STRUCTURAL CHANGE AND AGGREGATE EMPLOYMENT FLUCTUATIONS IN CHINA
Published date | 01 February 2021 |
Author | Xiaodong Zhu,Wen Yao |
DOI | http://doi.org/10.1111/iere.12486 |
Date | 01 February 2021 |
INTERNATIONALECONOMIC REVIEW
Vol. 62, No. 1, February 2021 DOI: 10.1111/iere.12486
STRUCTURAL CHANGE AND AGGREGATE EMPLOYMENT
FLUCTUATIONS IN CHINA∗
By Wen Yao and Xiaodong Zhu1
School of Economics and Management, Tsinghua University, China; Department of
Economics, University of Toronto, Canada; PBC School of Finance, Tsinghua University,
China
In developed countries, aggregate employment is strongly procyclical and almost as volatile as output. In
China, the correlation of aggregate employment and output is close to zero, and the volatility of aggregate em-
ployment is very low. We argue that the key to understanding aggregate employment fluctuations in China is
labor reallocation between the agricultural and nonagricultural sectors, and that the income effect plays an im-
portant role in determining the labor reallocation dynamics in both the long run and short run.
1. introduction
One salient feature of business cycles in developed countries is that aggregate employment
has a strong positive correlation with aggregate output (i.e., is procyclical) and is almost as
volatile as output. However, this is not the case in China, where the correlation of the cycli-
cal components of aggregate employment and output is close to zero, and the volatility of ag-
gregate employment is also very low. These puzzling facts about aggregate employment fluc-
tuations in China are present even after we correct for well-known measurement problems in
the official employment and GDP series, and they are robust to different detrending meth-
ods. In this article, we argue that the key to understanding aggregate employment fluctuations
in China is labor reallocation between the agricultural and nonagricultural sectors, and that
the income effect (i.e., the decline in the relative demand for agricultural goods with house-
hold income) plays an important role in the reallocation. Our argument is motivated by the
following three sets of empirical facts.
First, at the sectoral level, the cyclical properties of employment in China are similar to
those of developed countries. For both China and OECD countries, the volatility of sectoral
employment relative to the volatility of sectoral GDP is high and employment is strongly pro-
cyclical in the nonagricultural sector. In the agricultural sector, the relative volatility of em-
ployment is actually higher in China than in OECD countries, and employment is acyclical in
all the countries.
∗Manuscript received March 2020; revised July 2020.
1We would like to thank David Lagakos, Diego Restuccia, Tao Zha, and participants at the 2016 PBC-SAIF Con-
ference on Monetary Policy, 2016 Midwest Macro Meetings, 2017 Growth and Institution Program Meeting at Ts-
inghua University, 2017 and 2018 China Meeting of the Econometric Society, 2018 Annual Meeting of the Society
of Economic Dynamics, 2018 Workshop on Structural Transformation and Macroeconomic Dynamics at University
of Cagliari, 2019 Bank of Canada-Tsinghua PBCSF-University of Toronto Conference on the Chinese Economy, and
seminars at various universities and Federal Reserve Banks for valuable comments. We also thank the editor Dirk
Krueger and two anonymous referees for helpful comments and guidance. Wen Yao acknowledges financial support
from the National Natural Science Foundation of China (Grant No. 71603144). Please address correspondence to:
Wen Yao, Department of Economics, School of Economics and Management, Tsinghua University, Beijing, 100084,
China. E-mail: yaow@sem.tsinghua.edu.cn.
65
© (2020) by the Economics Department of the University of Pennsylvania and the Osaka University Institute of So-
cial and Economic Research Association
66 yao a nd zh u
Second, we show that disparities in aggregate moments between China and the developed
countries are explained by the nations being in different stages of structural change. Using a
panel data of 40 countries from the Groningen Growth and Development Center (GGDC),
we find that the comovement of aggregate employment and output at any point in time is
negatively related to the agricultural employment share at each point in time. Da-Rocha and
Restuccia (2006) document that average agricultural employment share has a negative effect
on the correlation between aggregate employment and output. However, we find that, even
after controlling for the average share, the agricultural employment share at each point in
time still matters. This dynamic effect of economic structure on aggregate employment fluctu-
ations is particularly relevant for China, where agricultural employment share declined from
71% in 1978 to 27% in 2017. Therefore, any theory for explaining aggregate employment
fluctuations in China should be able to match the secular trend of labor reallocation out of
agriculture.
Third, and most importantly, we find that almost all countries in the GGDC data set have a
ratio of agricultural employment to nonagricultural employment that is negatively correlated
with per capita GDP over the business cycles. Boppart (2014) and Comin et al. (2020) empha-
size that the income effect is important for understanding the secular trend of labor realloca-
tion from agriculture to manufacturing and services. Our new fact suggests that the income
effect is also important for determining labor reallocation between sectors at the business cy-
cle frequency.
Given these facts, we construct a two-sector growth model with nonhomothetic Constant
Elasticity of Substitution (CES) preferences recently used by Comin et al. (2020). In this
model, the income effect plays an important role in labor allocation both in the long run
and at the business cycle frequency. Using expenditure and price data of 40 countries and
a panel regression that is derived from our model, we first show empirically the presence
of a strong income effect. We then calibrate the parameters of our model so that it can ac-
count for China’s secular trend in labor reallocation from agriculture to nonagriculture. The
calibration reveals that the income effect is important in accounting for long-run structural
change in China. Without the income effect, the model would not match the structural change
in China in the long run. Finally, we examine the calibrated model’s implications for la-
bor market dynamics at the business cycle frequency. Fluctuations in this model are driven
by productivity shocks in the two sectors. We find that our model can indeed account for
China’s employment fluctuations at the sector level and in aggregate. At the business cycle
frequency, the income effect is also important for the model to match China’s business cy-
cle moments as, without the income effect, the model could generate neither the low correla-
tion between aggregate employment and output nor the negative correlation between relative
employment in agriculture and aggregate income in China. Our model also does a good job
at matching the structural change and aggregate employment fluctuations in developed coun-
tries such as the United States. In particular, our model implies a low employment–output
correlation for China and, at the same time, a high employment–output correlation for the
United States.
Our article contributes to the literature by documenting the importance of the income
effect for understanding aggregate employment fluctuations and constructing a model incor-
porating the income effect that can account for both long-run structural change and short-run
employment fluctuations in China. As such, our article is related to two strands of literature.
First, it is related to the literature on structural change; see, for example, Caselli and Coleman
(2001), Kongsamut et al. (2001), Ngai and Pissarides (2007), Acemoglu and Guerrieri (2008),
and Herrendorf et al. (2013). Most of the studies in this literature aim to understand the
sources of structural change in the long run; our article builds on this literature and studies
business cycle implications of the income effect to show that it is important for understanding
aggregate employment fluctuations in the short run. Our article is also related to the literature
on business cycles in China. Brandt and Zhu (2000) are one of the first to study business
employment fluctuations in china 67
cycles in China during the reform period. Their focus, however, is on the relationship be-
tween GDP growth and inflation over the business cycles in the 1980s and early 1990s. More
recently, Chang et al. (2016) focus on understanding the weak correlation between invest-
ment and consumption in China since the late 1990s. Neither of these studies examines the
relationship between aggregate employment and output. He et al. (2009) carry out an exercise
on business cycle accounting for China in the spirit of Chari et al. (2007). They find that most
of the fluctuations in aggregate employment can be accounted for only by variations in an
unobserved labor wedge, highlighting the inability of a standard one-sector business cycle
model to account for China’s employment fluctuations. Our article shows that a standard two-
sector model with nonhomothetic CES preferences can account for aggregate employment
fluctuations in China without introducing a time-varying labor wedge.
Two studies are closely related to our article. Da-Rocha and Restuccia (2006) are the
first to document the low correlation between aggregate employment and output in coun-
tries with a large agricultural sector. They use a two-sector real business cycle model to
examine the role of labor reallocation in accounting for the cyclical behavior of aggregate
employment. In order to focus on cyclical fluctuations, they assume that each country is
fluctuating around a steady state with a constant agricultural employment share.1Since struc-
tural change (i.e., the secular decline in agricultural employment share) is a very prominent
phenomenon in China during our period of study, and since the correlation between aggre-
gate employment and output fluctuations is affected by the agricultural employment share
at each point of time, not just the average of the share over a period of time, it is impor-
tant to have a unified model that can account for both the secular trend of structural change
and employment fluctuations around the trend. We provide such a unified model in this
article.
Another closely related paper is that by Storesletten et al. (2019) (hereafter referred to
as SZZ), who also use a two-sector model to account for both structural change and ag-
gregate employment fluctuations in China. Our article has four strengths. First, we show
that the income effect is empirically important at the business cycle frequency for a large
panel of countries and quantitatively important for accounting for aggregate employment
fluctuations in China. In contrast, SZZ emphasize capital deepening within agriculture in-
stead of the income effect as the driving force for labor reallocation between the two sec-
tors. Note that, although SZZ also consider the income effect using a generalized Stone–
Geary utility function, the income effect implied by the Stone–Geary utility function is
very special in that it disappears in the long run. As shown by Comin et al. (2020), a
model with a more general form of the income effect, that is, preferences represented by
a nonhomothetic CES utility function, performs much better than the generalized Stone–
Geary utility function in accounting for the secular trend of structural change across coun-
tries. In this article, we use the more general nonhomothetic CES utility function and show
that it performs well in accounting for labor reallocation over the business cycles and ag-
gregate employment fluctuations in China. Second, all important endogenous variables in
our article, such as sectoral employment and output, have empirical counterparts that can
be directly measured from available data. SZZ, however, assume that there are two sub-
sectors within agriculture, that is, traditional agriculture and modern agriculture, that can-
not be directly observed or identified in the data. Third, although SZZ assume an elastic-
ity of substitution between agricultural and nonagricultural goods that is greater than one,
we find in our estimation that this elasticity is less than one, which is consistent with the
values used or estimated in the literature on structural change (e.g., Ngai and Pissarides,
2007; Acemoglu and Guerrieri, 2008; Herrendorf et al., 2013; Comin et al. 2020). An elas-
ticity that is less than one implies that an exogenous increase in agricultural productivity
1Moro (2012) uses a similar method to examine the impact of reallocation from manufacturing to services on the
GDP volatility in the United States.
To continue reading
Request your trial