2.1. Data Sources and Sampling

The research data are obtained from CFPS, which is a nationwide, comprehensive, longitudinal social survey conducted by the Institute of Social Science Survey, Peking University [36,37]. The CFPS 2018 survey started in June 2018 and ended in May 2019. A total of 15,000 families were interviewed, and about 44,000 questionnaires were collected. The questionnaires covered 31 provinces in China, and the multi-stage sample strategy in CFPS is well-designed as stated in [37]; thus, CFPS can be considered nearly nationally representative.

Considering the particularity in terms of concepts, classifications, survey content, and statistical measures of agricultural work, agricultural workers are not involved in this study. Furthermore, since CFPS records the income of self-employment by family, it is difficult to separate personal income from the household income. Thus, this study only involves non-agricultural employees, and the self-employment observations are not included. From 37,354 self-administered questionnaires, an initial data set was firstly established containing all the employees that took the depression test completely and were aged between 18–65. Then, a tag variable was added to mark if one observation belonged to informal employment or not. The rules for distinguishing between formal and informal employment are shown as follows:

Employees who sign labor contracts and have social insurance (including endowment insurance and medical insurance) are considered to be formal employment.

Considering the Chinese unique housing provident fund system, employees working in state-owned enterprises, government agencies, and public institutions are regarded as having formal employment as long as they have the housing provident fund, whether they sign labor contracts or not. In fact, many lifelong employees in the public sectors did not sign labor contracts in the past; only new employees and temporary workers signed labor contracts, oppositely.

On the basis of the above screening rules, if an employee signs a labor dispatch or labor intermediary contract, which means the organization providing the labor contract is inconsistent with the organization he/she actually works for, this observation is regarded as informal and deleted from the formal data set.

Abnormal observations are excluded manually if the key variable is missing or exceeds normal range.

After the above steps, a total set of 8893 observations was established, among which 2984 observations were identified as formal employment, and the remaining 5909 observations were identified as informal employment. Comparing with other statistics reported in [3,4,34], the proportion of informal employment was a bit higher (66.4%) because the screening rules of formal employment in this study are relatively strict.