Cohort crowding: how resources affect collegiate attainment

Cohort Crowding: How Resources Affect Collegiate AttainmentAnalyses of college enrollment and college completion rates typically focus on demandsidefactors including the financial attractiveness of a college education and the availability offinancial aid. Implicit in this line of research is the assumption that colleges and universitiesadjust supply fully with changes in demand. However, there is good reason to believe that supplyis not even close to perfectly elastic. The higher education market in the United States isdominated by public and non-profit production, and colleges and universities receive considerablesubsidies from state, federal, and private sources. Indeed, Winston (1999) estimates that studentfees cover only about 12% percent of total educational costs at public colleges and universities inthe U.S. Parallel estimates of the average share of costs covered by student tuition and fees atprivate colleges and universities are substantially higher, but still less than 50%, and much lowerat the most selective institutions. Because consumers pay only a fraction of the cost ofproduction, changes in demand are unlikely to be accommodated fully by colleges anduniversities without commensurate increases in non-tuition revenue. For this reason, publicinvestment in higher education plays a crucial role in determining the degrees produced and thesupply of college-educated workers to the labor market.Because only part of any observed change in current expenditures is likely to beexogenous and expenditures translate to resources with long lags, direct assessment of the effectof resources per student on degree outcomes using measures of current expenditures or stateappropriations is difficult. As an alternative, we focus the effects of “crowding out” brought aboutby relatively large cohorts vying for a limited number of spaces in the higher education sector.Increases in the college-age population shift out the higher education demand function. Yet, in theabsence of complete per capita adjustment in educational subsidies, increasing enrollment islikely to reduce educational resources per student in aggregate. In this circumstance – when anincrease in cohort size does not produce a proportionate increase in resources – large cohorts areessentially crowded out of the higher education sector, affecting educational attainment. To thisend, variation in cohort size leads to an indirect test of how the availability of resources at thecollege and university level affects collegiate attainment.We are not the first to note that within-state growth in cohort size is negatively related tocollegiate attainment. Card and Lemieux (2000) have found similar results using Census andCPS data. The innovation of our research is in the analysis of possible mechanisms leading toreductions in college completion among large cohorts. We present what we believe to becompelling evidence that limited adjustment among colleges and universities on the supply-sideof the higher education market is what causes the crowding out of large cohorts. The negative2

association between state specific cohort size and college completion rates represents evidence ofthe importance of capacity constraints for understanding college graduation rates.The first section of this analysis sets forth the defining characteristics of the market forhigher education in the context of a rudimentary but illustrative model. In short, the high degreeof public subsidy, the mixed market of public and non-profit providers, and the differentiatednature of the product in higher education lead to adjustments along both quality and quantitydimensions in response to demand shocks, with the predicted change in undergraduate degreeattainment ultimately less than proportionate to the change in cohort size. The second sectionturns to the measurement of variation within states over time in cohort size, and the effect of thisvariable on collegiate attainment. The third section focuses more directly on the nature ofadjustments in higher education by type of institution and the link between current revenuesavailable to higher education and cohort size. In the fourth section, we assess the empiricalstrength of the alternative explanation that larger cohorts may have lower demand for college ifcollege preparedness (e.g., pre-collegiate achievement) is also connected to cohort size; but wefind such demand-side factors account for little of the reduction in college completion amonglarge cohorts. The basic finding is that the incomplete adjustment of resources in response topopulation changes leads to a substantial reduction in collegiate attainment rates.The conclusion is clear: resources matter in the production of higher education.Declines in the public investment in our colleges and universities will have a large impact on boththe quantity and the quality of college graduates in the country.I. The Market for Higher Education and Responses to Demand ShocksAlthough the scale of the higher education industry has increased substantially since theWorld War II era, the salient features of the industry – including a mix of public and non-profitprovision, substantial public subsidies, and a market differentiated by quality – have remainedconstant (Goldin and Katz, 1999; Winston, 1999). The substantial levels of private and publicsubsidies in the market for higher education have fundamental effects on the “quantity” ofcollegiate attainment and the “quality” (or resource intensity) of this product.Public colleges and universities, in particular, expanded markedly in the immediate postwaryears. The share of undergraduate degrees awarded by public institutions has increased fromabout 50% in 1947 to nearly 66% in the most recent academic year, while the share of totalcollege enrollment (including enrollment at sub-baccalaureate institutions) is even higher atnearly 80%. Within the public sector, nearly all states have differentiated systems of public higher3

education, including a mix of community colleges open to all students, comprehensive four-yearinstitutions with modest admissions standards, and flagship public research universities.State colleges and universities provide the majority of higher education enrollmentopportunities to their residents at a steep discount in price, made possible through substantialpublic subsidies. Table 1 summarizes the role of state appropriations in the current revenuestream of colleges and universities. Notably, state and local support accounted for 57.5% ofcurrent revenues at the public community colleges and more than 36% of revenues atcomprehensive four-year institutions. Although the share of state support (30%) is somewhatlower at public research universities, the levels tend to be higher, and these institutions also relyon a range of other non-tuition revenues including resources from private donations and researchsupport from the federal government. Overall, the data in this table show that tuition paymentsfall far short of covering total educational costs. 1As shown in the final columns of Table 1, there are substantial differences in tuitionprices between in-state and out-of-state tuition charges 2 and between private and publicinstitutions. It is these large differences in expected tuition prices that lead to a concentration onpublic colleges and universities as the institutions determining the choic e set within a state forstudents likely to be at the margin of college enrollment and completion.Private institutions, particularly the selective colleges and research universities, also relyheavily on non-tuition sources of revenue, including private contributions and income fromendowment. Nevertheless, the selective private institutions with substantial subsidies enroll avery small share of the undergraduate population. While the state-level is the appropriate unit ofanalysis for the consideration of the collegiate attainment of students likely to be at the margin ofdegree attainment, the most competitive level of colleges and universities (e.g., StanfordUniversity, Harvard University) define a highly integrated and national market that is unlikely toinclude students at the margin of college enrollment (Hoxby, 2000).Understanding how colleges and universities respond to changes in demand for highereducation, particularly those brought about by variation in the size of cohorts entering college,requires consideration of how non-tuition revenue affects the tradeoffs faced by these institutions.1 Estimates by Winston mentioned earlier place the degree of subsidy even higher, becauseconventional statements of revenues and expenses for colleges and universities fail to account for capitalcosts (both depreciation and the opportunity cost of funds), which account for an average of 25% ofeducational costs.2 The average ratio of in-state tuition to out-of-state tuition was .35 in 1996-97, amounting to anaverage tuition price difference of $6,145; examples of large differences include Colorado and Vermontwhere the difference between in -state and out-of-state undergraduate charges at the state universityexceeded $11,500 in 1996-97.4

education and requires that there be no residual shareholders. 8 Most generally, the nondistributionL nconstraint is: nT ( , ) − nc(q)+ L = 0 , 9 where c(q) reflects per-student costs as a functionn popof quality (or resources), with c’(q)>0 and total costs of n c(q). 10 It follows that the per-studentcost of the quality of education provided is simply equal to tuition plus the per-student subsidy(L/n). In this constraint, tuition revenue is endogenous and reflects the tuition rate determined byinverse demand function for college enrollment. This specification is straightforward for privateinstitutions, but it is less clear that the treatment of tuition as endogenous applies fully to publiccolleges and universities, as state legislators and governors may have considerable sway in thedetermination of tuition. 11The nonlinear constraint provides much of the intuition for analyzing the institutionaladjustment to a demand shock. The constraint, illustrated in Figure 1 for the case in which tuitionis determined exogenously, is nonlinear and asymptotes at the level of quality equal to tuition, asthe number of students enrolled goes to infinity. 12 At low levels of enrollment the price ofexpansion is high as the relative change in per-student resource subsidy is substantial. At higherlevels of enrollment (and lower levels of quality), the quantity-quality tradeoff is less pronouncedas per-student subsidy becomes a tiny piece of the cost of education for any individual student.Adding a form of subsidy that varied with enrollment would simply raise the constraint (and theasymptote). The constraint shifts with changes in subsidy. As the subsidy level decreases, there isan “income effect,” which implies a reduction in enrollment and quality, and a “price effect,”8 This model builds on Hansmann’s (1981) effort to illustrate the potential quality-quantity choicefor a performing arts organization when private donations depend on the quality of the presentation.9 In some states, a component of state funding for higher education depends on enrollment. Forsimplicity, we have ignored this fact. What is important is the L rises less than proportionately with n andthis is always true. The model explains the response of students and college administrations to changes inthe size of the college-age population. Additionally, one might imagine that the cost of producing a givenlevel of quality, c, would depend not just on the level of quality produced, but also on enrollment rates – asenrollment rates rise, selectivity falls. Letting c be a function of q and n/pop, rather than just q, wouldcomplicate without changing the basic story.10 Implicitly we are assuming constant returns to scale, which we think of as a reasonableassumption for the long run.11 A survey of State Higher Education Executive Officers finds that legislatures in10 statesexplicitly set tuition in practice or in statute (Christal, 1997). In other states, tuition determination isgenerally the responsibility of governing boards or state higher education authorities, with these authoritiesoften composed of political appointees (Kane, Orszag, and Gunter, 2003).12 The basic “shape” of the constraint would be the same with the endogenous treatment of tuitionthough the slope would be steeper and the function would “flatten” at a greater rate. With tuition fixed, theslope of the constraint is2− L / n , while endogenous tuition yields∂c(q)− L= T2 1+∂nn1Tpop2L−2n.6

elative to cohort size) was more severe before 1970s than it was after, though the difference wasnot statistically significant.Another variable of interest in measuring higher educational attainment is enrollment.Enrollment represents an “input” to educational attainment while degrees awarded are anoutcome. Table 3 replicates the specifications in Table 2, focusing on total enrollment rather thanthe baccalaureate degree outcome. First in both periods, and particularly for the longer seriesfrom 1954-1996, the estimates for elasticity with respect to cohort size are larger for enrollment(0.89 and 0.79) than for BA degrees conferred (.71 and .62), though they are still significantlyless than one in all of the specifications.The majority of the variation in the cohort size variables employed in these specificationsis related to broad shifts in the population, with regions in the South and West growing muchmore rapidly than those in the North East and North Central. Indeed 83% of the variation in ourcohort size variable can be explained by a simple linear trend. To eliminate these secular trendsfrom the data, we have estimated specifications that include state-specific linear trends (see evennumberedcolumns of Tables 2 and 3). 16 Doing so further attenuates the link between cohort sizeand degrees awarded substantially.For the purpose of interpreting the substantial differences between results with andwithout the inclusion of state specific trends, it is important to understand the source and natureof the variation in cohort size. The cross-state variation in changes in the size of the college-agepopulation is being driven by cross-state and cross- country migration flows. 17 Presumablysecular trends in the size of the college-age population are being driven by secular trends in thelocus of economic activity in the country. Indeed, the within-state-and-year correlation betweenthe size of the 18-year old population and employment (both measured in logs) is 0.7. Aftersecular trends have been removed this correlation drops quite dramatically to between 0.1 and 0.3depending on the time period used. We suspect that even here changes in the size of the collegeagepopulation are being largely driven by changes in the location of economic activity, but thatthe timing is off – young adults migrate in response to economic opportunities, but have collegeagechildren one or two decades later.To get a perspective on the time series properties or our population measures, followingBaker, Benjamin, and Stanger (1999), we decomposed our measured into orthogonal time seriescomponents using Fourier series techniques. With 43 years of data (1954-1996), we employ the16 We also experimented with quadratic trends, which show results very similar to those presented.17 While at the national level changes in cohort size during the second half of the twentieth centurywere driven by changes in cohort fertility rates, within state and year, there is essentially no associationbetween family size and cohort size.11

Results Based on Census DataAn alternative way to measure the effect of cohort size on degree completion employsCensus micro data on BA completion rates, organized by state of birth and birth cohort, as therelevant demographic variable. 19 Using alternative data sources serves to underscore thereliability of our results while necessarily answering somewhat different questions. Table 4presents estimates of the effects of cohort size (measured in terms of year and state of birth) ondegree completion using Census data, along with estimates of the effects on degrees conferredusing the institutional data for comparison. Because the dependent variable for the Census-basedestimates is the cohort share (fraction with a BA) while for the institutional-based results it is thenumber of BAs, subtracting one from the latter results will put these two estimates on the samefooting. Regressions of the share of the population born in each state and year with a BA degreeon cohort size (this time measured with cohort size at the year of birth) show a substantial declinein college completion shares with changes in cohort size. The magnitude of the estimated declinewhen using the Census data is somewhat smaller than the magnitude of the decline when usingthe institutional data – that is -0.26 or -0.24 versus -0.34 or –0.29, but the differences between thepoint estimates are not qualitatively important, nor are they statistically significant. 20 However,while including state-specific trends in the specifications using institutional data tends tostrengthen the evidence for “crowding out,” doing so significantly weakens the evidence in thespecifications using Census data.models using the institutional data restriction attention to census years. Results from the analysis producedresults that are completely consistent with results based on all years of data19 Note that when we use Census micro data organized by state and year of birth to measure thecollege completion rates, the regression estimates with the dependent variable specified as a completionrate (ln [BA/N]) are not equal to the specification in levels (ln[BA])minus one. The reason is that that thedenominator in the measure of the completion rate, a measure of cohort size calculated from the micro data,is not identical to the population measure from Vital Statistics sources. What we observe in the data is anegative correlation between the error (the difference in the measures) and the measure of population fromthe Vital Statistics source. While the simple correlation between the population measure from the VitalStatistics and the population measure from the Census micro data is 0.996, a regression of the Censusmeasure on the Vital Statistics measure (both measured in logs), produces a coefficient of .952 (.0127). Theresult is that regression estimates using the Census micro data in levels will produce estimates suggesting alower elasticity of completion with respect to population than related estimates with the dependent variablespecified as a completion rate.20 These estimates are quite similar to those presented in Card and Lemieux (2000). Depending onthe specification and cohorts used in their regressions, they estimate a 10% increase in cohort size isassociated with a decline in college completion of between 0.4 and 0.8 percentage points for men andbetween 0.3 and 0.6 percentage points for women. To convert these estimates into elasticities, one needs todivide by the average fraction receiving a BA in these cohorts – somewhat more than 0.2 for men andsomewhat less than 0.2 for women. Using different samples, somewhat different specifications, and vitalstatistics rather than census data to determine cohort size, we obtain results quite similar to those reportedby Card and Lemieux.13

Explaining the Divergence between Results Using Census and Institutional DataPossible explanations for the large difference between the results based on theinstitutional data versus the Census data with the inclusion of state-specific trends include eitherdifferences in the timing of the cohort size variable or differences in the nature of the measuresused.In terms of the timing of the cohort-based measure, cohort size at birth is used in theCensus regressions and cohort size at the typical college enrollment age (age 18) is used in theinstitutional measures. Given our maintained hypothesis that the primary reason for theassociation between cohort size and college completion rates involves cohort crowding atcolleges and universities, the right measure of cohort size is the cohort size during the typicalcollege going years. Cohort size at birth can then be thought of as an error-ridden proxy forcohort size 18 to 22 years later. Indeed, the correlation between year-of-birth measures of cohortsize and measures of the same cohort at age 18 after removing state and year effects is 0.84.Removing state-specific trends drops the simple correlation to 0.37. Thus, relatively short-runfluctuations in the size of cohorts measured in the year of birth are relatively weakly related to thesize of the college-age population in a state 18 years later. The modest degree of cohort crowdingshown in columns (2) and (4) of Table 4 is consistent with the notion that, once state specifictrends are included in the specification, cohort size at birth is a relatively poor proxy for cohortsize at college attendance age.Nevertheless, there are a number of reasons why the institutional measure of BAproduction regressed on cohort size at age 18 might tend to exaggerate estimates of cohortcrowding. It would not be too surprising to find that being born into a large (state-specific) cohortwould tend to increase the odds that a person went out of state to attend college or temporarilypostponed going to college. In either case our estimates based on the institutional data, wouldoverstate the magnitude of the effect of cohort size on college completion (crowding). TheResidence and Migration survey, which has been conducted periodically by the Department ofEducation since the 1940s, provides some information on enrollment by state of residence andcollege. We estimate the elasticity of total enrollment at the state level with respect to population[0.61 (0.11)] and the elasticity of enrollment of state residents with respect to population [0.62(0.08)]. 21 If increased out-of-state enrollment were a substantial part of the response to cohortexpansion, we would see larger elasticities (closer to 1) for enrollment of state residents than fortotal enrollment in the state. Yet, the elasticities for these two measures are virtually identical.21 Estimating equations include state and year fixed effects; the addition of state-specific trendsdoes not materially affect these results.14

The available empirical evidence does not support the notion that relatively large cohorts ofcollege students are “absorbed” by neighboring states. In fact, it appears that the enrollmentbehavior of state residents is closely tied to in-state college opportunitie s, with elasticities veryclose to the estimates using the full panel of BA degree data in Table 2. 22In addition, it seems plausible that time to degree may tend to rise with cohort size. Ifcollege-age individuals in large cohorts faced fewer resources, they may find college lessattractive, and marginal students might delay college enrollment. 23 What is more, fewer resourcesmight translate into slower progress while in school. Empirically, calculations using the CurrentPopulation Survey suggest that increasing cohort size does increase the age at degree receipt,though the magnitude of this effect is by no means large enough to explain the overall effect oncompletion rates. One concern is that, in the institutional measures of degree attainment, increasesin time to degree will tend to bias estimates of the effect of cohort size on the number of BAsproduced per cohort; yet, such errors should be offsetting as time to degree expands and contractswith cohort size. 24 All in all, we doubt these inter-temporal substitution effects have much effecton our estimates.To sum up, while the institutional measures are not, conceptually what we would like tohave (they are period, rather than cohort measures and represent graduates from institutionswithin a state rather than graduates of the resident population in a state), our evidence suggeststhat regressions that use these measure will produce estimates of the effect of cohort size oncollege completion that are close to the ones we would have obtained from conceptually22 Focusing on just the flagships s chools, we do find that out-of-state enrollment tends to fall in theface of large cohorts; however this effect is small enough and the flagship schools represent a small enoughshare of total enrollment, that this pattern hardly shows up when we combine data on all colleges anduniversities.23 If shifts in population size occurred at fairly high frequency, individuals who are part ofparticularly large cohorts might find it optimal to postpone college for several years. However, this kind ofsubstitution is unlikely to be plausible in large scale as most of the movements in cohort size observedoccur at relatively low frequency. Moreover, such inter-temporal adjustment would likely reduce thenumber of years over which an individual would get the benefits of a college education, thus reducing thereturn to the degree.24 A simple example will illustrate. Suppose that the fraction of a cohort ultimately receiving a BAremains constant, but cohort size grows at a rate of 1% per year. Also suppose that, as a result on theincrease in cohort size, the age at which individuals typically receive their BA rises from 22 to 23 yearsover a 10-year period. During this period we would observe an approximate 10% increase in the size of thepopulation, but only a 9% increase in the size of the number of BAs. As a result, we would underestimate?ln(BA) / ? ln(Pop) by about 10% – or estimate crowd out when none exists. This example might make itseem as if the data we are using will lead us to overestimate crowd out effects. However, if one runs thisthough experiment in reverse, one sees that during periods of declining cohort size (and decreased time todegree), we will tend to overestimate ?ln(BA) / ? ln(Pop). Since the variables we use in our regressions areall deviations from year and state means, some states will be experiencing increases in cohort size (relativeto the mean), while others are experiencing decreases in cohort size and the bias introduced by the fact thatwe are using period rather than cohort measures will tend to cancel out.15

appropriate measures. Where there are discrepancies between the Census measures and theinstitutional measures, they are due to the effects of differences in the age of observation for thecorresponding measure of cohort size. These results are consistent with the interpretation that the“crowding” that matters occurs at the collegiate level.III. Higher Education Resources and Cohort SizeIn considering the supply-side adjustment of colleges and universities to the reductionsand expansions in cohort size, colleges and universities face the potential margins of adjustmentof price (tuition), quantity (enrollment) and resources per student.While price is usually a natural margin for adjustment to a demand shock, both the fact that forpublic institutions tuition covers such a small fraction of total costs, and the unusual link betweenpolitical actors at the state level and public colleges and universities undercuts this potentialchannel for adjustment to market forces. 25 In fact, the link between cohort size and tuitioncharged by public institutions is not positive. Instead, the evidence from regressions of the tuitioncharged to in-state students by public colleges and universities on cohort size points clearly in theother direction: a 10% increase in cohort size is associated with a 2.7% decrease in tuition at thestate comprehensive schools and 4.2% decrease at flagship institutions. As in previousspecifications, these regressions include state and year fixed effects. Since there were secularincreases in tuition at state schools over the period under study, the right way to think about theseresults is that they suggest that tuition rose less rapidly in states that experienced populationgrowth over the period in question. It is only among the non-research private institutions that wesee the expected positive link between tuition and cohort size. When state and localappropriations are included in the regressions, the effect of cohort size on tuition approaches zero,with increases in state appropriations having a substantial negative effect on tuition prices atpublic universities. These results suggest that state universities raise tuition to cover expenses as alast resort in adverse budgetary environments. 26Beyond tuition levels, we consider the relationship between all sources of revenue andcohort size in the regressions presented in Table 5. For total education expenditures, state25 For example, in 1999, the Virginia legislature instituted a 20-percent reduction in tuition atpublic colleges as part of the fulfillment of campaign promises of then Governor James Gilmore. Withrecent reductions in state tax revenues, states have turned to cutting appropriations and allowing publicinstitutions to increase tuition.26 When either state employment or state and local appropriations are included in the regressionthe effect of cohort size on tuition approaches zero, with increases in either state employment of stateappropriations having a substantial negative effect on tuition prices. These results would seem to confirmthat states tend to raise tuitions when state economies are in trouble.16

appropriations, federal revenue, and tuition and fee revenue, regressions on the college-agepopulation produce a remarkably similar story with these resource measures failing to adjustproportionately with changes in cohort size. Since the overall adjustment in education and generalexpenditures is less than the change in enrollment with changes in cohort size, the inference isthat resources per student decline with increases in cohort size. State appropriations fail to keeppace, as well, indicating that subsidy per student does decline. Not surprisingly, federal resourcesshow the least adjustment and the coefficient on the population measure is indistinguishable fromzero with and without state trends. Tuition revenues do not appreciably fill the revenue gap, withchanges in cohort size associated with an elasticity of about 0.6. 27A central proposition of this analysis is that increases in total educational resources leadto increases in both enrollment and attainment. Including educational resources as an explanatoryvariable is problematic to the extent that increases in total expenditures come from additionaltuition revenue. State appropriations provide a more plausibly exogenous source of variation. InTable 6, we revisit the within-state regressions of collegiate attainment on cohort size with theaddition of a measure of expenditures instrumented with appropriations. In the models withoutstate-specific trends, we find that the coefficients on expenditures are 0.34 (0.10) for the totalenrollment outcome and 0.21 (0.09) for the BA outcome. When state -specific trends are includedthe estimated effect of expenditures (instrumented with appropriations) goes to zero. Theseresults tend to underline our concerns with using state appropriations as explanatory variables; wesuspect that after state-specific trends are included in the regressions, variation in current accountstate expenditures do not do a very good job reflecting variation in resources available tostudents.Variation in relative adjustments in resources and enrollment across institution types is tobe expected from the differences in the objective functions of colleges and universities.Community colleges and four-year institutions with modest admission requirements tend to place“access” at the center of their mission, attempting to provide enrollment opportunities for allapplicants who meet minimum qualifications. At the other extreme, research universities andliberal arts colleges are likely to emphasize the role of student quality and resources per student asthey make choices at the margin between quality and quantity. Thus, what the model predicts issubstantial accommodation among two-year institutions to changes in cohort size and littleaccommodation (and, perhaps, a ratcheting up of selectivity) among universities.27 Note that this tuition revenue measure is implicitly a weighted average of public and privatetuition as it represents total tuition revenue for all institutions in a state.17

analysis is whether results we attribute to adjustments on the supply side of the higher market areinstead related to demand-side explanations. Two related concerns surface. First, relatively largecohorts may be distinguished by adverse demographic or economic shocks that have direct effectson collegiate attainment. For example, if big cohorts are distinguished by low parental educationor large family size, such “compositional effects” might account for reduced college completionrather than crowding out on the supply side of the market. Secondly, membership in a relativelylarge birth cohort may dilute educational resources at the elementary and secondary levels, whichwould also reduce college preparedness. Both types of effects imply that the change in collegedemand may be far less than a change in cohort size.To understand how compositional factors such as race, parental education, and familysize are tied to cohort size, we examined 16 year olds and their parental characteristics by statefor the 1970, 1980, and 1990 decennial Census enumerations. 30 Regressions of a particularparental demographic characteristic (e.g., share black or number of children) on the size of the16-year-old population in a state and Census year, with year and state fixed effects, provides anindication of how changes in cohort size are tied to observed demographic characteristics. Morechildren in a family imply fewer resources per child, and educational attainment—both formaland informal—is likely to be negatively affected by a decrease in the time and financial resourcesavailable to each child within the family (Becker, 1981; Willis, 1973). Similarly, other familycharacteristics such as maternal education, family structure, and race may affect the educationalresources available outside of schools.For the most part, we find effects that are not statistically significant and economicallysmall. This applies to measures of family size, parental education, and ethnicity. When we weightby state size, the only significant effect is on Hispanic ethnicity, with a coefficient of .27indicating that a 2.7% increase in the share of college-age students of Hispanic origin wouldfollow a 10% increase in the size of the 16-year-old population. Even with the upper endestimates of the effect of Hispanic origin on college completion, these compositional effectscould not explain a sizable share of the reduction of college completion with increases in cohortsize.Individuals in relatively large cohorts may also face diminished resources in elementaryand secondary schools, with these resource effects reducing college preparedness and completion.Examination of the link between resources per student at the primary and secondary levels andcohort size helps to place this question in perspective. First, resources per student are negatively30 We focus on 16 rather than 18 year olds to try to minimize the extent to which individuals mayhave moved out of the households in which they grew up.19

for these cohorts. 36 Adding to the demographic pressure on higher education resources created byincreasing cohort size are persistently tight state budget circumstances in which higher educationcompetes with entitlement programs and the rising cost of healthcare (Kane, Orszag, and Gunter,2003).The impending collision of large cohorts and limited public resources in higher educationis not just a predicament for colleges and universities, but a potential crisis in economic growthfor decades to come if the flow of college-educated workers to the labor force is further curtailed.Because tuition is a small enough part of the total revenues of higher education, it would takeenormous increases in tuition to provide the necessary resources for colleges and universities toexpand and maintain quality, with such a shift adversely affecting the level and distribution ofenrollment. At issue is how colleges and universities, particularly in the public sector, can raiseresources from non-tuition sources to expand the pool of collegiate opportunities without dilutingquality. While states have, historically, been central to the development of infrastructure and theprovision of subsidy in higher education, the current fiscal constraints and restrictions againstdeficit funding may limit the capacity of states to make investments in human capital throughpublic colleges and universities.36 For example one report estimated that the number of students seeking to enroll at California’spublic colleges would rise by 36% between 2000 and 2010. Former University of California PresidentClark Kerr has dubbed this projected increase in demand “Tidal Wave II” as nearly three-quarters of theprojected growth is strictly attributable to population growth (Schmidt, 1999).23

ReferencesBaker, Michael, Dwayne Benjamin, Shuchita Stanger. 1999. “The Highs and Lows of the Minimum WageEffect: A Time-Series Cross-Section Study of the Canadian Law.” Journal of Labor Economics, Vol. 17,No. 2. (April), pp. 318-350.Becker, Gary. 1981. A Treatise on the Family. Cambridge: Harvard University Press.Card, David and Alan Krueger. 1992. “Does School Quality Matter? Returns to Education and theCharacteristics of Public Schools in the United States.” Journal of Political Economy , Vol. 100, No. 1.Card, David and Thomas Lemieux. 2001. “Can Falling Supply Explain the Rising Return to College forYounger Men?” Quarterly Journal of Economics (May), 705-746.Card, David and Thomas Lemieux. 2000. “Dropout and Enrollment Trends in the Post-War Period: What WentWrong in the 1970s?” in (J. Gruber, ed.) An Economic Analysis of Risky Behavior Among Youth.Chicago: University of Chicago Press, 439-482.Cheit, Earl. 1971. The New Depression in Higher Education. Berkeley: Carnegie Commission on HigherEducation.Christal, Melodie. 1997. “State tuition and fee policies, 1996-1997.” Denver: State Higher Education Exe cutiveOfficers.Finn, Chester. 1978. Scholars, dollars, and bureaucrats. Washington: Brookings Institution.Fortin, Nicole. 2003. “Higher Education Policies and Decelerating Wage Inequality: Cross-State Evidence fromthe 1990s.” Mimeo, June.Goldin, Claudia and Larry Katz. 1999. “The Shaping of Higher Education: The Formative Years in the UnitedStates, 1890 to 1940.” The Journal of Economic Perspectives, Vol. 13, No. 1 (Winter).Goldin, Claudia. 1999. “A Brief History of Education in the United States.” NBER Historical Paper 119.Graham, Hugh and Nancy Diamond. 1997. The Rise of American Research Universities. Baltimore: JohnsHopkins Press.Hansmann, H. 1981. “Nonprofit Enterprises in the Performing Arts.” Bell Journal of Economics.Hanushek, Eric; Steven G. Rivkin, Lori L. Taylor . 1996. “Aggregation and the Estimated Effects of SchoolResources.” The Review of Economics and Statistics. Vol. 78, No. 4. (Nov.), pp. 611-627.Hoxby, Caroline. 2000. “The Effects of Geographic Integration and Increasing Competition in the Market forCollege Education.” Mimeo.Kane, T. and C. Rouse. 1995. “Labor Market Returns to Two- and Four-Year Colleges: Is a Credit a Credit andDo Degrees Matter?” American Economic Review 85, 600-614.Kane, T.; Peter Orszag and David Gunter. 2003. “State Support for Higher Education, Medicaid, and theBusiness Cycle.” Mimeo.Krueger, Alan. 1999. “Experimental Estimates of Education Production Functions.” Quarterly Journal ofEconomics, Vol. 114, no. 2 (May): 497-532.Loeb, Susanna and John Bound. 1996. “The Effect of Measured School Inputs on Academic Achievement:Evidence from the 1920s, 1930s and 1940s Birth Cohorts.” The Review of Economics and Statistics, Vol.78, No. 4. (Nov., 1996), pp. 653-664.Poterba, James. 1997. “Demographic Structure and the Political Economy of Public Education.” Journal ofPublic Policy and Management.Schmidt, Peter. 1999. “Applications to California’s Public Colleges Predicted to Rise by 36% by 2010.” TheChronicle of Higher Education. October 1, 1999.Stapleton, David C. and Douglas Young. 1988. “Educational Attainment and Cohort Size.” Journal of LaborEconomics v6 n3 330-61.Welch, Finis. 1979. “Effects of Cohort Size on Earnings: The Baby Boom Babies' Financial Bust.” The Journalof Political Economy , Vol. 87, No. 5, Part 2: Education and Income Distribution. pp. S65-S97.24

Winston, Gordon. 1999. “Subsidies, Hierarchy and Peers: The Awkward Economics of Higher Education.”Journal of Economic Perspectives, Vol. 13, No. 1.Willis, Robert. 1973. “A New Approach to the Economic Theory of Fertility Behavior.” The Journal ofPolitical Economy , Vol. 81, No. 2, Part 2: New Economic Approaches to Fertility. (Mar. - Apr., 1973), pp.S14-S64.25

Figure 1: Public university adjustment to changes in cohort size absent changes in subsidya. Quality maximizationResources per student [c(q)]University indifferencecurve and q,n maximumDemand shifts in responseto change in cohort sizeStudent Enrollment [n]b. Enrollment maximizationResources per student [c(q)]Student demand increase inproportionto change in cohort size (?p)Change inenrollment (?e)Student Enrollment [n]Notes: Collegiate demand is drawn holding tuition constant, with the assumption of exogenoustuition setting.26

Figure 2: State-specific trends in cohort size and BA degrees conferredCaliforniaIowa11.6174B B B B BB BB B B P P P P P P P PP P P P P PPP PB B B B B B BBB B B B B BP B P PPPP P BP P P PB13.01589.78953B B B B B B B B B BP P P P P P PBB B B BP P P P B B B BB PB B BBBB B BPP10.9609Ln Inst BA degrees9.78313BBP PBP P B BBB PPBPBPB B B B PPBP PB BBBBP P P11.765Ln Population 18Ln Inst BA degrees8.3582BP P PB B BPBBPBPB BB BBPPBB PPBPBP PB B P PP P PPPPPP PP P PP P10.4265Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22MichiganNew YorkLn Inst BA degrees10.73019.25158P PB B B B BB BB BB B B BB B B BB B B B B B B B B BP P P P P P P PPPPBP P PPP P P P P P PBPPBPB BPBP B PPBPBBBBPBBP PBB B P P P P PB P BP P12.151311.3541Ln Population 18Ln Inst BA degrees11.496410.2844P P PB B B B B BB B P P P P P P P B B B B B B B B B BB PP B PB PB B BPPBP P PBPPB B P P P PBPPPPB PB BPBPBBPB BBBB BBPB B P P PBP P PB P PP12.693712.0947Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22LouisianaPennsylvaniaLn Inst BA degrees9.797528.22067B B B B B B BB B B B B B B B B B B B B BP P P PP P P P PP P B B BB B B PPPP PB PP P PPPB P P PPBP PB BB B PBPB B B B PPB B B PBB BP P PPPB PP PP P11.353510.6435Ln Population 18Ln Inst BA degrees11.08419.72609P PB B B BB B BB B B B B BB B B B B B B B B PB B BP P P P P P P PP PBPB B PPPPPB PPPP P P PBB B B B PPBPBPPPB B B PPBBP BBPBP P P P BP PBP12.33811.7675Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 2227

Table 1: Distribution of revenues, enrollment and tuition by type of institutionUndergr.Share of1st timeCurrent Fund Revenues (1996) Enr. Share students TuitionState Federal Private Endow. Tuition Aux. (% of Total) in-state In-state Out-of-& Local & Other 1996 1996 StateCommunity Colleges 57.5% 11.7% 1.0% 0.1% 20.2% 9.5% 37% 92.7% 1,814 4,362Other Public 36.3% 10.7% 4.0% 0.4% 18.3% 30.2% 33% 81.5% 2,725 6,981Flagship Public 29.0% 14.8% 6.4% 1.3% 17.2% 31.4% 9% 72.0% 3,493 9,998All Private 2.8% 10.3% 9.1% 5.1% 41.9% 30.8% 21% 54.2% 12,881Research I Private 2.3% 16.1% 9.5% 5.7% 22.9% 43.5% 2% 23.6% 19,814Liberal Arts Colleges 1.4% 3.0% 9.1% 10.5% 55.5% 20.5% 2% 43.2% 17,648Notes: Data are from authors’ tabulations using U.S. Department of Education, National Center for Education Statistics, Integrated PostsecondaryEducation Data System (IPEDS), "Finance, 1996-97", “Residence and Migration” and “Fall Enrollment, 1996" surveys.28

Table 2: Regression of BA degrees conferred on measures of cohort sizeDependent Variable: Ln (BA)All Institutions All Institutions Public InstitutionsAll Years Late Years Late Years(1954-1996) (1967-1996) (1967-1996)Explanatory (1) (2) (3) (4) (5) (6)Regression VariableA. No Weights Ln Pop 18 0.71 0.20 0.62 0.15 0.59 0.21(0.07) (0.08) (0.09) (0.12) (0.08) (0.13)B. Avg Pop Wt Ln Pop 18 0.63 0.20 0.56 -0.03 0.54 0.06(0.09) (0.07) (0.08) (0.10) (0.08) (0.17)State Fixed Effects Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes YesState Trends No Yes No Yes No YesNotes: Degree attainment measures are from institutional measures of degrees conferred.Population data are from single year of age tabulations of the Department of Census. See DataAppendix for details.29

Table 3: Regression of total enrollment on measures of cohort sizeDependent Variable: Ln (Total Enrollment)All Institutions All Institutions Public InstitutionsAll Years Late Years Late Years(1954-1996) (1967-1996) (1967-1996)Explanatory (1) (2) (3) (4) (5) (6)Regression VariableA. No Weights Ln Pop 18 0.89 0.21 0.79 -0.01 0.66 -0.02(0.15) (0.09) (0.18) (0.09) (0.16) (0.12)B. Avg Pop Wt Ln Pop 18 0.82 0.42 0.63 -0.18 0.49 -0.16(0.11) (0.18) (0.18) (0.11) (0.20) (0.14)State Fixed Effects Yes Yes Yes Yes Yes YesYear Fixed Effects Yes Yes Yes Yes Yes YesState Trends No Yes No Yes No YesNotes: Total enrollment measures are from institutional surveys. Population data are from singleyear of age tabulations of the Department of Census. See Data Appendix for details.30

Table 4: Regression of BA degrees per capita on cohort size,Comparison between Census and Institutional data1990 Census 2000 Census1954-87 Degree years 1954-96 Degree years1932-1965 Birth cohorts 1932-1974 Birth cohorts(1) (2) (3) (4)Census-based share born in state with BA on year of birth population variableLn Pop Coefficient -0.26 -0.11 -0.24 -0.13(0.05) (0.03) (0.03) (0.05)Institutional BA / Pop 18 on population 18 variableLn Pop Coefficient -0.34 -0.97 -0.29 -0.80(0.09) (0.09) (0.07) (0.08)State Effects Yes Yes Yes YesYear Effects Yes Yes Yes YesState Trends No Yes No YesNotes: In the top row, Columns (1)-(2) use observations by year and state of birth from the 1990Census and Vital Statistics measures of birth cohort size; Columns (3)-(4) use observations byyear and state of birth from the 2000 Census and Vital Statistics measures of birth cohort size.The second row corresponds to results in Table 2, with the difference that dependent variable ismeasured as the log of the ratio of BA degrees to cohort size rather than the log of the number ofBA degrees.31

Table 5: Regressions of revenue variables on populationCoefficient on Ln Population 18Explanatory Variable (1) (2)Ed & General Expenses 0.49 0.25(0.09) (0.10)State Appropriations 0.58 0.43(0.11) (0.13)Federal Revenue 0.27 0.04(0.19) (0.22)Tuition and Fee Revenue 0.61 0.26(0.08) (0.12)State Fixed Effects Y YYear Fixed Effect Y YState Trends N YNotes: Data extend from 1950-1996 (even years pre 1966). See Data Appendix for source notes.32

Table 6: Regressions of collegiate attainment on cohort size and resourcesRegressionIV EstimatesLn Total EnrollmentAll YearsLn Inst BAAll Years(1954-1996) (1954-1996)Explanatory (1) (2) (3) (4)VariableLn Education and 0.34 -0.12 0.21 -0.02& General Exp. (0.10) (0.09) (0.09) (0.10)Ln Population 18 0.76 0.16 0.60 0.23(0.14) (0.10) (0.09) (0.08)First StageLn State 0.35 0.26 0.34 0.24Appropriations (0.02) (0.02) (0.02) (0.02)State Fixed Effects Yes Yes Yes YesYear Fixed Effects Yes Yes Yes YesState Trends No Yes No YesNotes: Total enrollment measures and BA degrees conferred are from institutional surveys ofcolleges and universities. State appropriations are used as an instrument for education & generalexpenditures. Population data are from single year of age tabulations of the Department ofCensus. See Data Appendix for details.33

Table 7: Effects of cohort size on BA completion and enrollmentby type of institution, 1968-1996Ln UndergraduateFTE EnrLn BAon Ln Pop 18 on Ln Pop 18Community Colleges 0.82(0.37)Other Public 0.56 0.82(0.13) (0.22)Flagship Public 0.20 0.28(0.07) (0.05)All Private 0.99 1.20(0.42) (0.45)Research I Private -0.07 -0.19(0.36) (0.45)Liberal Arts Colleges 0.26 0.71(0.17) (0.17)Notes: Data on degrees conferred and enrollment by type of institution are tabulated frominstitution-level data from the HEGIS/IPEDs. Enrollment data are aligned with the populationage 18 in the state in the concurrent year; BA data are aligned with the population age 18 yearsold in the state 4 years prior, in accordance with the expectation of a four-year modal time todegree. See Data Appendix for further details. Each entry and associated standard error is thecoefficient on the log of the population measure in a regression that also includes state and yearfixed effects.34

Data AppendixThe primary sources of data for this analysis are: institutional surveys of colleges anduniversities, the decennial Census files, the decennial Census publications, population estimatesby the Census Bureau, and standard measures of labor markets characteristics.Population dataPopulation measures by state and age are primary to our analysis. There are two sourcesfor our population measures: age-specific estimates from the Census bureau and measures of birthcohort size from the Vital Statistics tabulations.Measures of birth cohort size for each state from 1928 – 1975 were entered from vitalstatistics data distributed by the National Center for Health Statistics. The original data camefrom birth registrations. (These are the cohorts that would have been 22 between 1950 and 1997).For the most recent three decades, data on population by state and single year of age areavailable through the Bureau of the Census website. For the years between Census enumerations,these numbers are estimates which take into account mortality and migration. See:http://eire.census.gov/popest/topics/methodology/stage98.txt for a discussion of thismethodology. For the years before 1970, we combine data on the state and single year of ageenumerations published in the U.S. Census Bureau State Volumes for 1950 (Table 51), 1960(Table 94) and 1970 (Table 19) and the total population in each state and year for single yearsfrom 1950-1970. Because the population measures between census enumerations are estimates,measurement error is a logical concern with these data. We discuss in the text why we do notbelieve this to be a significant problem.College enrollment and BA degree outcomesThe primary measures of collegiate attainment are collected from federal surveys ofcolleges and universities. The degree data are based on the annual “Earned Degrees Conferred”survey conducted by the National Center for Education Statistics (NCES), which records degreesawarded in the 12-month academic year from July to June. 37 The enrollment data are from the“Fall Enrollment” surveys which record the number of students enrolled in classes in the fall.Through 1986, these surveys were part of the larger NCES Higher Education General InformationSurvey (HEGIS), which was subsequently redesigned as the Integrated Postsecondary EducationData System (IPEDS) collection.The complete “Earned Degrees Conferred” survey records degrees and certificatescompleted by field and level, though this analysis makes use of only the total baccalaureatedegree series. Similarly, the “Opening (Fall) Enrollment” survey records enrollment by level(undergraduate/graduate) and full-time/part-time status. We work primarily with the totalenrollment variable as this is available from 1950 forward.Historical data (primarily in the years prior to 1966) are keypunched from publishedtabulations in the government document publications under the titles “Earned Degrees Conferred”37 In 1960-61, the survey began to separately delineate first professional and baccalaureatedegrees. Prior to this point, the two were combined, reflecting the fact that in the early part of the centuryfirst professional programs were concurrently undergraduate degree programs at some institutions.35

and “Opening Fall Enrollment.” Machine-readable data are employed after 1966 (1967 forenrollment), which allows for the distinction of institutions by control (public/private) andCarnegie Classification.Another source of data from institutional collections is the Residence and MigrationSurvey conducted periodically by the by the Office of Education and later the Department ofEducation. Data for the following years are available: 1949, 1958, 1963, 1968, 1972, 1975, 1979,1981, 1984, 1986, 1988, 1992, 1994, and 1996. This survey records first-time freshmanenrollment by state of residence and state of attendance.A final source of data based on institutional responses in the Annual Survey of Colleges,conducted each fall by the College Board. This proprietary survey asks colleges and universitiesabout the characteristics of the entering undergraduate class, admissions procedures, courseofferings and expectations for undergraduates. Machine readable data are available from 1986-2000,In addition to the data which record degrees awarded and enrollment in each year andstate by colleges and universities, the decennial Census enumerations record college attainment toindividuals by state of birth and age (or, implicitly, year of birth). As discussed in the text, theCensus data are conceptually different from the institutional data in that they do not record theyear or state of degree receipt. As is well known, the form of the Census question on educationalattainment changed to a degree-based question from an item that recorded years of attainmentwith the 1990 Census. When we employ data from the 1970 and 1980 enumerations, we treatcompleting 16 years as equivalent to BA degree receipt.Higher education finance variablesEach year as part of the institutional reporting to the federal government, colleges anduniversities complete a survey of institutional finances in which they report basic income andexpense items including the sources of revenues and expenses. We are particularly interested indistinguishing sources of public support from tuition and fee revenue, while also measuring thetotal level of educational expenditures. For this purpose, we focus on the collection of fourvariables:Revenues-Tuition and fees: Includes all tuition and fees assessed (net of refunds) againststudent for current operating purposes (except receipts from the Veterans’ Administration andFederal grants or contractual payments for research which are included in FederalAppropriations). Tuition and fees remissions or exemptions are assessed and reported as revenueeven though there is no intention of collecting from the student. An amount equal to suchremissions or exemptions is reflected as expenditures and classified in the category Scholarshipsand Fellowships or Staff Benefits (i.e. if for faculty, staff or their offspring’s tuition). Schools areinstructed to include tuition and fees collected by the institution, sent to the state and returned tothe institution in the form of appropriations.Revenues-State appropriations : Those monies received from or made available to aninstitution through acts of the state legislative body, except institutional fees and other incomereappropriated by the legislatures to the institution (i.e. tuition and fees collected by the institutionand returned to the institution in the form of appropriations are subtracted as they already appearas tuition and fees). This line item also includes Federal aid received through State channels andregional compacts.36

Revenues-Federal (Appropriations and Grants & Contracts): All other monies receivedfrom or made available to an institution by the Federal Government, excluding any income fromFederal land grants which is included in endowment earnings. Grants and contracts are revenuesfrom governmental agencies which are received or made available for specific projects orprograms. Examples are research projects or training programs.Expenditures-Education and General: Include the following categories of currentexpenditures:1) Executive and administrative offices, 2) Instructional departments , colleges andschools including office, equipment, laboratory expenses; and salaries of department heads,professors, and other instructional staff (including student assistants) technicians, secretaries,clerks, etc., 3) Extension and public services (e.g. non-degree courses, public lectures, or radiobroadcasts), 4) Libraries (salaries, wages, operating expenses, books, periodicals, binding, etc.),5) Operation and maintenance of the physical plant and 6) Organized research.In years prior to 1977, all state level financial data are from published tabulations, as wehave found the machine-readable data for early years (through Webcaspar) to be unreliable,presumably due to problems with imputations. For years 1950-58, the Biennial Survey ofEducation includes tables reporting these items at the state level. Beginning with 1959 the annualOffice of Education/Department of Education publication “Financial statistics of institutions ofhigher education” contains the tables with these data.We note that we have compared the measures of state appropriations with data reportedfrom state governments on appropriations and found these measures to be highly correlated. Datafrom state governments are compiled annually since 1960 by the Center for the Study ofEducation Policy at Illinois State in the Grapevine series. These data differ slightly frommeasures reported by institutions in that they also include state funds to non-collegiate highereducation institutions such as the administrative boards coordinating higher education.Data on tuition and fees, measured as a price, are available from 1970 to the current year.There are two sources for these data, which provide measures that are highly correlated. First, theDepartment of Education collects data on tuition and fees at the institution level as part of theInstitutional Characteristics section of the annual HEGIS/IPEDS surveys. In addition, theWashington Higher Education Coordinating Board conducts an annual survey of tuition and feesat public institutions, which includes data from 1972-73 to the present.Labor market variablesLabor market variables used in the analysis include the state specific unemployment rate,personal income and manufacturing wage bill. State-specific unemployment rates are availablefrom 1970 through the Bureau of Labor Statistics [http://www.bls.gov/lau/staadoc.htm]. Thepersonal income and manufacturing wage measures are from the U.S. Department of Commerce,Bureau of Economic Analysis, Regional Accounts Data and are available from 1958 to 2001[http://www.bea.doc.gov/bea/regional/spi/default.cfm#s2, the manufacturing wage bill is the s07series, industry code 400].In addition, we compute the state- and age-specific measures of the college wagepremium in the 1970, 1980, and 1990 Census enumerations. The adjusted average relative wagemeasures are computed as the return to exactly a BA Degree (or 16 years of completed education)from state-specific hourly wage regressions with a full set of controls for demographic and laborforce characteristics including, race, sex, part-time status.37

Primary-secondary school quality measuresMeasures of school quality experienced for students of different states and ages are fromthe data assembled by Card and Krueger (1992; see Appendix A of this paper) for the analysis ofthe effects of school quality on earnings. The original source for these data is the Biennial Surveyof Education in the United States and related materials. The specific measures employed in thisanalysis the pupil-teacher ratio, average term length, and average teacher salary from academicyear 1919-20 to academic year 1965-66 (coded by C-K as 1920-1960 for even years). To providea single measure of school quality for each single year birth cohort, we averaged the potentialschool quality experienced from ages 6 to 17 within a state. In examining the contemporaryrelationship between school resources and enrollments, we use data from Statistics of StateSchool Systems (1998) for 1970-1995, which records data on enrollments and resources (bothexpenditures and class size).38

Appendix Graphs: State-specific trends in cohort size and degrees conferredAlabamaArizonaLn Inst BA degrees9.95948.48177P P P PP PB B B B BPP PBPB B B B B PPB BB BB B B B B B B B B BP P PPP PPPB BP P P PPPPPBPP PBPPBB BPBP PPB B B B B B B B PP P BP PB B BB P11.23910.7051Ln Population 18Ln Inst BA degrees9.842787.36771BB BB BB B BP P P P P PP P P P P PP P P P P PB B B B B B B B B B B B B B B B B BP P BPPBP P P P P PBBB BPP PBPBPBB BBB P P P PB PBP PBPB B B BP P10.94569.35314Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22ArkansasCaliforniaLn Inst BA degrees9.115927.82285PBBB BPPPBB B PB BBP P P P P P PB B B B P PPB B B BBB BB B BBB B B B B B B B B B B B PPPPPP P P PPPP P PBPPPPB B B P PPBBPBBB PP PPP P10.652810.075Ln Population 18Ln Inst BA degrees11.61749.78313B B B B BP P P P P PP P P PB BP PB B B PB B P P P PB B B B B B B BB B B B P B PP PPP P BP P P PBBBP P BP P B BBB PPBPBB B B B P PPBP P B BB B B P P P13.015811.765Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22ColoradoConnecticutLn Inst BA degrees9.914728.22577PB B B BBP P P P P P PP PB B B B B B B B B B BB B B B B PB B B B PPP P P P P PB B P P P P PP PBP PBB B P P PB B PB PB B B B PB P PBBPPBP PBB B B PP P10.95779.81184Ln Population 18Ln Inst BA degrees9.629718.34806P PBB B B B B B B B PB BB BBB B B B B BB B B B BP P P P P P P P PPPP PPP P P PBPB P P PBPP PP P PBPBB B BBPBBB BPBB B P P P P PP PBBB B B P P10.963610.0771Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22DelawareFloridaLn Inst BA degrees8.394125.9162PPBB BB B BP PPBP PPBB BP P P P B PB B B B B B B BP P B PP PB BP P P PP P PB B B B B PPBP P PPB B B B B B B BB B PB B B B B BPPPB B PPPP PP P P9.479078.36869Ln Population 18Ln Inst BA degrees10.74238.39004B B B B B BP P P P P P PP P PP P P P P P P P PB B B B B B B B B BB B B B B B PPBB B PPP B P B P P PBP P B PPBB BP P BPP BPB B B B B B B P PBPBBBBPPP12.047110.5238Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 2239

GeorgiaIdahoLn Inst BA degrees10.21548.4855PB B B BP P P PP P P PP PPB PPPB B PPPB B B B B B B B B B B BB B B B B B B P P PPPPP PPBPBBP P P BPB BBBPB B B B B B BPB BBP PP P P PBPB P P11.587310.929Ln Population 18Ln Inst BA degrees8.409396.77194BB BBPB P PPB B B B B BBB BB B B B B B BBPP BBBB B P P P P PPPBBPBP P PPPPB BPBBPB B B B BPBB B PB BB B P P P P P PP P PP P P P P P P P PP9.826669.03181Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22IllinoisIndianaLn Inst BA degrees10.88249.51959P P P PPBB B B B B B BB BB B BB B B B B B B B BB B BP P P P P PPPBPPP BPPPPB PPP P P PBPPPP BB BP PBBPBB B B B BBP P PBPBBP P P PB B P P12.297111.5736Ln Population 18Ln Inst BA degrees10.35628.98695P PPB B P PB B BBBP P P P P PBB B B B BB B B B BB P B B B B B PB B PBP P P P PBP PPPP P PBP PB B BB PB PPBPBBPBBB PBBP P PBB P P P P PB B P11.61210.8758Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22IowaKansasLn Inst BA degrees9.789538.3582B B B BB BB B B B B B B B B B B B BP P P P PP PBP P P P PB B BBBB B BPPBPP P PB B BPPBBPPP P BPP P B BB BPBPPBB PPP PBPBP PB B P PP P PP10.960910.4265Ln Population 18Ln Inst BA degrees9.60738.27359B B B BP P P PPP PPBB BBB P PB B BB B B B B B B B BB PB B B B BBP P P P PPPBPB BPBBPPBPB B B B B P P PP PPPPBPBPBPB PB B B P P P P P P P P10.76310.1845Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22KentuckyLouisianaLn Inst BA degrees9.593838.21148P P PP PB B B B BPPB B B BB PPBBB PBB B BB B B B B B B BBB BP P P PPBP PPPPP PPP P PB BB BPP BPBPPB B B BPP B PPBPPPB PPBP PB B BP11.179410.6521Ln Population 18Ln Inst BA degrees9.797528.22067B B BB B B B B BB B B B B B B B B B BP P P P P P P P PBB B BB B B P PPPPP PBP P P P P PBP PPPBP PB BB BBPPB B B B PPB B B PBB BP P PPP P P PB P P11.353510.6435Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 2240

MaineMarylandLn Inst BA degrees8.695517.27932B B P P PP P B PB BPB B B B B B B B B B B B PB B B BBP P P PBBBBBP P P PP PBP P P P PBP P PPBP BPB PBBPPBBB BP BPPB PPBPBP PB B B PBP P9.990639.41632Ln Population 18Ln Inst BA degrees9.983088.17019P P P B B B B BB B B B BB B BB BB B B B B B P P P P P P PP PPP BBP P B B PPP P P PPB B P PPBPPBBP BPP PB BBPBB PB BB B B P P PB P PBB B P B PP P11.312810.3581Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22MassachusettsMichiganLn Inst BA degrees10.71569.46816P P PB B B B B B B B B B B B B B B B B B B B BP P P P PB BP P P P PPBPPP PBPP P PP PPB B B PBPPBP PB BPBB PPBP B BBB B PBBPB P P P P PB B P B P11.666810.9412Ln Population 18Ln Inst BA degrees10.73019.25158P PB B B B BB BB BB B B B B B B BB B B B B B B B B BP P P P P P P PPPPBP P PPP P P P P P PBPPBPB BPB PB PPBPBBBBPBBP PBB B P P P P PB P BP P12.151311.3541Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22MinnesotaMississippiLn Inst BA degrees10.1161B BB B BB B B B B B B BB B B B B B B B B B B B BP P P PP P BPP P P P P PPBPPPBP PP P P PPPBPB BP PB B P P PBB B PBBBPBBB P BPB B P P P P P P PP11.3293Ln Population 18Ln Inst BA degrees9.27547BBP BP P P PP P B BB P BBB B B B P PB B B B BB B B B B PB BB B PBPB PB PPPPPPPB PB BB P P PPPPBP PB PB B B PB B PPPBBP PPBB B PP P PP10.8599Ln Population 188.6700910.59897.90839B10.34391950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22MissouriMontanaLn Inst BA degrees10.2378.70384P P P P P B B B B BBB B B B B B B B B P P P PPB B B B B B B B BB B P PPPB B PP P PP PPBP PP PPBB BP PBPB PB B PB B PBBPBBPBPP PBB P P P P PB P11.47710.8413Ln Population 18Ln Inst BA degrees8.438586.77537P P PP P PPBB B BBB B BB B B B B BB B B B B B B BB B BBP P P P P PBPBPB B B B PP P P P P PBPBP P P P P PP PBBB B PBBBPBP PPPB PBPPB BP P P9.614548.97335Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 2241

NebraskaNevadaLn Inst BA degrees9.22079PB BB P BB B B PP P PBP B BPBP P PBB B B B B B B B B B B B P PPP B BB BBBPB PB B P P PBPPBP PB PP P PPPP PB B B PB10.3369Ln Population 18Ln Inst BA degrees8.13652P PB B B BP PP P P P PPPP P PBB BB B B B B B BP P P PB B B B B B P P P BB B B B BP P PP P P PB B P PBPBPB B BP B B P PBPBP B P9.63835Ln Population 187.79482P BB P P P PBPPB B P B P P9.749465.11199BPB B PB B BBP PBP7.531021950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22New HampshireNew JerseyLn Inst BA degrees8.943777.26333PBBB B B B B B B B B B B B B B B BPP P PP P P P P P P P P P PPB B BBPB B P BPPB B P PPBP P P PBB B BP PB B P PBPPB B B B B PP BBB B BPPP P PB P P9.799968.82513Ln Population 18Ln Inst BA degrees10.15898.66785PB B B B B B B B BB B P PBB B BBB B B B B B B B BP P PP P P P P P PPPBPPP PP P PBPBBP PPPP BPB B B B PP PBBB B B PBBP P P PBB PPBP PBBP P11.815410.984Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22New MexicoNew YorkLn Inst BA degrees8.707487.01031P B B BB B BB B B B B B B BB B B B B B B B B B BP P P PP P PPP P PBPP P P P P P P PB PBP P PPB P PPB B B PP PBBBP P P P PPB B BBB PBP PPBB B B B P10.19329.29271Ln Population 18Ln Inst BA degrees11.496410.2844P P PB B B B B B B B B B BB B B B B B B B B B B BP P P P P PPPP PPP BPBP P PBPPB B P P P PBPPPPBPB B PBBPBB BBPBB BBPPB BB B P P P P P P P P12.693712.0947Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22North CarolinaNorth DakotaLn Inst BA degrees10.398PB B B B BP PPP PPPBPP PPB BP PPB B B B B B B B B BB B B B B B PPPP PPPPP B B P PBBPPBPB BPBB PBP P11.6663Ln Population 18Ln Inst BA degrees8.46695BB B B B BB B BB B B B B B B BB B B B B B B B B BP P P P P P PBPP PPPBPPP PB B B P PPP PBBPPB B B PPPB BPPP P PB9.52289Ln Population 18PPP8.75211PBB BB B B PP B P PB B B P P P B PBP11.10126.80461PBBBB B BPP P PPPP P9.004961950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 2242

OhioOklahomaLn Inst BA degrees10.85239.49454B B B B BP PBBP P P BBBBP P B PP PB B BBBBBB B B B B BB B B B B B B B B B B B B BP P P P P PPPPP PBP P P P P PPBPPP P P PBPB B P PB PBP PP P P12.290811.5049Ln Population 18Ln Inst BA degrees9.663588.46968P PPBB B B B BBBB B B B B B B B B B BP P P P PBBP P P PB BB B BB PPPBPP P P P PPB P P P P PPBB BBPB P PBB B B B BPPBBP PP PBBPBP P PB P P10.950410.3531Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22OregonPennsylvaniaLn Inst BA degrees9.501148.04206PB B B B B B BPPPP P P P PB B B B B B B B B B BB B P P PB B BP B B B PP PBBP PP PP P PP P PB P PBBB BPP PB PB B BB B B PBB B B P P P P P P PP PB B P10.74539.84056Ln Population 18Ln Inst BA degrees11.08419.72609P PB B B B BBB BB B B B BB B B B B B B B B B P B BP P P P P P P PP PBP BPB PPB P P PP PPP P PBB B B B PPB PBPPP B B BPPBBP BPBPB P P P BPPBP12.33811.7675Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22Rhode islandSouth CarolinaLn Inst BA degrees9.148047.43426PB B B B B B BPBPPP P P PP P PB B B B B B B B B BB B B B B BP P PPP P P P PP PPP PBPBBBBPPP BP BB PPB B P PPPBB B PBB B B B PPB BB P P PB P9.876589.15074Ln Population 18Ln Inst BA degrees9.628988.05611PB BPBP P P P BP B BPPPB B B B B B B B BB B B B B P BPPB BP PPPPPPPPBPP P PBB B PPPBBP PPBB B B B PPB B B B B B B B PPPB B B B PPP P P P11.078410.4936Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22South DakotaTennesseeLn Inst BA degrees8.512386.8533B B B B BP P P P PP PP PBBB B BB B B B BPPBB B B B B B B B P B BB B BPBPPB BPB B B PPPPB B B BP PPBPPPPBP P PP P PPP PPBPBPBPB BP PP9.583149.04061Ln Population 18Ln Inst BA degrees9.935918.55468B BB BB B B B B P P P P P PBB B B B B B PB BBBB B PB B B B B PPPPPBPPPP PP P PP PPPBPPPBB BB PBPPPBPPB B B B BP BPBBP P PPB B B PP P11.374410.7978Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 2243

TexasUtahLn Inst BA degrees11.16719.55457P B B B BB BP P P P P P P PP PP PB B B B B B B B B B BB B B B B PB B PBP P PP P PPB PBP P P PB PBBB BP PBPPBBB B B B B PPBB PPPB P PB B B P PP P12.534311.6802Ln Population 18Ln Inst BA degrees9.633977.82684PP BB B PPP P P PPP PBP P P PPB B B B B BB B B B B B BB B B B B B B B PPP P P PPB BB PPB PBP PB BPBPPB PPBB B B BBPBB B PBP P P P P PB B P P10.35359.30002Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22VermontVirginiaLn Inst BA degrees8.456817.03703PB B B B B B BBBB BB P B P P P BB B B BB B B B BBP P P PB P P PP P PPBPPPBP P PBPPB P PPBPBBPBB P PPPB PPB B B BP P PBB PB BB B BP P P P9.3928.47305Ln Population 18Ln Inst BA degrees10.3398.39728P P PB B B B B B B B B B BB B B B B B B B B B B BPP P P PP P PPPPPP PPPPP P PPP PB B PPBBBBPPP PBP B B B PPBB B B B B PPBPBBBB B B P P P PP11.554110.8342Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22WashingtonWest VirginiaLn Inst BA degrees10.02098.54111P P PB B B BP P P P PP PB B P PP PPB B BB BB B BB B BB B B BP P P PP P PP P PB B B BBP P P BBBP BPBPB PBBPBB BB P P BP P PBB B PB BBP PP B P11.213110.2802Ln Population 18Ln Inst BA degrees9.109977.83439BP PBBPBB B BB B B PP PB B B B B B BP PPP P BBP B B B B P P B BB B P PP BPPPP PBP PPPPB B B B PPP P PBPPPPB B B P PPBBPBBB PPBBB B P10.472410.1044Ln Population 181950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 22WisconsinWyoming10.2295PB B BBB B B B B B B B B B B B BP P P P PPBP P P PB BB B B B BB PPB B PPP11.47867.52618P PB B P B BP PP B B B B B B B BBB B B P PBPBBPB B B B B B B B PPPPPP PP P PP P9.06277Ln Inst BA degreesB P P PPB B BBBPB PB B B P PBPP P P P PP PLn Population 18Ln Inst BA degreesPBB B PB BPBPB B PB B PP P PB BPPBLn Population 188.57622BBPBB BBP P P P P P PB P P P10.73985.90263P BPPBBP P PP PPBB B P8.208311950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 221950 1955 1960 1965 1970 1975 1980 1985 1990 1995Year When 2244