Financial Changes and Optimal Spending Rates
Among Top Liberal Arts Colleges
Roger T. Kaufman
Professor of Economics
Department of Economics
Northampton, MA 01063
Richard S. Volpert '56
Professor of Economics
Amherst, MA 01002
Financial Changes and Optimal Spending Rates
Among Top Liberal Arts Colleges
The changes in the financial situations among the top 49 liberal arts colleges between 1996 and 2001 are documented using Integrated Postsecondary Education Data System (IPEDS) financial and enrollment data. These data show large disparities in net assets per student, expenses per student, and the subsidies per full-paying student and the average student. Differences in comprehensive fees are considerably smaller. In addition the wealthiest six institutions have been able to maintain relatively greater spending and still increase their net assets faster than their peer institutions. Two views of the optimal rates of spending are then presented and these spending rates are calculated for each institution. Despite current claims of financial crisis, these calculations show that most of these institutions have spending rates well below the rate that would be required to achieve inter-generational equity.
In this paper we describe some of the dramatic changes in wealth and spending that occurred among the nation's top liberal arts colleges between 1996 and 2001. We find this exercise instructive for several reasons. First, there are substantial financial differences among these colleges that are often masked in the national rankings, and the financial hierarchy has undergone some interesting changes. Secondly, it is useful to examine the substantial subsidies that liberal arts colleges give to both full-paying students and, even more so, to those receiving financial aid. Thirdly, since current levels of wealth far exceed what was expected just a decade ago, it is useful to see how colleges have changed their spending and saving patterns and how well they prepared themselves for the most recent decline in financial markets.
Finally, we believe that these data shed light upon several additional questions about the future of liberal arts education at the top colleges in the United States. How will colleges adapt to their new financial positions? Are they spending too much or too little of their total wealth? What difficulties are colleges facing as some of their financial gains dissipate? Are we experiencing a kind of "arms race" among the top colleges and universities? Is it possible for colleges to accumulate "too much" wealth?
The Financial Status of Top Liberal Arts Colleges FY2000/2001
We selected our sample using the top 50 national liberal arts colleges as ranked by U.S. News and World Report in their September 23, 2002 issue. Since there was a three-way tie for last place, we include only the top 49 colleges in our analysis. We recognize that this is an imperfect ranking of the best schools. As we illustrate, however, the most dramatic changes occurred among the very top colleges, which we believe would be included in anyone's ranking of the top fifty. Our principal sources of financial and enrollment data are the Integrated Postsecondary Education Data System (or IPEDS) annual surveys, conducted by the U.S. Department of Education. Unfortunately, the accounting methods used by most liberal arts colleges changed significantly in 1996. In order to achieve consistency we are restricted to annual data beginning in the fiscal years 1996/97. Even in the five-year period ending in 2000/01, however, substantial changes occurred. We should also note several data inconsistencies in the IPEDS data base that are probably due to reporting errors.
In Table 1 we list the top 49 schools and several key variables. The schools are ranked according to net assets per full-time student at the end of the fiscal year 2000/01, typically June, 2001.1 The data in Column (1) reveal an enormous variation in wealth among the top 49 schools and a substantial gap between the top six colleges – Grinnell, Pomona, Swarthmore, Williams, Amherst, and Wellesley – and the remaining 43 schools. Since five of the wealthiest six are also consistently listed among the "best" 5 or 6 liberal arts colleges and Grinnell is typically ranked in the top ten, we find it useful to separate the data for these six schools from the others in much of our analysis. Therefore, at the bottom of our tables we present the averages for these six colleges and the quartile boundaries for the remaining 43. Thus, Q2 measures the median among all remaining 43 while Q1 and Q3 measure the medians of the top and bottom halves of these 43 colleges ranked in order of the variable in question.2 Net assets per student averaged $745,046 in the top six colleges, amounting to 294%, or almost three times, the median among the remaining 43. Thus, even among the top 49 schools there are substantial hierarchies at least with regard to wealth (see Winston (1999) for more on education hierarchies).
Colleges can use their wealth in a variety of ways. They can increase spending on, among other things, instruction, academic support, research, student services, and institutional support. Secondly, wealthier colleges could maintain spending but increase educational subsidies in one of two ways. They can reduce the "sticker price" or comprehensive fee they charge to full-paying students. Alternatively, they could increase financial aid by either attracting a less affluent student population and/or offering more generous financial aid packages. Colleges could also use their wealth to construct new buildings or purchase new land, equipment, and collections.3 Finally, colleges could let their financial wealth accumulate, which would eventually increase the differences in wealth between themselves and less affluent colleges.
In the second column of Table 1 we present data on total expenses per student during the fiscal year 2000/01. Operating and maintenance costs on plant and equipment are included in expenses, while construction expenditures on new buildings and other capital budget items are not. The largest component of expenses is employee payroll and benefits. Expenses will be greater for schools that have lower student-faculty ratios, more supporting staff per student, higher wages and salaries per employee, or a greater capital stock (per student) to maintain. According to the IPEDS data, spending among these 49 colleges ranged from $67,316 per student at Wellesley4 to $26,983 at Whitman.
The differences in spending, and hence the costs of education, among these colleges are considerably smaller than the differences in wealth. The average spending among the wealthiest six was $54,232 per student in 2000/01, 139% of the median of $39,138 among the remaining 43.5 While the difference in net assets per student between the poorest and richest college was 1,129 percent, the maximum difference in spending per student was only 149 percent. Furthermore, the standard deviation in net assets was 75 percent of the median among all 49 colleges, compared to a comparable figure of 22 percent for expenses.6
As the third column of Table 1 illustrates, the differences in the gross comprehensive fee, or "sticker price" among these schools is considerably smaller than the differences in expenses or wealth.7 Even if we exclude Grinnell, which ranks 1st in wealth and 45th in comprehensive fee, the average "sticker price" among the remaining top five schools was $33,230, which is only three percent more than the median sticker price among the remaining 43. Only four schools charged less than $26,000. Full-paying students pay similar amounts at almost all of these institutions even though the cost of their education (as reflected in expenses per student) varies considerably.
The operating subsidy provided to each full-paying students is equal to the difference between total expenses (per student) and the comprehensive fee.8 The wide variation in expenses but relatively similar comprehensive fees is reflected in Column (4) by large differences in the estimated subsidy per full-paying student. In general, the wealthier colleges provide larger subsidies to full-paying students. Although the estimated operating subsidy of $33,922 at Wellesley is even larger than its comprehensive fee, the subsidy at 16 of the 49 colleges is less than $3,000! Indeed, the average subsidy to full-paying students among the wealthiest six colleges is 358 percent of the median subsidy at the remaining 43.
While the wealthiest schools spend more per student but charge a comparable amount, they also use some of their greater wealth to keep their net comprehensive fees below those at many of their peer institutions. The net comprehensive fee may be viewed as the amount paid by the "average" student. It is approximated in Column (5) of Table 1 by adding the IPEDS figures for net tuition (plus fees) and net revenue from auxiliary enterprises. The former is calculated as the price for tuition and fees minus the total amount of aid (institutional and governmental) that is devoted to tuition relief. The latter consists of total charges for auxiliary enterprises, of which room and board are the largest components, minus the amount of aid devoted to auxiliary enterprises.
Based on other information about net comprehensive fees at several of the schools in our sample, we know that these data are only approximate and contain several errors. At Wellesley and several others auxiliary enterprise revenue consists of substantial amounts other than room and board.9 At Middlebury and Oberlin, furthermore, the calculated net comprehensive fees are greater than the comprehensive fees themselves despite the fact that both schools devote considerable amounts to financial aid. These discrepancies may also reflect large auxiliary expenses (such as Middlebury's Bread Loaf programs and Oberlin's Conservatory), but they also suggest reporting errors. Although these approximate estimates of net comprehensive fees vary considerably among institutions, the variations are not closely connected to wealth. Even excluding Wellesley, the average net comprehensive fee among the remaining wealthiest five colleges was $21,566, which is actually less than the median of $23,421 among the remaining 41 schools for which we believe we have reliable data, and the correlation between net assets per student and net comprehensive fee among all 46 colleges with reliable data is –0.21. Note that greater amounts spent on financial aid may reflect a less affluent student body and/or a more generous financial aid policy.
In the sixth column of Table 1 we calculate each college's "subsidy" for the average student, in terms of the amount it spends minus the amounts it receives from students and governments.10 This subsidy comes from the financial funds of the college by using gifts to the alumni fund or funds taken from the endowment. Consequently, we call it expenditure from financial funds, or EXFF. As we explain in the next section, it can be calculated readily from aggregate data in the IPEDS survey that appear to be reliable and avoid some of the aforementioned anomalies.
By providing a more expensive education and more financial aid, the wealthier schools give greater subsidies to the average student. Even in the top group, the variation is substantial. Wellesley provides an average subsidy of $36,543,11 while Amherst's is only $20,306. Even at Amherst and Pomona, however, the average subsidy is 161 percent of the subsidy at the median school.
The data in the first six columns of Table 1 suggest an interesting pattern. The wealthiest colleges have higher expenses, but the differences in expenses among these 49 institutions are considerably smaller than the differences in wealth. Because prices charged to full-paying students vary much less, the wealthier colleges provide greater subsidies for their full-paying students. The correlation between net assets per student and the estimated subsidy per full-paying student is 0.73. Since the wealthier colleges also provide more financial aid, the subsidy for the average student also varies directly with wealth, with a correlation of 0.80.
Even though wealthier schools have greater expenses and provide higher subsidies per student, their wealth is so much greater that spending as a percentage of net assets is inversely related to wealth. As illustrated in the last column of Table 1, the average subsidy among the wealthiest six for fiscal year 2000/01 was a much smaller percentage of net assets (3.77 percent) than the median subsidy (5.16 percent) among the remaining 43 schools. The correlation between net assets per student and EXFF as a percentage of net assets is –0.40.
Overall, these data indicate that among the elite liberal arts colleges:
There are enormous differences in wealth.
There are much smaller differences in comprehensive fees.
The wealthiest institutions spend more on their students and provide more financial aid, but these disparities are not as great as the disparities in wealth.
The wealthiest schools provide larger subsidies per student, but their spending as a percentage of wealth is smaller.
Changes since 1996
Most educational institutions have experienced dramatic changes in their financial positions during the last few years. The stock market boom and partial bust has led to substantial increases in wealth for most. In this section we document these changes in terms of some of the variables defined in the preceding section. We then extend our discussion to talk about long-run trends and sustainability.
In Column (3) of Table 2 we illustrate the dramatic changes in wealth among our 49 institutions between 1996 and 2001.12 Once again, the schools are listed in descending order of their wealth, or net assets per student at the end of fiscal year 2000/01, as in Table 1. The average percentage increase in net assets per student among the six wealthiest colleges over this five-year period was 93.18 percent, compared with a median of 55.98 percent among the remaining schools. Thus, the initial differences in wealth became even greater during this period.
Column (2) in Table 2 depicts the percentage changes in expenses per student from fiscal year 1996/97 through fiscal year 2000/01. While the correlation between the changes in wealth and expenses is 0.37,13 expenses at Grinnell and Amherst rose much less rapidly than the median increase of 23.88 percent among the other 43. During this same period, the higher education price index, or HEPI, rose by 15.7 percent. Consequently, most of these colleges either increased the amounts of resources they used or the price of their inputs rose more rapidly than elsewhere in higher education.
Finally, Column (3) presents the percentage changes in the subsidies per average student, or what we have called expenditures from financial funds. The mean increase among the six wealthiest schools was 51.07 percent, which is less than the median increase among the remaining 43. Williams was the only college among the wealthiest six that had faster growth in subsidies per average student than the median among the others, while Amherst had one of the slowest rates of increase.
Thus, the data in Table 2 indicate that between 1996/97 and 2000/01:
The wealthiest six schools enjoyed rapid increases in wealth.
Spending at the wealthiest schools increased, but not by as much as the increase in wealth.
Subsidies per student rose less rapidly at the wealthiest schools.
The Determinants of Saving
Educational institutions, like individuals can increase their net assets, or save, by spending less than their total income:
Saving = Gifts + Investment Income – EXFF (1)
Equation (1) illustrates that colleges can increase their wealth (i.e., save) in a variety of ways. First, they can receive gifts.14 Secondly, they can receive investment income from their endowments in the form of interest, dividends, other endowment income, and both realized and unrealized capital gains. Finally, they can spend less. EXFF represents what we have called expenditures from financial funds. It is equal to total expenses minus the sum of net comprehensive fees, government grants, and other revenue.15
To determine the affordability and sustainability of various spending patterns we express all the variables in equation (1) as percentages of net assets in Table 3. The numerators for the variables are the average annual financial flows for the 5 fiscal years 1996/97–2000/01. Net assets over the entire period are approximated by averaging net assets at the beginning of fiscal year 1996/97 with those at the end of fiscal year 2000/01.
Column (1) of Table 3 illustrates that gifts to wealthier schools represent a smaller percentage of their net assets than at their competitors during the 1996/97-2000/01 period. Some of these differences may be misleading because they reflect variations in the levels of net assets rather than total gifts. Both Swarthmore and Grinnell, for example, receive substantial gifts, but these are a relatively small percentage of net assets because these two schools are so wealthy.
Investment returns (in Column (2)), on the other hand, were higher among five of the wealthiest six colleges (all but Swarthmore) than all but two of the other 43. This was primarily the result of decisions by trustees of four of these five institutions to invest a considerable portion of their endowments in "alternative assets," which include real estate, private equity, venture capital, oil and gas projects, and hedge funds. In the fiscal year 1999/2000 alone, investment returns as a percentage of long-term investments were 40.8, 48.4, 41.3, and 39.7 at Pomona, Williams, Amherst, and Wellesley, respectively, compared with an average of 12.9 percent at the others.16 At Swarthmore, on the other hand, trustees were more conservative, perhaps because it began the period wealthier than any other liberal arts college. Yet even Swarthmore's annual return during this five-year period of 9.64 percent (of net assets) was greater than all but 11 of the remaining colleges.17 Another reason the wealthier colleges had higher investment returns on net assets is because financial wealth at these institutions comprise a larger fraction of net assets. Schools at which net assets consist primarily of physical capital in the form of college buildings and equipment will ceteris paribus have lower reported investment returns as a percentage of net assets.
The sum of gift and investment income, is represented in Column (3) of Table 3. The totals for the wealthiest six colleges are much closer to the median among the top half of the remaining schools, reflecting lower gift rates but higher investment returns. The last two columns of Table 3 depict the rates of spending from financial funds and saving. Although the wealthiest schools spent greater amounts than their competitors and gave greater subsidies to both the full-paying and average student, these generally comprised a smaller percentage of their net assets than at most of the other institutions (with the exception of Williams and Wellesley). Expenditures from financial funds averaged 5.31 percent among the wealthiest six, compared with a median of 5.86 percent among the remaining 43.
Recall from equation (1) that saving increases with gifts and investment returns and decreases with spending. As a result of high investment returns, five of the six wealthiest schools were able to save a substantially higher portion of their net assets than almost all of the remaining schools. Thus, the wealthiest schools were able to provide more services to their students (reflected by higher expenses), provide more financial aid, and still save a greater percentage of their assets, resulting in a widening gap between themselves and the others.
Recall that colleges can also use some of their increased wealth for capital spending, thereby converting financial assets into physical assets. The IPEDS survey provides data on the value of each of the following at the beginning and end of the fiscal year: land, buildings, and equipment, where the last category also includes the value of art, library, and other collections.18 The value of plant, land, and equipment among the six leading institutions rose by an annual average of 9.32 percent during the four-year period 1997-2001.19 These growth rates are slightly smaller than the median growth rate among the remaining 43 schools. Yet gross investment in physical capital as a percentage of net assets for these six colleges was 1.82 percent, less than the median percentage of 3.34 percent among the remaining 43, and less than the physical capital investment rates at all but 14. Since the wealthiest schools' savings rates were so much greater than the others, the fraction of total savings that went towards physical investment was considerably smaller among the wealthiest colleges. This suggests that the relative financial positions of the wealthiest institutions improved even more than their higher saving rates imply.20
Overall, the data in Table 3 indicate that the most affluent schools were able to save more during the 5 year period from 1996-2001 because:
In general, they enjoyed higher investment returns.
The affluent schools have so much greater wealth that their higher subsidies per student comprise a smaller fraction of net assets.
Sustainability and the Current "Crisis" in College Finances
As college endowments rose during the late 1990's in response to burgeoning financial markets, we have seen how many schools increased their expenses and the subsidies they provided to both the full-paying and average student. When financial markets fell, many colleges experienced declines in the amounts of money from endowment available for the operating budget, resulting in what many college officials characterize as a "crisis."21 Staff and, in some cases, faculty reductions have been announced at Oberlin, Smith, and Yale. Significant budget shortfalls have been declared at Stanford, Dartmouth, Mt. Holyoke, and Skidmore, among others. But what is the nature of this "crisis?" Did these institutions spend too much during the boom years? Do they need to slash spending now in response to declining endowments? What are the appropriate long-run saving and spending rates?
The financial ratios for fiscal year 2000/01 that are presented in Table 4 illustrate the current dilemma. Four of the wealthiest six colleges experienced negative investment returns in 2000/01, and the median investment return among the remaining 43 colleges was -2.19 percent of beginning-year net assets. Although positive gift rates buffered the negative investment returns, saving rates became negative at most schools, resulting in a decline in even nominal net assets. Negative saving rates are unsustainable, but the appropriate response depends critically upon one's view of college endowments.
On that issue we find two strains of thought that differ largely in their treatments of new gifts to endowment. Massy (1990), for example, describes two principles for spending and accumulation that are sometimes in conflict. The first principle is to maintain endowment's share of operating budget support. Since the financial resources of the college include both annual gifts and the endowment, we interpret this principle as requiring that the ratio of EXFF (or total spending from gifts and the endowment) to Total Expenses remain constant. A common justification of this principle is the desire to maintain equality of educational services across generations of students. Assuming no change in the intensity of resources required to provide these services, this principle requires the nominal value of net assets to increase by the same rate as the higher education price index, or HEPI. This would allow nominal spending to rise at the same rate without changing the percentage of net assets used for spending. Today, for example, the consumer price index is projected to increase by 2.5 percent for the next decade.22 If HEPI is expected to rise by an additional 2 percentage points23, the nominal value of the endowment should rise by 4.5 percent.
Recall that the growth rate of net assets is equal to the saving rate. The data in Table 3 indicate that saving rates during the 1996-2001 period exceeded 4.5 percent in all but four of the 49 colleges in our sample. The median saving rate was almost twice that amount! According to this model, the top liberal arts colleges had spending rates during this period that were too small rather than too large.
But the 1996-2001 financial climate was unusual. What is the appropriate long-run spending rate for a more typical period of "normal" investment returns? In order to determine whether current spending rates are sustainable we divide all the variables in Equation (1) by net assets (N.A.) and rearrange them to obtain:
EXFF/N.A. = Gifts/N.A. + Inv. Income/N.A. – Saving/N.A. (2)
Suppose annual gift rates are expected to continue unchanged at rate g, the long-run "normal" rate of return on net assets i is equal to the expected real rate of return r plus the expected rate of CPI inflation πe, and the desired saving rate is s, which may be greater or less than the aforementioned 4.5 percent. Equation (2) can then be re-arranged to calculate the long-run sustainable spending rate (EXFF/N.A.)s
(EXFF/N.A.)s = g + i - s = g + (r + πe ) – s (3)
In Columns (1) and (6) of Table 5 we reproduce the actual gift and spending rate data for the 1996/97-2000/01 period from Table 3, and in Column (5) we calculate the sustainable spending rates for inter-generational equity using a nominal investment return of 8 percent on long-term financial assets, a desirable saving rate of 4.5 percent, and the actual gift rate for each school during this period. The 8 percent return represents a real return of 5.5 percent, which is the rate Massy (1990) suggests for an endowment consisting of 70 percent stocks and 30 percent bonds, and CPI inflation of 2.5 percent.24 Each college's predicted rate of return on net assets is calculated by multiplying 8 percent by its ratio of long-term (financial) investments to net assets in June 2001. Note that the sustainable spending rates are independent of the projected rate of CPI inflation since changes in inflation would affect both i and s in equation (3).25
In Column (7) we calculate the difference between the calculated sustainable rate of spending and the actual spending rate during this period. Using our assumptions, the long-run sustainable spending rates were greater than the actual spending rates during this five-year period at all but 5 of the 49 colleges. At Swarthmore, for example, the sustainable spending rate would be 2.37+7.73-4.5=5.60 percent, which is 0.72 percentage points greater than its actual spending rate of 4.88 percent during this period. At Centre College, on the other hand, the gift rate was larger but its ratio of financial to total net assets was smaller, and its sustainable rate of 3.97 + 6.01 – 4.5 = 5.48 percent is 2.95 percentage points below its actual spending rate of 8.43 percent.26 Furthermore, re-examination of the data in Column (4) of Table 4 indicates that the actual spending rates in fiscal year 2000/01 were generally below the five-year averages. As a result, the actual 2000/01 spending rates exceeded the sustainable rates in only two schools. This is hardly suggestive of a crisis.
These calculations assume that each college's gift rate during this five-year period will continue, and the gift rates among these schools vary considerably. Gift rates, however, will be unusually high during capital fund drives. Some of these schools, like DePauw, have received unusually large gifts during this period. Consequently, in the last two columns of Table 5 we present another calculation of each college's sustainable rate of spending based on an estimate of each school's "normalized" gift rate. The normalized gift rate is the predicted rate of giving for each school using its actual wealth per student at the end of fiscal year 2000/01 and the coefficients from a cross-section regression among our 49 colleges during the 1996/97 – 2000/01 period in which we regress the logarithm of actual gift rates on a constant and the logarithm of net assets per student. In order to reduce the effect of extreme values (e.g., at Sarah Lawrence and Bard), we used a regression that minimized the sum of absolute deviations (called LAD, or least absolute deviations) rather than ordinary least squares.
In most cases the normalized sustainable spending rates are similar to the un-normalized rates, and the correlation between the two is 0.47. As expected, the actual gift rates at Sarah Lawrence and Bard far exceeded their "normalized" rates. The number of schools whose actual spending rates exceed their normalized sustainable rates increases to 15. Williams, Carleton, the University of the South, and Centre are the only colleges whose actual spending rates exceed their sustainable rates calculated both ways.
Nevertheless, the data in Tables 4 and 5 imply that most of these colleges have spent prudently during this five-year period. Even though their saving rates in fiscal year 2000/01 were negative in many cases, their overall spending rates are still lower than their sustainable rates. As a result, their net assets per student will continue to rise faster than HEPI and they will be able to hoard their wealth and/or increase the amount of resources they provide to future generations. Ceteris paribus, the ratios of expenditures from financial funds to their operating budgets will increase. In order to preserve inter-generational equity and keep the ratio of EXFF to Expenses constant over time, actual spending rates would have to increase!
Treasurers' View of Sustainability and the Arms Race
If the preceding analysis is correct and spending rates are not excessive, why do many college officials and trustees see a crisis? After all, the real value of net assets per student at most of these institutions had never been greater! We suggest that the answer lies in the alternative way in which some college treasurers and trustees view their obligation. Massy's second principle for spending and accumulation is to "maintain the purchasing power of each existing endowment fund."27 Although some gifts, like those to the alumni fund, are earmarked for the operating budget, this principle implies that the real endowment should grow in response to new gifts to endowment. As Tobin (1974) avers, real "consumption rises to encompass an enlarged scope of activities when, but not before, capital gifts enlarge the endowment."28 Consequently, the nominal value of new gifts to the endowment (as well as the existing endowment) should rise by the increase in HEPI.
One can compare the two views formally by distinguishing between new gifts (as a percentage of net assets) that are used for the operating budget, and gifts to the endowment. Under the treasurer's view, the sustainable spending rate in equation (3) should include only gifts for the operating budget and not gifts to the endowment. At this spending rate, existing endowment funds will grow at 4.5%, and the total endowment will grow at (i.e, the saving rate will equal) 4.5% plus the endowment gift rate.
In Table 6 we use the treasurers' view to present another set of sustainable spending rates. Based on conversations with college officials and a cursory examination of the treasurers' reports at several of the schools in our sample, we assume that approximately one-fourth of all gifts (mainly annual giving to the alumni fund) are meant to support the operating budget and three-fourths are intended to increase the amount of real educational services provided to future generations of students. As before, we assume a long-run real investment return on financial assets of 5.5 percent, CPI inflation of 2.5 percent, and HEPI inflation of CPI inflation plus 2 percent, or 4.5 percent. As in Table 5, one set of estimates uses the actual gift rates during this period and another uses the normalized rates from the aforementioned regression.
The implications of the treasurers' view are dramatic. Using either the actual or the normalized gift rates, we calculate that the sustainable spending rates lie below the actual spending rates at all 49 colleges except Grinnell, Colorado College, and Rhodes.29 The average difference between the actual and sustainable spending rates (using the normalized gift rates) is 1.84 percent of net assets among the wealthiest six colleges, and the median difference among the remaining 43 is 2.73 percent of net assets.30
Even in the absence of the stock market's decline, the treasurers' view suggests that financial restraint is necessary at almost all of the top liberal arts colleges assuming that the future real rate of return on long-term investments remains equal to 5.5 percent. If the trustees' objective at these colleges is to increase the purchasing power of the endowment in response to new gifts in order to provide additional services to future generations of students (or hoard the increase) and maintain the real value of each new gift to endowment, current spending rates are probably too high and need to be reduced.
It is important to understand the long-run implications of both views. Critics of the treasurers' view argue that it leads to what many people see as an arms race among the top liberal arts colleges (and universities).31 Although their wealth has grown during the past decade more rapidly than they ever expected, many colleges still feel poor because they compare themselves not only on the basis of the educational services they provide, but on the size of their endowments. Instead of allowing some of the principal from new gifts to the endowment to be used to improve the quality and quantity of educational services provided to current as well as future generations of students, these gifts must be added to an ever-increasing endowment.
Recall that the desired saving rate in the treasurers' view keeps net assets growing at the rate of increase of HEPI plus the endowment gift rate. The first component provides a constant stream of educational services to future generations of students while the additional growth in net assets resulting from new gifts to endowment could either be hoarded or used to increase the amount of educational services provided in the future. As an example, suppose three-fourths of all gifts are intended to increase the purchasing power of the endowment. This assumption along with the normalized gift data in Column (8) of Table 6 implies that the average annual increase in the purchasing power of net assets per student among the remaining 43 colleges would be ¾(4.76%) = 3.57 percent. At this rate the amount of educational services that these institutions could provide to future generations of students would double every 20 years!
Conversion to Endowment Takeout Rates
Some readers may find it useful to convert our sustainable spending rates to approximate endowment takeout rates. In the treasurers' view, the takeout rate would equal the long-run nominal rate of return on the endowment minus HEPI inflation. If, for example, HEPI grows at 4.5 percent and the nominal rate of return is 8 percent, the endowment takeout rate should be 8 – 4.5 = 3.5 percent, regardless of the gift rate. This would keep the existing endowment, exclusive of new gifts, growing at the rate of HEPI inflation. The real value of the endowment (deflated by HEPI) would then grow by any new gifts to the endowment. Most colleges currently have target takeout rates between 4 and 5 percent (Cambridge Associates (2000)). While these takeout rates allow the existing endowments to rise more quickly than the CPI, they will rise by less than HEPI (and therefore provide fewer educational services) in the absence of new gifts. This conclusion, of course, depends critically on our assumption that HEPI inflation exceeds CPI inflation by two percentage points.
Conversion from spending to takeout rates in the inter-generational equity model requires two adjustments. Recall that our spending variable includes spending from both gifts and endowment. Consequently, in order to calculate spending from endowment we need to subtract new gifts that support the operating budget, which are typically annual contributions to the alumni fund. We approximate this as one quarter of all gifts. Next, we need to express spending as a percentage of endowment, rather than net assets.
In Column (9) of Table 5 we calculated that the average normalized sustainable spending rate is 5.22 percent of net assets for the wealthiest six colleges using the inter-generational equity model. The average normalized gift rate among these six schools (in Column (8) of Table 5) is 2.33 percent, one quarter of which will be used to support the operating budget. Finally, the average ratio of net long-term investments to net assets for these schools was 92.4 percent at the end of FY 2000/01. Thus, our estimate of the average normalized sustainable endowment takeout rate for the wealthiest six schools would be (5.22 – 0.58)/0.924 = 5.09 percent, which is larger than the typical effective target takeout rates for most of these colleges.32
The comparison for the remaining 43 colleges is more dramatic. The median normalized sustainable spending rate is 7.14 percent of net assets in Table 5. The median normalized gift rate among these schools was 4.76 percent, and the median ratio of net long-term investments to net assets was 81.85 percent. Consequently, our estimate of the median normalized sustainable endowment takeout rate using the inter-generational equity model would be [7.14 – 0.25(4.76)]/.8185 = 7.27 percent. Obviously, this is substantially greater than the typical target takeout rate.
By providing a more expensive education than their peers without charging substantially higher comprehensive fees, the wealthier colleges among the top 49 liberal arts colleges provide greater subsidies to even full-paying students. They also provide more financial aid, resulting in even greater subsidies to the average student. Nevertheless, the differences in wealth exceed the differences in spending and subsidization. If current trends continue, the differences in wealth will widen.
Although many college treasurers and trustees believe their schools are in a crisis, the reality depends upon one's view of inter-generational equity and the treatment of new gifts to endowment. If the goal is to maintain the ratio of expenditures from financial funds to total expenses, there is no crisis. On the contrary, current spending rates for most schools are lower than their sustainable rates.
If, however, the objective is to increase the purchasing power of the endowment in response to new gifts in order to provide additional services to future generations of students and maintain the real value of each new gift to endowment, current spending rates are probably too high and need to be reduced. This policy, however, implies that college endowments will on average grow much faster than total expenses, and an ever-increasing share of total expenses will be financed from endowment. Since this will appear like an arms race, college trustees must be prepared to justify ever-increasing tuitions to an increasingly skeptical public. In either event, college trustees need to recognize the long-term implications of the spending rules they adopt.
America's Best Colleges, 2003 Edition. (September 23, 2002). U.S. News and World Report, p. 60.
Cambridge Associates, Inc. (2000). 1999 NACUBO endowment study. Washington, DC: National Association of College and University Business Officers.
College Entrance Examination Board. (2002). The college handbook. Washington, DC: College Entrance Examination Board.
Coiner, H. M. (1992). How large a fraction of university endowment may safely be
spent? Journal for Higher Education Management, 8 (1), 57-67.
Federal Reserve Bank of Philadelphia (November 22,2002). Survey of professional
forecasters. Philadelphia: Federal Reserve Bank of Philadelphia.
Hansmann, H. (1990). Why do universities have endowments? Journal of Legal
Studies,19 (1), 3-42.
Hopkins, D. & Massy, W. (1981). Planning models for colleges and
universities. Stanford, CA: Stanford University Press.
Kaufman, Roger T. & Woglom G. (2003). Incorporating non-financial wealth in
college and university investment strategies. Journal of Educational Finance,
29 (1), 61-82.
Massy, William F. (1990). Endowment: perspectives, policies, & management.
Washington, DC: Association of Governing Boards of Universities and Colleges.
National Center for Education Statistics (various dates). Finance survey of the NCES
integrated postsecondary education data system. Retrieved, March 9,2004, from
Swensen, David. (2000). Pioneering portfolio management. New York: The Free Press.
Tobin, James (1974). What is Permanent Income? American Economic Review. 64 (2), 427-32.
Winston, Gordon C. & Lewis, E.G. (1993). Physical capital and service costs in US colleges and universities: 1993. Eastern Economic Journal, 23 (2), 165-89.
Winston, Gordon C. (1999). Subsidies, hierarchy and peers: the awkward economics of higher education, Journal of Economic Perspectives, 13 (1), 13-36.
Table 1: Summary Financial Statistics FY 2000/01
Table 2: Percentage Changes in Financial Variables FY1996/97 and FY2000/01
Table 3:Financial Ratios Fiscal Years 1996/97-2000/01
Table 4:Financial Ratios FY 2000/01
Table 5: Sustainable Spending Rates
Table 6: Treasurers' View of Optimal Spending Rates
[These tables are on the Excel file, "Tables Mar 7 2004.xls" on the diskette accompanying the hardcopy of the paper]
1 Net assets are assets minus liabilities. The major component of the former at most schools is the endowment, but it also includes the value of life funds and the book value of plant, equipment, and collections. Prior to 1996 private colleges and universities completed the same IPEDS survey as public institutions (the GASB survey). In this survey respondents are asked to estimate the replacement value of their buildings and equipment. Since 1996, however, private institutions complete the FASB survey, which does not ask these questions. The major liability at most colleges is the value of the outstanding debt from bond issues. As denominator, we chose to use the total full-time student population. Undergraduates comprised more than 95 percent of the total in 2000/01 at 41 of these 49 colleges. At another three (Union, the University of the South, and Wesleyan), undergraduates comprise between 90 and 95 percent, and at the remaining five (Bard, Bryn Mawr, Sarah Lawrence, Smith, and Washington and Lee) they comprise between 80 and 90 percent of the full-time student body.
2 When quartiles do not divide evenly, EXCEL uses interpolation to calculate medians.
3 Since physical assets are illiquid, once they are built a college loses much of its flexibility in reallocating that wealth to other uses.
4 This is higher than the $60,510 figure for "educational and general costs per student” that is reported in the Wellesley College Annual Report
. Most of this difference reflects our decision to include the costs for auxiliary enterprises, most of which comprise the costs of room and board. We decided to include this category for several reasons
, one of which is the apparent practice of some schools to classify room and board costs as educational costs while others classify them as auxiliary enterprises. Consequently, the IPEDS data do not allow us to separate educational and general costs from other costs in a consistent and/or reliable manner. A small portion of our estimate of total expenses may also include costs for other, unrelated auxiliary enterprises.
5 More comprehensive measures of total costs, such as those developed by Winston and Lewis include an estimate of the implicit cost of physical capital (1993). Inclusion of these costs would increase the differences in our sample.
6 Although the coefficient of variation, calculated as the standard deviation divided by the mean, is more frequently utilized as a summary statistic, we used a summary statistic with the median as the denominator because a few of the extreme values in our sample skew the mean.
7 Comprehensive fee data for these 49 colleges were taken from The College Board College Handbook, published by the College Entrance Examination Board (2002).
8 Most, but not all, of this amount is a subsidy from the institution to the students. The cost of services paid by federal and state grants and contracts and the costs of running auxiliary enterprises, however, are included as expenses even though they may not be provided for (or paid by) the full-time students.
9 While imputed data for net tuition seem to be more accurate, several schools mistakenly reported their comprehensive fee as their tuition.
10 Our computed subsidy includes grants from private entities, like foundations, but these amounts are small for almost all of these liberal arts colleges.
11 Because of the substantial size of Wellesley's auxiliary enterprises, not all of this subsidy is spent on the full-time students.
12 To provide a slightly longer period we calculated the change in net assets per student from the beginning of fiscal year 1996/97 (which is usually July 1, 1996) until the end of fiscal year 2000/01, which is usually June 30, 2001. The starting point for the remaining variables in Table 2 is the fiscal year 1996/97.
13 Two colleges, Dickinson and Vassar
, reported suspicious declines in expenses per student. At both schools, however, enrollments rose -- by 28 percent at Dickinson and by 5.5 percent at Vassar. Thus, total expenses increased at both schools.
14 The IPEDS data do not separate gifts from private grants and contracts. Furthermore, the surveys include two items representing "other specific changes in net assets,” which often represent changes in institutions' accounting methods. Consequently, our saving variable does not always correspond with the reported change in net assets.
15 Other revenue includes government appropriations, revenue from hospitals and independent operations, and sales and services of educational activities and auxiliary enterprises that are not counted elsewhere. Much of the revenue from auxiliary enterprises is included in the room and board component of net comprehensive fees. An appendix that indicates how each of the variables in this paper was derived from the IPEDS data is available from the authors upon request.
16 These returns are similar to those earned by the top Ivy League institutions and the other top national universities that also devoted a large part of their endowment to alternative assets.
17 Incomplete data indicate that the wealthiest schools experienced larger declines in their endowments during fiscal year 2001/02 than most of the other top colleges, but none of the declines among the wealthiest six exceeded ten per cent.
18 According to the IPEDS instructions, accumulated depreciation is not included (or subtracted) in these estimates of the disaggregated categories of physical assets although it is subtracted in the calculation of total and net assets.
19 The IPEDS survey for fiscal year 1996/97 only provided data on the book value of plant, equipment, and collections for the end of the fiscal year. Thus, we only have four years of data.
20 This conclusion, however, may be premature if the wealthiest six colleges have recently embarked upon constructions project based on their newly acquired wealth.
21 Almost all of these colleges are obliged to run a balanced operating budget in most years. Our observations and discussions with college personnel indicate that colleges seem to fall into one of two categories concerning the relationship between spending from endowment and the operating budget. Some that have fixed takeout ratios calculate the amount that will be available to spend and then formulate their operating budgets based on this takeout. The formula for each school is given in the annual NACUBO Endowment Study, prepared by Cambridge Associates, Inc. for the National Association of College and University Business Officers. The most common formula allows for spending a specific percentage of the endowment's value, either last year's value or some rolling average. Colleges with more flexible takeout rates first construct their operating budgets (subject to some flexible constraints about spending from endowment) and then withdraw funds from the endowment sufficient to balance the budget (up to some limit). Consequently, another common formula allows for spending from endowment to grow at some fixed rate, often equal to the rate of inflation plus a small percentage.
22 Federal Reserve Bank of Philadelphia (2002).
23 It seems unlikely to us that HEPI inflation will be almost double that of CPI inflation for the indefinite future although this would accommodate 2% increases in real wages for continuing employees
, assuming no changes in productivity. Part of the additional two percentage points might also reflect requisite future increases in resources required by advancing technology, as discussed in Hopkins and Massy (1981) or trustees' beliefs that their college will need to increase the real (inflation-adjusted) subsidies provided to future students, either full-paying students or those on financial aid.
24 The expected rate of return would obviously depend upon the portfolio allocation of the endowment and would be higher at those institutions that have invested heavily in alternative assets. In Kaufman and Woglom (2003), we investigate this allocation in more detail for several of these colleges.
25 Throughout this paper we have ignored the effects of volatility and uncertainty. These effects have important implications for determining optimal spending rules, which are discussed in Coiner (1992). He shows that sustainable spending rates can be affected by as much as 1 percentage point because of volatility and uncertainty.
26 These numbers are slightly different that those in Column (7) because of rounding.
27 Massy, (1990), p. 47.
28 Tobin (1974), p. 427.
29 The correlation between net assets per student and the excess of the treasurers' sustainable spending rate over the actual rate is 0.32 (using the normalized gift rates). The sustainable spending rate, of course, is very sensitive to the assumption that one fourth of all gifts are earmarked for the operating budget.
30 Although the median (normalized) sustainable spending rate of 3.28 percent of net assets (among the remaining 43 colleges) may seem low, recall, that the median ratio of long-term investments to net assets among these schools is 0.82. Consequently, a spending rate of 3.28 percent of net assets is equivalent to 4 percent of long-term investments. This figure is much closer to the target endowment payout rates at most of the schools in our sample when adjustments are made to account for the effects of using rolling averages.
31 See, for example, Hansmann (1990).
32 Although many schools have statutory takeout rates in this range, they usually use a 2-3 year rolling average of the endowment as their base. When the endowment is rising over time, this results in lower effective takeout rates.