The Return Expectations of Institutional Investors

Transcription

1 The Return Expectations of Institutional Investors Aleksandar Andonov Erasmus University Joshua Rauh Stanford GSB, Hoover Institution & NBER January 2018

2 Motivation Considerable attention has been devoted to estimating investor beliefs about expected returns on asset classes as parameters of portfolio choice models Black and Litterman (1992); Pastor (2000); Avramov and Zhou (2010); Ang, Ayala and Goetzmann (2014) Little direct, large-sample evidence about the beliefs of institutional investors across a range of asset classes, and even less on the cross-sectional drivers of such beliefs Institutional investors appear to have persistent skill in some asset classes (Cavagnaro, Sensoy, Wang and Weisbach, 2016) but not in others (Goyal and Wahal (2008)) Extrapolation of past returns to future expectations has been documented among retail investors Vissing-Jorgensen, 2003; Malmendier and Nagel, 2011; Greenwood and Shleifer, 2014 Is extrapolation also important for institutional investors? Is extrapolation among institutional investors based on rational updating about skills or other factors? Challenge: How to observe/infer institutional investors actual expectations about future expected returns by asset class Andonov and Rauh The Return Expectations of Institutional Investors 2

3 Explaining the portfolio expected return We use newly required disclosures for U.S. public pension funds (around $4T under management) Must disclose by asset class the allocation and the expected return Null hypothesis the only determinants of the Portfolio ER are: The asset allocation chosen by the fund (more risky assets, higher Portfolio ER). Whether the Portfolio ER is stated on an arithmetic or geometric basis Alternative hypotheses: Past returns shape the Portfolio ER: Rational skill hypothesis past returns reflect genuine variation in the skill of pension plans in alternative asset classes Excessive extrapolation past returns affect expectations about asset classes in which they provide no information about future returns (Benartzi, 2001; Greenwood and Shleifer, 2014). Unfunded pension liabilities affect Portfolio ER strategic incentives to reduce the amount of recognized unfunded liabilities through optimistic return expectations (Brown and Wilcox, 2009; Novy- Marx and Rauh, 2011; Andonov, Bauer, and Cremers, 2017) Andonov and Rauh The Return Expectations of Institutional Investors 3

4 Preview of results Public pension funds extrapolate past performance in forming their expectations Past performance adds substantial explanatory power for portfolio expected returns in the crosssection even after controlling for asset allocation and risk-taking. Rational skill hypothesis cannot fully explain the extrapolation 1. Extrapolation occurs across many asset classes including public equity, where there is zero performance persistence for LPs 2. In private equity, the extrapolation of past returns is driven by the oldest investments, even though these are less informative about the future period. 3. Years of experience in private equity is negatively correlated with the return assumption: novices are assuming higher returns in the asset class State governments that face higher unfunded pension liabilities relative to their revenues and GSP assume higher portfolio returns, controlling for asset allocation Operates through higher asset inflation assumption, does not affect expected real returns on assets Strategic incentives to reduce the recognized magnitude of unfunded liabilities Andonov and Rauh The Return Expectations of Institutional Investors 4

5 New Disclosures: GASB 67 Purpose: explanation and justification of the plan s overall long-term rate of return assumption Effective for all fiscal years beginning with 2014 Building-block method: pension plans declare asset allocation and expected returns by asset class Specifically, systems must now disclose The assumed asset allocation of the pension plan s portfolio ; the long-term expected real rate of return for each major asset class ; and whether the expected rates of return are presented as arithmetic or geometric means, if not otherwise disclosed. Andonov and Rauh The Return Expectations of Institutional Investors 5

6 GASB guidelines on the required disclosure GASB provides guidelines with arithmetic real rates of return. In their example the Portfolio ER equals the Pension DR of 7.75%. Andonov and Rauh The Return Expectations of Institutional Investors 6

7 Example of GASB 67 reporting Three Connecticut plans: different asset allocation and expected returns by asset class. DR = 8.00% DR = 8.50% DR = 8.00% Authors calculation: Portfolio ER 7.937% 9.091% 7.636% Andonov and Rauh The Return Expectations of Institutional Investors 7

8 Data We collect the new disclosures for 231 U.S. public pension plans Time period ( 679 obsevations) Source: CAFRs or separate GASB 67 disclosure statements. Reporting basis dimension #1: nominal/real Around 89% report real expected returns, and then an inflation assumption separately The remaining 11% report nominal expected returns, and then an inflation assumption separately We convert all real disclosures into nominal ones with the plan s inflation assumption to allow for comparability We also separately analyze the inflation assumption and the expected real returns Reporting basis dimension #2: arithmetic/geometric 38% disclose on a geometric basis, 62% disclose on an arithmetic basis If returns are lognormally distributed, the difference between arithmetic and geometric would converge as T gets large to approximately σ 2 /2 Systems do not generally disclose assumed volatility Andonov and Rauh The Return Expectations of Institutional Investors 9

9 Portfolio composition and expected returns by asset class Geometric returns are 0.69% lower than arithmetic (implies volatility of 0.118). Andonov and Rauh The Return Expectations of Institutional Investors 10

10 Expected (nominal) returns by asset class Pronounced differences between arithmetic and geometric in the risky asset classes; no differences in fixed income and cash. Andonov and Rauh The Return Expectations of Institutional Investors 11

11 Portfolio ER generally does not match the Pension DR Contrary to GASB guidelines, the Portfolio ER ( dot product ) generally does not match Pension DR Mismatch between the Portfolio ER and the Pension DR for 93% of the arithmetic plans and 89% of the geometric plans. Variation in Portfolio ER opportunity to analyze the drivers of heterogeneity in the formation of return assumptions. Andonov and Rauh The Return Expectations of Institutional Investors 12

12 Pension discount rate and portfolio expected return Positive relation, but considerable variation in the Portfolio ER than is not explained by Pension DR. Example: 48 plans report the same Pension DR of 7.50% in 2014, but their arithmetic Portfolio ER range from 7.19% to 11.32%. Andonov and Rauh The Return Expectations of Institutional Investors 13

13 Components of Portfolio ER and historical averages For arithmetic systems, 371 of the 420 plans have a Portfolio ER that exceeds the past return. For geometric plans, this is the case for 238 of the 259 plans. Pension plans reporting on arithmetic basis Portfolio ER 8.301% Pension plans reporting on geometric basis Portfolio ER 7.608% Inflation rate Real return Inflation rate Real return 2.925% 5.376% 2.825% 4.782% Historical averages: Past arithmetic return 6.835% Historical averages: Past arithmetic/geometric return 6.776%/6.097% Andonov and Rauh The Return Expectations of Institutional Investors 14

14 Empirical analysis Dependent variable Portfolio ER. Indicator variable for geometric reporting. ω as a 5-vector of allocations to public equity, real assets, private equity, hedge funds and other risky assets (the omitted asset categories are fixed income and cash). Control for pension fund size. Add later variables capturing past returns, past standard deviation, and unfunded liabilities. When analyzing the relation with past returns, we control for reporting month fixed effects (pension funds have different fiscal-year ending dates). Include year fixed effects. Double cluster the standard errors by pension plan and year. Portfolo ER it = α t + β Geometric it + γ ω it + ε it Andonov and Rauh The Return Expectations of Institutional Investors 15

15 Explaining the portfolio expected return (1) Pension plans reporting on a geometric basis expect returns that are 72 basis points lower. The asset allocation variables explain 28.9% of the variation in the Portfolio ER and the positive coefficients indicate that pension plans that invest more in risky assets expect higher returns. The past return explains an additional 4.8% of the variation in the Portfolio ER: A one percentage point increase in the average arithmetic return in the previous 10- year period is associated with 32 basis points higher Portfolio ER. The 10-year standard deviation is insignificant indicating that pension plans do not set higher expected returns in response to the risk they took in the past. Andonov and Rauh The Return Expectations of Institutional Investors 16

16 Figure 3: Portfolio expected return and past return In Panel B, we calculate the residuals of Portfolio ER based on column (3) of Table 3. Andonov and Rauh The Return Expectations of Institutional Investors 17

17 Explaining the portfolio expected return (2) Fiscally stressed governments face pressure to maintain higher expected rates of return: An unfunded liability equal to an additional year of total government revenue raises the Portfolio ER by 14 basis points. A one standard deviation of Unfunded liability / GSP (or an increase of 0.083) increases the Portfolio ER by 9 basis points. Past returns and fiscal pressure are both important determinants of the portfolio return assumptions of pension systems. Andonov and Rauh The Return Expectations of Institutional Investors 18

18 Decomposing the portfolio expected return Expected inflation rate: We expect no relation to past returns. Could it be related to past experienced inflation, even if fund investments are diversified? Cross-sectional variation in the experienced inflation: in 2015, the lowest average 10-year inflation rate was in Michigan and equals 1.38, while the highest was in Hawaii and equals Inflation experiences affect inflation expectations at individual level (Malmendier and Nagel, 2016). Public pension funds tend to overweight local investments (Hochberg and Rauh, 2013; Brown, Pollet and Weisbenner, 2015). Though no evidence that performance depends on the cross-sectional differences in inflation rates. Expected real rate of return: Better candidate for extrapolating past performance. The extrapolation could be rational or excessive. Andonov and Rauh The Return Expectations of Institutional Investors 19

19 Expected inflation rate (component of Portfolio ER) Past returns are economically and statistically insignificant. No evidence that pension plans make differential inflation assumptions on the basis of past experienced regional inflation. Pension funds in states with large unfunded liabilities relative to their resources tend to justify higher return assumptions using higher inflation: The economic magnitude is large as 65 70% of the unfunded liability effects found in Table 3. Andonov and Rauh The Return Expectations of Institutional Investors 20

20 Expected real return (component of Portfolio ER) Past returns operate almost completely through increasing the pension fund s real expected return assumptions: The coefficients on past returns are around 90% of their values in Table 3. Geometric reporting and percentage allocated to risky assets are also significantly relater to the expected real rate of return. Andonov and Rauh The Return Expectations of Institutional Investors 21

21 Reasons why pension plans extrapolate their past returns Rational skill hypothesis: The extrapolation is primarily or exclusively in asset classes with performance persistence. Alternative assets: Private equity (Lerner, Schoar and Wongsunwai, 2007; Cavagnaro, Sensoy, Wang and Weisbach, 2016). Real estate mixed evidence (Andonov, Eichholtz and Kok, 2015) Hedge funds mixed evidence (Fung, Hsieh, Naik and Ramadorai, 2008; Jagannathan, Malakhov and Novikov, 2010; Dichev and Yu, 2011). Excessive extrapolation hypothesis: The extrapolation also occurs in asset classes where there is no evidence of performance persistence. Public equity: Pension funds cannot time the hiring and firing of asset managers in public equity (Goyal and Wahal, 2008). These asset managers display heterogeneity in performance, but they have only modest persistence (Busse, Goyal and Wahal, 2010). Andonov and Rauh The Return Expectations of Institutional Investors 22

22 Expected real returns by asset class The magnitudes of geometric coefficients also reflect the volatility of the different risky assets. Past performance is positively related to the expected real returns in all risky asset classes. Andonov and Rauh The Return Expectations of Institutional Investors 23

23 Private equity strongest potential for persistence Extrapolating private equity performance can be explained if pension plans display skill or have differential access to general partners (GPs) of a given quality: Persistent differences in skills and performance among institutional investors investing in private equity funds (Cavagnaro, Sensoy, Wang and Weisbach, 2016). Public pension funds are more likely to reinvest in the follow-on fund of the same GP (Lerner, Schoar and Wongsunwai, 2007). Persistence in performance on a GP level when considering consecutive funds (Kaplan and Schoar, 2005; Hochberg, Ljungqvist and Vissing-Jørgensen, 2013; Korteweg and Sorensen, 2015). Braun, Jenkinson and Stoff (2017): GP-level persistence has diminished over time. Rational extrapolation (skill) hypothesis: Persistence on a GP level could justify extrapolating recent past performance if the performance measures available for such a young fund are sufficiently informative so that a reinvestment decision could be made on the basis of such information. Andonov and Rauh The Return Expectations of Institutional Investors 25

24 Preqin data to test the rational extrapolation hypothesis 2017 Preqin database: pension plan private equity investments and performance. Three categories of investments in private equity funds based on age: Past PE IRR old funds average net IRR of investments made more than 13 years ago. Fully realized and liquidated; Performance does not depend on valuation of unrealized assets. Information from distant past, may not be relevant for investment decisions made today. Past PE IRR medium funds average net IRR of investments made 9 13 years ago. Sufficient time to incorporate cash distributions in the reported returns. Their performance may be more informative about current financial decisions. Past PE IRR recent funds average net IRR of investments made 3 8 years ago. Their performance would be most meaningful for predicting the likely performance of future funds, but only if such performance can be accurately measured. Reported returns depend primarily on the valuation of illiquid assets, though subject to FASB Statement of Accounting Standards 157. Hypothesis: recent and medium PE investments will be most informative about future performance; little reason to use the performance of old funds to develop expectations. Andonov and Rauh The Return Expectations of Institutional Investors 26

25 Expected real returns in private equity Pension plans do not extrapolate the performance of their recent and medium PE investments. The positive relation is driven by the positive relation between the return expectations and the performance of old PE funds. Liquidated and fully distributed. Less experienced pension plans expect higher returns: negative relation with the number of investments. Difficult to rationalize (Lerner, Schoar and Wongsunwai, 2007; Sensoy, Wang and Weisbach, 2014). Andonov and Rauh The Return Expectations of Institutional Investors 27

26 Conclusion Extrapolation of past performance cannot be fully justified by the rational skills hypothesis, seems more in line with excessive extrapolation Strong evidence of extrapolation in public equity, an asset class that comprises almost half of the portfolios and where there is no evidence of after-fees performance persistence at the LP level Even grounds for extrapolation in alternative asset classes is questionable, as it seems to be based on oldest investments and is higher for novices in the asset class Fiscal pressure of liability side leads to increased return assumptions through asset inflation Future work: consider deviations between asset inflation assumptions and actuarial/salary inflation assumptions Andonov and Rauh The Return Expectations of Institutional Investors 28