The Size and Distribution of Hidden Household Income in China

The Size and Distribution of Hidden Household Income in China
The Size and Distribution of Hidden Household Income in China

Xiaolu Wang
National Economic Research
Institute
China Reform Foundation
C510 Guo Hong Building,
No. A-11 Muxidi Beili
Xicheng District, Beijing,
100038, China
wangxiaolu@neri.org.cn

Wing Thye Woo
Economics Department
University of California
Davis, California 95616, USA

Central University of Finance
and Economics, Beijing, China
wtwoo@ucdavis.edu

The Size and Distribution of Hidden
Household Income in China*

Abstract
Official Chinese data on urban household income are seriously
flawed because of significant underreporting of income by respon-
dents and non-participation by the high income groups in official
household surveys. We collected urban household income and
expenditure data in a way that increased their reliability and the
coverage of the wealthy. We utilized the well-known relationship
between Engel’s coefficient and income level through two differ-
ent approaches to deduce the true level of household income for
each of the seven Chinese income categories (lowest income, low
income, lower middle income, middle income, upper middle in-
come, high income, and highest income). We found that the ratio
of our estimated income to official income increased from 1.12
for the lowest income group to 3.19 for the highest income
group. Total household disposable income in 2008 is RMB
14.0 trillion according to the official data but RMB 23.2 trillion ac-
cording to our estimate; and 63 percent of the unreported in-
come went to the wealthiest 10 percent of urban households.
The income of the wealthiest 10 percent of Chinese households
is really 65 times that of the poorest 10 percent instead of the
23 times reported in the official data. The Gini coefficient is clearly
much higher than the usually reported figure of 0.47.

In one of the estimations, we had to drop the 76 wealthiest
households (1.8 percent of our sample) from the analysis because
there were no super-rich in the official data for us to match char-
acteristics with. We therefore still understate the income of the
highest income households. As the amount of unreported income
indicates the degree of corruption, it is troubling that it grew
91 percent in 2005–08 compared to the 71 percent growth in
gross national income. Serious institutional reforms must be en-
acted if corruption is not to derail economic development and
social harmony.

* This article is part of a research project of the Chinese Research
Society for Economic System Reform. We thank the many indi-

Asian Economic Papers 10:1

Institute of Technology

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

1. Introduction

It is easy to sing the praises of China’s economic performance. An average annual
GDP growth rate of 10 percent during the 1978–2006 period has raised GDP per ca-
pita by almost nine-fold over the period. The prevailing expectation in 2006 was that
China would continue to register impressive growth for some time to come.1 It was
therefore a big surprise at the end of 2006 when the Plenum of the Central Commit-
tee of the Communist Party of China (CPC) did not repeat what every Plenum had
proclaimed since the famous 1978 Plenum that the chief task of the CPC was eco-
nomic construction. This 2006 Plenum proclaimed instead that the chief task of the
CPC was the establishment of a harmonious society by 2020.2 The obvious implica-
tion from this new party line is that the present major social, economic and political
trends within China might not lead to a harmonious society or, at least, not lead to a
harmonious society fast enough.

We believe that this switch in emphasis occurred because the CPC has concluded
that social stability requires not just a high economic growth rate to keep unemploy-
ment low but also a growth pattern that diffuses the additional income widely; and
that the increase in income inequality has been too rapid. In the 1985–87 period,
China’s Gini coefªcient was about 30 percent.3 However, according to the Asian De-
velopment Bank (2007), China’s Gini coefªcient climbed from 40.74 percent in 1993
to 47.25 percent in 2004 and overtook the four Asian countries (Thailand, the Philip-
pines, Malaysia and Turkmenistan) that had higher Gini coefªcients than China in

viduals and organizations who made this project possible. We also thank the readers of the
earlier Wang (2007) study and of earlier drafts of this report for their valuable comments. We
are solely responsible for the remaining mistakes in this article.

1 See, for example, the Goldman-Sachs report by Wilson and Stupnytska (2007), which pre-

dicted that China’s GDP would surpass that of the United States in 2027. For a review of the
debate on how to interpret China’s high growth in the post-1978 period, Woo (1999) and
Woo (2001).

2 The harmonious socialist society would be (1) a democratic society under the rule of law;

(2) a society based on equality and justice; (3) an honest and caring society; (4) a stable, vig-
orous and orderly society; and (5) a society in which humans live in harmony with nature;
see “CPC key plenum elevates social harmony to more prominent position,” People’s Daily
Online, 12 October 2006. What is revealing is that the ofªcial descriptions of the harmonious
society downplayed the prominence of achieving a prosperous society. Of the nine objec-
tives listed in the Communique of the 2006 Plenum, “the objective of building a moderately
prosperous society” was not only listed last, it was also qualiªed with the condition that the
prosperity should be shared “all-around.” And this qualiªer is actually a repetition because
the narrowing of income gaps had already been listed as the second objective.

3 Wu and Perloff (2005) put the rural and urban Gini coefªcients to be 27.2 percent and

19.1 percent respectively in 1985; and Benjamin, Brandt, Giles and Wang (2005) estimated
them to be 32 percent and 22 percent respectively in 1987.

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

1993–94. While Nepal had the highest Gini coefªcient in Asia in 2001–04, its value of
47.30 percent is statistically indistinguishable from China’s value of 47.25 percent. If
one combines this with the fact that China’s income ratio of the richest 20 percent to
the poorest 20 percent (11.37) is the highest in Asia and is signiªcantly higher than
the next highest income ratio (9.47 for Nepal), China is probably the most unequal
country in Asia today.

One severe difªculty with knowing the extent to which inclusive growth (bao rong
xing zeng zhang) has not been achieved is the widespread phenomenon of hidden in-
come. In 2007, one of us published a research report (Wang 2007) on unreported in-
come in China in 2005 based on an urban income survey he had conducted in 2005–
06. Wang (2007) estimated that unreported Chinese urban household income—the
difference between his estimate of household income and the level of household in-
come reported by the National Bureau of Statistics (NBS)—totalled RMB 4.8 trillion
in 2005, and that most of this unreported income belonged to the high income
classes. Wang called this unreported income “hidden income.” A correction of the
income statistics by including the hidden income showed that the ratio of the in-
come of the wealthiest 10 percent of households to the income of the poorest 10 per-
cent households in urban areas was 31:1 instead of the reported 9:1, and that the
same ratio on a nationwide basis, was 55:1 instead of the reported 21:1. In short, the
2007 report showed that China’s income inequality problem is much more severe
than what is usually reported.4

The Wang (2007) report also pointed out that the level of household income re-
ported in the NBS Household Survey was lower than the level reported in the NBS
Economic Census. The latter was still lower than his estimate of household income,
and he called the difference between the estimates of the Economic Census and his
survey “gray income.”

What has happened to income distribution since 2005? We conducted a second
survey on urban household incomes in 2009 to obtain the data for 2008, and this pa-
per reports the ªndings of this second survey. In what that follows, Section 2 de-
scribes the survey method, Section 3 explains the analytical techniques and presents
the results, and Section 4 estimates the levels of disposable incomes of urban resi-
dents. Section 5 discusses the sources of gray income, and Section 6 analyzes its

4 The 2007 report also conªrmed the veracity of our estimated household income by using

data such as family ownership of cars and housing, number of overseas trips, and amount
of private bank deposits.

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

impact on national income distribution. Section 7 concludes with some remarks on
our ªndings.

2. Survey method and sample distribution

2.1 Gathering reliable data
The NBS survey samples of urban and rural residents are determined by random
sampling that follows standard statistical procedures. We see two defects in the NBS
approach:

1) NBS random sampling is based on the principle of voluntary participation. A

considerable proportion of higher income residents, however, are unwilling to be
included in the survey. The samples are, therefore, deªcient in high income resi-
dents.

2) Among higher income residents in the sampling, many are reluctant to provide
true information about their income. They tend to report truthfully their regular
salaries, but are relatively untruthful about other types of income, especially the
“gray income” from unidentiªed sources.

In contrast, we obtained more reliable data about household income in our 2005–06
survey because we drew upon the methods of sociology. We asked our professional
survey staff in different regions to interview only the people they are familiar with,
namely, their relatives, friends, neighbors, and former schoolmates, whose family
ªnancial status they generally know. In the 2009 survey, we adopted the same
method, but implemented even stricter quality control measures and expanded the
sample size.5 It is important to note that our method is different from that of random
sampling and therefore our data cannot be used directly to extrapolate the general
distribution of urban household income.

Before the survey, we trained our survey staff at various locations on questionnaire
and survey methodology. To eliminate the interviewee’s fears, the questionnaire had
no information on the identity of the interviewee and the interviewee was assured
of the research purpose of the survey as well as the conªdentiality of his personal
data. We also took measures to lower sensitivity to the survey (e.g., we emphasized
that our main purpose was to study consumption structure rather than income lev-
els). The questionnaire is designed to inquire about consumption details before the
income details, and to inquire about different components of consumption and in-

5 The professional survey companies in the different regions employed a total of about

450 people to conduct the interviews.

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

come before the total amount. As for the sources of income, the questionnaire only
asked the interviewees to choose among a few simple categories, including wages
and salary, part-time job and service payments, entity-business return, gains from
capital and ªnancial markets, property rents, intellectual property royalties, transfer
income, and (unclassiªed) other incomes.

After the survey, the surveyors were required to report their relationship with each
interviewee and their personal judgment about the creditability of the survey result
(including their judgment on the direction and extent of possible deviations). Then,
in addition to making a thorough check for completeness of the information and
correctness of the survey locations, we also designed a set of screening procedures
to examine the rationality of the logic between answers to different questions and
the consistency between income, saving and expenditure data in each questionnaire.
We then omitted the “suspicious” questionnaires.

2.2 Distribution of survey samples
This survey was conducted in 64 cities of different scale in 19 provinces (including
cities under direct administration of the central government), as well as 14 county
towns and administrative towns. The provinces are Beijing, Shanghai, Shandong,
Jiangsu, Zhejiang, Guangdong, Shanxi, Henan, Hubei, Anhui, Jiangxi, Liaoning,
Heilongjiang, Sichuan, Chongqing, Yunnan, Shaanxi, Gansu, and Qinghai. This se-
lection of the provinces achieves balance among the East, Northeast, Central, and
West regions, and between north and south China. The cities are Beijing, Shanghai,
Jinan, Nanjing, Hangzhou, Guangzhou, Tai Yuan, Zhengzhou, Wuhan, Hefei,
Nanchang, Shenyang, Harbin, Chengdu, Chongqing, Kunming, Xi’an, Lanzhou,
Xining, Shenzhen, Qingdao, Suzhou, Datong, Anshan, Fushun, Qiqihar, Daqing,
Xuzhou, Yangzhou, Fuyang (in Anhui Province), Wuhu, Lu’an, Rizhao, Xiangfan,
Yichang, Dongguan, Zhongshan, Mianyang, Xinzhou, Kaifeng, Sanmenxia,
Zhumadian, Xiaogan, Yidu, Pizhou, Fuyang (in Zhejiang Province), Jinhua,
Shaoxing, Shaoguan, Chaohu, Chuzhou, Ganzhou, Ji’an, Jingdezhen, Jiujiang,
Dandong, Tieling, Mudanjiang, Xichang, Xianyang, Baiyin, Jiayuguan, Tianshui,
and Yuxi. Of these, 21 cities are either under direct administration of the central gov-
ernment, or provincial capitals, or “sub-provincial” cities, and 43 are smaller cities at
the prefecture and county levels. In this way, a generally balanced distribution was
kept among cities of different scales.

County towns and administrative towns include Fanzhi County in Shanxi Province,
Pei County in Jiangsu Province, Xiangshan County in Zhejiang Province, Ping Yuan
County and Qihe County in Shandong Province, Hua County in Henan Province,
Dawu County in Hubei Province, Zhijiang County, Kai County, and Zhong County

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

in Chongqing City, Liquan County in Shaanxi Province, Gaolan County and
Jingchuan County in Gansu Province, and Minhe County in Qinghai Province. The
geographical distribution of these towns was also generally balanced.

For this survey we chose a large number of cities with a relatively scattered distribu-
tion of observations in each city, mainly for two reasons. First, if we choose too
many observations in a city, we cannot ensure that our surveyors are familiar with
all the respondents, which is a prerequisite of the survey. Second, the wide geo-
graphical distribution achieved by the large number of cities ensures accurate repre-
sentation of urban China.

Our methodology does have shortcomings. A major problem is that the survey is
done in one interview, and all the income and consumption data of the interviewed
families are provided by the interviewee according to his memory (we have ex-
cluded those family members who are unfamiliar with their family income and con-
sumption). Compared with surveys that require a respondent to record his income
and expenditure every day, our methodology has greater data error. However, re-
quiring a respondent to record his information over an extended period of time is
more prone to systematic distortion because of the respondent’s sensitivity to some
survey questions. As the data errors in our survey were caused by inaccurate mem-
ory, they are mostly random instead of systematic. When the desired value is calcu-
lated by taking group averages, then the random errors should offset each other and
cause limited bias, but systematic bias cannot be offset by averaging. Our adoption
of this survey method is therefore rational.

Our survey covered 4,909 families. After strict inspection, 689 suspicious question-
naires were dropped and 25 negative income observations6 were excluded to arrive
at the ªnal (effective) sample size of 4,195 observations. Table 1 shows the informa-
tion about regional distribution of the total (collected) sample and the effective (ac-
tually used) sample, the scale of the cities, the age and household registration status
of the respondents, and the education level and the profession of the family member
with the highest income. The samples are generally evenly distributed across re-
gions and cities of different size, and among various age group and education
levels.

However, it seems from Table 1 that our survey sample is skewed toward people
living in larger cities, with better education, owning their business, or working in

6 Most of the negative income families are not normally low-income families. Their negative

income was commonly due to temporary losses in their family business.

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

Table 1. Sample distribution in various classiªcations

Total sample
(total
collected
observations)

Distribution
(%)

Effective
sample
(observations
actually used)

Distribution
(%)

1. Geographical location

Eastern region
Central and Northeast regions
Western region

Total

2. Distribution by the scale of the cities

Cities with more than 2 million in population
Cities with 1 to 2 million in population
Cities with less than 1 million in population
County towns and administrative towns

Total

3. Age of the interviewees

20–29
30–39
40–49
50–59
60 and above

Total

1,863
1,848
1,198

4,909

2,495
915
995
504

4,909

1,647
1,383
1,236
520
123

4,909

4. Household registration of the interviewees

Local urban resident
Non-local urban resident
Non-local rural resident
Forgot to answer

Total

4,457
276
156
20

4,909

37.95
37.65
24.40

100.00

50.83
18.64
20.27
10.27

100.00

33.55
28.17
25.18
10.59
2.51

100.00

90.79
5.62
3.18
0.41

100.00

5. Educational level of the family member with highest income

Elementary school and below
Junior middle school
Senior middle school (including equivalency)
University and college
Post-graduate and PhD
Forgot to answer or indeªnable

Total

165
970
1,833
1,822
82
37

4,909

6. Profession of highest income member of the family

General technical personnel
Middle and senior level technical personnel
Other professional (scientists, teachers, doctors,

performers, etc.)

Low level ofªcer of the Party, government,

army, etc.

Middle and senior level ofªcial of the Party,

government, army, etc.

Ordinary staff of enterprises and social

organizations

Middle and senior manager of enterprises and

social organisations

Service personnel
Worker
Family business or self employed
Owner, partner, shareholder of private

enterprises

Other occupations
Students, post-graduates
Jobless (including retired)
Forgot to answer or indeªnable

Total

Source: Our 2009 survey sample data.

396
262
339

193

561

327

317
659
1,008
317

73
20
349
36

4,909

Note: The scale of the city is measured by its regular urban population.

3.36
19.76
37.34
37.12
1.67
0.75

100.00

8.07
5.34
6.91

3.93

1.06

11.43

6.66

6.46
13.42
20.53
6.46

1.49
0.41
7.11
0.73

Asian Economic Papers

1,563
1,605
1,027

4,195

2,083
789
889
434

4,195

1,411
1,196
1,062
425
101

4,195

3,808
234
138
15

4,195

136
832
1,565
1,569
74
19

4,195

353
227
302

165

483

268

277
562
853
277

66
17
278
20

37.26
38.26
24.48

100.00

49.65
18.81
21.19
10.35

100.00

33.64
28.51
25.32
10.13
2.41

100.00

90.77
5.58
3.29
0.36

100.00

3.24
19.83
37.31
37.40
1.76
0.45

100.00

8.41
5.41
7.20

3.93

1.12

11.51

6.39

6.60
13.40
20.33
6.60

1.57
0.41
6.63
0.48

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

100.00

4,195

100.00

The Size and Distribution of Hidden Household Income in China

white collar positions. For example, 27.3 percent of our sample population received
tertiary level education while ofªcial data suggest that the proportion of urban pop-
ulation with tertiary education is less than 14.7 percent.7 This skewing of our survey
sample toward people living in large cities and having more education is actually an
outcome that we had deliberately created. According to Wang (2007), the under-
statement of urban household income in the ofªcial data mainly occurs with higher
income residents. To ensure a large enough sample of high income and very high in-
come households, we intentionally increased the number of observations for this
type of people. As will be explained later, the methods we employ to analyze the
data do not allow the sample distribution to inºuence our estimation of the income
distribution of the total urban population.

3. Estimation methods and results

3.1 Engel’s coefªcient method
Economists call the proportion of food expenses in the total consumption expendi-
ture of a family Engel’s coefªcient, and they have long established that the value of
Engel’s coefªcient declines with the rise in income. This happens because after the
basic demand for food by the family has been met, its members start spending in-
creasingly more on transportation and communication, luxury goods, higher level
education, and cultural entertainment. The growth rate of food expenditure be-
comes increasingly lower than the growth rate of consumption.

In presenting the household income data, the NBS divides the urban resident fami-
lies into seven income groups according to their per-capita incomes:

(1) lowest income
(2) low income
(3) lower middle income
(4) middle income
(5) upper middle income
(6) high income
(7) highest income

Groups (1), (2), (6), and (7) account for 10 percent each of all urban families. Groups
(3), (4), and (5) account for 20 percent each of all urban families. We calculated the

7 Table 3-12 in the China Statistical Yearbook 2009 (CSY2009) reports that 6.7 percent of the Chi-
nese population has at least a college-level education, and Table 3-4 in CSY2009 shows that
45.7 percent of the Chinese population are urban residents. The 14.7 percent is obtained if all
such educated people live in urban areas.

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

average Engel’s coefªcients of the seven income groups from the published NBS
data.

According to the information obtained from our 2005–06 and 2009 surveys, higher
income families in the NBS household survey usually underreport their incomes to
a large extent. Some of them also underreport their food expenditure and total con-
sumption expenditure (but to a much smaller extent on average) and the propor-
tions of the underreporting in these two items are roughly the same. We therefore
assume that each household group in the NBS survey underreports their food ex-
penditure and total consumption by the same proportion, which implies that the
Engel’s coefªcient for each income group that is calculated from the NBS survey is
the true value for that group of households, even though the income level could be
seriously underreported.

The important implication is that if we can obtain an independent estimate of the
true relationship between Engel’s coefªcient and income level in China, then we can
use the Engel’s coefªcient of each NBS income group to deduce the true level of in-
come in each NBS income group. The difference between the deduced income level
and the NBS-reported income level is the “hidden income” of the average family in
each income group.

As our samples were collected in a manner that encouraged respondents to report
their true income and true expenditure, we can use the 2008 sample to calculate the
true relationship between Engel’s coefªcient and income level. We calculated this re-
lationship in two ways by assuming, in turn, that the size of Engel’s coefªcient de-
pends:

• only on the per capita income in the family. The use of this particular Engel’s

coefªcient to estimate actual income is called the “simple-Engel approach”; and
• not only on income but also on a number of other variables (that we will identify
later). We call this more general view of Engel’s coefªcient the “supplemented-
Engel approach,” and this is our preferred approach.

It is important to understand that:

1. the estimation of the true multivariable Engel’s coefªcient equation can be done
without our data sample to be representative of the national population; and
2. (as will be shown) when we use the estimated multivariable Engel coefªcient
equation in combination with the national average values8 of the variables for

8 These national average values are not from our survey sample.

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

each income group and the NBS value of the Engel’s coefªcient for each income
group, we obtain the true level of the national average income for each income
group.

3.2 The simple-Engel approach
First, we calculate the per capita income and Engel’s coefªcients of all observations.9

Second, we sort all the valid study samples according to their households’ per
capita disposable income from the lowest to the highest. To group the samples, we
start at the lowest income and keep adding observations until we achieve an aver-
age Engel’s coefªcient that equals that of the “lowest income group” of the NBS sur-
vey. This chosen sample group is called the “lowest income group.” Then, we start
with the next observation above the cut-off income of the lowest income group and
use the same method as before to arrive at the upper cut-off income of the “low in-
come group,” that is, this group of observations in our sample has the same average
Engel’s coefªcient as the “low income group” of NBS data. This method is repeated
for the next higher income group. Our procedure of grouping does not require con-
sideration about the number of observations in each income group.

We had to leave out the 76 wealthiest observations from the “highest income group”
because the values of their Engel’s coefªcients are so low that their inclusion would
render the average value of the Engel’s coefªcient to be far below the NBS value of
the Engel’s coefªcient in the ofªcial “highest income group.” This suggests that the
NBS household sample does not contain the very rich families in China. The
76 excluded observations accounted for 1.8 percent of our survey sample and have
(a) a minimum annual per capita disposable income of more than RMB 400,000,
(b) a maximum per capita income of RMB 1.76 million, and (c) an average per capita
income of RMB 658,811.

Third, we calculate the average per capita income of each income group of the study
samples.

Fourth, we compare the per capita income of the part of our sample in each income
group with that of the corresponding ofªcial sample group, and discover the under-
statement of income in the ofªcial samples.

Table 2 shows the distribution of our survey sample and the NBS (ofªcial) sample
by income groups. The “high income group” and “highest income group” together

9 The unit of observation is the family.

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

Table 2. Distribution of our sample and ofªcial sample by income group

Group

Lowest income
Low income
Lower middle income
Middle income
Upper middle income
High income
Highest income
Excluded samples

Total

Study samples
range (RMB)
(cid:2)
1–7,000
(cid:2) 7,001–10,000
(cid:2) 10,001–17,000
(cid:2) 17,001–26,500
(cid:2) 26,501–34,000
(cid:2) 34,001–75,000
(cid:2) 75,001–400,000
(cid:2)400,000

Study samples
(households)

Study samples
proportion (%)

Ofªcial samples
proportion (%)

365
622
927
650
355
635
565
76

4,195

8.7
14.8
22.1
15.5
8.5
15.1
13.5
1.8

100.0

10
10
20
20
20
10
10
0

100.0

Source: Our 2009 survey result and statistics (NBS 2009).

Note: Altogether 65,000 urban households are included in the ofªcial samples.

make up 28.6 percent of our sample, whereas they make up only 20 percent of the
ofªcial sample. So we have achieved the desired skewing of our survey sample dis-
cussed earlier to encompass rich and super-rich households.

Table 3 shows the per capita incomes between our samples and the ofªcial NBS
samples. The per capita income of each income group of our survey sample is
always higher than that of the ofªcial samples. The gap expands for the higher in-
come groups. In the highest income group, the NBS survey shows a per capita
income of RMB 43,614 but our survey sample shows a per capita income of
RMB 164,034—nearly 3.8 times larger. The unreported income gap of this “highest
income group” accounted for about two-thirds of the total hidden income. These
ªndings coincide with the ªndings reported in our 2007 study. The general consis-
tency between the two studies is re-assuring about their credibility.

3.3 The supplemented-Engel approach
We see ªve other variables (beside income) to also be determinants of the size of
Engel’s coefªcient. First, prices of consumption goods vary from city to city. For in-
stance, food prices tend to be higher in large cities than in small cities, and so
Engel’s coefªcient in large cities is likely to be higher. We use a city scale variable
(henceforth city) to catch this price effect. Extra large cities (population of more than
2 million), large cities (population between 1 and 2 million), small and medium cit-
ies (population of less than 1 million), and county towns are given the values of 1, 2,
3, and 4, respectively.

Second, residents in different places have different dietary habits. Because people in
some regions may spend more on food than others, we insert region-speciªc dummies
to the regression equations. An analysis of our survey data shows that, under the
same circumstances, Engel’s coefªcients in Shanghai, Jiangxi, and Sichuan are

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

Table 3. The simple-Engel approach: Comparison between estimated and ofªcial per capita
urban income in 2008 (RMB)

Estimated data

Ofªcial data

Comparison between
two samples

Group

Engel’s
coefªcient

Per-capita
income (RMB)

Engel’s
coefªcient

Per-capita
income (RMB) Gap (RMB)

Divergence
(%)

Lowest income
Low income
Lower middle income
Middle income
Upper middle income
High income
Highest income
Excluded observations

0.4816
0.4595
0.4297
0.4065
0.3790
0.3437
0.2908
0.2241

5,685
8,646
13,392
20,941
29,910
47,772
164,034
658,811

Source: Our 2009 survey result and statistics (NBS 2009).

0.4814
0.4594
0.4289
0.4042
0.3787
0.3403
0.2918

4,754
7,363
10,196
13,984
19,254
26,250
43,614

931
1,283
3,196
6,957
10,656
21,500
120,420

19.6
17.4
31.3
49.7
55.3
82.0
276.1

Note: The tiny deviation between Engel’s coefªcients of study samples and correspondent ofªcial samples has little inºuence upon the

analysis and therefore is treated as equal. “Gap” refers to the amount that estimated income exceeds the ofªcial income. “Divergence”

refers to the proportion of income gap as percent of the ofªcial income.

noticeably higher than the average level of all provinces. Dummy variable H1 is
used to represent these three provinces. Engel’s coefªcients in Beijing, Shandong,
Hubei, Guangdong, Chongqing, and Henan are moderately higher than the aver-
age, and they are presented by dummy variable H2. Engel’s coefªcients in Liaoning
and Shanxi are lower than the average, and they are presented by dummy variable
L1. Observations from the other provinces (including Jiangsu, Zhejiang, Anhui,
Heilongjiang, Yunnan, Shanxi, Gansu, and Qinghai) constitute the reference sample.

Third, family size may have impact on Engel’s coefªcient, because bigger families
tend to buy food in bulk to save on food expenses. A family variable is used to repre-
sent the number of family members.

Fourth, the education level may affect Engel’s coefªcient, because residents with a
higher educational background may consume more communication, education, and
cultural entertainment services, whereas residents with lower education back-
grounds may consume more food, cigarettes, and drinks. A variable edu18 is set to
represent the average education level for family members at or above age of 18. The
variable edu18 is valued from 1 to 5 to refer to:

junior middle school

1. elementary school and below
2.
3. senior middle school and vocational school
4. college and university
5. post-graduate and doctoral studies

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

Fifth, Engel’s coefªcient may be affected by the family’s employment ratio (the pro-
portion of employed family members in the whole family, the emp variable). On one
hand, more family members being employed may mean fewer food expenses be-
cause they may eat at their workplaces and enjoy food subsidies there. On the other
hand, however, they may prefer eating in regular restaurants and hence incur more
food expenses.

To take non-linearity into account, we estimated the following four speciªcations of
Engel’s coefªcient (eng) equation:

eng (cid:3) C1

(cid:4) a1lnY (cid:4) a2city (cid:4) a3family (cid:4) a4edu18 (cid:4) a5emp (cid:4)

a6H2 (cid:4) a7H1 (cid:4) a8L1

eng (cid:3) C2

(cid:4) b1lnY (cid:4) b2city (cid:4) b3family (cid:4) b4edu18 (cid:4) b5emp (cid:4)

b6H2 (cid:4)b7H1 (cid:4) b8L1 (cid:4) b9(lnY)2

eng (cid:3) C3

(cid:4) c1Y (cid:4) c2city (cid:4) c3family (cid:4) c4edu18 (cid:4) c5emp (cid:4) c6H2 (cid:4) c7H1 (cid:4)

c8L1 (cid:4) c9Y2 (cid:4) c10city2 (cid:4) c11family2 (cid:4) c12edu182 (cid:4) c13emp2

(1)

(2)

(3)

eng (cid:3) C4

(cid:4) d1Y (cid:4) d2city (cid:4) d3family (cid:4) d4edu18 (cid:4) d5emp (cid:4) d6H2 (cid:4) d7H1 (cid:4)

(4)

d8L1 (cid:4) d9Y2 (cid:4) d10city2 (cid:4) d11family2 (cid:4) d12edu182 (cid:4) d13emp2 (cid:4)
d14Y3 (cid:4) d15city3 (cid:4) d16family3 (cid:4) d17edu183 (cid:4) d18emp3

In preliminary regressions not reported here, the squared and cubic terms of some
variables in equations (3) and (4) were found to be statistically insigniªcant at the
10 percent level. These variables were omitted from the ªnal speciªcations. The re-
gression results of the ªnal speciªcations are shown in Table 4. We see that although
the adjusted R2 of the four models are not high, most of the variables found strong
statistical support. As model (2) has the highest adjusted R2, it is our preferred
model, and we will use it in the subsequent gray income estimations.

For each income group, we assign values to each control variable that equal to the
national average values of those variables in that income group. Speciªcally, the na-
tional average value for all income groups of the:

• city variable is approximately 2.5. Because we know from our 2006 and 2009 sur-
veys and from international experience that the richest households tend to live in
the bigger cities, and the poorest household tend to live the smaller towns, we as-
sume that city (cid:3) 1.3 for the highest income group, and city (cid:3) 3.3 for the lowest in-
come group; and that the values for the city variable for the other ªve income

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

Table 4. Estimating four speciªcations of the supplemented-Engel approach

(1) Semi logarithm
function

(2) Semi logarithm
quadratic function

(3) Quadratic
function

(4) Cubic function

coefªcient
(cid:5)0.05739

coefªcient

t-ratio
(cid:5)28.66*** (cid:5)0.12004
(cid:5)0.00295

t-ratio
(cid:5)4.63***
2.42**

coefªcient

t-ratio

coefªcient

t-ratio

(cid:5)7.67E-07
(cid:5)5.44E-13

(cid:5)0.00664

(cid:5)3.50*** (cid:5)0.00677

(cid:5)3.57*** (cid:5)0.00385

(cid:5)0.01116

(cid:5)4.35*** (cid:5)0.01066

(cid:5)0.01427
(cid:5)0.01585
(cid:5)0.07106
(cid:5)0.02557
(cid:5)0.03938
(cid:5)1.06077
(cid:5)0.2463

(cid:5)6.41*** (cid:5)0.01423
(cid:5)1.95** (cid:5)0.01350
11.47*** (cid:5)0.07078
5.66*** (cid:5)0.02544
(cid:5)6.06*** (cid:5)0.03979
49.76*** (cid:5)1.38627
(cid:5)0.2472

(cid:5)4.15*** (cid:5)0.03194
(cid:5)0.00117
(cid:5)6.40*** (cid:5)0.01559
(cid:5)1.65*
(cid:5)0.03781
11.43*** (cid:5)0.07601
5.62*** (cid:5)0.02615
(cid:5)6.13*** (cid:5)0.03298
10.19*** (cid:5)0.5790
(cid:5)0.1973

(cid:5)20.8***

(cid:5)1.97**

(cid:5)1.24E-06
13.88*** (cid:5)1.93E-12
(cid:5)7.49E-19
(cid:5)0.12508
(cid:5)0.05612
(cid:5)0.00774
(cid:5)6.80*** (cid:5)0.02741
2.84*** (cid:5)0.00098
(cid:5)6.78*** (cid:5)0.01498
(cid:5)4.53*** (cid:5)0.03164
11.89*** (cid:5)0.07543
5.58*** (cid:5)0.02858
(cid:5)4.93*** (cid:5)0.03149
37.80*** (cid:5)0.64580
(cid:5)0.2130

(cid:5)19.31***
12.15***
(cid:5)8.99***
(cid:5)2.21**
2.22**
(cid:5)2.28**
(cid:5)5.83***
2.39**
(cid:5)6.54***
(cid:5)3.82***
11.89***
6.12***
(cid:5)4.74***
16.57***

lnY
(lnY)2
Y
Y2
Y3
city
city2
city3
edu18
edu182
family
emp
H1
H2
L1
C
Adj.R2

Observations

4,195

Note: *Statistically signiªcant at the 10 percent level. **Statistically signiªcant at the 5 percent level. ***Statistically signiªcant at the
1 percent level. In every case, Prob.>.F is 0.000.

groups lie proportionally within this range (e.g., city (cid:3) 2.3 for the middle income
group).

• education level of urban residents above 18 years of age is around 3. Again be-

cause edu18 is closely related to income, we assume that edu18 (cid:3) 3.8 for the high-
est income group, and edu18 (cid:3) 2.6 for the lowest income group; and that the val-
ues for the edu18 variable for the other ªve income groups lie proportionally
within this range (e.g., edu18 (cid:3) 3.2 for the middle income group).

• family employment ratio is around 0.5. From the NBS household survey, we

know the values for the emp variable in each income group (e.g., emp (cid:3) 0.62 in the
highest income group and emp (cid:3) 0.38 in the lowest income group).

• family size variable is 2.9. From the NBS household survey, we know the values
for the family variable in each income group (e.g., family (cid:3) 2.6 in the highest in-
come group and family (cid:3) 3.3 in the lowest income group).

• regional dietary effect is about 0.01, which is obtained by multiplying the national
average value of each regional dummy variable with its estimated coefªcient, and
then adding them up.

Based on the estimated parameter values of regression equation (2), the given values
for the ªve control variables for each income group, and the NBS value of Engel’s
coefªcient (eng) for each income group, equation (2) for each income group is re-
duced to a quadratic equation in lnY:

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

b9(lnY)2 (cid:4) a1lnY (cid:4) [(C2

(cid:4) b2city (cid:4) b3family (cid:4) b4edu18 (cid:4) b5emp (cid:4)

(5)

b6H2 (cid:4) b7H1 (cid:4) b8L1) (cid:5) eng] (cid:3) 0

From equation (5) it is straightforward to compute the value of lnY using the qua-
dratic formula, and then of Y for each income group; see Table 5. This supplemented
Engel approach allows us to estimate the national average income for each of the
seven income groups without requiring that our survey sample be a representative
national sample.

Table 5. The supplemented-Engel approach: Comparison between estimated and ofªcial
per-capita urban income in 2008 (RMB)

Group

Engel’s coefªcient

Ofªcial income

Estimated income:
simple-Engel

Estimated income:
supplemented-Engel

Lowest income
Low income
Lower middle income
Middle income
Upper middle income
High income
Highest income
All urban residents
Left-out observations

0.481
0.459
0.429
0.404
0.379
0.340
0.292
0.379
0.224

4,754
7,363
10,196
13,984
19,254
26,250
43,614
16,885

5,685
8,646
13,392
20,941
29,910
47,772
164,034
35,462
658,811

5,350
7,430
11,970
17,900
27,560
54,900
139,000
32,154

Source: NBS (2009), and authors’ estimation.

Note: The RMB 16,885 urban income is an weighted average from the ofªcial sample groups, while the published average by NBS is

RMB 15,781. The estimated income of all urban residents does not include the left-out samples.

4. Estimating the true income of urban residents

4.1 The estimated urban household income by group
Table 5 compares the estimated results from the supplemented-Engel approach with
the results from the ofªcial data and our simple-Engel approach. It shows that esti-
mated incomes derived from the supplemented-Engel approach for the two low-
income groups are only marginally higher than the ofªcial incomes. The gap be-
tween estimated income and the ofªcial income becomes signiªcantly greater for
the middle-income groups and above. The greatest difference lay within the highest
income group, with per-capita income at RMB 164,034 according to the simple-
Engel approach and RMB 139,000 according to the supplemented-Engel approach,
which are 3.76 and 3.19 times ofªcial income, respectively. Driven by the high esti-
mated income of the high income and highest income groups, the average per-
capita income of all urban residents is nearly double the ofªcial income—that is,
RMB 35,462 according to the simple-Engel approach and RMB 32,154 according to the
supplemented-Engel approach instead of RMB 16,885 according to the NBS survey.

Our judgment is that the supplemented-Engel approach to estimating the true
income level is preferable to the simple-Engel approach because of the strong

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

Table 6. Ratio of estimated income to ofªcial incomes in 2005 and 2008

Group

Lowest income
Low income
Lower middle income
Middle income
Upper middle income
High income
Highest income
All urban residents

Ratio of estimated
income to ofªcial
income in 2005 (%)

Ratio of estimated
income to ofªcial
income in 2008 (%)

Distribution of
hidden income
in 2008 (%)

99.1
101.8
106.9
114.0
130.6
138.7
337.6
177.7

112.5
100.9
117.4
128.0
143.1
209.1
318.7
190.4

0.4
0.0
2.3
5.1
10.9
18.8
62.5
100.0

Source: NBS (2006, 2009) and authors’ estimation.

Note: The estimated incomes are based on the supplemented-Engel approach.

statistical signiªcance of the other variables reported in Table 4. Hence, from this
point onward, we will use the income estimates obtained from speciªcation (2) of
the supplemented-Engel approach in all calculations.

Table 6 shows the ratios between the estimated income and ofªcial data in 2005 and
2008. We ªnd that in the high income group, the gap between the estimated income
and ofªcial data has widened most signiªcantly, from 138.7 percent in 2005 to
209.1 percent in 2008. The greatest deviation still occurs at the highest income group,
337.6 percent in 2005 and 318.7 percent in 2008.

We want to reiterate that the fact that we had to exclude the 76 wealthiest observa-
tions when we employed the simple-Engel approach means that the NBS household
survey sample most probably has few (or even no) households with per capita dis-
posable income greater than RMB 400,000. Therefore, strictly speaking, the “highest
income group” category in Tables 5 and 6 does not really capture the truly top-
income people. Because we do not know the proportion and income level of the
missing super–high income households from the ofªcial samples, we cannot correct
this distortion in the ofªcial data. What we have done in this paper, therefore, is to
correct only the understatement of income in six of the seven income categories in
the ofªcial data. Although we still understate the income level of the highest income
households, we are conªdent that our income estimates for this group is far closer to
the true level than the ofªcial income statistics.

4.2 How large is the hidden income?
As indicated in Table 6, the hidden income of the highest income families accounts
for 63 percent of all hidden income, and this makes the income gap between the top
and bottom 10 percent of urban families 26:1 rather than 9:1 according to the ofªcial
data. Together with the hidden income of the high income group, the wealthiest
20 percent of the urban population takes up more than 80 percent of total hidden

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

Table 7. Income changes between 2005 and 2008 from ofªcial data and supplemented-Engel
approach

Per-capita urban disposable income (RMB, ofªcial)
Per-capita urban disposable income (RMB, estimated)
Urban population (million)
Per-capita rural net income (RMB)
Rural population (million)
Total household disposable income (RMB billion, ofªcial)
Total household disposable income (RMB billion, estimated)
Estimated hidden income (RMB billion)
GDP (RMB billion, ofªcial)

Source: NBS (2006, 2009), Wang (2007), authors’ estimations.

2005

11,100
19,730
562
3,537
745
8,876
13,727
4,851
18,322

2008

16,885
32,154
607
5,171
721
13,974
23,237
9,263
31,405

Change %

52.1
63.0
7.9
46.2
(cid:5)3.2
57.4
69.3
91.0
71.4

Note: The ofªcial data of urban and rural disposable income per capita are derived from the group statistics as weighted averages,

which are slightly higher than the ofªcial average income data published by the NBS. GDP data are not adjusted.

income. Because hidden income occurs mostly in urban areas, if we use the wealthi-
est 20 percent urban families and poorest 20 percent rural families to represent the
nationwide top and bottom 10 percent families respectively,10 the income gap is
65 times instead of the 23 times proposed in the ofªcial data.

Using our estimated urban income, we derive an approximate total household dis-
posable income of RMB 23.2 trillion in 2008 compared to less than RMB 14 trillion in
the ofªcial household statistics. This means that the total hidden income in China in
2008 is RMB 9.26 trillion, almost double the RMB 4.85 trillion in 2005 (up by 91 per-
cent). As nominal GDP had increased by only 71.4 percent in the same period, hid-
den income had expanded at a much faster pace than GDP.

Table 7 reports the changes in some key indicators between 2005 and 2008 according
to the ofªcial data and according to our corrected ofªcial data. After including the
hidden income, total household disposable income increased by 69.3 percent from
2005 to 2008, similar to nominal GDP growth. According to the ofªcial statistics (ex-
cluding hidden income), total household disposable income had increased by only
57.4 percent over the 2005–08 period, causing it to decline from 48.4 percent of GDP
in 2005 to 44.5 percent in 2008.

4.3 Cross-checking the size of our estimated hidden income
There have been discrepancies between the ofªcial household income statistics and
other ofªcial data series for a long time. However, as we shall see, some of these
contradictions disappear once we include the estimated hidden income into the
ofªcial household income. In this section, we check our estimate of RMB 9.3 trillion

10 This is because half of the Chinese population are urban residents and that average rural per

capita income on average is only one-third of average urban per capita income.

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

in hidden income in 2008 in three ways, by using (a) consumption and saving data,
(b) property prices, and (c) private ownership of automobiles.11

Consumption and savings According to the ofªcial data on urban and rural
households, total savings (the difference between disposable income and consump-
tion) in the nation should have been RMB 3.55 trillion in 2008. We can check the
plausibility of this number by calculating the approximate total household savings
from the amount of household savings that was put in each of the following six in-
vestment vehicles in 2008.

(1) Household savings deposits in the banking system increased by RMB 4.54 trillion
in 2008, which is more than the RMB 3.55 trillion in total household savings cal-
culated from the ofªcial household statistics.

(2) New (i.e. excluding second-hand) residential property sales in 2008 were RMB
2.12 trillion. After deducting the RMB 300 billion increase in mortgage loans,
RMB 1.82 trillion of household savings was used in property purchases.

(3) In 2008, RMB 371.1 billion was spent on private housing construction in rural ar-
eas. It is also common for urban residents to build their own houses, with self-
built houses accounting for 15–16 percent of self-owned property. As most resi-
dents use personal savings instead of loans from banks, we estimate that RMB
700 billion of household savings was spent on private housing construction.
(4) Equity in private industrial enterprises (excluding micro-businesses) increased
by RMB 1.09 trillion in 2008, and this increase basically came from the owners’
savings. Because equity increases in the services sector are estimated to be no less
than that in the industrial sector, private savings provided RMB 2.5 to 3.0 trillion
of the overall private investment in industrial and service sectors.

(5) The negotiable market value of A-shares shrank by only 50.9 percent in 2008

when the Shanghai Composite Index and Shenzhen Composite Index dropped
by 65.4 percent and 62.4 percent, respectively. This approximately 13 percentage-
point gap between the fall in market value and the fall in share prices means that
there was a net investment of RMB 1.35 trillion in the stock market in that year.
There was also a RMB 1.7 trillion net increase in treasury and corporate bonds
during that year. If we make the conservative assumption that one-third of the
investment in bonds and stocks came from household savings, then the amount
was around RMB 1 trillion.

(6) It is estimated that net private investment in commodity futures, gold, foreign
exchange, ªnancial derivatives, cash, and deposits in overseas banks together
amounted to RMB 500 billion in 2008.

11 All the data to be adjusted are from various issues of the NBS China Statistical Yearbook.

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

These six estimations were based on ofªcial statistics, and together they imply total
household savings of at least RMB 11–11.5 trillion in 2008, dwarªng the RMB 3.5 tril-
lion computed from the ofªcial household survey data. Now, NBS household data
put total household consumption to be RMB 9.46 trillion, which is RMB 1.4 trillion
less than the household consumption from the GDP statistics. If we make the con-
servative assumption consumption in ofªcial household survey data was under-
reported by RMB 2 trillion, then the total hidden income in 2008 should be at least
RMB 9.5–10 trillion.12 This number coincides with the RMB 9.3 trillion in hidden in-
come estimated from the supplemented-Engel approach.

Property price and income International experience tells us that housing prices
are usually three to ªve times that of annual household income to be affordable. In
recent years, China’s housing prices have been about 10 times the average urban
household income, which is well above the affordability of urban residents. But
then, the real estate market has been booming in the past years, total residential
property sales reached RMB 3.8 trillion in 2009, sharply up from RMB 2.1 trillion in
2008. These events imply that the true average urban household income should be
at least double the income level in the ofªcial data. This amount is about what we
had found using the supplemented-Engel approach; see Table 6.

According to the ofªcial household survey, the wealthiest 20 percent of urban
households had an average household income of RMB 89,425 in 2008. Because the
average property price in the primary market is about ªve times that amount, this
means that high income families were barely capable of purchasing property. How-
ever, this is not consistent with what have we observed.

During the 20 years between 1990 and 2009, more than 46 million apartments were
sold in the open (commercial) market when the richest 20 percent urban families
amounted to only 41 million households. This probably meant that some of these
apartments were bought by middle-income families, and that some high-income
families bought more than one apartment because our survey data show that more
than one-third of high-income families did not buy property from the real estate
market. This is because some of them live in properties provided either by their
companies or by the government; and the others had bought the property at non-
market prices during the housing reform era in the late 1990s. The high-income fam-
ilies who did purchase property from the market usually paid prices that were
much higher than the market average; and that around one-third of high-income

12 This range is from (11.0 (cid:5) 3.5 (cid:4) 2.0) trillion, and from (11.5 (cid:5) 3.5 (cid:4) 2.0) trillion.

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

families owned at least two residences. These facts show that the income of the
high-income group must be much higher than reported in the ofªcial data.

Holding of private automobiles According to the ofªcial data on car registration,
individuals owned 28.14 million private sedans in 2008. Assuming 90 percent of
these belong to urban residents, there are 12.1 cars for every 100 urban households,
which means that a majority of the 20 percent richest urban families probably own a
car. The catch is that the ofªcial household survey data in 2008 show only 8.8 pri-
vate automobiles for every 100 households. This discrepancy may indicate that
nearly one-third of high-income families are missing from the ofªcial survey data.

As the price of an average private sedan is about RMB 100,000, with RMB 20,000 re-
lated expenses each year (on fuel, maintenance, insurance, annual inspection, park-
ing, and tolls), it is reasonable to expect that families who can afford to own a car
would have an annual household income of not less than RMB 200,000. According
to the ofªcial data, the annual disposable income of the wealthiest 20 percent house-
holds was only RMB 89,425, which means that most of them cannot afford to buy a
car. In contrast, our analysis indicates that the wealthiest 20 percent of urban house-
holds have an actual annual income of RMB 248,192, which means that most of
them can afford a car.

5. Gray income and its sources

5.1 What does this huge hidden income tell us?
Gray income (the difference between our estimated income level and the income
level from the NBS Economic Census) is income that cannot be clearly deªned as le-
gitimate or illegitimate. For instance, presents and gift money received during wed-
dings are permitted by law, and some ofªcials collect huge amounts of money at the
weddings of their children and relatives. Some government organizations and state-
owned enterprises also provide their staff with big bonuses and welfare beneªts, far
above normal market practices. Tax evasion is one of the major reasons for the gray
income phenomenon.

Under the current circumstances, gray income is usually connected with the follow-
ing four phenomena.

(1) Abuse of power for personal gain A survey in 2006 covering 4,000 enterprises
in China included such a question: “How much did your company informally pay
ofªcials of government and regulatory agencies?” Only 19.8 percent of the managers
replied “none,” whereas 80.2 percent replied “a little,” “quite a lot,” and “a lot.”
Within which, those who answered “quite a lot” and “a lot” accounted for 18.1 per-

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

cent. The situation is worse in industries related to natural resources and monopo-
lies, and in industries under intensive supervision by state authorities. The propor-
tion of managers saying that informal payment is “quite a lot” or “a lot” was
35.2 percent in the mining industry, 24.3 percent in power and gas supply, 23 per-
cent in the real estate sector, and 24.2 percent in the chemicals industry (Wang 2006).

The embezzlement of public resources is also common. According to the National
Audit Ofªce (2010) report on the central government budget in 2009, 5,170 fake in-
voices worth RMB 142 million were found in the 29,363 doubtful invoices already
reimbursed by 56 central government departments. This ªnding is not surprising
because it is common to see people selling fake invoices in the streets and to receive
such advertisements in short message service and e-mails.

Another “emerging industry” that reºects the fast growth in gray income is the gift
purchase trade. In many cities, there are an increasing number of traders in the busi-
ness of buying expensive cigarettes, wine, medicine, jewelry, and gift coupons from
households at discount prices. It is certainly strange for households to be buying
luxury products and consumer coupons at high prices from regular shops and then
re-selling them to these traders at lower prices. There can only be one explanation
for this strange phenomenon—namely, many high-income households have re-
ceived such items as gifts and were selling them for cash. A key reason for such a
rampant gift-giving culture is that it is a safer form of corruption than receiving
cash.

(2) Public investment and corruption Public investment is another source of gray
income. Two recent examples are the Beijing–Shanghai express railway project and
the western section of the West-to-East Natural Gas Transmission Project. When
these projects were audited, overcharging of RMB 815 million in project construction
was found, in addition to RMB 1.794 billion on irrelevant fees in construction and
reimbursements of fake invoices. Furthermore, 80 percent of the work in the con-
struction contract of RMB 3.6 billion for the West-to-East Natural Gas Transmission
Project was awarded without public bidding procedures (National Audit Ofªce
2010).

(3) Leaking of land revenue Because many local governments do not have ade-
quate budgetary resources under the current ªscal system for infrastructure con-
struction and provision of public services, they rely heavily on the sale of land. In
2009, government revenue from land transfer fees reached RMB 1.5 trillion13 (20 per-
cent of government budgetary revenue) but this revenue is excluded from the

13 See Xinhuanet (2010).

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

formal government budget. Due to poor management of this revenue, it has become
a major source of gray income for local government ofªcials in some areas. In addi-
tion, the power of local authority to exempt land transfer fees could be another
source of corruption. The National Audit Ofªce revealed that in 2009, RMB 68.4 bil-
lion of land transfer fees in 11 provinces was not placed under the budgetary super-
vision system, and RMB 38.1 billion was not collected. One can only imagine how
much of the fee exemption ended up in the pockets of local ofªcials.

The government has the authority for land approval, expropriation, and sales, and
this confers monopoly status on the supply of land and on the related real estate sec-
tor. It is estimated that in 2009, the proªts of the real estate sector was RMB 1.7 tril-
lion, which is more than half of total industrial proªts, when the number of employ-
ees in the real estate sector was only 1.3 percent of those working in the industrial
sector, and the assets of the real estate sector was worth only 6 percent of industrial
assets.14 Of China’s top 30 wealthiest billionaires in Forbes 200915 list, 11 were in
property; and of the 36 persons (some tied) listed as China’s top 30 wealthiest bil-
lionaires in the Hurun 201016 list, 17 were real estate developers. Real estate is clearly
a most proªtable industry.

However, the RMB 1.7 trillion proªt of the real estate industry does not all go to real
estate developers. To acquire good pieces of land from local governments, real estate
developers sometimes need to “contribute” signiªcantly to people who have the
authority to approve land development. The proªt of the real estate sector is actu-
ally divided between property developers and those who have permit approval
power.

(4) Distribution of monopoly proªts The national wage statistics of 2008 and
2009 show that the average wage rate in highly monopolistic industries (such as oil,
tobacco, power generation and supply, telecommunication, banking, and insurance)
is about twice the national average. These data, however, fail to fully reºect the real
gap between different industries. First, the actual per capita income of workers and
staff members in monopolistic industries is far more than reported income in ofªcial
data. According to Bu Zhengfa, former Vice Minister of Labor and Social Security,
the actual per capita income gap between these industries and other sectors is be-

14 This observation was ªrst made by Chen Wanzhi, a member of National People’s Congress;
see East Morning Paper (4 March 2010). The ªgures here were calculated by the authors from
updated NBS (2010) data.

15 Available at www.forbes.com/lists/2009/74/china-billionaires-09_The-400-Richest-Chinese

_Rank.html.

16 See “Hurun Rich List 2010 sponsored by Hainan Clearwater Bay” at www.hurun.net/

listen186.aspx.

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

tween ªve to ten times.17 Second, within some of the monopolistic industries, there
is a much wider income gap between ordinary workers, staff, and senior manage-
ment than in the normal cases.

6. Recalculating the distribution of national income

Based on the NBS household income survey, household disposable income in 2008
is slightly below RMB 14 trillion, which is 44.2 percent of the Gross National Income
(GNI). However, based on the NBS Flow of Funds (FOF) accounts, household dis-
posable income was RMB 17.9 trillion18 (roughly RMB 4 trillion higher). The house-
hold income from the FOF data is, however, still RMB 5.4 trillion lower than our es-
timate of RMB 23.2 trillion (see Table 7). We call this RMB 5.4 trillion gap “gray
income.”

According to the FOF data, disposable household income in 2008 was 56.5 percent
of GNI. Of this, compensation of employees accounted for 46.7 percent of GNI, and
non-labor income accounted for 9.9 percent. The disposable income of the corpora-
tion sector (including both ªnancial and non-ªnancial corporations) and the govern-
ment sector accounted for 17.7 percent and 25.9 percent of total disposable income,
respectively.

It is reasonable to expect that gray income does not generally come from wages (as
the respondents have no reason to hide it). Thus the difference between estimated
household income from this study (and Wang’s (2007) study on 2005 data) and FOF
accounts—RMB 5.4 trillion in 2008 and RMB 2.7 trillion in 2005—are treated as non-
wage income in household income.

After the adjustment in household disposable income, there should also be some
corresponding adjustment to GNI and GDP. For instance, it is common to see some
companies report their irregular payment to various parties (e.g., bribes to people
outside the company, non-reported payments to managers to evade taxes) as pro-
duction costs. This practice understates the companies’ value-added, and if such un-
derstatement is widespread then GDP will be understated substantially.

Another source of gray income is leakage of public funds and public assets, and
transfer payments (e.g., bribes) among people. This part of gray income does not

17 Yangtse Evening Paper, 15 May 2006.

18 The FOF data are from the National Economic Census, which surveys the entire enterprise
sector and not just a sample of households. The NBS has not published the 2008 FOF data.
This number is derived from a linear projection of FOF data from 2005 to 2007. Data are
from NBS (various years).

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

Table 8. GNI before and after two adjustments for hidden income in 2005 and 2008 (RMB
trillion)

Household sector
Compensation of employees (wage income)
Non-wage income
Enterprise sector
Non-ªnancial
Financial
Government sector
Gross National Income

Source: NBS (2006, 2009), authors’ estimations.

Before adjustment

After adjustment

2005

11.06
9.28
1.78
3.73
3.60
0.13
3.83
18.41

2008

17.87
14.75
3.12
5.61
5.20
0.41
8.20
31.62

2005

13.73
9.28
4.45
3.20
3.09
0.11
3.29
20.01

2008

23.24
14.75
8.49
4.74
4.39
0.35
6.92
34.84

Note: The national income component data in 2008 before adjustment are estimated by linear projection from previous Flow of Funds

Accounts (NBS, 2005–07) with certain price adjustment. The data after adjustment are obtained by allocating the estimated hidden

income into each sector.

result in an understatement of GDP, but only increases household income (though
only a few would beneªt), at the expense of income distributed to government and
state-owned enterprises. This type of gray income reduces the income of some
groups but increases those of others.

We recalculated the GNI in 2005 and 2008 by assuming that 60 percent of the RMB
5.4 trillion of gray income was an understatement of the value-added of enterprises
and individual business; and that the remaining 40 percent was transfer from enter-
prise and government incomes to individuals. Tables 8 and 9 show the distribution
of national income to households, enterprises, and government before and after
these two adjustments.

Table 9 shows that before the adjustment, household income accounted for only
56.5 percent of national income in 2008. This share rises to 66.7 percent after the ad-
justment, an increase of 10 percentage points. Before the adjustment, the GNI share
of household income dropped by 3.6 percentage points between 2005 and 2008; but
with the adjustment, the decline was only 1.9 percentage points. This means that
when gray income is taken into consideration, the share of household income in
GDP is not that low, and has not been declining so rapidly.

Because most of the gray income is concentrated in the richest 10–20 percent of
households, China’s income inequality is much worse than shown in the ofªcial
data. And as the bulk of gray income is likely to have come from the diversion of
enterprise and government income and from the expropriation of the income (and,
sometimes, property) of politically weak households, such embezzlements could
cause social conºict and instability as well as economic inefªciency.

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

Table 9. Structure of GNI before and after two adjustments for hidden income in
2005 and 2008 (%)

Household sector
Compensation of employees (wage income)
Non-wage income
Enterprise sector
Non-ªnancial
Financial
Government sector

Sum as Gross National Income

Sources: NBS (2006, 2009), this study.

Before adjustment (%)

After adjustment (%)

2005

60.1
50.4
9.7
20.3
19.6
0.7
20.8

101.1

2008

56.5
46.7
9.9
17.7
16.4
1.3
25.9

100.2

2005

68.6
46.4
22.2
16.0
15.5
0.5
16.4

101.0

2008

66.7
42.3
24.4
13.6
12.6
1.0
19.9

100.2

Note: Income by sectors is disposable income, and their sum differs slightly from GNI. This explains why the sum of these ratios differs

slightly from 100 percent.

7. Final remarks

This paper analyzed the NBS statistics on household income alongside the income
and expenditure data collected by us, and found that the wealthiest 10 percent of ur-
ban households have a per capita disposable income of RMB 139,000 in 2008 instead
of the reported RMB 43,614, and that the second wealthiest 10 percent households
have a per-capita disposable income of RMB 54,900 instead of the reported RMB
26,250. The Gini coefªcient is probably much higher than the 0.47 to 0.50 calculated
by different experts.

This concentration of hidden income in the high-income groups is due to institu-
tional defects. Gray income has its origins in the misuse of power and is closely con-
nected with corruption. The widespread existence of gray income reveals that insti-
tutional reforms have lagged far behind economic reforms. Unless the government
could stay largely uninºuenced by the rent-seeking lobbying of capital owners and
other special interest groups, the free competition of the market economy would in-
evitably be replaced by the monopolistic practices of crony capitalism. Such a devel-
opment would accentuate income inequality, economic inefªciency, and social
conºict. To avoid these serious threats to economic development and social har-
mony, institutional reforms are essential, especially in the public ªnance system and
the government administrative system.19

19 The threat to economic development and social harmony in China comes from more than
just the lag in reforming its administrative system. Increasing tensions in relations with
other countries and deterioration of the natural environment are also becoming serious
threats; see Woo (2007).

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3

The Size and Distribution of Hidden Household Income in China

It might seem surprising to hear that China needs such fundamental and compre-
hensive reform when it has experienced high growth for almost 30 years. Why med-
dle with success? Why ªx the economy if it is not broken? The frank answer is that
the economy in 1978 was a broken economy and the story of the last 30 years has
been a story of successful repair. Post-1978 growth has stayed high because the gov-
ernment has continually changed policies to keep marketizing the economy, deep-
ening its integration into the international economy, and reducing the discrimina-
tion against the private sector. In short, policy changes and institutional reforms
were the reason for keeping growth high in 1979–2009, and the reforms process will
have to continue if future growth is to remain high.

References

Asian Development Bank. 2007. Key Indicators: Inequality in Asia.

Benjamin, Dwayne, Loren Brandt, John Giles, and Sangui Wang. 2005. Income Inequality Dur-
ing China’s Economic Transition. Manuscript.

National Audit Ofªce. 2010. Annual Auditing Report on Implementation of 2009 Central Gov-
ernment Budget and Other Budgetary Revenue and Expenditure. Available at
www.xinhuanet.com/

National Bureau of Statistics. Various years. China Statistical Yearbook. Beijing: China Statistics
Press.

Wang, Xiaolu. 2006. A Background Paper for the China Entrepreneur Survey System 2006 Re-
port. Working paper, National Economic Research Institute, Beijing.

Wang, Xiaolu. 2007. Grey Income and Income Inequality in China [Chinese]. Comparative
Studies 31. Beijing: CITIC Press.

Wilson, Dominic and Anna Stupnytska. 2007. The N-11: More Than an Acronym. Global Eco-
nomics Paper No. 153. Goldman Sachs. 28 March 2007.

Woo, Wing Thye. 1999. The Real Reasons for China’s High Economic Growth. The China Journal
41:115–137.

Woo, Wing Thye. 2001. Recent Claims of China’s Economic Exceptionalism: Reºections In-
spired by WTO Accession. China Economic Review 12(2/3):107–136.

Woo, Wing Thye. 2007. The Challenges of Governance Structure, Trade Disputes and Natural
Environment to China’s Growth. Comparative Economic Studies 40(4):572–602.

Wu, Ximing and Jeffrey M. Perloff. 2005. China’s Income Distribution, 1985–2001. Manuscript.

Xinhuanet. 2010. China Index Research Institute: Countrywide Land Remise Fee Reached RMB
1.5 Trillion in 2009. Available at www.xinhuanet.com/

Asian Economic Papers

D
o
w
n
o
a
d
e
d

f
r
o
m
h

t
t

:
/
/

d
i
r
e
c
t
.

i
t
.

e
d
u
a
s
e
p
a
r
t
i
c
e
–
p
d

f
/

1
0
1
1
1
6
8
2
6
8
6
a
s
e
p
_
a
_
0
0
0
6
4
p
d

b
y
g
u
e
s
t

o
n
0
7
S
e
p
e
m
b
e
r
2
0
2
3
Download pdf