This article examines the export-led growth hypothesis for Korea in five-variable vector autoregressive and vector error correction models from 1973:1 to 1994:4. Results of the vector autoregressive models indicate economic growth Granger-causing export growth, regardless of the sample period. However, results of the vector error correction models show bidirectional causality between export growth and economic growth when the multivariate generalization of the Granger causality tests are used. In the variance decompositions, the real exchange rate contains most information regarding future fluctuations in economic growth and export growth followed by money supply and government expenditure in the subsample and the full sample, with economic growth as a dependent variable. However, when the dependent variable is export growth, then the order of the magnitude in the full sample becomes the exchange rate, followed by economic growth, government spending, and money supply. The findings in this paper suggest that the omitted variables have masked or overstated the effect of exports on income or income on exports in prior studies.