures 40 60 80 100 120 140 F IGURE 1: Nationwide trends in per capita public educational expenditures (1982-1984 dollars) 1910 1920 1930 year Per capita public educational expenditures Source: Authors’ calculations, Carter et al. (2006). 25 1940 0 Per capita public educational spending 50 100 150 F IGURE 2: Per capita educational spending in four Southern states, 1910-1940 1910 1920 1930 1940 year Source: Authors’ calculations and annual reports of states’ Department of Education or equivalent office in AL, GA, SC, and LA. 26 0 10 20 30 40 F IGURE 3: Per capita spending in four Southern states by race, 1910-1940 1910 1920 1930 1940 year White spending per white adult Black spending per black adult White spending trend, quadratic fit Black spending trend, quadratic fit Source: Authors’ calculations and numerous annual reports of states’ Department of Education or equivalent office. 27 TABLE 1: Summary statistics: county schooling expenditures and female populations, 1910-1940 (1) (2) Dependent variables Mean Standard deviation ln(per-adult educational expenditures) ln(per-adult white educational expenditures) ln(per-adult black educational expenditures) ratio of black to white per-adult expenditures White term length (days) Black term length (days) White teachers per thousand white adults Black teachers per thousand black adults ln(average white teacher salary) ln(average black teacher salary) 2.371 2.742 1.030 0.246 146.270 111.996 15.207 11.255 6.494 5.281 (0.661) (0.641) (0.707) (0.621) (27.237) (29.915) (4.712) (8.274) (0.458) (0.785) Suffrage treatment measures Mean Zt (year≥ 1921) Percent of 1921 electorate female Percent of 1921 voters female 0.684 49.80 28.50 Socioeconomic control variables Mean Crop value per capita Percent of land devoted to agriculture Black-white ratio Population (in thousands) 111.686 0.618 1.052 0.031 n (county-years) Standard deviation (0.465) (2.335) (6.332) Standard deviation (54.846) (0.185) (0.998) (0.039) 8,029 Source: Authors’ calculations and numerous annual reports of four Southern states’ Department of Education or equivalent office. 28 TABLE 2: Estimated changes in local educational spending per capita, after the Nineteenth Amendment Proxy for VcF Percent of Electorate Female Percent of Voters Female (2) (3) (1) Year ≥ 1921 0.507*** (0.025) -0.778** (0.376) 0.680*** (0.194) Year ≥ 1921*(t - 1921) 0.155*** (0.010) 0.163 (0.218) 0.175*** (0.065) Year ≥ 1921*(t - 1921)2 0.006*** (0.001) 0.014 (0.018) 0.007* (0.004) Year ≥ 1921 * VcF 0.026*** (0.008) -0.006 (0.006) Year ≥ 1921*(t - 1921)* VcF -2.2E-04 (0.004) -0.001 (0.002) Year ≥ 1921*(t - 1921)2 *VcF -1.7E-4 (3.6E-4) -6.0E-5 (1.2E-4) 8,029 295 0.70 8,029 295 0.70 Observations Number of counties Adjusted R-squared 8,029 295 0.70 Notes: Coefficient estimates of Equation 1 with per capita expenditures (in natural logs) as the dependent variable. Robust standard errors are in parentheses below each coefficient. Standard errors are estimated by bootstrap for Column (3) results. *** indicates statistical significance at 99% confidence (with respect to zero), ** at 95%, and * at 90%. 29 −.4 −.2 0 .2 .4 .6 F IGURE 4 1915 1917 1919 1921 1923 1925 year discrete break from spending trends, by year 95 percent confidence interval Source: Authors’ calculations. Coefficient estimates of Equation 1 with per capita expenditures (in natural logs) as the dependent variable and different years standing in for the break point. 30 TABLE 3: Estimated changes in local educational spending after the Nineteenth Amendment, by race, and to the racial gap in education funding (1) Outcome (2) Log white spending (3) (4) (5) Log black spending Ratio of black spending to white spending Year ≥ 1921 0.475*** (0.030) -0.804* (0.441) 0.324*** (0.026) -0.058 (0.561) -0.066*** (0.025) Year ≥ 1921*(t - 1921) 0.144*** (0.012) 0.153 (0.220) 0.070*** (0.013) 0.029 (0.465) -0.040*** (0.014) Year ≥ 1921*(t - 1921)2 0.006*** (0.001) 0.017 (0.021) 0.003*** (0.001) 0.045* (0.026) -0.004** (0.002) Year ≥ 1921 * VcF 0.026*** (0.009) 0.008 (0.011) Year ≥ 1921*(t - 1921)* VcF -2.2E-4 (0.004) 0.001 (0.009) Year ≥ 1921*(t - 1921)2 *VcF -2.3E-4 (4.2E-4) -0.001 (0.001) Observations Number of counties Adjusted R-squared 8,029 295 0.65 8,029 295 0.65 8,029 295 0.59 8,029 295 0.60 8,029 295 0.19 Notes: Coefficient estimates of Equation 1 for black and white school per capita school expenditures (in natural logs), separately, and for the ratio of black to white per pupil expenditures (in natural log). Robust standard errors are in parentheses below each coefficient. *** indicates statistical significance at 99% confidence (with respect to zero), ** at 95%, and * at 90%. 31 TABLE 4: Estimated changes in local segregated educational resources after the Nineteenth Amendment (1) Outcome (2) Term length (3) (4) (5) (6) White AfricanAmerican Number of teachers AfricanWhite American Log average teacher salaries AfricanWhite American Year ≥ 1921 1.618 (1.216) 3.208*** (1.218) 0.932*** (0.151) 0.030 (0.229) 0.352*** (0.021) -0.245*** (0.083) Year ≥ 1921*(t - 1921) -1.754** (0.745) -3.060*** (0.717) 0.225 (0.157) 0.366** (0.156) 0.128*** (0.010) 0.121*** (0.013) Year ≥ 1921*(t - 1921)2 -0.279*** (0.056) -0.161*** (0.054) -0.018*** (0.007) -0.016 (0.012) 0.005*** (0.001) 0.007*** (0.001) Observations Number of counties Adjusted R-squared 6,498 294 0.49 6,498 294 0.50 8,026 295 0.20 8,026 295 0.10 6,497 295 0.49 6,497 295 0.32 Notes: Coefficient estimates of Equation 1 for school resources with no interactions with VcF . Standard errors are in parentheses below each coefficient. ** indicates statistical significance at 99% confidence (with respect to zero), ** at 95%, and * at 90%. 32 F IGURE 5: Term length by race, 1910-1940 80 120 130 100 140 120 150 140 160 160 Black term length (days) 170 White term length (days) 1910 1920 1930 1940 1910 1920 year n = 6498 county−years 1930 1940 year Population−weighted average Quadratic trend Population−weighted average Quadratic trend n = 6498 county−years Source: Authors’ calculations and numerous annual reports of states’ Department of Education or equivalent office. 33 F IGURE 6: Teachers per thousand adults by race, 1910-1940 8 13 14 10 15 12 16 14 17 16 Black teachers per-capita (000s) 18 White teachers per-capita (000s) 1910 1920 1930 1940 1910 1920 year n = 8026 county−years 1930 1940 year Population−weighted average Quadratic trend Population−weighted average Quadratic trend n = 8026 county−years Source: Authors’ calculations and numerous annual reports of states’ Department of Education or equivalent office. 34 F IGURE 7: Average teacher salary by race, 1910-1940 Black average teacher salary 300 200 100 400 600 800 400 1000 500 1200 White average teacher salary 1910 1920 1930 1940 1910 1920 year n = 6572 county−years 1930 1940 year Population−weighted average Quadratic trend Population−weighted average Quadratic trend n = 6520 county−years Source: Authors’ calculations and numerous annual reports of states’ Department of Education or equivalent office. 35 TABLE 5: Spending and suffrage summary statistics, four-state sample versus all states (1870 - 1940) (1) (2) All states Four-state subsample ln(total expenditures) 2.816 (0.963) 2.212 (0.979) ln(education expenditures) 1.312 (1.212) 0.949 (1.311) Y EARSIN CEst ∗ pctf em 0.200 (0.238) 0.164 (0.237) Female suffrage (0,1) 0.415 (0.493) 0.326 (0.470) n (state-years) 1,882 161 Variable Source: Authors’ calculations and Lott & Kenny (1999). Standard deviations are in parentheses. Y EARSIN CEst is equal to the number of since elapsed since full suffrage. pctf em is the share of the adult population that is female. SU F Fst is an indicator for full suffrage in state s in year t, which is achieved for all states by 1920. 36 TABLE 6: Estimated changes in state education expenditures after women’s suffrage, four-state sample versus all states (1) Suffrage proxy Time trend controls SU F F RAGEst Y EARSIN CEst *pctfem No -0.132 (-0.178) SU F F RAGEst ∗ F OU RST AT Es Observations Adjusted R-squared (2) No -0.136 (-0.180) (3) SU F Fst Yes Yes 0.143* (-0.086) 0.151* (-0.086) 0.049 (0.350) 1,827 0.71 1,827 0.71 (4) -0.136 (0.212) 1,827 0.78 1,827 0.78 Notes: Selected coefficient estimates from Equation 3 for state-level public expenditures. Y EARSIN CEst is equal tot he number of since elapsed since full suffrage. pctf em is the share of the adult population that is female. SU F Fst is an indicator for full suffrage in state s in year t, which is achieved for all states by 1920. Time trend controls (α(t) + θs ∗ α(t) in Equation 3) are excluded in Columns (1) and (2). Standard errors are in parentheses below each coefficient. Additional controls include socioeconomic variables listed in the text, state fixed effects, and year fixed effects. *** indicates statistical significance at 99% confidence (with respect to zero), ** at 95%, and * at 90%. 37 F IGURE 8: Graphical representation of the model. Rectangular nodes indicate observed variables. Prior distributions for mu3, tau3, mu5 and tau5 (female groups) are uninformative. Priors for mu4, tau4, mu6 and tau6 (male groups) are based on 1916 vote turnouts. 38
© Copyright 2026 Paperzz