Interval Parameter Estimation of Quantile Regression Using Bca-Bootstrap Approach with Application to Open Unemployment Rate Study

Abstract

Parameter estimation in the linear regression model using ordinary least square (OLS) method is less precise to analyze data containing outliers. It is because outliers can cause unstable parameter estimate. In addition, the existence of outliers causes residuals to be larger so that the residual’s variance is not constant (heteroscedasticity). One model that is able to overcome the effect of outliers is quantile regression because it can accommodate the non-homogeneous variances in modeling. In this study, the confidence interval of the parameter estimate in the quantile regression model was obtained, i.e., the Bias-Corrected and accelerated (BCa) bootstrap method. The proposed method was applied in modeling the open unemployment rate in Indonesia in 2017. The quantile value used in this study is quantile 0.05, 0.5, and 0.95 with 1500 resampling in BCa-bootstrap approach. The empirical result shows that the best quantile regression model is obtained at the value of quantile 0.95 which has a Pseudo R2 value is 60.45 percent. The model at quantile 0.95 shows that the percentage of youth, economic growth rate, and labor force participation rate have a significant effect on the open unemployment rate in Indonesia.