17.3 The Econometrics of Public Pension Funds
17.3.3 The Data Generating Process and Financial
Collectively speaking the econometric studies discussed above portray a compelling and consistent view of a data generating process underlying the ‘‘production’’ of financial performance. Specifically, the investigations suggest that the ways in which PERS are governed have a direct bearing on investment decisions and these investment decisions in turn affect the financial performance of public pension funds with little direct impact from governance practices.20 Figure 17.1 visually portrays this proposition. Ignoring path c for the moment, investment decisions are depicted as mediating (accounting for) the relationship between governance practices and financial performance. Again, referencing Tables 17.1 and 17.2, this acknowledgment appears to be particularly true for ETI policy and asset allocation decisions with respect to total rate of return; the most frequently used variant of financial performance. Yet, a central question emerges as to how this overall conclusion should be interpreted given that there are both theoretical and practical issues to consider.
Table 17.2 Two Econometric Studies Concerning the Effects of Governance Practices on PERS Investment Decisions, 1992*
Nofsinger (1998) Useem and Hess (2001)
Investment Decision ETI Policy Percent Equities Foreign Investment Tactical Investing Equity Indexing Governance policies
Constitutional restrictions , significant , insignificant , significant , significant Prudent standard , significant
Performance evaluation þ, significant þ, significant þ, insignificant þ, significant Board purview
Board sets allocation þ, insignificant -, insignificant , insignificant þ, insignificant Board directly invests þ, insignificant þ, insignificant , insignificant þ, significant Board composition
Total board members þ, significant þ, significant þ, insignificant , insignificant Percent elected , insignificant , insignificant , insignificant þ, significant þ, insignificant
*Conclusions based on final regressions in Nofsinger’s Table 4 and Useem and Hess’ Table 7 and Table 8.
AnEconometricAssessmentofStateandLocalGovernmentg473
Theoretically, modern investment theory asserts that meaningful dis- criminations between superior and inferior financial performance across units of analysis (e.g., individual investors, institutions, etc.) require the use of risk adjusted performance measures (Haugen, 1997). The reason for this is that total rates of return are more likely to depend on targeted levels of risk and the performance of markets than on ‘‘other’’ characteris- tics (e.g., superior/deficient skills, knowledge, governance practices, etc.).
To illustrate, PERS with fewer near term liabilities may be in a position to try and capture higher returns associated with a bullish equity market while those with pending liabilities may not. Somewhat paradoxically, the financial investment literature is replete with theoretically acceptable measures of risk adjusted financial performance such as the Jensen or Treynor Indexes.21However, the lack of appropriate data in all likelihood hinders their employment for the foreseeable future.22
Given these inherent limitations on available information, pension fund researchers have tried to account for risk in a number of ways. Fre- quently the allocation of a system’s portfolio to equities is cited as a con- trol variable for risk (e.g., given that stocks are generally more risky than bonds, cash, etc.). While not shown in Table 17.1, Useem and Mitchell (2000) do include the standard deviation of a system’s annual rate of return over five years (1988–1992) as a risk control. They find a positive yet statistically insignificant coefficient for this variable.
While these exogenous measures are certainly useful and creative econometric constructs a couple of other practical and substantive issues remain. For example, since asset allocation strategy decisions appear to correlate highly with total returns there is a possibility that raw total fund returns may be too noisy for comparative measurement purposes. Indeed, this was part of Ambachtsheer’s (1994) thesis in his analysis of the char- acteristics of 184 public and corporate pension plans during 1990–1993.
If his assertion continues to be correct then the possibility exists that any direct governance practice effects, and/or their statistical significance, may Figure 17.1 Conceptual framework relating governance practices and invest- ment decisions to financial performance.
essentially be hindered by noise. Therefore, the econometric assessment of path c in Figure 17.1 may be problematic.
Beyond this aspect there is also the chance that total return and risk adjusted measures of financial performance, should the latter be calculable, could lead to different findings and subsequent recommendations in terms of public policy and administration. For example, while Table 17.2 indicates that independent investment performance evaluations tend to result in public pensions placing higher percentages of assets in equity investments, and Table 17.1 indicates that marginal increases in equity investments result in positive changes for total rates of returns, neither table indicates that independent performance evaluations result in superior financial performance on a risk adjusted basis.23
So the question becomes ‘‘How can these possibilities be examined with the information that is currently available?’’ The best answer to this ques- tion appears to reside with Ambachshteer (1994) and Nofsinger (1998), who both perform an analysis that compares a pension’s total rate of return against a simple benchmark return that attempts to account for retirement system portfolios with lower, similar, and higher levels of risk. Again, referencing Table 17.1, this result is what Nofsinger (1998) refers to as Abnormal Return.24Note that the absence of results in the third column of the table hints that the relationships between the previously unexamined predictors (listed in the first column) and the abnormal rate of return measure of financial performance are a logical next step in terms of econometrically oriented pension fund investment research.