Handling multi-collinearity using principal component analysis with the panel data model

Abstract

When designing a statistical model, applied researchers strive to make the model consistent, unbiased, and efficient. Labor productivity is an important economic indicator that is closely linked to economic growth, competitiveness, and living standards within an economy. This paper proposes the one-way error component panel data model for labor productivity. One of the problems that we can encounter in panel data is the problem of multi-collinearity. Therefore, multi-collinearity problem is considered. This problem has been detected. After that, the principal component technique is used to get new good unrelated estimators. For the purposes of our analysis, the multi-collinearity problem between the explanatory variables was examined, using principal component techniques with the application of the panel data model focused on the impact of public capital, private capital stock, labor, and state unemployment rate on gross state products. The analysis was based on three estimation methods: fixed effect, random effect, and pooling effect. The challenge is to get estimators with good properties under reasonable assumptions and to ensure that statistical inference is valid throughout robust standard errors. And after application, fixed effect estimation turned out to play a key role in the estimation of panel data models. Based on the results of hypothesis testing, the real data result showed that the fixed effect model was more accurate compared to the two models of random effect and pooling effect. In addition, robust estimation was used to get more efficient estimators since heteroscedasticity has been confirmed