Decentralized Investment Management: Evidence from the Pension Fund Industry
David Blake, Allan Timmermann, Ian Tonks and Russ Wermers
Decentralized investment management is widespread throughout the institutional investment industry and, in particular, the pension fund industry. Yet, despite the huge economic importance of this practice, very little is known about the economic motivation for decentralizing or about how fund performance and risk-taking behavior are affected by decentralization.
We used a unique proprietary dataset to study decentralization in investment management in the UK pension fund industry from 1984 to 2004. Over this time period, most pension fund sponsors shifted a) from employing a single balanced manager, who invested across all asset classes, to specialist managers, who specialize mostly within a single asset class, and b) from a single manager (either balanced or specialist) to competing multiple managers (balanced, specialist, multi-asset or combinations thereof) within each asset class.
This secular shift from single balanced managers to multiple specialist managers carries significant decentralization costs. Decentralization involves suboptimal risk-taking at the portfolio level, due to the problem of coordinating different managers through incentive contracts. The hiring of multiple managers also increases total fund management fees. We examined whether these shifts have been rational; that is, whether fund sponsors have experienced increased performance to compensate for the suboptimal diversification.
We first examined whether the performance of specialist mandates is better than that of balanced mandates. We found that, after conditioning on fund manager mandates, specialist managers did generate superior performance, particularly in respect of stock selection or alpha. By contrast, balanced fund managers failed to display any significant stock-selection or market-timing skills, either in the form of strategic or tactical asset allocation skills. There was also evidence of persistence in performance over time by specialists, especially in UK equities, the most important asset class held by UK pension funds.
We then examined the effects on performance and risk-taking from employing multiple managers. We found mild evidence to support the conjectures that competition between multiple managers produces better performance – but this held only in the case of competing specialist managers in UK equities – and that pension fund sponsors react to the coordination problem by controlling risk levels: total pension fund risk (and, in particular, alpha risk) is lower under decentralized investment management.
We found that the switch from balanced to specialist mandates and the switch from single to multiple managers were preceded by poor performance. In the latter case, part of the poor performance was due to the fund becoming too large for a single manager to manage: sponsors properly anticipated diseconomies-of-scale as funds grew larger and added managers with different strategies before performance deteriorates significantly. Interestingly, competition between multiple specialist managers also improves performance, after controlling for size of assets and fund management company-level skill effects.
We also studied changes in risk-taking when moving to decentralized management. Here, we found that sponsors appeared to rationally anticipate the difficulty of coordinating multiple managers by allocating reduced risk budgets to each manager, which helped to compensate for the suboptimal diversification that results from decentralization.
Overall, our findings help to explain both the shift from balanced to specialist managers over the sample period – pension funds benefited from superior performance as a result of the shift – and the shift from single to multiple managers – pension funds benefited from risk reduction and from avoiding fund-level diseconomies-of-scale by employing multiple managers. We interpret these shifts as being rational by pension fund sponsors, despite the greater coordination problems and diversification loss associated with increased decentralization.
Finally, we note that, following the end of our sample period in 2004, further specialization of skills in pension fund management has occurred. For example, the emergence of diversified growth funds which, in addition to the standard asset classes considered in our paper, offer investments in such “alternatives” as private equity, hedge funds, commodities, infrastructure, currencies and emerging market debt. While the objective of such funds is to generate stable absolute returns (e.g., inflation + 5% per year) over an investment cycle (e.g., 5 years) with lower volatility than an all-equity fund (e.g., one-third lower), it is clear that the trend documented in our study of pension funds employing multiple asset managers with specialist knowledge appears to be continuing, even if the specialist managers are employed within the same fund management company.