How do sell-side analysts obtain price-earnings multiples to value firms?

Abstract

Previous studies of analysts’ valuation methods show that sell-side analysts often rely on multiples-based relative valuation methods in deriving target price forecasts, predominantly earnings-based multiples. However, little is known about how analysts actually arrive at the earnings multiples that they apply in their valuations. Based on extant valuation theory, we analyse three benchmarks/reference points that analysts use to select these multiples using U.S. data. By mimicking analysts’ relative valuation processes, we show that analysts tend to assign earnings multiple premiums (discounts) to those firms expected to have growth premiums (higher risk levels) relative to comparable firms. We provide evidence that analysts use firms’ historical earnings multiples as benchmarks, and assign firms that are expected to have more (less) attractive fundamentals than they have had in the past earnings multiples that are at a premium (discount) relative to the average historical earnings multiples at which they traded. The forward price-earnings multiple for the broad U.S. market index signals the market’s expectations about the growth prospects of the U.S. economy and future economic conditions and we also find that changes in this multiple affect analysts’ choices of firm-specific earnings multiples.

Keywords:

• relative valuation
• analyst target P/E multiple
• P/E multiple premium
• growth and risk premiums
• deviation from long-run average

Acknowledgements

We thank the editor, Vivien Beattie, guest editors Mark Clatworthy and Edward Lee, and the two anonymous reviewers for their detailed comments and suggestions. We also thank Steve Young, Lisa Kutcher, and seminar participants at the 2013 American Accounting Association Annual Meeting and the 2011 American Accounting Association Western Region Meeting for their comments. All remaining errors are our own.

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCiD

Yuan Yin http://orcid.org/0000-0003-1569-6154

Notes

1. Following Yin et al. (2014), we use the term ‘analyst target P/E multiple’ to describe the forward P/E multiple which the analyst applies to value the target firm.

2. An additional complication that study faced is that since the data item for long-term growth is not provided in a significant portion of the sample broker reports, the measure was estimated using only two or three years’ (rather than the usual five years’) forecasted income statements.

3. These ideas have played a part in research in the intervening years. For example, Beaver and Morse (1978) and Barker and Imam (2008) suggest that the quality of earnings per share used in P/E-based valuations is important and that transitory items should be excluded. Existing evidence suggests that analyst earnings forecasts exclude transitory elements and reflect their assessments of firms’ sustainable future earnings (e.g. Bradshaw and Sloan 2002, Barker and Imam 2008).

4. captures the usual measure of EPS growth in FY2. It also reflects an adjustment for foregone earnings as a result of dividends distribution (dps₁).

5. For a firm that pays out all its earnings as dividends, it can be shown that (γ − 1) equals the long-run growth rate in the firm's expected earnings per share. The Gordon and Shapiro (1956) constant growth model is commonly used to establish theoretical linkages between the P/E multiple and growth and risk factors. The constant growth model, however, relies on an assumption of a 100% payout ratio that equates growth rates in earnings per share and dividends per share. As Equation (2) makes clear, this assumption is partially relaxed in the OJ model, by making a distinction between near-term growth, g₂, and long-term growth, γ − 1, which is assumed to be constant.

6. See Pfizer Inc. report, Credit Suisse First Boston Corporation, 9 January 2005.

7. Economic theory suggests that under competitive market conditions, high profitability will revert to the mean over the long run (Stigler 1963). We recognize that the economic rule of profitability mean reversion may impact analysts' earnings expectations, and thus their choices of P/E multiples.

8. As an untabulated sensitivity test, we re-estimate the regressions in the study using analyst target P/E multiples (as opposed to the inverse E/P measures used in our main tests) as the dependent variables. We eliminate observations with negative EPS₁ forecasts. We also eliminate a small portion of observations with the lowest (2%, 3%, and 5%) EPS₁ forecasts in an attempt to minimize the scaling problems (i.e. very large outliers) that might arise when the scaling variable EPS₁ approaches zero in the calculation of analyst target P/E multiples. The regression results are consistent with those based on analyst E/P multiples, and all inferences are the same. One notable difference, to be expected given the volatile nature of the P/E multiple variable, is that the adjusted R²s of the regressions are significantly lower than those based on the E/P multiples.

9. The review of broker reports and evidence in Peasnell and Yin (2014) suggests that analysts frequently use the gross margin ratio to forecast future earnings by multiplying forecasted sales by the gross margin ratio to obtain forecasted gross profit. Based on this evidence, we use gross margin ratio to measure past profitability in this study. Return on Equity (ROE) and the operating gross margin ratio would also be possibilities here. However, ROE exhibits a strong mean reversion tendency over the long run due to competition (e.g. Freeman et al. 1982, Fama and French 2000). Hence, it may act as a proxy for the expected future profitability in our tests, and therefore may not be suitable for testing GD's assertion relating to the positive effect of past profitability on target P/E multiples. The operating gross margin ratio exhibits weaker mean reversion tendency than ROE, possibly due to the fact that technology and cost structure differ across industries (Nissim and Penman 2001), but it is also less stable and less useful than the gross margin ratio for forecasting earnings. As a sensitivity check, we re-ran all our empirical tests using both ROE and the operating gross margin ratio to measure past profitability. We find that analyst E/P multiple premiums are positively associated with the industry average-adjusted ROE and the operating gross margin ratio, suggesting that those measures acted more like proxies for future profitability. We interpret the results as suggesting that analysts expect future profitability of firms with high past ROE and operating gross margin ratio to revert to the mean (decay) over the long run and therefore issue higher target E/P multiples. We find similar results when ROE and the operating gross margin ratio are used in the historical average-adjusted tests.

10. Analyst target E/P multiple is similar to an ex-dividend yield. Instead of correcting analyst target E/P multiples by adding the dividend yield, we prefer to include the dividend yield as a control variable.

11. Our reading of broker reports suggests that, in estimating industry average multiples, analysts do not appear to use the ‘out-of-sample’ approach. We therefore follow this practice and adopt the ‘in-sample’ approach for our empirical analysis here.

12. Richardson et al. (2004) suggest that analysts tend to issue optimistic EPS₁ forecasts at the beginning of the fiscal year, and they revise down the upward bias in their forecasts as the fiscal year end approaches. Given that the majority of firms have December 31 year-ends, in our tabulated results we use December EPS₁ consensus forecasts in order to reduce the influence of time-dependent factors that might introduce potential noise. Similarly, we use December consensus EPS₁, EPS₂ and LTG forecasts for the calculation of 10-year historical averages of G₂ and LTG. As a sensitivity test, we used consensus (EPS₁, EPS₂ and LTG) forecasts released in the months in which individual firms’ fiscal years end for the empirical analysis and obtained essentially the same results.

13. Analysts issue multiple EPS₁ forecasts for a given firm each year and eliminating negative EPS₁ forecasts does not lead to loss of a significant number of firm-year observations or sample firms. This procedure does not affect the inferences of our study.

14. Where reference is made to variables in Equations (4) and (5), time and firm subscripts are suppressed for compactness.

15. It is important to note that this study does not attempt to compare the effects of G₂ and LTG on analyst target P/E multiples. One needs to be cautious about making such a comparison for two reasons. First, there is not sufficient theoretical support for the argument that analysts place more weight on G₂ than LTG. Existing evidence (e.g. Bradshaw 2004) suggests that long-term growth forecasts play an important role in analysts’ stock recommendation decisions. Second, analysts issue long-term growth forecasts much less frequently than target price forecasts. We match target P/E multiples with long-term growth forecasts of less than 365 days to avoid significant loss of observations. We expect this procedure and the stickiness of LTG forecasts to impact the strength of the statistical association between LTG and target P/E multiples and limit our ability to make a valid comparison.

16. The sample size of the telecommunication services is not large enough (28 observations) for making reliable inferences. The regression analysis for that subsample is therefore omitted.

17. Prior literature and broker reports suggest that the earnings multiples represent a common measure of how expensive (cheap) stocks are in the market (e.g. Morgan Stanley 2012, Hsu et al. 2013). When analysts use valuation methods such as price-to-book value or the dividend yield model, for example, to value financial and utility firms, the results in the tables provide insight into their opinions on how many times forecasted earnings the stocks should trade.

18. The accuracy of analysts' long-term growth forecasts is weaker compared with their near-term forecasts due to factors such as greater uncertainty associated with longer forecast horizons and significant optimism in long-term growth forecasts (e.g. Bradshaw et al. 2012). However, valuation theory suggests that it is critical for analysts to forecast long-run future earnings. Moreover, existing evidence suggests that the market appears to reward analysts' efforts to forecast firms' long-term performance and such efforts also improve the performance of analysts' stock recommendations (Jung et al. 2012, Peasnell et al. 2016).