The past attempts to investigate whether the Marshall-Lerner condition is fulfilled by using aggregate data in Bangladesh suffer from aggregation bias. This paper estimates trade elasticities using bilateral data between Bangladesh and its major trading partners. The results, using data covering 1973-2009, confirm long run relationships of volumes of export and import with real exchange rate and real income. The study unveils that the Marshall-Lerner condition holds only in case of the United States. As such, the depreciation of real exchange rate may not be effective in improving the trade balance of Bangladesh in the long run.