ICPM-sponsored research: Agency Costs Measurement. October 2005

Transcription

1 ICPM-sponsored research: Agency Costs Measurement October 2005

2 Agenda Overview of the intended agency costs project Presentation of the project team Presentation and discussion of the main research questions and hypotheses Overview of the current literature Introduction of the CEM data base Feedback from you is appreciated!

3 Project deliverables An ICPM working paper on (1) the main drivers of pension fund performance and (2) a direct comparison of pension fund performance with mutual fund performance To be presented at next year s workshop or colloquium Seek direct exposure in both the academic and professional literature

4 Project team A joint venture of investors, consultants and academics Keith Ambachtsheer (KPA and ICPM) Hubert Lum (CEM) Rob Bauer (ABP and University Maastricht) Rik Frehen (University Maastricht) Roger Otten (University Maastricht)

5 Two barriers to lifetime consumption utility maximization 1. Bounded rationality leads to personal decision dysfunction (human foibles) 2.Principal-agent and information asymmetry problems drive wedge between retirement savers and their money See presentation by Keith Ambachtsheer from the ICPM June Workshop: DB Plans under Siege: Can They Survive?

6 How to address human foibles and agency costs? Keith Ambachtsheer during the last workshop in May (EPPAM): Design an auto-pilot savings and investments process and subsequent conversion of financial capital into life annuities Create single purpose pension co-ops, foster good governance and organization design and build economies of scale But, what can we learn from the past?

7 Possible research questions Did pension funds (DB or DC) perform better or worse than mutual funds ( the alternative )? Did DB plans perform better or worse than DC plans? Which impact do pension fund characteristics have on pension fund performance? Moreover, why?

8 Which performance? Total returns excluding and including costs But, how to integrate the heterogeneity between funds (e.g. mature versus immature funds)? Active returns (net value added) excluding and including costs How to treat the differences in benchmarks?

9 And which data? Data! Data! Data! he (Sherlock Holmes) cried impatiently. I can t make bricks without clay. Arthur Conan Doyle, The Copper Beaches

10 Pension fund characteristics? Examples (of a long list ): Size (related to economies of scale) Costs (in various dimensions) Country Plan type (DB versus DC, Public versus Corporate, 401(k), 403(b) etc.) Pension fund governance Internal versus external management Active versus passive Etc.

11 PF: academic evidence I U.S. Pension fund performance: Lakonishok, Shleifer and Vishny (Brookings Papers, 1992) 769 US DB plans over DB plans underperform S&P 500 by 250 basis points They doubt whether pension fund industry can continue to exist

12 PF: academic evidence II Dutch pension fund performance: Schweitzer (Ph.D. thesis UM, 1999) 235 Dutch DB plans over DB plans under-perform benchmark by by 200 bps. Some correction for asset allocation and risk Excluding ABP and PGGM

13 PF: academic evidence III UK pension fund performance: Blake, Lehman and Timmermann (Journal of Business, 1999) 306 funds over Managed accounts No performance difference after correction for asset allocation and risk

14 CEM Seminar 2003 US/CND pension fund performance: Based on CEM database: Net Value Added (NVA) is 10 bps. NVA is not equally high in all assets (large negative NVA in domestic large cap) Some indication in standard univariate analysis that size and costs matter

15 Summary pension fund evidence Mixed results on (limited) empirical research into Pension Fund Performance At most performance equal to indices after fees Some signs of size and costs to predict performance Performance models used differ substantially

16 What about mutual funds? Overactive US equities, average annual returns, , %) Sources: DALBAR; Bogle Financial Centre S&P 500 Average Fund Average Fund Investor

17 MF: academic evidence I U.S. mutual fund performance: Consensus in this widely researched topic is that US active mutual funds on average underperform their benchmark by 200 bps. (after fees). Relevant corrections for styles (or risk) have to be made Performance is related to size (+) and costs (-) Performance persistence is present, but mainly icy hands.

18 MF: academic evidence II European mutual fund performance: Scarce literature, but recently Otten and Bams (European Financial Management, 2002) studied 5 countries in Europe On average European mutual funds show positive after cost alpha s (i.e. NVA insignificantly different from 0) Small cap funds outperform in all countries. Adding back management fees makes 4 out 5 countries significant out-performers UK: strong short term persistence (7.28%)

19 MF: academic evidence III A brief wrap-up Europe US NVA (a.c.) positive sig. negative NVA (b.c.) sig. positive zero Importance (*) +/- 10% +/- 40% (*) defined as (e.g. mutual funds / mc stock market )

20 Final scorecard Pension Funds Mutual Funds US -/0 - EU -/0 + Note: focus on active returns

21 CEM 2004 pension fund data base 238 DC funds, Canada, US 716 DB funds, Canada, Europe and US DC: DB: Data on pension fund characteristics, e.g.: 1. Total and active returns 2. Costs and various breakdowns 3. Asset allocation and various breakdowns, etc.

22 Project focus I In the first stage, the Project will focus on comparing US and Canadian pension fund equity and bond returns with matched mutual fund equity and bond returns (using the CRSP Mutual Fund data base) Next, for both pension funds and mutual funds we will try to determine the main drivers of performance (total and active returns)

23 Project focus II Project will combine the strength and richness of the CEM data base with knowledge on newly developed tools in econometrics (e.g. panel data models) and finance (e.g. performance evaluation techniques)

24 Why do we need these tools? To deal with several issues in the empirical analysis: Selection bias, survivor bias etc. Multicollinearity Heterogeneity Unbalanced panel (PFs come and go) Furthermore, we would like to study the impact of pension fund characteristics on performance in a multivariate context

25 Looking ahead We will first spend a lot of time to study the data bases (CEM and CRSP) Then we will formulate several hypotheses and first test these using standard methods and subsequently, in more sophisticated panel data models We expect to have our results ready to write down and discuss with you at the ICPM June Workshop

26 And finally, be careful with the data I have, said he (Sherlock Holmes) come to an erroneous conclusion, which shows, my dear Watson, how dangerous it always is to reason from insufficient data. Arthur Conan Doyle, The Speckled Band