Giulia Mascagni (IDS)

One size does not fit all:  A field experiment on the drivers of tax compliance and on delivery methods in Rwanda 

Abstract:  Although field experiments of tax compliance represent a growing field of research, the literature has so far focussed exclusively on high and middle-­‐income countries.  This paper starts filling this gap by reporting the results of a tax field experiment in Rwanda, while also highlighting some characteristics that may be common to other low-‐income countries.  We evaluate an intervention carried out by the Rwanda Revenue Authority (RRA), which involved sending messages to taxpayers to nudge their declaration behaviour during the filing period of January-­March 2016.  Our study focuses particularly on declarations for business profits tax, filed by both corporations and unincorporated firms.  Building on the existing theoretical and empirical literature, our study is designed to address two inter-‐related questions.  First, what are the key drivers of compliance in Rwanda? Second, what is the best delivery method to reach taxpayers with messages designed to improve compliance?  Although other studies have explored delivery methods in the context of taxpayer communication, our study is the first one to interact them with different message contents. As a result, we evaluate a set of nine treatments that combine three message contents (deterrence, fiscal exchange, reminders) and three delivery methods (letters, SMS, emails) – as compared to a control group that received no message.  As far as the drivers of compliance are concerned, we find that a friendly approach to taxpayers, including both information about how tax revenues are spent (fiscal exchange) and gentle reminders of deadlines (control message), is generally more effective than deterrence.  However, small taxpayers are still quite responsive to the possibility of being fined and prosecuted (deterrence).  As regards delivery methods, we show that cheaper methods like SMS and emails can be highly effective and result in economically and statistically significant revenue gains for the RRA, as compared to letters. Overall, we estimate that our experiment generated about nine million USD as extra revenue for the RRA.