Spending Policies, Asset Allocation, and Investment Performance of University Endowments Keith C. Brown University of Texas

Cristian Ioan Tiu University at Buffalo

Endowment Asset Management Conference: Institut für Strategische Kapitalmarktforschung Vienna University of Economics and Business Hertenstein, Switzerland 25 August 2014

University Endowment Funds: Issues and Operations •

University endowments have stimulated a considerable amount of recent research on a variety of topics – Existence of endowment: Hansmann (1990); Merton (1993); Hoxby (2014); Cejnek, Franz, and Stoughton (2014) – Asset allocation: Black (1976); Brown, Garlappi, and Tiu (2010); Brown and Tiu (2010); Barber and Wang (2013) – Other issues: Brown et al (2010); Reed, Tiu, and Yoeli (2011), Dimmock (2012); Goetzmann and Oster (2014); Brown, Dimmock, Kang, and Weisbenner (2014); Ang, Ayala, and Goetzmann (2014)

•

Generally, endowment funds need to balance the current spending demands of the institution with the desire to preserve the security and long-term viability of future generations – This is the essence of the inter-generational investment “problem” that endowments face and must solve through their payout and investment decisions – Specifically, endowment boards are responsible for developing two important policies that define how the fund should operate: • (i) Permanent Spending Policy 1 • (ii) Investment Policy

University Endowment Funds: Issues and Operations (cont.) •

Structurally, endowment organization represent a sometimes complex network of both internal (Board, Investment Staff) and external (Consultants, Sub-Managers) personnel

2

The Role of the Endowment Spending Policy • Given how central an endowment’s spending policy is to defining the mission of the institution, it is surprising how little research there has been on the topic – An endowment’s spending policy essentially captures the institution’s attitudes on the inter-generational investment “problem” it faces – An endowment’s investment policy provides the framework for how that investment problem should be solved in an optimal manner

• The prior literature that has considered spending policy issues has tended to focus on two conceptual questions: – Question 1: When should the spending policy be determined relative to decisions about the creation and implementation of the endowment’s investment policy? – Question 2: How often and when should an endowment’s spending policy be changed or otherwise adjusted? 3

The Role of the Endowment Spending Policy (cont.) • There are two prevailing beliefs with regard to the first question on the optimal interaction between endowment spending and investment policies (Question 1): – Endowment spending should be defined so as “…to make investment management independent of the spending decisions of the university”. • Litvak, Malkiel, Quandt (1974); Garland (2005) • Tobin (1974): Inter-generational equity

– Spending policy and investment policy (i.e., asset allocation) should be determined jointly • Dybvig (1999); Blume (2010) • Gilbert and Hrdlicka (2014): Inter-generational fairness

• There appears to be little theoretical controversy over how often spending policies should be changed (Question 2): – The optimal spending rule for a given endowment should not vary over time absent a significant revision in the university’s mission (i.e., utility function) • Merton (1993); Woglom (2003) 4

General Premise of this Study:

• Given the long-term (i.e., infinite-lived) and relatively static investment problem institutions face, endowments should only need to make infrequent changes to their permanent spending policies. • Changes in spending policies may be related to myriad characteristics of the endowment organization (e.g., fund size, budget responsibility), but asset allocation adjustments should be caused by, rather than be the cause of, spending rule modifications • Changing spending policies should not result in differences in risk-adjusted returns 5

Specific Research Goals: • Primary Goal: Examine how the form and level of the payout policies employed at more than 800 university endowment funds are determined and how those spending policies interact with the organization’s asset allocation strategy and investment performance • Additional Goal: Document how both spending rules and rates at endowments change over time and consider whether these changes increase or decrease measurably due to economic conditions (e.g., the 2008 financial crisis) or performance 6

Elements of Endowment Spending Policies • Endowment spending policies have two main features: – Spending Rule: The general process by which the annual payout amount is determined – Pre-Specified Payout Rate: The specific percentage applied within the spending rule formula

• Example: In 2007, the Texas Christian University Endowment decided its payout as a “Hybrid” of: – 50% of previous year’s payout, increased by Higher Education Price Index (HEPI) level – 50% of an amount equal to 5.0% of the endowment’s average market value in the previous four quarters 7

A Taxonomy of Spending Policy Rules 1. Determine on an Appropriate Annual Basis 2. Increase Prior Year’s Spending by a Percentage – Designated percentage often tied to inflation

3. Spend a Percentage of a Moving Average of Past Market Values – Often based on quarterly or annual endowment values over a three- to five-year period

4. Spend a Percentage of Current Yield – Yield defined in various ways as income or earnings

5. Spend a Percentage of Beginning-of-Period Assets Under Management (AUM) – Includes the possibility of no spending at all

6. Hybrid Formulas – Combines two or more categories; Yale or Stanford rules

7. Other Rules 8

Summary of Main Findings • The vast majority of universities (~ 55-75%) use a Moving Average spending rule formula – Moving Average rules most commonly use annual or quarterly reset frequencies over 3-5 years – There has been a slight trend toward using different rules (Decide Annually, Hybrid), particularly at the largest endowments

• Endowments altered their spending policies far more often than predicted during the 2003-2013 sample period – More than half of the endowments changed their spending rules at least once during the period – On average, about 25% of the endowments changed their permanent payout policies in any given year 9

Summary of Main Findings (cont.) • There are significant patterns as to which endowments tend to change their payout rules and in the economic determinants of those changes – Large endowments with the poorest past returns, the lowest past payout rates, and the greatest responsibility for funding the operating budget were more likely to modify their spending policies – Endowments with the ability to use temporary spending appropriations were slightly less likely to modify their permanent spending rules

• Spending policy changes and asset allocation changes are related in an interesting way – Both spending and investment policy changes show a strong tendency to persist over time – Spending policy changes strongly tend to precede asset allocation changes, but not vice versa 10

Summary of Main Findings (cont.) • There appears to be no difference in benchmarkadjusted investment performance between endowments that either did or did not adjust their spending rules during the 2003-2013 sample period – Both the “Change” and “No Change” endowment samples produced positive (although statistically insignificant) alphas – The economic and statistical difference in these alphas between groups was negligible

11

Endowment Fund Data Base • We use the NACUBO-Commonfund Study of Endowments panel data set – Survey-based, self-reported annual observations on payout rules, payout rates, asset allocation weights and investment returns for more than 800 college and university endowment funds

• Available data for each variable: – – – –

Fund-specific return performance: 1984-2013 Fund-specific asset allocation weights: 1989-2013 Fund-specific actual payout rates: 1994-2013 Fund spending policy rules and rates: 2003-2013

• For every endowment in each survey year, we analyze the data base in order to match and categorize: – Nature of spending policy (i.e., rule and rate) – Asset allocation weights in 10 asset class categories – Other characteristics (e.g., assets under management, investment performance, percentage of budget delivered)

12

Endowment Fund Sample: Summary Statistics Assets Under Mgmt ($ Mil)

Annual Return (%)

Annual Payout (%)

Year

# Fnd

Avg

Max

Min

Avg

Max

Min

Avg

Max

Min

2013

795

563.0

32,334.3

0.7

11.7

27.6

1.2

4.4

16.1

0.0

2011

809

504.3

31,728.1

0.6

18.6

31.8

-4.2

4.5

85.0

0.0

2008

704

444.7

36,556.3

0.6

-3.1

12.1

-22.6

4.6

45.6

0.0

2004

703

320.9

22,143.7

1.9

15.0

32.1

-0.6

4.9

18.4

0.0

1999

470

407.6

14,256.0

7.2

10.9

61.0

-15.8

4.8

13.0

0.1

1994

375

189.1

7,045.9

0.5

3.3

43.0

-13.0

5.2

18.2

1.8

1989

280

164.8

4,479.0

1.9

14.6

163.0

-6.2

---

---

---

1984

200

127.0

2,486.3

1.0

-1.4

209.0

-17.0

---

---

--13

Spending Rule Adoption: Over Time •

Moving Average rules dominate adoption, but there is a slight trend toward using Decide Annually and Hybrid rules

14

Spending Rule Adoption: By Fund Size •

The most substantial reduction in using Moving Average rules (and increase in Hybrid rules) came from the largest endowments

15

Asset Allocation Decision: Over Time •

Endowments have significantly increased Alternatives investments over time by decreasing Cash and Fixed-Income investments

16

Asset Allocation Decision: By Fund Size •

Largest endowments have reduced traditional Fixed-Income and increased Alternatives investments by the most

17

Frequency of Spending Rule Changes 611 endowments reported spending rule data over at least one pair of consecutive years – 285 (46.6%) of those endowments did not change their spending rule – 326 (53.4%) of those endowments did change their spending rule at least once

300

250 Number of Endowments

•

200

150

100

50

0 0

1

2

3

4 5 6 Frequency of Changes

7

8

9

10

18

Spending Policy Changes: Year by Year •

Approximately 25% of funds changed some aspect of their spending policy (rule or rate) from one year to the next – This is far greater than expected for a “random chance” event

•

More than 17% of funds changed their spending rules annually

•

Spending rule changes appear to have decreased after the 2008 economic crisis – 2009-12: 13.2% – 2003-08: 19.8%

•

Spending policy changes were abnormally low in 2006 (UPMIFA adoption) 19

Spending Rule Transition Patterns • Only Moving Average and Hybrid rule adoption appear to be stable from one year to the next – Moving Average rules were retained by 89% of funds – Hybrid rules were retained by 83% of funds

20

Spending Rule Retention: Before and After the Economic Crisis

• Every spending rule category was retained with a greater frequency following the 2008 economic crisis • This suggests greater stability in permanent spending rule adoption following the 2008 market crash

20032008

2009 2013

45.4%

61.5%

Increase Pct.

48.4

81.0

Moving Avg.

86.8

93.0

Pct. of Yield

30.4

62.9

Pct. of AUM

44.8

66.0

Hybrid

68.2

86.6

Other

42.7

56.4

Rule Decide Annually

21

Use of Special Spending Appropriations in the Post-2008 Period • Roughly one in four endowments used special spending appropriations in addition to their permanent rules following 2008 • Even five years following the market crash (i.e., in 2013), funds still used special appropriations to the same extent as immediately after the downturn

Rule

2009

2013

28.2%

28.4%

Increase Pct.

34.5

22.5

Moving Avg.

18.1

22.4

Pct. of Yield

21.2

12.5

Pct. of AUM

3.0

22.2

Hybrid

27.7

29.8

Other

19.7

28.1

Decide Annually

22

Economic Determinants of Spending Rule Changes • We define POLCHGT+1 as equaling 1 if an endowment changed its spending rule between Years T and T+1, 0 otherwise • We then estimated several forms of a probit regression based on the following potential economic determinants observable at Year T: Determinant

Variable

Predicted Impact on POLCHG:

Past Fund Return

RET

- (i.e., less likely to change rule)

Fund Return Volatility

VOL

+

Past Payout Rate

PAYOUT

-

Fund Size

LOGAUM

+/-

ALTINV

+/-

DONATE

-

Budget Responsibility

PCTBDGT

+

Special Appropriations

APPROP

-

Alternative Asset Invest Donations to Fund

23

Economic Determinants of Spending Changes (cont.)

• Past Returns, Past Payout rates, and Fund Size are all significant in the predicted direction – More likely to modify spending rules: • Larger funds • Funds with lower past returns • Funds with lower past payout

• The negative impact of past returns becomes did not become materially stronger in the post-2008 period • Interestingly, past donations is not a significant determinant of future spending rule changes 24

Economic Determinants of Spending Changes (cont.) • We repeat the analysis for funds separated into quartiles by the percentage of the institution’s budget (PCTBDGT) the fund is responsible for delivering • There are some important differences between results in the PCTBDGT quartiles in their tendency to adjust spending rules – Funds with lower payouts remain more likely to alter spending policy regardless of budget contribution percentage – Large funds with low past returns are only more likely to implement a spending rule change if they need to deliver a large percentage of the institution’s budget 25

Economic Determinants of Spending Changes (cont.)

• We also repeat the analysis for funds that either did or did not use special spending appropriations (APPROP) during 2009-2013 • Endowments that did not— and perhaps could not—use special appropriations showed a very slight tendency to change their permanent spending rules more frequently – Endowments that did use special annual adjustments had a somewhat less predictable pattern in the permanent spending rule adjustments

26

Budget Contribution and Asset Allocation: Over Time •

Higher Budget Contribution funds have invested more heavily in Alternatives investments

27

Spending Rule Changes vs. Asset Allocation Changes: Persistence and Lead-Lag Relationships • In addition to POLCHG to measure spending rule changes, we define the following measure of asset allocation changes for each Fund E:

• We then estimate a vector auto-regression (VAR[1]) model of the following form:

28

Spending Rule Changes vs. Asset Allocation Changes (cont.) •

Spending policy changes and asset allocation changes both tend to persist over time – Endowments that alter either policy tend to do so repeatedly from one year to the next

•

There is strong evidence that spending rule changes tend to precede asset allocation changes and not the other way around – This is consistent with the logic that the investment problem (i.e., spending policy) be established before the investment solution (i.e., investment policy) is adopted – In particular, it is clear that endowment boards do not appear to treat spending and investment policies as optimally determined in a joint manner

29

Spending Rule Changes and Investment Performance • For each Fund E, we calculate:

where wi is the investment weight for the i-th asset class and RB is the return to the benchmark portfolio

• Over 2003-2013, we calculate the average ALPHA for equal- and value-weighted versions of two portfolios: – Endowments that did not change their spending rules – Endowments that did change their spending rules at least once 30

Spending Policy Changes and Investment Performance (cont.) • Overall, endowment fund managers appear to be good investors (i.e., large, positive average alphas that are marginally statistically significant) • There is no meaningful difference in risk-adjusted performance between endowments that either do or do not change their permanent spending rules

31

Concluding Thoughts: • This study addresses several issues about how endowments set their permanent spending policies and adjust them over time – The most surprising result is that endowments alter their payout policies far more frequently than predicted given the long-term nature of their investment problem • This could suggest that universities change their utility functions more often than expected or that the range of spending rules used in practice is insufficient to describe the solution to their payout problems

– Large endowments that have underperformed and had low past payout rates are particularly likely to adjust their spending rules • These effects are magnified for those endowments responsible for delivering a larger percentage of the institution’s operating budget • The ability to use temporary spending appropriations makes the need to change permanent rules somewhat less likely

– Spending policy changes demonstrate a strong tendency to lead asset allocation adjustments (i.e., the two policies are not determined simultaneously) – There is no difference in risk-adjusted performance between endowments that either do or do not alter their payout policies, although both groups produce positive abnormal returns 32

Concluding Thoughts (cont.):

• Questions for future research – Does changing the spending policy too frequently represent sub-optimal behavior for an endowment? – Does the nature of the institution’s investment problem (i.e., utility function) change more than would be predicted by the long-term nature of its investor profile? – If there are no investment performance costs, what might be the other economic costs of altering the permanent spending policy? – What, if any, are the long-term incremental benefits of spending rule adjustments? – Are there alternative definitions of spending rule formulas that better capture the nature of the university’s investment problem? 33