Do Business Cycles Affect State Appropriations to Higher Education?

Humphreys, Brad R.

Brad R. Humphreys [*]

Spending on higher education constitutes an important and increasing portion of state government spending and a major source of operating funds at public institutions of higher education. Anecdotal evidence suggests that state appropriations are subject to cyclical variation. An analysis of state appropriations to higher education, enrollment in two- and four-year public colleges and universities, and state-specific measures of the business cycle for all 50 states over the period 1969-1994 shows that state appropriations to higher education are highly sensitive to changes in the business cycle. A 1% change in real per capita income was, on average, associated with a 1.39% change in real state appropriations per full-time equivalent student enrolled. This implied decline in state government funding, coupled with the increase in enrollment in higher education during recessions reported by Betts and McFarland (1995), suggest that public institutions of higher education may experience fiscal stress during econom ic downturns. These results also suggest that state legislators and education policymakers should reconsider their higher education funding policies during recessions in order to allow public colleges and universities to provide dislocated workers with access to quality education and training during these periods.

1. Introduction and Motivation

Business cycles affect higher education in a number of important ways. Anecdotal evidence from college admissions officers and administrators, as well as empirical evidence on the behavior of students and institutions of higher education, suggest that enrollment in higher education is countercylical. For example, Betts and McFarland (1995) recently showed that enrollments in community colleges are highly responsive to changes in local labor market conditions. These authors found a 1% increase in the unemployment rate among all adults associated with a 4% increase in enrollment at community colleges in a large panel of community colleges across the United States. Leslie and Ramey (1986) found a similar relationship between economic conditions and enrollment using data aggregated to the regional level.

This paper examines the relationship between state appropriations to higher education and state-specific measures of business cycle conditions in order to better understand the effects of the business cycle on government funding of higher education. [1] I further investigate the relationship between the business cycle and state government appropriations to higher education reported by Betts and McFarland (1995) and Leslie and Ramey (1986). Both of these studies aggregated state and local government appropriations to higher education to the regional and national level. While data aggregated to the regional and national levels are readily available, decisions about government funding to higher education are made at the state and local levels, and these decisions may be affected more by local economic conditions than by aggregate economic conditions. If the timing of turning points in the business cycle varies across states, then aggregating government appropriations data across states may obscure important sta te-specific phenomena.

Most research has focused on the effects of economic conditions on demand for higher education. One notable exception is Leslie and Ramey (1986), who examined the relationship between enrollment, business cycles, and state appropriations to higher education using regional appropriations and enrollment data and a national measure of business cycle conditions. That study found a procyclical relationship between the NBER coincident index and regional-level government appropriations to higher education. For one region, the estimated elasticity of appropriations with respect to changes in the coincident index was three, suggesting that a 1% change in the NBER coincident index was associated with a 3% change in appropriations to higher education in that region. This study also highlighted the complex relationship between state funding for higher education, enrollment, and business cycle conditions by documenting a statistically significant relationship between enrollment and state appropriations at the regional le vel, although they found that the appropriations--enrollment relationship weakened after 1977.

Betts and McFarland (1995) also describe, in an informal way using aggregate data, a procyclical relationship between the business cycle and funding for higher education. These authors report that aggregate state and local government appropriations to higher education fell in the years 1971, 1975, 1980-1982 as well as in several years in the early 1990s; all of these years are either coincident with, or immediately following, years when the NBER Dating Committee has identified a cyclical downturn in the economy. The transmission mechanism linking business cycles and government appropriations to higher education during recessionary periods seems clear: recessions reduce tax revenues; state and local governments, most of which operate under legally required balanced budget restrictions, reduce appropriations to institutions of higher education, as well as other types of expenditures, in order to balance their budgets.

Betts and McFarland point out that these cyclical patterns of government support for higher education may place tremendous financial pressure on institutions of higher education during economic downturns, because the cuts in government funding happen at the same time enrollments are increasing. These financial crises affect community colleges because the open admissions policies of these institutions make them attractive to unemployed workers displaced by the recession. Public four-year institutions of higher education may also be adversely affected because these institutions typically rely on government appropriations to offset a large fraction of their operating and general expenditures.

An alternative explanation for rising enrollments and declining government appropriations discussed by Betts and McFarland is increasing returns to scale. If institutions of higher education operate at levels of enrollment where there are economies of scale, then these institutions could lower average operating costs by increasing enrollments that could in turn lead to lower government appropriations. In the case of community colleges, Betts and McFarland report that interviews with community college presidents revealed that these institutions operate at or above capacity nearly all of the time and that increases in enrollments were typically associated with increases in average operating costs. In the case of four-year colleges, the existing empirical evidence suggests that most institutions of higher education operate at enrollment levels where there are constant returns to scale, indicating that increases in enrollment would not lead to decreases in average operating costs, although there is a great deal of disagreement in this extensive literature. However, recent research in Getz, Sigfried, and Zhang (1991) and Koshal and Koshal (1999) find no evidence of increasing returns to scale in institutions of higher education. This evidence makes it unlikely that increasing returns to scale can explain the behavior of government appropriations to higher education during periods of rising enrollments.

The results presented in this paper suggest that state appropriations to higher education are quite sensitive to business cycles. The estimated elasticity of state appropriations per Mi-time equivalent (FTE) student is 1.39, suggesting that a 1% change in real personal income per capita is associated with a 1.39% change in real appropriations per FTE. Further, after accounting for the countercyclical behavior of enrollments, it appears that real appropriations fall more during recessionary years than they rise during expansionary years.

2. Empirical Methodology and Data

The financing of higher education takes place in a complex, dynamic setting. Institutions of higher education raise revenues from tuition and fees, grants from all levels of government, research support from public and private grantors, and other sources. Revenues generated from tuition and fees depend on enrollment, which in turn depends on complex decisions made by households. Government appropriations to higher education are determined by legislators and bureaucrats who have complex objectives and agendas.

Considerable variation exists in the process by which states determine the funding of public higher education. The National Association of State Budget Officers (1996) surveyed states in order to document this process. Based on the results of this survey, the processes by which states decide on appropriations to higher education appear to be idiosyncratic. Few patterns emerge. Eleven states reported operating on a biennial budget cycle, and initial budget requests originated from a variety of sources. Institutions made the initial budget request in 12 states; higher education regulatory bodies like boards of regents or boards of higher education made the initial budget request in 15 states; state budget offices made the initial budget request in five states; the governor made the initial budget request in six states. The remaining states were not included in the survey. In each case, the budgeting process involved oversight by many other areas of state government. Several states reported that the governor ex ercised line-item veto power over appropriations to higher education.

"Formula-based funding" represents one commonly reported feature of the budgeting process. [2] Initiated in Texas shortly after the second World War, funding formulas were used in 16 states in 1964. By 1992, 33 states reported using some sort of funding formula as part of the budget process. However, this increase in the use of formulas has not been a process of steady conversion. Eight states that used formulas in the 1980s had discontinued their use by 1996. Some other states, like Louisiana and Pennsylvania, appear to have used them intermittently over the past 20 years.

Despite the name, funding formulas rarely, if ever, completely determine total appropriations to higher education in a state. No state uses a single formula, or a formula with a single variable like enrollment. McKeown (1996) identified eight different functional areas in higher education funding formulas used by states: instruction, research, service, academic support, student services, instructional support, scholarships and fellowships, and plant operation. The base units in these formulas vary considerably and credit hours appear more commonly than enrollments. Many of these functional areas use formulas based on factors like the level, type, or mission of the campus, the number of faculty, and the square footage or acreage of the campus.

Most importantly, funding formulas appear to serve as a starting point for the process by which appropriations to higher education are determined. The results of funding formulas are subject to adjustment by many other areas of state government, including governors, legislative committees, and higher education regulatory boards. The primary goal of these formulas seems to be the allocation of state appropriations among the institutions of higher education and not a method of determining the total size of appropriations to higher education from year to year.

The frequent changes in the use, composition, and weights of formulas suggests that state government officials have discretionary power over total appropriations to higher education. Some evidence suggests that cuts in appropriations to public higher education made during recessions are mainly discretionary. Lewis and Farris (1995) report that in fiscal years 1991-1993 between 42% and 55% of all public institutions of higher education experienced cuts in their operating budgets after these budgets had been approved. The overwhelming reason given for these cuts was the effect of recession on government appropriations. The idiosyncratic and everchanging nature of the process by which state governments determine appropriations to higher education suggests that these factors would be difficult, if not impossible, to formally incorporate into an empirical model.

The decisions made by households, higher education administrators, and government decision makers are interrelated and affected to one degree or another by the business cycle. Given the complex, interrelated nature of the behavior of these agents, I adopt a relatively simple reduced form econometric model to analyze the relationship between government appropriations to higher education and the business cycle. This reduced form empirical model is

[A.sub.i,t] = [[gamma].sub.i] + [beta][X.sub.i,t] + [[epsilon].sub.i,t] (1)

where [A.sub.i,t] is a measure of government appropriations to higher education in state i during year t, [[gamma].sub.i] is a vector of state-specific factors that affect government appropriations to higher education and do not change over time, [X.sub.i,t] is a vector of variables that can explain variation in appropriations to higher education, including variables that reflect the state of the business cycle, in state i during year t, and [[epsilon].sub.i,t] is an unobservable equation error associated with state i during year t. [X.sub.i,t] in general can include variables that do not vary across states but vary over time as well as lagged values of variables that reflect the current state of the business cycle. [beta] is a vector of parameters to be estimated.

The changing nature of the funding process used by states and the discretionary power exercised by state legislators and bureaucrats suggests that a reduced form empirical model is appropriate for this analysis. The parameters estimated from a reduced form model reflect the net effect on appropriations of all decisions made by legislators, bureaucrats, and regulatory bodies over the five business cycles in the sample. Some structural detail will be lost by adopting a reduced form empirical model, but a clear picture of the net effect of the many and various decisions made about the funding of higher education in the states will emerge.

[[gamma].sub.i] captures state-specific factors that affect government appropriations to higher education and do not change over time. Several studies have examined the relationship between government appropriations and the economy using cross-sectional data. Clotfelter (1976) examined the effect of out-migration of recent college graduates, the ability of governments to raise tax revenues, and general economic conditions like wages and per-capita income on both expenditure by public institutions of higher education and government aid to higher education. Clotfelter found increased enrollment in higher education and higher per-capita income associated with larger appropriations per capita to higher education, although the variation in income in this study may not be due to business cycle effects as the data are cross sectional. Quigley and Rubinfeld (1993) undertook a study of the effects of political and economic factors on appropriations, expenditure and enrollment in public higher education using data fro m 1985. This study identified a variety of state-specific institutional, political, and social factors that affect the level of public support for higher education and the mix of public and private provision of higher education in states, including geographic location, institutional factors captured by the order of statehood, the availability of private higher education in each state, and the quality of public higher education in each state. These studies suggest that state-specific factors have an important effect on funding and provision of public higher education at the state level and suggest that controlling for these differences across states is important in this empirical analysis.

Leslie and Ramey (1986) used contemporaneous observations of an aggregate measure of business cycle conditions, the coincident index of economic indicators, to explain observed variation in government appropriations to higher education in eight regions of the United States. The coincident index accurately reflects current business cycle conditions for the entire economy, but it may not reflect the current business cycle conditions in each state at a particular time, and it may not accurately reflect the impact of business cycle conditions on tax revenues in each state. This index is constructed from a number of macroeconomic series and reflects changes in personal income, production, employment, and sales in the national economy; business cycle conditions in individual regions of the country are reflected in this index only to the extent that there is homogeneity among regions of the United States in terms of business cycle fluctuations and that these regional effects are contemporaneously correlated.

Hoenack (1983) and Hoenack and Pierro (1990) developed structural models of the legislative demand, household demand, and institutional supply of higher education and tested these models using detailed time-series data from Minnesota. In these papers, state appropriations per voter are affected by the level of state and federal financial aid per high school graduate in Minnesota, the marginal cost of enrollments in higher education, average tax revenues, and enrollment in higher education per voter. The reduced form equations estimated in this paper are closely related to these structural models to the extent that changes in the business cycle affect tax revenues and enrollments. The estimates reported by Hoenack and Pierro (1990) indicate that appropriations and tax revenues rise and fall together, and to the extent that changes in tax revenues and enrollments are due to business cycle effects, the results in this paper support their conclusions about the determinants of state appropriations to higher educa tion. [3]

This study uses total state personal income as a measure of state-specific business cycle conditions. Total state personal income offers several potential advantages over more aggregated measures of business cycle activity. First total state personal income, or any state-specific indicator of business cycle conditions, will reflect state-specific changes in the business cycle better than aggregate indicators of business cycle conditions if turning points in the aggregate economy differ from turning points in states. Nardinelli, Wallace, and Warner (1988) identified 13 states where income fluctuations differ from fluctuations in aggregate income over the period 1956-1982, indicating that the turning point in the business cycle may differ across states.

Second, total state personal income reflects state-specific changes in household income. Sales taxes and to a lesser extent personal income taxes represent an important source of state and local government tax revenues. Over the period 1969-1994, the sample period in this paper, sales taxes and personal income taxes accounted for 32% of total state and local government tax revenues. [4] If spending cuts motivated by budget balancing in response to decreases in tax revenues made by state legislators and bureaucrats are the transmission mechanism through which business cycles affect state spending on higher education, then state-specific changes in personal income will be a good explanatory variable for this analysis. Aggregated measures of business cycle activity like the NBER coincident index reflect changes in many different measures of aggregate economic activity like total employment and production that may not accurately reflect state-specific business cycle conditions.

This discussion raises the point that state tax revenues may be a better explanatory variable than total state personal income, because tax revenues are the transmission mechanism by which business cycles affect appropriations. However, state tax revenues would also be affected by factors like tax relief measures and changes in income or sales tax rates. Adjusting state tax revenues for these effects would require an immense amount of data collection as no comprehensive compilation of these policy changes exists. Also, income is a widely used, but by no means definitive, indicator of business cycle conditions and should be correlated with changes in tax revenue but not correlated with tax relief measures or changes in the tax rate.

The timing of the measure of state-specific business cycle conditions may also be an important determinant of the effect of business cycles on state government spending on higher education. Appropriations to higher education are set by legislatures as part of the budgetary process. These decisions are made at regular intervals during the fiscal year, according to legislative schedules. Thus the timing of turning points in the business cycle relative to annual state budgetary cycles might produce lags in this relationship. In particular, state legislatures may not learn of unexpected changes in tax revenues until well after cyclical turning points have occurred. State legislatures may not respond to short-run economic fluctuations, either because they are not in session when these events occur, or because appropriations bills become effective only after they are passed. Finally, using variables from year t to explain observed variation in government appropriations to higher education in year t may lead to sim ultaneity bias in the empirical estimates if unmeasured or omitted factors are causing variation in both variables. For these reasons, I use lagged income as a measure of state-specific business cycle conditions.

Differences in population and preferences for the provision of public higher education in states may also affect the relationship between income and state appropriations to public higher education. One way to control for this is to normalize the appropriations variable by a measure of enrollment in public institutions of higher education and personal income by a measure of population. This normalization has intuitive appeal because both expenditure per student and income per person are easier to compare than total appropriations and aggregate state personal income. Thus, one part of the empirical analysis defines [A.sub.i,t] as real state appropriations to higher education per FTE enrolled in two- or four-year public institutions of higher education and all income variables measured in per capita terms. [5]

This normalization has some drawbacks. Although the numerator of this fraction reflects changes made by state government officials, the denominator also changes in response to many factors, including business cycles. The exact cyclical behavior of enrollments in public higher education in this sample is difficult to determine, because of the lack of state-specific reference cycles. However, an approximate measure of the cyclicality of enrollments can be obtained. I regressed FTE enrollments on a constant and a time trend for each state

[FTE.sub.i,t] = [[beta].sub.0,i] + [[beta].sub.1,i][t.sub.i] + [[eta].sub.1,i,t]

and real income on a constant and time trend for each state

[INC.sub.i,t] = [[gamma].sub.0,i] + [[gamma].sub.1,i][t.sub.i] + [[eta].sub.2,t]

where [t.sub.i] is a state-specific time trend. The contemporaneous correlation between the residuals from these two regressions was -0.26, indicating that when real income is above trend, FTE enrollments tend to be below trend. This result suggests that enrollments are countercyclical much like the findings of Betts and McFarland (1995).

In order to further investigate the effect of business cycles on state appropriations to higher education, Equation 1 was estimated with [A.sub.i,t] defined as the growth rate of real state appropriations to higher education and all state-specific income variables defined as growth rates. Growth rates are commonly used in the empirical business cycle literature; estimating the relationship between the growth rate of real income and the growth rate of real appropriations will make the results in this paper more comparable to this existing literature. Transforming the levels data to growth rates may also correct for the effects of heteroscedasticity, if present.

3. Empirical Results and Discussion

Table 1 shows the results obtained by estimating Equation 1 for the entire panel under several specifications of [X.sub.i,t]. [6] The dependent variable is real state appropriations per FTE for all specifications reported on Table 1. Specification 1 includes a single variable, real per capita income per person lagged one year, as the measure of state-specific business cycle conditions. Year-specific dummy variables were also included in this specification to capture unmeasured factors that affect government spending on higher education in states over time. The year dummies and state-specific effects were all significant for this specification. [7] The parameter on lagged real per capita income is significant at the 1% level. The implied elasticity of real state appropriations to higher education per FTE with respect to changes in real per capita income in the previous year, based on this point estimate and the means of the variables is 1.39, implying that a 1% change in real per capita income in the precedin g year is associated with a 1.39% change in real state appropriations per FTE to higher education in the following year; state appropriations per FTE to higher education are highly sensitive to state-specific changes in the business cycle in this specification.

Specification 2 includes two lags of real per capita personal income as well as year dummies. Including a second lag of per capita income allows this specification to capture effects of the business cycle on state appropriations to higher education that occur over a longer period of time than those in specification 1. The coefficient on real personal income lagged one year is statistically significant, but the coefficient on real per capita income lagged two years is not. Specification 1 is nested in specification 2 under the condition that the parameter on the second lag of real per capita income is equal to zero. The F-statistic for this restriction has a value of 42.5, which suggests that two lags of real per capita income belong in the empirical model, even though the second lag is not statistically significant. [8]

A two-year lag is consistent with the results in Holtz-Eakin, Newey, and Rosen (1989) regarding the relationship between government revenues and government expenditure. It is also consistent with biennial budget cycles that are used in some states. The overall impact of business cycle effects on real state government appropriations to higher education is very close to the one-year effect estimated from specification 1. The cumulative impact has an elasticity of 1.39 at the means of the variables, which again suggests that variations in real per capita income have a more than proportional effect on real appropriations per FTE.

Specification 3 investigates the possibility that the effect of business cycles on state appropriations to higher education varies across legs of the business cycle. In order to test this hypothesis, a state-specific measure of the turning points in the business cycle was constructed, based on the annual growth rate of real personal income in each state in each year in the sample. Years in which this growth rate was negative were defined as recessions, and years in which this growth rate was zero or positive were defined as expansions. This method is roughly consistent with the method used by the NBER Dating Committee to determine the turning points in the business cycle for the entire U.S. economy, although the NBER uses quarterly data. Using this procedure, 18% of the state-years in the sample were recessionary.

Based on this criterion, two dummy variables were created, reflecting the expansionary and recessionary years for each state. The dummy variable for recessionary years, [[D.sup.r].sub.i,t] was constructed using the criteria

[[D.sup.r].sub.i,t] = {0 if % change in real per capita personal income [greater than or equal to] 0 {1 otherwise.

The dummy variable for expansionary years was constructed using

[[D.sup.e].sub.i,t] = {1 if % change in real per capita personal income [greater than or equal to] 0 {0 otherwise.

Specification 3 interacts these two dummy variables with real per capita personal income lagged one year. The parameter on [[D.sup.e].sub.i,t-1] [X.sub.i,t-1] captures the impact of expansionary years on real state government appropriations per student and the parameter on [[D.sup.r].sub.i,t-1] [X.sub.i,t-1] captures the impact during recessions.

The parameters on these two variables are statistically significant and close to the same size. Specification 1 is nested in specification 3 under the restriction [[beta].sub.e] = [[beta].sub.r]. An F-test on this restriction has a value of 0.324, accepting the null hypothesis that the effect of expansions and recessions on real state appropriations to higher education, is roughly the same.

Table 2 shows the average impact of a recession, here defined as a 2.25% decline--roughly the median decline in the recessionary years in the sample--in real per capita personal income, on state appropriations to higher education based on the results for specification 2 on Table 1 for several states. These states represent the extremes of the distribution of state provision of higher education in the sample for several definitions of provision. The reported impact is calculated at the average appropriations per FTE over the period 1969-1994 in 1984 dollars and the average FTE enrollment in two-year and four-year public institutions of higher education in each state. New Hampshire (average appropriations per FTE $1,982) and New York ($4,818) represent the two extremes in terms of average appropriations per FTE. [9] Nevada (2.6%) and Utah (5.5%) have the lowest and highest average FTE enrollment in public two- and four-year institutions of higher education per capita in the sample. Vermont (14,073) and Califor nia (924,143) have the lowest and highest average FTE enrollment, respectively, in public two- and four-year institutions of higher education in the sample. These estimated impacts suggest that the effects of a recession on state appropriations to higher education are quite substantial in magnitude.

As further evidence of the impact of these effects, consider the case of Maryland's appropriations to public higher education during the early 1990s. The impact of the recession during this period on state funding for higher education in Maryland has been documented by Eaton, Miyares, and Robertson (1995). Maryland went through a two-year-long recession in the early 1990s. The results in Table 2 suggest that the average impact of this recession on state appropriations to higher education would have been about $26.5 million 1984 dollars. The actual cuts, which took place over two years, totaled $91 million 1984 dollars, or $126 million current dollars. The actual decline in real appropriations was over three times the estimated decline on Table 1, suggesting that either the economic downturn in the early 1990s in Maryland was larger than average, or the state legislators and bureaucrats who determined the size of the spending cuts were unusually harsh. The decline in real per capita income in Maryland during this recession was roughly equal to the median decline across all states during recessions over the sample period. The fraction of state government expenditure on Medicaid rose from 10% to just below 14%, and the fraction spent on public safety increased from about 11.5% to 12.5% during the same period.

Eaton, Miyares, and Robertson argue that the cuts in appropriations to higher education in the early 1990s in Maryland were extraordinarily large, relative to both prior experiences in Maryland and to experiences in other states. The results from Tables 1 and 2 suggest that the decline in appropriations to higher education in Maryland were considerably larger than the average decline across all states and recessions in the sample, which supports their claim. State policy makers in Maryland appear to have performed a considerable amount of budget-balancing at the expense of funding to higher education during this recession. The overall impact of these funding cuts on public higher education was the discontinuation or reduction in size of 60 bachelors programs and 35 masters and doctoral programs, 22 departments consolidated into 10, and the elimination of two schools.

The results presented in Tables 1 and 2 suggest that state business cycle conditions, as reflected in variation in real per capita income, have a significant effect on real state government appropriations per FTE enrollment. The estimates from specification 3 also suggest that the elasticity during recessionary years is equal to the elasticity during expansionary years. Although appropriations per student fall more than proportionately during recessions, they also rise more than proportionately during expansions. This symmetric effect could be interpreted as evidence that state governments cut back on appropriations to higher education during recessions and then give back these funds during expansions.

Another explanation is that changes in enrollments, which are countercyclical, are contributing to the symmetry of the estimated effects of recessions and expansions on appropriations per FTE. If enrollments are countercyclical, then appropriations per FTE will rise (fall) during expansions (recessions) because the denominator falls (rises), even if appropriations remain unchanged.

One way to investigate the relationship between appropriations to higher education and the business cycle and control for differences among states in size and preferences for the provision of public higher education, while removing the effects of variation in enrollments, is to use growth rates of the variables. Table 3 shows estimates of Equation 1 when the dependent variable is the growth rate of real state appropriations and the business cycle variables are defined as the growth rate of real income.

The results are similar to those reported on Table 1 in terms of the sign and significance of the variables. The first lag of growth in real income is significant in specification 1 and although the second lag is not significant in specification 2, an F-test on the implied restriction rejects specification 1 in favor of specification 2. The elasticity of the growth rate of real appropriations with respect to changes in the growth rate of real income implied by specification 2 at the means of the variables is 0.4. This equates to an elasticity of 2.9 when the variables are expressed in levels. The implied impact of business cycles on state appropriations for this specification is considerably larger that the impact implied by the results on Table 1.

However, the estimates of specification 3 are strikingly different in that the effect of recessions on the growth rate of real state appropriations to higher education are considerably larger than the effect of expansions. The elasticities, 1.14 for recessions and 0.15 for expansions, are also different. These estimates imply a much different effect of recessions on appropriations to higher education than expansions. Based on these estimates, the cuts in appropriations to higher education during recessions are more than proportional to the decline in real income, while the increases during expansions are less than proportional to the increases in real income. Note that this does not necessarily imply that total funding to higher education has fallen as a result of recessions because expansions last considerably longer than recessions and the increase in income during expansions are generally larger than the decline during recessions. Still, these estimates imply that funding to higher education may take quite some time to return to prerecession levels following a period of recession-induced budget cuts.

One explanation for the difference between the estimates of specification 3 reported in Table 1 and those reported in Table 3 is the effect of enrollments on the dependent variable in Table 1. If the effect of expansions on enrollments is relatively stronger than the effect of recessions on enrollments, then changes in appropriations per FTE will be larger in absolute value during expansions than during recessions. This would tend to increase the estimated impact of expansions on appropriations per FTE and decrease the estimated impact of recessions.

If appropriations to higher education fall more during recessions than they rise during expansions, then this ratio should exhibit cyclical variation. An examination of plots of this ratio by state showed a considerable amount of variation over the business cycle, confirming the empirical results shown in Table 3. This ratio fell sharply during both the 1974 recession and the two closely spaced recessions in the early 1980s and rebounded during the long expansion in the 1980s in most states. A similar pattern could be seen for the 1990 recession, although differences in the timing of that recession across states make it harder to clearly discern.

In summary, the results reported in Tables 1 and 3 are consistent. Both suggest that business cycles affect both state appropriations per student and the growth rate of state appropriations to higher education. Both also suggest that the impact of business cycles on appropriations to higher education during recessions is proportionately larger than the change in the economic climate, as measured by declines in personal income. However, the countercyclical behavior of enrollments may mask asymmetries in the effect of recessions and expansions on state appropriations to higher education when appropriations are weighted by enrollments.

4. Conclusions

The results presented in this paper suggest that state-specific changes in the business cycle have significant effects on state government appropriations to higher education. A 1% decline in real per capita income was associated, on average with, a 1.39% decline in state appropriations to higher education per student in the following year. The effects are statistically significant as long as two years after a change in real per capita personal income. Because enrollments rise during recessions, a decline in funding may place significant fiscal stress on public institutions of higher education.

Consider the implications of these results, along with the previously discussed enrollment effects reported by Betts and McFarland (1995), in community colleges during an economic downturn, If the local unemployment rate rose 2% during a recession, and real per capita income fell by 2.25%, a typical community college would experience an 8% increase in enrollment and a 3.4% decrease in state appropriations per student at roughly the same time, assuming that the estimates reported on Table 1 reflect on average the same effect on state appropriations to four-year colleges and community colleges. [10] A community college facing such fiscal stress might find it difficult to provide quality higher educational services to its student body. Worse, the additional students enrolled during this period are primarily people who were laid off or fired during the economic downturn and are attempting to acquire new skills in order to regain employment. Outcomes like this are clearly not consistent with the main goals of pub lic higher education.

These results have clear policy implications. In order to avoid placing severe fiscal burdens on institutions of higher education during economic downturns, state policy makers should consider extending some budgetary protection to higher education during these periods. The evidence in this paper suggests that under current policies, funding to institutions of higher education may be an attractive target for policy makers seeking the path of least resistance when cutting spending to balance state budgets. Unfortunately, the timing of these budget cuts places a considerable amount of fiscal stress on public institutions of higher education because they coincide with increases in enrollment.

Countercyclical economic policies are typically associated with monetary and fiscal policies undertaken by the federal government. However, public higher education plays an important role in educating workers who lose their jobs during economic downturns, and thus plays an important, although indirect role in stabilization policies. Because of the importance of education to long run economic growth and stability, state government policy makers should, in this case, be encouraged to pursue their own stabilization policies by not cutting appropriations to higher education during economic downturns. By maintaining funding to higher education during recessions, public institutions of higher education will be better able to provide quality education to students during recessions. This may reduce the duration of these spells of unemployment and should make these students more productive when they return to the labor force.

(*.) Department of Economics, 1000 Hilltop Circle, University of Maryland Baltimore County, Baltimore, MD 21250, USA; E-mail humphrey@umbc.edu.

Thanks to Bob Black, Kathleen Carroll, Dennis Coates, Bill Lord, Clark Nardinelli, and two anonymous referees for helpful comments. Ryan Mutter provided excellent research assistance; Karl Kendell at the Census Bureau Governments Division, David Smith at the Bureau of Labor Statistics, and the Center for Higher Education and Educational Finance for kindly provided data. All remaining errors are my own.

Received December 1998; accepted January 2000.

(1.) McGranahan (1999) examined the effect of business cycles on state government budgets and found that many revenue and expenditure categories are sensitive to changes in the business cycle. This study found a similar relationship between appropriations and the business cycle using a different measure of business cycle effects, the unemployment rate.

(2.) See McKeown (1996) for a detailed discussion of the development and use of formulas in the funding of education.

(3.) Although this study omits the effects of state and federal financial aid on state appropriations to higher education, the marginal cost variable in these models depends in part on enrollments, and thus may also be affected by business cycles.

(4.) Based on data from table B-87 in the Economic Report of the President (1998).

(5.) It would be more appropriate to weight income by the population age 16 and older. However, annual estimates of state population age 16 and older are not readily available for the early part of the sample.

(6.) Regression diagnostics indicated that the equation errors for these regressions were both heteroskedastic and autocorrelated. The procedure suggested by Newey and West (1987) was used to correct the standard errors; the reported standard errors are robust to heteroskedasticity and autocorrelation of order AR(1).

(7.) These results are available on request from the author.

(8.) This F-statstic is based on the sum of squared errors from an ordinary least squares (OLS) regression, and not on those from the Newey-West correction. It is not comparable to the t-test implied by the standard errors reported in Table 1. Because an additional lag reduces the sample size by 50, the F- and t-tests based on the OLS standard errors are not strictly comparable. This also holds for the F-test for the growth rate regressions reported below.

(9.) In the lower 48 states, Alaska and Hawaii both have higher appropriations per FTE, but I have excluded them from this example but not the sample.

(10.) Unfortunately, state appropriations data were not available disaggregated by level of institution.

References

Betts, Julian R., and Laurel L. McFarland. 1995. Safe port in a storm: The impact of labor market conditions on community college enrollments. Journal of Human Resources 30:741-65.

Clotfelter, Charles T. 1976. Public spending for higher education: An empirical test of two hypotheses. Public Finance 32:177-94.

Eaton, Gertrude, Javier Miyares, and Ruth Robertson. 1995. Statewide planning during declining state support. Planning far Higher Education 23:27-34.

Getz, Malcolm, John J. Siegfried, and Hao Zhang. 1991. Estimating economies of scale in higher education. Economics Letters 37:203-8.

Hoenack, Stephen A. 1983. Economic Behavior Within Organizations. New York, NY: Cambridge University Press.

Hoenack, Stephen A., and Daniel J. Pierro. 1990. An econometric model of a public university's income and enrollments. Journal of Economic Behavior and Organization 14:403-23.

Holtz-Eakin, Douglas, Whitney Newey, and Harvey Rosen. 1989. The revenues-expenditures nexus: Evidence from local government data. International Economic Review 30:415-29.

Koshal, Rajindar K., and Manjulika Koshal. 1999. Economies of scale and scope in higher education: A case of comprehensive universities. Economics of Education Review 18:269-77.

Leslie, Larry L., and Garey Ramey. 1986. State appropriations and enrollments, does enrollment growth still pay? Journal of Higher Education 57:1-19.

Lewis, Laurie, and Elizabeth Farris. 1995. Statistics in brief: Higher education finances and services. Washington, DC: United States Department of Education, National Center for Education Statistics.

McGranahan, Leslie. 1999. State budgets and the business cycle: Implications for the federal balanced budget amendment debate. Federal Reserve Bank of Chicago Economic Perspectives 23:2-17.

McKeown, Mary P. 1996. State funding formulas: Promise fulfilled? In Struggle to survive: Funding higher education in the next century, edited by David S. Honeyman, James L. Wattenbarger, and Kathleen C. Westbrook. Thousand Oaks, CA: Corwin Press, pp. 49-85.

Nardinelli, Clark, Miles Wallace, and John Warner. 1988. State busines cycles and their relationship to the national cycle: Structural and institutional determinants. Southern Economic Journal 54:950-60.

NASBO Information Brief. 1996. State innovations in higher education finance and government. Washington, DC: National Association of State Budget Officers.

Newey, Whitney K., and Kenneth D. West. 1987. A simple, positive semi-definite, heteroskedasticity and autocorrelation consistent covariance matrix. Econometrica 55:703-8.

Quigley, John, and Daniel Rubenfeld. 1993. Public choices in public higher education. In Studies of supply and demand in higher education, edited by Charles Clotfelter and Michael Rothchild. Chicago, IL: University of Chicago Press, pp. 243-83.

United States Department of Education. 1990. State higher education profiles. Washington, DC: U.S. Government Printing Office.

United States Department of Education. 1996. Digest of educational statistics. Washington, DC: United States Department of Education.

United States Government Printing Office. 1998. Economic report of the president. Washington, DC: United States Government Printing Office.

Data Appendix

State-specific data on total personal income and population for the period 1969-1994 were taken from the U.S. Department of Commerce Regional Economic Information System (REIS) CD-ROM, which is available from the Bureau of Economic Analysis, Regional Economic Measurement Division. Data on state government appropriations to higher education were provided by the Center for the Study of Education Finance at Illinois State University and can also be found in past editions of the higher education newsletter Grapevine, which is published by the Center for the Study of Educational Finance, 340 DeGarmo Hall, Illinois State University, Normal, IL 61761-6901. Some of the year-to-year variation in enrollment is due to the closing of existing institutions of higher education and the opening of new institutions of higher education. Although such events are relatively rare, they do occur in the sample. The year-specific dummy variables described above also serve as a control for these year-to-year changes in the total num ber of institutions of higher education in the states.

The appropriation data were deflated using the higher education price index (HEPI): the personal income data were deflated using the consumer price index (CPI). The CPI and HEPI data were taken from the Digest of Educational Statistics (1996), table 38. These data are available on line at http://nces.ed.gov/pubs/digest97/d97t038.html, the National Center for Education Statistics website. Deflating the appropriations data by the CPI had no substantial effect on the empirical results.

The enrollment data were compiled from two sources. Total and FTE enrollment by state from 1969 through 1990 were published in State Higher Education Profiles (SHEP) (1990). For the period 1991-1994, similar measures of total and FTE enrollment were calculated from the publicly available Integrated Postsecondary Education Data System (IPEDS) Fall Enrollment surveys. The 1991-1994 enrollment data were constructed using the same set of public institutions that comprised the 1990 SHEP sample; the 1991-1994 data are comparable to the pre-1991 data.

The mean values of the key variables for each state for the period 1969-1994, in 1984 dollars, are shown on Appendix Table 1. There is a good deal of variation in state appropriations per FTE in public institutions in the sample, as well as the preferences for the provision of public higher education, as measured by FTE enrollment per capita. Note when comparing appropriations per FTE and income per person that the denominators are quite different, as shown by column three.

Appendix Table 2 shows the sample means by state for the data expressed in growth rates, which yield the estimates shown on Table 3. One interesting feature of this table is the negative growth rates of enrollment per capita in the group of predominantly western states (Arizona, California, Hawaii, Oregon, and Washington, along with Minnesota). It appears that the population in these states was growing faster than the capacity of public institutions of higher education over the sample.