Fear of the state in governance surveys? Empirical evidence from African countries

Thomas Calvo (1, 2) , Mireille Razafindrakoto (1, 2) , François Roubaud (1, 2)

1 DIAL - Développement, institutions et analyses de long terme
2 LEDa - Laboratoire d'Economie de Dauphine

The need to collect data on governance-related issues has been growing since the 1990s. Demand gained momentum in 2015 with the adoption of SDG16 worldwide and Agenda 2063 in Africa. African countries played a key role in the adoption of SDG16 and are now leading the process of collecting harmonised household data on Governance, Peace and Security (GPS). Yet the possibility has recently been raised that sensitive survey data collected by government institutions are potentially biased due to self-censorship by respondents. This paper studies the potential bias in responses to what are seen as sensitive questions, here governance issues, in surveys conducted by public organisations. We compare Afrobarometer (AB) survey data, collected in eight African countries by self-professed independent institutions, with first-hand harmonised GPS survey data collected by National Statistics Offices (NSOs). We identify over 20 similarly worded questions on democracy, trust in institutions and perceived corruption. We first compare responses from AB survey respondents based on who they believe the survey sponsor to be. No systematic response bias is found between respondents who believe the government to be behind the AB survey and those who consider it to be conducted by an independent institution. Our estimations suggest that the observed residual differences are due to a selection bias on the observables, which is mitigated by propensity score matching procedures. The absence of a systematic self-censorship or attenuation bias is further evidenced by means of an experimental design, whereby responses from GPS surveys conducted by NSOs (the treatment) are compared with AB surveys sponsored by reportedly independent bodies. Our results provide evidence, at much higher levels of precision than other existing data sources, of the capacity and legitimacy of government-related organisations to collect data on governance as a matter of national interest and sovereignty.

Mots clés: Survey sponsor effect; Governance data; Official statistics; SDG16; Africa

Domaines: Économie et finance quantitative [q-fin]

Thomas Calvo, Mireille Razafindrakoto, François Roubaud. Fear of the state in governance surveys? Empirical evidence from African countries. World Development, 2019, 123 (November 2019). ⟨hal-03355118⟩