Priming in inflation expectations surveys

Abstract

Since the global financial crisis of 2007/2008, there has been increased attention on inflation expectations and the use of central bank communication as a tool to achieve a central bank’s objective for inflation. However, much of this research analyses the survey data with limited consideration of the survey design that generated the data, or the differences across surveys and countries. In this research note, we focus on one element of South Africa’s Bureau of Economic Research household inflation expectation survey question – the inclusion of a historical inflation number in the survey question. Using a dataset created by Pienaar (2018), we are able to evaluate the impact of its inclusion on the data created. We find that the inclusion of a historical inflation number into the survey question, distorts survey responses, particularly a group considered to be relatively ‘less rational’. We do not investigate whether this bias is caused by anchoring (Tversky & Kahneman (1974), learning (Cavallo, Cruces, & Perez-Truglia, 2017), or any other theory, but we do argue that the observed bias should raise concern about the interpretation of surveys, where the question includes any form of extra information (priming). The impact not only distorts the level of the response, it also leads to changes in the distribution.