Hansen (2000) reports that the p-value for threshold model using initial per capital output was 0.088 and that for the threshold model using initial literacy rate is 0.214. Hence, Hansen concluded that there might be a sample split based on output. In our framework, using 500 bootstrap replications, the p-value of test F1 for the threshold effect with an unstable threshold using initial per capital output is significant at 0.029 and that for the threshold effect with an unstable threshold using initial literary rate is 0.141. This suggests that the conclusion in Hansen (2000) is robust, and hence it is reasonable to split the sample based on output.