Economies of Scale, Lack of Skill, or Misalignment of Interest? 24 th October, 2006 Colloquium ICPM

Transcription

1 Economies of Scale, Lack of Skill, or Misalignment of Interest? 24 th October, 2006 Colloquium ICPM

2 The Project Participants The instigator: Keith Ambachtsheer The researchers: Rob Bauer (Maastricht University and ABP) Rik Frehen (Maastricht University) Hubert Lum (CEM Benchmarking Inc.) Rogér Otten (Maastricht University) Funding from ICPM and Netspar, CEM Benchmarking Inc. provided pension data

3 Agenda I. Introduction, scope and a teaser II. The performance debate: presentation of the main research questions and hypotheses III. Data and research methods IV. Discussion of empirical results

4 I. Introduction, scope and a teaser

5 Introduction and Scope We document the comparative performance of DB, DC and mutual funds We focus on US (and Canadian) domestic equity portfolios Key question: what determines performance differences between pension and mutual funds? Why relevant?

6 Why Relevant? Keith Ambachtsheer (FAJ, 2005): In a world where the clients of (.) financial services organizations are millions of remote, faceless individuals, will the boards and managers (.) they hire serve the financial interests of the beneficiaries? Or will they use their power to serve their own interests?

7 Why Relevant? David Swensen (CIO Yale Endowment), in his recent book Unconventional Success (2005): The for-profit mutual fund industry consistently fails the average investor. From excessive management fees to the frequent churning of portfolios, the relentless pursuit of profits by mutual fund management companies harms individual clients.

8 Why Relevant? David Swensen proposes: A contrarian investment alternative that promotes well-diversified, equity-oriented, market-mimicking portfolios.. which should be implemented by not-for-profit investment companies such as Vanguard and TIAA-CREFF

9 Why Relevant? John Bogle in The Battle for the Soul of Capitalism (2006) : In mutual funds, you get what you don t pay for In the June 2006 workshop, after showing him some preliminary results of today s paper, Mr. Bogle signed my copy with: Press on, regardless!

10 What You Do Not Pay For Source: DALBAR; Bogle Financial Market Research Center

11 Economies of scale, (lack of) skill or misalignment of interest? A few teasers

12 Economies of Scale? Regressing total equity costs per dollar under management (CEM database) on size of the equity holdings:

13 Lack of Skill? Bauer, Derwall and Otten study Canadian mutual fund performance (forthcoming in: Journal of Business Ethics, 2006): Canadian funds on average underperform the broad market with almost 3% p.a.

14 Misalignment of Interest? David Swensen lists potential causes of poor mutual fund performance Obvious causes: Mutual fund fees, including sales loads, management,distribution, incentive and intermediary fees. Portfolio turnover, including trading costs in all its dimensions, tax costs of turnover, and investor-inflicted (!) turnover costs. Hidden causes: Pay-to-play, pricing games and soft dollars

15 Misalignment of Interest? Lakonishok, Shleifer and Vishny (1992) find that DB plans underperform broad market indices and matched mutual funds They partially explain this by pointing at potential agency conflicts between companies, pension treasurers, money management firms and plan participants However, they use broad market benchmarks and fixed cost levels instead of fund-specific information!

16 II. The performance debate

17 The Performance Debate: Mutual Funds Four decades of equity mutual fund research generally show the inability of fund managers to beat the market index Good examples are Malkiel (1995) and Gruber (1996) Some studies find persistence in performance (over short horizons), in most cases so-called icy hands : losers tend to remain losers Carhart (1997) shows that the performance persistence is mainly attributable to simple momentum strategies

18 The Performance Debate: DB Pension Funds Due to data availability (there is no CRSP for pension funds), studies on the performance of DB plans are limited in both number and sophistication. Majority is focused on performance of delegated equity portfolio management instead of performance at total plan level Benchmarks are approximated by broad market indices Costs are assumed to be a fixed number of basis points in virtually all studies Empirical evidence is mixed

19 The Performance Debate: DC Pension Funds Data is scarce for DC plans as well: Virtually no empirical evidence of equity performance at total plan level An exception is a recent study by Elton, Gruber and Blake (2006), who investigate 43 DC plans in the period (!) The average mutual fund offered in the 401(k) plan underperforms the market by 0.31% p.a., but outperforms matched mutual funds not offered in the DC plan. Hence, DC plan sponsors seem to be able to pick the best performing funds for their participants!

20 Contribution of Our Study We compare DB, DC and mutual fund equity performance (and persistence) at various aggregation levels: at the total plan and lower aggregation levels (e.g. active versus passive etc.) The CEM database allows us to incorporate benchmarks and costs in the comparison fairly and accurately We adjust for risk and style deviations, despite the low (annual) data frequency

21 III. Data and research methods

22 Mutual Fund Database We use the CRSP mutual fund database Covers all US mutual funds for our sample period ( ), survivorship bias free Contains fund-specific variables, e.g. expense ratio, NAV, investment style, turnover etc. We select all fund information with the investment objective US Equity We retrieve investment styles and match it with corresponding style benchmarks if necessary (see section 3.1 in the paper) Globefund.com is used for Canada

23 Pension Fund Database CEM benchmarking Inc. (CEM) provided us with data on DB and DC (equity) portfolios Pension funds report gross performance, benchmarks and cost information to CEM CEM database contains info on different fund types (DB, DC) and classifications (active, passive), countries and regions We remove outliers ( > 3 st. deviations) and funds reporting less than 2 years

24 CEM Pension Database 238 DC funds, Canada, US 716 DB funds, Canada, Europe and US DC: DB: Data on pension fund characteristics, e.g.: Total and active returns Fund-specific costs and various cost breakdowns Fund-specific benchmarks Caution: possible selection and selfreporting bias

25 Data on DB Pension Funds

26 CEM Database Structure Level 4 Level 3 All LC SC Act Pas Level 2 Level 1 (DB only) Act LC Pas LC Act SC Pas SC In Act LC In Pas LC In Act SC In Pas SC Ex Act LC Ex Pas LC Ex Act SC Ex Pas SC

27 Net Value Added (NVA) Pension Funds R fund-specific BMR fund-specific C fund-specific Mutual Funds R fund-specific BMR style-specific C fund-specific

28 Standard Analysis Compute time-series mean per fund Cross-sectional average of time series mean (equally weighted) [ Mean ] Cross-sectional standard deviation of time series means [ s.d. ] T-test based on cross-sectional mean and standard deviation [ t-stat ]

29 Persistence Tests Chi-squared Ranking and evaluation period 1 year Split up Winners (W) and Losers (L) in ranking period Compute transition probabilities in evaluation period (WW, WL, LW, LL) Perform Chi-squared test Portfolio-test Ranking and evaluation period 1 year Form deciles (D1 D10) in ranking period Portfolio: D1 - D10 T-test on time series of portfolio in evaluation periods

30 Example Persistence Results (DB) p-value Chi-squared p-value Portfolio test

31 Risk and Style Adjustments a 0 can be interpreted as risk-adjusted NVA

32 IV. Discussion of empirical results

33 DB: Standard Analysis (NVA in bp.) Adding back costs makes GVA positive

34 DC: Standard Analysis (NVA in bp.) Adding back costs makes GVA positive

35 MF: Standard Analysis (NVA in bp.) Adding back costs does not make GVA positive

36 Size Matching Mutual Funds (Q10) ALL Q1... Q9 Q10 DB DC Average Size Equity Holdings (mln.) 0, , , ,79 617,32 Average Total Equity Costs (bp.)

37 Q10: Standard Analysis (NVA in bp.) Adding back costs does not make GVA positive

38 Lower Aggregation Level Results Same conclusion

39 Persistence Results No evidence of persistence in pension fund equity performance Mixed evidence of persistence in mutual fund equity performance (chi-squared test shows significant persistence, portfolio test does not) As a result of yearly return frequency? Or, because equity portfolios at plan level are a combination of institutional mandates?

40 Risk and Style Adjustment: DB

41 Risk and Style Adjustment: DC ALL LC SC Act Pas

42 Risk and Style Adjustment: MF Q10 ALL LC SC Act Pas

43 Discussion of Results Active (net) equity returns of DB and DC pension funds are in line with pre-determined benchmarks Mutual funds (as expected) strongly underperform benchmarks. If we include entry and exit loads, the picture would be even more dramatic. We do not find clear evidence of persistence Risk and style adjustment does not change the conclusion

44 Concluding Comments Economies of scale? Yes, but costs cannot account for the full performance difference Lack of skill? We cannot find persistent skill in equity portfolios of both pension and mutual funds. In most cases, they are managed by the same group of institutional asset managers (see next slide) Misalignment of interest? Agency costs are difficult to quantify, but remain a usual suspect

45 Check this out! correlation DB DC MF DB 1,00 DC 0,80 1,00 MF 0,65 0,77 1,00

46 Future Research? Identify (measurable) proxies for agency costs Extend the analysis to other countries and asset classes More focus on marginal impact of different cost components

47 Dessert: Canada DB (NVA) Canadian pension funds on average (only DB!) outperform the benchmark Canadian mutual funds underperform broad market benchmarks, except for small cap funds. No evidence of persistence in performance Be careful: mutual fund data less reliable!

48 Dessert: Canada DB (NVA)

49 Dessert: Canada MF (NVA)

50 V. Appendix

51 Data on DC Pension Funds

52 Internal versus External?

53 Table 9: Persistence Tests