*4.2. Housing Added-Value Model*

The housing added-value model adopts the same independent variables and methodology as the housing value model. The basic condition of the model and control variables is presented in Table 7.



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

The coefficient of determination in the OLS model is 0.533, which indicates that the model fits the data well. The model of control variables corresponds to the housing value model. Regarding time factor, the housing added-value model is different compared with the housing value model. The housing condition from 2004 to 2008 is not significant in Table 8. As houses before 1998 were relatively cheap and held for a longer period, new houses after 2009 had a lower added value than those before 1998. However, houses between 1999 and 2003 had a higher added value than houses before 1998. Therefore, Hypothesis 4b has been proven to be partly false.


**Table 8.** Time factor of housing added-value model.

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

From the distribution of quantile regression coefficients in Figure 5, small differences are observed in high-return housing after 2009 compared with high-return housing of 1998. This result indicates that low-price housing after 2009 has not fully appreciated and lags behind housing held for a longer time. Another discovery is that housing between 1999 and 2003 had a higher added value than housing before 1998. Taking the historical background into account, 1999 was the starting point of the commodification reform of housing, and premium housing resources that had been prevented from transaction were released into the redistribution system. Therefore, the purchase or construction of property at this time would result in a first-mover advantage. These have become the most value-added properties, even if they are not those selling at the highest price. Such an advantage is most remarkable at quantile levels from 0.05 to 0.15, indicating that, in low-return housing, the added value of purchase or construction from 1999 to 2003 is 23.7% to 61.4% higher than that of purchase or construction before 1998. Due to the commodification of housing, a number of people with low added-value housing have acquired the trading right of housing without much cost, allowing them to gain substantial added-value advantages. Additionally, such an advantage results in 20% more added value in other quantiles and rarely decreases as quantiles increase.

**Figure 5.** Quantile regression coefficient line: time factor of housing added-value model.

In terms of work unit type, from the overall OLS model in Table 9, collective enterprises have the highest value-added; this is similar to the total housing value model, which followed party and government organs and public institutions. However, the *p*-value of party and government organs is between 0.05 and 0.1, which is not steady. Regarding the level of added value, the lower the marketization level, the higher the added value. This phenomenon applies to all types of work units except for collective enterprises. In other words, the work units that do not have a fully developed marketization similar to party and government organs and public institutions tend to have higher added value. After considering the exception of collective enterprises, Hypothesis 3b is proven to be partly true.


**Table 9.** Organizational factor of housing added-value model.

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

From the changing pattern of quantile regression coefficients in Figure 6, a similar pattern can be observed. The group of self-employed individuals is the work unit type with the highest level of marketization and the lowest added value. In quantiles after 0.45, the middle- and high-return housing with high added value do not fit into the group of self-employed individuals. Although there is a breakpoint in coefficients of party and government organs, these factors remain advantageous in low- and middle-return housing on quantile levels from 0.35 to 0.45. The income of non-corporate organizations, such as party and government agencies and public institutions from real-estate appreciation, is between 0.25 and 0.85 points, which is greater than private and state-owned enterprises. In general, in housing quantiles with higher returns, the difference between non-enterprise organizations and collective enterprises tends to be smaller.

**Figure 6.** Quantile regression coefficient line: organizational factor of housing added-value model.

Regarding the human capital factor, from the OLS model in Table 10, the higher the education level, the higher the income and the higher the housing added value. However, the advantage of a college education is less remarkable than it is in the housing value

model. The housing added value of high-school education and college education is 68.3% and 58.4% higher, respectively, than elementary school education and below.


**Table 10.** Human capital factor of housing added-value model.

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

In the quantile regression model in Figure 7, the added value of college education is higher than that of other education levels at high quantile levels from 0.6 to 0.95. At lowquantile levels from 0.05 to 0.35, the difference between college education and high-school education is not substantial. The added value of high-school education is even higher than that of college education, which probably reflects that the human capital factor plays an insignificant role in low-return housing. Therefore, high-school education can also result in high returns on low-return housing. However, college education is a must to gain returns on high-return housing. With the increase in quantiles, the influence of income on housing added value decreases, which is the same as in the housing price model.

**Figure 7.** Quantile regression coefficient line: human capital factor of housing added-value model.

In Table 11, any variables in occupation and political capital factor are not significant, but the returns of organization clerks still rank the highest in terms of housing added value. The difference between the occupation of section chief and above and the occupation of organization clerk is smaller than that in the housing price model. The housing added value of section chief and above and organization clerk is, respectively, 35.1% and 44.9% higher than that of an ordinary worker.


**Table 11.** Political capital factor of housing added-value model.

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

In the quantile model in Figure 8, the *p*-value of party membership is partly negative at the low and middle quantile levels, which is similar to the housing value model. Therefore, Hypothesis 2b is proven to be partly false. The difference in the returns among various occupations is not large and decreases with the increase in quantiles. However, in the housing price model, the regression coefficients of various occupations in different housing prices are more or less steady. This result indicates that a negative correlation between the returns from the high-return property market and occupation, which is especially remarkable at quantile levels from 0.1 to 0.25.

**Figure 8.** Quantile regression coefficient line: political capital factor of housing added-value model.
