Undercover Economist

To tweet or not to tweet?

Economist Justin Wolfers runs a controlled experiment to test how Twitter is affecting his productivity

I don’t normally hold with the traditional New Year’s resolution of quitting some objectionable habit – even though my favourite economist, Thomas Schelling, has written very thoughtfully on the subject. (Schelling, a brilliant game theorist and long-time smoker, used a variety of game theoretic tricks to outwit a formidable opponent – his addicted self.)

But as 2011 drew to a close, I had been wondering about my addiction to Twitter, the service that allows users to publish online short messages – grumbles, aphorisms and most often, links to recommended articles. Other users can choose to whose messages they will subscribe and unlike on fully-fledged social networks, such as Facebook, this is not necessarily a reciprocal relationship. (Facebook users have friends; Twitter users follow and have followers.) My Twitter habit has the pernicious consequence of being rather time-consuming – but it has plenty of benefits too. Should I quit? Cut down? Or should I resolve only to stop feeling guilty?

Part of the problem, I realised, is the difficulty of measuring the costs and benefits of the habit. Imagine my curiosity, then, when I noticed that the economist Justin Wolfers – a self-described Twitter cynic – had joined the club and was running an experiment to test how Twitter was affecting his productivity.

“Every morning I would flip a coin,” he explained to me. “Heads, I would sign on to Twitter, tails, I would simply tweet ‘Tails: goodbye for another day Twitter.’” It might seem strange to run an experiment with a single subject, but that all depends. If the aim is to discover the effect of Twitter on the productivity of Justin Wolfers, the experimental design looks just fine.

The challenge, of course, is to interpret the results. “I tried to be scientific,” said Wolfers, who installed software on his computer to record his use of different programmes and websites, while also using Google alerts to track whether his tweets were having much impact on web chatter. “I’m not sure I succeeded.”

Wolfers rated his productivity levels at the end of each day – revealing, and “also a total bummer” – rarely topping six out of 10. But that’s not unusual: a persistent anxiety that each day has been poorly spent is, I feel, the sign of many a productive person.

Ultimately the formal experiment broke down: “After a while, I got tired of flipping coins.” Wolfers has his data; he has never bothered to analyse it. He has decided that Twitter works for him. The informal experiment of giving Twitter a try to see how it worked out was, it seems, of far more practical use than the formal experiment of randomising days on and days off.

This makes some sense. I’ve become convinced that most of us do not experiment enough with new experiences. (The first 20 years or so of life are an exhausting but stimulating exception to this rule.) Yet few of the experiments we could be trying are conducive to a proper randomised trial.

Somehow this is a great disappointment to my inner nerd. Both Justin and I like the idea of running controlled experiments in everyday life – gut feelings can be so misleading – but he warns that to do it right takes more discipline and time than many of us might want to deploy.

Yet Justin Wolfers’s experiment has inspired an unexpected insight. The toss of a coin might not have generated data that anyone cared to use, but it had the obvious consequence of reducing the days spent on Twitter by about half.

Of course one could simply decide to spend less time on Twitter, but the arbitrary dictates of the coin have a curious power. (Yes, I have read The Dice Man.) So I do not think I’ll be quitting Twitter this year; I will be using the toss of a coin to help me cut back a little.

Also published at ft.com.