The Chinese GDP Growth Rate Puzzle - 麻省理工学院人工智能研究专业

The Chinese GDP Growth Rate Puzzle
The Chinese GDP Growth Rate Puzzle

Harry X. 吴
School of Accounting and
Finance
Faculty of Business
The Hong Kong Polytechnic
大学
Kowloon
Hung Hom
香港, SAR, 中国
afhxwu@inet.polyu.edu.hk

Asian Economic Papers 6:1

The Chinese GDP Growth Rate Puzzle:
How Fast Has the Chinese Economy
Grown?*

抽象的
The Chinese statistical authorities recently revised the Chinese
GDP level and real growth rate for the period 1993–2004 follow-
ing China’s first national economic census for 2004. 然而, 这
methodology used in their revision is opaque. Using a trend-
deviation interpolation approach, this study has managed to repli-
cate the basic procedures of the revision and reproduced the offi-
cial estimates. Through this exercise, we have found that the esti-
mates that could be obtained by the straightforward interpolation
procedures were significantly modified. Based on a political econ-
omy argument, we attempt to explain why the revision had to
leave the growth rate of 1998 intact and why it had to bypass the
price issue and directly work on the real growth rate revision.
Based on previous studies and other observations, we also ques-
tion the census results on non-service industries.

1. 介绍

The long debate about the problem with China’s GDP sta-
tistics is in essence not a question of accuracy—indeed
there is no such thing as perfect national accounts in any
country that could produce faultless GDP ªgures—but a

* The earlier versions of this paper were presented at the Asian

Economic Panel Meeting held at the Korea Institute for Interna-
tional Economic Policy, Seoul, 20–21 March 2006, and at the 19th
General Conference of International Association for Research in
Income and Wealth (IARIW), Joensuu, 芬兰, 20–26 August
2006. Helpful comments and suggestions made by Bart van Ark,
Angus Maddison, Ren Ruoen, Huang Yiping, Kim Si Joong, 和
Wing Thye Woo, as well as the participants of the conferences,
are gratefully acknowledged. The author is responsible for any
error and omission. Please send correspondence to the author
via email.

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

question of how institutional or methodological problems may cause data fabrica-
tion or distortion and, more importantly, how institutional constraints may affect
the improvement of statistical methodology (吴 2000, 2002; Maddison 1998; Ren
1997; Woo 1998; Keidel 1992). These data problems are not only seen in the ofªcial
annual estimates1 but also in surveys and censuses. Without a doubt, allocating
more public resources to conduct surveys or censuses may improve statistical cover-
年龄, hence increasing the accuracy of statistics, but it alone will not be the solution
to the data problem. 在 31 十二月 2004, as one of the signiªcant efforts to im-
prove its growth estimates, China’s National Bureau of Statistics (NBS) 实施
China’s ªrst National Economic Census that covered all nonagricultural (中学
and tertiary) 活动, which totally mobilized 13 million personnel with an input
of nearly 2 billion yuan (People’s Daily online, 22 十二月 2005). Although the cen-
sus has discovered serious underreporting problems in service activities, the census-
based revised annual GDP estimates are not less questionable than what we had be-
fore.

在 20 十二月 2005, after about one year’s work on the census data, China’s NBS
announced that the census-based GDP estimate for 2004 was RMB 15,988 十亿.
This raises the nominal GDP level in 2004 by RMB 2,300 十亿, 或者 16.8 百分,
compared with the original estimates based on the annual statistics (13,688 十亿)
available in the China Statistical Yearbook (CSY) (见表 1). Of this upward revi-
锡安, 92.6 percent is attributed to services, which appears to support NBS’s long con-
cern about the under-coverage problem in the accounting of value-added by ser-
恶习 (see Xu 2002; Yue and Zhang 2002), 和 10.4 percent to industry (the Chinese
standard of industry includes mining, manufacturing, and utilities, IE。, 二 (中号) 在
桌子 1), which seems surprisingly small given the problems found in the Chinese
industrial statistics (see Wu 2002). There is also a 0.8 percent upward revision that is
attributed to agriculture, which does not seem to ªt into this nonagricultural activ-
ity–focused census. 然而, this revision has resulted in a positive 3.8 percent dis-
crepancy (IE。, 92.6 百分 (西德:2) 10.4 百分 (西德:2) 0.8 百分 (西德:3) 103.8 百分), 或者 88 曾是-
lion yuan, that has to be “made up” by an unexplained downward adjustment for
construction output (see panel E, 二 (C), 桌子 1).

在 9 一月 2006, NBS released its revision of China’s GDP level at current prices
and real GDP growth rate for the period 1993–2004. This revision did not cover the
pre-1993 period because of a previous output revision for the period 1978–92 after

1 Also known as “usual statistics” as used in the Chinese terminology, referring to statistics

based on data collected through the state statistical reporting system (made of the NBS chan-
nels plus statistical ofªces run by various ministries) that was developed during the central
planning period.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

桌子 1. The recent official revision of China’s nominal GDP level as a result of the 2004
人口普查: Confrontation of revised and original estimates (1992–2004年)

全部的

我

二 (中号)

二 (C)

三、

全部的

我

二 (中号)

二 (C)

三、

(A) Revised level (billion yuan)

(D) Revised growth rate (%)

1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
1992–2004年

2,664
3,533
4,820
6,079
7,118
7,897
8,440
8,968
9,922
10,966
12,033
13,582
15,988

580
689
947
1,202
1,389
1,427
1,462
1,455
1,472
1,552
1,624
1,707
2,096

1,028
1,419
1,948
2,495
2,945
3,292
3,402
3,586
4,003
4,358
4,743
5,495
6,521

142
227
297
373
439
462
499
517
552
593
647
749
869

914
1,199
1,628
2,009
2,346
2,717
3,078
3,410
3,894
4,463
5,020
5,632
6,502

23.2
32.6
36.4
26.1
17.1
11.0
6.9
6.2
10.6
10.5
9.7
12.9
17.7
14.6

9.7
18.7
37.5
26.9
15.5
2.7
2.5
(西德:4)0.5
1.2
5.4
4.7
5.1
22.8
11.2

27.2
38.0
37.3
28.1
18.0
11.8
3.3
5.4
11.6
8.9
8.8
15.8
18.7
16.3

39.4
60.1
30.8
25.8
17.6
5.4
7.9
3.7
6.8
7.4
9.0
15.9
16.1
17.3

(乙) Original (CSY) 等级 (billion yuan)

(乙) Original (CSY) growth rate (%)

2,664
3,463
4,676
5,848
6,789
7,446
7,835
8,207
8,947
9,732
10,517
11,739
13,688

580
688
946
1,199
1,384
1,421
1,455
1,447
1,463
1,541
1,612
1,693
2,077

1,028
1,414
1,936
2,472
2,908
3,241
3,339
3,509
3,905
4,238
4,598
5,309
6,282

142
229
301
382
453
481
523
547
589
638
701
818
957

914
1,132
1,493
1,795
2,043
2,303
2,517
2,704
2,991
3,315
3,608
3,919
4,372

23.2
30.0
35.0
25.1
16.1
9.7
5.2
4.8
9.0
8.8
8.1
11.6
16.6
14.6

9.7
18.7
37.4
26.8
15.4
2.7
2.4
(西德:4)0.5
1.1
5.4
4.6
5.0
22.7
11.2

27.2
37.5
36.9
27.7
17.7
11.4
3.0
5.1
11.3
8.5
8.5
15.5
18.3
16.3

39.4
61.5
31.9
26.8
18.6
6.2
8.7
4.6
7.6
8.3
9.9
16.8
17.0
17.3

26.5
31.2
35.8
23.4
16.7
15.8
13.3
10.8
14.2
14.6
12.5
12.2
15.4
13.9

26.5
23.9
31.8
20.2
13.8
12.7
9.3
7.4
10.6
10.9
8.8
8.6
11.6
13.9

(F) Change in growth rate (%) ( D E)

0
70
144
232
329
451
606
761
975
1,234
1,516
1,843
2,300

0
1
1
3
4
5
7
8
9
10
12
14
19

0
4
12
23
37
51
63
77
99
121
146
185
240

0
(西德:4)2
(西德:4)5
(西德:4)9
(西德:4)14
(西德:4)19
(西德:4)25
(西德:4)30
(西德:4)37
(西德:4)44
(西德:4)54
(西德:4)69
(西德:4)88

0
67
135
215
303
414
561
706
904
1,147
1,412
1,713
2,130

0.0
2.6
1.4
1.1
1.0
1.3
1.7
1.5
1.6
1.8
1.7
1.3
1.1
1.5

0.0
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1

0.0
0.4
0.4
0.4
0.4
0.3
0.3
0.3
0.4
0.3
0.3
0.4
0.4
0.4

0.0
(西德:4)1.3
(西德:4)1.0
(西德:4)1.0
(西德:4)0.9
(西德:4)0.8
(西德:4)0.9
(西德:4)0.9
(西德:4)0.9
(西德:4)0.8
(西德:4)0.9
(西德:4)0.9
(西德:4)0.9
(西德:4)0.9

0.0
7.3
3.9
3.2
2.9
3.1
4.0
3.4
3.6
3.7
3.7
3.6
3.9
3.8

来源: The revised estimates are published at the NBS website (www.stats.gov.cn/tjdt/zygg/t20060109 _402300176.htm) following

the official announcement on 9 一月 2006. The former or CSY annual estimates are available from China Statistical Yearbook

(NBS 2005 and earlier issues).

Notes: 1992 is used as the initial benchmark that is not included in the revision.

我

primary; 二

secondary that includes II (中号) (manufacturing, 矿业, utilities) and II (C) (建造); 三、

第三.

The implicit GDP deflator is expressed as percent change from the previous year. It is derived by the definition: P V/Q, where P is

price index, V is value index, and Q is volume index.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

China’s ªrst tertiary census for 1992. 桌子 1 compares the 2004 census-based revi-
sion of the nominal GDP ªgures with the former estimates by major sector in 1992–
2004. It shows that the revision has raised the nominal GDP growth rate from 13.9
到 17.8 percent per annum for services (Sector III), 从 16.3 到 16.6 for industry (二
(中号)) 和来自 11.2 到 11.3 for agriculture (我). As for construction (二 (C)), the revision
有, 然而, lowered its nominal growth rate from 17.3 到 16.3 percent per annum.
因此, the nominal growth rate of the total GDP has been revised from 14.6 到
16.1 percent per annum for this period.

表中 2, we further compare the revised real GDP growth rates and their implicit
deºators with the original estimates. It appears that NBS has attributed the entire
upward revision of real output to services, which raises the real growth rate of the
service output from 8.6 到 10 percent per annum. 因此, China’s real GDP
growth rate has been raised from 9.4 到 9.9 percent per annum. Taking into account
the nominal adjustment as reported in Table 1, this revision implies that the implicit
inºation rate over this period has been raised from 4.8 到 5.7 percent per annum for
the economy as a whole and from 4.9 到 7.0 percent per annum for services.

然而, NBS does not explain why the 7.4 percent of the nominal revision for 2004
that is attributed to all the non-service sectors should be treated as a pure price ef-
影响. Strikingly, the real GDP growth rate for 1998 remained unadjusted at 7.8 每-
分, reºecting the much-disputed growth performance of the Chinese economy at
the height of the Asian ªnancial crisis. 数字 1 depicts the impact of the 2004 Cen-
sus-based revision on the real growth performance of China’s service output and
total GDP.

在本文中, we raise some questions about NBS’s 2004 census-based GDP revision
and explore their likely implications for China’s growth performance. Our objective
is to be constructive to the process of improving the accuracy of the Chinese na-
tional accounts by examining methodological issues.

Our ªrst methodological inquiry is to identify the procedures that were used by
NBS in the revision. Because the information about the approach used in the revi-
sion suggests that NBS followed the standard interpolation procedures in the revi-
锡安, we will try to duplicate what NBS did to see whether there are any departures
from the standard procedures.

Our second methodological inquiry focuses on the underlying price problem. 这
type of census by nature cannot obtain information on price changes. Even if the
1992 benchmark could be assumed problem-free and thus the under-coverage

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

桌子 2. The recent official revision of China’s real GDP growth rates and GDP deflators as
a result of the 2004 人口普查: Confrontation of revised and original estimates (1992–2004年)

全部的

我

二 (中号)

二 (C)

三、

全部的

我

二 (中号)

二 (C)

三、

(A) Revised real growth rate (%)

(D) Revised implicit deflator (%)

14.2
14.0
13.1
10.9
10.0
9.3
7.8
7.6
8.4
8.3
9.1
10.0
10.1
9.9

4.7
4.7
4.0
5.0
5.1
3.5
3.5
2.8
2.4
2.8
2.9
2.5
6.3
3.8

21.2
20.1
18.9
14.0
12.5
11.3
8.9
8.5
9.8
8.7
10.0
12.8
11.5
12.2

21.0
18.0
13.7
12.4
8.5
2.6
9.0
4.3
5.7
6.8
8.8
12.1
8.1
9.1

12.4
12.1
11.0
9.8
9.4
10.7
8.3
9.3
9.7
10.2
10.4
9.5
10.0
10.0

7.9
16.4
20.6
13.7
6.4
1.5
(西德:4)0.9
(西德:4)1.3
2.1
2.1
0.6
2.6
6.9
5.7

4.7
13.4
32.2
20.9
9.9
(西德:4)0.7
(西德:4)1.0
(西德:4)3.2
(西德:4)1.2
2.6
1.7
2.5
15.5
7.2

4.9
14.9
15.5
12.3
4.9
0.4
(西德:4)5.1
(西德:4)2.8
1.7
0.1
(西德:4)1.1
2.7
6.4
4.0

15.2
35.7
15.1
11.9
8.4
2.7
(西德:4)1.0
(西德:4)0.5
1.0
0.6
0.2
3.4
7.4
6.6

12.5
17.1
22.3
12.4
6.7
4.6
4.6
1.3
4.1
4.0
1.9
2.5
5.0
7.0

(乙) Original (CSY) real growth rate (%)

(乙) Original (CSY) implicit deflator (%)

14.2
13.5
12.6
10.5
9.6
8.8
7.8
7.1
8.0
7.5
8.3
9.5
9.5
9.4

4.7
4.7
4.0
5.0
5.1
3.5
3.5
2.8
2.4
2.8
2.9
2.5
6.3
3.8

21.2
20.1
18.9
14.0
12.5
11.3
8.9
8.5
9.8
8.7
10.0
12.8
11.5
12.2

21.0
18.0
13.7
12.4
8.5
2.6
9.0
4.3
5.7
6.8
8.8
12.1
8.1
9.1

0.0
0.5
0.5
0.4
0.4
0.5
0.0
0.5
0.4
0.8
0.8
0.5
0.6
0.5

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

12.4
10.7
9.6
8.4
7.9
9.1
8.3
7.7
8.1
8.4
8.7
7.8
8.3
8.6

0.0
1.4
1.4
1.4
1.5
1.6
0.0
1.6
1.6
1.8
1.7
1.7
1.7
1.5

7.9
14.6
19.9
13.2
5.9
0.8
(西德:4)2.4
(西德:4)2.2
0.9
1.2
(西德:4)0.2
1.9
6.5
4.8

4.7
13.3
32.1
20.8
9.8
(西德:4)0.8
(西德:4)1.1
(西德:4)3.3
(西德:4)1.3
2.5
1.6
2.5
15.4
7.2

4.9
14.5
15.1
12.0
4.6
0.1
(西德:4)5.4
(西德:4)3.1
1.4
(西德:4)0.2
(西德:4)1.4
2.4
6.1
3.6

15.2
36.9
16.0
12.8
9.3
3.5
(西德:4)0.2
0.3
1.8
1.4
1.0
4.2
8.2
7.5

(F) Change in implicit deflator
(%) ( D E)

0.0
1.8
0.7
0.6
0.5
0.7
1.5
0.9
1.1
0.9
0.8
0.7
0.4
0.9

0.0
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1
0.1

0.0
0.4
0.4
0.4
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3
0.3

0.0
(西德:4)1.1
(西德:4)0.9
(西德:4)0.9
(西德:4)0.9
(西德:4)0.8
(西德:4)0.8
(西德:4)0.8
(西德:4)0.8
(西德:4)0.8
(西德:4)0.8
(西德:4)0.8
(西德:4)0.9
(西德:4)0.9

12.5
11.9
20.3
10.9
5.5
3.3
0.9
(西德:4)0.3
2.3
2.3
0.1
0.8
3.0
4.9

0.0
5.1
2.0
1.5
1.2
1.3
3.7
1.6
1.8
1.7
1.8
1.7
1.9
2.1

1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
1992–2004年

Source and Notes: 见表 1.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

数字 1. China’s growth performance of real GDP: Revised versus original estimates
(percent per annum)

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

来源: 桌子 2, panels (A) 和 (乙).

problem is entirely due to new services and products that emerged after 1992, 这
Chinese statistical authorities still face complicated price problems. 因为
prices of new services and products are usually high at earlier stages and decline
quickly throughout the stages of maturing, it is apparently difªcult to make plausi-
ble assumptions for price changes in the revision. 此外, leaving the real GDP
growth rate for 1998 completely unadjusted suggests that the level amendment
for this year is a pure price effect, which appears to be inconsistent with the system-
atic adjustment of data over the entire period. Our question about the underlying
price problems can be illustrated by Figure 1. Users of the Chinese GDP estimates
would naturally expect an explanation about the assumptions that were used for
price changes across individual sectors that could warrant such a revision of the real
输出.

最后, it is difªcult to accept the ªndings of the 2004 census that the ofªcial annual
output estimates for manufacturing and mining industries are basically free from
问题. How could we reconcile such ªndings with numerous disclosures of data
fabrications in industrial output made by local ofªcials, SOE managers, 乃至
private ªrms (with different incentives)? 实际上, it has been reported that data fabri-
cation to exaggerate local performance could be even more serious in a national
event like a census than in the usual reporting exercise that supports the annual esti-
运动. Taking into account these possibilities, we would like to propose in the con-
clusion some conjectures about the likely real GDP performance over this period.

Asian Economic Papers

The Chinese GDP Growth Rate Puzzle

2. The basic approach used in the nominal GDP revision

To the best of our understanding of the brief explanation in the ofªcial announce-
蒙特 (NBS 2006), the basic approach used by NBS in the 2004 census-based revision
of the nominal GDP estimates contains three major steps:

Step 1: Calculating the (简单的) deviation of the NBS annual GDP estimates, 酒吧-

lished in the China Statistical Yearbook (hereafter the CSY GDP series), 从一个
“historical trend” that is also derived from the CSY GDP series.

Step 2: Deriving a “new trend” for the same period using the same GDP estimates
为了 1992 (as in Step 1) and the census-based GDP estimates for 2004, 哪个
generates a series of trend-value GDP estimates for this period.

Step 3: Interpolating GDP for individual years between the two benchmarks of

the new trend by adjusting the trend-value GDP estimates obtained in Step
2 by the annual deviation value obtained in Step 1 based on the CSY GDP
系列.

This approach may be best described as the trend-deviation interpolation method
often used by national account statisticians. Based on what is explained by NBS and
the standard procedures of the trend-deviation interpolation method,2 我们有
managed to repeat the likely procedures used by NBS in the revision. Our ªndings
show that although NBS in principle adopted the standard interpolation procedures
in revising the nominal GDP, they allowed discretionary modiªcations in the exer-
cise. 下文中, we will ªrst present the standard interpolation procedures of
the trend-deviation method, then derive nominal GDP estimates following the pro-
cedures, and ªnally compare our results with the NBS estimates and discuss their
implications.

Strictly speaking, the trend-deviation interpolation method requires an indicator
(我) that is considered highly correlated with the variable (X) to be estimated. 这
indicator is an existing time series, and the variable to be estimated has only two
benchmark values based on surveys or censuses. The indicator is used to obtain the
deviation of its actual value from its trend value (IE。, trend-deviation ratio) 为了
every time point of the period concerned. By applying the indicator’s trend-
deviation ratio to the variable to be estimated, it allows the annual movement pat-
tern of the variable to follow that of the indicator.

2 See an introduction to the procedures by Derek Blades (2004), the former chief of national ac-
counts at OECD, prepared for the NBS/Asian Development Bank Workshop of “Improving
Service Statistics in China” in Shanghai in November 2004.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

第一的, let us assume that both I and X generally follow an exponential trend, 然后我们
could estimate their trend growth rates over a given period n, beginning from the
time point 0 to the current time T (时间 (西德:4) 0 (西德:3) n), using the following equations for I
and X, 分别:

和

我
r
趋势

(西德:2)

经验值

我
ln

时间

(西德:4)
(西德:5)(西德:6)

(西德:3)

我
ln 0

(西德:7)
(西德:8)(西德:9) (西德:3)

X
r
趋势

(西德:2)

经验值

时间

(西德:4)
(西德:5)(西德:6)

(西德:3)

X
ln 0

(西德:7)
(西德:8)(西德:9) (西德:3)

(1A)

(1乙)

第二, we could use the estimated growth rate of the trend to calculate the trend
value at time t over this period for I and X, 分别:

和

我

趋势
t

(西德:2)

我

0 1(

(西德:10)

我
r
趋势

)

趋势
t

(西德:2)

0 1(

(西德:10)

X
r
趋势

t
) .

最后, the X series over the period could be estimated by multiplying the trend
value of X by a parameter D based on the indicator I:

(西德:2)

我
D X
t

趋势
t

(2A)

(2乙)

(3)

(西德:2)

我
t

actual
where D
我
t
its own exponential trend.

趋势
t

我
/

, IE。, the deviation of the actual value of the indicator I from

In the current case of China, the indicator is the CSY GDP series rather than any
other non-GDP indicator that is closely associated with the variation of GDP.
恰恰, the revised GDP series is obtained by adjusting the 1992 和 2004 岑-
suses-based trend values (修改) by the deviations derived from the CSY GDP
系列 (original), 那是,

修改
GDP
t

(西德:2)

original
t

trend revised
,
GDP
t

(4)

(西德:2)

original
t

original
GDP
t

trend original
,
GDP
/
t

the deviation of the actual value from the
where D
,
trend value estimated based on the CSY series. One of the most obvious merits of
this interpolation method is to make use of all available information in the existing
GDP estimates (the CSY series) and from the newly conducted census.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

桌子 3. Comparison of straightforward estimates by equation (4) and the NBS revised
估计 (billion current yuan)

(A) Results of interpolation by eq. (4)

(乙) Nominal gap
( NBS revised estimates

eq. (4) 估计)

全部的

2,664
3,505
4,787
6,052
7,104
7,898
8,432
8,966
9,923
10,967
12,035
13,589
15,988

我

580
689
947
1,202
1,389
1,426
1,462
1,455
1,472
1,552
1,624
1,707
2,096

1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004

二 (中号)

二 (C)

三、

全部的

我

二 (中号)

二 (C)

三、

1,028
1,419
1,948
2,495
2,945
3,292
3,402
3,586
4,003
4,358
4,743
5,495
6,521

142
227
297
373
439
462
499
517
552
593
647
749
869

914
1,170
1,595
1,982
2,332
2,717
3,070
3,408
3,896
4,465
5,021
5,638
6,502

0.0
28.7
33.0
27.6
13.9
(西德:4)0.4
8.1
1.4
(西德:4)1.9
(西德:4)2.0
(西德:4)1.6
(西德:4)6.3
0.0

0.0
0.0
0.0
0.0
0.0
0.1
0.0
0.0
0.0
(西德:4)0.1
0.1
0.0
0.0

0.0
0.0
0.0
0.1
0.0
0.0
(西德:4)0.1
0.0
0.1
0.0
0.0
0.0
0.0

0.0
(西德:4)0.1
0.0
0.0
0.0
0.0
0.1
0.0
0.0
0.1
0.0
0.1
0.0

0.0
28.7
33.0
27.5
13.9
(西德:4)0.5
8.1
1.4
(西德:4)2.0
(西德:4)1.9
(西德:4)1.7
(西德:4)6.3
0.0

来源: Author’s calculation using equation (4). NBS revised nominal GDP data are from Table 1, panel (A).

We have produced new estimates for every component of GDP using equation
(4) and reported them in panel (A) of Table 3. In panel (乙), we have calculated the
nominal value gap between our results and the NBS revised estimates to see if the
standard interpolation procedures of the trend-deviation interpolation method were
followed by NBS. If the standard procedures were strictly followed, the expected
value of the “nominal gap” should be zero. 桌子 3 shows that NBS did indeed
adopt the standard procedures of the trend-deviation interpolation method for all
the non-service sectors, but introduced some unexplained modiªcations to the re-
sults for services obtained by the standard procedures.

数字 2 compares the NBS revised nominal GDP level with both the original CSY
estimates and our equation (4)–based results. The sector and industry codes are
deªned as those in Table 1, namely primary (我), 中学 (二), and tertiary (三、). 这
secondary sector is further divided into two subgroups: manufacturing, 矿业, 和
utilities, 二 (中号); and construction, 二 (C). Setting the levels in 1992 成为 1, panel (A)
图中 2 shows that all sectors are affected by the revision, with the tertiary sector
affected most and positively, followed by the construction industry, but negatively.

控制板 (乙) of Figure 2 shows whether the revision exactly followed the standard pro-
cedures of the trend-deviation interpolation method. If the NBS revision had strictly
followed the standard procedures, its results would have completely overlapped
with our results, 那是, the ratio for all sectors would have been equal to one at all
time points. Although this is basically true for the non-service sectors, it is not the
case for the tertiary sector. One can see that the NBS revised estimates for the ter-

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

数字 2. How much has been adjusted and how close to the standard procedures? (nominal
GDP indices)

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

来源: Indices in panel (A) are calculated using the data in panels (A) 和 (乙) of Table 1. Indices in panel (乙) above are calculated using

the data in panel (A) of Table 1 and the data in panel (A) of Table 3.

tiary sector drift from the baseline at most time points with a rather irregular pat-
tern, which conªrms that NBS did introduce adjustments to the results obtained
through standard procedures.

It is difªcult to understand why these adjustments were introduced into the nomi-
nal GDP revision. 实际上, as we will discuss below, any ad hoc amendment to the
nominal GDP estimates obtained by the standard procedures was really unneces-
sary because NBS (rather unexpectedly) bypassed the price problem in the revision
of the real GDP growth estimates. Another question related to panel (乙) 图中 2
is that if the 1992 benchmark were indeed problem-free as assumed by NBS in this
revision, why was the nominal service GDP in 1993 obtained through the standard
procedures substantially upward adjusted?

3. The revision of the real GDP growth rate estimates

To obtain the estimates for the real GDP growth rates, one needs proper deºators.
然而, China’s ªrst economic census in 2004 did not include any survey on
价格, which is not a usual task in this type of census. All output and income data
collected in the census are stock information at the time of the census and only in

Asian Economic Papers

The Chinese GDP Growth Rate Puzzle

数字 3. Implicit GDP deºators for the service and construction sectors: Revised versus
original estimates (1992

100)

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

来源: Author’s calculation based on data from panels (D) 和 (乙), 桌子 2.

nominal terms. 所以, in the current context, the price issue is basically inde-
pendent. 然后, what assumptions were applied to the price changes of individual
sectors that could warrant the revision of the real GDP estimates as shown in Fig-
乌尔 1?

3.1 Changes in implicit deºators
Our investigation starts with the implicit GDP deºators of individual sectors re-
ported in panels (D) 和 (乙) of Table 2. Both the original and revised estimates-
derived implicit deºators were converted into 1992-based indices to demonstrate
the impact of the revision on price change over time. The results for services and
construction are shown in Figure 3. We choose these two sectors because they were
most affected in the revision, especially in nominal terms and hence in price de-
ºators, and because the revision resulted in an opposite effect on the two sectors.
The revision has resulted in higher price levels for services and lower price levels
for construction over the whole period in question. 一般来说, the underlying trend
has changed substantially in both cases, but the annual pattern of the movement re-
mains similar. 然而, a closer examination shows that in each case the revision of
the price index for 1998 created an obvious “outlier” that was not in line with the
original annual pattern and largely responsible for the slope change of the trend.

Asian Economic Papers

The Chinese GDP Growth Rate Puzzle

这一观察, together with Figure 1, which shows no revision to the real GDP
growth rate in 1998, takes us back to the hot debate in the early 2000s about whether
the statistical authorities had arbitrarily raised the growth rate for 1998 when China
was badly hit by the Asian ªnancial crisis. The ofªcial estimate of the real GDP
growth rate for 1998 曾是 7.8 百分, 仅有的 0.2 points lower than the government’s
8 percent growth target for that year, suggesting that the target was virtually met in
a difªcult economic situation. 然而, 这 1998 growth rate has been challenged
by many researchers as an overestimate of China’s real growth performance in 1998.
例如, based on the change of energy consumption for 1997–99, Rawski
(2001) suspected that China’s real GDP growth in 1998 ranged from (西德:4)2 到 (西德:2)2 每-
分. But his estimation was criticized by Ren (2002) among others for lacking sound
empirical support. Other researchers have used the expenditure approach (骗子-
trast to the NBS’s value-added approach) to arrive at different results. Keidel (2001)
found the growth rate in 1998 to be between 6.9 和 7.2 百分, whereas Shiau
(2005) showed that it was between 2.6 和 4.7 百分. Such variations in estimates
are largely due to different choices of deºators.3 Without a proper justiªcation on
the choice of deºators, the census-based revision is surely in no position to face
these challenges.

很遗憾, the infamous “7.8 percent” for 1998 is a big dilemma that NBS could
not easily bypass when revising China’s real GDP growth rate. 一方面,
NBS could not systematically raise the 1998 growth rate together with the overall
upward revision for the whole period because that would invite further interna-
tional criticisms. 另一方面, it could not take this chance to make a reason-
able downward revision for 1998 because that would amount to the admission that
the original estimate was a mistake, whose implications would be by no means
purely technical to the authorities. Although leaving 1998 intact in this overall up-
ward revision means that the growth rate of 1998 is in fact relatively lowered, 这样的
an arbitrary treatment has made the whole data revision less credible.

这 1998 anomaly suggests that there were certainly ad hoc modiªcations in the re-
想象, but one cannot be sure whether they were made on prices or real growth
费率. 尽管如此, by assuming that the 1992 benchmark was problem-free, NBS
faced more complicated price problems. This assumption means that the under-

3 It should be mentioned that there are also different views. Using the principal component

分析, Klein and Ozmucur (2002) found that the variation of the ofªcial GDP growth was
well associated with the variation of 15 major macroeconomic indicators, suggesting that the
ofªcial GDP estimates were not outliers. 尽管如此, because the major indicators are from
the same ofªcial sources that generate the information for the GDP estimation, surely no sen-
sible inference can be made from their ªndings in the context of this debate.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

coverage problem is mainly due to the new products and services that emerged af-
特尔 1992. Because the prices of new products/services are usually high at the early
stages and decline quickly throughout the stages of maturing, it is therefore almost
impossible for NBS to introduce a new trend to adjust the original price changes in
the absence of necessary price information for new products and services. 然后,
how did NBS solve the price problem in the revision? Our working hypothesis is
that NBS did not directly work on prices; instead it began with a new GDP growth
trend that could satisfy a certain growth target for the period in question and then
followed the trend-deviation interpolation method to adjust the original annual real
growth rates. 换句话说, the new (修改) deºators as shown in Figure 3 是
merely indirect results of the revised real growth rate estimates rather than actual
price changes that are independent of the revision.

3.2 The government growth target
What is the government growth target? How could that target affect NBS’s revision
of the real growth estimates? The “state of the art” of the planning administration in
China during the central planning era was “leaving room” [liu you yu di] for ful-
ªlling annual or ªve-year plans (FYPs). Because being below the planned target was
politically unacceptable, economic authorities at all levels tended to leave enough
“room” so that they could easily meet or even exceed their targets. As Table 4
节目, there was virtually no target undershooting in any of the FYPs since the
1980s.4 However, it shows that the extent to which the target was exceeded varied
greatly. Our hypothesis is that this outcome is largely the consequence of a game be-
tween the lower level and the upper level within the planning hierarchy. A great ex-
cess above the planned target (as seen in the 6th and 8th FYPs, 桌子 4) often sends a
signal to the upper authorities indicating that lower authorities might have deliber-
ately left too much “room” in the current plan. 因此, in the next FYP the central au-
thorities tend to set a higher growth target to tap the potentials maximally. 这可能
leave little room for overfulªlling the plan, 然而. In such a case the plan may be
just met or marginally overshot (as seen in the 7th and 9th FYPs). One could also in-
terpret this result as a warning from the lower authorities that if such a high target is
to be maintained, there may be a good chance to miss it. That is why we could see
that a marginal excess of the target is often followed by a downward adjustment of
the planned target in the next FYP (as seen in the 8th and 10th FYPs).

In such an interactive process, changes in national plans can be rather erratic, 和
hence the macroeconomic performance tends to be volatile. For the current discus-

4 This is also the case during the pre-reform period, but that period is beyond the scope of this

学习.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

桌子 4. Growth target and actual growth achieved of China’s five-year plans (FYPs) (在
百分)

Five Year Plan
(period covered)

生长
目标

Target adjusted
compared with
the last FYPa

生长
速度
达到了

Growth rate
达到了
(修改)C

Growth target
overshot or
undershotb

Growth target over-
shot or undershotb
(修改)C

The 6th Five-Year
Plan (1981–1985)
The 7th Five-Year
Plan (1986–1990)
The 8th Five-Year
Plan (1991–1995)
The 9th Five-Year
Plan (1996–2000)
The 10th Five-Year
Plan (2001–2005年)
The 11th Five-Year
Plan (2006–2010)

4.0

7.5

6.0

8.0

7.0

7.5

n.a.

87.5

(西德:4)20.0

33.3

(西德:4)12.5

7.1

10.7

7.9

12.0

8.3

8.8

n.a.

12.3

8.6

9.4

n.a.

167.5

5.3

100.0

3.8

25.7

n.a.

105.0

7.5

34.3

n.a.

来源: Information on the growth target of various five-year plans is available from ECACE (Editorial Committee of Almanac of

China’s Economy) (various volumes) and the Web site of NDRC (National Development and Reform Commission) (http://

ghs.ndrc.gov.cn/). The GDP growth rate for 2005 是 9.9 百分 (as reported by the New China News Agent, 6 行进 2006), 哪个

is used for calculating the average growth rate of the 10th FYP.

笔记: A. Calculated as (current target rate/previous target rate

100.

乙. Calculated as (actual rate/target rate

100.

C. Calculated using the recent 2004 census-based revision of growth rates.

n.a.: not available.

锡安, it is important to note that the situation began to change in the early 1990s
when the market was allowed to play more important roles in economic decisions.
Around the mid 1990s, authorities began to exercise monetary and ªscal policies to
replace the traditional planning administration, aiming to smooth out aggregate vol-
atility. The highly volatile or stop-go macroeconomic performance5 under the Zhao
Ziyang and Li Peng administrations from the mid 1980s to the mid 1990s has taught
later governments some important lessons: growth should not be too fast so that the
necessary macroeconomic balances can be maintained, especially in the energy, min-
erals, and transportation markets. 然而, growth has to be fast enough to create
jobs and hence reduce the pain of the reform of the state sector.

然而, how fast is fast enough? The leadership’s “rule of thumb” is close to but not
多于 10 percent a year. This has been seen in discussions of various central
governments’ think tanks. 例如, 刘 (1999) described an “ideal policy goal”
as a “high growth rate that could be nine but below ten percent a year and low
inºation that should not exceed three percent a year.” Given this “rule of thumb,”
and the fact that high growth had already been achieved, there was really not much
room for NBS to play in the revision. 一方面, NBS strongly believed that
the service GDP had been indeed underestimated (which would have some positive

5 See Woo (2006) for a recent analysis of the stop-go characteristic of the Chinese economy.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

impact on the overall growth), and on the other hand, NBS found it politically
difªcult to adjust downward the real growth rate of any other sector. NBS might
have tried several scenarios in the revision, but it looks to us that an annual growth
率 9.9 percent was the most acceptable rate for the period 1992–2004. 因为
the impact of this upward revision on the entire reform (post-1978) period is merely
0.2 百分点 (从 9.4 到 9.6 as announced when NBS released their re-
vised growth estimates), this result may not signiªcantly aggravate international
critics who believe that China’s post-reform GDP growth performance has already
been exaggerated (例如, Maddison 1998, 2006). 尽管如此, it indeed presents a
much better FYP performance for the current administration, 那是, the revised an-
nual real GDP growth rate for the 10th FYP period (2001–05) 是 9.4 而不是 8.8
百分 (桌子 4)!

3.3 What if the original deºators were adopted?
To support our growth target hypothesis, we have to see what would be the effect
on China’s GDP growth rate if NBS just used the original GDP deºators (panel (乙),
桌子 2), which is certainly the most logical choice for NBS. Our results show that
if the original NBS price deºators were used, the annual growth rate of China’s GDP
之间 1992 和 2004 将会 10.8 per annum, instead of the revised rate of 9.9
百分, exceeding the “rule of thumb” target rate of 10 百分.

数字 4 shows the effect of the original and the “new deºators” on the real growth.
The NBS newly revised estimates for the total and service GDP in nominal terms are
deºated by the two deºators, respectively.6 We want to highlight three points about
数字 4. 第一的, the already overheated economy in 1993 as seen by the original data
would be much worse (more out-of-control growth). 第二, the Li Peng adminis-
tration had claimed a “soft landing” in 1996 from the high growth of 1993. 然而,
the use of original deºators would show GDP growth dropping from 15.8 百分比在
1993 到 10.5 百分比在 1996 rather than from 14.0 到 10.0 百分. 第三, the widely
criticized overstatement for the performance in 1998 when China was hit during the
Asian ªnancial crisis would look even more exaggerated (9.5 percent instead of
7.8 百分). All these ªndings are politically difªcult to accept. 所以, 这是真实的-
sonable to believe that in the absence of “satisfactory” deºators, it was inevitable for
NBS to work directly on the revision of the real growth estimates to ensure that the
government’s “rule-of-thumb” growth target would be met.

6 Note that the series deºated by the new deºators are the same as those shown in Figure 1

(IE。, the one labeled with “revised” in Figure 1), whereas the series deºated by the original
deºators are from our calculation.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

数字 4. What if the original deºators were adopted? (percent per annum)

来源: Nominal GDP are deºated by the implicit deºators. Data for nominal GDP are from panels (A) 和 (乙) of Table 1. Data for

deºators are from panels (D) 和 (乙) of Table 2.

3.4 Revising real growth estimates by interpolation
To investigate the actual procedures adopted by NBS to work out the real GDP
growth we begin with the new “trend rate” of 9.9 百分 (参见方程 (1乙)) 为了
whole economy in 1992–2004. 然后, we use this growth rate to generate a series of
“trend values” as a 1992-based index (based on equation (2乙)). 下一个, we adjust this
index by a set of deviation parameters (D, in equation (4), equivalent to equation
(3)) obtained from the original real output index also with 1992 as the benchmark.
The same exercise is done for every sector. 然而, we follow NBS to assume that
there is only pure price effect on the output of the non-service sectors, 那是, no ad-
justment to the CSY estimates for 1992 和 2004 (see panel (C), 桌子 2).

控制板 (A) of Table 5 reports our results by the standard procedures. In panel (乙) 我们的
results are compared with those of NBS with the latter as the base ((西德:3)1). The compar-
ison conªrms our expectations. 第一的, for all non-service sectors our results have ex-
actly replicated what was reported by NBS, supporting our hypothesis that NBS had
bypassed the price problem and worked directly on the revision of the real growth
费率. 第二, for services, we again observed some ad hoc downward modiªca-
系统蒸发散 (as the discrepancies are all positive—panel (乙)) made to the results obtained
through the standard procedures. Logically, such modiªcations create what we call
“real value gaps” that should be compensated by opposite price effect.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

桌子 5. Confrontation of real GDP growth indices based on equation (4) 以及那些
constructed by NBS revised estimates (1992

100)

A) 方程 (4)–derived growth index

乙) Growth index ratio
(NBS Index

1.000)

全部的

我

二 (中号)

二 (C)

三、

全部的

我

二 (中号)

二 (C)

三、

1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004

100.0
114.0
129.0
143.1
157.6
172.2
186.5
200.6
217.6
235.0
255.7
281.2
309.3

100.0
104.7
108.9
114.3
120.2
124.4
128.7
132.3
135.5
139.3
143.3
146.9
156.2

100.0
120.1
142.8
162.8
183.1
203.8
222.0
240.8
264.4
287.5
316.2
356.7
397.7

100.0
118.0
134.2
150.8
163.6
167.9
183.0
190.9
201.7
215.4
234.4
262.8
284.1

100.0
112.2
124.6
136.9
149.6
165.4
181.6
198.1
217.1
238.4
262.6
286.9
314.9

来源: Author’s calculation for Panel A.

笔记: NBS revised estimates are from Tables 1 和 2.

1.000
1.000
1.000
1.001
1.002
1.002
1.006
1.006
1.007
1.004
1.001
1.001
1.000

1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000

1.000
1.001
1.001
1.002
1.001
1.000
1.013
1.012
1.010
1.007
1.005
1.002
1.000

下一个, 表中 6, using our estimates of both the nominal GDP (桌子 3) and the real
growth rates (桌子 5), we work out the implicit GDP deºator for individual sectors,
then compare it with the implicit deºators derived from the NBS revised estimates
(panel (D), 桌子 2) with the latter as the benchmark ((西德:3)1). If the ratio differs from
一, it indicates some “price effect” created by ad hoc modiªcations to the estimates
that could be obtained through the standard interpolation procedures as given by
方程 (4). We shall call such an effect in this context a “price gap.” In the current
案件, a “price gap” should be less than one as we expected. Our results indeed show
that it is the case for the tertiary sector. We depict both “real gaps” and “price gaps”
图中 5. Without “nominal gap,” one would expect to see the two gaps mirroring
彼此. Apparently, this is not the case, a situation that deserves further investi-
gation.

Taking 1993 as an example, the real GDP ratio is 0.1 percent above the benchmark
[(西德:3) (1.001 (西德:4) 1) (西德:5) 100], 但是 (implicit) deºator ratio is 2.5 percent below the bench-
标记 [(西德:3) (0.975 (西德:4) 1) (西德:5) 100]. How can this be explained? 实际上, what is missing here
are NBS’s ad hoc modiªcations to the nominal values that we discovered earlier (看
桌子 3). By adding the “nominal gap” that could be derived from Table 3, also tak-
ing NBS ªgures as the base ((西德:3)1), the “real gap” in Figure 5 can now be explained.
This could be double-checked by using the implicit deºator to calculate a new set of
the “real gap,” which we found is exactly the same as the one obtained by directly
comparing our results (standard procedures) with the NBS results.

实际上, 数字 5 demonstrates what arbitrary modiªcations in nominal or real terms
that NBS had to make in order to arrive at their desired growth rate estimates.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

桌子 6. Confrontation of implicit deflators based on equation (4) results and those by NBS
revised estimates

A) Implicit GDP deflator

乙) Deflator ratio (NBS index

1.000)

全部的

我

二 (中号)

二 (C)

三、

全部的

我

二 (中号)

二 (C)

三、

1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004

100.0
115.4
139.4
158.7
169.2
172.2
169.7
167.8
171.2
175.2
176.7
181.4
194.0

100.0
113.4
150.0
181.3
199.2
197.7
195.8
189.6
187.2
192.1
195.3
200.3
231.3

100.0
114.9
132.6
149.0
156.3
157.0
149.0
144.8
147.2
147.4
145.9
149.8
159.4

100.0
135.8
156.2
174.8
189.5
194.6
192.5
191.5
193.5
194.6
194.9
201.5
216.3

100.0
114.2
140.1
158.5
170.5
179.7
185.0
188.2
196.4
204.9
209.2
215.0
226.0

来源: Author’s calculation based on Table 5 for panel (A).

笔记: NBS revised estimates are from Tables 1 和 2.

1.000
0.992
0.993
0.994
0.996
0.998
0.993
0.994
0.993
0.996
0.999
0.999
1.000

1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000
1.000

1.000
0.975
0.978
0.985
0.993
1.000
0.984
0.988
0.990
0.993
0.996
0.999
1.000

4. What if either of the two benchmarks is unreliable?

So far we have not challenged either of the two benchmarks based on the 1992 和
2004 census results that were used by NBS in the revision. Apparently, the revision
based on the deviation of the old trend from the new one is acceptable only if the
new trend is accurate. NBS’s revision improperly assumes that the estimates based
在 1992 tertiary census are accurate, which is contrary to what some NBS statis-
ticians believe. They feel that some services were not sufªciently covered in the 1992
census and afterward. One may argue that in an extreme case, if the degree of the
undercoverage or underreporting was more or less the same back in 1992, 有
no justiªable reason for revising the existing real growth rate estimates. 然而, 作为
many may argue, with continuous efforts made by NBS over the past decade to im-
prove its statistical work including statistical coverage, the undercoverage problem
might have been improved over time. 所以, the growth rate should be adjusted
downward rather than upward. 事实上, it is quite likely that there is a combined
effect of the two forces.

Our next question is whether the 2004 census results are reliable. There are a num-
ber of important problems observed in the annual estimates based on the NBS sta-
tistical reporting system that were not fully addressed in the conduction of the cen-
他们的, which further substantiates our skepticism. The ªrst problem is the serious
discrepancy between local and national accounts. China’s regional GDP estimates
have been persistently higher than national estimates, which are largely driven by
the political incentives of localities. As disclosed by Li Deshui, the former head of
NBS, 在 2005 China People’s Political Consultation Congress (CPPCC) in Beijing,

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

数字 5. How much has been arbitrarily adjusted as suggested by the standard
interpolation procedures?—“Gaps” compared with the NBS nominal, 真实的, and implicit
price estimates

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

来源: Derived based on data from Tables 3, 5, 和 6.

if regional estimates were used instead, China’s GDP in 2004 将会 2,658 十亿
more and its growth rate would be 3.9 percentage points higher than the NBS
ªgures estimated based on the information through the state statistical reporting
系统 (reported by a Chinese newspaper Southern City Herald, 8 行进 2005).
讽刺地, 这 2,658 billion yuan of likely data inºation is coincidently close to the
2,300 billion of underestimation discovered by the 2004 人口普查.

Li’s point was made in March 2005 when NBS started processing the 2004 人口普查
data including crosschecking. 然而, NBS has not provided any information on
how serious the local over-reporting problem is as discovered in this census. 作为
serious discrepancy between regional and national GDP accounts has been hanging
there for over a decade, it is reasonable to assume that NBS would have taken this
census as a good opportunity to investigate the likely causes of the discrepancy.
尽管如此, the revised estimates by NBS appear to conªrm that the only possible
problem of China’s GDP estimation has been the underestimation of service output.

此外, it is difªcult to accept that other sectors are problem free. The state sta-
tistical reporting system has been widely criticized for misreporting for manufactur-
ing industries, and one of the main purposes of the national economic census in

Asian Economic Papers

The Chinese GDP Growth Rate Puzzle

2004 was to compensate for the deªciencies of that system. According to Xu (2002),
NBS has had to adjust rural industrial output downward since the 1990s after they
discovered in the 1995 industrial census that 40 percent of the rural industrial out-
put was overstated. 实际上, some serious data fabrication cases were found even
during the 2004 人口普查,7 suggesting that over-reporting of manufacturing output at
or below township level could be a big problem. One would reasonably question
why the census eventually did not discover any signiªcant output overestimation in
manufacturing industries.

In an earlier study, 吴 (2002), based on major industrial products with the 1987
weights derived from China’s 1987 Input-Output Table, constructed a real output in-
dex for Chinese industry, which could bypass the “comparable price” problem (IE。,
the segmented weights problem discussed in Maddison (1998)). Wu’s estimates
were signiªcantly different from the ofªcial estimates. The latter might have over-
stated the industrial growth for the period 1978–97 by 3.3 百分点. 数字 6
reports a preliminary update of these earlier estimates for the period 1978–2002.
Over this period our estimates show that China’s industrial growth was 16.2 百分
per annum compared with the ofªcial ªgure of 20.4 percent per annum.

此外, our estimates also suggest negative industrial growth in 1996 ((西德:4)3.2 每-
分, at the time when the authorities claimed a “soft landing”) 和 1998 ((西德:4)7.1 每-
分), which is in line with overall macroeconomic performance in China and the sit-
uation in the world economy, especially in the Asian ªnancial crisis. 数字 6 还
demonstrates that the ofªcial series of annual growth since the early 1990s is
signiªcantly less volatile than our series, suggesting that the actual volatility in in-
dustrial production might have been smoothed out to show more stable growth
(because of “successful” macroeconomic management).

5. Concluding remarks

The census is unquestionably a serious effort by the Chinese statistical authorities’
to improve the quality of China’s national accounts. Our investigation in this paper

7 例如, one of the serious data fabrication cases disclosed is that the authorities of

Maiwang Town, Hubei Province, assigned village ofªcials with “income quotas” which were
supposed to be ªlled into the census questionnaires. 因此, 80 million yuan from busi-
ness income by private ªrms was blown up to 1,009 百万, 或者 12.6 times the actual value
that was discovered by a NBS task force after an anonymous informant reported the infor-
mation to the census authorities (Southern Weekly, 2 六月 2005). One may reasonably wonder
if this case is only the tip of iceberg.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

数字 6. Ofªcial estimates may have exaggerated China’s industrial growth and smoothed
out industrial volatility (1978–2002)

来源: Author’s preliminary estimates that update his results published in Review of Income and Wealth, 48:2 (2002). Refer to that

paper for the methodology used in the estimation.

shows that the legacies of the traditional system and political constraints might have
limited its effect. We hope that these ªndings will help initiate a constructive dialog
among researchers to contribute to making national accounting in China method-
ologically sounder.

We close by touching on the always tempting question of “How fast has the Chinese
economy grown?” (to quote Angus Maddison’s (2006) recent update of his estimate
for China’s GDP growth). Maddison (2006) combines his work on agriculture and
services with our work on industrial output, including the above-cited preliminary
update to our earlier estimates of China’s industrial growth. By converting the nom-
inal values into the 1990 international Geary-Khamis dollars, Maddison shows that
China grew at 7.9 percent per annum in the 1990–2003 period compared with the
ofªcial rate of 9.9 percent per annum in the 1992–2004 period. He has not taken into
account NBS’s recent upward revision by 0.5 百分. If we can accept the NBS re-
vised estimates, Maddison’s estimate should be raised to 8.4 percent per annum,
which is still 1.5 percentage points below the ofªcial growth rate.

If the growth rate estimated by Maddison is closer to reality than the ofªcial esti-
伙伴, then the existing growth accounting results on the productivity of the Chi-
nese economy based on the ofªcial growth rate estimates are seriously ºawed.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

参考

Blades, Derek. 2004. Interpolation Methods. Paper presented at the NBS/ADB Workshop on
Improving Service Statistics, 17–18 November, Shanghai.

Editorial Committee of Almanac of China’s Economy (ECACE) (1996, 1998, 2001, 2003, 2004).
Almanac of China’s Economy Volume 1996: 16–25; 体积 1998: 23–29; 体积 2001: 3, 10–11;
体积 2003: 27–28; 体积 2004: 3–5. 北京: Almanac of China’s Economy Press.

Keidel, 阿尔伯特. 1992. How Badly Do China’s National Accounts Underestimate China’s GNP?
Rock Creek Research Inc. Paper No. E-8042.

Keidel, 阿尔伯特. 2001. China’s GDP Expenditure Accounts. China Economic Review 12 (4):355–67.

克莱因, Laurence R., 和S. Ozmucur. 2002. The Estimation of China’s Economic Growth Rate.
Journal of Economic and Social Measurement 28 (4):187–202.

刘, Yingqiu. 1999. Macroeconomic Policy Choice at the Semi-high Growth Stage. Working
纸. 北京: Institute of National Land and Regional Development.

Maddison, Angus. 1998. Chinese Economic Performance in the Long Run. 巴黎: OECD Develop-
ment Centre.

Maddison, Angus. 2006. Do Ofªcial Statistics Exaggerate China’s GDP Growth? A Reply to
Carsten Holz. Review of Income and Wealth 52 (1):121–26.

National Bureau of Statistics (NBS). 2005. China Statistical Yearbook 2005. 北京: China Statisti-
cal Press.

———. 2006. Announcement of the Revision of the Historical Statistics of China’s Gross Domestic
产品. 北京: National Bureau of Statistics.

Rawski, Thomas G. 2001. What is Happening to China’s GDP Statistics? China Economic Review
12 (4):347–54.

Ren, Ruoen. 1997. China’s Economic Performance in An International Perspective. 巴黎: 经合组织
Development Centre.

———. 2002. How Inºated Is China’s GDP Statistics?—Comment on Methodological Issues of
Two Studies on China’s GDP Statistics. Economics Quarterly 2 (1):37–52.

Shiau, 艾伦. 2005. Has the Chinese Government Overestimated the Economic Growth? 在
China’s Economic Growth: Studies and Debates, edited by Yue Ximing, Zhang Shuguang, 和
Xianehun Xu, PP. 297–355. 北京: CITIC Publication.

Woo, Wing Thye. 1998. Chinese Economic Growth: Sources and Prospects. In The Chinese Econ-
奥米, edited by Michel Fouquin and Francoise Lemoine, PP. 17–47. 伦敦: Economica.

Woo, Wing Thye. 2006. The Structural Nature of Internal and External Imbalances in China.
Journal of Chinese Economic and Business Studies 4 (1):1–20.

吴, Harry X. 2000. China’s GDP Level and Growth Performance: Alternative Estimates and
the Implications. Review of Income and Wealth 46 (4):475–99.

———. 2002. How Fast Has Chinese Industry Grown? Measuring the Real Output of Chinese
Industry, 1949–97. Review of Income and Wealth 48 (2):179–204.

Asian Economic Papers

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

The Chinese GDP Growth Rate Puzzle

徐, Xianchun. 2002. Study on Some Problems in Estimating China’s Gross Domestic Product.
Review of Income and Wealth 48 (2):205–15.

Ximing, Yue and Shuguang, 张. 2002. Problems in Measuring China’s Service Value
Added. Journal of Economic Research [Jing Ji Yan Jiu]. 12:51–59.

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
A
s
e
p
A
r
t
我
C
e
–
p
d

我

F
/

6
1
1
1
6
8
2
3
0
9
A
s
e
p
2
0
0
7
6
1
1
p
d

乙
y
G
你
e
s
t

哦
n
0
8
S
e
p
e
米
乙
e
r
2
0
2
3

Asian Economic Papers
下载pdf