The Trouble with Dividend Forecasting

Predicting the future of dividends in a volatile market can be challenging. When so many forecasts are based on inaccurate or incomplete information, it’s nearly impossible. Rather than trying to center an investment strategy around such unpredictable numbers, investors would be wise to diversify their portfolios at both the sector and security levels.

Copious amounts of ink have been spilled on the long-term benefits of dividend investing. While it’s easy to determine which companies have paid consistent dividends, what’s more enticing is to know which companies will continue to produce and grow their dividends. The inherent problem with any forecast or prediction is that, at their core, they’re just guesses. 

Nevertheless, forecasting can still be valid, especially when we have loads of historical data. Dividends are no exception but do come with some unique wrinkles. Let’s dissect some of the nuances of dividend forecasting and discuss some unique challenges.

Consider the source

First, let’s consider what drives dividend forecasts. Readily available dividend forecasts come from sell-side analysts, typically deep subject matter experts on a set of companies in a given sector. They create detailed financial models, including forecasts for items like earnings, free cash flow, and dividends. In addition, financial data providers like Bloomberg and FactSet aggregate these results, which are then used by other financial professionals when making investment decisions. 

However, dividends often get short shrift when it comes to forecasting, which you can see with one look at S&P 500^® constituents. These companies tend to be widely covered, meaning they have estimates from multiple analysts. Yet, even here, far fewer companies have dividend estimates than earnings estimates, and the number of estimates per company tends to be lower.

The table below summarizes the constituents of the S&P 500^® and whether they pay dividends and receive published forecasts. These estimates are aggregated into a single, unified number commonly called consensus estimates. Some lightly followed securities may have only a handful of analysts that produce dividend estimates; in the case of quarterly numbers, it may be just one or two predictions. Moreover, around half (55.4%) of sell-side analysts publish an annual dividend estimate, and only 26.6% are on record with quarterly projections. Suffice it to say, for many stocks, dividend forecasts can be slim pickings.

Sources: FactSet, Parametric, 10/31/2021. For illustrative purposes. Not a recommendation to buy or sell any security. Past performance is not indicative of future results. All investments are subject to risks, including the risk of loss.

Dividend payers

As of October 31, 2021, 394 companies in the S&P 500^® paid a regular quarterly dividend. Interestingly, nearly all companies (495) have a consensus estimate annual dividend, and most (461) have quarterly estimates. This means some analysts publish dividend estimates of zero. Take the newer, non-dividend-paying index member Tesla as an example. Of the 35 analysts who follow the stock, 11 publish an annual dividend estimate, and three even publish quarterly estimates—all zeros across the board. Forecasting dividends is easy if the company doesn’t pay one. Unfortunately, that’s where the simplicity ends. Next, let’s look at the alignment between predictions and reality.

Without a doubt, there is noticeable variability when comparing actual results with forward-looking estimates. The chart below shows the surprise factor, or how different the published results are from consensus estimates. Two points of order before dissecting the numbers: First, the numbers are absolute values since we’re not concerned about over- or underestimating dividends, just the delta; second, the chart excludes companies that don’t pay dividends, including those that suspend a payout. The latter parameter helps us filter outliers (100% cuts) that skew the data even more than negatives.

Sources: FactSet, Parametric, 10/31/21. For illustrative purposes. Not a recommendation to buy or sell any security. Past performance is not indicative of future results. All investments are subject to risks, including the risk of loss.

What are the results telling us? On an average annual basis, dividend estimates for S&P 500^® constituents are typically off by 6% to 7%. Approximately 11% are off by 10% or more. Not surprisingly, the quarterly numbers exhibit far more variability. This is because if the consensus estimate doesn’t factor in a dividend payout change, the surprise factor is more pronounced. That point rings true when we compare the most significant surprises on annual and quarterly bases. Of course, the other three quarters can buffer some of the outsize surprises a meaningful dividend raise or cut can produce, but on a quarterly basis, everything is laid bare.

Why do estimates and actuals vary to such a degree? One reason is that some analysts don’t forecast dividends at all. Unlike other financial items they analyze and model carefully, many analysts use the current dividend payment as a future estimate and revise it only when a company makes a change. More detailed and dividend-focused analysts may elect to factor growth into their forward-looking numbers. Whether it’s applying a company’s historical growth rate, market growth rate, or even inflation, they use a factor to manipulate the current dividend and predict a future one more accurately. Last, some analysts go even deeper with their predictions. Unfortunately, there are nearly 4,500 individual annual estimates (approximately 3,900 excluding zero-payers) across the entire S&P 500^®. Looking at each individual estimate and model would be a herculean task. In the end, we’re left with a handful of decent or pretty good predictions using simple methods; the remaining estimates tend to vary significantly.

Last year provided a stress test for companies as the COVID-19 pandemic upended life as we knew it. The impact on dividends was also pronounced. One example of such hardship was Marriott International (MAR). The hotel chain was among the first wave of dividend cutters on March 18, 2020, coinciding with businesses around the globe that shut their doors to minimize virus spread. At that time, 20 analysts covered the company, and 18 published an annual dividend estimate (two-thirds with quarterly dividends). At quarter end, not a single analyst had cut their forecasted dividends according to FactSet. Think about that for a moment. It is striking that consensus estimates didn’t reflect this change, even after Marriott publicly announced a dividend cut. By the second half of March, 13 S&P 500^® companies had suspended their dividend payments, and of those 13, all retained positive consensus estimates. These aren’t lightly followed companies either; the average annual dividend estimates was 12, with around seven analysts issuing quarterly forecasts. If the professional financial prognosticators can get estimates wrong, even after definitive company announcements, what chances do the rest of us have to get them right?

Our point here isn’t to bash our friends on the sell side or to demean dividend forecasting. Instead, we mean this as a cautionary tale so that investors can better understand the limitations that exist when attempting to provide forward-looking estimates. The market is unpredictable, and dividends are no exception. A strong track record can be a meaningful predictor of future results, but that road has pitfalls even for companies with a long dividend history. So how can investors allocate capital to dividend-paying stocks without the consternation of trying to predict changes? The straightforward answer is diversification. Rather than cherry-picking favorites based on imprecise forecasts, investing in a pool of diverse securities drastically reduces idiosyncratic risk and provides less variability of returns. Fortunately, Parametric offers transparent, rules-based portfolio solutions that offer diversification at the sector and security levels while helping investors avoid the pitfalls of predictive market dynamics.

References to specific securities and their issuers are for illustrative purposes only and are not intended to be, and should not be interpreted as, a recommendation to purchase or sell such securities.  It should not be assumed that any of the securities referenced will be profitable in the future or will equal their past performance. All investments are subject to risks, including the risk of loss. 

S&P Dow Jones Indices are a product of S&P Dow Jones Indices LLC (“S&P DJI”) and have been licensed for use.  S&P^® and S&P 500^® are registered trademarks of S&P DJI; Dow Jones^® is a registered trademark of Dow Jones Trademark Holdings LLC (“Dow Jones”); S&P DJI, Dow Jones, and their respective affiliates do not sponsor, endorse, sell, or promote the strategy(s) described herein, will not have any liability with respect thereto, and do not have any liability for any errors, omissions, or interruptions of the S&P Dow Jones Indices.