PsyFi Search

Thursday 30 July 2009

Real Fortune Telling: Dividend Forecast Indexing

Fallible Analysts, Reliable Dividends

Unfortunately investment analysts are just as prone to the standard behavioural biases as the rest of us and are inevitably more affected by incentives designed to promote short-term outperformance. In particular when market conditions become turbulent most earnings forecasts become slightly less useful than an unsinkable submarine.

Dividends, on the other hand, are less volatile than earnings. Even better, analysts' dividend forecasts are much more reliable than their earnings forecasts. So if dividends are a sign of healthy companies then using analyst dividend forecasts as a forward signalling mechanism may offer a way of obtaining market beating returns at little excess risk. Although, as it turns out, whether this works or not depends on whether your country’s legal system was created by a bunch of Vikings masquerading as French nobles rather than the descendents of Julius Caesar.

Earnings Forecasts, Accounts and Other Hopeless Ideas

There’s been a copious amount of research done on the accuracy of earnings forecasts and the resulting analyst recommendations. The findings, broadly, appear to be that analysts as a group aren’t much better at predicting long-term future earnings than your average tarot card wielding mystic. Even so, skepticism must be tempered with the recognition that companies do their best to guide analysts to an accurate forecast because missing targets tends to result in nasty stuff happening to share prices followed in short order by many other things that company executives hold dear: share options, executive bonuses and future remuneration awards, for instance.

As an example, Brown and Higgins looked at this in “Managers’ forecast guidance of analysts: International evidence” and reported that “managers in strong-investor-protection countries are more likely to utilize forecast guidance to avoid negative earnings surprises than managers in weak-investor-protection countries”. Note the caveat on legal systems, which we’ll return to later.

All of this leads us, sadly and inevitably, to the conclusion that companies are wont to massage their figures to come in around the forecast targets. Most modern corporations are simply too complex to forecast anything with pin-point accuracy. However, a set of accounts is simply a snapshot in time and accountants are master number wranglers: even within modern accountancy standards it’s possible to manipulate earnings enough to stay within forecast ranges – most of the time. Of course, taken to the extreme, this is how many of the more dubious financial frauds have been allowed to occur.

Earnings Forecasting – Art or Science?

While these type of practices work quite well in normal times things can get difficult when the economy takes a nasty dive. Those companies who’ve been the more aggressive manipulators of their figures tend to find it impossible to continue to meet earnings forecasts. At about this point analyst forecast accuracy also declines rapidly. Essentially, exactly when you need decent earnings forecasts you instead get the equivalent of a soothsayer with amnesia and no chicken entrails.

Worse than this, analysts tend to be pretty slow to revise earnings at the individual company level even when it’s absolutely obvious that the market’s taking a tumble. Generally analyst estimates lag the market on the way down, then overshoot and miss the rebound. Which looks uncannily like the way most private investors invest.

Although earnings forecasting is a bit more like an art than a science analysts tend to do better when it comes to dividends. There are a number of reasons for this, but the top of the list is that companies have more control over dividend payments: basically they can choose to retain less of their earnings in order to maintain or increase dividends, even as profits are declining.

Dividend Forecast Accuracy

Philip Brown, Janice How and Peter Verhoeven in "The accuracy of analysts' dividend forecasts around the world" have analysed, unsurprisingly, the accuracy of analysts’ dividend forecasts around the world. One of the slightly odder findings of the research is that dividend forecasts are less prone to error in countries like the US and the UK which have common-law based legal systems, as developed under the Normans in medieval England, as opposed to those like France and Germany which have a civil law system based on the Roman model.

Their theory is that the penalties under common-law of breeching company confidentiality and guiding analysts and inside investors directly is much higher; so in countries with such a legal system companies are less likely to directly provide information to restricted subsets of insiders. The companys' alternative is to use dividends as a signalling mechanism, to indicate managers' long-term confidence, or otherwise, in the business. The result is, in countries governed by common-law, that dividends tend to be “sticky”. Note that this research doesn't really cover the US, but, aside from issues with share buy-backs which tend to distort the results, there's no obvious reason it wouldn't apply there, too.

Market Cap Or Dividend Weighted Indexing?

Why is any of this important, you might reasonably ask? Well, as we know, the behavioural bias of investors accompanied by the perverse short-term incentivisation of many of the players in the markets, make prediction of any kind virtually impossible. For the most part the best most of us can do is invest regularly in broad based index trackers.

Yet, as we’ve also seen, while these investment vehicles have huge advantages over mutual funds in terms of lower fees and removing human error they also capture the underperformance of market capitalisation weighted indexes. In essence, overvalued stocks have bigger market caps and undervalued stocks have smaller ones. Over time you’d expect market cap weighted index trackers to underperform the general economy even as they outperform the majority of actively managed portfolios.

Which is why investing based on dividend forecasts is an interesting concept: we know that the majority of long-term investment returns come from dividends – indeed, as has recently been shown, if you take away dividends there’s been no growth from shares for the last quarter century. So focusing on dividend payers seems to be logical and focusing on dividend payers who intend to keep on paying dividends makes even more sense. That short-term analyst dividend forecasts are reasonably accurate is essential.

Dividend Forecast Index Trackers

Based on this it ought to be possible to construct an index tracker weighted on short-term dividend forecasts which will capture the majority of dividends – and therefore the majority of standard market returns. Oddly enough, although there are a lot of funds around looking at weighting indexes by other than market cap there aren’t very many looking at doing so using dividend forecasts.

In fact, the only one I can find is a UK based indexer, The Munro Fund, which does exactly what the theory might suggest, weighting its holdings by dividend forecast. It’s only being going a couple of years, investing into the teeth of the global gale, so it’s too early to judge the effect in practice. However, the idea both captures company and analyst incentives and removes investor behavioural bias: over reasonable periods it’s hard to see how this can’t outperform the indexes.

Of course, it’s not immune to surprise dividend cuts and irrational analyst forecasts but in an imperfect world that’s something we just have to accept. As the fund’s manager, Rob Davies, puts it when comparing the approach to using market cap weighting: they prefer to be roughly right about the future than precisely wrong about the past.

Related Articles: Fundamental Indexing Can't Save You From Aliens, Arnott: 40 Years of Bonds Beating Equities, Jack Bogle and the Bogleheads


  1. I think the IUKD ETF is a dividend-weighted fund. It tracks the FTSE UK Dividend Plus index.

    Now, can anyone tell me why it's done so disastrously during the credit crunch, even relative to other shares?

  2. Because it is weighted on yield there is price component in the portfolio construction process. As the market depresses a share in expectation of a dividend cut the fund maintains, or even increases its position. So it gets caught in a value trap.
    It has an arcane process to calculate weights that is not disclosed.
    The Munro Fund uses gross cash dividends, i.e. the total amount a company is forecast to pay out to construct its portfolio.
    So Vodafone paying out £4.3b next year out of the total of £58b for the FTSE 350 gives it a weight of 7.6%.
    Eliminating the price component is the secret sauce.

  3. I think dividend forecasting is more reliable simple simply because dividends are more sticky.

    It's like the weather. In the absence of any other information, the best way to predict tomorrow's weather is to say "like today's weather"! ;)