7 November 2017

Academic insight adds a useful dimension to the executive pay debate

Three academics - Radhakrishnan Gopalan, John Horn, and Todd Milbourn - have published an article called 'Pay Your Executives What They are Really Worth - Comp Targets That Work' in the current issue of Harvard Business Review. They set out four principles for compensation target design. In brief:

1. Use multiple metrics. Because individual metrics can be gamed it is important to create a balance of three to five metrics that reflect different dimensions of business performance.

2. Increase pay-outs at a constant rate, adjusting for risk. (Translation: avoid thresholds, steps, and kinks in pay-out curves). Steps in pay-out curves cause executives to game performance to optimise pay-outs.

3. Reward performance relative to competitors. Internal targets encourage an inwards-looking focus that keeps executives in their comfort zone, and the target setting process can be captured by CEOs resulting in soft goals. Relative measures mitigate these risks.

4. Include non-financial targets. These reflect long-term investments that take time to show up in financial measures, and are often externally benchmarked and harder to manipulate.

Academic insights into executive pay are worth looking at. Executive pay is an area where bogus studies are endemic. As we highlighted in our paper Executive pay in a world of truthiness, myths abound because of selective use of "evidence" that is often based on poorly framed studies As a result, confirmation bias is widespread, and a multitude of studies exists to prove commentators' worst fears about executive pay. What the market demands, the market will surely supply.

Academics are not immune from this, but at their best bring the rigour of large-scale evidence to the table to help sift myth from reality. Of course not everything can be proven by academic studies. Judgement and experience is also required to determine the best approach for a given company. There can be problems separating correlation from causation. But there are three reasons why it's worth taking seriously what Gopalan, Horn, and Milbourn have to say. First, they are basing their conclusion on systematic research evidence, carried out with co-workers Benjamin Bennett and J. Carr Bettis. This work applies carefully designed controls to avoid false conclusions. Second, that research has been published in a peer reviewed academic journal. Third, the Journal of Financial Economics, where it was published, is one of the top journals with a very high rejection rate - the peer review process for this journal provides a powerful process for sorting the wheat from the chaff.

The problems of target-based pay have been well documented. We discussed them in the Purposeful Company Executive Remuneration Report, which I co-authored with Professor Alex Edmans from London Business School. The work by Bennett and co-workers adds to the significant body of evidence that incentive pay works, but sometimes you might wish it didn't - targets can be met in ways that were not intended. They find that basing pay on internal accounting metrics that are directly within the control of management increases the risks of short term behaviour to meet the target (they add to the body of evidence that to meet such targets executives may at times cut R&D or investment, buy back shares, manage accruals - in extreme cases even commit fraud). With such large sums now at stake in executive pay, it is important to be very thoughtful about the target on which pay is based.

Of course, just because there are problems with an approach on occasion doesn't mean that it should be discarded. The situations where targets incentivise improved performance may outweigh those where they cause myopic behaviour. Sometimes cutting costs and investment, or buying back shares, makes sense, if investment opportunities outside the firm are better than those inside it. Given their findings, the recommendations by Gopalan et al make a lot of intuitive sense, working with the grain of current performance-pay practices while managing the risk. But would they work in practice? I have two concerns.

First, their response to the problems with incentive targets is to use more of them. This is indeed what most companies now do. Most UK plcs operate two to four measures in their LTIP and a similar number in the bonus. But while this does inhibit the ability to game the system in aggregate, it comes at the cost of complexity, and executives may just focus on the measures where they have most control.

Second, the idea of using relative performance is well-established but problematic. Comparing financial metrics is fraught with difficulty in terms of comparability of definitions, timing of disclosure, and the fact that absolute levels of performance can be significantly affected by business model (as a simple example, hotel company running a franchise model will have much higher Return on Capital than a company that owns and manages its own hotels so relative performance is difficult to assess). These practical difficulties means that relative total shareholder return (TSR) has been the only relative measure that has achieved widespread adoption.

Relative TSR measures are problematic because investors have systematically ignored the second, very sensible, recommendation of Gopalan et al. Relative TSR measures typically have a threshold at median where 20% or 25% of the award pays out (and nothing below), with the whole award vesting for upper quartile performance. This is a test case of their concerns. Our analysis has shown that across a wide range of industries this construct too often rewards volatility of performance rather than sustained above average performance. Companies that crash around between the upper and lower quartiles do better than those coming consistently a bit above median. To make relative performance work effectively, investors need to allow progressive vesting below median - perhaps on a widened scale where vesting starts at zero for lower quartile with maximum at upper decile. This is likely a step too far for UK investors.

Gopalan et al have overlooked another possible solution: no targets at all. Solving a problem by doing even more of the same thing that caused it could be seen as a strange idea. It's a bit like communists who claim the USSR wouldn't have failed if they'd only done communism properly. The problem with targets is that they are never comprehensive, even if you have lots of them. However, the long-term share price is. Although intangible investments can take a number of years to feed through into the share price, over the long-enough term (five to seven years) the share price captures most information about the performance and prospects of the company, and over the long-term is difficult to game.

Designing remuneration to make executives significant long-term shareholders in the company is a very powerful way of incentivising long-term performance. This is why I think that long-dated stock awards with phased vesting and high shareholding requirements (including beyond retirement) should be part of the picture. Target-based incentives work superbly for some companies and situations, if carefully designed. But there's a place for simpler alternatives like restricted stock or deferred share awards, in the right circumstances. In any event, regardless of how the incentives are designed, they should lead to the build-up of large and long-term holdings of stock.

I've been arguing this for over a decade, and finally the market seems to be moving this way. It's right the shareholders should consider radical changes to design skeptically - it is up to companies to make the case. But at the same time shareholders (and proxy agencies) should be open-minded. Increasingly they are. This is good news, and will help remuneration committees to design packages that, in the words of Gopalan et al, pay their executives what they are really worth.

I wrote this article while a Partner at PwC, and so references to ‘us’ and ‘our’ should be read as referring to PwC