This paper uses annual US data to examine the causal relationship between immigration and real GDP. Despite its implications for policy, a statistically robust relationship between these two series has been difficult to pin down. Our tests reveal that both the series are break-stationary. Therefore, we apply the Gregory-Hansen (1996) residual based cointegration approach to these series to establish a long run relation between them in the presence of regime shifts. Standard Granger causality test shows that the relation flows from economic growth to immigration in the short run, but not the reverse. However, the Error Correction Models within Vector Error Correction framework shows a bidirectional feedback relationship in the long run which is intuitively more appealing.