How should investors forecast earnings growth?

Tuesday, Feb 14 2012 by
How should investors forecast earnings growth

Unless you're a deep value / bargain investor focused on buying £1 worth of assets for 50p, a company's future growth rate is likely to be of some interest to you. Whether you are doing a DCF valuation, or simply trying to figure out what the PEG ratio looks like, forecast growth is a key input to help determine your target price or stock selection criterion. So what are the options for good and reliable forecasting? To get wiser in this area, it's definitely worth taking a close look at Professor Damodaran's "The Dark side of Valuation" (available on Amazon). He's by far the clearest thinker we've come across in this space and, in broad terms, he argues that there are two easy but bad ways to forecast (i.e. relying on the past or other people) and one good but hard way (i.e. forecasting from fundamental determinants).

Can you trust the past?

 The first thing that investors tend to do when forced to consider the future is to look at the past. Unfortunately, this is an approach that is likely to lead nowhere in terms of forecast accuracy for a number of reasons. First, historical growth rates tend to involve considerable "noise" - this is especially true of earnings due to the number of adjustments involved versus sales. In a famous study, Little (1960) coined the term 'Higgledy Piggledy Growth" because he found almost no evidence that firms that grew fast in one period continued to grow fast in the next period. In addition, historic growth measurement is not straightforward - rates may be complicated by the presence of negative earnings, or they may differ greatly depending the period or methodology selected (e.g. an arthmetric average, a geometric average, a log-linear regression model or more sophisticated statistical techniques like time-series models). Even time-series models (despite requiring a lot of historical data) don't seem to be that accurate if you're looking to forecast over longish time-periods. 

Can you trust the management?

If not the past, how about the revenues and earnings forecast provided by the company management? While this practice has the advantage of simpliciity or indeed laziness, it is also flawed. The main issue is bias - clearly, company management are not going to be neutral about the company’s future prospects and their own skill. Management forecasts may represent wish lists, or alternatively management may play down expectations if their compensation is tied to beating the forecasts provided. Finally, management forecasts can be inconsistent (e.g. by assuming revenue growth of 10% a year for the next 10 years but with little or no new capex over the period). While there is clearly useful information in these management estimates, it needs to handled with care and an independent perspective that checks for internal consistency. 

Can you trust the analysts?

So if neither of those approaches work, what about trusting analysts that follow the firm to come up with the right growth estimate? Unfortunately, the evidence suggests that analysts are hopeless forecasters, especially over the long-term. Analysts often make significant errors, either by relying on the past (see above), using erroneous or misleading data sources or by overlooking significant shifts in a firm's prospects. Furthermore, when managers provide forecasts, analysts usually jump at the opportunity by arguing that managers know more about the company than they do without recognising the bias issues mentioned above. 

David Dreman discusses the appalling track record of analyst forecasting in his book, Contrarian Investment Strategies. More recent work by James Montier found that:

In the US, the average 24-month forecast error is 93%, and the average 12-month forecast error is 47% over the period 2001-2006. The data for Europe are no less disconcerting. The average 24-month forecast error is 95%, and the average 12-month forecast error is 43%. To put it mildly, analysts don’t have a clue about future earnings.

According to Damodaran, the general consensus is that analyst forecasts are better than extrapolating from historical growth over the very short-term (up to 1 year) but not over longer time frames. He suggests several factors/questions that can help determine the weight assigned to analyst forecasts:

  1. How much recent firm-specific information is there? Have there been significant changes in management or business conditions in the recent past, for example, a restructuring?
  2. How many analysts follow the stocks? Generally speaking, the more there are, the better informative the consensus (although there is also a risk of "herding")
  3. What much disagreement is there? Because of this herding phenomenon, the extent of disagreement between analysts, for example measured by the standard deviation in growth predictions, is also a useful measure of the reliability of the consensus forecasts.
  4. Quality of analysts following the stock (although this can be difficult to judge in many cases).

The Answer: Listen to the Business Fundamentals

Given the inherent limitations of these approaches, Damodaran suggests another interesting approach which leverages the Sustaintable Growth Model. Rather than treating growth as being divorced  from  the  operating  details  of  the  firm, he argues that it's better to make it "endogenous", i.e., to make it a function of how much a firm reinvests for future growth and the quality of its reinvestment. 

He points out that the expected growth in operating earnings for a stable firm should be a product of the reinvestment rate (the proportion of after-tax operating income that is reinvested back into new assets) and the return on capital the firm makes on its investments:

Expected growth in EBIT = Reinvestment Rate * ROC.

In the same way, growth in earnings per share or net income is a function of the proportion of net income that is invested back into the business (the retention rate), and the return made on just the equity investment (the return on equity):

Expected growth in Earnings = Reinvestment Rate * Return on Equity.

So if the company’s retention rate is 20% and its return on stockholders’ equity is projected to be 25%, then its growth for the coming year should be 5%.

He argues that forecasting this way is useful at two levels: firstly, it recognises that growth is never costless, i.e. to grow faster, you have to reinvest more, which means less is available for dividends. Secondly, it allows you to distinguish between growth that creates value vs. that which destroys value. Of course, this approach then just creates the problem of having to estimate a company's reinvestment rates and marginal returns on equity and/or capital in the future. While this is difficult to do, the point is  - how can growth-based valuation to be meaningful without first thinking hard about the firm's reinvestment policy and its return rates? As Damodaran notes:

"I do not see how any company can be valued without making assumptions about these variables. Note that those who use analyst or historical growth rates are implicitly assuming something about reinvestment rates and returns, but they are either unaware of these assumptions or do not make them explicit".

Don't Forget Mean Reversion

Finally, whether considering growth rates on their own, or the fundamental determinants of growth like ROE, it's critical to factor in mean-reversion. Mean-reversion refers to the fact that above average performance tends not to last. The human tendency is to look at fast-growth firms and wrongly assume that they are inherently fast-growth firms or vice versa - this is an example of “anchoring bias” where we overweigh the significance of recent observations when forming expectations about the future. It also ignores the competitive pressure which is at the heart of the capitalist system.

Even for a stellar company, the growth rate cannot forever be higher than the GDP growth rate. In the US, historic GDP growth has been 5.3% from 1790 to 2010 and 6.6% from 1945 to 2010. Even these numbers may be optimistic if the US economy itself mean-reverts! One piece of research, based on five decades of data, showed that only 10% of large U.S. companies had increased their earnings by 20% for at least five consecutive years, only 3% had grown by 20% for at least 10 years straight, and not a single one had done it for 15 years in a row. According to Montier, most studies show mean reversion in profitability at around 30% to 50% a year...

We will be building some of this thinking on forecasting into Stockopedia Premium (learn more here), so any comments are of course welcome!

Further Reading

Filed Under: Growth Investing, Growth,

About the Author's Investment Research

Stockopedia Features Profile Image Promotional
Stockopedia PRO

Stockopedia PRO is now LIVE - providing 2300 Stock Reports that provide everything that the fundamentally oriented investor needs to know to make actionable decisions. In Q4 2011, the service opened on an invitation only basis, with a full launch in Q1 2012.  Sign up here: more or visit website »


As per our Terms of Use, Stockopedia is a financial news & data site, discussion forum and content aggregator. Our site should be used for educational & informational purposes only. We do not provide investment advice, recommendations or views as to whether an investment or strategy is suited to the investment needs of a specific individual. You should make your own decisions and seek independent professional advice before doing so. Remember: Shares can go down as well as up. Past performance is not a guide to future performance & investors may not get back the amount invested.

Do you like this Post?
6 thumbs up
1 thumb down
Share this post with friends

6 Comments on this Article show/hide all

Odlanier 8th Mar '12 1 of 6


I'm trying to make sense of this, as I don't have a financial background.


Expected growth in Earnings = Reinvestment Rate * Return on Equity.

Do I assume that a company with no dividend has a Reinvestment Rate of 1? In that case Expected growth in Earnings = 1 x Return on Equity (which is essentially saying:

Expected growth in Earnings = Expected growth in Earnings.

And in either case (with or without dividend), wouldn't Return on equity be something that would have to be calculated from historical values anyway(which seems to be what this approach is trying to avoid)?

| Link | Share | 3 replies
Edward Croft 8th Mar '12 2 of 6

In reply to Odlanier, post #1

1. Yes a company with no dividend has a reinvestment rate of 100% of earnings
2. The Return on Equity (ROE) is the Net Profit / Equity (Book Value). So if all profits can be reinvested back into the business then the equity base will grow proportionally by the ROE. If the management can produce the same ROE on that larger capital base then yes the expected growth in earnings would be equal to the ROE.

In practice though many companies with high ROEs and low dividend payouts fall prey to poor management decisions. They tend to be flush with cash and as a result management are prone to indulge in empire building and bad capital allocation decisions.

I'm going to write an article on this, but a great paper on the topic is here by Rob Arnott - I can't recommend it enough ( Basicallly its been shown that companies with lower payout ratios end up having higher earnings growth - which is completely counter intuitive to the equation shown above.

On the note querying the use of the historic ROE - many investors use a 'cyclically adjusted' ROE figure (i.e over 5 or 10 years) rather than using the latest ROE as its more indicative of longer term performance and thus probably more indicative of future returns. You have to estimate earnings growth somehow and historical figures are sometimes the best guide. As has been shown brokers tend to be over-optimistic!

Not sure if that answers the question at all but might help a bit.

Blog: Follow @edcroft on Twitter
| Link | Share
Odlanier 8th Mar '12 3 of 6

Thanks, you've cleared up quite a bit actually.

It's frustrating because I'm trying to learn this stuff and I think I just have ideological issue with making a forecast on the basis of some sort of regression, average, extrapolation etc.

I think I need to continue my research until I'm able to create a model that incorporates factors such as product line, market share, market saturation etc.

Also I would have thought that macro-economic factors such as gdp would be considered.

A little while ago I had a job interview at Bloomberg and they asked me how I would improve one of their publications. I said charts should be marked with 'milestones' and was told that it was already under consideration.

Eg. How could someone ever forcast Apple's earnings growth using a chart that contains figures from both before and after the launch of the ipod. I know the chart will swing upwards and some will weight any averages towards more recent data but if you dont know the nature of the event that caused the upswng its still a stab in the dark.

Thanks for your help and I will read that paper.

| Link | Share | 1 reply
Edward Croft 9th Mar '12 4 of 6

In reply to Odlanier, post #3

Odlanier - pre-iPhone admittedly with historical data you couldn't have predicted its success. But that was a game changing product launch for them and more like startup economics than typical investment economics. But post iPhone launch as soon as you saw the ROE that Apple was generating, suddenly you had some stats to work off. If you look at the recent ROE trend for Apple you can see how even if you were conservative you could predict continued strong earnings growth as the sustained ROE was so strong . If you also marry that with some thought into Apples growing economic moat due to network effects ( and think about their impressive ability to maintain margins then predicting future ROE with historic ROE clearly shouldn't pose issues. You are never going to predict a blowout quarter like the last one which was totally phenomenal but you'll sleep easy at night knowing that your thinking is sound.

The best blog on AAPL is asymco - and they have some tremendous posts such as this one which clearly show how the latest surge in the stock was predictable.

The thing is if you don't work off regressions, averages and extrapolations what have you got? Blind Speculation ? Sure you can add in some macro and sector industry analysis to stock selection but everyone is subject to confirmation bias - only building evidence into models that confirm their bias. Warren Buffett only buys predictable businesses when they are cheap - i.e. almost every stock he buys has a consistent operating history and predictable ROE/ROCE trend - he wants to buy sure fire success and you don't get that by swinging for the fences and predicting the future.

On Information Overload
PS - regarding product lines / market shares / market saturation etc... This may sound like heresy but  I just don't see the benefit unless you are managing a company internally. Its been shown in countless studies that we don't make better investment decisions by increasing the amount of information we have on a stock beyond a handful of metrics. On the contrary we make slightly less accurate decisions as our confidence level goes off the scale when information levels rise. That's a dangerous recipe! Check out James Montier on the subject - - one massive reason NOT to listen to gurus.  It's for these reasons that we've been building the checklist tools on this site as an aid to better decision making and reducing information overload.

Blog: Follow @edcroft on Twitter
| Link | Share
jeavom 5th Sep '12 5 of 6

When forecasting earnings growth I have usually used the lower of: 10 year CAGR earnings, 5 year CAGR earnings, 10 year CAGR sales or 5 year CAGR sales. I also use the above method as an alternative with ROE being chosen as the lower 10 year or 5 year average ROE. This has worked pretty well for me over the past decade, although I don't use forecast earnings for finance or property companies, preferring to use asset projections based on a similar methodology.

I use the broker forecasts to get a feel for "reasonable" price ranges expected and try to buy at the lower end for a bit of extra safety. However, I never use prospective PE for this. I prefer to use price to sales.

I would love to see this earnings growth approach used in the valuation section. I would like to see sales, asset and earnings based fair value stick price calculators being made available. (And possibly a hybrid of these). Ed - is there an email I can drop proposed calcuations for these extra methods? I also have a suggestion on an appropriate investor discount rate that is share specific.


| Link | Share | 1 reply
Murakami 5th Sep '12 6 of 6

In reply to jeavom, post #5

Useful suggestions, thanks - if you hit the Green Feedback button on the right hand side of the screen, that's the best way to give feedback on the data side of the site/application, and we can discuss it with you there in detail.

| Link | Share

What's your view on this article? Log In to Comment Now

You can track all @StockoChat comments via Twitter

About Stockopedia Features

Stockopedia Features

Stockopedia Features covers in-depth stories on strategies, companies and themes that are relevant to online investors. Investing is hard work. We don't try to over-simplify complex concepts - we prefer to try to help you navigate the detail.  more »

Stock Picking Tutorial Centre

Let’s get you setup so you get the most out of our service
Done, Let's add some stocks
Brilliant - You've created a folio! Now let's add some stocks to it.

  • Apple (AAPL)

  • Shell (RDSA)

  • Twitter (TWTR)

  • Volkswagon AG (VOK)

  • McDonalds (MCD)

  • Vodafone (VOD)

  • Barratt Homes (BDEV)

  • Microsoft (MSFT)

  • Tesco (TSCO)
Save and show me my analysis