Impact of survey design in the estimation of habitual food consumption A study based on urban households of Mongolia

This paper evaluates the impact of survey design on food consumption measurement using data from the Mongolia Household Socio-Economic Survey (MSES) 2007-2008. We exploit the fact that the MSES uses a diversified approach to collect food data by measuring consumption, acquisition and stocks using both, diaries and recall interviews. First, we estimate the effect of diary length on reporting of food consumption, finding a significant decrease of reported food over time. We conclude that shorter diaries would increase the quality of estimates and lower the costs of inquiry. Second, we show that recall interviews, combined with a measurement of stocks, perform well in measuring household food consumption compared to diaries. Third, we find cyclical variation in food consumption and acquisition not only between months, but also within months and weeks. Especially for food acquisition, we detect large difference within months, which can lead to biased estimates of food acquisition. This highlights the need for enumeration to be spread over time to overcome cyclicality on food acquisition and consumption.