Monday, March 31, 2014

The BIT's tax experiment is on NBER

Michael Hallsworth et al. have a new NBER working paper called The Behavioralist As Tax Collector: Using Natural Field Experiments to Enhance Tax Compliance*. This is an interesting and welcome development.

To put this paper in context, one of the biggest successes of the Behavioural Insights Team since their formation in 2010 has been their tax experiments using social normative messages like "9 out of 10 people pay their tax on time" in communications with taxpayers. Their findings were first highlighted in their 2012 report on 'Fraud, Error and Debt':

"The graph shows a 15 percentage point increase from the old-style control letter which contained no social norm and the localised social norm letters. HMRC estimates that this effect, if rolled out and repeated across the country, could advance approximately £160 million of tax debts to the Exchequer... This would free up collector resource capable of generating £30 million of extra revenue annually."

Those figures are eyebrow-raising, to put it mildly**. Given the low implementation costs of adding a sentence like "9 out of 10 people in your neighbourhood pay their tax on time" to a tax letter that is going to be sent anyway, those sort of returns potentially mean enormous benefit:cost ratios. It's a really striking result, one that the BIT themselves often cite in interviews and presentations and which has inspired similarly successful efforts in Australia and Ireland (see #6).

The problem is that this graph, while intriguing, is from a paper that is off the (academic) reservation. I don't mean that as criticism - I admire the clarity of the BIT's white papers on health, charitable giving and so on and they are clearly aimed at policy-makers rather than academics. As an academic working in Behavioural Science, who is interested in behavioural change interventions, I am however in the position where I'm not comfortable citing some of the most interesting work being done in the area because it hasn't been peer-reviewed.

So it's great to see this NBER working paper which looks under the hood of two of the taxation field experiments conducted in the UK (N=200,000). As implied by the descriptive statistics in the figure below, the social normative messages were effective at increasing tax payments.

I hope to see more of the BIT's work in this format in coming years.

*Update - I may have mistakenly implied that NBER working papers are peer-reviewed. This is not the case, as Paul Krugman explains here.

*The title of the paper is in the same vein as John List's "The Behavioralist goes to the factory / the market / school" series.
**This is partly a function of the fact that the BIT work on a national scale and phrase their findings accordingly, but a 15 percentage point increase is also a very large effect size. The economics literature has taken a few stabs at social normative messages in tax compliance communications before and the average effect size has been basically zero.

No comments: