*4.1. Housing Value Model*

Because independent variables are mostly dummy variables, the basic condition of the model and the result of control variables are displayed first in Table 2.

In general, because of the proper selection of the covariate, R2 (coefficient of determination) in the OLS model is over 0.5, which indicates that the model fits the data well. Because the values of different variables in the questionnaires are not exactly the same, 1648 cases were eventually selected for the model. In contrast to the life course theory, age and housing value do not increase first and then decrease. Age basically exerts zero influence on the housing price, which probably occurs because marriage status and family member numbers are controlled. In terms of gender, the income returns on housing for males compared with females is lower. A possible reason for this result is that the gender advantages of males are demonstrated in other variables, which—in turn—serve as the control variables of gender. Regarding the difference among provinces, the municipalities of Shanghai and Guangdong Province have a greater advantage on housing price than other provinces, and Jilin Province ranks lowest. The model results are listed, respectively, according to independent variables in Table 3.


**Table 2.** Basic condition of the housing price model and control variables.

\*\*\* *p* < 0.001, \*\* *p* < 0.01, \* *p* < 0.05, ! *p* < 0.1 Double underline is *p* < 0.05. Single underline is *p* < 0.1.

**Table 3.** Time factor of housing price model.


\*\*\* *p* < 0.001, \*\* *p* < 0.01, \* *p* < 0.05.

As can be generally observed from the OLS model (Table 3), the time of purchase or construction continues to have a significantly positive effect on housing value after controlling other variables. Compared with individuals who entered the property market before 1998, individuals who entered the property market from 1999 to 2003, 2004 to 2008, and after 2009 have an increased housing value of 39.0%, 40.6%, and 48.9%, respectively. In other words, the later an individual acquires housing, the higher the housing price. This phenomenon can be attributed to the ever-increasing housing price. Due to the ongoing process of China's market transition, the more fully developed the marketization, the higher the housing value. Hypothesis 4a is thus proven true.

Additionally, such a positive effect varies at different quantile levels. To study the difference and changing trend of the effect in question, a quantile regression coefficient line graph is used. In Figure 1, 19 quantiles are at an interval of 0.05 on the *x*-axis. There are quantile regression coefficients on the *y*-axis. Only those regression coefficients with a *p*-value less than 0.1 are marked.

At the quantile levels from 0.1 to 0.25, which are mainly low-price houses, individuals who enter the real-estate market later obtain higher returns. One possible explanation for this result is that low-price houses are mostly government-subsidized housing units. Due to the commodification reform, houses acquired after 1999 can better satisfy the low-end living needs than before. With the increase of quantile, the value of low-price houses decreases gradually. At the quantile levels from 0.4 to 0.5, the housing value of individuals entering the property market from 2004 to 2008 is slightly lower than that from 1999 to 2003. This result shows that, in the early days of market transition, before a substantial amount of private capital flowed into the property market, the acquisition of middle-end housing could result in some advantages. At the quantile levels from 0.85 to 0.9, a number of high-price houses emerge after 2009 compared with 1998. Such a phenomenon is rare before 2009.

**Figure 1.** Quantile regression coefficient line: time factor of housing price model.

In terms of organizational factor, in Table 4 and Figure 2, the OLS model shows that different work units continue to exert influence on the distribution mechanism of housing, if not a great influence in the redistribution era. Unexpectedly, the housing value of collective enterprises ranks the highest, and the regression coefficients of other enterprises are negative. This is because of the many personal factors under control, and it is different from the results of early empirical studies. Based on the statistics from 1999 [26,36], the income of those from collective enterprises was substantially lower than from private enterprises, the party, and administration organs because collective enterprises were stuck between redistribution and the market. However, in housing distribution in recent years, the identity of collective enterprise has become an advantage rather than a disadvantage. Because party and administration organs do not show significance, and the difference between state-owned enterprises, public institutions, and private enterprises is small, Hypothesis 3a is not directly proven.


**Table 4.** Organizational factor of housing price model.

\*\*\* *p* < 0.001, \*\* *p* < 0.01, \* *p* < 0.05, ! *p* < 0.1.

Although the difference in the regression coefficients among work organizations is small in the OLS model and at quantile levels from 0.4 to 0.55, the regression coefficient of state-owned enterprises is remarkably lower than other types of organization at the quantile levels from 0.1 to 0.3. In general, the regression coefficient of state-owned enterprises is lower than that of private enterprises. Additionally, the regression coefficient of public institutions is higher than that of state-owned enterprises and lower than that of private enterprises. The *p*-values of party and government organs, as well as self-employed individuals, are relatively low; thus, an overall pattern is difficult to observe. In general, Hypothesis 3b is proven partly false even though the housing prices of party and government organs and public institutions are higher than those of other organizations at high quantile levels from 0.65 to 0.75; this is illustrative of institutional inertia. However, the regression coefficient of state-owned enterprises ranks the lowest among low-quantile housing, which reflects the brunt of market transition toward old institutional arrangements. As state-owned enterprises have borne most of the brunt, party and government organs and public institutions remain advantageous in middle- to high-price housing.

**Figure 2.** Quantile regression coefficient line: time factor of housing price model.

In terms of human capital factors, Table 5 and Figure 3 show that, the higher the education level, the higher the housing value. There is a remarkable advantage in college education compared with other levels of education, which corresponds to the conclusion of the OLS model. Therefore, Hypotheses 1a and 1b are proven true. With the increase in quantile, the two variables—income and education level—tend to decrease. This result demonstrates two things. First, as a type of investment, houses, especially high-price houses with a quantile over 0.7, are not directly related to income but related to the investment ability of investors. Second, regarding the acquisition of low-price and middleprice houses with a quantile below 0.65, high-school education has a greater advantage than secondary school education. However, such an advantage becomes less significant when the quantile is over 0.7, and sometimes secondary school education tends to be more rewarding. In terms of human capital, the returns on education mainly differ between individuals with or without a college education. The returns on income are higher in low-value and middle-value houses.


**Table 5.** Human capital factor of housing price model.

\*\*\* *p* < 0.001, \*\* *p* < 0.01, \* *p* < 0.05.

**Figure 3.** Quantile regression coefficient line: human capital factor of housing price model.

In terms of political capital in Table 6, the housing value of individuals with a title of section chief or above is 32.0% higher than that of ordinary workers, 15.5% higher than that of senior technicians, and 27.8% higher than that of junior management. Hypothesis 2a is thus proven true. Because the *p*-value of party membership is mostly low and even negative in some quantiles, the influence of party membership on housing value is exerted through other factors, such as occupation, work unit, and education level. When the aforementioned factors are under control, party membership is not as significant as in models with fewer variables. Sometimes, the *p*-value of party membership in the quantile regression is negative.


**Table 6.** Political capital factor of housing price model.

\*\*\* *p* < 0.001, \*\* *p* < 0.01, \* *p* < 0.05, ! *p* < 0.1.

The returns of housing value for senior cadre rank are higher than those for junior cadre rank, but the difference is not substantial. The influence and changing trend of political capital in different quantiles can be clearly seen in Figure 4. The occupation variable levels off in different quantiles. However, at the quantile levels from 0.75 to 0.9 and

from 0.2 to 0.55, organization clerks have a higher return than other occupations, which has rarely been observed in the literature.
