How to improve the ISS P4P model

24 October 2017

Shareholding provides the key for linking pay to performance

As I’ve written elsewhere, executive pay is plagued by enduring myths that define the policymaking debate. One of these myths is that there is no link between pay and performance. This is an important accusation; one that brings executive pay, and by extension big business, into disrepute.

There are indeed many problems with how incentive plans are designed and operated. Reform of executive pay design is certainly needed. But most analysis that compares pay with performance is deeply flawed, mainly because it ignores the impact of executives’ existing shareholdings. We analyse this issue in detail in Paying for performance and I set out some of the significant implications in this blog.

Most analysis takes as its starting point the single figure of pay, which broadly represents the amount of pay that is awarded based on performance and service in that year. This single figure, perhaps averaged over several years, is then typically compared with the total shareholder return (TSR) delivered by the company, as either the absolute £ value added, or the % return.

This type of analysis is the basis of CEO “value for money” analysis produced by UK consultants and was the basis of one the most quoted studies of recent times: MSCI’s paper Are CEOs paid for performance? Evaluating the effectiveness of equity incentives, which purported to show a small negative correlation between pay and performance in US companies over the period 2006 to 2015, and which has repeatedly been used to support the proposition that pay for performance is a bad joke. A similar approach underlies the Pay for Performance (P4P) quantitative screening methodology introduced into the UK by ISS, the prominent proxy agency, in 2016.

Unfortunately these analyses, in common with most of this type, suffer from (at least) two fatal flaws, which render their results largely worthless. In the case of ISS this is a concern, as their methodology is increasingly influential as a screening process in their proxy analysis and voting recommendation process.

Flaw 1: the analysis is not size adjusted

One of the best documented relationships in executive pay is that more valuable companies (in terms of market capitalisation) pay more than smaller ones. The rationale for this is simple. We don’t have to believe that it is harder to run a big company - just that because the value stakes are higher it’s worth paying more to get a better CEO. There is a strong theoretical and empirical literature that supports this relationship between company value and pay.

For this reason, it's important to be careful that any conclusions that compare pay with performance are not just picking up a size effect. For example, over the period of the MSCI study, small-cap companies outperformance of large-caps in the US, in percentage TSR terms. Because smaller companies pay less, this will automatically show a negative relationship between pay and performance.

The ISS methodology partially mitigates this as size is one of the factors they use for determining peer groups. But size is measured by revenue in most cases, rather than market capitalisation, which is shown to have the most robust theoretical and empirical relationship with pay levels. And sector plays a bigger role than size in peer selection, meaning that groups can be broad and in some cases the analysed company can be some way from the median of the peer group. Given that a doubling of market capitalisation on average leads to a c. 30% total compensation difference, this can create significant distortions.

Flaw 2: the analysis ignores existing shareholdings

In the field of academic research into executive pay, it is widely recognised that a critically important, and often dominant, part of a CEO’s incentives are the indirect effects from changes in the value of their shareholdings rather than the “flow” measure of compensation vesting in the year. Analysing the performance sensitivity of pay without taking into account shareholdings is like studying investment returns based on dividends but ignoring capital gains. In other words, it makes no sense, and misses a crucial component in understanding the incentives a CEO faces.

The results are profound. In our research we found that using just the single figure of pay resulted in an R-squared coefficient of less than 20% between realised pay rank and TSR rank across the FTSE-100 - a weak correlation. This is a similar conclusion to that reached in the MSCI study. However, adjusting for size and changes in value of existing shareholdings increases the R-squared to nearly 80%, suggesting that over three quarters of the variation in pay rank is explained by performance - a very strong result.

This is an important problem for two of ISS’s tests. The Relative Degree of Alignment test, which compares ranking against the peer group of pay and TSR over three years; and the Pay-TSR Alignment test, which compares the trend of pay with the trend of TSR for the company over five years.

Lessons to be learned

Failing to adjust for size and existing shareholdings would immediately kill the chances of any academic study of pay versus performance getting published in a top quality peer-reviewed journal. The radical differences that arise when the correct analysis is undertaken suggests that the ISS P4P screening is likely to be throwing up deeply flawed results.

Instead, the model should adjust for size and should include the change in value of vested shares and of unvested deferred awards that have already been recognised in the single figure in an earlier year (for example deferred bonuses). The data on company size and shareholdings is available to enable the methodology to be improved. There are debates to be had on the details: should all shareholdings be included or just those up to the shareholding requirement? Should gains be counted as equivalent to losses, or is it the penalty of wealth-at-risk that we should be interested in? But the principle of including impact of shareholdings must be reflected, otherwise a key factor is missed.

BEIS could help by requiring disclosure of the change in value of these vested shares and unvested deferred awards alongside the ten year history of CEO pay. This would improve stakeholder understanding of the extent to which executives are aligned to performance. Few people realise that a 20% share price fall is equivalent to a £2m pay cut for the typical FTSE-100 CEO. It would also remove a bias against restricted stock and deferred bonus awards in the current disclosure regulations: these awards currently appear to be lacking in performance linkage, as they are recognised in the single figure at award rather than vesting.

Even if BEIS do not help in this area (which seems probable given that such disclosure was not mentioned in their response to the Green Paper), companies could voluntarily disclose this item, which would be easy for them to calculate.

There’s a big difference in alignment between two identically paid CEOs, one of whom owns no shares and the other of whom owns 10x salary. Research shows that building large and long-term shareholdings is the best way to align pay with performance over the long term. Executive pay package design should be centred around this objective, disclosure should report it, and ISS’s pay-performance analysis should reflect this truth.

I wrote this article while a Partner at PwC and so reference to ‘we’ and ‘our’ should be taken to refer to PwC

---