Is past success a predictor of more success to come? That's a question managers should be thinking more deeply on following a study that found those managers who issued a series of accurate short-term forecasts for their firms were more likely to then issue a less accurate forecast.
That counterintuitive finding emerged in a study by Gilles Hilary and Charles Hsu, who attribute the poorer result to overconfidence on the part of the manager.
"The self-attribution principles suggests that managers who successfully forecasted earnings in the past attribute too much of their success to their superior ability and too little of it to luck," they say.
"The resulting overconfidence leads to suboptimal behavior whereby they place too great a weight on private information and too little on public signals such as market prices and financial analyst forecasts."
They confirm this bias in an investigation of 5,768 management forecasts issued between 1994 and 2007.
Managers who experienced a streak of accurate predictions subsequently issued less accurate forecasts such that increasing the streak by one standard deviation increased error in the subsequent forecast by 14 per cent of its median value.
Moreover, managers on a streak issued more precise forecasts suggesting they were suffering dynamic overconfidence.
"Overconfidence is not a fixed characteristic but rather a recurring and dynamic phenomenon that varies in intensity over time," and subsequent inferior performance can lead managers to revise their perception of their skill downward, thus reducing overconfidence.
The authors were also interested in whether the market responded to fluctuating overconfidence and found that analysts and investors did indeed seem to be aware of the problem and to discount overconfident reports. For every increase in a streak of one standard deviation, the sensitivity of the three-day return to the forecast revision was reduced by about 40 per cent of the median sensitivity, and the sensitivity of the analyst reaction fell by more than 50 per cent of the median sensitivity.
"This suggests that both investors and analysts recognize that the forecasts issued by overconfident managers are less accurate and, as a consequence, react less strongly to these forecasts than to those issued by managers who have not had a recent run of successful predictions," they say.
There is a note of caution, however, that the findings do not mean overconfident managers necessarily underperform compared to other managers.
"Our framework does not predict nor preclude that managers who have experienced a series of successful predictions will subsequently make lower quality forecasts than those who haven't. Rather, it predicts overconfident managers will issue lower quality forecasts than their skill and environment would otherwise predict in the absence of overconfidence," they say.
BizStudies
Overconfidence can Put a Halt to Winning Streaks