We estimate the short-run and long-run determinants of the trade balance of Turkish Mining with EU (15) countries as well as impact of Customs Union (CU) agreement using the bounds testing approach to the cointegration and the error correction modeling. In selecting the optimal model, we follow Yazici and Islam (2011a, 2011b, 2012) and Yazici (2012) and adopt their model selection strategy where selection is made from the set of those models that satisfy both diagnostic tests and cointegrtion, thus ensuring the selection of a statistically reliable and cointegrated model. Estimation results based on the data for 1988-I to 2008-IV period indicate that in the determination of mining trade balance in the short-run only real domestic income matters. Long-run results indicate that real exchange rate and real domestic income variables have coefficients with expected signs but they are not statistically significant. Only statistically significant long-run determinant of mining trade balance is real EU(15) income. Dummy variable for the customs union agreement does not have a statistically significant coefficient, meaning that the agreement does not have a significant long-run effect on mining trade balance of Turkey with EU(15).