This is my last post before my summer vacation. The next two weeks I will spend in a landmark cottage in Scotland without wifi and probably hardly any mobile phone reception. It is going to be perfect, and perfectly relaxing.
If you work in banking, you know that time off comes at a premium since we all tend to work long hours. I used to joke that holidays are when I work less than 45 hours a week. But working long hours is really not helpful. Being stressed or sleep-deprived increases your risk of depression, anxiety and most certainly reduces your ability to perform complex analytical tasks. On the other hand, when you are well-rested, your ability to perform such tasks is enhanced.
This can be seen in the data…
Sima Jannati and Sarah Khalaf looked at the forecasts of equity analysts in the United States make and tried to find out if well-rested analysts make better forecasts. To do this, they looked at the accuracy of the forecasts an analyst makes in the three days after a public holiday, vs. the accuracy of forecasts made before the public holiday. They also looked at the accuracy of forecasts on the first day after the switch to daylight savings time each spring (when we all get one hour less sleep) and after the switch back in autumn (when we all get one extra hour of sleep).
Making forecasts about company earnings is a complex task that requires the analyst to weigh many different factors. A stressed or exhausted analyst has reduced mental bandwidth and is likely to engage in oversimplified reasoning and flawed forecasts. Meanwhile, a well-rested analyst is able to weigh different factors more carefully and should thus make better forecasts. And indeed, forecasts made by analysts in the first three workdays after a public holiday are more accurate than forecasts made in the three days or the thirty days before the public holiday. The estimation error for earnings growth shrinks by 0.5% to 0.6%. If that sounds like nothing, be aware that it is roughly the same reduction in estimation error an analyst achieves with one additional year of professional experience and roughly seven times the difference in forecast error between the top 25% most experienced analysts and the 25% least experienced analysts covering a company.
The last sentence of the previous paragraph should give you pause (and if it doesn’t I’d say it is a sign that you are exhausted and need some vacation, too). How can it be that the accuracy of an analyst’s forecasts declines on average by 0.5% per year but the difference between the most experienced analysts’ forecasts and the least experienced analysts’ forecasts is just one seventh of that or 0.07%? Surely, the most experienced analysts aren’t just one seventh of a year older than the least experienced analysts.
The reason for this seeming discrepancy is that the holiday effect is not the same across analysts. Female analysts tend to make better more precise forecasts than male analysts and they also exhibit less of a holiday effect than male analysts. And on average female analysts tend to have fewer years of experience than male analysts, thus skewing the results. Similarly, all-star analysts that make the most accurate forecasts tend to show less of a holiday effect. Their forecasts are always more accurate than that of the average analyst. And this also skews the results and helps explain the seeming contradiction.
But in all, the results are that if you are an analyst, taking a few days off will improve your forecasts. In fact, even getting one hour of additional sleep will help. The study found that on the Monday after daylight savings time falls back in autumn, forecast errors drop meaningfully compared to the Mondays two or three weeks later.
All of this is my convoluted way of saying, I am off on vacation and you can look forward to better forecasts in September when I am back in action.