The Study of Statistical Process in Sample Selection Problem
|School||Zhejiang Technology and Business University|
|Keywords||Sample Selection Problem Statistical Process Applicability Theoretical Comparison Empirical Test|
With the development of social economy, people pay more attention to decision making about individuals, families, enterprises and other micro subjects. This generates a large number of micro databases. However, when it comes to study specific econometrics, a sample selection problem may sneak in it, namely, the micro data obtained from research is not a random sample in the population of study. If the traditional econometric model is still adopted to do the research, it could have biased in the estimations. Currently, when doing econometric studies with micro data, the sample selection problem is not necessarily discussed by many scholars. While in some literatures those have discussed the problem, the methods’ applicability of sample selection problem is not sufficiently concerned. Therefore, it is worthy to study the sample selection problem and the differences in its corrections.Firstly, building on the universality of sample selection problems, this paper summarizes three aspects that easily lead to sample selection problems, which are the research topic, the characteristics of study design and control deficiencies of sampling, respectively. And the paper presents some examples to illustrate that the sample selection problem has not been paid enough attention. We try to relearn the sample selection problem from its reasons, nature and data foundation.Secondly, for the five kinds of statistical processing methods commonly used at present, this paper has done some analysis and comparison systematically from their specifications and estimations to illustrate the differences from five aspects:setting features, assumptions, estimation methods, the path to solve deviation and whether to use instrumental variables to identify. Therefore, we should pay close attention to the applicability of the methods adopted in study when dealing with a certain sample selection problem.Finally, this paper makes an empirical test further which uses a micro data set from the Stata12.0software to verify the view above. According to the condition provided by the data set, we conduct a study named "estimation of private sector wages equation". This article indicates the sample selection problem in this study through theoretical analysis and descriptive statistical analysis of the data. The standardized Heckman model, Switch model and treatment effect estimation method are employed to correct the sample selection bias, respectively. Through comparing and analyzing the differences in estimations of these three methods, this article explains and verifies that different statistical processing methods of sample selection problem can have a significant impact on the results. For this reason, we further emphasize the importance of statistical processes’ applicability in sample selection problem in economic research.