4.1 Introduction

When it comes to corporate voting, institutional investors have long been accused of rubber-stamping management's proposals. Faced with mounting pressure to actively monitor portfolio firrns, many investors have resorted to proxy advisors. Proxy advisors provide three main services: they rate the `quality' of portfolio firrns' governance just as credit rating agencies rate the `quality' of debt; they offer a logistics platform via which investors can easily execute votes; and most importantly for our purposes, they issue recommendations on how investors should vote their shares in the hundreds, sometimes thousands, of firrns included in their portfolio.

There is increasing evidence that proxy advisors are influential. Anecdotal evidence suggests that, for example, the voting recommendation by ISS, the largest proxy advisory firrn, is the explanatory factor for why Hewlett-Packard shareholders approved the controversial acquisitin of Compaq in 2002 (Burrows & Park 2002). Empirical evidence suggests that voting recommendations correlate with voting outcomes; one study found that ISS recommendations unfavourable to management are associated with up to 20.6% fewer votes cast in favour of management (Bethel & Gillan 2002). There is also some evidence of a causal relationship between voting recommendations and voting outcomes: studies controlling for underlying factors known to impact voting outcomes, such as firrn performance, still find that proxy advisors have significant influence (Cai et al. 2009).

The apparent influence of proxy advisors, coupled with questions about conflicts of interest and flawed methodologies, has raised concerns among company managers as well as policymakers. Both the US Securities and Exchange Commission (SEC) and the European Commission are contemplating regulating the proxy advisory business.1 Some policymakers have already responded. The NYSE Commission on Corporate Governance, for example, has stated that even if institutional investors obtain outside proxy advice, they remain responsible for ensuring that their votes are cast only after the exercise of thoughtful judgment.2

Just as a low rate of deviation from management's voting recommendations is sometimes interpreted as evidence of limited monitoring activity, so is a low rate of deviation from proxy advisors' voting recommendations. For example, a recent survey finding that institutional investors rarely deviate from proxy voting recommendations was met with disapproval by the Dutch Corporate Governance Monitoring Committee (2010), which apparently interpreted the finding as evidence that investors blindly follow voting recommendations and in response noted that investors ought to ensure that they vote on the basis of their own insight. A low rate of deviation, however, need not be caused by investors blindly following these recommendations. Indeed, empirical studies finding that the voting behaviour of investors is associated with proxy voting recommendations are unable to explain what accounts for this association (see e.g. Cotter et al. 2009; Iliev et al. 2010). Institutional investors, meanwhile, claim that they do verify the accuracy of voting recommendations, in particular those in respect of domestic portfolio firrns (Paape & Lachotzki 2002). They also claim that proxy advisors are far less influential than company directors believe them to be (Australian Institute of Company Directors, 2011).

To resolve this puzzle, we analyze proprietary data made available by four fund managers jointly managing over USD 600bn in assets. We use this data to examine two possible explanations for why institutional investors may be devoting resources to verifying the accuracy of voting recommendations and at the same time rarely end up deviating from those recommendations. One explanation could be that investors, upon receiving voting recommendations, do verify the accuracy of all or at least some of these recommendations yet reach the same conclusion as the proxy advisor most of the time. Another explanation could be that causality runs in the opposite direction: voting recommendations are based on institutional investors' governance preferences as communicated to the proxy advisor in advance, so there is linie need to verify whether these recommendations, once issued, conform to the investor's preferences. Our results offer support for each of these explanations.

The remainder of the paper is organized as follows. Section 4.2 briefly surveys the related literature. Section 4.3 describes our data and methodology. Section 4.4 explores the possibility that investors, upon receiving voting recommendations, verify the accuracy of such recommendations and then reach the same conclusion as their proxy advisor most of the time. As described in further detail below, our data shows that the funds rarely deviate from voting recommendations: on average, they deviate 3.7% of the time. Therefore, if and to the extent the funds conduct a verification exercise, they must reach the same conclusion as the proxy advisor most of the time. The remaining question is, do they actually conduct a verification exercise? That is, do they verify that the proxy advisor has applied the same criteria as the investor would, and that these criteria have been applied correctly in the case at hand? Or do they just follow voting recommendations blindly?

While we cannot measure the resources the funds devote to verifying the accuracy of voting recommendations in absolute terras, we can estimate the resources devoted to verifying recommendations on certain proposals relative to the resources devoted to verifying recommendations on other proposals, by using the funds' propensity to deviate from voting recommendations as a proxy. To use this proxy, we need to make only two basic assumptions: First, that if subsequent to receiving a voting recommendation, the funds conduct no verification exercise at all, they will vote in line with the voting recommendation (i.e., voting according to the recommendation is the default option).3 Second, we need to assume that if the funds do conduct a verification exercise, then, as the amount of resources they devote to this exercise increases and all else remains equal, they become more likely (up to a certain point) to reach a different conclusion than their proxy advisor, and therefore more likely to deviate from the proxy advisor's recommendation.

If these assumptions are correct, then the finding that funds sometimes do deviate from their proxy advisor's voting recommendations would suggest that these funds devote at least some resources to verifying the accuracy of voting recommendations. More interestingly, the finding that funds' propensity to deviate varies across different types of voting decisions would suggest that the funds prioritize when deciding on the allocation of the limited resources available for verifying the accuracy of voting recommendations. We hypothesize that funds devote most resources to voting decisions that potentially have the greatest impact on portfolio performance. Our results are consistent with this hypothesis: we find that the funds deviate from voting recommendations more often when they hold a large stake in the portfolio firm, when the firm performs poorly (and as a result there is an increased need for monitoring) and when the proposal has potentially significant value implications. In section 4.5, we explore the hypothesis that investors receive proxy voting recommendations that already reflect their stated preferences, which would imply a lesser need to verify the accuracy of these voting recommendations. An important channel through which investors can signal preferences to their proxy advisor is by `customizing' the voting policy on the basis of which the proxy advisor issues recommendations to that investor. To determine whether customized voting policies matter, for the two funds in our sample that receive customized voting recommendations, we measure the frequency of differences between recommendations based on the proxy advisors' general voting policy and recommendations based on the customized voting policy. We find that the results for the two funds are quite different, suggesting that as a general matter the degree of customization, and as a result the significance of customized voting policies, may vary across funds.

In section 4.6, we briefly discuss the human resources devoted to corporate governance by the funds in our sample. Section 4.7 discusses policy implications and concludes.