To Err is Human: Inconsistencies in Food Conversion Factors and Inequality in Malawi

One of the key deficiencies of household survey data for measuring poverty and inequality is that survey nonresponse depends on the income of the respondent, whereby rich people are less likely to cooperate with household surveys than poor people. This nonrandom nonresponse may hide the true level of poverty and inequality. Another potential problem with household survey data which has received scant attention in the literature is the quality of food conversion factors. Nonrandom errors and inconsistencies in food conversion factors can potentially have a nontrivial impact on measured poverty and inequality.The paper looks at the impact of errors and inconsistencies in food conversion factors on measured consumption inequality. Malawi has been used as case study with data from the Second and the Third Integrated Household Surveys (IHS2 and IHS3). Two consumption aggregates are used; an official aggregate which is contaminated by errors and inconsistencies in food conversion factors and a new aggregate which cleans out these problems. The paper finds that the inconsistencies and errors in the conversion factors were not random in that they affected the richest households more than the poorest households. Consequently, the official aggregate understates the level of inequality as measured by the Gini coefficient. Inequality is underestimated by 4.4 and 2.3 Gini points for 2004/5 and 2010/11 respectively. The disparities are not only sizable but they are also statistically significant. I also find that the official aggregate progressively underestimates the share accruing to higher percentiles. Nonparametric tests for Lorenz dominance confirm that these differences in measured inequality are robust. Using the new aggregate, the paper also finds that inequality is not worsening overtime as the official aggregate suggests. All this implies that the quality of food conversion factors is critical for the accurate measurement of levels of and trends in inequality