Active strategies that tilt on dividend yield have suffered mightily during the Covid-19 pandemic. Dividend yield ETFs, for example, have strongly underperformed the broader US market, as investors lost confidence in companies’ ability to pay dividends. At the same time, the active risk of these ETFs has surged.
Not surprisingly, Axioma’s Dividend Yield factor in the US4 model is down over 3% year-to-date, more than two standard deviations below its long-term average. The pandemic crisis precipitated the negative return, but even as stocks have recovered the factor continued to produce a negative return, falling more than 1% so far this quarter.
If the market is, in fact, still worried about further cuts in dividends, we thought it would be instructive to test two scenarios, where companies that once paid strong dividends either (1) reduced or (2) cancelled them, in order to see the potential changes in the risk characteristics of dividend yield ETFs.
In doing so, we revealed a bit of a silver lining for investors going forward. By subjecting dividend yield ETFs to stress-testing scenarios simulating widespread dividend cuts, it resulted in lower levels of active risk for these funds, thereby freeing-up some of the risk budget for other, perhaps more profitable strategies.
The dismal performance of dividend yield ETFs
When interest rates are low, investors typically seek high dividend stocks as a more profitable alternative to low bond yields. This, of course, only works if investors have confidence in the forecasted dividend payouts of those companies. In the current crisis, investors are concerned not only about companies being unable to pay dividends this year due to losses incurred as a result of the lockdowns, but also about not being able to rely on companies’ dividend forecasts for future years. This lack of confidence even led some to price in bankruptcy risk. So, we set out to look at the performance and risk characteristics of popular dividend yield ETFs, to see what we could find.
In this study we examined four dividend yield ETFs:
- FVD: First Trust Value Line Dividend Index Fund
- HDV: iShares Core High Dividend
- SDY: SPDR S&P Dividend ETF
- VYM: Vanguard High Dividend Yield ETF
While all four funds mirrored the market plunge in February-March, none was able to recover to the extent the US market did in the following months. All four ETFs have recorded year-to-date cumulative losses of between 8% and 11%, underperforming the STOXX USA Total Market index by at least 14% as of August 19. HDV was the biggest underperformer, with an active cumulative year-to-date return of close to -17%.
The variance in performance among these ETFs may be attributable to a number of different characteristics in each fund, such as their dividend yield ratios, sector tilts, size tilts, and the number of stocks held, as elaborated in the paper What’s in a Name? In the Case of Smart Beta, It’s Hard to Tell.
Even over a longer period, dividend yield strategies have not had a great run, but at least they recorded positive total returns. Over the three years from 2017 and 2019, the ETFs recorded annualized total returns between 10% and 12%. While their active returns versus the STOXX USA Total Market were negative then (between -2% and -5%), relative performance has since deteriorated, with active losses year-to-date in 2020 that are four-to-five times bigger than their annualized 2017-2019 active returns.
Information ratios tell a similar story, being three-to-five times more negative in 2020, compared with the 2017-2019 period for each ETF, indicating much higher levels of additional losses per unit of increased risk versus the US market.
The Surge in Active Risk
Active risk surged in March of this year and has continued upward ever since. By August 19, active risk had doubled from its level at the end of 2019 for three of the ETFs, and tripled for SDY. The current level of active risk for each ETF is also two to three times the median level of active risk during the prior three years (2017-2019), as measured by Axioma’s US4 medium-horizon fundamental model. We suspect the investors seeking high dividend yield did not expect such high active risk.
A point-in-time risk decomposition (on August 19) revealed that the Dividend Yield factor accounted for little of the risk allocation, i.e., only 8% to 11% of the active risk in each ETF. Instead, the largest proportion of active risk (22% to 30%) came from the funds’ big negative exposure to Medium-Term Momentum. In other words, these funds had large (intended or unintended) negative Momentum tilts, and it was this contrarian bet that drove their active risk up and hurt year-to-date returns.
Ubiquitous Dividend Cuts Would Drive Down Tracking Error
In order to preserve cash and weather this crisis, many dividend-paying companies are being compelled to cut dividends by unprecedented amounts. While some have done so already, investors seem to be concerned about other shoes yet to drop.
To get a sense of the potential impact of further cuts, we ran two stress tests to examine the risk implications for a dividend yield ETF, should 1) all companies in the fund (which were initially selected for their high dividend yield) severely reduce dividend payments, or 2) cancel them[1]. The latter stress test is clearly a worst-case scenario that should provide some insights on how ETF risk would be affected directionally. While we cannot say what their performance might be under these scenarios, we can certainly evaluate the risk impact.
In the first stress test, for each ETF, we set each company’s exposure to the Dividend Yield factor in the US model to be equal to the average exposure in the model universe, keeping everything else unchanged in the risk model. That is, each company is no longer paying a high dividend yield, but rather an average yield.
In the second stress test, for each ETF, we set each company’s exposure to the Dividend Yield factor to be equal to the exposure of companies that typically do not pay dividends, keeping everything else unchanged in the risk model. The “no dividend payment” exposure corresponds to the minimum exposure in the US model universe and it is largely negative.
Obviously, the change in direction of the exposure to the Dividend Yield factor was not surprising, but what stands out is the change in the magnitude of the exposure.
Originally, as we would expect, all four ETFs had positive (and high) active exposures to the Dividend Yield factor in the US model, due to their stated mandate of investing in companies offering high dividend yield. (The STOXX USA Total Market’s exposure to Dividend Yield was close to zero.)
Reducing the dividend payment to an average level (in the first stress test) resulted in an aggregated Dividend Yield active exposure of close to zero. As anticipated, the ETFs were no longer in the “high dividend yield” category, but rather became more like the US market in terms of dividend payments. Still, actual dividend yields were positive and, under this scenario, may be the best that investors can expect.
When canceling dividends completely (as in the second stress test), the aggregate Dividend Yield exposure flipped sign for each ETF, turning from positive to negative, as expected.
However, the magnitude of the active exposure to the Dividend Yield factor decreased significantly because of the asymmetrical nature of dividend yield.
Since exposure was close to zero in the first test, the Dividend Yield style factor added little to the active risk of each ETF, compared with the original contribution to active risk versus the STOXX USA Total Market.
With its negative exposure in the second test, the Dividend Yield factor no longer added to the active risk of the ETFs. In fact, it detracted from the active risk of three of the funds (VYM, SDY and FVD), while for HDV it was very close to zero. In other words, it became a diversifying factor.
Importantly, Medium-Term Momentum remained the major contributor to active risk in both scenarios.
The stress tests also showed a decline in the total predicted risk for each ETF, as estimated by Axioma’s US4 medium-horizon fundamental model, should companies in each fund decide or be forced to reduce or cease paying dividends.
Active risk (i.e., tracking error) in these scenarios also went down for each fund. However, it remained far higher than it was at the beginning of this year.
Conclusions
Dividend yield ETFs have substantially underperformed the market during this crisis—even more than they did over the past three years—while their active risk surged. However, in each fund, it was not the positive exposure to Dividend Yield, but the negative exposure to Medium-Term Momentum that was the main contributor to active risk. This negative exposure to Momentum also drove the year-to-date return down. The negative relationship between (supposedly) high dividend-paying companies and Momentum tells us that investors are very skeptical about dividend payments after this crisis, preferring instead to bet on companies with other, more reliable attributes.
In a scenario where companies in high dividend yield ETFs only pay an average level of dividends, the funds’ exposure to Dividend Yield plunges close to zero, and therefore the factor ends up contributing little to the active risk of an ETF. This scenario seems much more plausible than our more extreme possibility.
Stressing testing an extreme case, where all companies in an ETF would stop paying dividends, we have shown that dividend cancellations would flip the sign of a fund’s exposure to the Dividend Yield style factor from positive to negative, while the magnitude of the exposure would decrease substantially. Dividend Yield would end up reducing the active risk of the fund in this scenario, while currently it has a positive contribution.
Further cuts in dividends have the potential of lowering the tracking error for dividend yield ETFs, although not substantially, as active risk would remain high relative to historic levels even in an extreme scenario where all companies cancel dividends. Still, fund managers with more room in their risk budgets could re-focus their investments on more profitable strategies, should their mandates permit it.
Of course, these ETFs will rebalance at some point, too. If they shift to the “new” high yield companies, that may result in a higher level of risk as well, both from the higher exposure to Dividend Yield, but also because these companies are likely to have even more negative Momentum exposures. After all, those high yields may be the result of plunging prices, accompanied by more questions about the viability of the dividend payment.
[1] Many thanks to our colleague Walter Wang for creating these stress tests and running the detailed analyses.