Idea in Brief
Total shareholder return (TSR) has become the definitive metric for assessing company performance. But it conflates performance associated with operations and strategy with that arising from cash distributions.
A Better Measure
A new measure for assessing performance, core operating shareholder return (COSR), emphasizes value created through operations and does not penalize or reward managers for their dividend and buyback decisions.
The adoption of COSR promises to bring about an increased emphasis on operational performance, a decreased reliance on financial engineering and cash distributions, and a more credible compensation process.
Total shareholder return (TSR) has become the definitive performance metric for public companies. As executive compensation has shifted over the past two decades away from grants of stock and options that vest with time to grants that vest with performance, TSR has become a critical element of governance and compensation and, therefore, of how firms are managed. TSR is sold as a neutral, market-based measure that captures value creation and can’t be manipulated by managers using accounting maneuvers. Are those claims justified?
Unlike measures such as revenue growth or earnings per share that reflect the past, TSR is based on stock prices and therefore captures investor expectations of what will happen as a result of management decisions. This is its chief attraction. The problem is that TSR conflates performance arising from managers’ strategic and operating decisions with performance associated with cash distributions—namely, dividends and buybacks. Because firms have moved from distributing roughly half of their operating earnings to shareholders (largely through dividends) to distributing nearly all operating earnings today (largely through buybacks) that conflation takes on a new significance.
In the following pages, we explain the flaws of TSR as a measure of strategic and operational performance, demonstrate the scale of the distortions it introduces, and propose a new metric to replace it: core operating shareholder return (COSR). This metric emphasizes the value created through operations and does not penalize or reward managers for their dividend and buyback decisions. At a minimum, COSR can complement TSR by providing a more precise measure of operational performance. We also provide a comprehensive assessment of the buyback revolution by identifying which firms reallocated the most value—both positively and negatively—through buybacks and how much total value buybacks have generated or lost for investors over the past 20 years. The verdict is quite damning.
TSR’s Flawed Assumptions
TSR calculates the return for a buy-and-hold investor over a specified period by considering the effects of stock price changes and dividends. Over one year, for example, a stock that begins at $100, ends at $110, and pays a $2 dividend has a TSR of 12%: (10 + 2) / 100 = 12%. What could be wrong with that?
Calculating TSR over multiple years requires one to make an assumption about what investors do with the dividends they receive. TSR extends the buy-and-hold logic to the dividend payout and assumes that shareholders will reinvest it in the company’s stock: Over a given holding period, TSR equals the stock price gain plus the compounded value of dividends reinvested in the company stock, divided by the starting stock price. Shareholders, however, very rarely reinvest dividends in the company stock; thus the TSR assumption reinforces the effect of the stock’s performance relative to the market in a way that is divorced from reality.
Our analysis suggests that managers time their buybacks poorly, and persistently so, resulting in losses for long-term investors.
Suppose a stock pays a dividend and its subsequent performance is lower than the market average. The stock’s underperformance will be compounded by TSR because it is assumed that the dividends will be reinvested back in the company rather than in a higher-performing alternative. By contrast, the TSR of a market outperformer is inflated by the assumption that shareholders will reinvest in the company’s high-return stock. That assumption, however, runs counter to the actual behavior of shareholders who routinely reallocate dividends to other investments.
The problems with TSR go beyond assumptions about dividend reinvestment. The broader issue is that cash distributions of both kinds—dividends as well as buyback programs—are mischaracterized by TSR as a source of value creation. But buybacks are merely distributions of cash used by firms to purchase their own stock. Although companies that do buybacks hope they will drive the share price higher (partly by signaling management’s conviction that the stock is undervalued), buybacks cannot fundamentally create value. They are more akin to an investment in a security.
Although buybacks don’t directly create value, they can transfer it across shareholders, because share price does not always reflect fundamental value. When a firm conducts a buyback program, shareholders can choose to sell or not. If a firm buys shares when the price is $90 but the fundamental value is $100, the buyback will transfer value from selling shareholders to nonselling shareholders. In contrast, a poorly timed buyback (a firm buys its shares for $110 when the fundamental value is $100) transfers value from nonsellers to selling shareholders.
Whether the timing of a buyback is good or bad can be determined only after the fact: The price of the shares eventually either goes up or goes down relative to the market. TSR conflates those timing effects with operating and strategic decisions. In the process, TSR effectively rewards or punishes managers for their timing of repurchases as well as for their strategic and operating decisions.
Our COSR metric overcomes these issues by cleansing TSRs of these distortions. First, dividends are no longer assumed to be reinvested in company stock. Instead, they are assumed to be reinvested in a broad-market benchmark. This enables our second modification to TSR: Cash used to buy back shares is also assumed to be invested in the broadmarket benchmark. While the first change to TSR simply replaces a faulty assumption about dividend reinvestment with a better one, the second change is more significant. It imagines a counterfactual world where the cash used for buybacks is treated not like an investment in a security but as a dividend payout to shareholders.
The mechanics of this are relatively simple: To calculate COSR for a given holding period, sum up the ending stock price, the compounded value of dividends that have been reinvested in the S&P 500, and the compounded value of cash deployed as buybacks now assumed to be reinvested in the S&P 500. Then divide that sum by the starting stock price. With TSR, all the shareholder value is tied up in how the stock price moves. With COSR, some of the value is tied up in the stock’s movement (start and end price), but the performance of cash distributions is separated from the stock price.
To see how big an impact these changes can have, consider the case of IBM. Its TSR of 132% for the past two decades compares badly with the S&P 500 return of 332%. Simply changing the dividend assumption to assume reinvestment in the S&P 500 index, instead of IBM stock, adds an extra 94% to investors’ returns on their IBM shares. Adjusting for the poor performance of its buybacks contributes a further 237%, yielding a COSR of 463%. In short, the assumptions embedded in the TSR metric obscured the remarkably good performance of IBM’s managers in generating healthy cash flows from their business.
But TSR won’t always look worse than COSR. For example, the TSR for Apple from 2010 to 2020 was an eye-popping 1,233%, but its COSR was 876%. That’s because TSR rewarded Apple for dividends paid during that decade by assuming they were reinvested in Apple’s outperforming stock; TSR also rewarded Apple for its well-timed and massive buybacks. Similarly, a further analysis of IBM’s performance for each of the past two decades demonstrates that while its COSR was quite similar for the two periods, its TSR was only slightly higher than its COSR in the first decade and considerably lower in the second, indicating that although operating performance was comparable, the effects of dividends and buybacks considerably affected its TSR.
The Difference Between TSR and COSR
COSR and TSR differ depending on a firm’s cash distribution policies and its market performance over a given period. Of course, firms adjust their policies and mix them over time, but it is useful to consider their effects separately.
For companies that do not distribute cash in the form of dividends or buybacks, there is no difference between COSR and TSR. This is typically the case for young, high-growth firms.
COSR and TSR numbers will diverge for a dividend-paying firm depending on the relative performance of its stock after the dividend is paid versus a market benchmark. A market outperformer will most likely have a TSR that is higher than its COSR, given the assumption that shareholders reinvest the dividends in the stock. Firms whose stock performance is tracking with the market will see no difference between the two measures. Many dividend payers are more-mature firms with shareholders looking to reinvest dividends in higher-growth firms. For such firms, TSR reinforces their weaker stock performance; thus their COSR numbers tend to be higher.
TSR and COSR in firms that do buybacks diverge for two reasons. First, regardless of whether firms are good at timing their buybacks, COSR will diverge from TSR for the same reasons that they diverge for dividend paying firms—distributed cash under COSR earns the alternative market return rather than the return on a single stock. Second, managers are given credit for their timing ability in executing buybacks in TSR but not in COSR. As a consequence, firms with good timing ability and high-performing stocks will have a TSR that’s higher than its COSR. In contrast, firms with poor timing of buybacks will generally deliver a COSR that’s higher than its TSR—because the COSR ignores timing effects and assumes reinvestment in the higher-performing benchmark portfolio.
At the majority of firms in the S&P 500, COSRs have been higher and less variable than TSRs over the past 20 years across all sectors. This suggests that the quality of management performance in most companies and industries has been obscured by the effects of payout policies, characterized by an enthusiasm for buybacks. (See the exhibit “TSR Understates Operating Performance for the S&P 500.”) And as the exhibit “Identifying the Sources of Bias in TSR” demonstrates, the timing effect is correlated with the dividend effect: Firms with poorly timed buybacks also have depressed TSRs because shareholders are assumed to be reinvesting dividends in the firms’ own poorly performing stocks.
We can dive deeper with this analysis and compare companies within a sector. In the IT industry, some firms (such as IBM) have higher COSRs than TSRs because of poorly timed buyback activity and dividends that are assumed to compound at low rates. Other firms, like Microsoft and Apple, feature higher TSRs than COSRs because of propitious buyback activities and dividends that are assumed to compound at high rates. In this sense, TSR magnifies differences in operational performance across firms, whereas COSR allows operational performance to be cleanly isolated from the compounding effects of distribution policies.
The pharmaceutical industry is a particularly compelling example, given the magnitude of its cash distributions. A few firms—such as Regeneron and Vertex—see little difference between COSR and TSR because of their limited cash distributions. Many others have comparable COSRs but widely divergent TSRs. Consider Gilead and Lilly, whose COSRs are very similar. Gilead’s much lower TSR reflects its poor record of value-destroying buybacks; Lilly has engaged in few buybacks. And although the stock price performance of Lilly was not favorable relative to the stock market over the entire period, extreme outperformance at the end meant that the sizable dividends paid out during those years dramatically inflated its TSR. Biogen comes out as a particularly poor performer with an average annual TSR for the decade that is lower by 3% per year than its COSR, purely because of ill-timed buybacks. Pfizer, a mid-performing firm on the basis of COSR, is a poorly performing firm in terms of TSR.
By comparing TSR and COSR, we can make an overall assessment of the impact of buybacks on returns for long-term investors—a topic that has generated much heated debate over the past decade. Some see buybacks as a way to avoid corporate overinvestment, while others see it as shortchanging long-term investors and even damaging the national interest.
Our analysis looks at buyback performance of S&P 500 firms over the past 20 years. We find that firms did not succeed in timing buybacks well in either decade but that performance deteriorated further in the second decade, suggesting that buybacks have become overused.
The firms most proficient at timing their buybacks have annual TSRs that are 2% to 3% higher than their COSRs, while the worst firms have annual TSRs that are 7% to 9% lower than their COSRs. The firms whose buybacks have performed the best according to our analysis—such as Apple, Microsoft, Nvidia, and Mastercard—have reallocated as much as 30% of their current value to their nonselling shareholders. The firms with the worst-performing buybacks—including ExxonMobil, Wells Fargo, and IBM—have transferred more than $100 billion in value away from their long-term shareholders. The absolute value of the reallocations is staggering. Aggregating across all S&P 500 firms, long-term shareholders lost out on more than $1 trillion in value from 2011 to 2020 and, if Apple’s stunning buybacks were excluded, that loss would be closer to $1.8 trillion.
It’s useful to consider whether buyback performance is associated with luck or skill. Any effort to cleanly tease out one from the other is hazardous, but the persistence of the timing ability of managers offers a hint. If timing of buyback activity is skill based, it would be somewhat persistent. For example, one can look at two periods (2001 to 2010 and 2011 to 2020) and ask what fraction of S&P 500 firms might be expected to have positive timing ability in both decades, in neither decade, or in one decade. If luck were driving the results, one would expect 25% of firms to have positive buyback results in both periods, 25% in neither period, and 50% in only one period. Instead, only 14% of firms have positive buyback ability in both periods, and 44% of firms have negative buyback activity in both periods. We get similar findings when we look at five-year periods. This suggests that managers time their buybacks poorly, and persistently so, resulting in losses for long-term investors.
Is There a Catch with COSR?
TSR has proliferated widely without a serious examination of its underlying assumptions. We should not make the same mistake with COSR. Several possible objections to COSR exist, and we consider them here:
Shouldn’t managers get credit for their timing ability with buybacks?
TSR includes returns that are driven by managers’ ability to time the market (or lack thereof), but it doesn’t shed any light on how important those returns are relative to operating performance. It would be instructive to learn, for example, if managers were simply destroying large fractions of their operating value by making ill-timed buyback decisions. At the very least, calculating COSR in addition to TSR gives you more insight into where your managers’ skills lie and where your company’s value is coming from. Given the evidence on managers’ poor timing ability, using COSR instead of TSR will give managers less incentive to try to do buybacks.
Is COSR’s method of treating buybacks and dividends realistic?
Neither COSR nor TSR is fully realistic. TSR makes the assumption that dividends are reinvested in the stock itself, which is contrary to the real-world practice of investors and the decision of managers to distribute cash out of the firm. COSR assumes that the cash deployed in buybacks is distributed as dividends and invested in a reasonable alternative alongside dividends that are also deployed in that manner. Given that cash distributions are, in fact, distributions out of the firm, we think the reinvestment assumption of COSR is more realistic.
The critical question in choosing between TSR and COSR is what kind of performance should serve as the basis for rewarding or punishing managers. COSR measures the impact of the strategic and operating decisions managers make by explicitly removing buyback timing incentives and the misattribution of returns from dividend reinvestment. It is a superior mechanism for focusing the attention of company leaders on the most important things under their control.
Under COSR, shareholders’ investment of dividends in outside stocks is taken into account. Isn’t it wrong to combine two different assets under one return?
No. Our view is that cash distributions from firms are, in fact, distinct from the firm itself and should be treated as such. The presumptive reinvestment of dividends in the stock is in fact the odd choice given that investors redeploy cash from dividends in the best alternative investment outside the firm. COSR reflects that reality.
What determines the relevant outside investment used in COSR?
The question of what investment to choose can be quite nuanced. One could consider, for instance, making it a portfolio of stocks of industry peers. We have deployed a broad-market index as this is a fairly straightforward solution. We imagine that large shareholders and governance firms will converge on the suitable investment opportunity for any given firm.
. . .
In an era of rising expectations for businesses and managers, it is critically important that the metrics by which we judge corporate performance and set managerial compensation are not distorted by non-value-creating factors and financial engineering. Performance metrics should focus the attention of managers on what matters: core operations. The adoption of COSR promises to bring about an increased emphasis on operational performance, a decreased reliance on financial engineering and cash distributions, and a more credible compensation process.