STATA Quantile Regression Analysis -- 2

This project seeks a highly qualified professional with extensive experience in econometrics and statistical regression modeling, particularly utilizing STATA software. The project involves a comprehensive analysis of employee wage data using quantile regression and decomposition techniques, both with sample selection adjustments. OBJECTIVE 1.) Determine the impact of various characteristics of workers on their wage in the Sectors (regular and casual separately) using the BEST sample selection adjustment technique along a conditional quantile regression model, and then decompose the gender wage gap (between males and females) with the same sample selection adjustment technique using Melly’s Machado-Mata decomposition for regular and casual separately. 2.) Determine the impact of various worker characteristics on their wage in the Sectors (regular and casual separately) using centred regression for different percentiles. Then, the gender wage gap (between males and females) will be decomposed using Oaxaca-Blinder Decomposition (or Machado-Mata-Melly Decomposition) using this centred regression coefficient for different percentiles for regular and casual sectors separately. WORK EXPLANATION 1.) Sample Selection Adjustment: It addresses a potential bias, i.e., we only observe wages for people who are employed. But factors influencing labor force participation (like childcare responsibilities) might also affect wages. Women might be more likely to be out of the workforce for these reasons, leading to a biased sample of employed women with potentially lower wages. This can underestimate the true gender wage gap, particularly at certain points in the wage distribution. Methodology for incorporating sample selection adjustment in quantile regression: A separate model is estimated to predict the probability of being employed, considering factors like education, marital status, and the presence of young children. The estimated probability of employment from the first model is then used to weigh the wage regression in the second model. This adjusts for the underrepresentation of certain groups (like women with young children) who might have lower wages but are not observed in the wage data. Note: The variable(s) required for the sample selection can be determined from the variable’s description. The concept here is to adjust the sample based on entering the labor force, like marital status affecting the choice of females to enter the labor force. 2.) Centred Regression Model explanation The attached Research Paper explains an alternative estimation of the marginal effects for Recentered Influence Functions (RIF)-Ordinary Least Square (OLS) for interpreting RIF regressions. In this, post-estimation restricted least squares (RLS) regression analysis is combined with centered continuous variables to provide a regression output that should be easier to understand and interpret. Also, the RIFs are used to analyze unconditional partial effects on unconditional quantiles in the regression analysis framework. WORK PLAN 1.) For the first objective, the effectiveness of these three sample selection adjustment methodologies in quantile regression modelling framework is to be tested, and the best one is to be further used in “decomposition” analysis. The three types of sample selection adjustment methodologies are: • Albrecht et al. (Research Paper): This procedure combines a semiparametric binary model for the participation equation with a linear quantile regression model for the wage equation and follows Buchinsky's (1998) approach. The research paper is based on this approach. • Arhomme (Stata Package): ssc install arhomme • Qregsel (Stata Package): ssc install qregsel To determine the impact of various characteristics of workers on their wage in the regular and casual sectors using “sample selection correction for conditional quantile regression”: Run a quantile regression model with sample selection adjustment for the two Sectors (Regular and Casual separately). The code for these is: [Sectors = 1 and 2]. To decompose the gender wage gap (between males and females) with the same sample selection adjustment technique for regular and casual separately: Run Melly’s Machado-Mata Decomposition (available on the page: [login to view URL]) incorporating the best-suited sample selection correction adjustment procedure. 2.) For the second objective, the regression methodology proposed in the attached research paper is to be followed to determine the impact of various characteristics of workers on their wage in the Sectors (regular and casual separately) using centred regression on different percentiles (10th, 25th, 50th, 75th and 90th can be considered). As a next step, you need to get the Oaxaca-Blinder decomposition technique (or Machado-Mata-Melly decomposition) for each percentile using centered regression models. A somewhat similar methodology, i.e., Oaxaca-Blinder decomposition using RIF (through a user-defined oaxaca_rif function in STATA), is explained in this publicly available paper: [login to view URL] For my work, a similar type of decomposition is to be done, but using the “centered regression model” instead of RIF regression. Dependent Variable: Log_Wages Predictors/Independent Variables: Sex, Residential Area, Age, Religion, Social Group, Marital Status, General Education, Technical Education, Occupation Type, Industry, Employment Type, Enterprise (These can be increased/decreased depending on the model efficiency.) ATTACHMENTS: Complete and detailed workplan to follow for the project, and supporting research papers. DELIVERABLES: STATA code for the above and output with some interpretation. EXPECTED DEADLINE: Approximately one week. It can be extended based on the work requirements. Please feel free to message me for further details regarding the work. I will be fully committed to working side-by-side. The budget and timeline can be discussed further.