Qiang Li
Housing Space of Chinese People in 1993
Research question:
What were the factors that influenced individual’s housing space in China in early 1990s? Was individual income and employment status (economic status), individual’s education (human capital), individual’s work unit characteristics (institutional characteristic), or individual’s administrative and political status more important in determining one’s living condition, such as housing space?
Data:
China Housing Survey 1993, obtained from Inter-University Consortium for Political and Social Research (ICPSR).
Models:
Dependent variable: Space of housing (Square meters)
Quantitative Predictors:
1. Income determines housing space: regress housing space on income, control for number of people living in the house.
Nominal and Ordinal Predictors:
2. Individual’s education and employment status determine housing space: ANOVA analysis.
3. Individual’s political affiliation and administrative position determine housing space: ANOVA analysis.
4. Work unit characteristics (nature, industry, administrative rank, and government juris) determine one’s house spacing:
Data check:
1. For regression analysis, check the normality of variables, especially the dependent variable:
The results show that housing space and earnings are highly skewed. Therefore I made log transformation to make them normal.
2. For ANOVA Analysis, check 1) whether the dependent variable is normally distributed in every category within each predictors or 2) whether the sample size of every category is large to make the CLT applicable if the distribution is not exactly normal. 3) if largest sample standard deviation is no more than twice the smallest sample:
The results show that in most cases, log transformation of housing space is roughly normally distributed among each category of each predictor and the standard deviation of largest sample is no more than that of the smallest sample.
Only the variable of industry nature of present work unit has too many categories, and each category has small sample size, which is not always normally distributed. So I exclude it from the ANOVA analysis.
Analysis and Results:
1. Correlation:
Pearson correlation coefficient between log of housing space and log of earning is 0.07489, which is significant at .0008 level. (See output)
2. Regression:
Model: log of housing space = ß0 + ß1(number of Cohabiters)+ ß2(Log of earnings) + Error.
Regression coefficient of log of earning is .05143, and it is significant only at .0467 level. R square of the model is only .0772, which means that individual’s earning\ is not a strong predictor of one’s housing space. (See output)
3. ANOVA:
Except employment status, the overall test of the predictors of education, political affiliation, nature of work unit, administerial rank of work unit, and government juris of work unit are significant, which means that both individual’s education and party affiliation, and work unit related characteristics have significant influence on one’s housing space.
Based on pairwise means comparison test, the following list the significant different mean pairs among each predictor:
1) Education: between college degree and either no formal education, or elementary school, or high school degrees.
2) Political Affiliation: between communist party members and those without political affiliation.
3) Nature of work Unit: between state institutions and state enterprises.
4) Administerial rank of work unit: between no rank work unit and all other higher rank work units; between bureau or higher rank and sector or lower rank work unit.
5) Government juris of work unit: between central ministries and either municipal bureau, municipal company, or district company; between municipal government and district company, between district bureau and district company.
Conclusion:
In early 1990s, not long after Tian-an Men Square Accident in 1989, China still allocated housing through a planned system rather than a market system.
Even as a fact of today, the political reform is still a blurry construction in mind which lags far behind the fast development of economy. People are now obviously feeling the nervousness in all kinds of anomy. Even the CPC and the elites of China are struggling in themselves on how to balance the conflicting interests of the gainers and the losers so that to maintain the main political constitutions and social well-being, on the other hand, to steadily decrease all kinds of outdated restriction on individual freedom and civil rights.
In such a situation, as a result of the political legacy, the excessive control on social life in Mao’s era is reflected on the allocation of private spaces. Individual’s income (pure economic status) does not have a strong influence on the housing space. Most of individuals were still highly organized and controlled by work unit *in China at that time, and their housing was also allocated through their work unit. Thus, the characters of work unit matter a lot for one’s housing space. In addition, at that time in China, work unit was not simply an economic unit, it was rather a political, social and administerial unit, therefore, what really matters was the nature of work unit, whether it was state institution or state company, the administerial rank of the work unit, whether it had a rank or not, whether it had the highest rank or not, whether it was the governmental juris of work unit or not, and whether it was the most centralized or less centralized unit. Since some fields then had already been facing the impact of free economy, the difference (inside or outside of), level in or distance from the communist system were important, even are no less important today one can experience in other cases.
*work unit: Dan Wei in Chinese, as a unit on which political, economic and all other kinds of controls and resources were centralized, and to which individuals had to adhere themselves so that they could make their lives. It was also excessively powerful in its functioning in Mao’s era since it was the organizer of all aspects of social life.
SAS Program
*Dependent Variable: V1003 = "M sq, present housing" ;
** Descriptive Statistic: check for distribution;
proc freq data = china;
tables v1003;
run;
proc univariate plot data=china;
var v1003;
run;
*Quantitative Predictors;
**V0618 = "Present estimate, monthly earnings";
proc univariate plot data=china;
var v0618;
run;
**Create cohab = "Number of people living in present house", make log tranformation of housing space and earnings;
data newchina;
set china;
if v0140~=. then cohab=1 ;
if v0147~=. then cohab=2 ;
if v0203~=. then cohab=3 ;
if v0210~=. then cohab=4 ;
if v0217~=. then cohab=5 ;
if v0224~=. then cohab=6 ;
if v0231~=. then cohab=7 ;
if v0238~=. then cohab=8 ;
if v0245~=. then cohab=9 ;
lv1003=log(v1003);
label lv1003='log transformation of housing space';
lv0618=log(v0618);
label lv0618='log transformation of earnings';
run;
proc freq data=newchina;
tables cohab;
run;
proc univariate plot data=newchina;
var cohab;
run;
proc univariate plot data=newchina;
var lv1003;
run;
proc univariate plot data=newchina;
var lv0618;
run;
proc print data=newchina label;
var lv1003 cohab lv0618;
run;
**Correlation and Regression Analysis: Space of house (log) = monthly earnings (log), control for number of cohabitors;
options linesize=79 pagesize=60;
proc plot data=newchina;
plot lv1003*cohab='.';
plot lv1003*lv0618='.';
run;
proc corr data=newchina;
var lv1003 cohab;
var lv1003 lv0618;
run;
proc reg data=newchina;
model lv1003=cohab lv0618/clb;
run;
**predicted Values and residuals;
proc reg data=newchina lp;
model lv1003=cohab lv0618/p;
plot residual. * predicted./symbol = '.' hplots =2 vplots = 2;
run;
*Norminal and Ordinal Predictors;
** V0603 = "Present education"
V0604 = "Present employment status"
V0609 = "Present political affiliation"
V0610 = "Present adm position"
V0611 = "Present nature of work unit"
V0613 = "Present industry of work unit"
V0615 = "Present adm rank of work unit"
V0616 = "Present govt juris of work unit";
proc freq data=newchina;
tables v1509 v0603 v0604 v0606 v0609 v0610 v0611 v0613 v0615 v0616;
run;
*** ANOVA analysis: "Space of house" = v0603 "education";
1
Qiang Li
options linesize=60;
proc sort data=newchina;
by v0603;
run;
proc univariate plot data=newchina;
var lv1003;
by v0603;
run;
proc means data=newchina;
var lv1003;
by v0603;
run;
proc anova data=newchina;
class v0603;
model lv1003=v0603;
run;
proc anova data=newchina;
class v0603;
model lv1003=v0603;
means v0603/bon alpha = .05;
run;
*** ANOVA analysis: "Space of house" = v0604 "employment status";
proc sort data=newchina;
by v0604;
run;
proc univariate plot data=newchina;
var lv1003;
by v0604;
run;
proc means data=newchina;
var lv1003;
by v0604;
run;
proc anova data=newchina;
class v0604;
model lv1003=v0604;
run;
proc anova data=newchina;
class v0604;
model lv1003=v0604;
means v0604/bon alpha = .05;
run;
*** ANOVA analysis: "Space of house" = v0609 "Present political affiliation";
proc sort data=newchina;
by v0609;
run;
proc univariate plot data=newchina;
var lv1003;
by v0609;
run;
proc means data=newchina;
var lv1003;
by v0609;
run;
proc anova data=newchina;
class v0609;
model lv1003=v0609;
run;
proc anova data=newchina;
class v0609;
model lv1003=v0609;
means v0609/bon alpha = .05;
run;
*** ANOVA analysis: "Space of house" = v0610 "Present adm position";
proc sort data=newchina;
by v0610;
run;
proc univariate plot data=newchina;
var lv1003;
by v0610;
run;
proc means data=newchina;
var lv1003;
by v0610;
run;
proc anova data=newchina;
class v0610;
model lv1003=v0610;
run;
proc anova data=newchina;
class v0610;
model lv1003=v0610;
means v0610/bon alpha = .05;
run;
*** ANOVA analysis: "Space of house" = V0611 "Present nature of work unit" ;
proc sort data=newchina;
by v0611;
run;
proc univariate plot data=newchina;
var lv1003;
by v0611;
run;
proc means data=newchina;
var lv1003;
by v0611;
run;
proc anova data=newchina;
class v0611;
model lv1003=v0611;
run;
proc anova data=newchina;
class v0611;
model lv1003=v0611;
means v0611/bon alpha = .05;
run;
*** ANOVA analysis: "Space of house" = V0613 "Present industry of work unit" ;
proc sort data=newchina;
by v0613;
run;
proc univariate plot data=newchina;
var lv1003;
by v0613;
run;
proc means data=newchina;
var lv1003;
by v0613;
run;
proc anova data=newchina;
class v0613;
model lv1003=v0613;
run;
proc anova data=newchina;
class v0613;
model lv1003=v0613;
means v0613/bon alpha = .05;
run;
*** ANOVA analysis: "Space of house" = V0615 "Present adm rank of work unit" ;
proc sort data=newchina;
by v0615;
run;
proc univariate plot data=newchina;
var lv1003;
by v0615;
run;
proc means data=newchina;
var lv1003;
by v0615;
run;
proc anova data=newchina;
class v0615;
model lv1003=v0615;
run;
proc anova data=newchina;
class v0615;
model lv1003=v0615;
means v0615/bon alpha = .05;
run;
*** ANOVA analysis: "Space of house" = V0616 "Present govt juris of work unit" ;
proc sort data=newchina;
by v0616;
run;
proc univariate plot data=newchina;
var lv1003;
by v0616;
run;
proc means data=newchina;
var lv1003;
by v0616;
run;
proc anova data=newchina;
class v0616;
model lv1003=v0616;
run;
proc anova data=newchina;
class v0616;
model lv1003=v0616;
means v0616/bon alpha = .05;
run;
1
Qiang Li
SAS Output
The UNIVARIATE Procedure
Variable: V1003 (M sq, present housing)
Histogram # Boxplot
97.5+** 18 *
.* 4 *
.* 1 *
.* 6 *
.* 3 0
.** 17 0
.** 11 0
.**** 31 0
.** 16 0
.******** 64 |
.******* 55 |
.************** 124 |
.************** 122 |
.***************** 147 +-----+
.****************************** 269 | + |
.******************************************** 394 *-----*
.***************************************** 365 +-----+
.**************************************** 356 |
.********** 82 |
2.5+* 8 |
----+----+----+----+----+----+----+----+----
* may represent up to 9 counts
The UNIVARIATE Procedure
Variable: lv1003 (log transformation of housing space)
Histogram # Boxplot
4.7+** 11 0
.** 12 0
4.3+***** 31 |
.******** 53 |
3.9+****************** 120 |
.************************** 181 |
3.5+******************************* 211 +-----+
.**************************************** 279 | |
3.1+***************************************** 284 *--+--*
.****************************************** 293 | |
2.7+******************************************** 304 +-----+
.******************** 135 |
2.3+************** 98 |
.********* 58 |
1.9+** 10 |
.* 5 |
1.5+
.* 4 0
1.1+* 1 0
.
0.7+* 3 0
----+----+----+----+----+----+----+----+----
* may represent up to 7 counts
The UNIVARIATE Procedure
Variable: V0618 (Present estimate, monthly earnings)
Histogram # Boxplot
2050+* 1 *
.
1850+
.
1650+
.* 1 *
1450+
.* 1 *
1250+* 2 *
.* 2 *
1050+* 12 0
.* 3 0
850+* 8 0
.** 16 0
650+**** 47 |
.*********** 141 |
450+******************** 280 +-----+
.******************************** 435 | + |
250+************************************************ 662 *-----*
.************************* 346 |
50+*** 39 |
----+----+----+----+----+----+----+----+----+---
· may represent up to 14 counts
The UNIVARIATE Procedure
Variable: lv0618 (log transformation of earnings)
Histogram # Boxplot
7.7+* 1 0
.
.* 2 0
.* 4 0
.** 15 |
.** 9 |
.**** 26 |
6.3+*********************** 178 |
.************* 101 |
.********************************************* 357 +-----+
.******************************************* 338 *--+--*
.**************************************** 318 | |
.***************************************** 321 +-----+
.************************** 203 |
4.9+****** 44 |
.***** 40 |
.* 8 |
.** 11 0
.* 3 0
.** 9 0
.* 3 0
3.5+* 3 0
----+----+----+----+----+----+----+----+----+
· may represent up to 8 counts
Plot of lv1003*cohab. Symbol used is '.'.
‚ ‚
5.0 ˆ
‚
‚
‚
‚ . . . . .
l 4.5 ˆ . .
o ‚ . . . .
g ‚ . . . .
‚ ......
t ‚ ......
r 4.0 ˆ ......
a ‚ ......
n ‚ ......
s ‚ ......
f ‚ ......
o 3.5 ˆ ......
r ‚ ......
m ‚ ......
a ‚ ......
t ‚ ......
i 3.0 ˆ ......
o ‚ ......
n ‚ . . . . .
‚ ......
o ‚ . . . . .
f 2.5 ˆ . . . . .
‚ . . . .
h ‚ . . . . .
o ‚ . . .
u ‚ . . .
s 2.0 ˆ . .
i ‚ . .
n ‚ . .
g ‚
‚
s 1.5 ˆ
p ‚ . .
a ‚
c ‚
e ‚ .
1.0 ˆ
‚
‚
‚ .
‚
0.5 ˆ
‚
Šƒƒˆƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒˆƒƒƒƒƒƒƒˆƒƒ