America's Graduation from High School: TheEvolutionand Spreadof Secondary Schooling in the Twentieth Century CLAUDIAGOLDIN Secondary-school enrollmentand graduation ratesincreasedspectacularly much in of the United States from 1910 to 1940; the advancewas particularly rapidfrom 1920 to 1935 in the nonsouthern states.This increasewas uniquelyAmerican;no othernationunderwent equivalent an changefor severaldecades.Statesthatrapidly expanded theirhigh school enrollments earlyin the periodhad greaterwealth,more homogeneity of wealth, and less manufacturingactivity than others. Factors promptingthe expansioninclude the substantialreturnsto educationearly in the centuryand a responsive"state." This work is based on a newly constructed statelevel dataset. American economicgrowthcomes fromthe now-familiar decomposition of the growthresidual. From1929to 1982nationalincomeper worker grew annually at a 1.48 percent average. Conventionalfactor growth (changesin laborhours,physicalcapitalperworker)can accountfor a mere 5 percent,leavingthe remainder residual.Of thatresidual,accordingto as EdwardDenison, 28 percentcan be explainedby increasesin years and quality of formal educationper Americanworker.'Because much of the remainingresidualis likely to have originatedin advancesin knowledge The Journal of Economic History. Vol. 58, No. 2 (June 1998). C The Economic History Association. All rights reserved. ISSN 0022-0507. Claudia Goldin is Professor, Department of Economics, Harvard University, 215 Littauer Bldg., Cambridge, MA 02138 and Research Associate and Program Director, National Bureau of Economic Research. This research was supportedby The Brookings Institution,the National Science Foundation, and the Spencer Foundation (Grant 199600128). The data presented, the statements made, and the views expressed are solely the responsibility of the author. Linda Tuch provided extraordinary research assistance on the project when I was at The Brookings Institution. Jessica Wolpaw very ably assisted with the 1940 PUMS. Helpful comments and suggestions were offered by Richard Easterlin, Edward Glaeser, Lawrence F. Katz, Robert Margo, Stefanie Schmidt, and the participants of the All-UC Conference on Capital Markets, the Greater D.C. Area Economic History Seminar, and workshops at The Brookings Institution, Columbia University, Rutgers University, the University of Texas at Austin, and Yale University. Two anonymous referees and Gary Libecap made the article a better one. An appendix, describing the methods used to produce the state-level public and private school data, is available as NBER-DAE Working Paper No. 57. This article draws on Claudia Goldin, "How America Graduatedfrom High School: 1910 to 1960," NBER Working Paper No. 4762 but differs substanitially from it. l See Denison, Trends, table 8-3, p. 113. The role of formal schooling in economic growth is virtually the same for most other subperiods of this century for which the calculation has been made, including the 1929 to 1948 period, of particularrelevance here. On the subject of economic growth and human capital, see also Barro, Determinants, for evidence of a relationship between education and economic growth across more than 100 countries; and Jorgenson, Productivity, for the relationship between productivity and education within the United States. the of An oft-citedstatisticdemonstrating importance humancapitalto 345 346 Goldin TABLE 1 COMPUTING THE IMPACT OF INCREASED SECONDARY SCHOOLING ON THE EDUCATIONAL ATTAINMENT OF THREE COHORTS OF MEN, BORN 1886 TO 1950 Percentage Distribution of Highest Grade Completed by Grade Cohort Born in Year 1946-1950 1926-1930 1886-1890 ?8 6.6 17.4 72.5 9-11 12.5 21.2 10.7 12 45.6 42.3 8.3 >12 35.3 19.1 8.6 All Grades 100 100 100 Mean Years in Grade Interval <8 1946-1950 1926-1930 1886-1890 7.7 7.2 5.8 9-11 11.5 10.5 9.8 12 12.0 12.0 12.0 >12 15.3 15.2 15.2 Overall Mean 12.82 11.46 7.58 Mean Without Increase to Eighth Grade 9.81 10.47 Counterfactuals If the 1886-1890 cohort had high school attainment of the 1926-1930 cohort If the 1886-1890 cohort had high school attainment of the 1946-1950 cohort Mean 10.27 11.36 Notes: The counterfactual figure of 10.27 years for the 1886-1890 cohort is computed by assuming that 21.2 percent of the cohort (ratherthan 10.7 percent) attains grades 9-1 1; 52.8 percent (rather than 8.3 percent) completes high school but does not continue to college; and 8.6 percent completes high school and continues to college. Increasing those in grades 9 to 12 necessitates decreasing the group below ninth grade. Because 31.3 percent of the 1886-1890 cohort had already completed the eighth grade, 23.7 percent = [(21.2 + 52.8 - 10.7 - 8.3) - 31.3] are required to have their primary-grade education increased to advance them to completion of grade eight. The mean for the lowest group is reduced because those who complete the highest elementary school grades are moved on to high school. Mean years in each interval are: 2.3, 10.5, 12, 15.2, where the 2.3 figure is computed from the 1940 PUMS for the truncated distribution.The mean years per student required to bring the group to ninth grade is 1.944 (calculated using the 1940 PUMS). The increase in primary school years is 0.461 (1.944 . 0.237). The column "mean without increase to eighth grade" subtracts this amount from the counterfactual mean. A similar procedure is used to produce the counterfactual using the 1946-1950 high school attainment.It should be noted that the data use all males in 1940, not just the native-born, but that the 1940 census overstates the proportion who complete twelfth grade. See the text. Sources: Percentage distribution of highest grade completed and mean years in grade interval for cohorts born 1926-1930 are 1946-1950 from Smith and Ward, Women's Wages. Data for the cohort born 1886-1890 from the 1940 Public Use Micro-data Sample (PUMS, entire 1/100 sample); see ICPSR, Census. of resultingin technologicalchangeand its diffusion,the contribution humancapitalfornationto economicgrowthmusthavebeen even greater than the 28 percentestimated.Althoughone can quibblewith the details of the Denison calculation,its fundamental conclusionis likely to be unaffected. Humancapitalaccumulation technological and changewereto the twentieth centurywhat physical capitalaccumulation was to the nineteenthcentury -the engine of growth. America's Graduation from High School 347 Among the possible contributors the increasededucationalstock of to Americans duringthe firstthree-quarters the twentiethcentury,changes of at the secondary schoollevel were quantitatively most significant. the About 70 percentof the increase,from 1930 to 1970,in the educational attainment of U.S. males 40 to 44-yearsold was due to the rise in secondaryschool attendance.This important can be demonstrated follows. Hold the fact as fraction going to college (havingmore than 12 years of schooling) at the level of an older cohort(in this example,men 40 to 44-yearsold in 1930, bornbetween 1886 and 1890) but allow high school graduation attenand danceto rise to thatachievedby a youngercohort(men 40 to 44-yearsold in 1970, born between 1926 and 1930). This experimentresultsin an increase of 2.7 years in mean years of schooling (see Table 1 for details). Yearsof schooling, as measuredby the 1940 U.S. populationcensus, increased by 3.9 years between these cohorts. Therefore70 percentof the increase(2.7 / 3.9) in schoolingwas due to increased high schoolattendance and graduation.2 Thus advancesin secondaryschooling may provide the singlemost important measurablereasonfor per capitaincome growthfor much of this century.3 The diff-usion high school educationacrossthe entireUnited Statesis of shown in Figure 1, which graphstotal secondaryschool enrollmentas a of percentage 14 to 17-yearolds andgraduates frompublicandprivatehigh schools as a fractionof 17-yearolds. The high school enrollmentraterose from 18 percentto 73 percent and the graduation rate increasedfrom 9 percent to 51 percentduringthe threedecadesafter 1910. The rate of increase was nothing short of spectacular and the levels attainedwere unothercountry untilmuchlaterin the century. equaledby any However,these datahide largeandimportant aggregate differences acrossthe variousstates and regions. Giventhe generallyacceptedrole of educationin economic growth,it is unfortunate educational that stock datafor the United Stateswere not collected until 1940, when a questionon highestgradecompletedwas added 2 The figure would decrease to 57.5 percent if the 0.461 years needed to advance those from grades five to seven to grade eight were subtracted.But it would rise to 85 percent if the mean for the primary school (only) group was kept at 5.8 years, rather than falling. (See Table 1 for details.) Note that by computing the educational advance of men 40 to 44-years old between 1930 and 1970, I am capturing men who would have been 17-years old from 1903 to 1947. Because the calculation uses the 1940 census, it understatesthe role of secondary schooling to the increase in the human capital stock (see the section titled "Contemporaneous and Retrospective Education Data"). But because the foreign-born population is included in the 1886-90 cohort, the calculation may overstate the impact of secondary schooling on the native-born population. 3 By "educational stock" I mean years of education, although the quality of education is also considered in, for example, Denison, Trends. If the return to a year of college is greater than that to a year of secondary school, the role of college in augmenting income would be larger than its influence on the stock of human capital. But, for the period considered here, the returnto a year of secondary schooling was about as great as that to a year of college (see Goldin and Katz, "Human Capital"). 348 0. 1 Goldin Total Enrollment Rate 0.8 ~0.7- , 0.6 03- z - 2 A f X t K~~~~~~~rauation Rate 0.23, 00.2 1890 1900 1910 1920 1930 Year 1940 1950 1960 1970 FIGURE 1 SECONDARY SCHOOL ENROLLMENT AND GRADUATION RATES: ENTIRE UNITED STATES Notes: Enrollment figures are divided by the number of 14 to 17-year olds; graduation figures are divided by the number of 17-year olds. The total includes both males and females in public and private schools. Source: U.S. Department of Education, 120 Years, tables 9 and 19. to the U.S. federal population census.4 But contemporaneous evidence on graduation and enrollment allows one to obtain educational flow data for periods when the stock equivalents do not exist, and of equal importance, they enable checks on the stock data for the native-born population. Such data can be obtained from the reports of schools, school districts, and states and are those on which I primarilyrely. They reveal that the growth of U.S. secondary schooling from 1910 to 1940, known as the period of the "high school movement," was considerably faster in certain regions than Figure 1 shows for the entire United States.5 They are also used to expose various deficiencies in the stock data derived from the 1940 U.S. population census. 4 Only the Iowa (1915, 1925) and South Dakota (1915) state censuses asked questions on the educational stock before 1940. 5 The term "movement" might make the increase in secondary schools appear to be a coordinated crusade. Those who use the term today and in the past-it was used in many state school reports in the 1910s and 1920s-may ascribe to such a belief and various national associations (such as the National Educational Association and college associations) did spread the "gospel" of secondary school education and helped coordinate curriculumchange. But in 1910 the more than 125,000 school districts in America (U.S. Bureau of the Census, Historical Statistics,series H 412) engaged primarily in a grassroots level change. It was clearly affected by national propaganda and facilitated by various state laws, but its spread was less coordinated than the term "movement" would imply. America's Graduation from High School 349 Because the United States underwent a rapid increase in education in the several decades after 1910, it might be conjectured that other industrialized countries did as well. But the countries whose per capita incomes were closest to that of the United States in 1910 did not undergo an equivalent transformation at that time. Rather, their high school movements did not materialize for anotherthirtyor more years.6Because each country had (and often still has) a somewhat different educational system, and because even the meaning of "secondary school" varies among them, educational data must be put on a comparablebasis by using student ages ratherthan grades of school. Britain, France, and even Germany had attendancerates in the 14 to 19-year old range that were considerably lower than in the United States for much of the twentieth century.7Thus the "high school movement" in the United States was a unique educational advance for the period. The modern American high school was born at the turn of this century and maturedfully during its next few decades.8 Its spectacular growth from 1910 to 1940 was central to the increase in U.S. educational attainment during much of this century. But the subject of secondary schooling has been a much-neglected part of American economic history, due to the absence of data on enrollment and graduationat other than the national level.9 I have constructed such data for public and private high schools covering the period since 1910. The statistics are the first of their kind to be produced and reveal surprising aspects of the spread of America's high schools. The increase in secondary school education after 19 10 was rapid across the entire nation. But large differences even among nonsouthem states appeared early and suggest some of the reasons why educational advance occurred. In 1924, for example, the graduationrate in Nebraska or California was twice that in 6 For example, in 1944, as President Roosevelt signed the G.I. Bill of Rights to enable America's returning soldiers to further their education, Prime Minister Churchill and the Labor Party had just achieved passage of a law that guaranteed a tuition-free secondary school education for all British children. 7For estimates of attendance by age for these countries, see Ringer, Education. In 1936/38, for example, the attendance rate of 14-year olds in Britain was 38 percent and that of 17-year olds was just 4 percent. For the United States in 1936/38 the enrollment rate of 14 to 17-year olds was 68 percent. If attendance rates were 80 percent of enrollment rates (which is what they were nationally for elementary and secondary students), then the attendance rate for the 14 to 17-year old group in the United States would have been 54.4 percent. That for a comparable group in Britain could not have exceeded 20 percent. Even in 1960/62 the attendance rate for 17-year olds in Britain was just 15 percent. Including enrollment in full-time technical schools in Britain would not serve greatly to close the gap. See Goldin and Katz, "Why the United States." 8 See Reese, Origins, for an excellent summary of the predecessors of the modem U.S. high school. The first U.S. public secondary school (Boston Latin) was founded in 1635. See Katz, Irony, and Vinovskis, Origins, on the protest over the early establishment of a public high school in Beverly, MA. 9 In the historical literaturein general, see, for example, Krug, Shaping; Labaree, Making; Perlmann, Ethnic Differences; Reese, Origins; Tyack, One Best System; and Ueda, Avenues, to mention but a few of the history books on the subject. The sociology literature is equally vast, although less focused on the high school. See, for example, Fuller, "Youth Job Structure"; Walters and O'Connell, "Family Economy"; and the oft-cited Trow, "Second Transformation." 350 Goldin New Jersey or New York. The subject to be explored, therefore, is where and when high school enrollment and graduationrates advanced.10The task at hand is to understandthe manner in which secondary schooling diffused spatially and over time, and thus unpacking the data in Figure 1 by state and region is a first step in understandingthe American high school movement. INSTITUTION A UNIQUELY-AMERICAN Not only was the high school movement from 19 10 to 1940 a uniquelyAmerican phenomenon, the secondary school as we know it today was a uniquely-American invention. As an institution, it was rooted in egalitarianism and was often a by-product of the extensive state university systems in the United States. Not until 1902 was there a standardformat established and prescribed for The America's high schools, requiringfifteen units for graduation.11 modem form evolved in the early 1900s and diffused so rapidly that the high school of the 1920s bears a far greater resemblance to that of the late twentieth century than it does to that of 1900. Precollege education in the United States was a very local affair in the period under study and remains so today. Federal legislation plays practically no part in the story of expansion. State funding accounted for a minor fraction of the expenditures on education by localities in the 1910 to 1940 period. In 1925, for example, just 16 percent of all kindergarten through twelfth-grade educational funds came from the states, and it was mainly in the South that the fiscal role of the states was the largest.12Whatever decisions led to the increase in secondary schooling, they came primarily, although not exclusively, from a grass-roots movement that was funded by local propertytaxes and in which school building and curriculum decisions were made in the more than 125,000 school districts nationwide.'3 In the 1910s about 50 percent of all public high-school graduatessaid they intended to continue to college or another institution of higher leaming without delay, and 55 percent of male public high-school graduatessaid they intended to (see Table 2). By 1923 the number had slipped to 44 percent, and by 1933 only 25 percent planned to continue their studies upon high 10 See Goldin and Katz, "Why the United States," on the causes of the high school movement and "Decline," on some of the consequences. "' A unit was four to five periods, each 45 minutes per week during a 35-week semester. The standard was set by a commission including many college administrators. See Krug, Shaping of the American High School 1880-1920, p. 153. 12 U.S. Bureau of Education, Biennial Survey, 1924-1926, p. 593. The percentage means that 16 percent of all revenue receipts for education were received and distributed by state governments. The federal government, which provided funds through the Smith-Hughes Act for vocational education, is included in the state percentage, although it accounts for a very small fraction. 13 The degree to which all these school districts were truly independent fiscally is still open to investigation. America's GraduationfromHigh School 351 TABLE 2 PERCENTAGE OF HIGH SCHOOL GRADUATES CONTINUING TO COLLEGE AND OTHER INSTITUTIONS, 1901 TO 1937 Percentage Intending to Continue Education among High School Graduates in Year 1901 Continuing to college from: Public high schools Public high schools, males only Public and private high schools Continuing to college and other institutions' from: Public high schools Public high schools, males only Public and private high schools a b "Other institutions" probably 1910 1914 1923 1933 1937 31 40 33 34 45 35 35 45 35 31 37 32 21 23 23a 24 26 - 49 55 49 50 55 50 44 49 46 25 25 25a 29 30 Private school graduation numbers are for 1932. include normal, nursing, and library schools. Notes: "Percentage continuing" was reported by the commissioner of education who gathered the information from the reports of school principals. It is the percentage who intended to continue, probably in the immediate future. Some did not eventually do so, others went back to school at a later date having never indicated an intent to continue their education. These figures probably do not include students in the preparatorydepartmentsof colleges and universities. Their inclusion would increase the continuation percentages in the earlier years. Sources: Office of Education, Report of the Commissioner of Education (various years); and Office [Bureau] of Education, Biennial Survey of Education (various years). school graduation.14 proportionof all American youths entering college The had not decreased. It had expanded. Rather, fewer of the new entrants treatedhigh school as a prelude to college. Thus the fraction continuing after graduation declined with the rise in high school enrollments. Only in the 1970s did the fraction continuing reach the levels attained before 1920.'" Thus a primary reason to attend high school around 1900 was to gain entrance to college and most high schools offered the classical (or Latinscientific) curriculum,requiredby many colleges for admission. The courses of the classical curriculum included Latin, possibly Greek, often French or German, English, history, mathematics, and some science. Secondary schools began as preparatory institutions, but as long as they primarily 14 The proportion of males and females who actually continued to a B.A.-granting institution in 1910 was about 40 percent. The calculation is complicated by deficiencies in the data on higher education enrollments. In 1910 there were about 156,000 public and private secondary school graduates (U.S. Bureau of the Census, Historical Statistics, series H 598). There were 174,213 collegiate students in 1910 (U.S. Bureau of Education, Biennial Survey, 1928-1930, p. 338). If 35 percent of these students were in their first year, then 39 percent of the graduatingclass would have continued to a B.A. granting institution. Note that this does not include a host of non-B.A. granting institutions of higher education in 1910 but may include students enrolling in several colleges during the academic year. Thus the figure for the "intent to continue" in 1910 is close to the estimated actual one. '5 Using data on educational attainment from the federal population census, Smith and Ward show that the fraction of male high school graduates who continued to college did not reach that achieved by the 1886-90 birth cohort until that born 1946-50. See Smith and Ward, Women's Wages, p. 39. 352 Goldin 16 trainedyouths for college entrancethey would have limited appeal. The high school movement transfornedsecondaryschools from preparatory institutionsto schools thatawardedterminaldegreesto the vast majority. high schoolaround1910 to 1940 often who entered The flood of students sought an educationthatwould lead directlyto employment,not college. jobs that The economyhadbegunproducinglargenumbersof white-collar demanded formal educationin excess of that provided by the common occuby schoolbutless thanthatfurnished college.7 Even someblue-collar by the pationsdemanded cognitiveskills furnished a high school education, use blue-prints, complex forsuch as the abilityto readmanuals,interpret and of the mulas, and understand fundamentals geometry,chemistry, elecI8 tricity. Increased demandfor a high school educationby studentsnot boundfor If curricula. of college led to a questioning the classicalandLatin-scientific high schools were to preparestudentsfor life, ratherthan for college, the by wouldhaveto changefromthatrequired college. The English curriculum was the most popularsuccessorto the classical;it droppedthe curriculum and Greekrequirement gave studentsmorechoice in foreignlanguages.In were enrolledin Latin;by 1934 1900, 51 percentof all high school students The only 16 percentwere takingLatin.19 new issue was whetherhigh school in should trainstudentsfor employmentby offering instruction a host of technicalor manual,and (includingcommercial), arts.Vocational practical into incorporated most high school curricula. industrial courseswererapidly By 1934, 10 percentof high school studentswere enrolledin bookkeeping, 17 percentin typing,and 9 percentin shorthand.20 were formedto studythe effectsof secondary Then,as now, commissions competitiveness.One of these, the schools on Americanmanufacturing headedby CarrollWright,the formerU.S. CommisDouglasCommission concludedin 1906 thatthe currentschool system sionerof LaborStatistics, countries extenhad was not meetingthe needs of industry. Some European sive industrialand manual trainingprograms.By implication,then, the UnitedStateswould lose out in worldmarketsunless somethingwas done 16 See Krug, Shaping of the American High School 1880-1920, pp. 62-63, for a discussion of the famous Committee of Ten Report issued in 1893 which advised secondary schools to offer alternatives to the classical and Latin-scientific programs. None of the alternativeproposals, however, included the commercial and vocational courses that were to become part of the modern high school. '7 See Goldin and Katz, "Decline." 18 On the increased demand for educated youths in blue-collar occupations, see Goldin and Katz, "Origins" and "Technology." It has also been claimed that credentialization, or "signaling," evolved in this period and that a high school diploma was the first credential of that type. See Labaree, Making. 19 U.S. Department of Education, 120 Years, table 16. Krug, Shaping of the American High School, volume 2, notes that the classical curriculum was difficult to dislodge even in the 1920s, possibly because high schools wanted to ensure that some of their students could continue in college. Many secondary school programs were severely criticized as being elitist. 20 Figures on curriculum are all from Department of Education, 120 Years, table 16. America's Graduation from High School 353 about trainingyouths for manufacturing jobs.2"Many looked to the apprenticeship system in Germany,in which students combined academic subjects and industrialtraining. Such an industrialtrack, however, was quickly abandoned, a casualty of a lack of cooperation between firms and unions and a suspicion that tax dollars would support particular firms and industries. STATE-LEVEL STATISTICS SECONDARYSCHOOLENROLLMENTAND ON GRADUATION The widely cited high school enrollment and graduation numbers for the entire United States, given in Figure 1, conceal cross-sectional differences and time-series changes of enormous consequence in understanding the diffusion of secondary school education in America. The South, for example, had meager high school enrollment and graduation rates, for white as well as black youths, early in the centurybut its high school rates converged on the rest of the country beginning in the 1950s. An illusion is thereby created that secondary schooling rates, from 1910 to 1970, increased gradually and continuously across the United States when, in fact, most nonsouthem regions show a burst of growth from 1910 to 1935 and then little change until the late 1950s. A time-series compilation of secondary-school data at the state and regional levels has not previously been attempted. The reasons primarily concern the large number of sources required to construct the statistical history of secondary schools. Brief descriptions of the data, the statistical problems, and the methods used to overcome them are given because the data are at variance with the educational stock data in the 1940 U.S. population census, even though they are fully consistent with the flow data for the entire nation. Some information concerning how the data were constructed is needed to understand the reasons for the important differences. Construction of State-Level Statistics22 The state-level secondary schooling data come from several sources, and among the most important are the annual Reports of the Commissioner of Education for the years before 1918, and the Biennial Surveys of Education for the years 1917 to 1958. Other sources used include the reports of various state commissioners of education, documents pertainingto Catholic schools, and state and national surveys of colleges and universities regarding their preparatorydepartments. 21 See Massachusetts Commission on Industrial and Technical Education, Report. 22 Figures 2 through 6, the regressions reported in Table 3, and the discussion in the article are based on the data set, which is not included due to lack of space. The data are available from the author. A more extensive discussion on the construction of the data set can be found in Goldin, "Appendix." 354 Goldin The secondary school group in the United States has almost always been consistently defined as grades nine throughtwelve.23Two measures of education are developed and discussed here: the proportion of youths enrolled in the four secondary school grades (nine through twelve), and the proportion graduatingfrom high school (as defimedby the several states).24Attendance data do not exist for the states at the national level, but some states, in their separateeducational bulletins, reportedboth enrollment and average daily attendance;and attendance data for youths in cities were collected by the federal government.25 Graduationrate data, as well as enrollment retention rates from year to year, are the most reliable measures of secondary schooling for the entire United States.26 State-level data on the number of individuals enrolled in and graduating from public and private secondary schools have been regularly collected by the U.S. Commissioner of Education since 1909. The commissioner requested such data from each secondary school on record with the U.S. Bureau of Education (later the U.S. Office of Education, and then the Department of Education).27 After 1920, the school surveys were supplemented by the bureau with those undertaken by the various states. Undercounts between 1910 and 1922 required adjustments to the official U.S. Bureau of Education state-level statistics. The preparatorydepartmentsof colleges and universities were never included in the bureau's secondary-school surveys and have been added. Missing data,particularlyfor private schools, required other adjustments. 23 Each state formulated it own definition of what constituted enrollment in various grades and graduation from secondary school. In most states, secondary school graduation enabled admission to the state university, and, for that reason and others, the state took great interest in the accreditation of secondary schools. Attendance in common schools for more than eight years generally did not constitute secondary-school participation. See Goldin and Katz, "Human Capital," for evidence that common schooling beyond nine years had no pecuniary return. 24 Quality of education is another dimension. Objective measures of educational quality include days per year, hours per day, students per teacher, and resources per pupil. Also important are curriculum and performance measures. On curriculumchanges in U.S. high schools from 1928 to 1990 see Angus and Mirel, "Rhetoric." 25 The difference between enrollment and attendance is twofold. First, some students who enroll in September will never attend school, and second, students are absent for various reasons throughout the year. A sampling of various state reports in the 1920s indicates that the ratio of average daily attendance to enrollment exceeded 0.80. The ratio of average daily attendance to enrollment across U.S. cities in the 1920s was 87 percent. The city data are described in Goldin, "How America." 26 The retention rate data are not presented here, but can be produced from the enrollment-by-grade data. Summer school and night school enrollments, it should be noted, are believed to be omitted in all of the state data. I have examined the possibility that the graduation data are understated by the omission of these numbers and find that, if they are, it would entail a trivial bias. 27 The Bureau of Education, the forerunnerto today's Department of Education, was established in 1867 and became the Office of Education in 1869, an agency of the Department of the Interior where it stayed for 70 years. It was known as the Bureau of Education for those 70 years, but in 1929 it was renamed the Office of Education. In 1939 it became part of the Federal Security Agency and was, in 1953, included in the new agency of Health, Education, and Welfare (HEW). The Department of Education became a separate cabinet-level agency in 1980. America's Graduation from High School 355 Sometime in the 1920s, the bureau began to accept the state data and adopted a method for making revisions to the national data from the late nineteenth century.The procedure was never fully described by the bureau. For all years from 1910 to 1962, the total enrollment and graduation data I have produced by state are virtually identical to the national aggregates constructed by the Department of Education and contained in the widely used series in U.S. Bureau of the Census, Historical Statistics.28The difference is that my biennial or annual estimates are by state, sex, and grade; the bureau never published detailed estimates, if indeed it ever produced them. The data presented here begin with 1910 because the proportion of secondary schools responding to the Bureau of Education's request was low in prior years and evidence for all the states would be difficult to obtain.29 Although the adjustmentsmade to the original state-level data are many, most are confined to the years between 1910 and 1922 or concern the inclusion of private-school students and those in the preparatorydepartments of colleges and universities. Undercountspresent the greatestpotential problem with the school survey data collected by the U.S. Bureau of Education. Before 1920 the bureau relied solely on its own school surveys. Enrollment data in various state reports exceeded those in the bureau's school survey data. But there is scant commentary in the annual reports of the bureau regardingthe possibility of an undercount.Mention was made that the number of responding schools was about 85 percent of the total in the 1910 to 1920 period but that because most were small the undercount of students was far smaller than that of schools. Beginning in 1920 and continuing to 1938, the U.S. Bureau of Education attemptedto reconcile their data with those reportedby the states by requesting information from the states and publishing it in a separate section of the Biennial Surveys, "Statistics of State School Systems." Thus the Biennial Surveys from 1920 to 1938 contain two series for both the public and private schools. One was obtained from the school surveys and contains detail on students, teachers, and schools. The other, from the states, consists of only The adjustmentsmade by the Bureau of Education prior to 1930 were poorly documented in the bureau's reports. So meager was the documentation that extremely able personnel at the current Department of Education made fundamental errors in interpreting them. An otherwise informative publication, U.S. Department of Education, 120 Years, gives a largely inaccurate historical series on high school graduation rates. The revised data (such as that found in Historical Statistics) are used for the total, but graduates from private high schools are computed by subtracting from these the number graduating from public high schools, as given in the original Annual Reports and Biennial Surveys. This procedure results in an extremely large, and erroneous, figure for the private graduation numbers for the period to the mid-1920s. The public graduation numbers were largely revised by the U.S. Bureau of Education on the basis of state survey data. Therefore the total graduation data in Historical Statistics already reflect revisions to the published data of the U.S. Bureau of Education. 29 My data series continues to the present but is used in this article to 1962 to highlight the most important historical period, 1910 to 1940. I include some years in the 1960s to show consistency because the sources change with 1958. 28 356 Goldin the enrollment numbers by grade. The state data have been treated as more accurateby subsequent researchersat the Bureau of Education. In constructing my series, I combine the evidence, using the school survey data for the detail and the state data for the totals. In the early 1930s the U.S. Bureau of Education began to augment their school survey informationwith data from the records of the various state departments of education and, in consequence, the enrollment data from the two surveys are virtually identical with that report and after.30 To summarize, the data on enrollments and graduation before 1920 were obtained by the U.S. Bureau of Education through their school surveys. From 1920 to 1938 the bureau obtained data both from schools and from the states. The data obtained from the states contain only enrollments, not graduates,although enrollments are given by grades. Therefore, for the 1920 to 1938 period the state data can be used to revise those from the school surveys, but there are no easily obtainable state-level data for the period before 1920. The bureau, it appears, later adjustedthe national enrollment data for the 1910 to 1920 period by a factor of 1.175, under the assumption that their data were 85 percent those of the states. The bureau also applied the same correction to the graduationnumbers for both public and private secondary schools.3' (The widely used U.S. graduationrate data in Historical Statistics, series H-598, contain these numbers.) I have, instead, adjusted the 1910 to 1922 data using the ratio of the enrollment reportedby the state to that from the school reports for the first reliable subsequent year, 1924. Any resulting state enrollment and graduation numbers that seemed anomalous were checked, when possible, against the individual state reports, and any differences were reconciled. It is likely that my procedures, if they err, overstate enrollment in and graduationfrom public high schools in the early period.32Because I emphasize the large increase over time in secondary school enrollment and graduation, a bias that increases schooling numbers in the 1910s is preferredto one that lowers such numbers. Particularly in the comparison with the educational attainment data from the 1940 census, the preferred bias in the contemporaneous data is one that creates an upper bound to the actual numbers. Private school data on enrollments and graduates at the state level also had to be revised for the 1910 to 1922 period. The adjustmentsemployed are Any differences seem due to the exclusion of data from schools with fewer than 10 pupils. Although there is no mention that this was the procedure used, one can virtually duplicate the national graduation numbers with such a procedure. But there is no reason that the undercount of enrollment should have been the same as the undercount of schools and, similarly, that the undercount of graduates should have been the same as the undercount of enrollments. 32 The implied undercount of students is almost identical to the percentage of schools not reporting, yet the schools that did not report were smaller than average. 31 30 America's Graduation from High School 357 similar to those used for the public school numbers. The final undercount of importanceconcerns students in the preparatorydepartmentsof colleges and universities.33 The students in these preparatory departments were not included by the Bureau of Education in the high school numbers. Rather, they found their way into the college category because schools, not students, were surveyed by the commissioner. In 1910, preparatory students in U.S. colleges and universities were 31 percent of all private high school students and in 1920 they were 22 percent. Although private schools were a substantial fraction of all secondary students before 1910, the fraction fell with the expansion of public high schools. By 1910, the year this study begins, the fraction of all secondary school graduates from private schools nationwide was 18.2 percent. It declined furtherto 12.8 percent in 1920 and then to 10 percent in 1930. Large variations exist in the percentage by region and over time. Whereas the greatest fractions in private secondary schools before 1920 were in the South, the greatest fractions since the 1930s have been in the New England and Middle Atlantic states.34 The data constructionjust described produces graduation and enrollment statistics by state and, when aggregated, by region. To obtain the more useful graduation and enrollment rates, the relevant population (14 to 17year olds for enrollment; 17-year olds for graduation)is used as the denominator.35 Two other sources could potentially yield national information on both school going and educational attainmentfor the period under consideration. Both are from the federal population censuses. Prior to 1940, the population censuses gave contemporaneous informationon school going, derived from a question concerning whether the individual attended school during the previous year. Although the data correlate well by state with those I have constructed, the levels are considerably higher. The reason concerns the question asked. The census requested information on attendance at any school, for even one day in the previous year. The school could have been a music, technical, vocational, or academic institution and could have operated at night or during the day, for the summer or school year. The other source is the 1940 and subsequent censuses which asked information on highest grade completed. These data have been used by many researchersto 33 Prior to the expansion of public secondary schools, colleges and universities in many states could not depend on high schools to produce sufficient numbers of properly trained students. When the University of Nebraska opened its doors in 1871, for example, just one public high school existed in the state. Many colleges and universities, therefore, instituted their own preparatory departments and some still exist. 34 Data on private schools are not presented due to space limitations and are available from the author. 35 The convention is to use these age groups. The population data were estimated by a constant growth-rate interpolation between the decennial census years. 358 Goldin But the data for many older cohorts calculate completed grades by cohort.36 produced from this source are at variance with the contemporaneous data from secondary schools. More comments on this topic follow later. WHEREAND WHEN DID AMERICAGRADUATEFROM HIGH SCHOOL? Graduation Data The state-level high school data constructedby the methods just outlined can document educational advances during the period of the high school movement. Even though the procedures produce state-level estimates, there are compelling reasons to aggregate to the standardnine census divisions. Most important is that within-region levels and movements are similar across states and thus that aggregation allows one to summarize the movements of 49 separate series. Another matter for consideration is whether enrollment or graduationrates should be used. For most of the states the two move together, but graduation data are far cleaner in the sense that they contain fewer biases and potential inconsistencies. The public and private graduation rates (number of graduates/17-year olds) by region are graphed in Figure 2. The upper panel contains the three northern and midwestern regions (New England, Middle Atlantic, East North Central) and the lower panel has the three western regions (West North Central,Mountain, Pacific), all for 1910 to 1962. Figure 3 graphs the same for two regions of the South (South Atlantic, East South Central) and for white youths (in the South Atlantic) for 1930 to 1954, the period of data availability.37 In all nonsouthernregions, the total (public and private) graduation rates were at or below 10 percent in 1910 although each series rose to 50 percent or greaterby the mid-1930s. In 1935 the median 17-year old in these regions was a high school graduate,althoughjust 15 years earlier in 1920 the youth stood only a 20 percent probability of having a high school diploma. The increase in graduationrates from 1920 to 1935 was extremely rapid, particularly so during the early years of the Great Depression in the industrialareas of the northeast and Midwest. Graduationrates, in all regions, were substantially lowered immediately before and during American involvement in World War II, although the G.I. Bill later increased high school rates to abnormallyhigh levels.38Even when private schools are added to the total, graduation rates in the 1950s are no See, for example, Smith and Ward, Women 's Wages. For clarity, only two of the three southern regions are graphed. 38 The reason they are increased above normal is because the denominator here is the number of 17year olds and all returning G.I.'s would have been older than 17. The G.I. Bill granted stipends to returning soldiers and covered high schools, as well as colleges. See Olson, GI. Bill. 37 36 from High School America's Graduation 0.8 - 359 New England -9 Middle Atlantic East North Central , 0.7 . 0.6 0.5 0 .4 C704 0 0.3 .~0.2 0.1 0 1910 1915 1920 1925 1930 1935 1940 1945 1950 1955 1960 Year 0.8- 0 West North Central Mountain Pacific 4 0.7 0 IjI ~ ~ '- 0.1 ' 191O 1915 1920 1925 1930 1935 1940 1945 1950 1955 1960 Year 2 FIGURE TOTAL(PUBLICAND PRIVATE)HIGHSCHOOLGRADUATIONRATES is Notes: Males and females are combined.The numberof graduates divided by the approximate dataaremade of growthrateinterpolations population numberof 17-yearolds in the state.Constant census years. betwveen 360 Goldin FIGURE 2-continued Sources: U.S. Bureau of Education, Report, (various years); U.S. Bureau [Office] of Education, Biennial Survey, (various years); U.S. Departmentof Health, Education, and Welfare, Statistics of State School Systems and Statistics of Non-Public Secondary Schools, (various years); and National Catholic Welfare Conference, Summary, (various years). See also Goldin, "Appendix." Population data are from the U.S. Population censuses (1910 to 1970). higher than are those of the late 1930s.39In most nonsouthem regions, high ratesdid not increase,or increasedonly slightly,during school graduation rapid increase in high The extraordinarily the twenty years after 1938.40 ratesin the 1930s is probablythe reason. school graduation rates duringWorldWarII occurredin all The reductionin graduation regions but was largestin the New Englandand Pacific states,the sites of due many war-timeindustries.The decreasewas not primarily to the miliof taryrecruitment high schoolboys for most wouldhave been below drafthigh schoolyouthsmovedintocivilianjobs as wages rose able age. Rather, WorldWarI at disproportionately the lower end of the wage distribution. it hada similarimpact,although is barelynoticeablein Figure2 becausethe levels are lower. Femaleyouthsalso left high school duringWorldWarII, althoughless thandid males (see Figure4). Teenagedgirls were attracted by many of the same factorsthat enticedboys to leave high school-the increasein the relativewage of unskilledlabor.4" The threewesternregions-West North Central,Mountain,and Pacific ratesduring1910 to 1940 thandid the increasesin graduation -had greater regions of the East and the increaseswere more continuous.Ratherthan gettingoff to a slow startandthenhavinga largespurtfrom 1920 to 1940, as did the industrial statesof the East and Midwest,the statesof the West rates that grew more rapidly and the Plains had high school graduation period.42 duringthe pre-Depression fromthe othersin the considThe regionsof the Southaredistinguished erably lower initial secondaryschoolinglevels and theirmore continuous increaseto the end date(see Figure3). Nor arethe datamuch alteredwhen is as only the whitepopulation considered, it can be for the yearsfrom 1930 3 Private school graduation data are available after 1934 for only a few years in the 1940s and 1950s. The data for the 1950s have been constructed from Catholic school reports (National Catholic Welfare Conference, Summary), not readily available for the late 1930s. 40 The Office of Education recognized in the 1950s that their method of collecting enrollment data led to numbers that were inflated by about 2 percent because students could be enrolled any time during the school year. They switched to a fall enrollment basis in the 1960s, thus reducing the potential for double-counting. The change does not affect the graduation data. 41 Young, high-school aged women were also enticed to leave school by the chance to work in jobs they could not have entered before the war. 42 Graduation and enrollment rates in the Pacific region increased so greatly during the 1930s that they exceeded levels attained in the early 1950s. The reason does not appear to be the inclusion of parttime students but may, instead, be due to the effects of the Depression combined with the generosity of its public education system. America's Graduation from High School 0.6 South Atlantic +South Atlantic White 9East 361 South Central = 0 0.5 *~0.4 co 03 0 ci0.2 ~0.1 O 1910 1915 1920 1925 1930 1935 1940 1945 1950 1955 1960 Year FIGURE 3 TOTAL (PUBLIC AND PRIVATE) HIGH SCHOOL GRADUATION RATES FOR THE SOUTH ATLANTIC (TOTAL AND WHITE) AND EAST SOUTH CENTRAL REGIONS Notes: Males and females are combined. The number of graduates is divided by the approximate number of 17-year olds in the state. Constant growth rate interpolations of population data are made between census years. Sources: See Figure 2. to 1954. There is a discernible increase during the Depression, but the timing is different from that in other parts of the nation. The revolution in high schools did not take place in the South during the 1920s and 1930s but was drawn out over a much longer period of time. But even though the South lagged the rest of the nation in educational attainment,its rates of secondary school enrollment and graduationwere still higher than were those of many nations at the time. In all years and regions, although just two (Middle Atlantic and East North Central) are shown in Figure 4, the female graduationrate was higher than that for males.43It was considerably higher in New England, the West North Central, and the Mountain states. That differences by sex are found in all states suggests that the return to high school education, relative to grammar school, was greater for female than male youths. A high school education for a young woman meant entree to office jobs, whereas its ab43 The data in Figure 4 are for public high schools only because the data for private schools are not always separated by sex. The fraction in private schools declined in the 1930s and rose in the 1950s, accounting for the apparent decrease in Figure 4 in graduation rates after 1950. 362 0.7 Goldin 0.6 = 0.5 0.4 '~0.3 ~4 0.202 0. I Females 1910 1920 1930 1940 1950 1960 Middle Atlantic 0.70.60 0.5 0 0 0.3 0.2 0. 1 0 1910 1920 1930 1940 1950 1960 East North Central FIGURE 4 PUBLIC HIGH SCHOOL GRADUATION RATES BY SEX: MIDDLE ATLANTIC AND EAST NORTH CENTRAL REGIONS America's Graduation from High School 363 FIGURE 4-continued Notes: The number of public-school graduates is divided by the approximate number of 17-year olds in the state. Constant growth rate interpolationsof population data are made between census years. The dotted line is the public-school graduation rate for both sexes. Sources: See Figure 2. At sence meant a low-wage manufacturing as an operative.44 a time job when physicalstrength mattered jobs in variouscraftswere strictly still and limitedto males, secondaryschoolingfor girls had a higherrelativereturn thanfor boys.45 shouldbe realized,however,thatthe internal of return It rate was lowerfor girlsbecauseof theirlesserlife-timelaborforceparticipation. Enrollment Rates Anothermeasure high schoolparticipation attainment the enrollof and is mentrate.Figure5 contains publicandprivatesecondary-school enrollment datafor the New England, MiddleAtlantic,EastNorthCentral,and Pacific number youths 14 to regions.The enrollment dividesthe enrollment rate by 17-years old.46The graphshows similartrendsto those using graduation rates.47 All nonsouthern increasesin secondary regionsexperiencedsubstantial school educationduringthe 1910 to 1940 period, althoughthose of the industrial Northinitiallylagged.Certain regions,mainlyin the Westandthe the Plains,underwent mostrapidexpansionandin those areasboth graduation andenrollment rateswere high even beforethe onset of the GreatDepression. Most northernstates experienced an expansion in secondary schooling duringthe 1930s that was far greaterthan that experiencedby other regions. Growthwas so rapidin all non-Southregions that rates of enrollmentand even graduation the 1930s remainedin place for more in than twenty years. Finally,teenagedgirls enrolledin and graduated from high school at far greaterratesthandid boys. of Soaringenrollments the magnitude experienced-increasing by three to fourtimesin the 15 yearsfrom 1920to 1935-meant thatschooldistricts, acrossthe UnitedStates,hadto contendwith vastly increasedexpenditures andcorresponding burdens. tax Eachhigh school student cost twice as much to educate per year, in terms of variablecosts, than did each elementary 44 45 See Rotella,FromHome;and Goldin,Understanding. Goldin and Katz, "Human Capital," estimate that the return to a year of high school was 12.5 percent for (unmarried)females and 1 1.9 percent for males 18 to 34 years old. The return to a year of granmmar school was just 6.0 percent for females but 10.1 percent for males. Thus the only way that a young woman could reap a reasonable return from schooling was to continue with high school. 46 Enrollments are probably "fall," although each state was allowed to employ its own definition, such as enrollments throughout the year. The U.S. Department of Health, Education, and Welfare systematized the data in the 1960s to make them all fall enrollments. 4' Data on enrollments are not given due to lack of space. 364 Goldin 0.8 0.6 0.4 bD ~ 0.2 ~ -e0 1910 1915 1920 New England East North Central ~ - Middle Atlantic Pacific 1925 1930 1935 Year 1940 1945 1950 1955 1960 FIGURE 5 TOTAL (PUBLIC AND PRIVATE) SECONDARY SCHOOL ENROLLMENT RATES IN FOUR REGIONS Notes: Males and females are combined. The number of graduates is divided by the approximate number of 17-year olds in the state. Constant growth rate interpolations of population data are made between census years. Sources: See Figure 2. school student, and more high-school youths also meant the building of more schools.48It should not be surprising,therefore, that the states with the greatest wealth per capita-those of the Far West and Great Plains-were those with the earliest and most rapid increases in secondary education rates. Contemporaneous and Retrospective Education Data The data just discussed are contemporaneous flow data, whereas the growth calculation that I began with uses educational stock data. The stock data can be built up from flow data or, alternatively, can be directly estimated using the U.S. federal population census ever since 1940. But the educational stock data for certain cohorts are largely inconsistent with the flow data, often included in the same statistical reference sources (such as U.S. Bureau of the Census, Historical Statistics). The numbers I have produced allow the differences to be further explored by sex, geographic area, and secondary school grade and thereby point to the factors that may ac48 The estimate of the ratio of high school to elementary school variable costs comes from various state reports. It can also be calculated indirectly from the data on pupil-teacher ratios in U.S. Bureau of the Census, Historical Statistics (series H 423, 428) and teachers' wages. The city-level data from Goldin, "How America," produce a similar estimate. from High School America's Graduation 365 count for the disparities. In brief, the stock data from the 1940 federal population census for individuals older than 35 years greatly overstate the educational attainmentof the U.S. population, particularlyhigh school graduation. The contemporaneous flow data show that there were insufficient spaces in high schools to have accommodated the large numbers of Americans who claimed, in their older years, to have completed twelve years of schooling, a claim that has been, at times, erroneously translated into high school graduation. In comparing educational attainment by cohort across censuses, many researchershave noticed that cohorts gain schooling over time even if they did not go back to school.49People, it is claimed, simply inflate their educational attainment, and, moreover, this inflation is larger during periods of substantialgains in schooling.50It should be apparentfrom Figures 2, 3, and 5 that educational attainment increased enormously from 1920 to 1940. If there were ever a census when older cohorts would have seen reason to inflate their educational attainment, 1940 would have been it. There are also other reasons why the 1940 educational stock data might not be consistent with the flow data. Many older Americans, particularly those who grew up in ruralareas, had been students in ungraded elementary schools. Others were left back, and thus attended school for more years than grades. Although the census question was framed in terms of grades and not years, it may not have been fully understood. Because the 1940 census was the first to request informationon educational attainment,census takers may have been insufficiently trained to elicit accurate responses in a population with so many who had attended ungraded schools. Large differences are apparent between the 1940 census data on high school graduation and those from the contemporaneous series in Figure 6, either for all males (upper panel) or those born in nonsouthem states (lower panel). The graduation rate computed from the 1940 census is the fraction of males at each birth year who claimed to have completed twelve grades, translatedin the 1940 census as graduatinghigh school. The contemporaneous data are lined up with the 1940 census data by assuming that graduation occurred at age 18, but the results are largely unaffected if graduation occurred at seventeen or nineteen. The top two lines in each panel of Figure 6 are derived from the 1940 census. One gives the graduationrate for males born in the given area (either 49 See Denison, Trends; and Folger and Nam, Education, for comments that the retrospective schooling data from the 1940 federal population census do not agree with the contemporaneous data from the Commissioner of Education reports. 50 Consider an individual who finished tenth grade in 1910 but who did not graduate high school because no school in his district went through grade 12. If five years later there was such a high school and individuals five years younger, but identical to him in other ways, had graduated, he could legitimately feel that he did as well. 366 0.55 Goldin males -1940 -~1940 census, native-born census, native- & foreign-bornmales data,constructed 0.4 ~ 0.45 0.35 data,official contemporaneous *contemporaneous 20.4 0 0.3 '~0.25 e~ 0 0.2 CA~ 0.1 0.05 1870 1880 1890 1900 Birth Year 1910 1920 '~0.55 ~0.45 20.40.352.4 0. -.1940 census, native-born males, non-South census, native- & foreign-bornmales, non-South -1940 contemporaneousdata, constructed,non-South 0.3 ~0.25 0.2 -0.15 0. 0 z 1870 1880 1890 1900 Birth Year 1910 1920 FIGuRE 6 CONTEMPORANEOUS AND RETROSPECTIVE HIGHSCHOOLGRADUATION DATA FOR COHORTS BORN 1872 TO 1919, ENTIREUNITEDSTATES AND NON-SOUTHERN STATES Notes:1940 census,native-born males is the proportion menbornin the UnitedStates(or bornin of nonsouthern twelveor moreyearsof education who hadcomipleted twelfth and the states)who reported grade. The birth year is (1940 minus age). Southernstates are those in the census regions South Atlantic, East South Central,and West South Centraland includethe Districtof Columbia.1940 males is the sameas above exceptthatthe foreign-~born included census, native-&.foreign-born are andareassigned theirareaof current to residence. are data, Contemporaneous constructed the number of malegraduates frompublicandprivatehigh schools dividedby the numberof 17-yearolds. America's Graduation from High School FIGURE 6-continued 367 Sources: 1940 census, native-born males and 1940 census, native- & foreign-born males are from 1940 Public Use Micro-data Sample (entire 1/00 sample). See ICPSR, Census. Contemporaneous data, official are from U.S. Department of Education, 120 Years, table 19. Contemporaneous data, constructed are from Bureau of Education, Report (various years); and Bureau [Office] of Education, Biennial Survey (various years). See Figure 2 for additional sources. the entire United States or the non-South) and thus excludes the foreign born. The other line (that directly beneath it) includes the foreign born currently living in the area. If the foreign born did not continue their formal education once they entered the United States, then the upper line is correct for comparison with the contemporaneous data. If, instead, all the foreign born in an area received their high school education there, the lower line is correct. The true line is probably closer to the top one. The third and fourth lines (upper panel only) come from the contemporaneous graduation data just discussed. In sum, the graduation rate data from the 1940 population census for cohorts that were 18-years old in the 1910 to 1920 period are overstated, possibly by a factor of 1.5 or 2.51 The implication of the finding for the growth calculation with which I began is that it is underestimated. The growth of the actual educational stock is greaterthan that given by the population census stock estimate. The implication for estimates of the returns to education for the older cohorts is more complicated.52The new state-level data on graduation and enrollment suggest caution in using the stock estimates of schooling for the older cohorts of this century derived from the 1940 U.S. population census.53 EXPLAINING VARIATION IN SECONDARY SCHOOLING ACROSS STATES That a unique and major educational change swept much of the United States in the period from 1910 to 1940 is clear. Also apparentis that it was very much a grass-roots movement, albeit with some coordination through state legislation and the prompting of various national educational associa51The results also hold for females and for each region. The differences are less, however, for the higher-education regions, as can be seen in Figure 6 in the comparison between the entire United States and the non-South. 52 The bias in the return to education in the 1940 census, for example, could go in either direction. If the more economically successful inflate their educational status, then the return to education will be biased upward. But if the bias in reporting is independent of economic success, the return to a year of education will be biased downward. 53 Similar calculations with regard to grade ten completion show that it was mainly high school graduation that was overstated. The percentages completing grade ten in the stock and flow data are much closer. Because the high school completion data are overstated in the 1940 census, it is likely that the college data are as well. The 1950 population census probably also contains the same bias for older cohorts, but by then various cohorts had returned to school under the G.I. Bill and, furthermore, the G.E.D. (General Educational Development) examination existed. 368 Goldin tions. Why the movementoccurredandwhatexplainshow it was differentially receivedacrossAmericais a long andinvolved tale.54 Althoughthe UnitedStateswas uniquein the developedworldin embracing universaland free (nonvocational) secondaryschool educationin the therewere vast differencesacrossAmericain the earlypartof this century, timing of change(see Figures2, 3, and 5). One of the most intriguingasand data pects of the state-levelenrollment graduation is thatthe areasof the the UnitedStatesthatembraced high schoolmovementearliestandin which it grewthe fastestwere oftenfarming, and ranching, miningstates.But they shared muchin commonfor theywererelatively wealthyandmorehomogeneous in this respectthanwere most of the easternstates. Eventhougheducational were advances, to the earlytwentieth up century, the most rapidin New England, high schoolmovementin the 1920s spread belt." clear across the nationtracingout what I will termthe "educational MaineandMassachusetts, some otherstatesof New England,had high and rates of secondaryschool enrollment 1910. But by the early 1920s the by stateswith the highestrateswerethoseof the FarWestandthe Plains.It was and Oregon to Indiana,Iowa, Kansas,Nebraska,California,Washington, that the high school movement next spread, literally leaping across the countryfrom its birthplacein New England.The westernstates had high levels of wealthper capitaandtheircitizenswere often more homogeneous in their income and wealth thanwere those in the industrialized East and were also statesin which vast distanceswould have to be Midwest. They traversed bringstudentsto high schools. to High schools were generallymore thanfour times largerthanruralelethe schoolsandoftenrequired consolidation union of thatmany or mentary districts.The automobile, schoolbus, andimprovedroadswere of critithe cal importance secondary to school educationin Americaandmay account for why the movementtook off so rapidlyin the 191Osand 1920s. High levels of income and wealth, and its relativehomogeneityacross families, were not the only factors that led to differencesin secondary schooling across states. Industrial employmentis another.Many types of servedto reducethe level of high school educamanufacturing employment tion by increasing opportunity of schooling.I do not mean to imply the cost thathigh schooleducation not havea substantial of return youths did rate for facedwith manufacturing opportunities close to home. But I do want to job suggest that young people and possibly their parentshad a shortertime horizonwhenjobs for teens were close at hand.The role of manufacturing jobs in slowing the high school movementmeshes with the fact that the onset of the Great Depression produceda large increase in secondary 54 For further discussion see Goldin and Katz, "Why the United States." America's GraduationfromHigh School 369 schooling in the more industrial states and that World War II decreased graduation rates in all regions. The variables that are thought to determine educational outcomes involve both the individual (the demand side) and the educational jurisdiction (the supply side). For the individual, the factors are generally summarized in the expected returns from schooling. Returns should incorporate the costs of education, most importantlyopportunity costs. Parental resources will also detennine schooling decisions, to the extent that the capital market is imperfect, which it presumably was for these youths.55Finally, the financial resources of the school district will determine its ability to tax and to borrow. Two sets of economic factors, other than the North-South divide, are of overwhelming importance in explaining cross-state variation in secondary schooling just before the Great Depression. They are per capita wealth or income (or agriculturalincome per agriculturalworker), and the manufacturing wage and the extent of manufacturingemployment. The following analysis is offered as suggestive of the factors that encouraged the high school movement to spread and it should be thought of as an exploration in reduced-form mode.56 The high school graduation rate by state is regressed on the two sets of variables and some characteristics of the states' population (percentages foreign born and urban) as well as a South dummy. One set of variables includes the percentage of the work force in manufacturing,the manufacturing wage, and an interaction between the two. The other set is either per capita wealth or agriculturalincome per agriculturalworker. The year chosen for the graduationrate data is 1928, although most any other year in the 1920s will do just as well; the years available for independent variables range between 1920 and 1930.57The results are given in Table 3. There are two defensible assumptions in the estimation. One is that manufacturing wages were largely unaffected by the labor supply decisions of youths. Youths were but a small percentage of the manufacturing labor market. The second assumption is that the wage of educated workers (say, that of office workers) did not vary significantly across states. Evidence indicates that non-production-worker earnings varied far less than did production-worker wages.58Given the two assumptions, the regressions in 5 See Parsonsand Goldin,"Parental Altruism," a model of educational for investment. Also see GoldinandKatz,"Whythe UnitedStates"on incorporating publicchoice into suchmodels. 56 The robustness these resultsis demonstrated Goldinand Katz, "Whythe United States," of in whichalso includesfactors accounting statewealthdistribution, for religion,communitystability,and public universityavailability, amongothers.Similarresultsare shown to hold at the city level. 5 A full panelcannotbe assembled becausetherearetoo few variablesthataretime variantat the samefrequency the educationdata.But see Goldinand Katz,"Whythe UnitedStates"for various as regressions explainingthe changein graduation ratesbetween 1910 and 1928, and 1928 and 1938 58 Data assembled fromthe census of manufactures 230 non-Southcities revealthatalthough for manufacturing wagesdid varysubstantially acrossregions,thewagesformore-educated workers(cler- 370 Goldin Table 3 can be viewed as tracing out a demand curve for schooling as a function of the opportunitycost of school (the manufacturingwage) holding the tax base (wealth or agricultural income) and other factors constant. Both the manufacturing wage and per capita wealth (or, alternatively, agriculturalincome per farn worker) have positive coefficients, as does the percentage of the work force in manufacturing. But the interaction term between the two variables (manufacturingpercentage and wage) shows that when the percentage in manufacturing is somewhat greater than its mean (when it exceeds 0.272, using column 1), the relationship between the graduation rate and the wage is negative. Similarly, when the manufacturing wage is a bit largerthan its mean (when it exceeds $1,241 using colunm 1), the relationship between the graduationrate and the percentage of the labor force in manufacturing is also negative. The findings can be interpreted in the following manner. Wealth per capita and agriculturalincome per agriculturalworker are reasonable measures of the tax base for most states in the period.59But as the proportion of the work force in manufacturingincreased and as the wages of manufacturing workers did, the opportunity cost of schooling rose. Thus when the manufacturingsector was large enough, an increase in its wage served as a potent drag on the education of youths. When the manufacturing sector is sufficiently large, the two sets of variables produce effects akin to the classic ones of income and substitution. Thus states with high levels of education were those that were wealthy (income effect), but were also those without large manufacturing sectors (substitution effect). But because agriculturewas a declining sector by employment, states rich in agriculture, yet poor in manufacturing, schooled their youth to become educated migrants.60States with moderate to high, agricultural income per capita yet small urban and manufacturing sectors (for example, Kansas, Nebraska, Nevada) eventually exported their educated populations to some other states. States at the other end of the income spectrum, such as those in much of the South with poor agriculturaland manufacturing sectors, were ical workers) did not. The ratio of the highest to lowest mean clerical earnings across the six non-South census regions was 1.089 in 1915, but was 1.366 for production workers (both weighted by the relevant number of workers). For a description of the data see Goldin and Katz, "Decline." 59 Total per capita income is less strongly related to education than is income from agriculture or wealth. The reason may have something to do with the role of homogeneity of income rather than its level and the negative impact of income from certain sources, such as manufacturing. 60 Many state education reports openly acknowledged that the educated children of the farm population would leave ruralareas. That for Kansas c.1917 noted: "One theory holds that ... ruralhigh school [youth] ... should be given chiefly such work as will best fit [them] for rural vocations and for citizenship in a ruralcommunity. The other holds that ... the chief reason for establishing high schools in the country communities is to enable the boys and girls ... to enjoy the same sort of high-school advantages as are enjoyed by those who live in the larger centers. The latter point of view seems to be the prevailing one" (Kansas Superintendent of Public Instruction, Biennial Report, 1916-1918, pp. 61-63). America's Graduation from High School 371 also net exporters of their populations. But in this case, the populations, in comparison, were less well-educated. States somewhere in the middle (for example, California, Ohio, Wisconsin, Michigan) were the net receivers, and thus the relationship between education and net migration, across all states, is an inverted u-shaped one.6" CONCLUDINGREMARKS The period from 1910 to 1940 witnessed the second great transformation of American schooling: the rise of the public high school. The first transformation was the spread of the common school in the mid-nineteenth century and the third was the rise of college education in the post-World War II period. Neither was as spectacular in its increase as was the rise of the public high school. In 1938 the median eighteen-year-old in the East North Centralregion, for example, was a high school graduate, although in 1920 he had stood a 15 percent chance of having a diploma. Much of the gain in educational attainmentfrom 1900 to 1970 came from increased high school, not college, attendance. High-school enrollment and graduationrates in the North and West rose rapidly between 1920 and 1935. Rates achieved by the mid-1930s were maintained for the next 15 to 20 years before increasing again to recent levels. At the state level, higher per capita wealth or income (or agricultural income per worker) had a strong positive effect on secondary schooling, a likely result of the richer tax base and higher income parents. Manufacturing labor demand, primarilythe availability of operative positions in certain industries, was a drag on education, and the Great Depression provided an ironic fillip to schooling through its elimination of many jobs, particularly those for teenage males. The rapid increase in high-school enrollments and graduation rates was the response of a latent demand for more schooling to the building of public high schools, a decrease in transportcosts, and a change in the curriculum. School propagandacampaigns cannot be ruled out as an influence, but they probably operated more like informative, ratherthan deceptive, advertising. The returnsto education at the startof the high-school movement were high, although the rewards mainly accrued to those who shifted from manualjobs, in manufacturing and agriculture, to white-collar positions.62 The demand 61 A regression of the net migration rate, between 1920 and 1930 or between 1930 and 1940, on the graduation rate in 1928 (and its square) shows the quadraticrelationship. The relationship is unaffected by the inclusion of per capita income in 1929 and the share of the labor force in manufacturing in 1930. When estimated across only the northernand Midwestern states the relationship is negative. States with higher levels of education earlier in the twentieth century had a larger out-migration or smaller inmigration. The migration data are from Lee, Miller, Brainerd, and Easterlin, Population Redistribution. 62 Male clerical (office) workers in the Northeast and Midwest earned 1.5 times what male production workers did in both 1910 and 1915. The ratios were higher for females. Male clerical and su- 372 Goldin just declinedsubstantially priorto the rapid forjuveniles in manufacturing increasein high schools andthe demandfor white-collarworkers,particuA larlyin offices, rose.63 decadeor two later,in the 1920s, the demandfor high school-educatedblue-collar workers would increase in the more of industries the day.64 and capital-intensive higher-technology States thatpulled aheadof the pack in secondaryschooling were most areas with high, and narrowlydistributed,per often in rich agricultural activity.Income convergence income but moderatemanufacturing capita But strongin the 1929 to 1947 period.65 amongthe stateswas exceptionally the states with higher secondary-schoolgraduationrates in the 1920s achieved income levels significantlygreaterthan a simple convergence at Educationmattered the individuallevel and at process would predict.66 causedthe United Statesto thatof the state.Thathigherlevels of education grow more rapidlythan other countriesis a more difficultpropositionto prove. One fact is certain,however.In the 1910 to 1940 period America and economiesin botheducation income. pulledfaraheadof otherindustrial from high school; the people in virtuallyno other Americahad graduated countryhad. pervisoryworkersearnedabout twice what male productionworkersdid from 1890 to 1915. See white-collar workers,1890 to 1939. Goldinand Katz,"Decline,"for wage seriesby sex for ordinary in 63 The demandfor juveniles decreasedin the early 1900s, particularly those industriesthat laborplus capitalfor craftworkersplusjuveniles. See adultforeign-born and mechanized substituted of of Started.One piece of evidencein support this claim is thatthe proportion all Getting Osterman, sectordeclinedfrom 1900 to 1910 relativeto workingin the manufacturing maleteenagers employed table7. for the sameproportion male workersof all ages. See Goldin,"HowAmerica," 64 See Goldinand Katz,"Origins." 65 "Convergence." See, for example,Barroand Sala-l-Martin, 66 table8. See Goldin,"HowAmerica," REFERENCES Angus, David, and Jeffrey Mirel. "Rhetoric and Reality: American High School Course Taking, 1928-1990." In Learning from the Past: What History Teaches Us About School Reform, edited by Diane Ravitch and Mans A. Vinovskis, 295-328. Baltimore, MD: Johns Hopkins University Press, 1995. Barro, Robert. Determinants of Economic Growth:A Cross-County Empirical Study. Cambridge, MA: MIT Press, 1997. Barro, Robert, and Xavier Sala-I-Martin. "Convergence across States and Regions," Brookings Papers on Economic Activity, Vol 1. Washington, DC: The Brookings Institution, 1991: 107-82. Denison, Edward F. Trendsin American Economic Growth, 1929-1982. Washington, DC: The Brookings Institution, 1985. Folger, John K., and Charles B. Nam. Education of the American Population. A 1960 Census Monograph. Washington, DC: GPO, 1967. Fuller, Bruce. "Youth Job Structure and School Enrollment, 1890-1920," Sociology of Education 56, no. 3 (1983): 145-56. America's Graduation from High School 373 the Goldin, Claudia. Understanding Gender Gap: An Economic History of American Women. New York:OxfordUniversityPress, 1990. . "How America Graduatedfrom High School: 1910 to 1960," National Bureau of EconomicResearchWorking PaperNo. 4762, June 1994. . "Appendix to How America Graduatedfrom High School: An Exploratory Study, 1910to 1960,"NationalBureauof EconomicResearch-Development the American of EconomyWorking PaperNo. 57, June 1994. F. Goldin,Claudia,andLawrence Katz."TheDecline of Non-Competing Groups: Changes in the Premium Education,1890to 1940,"NationalBureauof EconomicResearch to WorkingPaperNo. 5202, August 1995. . "Technology, Skill, and the Wage Structure: Insightsfrom the Past,"American EconomicReview 86, no. 2 (1996): 252-57. __ . "Why the United States Led in Education: Lessons from Secondary School Expansion, 1910 to 1940,"NationalBureauof Economic ResearchWorkingPaper No. 6144, August 1997. __ . "The Origins of Technology-Skill Complementarity," Quarterly Journal of Economics,forthcoming. . "Human Capital and Social Capital: The Rise of Secondary Schooling in America, 1910 to 1940,"JournalofInterdisciplinary History,forthcoming. The WageStructure in Goldin,Claudia,andRobertA. Margo."TheGreatCompression: the UnitedStatesat Mid-Century," Journalof Economics107, no. 1 (1992): Quarterly 1-34. for Inter-university Consortium PoliticalandSocialResearch(ICPSR).Censusof Population, 1940 (UnitedStates): Public Use MicrodataSample. Ann Arbor,MI: ICPSR, 1984. 2 MA: MITPress, 1995. Jorgenson,Dale W. Productivity. volumes. Cambridge, KansasSuperintendent Public Instruction. of BiennialReportof the State Superintendent of Public Instruction.Topeka,KS, 1916-1918. Katz, Michael B. The Irony of Early School Reform:EducationalInnovation in MidNineteenthCentury MA: Harvard Massachusetts. Cambridge, UniversityPress, 1968. Krug, EdwardA. The Shaping of the AmericanHigh School: 1880-1920. Madison: Universityof WisconsinPress, 1964. . The Shaping of the American High School: Volume 2 1920-1941. Madison: Universityof WisconsinPress, 1972. Kuznets,Simon,Ann RatnerMiller,and RichardA. Easterlin. PopulationRedistribution and Economic Growth:UnitedStates, 1870-1950. Vol. 2. Analyses of Economic PA: Change.Philadelphia, The AmericanPhilosophicalSociety, 1960. DavidE TheMakingof an AmericanHigh School: The Credentials Marketand Labaree, the Central High School of Philadelphia, 1838-1939. New Haven, CT: Yale UniversityPress, 1988. Lee, Everett S., Ann Ratner Miller, Carol P. Brainerd, and Richard A. Easterlin. PopulationRedistribution EconomicGrowth:UnitedStates, 1870-1950. Vol. 1. and Methodological Considerations and Reference Tables. Philadelphia, PA: The AmericanPhilosophicalSociety, 1957. on Education. Massachusetts Commission Industrial Technical and Reportof the Commission of Industrialand Technical Education.New York:TeachersCollege, Columbia University,1906. of National CatholicWelfareConference,Department Education.Summaryof Catholic Education[variousyears].Washington, DC: NationalCatholicWelfareConference, variousyears. 374 Goldin and Olson,KeithW. TheG.I.Bill, the Veterans, the Colleges. Lexington:The University Press of Kentucky,1974 Labor Market.Cambridge,MA: The MIT Osterman,Paul. GettingStarted: The Youth Press, 1980. Altruismand Self Interest:Child Labor Parsons, Donald, and ClaudiaGoldin. "Parental CenturyAmericanFamilies,"EconomicInquiry27, no. 4 among Late-Nineteenth (1989): 637-59. among the Irish, ItalPerlmann, Joel.EthnicDifferences:Schoolingand Social Structure ians, Jews, and Blacks in an American City, 1880-1935. Cambridge:Cambridge UniversityPress, 1988. Reese, William J. The Origins of the AmericanHigh School. New Haven, CT: Yale UniversityPress, 1995. Ringer,FritzK. Educationand Societyin ModernEurope.Bloomington:IndianaUniversity Press, 1979. at 1870-1930. Ann Arbor,MI: Rotella, Elyce. FromHome to Office: U.S. Women Work, UMI ResearchPress, 1981. Smith, James P., and Michael P. Ward. Women'sWagesand Workin the Twentieth 1984. Century.SantaMonica, CA: The RandCorporation, Educationin the UnitedStates.New York:MacMillanCo., Smith,WilliamA. Secondary 1932. of InternaTrow,Martin."TheSecond Transformation AmericanSecondaryEducation," tional Journalof Comparative Sociology 2, no. 2 (1961): 144-66. MA: Harvard UniversityPress, 1974. Tyack,David. TheOneBest System.Cambridge, Ueda, Reed. Avenuesto Adulthood:The Originsof the High School and Social Mobility in an AmericanSuburb.New York:Cambridge UniversityPress, 1984. U.S. Bureauof the Census.FifteenthCensusof the UnitedStates: 1930. Population.Vol. DC, GPO: 1932. 3. Washington, _ Historical Statistics of the United States. Washington, DC: GPO, 1975. U.S. Bureauof Education. of Reportof the Commissioner Education[for the year].Washington,DC: GPO, variousyears. DC: U.S. Bureau[Office]of Education. BiennialSurveyofEducation[years].Washington, GPO, variousyears. U.S. Department of Commerce. Statistical Abstract of the United States, 1925. DC: GPO, 1926. Washington, . Statistical Abstract of the United States, 1938. Washington, DC: GPO, 1939. . Bureau of Economic Analysis. State Personal Income by State: Estimates for DC: GPO, 1984. 1929-1982. Washington, U.S. Departnentof Education.120 Years ofAmericanEducation:A StatisticalPortrait. DC: GPO, 1993. Washington, U.S. Departmentof Health, Education,and Welfare.Statistics of State School Systems DC: GPO,variousyears. [years].Washington, . National Center for Education Statistics. Digest of Education Statistics [year]. DC: GPO, variousyears. Washington, . Statistics of Non-Public Secondary Schools [years]. Washington, DC: GPO, vari- ous years. of Vinovskis,Matis A. TheOriginsofPublic High Schools:A Reexamination the Beverly Madison,Universityof WisconsinPress, 1985. High School Controversy. PamelaB., andPhilipJ. O'Connell."TheFamilyEconomy,Work,and Education Walters, in Journalof Sociology93, no. Participation the UnitedStates,1890-1940,"American 5 (1988): 1116-52.