The study is an attempt to investigate long-run and short-run relationships between gross domestic product (GDP) and agricultural GDP of Bangladesh applying the Vector Error Correction Model (VECM). The data published by the BBS for the period 1972-73 to 2015-16 was used. As preconditions of VECM, the stationarity of the time series data and co-integration of the variables were tested. The GDP and agricultural GDP in logarithm forms were found integrated of the order one (d=1) and co-integration of rank one (r=1).The coefficient (beta) of estimated long-run equilibrium equation was positive and significant implying a long-run equilibrium relationship between GDP and agricultural GDP. A positive and significant time trend was also observed in the co-integrating equation. Both coefficients (alpha) of the error correction terms (ECTs) in the VECM, which indicate the rates of short-run adjustments towards the long-run equilibrium, were negative and significant confirming consistency with the theory. Since both ECTs were negative and significant, there was a long-run bi-directional causality running from agricultural GDP to GDP and vice-versa. Responding to a positive shock in agricultural GDP, the GDP gradually increases while due to shock in GDP, the agricultural GDP decreases in the initial stage and increases gradually thereafter. Various post-estimation specification tests were performed and the estimated parameters of the model were found consistent and efficient.