Do financial traders make better returns in the stock market when they are well rested? You would intuitively assume that a trader’s level of sleep would affect their decision making.
Several studies have certainly shown that sleep affects the ability of people to make decisions in general. Though admittedly based on small samples of participants, these studies show that those who are short on sleep tend to have relatively low attention to detail, poor memory, poor performance and significant mood swings.
But when it comes to whether sleep affects financial decisions, the evidence has been mixed. The only measure of sleepiness that has been used is the annual clock changes for daylight saving that take place in many countries, since they disturb many people’s sleep. A few studies have used this to look at how stock market returns are affected on the Mondays directly after the clocks go back or forward by an hour.
One such study in 2000 concluded that returns were relatively low when traders lacked sleep, and suggested that the lack of sleep might make them more risk-averse because they were anxious and struggling to concentrate. But later studies, such as this one from 2002, suggested that the correlation between sleep and cautious investing might not be as strong empirically as initially thought.
Daylight-saving time changes have the advantage that we all have to adjust them, but they are far from an ideal proxy for sleep since they only occur twice a year, and the impact on people’s sleep is relatively small since the clock only changes by an hour. This might explain why the research evidence has been mixed in this area.
To try and improve our understanding in this area, I undertook a pilot study of a fund manager in England, analysing his investment transactions in the context of sleep data that he recorded in a diary.
I found that his sleep patterns did indeed influence his investment decisions. In line with the theory from the 2000 study, the fund manager made fewer transactions when he was short on sleep.
To see whether there was a wider correlation, I sought to develop a new proxy for sleep. We know that around 80% of people search for information online about their health issues, and there is no reason to believe that investors behave any differently. I also knew that Google data has been used by researchers to measure investor attention to individual stocks.
I therefore created a sleepiness index based on the extent to which people in the US were searching Google for 28 relevant terms including “sleep deprivation”, “sleeping pills” and “jet lag cure”. Some of these terms came from allowing the Google algorithm to offer up potential sleepiness terms based on suggested autocompletes.
The more that people searched for things to do with sleepiness, the greater the indication of sleep difficulties. Unlike the time changes from daylight saving, my index has the advantage of being based on daily data, and can measure a much wider range of sleepiness. To test its validity, I checked the index against times that we would normally associate with sleepiness, including daylight-saving time changes and also sunrises and sunsets. Sure enough, sleepiness-related Google searches increase at these times.
The index confirmed that stock-market returns are indeed quite low on days that traders are short on sleep. For every 1% daily increase in sleep difficulties across the population, stock-market returns fell by 0.14%. I also found that these patterns reversed on subsequent days, which may mean that traders realise that their initial decisions were poor and take steps to correct them.
What next from a research point of view? Researchers could potentially use the data from sleep apps to get more accurate measures of the relationship between stock market returns and the population’s sleepiness over time. No doubt the better we understand this, the more that traders will be able to use it to their advantage.
My work is another example of how online search data can shed new light on old research subjects. There are surely lots of other ways in which the academic community can use it to understand other factors that influence our decisions.