# Results of the linear regression model

We used linear regression models with overall life satisfaction as a continuous outcome variable, and sequentially introduced market competition, collective consumption, environment, and individual quality of life factors into the explanatory variables to examine the changes in model explanatory power and independent variables. The model is also weighted for age and sex to render the sample more consistent with the age and sex distribution of the sixth national census. The linear regression model is shown in Table 11.2. The results show that the variance in life satisfaction explained by the model is significant from model 1 with only absolute income to model 6 with all variables included, and that the variance in life satisfaction explained by the model increases sequentially.

First, the total household income is introduced to measure the absolute income, where Model 1 is able to explain 3.7% of the variance in life satisfaction, the effect of total household income is significant, and a higher household income corresponds to a higher life satisfaction of residents. The estimation of the model uses an adjusted R-square, which allows examining of the changes in the explanatory power of the model as variables are introduced. Relative income variables are then introduced, that is, economic status in comparison to one’s own past and, in comparison, to those around oneself. The results show that absolute income is still significant, and both measures of relative income are significant. The more the standard of living improves, the more satisfied respondents were with life. Higher self-evaluation of local socioeconomic status also corresponds to greater

 Variable Variable value Sample size Mean value Standarddeviation Individual level Overall life satisfaction 1 ~ 10,1 = very unsatisfied; 10 = very satisfied 5,574 6.842 1.819 Household income, previous year (yuan) 0- 10,100,000 4,083 76,820.430 197,255 Natural log of household income from previous year 4.277- 16.130 4,055 10.817 0.934 Change in standard of living compared to 5 years ago 1 = decreased greatly; 2 = decreased somewhat; 3 = no change; 4 = increased somewhat; 5 = increased greatly 5,542 г.126 0.950 Self-evaluation of individual economic status in local area 1 = lower stratum; 2 = lower-middle stratum; 3 = middle stratum; 4 = upper-middle stratum; 5 = upper stratum 5,490 2.357 0.893 Healthcare satisfaction level 1 = very unsatisfied; 10 = very satisfied 5,278 6.553 2.371 Evaluation of government's job in providing economic and affordable housing for lower-middle income earners 1 = very poor; 2 = not very good; 3 = relatively good; 4 = very good 4,467 2.406 0.786 Degree of fairness in distribution of wealth and income 1 = very unfair; 2 = not very fair; = relatively fair; 4 = very fair 5,171 2.149 0.732 Level of satisfaction for environment of place of residence 1 - 10,1 = very unsatisfied; 10 = very satisfied 5,582 6.250 2.009 Stress from burdensome medical expenses 0 = not stressed; 1 = stressed 5,549 0.278 0.448
 Stress from job loss/ family member unemployment /unstable employment 0 = not stressed; 1 = stressed 5,556 0.289 0.453 Gender 1 = male; 2 = female 5,583 1.542 0.498 Age 18-72 years 5,583 43.554 13.600 Marital status 0 = single (unmarried, divorced, or widowed); 1 = married (1st marriage, subsequent marriage, or cohabitation) 5,575 0.816 0.387 Education level 0 = elementary school/no education; 1 = middle school; 2 = higli-school/trade school/vocational school; 3 = junior college; 4 = undergraduate or higher 5,575 1.553 1.261 Household registration type 1 = local urban; 2 = non-local urban; 3 = local rural; 4 = non-local rural 5,562 2.111 1.132 Provincial level Per capita GDP (yuan) 2.315-10.011 5,583 5.054 1.904 Healthcare coverage 0.559 - 1 5,583 0.871 0.046 S02 emissions (unit: ten thousand tons) 0.419 - 164.497 5,583 81.560 39.011
 jModel 1 Model 2 ModeI 3 Model 4 Model 5 Model 6 Market competition factors: absolute income and relative income Natural log of total household income 0. 376*** (0. 033) 0. 183*** (0. 032) 0. 257*** (0. 038) 0. 266*** (0. 037) 0. 215*** (0. 037) 0. 163*** (0. 039) Change in standard of living compared to the past 0. 409*** (0. 034) 0. 305*** (0. 035) 0. 276 (0. 035) 0. 252*** (0. 035) 0. 250*** (0. 035) Self-evaluation of economic status in the local area 0. 517*** (0. 035) 0. 420*** (0. 039) 0. 377*** (0. 038) 0. 323*** (0. 038) 0. 290*** (0. 038) Collective consumption factors: social security, public services and social equity Healthcare satisfaction level 0. 147*** (0. 015) 0. 120*** (0. 014) 0. 109*** (0. 015) 0. 107*** (0. 014) Evaluation of government job on housing security 0. 272*** (0. 040) 0. 204*** (0. 040) 0. 198*** (0. 040) 0. 198*** (0. 040) Degree of fairness in distribution of income 0. 260*** (0. 043) 0. 209*** (0. 043) 0. 191*** (0. 043) 0. 207*** (0. 043) Environmental factors: evaluation of environment Satisfaction level towards environment of place of residence 0. 188*** (0. 017) 0. 182*** (0. 017) 0. 178*** (0. 017) Individual life quality: health and employment Stress from large medical expenses - 0. 387*** (0. 073) - 0.417*** (0. 073) Stress from job loss or unemployment of family member - 0. 310*** (0. 073) - 0. 237*** (0. 072) Control variables

1

 Female 0. 152[1] [2] (0. 056) Age - 0.081[2] (0. 016) Age squared 0. 001[2] (0. 000) Married 0. 307[2] (0. 090) Education level (with “elementary school or no education” as the reference) Middle school 0. 127 (0. 093) High-school, trade school or vocational school 0. 292[2] (0. 105) Junior college 0. 509[2] (0. 118) Undergraduate or higher 0. 655[2] (0. 124) Household registration type (with “non-local rural” as the reference) Local urban 0. 040 (0. 098) Non-local urban - 0. 002 (0. 139) Local rural 0. 071 (0. 103) Constant 20. 748[2] (0. 368) 20.108[2] (0. 347) - 0. 262 (0. 422) - 0. 864** (0. 418) 0. 276 (0. 438) 10. 746[2] (0. 558) Sample size R: 4,049 0. 037 3,980 0. 172 3,038 0. 257 3,037 0. 296 3,018 0. 309 3,001 0. 329

Note: Numbers in parentheses represent stable standard errors of the coefficients.

life satisfaction. Model 2 has significantly improved explanatory power, and can explain the 17.2% difference in life satisfaction, indicating that relative income has a very important impact on the life satisfaction of urban residents.

Model 3 introduces three variables to measure collective consumption, namely the level of social security, the level of public services and the degree of social equity, and shows a significant improvement in explanatory power, with a 9% increase in explanatory power over the model that includes only the market competition factor. The original market competition factor remains significant, while the newly introduced collective consumption factors have also produced significant positive effects. The more satisfied respondents were with health care, the higher their evaluation of government work for housing security, and the more equitable the income distribution, the higher the satisfaction level of the residents.

Model 4 introduces variables that measure environmental factors. In this model we see that the higher the environmental satisfaction with the area of residence, the higher the overall satisfaction with life, and the difference is significant. The introduction of environmental factors increased the explanatory power of the model so that it was able to account for 29.6% of the variance. Model 5 incorporates individual quality of life factors, measured in terms of health and employment pressures on oneself and one's family, respectively. The effects are significant, with stress stemming from health care expenses and stress from having family member who is unemployed or who experienced job loss having a negative impact on life satisfaction.

Therefore, the model used in this article, combining market and non-market factors (collective consumption, environment, and individual quality of life), can explain nearly 31% of the variance, and the explanatory power of the model increases with the introduction of variables, suggesting that these factors all have an important impact on the life satisfaction of urban residents. In addition, the model results also demonstrate the robustness of the analytical framework constructed in this article.

Model 6 includes all explanatory variables and control variables; the explanatory power is increased to 33%, and all explanatory variables are significant at the 1% level. With the exception of household registration type, all of the other control variables—sex, age, age squared, marital status, and education level— were significant. Women had higher life satisfaction than men; those who were married or had partners had higher life satisfaction; those with an education level of junior high-school, high-school, junior college or undergraduate or above had higher life satisfaction than those with no education or only primary school level education. There is a U-shaped relationship between age and life satisfaction; calculations show that the critical point of the age effect is 41 years old, which is essentially at the threshold for middle-aged: older than young people yet younger than the elderly. As this threshold approaches, stress gradually increases, and life satisfaction gradually decreases. After the threshold is crossed, with the accumulation of experience and wealth and stability of family and career, life satisfaction gradually rises again. The impact of demographic characteristics is essentially consistent with the existing literature (e.g., Applelon and Song, 2008; Guan, 2010; Liu, et al„ 2012).

Market competition, collective consumption and environmental quality 199