Long-term incentives: Don't pay for index outperformance

The love for stock returns is not exclusive to shareholders. Boards use them more and more as a performance indicator for executive pay. But there are hidden risks of stock price pay as we have shown last week. Better measure stock returns relative to peers, so called relative Total Shareholder Return, rTSR. But there are even deeper hidden risks with rTSR, as we show here.

Stakeholder: Executives, boards, shareholders

Risk: Retention and Reputation

Management: Use the Rank Method for rTSR in executive pay

Companies use long-term incentive plans (abbreviated "LTI" or "LTIP") to align executives with shareholder interests. These compensation plans are often based on performance criteria where Total Shareholder Return is one of the most popular metric. In other words, the payout from these compensation plans depends on the performance metric Total Shareholder Return, short TSR. TSR is the increase in the share price plus re-invested dividends in relation to the beginning share price.

How well is TSR suited as the basis of LTI payouts? There are risks to avoid.

Total Shareholder Return is too volatile

Anyone following stock prices knows that TSR is volatile. In recessions, stock prices can easily drop by 20% to 30%. In the financial crises, stock markets dropped on average by half. In good times, stock prices can double or triple and sometimes perform even better.

Because of this volatility, TSR is a risky performance metric. That makes TSR target setting for executive compensation a challenge. Where should you set the target for your LTI compensation plan? Is 5% shareholder return reasonable? Or better 8%? 10%?

No matter where you set your target, actual TSR results will be far off in most situations because of volatility. This means that LTI payouts are either at the maximum, the LTI pay cap, or zero. Only in the minority of situations, the payout will be in the sweet spot of 50% to 150% of target pay. That's a problem. You don't want your executive compensation payout oscillate between zero and double the target and only rarely be more reasonable. People will object to such results, both executives and shareholders alike.

Relative TSR - rTSR - is the better metric

Because of the high volatility of TSR, most LTI plans today are based on relative performance measurement, that is a company's TSR relative to it's peers TSR. This metric is often abbreviated as rTSR. Relative TSR has the advantage of neutralizing the ups and downs of the stock market. This makes compensation results more resilient without sacrificing incentive strength. On the contrary, relative performance measurement keeps target outperformance challenging in up-cycles and still reachable in down-cycles, maintaining the incentive to perform in any market situation much better than absolute TSR.

But relative performance measurement is easier said than done as we show below (and more so in future contributions in this newsletter).

How to measure rTSR

Relative performance can be measured by two quite different methods:

1. First, as the difference of the company metric value versus the average or median of the peer company values. This methodology is called the Alpha Method. A weighted average can also be used, as is typically the case when public stock indexes are used as the comparison benchmark.
2. Second, as the rank of the company metric value within the group of peer company values. This methodology is called the Rank Method. While a weighting is mathematically possible, it is very complicated, both in calculation and in understanding. Most companies use unweighted ranks (which happens to be the common practice of performance measurement in sports). To further simplify, percentile ranks are calculated which expresses the percentage of surpassed competitors. Percentile ranks are independent of the number of peers and therefore also work if a peer competitor is delisted or invalid for other reasons.

The two methods of relative performance measurement are explained in the following chart.

This chart nicely illustrates the fundamental differences between the Alpha and the Rank Method.

The Alpha Method

Using the Alpha Method, a company with the metric value of 15% (left vertical axis in the chart) is 6% better than the average (9%) and 8% better than the median (7%). The Alpha Method expresses the relative performance in this case as 6%, respectively 8%, depending on the selected middle value (the mean in former and the median in the latter).

The Rank Method

Using the Rank Method, the same company with the same metric value of 15% has a percentile rank of 81% (right vertical axis in the chart) and thus is better than 81% of the comparison companies. The Rank Method measures relative performance in this case as the 81st percentile rank.

In other words, relative performance for the sample company above can be expressed as 6% better than the average, 7% better than the median or a performance at the 81st percentile rank. If we include the difference to a weighted average, as is the case for stock market indexes, we have four different values for relative Total Shareholder Return for the exact same relative performance. Each of the four methods requires its own target setting process and has different risk profiles which means that high and low values have different probabilities in the different methods.

Which method you should use for your LTI depends on your preference for target setting effort and payout risk.

Alpha versus Rank

The two methods provide fundamentally different results. In the Alpha Method, the performance result for a rTSR metric can take values from -100% to theoretical infinity, whereas in the Rank Method the values can be percentile ranks from 0% to 100%. In other words, the Rank Method has a scale of values from 0% to 100% that is the same for any metric type. rSales, rEBIT, rROCE, and all other relative performance metrics, can only have values of 0% to 100%. This makes performance measurement far easier than in the Alpha Method.

The Alpha Method basically has no upper limit. Also, the value scales are different for each measure. For example, an ROCE metric can fall to less than 0%, while a return on sales metric figure cannot rise above 100%.

The Alpha Method thus requires different targets and different caps and floors (upper and lower payout limits) for each metric, while the Rank Method can be applied in the same way to any metric, even to ESG metrics and metrics measuring strategic projects. The rank result can never exceed 100% (a percentile rank of 100) and never be lower than 0% (a percentile rank of Zero).

Ranks also allow for direct interpretations: A percentile rank of 25 is always a performance better than 25% of all companies or worse than 75% or all companies, for any metric type.

The Alpha Method only allows for limited interpretations: Positive values are typically (not always) good, negative values are typically bad. But "how good" or "how bad" the alpha value actually is, cannot be answered right away. It requires additional knowledge.

By now we know that percentile ranks are far easier to use and require far less target setting effort then alpha values. But the most relevant difference is the risk profile of the two methods.

The Alpha Method has far higher risk

Due to the larger ranges for upper and limits, the Alpha Method is significantly riskier than the Rank Method. This means that extreme payouts are significantly more likely in the Alpha Method than with the Rank Method.

In the risk analysis below, the Alpha Method and the Rank Method were applied on hypothetical 3-year TSR results of approximately 30 peer companies for the company ABB in the past twenty years. In other words, the two methods were applied 30 times on 17 periods of 3 years. Actual TSR values were used for ABB and all its peers. In total, we calculated more than 500 LTI plan payouts for ABB's sector.

The chart above shows the probabilities of these 500 LTI payouts. For each LTI, a "vesting payout multiple in relation to the grant value" was calculated. A vesting payout multiple of 1.0x means that the beneficiary gets a payout of one time grant value.

A vesting payout multiple value of ">0.4x-0.8x" on the horizontal axis in the chart means that vesting was between 40% and 80% of the grant value. The probability of this payout (>0.4x-0.8x) is marked on the vertical axis. In the 500 LTI periods calculated, we observed a vesting from 0.4x to 0.8x (times) grant value in about 7% of cases using the Alpha Method. This is the value of the yellow bar on the left side of the chart. For the Rank Method, on the right side of the chart, the probability of this payout (>0.4x-0.8x) was roughly the same at around 8%.

However, all other probabilities are quite different. From the left chart we see a zero payout (0.0x) has a probability of 23% for the Alpha Method. In the Rank Method, a zero payout happens only in about 2% of all cases (right chart).

The Rank Method is much safer for the beneficiaries. It is also safer for the company's reputation because very high payouts, which can cause reputation problems, are also unlikely. Companies using the Alpha Method in their LTIs have a high risk of almost 35% that the maximum vesting is reached. Companies using the Rank Method reduce this risk again to about 2%.

Both zero pay and maximum pay are substantial risks to all parties involved and thus not in the interest of the company, nor the beneficiaries. That's why these bars are colored red in the graph. The Alpha Method produces risky outcomes in 58% of all cases (23% zero pay and 35% maximum pay). Companies using the Rank Method can reduce this risk to less than 5% (2% plus 2%). It can still happen but it is rare. This means that the Rank Method has significantly more payoffs in the green range, where the payouts are between 0.4x and 1.6x of grant value, which is preferred by the plan participants, the company and the shareholders.

In summary, the Rank Method has more than 75% probability of payouts that are desirable, green bars, while the same green payout range has a probability of less than 30% in the Alpha Method.

How Credit Suisse got the Alpha Method wrong

The Alpha Method is not common practice anymore. But that was not always the case. The former 瑞信 used the Alpha Method for a long-term compensation plan that spanned the five-year duration from 2004 to 2009.

While the exact peer group was not public, it was known that it was small. The handful of relevant peers included UBS, Deutsche and Lehman Brothers. The result was that Credit Suisse had to pay 71 million Swiss Francs to its CEO Brady Dougan in the middle of the 2009 credit crisis and recession. During this LTI period, the Credit Suisse’s stock price barely moved as shown in the graph below as the red line.

What could justify a 71 million Swiss Francs payout based on TSR when the share price didn't increase during that period?

To understand the risks of the Alpha Method, the stock prices of Credit Suisse and 瑞银集团 have been plotted against each other. While the stock price of Credit Suisse was roughly at the same level after this five-year period, the stock price of UBS fell by almost two thirds. This meant that Credit Suisse outperformed in the high triple digit percentages, possibly as much as 183% as calculated for the two banks for 2009. Deutsche performed similar to UBS while Lehman, of course, did far worse. The alpha outperformance of a bankrupt company is mathematically infinite, a meaningless value for the purpose of executive compensation.

It is understandable that a rTSR alpha value of 183% versus UBS must have felt like an extraordinary performance of Credit Suisse at the inception of the LTI plan in 2004 when it was conceived by its board of directors. "Certainly, such a high peer outperformance could be celebrated with high compensation for the Credit Suisse CEO Brady Dougan" must have been the justification behind it. What makes sense at the beginning of the period when Credit Suisse was in a difficult situation compared to UBS, looks rather unreasonable by the end of the period when a flat share price was sufficient to trigger a 71 million Swiss Francs payout.

How the Rank Method could have saved Credit Suisse

To assess the Brady Dougan CHF 71 Mio. LTI payout, we calculated that the Rank Method would have resulted in a performance at the 66th percentile rank over the 2004 - 2009 LTI period (not visible in the chart). Had Credit Suisse used the Rank Method, the reputation damage would have been avoided because no board would ever agree to a 71 million payout for a 66th percentile rank performance.

This is an extraordinary example and used here mainly for illustration purposes, but we must not forget: that it actually happened. It shows that the Alpha Method has serious risks in pay-for-performance contracts because it has an excessively high risk of extremely high or maximum payouts which is a reputation risk and an excessively high risk of low or zero payouts which is a retention risk. Neither are in the interest of the beneficiaries, nor the company or shareholders.

How boards should work with rTSR in LTI plans

Despites its flaws, relative Total Shareholder Return is a relevant metric for long-term executive compensation plans. It's the only financial metric looking towards the future. All other financial metrics only look backwards. For this reason, boards can't avoid it in performance based pay. But they can avoid many of its pitfalls if they follow some basic rules:

1. Never set absolute Total Shareholder Return targets. Always use relative performance measurement to neutralize the impact of economic cycles and changes in market sentiment.
2. Never use the Alpha Method for relative TSR performance measurement. Always use the Rank Method.
3. Always compliment rTSR with other performance metrics to reduce the volatility in the payout results. Ideally, measure these other metrics also relative to peers for the same benefit of neutralizing the impact of economic cycles, maintain retain and motivation in both up and down-cycles and increase the probability of payouts in the desirable "green zone" of the charts above.