This article shows that the effect of remittances on economic growth involves a U‐shaped pattern, which is negative initially but later becomes positive. The analysis differs significantly from earlier studies in that it examines important methodological issues on the specification and estimation of the long‐run growth effects of remittances by estimating their impact on total factor productivity (TFP) rather than on the growth rate of GDP, using time series data from Bangladesh. The use of single‐equation cointegration methods shows that remittances’ effect on long‐run growth in Bangladesh is negative and falling until the remittances‐to‐GDP ratio is roughly eight per cent. The benefits of remittances receipts outweigh their costs and their net effects start to become positive when the ratio exceeds 14 per cent.