The empirical validity of Okun's law has been well documented in the literature but the magnitude of the Okun's coefficient does not seem to reach a consensus. This study provides a modeling strategy to allow for the Okun's coefficients being different in distinct regimes using the multiple structural change model. The approach considered here has the advantages that it can directly determine the number of breaks and, in the same time, estimate the break dates along with the corresponding confidence intervals. Using data of U.S. and Canada, we have the following findings when four filtered methods are considered. First, overwhelming evidence in support of structural changes is found irrespective of countries considered and filtered approaches used. Second, the estimated break dates are nearly identical in most cases, implying that the results are robust to techniques used to extract the cyclical data. Third, the process of the unemployment rate exhibits positive persistence but no indication of nonstationarity is found. Fourth, the estimates of the contemporaneous Okun's coefficient show a substantial disparity in different regimes but the values are all negative in the right direction and mostly are statistically significant. Last, the long-run estimates of Okun's coefficient are always larger than (in absolute values) the short-run counterparts, suggesting that the (un) employment is more responsive to economic growth in the long run.