Forecast EPS does not come from the company, although they may provide guidance. It is the average of the earnings expected to be declared for the current accounting period, as forecast by the analysts who cover the company. For a large company in the FTSE 100 this will mean a dozen or more analysts’ forecasts being available at any one time.
Since we are interested in the returns to be had from the company in the future, not the recent past, why do we not always use forecast EPS and forget about historic EPS? The major problem here is that forecast EPS will usually be quoted only if three or more analysts follow the company. Forecast earnings are thus available only for the few hundred largest UK companies – effectively, members of the FTSE-350 Index plus a few others. Of the roughly 1,300 trading companies quoted in London, there are hundreds with a market capitalisation of less than £50m. These are unlikely to have even three analysts covering them, particularly if they are quoted on AIM. Such companies have no forecast earnings and thus no prospective P/E figure.
Because of this limited coverage offered by forecast earnings, academic research to back-test particular trading rules invariably uses historical EPS, not forecast EPS. (The fact that you have to pay for a higher DataStream subscription level to get the analysts’ forecasts, and few universities can afford to do so, may also be relevant!)
Problems with earnings figures
There is a significant problem near the top of the profit and loss account. Operating income, being one large number (sales) minus another large number (costs), may be highly variable. Depending on the industry and business model, perfectly healthy companies may nevertheless have wafer-thin profit margins.
For example, Inchcape import and distribute thousands of cars every year. Given the huge volume of metal moving through their distribution channels, a 1% or 2% profit margin is very healthy. However a small percentage change in either sales or costs leads to a very large change in the operating income. Forecasting earnings figures for such companies is therefore a particularly error-prone business. Forecast earnings depend acutely on the assumptions made about how gross sales and total costs will change in the future.
Another sticking point when calculating earnings is that sales in any one year, and hence earnings, are easily manipulated. It was mentioned above that what is classified as an exceptional cost, and thus to be excluded from adjusted EPS, is a matter of opinion. The company’s managers are likely to take a more liberal view of what costs are exceptional than a potential investor. Some large companies seem to need to close down an unprofitable division or operations in some particular county every year, but they still book it as exceptional. When this seems to happen year after year, one is entitled to ask how exceptional it really is.
A whole branch of academic finance literature exists on the manipulation of earnings and how it may be identified. The dividing line between current managers wanting to present their actions in as positive a light as possible, and fraud, is not necessarily easy to draw. Any investor who follows the market for any length of time will have read about scandals of companies booking sales before they are absolutely definite, so as to enhance the apparent performance of the sales managers responsible and thus their bonuses. A recent well-publicised case is Findel, who run Kleeneze among other businesses. In 2010 their educational supplies division was found to have been making “unsubstantiated accounting entries”. Earnings had to be restated, the shares of the Group fell precipitately and have not yet recovered.
One final problem is the practical indeterminacy of EPS figures. Trying to pin down the precise basis of any particular basic, diluted, adjusted, forecast, rolling or historical EPS figure offered you is a real can of worms. Three different financial data sources, such as DataStream, the Financial Times and Hemmington Scott’s Company REFS will likely provide three different figures for EPS. However closely you read the online help files, unless you work as one of their data analysts and are able to read the computer code you will probably never know what exactly they include and exclude in their EPS figure. Even when I have sat down with a copy of an annual report and tried to work out exactly where an EPS figure came from, I have usually been unsuccessful. However, as long as you are using the same definition for all the companies you are considering investing in, it is unlikely to be a decisive factor.
Chapter 3. The Price-Earnings Ratio (P/E)
Having covered earnings in considerable detail, we are finally able to use the earnings figure arrived at as one of the two inputs into the P/E calculation.
The other input is of course the share price. At the level of detail required for the P/E, the price is quickly dealt with. The figure used in the quoted P/E statistics is invariably the previous day’s closing price, so unlike the EPS figure it will not vary from source to source. However, as with EPS figures there is still a certain amount of indeterminacy involved: bid, mid, offer and closing prices may all be slightly different and the method of calculation varies depending on the trading platform. The closing price may not necessarily be the price at which any shares actually changed hands. If you try to buy some shares today, the price asked by the market, and hence the P/E available, will have moved on again.
Understanding the P/E ratio
As mentioned in a prior chapter, early in the twentieth century US investors started comparing share prices to profits: dividing the company’s market capitalisation by last year’s declared profits. Equivalently, at the per-share level they could divide the share price by the EPS. This gave them a rough-and-ready estimate of how many future years’ earnings they were paying in order to own part of the company.
Typical P/Es encountered when the market is neither particularly high or low are in the range of 8-12, i.e. you are paying 8-12 years’ worth of future earnings in order to own one share now. As we shall see in a later chapter, however, this varies according to several factors:
The average P/E of the market itself waxes and wanes with overall market confidence, as Figure 1 shows. The FTSE 100’s P/E peaked at 30.5 in the technology stock boom in early 2000, but fell to a low of 7.2 at the nadir of the banking crisis in early 2009.
Figure 1: FTSE 100 P/Es, 1993-2012
Large companies generally have higher P/Es than small companies. I am not aware of any accepted explanation for why this is. In my opinion it is probably due to the fact that fund managers investing billions have little choice but to invest in the largest companies: researching hundreds of sub-£50m companies simply isn’t worth their time.
Figure 2 is a scatter plot of the market capitalisations of individual companies versus their P/Es. This is for all 974 quoted UK companies that had a P/E on one day in 2012. Due to some extreme values I have taken logarithms of all the data points.
Figure 2: Market capitalisations versus P/Es for all UK companies. (7 February 2012)
As can be seen, the P/Es do cluster around the 10-20 mark. Given that the scatter plot shows almost 1000 companies, very few have a P/E over 100 (log10 of P/E > 2). The red line is the line of best fit: other things being equal, you would expect a company with a market capitalisation of £10m to have a P/E of 10.4. Since the logarithms are base 10, this company would be near point (1,1) in the scatter plot. On the other hand, there are three giant companies (circled) with market capitalisations close to £100bn, or £105m. According to the line of best fit you would expect these to have P/Es of 21.3, but in fact they have relatively low P/Es for their size, of 9.9 (HSBC), 6.3