This paper examines the current state of food insecurity in Bangladesh and its socio-economic drivers using data from the latest Household Income and Expenditure Survey (HIES 2022). Unlike previous studies that relied on less precise measures of food insecurity, such as food expenditure, diversity, and calorie intake, this study employs the internationally recognized Food Insecurity Experience Scale (FIES) and Rasch model-based thresholds to classify households as food secure or insecure. Multilevel logistic regression is used to identify significant predictors of moderate and severe food insecurity, considering the hierarchical structure of the data, with households nested within geographical clusters. Key factors found to be significantly associated with food security include the wealth index, land ownership, education of the household head, family size, remittance income and exposure to shocks. A classification tree, a popular machine learning method, is also applied to explore important interactions among these determinants. The tree analysis confirms the importance of several regression-based predictors and identifies households at the highest risk of food insecurity through variable interactions. Factors such as poverty, lack of land ownership, low education levels, and high dependency ratios collectively increase a household's vulnerability to moderate food insecurity to around 51% while the national prevalence is 19%. District-level maps of food insecurity prevalence reveal significant regional disparities, underscoring the need for targeted, district-specific interventions to effectively combat food insecurity. More broadly, policies promoting education and family planning, training in better shock management, and facilitating remittance flows through simplified processes may contribute to addressing the food insecurity challenge.

The potential socioeconomic benefits of remittances remain untapped as a large portion of remittances is transferred through informal channels worldwide. In this paper, we explored the effect of the financial sector development in the home country on the preference for informal remittance channels using a nationally representative household dataset from Bangladesh along with subdistrict level data on bank branches. However, the identification of the effect of the banking network on the choice of remittance channel is challenging due to the unobserved heterogeneity problem. We circumvent this problem by using an instrumental variable (IV) approach. We found that greater accessibility to the banking network at the subdistrict level significantly reduces the use of the informal remittance transfer channels: a 10% increase in the availability of the total number of bank branches reduces the probability of using the informal channels, at least once, by 1.79% and hundi by 2.3% on average. The results are consistent and robust across different specifications and estimation methods. The main policy implication of this result is that the expansion of bank branches can enhance the inflows of remittances through formal channels and thereby can magnify the macroeconomic benefits of remittances globally.