You can smoke this study

Friday posts are about the weird and wonderful side of econ research and the studies that are so clearly wrong that you wonder how anyone ever thought it would be a good idea to post them online.

Case in point, the smoke and mirrors unleashed by Evangelos Vasileiou from the Hellenic Mediterranean University in Greece. He tried to find a market anomaly driven by psychological stress that hadn’t been described in the literature, so he looked at the UK’s ‘No Smoking Day’. This is a campaign in the UK, where on one Wednesday of the year, smokers are urged to give up smoking for the day.

Now, I have lived in the UK for ten years now, and this was the first time I ever heard of this campaign. In fact, the campaign is clearly so well-funded and organised that its website still announces the No Smoking Day of 12th March 2025(!?). But in case you are eager to participate, the next No Smoking Day is going to be 10 March 2027.

Anyway, Vasileiou analysed the returns on the Wednesday designated as this awareness day nobody ever heard of and compared them to any other Wednesday that wasn’t designated a No Smoking Day. This gave him a difference in returns of the FTSE 100, where the No Smoking Wednesdays had significantly lower returns than the Smoking Wednesdays (presumably because investors on nicotine withdrawal are more risk averse)

Mind you, the difference was only statistically significant at the 10% level, so he performed a break point analysis and identified 1991 as a structural break. From 1984 to 1991, there was no statistical difference in returns, while since 1991, there has been (though the difference remains significant only at the 10% level).

What really made a difference was when he added additional control variables to the statistical tests. In this case, he added the proportion of people in the UK who smoke and the proportion of smokers in the UK who quit smoking. Now, let’s ignore for a moment that these two variables are highly correlated (the share of smokers will mostly decline because people quit smoking), which invalidates the entire statistical approach. He also added the share of people in the UK who own shares directly. Why this should matter is anyone’s guess.

But what these variables do is add two steadily falling trends (share of smokers and share of people who own stocks) and one rising trend (share of smokers who quit). And all of a sudden, the difference in returns was statistically significant at the 5% level, which is what every researcher wants to see because now you can submit it to a journal for publication…

IK am not a smoker, but I can see some letters forming in the haze of my colleague smoking a fag right now: d-a-t-a m-i-n-i-n-g.

Average returns on a No Smoking Wednesday vs a Smoking Wednesday