OBJECTIVE: The aim of this study was to examine the drivers of dietary change over time and the roles agriculture and economic development have played.

METHODS: We used principal component analysis to derive dietary patterns from 7 cross-sectional rounds of the Bangladesh Household [Income and] Expenditure Survey. We then used linear probability models to estimate associations of adherence to dietary patterns with socio-economic characteristics of households, and with agricultural production on the household and regional level. For dietary patterns that increased or decreased over time, Blinder-Oaxaca decomposition was used to assess factors associated with these changes.

RESULTS: Seven dietary patterns were identified: modern, traditional, festival, winter, summer, monotonous, and spices. All diets were present in all survey rounds. In 1985, over 40% of households had diets not associated with any identified pattern, which declined to 12% by 2010. The proportion of the population in households adhering to the modern, winter, summer, and monotonous diets increased over time, whereas the proportion adhering to the traditional diet decreased. Although many factors were associated with adherence to dietary patterns in the pooled sample, changes in observed factors only explained a limited proportion of change over time due to variation in coefficients between periods. Increased real per capita expenditure was the largest driver of elevated adherence to dietary patterns over time, whereas changes in the agricultural system increased adherence to less diverse dietary patterns.