The Social Safety Net (SSN) programmes play a key role in Bangladesh to protect the poor households from food insecurity. This study examines the effect of these programs on calorie consumption of poor households using the 2005 Household Income and Expenditure Survey data. Three treatment effect evaluation designs are applied to compare the estimated effects. Mean difference and matching estimators that do not consider endogeneity of treatment dummy produce significant negative effects when applied to the whole sample. Unconfoundedness and overlap assumptions do not exist and the assumptions are satisfied after dropping some observations using the criteria of propensity score. The effect of the SSN programmes on calorie consumption is estimated in the reduced sample using the same econometric methods, and it is found that there are insignificant positive effects in all cases. However, the treatment dummy has serious endogeneity problem, as selection for treatment is also determined by some unobserved factors such as corruption. In this case, instrumental variables regressions taking regional dummies as instruments that do not have relation with calorie consumption are applied, and produce significant positive average treatment effect.