Analyst Forecasts: Sales And Profit Margins
Hong Kong Polytechnic University – School of Accounting and Finance
Hong Kong Polytechnic University – School of Accounting and Finance
Hong Kong Polytechnic University – School of Accounting and Finance
January 2016
Abstract:
Equity analysts forecast sales (S) in addition to eps. We study the two components of eps forecasts, PM and S where PM=eps/S by definition. We use I/B/E/S data and focus on 12 months forecasts. Well-established prior research in analyst forecasts (AF) motivates the three initial issues in our study: the potential upward AF bias, the relative accuracy of AF compared to benchmark models, and the sub-optimality of AF. For these questions one can hypothesize that the two components produce similar results, much like eps. We question whether results related to S are different from results related to PM. Our main findings show (i) the upward bias of PM exceeds that of S, (ii) eps forecast errors depend more on the uncertainty in PM than that in S and (iii) reported S has greater impact on forecast revisions than reported PM. Related to (iii), results show that AF errors of S correlate positively with subsequent changes (growth) in S; PM forecast errors, by contrast, show no such correlation. We also show existence of AF sub-optimality, especially when firms are small.
Analyst Forecasts: Sales And Profit Margins – Introduction
Research on analysts’ forecasts (AF) can be refined by focusing on the two underlying components of eps: sales (S) and profit margins (PM). We study the extent to which AF of S and PM are similar or distinct in terms of their respective statistical properties. The topic is of interest due to the two components’ economic dissimilarity. PM is a ratio while S is a dollar amount (per share). S is similar to eps because it is also a dollar amount and tends to grow over time. Both eps and S can reflect results of underlying investments and from the changes in investments. But S by itself provides no information about value creation or economic performance while eps at least indicates if a firm has a profit or a loss. PM reflects a firm’s competitive conditions and operating efficiency. It connects with eps insofar that both have the same sign, positive (income) or negative (losses). But PM is not expected to grow over time. Of all of these aspects, the central distinction refers to a simple fact: S captures growth, PM does not.
We start out addressing three well-established topics in analyst forecasts of eps research. First, because AF of eps is generally viewed as too optimistic, an obvious question is: do (S, PM) both contribute, more or less equally, to this upward bias? Second, the literature recognizes that AF of eps tends to be more accurate than benchmark schemes based on random walk concepts. We ask: is it true that AF of both S and PM are more accurate than benchmark models to a similar degree? Third, the literature on AF of eps addresses whether more sophisticated forecasting schemes can perform better than analyst forecasts. Both (S, PM) may, or may not, be similar in this regard. While one can assume the null that what the literature document regarding the eps from the perspectives of bias, relative accuracy and sub-optimality can be equally attributed to S and PM, we want to emphasize that this may not be the case. The eps forecast derives from the product of two components with distinctly asymmetric statistical properties: the growth in expected eps on average rests on the growth in S, not the change in the PM. The difference between S and PM goes beyond growth. As to accounting rules, many expense items tend to be relatively arcane as compared to S. Accrual expenses like (parts of) taxes, pensions, and more or less hidden non-recurring expenses tend to be sensitive to subjective accounting estimates, and more so than revenues insofar that revenues are harder to manipulate. Research recognizes this issue of “soft” accruals – or discretionary expense accruals – and the research deriving from Jones’s model illustrates: it presumes that as a starting point sales need no adjustment for “soft” accruals. The so-called modified Jones model acts as a check just in case S, too, is somewhat on the “soft” side. This view suggests S is a more reliable number compared to PM.
These two points capture the core difference in the S and PM characteristics that informs the research in our paper: sales and sales forecasts reflects growth whereas PM and PM forecasts reflect partially transitory up and down ticks. An important aspect stands out: it is likely that S is relatively less uncertain (or more informative) than PM. S has a growth trend which may be reliably estimated from current growth and investment. PM does not have a natural trend component and can only rely on a random walk. The randomness of PM is intensified as accounting rules for PM are relatively more arcane as compared those for S. This relative uncertainty leads us to predict that AF of S may perform better than analyst forecasts of PM. We discuss our predictions below.
We predict that PM optimism exceeds that of S. Analysts optimism tendency can be affected by the concern of their clienteles. Because the expense accounting is arcane and relative hard to “understand”, analyst clienteles may be less likely to react adversely when a PM forecast goes wrong. In other words, an analyst who wants to sell an upbeat story (firm X is a strong buy) tends to feel less exposed to adverse “feedback” if he/she has inflated PM as opposed to S.
See full PDF below.
