Room for improvement
For some, investing is just a game. Pension investments, however, have such a social impact and so many stakeholders that they must be monitored carefully. Fortunately, pension funds can obtain almost any statistic for their investments, including a performance attribution down to the lowest decision level. But does this mean proper use of performance numbers? And are investment practitioners always dealing ethically with performance data? We first give three examples where monitoring is not yet optimal.
Example 1. Data not matching process
The oldest rule in performance measurement is that performance reports must match the portfolio decision process. Many funds, however, do not track the actual decisions leading to their investment portfolios.
Consider a pension fund managing all portfolios internally. One of the portfolios invests in Asian equities, another in global indirect real estate. Both portfolio managers can invest in Hong Kong property companies. The accounting system, however, does not track which manager bought or sold the shares. The total holding in each property company is reported as a single line. Attribution of decisions to each of the portfolio managers then becomes impossible.
The solution, subdividing the portfolio into baskets for each manager, has many more advantages than just the evaluation of their separate performances. Each basket can be assigned a unique manager, benchmark, track record, risk budget, and target weight.
Example 2. Conflicting benchmarks and guidelines Consider an internal portfolio invested in UK and US equities. The pension fund decides to rebalance all benchmarks every quarter. The benchmark for the manager is defined as 50% MSCI UK + 50% MSCI USA, rebalanced quarterly. The manager also has strict guidelines: each region must be within a 45–55% range. During one of the quarters, the UK equity market outperforms the US equity market by so much, that the 50% portion of the benchmark initially allocated to UK rises to as much as 60% of the benchmark portfolio, bringing the US benchmark weight down to 40%.
The manager believes that the outperformance of the UK will continue, so he wants to keep UK overweight relative to the benchmark. The imposed ranges, however, force the manager to sell UK equities and buy US equities. This means that he cannot implement his preferred strategy but is forced to implement the exact opposite. The manager will face a negative performance if he turns out to be right! The benchmark will then outperform the portfolio, to the detriment of his bonus and reputation.
This problem can be solved by assuring that the guidelines and benchmark are not conflicting, for example by using floating ranges (5% deviation from benchmark at any time). Another solution is to use daily rebalancing (but this may conflict with overall strategy). Either solution will allow both implementation of the preferred strategy and proper evaluation.
Example 3: Overpaying performance fees Performance fees are often based on the portfolio return and benchmark return over a certain period. Outperformance results in a performance fee. If, however, during the period an additional amount was contributed to the portfolio by the client, the outperformance may have originated in the subperiod before the contribution. After the contribution, the manager may even have underperformed the benchmark. This way, a manager may be rewarded for losing pension fund value.
Each subperiod between cash flows should be evaluated separately in order to align the interests of portfolio manager and pension fund.
Case study: regulations fuel unethical behaviour Dutch legislation was passed in 1998 to improve pension market transparency. Part of the regulations is that industry-wide pension funds must disclose their performance annually. A risk-adjusted performance measure is calculated to reflect investment management skill. A fund showing significant lack of skill over a five-year period must allow all companies in the industry to make their own pension arrangements outside the industry-wide fund.
Although the set-up looks promising, the risk adjustment and focus on short-term rather than long-term performance have been criticised. We will not go into these drawbacks, but show how two issues give rise to performance manipulations. We suggest improvements.
Issue 1: Rebalancing frequency Regulation requires pension funds to define, in advance, the weights and indices for each asset class in the benchmark. If a fund specifies that its benchmark consists of 50% equities with index MSCI EMU, and 50% bonds with index EFFAS EU > 1 year, the law considers the benchmark properly specified. Performance professionals know, however, that the rebalancing frequency is missing. Funds could take advantage of this loophole by simulating after year-end the rebalancing frequency that would have given the lowest benchmark return, then choosing this benchmark return for their performance evaluation.
Consider sample funds A, B and C, each using the above 50/50 benchmark. Suppose fund A rebalances annually, fund B quarterly and fund C every month.
The graphs show the weights of the equities portions of the benchmarks for each fund, using index returns for 2001 (not shown). For fund A the weights of equities gradually declined. Funds B and C brought the weights of equities back to 50% at the beginning of each quarter and month, respectively. The bottom graph shows the changes in equities due to rebalancing. The benchmark of fund C ‘purchased’ equities after every month, in exchange for bonds, whenever equities yielded less than bonds. The benchmark of fund B purchased its most significant amount after the low equity returns in the third quarter. Over 2001, the benchmarks achieved returns of –5.97% (A), –5.42% (B) and –6.05% (C). The low return for fund C resulted mainly from the ‘purchases’ of equities during the third quarter, each of which declined further in value in subsequent months. Benchmark B did best, because most equities were purchased at the bottom of the market.
Selecting monthly rebalancing after year-end would allow a fund to overstate its performance by as much as 63 basis points. It is obvious that pension funds must be required to state their rebalancing frequencies as part of the annual benchmark.
Issue 2: Risk adjustment Another part of the regulations is the risk adjustment. The parameters to be used are (in terms of the above example) the benchmark weights of equities and bonds. Current law would allow each of the three funds to use a weight of 50% for each asset class in this risk adjustment. It is obvious that the three funds did not all have a 50/50 allocation to each asset class in 2001. This means that a risk comparison of the three funds would be an apples and oranges comparison. At least one of the funds would be overstating or understating risk adjusted performance.
It is not trivial which weights to use instead, but the best candidate seems to be the average invested capital, the basis of performance attribution using money weighted returns. This would give the result of a 50/50 allocation for fund A, and a 53.7/46.3 split for fund C, the highest weight being for equities. This result seems counterintuitive considering the first three graphs, but is consistent with the additional capital invested in equities, shown in the last. Fund A did not invest additional capital into either asset class during the year.
The performance measurer should not be at the end of the pipeline, but must serve as the professional feedback mechanism in the investment process. He or she should be invited to play a consultant’s role and implement improvements in order to make the process truly consistent and transparent. As long as ethical behaviour is discouraged, regulations and performance standards must be enforced and improved to avoid fraudulent practices.
When the stakes are so high, the rules of the game must be clear! Professionalism, integrity and regulations must help in removing aces left in anyone’s sleeves.
Hans Braker is senior consultant at Aon Investment Consulting in Rotterdam. firstname.lastname@example.org