Abstract

This study examines how private colleges and universities choose to spend versus reinvest resources in endowment funds that have suffered investment losses. The analysis takes advantage of a market downturn and public policy shift, which together revealed how colleges define prudent spending. Investment losses during the financial crisis of 2008 left many endowment gift funds below their original donated values, or “underwater.” Colleges in some states were legally required to cut spending from underwater funds. Other states had recently enacted the Uniform Prudent Management of Institutional Funds Act, which allows prudent spending from underwater funds. The act loosened financial constraints, and affected colleges responded by spending 22 percent more from their endowments in the fiscal year after the financial crisis. Constrained colleges did not increase spending from unrestricted parts of their endowments to offset reduced spending from underwater funds.

Only a year ago, universities, with their $400 billion in endowment money, faced a congressional inquiry because of widespread concern they needed to spend more of their savings on financial aid for students. Now, colleges are finding that state laws are not letting them spend enough, [after] taking a beating from the recession and the collapse in stock prices… The growing movement [toward a new state law] is sparking a debate about the value of freeing up emergency cash versus the danger of further depleting key financial reserves, potentially shortchanging future generations.

The Wall Street Journal, 11 February 2009

1.  Introduction

This study informs the debate that is central in the extracted quote, by measuring changes in private college endowment spending caused by a state law called the Uniform Prudent Management of Institutional Funds Act, or UPMIFA (pronounced up-MIFF-uh).

In the early 2000s, endowments grew quickly in size and importance. In the five-year period leading up to 2008, the average private college endowment gained 46 percent in real value from a combination of positive investment returns and new donations, net of spending. By spending roughly 5 percent of a moving average of its total endowment value in each year, the average private college covered 10 percent of its yearly expenses, though some covered well over half. Endowment spending typically supports scholarships and faculty salaries.

During the financial crisis of 2008, the average private college endowment lost 22 percent of its value, jeopardizing financial support for students and faculty (NACUBO-Commonfund 2009). Many recently donated endowment gift funds were “underwater,” worth less than their original value at donation. Expecting declines in revenue from tuition and other sources (Nelson and Goodman 2009), each college had to choose how much to spend from smaller, and in many cases underwater, gift funds. In many states, the law prohibited spending from underwater funds, imposing large cuts to endowment revenue, which would be difficult to replace from other sources.

UPMIFA allows for spending from underwater funds where prior law did not. The law imposes a prudent person standard, which simply requires colleges to consider several economic factors before spending. Introduced in 2006, UPMIFA was enacted in all states by 2012.1 Small differences in timing of enactment, combined with the prevalence of underwater funds after the financial crisis, led to large differences in spending constraints on otherwise similar colleges.

To assess the effect of spending constraints, I compare endowment spending before and after UPMIFA. I focus on private, nonprofit, endowed, four-year colleges and universities in the United States, using panel data over fiscal years 2005 to 2013. During this time, state laws vary within each college across years, and across colleges within each year. Therefore, it is possible to identify the law's effect on spending while controlling for college and year fixed effects in a difference-in-differences approach, estimating increases in spending after enactment of UPMIFA. Colleges vary in their level of underwater funds at the time their state enacted UPMIFA. I exploit this variation to introduce a third difference, which relies on the sudden onset of the financial crisis for identification.

I find that UPMIFA allows colleges to spend 6.12 cents more per dollar in underwater funds. This implies that the average college under UPMIFA spent a typical amount from underwater funds, whereas colleges under prior law spent little to nothing from underwater funds. These findings imply that from the effect of legal constraints alone, the average private college under prior law experienced a 22 percent decrease in endowment revenue in fiscal 2010, on top of a roughly 6 percent decrease in endowment revenue from absorbing the investment losses of fiscal 2009. Colleges do not appear to have increased spending from unrestricted parts of their endowment to offset these losses. The results appear to be driven by larger endowments.

This study contributes to a growing empirical literature on endowment spending decisions. Most colleges use similar, simple formulas to plan spending (Brown and Tiu 2015), but colleges systematically deviate from their plans, usually to spend less (Brown et al. 2014). Legal constraints on endowment spending have received little attention as an explanation for these deviations, for two reasons: Nearly all states used the same uniform law for decades, and the law restricted spending only in extreme circumstances (Budak and Gary 2010). UPMIFA represents a new and less restrictive endowment law that was introduced just before the extreme investment losses of the late 2000s recession.

A few other studies address UPMIFA, but do not test for its impact by comparing spending across groups of colleges. The calibrated model in Gilbert and Hrdlicka (2015) predicts that external constraints on endowment spending will lead to larger and more risky endowments. However, the study does not test this prediction using the loosening in constraints provided by the adoption of UPMIFA. Bass (2010) surveys a set of colleges in early-enacting states about how they will respond to UPMIFA. The majority of colleges plan to take advantage of the flexibility to spend from underwater funds under UPMIFA but many also report spending from underwater funds under prior law, suggesting a muted effect of the change to UPMIFA. The present study provides information beyond these targeted surveys of college plans after UPMIFA, with a larger panel including observations of actual spending by colleges under different laws.

The comparisons in this study provide an empirical definition of “prudent management,” using colleges’ behavior to reveal how they interpret the legal definition provided in UPMIFA. The analysis focuses on spending from underwater funds, which under UPMIFA should be representative of a college's general approach to weighing current spending versus investment for the future.2

The remainder of the study is divided into six sections. Section 2 provides additional background, which informs the empirical strategy laid out in section 3. Section 4 describes the data used to implement the empirical strategy. Section 5 reports the main results, section 6 provides supplementary results, and section 7 includes some robustness checks. Section 8 concludes with a discussion of the results.

2.  Endowment Spending and State Laws

Endowment donors stipulate that the college invest their gifts and spend only the investment income for a particular purpose, like student financial aid or a faculty chair. Within these constraints, the college chooses how to invest gift funds together with other savings in its endowment, and how much to spend each year on particular students or faculty.3 On behalf of donors and taxpayers, state laws (as enforced by the state attorney general) guide the choices made by nonprofit, tax-exempt college endowments. This section discusses how endowment spending works in practice, and how state laws affect this process.

The vast majority of colleges plan endowment spending using simple formulas that combine future expectations with recent history (Brown and Tiu 2015). Under a moving-average rule—the most common formula—a college sets a constant “policy spending rate” near 5 percent. This rate corresponds to its long-term target for real rate of return on investments (NACUBO-Commonfund 2014). The college multiplies that rate by a moving average of the total endowment value. Most often the formula averages endowment values over the past three to five years at a yearly or quarterly frequency (Brown and Tiu 2015).

Total spending during a fiscal year divided by total endowment value at the beginning of the fiscal year is called the “effective spending rate” or simply the spending rate. Whereas the policy spending rate is constant, the effective spending rate moves counter to changes in endowment value coming from new donations, investment gains, or investment losses. These changes in value are incorporated slowly into spending levels.

Colleges often change the parameters of their spending formulas, and actual spending varies around planned spending (Brown and Tiu 2015). This is not surprising because, in theory, a simple spending rule is unlikely to solve the complicated intertemporal allocation problem each college faces (Hansmann 1990; Hoxby 2015). In practice, college spending appears to respond to the source of capital gains and to respond asymmetrically to gains and losses, which are not components of the typical spending rule (Woglom 2003; Brown et al. 2014). Still, the prevalence of the moving-average formula illustrates the importance of sustaining the value of the endowment while providing a steady and growing stream of payouts.

A college endowment is commonly discussed as a whole but is actually made up of many separate donations and deposits. Endowed gifts each create a permanent fund. In addition to gift funds, sometimes called “true endowment” funds, approximately one third of endowment value is in quasi-endowment funds. These are assets kept in the endowment by a resolution of the college's governing board. The college has only a reversible internal commitment, and no external legal obligation, to permanently maintain the value of quasi-endowment funds.

Most of the time, the legal distinction between the endowment as a unit and the endowment as a collection of funds is not economically important. Investment and spending decisions are made at the level of the endowment, so that each fund experiences the same net percent growth. Purpose restrictions of individual gifts are not economically binding except in the rare case where investment income from a gift fund provides the last dollar a college spends on a particular purpose (Ehrenberg 2009).4

Sometimes, legal constraints on spending from individual gift funds bind and become economically important. Agreements with donors are made at the fund level but rarely include explicit instructions for spending. State laws interpret donor intent, defining income intended to be spent versus principal intended to be saved. The definition has evolved over time and been codified in two versions of a uniform state law. Both versions were promulgated by the National Conference of Commissioners on Uniform State Laws (NCCUSL, now called the Uniform Law Commission or ULC), a nonpartisan, national association of lawyers seeking to unify laws across state boundaries whenever appropriate.

Before 1972, the working definition of endowment income was limited to trust income, including interest, dividends, rents, and royalties from investments bought with the original gift principal (Budak and Gary 2010). In 1972, the NCCUSL published the Uniform Management of Institutional Funds Act or UMIFA (NCCUSL 1972). This predecessor to UPMIFA expanded the definition of endowment income by expressly allowing endowments to pursue and spend capital gains. To protect some concept of permanent principal, the law limited spending to capital gains above “historic dollar value.” The law defined historic dollar value as the original gift amount plus any subsequent donations to the fund, not adjusted for inflation. This created the concept of an underwater fund, worth less than its historic dollar value and therefore yielding no capital gains to spend. The law was enacted nearly universally.

Newer funds with little net nominal growth will fall below their historic dollar values after investment losses. Designating them as underwater, the old uniform law limits spending only from these newer funds even though they tend to have lost the same percent of their real value as older funds. Mainly to address this inconsistency, in 2006 the NCCUSL published UPMIFA (NCCUSL 2006). UPMIFA ignores underwater funds per se and ensures the preservation of all endowment funds using a prudent person standard rather than a sharp, nominal, backward-looking spending limit.5

Because there is no clear consensus on endowment management objectives, it is not clear that prudent management of underwater funds implies spending from those funds.6

Just after the financial crisis of 2008 led to a large increase in underwater funds, twenty-five states had not yet enacted UPMIFA. Colleges in these states therefore saw a gap between what their normal spending policies would indicate for the following year's spending and what prior law allowed. The gap varied across colleges by the amount in underwater funds, which ranged from very little of the endowment to over half of the endowment.

Colleges under prior law had a few options for filling the gap between planned spending and legally allowed spending, including contacting donors to modify agreements, seeking legal release from agreements with unavailable donors, or limiting spending to trust income (Sare 2009). Each of these options requires a fundamental change in either the spending or investment plan for each underwater fund. A college could also legally spend more from unrestricted quasi-endowment funds, which can never be underwater. Besides these options within the endowment, the college could seek other revenues or make budget cuts. Colleges in UPMIFA states could, in addition to all these more difficult options, pursue their definition of prudent spending from underwater funds.

The following section lays out a plan for measuring the impact of UPMIFA by comparing spending across colleges.

3.  Empirical Model

Main Specification

The key parameter to estimate is the difference in spending rates caused by UPMIFA. This could ideally be estimated by observing a college's spending rates from underwater funds under UPMIFA versus prior law, all else held equal. The following empirical model approximates this ideal comparison using the available data and the natural variation in timing of enactment of UPMIFA.

Survey data, which I detail in section 4, do not include spending rates from individual gift funds or from types of funds. They do include aggregate endowment values by type of fund, such as the total value in the endowment, the value in underwater funds, and the value in quasi-endowment funds. The data include aggregate flows at the level of the endowment, such as spending and investment returns. Colleges are observed once per year. I use this information to infer spending rates by estimating the coefficients of the following panel regression.
Spendit=j=0,1,2αjBeg.valueit-j+θUPMIFAit+ζUnderwaterit+ηUPMIFAitUnderwaterit+Xit'β+γi+δt+ɛit.
(1)

This regression estimates average spending rate slopes α0, α1, α2, ζ, and η across many colleges i and fiscal years t. γi, δt, and θ allow for different spending intercepts by college, year, and law. ɛit is an idiosyncratic error term. I report results with and without Xit, which includes investment returns broken into positive and negative components, as well as lagged expenses (these measures make up the endowment shocks in Brown et al. 2014).

This regression mirrors the moving-average formula, where spending is a linear function of current and lagged endowment values. Therefore, I expect the sum of the coefficients on lagged endowment values α0+α1+α2 to approximate the policy spending rate, around 5 cents per dollar. For colleges that do not use a moving-average formula, current and lagged endowment values and current value in underwater funds control in a simple and flexible way for a college's recent financial history.

Dollars in underwater funds reduce the base available for spending, but only prior to UPMIFA. Therefore I expect the coefficient ζ to equal the negative of the effective spending rate, around negative 5 cents per dollar. The coefficient η on the interaction term represents how much of this reduction is offset under UPMIFA, and is the focus of this analysis.

η is a triple-difference parameter, θ captures the difference-in-differences in spending for colleges in states enacting UPMIFA when underwater funds are zero, and η captures how that difference-in-differences varies by the amount in underwater funds at the time of enactment. If the effect of UPMIFA operates only through underwater funds, then θ0.

Identification

To interpret η as the causal impact of UPMIFA on spending from underwater funds, there must be parallel differences-in-differences in spending, absent the policy change, at all levels of underwater funds. This assumption can fail if the effect of UPMIFA is heterogeneous and associated with the timing of enactment. In section 4, I provide some evidence for similarity across colleges in states that enact UPMIFA at varying times, and for parallel trends in spending, but the parallel differences in differences assumption is not directly testable.

The assumption breaks down under policy endogeneity or any sort of selection into enactment timing that is associated with potential differences-in-differences. Suppose that, after the drafting of UPMIFA, some colleges advocate more strongly for enactment because they are more inclined to take advantage of UPMIFA's added flexibility and to spend more from underwater funds. If they succeed and their states enact UPMIFA earlier, then there are no longer parallel differences-in-differences across early- and late-enacting states.

This story seems unlikely for a few reasons. First, advocacy by colleges was not the main determinant of variation in enactment timing. Members of the drafting committee of UPMIFA emphasized that the law was typically disseminated to legislators by lawyers and national organizations, not by colleges.7 State legislatures meet for different periods of the year, have different non-endowment-related priorities, and follow idiosyncratic conventions in setting the effective dates of new laws.8

Second, earlier enactments need not reflect variation in colleges’ preferences for spending under UPMIFA. Earlier enactments could instead reflect the preferences of state legislatures. The complexity of the issue was on display in Massachusetts, a state where lawmakers had previously criticized endowments for spending too little, but where UPMIFA, which allows for more spending, was enacted relatively late (Hechinger 2008).

Third, successful advocacy by colleges, if it occurred, could be a function of colleges’ size and resources, rather than any differential preference for additional spending under UPMIFA. Then states with more influential colleges would enact UPMIFA earlier. However, there would be no bias introduced because the effect of the policy at these colleges would be representative of the effect at smaller colleges in later-enacting states.

Most states saw fit to enact UPMIFA before the financial crisis, when underwater funds were nearly zero, and all states enacted it at some point. With such a short time span for all state legislatures to enact UPMIFA, small differences in timing of enactment could arise even without any differences in preferences across colleges or state legislatures.9

Alternative Specifications

There may be differences across colleges in the impact of UPMIFA that are not associated with the timing of enactment. In a supplemental analysis, I allow for heterogeneity in effects by observable characteristics of colleges that may lessen the impact of UPMIFA. I add terms to equation 1 to measure heterogeneity in the effect along these dimensions.
Spendit=j=0,1,2ωjBeg.valueit-j+ξUPMIFAit+χUnderwaterit+νUPMIFAitUnderwaterit+πPiUPMIFAit+φPiUnderwaterit+ψPiUPMIFAitUnderwaterit+Xit'μ+ιi+τt+εit.
(2)

The added terms are interactions between the key underwater and legal terms and a time-invariant measure Pi. Pi can represent the percent of college i’s operating expenses covered by endowment spending, as colleges that rely more heavily on endowment spending may make smaller cuts in spending under prior law. Pi can also be an indicator for the use of special appropriations from the endowment. Colleges sometimes draw special appropriations outside the normally reported spending rate. Because many colleges never do this, Brown and Tiu (2015) interpret colleges that do draw special appropriations as having more flexible control over spending. Pi can also represent the percent of endowment dollars in quasi-endowment funds, which are not subject to the constraints imposed by state laws.

In this extended regression ξ, χ, and ν represent impacts for colleges with Pi=0. π, φ, and ψ represent the differential impact for a unit increase in Pi. In all cases, I expect higher Pi to shrink the effect of underwater funds and therefore shrink the gap in spending across colleges under different laws. Then φ>0 and ψ<0.

The empirical models in equations 1 and 2 make particular choices that can be relaxed in robustness checks. In section 7, I allow for varying effects of UPMIFA in the post-enactment period, and use alternative measures of spending. Estimating all of these regressions requires data on college endowments during a period when variation in underwater funds and legal changes allow for identification.

4.  Data and Descriptive Statistics

This study focuses on private, nonprofit, four-year colleges and universities. These are the colleges where endowments are the most important to the budget and where endowments commonly support a single college campus. In contrast, public colleges and universities receive state funding determined by the same central authority for many colleges in a state, often for multiple years at a time. If budgets set at the state and year level affect public college and university endowment spending, then that variation in budgets could be confounded with the variation in state laws that I seek to isolate.

Data come from the National Association of College and University Business Officers (NACUBO), the Commonfund Institute, and the Integrated Postsecondary Education Data System (IPEDS). The NACUBO-Commonfund Study of Endowments (NCSE) provides data on spending, investment returns, and asset valuations for each college's endowment each fiscal year.10 IPEDS adds general college characteristics and finances. The NCSE began measuring underwater funds at the beginning of fiscal year (FY) 2005, so the sample for this study begins then and extends through FY 2013, covering the final state's enactment of UPMIFA. A total of 522 colleges have sufficient data to appear in the panel sample. The sample is skewed toward colleges with larger endowments and the resources to provide data. Appendix B includes further details on the data sources and sample, and Appendix C includes detailed data definitions.

The average endowment in the panel sample is summarized in table 1, in the first column, labeled All. The average college held $368.6 million in its endowment leading up to the financial crisis.11 An average spending rate of 4.8 cents per dollar covered an average of 9.8 percent of the college's operating expenses.12 Following the 2008 financial crisis, average investment returns over FY 2009 were –19.0 percent (including a rebound in stock prices late in the fiscal year). This led to 17.4 percent of endowment value being in underwater funds by the end of FY 2009. The amount in quasi-endowment funds was first reported in 2009, so I include the post-crisis average here as a measure of quasi-endowment for each college. The average endowment was made up of 31.4 percent quasi-endowment. Finally, covering all years in the sample, 39.3 percent of colleges ever drew special appropriations.

Table 1.
Mean Values for Uniform Prudent Management of Institutional Funds Act (UPMIFA) Enactment Cohorts
Enactment Cohort All 2008 2009 2010 2011 2012 
Number of colleges 522 65 106 142 146 63 
Pre—financial crisis, 2005—08 average       
Endowment value ($M) 368.6 226.8 242.3 367.7 478.7 393.9 
Effective spending rate (%)*** 4.8 4.3 5.4 4.7 4.5 5.1 
Operating expenses covered (% of operating expense) 9.8 7.9 11.8 9.3 8.9 11.3 
Post—financial crisis       
Investment return fiscal 2009 (%) −19.0 −18.9 −18.2 −18.9 −19.5 −19.2 
Underwater funds, end 2009 (%) 17.4 15.9 18.1 18.2 15.4 20.4 
Quasi-endowment 2009—13 average (%)*** 31.4 36.2 25.5 25.3 37.0 35.3 
Over entire sample       
Ever used special appropriations (% of colleges)* 39.3 27.7 45.3 41.8 34.2 47.6 
Enactment Cohort All 2008 2009 2010 2011 2012 
Number of colleges 522 65 106 142 146 63 
Pre—financial crisis, 2005—08 average       
Endowment value ($M) 368.6 226.8 242.3 367.7 478.7 393.9 
Effective spending rate (%)*** 4.8 4.3 5.4 4.7 4.5 5.1 
Operating expenses covered (% of operating expense) 9.8 7.9 11.8 9.3 8.9 11.3 
Post—financial crisis       
Investment return fiscal 2009 (%) −19.0 −18.9 −18.2 −18.9 −19.5 −19.2 
Underwater funds, end 2009 (%) 17.4 15.9 18.1 18.2 15.4 20.4 
Quasi-endowment 2009—13 average (%)*** 31.4 36.2 25.5 25.3 37.0 35.3 
Over entire sample       
Ever used special appropriations (% of colleges)* 39.3 27.7 45.3 41.8 34.2 47.6 

Source: NACUBO-Commonfund Study of Endowments (NCSE).

Notes: Enactment cohorts are defined by the first fiscal year UPMIFA was effective in their state in time to affect spending plans. The 2008 cohort also includes earlier enactments, and the 2012 cohort also includes later enactments. Unless noted otherwise, % values are percent of beginning-of-year endowment value in dollars.

*p < 0.10; ***p < 0.01 for F-test of equality of means across cohorts.

Figure 1 shows the progression of investment, spending, and underwater funds over time, averaging over all colleges in the sample each fiscal year. As investment returns fluctuate, spending stays relatively smooth. Underwater funds increase with a lag after investment losses, and decrease with a lag after investment gains. Although these aggregate trends are consistent with stated spending formulas, actual spending often varies around planned spending. Central to this study, legal constraints can be one reason for this variation.

Figure 1.

Trends in Mean Endowment Rates, 2005—13

Source: NCSE.

Figure 1.

Trends in Mean Endowment Rates, 2005—13

Source: NCSE.

Every state during 2005 to 2013 was subject to one of two rules on spending from underwater funds. Table 2 lists the effective date of the move to UPMIFA for all states. It also lists the lag between signing (when UPMIFA is passed by the legislature and signed or otherwise approved by the governor) and the effective date of the law.13

Table 2.
Effective Dates of UPMIFA, and Lag in Days Between Signing into Law and Effective Date
State Year Month Day Lag State Year Month Day Lag 
Alabama 2009 Jan 245 Montana 2007 Oct 151 
Alaskaa 2010 Sep 61 Nebraska 2007 Sep 150 
Arizona 2008 Sep 26 165 Nevadaa 2007 Oct 140 
Arkansas 2009 Jul 30 154 New Hampshire 2008 Jul 41 
California 2009 Jan 93 New Jersey 2009 Jun 10 
Colorado 2008 Sep 133 New Mexico 2009 Jul 85 
Connecticut 2007 Oct 188 New York 2010 Sep 17 
Delawarea 2007 Jul 31 13 North Carolina 2009 Mar 19 
D.C. 2008 Jan 23 65 North Dakota 2009 Apr 21 
Florida 2003 Jul Ohio 2009 Jun 146 
Georgia 2008 Jul 56 Oklahoma 2007 Nov 181 
Hawaii 2009 Jul 12 Oregon 2008 Jan 193 
Idaho 2007 Jul 97 Pennsylvania 1998 Dec 21 
Illinois 2009 Jun 30 Rhode Island 2009 Jun 30 
Indiana 2007 Jul 51 South Carolina 2008 Jul 20 
Iowa 2008 Jul 81 South Dakota 2007 Jul 121 
Kansas 2008 Jul 96 Tennessee 2007 Jul 44 
Kentucky 2010 Jul 15 112 Texas 2007 Sep 78 
Louisiana 2010 Jul 22 Utaha 2007 Apr 30 54 
Maine 2009 Jul 12 Vermont 2009 May 
Maryland 2009 Apr 14 Virginia 2008 Jul 120 
Massachusetts 2009 Jun 30 −2 Washington 2009 May 11 
Michigan 2009 Sep 10 West Virginia 2008 Jun 68 
Minnesota 2008 Aug 113 Wisconsin 2009 Aug 15 
Mississippi 2012 Jul 76 Wyominga 2009 Jul 111 
Missouri 2009 Aug 28 49      
State Year Month Day Lag State Year Month Day Lag 
Alabama 2009 Jan 245 Montana 2007 Oct 151 
Alaskaa 2010 Sep 61 Nebraska 2007 Sep 150 
Arizona 2008 Sep 26 165 Nevadaa 2007 Oct 140 
Arkansas 2009 Jul 30 154 New Hampshire 2008 Jul 41 
California 2009 Jan 93 New Jersey 2009 Jun 10 
Colorado 2008 Sep 133 New Mexico 2009 Jul 85 
Connecticut 2007 Oct 188 New York 2010 Sep 17 
Delawarea 2007 Jul 31 13 North Carolina 2009 Mar 19 
D.C. 2008 Jan 23 65 North Dakota 2009 Apr 21 
Florida 2003 Jul Ohio 2009 Jun 146 
Georgia 2008 Jul 56 Oklahoma 2007 Nov 181 
Hawaii 2009 Jul 12 Oregon 2008 Jan 193 
Idaho 2007 Jul 97 Pennsylvania 1998 Dec 21 
Illinois 2009 Jun 30 Rhode Island 2009 Jun 30 
Indiana 2007 Jul 51 South Carolina 2008 Jul 20 
Iowa 2008 Jul 81 South Dakota 2007 Jul 121 
Kansas 2008 Jul 96 Tennessee 2007 Jul 44 
Kentucky 2010 Jul 15 112 Texas 2007 Sep 78 
Louisiana 2010 Jul 22 Utaha 2007 Apr 30 54 
Maine 2009 Jul 12 Vermont 2009 May 
Maryland 2009 Apr 14 Virginia 2008 Jul 120 
Massachusetts 2009 Jun 30 −2 Washington 2009 May 11 
Michigan 2009 Sep 10 West Virginia 2008 Jun 68 
Minnesota 2008 Aug 113 Wisconsin 2009 Aug 15 
Mississippi 2012 Jul 76 Wyominga 2009 Jul 111 
Missouri 2009 Aug 28 49      

Source: State legislative records.

Notes: Lag = the number of days after the bill becomes law, before it becomes effective. A negative lag indicates a retroactive effective date. The dates for Florida and Pennsylvania refer to the first appearance of a law with similar provisions to UPMIFA. In Florida, UPMIFA itself was later enacted.

aNo colleges from the state appear in the panel sample.

A key decision for this analysis on yearly data is how to define UPMIFAit to accurately measure when the law impacts colleges. State laws apply at the time of appropriation for spending, a planning period that precedes actual expenditure (Budak and Gary 2010). Rogers (2012) surveyed twenty-three private liberal arts colleges and found that they appropriate an amount to be spent from the endowment prior to the beginning of the fiscal year, and do not change that amount. This happens far enough in advance of the new fiscal year that some forecasting of investment returns and donations is required to know the beginning-of-year endowment value. In contrast, Brown et al. (2014) find that endowment spending at doctoral universities reacts to investment returns during the same fiscal year, particularly investment losses. Neither study directly addresses the process required to react to a new law.

Members of the drafting committee of UPMIFA expected there would be heterogeneity across colleges in reaction time.14 Large endowments would be most aware of legal developments, as they have greater resources devoted to managing the endowment and greater dependence on endowment revenue. But larger endowments also tend to be older and less exposed to underwater funds. Smaller endowments typically heard about the law through nonprofit trade publications or their legal counsel, months after it was enacted in their state.15 In all cases, UPMIFA provides less guidance than prior law, and should initiate a rethinking of what prudent management means. Therefore, it could require time to incorporate UPMIFA into practice. However, colleges may predict the law with certainty in cases where there is a long lag between the date of signing and the effective date.

Taking all of this into account, I define UPMIFAit to indicate that the law is effective on or before 1 July, the first day of the academic fiscal year, and is signed into law at least three months prior (lag > 90 days).16 Given the heterogeneity across colleges, any setting will have some measurement error, which will attenuate estimated impacts of UPMIFA. In Appendix A, I present results with an alternative definition that ignores the signing date.

States can be divided into enactment cohorts by their first year with UPMIFA effective under this definition. In the regression analysis, college fixed effects control for any level differences across cohorts. However, it can be instructive to look for major differences across cohorts, which could suggest policy endogeneity. For the following descriptive statistics, the 2008 cohort also includes earlier enactments and the 2012 cohort also includes later enactments.

In table 1 there are some differences across cohorts. An F-test of equality in means across cohorts is rejected for spending rate and the percent in quasi-endowment. However, there are no monotonic trends by enactment year, and in a regression predicting enactment year, only percent in quasi-endowment is significant. It has a positive relationship with cohort year, indicating colleges with more quasi-endowment funds were in later-enacting states. The lack of any significant variation across cohorts in either average investment rate of return or average percent in underwater funds suggests that the financial crisis had a consistent impact on the average college in each cohort.

In the United States map in figure 2, darker shades of gray mark later-enacting states. The map shows that each cohort is geographically diverse. Again, this shows differences across cohorts but no clear systematic trends.

Figure 2.

Map of UPMIFA Enactments by State

Notes: Enactment cohorts are defined by the first fiscal year UPMIFA was effective in their state in time to affect spending plans. The 2008 cohort also includes earlier enactments, and the 2012 cohort also includes later enactments.

Figure 2.

Map of UPMIFA Enactments by State

Notes: Enactment cohorts are defined by the first fiscal year UPMIFA was effective in their state in time to affect spending plans. The 2008 cohort also includes earlier enactments, and the 2012 cohort also includes later enactments.

Figure 3 reports the mean spending by year and enactment cohort in the panel sample. There is some evidence for parallel trends in the pre-UPMIFA period. However, pre-UPMIFA changes in spending from FYs 2008 to 2009 are not parallel across cohorts, with the latest cohort increasing spending. This is potentially consistent with a situation where prior law reduced spending less for some states, allowing them to enact UPMIFA later. This situation would bias the estimates of the effect of UPMIFA downward, relative to what would be estimated with completely exogenous assignment of timing of enactment. Colleges in later-enacting states being apparently less constrained means that they spend more under prior law, and this decreases the measured difference in spending between colleges in early-enacting states already under UPMIFA, and colleges in later-enacting states under prior law.

Figure 3.

Trends in Spending Levels by Enactment Cohort, 2005—13

Note: Dotted lines represent fiscal years under UPMIFA.

Figure 3.

Trends in Spending Levels by Enactment Cohort, 2005—13

Note: Dotted lines represent fiscal years under UPMIFA.

The data are generally not well suited to provide strong evidence of parallel trends in support of the main identifying assumption. The required assumption is parallel trends across colleges in spending from underwater funds, absent the policy change, but this is difficult to show directly. I infer spending from underwater funds by estimating the coefficient in a regression of dollars of spending on the dollar value in underwater funds. These coefficients are noisy if only using pre-crisis data where underwater funds are very low. Some of the noise in figure 3 also comes from colleges entering and leaving the sample because of item nonresponse on key survey measures.

The regressions in the next section focus on underwater funds as a mechanism for UPMIFA's effect.

5.  Effects of UPMIFA and Underwater Funds

Overall, I find that colleges cut spending from underwater funds when legally required to, but UPMIFA allowed colleges to avoid these cuts. These conclusions are based on statistically significant estimates that are robust to adding covariates.

Regression results appear in table 3. Standard errors in parentheses are clustered at the state level to allow for arbitrary correlation within each state. This accounts for two features of this analysis. Legal changes happen at the state level, and serial correlation of error terms within a state is likely (Moulton 1986; Bertrand, Duflo, and Mullainathan 2004).

Table 3.
Effects of UPMIFA and Underwater Funds on Endowment Spending
 Coefficient (SE) Coefficient (SE) 
Dependent Variable: Endowment Spending ($M) (1) (2) 
Beginning-of-year endowment value ($100M)     
Current year 2.59*** (0.65) 3.44*** (1.03) 
Lagged one year 1.73*** (0.28) 1.59 (1.21) 
Lagged two years 2.05*** (0.32) 1.38*** (0.28) 
UPMIFA −0.38 (0.51) −0.32 (0.43) 
Underwater ($100M) −4.71*** (1.60) −5.11*** (1.75) 
UPMIFA × underwater 6.07*** (2.12) 6.12*** (2.13) 
Investment returns ($100M)     
Current year, positive   2.28** (1.02) 
Current year, negative   −1.69** (0.70) 
Lagged one year, positive   1.88 (1.41) 
Lagged one year, negative   −2.66** (1.17) 
Total expenses ($100M)     
Lagged one year   2.13 (1.58) 
Lagged two years   −0.84 (2.35) 
College and year fixed effects Included Included 
 Coefficient (SE) Coefficient (SE) 
Dependent Variable: Endowment Spending ($M) (1) (2) 
Beginning-of-year endowment value ($100M)     
Current year 2.59*** (0.65) 3.44*** (1.03) 
Lagged one year 1.73*** (0.28) 1.59 (1.21) 
Lagged two years 2.05*** (0.32) 1.38*** (0.28) 
UPMIFA −0.38 (0.51) −0.32 (0.43) 
Underwater ($100M) −4.71*** (1.60) −5.11*** (1.75) 
UPMIFA × underwater 6.07*** (2.12) 6.12*** (2.13) 
Investment returns ($100M)     
Current year, positive   2.28** (1.02) 
Current year, negative   −1.69** (0.70) 
Lagged one year, positive   1.88 (1.41) 
Lagged one year, negative   −2.66** (1.17) 
Total expenses ($100M)     
Lagged one year   2.13 (1.58) 
Lagged two years   −0.84 (2.35) 
College and year fixed effects Included Included 

Sources: NCSE and IPEDS, FYs 2005—13; 2,379 college-year observations across 522 colleges.

Notes: Standard errors (SE) in parentheses are clustered at the state level. Current dollars. Coefficients represent spending rates in cents per dollar, except in the case of the UPMIFA indicator, which represents spending changes in millions of dollars.

**p < 0.05; ***p <0.01.

The dependent variable of spending is measured in millions of dollars. Therefore, the coefficient on the indicator term UPMIFAit represents the million-dollar change in spending level when the indicator is on. Continuous independent variables are measured in hundreds of millions of dollars so that their coefficients represent the cent change in spending, in response to a dollar increase in the independent variable. This is analogous to a spending rate in percentage points.

In column 1, summing the coefficient estimates on the endowment values suggests that colleges spend an average of 6.37 cents per dollar of a moving average, higher than what colleges report for policy spending rates. This is conditional on college and year fixed effects, which take out level differences in spending. Similar sums appear in all of the reported regressions.

The coefficient estimate on underwater funds implies a 4.71-cent decrease in spending for each dollar in underwater funds, nearly equal to the estimated moving-average spending rate. This is consistent with colleges complying with the old uniform law and not spending from underwater funds.

UPMIFA causes these cuts to be offset completely by a 6.07-cent per dollar increase in spending from underwater funds. The coefficient on underwater funds and the coefficient on underwater funds interacted with UPMIFA sum to 1.4, but I cannot reject that this sum is zero. Therefore, colleges under UPMIFA spend no more or less in years with underwater funds. UPMIFA takes away the legal importance of underwater funds relative to prior law, and this is borne out in spending behavior. As expected, the effect of UPMIFA at zero underwater funds is small and not statistically significant.

Adding covariates in column 2 leaves the main results intact. The coefficients on the covariates suggest higher spending following investment fluctuations. Lagged expenses do not appear to have a clear relationship with endowment spending.

In the specification with covariates, the estimated increase in spending from underwater funds under UPMIFA is 6.12 cents per dollar. This is 28 percent larger than the average effective spending rate in the sample. Therefore, when underwater funds make up 17.4 percent of the endowment as they did at the beginning of fiscal 2010, UPMIFA allowed for 22 percent more spending (128 percent times 17.4 percent).

The following section explores differences in impacts of UPMIFA along some observable measures.

6.  Heterogeneous Effects of UPMIFA and Underwater Funds

Increased spending from underwater funds under UPMIFA means that prior law was a binding constraint. In this section, I explore characteristics that would lower the shadow price of that constraint to see whether colleges with those characteristics reacted differently to UPMIFA.

As discussed in section 2, some colleges have access to other sources of spending within the endowment, such as special appropriations and spending from quasi-endowment funds, to offset the legally required cuts in spending from underwater funds. Colleges that rely more heavily on their endowments to cover operating expenses may also be more likely to find offsetting sources of endowment revenue.

Table 4 shows the results from estimating equation 2. (The definitions and means of the Pi variables are reported in table 1.) The three variables identify distinct features of colleges, as they are not highly correlated with each other or with endowment value.

Table 4.
Heterogeneous Effects
 Coefficient (SE) Coefficient (SE) Coefficient (SE) 
Dependent Variable: Endowment Spending ($M) (1) (2) (3) 
Beginning-of-year endowment ($100M) 
Current year 3.41*** (1.07) 3.41*** (1.04) 3.33*** (1.11) 
Lagged one year 1.62 (1.23) 1.66* (1.22) 1.88 (1.18) 
Lagged two years 1.40*** (0.28) 1.37*** (0.29) 1.48*** (0.24) 
UPMIFA −0.74* (0.40) −0.58 (0.46) −0.68 (0.68) 
Underwater ($100M) −15.23*** (3.88) −7.76** (3.43) −4.32* (2.55) 
UPMIFA × underwater 14.21*** (4.26) −9.15** (3.63) −0.94 (3.25) 
Investment returns ($100M)       
Current year, positive 2.22* (1.10) 2.34** (1.03) 1.75 (1.08) 
Current year, negative −1.69** (0.76) −1.70** (0.70) −1.26* (0.91) 
Lagged one year, positive 1.92 (1.43) 1.99 (1.42) 1.76 (1.42) 
Lagged one year, negative −2.46** (1.20) −2.69** (1.17) −0.48 (1.13) 
Total endowment ($100M)       
Lagged one year 2.09 (1.65) 2.16 (1.51) 1.78 (1.67) 
Lagged two years −0.74 (2.31) −0.87 (2.29) −0.90 (2.25) 
Operating expenses covered (%) × 
UPMIFA 0.04 (0.02)     
Underwater 0.78*** (0.27)     
UPMIFA × underwater −0.59 (0.44)     
Ever used special appropriations ×      
UPMIFA   0.58 (0.50)   
Underwater   4.53 (4.35)   
UPMIFA × underwater   −5.66 (4.42)   
Quasi-endowment (%) ×       
UPMIFA     −0.02 (0.01) 
Underwater     0.02 (0.11) 
UPMIFA × underwater     0.19 (0.14) 
College and year fixed effects Included Included Included 
Number of colleges 342 521 467 
Number of college-year observations 1,914 2,378 2,225 
 Coefficient (SE) Coefficient (SE) Coefficient (SE) 
Dependent Variable: Endowment Spending ($M) (1) (2) (3) 
Beginning-of-year endowment ($100M) 
Current year 3.41*** (1.07) 3.41*** (1.04) 3.33*** (1.11) 
Lagged one year 1.62 (1.23) 1.66* (1.22) 1.88 (1.18) 
Lagged two years 1.40*** (0.28) 1.37*** (0.29) 1.48*** (0.24) 
UPMIFA −0.74* (0.40) −0.58 (0.46) −0.68 (0.68) 
Underwater ($100M) −15.23*** (3.88) −7.76** (3.43) −4.32* (2.55) 
UPMIFA × underwater 14.21*** (4.26) −9.15** (3.63) −0.94 (3.25) 
Investment returns ($100M)       
Current year, positive 2.22* (1.10) 2.34** (1.03) 1.75 (1.08) 
Current year, negative −1.69** (0.76) −1.70** (0.70) −1.26* (0.91) 
Lagged one year, positive 1.92 (1.43) 1.99 (1.42) 1.76 (1.42) 
Lagged one year, negative −2.46** (1.20) −2.69** (1.17) −0.48 (1.13) 
Total endowment ($100M)       
Lagged one year 2.09 (1.65) 2.16 (1.51) 1.78 (1.67) 
Lagged two years −0.74 (2.31) −0.87 (2.29) −0.90 (2.25) 
Operating expenses covered (%) × 
UPMIFA 0.04 (0.02)     
Underwater 0.78*** (0.27)     
UPMIFA × underwater −0.59 (0.44)     
Ever used special appropriations ×      
UPMIFA   0.58 (0.50)   
Underwater   4.53 (4.35)   
UPMIFA × underwater   −5.66 (4.42)   
Quasi-endowment (%) ×       
UPMIFA     −0.02 (0.01) 
Underwater     0.02 (0.11) 
UPMIFA × underwater     0.19 (0.14) 
College and year fixed effects Included Included Included 
Number of colleges 342 521 467 
Number of college-year observations 1,914 2,378 2,225 

Sources: NCSE and IPEDS, FYs 2005—13. Samples differ across columns because of missing data.

Notes: Standard errors (SE) in parentheses are clustered at the state level. Current dollars. Coefficients represent spending rates in cents per dollar, except in the case of the UPMIFA indicator, which represents spending changes in millions of dollars. Percent operating expenses covered by endowment is averaged over pre—financial crisis fiscal years 2005—08. Percent in quasi-endowment is averaged over post-financial crisis FYs 2009—13.

*p < 0.10; **p < 0.05; ***p < 0.01.

In column 1, colleges that cover more of their expenses with endowment spending make significantly smaller cuts to spending from underwater funds (φ=0.78), but this does not significantly differ based on UPMIFA (ψ=-0.59). In column 2, colleges that draw special appropriations are also estimated to make smaller cuts to spending from underwater funds, which is offset by UPMIFA, but neither estimate is significant (φ=4.53 and ψ=-5.66).

In column 3, there is no evidence of a lessened impact of UPMIFA by percent in quasi-endowment funds. The sign on ψ is positive, suggesting that colleges with more quasi-endowment funds are actually more affected by differences in state laws. However, ψ is small and not significantly different from zero. In all three of these specifications, for a college at the sample average value of Pi, there is still a significant increase in spending from underwater funds caused by UPMIFA.

Spending down quasi-endowments is a legal way to avoid reduced overall spending under prior law, yet colleges do not appear to do so. Colleges that have more quasi-endowment value may be generally more inclined to save. They treat unrestricted assets as permanent funds by placing them in their endowments, and they appear to continue to preserve the value of quasi-endowment funds at the expense of lost endowment revenue.

Keeping unrestricted funds saved in the endowment even when other revenues are down is consistent with building the endowment for its own sake, or endowment hoarding (Hansmann 1990; Brown et al. 2014). Keeping unrestricted funds in the endowment could also be consistent with the model in Hoxby (2015), where colleges accumulate endowment funds and expend them based on the flow of profitable opportunities to invest in intellectual capital, if market investments were seen to have relatively high growth potential after the financial crisis.

7.  Robustness Checks

The main results are generally robust to alternative definitions of the UPMIFAit variable. The main results also hold when spending is measured in terms of deviations from planned spending, as calculated from reported spending formulas. However, the main results do not hold when dollar measures are normalized to estimate the regression in terms of rates.

The effects of UPMIFA could vary over time. For example, spending could grow over time as more colleges implement new spending policies under UPMIFA. Alternatively, increased spending could be localized to the first year after enactment, if colleges decide it is prudent to temporarily spend more from underwater funds to compensate for foregone spending in years under prior law. I estimate how the effects are distributed over the years after UPMIFA's enactment in an event study, expanding the UPMIFAit term from equation 1 into multiple terms UPMIFAitk. For k=1,2,3,4+, UPMIFAitk is an indicator that year t is the kth year UPMIFA has been in effect in the state of college i. These terms are also interacted with underwater funds in year t.

Table 5 shows the results of estimating this specification. There are positive and significant impacts of UPMIFA in all years after enactment. Overall there does not appear to be a trend in impacts, as the coefficients on the interaction terms are not statistically significantly different from each other. There does not appear to be extra catch-up spending in the first year after enactment. Appendix table A.1 reports the results of estimating the event study using an alternative definition of timing of enactment.

Table 5.
Event Study Defining UPMIFA Relative to First Year Effective
Dependent Variable: Endowment Spending ($M) Coefficient (SE) 
Beginning-of-year endowment ($100M)   
Current year 3.40*** (1.08) 
Lagged one year 1.62 (1.25) 
Lagged two years 1.39*** (0.29) 
UPMIFA (1st year) −0.45 (0.45) 
UPMIFA (2nd year) −1.45** (0.62) 
UPMIFA (3rd year) −1.25* (0.69) 
UPMIFA (4th year+) −1.51 (0.91) 
Underwater ($100M) −5.25*** (1.78) 
UPMIFA (1st year) × underwater 5.27** (2.35) 
UPMIFA (2nd year) × underwater 9.50*** (3.37) 
UPMIFA (3rd year) × underwater 6.44*** (2.05) 
UPMIFA (4th year+) × underwater 6.76*** (2.25) 
Investment returns ($100M)   
Current year, positive 2.26** (1.02) 
Current year, negative −1.70** (0.71) 
Lagged one year, positive 1.86 (1.43) 
Lagged one year, negative −2.61** (1.19) 
Total expenses ($100M)   
Lagged one year 2.10 (1.46) 
Lagged two years −0.95 (2.19) 
College and year fixed effects Included 
Dependent Variable: Endowment Spending ($M) Coefficient (SE) 
Beginning-of-year endowment ($100M)   
Current year 3.40*** (1.08) 
Lagged one year 1.62 (1.25) 
Lagged two years 1.39*** (0.29) 
UPMIFA (1st year) −0.45 (0.45) 
UPMIFA (2nd year) −1.45** (0.62) 
UPMIFA (3rd year) −1.25* (0.69) 
UPMIFA (4th year+) −1.51 (0.91) 
Underwater ($100M) −5.25*** (1.78) 
UPMIFA (1st year) × underwater 5.27** (2.35) 
UPMIFA (2nd year) × underwater 9.50*** (3.37) 
UPMIFA (3rd year) × underwater 6.44*** (2.05) 
UPMIFA (4th year+) × underwater 6.76*** (2.25) 
Investment returns ($100M)   
Current year, positive 2.26** (1.02) 
Current year, negative −1.70** (0.71) 
Lagged one year, positive 1.86 (1.43) 
Lagged one year, negative −2.61** (1.19) 
Total expenses ($100M)   
Lagged one year 2.10 (1.46) 
Lagged two years −0.95 (2.19) 
College and year fixed effects Included 

Sources: NCSE and IPEDS, FYs 2005—13; 2,379 college-year observations across 522 colleges.

Notes: Standard errors (SE) in parentheses are clustered at the state level. Current dollars. Coefficients represent spending rates in cents per dollar, except in the case of the UPMIFA indicator, which represents spending changes in millions of dollars.

*p < 0.10; **p < 0.05; ***p < 0.01.

UPMIFA's impacts may manifest differently across colleges that use different spending formulas. I argue that because spending formulas change often and need not be followed, the most important outcome is a direct measure of actual spending. Looking instead at survey responses about spending formulas, colleges that consistently report using a moving-average formula represent 55 percent of colleges in the sample. They tend to be smaller than others in size, but similar in pre–financial-crisis spending rates. Columns 1 and 2 of table 6 report the results of estimating equation 1 when dividing the sample into moving-average formula users and others. The conclusions in both groups are similar to the main conclusions in section 5. However, the reduction in spending from underwater funds under prior law is significantly weaker among the group using a moving-average formula. This is consistent with moving-average users being more likely to follow their formulas even in the presence of underwater funds.

Table 6.
Robustness Checks Related to Spending Formulas
 Moving-Average User Nonuser Deviation from Planned 
 Coefficient (SE) Coefficient (SE) Coefficient (SE) 
Dependent Variable: Endowment Spending ($M) (1) (2) (3) 
Beginning-of-year endowment ($100M)     
Current year 2.84** (1.09) 3.38** (1.44) 2.78*** (0.32) 
Lagged one year 2.68* (1.42) 1.44 (1.68) −0.61 (0.84) 
Lagged two years 1.87*** (0.41) 1.06*** (0.38) −0.10 (0.19) 
UPMIFA −0.49 (0.37) 0.13 (0.66) −0.67** (0.31) 
Underwater ($100M) −0.93 (1.87) −6.31** (2.86) −5.13** (2.36) 
UPMIFA × underwater 3.58*** (1.30) 6.18** (2.92) 4.95** (2.24) 
Investment returns ($100M)       
Current year, positive −0.19 (1.40) 3.05*** (1.11) 2.20** (1.07) 
Current year, negative −0.95 (1.07) −1.30 (0.96) 0.72 (0.97) 
Lagged one year, positive 0.51 (1.81) 2.16 (2.25) −0.97 (0.88) 
Lagged one year, negative 1.46 (1.97) −3.12** (1.24) 0.87 (0.97) 
Total expenses ($100M)       
Lagged one year −2.49 (3.40) 1.23 (1.28) 4.71 (3.06) 
Lagged two years 0.62 (3.39) 1.12 (2.49) −5.12* (2.67) 
College and year fixed effects Included Included Included 
Number of colleges 287 235 439 
Number of college-year observations 1,320 1,059 1,731 
 Moving-Average User Nonuser Deviation from Planned 
 Coefficient (SE) Coefficient (SE) Coefficient (SE) 
Dependent Variable: Endowment Spending ($M) (1) (2) (3) 
Beginning-of-year endowment ($100M)     
Current year 2.84** (1.09) 3.38** (1.44) 2.78*** (0.32) 
Lagged one year 2.68* (1.42) 1.44 (1.68) −0.61 (0.84) 
Lagged two years 1.87*** (0.41) 1.06*** (0.38) −0.10 (0.19) 
UPMIFA −0.49 (0.37) 0.13 (0.66) −0.67** (0.31) 
Underwater ($100M) −0.93 (1.87) −6.31** (2.86) −5.13** (2.36) 
UPMIFA × underwater 3.58*** (1.30) 6.18** (2.92) 4.95** (2.24) 
Investment returns ($100M)       
Current year, positive −0.19 (1.40) 3.05*** (1.11) 2.20** (1.07) 
Current year, negative −0.95 (1.07) −1.30 (0.96) 0.72 (0.97) 
Lagged one year, positive 0.51 (1.81) 2.16 (2.25) −0.97 (0.88) 
Lagged one year, negative 1.46 (1.97) −3.12** (1.24) 0.87 (0.97) 
Total expenses ($100M)       
Lagged one year −2.49 (3.40) 1.23 (1.28) 4.71 (3.06) 
Lagged two years 0.62 (3.39) 1.12 (2.49) −5.12* (2.67) 
College and year fixed effects Included Included Included 
Number of colleges 287 235 439 
Number of college-year observations 1,320 1,059 1,731 

Sources: NCSE and IPEDS, FYs 2005—13. Samples differ across columns because of missing data.

Notes: Standard errors (SE) in parentheses are clustered at the state level. Moving-average users always use a moving-average formula, and nonusers represent the rest of colleges. In the third column, Deviation from Planned spending is the dependent variable. It represents the difference in dollars between actual spending and planned spending, calculated from NCSE survey responses.

*p < 0.10; **p < 0.05; ***p < 0.01.

Where there is enough information, I use lagged endowment values and reported spending formulas to calculate the planned spending implied by moving-average formulas. In this process, I ignore the potential effects of underwater funds. Similar to table 5 in Brown et al. (2014, p. 952), I calculate the difference between actual spending and an estimate of planned spending. The median observable deviation is zero dollars, and half of all deviations are between negative 0.5 percent and 0.5 percent of beginning endowment value. Column 3 of table 6 reports estimates of the main regression specification with this alternative dependent variable. The key coefficients have the same signs and similar magnitudes to the main results in table 3. This supports the main finding that underwater funds caused reductions in spending, but only under prior law.

All regression analyses so far have measured endowment stocks and flows in terms of dollars. Equation 1 then reflects the way that spending is commonly chosen in practice, by applying constant policy spending rates to current and lagged endowment values. The regression averages across colleges with varying spending rates, and recovers the difference in spending rates from underwater funds between UPMIFA and prior law.

Another way to estimate the equation is to look for UPMIFA's impact on spending rates directly. This approach divides spending and underwater funds by the current endowment value, instead of including current and lagged values in the regression. Column 1 of table 7 shows the results of estimating equation 1 with scaled measures. The interpretation of coefficients is not as clear, as with dollar measures, but coefficients should be of the same signs and similar magnitudes. For example, a coefficient of −4.8 on the fraction of endowment value in underwater funds would imply that if all funds were underwater, then the spending rate would be reduced by 4.8 percentage points (to zero for an average college). The estimated coefficient of −0.8 to −0.6 on the fraction underwater is about 15 percent as large as expected. The estimated coefficient on the interaction term with UPMIFA is smaller still, and not statistically significant.

Table 7.
Robustness Checks Related to Spending Rates
 Coefficient (SE) Coefficient (SE) 
Dependent Variable: Spending Rate (% of Endowment Value) (1) (2) 
UPMIFA −0.07 (0.15) −0.07 (0.15) 
Underwater (fraction of value) −0.80** (0.33) −0.61* (0.35) 
UPMIFA × underwater 0.30 (0.50) 0.21 (0.50) 
Endowment shocks     
Current year, positive   −0.13* (0.09) 
Current year, negative   −0.42*** (0.12) 
Lagged one year, positive   0.34** (0.14) 
Lagged one year, negative   −0.08 (0.12) 
College and year fixed effects Included Included 
 Coefficient (SE) Coefficient (SE) 
Dependent Variable: Spending Rate (% of Endowment Value) (1) (2) 
UPMIFA −0.07 (0.15) −0.07 (0.15) 
Underwater (fraction of value) −0.80** (0.33) −0.61* (0.35) 
UPMIFA × underwater 0.30 (0.50) 0.21 (0.50) 
Endowment shocks     
Current year, positive   −0.13* (0.09) 
Current year, negative   −0.42*** (0.12) 
Lagged one year, positive   0.34** (0.14) 
Lagged one year, negative   −0.08 (0.12) 
College and year fixed effects Included Included 

Sources: NCSE and IPEDS, FYs 2005—13; 2,379 college-year observations across 522 colleges.

Notes: Standard errors (SE) in parentheses are clustered at the state level. Endowment shocks are investment returns as a fraction of lagged college expenses.

*p < 0.10; **p < 0.05; ***p < 0.01.

Table A.1.
Event Study with Alternative Definition of UPMIFA
Dependent Variable: Endowment Spending ($M) Coefficient (SE) 
Beginning-of-year endowment ($100M)   
Current year 3.40*** (1.04) 
Lagged one year 1.63 (1.22) 
Lagged two years 1.37*** (0.31) 
UPMIFA (1st year) −0.37 (0.47) 
UPMIFA (2nd year) −0.29 (0.55) 
UPMIFA (3rd year) −0.38 (0.47) 
UPMIFA (4th year+) −0.26 (0.49) 
Underwater ($100M) −3.57*** (1.02) 
UPMIFA (1st year) × underwater −0.32 (2.21) 
UPMIFA (2nd year) × underwater 4.85** (2.06) 
UPMIFA (3rd year) × underwater 5.24 (3.43) 
UPMIFA (4th year+) × underwater 5.06** (1.88) 
Investment returns ($100M)   
Current year, positive 2.27** (1.05) 
Current year, negative −1.69** (0.72) 
Lagged one year, positive 1.91 (1.41) 
Lagged one year, negative −2.61** (1.17) 
Total expenses ($100M)   
Lagged one year 2.10 (1.50) 
Lagged two years −0.86 (2.21) 
College and year fixed effects Included 
Dependent Variable: Endowment Spending ($M) Coefficient (SE) 
Beginning-of-year endowment ($100M)   
Current year 3.40*** (1.04) 
Lagged one year 1.63 (1.22) 
Lagged two years 1.37*** (0.31) 
UPMIFA (1st year) −0.37 (0.47) 
UPMIFA (2nd year) −0.29 (0.55) 
UPMIFA (3rd year) −0.38 (0.47) 
UPMIFA (4th year+) −0.26 (0.49) 
Underwater ($100M) −3.57*** (1.02) 
UPMIFA (1st year) × underwater −0.32 (2.21) 
UPMIFA (2nd year) × underwater 4.85** (2.06) 
UPMIFA (3rd year) × underwater 5.24 (3.43) 
UPMIFA (4th year+) × underwater 5.06** (1.88) 
Investment returns ($100M)   
Current year, positive 2.27** (1.05) 
Current year, negative −1.69** (0.72) 
Lagged one year, positive 1.91 (1.41) 
Lagged one year, negative −2.61** (1.17) 
Total expenses ($100M)   
Lagged one year 2.10 (1.50) 
Lagged two years −0.86 (2.21) 
College and year fixed effects Included 

Sources: NCSE and IPEDS, fiscal years 2005—2013; 2,379 college-year observations across 522 colleges.

Notes: Standard errors (SE) in parentheses are clustered at the state level. Current dollars. Coefficients represent spending rates in cents per dollar, except in the case of the UPMIFA indicator, which represents spending changes in millions of dollars. The definition of UPMIFA here ignores signing date. **p < 0.05; ***p < 0.01.

Table B.1.
Panel Sample Compared with Population of Private College Endowments
 IPEDS Population Ever in Panel Sample 
Number of colleges 1,336 522 
As % of IPEDS colleges 100 39.1 
As % of IPEDS dollar value 100 61.3 
Mean endowment ($M) 207.8 327.5 
Median endowment ($M) 20.5 73.9 
Mean endowment/expense ratio 1.69 2.23 
Baccalaureate/masters/doctoral/other (%) 36/8/26/30 41/14/34/11 
 IPEDS Population Ever in Panel Sample 
Number of colleges 1,336 522 
As % of IPEDS colleges 100 39.1 
As % of IPEDS dollar value 100 61.3 
Mean endowment ($M) 207.8 327.5 
Median endowment ($M) 20.5 73.9 
Mean endowment/expense ratio 1.69 2.23 
Baccalaureate/masters/doctoral/other (%) 36/8/26/30 41/14/34/11 

Source: IPEDS.

Note: Measures are averages within each college over fiscal years 2005—08.

These results may indicate that large endowments have an undue influence on the main results. The distribution of endowment size is highly skewed, with the largest endowments over 100 times larger than the median endowment. The main regression conditions out current and lagged endowment values, college fixed effects, year fixed effects, and covariates. However, there are both positive and negative outliers driven by large endowment values. The distribution of spending rates has much tighter bounds and few outliers.

The much smaller estimates in the normalized regression are therefore consistent with the largest endowments driving the estimated effect of UPMIFA. When estimating the regression in various subsamples by endowment size, the coefficients do not have a clear monotonic pattern, but are generally larger among large endowments. This could be explained by larger endowments having more resources to shift spending in compliance with prior law, or more monitoring that would enforce compliance with prior law.

Normalizing dollar measures makes the specification more comparable to that of Brown et al. (2014). I include their endowment shock measures as controls in column 2 of table 7. Brown and colleagues estimate that both positive and negative endowment shocks lead to decreases in spending rates among doctoral universities before 2008. I find the opposite among private colleges and universities during the period of this study.

Brown et al. (2014) show that endowment shocks predict spending rates under a range of assumptions about a college's treatment of underwater funds before UPMIFA. In column 2 of table 7, the effects of underwater funds, UPMIFA, and endowment shocks can be directly compared. The effect of underwater funds under prior law has the highest point estimate, but the differences in magnitudes are not statistically significant. This suggests that underwater funds are at least as important an explanation for decreased spending as endowment shocks in this sample.

8.  Discussion and Conclusion

UPMIFA afforded colleges greater flexibility and responsibility to manage their endowments during the financial crisis. Taking account of the effects of the financial crisis on endowment spending in fiscal 2010, the effect of underwater funds were larger for the average college than the direct effect of investment losses. For a typical college that plans spending using a three-year moving-average rule, the 19 percent drop in asset values in FY 2009 would lead to a 6 percent drop in spending in FY 2010, smaller than the 22 percent drop imposed by underwater funds under prior law. For the average college, 22 percent of endowment spending translates to roughly 2 percent of the yearly budget. Spending gaps across states soon closed as proportions in underwater funds decreased and all states enacted UPMIFA.

The most likely reason that prior law was binding is that UPMIFA, like the prior uniform law when it was introduced, represented an overdue update. The gradual development of prudent management in practice was captured in an abrupt change in law.

This episode provided an opportunity to learn how colleges manage resources under constraints. Colleges that had placed larger amounts in quasi-endowment funds were no less likely, and perhaps even more likely, to feel the effect of legal constraints. This suggests that colleges view quasi-endowment funds as permanent funds themselves, not as sources of insurance against unavailability of other endowment income. This is part of the empirical definition of prudent management of endowment funds.

Although endowment decisions reveal part of how colleges manage resources, this study does not address how other revenues and expenses outside the endowment react to changes in endowment revenue. During the brief period when legal variation and underwater funds coincided, many other changes were occurring, making it empirically difficult to trace the redistribution of a few percentage points of the larger budget.

Brown et al. (2014) show how doctoral university operations were affected by shocks to endowment revenue in the 1990s and early 2000s. Universities tended to cut support staff and tenure-system faculty after investment losses, without increasing them after investment gains. Brown, Dimmock, and Weisbenner (2015) show that colleges and universities also tend to receive more donations after investment losses. An important avenue for future research is to show the effects of colleges’ financial and hiring inputs on student achievement outcomes.

Policy makers and the public who are concerned about student outcomes often disagree with how private colleges manage endowment funds. Typically, these observers call for increased spending to help lower the price or increase the quality of education for today's students, without providing a normative model of how endowments and other resources should be allocated across all generations of students. UPMIFA removed one barrier to increased spending in the present—the zero percent ceiling on spending rates from underwater funds. Other proposals have called for a 5 percent floor on effective spending rates from all endowment funds (Wolf 2011). Such proposals would move away from the prudent person standard, would be unlikely to solve a college's complicated intertemporal allocation problem, and could prevent colleges from upholding donor agreements to maintain permanent funds.

All state legislatures now agree on the definition of prudent management of endowments. But at a crucial moment, the old definition limited spending more so than did colleges’ notions of prudence, and briefly prevented colleges in late-enacting states from providing for scholarships, salaries, and other expenses supported by endowments.

Appendix A:  Additional Results

Section 4 discusses the definition of the UPMIFAit indicator, which is equal to 1 in fiscal year t if UPMIFA is effective by the beginning of the fiscal year, and was signed into law at least three months prior. Table A.1 reports the results of estimating the event study using an alternative definition that removes the condition on the signing date. The effect of UPMIFA on spending from underwater funds is smaller on average, and is nearly zero in the first year. This suggests that colleges in late-signing states may not react to the law in the first year it is effective, when it is signed into law too late to affect budgeting for the fiscal year. This supports the choice of definition in the main analysis.

Appendix B:  Data Sources and Sample Selection

Under a restricted-use data agreement, NACUBO and Commonfund provided college-identified data from their yearly surveys of endowments. These surveys include the NACUBO Endowment Study and the Commonfund Benchmarks Study series of Educational Endowment Reports. In fiscal 2009 these series combined to form the NACUBO-Commonfund Study of Endowments (NCSE). For brevity I refer to all of these surveys as the “NCSE.” For more information on data collection and definitions, see NACUBO-Commonfund (2014).

IPEDS is publicly available. Reporting to IPEDS is mandatory for any college with students who receive federal financial aid, and IPEDS therefore achieves a larger sample coverage than the endowment surveys.

Using data from IPEDS, table B.1 compares the panel sample used in this study to the population of endowments at private, nonprofit, four-year colleges and universities in the United States. Relative to the IPEDS Population in the first column, the Panel Sample in the second column is skewed toward larger endowments both at the mean and median. The sample covers 39.1 percent of colleges in the population, but 61.3 percent of the endowment dollar value in the population. The endowment-to-expense ratio in the sample is higher, at 2.23 versus 1.69. The panel contains a much smaller proportion of colleges in the Other Carnegie classification, which includes medical colleges, engineering schools, and theological seminaries. The remaining colleges are distributed in similar proportions across baccalaureate, masters, and doctoral institutions.

Colleges move in and out of the main analysis sample because of item nonresponse on survey measures. However, missing data within colleges do not appear to bias results. Estimating equation 1 with a balanced sample of colleges that report all measures in all years, I draw similar conclusions (results not shown). There is no evidence that a college-year observation appearing in the sample is associated with UPMIFA, conditional on endowment values and college and year fixed effects (regression not shown).

Appendix C:  Selected Variable Definitions

Endowment value consists of permanent funds, including cash, securities, and property, whose income finances operations. This includes true endowment funds perpetually bound by gift instruments, as well as quasi-endowment funds and term endowment funds. Quasi-endowment funds function as endowment by the vote of the governing board. Term endowment funds function as true endowment funds for a period of time, after which they can be spent or remain in the endowment as quasi-endowment funds. This study does not distinguish between term endowment funds and true endowment funds, nor between expired term endowment funds and other quasi-endowment funds.

About 80 percent of colleges reporting to the NCSE in any given year use a moving-average spending formula. Dollars spent (Spendit) is a linear function of current and lagged endowment values. If the college uses a yearly frequency for n years and a constant policy spending rate s, then the function is Spendit=s1nj=0,1,,n-1Beg.valueit-j.

An underwater fund is a donated gift fund whose current value is below its historic dollar value (HDV), the nominal value at donation. The total value in these funds constitutes Underwaterit=fEndmt.itBeg.valueft1{Beg.valueft<HDVf}, for all gift funds f in the endowment of college i at the beginning of fiscal year t, where 1{·} is an indicator function equal to 1 if the inequality is true and 0 otherwise.

Investment returns are net of management and administrative costs. Investment returns during fiscal year t can be broken into a positive component Returnt+=max{0,Returnt} and a negative component Returnt-=min{0,Returnt}.

Operating expenses include direct costs to provide education (e.g., paying salaries) but do not include taxes or debt service.

Special appropriations are defined by each college. Some colleges may spend an unusually large amount but simply report it in their effective spending rate without designating it a special appropriation.

Acknowledgments

For all their help, I would like to thank Associate Editor Stephanie Riegg Cellini, anonymous referees, Luke Anderson, David Bass, Christine England, Susan Gary, Jesse Gregory, William Jarvis, Sheldon Kurtz, Ken Redd, John Karl Scholz, Christopher Taber, Larry Tavares, and the public economics reading group at the University of Wisconsin–Madison.

REFERENCES

Alaska State Legislature
.
2010
.
House bill 416, 26th legislature bill history/action for legislature
.
Available
www.akleg.gov/basis/Bill/detail/26?Root=hb416.
Accessed 27 June 2018
.
Arnold
and
Porter
,
LLP
.
2009
.
UPMIFA and UMIFA enactment
.
Available
https://files.arnoldporter.com/upmifa%20chart%20(nov.%206,%202009).pdf.
Accessed 21 August 2018
.
Bass
,
David
.
2010
.
Spending and management of endowments under UPMIFA: Findings of a 2010 survey of colleges, universities, and institutionally related foundations conducted by AGB in partnership with Commonfund Institute
.
Available
https://www.agb.org/briefs/spending-and-management-of-endowments-under-upmifa.
Accessed 11 October 2018
.
Bertrand
,
Marianne
,
Esther
Duflo
, and
Sendhil
Mullainathan
.
2004
.
How much should we trust differences-in-differences estimates
?
Quarterly Journal of Economics
119
(
1
):
249
275
. doi:10.1162/003355304772839588.
Black
,
Fischer
.
1976
.
The investment policy spectrum: Individuals, endowment funds and pension funds
.
Financial Analysts Journal
32
(
1
):
23
31
. doi:10.2469/faj.v32.n1.23.
Brown
,
Jeffrey R.
,
Stephen G.
Dimmock
,
Jun-Koo
Kang
, and
Scott J.
Weisbenner
.
2014
.
How university endowments respond to financial market shocks: Evidence and implications
.
American Economic Review
104
(
3
):
931
962
. doi:10.1257/aer.104.3.931.
Brown
,
Jeffrey R.
,
Stephen G.
Dimmock
, and
Scott
Weisbenner
.
2015
.
The supply of and demand for charitable donations to higher education
. In
How the financial crisis and great recession affected higher education
,
edited by
Jeffrey R.
Brown
and
Caroline M.
Hoxby
, pp.
151
174
.
Chicago
:
University of Chicago Press
.
Brown
,
Keith C.
,
Lorenzo
Garlappi
, and
Cristian
Tiu
.
2010
.
Asset allocation and portfolio performance: Evidence from university endowment funds
.
Journal of Financial Markets
13
(
2
):
268
294
. doi:10.1016/j.finmar.2009.12.001.
Brown
,
Keith C.
, and
Cristian Ioan
Tiu
.
2015
.
The interaction of spending policies, asset allocation strategies, and investment performance at university endowment funds
. In
How the financial crisis and great recession affected higher education
,
edited by
Jeffrey R.
Brown
and
Caroline M.
Hoxby
, pp.
43
98
.
Chicago
:
University of Chicago Press
.
Budak
,
Susan E.
, and
Susan N.
Gary
.
2010
.
Legal and accounting challenges of underwater endowment funds
.
Probate & Property Magazine
24
(
1
):
29
30
.
Dimmock
,
Stephen G.
2012
.
Background risk and university endowment funds
.
Review of Economics and Statistics
94
(
3
):
789
799
. doi:10.1162/REST_a_00180.
Ehrenberg
,
Ronald G.
2009
. Demystifying endowments.
New York
:
TIAA-CREF
.
Florida State Legislature
.
2003
.
Chapter 2003–399: Senate bill no. 4-A.
Available
laws.flrules.org/2003/399.
Accessed 21 August 2018
.
Fisman
,
Raymond
, and
R. Glenn
Hubbard
.
2003
.
The role of nonprofit endowments
. In
The governance of not-for-profit organizations
,
edited by
Edward L.
Glaeser
, pp.
217
234
.
Chicago
:
University of Chicago Press
. doi:10.7208/chicago/9780226297866.003.0008.
Gilbert
,
Thomas
, and
Christopher
Hrdlicka
.
2015
.
Why are university endowments large and risky
?
Review of Financial Studies
28
(
9
):
2643
2686
. doi:10.1093/rfs/hhv031.
Goetzmann
,
William N.
, and
Sharon
Oster
.
2015
.
Competition among university endowments
. In
How the financial crisis and great recession affected higher education
,
edited by
Jeffrey R.
Brown
and
Caroline M.
Hoxby
, pp.
99
126
.
Chicago
:
University of Chicago Press
.
Hansmann
,
Henry
.
1990
.
Why do universities have endowments
?
Journal of Legal Studies
19
(
1
):
3
42
. doi:10.1086/467841.
Hechinger
,
John
.
2008
.
College endowment tax studied
.
The Wall Street Journal
,
9 May
.
Hechinger
,
John
, and
Jennifer
Levitz
.
2009
.
Battered nonprofits seek to tap nest eggs
.
The Wall Street Journal
,
11 February
.
Hoxby
,
Caroline M.
2015
.
Endowment management based on a positive model of the university
. In
How the financial crisis and great recession affected higher education
,
edited by
Jeffrey R.
Brown
and
Caroline M.
Hoxby
, pp.
15
41
.
Chicago
:
University of Chicago Press
.
Lerner
,
Josh
,
Antoinette
Schoar
, and
Jialan
Wang
.
2008
.
Secrets of the academy: The drivers of university endowment success
.
Journal of Economic Perspectives
22
(
3
):
207
222
. doi:10.1257/jep.22.3.207.
Merton
,
Robert C.
1993
.
Optimal investment strategies for university endowment funds
. In
Studies of supply and demand in higher education
,
edited by
Charles T.
Clotfelter
and
Michael
Rothschild
, pp.
211
242
.
Chicago
:
University of Chicago Press
.
Moulton
,
Brent R.
1986
.
Random group effects and the precision of regression estimates
.
Journal of Econometrics
32
(
3
):
385
397
. doi:10.1016/0304-4076(86)90021-7.
National Association for College and University Business Officers and Commonfund Institute (NACUBO-Commonfund)
.
2009
.
NACUBO-Commonfund endowment study follow-up survey
.
Washington, DC
:
National Association of College and University Business Officers and Commonfund Institute
.
National Association for College and University Business Officers and Commonfund Institute (NACUBO-Commonfund)
.
2014
.
2013 NACUBO-Commonfund study of endowments
.
Washington, DC
:
National Association of College and University Business Officers and Commonfund Institute
.
National Conference of Commissioners on Uniform State Laws (NCCUSL)
.
1972
.
Uniform Management of Institutional Funds Act
.
Available
www.uniformlaws.org/shared/docs/management%20of%20institutional%20funds/umifa1972%20complete.pdf.
Accessed 21 August 2018
.
National Conference of Commissioners on Uniform State Laws (NCCUSL)
.
2006
.
Uniform Prudent Management of Institutional Funds Act
.
Available
www.uniformlaws.org/shared/docs/prudent%20mgt%20of%20institutional%20funds/upmifa_final_06.pdf.
Accessed 21 August 2018
.
Nelson
,
John C.
, and
Roger
Goodman
.
2009
.
U.S. colleges and universities rating roadmap: Focus on special risks during recession & credit crisis. Special comment.
Atlanta, GA
:
Moody's Investors Service
.
Pennsylvania General Assembly
.
1998
.
1998 Act
141
.
Available
www.legis.state.pa.us/cfdocs/legis/li/uconsCheck.cfm?yr=1998&sessInd=0&act=141.
Accessed 21 August 2018
.
Quinn
,
Patrick
.
2009
.
UPMIFA Passes in Massachusetts and Rhode Island
.
Not-for-Profit News
,
16 July
.
Rogers
,
Fred
.
2012
.
Endowment spending: A look back
.
Wilton, CT
:
Commonfund Institute
.
Rosen
,
Harvey S.
, and
Alexander J. W.
Sappington
.
2016
.
What do university endowment managers worry about? An analysis of alternative asset investments and background income
.
Education Finance and Policy
11
(
4
):
404
425
. doi:10.1162/EDFP_a_00193.
Sare
,
John
.
2009
.
Underwater endowments: Understanding your options
.
Available
https://www.pbwt.com/content/uploads/2015/07/EO_NL_UnderwaterEndowments_Mar2009.pdf.
Acces-sed 24 June 2018
.
South Dakota State Legislature
.
2000
.
South Dakota Code
,
chap. 230
.
Thaler
,
Richard H.
, and
J.
Peter Williamson
.
1994
.
College and university endowment funds: Why not 100% equities
?
Journal of Portfolio Management
21
(
1
):
27
37
. doi:10.3905/jpm.1994.409492.
Tobin
,
James
.
1974
.
What is permanent endowment income
?
American Economic Review
64
(
2
):
427
432
.
Weisbrod
,
Burton A.
,
Jeffrey P.
Ballou
, and
Evelyn D.
Asch
.
2008
.
Mission and money: Understanding the university
.
New York
:
Cambridge University Press
.
Winston
,
Gordon C.
1999
.
Subsidies, hierarchy and peers: The awkward economics of higher education
.
Journal of Economic Perspectives
13
(
1
):
13
36
. doi:10.1257/jep.13.1.13.
Woglom
,
Geoffrey
.
2003
.
Endowment spending rates, intergenerational equity and the sources of capital gains
.
Economics of Education Review
22
(
6
):
591
601
. doi:10.1016/S0272-7757(03)00026-8.
Wolf
,
Alexander M.
2011
.
The problems with payouts: Assessing the proposal for a mandatory distribution requirement for university endowments
.
Harvard Journal on Legislation
48
(
2
):
591
622
.

Notes

1. 

“All states” includes the District of Columbia, as well as Pennsylvania, which adopted the key provisions of UPMIFA in 1998, although it is the only state not to have enacted the 2006 version (Pennsylvania General Assembly 1998).

2. 

UPMIFA had a measurable impact on endowment spending, but I do not measure its indirect impact on endowment asset allocation or on other revenue sources, such as tuition. Many empirical studies of endowments take spending rules as given and examine investment choices (Black 1976; Thaler and Williamson 1994; Lerner, Schoar, and Wang 2008; Brown, Garlappi, and Tiu 2010; Dimmock 2012; Goetzmann and Oster 2015; Rosen and Sappington 2016). In contrast, Merton (1993), Hoxby (2015), and Gilbert and Hrdlicka (2015) lay out full models of both investment and spending, taking into account the college's other revenues and expenses.

3. 

Endowment management decisions are made by a combination of a college's or university's governing board and investment committee (NACUBO-Commonfund 2014). Throughout this study I refer to these entities together as the “college.”

4. 

Similarly, endowment contracts are not a binding reason for colleges to hold savings, because most colleges save additional unrestricted assets in quasi-endowment funds. This study therefore focuses on the choice of how much investment income to spend, rather than the choice of which, if any, endowment contracts to accept. A series of studies, including Tobin (1974), Hansmann (1990), Winston (1999), Weisbrod, Ballou, and Asch (2008), and Hoxby (2015), discuss objectives that would lead colleges to accumulate endowment assets.

5. 

Gilbert and Hrdlicka (2015) discuss the clause in many states’ enacted versions of UPMIFA that sets a sharp limit by establishing a rebuttable presumption of imprudence if a college spends more than 7 percent of some measure of current endowment value. Under current practice however, this 7-percent ceiling will rarely be hit as colleges spend closer to 5 percent of current value.

6. 

The law says, “In making a determination to appropriate or accumulate, the institution shall act in good faith, with the care that an ordinarily prudent person in a like position would exercise under similar circumstances, and shall consider, if relevant, the following factors: (1) the duration and preservation of the endowment fund; (2) the purposes of the institution and the endowment fund; (3) general economic conditions; (4) the possible effect of inflation or deflation; (5) the expected total return from income and the appreciation of investments; (6) other resources of the institution; and (7) the investment policy of the institution” (NCCUSL 2006).

7. 

E-mail message to author from Susan Gary, University of Oregon School of Law, reporter for the UPMIFA drafting committee, 22 November 2011.

8. 

For example, Oregon's enactment was moved along by a ULC member who filed it as an Oregon Law Commission bill rather than the more typical route of filing it as a State Bar Association bill, which, given the timing of ULC's publication of UPMIFA, would have required waiting until the next legislative session (Susan Gary, e-mail message, 22 November 2011).

9. 

The error term ɛit also contains omitted factors such as idiosyncratic opportunities for “intellectual venture capital investments” that compete for resources with market investments, making colleges want to spend resources rather than keep them in the endowment (Hoxby 2015). There is no reason to believe that these opportunities would arrive at differential rates that coincide with the enactment of UPMIFA and underwater funds. Many other omitted factors affect endowment management choices, including how a college's close competitors manage their endowments (Goetzmann and Oster 2015). These are also unlikely to be timed coincidentally with UPMIFA. If a college's in-state competitors are affected by UPMIFA itself, which in turn affects the college, this can be thought of as a component of the reduced-form effect of UPMIFA being effective in the state.

10. 

UPMIFA also governs charitable endowments outside of higher education. Endowments of churches, museums, and community organizations tend to be smaller than those of colleges (Fisman and Hubbard 2003). Smaller endowments tend to be made of up newer funds, meaning they had higher concentrations of underwater funds after the financial crisis and potentially felt larger effects of UPMIFA. Without a targeted survey like the NCSE, these organizations are harder to study.

11. 

All measures throughout this study are in current dollars. The results are robust to estimating in constant 2013 dollars using the Higher Education Pricing Index for private colleges.

12. 

A college's operating expenses include direct costs to provide education (e.g., paying salaries) but do not include taxes or debt service.

13. 

Only Alaska, South Dakota, and Pennsylvania did not enact the old uniform law (Arnold and Porter, LLP 2009; Budak and Gary 2010; Alaska State Legislature 2010; South Dakota State Legislature 2000). No colleges from Alaska appear in the panel. South Dakota prohibited spending from funds below historic dollar value, and is therefore coded as non-UPMIFA in the regressions until it enacts UPMIFA. Pennsylvania never enacted a law with reference to historic dollar value but enacted a law in 1998 with similar provisions to UPMIFA (Pennsylvania General Assembly 1998). In 2003, Florida enacted legislation ignoring historic dollar value and imposing a prudence standard for spending similar to that of UPMIFA (Florida State Legislature 2003). Even though Florida later enacted UPMIFA during the sample, there was no substantive change to law or practice. Pennsylvania and Florida are therefore coded as always under UPMIFA in the regressions, and are grouped with the earliest-enacting states in the tables and figures.

14. 

Personal communication, Sheldon Kurtz, University of Iowa College of Law, UPMIFA drafting committee, 4 February 2016.

15. 

For example, Quinn (2009) points out the immediate accounting implications of UPMIFA, but also says that changing spending in response would require action by the governing board.

16. 

According to NCSE responses, 77 percent of college-year observations in the panel analysis sample begin the fiscal year on 1 July, and 20 percent begin on 1 June.