Testing the Effectiveness of Mobile Phone Data Collection for Microenterprises in Africa

This project proposes a Randomised Controlled Trial to test the effectiveness of using mobile phones to collect data on microenterprises. The researchers intend to track microenterprises over a 16-week period, with random assignment to alternative survey methods. The trial will run in Soweto, a low-income neighbourhood in South Africa, where the team is working with a partner NGO to prepare a randomised evaluation of an entrepreneurship training program. This presents an ideal opportunity to explore innovative data collection methods for microenterprises in Sub-Saharan Africa.

The researchers will conduct a listing exercise in Soweto in order to build a representative sample of 900 enterprises, which will then be divided between three data collection methods, using stratified random assignment:

  • conventional face-to-face interviews at four-weekly intervals
  • conventional face-to-face interviews at weekly intervals; and
  • mobile-phone based interviews at weekly intervals.

This will allow the team to investigate (i) the relative accuracy of face-to-face and mobile phone-based interviews, and (ii) the degree of volatility in microenterprise outcomes omitted from low-frequency (monthly) interviews. All firms in the sample will also receive face-to-face baseline and endline interviews, in order to improve the comparability of the data collected across the three sub-groups.

If the results from this study suggest that we can accurately collect high-frequency data on microenterprise profits and sales using mobile phones, researchers will have a powerful and cheap new method for measuring volatility and dynamics in microenterprise performance. This is directly relevant for a very wide range of policy questions concerning microenterprises. For example, it is often argued that one key advantage of microfinance is that it helps to smooth shocks. It remains unclear whether microfinance does this by implicitly insuring microenterprises against unanticipated shocks, or by helping microenterprises to deal more effectively with ‘lumpy’ costs (such as end-of-month rent payments). This is a critical distinction for understanding how microfinance products can be improved, but is very difficult to identify without high-frequency data. If mobile phones can be used to reliably measure microenterprise performance, researchers will be able to characterise this volatility much more precisely, and better understand how volatility shifts over time.


Simon Quinn

University of Oxford

Kate Orkin

University of Oxford

Robert Garlick

Duke University