5 December 2020

ISS voting recommendations are influential. Academic evidence suggests that’s probably not a good thing. Investors should revisit how they use them.

Several people have asked me to do a short version of my blog about the role of ISS, which would be more suitable for RemCo consumption. The academic evidence is detailed and nuanced, and I wanted to make sure I was scrupulously fair to ISS, whom I sometimes think get an unfair rap. I’ve cut a few corners in the academic explanations in this shorter version, so it should be read in conjunction with the longer blog for context.

The case to answer

In the blog I looked at what academic research has to say about four questions (spoiler alert: summary conclusions in brackets):

• Does ISS influence the vote? (Yes)

• Which shareholders do they influence? (Investors who are the least likely to be informed themselves)

• Does ISS make good recommendations? (It seems not in terms of link to shareholder value)

• Does ISS impose a one-size-fits all approach? (Probably yes, although some mixed evidence)

Does ISS influence the vote?

An ISS Against recommendation is typically associated with a 30% to 35% vote against the resolution. But is this correlation or causation? It could be that ISS just correctly anticipates the views of many investors much of the time - the relationship between voting outcomes and ISS recommendations is then simply a correlation.

Addressing this question of correlation versus causation is notoriously difficult as it is generally not possible randomly to assign negative ISS recommendations to a test group of companies. ISS used to have a screening method based on TSR performance that would cause them to adopt greater scrutiny of companies below certain performance thresholds. Malenko and Shen studied the causal impact of ISS voting recommendations by looking at differences in voting outcome for companies falling just below and above the threshold (which is essentially random). They estimated an ISS Against recommendation for a remuneration proposal has a causal impact of 25% points on the vote.

This estimate is in the same ball-park as other findings. Ertimur, Ferri, and Oesch look at the relationship between say on pay voting patterns in 2011 and the nature of the investor base. They estimate a range for ISS’s influence on non-blockholder institutional investors at between around 10% and 34% points for an Against recommendation. In another paper, Iliev and Lowry looked at the voting practices of different types of investor across a range of resolutions over the 2006 to 2010 period and found that more than 25% of investors (NB not votes) tend to automatically follow ISS Against recommendations. When I was at PwC we examined whether these findings could be extended to the UK looking at say on pay votes over 2015-2017. We exploited a difference between the voting practices of the same institutions depending on whether or not they were a top ten investor to infer a 10%-15% causal impact (out of an average 33% level of opposition) when ISS make an Against recommendation on pay resolutions.

In social sciences it is hard to conduct perfect experiments to identify causality. However, looking at a range of studies the evidence strongly points to two conclusions:

• ISS recommendations do have a causal impact on shareholder voting outcomes

• The causal impact of an ISS Against recommendation is material, at 10% to 25% points, and is responsible for around half the average level of negative votes in such cases

So both sides have a point. The evidence seems strong that ISS recommendations shift the dial in a causal manner. On the other hand, companies need to be careful about complaining too much. Very low votes, below 75% to 80% in favour, will normally only occur if some large shareholders, who do their own research, also agree with ISS.

Which shareholders do they influence?

In the PwC study it was found that the majority of ISS’s voting impact came from the tail of a company’s register. This reflects the intuition that a significant part of the impact of ISS comes from a less actively engaged tail of smaller shareholders. Ertimur, Ferri, and Oesch also find that the tendency to follow ISS recommendations is lower when firms have larger holdings, and so have greater incentives to do their own research.

Iliev and Lowry’s analysis looks at these issues in detail for the US market. They analyse the characteristics of investors that do and do not follow ISS recommendations. They identify a spectrum from engaged to disengaged voters. The most engaged voters do not rely on ISS at all. Entirely disengaged voters follow ISS all of the time. They design a model that predicts the extent to which a fund will be engaged, which they find is strongly associated with actual voting behaviour. Their model suggests that funds will invest in their own research if the benefits of bespoke research outweigh the costs. They identify six factors, which they expect to create a positive net benefit from voting: fund family size, fund size, % of fund net assets invested in the firm, % of the firm held by the firm, whether the fund is located in a city with lots of other fund managers, and fund turnover. These are mostly intuitive but see my longer blog for an explanation.

They find that these factors do indeed strongly predict voting behaviour. They find that firms predicted to have a low net-benefit of informed voting are 3x as likely to follow ISS recommendations. By contrast, the most engaged investors will ignore an ISS Against recommendation 78% of the time - given that these investors have the biggest stake in getting the vote right, this raises questions about the value of ISS recommendations.

The available evidence suggests, in line with common sense, that ISS recommendations are most influential with firms that have the least incentive to invest in becoming informed themselves. This doesn’t mean that they are uninformed - the researchers can’t measure that directly. But it seems highly probable. These tends to be firms or funds that are smaller or have a smaller holding in the firm (both relative to the fund size and the firm size).

So the ISS voting recommendations are influential, and most influential with the least informed investors. But are they any good?

Does ISS make good recommendations?

The evidence on the quality of ISS voting recommendations is mixed. I’ve already discussed the fact that Iliev and Lowry find that investors with the greatest incentive to be informed are far less likely to follow ISS Against recommendations. On the face of it this is troubling: investors with the biggest incentive to get it right ignore ISS Against recommendations most of the time. The authors undertake a further analysis of the market reaction to close votes (on the basis that such a vote is hard to predict and so the outcome is new information to the market). When a vote that is supported by the most engaged investors just passes, there is a positive abnormal share price return. Conversely when such a vote just fails, there is a negative abnormal share price return. The authors find no correlation between the ISS recommendation and abnormal returns whether close votes are passed or failed. Taken together these findings suggest that voting behaviour of actively engaged owners is informative in relation to shareholder value whereas ISS recommendations are not.

Ertimur, Ferri, and Oesch address the informativeness of ISS recommendations in two ways. First they look at share price returns around the announcement of ISS recommendations. They find a negative abnormal share price return on announcement of an ISS Against vote. The negative return is higher for companies with no history of say on pay dissent (indicating that the recommendation was less expected). On the face of it, this may indicate that the ISS recommendation is giving significant information about the quality of the compensation plan. However, it could also be that, in line with Iliev and Lowry, engaged investors are be concerned that the ISS recommendation place pressure on the company to make suboptimal changes to compensation.

However, they explore the issue further by examining situations where a company responds to a negative ISS recommendation on a compensation plan by making changes for the next year’s AGM. The authors found no positive share price reaction to the announcement of changes to compensation that were introduced to address ISS’s concerns raised in the prior year. This suggests that investors did not see in value in firms changing compensation designs to satisfy ISS, and indeed the earlier share price fall on announcement of the Against recommendation may have been due to fear of suboptimal changes to appease the proxy advisor.

Larcker, McCall, and Ormazabal analyse the 2011 AGM season, looking at announced changes in firms’ compensation practices in the eight months prior to the proxy statement release date. They find that changes introduced to align with proxy advisor preferences lead to negative abnormal returns whereas other announced changes lead to neither positive nor negative abnormal returns. They conclude that compensation changes desired by proxy advisors such as ISS may come at a net cost to shareholders. They also found that the negative impact of proxy-advisor-aligned compensation changes were largest in firms that exhibited the greatest difference from industry norms across a number of characteristics. The authors’ interpretation is that the costs of following proxy advisor recommendations are greatest in firms where the optimal contract is less likely to conform to a standard model.

A recent unpublished paper by Alberquerque, Carter, and Gallani is more encouraging for ISS. It is important to highlight the caveat that this paper has not been accepted for publication and so has not completed the peer review process. But subject to that caveat, the results are interesting. Their regressions find a statistically significant association between an ISS Against recommendation on pay and subsequent underperformance. This does suggest that the ISS Against recommendation may have informative value about the quality of compensation, which then feeds into performance. However, further analysis shows that this relationship only holds for non-December year ends. For December year ends, which comprise 75% of the total, the ISS recommendation has no predictive power. The authors attribute this to a workload compression issue. In the busy season, more analysis is outsourced to temporary, potentially less expert, resource and the availability of experienced senior managers, who overlay judgement, is constrained.

ISS research may be excellent and extremely useful in providing standardised information to investors with economies of scale. However, the evidence suggests that their voting recommendations are of limited use in promoting shareholder value. The picture is distinctly mixed with most studies showing that ISS voting recommendations do not provide shareholders with useful information. Even in the most favourable study, the recommendations are only found to provide useful information outside of the busy season, and so not for the vast majority of companies.

Does ISS impose a one-size-fits-all approach?

We can turn again to the study by Iliev and Lowry for insight. Studying ‘standard’ shareholder resolutions across a large number of company meetings they find that in the vast majority of cases ISS supports the resolution, against management’s recommendation. These are proposals such as declassifying boards, amending articles and calling special meetings. This suggests, across many of these issues, application of what is tantamount to a one-size-fits-all approach. This does not matter if ISS is simply reflecting the settled consensus of informed investors. However, the most engaged active voters identified by Iliev and Lowry show much greater diversity of approach. These engaged investors supported management over half the time on six out of the 11 issues. Only in two cases - declassification of boards and reducing supermajority vote requirements - did these investors support management less than 20% of the time. This suggests that ISS may adopt one-size-fits all policies that do not reflect the approach of more engaged investors.

Counter-evidence is found in Ertimur, Ferri, and Oesch. This paper focusses on say-on-pay evaluations and finds that a range of potentially controversial pay practices can lead to differing levels of censure from ISS, suggesting that they take company context into account. In most cases these authors found that presence of a specific negative compensation feature would not automatically lead to a negative recommendation with ISS taking into account the overall context and quality of the compensation plan (UK readers who have tried to adopt restricted stock with a discount of less than 50% or alter pro-rating provisions on LTIPs might beg to differ). This finding might reflect the non-binary nature of many of the compensation features identified by these authors.

A recent unpublished paper by Felipe Cabazon identified an increased convergence in CEO pay in the US, with 25% of the divergence in pay structures disappearing over the ten years to 2016. It also finds that firms are gradually converging towards a pattern of pay reflecting ISS recommendations. This paper must be subject to the same caveat that applies to the other unpublished paper referenced above: it has not been subject to peer review and so its results must be viewed with caution at this stage. However, while it does not establish causality, it does add further weight to the view that proxy advisor evaluation frameworks may lead to one-size-fits-all approaches, which certainly accords with the experience of many practitioners.

Is voting all it’s cracked up to be?

The academic evidence on ISS isn’t all one way. But it does suggest that:

• ISS has a causal impact on voting outcomes, probably amounting to 10% to 25% points for an Against recommendation

• This impact primarily comes through ISS’s influence on investors who have less incentive to do their own research and who are presumably less informed

• ISS recommendations differ markedly from those of the most engaged investors and their Against recommendations aren’t that good at identifying bad compensation practices, as measured by impact on shareholder value - indeed amending compensation to meet ISS concerns may be harmful to shareholder value

• ISS recommendations contribute to an increasingly one-size-fits all environment for governance and compensation

So ISS voting recommendations have impact, are most influential with less informed investors, are of questionable value, and contribute to one-size-fits-all governance. This all adds up to a real problem with voting recommendations and the way they are used by investors.

If this sounds like a hatchet job on ISS, that’s really not what’s intended. My own experience as a practitioner was that ISS produces excellent, and generally accurate, research in almost impossibly short timeframes. It isn’t surprising that many investors subscribe for this service to obtain economies of scale and prevent duplication of standard data gathering in a compressed AGM season. The problem comes with the voting recommendations and how they are used.

How do we get out of this bind?

For sure there’s more that proxy agencies and investors could do (see the longer blog for some ideas and more in-depth discussion) but perhaps the biggest problem of all is in attitudes to voting. Voting in all circumstances has come to be seen as a moral, or in some places a legal, obligation and the sign of a good steward. But good stewardship comes in many shapes and sizes and need not always involve voting. Mandating voting (whether legally or in practice) inevitably creates circumstances where the cost of becoming an informed voter outweighs the benefits to the investor. Voting recommendations fill the gap. But as we’ve seen it may be better for such investors not to vote at all.

Perhaps the emphasis in stewardship codes could shift from voting to informed voting, with investors explaining the circumstances under which they are equipped to cast their vote responsibly and with knowledge, and only voting in those cases. The evidence presented here suggests that no vote may be better than a vote cast in an unthinking manner in line with an ISS voting recommendation. Further nudges could be given by discouraging robo-voting, the automatic population of proxy voting records using ISS recommendations without any intervening input from the asset manager. Disclosure requirements could shine a light on the prevalence of this practice. But as long as voting is deemed to be compulsory, the market for voting recommendations will remain.

Investors should be accountable for deciding how to discharge their voting rights. If there’s a perceived problem with how they are doing that, then it seems preferable to take the issue up with the principal rather than with a mere service provider. But perhaps the main problem is the perception that voting is always a good thing. Informed voting, yes. But it’s better not to vote than to vote in ignorance.

This blog is a summary of my longer piece here, which contains a fuller description of the academic evidence.