
Two managers issue earnings forecasts; manager A’s forecasts are consistently three cents below realized earnings, while manager B’s forecasts are one cent below half of the time and one cent above the rest of the time. Do investors prefer A’s forecasts, even though B’s are more accurate, and does this preference depend on the type of investor? These were the questions posed by a team of three researchers, Gilles Hilary, Charles Hsu & Rencheng Wang.
The voluntary disclosure of financial information through management forecasts is an important part of a firm’s communication activities. Managers issue information about expected earnings to set or alter market earnings expectations, to pre-empt litigation concerns, and to be seen as being transparent. The extent to which management forecasts affect price formation as well as managers’ career development has been extensively studied: forecast accuracy is used to assess forecast quality and even management ability, with the conclusion that more accurate managers and firms exert a greater influence on price and analyst opinions, and lower CEO turnover.
The researchers argue, however, that the usefulness of a management forecast is based not on its accuracy, but on its consistency. In addition, they contend that managers who make consistent forecast errors have a greater ability to move prices and analyst forecast revisions. They also examined differences in reactions to management forecasts by examining whether the degree of users’ sophistication affects their understanding of forecast properties.
To return to the example: although the forecasts of manager A are less accrurate than those of manager B, the forecasts of manager A have greater implication for earnings predictability. Rational users of A’s forecast will adjust their own forecasts upwards because manager A exhibits a systematic bias in her forecasts – if A issues a forecast of 48 cents per share, they will update theirs to 51 cents. With B, they will find it difficult to unravel biases in management forecasts and thus dampen their response.
The research team was thus able to come up with testable hypotheses. They took a sample from the management forecasts of quarterly EPS (earnings per share) in the First Call database, and acquired forecasts over at least six financial quarters from CEOs selected from ExecuComp database.
The findings show that managers with higher forecast consistency have a greater ability to move prices and influence analyst revisions. There was some support for the notion that accuracy influences market reactions and analyst revisions, but effects are generally weaker than those of consistency. Indeed, the effect of accuracy often disappears with control for consistency. In contrast, the effect of consistency is robust to a host of specification checks. It is easier to predict the bias when managers are more consistent, and investors and analysts filter systematic bias in management forecasts more easily when forecasts are more consistent.
Institutional investors and experienced analysts – in other words, sophisticated users -- react more to consistent forecasts than retail investors and inexperienced analysts.
This research makes a number of contributions. Existing research focuses on absolute forecast error, but this study shifts it from the magnitude of error, i.e., accuracy, to the volatility, i.e., consistency. It shows that less sophisticated users value accuracy more than consistency; this is valuable information for regulators interested in understanding the tradeoffs associated with biased forecasts among different types of users.
The study contributes to the literature on “downward biases” in management forecasts. The large majority of managers are downward biased in their quarterly forecasts. Previous studies have shown that managers guide analysts’ expectations downwards by issuing pessimistic quarterly forecasts, but there is little research on how managers trade off forecast accuracy for “lowballing” without compromising the quality of the forecasts. This research shows that the bias is not necessarily detrimental to the influence of their forecasts as long as it is identifiable and predictable where users are concerned.