Education, Experience or Discrimination? A Path Analysis of Mobile Population Wa
开篇：润墨网以专业的文秘视角，为您筛选了一篇Education, Experience or Discrimination? A Path Analysis of Mobile Population Wa范文，如需获取更多写作素材，在线客服老师一对一协助。欢迎您的阅读与分享！
Abstract
For the purpose of estimating the determinants of mobile population wage gaps in China’s urban labor market， this thesis conducts a path analysis on the variables which have significant effects on China’s labor market using the date provided by the CHIP. The empirical analysis shows that the main factors affecting labor income are workers’ education， experience， gender， marital status， industry and occupation. Mobility does not directly affect personal income. But it can impact on the possibility of workers entering monopoly industry and formal occupation， and consequently have an indirect effect on income. Labor market discrimination against mobile population is a significant factor which causes the wage gap in China’s urban labor market.
Key words： Education； Experience； Discrimination； Mobile Population； Income Disparities
Yang Yanjun （2013）. Education， Experience or Discrimination？ A path analysis of mobile population wage Gaps in China’s Urban Labor Market. Higher Education of Social Science， 5（1）， -0. Available from： URL： http：///index.php/hess/article/view/j.hess.1927024020130501.1083
DOI： http：///10.3968/j.hess.1927024020130501.1083
INTRODUCTION
Since the reform and opening up， the operating mechanism of China’s urban labor market has experienced a significant change and the size of the mobile population increased significantly. The mobile population has become a real part of the urban population. In this process， their employment situation has also been improved， but some problems， such as low quality of employment， low income levels remain widespread. This thesis aims to discuss the determinants of mobile population wage gaps in China’s urban labor market.
There are different theoretical models which may be relevant for the analysis of the determinants of labor income. The most commonly used is the Mincer Model which presented by Mincer （1958， 1970）. The standard equation of the model shown as follow：
Where the variable Inc represents the annual income of workers， Sch represents years of education， Epr represents the work experience of workers， a is the intercept of the equation， b is the return to human capital， c and d are two different rates of return to people’s work experience， and is the corresponding disturbance vector.
According to the current literature， we see that most of empirical researches about the worker income are based on multiple linear regression of the Mincer equation. However， this method has two deficiencies： on the one hand， such studies ignore other income-related variables， such as personal capacity and jobs， on the other hand， multiple linear regression can only reflect the direct relationship between the dependent and independent variables.
In fact， the variables that affect the income of workers in many cases are not independent of each other. The independent variable can not only affect the dependent variable by acting directly， but also can have an indirect effect on the dependent variable. To clarify the direct and indirect effects between these variables， we can use the path analysis.
1. THEORY FRAMEWORK
Path analysis is an extension of the regression model， used to test the fit of the correlation matrix against two or more causal models which are being compared by the researcher. The model is usually depicted in a circle-and-arrow figure just as Figure 1.
Figure 1
Path Analysis of the Causal Relationship Between the Variables
The single-headed arrows in the figure indicate causation. And the researcher will do a regression for each variable in the model as a dependent on others which the model indicates are causes. The regression weights predicted by the model are compared with the observed correlation matrix for the variables， and a goodness-of-fit statistic is calculated. The best-fitting of two or more models is selected by the researcher as the best model for advancement of theory.
A path model is a diagram relating independent， intermediary， and dependent variables. And the arrows between the variables are causal paths. Single arrows indicate causation between exogenous or intermediary variables and the dependent（s）. Double arrows indicate correlation between pairs of exogenous variables.
The model in Figure 1 is a standard path model which has correlated exogenous variables F1， F2 and F3， and endogenous variables D. The causal paths relevant to variable D are the paths from F1 to D， from F2 to D， from F3 to D which represent the direct causes， and the paths reflecting indirect causes which include the paths from F2 to F3 to D， from F2 to F1 to D. This model is specified by the following path equations：
Equation1： D=a11F1+ a12F2+ a13F3+e1
Equation2： F1= a21F2+ a22F3+e2
Equation3： F3= a31F2 +e3
Where aij are the regression coefficients and their subscripts are the equation number and variable number， thus a12 is the coefficient in Equation 1 for variable 2， which is F2.
2. EMPIRICAL ANALYSIS
2.1 The Date and Variables
The data we used in the paper is provided by the CHIP （Chinese Household Income Project）. The purpose of this project was to measure and estimate the distribution of income in both rural and urban areas of the People’s Republic of China. Data were collected through a series of questionnaire-based interviews conducted in rural and urban areas in 1988， 1995， and 2002. Individual respondents reported on their economic status， employment， level of education， sources of income， household composition， and household expenditures. To meet the needs of the analysis， we filter the survey samples based on the worker’s age， job and residence before the empirical study. Then， we delete the samples which information are missing and finally get 9127 samples.
Table 1
Income Comparison Between Mobile Population and Local Population
Variables Minimum Maximum Mean Std.Dev. Percentiles
25 50 75
Mobile population 0 100000 9004.77 9160.378 4407 6955 10575
Local population 0 130000 12087.84 8697.796 6400 10198 15000
Table 1 shows the income comparison between mobile population and local population of 9127 samples from CHIP， which means that the average income level of the mobile population is significantly lower than local residents. And even more importantly， the income disparity within mobile population is also significant.
Table 2
The Main Characteristics of the Selected Variables
Variables Meaning Type Value
Income Annual income of workers Continuous
Education Years of education of workers Continuous
Experience Work experience Continuous
Marriage The marital status of workers Dummy Unmarried=1
Married=0
Gender Workers' gender Dummy Male=1
Female=0
Mobility Population movements Dummy Local=1
Mobile=0
Area The area of population outflow Dummy Urban=1
Rural =0
Industry Industry of workers Dummy Monopoly=1
Competitive=0
Occupation Occupation of workers Dummy Formal =1
Informal =0
To estimate the determinants of the income disparities， we select several variables include workers’ education， experience， marital status， industry， occupation， gender and area， which may have significant affect on workers’ income. The main characteristics of the selected variables are shown in Table 2.
Multiple Linear Regression of the Mincer Equation
To estimate the direct relationship between the variables， we take stepwise multiple linear regression of the Mincer equation. The independent variable is worker’s annual income， the dependent variables as shown in Table 2 which include worker’s education， experience， industry， occupation， gender， marital status and mobility. The result shows as Table 3. Model 1 is the directly regression analysis results of Mincer equation and model 2 is the results of stepwise multiple linear regression of extended Mincer equation.
Table 3
Multiple Linear Regression of the Mincer Equation
Model B Std. E Beta t Sig. 95% Confidence Interval for B
Lower Upper
1 Constant 7.771 0.036 216.017 0.000 7.701 7.842
Education 0.080 0.002 0.358 37.024 0.000 0.076 0.084
Experience 0.030 0.003 0.423 11.679 0.000 0.025 0.035
Experience 2 0.000 0.000 -0.112 -3.089 0.002 0.000 0.000
2 Constant 7.795 0.036 217.245 0.000 7.725 7.866
Education 0.057 0.002 0.255 25.245 0.000 0.053 0.062
Experience 0.022 0.003 0.309 7.577 0.000 0.016 0.027
Experience 2 0.000 0.000 -0.082 -2.121 0.034 0.000 0.000
Industry 0.265 0.014 0.193 19.285 0.000 0.238 0.292
Occupation 0.158 0.015 0.106 10.665 0.000 0.129 0.187
Gender 0.145 0.013 0.108 11.539 0.000 0.121 0.170
Marriage 0.085 0.024 0.041 3.580 0.000 0.038 0.132
From statistical indicators of model 1 and model 2， we can see model 2 is much better than model 1. And model 2 shows that among the factors listed in Table 2， mobility and area have no direct impact on the annual income of workers. Industry， occupation and gender attributes of workers are the most significant factors that directly affect their income.
2.3 Determinants of Workers’ Industry and Occupation
From the analysis above， we can get the conclusion that mobility of the population does not directly affect their income， which is inconsistent with the relevant data in Table 1. Accordingly， we believe there may be intermediate factors， through which the attribute of mobility can have an indirect effect on workers’ income. However， industry and occupation can only be used as intermediate variables in all factors. So we analyze the determinants of workers entering monopoly industry and formal occupation respectively.
Table 4
Logistic Model on the Determinants of Workers Entering Monopoly Industry
Variable Model 1 Model 2 Model 3 Model 4
Constant -4.471***
1088.420 -4.890***
600.158 -4.891***
600.280 -5.088***
125.167
Educaiton 0.302***
1053.902 0.304***
1062.637 0.303***
1056.151 0.302***
1042.567
Experience 0.025***
98.228 0.011**
4.397 0.011**
4.326 0.011**
3.894
Gender 0.015
0.047
Mobility 9.471**
Mobility 1 0.229**
4.145
Mobility 2 0.089
0.488
Mobility 3 -1.377
0.307
Overall Percentage 69.1 69.0 69.0 69.1
Chi-square 1308.721 1317.038 1317.145 1330.633
-2 Log likelihood 10917.602 10909.285 10909.178 10895.690
Cox & Snell R2 0.134 0.134 0.134 0.136
Nagelkerke R2 0.181 0.182 0.182 0.184
HL Test 26.405 21.201 23.290 22.138
Note： The first line of each cell is the estimated parameter of corresponding variable， the second line is Wald statistic， and symbol ***， **， * respectively indicates that the parameter is significant at 1%， 5% and 10% level.
Mobility is a multi-categorical variable which representative of different types of population which includes local rural population， local urban population， mobile urban population and mobile rural population. Therefore， for this variable， we set up three dummy variables include moblie1， mobile 2 and mobile 3， which representative of local rural population， local urban population and mobile urban population respectively.
Table 4 shows the results of Logistic model on the determinants of workers entering monopoly industry. From the statistical indicators of each model， we can learn that Model 4 is the best model fitting the data relationships， which means mobility have significant effects on workers entering monopoly industry. From the analysis above， we believe that workers’ industry directly affect the level of income. Consequently， mobility can have an indirect effect on personal income by this way.
Table 5
Logistic Model on the Determinants of Workers Access to Formal Occupation
Variable Model 1 Model 2 Model 3 Model 4 Model 5
Constant -2.697***
445.919 -4.348***
96.336 -4.457***
100.612 -4.336***
95.478 -3.494***
53.751
Educaiton 0.068***
598.481 0.065***
544.374 0.062***
487.299 0.054***
286.359 0.075***
192.358
Experience 0.207***
525.336 0.198***
474.700 0.196***
459.468 0.199***
469.672 0.194***
439.139
Mobility 81.890*** 82.322*** 89.341*** 87.189***
Mobility 1 1.834***
18.119 1.866***
18.683 1.936***
20.183 1.908***
19.328
Mobility 2 0.299
0.354 0.347
0.474 0.338
0.453 0.341
0.454
Mobility 3 -0.408
0.480 -0.400
0.458 -0.390
0.439 -0.451
0.583
Gender 0.299***
35.541 0.321***
40.561 0.332***
43.152
Marriage 0.489***
36.762 0.598***
51.153
Overall Percentage 74.3 74.6 74.5 74.6 74.7
Chi-square 1022.570 1126.692 1162.212 1198.615 1223.318
-2 log likelihood 9904.257 9800.136 9764.671 9728.214 9703.510
Cox & Snell R2 0.106 0.116 0.120 0.123 0.125
Nagelkerke R2 0.152 0.166 0.171 0.176 0.180
Note： The first line of each cell is the estimated parameter of corresponding variable， The second line is Wald statistic， and symbol ***， **， * respectively indicates that the parameter is significant at 1%， 5% and 10% level.
Table 5 shows the results of Logistic model on the determinants of workers access to formal occupation and Model 5 is the best model fitting the data relationships. Model 5 shows that mobility have significant effects on workers access to formal occupation， and affect personal income in the same way of industry.
2.4 The Path Analysis of the Mobile Population Wage Gaps
The preceding analysis means： some of the variables which influence personal income diversity are direct influence factors while another part of them are indirect factors. To clarify the direct and indirect effects between these variables， we can use path analysis.
Table 6
The Result of Path Analysis Model on Workers’ Income
Varaibles Edu Exp Gen Mar Occu Ind Mob
Direct effect Ind A11 A12 A17
Occu A21 A22 A23 A24 A27
Income A31 A32 A33 A34 A35 A36
Indirect effect Income A11，A36 A12，A36 A17，A36
A21，A35 A22，A35 A23， A35 A24，A35 A27，A35
According to the models in Table 3， Table 4 and Table 5， we draw a table which clearly shows the relationship between the variables just as Table 6. The model in Table 6 is a path model which has correlated exogenous variables Education， Experience， Gender， Marrage， Occupation， Industry， Mobility， and endogenous variables Income. The causal paths relevant to variable Income are the paths from Education， Experience， Gender， Marriage， Occupation， and Industry to Income respectively which represent the direct causes， and the paths reflecting indirect causes which include the paths from Mobility to Industy to Income， from Mobility to Occupation to Income. This model is specified by the following path equations：
Equation1： Ind=A11Edu+ A12Exp+ A17Mob
Equation2： Occu=A21Edu+ A22Exp+ A23Gen + A27Mob
Equation3： Income=A31Edu+ A32Exp+A33Gen+ A34Mar+ A35Occu+ A36Ind
Where Aij are the regression coefficients and their subscripts are the equation number and variable number， thus A11 is the coefficient in Equation 1 for Edu. From the model， we can see： Although mobility does not directly affect personal income， it can impact on the possibility of workers entering monopoly industry and formal occupation， and consequently have an indirect effect on income.
CONCLUSION
As we noted at the outset， although the mobile population has become a real part of the urban population， some problems， such as low quality of employment， low income levels remain widespread. According to the current literature， most empirical researches about the worker income are based on multiple linear regression of the Mincer equation. In fact， the variables that affect the income of workers in many cases are not independent of each other. To clarify the direct and indirect effects between these variables， we use the path analysis using the date provided by the CHIP.
The empirical analysis results indicate that the main factors affecting labor income are workers’ education， experience， gender， marital status， industry and occupation. Mobility does not directly affect personal income. But it can impact on the possibility of workers entering monopoly industry and formal occupation， and consequently have an indirect effect on income. Labor market discrimination against mobile population is a significant factor which causes the wage gap in China’s urban labor market.
REFERENCES
Alicia Adsera， & Barry R. Chiswick （2006）. Are there gender and country of origin differences in immigrant labor market outcomes across European destinations？ Journal of Population Economics， （3）， 495-526.
Alwin， Duane F.， & Robert M. Hauser （1975）. The decomposition of effects in path analysis. American Sociological Review， 40（Feb.）， 37-47.
Amelie Constant， & Douglas S Massey （2005）. Labor market segmentation and the earnings of German guestworkers. Population Research and Policy Review， （5）， 489-512.
Baron， R. M.， & Kenny， D. A. （1986）. The moderator-mediator variable distinction in social sociological research： Conceptual， strategic， and statistical considerations. Journal of Personality and Social Psychology， 51， 1173-1182
Becker， Gary S.， & Chiswick， Barry R. （1966）. Education and the distribution of earnings. The American Economic Review. 56（1/2）， 358-369.
Ferrer， Ana， & Riddell， W. Craig （2008）. Education， credentials， and immigrant earnings. Canadian Journal of Economics， 41（1）， 186-216.
Lynch， L. M. （1989）. The youth labor market in the eighties： determinants of re-employment probabilities for young men and women. The Review of Economics and Statistics， 71（1）， 37-45.
Nina Martin， Sandra Morales， & Nik Theodore （2007）. Migrant worker centers： contending with downgrading in the low-wage labor market. Geo Journal， （2）， 155-165.
Mincer， Jacob （1958）. Investment in human capital and personal income distribution. Journal of Political Economy， 66（4）， 281-302.
Mincer， Jacob （1970）. The distribution of labor incomes： a survey with special reference to the human capital approach. Journal of Economic Literature， 8（1）， 1-26.
Helena Skyt Nielsen， Michael Rosholm， Nina Smith， Leif Husted （2004）. Qualifications， discrimination， or assimilation？ An extended framework for analysing immigrant wage gaps. Empirical Economics， （4）， 855-883.
Prentice， Ross （1976）. Use of the logistic model in retrospective studies. Biometrics， 32（3）， 599-606.
Schultz， T. Paul （1982）. Lifetime migration within educational strata in Venezuela： estimates of a logistic model. Economic Development and Cultural Change， 30（3）， 559-593.
Smith， Peter M.， & Mustard， Cameron A. （2010）. The unequal distribution of occupational health and safety risks among immigrants to Canada compared to Canadian-born labour market participants： 19932005. Safety Science， 48（10）， 1296-1303.
Solomon William， Polachek （1981）. Occupational self-selection： A human capital approach to sex differences in occupational structure. The Review of Economics and Statistics， 63（1）， 60-69.
Wright， S. （1921）. Correlation and causation. Journal of Agricultural Research， 20， 557-585.
Wright， S. （1934）. The method of path coefficients. Annals of Mathematical Statistics， 5， 161-215. The seminal article.