High Frequency Performance Monitoring

1
High Frequency Performance Monitoring

• Elroy Dimson Andrew Jackson
• London Business School

2
Outline

• Motivation
• Investment agreementFeatures of
  agreementConsultants control chartMonitoring
  process
• Illustrations
• An extreme event Frequent monitoringManager
  termination Probability of a -2s event.
• Analysis of overlapping returns
• Extensions
• Impact on client profitability Non-normal
  returns
• Conclusion

3
Investment Agreement

• Extract from Statement of Investment Principles
• Managers are chosen on the basis of their
  capability to add value through superior
  long-term performance. Managers are expected to
  follow a consistent investment process with risk
  exposure that is stable over time
• Typical Investment Management Agreement
• The objective is to produce a return from
  inception of 1 per annum above the benchmark,
  subject to a minimum time period of three years.
  The return will not fall more than 3 below the
  benchmark in any 12-month interval.

4
Features of Agreement

• Skilled managers (Goetzmann Ibbotson 94)
• No gaming (Khorana 96)
• Gains from good performance (Sirri Tufano 98)
• Penalty from underperformance (Brown Statman
  97)
• Conflict between target (1 pa) and floor (-3)

5
Features of Agreement

• Skilled managers (Goetzmann Ibbotson 94)
• (We no longer have a consensus in industry that
  managers are not skilled. They find that past
  winners do repeat, adjusting for risk,
  survivorship bias and cross-sectional dependence
  between fund returns. Mainly true for the worst
  performers).
• No gaming (Khorana 96)
• (Managers in the bottom decile are four times as
  likely to be replaced as those in the top decile.
  Performance is the period closest to replacement
  month is more significant than earlier
  performance. They theorise that poor performers
  attempt to prevent dismissal by window dressing
  and herding. This leads to higher turnover for
  managers that get dismissed. The evidence shows
  that portfolio turnover is indeed higher for
  managers who are dismissed. They find only weak
  evidence that the manager increases the risk of
  the fund prior to being replaced in an attempt to
  win back losses). Brown Harlow Starks find that
  managers in the middle of the pack by mid-year
  increase the risk of their fund in the second
  half of the year.
• Gains from good performance (Sirri Tufano 98)
• Find that mutual fund inflows are very strongly
  related to strong past performance. However,
  mutual fund outflows are only weakly related to
  poor past performance. The relationship is
  asymmetric
• Penalty from underperformance (Brown Statman
  97)
• This article argues that excessive portfolio
  churning and switching is generated because
  investors cannot distinguish between good luck
  and skill. They have short horizons, and do not
  allow for statistical noise in performance
  properly, so chase recent winners
• Conflict between target (1 pa) and floor (-3)

6
Consultants Control Chart
7
Monitoring Process

• Monitor tracking error (Roll 92)
• Specify minimum acceptable return (Sortino 99)
• Check performance frequently (Brown, Harlow
  Starks 96)
• Compare return relative to tracking error
  (Grinold Kahn 95)
• Interpret performance correctly (Marsh 91)

8
Monitoring Process

• Monitor tracking error (Roll 92)
• Unfortunately asset returns are exceedingly
  noisy, and a long time can elapse before the
  average performance is known and the fund manager
  can be statistically assured that the manager is
  adding (or subtracting) value. This has led many
  sponsors to focus on the volatility of tracking
  error. Fund managers are reviewed at least
  annually based on their performanceif a manager
  has had a run of back luck, say six consecutive
  months of underperformance, the manager is
  questioned as to whether a new manager should be
  chosen, regardless of whether the review period
  is of sufficient length to allow reliable
  assessment of manager quality. (It almost is
  never long enough)
• Specify minimum acceptable return (Sortino 99)
• Discusses Minimum Acceptable Return (MAR). They
  advocate measuring performance relative to the
  MAR for absolute returns, we modify this to
  relative returns.
• Check performance frequently (Brown, Harlow
  Starks 96)
• Brown Harlow Starks find that managers in the
  middle of the pack by mid-year increase the risk
  of their fund in the second half of the year.
• Compare return relative to tracking error
  (Grinold Kahn 95)
• The information ratio is commonly used to compare
  managers that may have different levels of
  tracking error.
• Interpret performance correctly (Marsh 91)
• Marsh(1991) states that it makes sense to monitor
  funds performance on a high frequency basis. He
  states that there is nothing intrinsically wrong
  with high frequency monitoring, only with the use
  that it could potentially be put to by those who
  do not understand how such figures should be
  interpreted.

9
An Extreme Event

• Assume Tracking error 4, MAR -10

10
Intra-Year Monitoring

• Same example Tracking error 4, MAR -10

11
Manager Termination

• Termination Rule Terminate if 2s below target
  over one year

12
Probability of a -2s event
13
Probability of a -zs event
14
Analysing Overlapping Returns

• Analytic solution is difficult due to overlapping
  periods.
• Method 1 Simulation from multivariate normal
  distribution.
• Method 2 Analytic approximation formula.

15
Effect on Client Profitability

• Investment managers incentive
• Assume the termination rule we saw earlier - ie
  fire if rolling one-year return is lt-2 standard
  deviations below target.
• (net fee 40bp, 500m FUM)

16
Extension Non-Normal Returns

• If excess returns are non-normally distributed,
  then effect is magnified.
• Assume t-distribution (5df) to get the following
  results.

17
Conclusion

• Failure to adjust for high frequency performance
  monitoring may lead to costly actions such as
  strategy revisions or manager termination.
• This paper suggests some methods to make
  appropriate adjustments.
• We show there are big incentives for managers to
  educate investors about the implications of
  high-frequency monitoring.
• Full paper is in JPM Winter 2001, and may be
  downloaded from ww.ssrn.com