Misunderestimating Corruption

Kraay, Aart ; Murrell, Peter

World Bank, Washington, DC

关键词: BENCHMARK;    BIASES;    BINOMIAL DISTRIBUTION;    CLASSIFICATION;    COMPUTERS;   

DOI  :  10.1596/1813-9450-6488
RP-ID  :  WPS6488

学科分类：社会科学、人文和艺术（综合）

来源: World Bank Open Knowledge Repository

【 摘 要 】

Estimates of the extent of corruptionrely largely on self-reports of individuals, businessmanagers, and government officials. Yet it is well knownthat survey respondents are reticent to tell the truth aboutactivities to which social and legal stigma are attached,implying a downward bias in survey-based estimates ofcorruption. This paper develops a method to estimate theprevalence of reticent behavior, in order to isolate ratesof corruption that fully reflect respondent reticence inanswering sensitive questions. The method is based on astatistical model of how respondents behave when answering acombination of conventional and random-response surveyquestions. The responses to these different types ofquestions reflect three probabilities -- that the respondenthas done the sensitive act in question, that the respondentexhibits reticence in answering sensitive questions, andthat a reticent respondent is not candid in answering anyspecific sensitive question. These probabilities can beestimated using a method-of-moments estimator. Evidence fromthe 2010 World Bank Enterprise survey in Peru suggestsreticence-adjusted estimates of corruption that are roughlytwice as large as indicated by responses to standardquestions. Reticence-adjusted estimates of corruption arealso substantially higher in a set of ten Asian countriescovered in the Gallup World Poll.