Institutional Investors’ Expectations, Manager Performance, And Fund Flows
University of Oxford, Saïd Business School
University of Connecticut
May 15, 2015
Using survey data we analyze institutional investors’ expectations about the future performance of fund managers and the impact of those expectations on asset allocation decisions. We find that institutional investors allocate funds mainly on the basis of fund managers’ past performance and of investment consultants’ recommendations, but not because they extrapolate their expectations from these. This suggests that institutional investors base their investment decisions on the most defensible variables at their disposal, and supports the existence of agency considerations in their decision making.
Institutional Investor Expectations, Manager Performance, And Fund Flows – Introduction
Academic studies have traditionally found a strong relationship between past performance and investor-directed fund flows (see, for instance, Del Guercio and Tkac, 2002 and Goyal and Wahal, 2008, for institutional investors; or Ippolito, 1992, Chevalier and Ellison, 1997 and Sirri and Tufano, 1998, for individual investors). 2 Such studies have frequently attributed this relationship to an assumption by investors that past performance will persist (see Lynch and Musto, 2003). Berk and Green (2004) provide a slightly different account of this relationship: in their model, good past performance signals an asset manager’s superior ability, but the combination of decreasing returns to size and investors rationally chasing past performance means that no differences in future performance are expected. An alternative explanation for the importance of past performance is that plan sponsors are hijacked by conflicts. Thus, Lakonishok et al. (1992) see the responsiveness of plan sponsors to past performance as possible evidence of agency problems within the sponsors’ organizations. Plan sponsor officials, as fiduciaries, have reasons to value manager characteristics that are easily justified to superiors or a trustee committee. One of the most important of these characteristics is an asset manager’s past performance, which is readily observable by the stakeholders of the plan to whom plan sponsor officials are answerable.
In this paper we provide evidence, for the first time, of the actual expectations of plan sponsors. Using thirteen years of survey data from Greenwich Associates covering plan sponsors with half of the institutional holdings of U.S. equities, we establish a measure of the future performance which plan sponsors expect from their asset managers. We then analyze these expectations as a function of three sets of possible determinants: first, the past performance of the asset managers; second, various non-performance attributes which plan sponsors identify in those asset managers; and third, the recommendations of asset managers by investment consultants. Next we set plan sponsors’ expectations, and the possible drivers of these expectations, against the actual future performance of the asset managers. Finally, we compare plan sponsors’ expectations of asset manager performance, as well as their past performance, consultants’ recommendations, and other factors, with the fund flows in and out of asset managers.
This analysis allows us to test whether the well documented correlation between fund flows and past performance results from investors extrapolating future performance from past performance, or from agency problems, or both. If the correlation between fund flows and past performance results from plan sponsors extrapolating future performance from past performance (as implied by Lynch and Musto, 2003 or Ippolito, 1992), then any influence of past performance on flows should be channeled through its effect on the expectation of future performance and, to the extent that these measures disagree, only expected future performance should matter. Money should not flow to funds with good past performance unless investors expect these funds to do well in the future as well, and as a result only expected future performance, not past performance should be significant in a multivariate regression. On the other hand, suppose that flows respond to past performance for agency reasons, as suggested by Lakonishok et al. (1992), with plan sponsors herding into these measures to avoid or deflect personal/career risk, rather than using past performance to form their expectations of future performance: in this case we should observe flows responding to past performance rather than to expected future performance (and to that extent only past performance should be significant in a multivariate analysis).
See full PDF below.