This study attempts to uncover the biases in the impact evaluation of remittances when the problems relating to selection bias and counter factual are not taken into account. Taking migration as an intervention and foreign remittances as an input, the study measures the socioeconomic impact using an approach which yields more accurate non-experimental estimates in self-select cases through multiple output and outcome indicators such as income, expenditure, saving, and capital accumulation which, directly and indirectly, affect households’ welfare, poverty incidence and growth prospects of a country. Using PIHS data, the study first calculates the difference in socioeconomic characteristics of treated or remittances beneficiary households (RBH) and control or remittances non-beneficiary households (NRBH) ignoring endogeneity and observable differences. Second, it calculates the propensity score and evaluates the impact using data from common support area for both RBH and NRBH households. Third, it evaluates the impact using the propensity score matching approach which replicates the experimental benchmark. The difference in the first and the third estimates reveals the bias originating from the issues of selection and difference in observable characteristics. The results show that after controlling for observable characteristics of households, regional difference, networking and applying the selection correction technique, the average impact of remittances is significantly reduced. A disaggregated analysis shows that the socioeconomic impact of remittances differs by the level of skills. The impact is significant for relatively low skilled poor households but for high skilled households it remains significant only in case of bank deposits. The paper concludes that estimates are biased upward if the selectivity issue and endogeniety problems are ignored which may lead to wrong policy implications.