Presentation on theme: "7 - 0 Second Investment Course – November 2005 Topic Seven: Investment “Tournaments” & Manager Compensation."— Presentation transcript:
7 - 0 Second Investment Course – November 2005 Topic Seven: Investment “Tournaments” & Manager Compensation
7 - 1 Some Background Past studies (e.g., Goetzmann, Greenwald, and Huberman (1992)) have shown that mutual fund investors focus primarily on published rankings of relative performance when making their investment decisions. Other studies (e.g., Sirri and Tufano (1992)) have shown that these allocation decisions are asymmetric in that funds with good relative performance experience net cash inflows while those with poor relative performance do not experience significant outflows. From these facts we suggest that the mutual fund industry can be viewed as a tournament in which all funds with a similar objective compete with one another during the year.
7 - 2 The Research Premise From these facts we suggest that the mutual fund industry can be viewed as a tournament in which all funds with a similar objective compete with one another during the year. This tournament structure, where cash flows into the funds and, ultimately, the manager’s compensation depends on relative performance, can provide incentives for managers to alter the investment characteristics of their portfolios. Specifically, managers of those funds most likely to be “losers” at the end of the tournament will have the incentive to increase the risk of their portfolios more than those managing funds likely to be “winners”. The study titled “Of Tournaments and Temptations: An Analysis of Managerial Incentives in the Mutual Fund Industry,” (by K. Brown, V. Harlow, and L. Starks) was published in Journal of Finance in 1996.
7 - 3 The Economics of Tournaments Many compensation and reward structures can be viewed as tournaments Tournaments are most appropriate in situations where an agent’s effort it not observable and performance of all agents depend on a common economic “shock” - Relative performance measures help separate the agent’s contribution from that due to the state of nature Tournament structures can outperform other reward schemes in mitigating moral hazards - Conditions for this include risk averse participants, a common shock component and a large number of agents Little empirical evidence exists on how tournaments are organized and how they operate
7 - 4 The Central Hypothesis Central Hypothesis: Interim Loser ( / ) > Interim Winner ( / ) where is a risk measure for the first part of the tournament;, the last part of the tournament Secondary Hypothesis: Fund characteristics will affect the incentives and the ability to increase risk. -- Size -- Age -- Marketing channel (load / no-load) 00 11 11 00 00 11
7 - 5 Data for the Study Monthly returns for 334 growth-oriented equity mutual funds from data base maintained by Morningstar for the period 1976 to 1991 A fund is only included if it has return data for the entire year We also updated the sample through 1996, which includes 478 funds
7 - 6 Growth-Oriented Equity Mutual Funds, 1976-1991 Number of Funds
7 - 8 Methodology For a particular year (i.e., tournament) consider the following assessment periods: where month “M” is the interim assessment month Calculate the interim cumulative return (RTN) for the j-th fund as follows: Calculate the Risk Adjustment Ratio (RAR) for each fund as follows: where and are the respective variances computed in the pre- and post-assessment periods RTN jMy = [(1 + r jly )(1 + r j2y )].... (1 + r jMy )] - 1 0M12 Pre-Assessment Period Post-Assessment Period RAR jy = ( ) ( ) MM (12-M) MM
7 - 9 Methodology (cont.) For each year, rank fund sample from highest to lowest by RTN variable. Classify interim “winners” and “losers” by whether they are above or below median value, respectively. For interim winner and loser funds, classify again according to whether RAR is above or below its median value. These classifications lead to a 2 x 2 contingency table: (i) interim winners and losers; and (ii) high or low volatility ratios.
7 - 10 Methodology (cont.) No requirements to specify an appropriate benchmark portfolio Market-timing assessment problems do not arise Mean-variance efficiency of a benchmark is not an issue Survivorship bias is not a problem (works against the central hypothesis) Advantages of Tournament Approach
7 - 11 Developing the Risk Change Hypotheses Null Hypothesis: High Risk Ratio 25.0 % Low Risk Ratio Interim Loser Interim Winner
7 - 12 Developing the Risk Change Hypotheses (cont.) Predicted Alternative Hypothesis: High Risk Ratio >25.0 % <25.0 % Low Risk Ratio Interim Loser Interim Winner
7 - 13 Risk Change Results 1980 - 1991 (2,484 observations) High Risk Ratio 27.7 % 22.4 % 22.2 % Low Risk Ratio Interim Loser Interim Winner (p-value 0.000)
7 - 15 Risk Change Results (cont.) 1989 - 1991 (932 observations) High Risk Ratio 31.2 % 18.8 % Low Risk Ratio Interim Loser Interim Winner (p-value 0.000)
7 - 16 Developing the Secondary Hypotheses Extreme winners and losers - Classify by upper and lower quartiles of RTN Window dressing effects - Analysis with and without December returns “New” and “entrenched” funds Small and large funds Load and no-load funds Influence of cumulative performance - Multi-period tournaments
7 - 17 Extreme Winners and Losers (1980-1991) Base Case (Median Ranks) High Risk Ratio 27.7 % 22.4 % 22.2 % Low Risk Ratio Interim Loser Interim Winner (p-value 0.000) High Risk Ratio 28.3 % 27.9 %22.2 % 21.6 % Low Risk Ratio Interim Loser Interim Winner (p-value 0.000) Extreme Upper and Lower Quartiles
7 - 18 Window Dressing Effects (1980-1991) Without December Returns High Risk Ratio 27.7 % 22.4 % 22.2 % Low Risk Ratio Interim Loser Interim Winner (p-value 0.000) High Risk Ratio 27.8 % 27.7 %22.4 % 22.1 % Low Risk Ratio Interim Loser Interim Winner (p-value 0.000) With December Returns
7 - 19 “New” and “Entrenched” Funds (1980-1991) New Funds High Risk Ratio 29.9 % 28.5 %20.9 % 20.7 % Low Risk Ratio Interim Loser Interim Winner (p-value 0.000) High Risk Ratio 25.8 % 26.9 %23.7 % 23.6 % Low Risk Ratio Interim Loser Interim Winner (p-value 0.000) Entrenched Funds
7 - 20 Small and Large Funds (1980-1991) Small Funds High Risk Ratio 31.3 % 25.2 %18.8 % 24.7 % Low Risk Ratio Interim Loser Interim Winner (p-value 0.000) High Risk Ratio 24.8 % 29.6 %25.3 % 20.3 % Low Risk Ratio Interim Loser Interim Winner (p-value 0.000) Large Funds
7 - 21 Load and No-Load Funds We would expect no-load funds to be more sensitive to performance rankings Simple tests indicate a significant tendency for no-load losers to increase portfolio risk in second part of year However, no-load funds tend to be new funds Controlling for other characteristics, no significant differences found between load and no-load funds
7 - 22 Influence of Cumulative Performance Current and past-year performance important in explaining new fund inflows (Sirri and Tufano (1992)) Viewed as a multi-period game, cumulative performance may be important in influencing portfolio risk changes - Three-year relative performance - Five-year relative performance
7 - 23 Influence of Cumulative Performance (1980-1991) Base Case (1 Year Ranking) High Risk Ratio 27.7% 22.4% 22.2% Low Risk Ratio Interim Loser Interim Winner High Risk Ratio 29.1% 23.4%24.9% 22.7% Low Risk Ratio Interim Loser Interim Winner 1 and 3 Year Rankings
7 - 24 Influence of Cumulative Performance (1980-1991) Base Case (1 Year Ranking) High Risk Ratio 27.7% 22.4% 22.2% Low Risk Ratio Interim Loser Interim Winner High Risk Ratio 31.1% 20.6%26.0% 22.5% Low Risk Ratio Interim Loser Interim Winner 1,3 and 5 Year Rankings
7 - 25 Influence of Cumulative Performance (1980-1991) The relationship between risk adjustments and past performance was investigated using logistic regressions Results indicate that cumulative past performance is almost as important as performance in the current tournament
7 - 26 Important Question #1 Active decision Passive -- loser portfolios are inherently riskier How Do Managers Alter Portfolio Risk?
7 - 27 How Do Managers Alter Portfolio Risk? Through simulation experiments, we assess whether the results could be caused by the increase in risk occurring at the asset level rather than by the portfolio managers’ decisions. Control portfolio samples - 250 simulated portfolios from CRSP database - 75-stock and 150-stock portfolios - Different periods within the 1980-1991 interval Results strongly support active decision - All tests significant at the 0.001 probability level
7 - 28 Important Question #2 Traditional objective classification categories such as ‘growth” are not necessarily accurate indicators of fund style or future performance All funds in the “growth” category may not be viewed by investors as being within the same performance tournament Could Investment Objective Misclassification Cause Spurious Results?
7 - 29 Objective Misclassification In order to assess the potential effects of objective misclassification, volatility-based subtournaments were investigated Subgroups formed within the sample based on: - Systematic risk (beta) - Total risk (volatility) Results suggest misclassification is not a source of the differences between winner and loser portfolios - All tests significant at the 0.021 level or better
7 - 30 Important Question #3 Risk change versus rank change Strategic response by interim winners Are Interim Losers Able to Change Their Ultimate Tournament Standing?
7 - 31 Terminal Return Hypotheses Null Hypothesis: Terminal Loser 50.0 % (less random error) Terminal Winner Interim Loser Interim Winner 0.0 % (plus random error) 50.0 % (less random error) 0.0 % (plus random error)
7 - 32 Terminal Return Hypotheses Predicted Alternative Hypothesis: Terminal Loser < 50.0 % (less random error) Terminal Winner Interim Loser Interim Winner > 0.0 % (plus random error) < 50.0 % (less random error) > 0.0 % (plus random error)
7 - 33 Terminal Return Results Average Spearman Rank Correlation coefficient between the interim and terminal rankings for the twelve annual tournaments was 0.81. Chi-squared tests against the null hypothesis for all tournaments were statistically significant. Logistic regression of the form (Rank Change) = f(RAR) had positive coefficient on RAR and significant at 0.001 level. Typical contingency table: High Risk Ratio 41.0 % 9.0 % Low Risk Ratio Interim Loser Interim Winner
7 - 35 Growth - Oriented Equity Mutual Funds, 1979-1996 Interim Losers Which Increase Risk 1 Year Ranking 3 Year Ranking 5 Year Ranking Null Hypothesis
7 - 36 Conclusions Interim losers alter the volatility of their funds during the latter part of a year to a significantly greater extent than do interim winners. This effect became significantly stronger during the last half of the 1980 - 1991 sample period when the number of new funds in the industry increased dramatically. This tendency existed for all funds but was somewhat more pronounced for newer funds and for smaller funds. Cumulative performance has almost as large an impact on the risk decision as does the interim return in the current tournament. Analysis of a simulated set of unmanaged stock portfolios confirm that the observed risk changes were due to explicit managerial actions. The difference in the interim, post-assessment period, and final annual rankings suggest that the mid-year volatility adjustments on the part of the interim losers did, in part, have the desired effect of increasing their rankings.