Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Contents lists available at ScienceDirect

Energy Policy
journal homepage: www.elsevier.com/locate/enpol

International oil shocks and household consumption in China

Dayong Zhang a, David C. Broadstock a,b,n, Hong Cao c
a
Team for Integrated Energy and Environmental Research Studies (TIERS), Southwestern University of Finance and Economics, China
b
Surrey Energy Economics Centre, Department of Economics, University of Surrey, UK
c
Beijing Institute of Technology, School of Management and Economics, China

H I G H L I G H T S

We study the impact of oil price shocks on residential consumption in China.

The most immediate effect passes through expenditure on transportation.

Effects also appear for health, education and food and clothing expenditure.

Existing price regulation offers no great benefit.

We argue that a compelling case for removing current price regulation exists.

art ic l e i nf o a b s t r a c t

Article history: We investigate the impacts that oil price shocks have on residential consumption in China. While it is
Received 14 April 2014 well understood that oil prices affect consumption in a multitude of ways, the timing and directness of
Received in revised form these effects on specific consumption categories is not clear. We demonstrate that the most immediate
22 August 2014
and direct effect passes through transportation consumption, as might be expected. But we also show
Accepted 25 August 2014
that significant effects pass through consumption in other sectors—including “food and clothes”,
“medical expenditure”, and other general “living expenditure”—with less immediacy. Given the results,
Keywords: particularly observed asymmetries with respect to rises and falls in international oil prices, we discuss
Oil shocks some implications for future adjustments to domestic price policies, in particular the case for removal of
Consumption
domestic price regulation.
Residential sector
& 2014 Elsevier Ltd. All rights reserved.
Oil price pass-through

1. Introduction forth policy objectives designed to continue recent economic

trends. 2 There is a general consensus that to complete this process
“China is now at such a crucial stage that without structural and further stimulate domestic demand, and by implication the
transformation and upgrading, we will not be able to achieve a level of household consumption, China will undergo significant
sustained economic growth. In readjusting the structure, the most structural transformation. On the surface, this appears to be a
important aspect is to expand domestic demand ….” natural progression for the Chinese economy. Growth in the
demand for all goods and services, however, will necessarily
(Li Keqiang, Summer Davos opening ceremony, September increase the level of energy consumed in the economy. This
11th, 2013)1 follows immediately from two facts: (a) energy is a critical factor
China has experienced more than 20 years of persistent high- of production and (b) transportation is required, to a greater or
paced economic growth, driven among other things by continued lesser degree, for the consumption of all goods and services, and
corporate and government investment, high levels of exports, and transportation is an energy-intensive (and emissions-intensive)
historically cheap labor. The Chinese government has openly set activity. In this regard, the stated objective of growing domestic
demand is somewhat at odds with targets on the reduction of
n
Correspondence to: Research Institute of Economics and Management, emissions and energy consumption that were committed to in the
Southwestern University of Finance and Economics, 55 Guanghuacun Street, Chinese government's 12th five-year plan.
Chengdu 610074, China. Tel.: þ 86 152 0834 0910.
E-mail address: davidbroadstock@swufe.edu.cn (D.C. Broadstock).
1 2
The full speech is available from the Xinhua News Agency at: 〈http://news. This is, for example, a stated objective in the recent 12th five-year plan of the
xinhuanet.com/english/china/2013-09/12/c_125371685.htm〉. Chinese government.

http://dx.doi.org/10.1016/j.enpol.2014.08.034
0301-4215/& 2014 Elsevier Ltd. All rights reserved.

Please cite this article as: Zhang, D., et al., International oil shocks and household consumption in China. Energy Policy (2014), http://dx.
doi.org/10.1016/j.enpol.2014.08.034i
2 D. Zhang et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Growth of the Chinese economy during recent years has well as consumption within specific expenditure categories,
coincided with huge increases in the consumption of oil, under- including “transportation and communication,” “food and cloth-
pinned by a surge in the rates of private car ownership, which ing,” “medical expenditure”, “education and entertainment,” and
have increased more than 30-fold, from 0.6 cars per 100 urban general “living expenditure.” In this regard, we follow a series of
households in 2000 to 18.28 in 2011. The consumption of oil used studies by Mehra and Peterson (2005), Edelstein and Kilian (2009),
for transportation in China doubled in the decade between 2000 Odusami (2010), and Wang (2013), which define the linkages
and 2010.3 The growth in car ownership increases demand for and between international oil shocks and household consumption
consumption of oil, posing genuine policy concerns since the under a permanent income hypothesis (PIH).
domestic consumption of oil far exceeds domestic production. Considering the results carefully, they indicate that household
Although China does produce oil, for many years now the consumption expenditure is not adversely affected by rising oil
economy has been a net importer and it is already among the prices; while falling oil prices seem to stimulate overall consump-
top three global importers of oil, with 68% of total oil consumed in tion to increase. The nuances of these effects, including timing,
2010 being from imported sources. With an oil supply gap that asymmetry and transmission routes (through alternative con-
continues to grow, China is unavoidably affected by international sumption categories) are all interesting, and each discussed in
energy markets, meaning that international oil price shocks can the main results. But more interesting is that, taken together, our
pass through to domestic activity. results point towards the conclusion that lifting the domestic oil
That rising oil prices can impact transportation costs is quite price policy is a very serious option that domestic policy makers
straight-forward, however there are further mechanisms by which should entertain.
changing international oil prices can impact upon the wider range The paper is organized as follows. Section 2 presents a general
of prices that consumers see. This is discussed in some detail for background discussion, outlining the nature of the oil pricing
the Chinese context by Tang et al. (2010) who, following Brown system in China and offering discussion of the literature on oil
and Yucel (2002), attribute the transmission mechanisms in to six shocks and the economy. Section 3 describes the methodology,
general channels of effect: supply side effects; wealth transfers; establishing the context of the planned consumption model under
general price inflation; real balance effects (as a result of changes a PIH along with the econometric formulation. Data are presented
in the demand for money); sector adjustments/re-structuring; and in Section 4, with the analysis results and discussion offered in
effects from increased uncertainty in the oil price. For example, a Section 5. The paper concludes in Section 6.
rise in the price of oil generates inflation and pushes up the
producer price index—the price faced by industrial consumers—
which is in general then passed on to the consumer via an increase 2. Background
in the consumer price index—the price that consumers pay. The
price changes will intuitively have some consequence upon the Despite an ongoing debate as to the specific role of oil in
consumption expenditures of households in the short-run. In shaping household consumption, the general mechanisms by
addition, the changing prices will further impact upon firm profits which it contributes to the economy are reasonably straightfor-
and investment, leading to lower levels of activity throughout the ward. Consider a rise in the price of oil. Scholars widely agree that
economy in the long-run. oil is needed to support most economic activities, underpinning
To shield domestic consumers against often volatile interna- the energy requirements of transportation needed to move both
tional oil prices, domestic retail oil prices in China are regulated. goods and the individuals who provide services. Bhattacharyya
The pricing system currently follows a form of floating-peg (2011), for example, provides a succinct review of these mechan-
regulation (see next section for further detail) against a bundle isms, highlighting that a rise in the price of oil effectively increases
of international market prices. The trend of regulation has been to the costs of all consumption, thereby reducing the quantity
increasingly normalize domestic prices against a bundle of inter- demanded by consumers for all goods and services. At the same
national prices, moving from a centrally controlled price in the time, an increase in energy prices makes energy a less desirable
1970s toward something today that quite closely reflects a market factor of production for firms, causing firms to substitute energy
mechanism. In light of the rapid growth of car ownership, it stands for other factor inputs, such as capital investment or more labor
to reason that international oil shocks could play an increasingly intensive production processes, see for example Broadstock et al.
important role in domestic household consumption decisions. (2007) for a global review of capital-energy substitution or Su
Investigating the nature, strength, and timing of these price et al. (2012) for a China specific example looking at capital, labor
pass-through effects therefore seems of interest and relevance to and energy, where further discussion on such types of substitution
policy debates, particularly regarding price regulation. effects is available.
A number of studies, both for China (e.g., Fan et al., 2007; Du To establish the importance of oil shocks on consumption in
et al., 2010) and elsewhere (e.g., Brown and Yucel, 2002; Hamilton the household sector of China, we first give some context to the
and Herrera, 2004; Kilian, 2008), have already established the Chinese oil pricing system. In so doing, we firmly establish the
predominantly negative influence of oil shocks to the macroecon- nature of price regulation and that international oil shocks do have
omy. However, largely due to data availability, the impact of oil a route through to domestic oil prices. We then provide a more
shocks to household consumption in China has remained an general discussion of the existing literature on how oil shocks pass
under-researched phenomenon. This is a significant omission throughout the economy.
since it is important to understand the household sector when
considering the welfare implications of exposure to international
oil shocks. In this paper, we therefore aim to assess whether and to 2.1. The oil pricing system in China: an overview
what extent oil shocks pass through to consumption by the
Chinese household sector, looking at aggregate consumption as China's oil pricing system has gone through three broad phases
since the People's Republic of China was established. The first
3
phase, which was in place primarily during the purely central
Similarly, ownership of air conditioning units has increased fourfold, from 30
units per 100 households in 2000 to 121 in 2011 (China statistical yearbook, 2011),
planned political system, ended in 1981. In this phase, the price of
reflecting the increased desire of Chinese households to complement their growing oil was set by the central government, with no scope for interna-
incomes with high energy–consuming luxury items. tional prices to pass through.

Please cite this article as: Zhang, D., et al., International oil shocks and household consumption in China. Energy Policy (2014), http://dx.
doi.org/10.1016/j.enpol.2014.08.034i
 D. Zhang et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 3

Phase two came into force at about the same time that China The evidence behind this and similar claims is grounded in a
initiated market reforms and introduced the famous “open-door” stream of influential studies that have followed from the early
policy in 1978. Since the reform, all parts of the economic system empirical contribution of Hamilton (1983). Over the decades since,
in China, including the domestic oil market, have continued to numerous studies—using different methods and concentrating on
shift toward a unique system mixing market-based and centrally different countries from across the world—all reinforce the under-
planned elements. From 1981 to 1998, a “dual-pricing” system was lying notion that oil shocks are important (see, for example, Mork,
used. In this mechanism, a base level of oil production was 1989; Mork et al., 1994; Lee et al., 1995; Bernanke et al., 1997; Lee and
required at a fixed price set by the central government; production Ni, 2002; Hamilton, 2003; Kilian, 2008; Kilian and Vigfusson, 2011).
beyond this base level enabled domestic oil producers to choose The point of departure among these studies comes by way of debate
between (a) selling any remaining oil on the international markets around the ways in which oil is important, particularly regarding the
(exporting) and accepting the international price or (b) accepting a role of asymmetry in shocks (e.g., positive and negative shocks).
regulated domestic price. This phase of the system appears to have Following for example Broadstock et al. (2014)—who study the
been aimed at providing protection for domestic oil producers to relationship between oil prices and financial market performance
help them establish and grow. in the Asia Pacific Region—we argue that the transmission
The final phase represented a shift away from a strong protec- mechanism by which international oil price shocks transmit to
tionist regime and toward a form of floating peg against a bundle domestic consumption expenditure can go through both a direct
of international prices. This phase has been in place since 1998 but channel and an indirect channel. Each of these effects intuitively
has seen some revisions to the definition of the international price revolves around the implied transportation cost resulting from an
peg, see for example Li and Ma (2011, in Chinese). From 1998 oil shock, but account for the nature of consumption across other
onward, price adjustments have been announced by the National product types also. The direct effect derives from the recognition
Development and Reform Commission (formerly known as the that people need to travel (and hence use petroleum, which is
National Planning Committee). From June 1998 until 2001, the made from oil) to travel and work and other places of recreation,
Singaporean market price was used as the benchmark peg for either in a car or for example as a passenger on a bus. An increase
Chinese domestic prices. During 2001, the peg remained attached, in the price of oil, and hence oil related products, will directly
in part, to the Singaporean price, but the basket of oil prices was increase the transportation cost and in turn alter the quantity
expanded to also include Rotterdam and Minas (Indonesian oil) demanded/consumption for goods that directly involve transport.
prices. Most recently, in 2008, the basket was again adjusted to That is to say, following a rise in the cost of the service, a fall in
track against European Brent, Dubai, and Minas oil prices. The consumption expenditure would derive directly from a reduction
domestic price in China therefore depends on the dynamics of in the quantity demanded for the service itself.
international crude oil, including taxes and other fees. The indirect channel, is however more complicated, and may
The rule by which price adjustments are made is as follows: when manifest in alternative ways. The first indirect effect may come
the 22-business-day moving (weighted) average of the basket of through inflationary concerns and/or general income effects.
international oil prices changes by more than 4%, domestic Chinese Bernanke et al. (1997) and Hamilton and Herrera (2004) for
prices are revised upward or downward accordingly. The nature of example study the connections between oil shocks and monetary
this mechanism means that shocks must, in general, be long-lived to policy. Their works have stimulated debate on how oil shocks pass
substantially shift the 22-day moving average; thus, a certain resi- through the real economy. The general idea is that rising oil prices
lience to oil shocks should exist. But if either (a) the shocks are lead to overall price inflation across all goods and services
sufficiently large or (b) multiple increases (or decreases) are sustained, (Bernanke et al., 1997), which can motivate the monetary authority
they can pass through to the economy within less than a month.4 to respond with contractionary measures and in turn cause further
Some have argued that exposure of China to shocks from depression in the economy e.g. a spiral effect. In such circum-
international oil markets would not necessarily be a bad thing, as stances consumption expenditure would be negatively affected
it may force the industry to be more responsible in cost manage- also. The second source of indirect effect will manifest as a
ment and create further benefits for industrial structural reform substitution effect resulting from a rise in the price of oil.
(Lin and Mou, 2008). However, one must also recognize that, since Increasing oil prices will cause transportation costs to raise as
1993, China has been a net oil importer, and the pricing policies mentioned earlier, accordingly the quantity demanded for goods/
discussed above will have played a role in ensuring stable economic services that have a high transport cost will decrease and in many
growth and protection for domestic Chinese oil producers. cases so will the level of expenditure on these items. For instance
consumers may allocate a higher share of expenditure towards
2.2. Oil shocks and the Chinese economy: existing evidence entertainment in the home (e.g. TV's or movies) as a result of
higher transportation costs.
Numerous studies characterize the general impact of oil shocks To date, only a handful of studies concentrate on the household
to the economy, with a growing number dedicated to under- sector of the economy and the role that oil shocks may have on it.
standing the specific Chinese context. The purpose here is not to Recent works include, for example, Mehra and Peterson (2005),
provide a complete review, but rather to recapitulate some of the Edelstein and Kilian (2009), Odusami (2010), and Wang (2013). As
key insights of the literature. one of the most important economies and largest importers of oil
The following quote from Hamilton (2009) neatly summarizes in the world, China inevitably attracts a large amount of research
what might be referred to as a stylized fact of oil and the economy: interest. However, due to limited availability of long-duration time
series data in China, empirical studies on the Chinese context have
“… a slowdown in overall consumer spending and a big drop in been relatively limited until recently.
consumer sentiment [is] again very much consistent with what Fan et al. (2007) provide one the earliest research studies to
was observed after earlier oil shocks.” concentrate on China. Using a computable general equilibrium
model of the economy, they show that international oil shocks
(Hamilton, 2009, pp. 251) have a generally negative impact on the Chinese economy,
including reductions in gross domestic product, investment, con-
4
In our empirical work, we use quarterly data; thus, shocks can reasonably be sumption, and exports. A number of studies have followed,
expected to occur with the same time period. offering generally supportive evidence. For example, Huang and

Please cite this article as: Zhang, D., et al., International oil shocks and household consumption in China. Energy Policy (2014), http://dx.
doi.org/10.1016/j.enpol.2014.08.034i
4 D. Zhang et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Guo (2007) find that oil shocks affect the long-term exchange rate by Palumbo et al., 2006) to describe consumption by the household
in China, while Faria et al. (2009) suggest that the export balance sector of the economy. Defining consumption, income, wealth, and
in China is positively linked with international oil prices. Studies the interest rate in real terms as C t , Y t ,W t , and r t , respectively, we
have also begun to highlight the connection to Chinese financial can write the household budget constraint as follows:
markets. Cong et al. (2008) and Broadstock et al. (2012), for W t þ 1 ¼ ð1 þ r t ÞðW t þ Y t  C t Þ; ð1Þ
example, highlight a dynamic relationship between international
oil shocks and stock market behavior in China. such that next-period wealth equals the discounted value of
Tang et al. (2010) offer a more focused discussion, aiming to current-period wealth plus earned income minus any con-
discern explicitly how oil shocks transmit throughout the Chinese sumption expenditure. Assuming a constant real interest rate
economy and to establish whether the six channels suggested by ðr t ¼ r t þ 1 ¼ r Þ and imposing the condition that lim W t þ i =
i-1
Brown and Yucel (2002)—the supply-side shock effect, wealth
ð1 þ rÞi Þ ¼ 0, then, by repeated substitution of the budget constraint,
transfer effect, inflation effect, real balance effect, sector adjust-
current-period wealth is obtained as follows:
ment effect, and unexpected effect—hold for China. Tang et al.
1 Ct þ i 1 Yt þi
(2010) establish a structural vector auto-regression model to show Wt ¼ ∑  ∑ : ð2Þ
i i
that oil shocks reduce output and investment but increase both i ¼ 0 ð1 þ rÞ i ¼ 0 ð1 þ rÞ
inflation and interest rates. They further claim that the effect of an Using the result from Hall (1978) that consumption follows a
oil shock on China's real economy persists for a long time due to martingale process gives EðC t þ 1 Þ ¼ C t ; then, taking the expecta-
the price control policies in place in China. tions of Eq. (2) results in the common form of the PIH:
Du et al. (2010) look more specifically at the impacts of oil

r 1 E Y r
shocks on economic growth. Their analysis demonstrates that t þi
Ct ¼ ∑ þ Wt: ð3Þ
shocks in international oil prices do affect economic growth as ð1 þ rÞ i ¼ 0 ð1 þ rÞi ð1 þ rÞ
well as price inflation in China. Moreover, they provide evidence Assuming a constant growth rate of real income, g, we have
that policy reforms in the price mechanism in China have had EðY t þ 1 Þ ¼ ð1 þ g ÞY t þ ηt þ 1 , where ηt þ 1 is a white noise process.
significant impacts on this relationship, causing structural instabil- Then we have
ity in their empirical model. Ou et al. (2012) offer one of the more
comprehensive empirical studies taking into consideration a total r r 1 ηt þ i
Ct ¼ Yt þ Wt þ ∑ : ð4Þ
of 71 macroeconomic indicators. Using a structural dynamic factor r g ð1 þ rÞ i ¼ 1 ð1 þ rÞ
i

model, they reveal the most probable transmission mechanisms of The derivation to this point establishes that a long-run relation-
oil shocks to the macroeconomy, suggesting that the inflation ship exists between consumption, income, and wealth. Mehra and
effect comes first, followed by supply-side effects and then real Peterson (2005) refer to this as the planned level of consumption,
balance effects. To the best of our knowledge, however, no studies C pt , expressing it in a simpler form by first taking expectations of
have explicitly targeted the effect of oil shocks upon consumption the error term and adding a constant term, leading to the
expenditure by the residential sector of the economy. estimable long-run relationship
Our research here is in effect a conceptual mix of the works of
Mehra and Peterson (2005) and Edelstein and Kilian (2009). We C pt ¼ a0 þ a1 Y t þ a2 W t ; ð5Þ
ultimately prefer the assumptions/modeling approach of Mehra where a1 ¼ ðr=r gÞ and a2 ¼ ðr=ð1 þ rÞÞ. Actual consumption, how-
and Peterson (2005) which are consistent with the idea that ever, differs from planned consumption (Campbell and Mankiw,
households, when faced by a rise in oil prices, may reallocate 1989). The short-run dynamics of consumption can therefore be
their consumption patterns within the context of a planned written in the form of an error correction model:
consumption framework, allowing for important dis-equilibrium

k
levels of expenditure to appear. One limitation of their study is the ΔC t ¼ b0 þ b1 C pt 1 C t  1 þ b2 ΔC pt 1 þ ∑ b3s ΔC t  s þμt : ð6Þ
failure to reflect the existence of the “reallocation” effect which s¼1

may result from a sudden price change. This was one of the major Substituting Eq. (5) into (6), we have
empirical contributions of Edelstein and Kilian (2009), and the
ΔC t ¼ b0 þ b1 ða0 þ a1 Y t þ a2 W t  C t  1 Þ þ b2 Δða0 þ a1 Y t  1
core motivation for wanting to look across the consumption k
categories. The contributions in our study are therefore two-fold: þ a2 W t  1 Þ þ ∑ b3s ΔC t  s þ μt : ð7Þ
first is the conceptual union of the two analytical approaches s¼1

discussed above; second is the application to the Chinese context, Assuming that future income grows constantly relative to the
which is desperately in need of empirical evidence. The latter current level, and that consumers have rational expectations, the
contribution is quite important, since China is markedly different expected value of accumulated and discounted future income
from US where: price regulations in China are much more severe streams is proportional to the current income. Therefore, the
and less transparent; and also because China is undergoing a model can be simplified to the following equation:
gradual process of market liberalization, but at a cautious pace.

k
ΔC t ¼ β0 þ β1 C pt 1  C t  1 þ β2 ΔY t  1 þ β 3 ΔW t  1 þ ∑ β4s ΔC t  s þ μt : ð8Þ
s¼1

3. Methodology and empirical framework Eq. (8) is the baseline model used in our analysis to capture the
dynamics of consumption changes. Following Mehra and Peterson
The model used here is based on the “planned consumption” (2005), oil prices are augmented into the short-run equation5:
framework used by Mehra (2001) and Mehra and Peterson (2005).

ΔC t ¼ β0 þ β1 C pt 1  C t  1 þ β2 ΔY t  1 þ β3 ΔW t  1
The empirical framework begins with a general/standard macro- k k
economic specification of (per capita) household consumption, þ ∑ β4s ΔC t  s þ ∑ β5s Δoilt  s þ μt : ð9Þ
where the level of consumption in an economy, C t , is affected by s¼1 s¼1

the existing level of wealth, W t , as well as current and discounted

expected future income, Y t and EðY t þ i Þ, respectively, where 5
Mehra and Peterson (2005) also include interest rates in their model. We also
i ¼ 1; :::; 1. In this regard, the approach therefore applies the considered interest rates but found that they were insignificant in all specifications,
commonly used PIH, which has been used recently (for example, so we do not discuss them further here.

Please cite this article as: Zhang, D., et al., International oil shocks and household consumption in China. Energy Policy (2014), http://dx.
doi.org/10.1016/j.enpol.2014.08.034i
 D. Zhang et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 5

Table 1 it is possible to construct a regulated domestic price series,

Variables used in the analysis. however in practice this is not feasible since the finer details on
the regulation schemes (e.g. the weightings attached to the basket
Variable Details
names members) and price revisions made by the NDRC are not perfectly
transparent. Accordingly to create any such series would involve
ΔC t Real consumption growth rate analyst judgment and make room for incidental errors that could
ΔY t  1 Lagged real disposable income growth rate impact on the statistical results, which is best avoided. Panel A of
ΔW t  1 Lagged real wealth growth rate
dit Changes in the short-term interest rate
Fig. 1 plots a selection of the oil prices that have constituted the
oilt Log real Brent price regulated basket over the years, including Minas, Cinta, Brent,
goilt Δoilt ; Growth rate of real oil price Dubai and for comparison the Chinese Daqing oil price. The
þ
goilt max{0, goilt }, Positive oil price growth correlations among these series are extremely high, each at 0.99,

goilt min{0, goilt }, Negative oil price growth with their movements being almost identical in all periods, with
ECTt The “error correction term”, that is produced as the residual from
the exception of some modest deviations in 2011–2012. Any
estimation of the long-run consumption equation (Eq. (5)).
weighted combination, which would require analyst assumptions,
would clearly be virtually identical to any of the individual series,
which require no strong analyst assumptions. Accordingly the use
Eqs. (5) and (9) establish the main equations for the empirical analysis. of Brent prices as an empirical measure of the oil prices relevant to
The variables used in the analysis are summarized in Table 1. China does not seem unreasonable.
Panel B of Fig. 1 shows the regulated gasoline price for China's
RON 95 gasoline7 sold in Beijing. Consistent data on regulated
gasoline prices do not cover a sufficiently long time span to include
4. Data
into our empirical framework, being available from only 2005
onwards as far as we could obtain the data. Nonetheless looking
The data on consumption expenditure ðC t Þ, income ðY t Þ, and
at these data alongside the oil price series it can be seen that oil and
wealth ðW t Þ for China are collected from the DATASTREAM
gasoline prices share some similar trends. Between 2005 and 2008
database,6 seasonally adjusted, and then deflated using consumer
both series are rising fairly steadily. The 2008 oil price collapse is
price index (base year: 2000). Wealth is defined as per-capita net
reflected in the gasoline prices by a fall also, though admittedly the
worth in constant prices, while labor income captures the current
fall in the gasoline price is much less dramatic than that of oil
period disposable income again in per-capita terms, these defini-
prices.8 From 2008 onwards gasoline price revisions are evidently
tions are the same as Mehra and Peterson (2005). For consumption
more frequent than before 2008, indirectly revealing the additional
expenditure, in addition to the per-capita total consumption (i.e.,
flexibility that resulted from the oil pricing scheme revisions that
across all goods and services combined), in urban households, we
took place in 2008 also. From 2011 to the end of the sample both oil
also have consumption expenditure for some specific consumption
and gasoline prices fluctuate mildly around a fairly constant value,
categories, including transportation and communication, food and
with no strong pattern of price growth or decline. Hence the
clothes, medical expenditure, education and entertainment, and
variation in gasoline prices are broadly reflected by variation in
other living expenditure. The data are collected quarterly, and the
international oil prices, as would be expected.
sample covers the first quarter of 2000 through the third quarter
Summary statistics for the untransformed variables are given in
of 2012.
Table 2, and each of the series is plotted in Fig. 2. Table 2 shows
Consistent with most existing studies, for oil prices, we use the
that income has roughly tripled over the 12-year period, and
European Brent crude spot price reported in the Energy Informa-
wealth is five times larger, on average. Total consumption has
tion Administration's online database. To ensure that the effects of
increased around two and one-half times, with the most signifi-
exchange rates are controlled for, the international oil price is
cant increase for consumption of transportation, which is almost
converted from US dollars into Chinese renminbi. After deflating,
five times greater at the end of the sample than at the beginning.
the series is then transformed into a growth rate by taking log
This rapid growth in transport expenditure implies some support
differences. Although the growth rate of oil prices is a commonly
to our earlier claims that oil shocks are becoming increasingly
used measure of oil shocks (following Hamilton 2003), to reflect
more relevant.
the potential nonlinearity (asymmetry), we also separate positive
oil price growth and negative oil price growth in our analysis. The
simple logic behind such decompositions is that consumers tend
to react quickly to price falls, being eager to boost their consump- 5. Empirical results
tion, however as prices rise consumers can be slower to reduce
their consumption, favoring instead to maintain their current level This section presents and discusses the empirical results,
of consumption (and the lifestyle attached to it) as long as possible proceeding first by confirming the nature of cointegration among
due to habit formation or lock-in effects, see for example Dargay the variables and thereby justifying the error correction model for
and Gately (1995). estimation. Following this, aggregate consumption (i.e., making no
Regarding the use of European Brent crude prices, from earlier distinction among the consumption categories) is modeled using
discussion we know that the actual regulated price of oil in China the short-run equation in (9), and the importance of international
includes Brent prices only since 2008, and that other international oil shocks is confirmed. The results are then extended to distin-
prices are also used in determining the domestic price. In principle guish among the specific consumption categories, offering a richer
image of the location and timing of oil shocks.
6
DATASTREAM is a commercial database that combines a large array of
7
national and international data collected from official data sources, such as national RON stands for research octane number.
8
statistical bureaus. It is one of the most comprehensive sources of economic and The reasons for this are not immediately clear, but it seems that the timing of
financial data compiled in a consistent and easy to use manner, which has made it a some limited gasoline price reductions coincide with the end of the oil price
widely used source of information in academic research. The DATASTREAM is collapse—it might be argued that the regulators exercised some discretion during
provided by Thomson-Reuters and further information can be found at https:// this period of oil price collapse, and waited to see how far prices would fall before
forms.thomsonreuters.com/datastream/. revising the domestic price.

Please cite this article as: Zhang, D., et al., International oil shocks and household consumption in China. Energy Policy (2014), http://dx.
doi.org/10.1016/j.enpol.2014.08.034i
6 D. Zhang et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

RON 93 Gasoline price and the Brent crude oil price.

160

Brent oil - Chinese RMB/barrel.

700

Gasoline - Chinese RMB/Liter.

Cinta 10
140
Daqing
Minas
120 Brent 9
600
Dubai
100
8

80 500
7
60
6 400
40

20 5
300

0
99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 2006 2008 2010 2012
Gasoline = Solid line; Oil = Dashed line (second y-axis)

Fig. 1. (A) International oil price comparisons and (B) Domestic gasoline prices in China. (sample periods differ due to data availability). Note: panel A shows the Cinta,
Daqing, Minas, Brent and Dubai international oil prices from 1999 to 2014. Panel B show the Chinese RON 95 gasoline price from 2005 to 2013 and for comparison includes
the (quarterly) Brent oil price series used for estimation.

Fig. 2. Main variables used in the analysis. Note: these plots are for de-trended data in levels and not the estimation-transformed variables.
 D. Zhang et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 7

Table 2
Descriptive statistics.

Variable Initial value Final value Avg. growth Mean Min. Max.

Income (Y) 1522.79 4809.01 2.26% 2897.54 1552.79 4809.01

Wealth (W) 9749.09 49,553.96 3.25% 25,615.22 9749.09 49,553.96

Consumption variables
Total 1229.07 3220.41 1.93% 2110.65 1229.07 3220.41
Transportation 97.78 484.78 3.20% 269.61 97.78 484.78
Food and clothing 616.86 1521.89 1.81% 996.69 616.40 1521.89
Medical expenditure 77.12 204.71 1.95% 144.89 77.12 207.33
Educ. and ent. 151.33 384.71 1.87% 273.41 151.33 399.84
Other living expenses 108.05 288.45 1.96% 211.54 101.01 292.65

Oil price variables

oilt 222.23 535.99 1.76% 389.56 161.42 716.71
goilt 8.82 1.44 1.90  72.85 28.45
þ 8.82 1.44 6.41 0.00 28.45
goilt

goilt 0 0  4.51  72.85 0.00

Notes: All variables are expressed in real monetary terms except for oil growth, measured in Chinese RMB. Statistics were calculated for data from the first quarter of 2000
through the third quarter of 2012.

Table 3 expressed in a standard linear regression form as follows:

Cointegration and break test results for variables used in the analysis.
yt ¼ μ1 þ μ1 DBt þ α1 xt þ α2 DBt xt þξt ; ð10Þ
Panel I. Cointegration test
Engle and Granger residual-based test where the date break indicator variable, DBt , is the unknown break
point to be estimated and takes the value 1 on and after the
Test statistics Cointegrating vector chosen break date and 0 for all other periods. Gregory and Hansen
nnn
(1996) then suggest a sequential search procedure to identify the
Tau statistic  5.2439 Income 0.9316
Z-statistic  38.0149nnn Wealth 0.0451 most likely breaking point using ADFn and/or Z n statistics, where
  
Johansen approach ADFn ¼ Inf ADFðτÞ ; τ A kmin ; kmax ;

No. of CI vectors Max-Eigen Trace Cointegrating Vector and

  
None 25.9031 nnn
34.4032 nn
Income 1.0622 nn
Z n ¼ Inf Z ðτÞ ; τ A kmin ; kmax ;
At most 1 8.4646 8.5001 Wealth 0.1322nn
At most 2 0.0355 0.0355 and where kmin and kmax are the trimming points from the start
Panel II. Gregory and Hansen break test and end of the full sample, reflecting the fact that a break cannot
happen in either the beginning or ending period of the sample.9
Statistics Breaking point Critical value Gregory and Hansen (1996), Table 1 provide critical values for both
nnn
test statistics. While the above test procedure is not dissimilar
Inf ADF  7.2553 2008Q2 5%  5.96
Inf Z  6.5993nnn 2007Q4 1%  6.45 from alternative procedures, such as Andrews–Ploberger types of
tests, the refinements by Gregory and Hansen (1996) ensure that
nnn
and nn denote significance at the 1% and 5% levels respectively. Note: the test retains power in the context of cointegrating variables/
CI ¼ cointegrating. relationships, making it preferable in our context.
The structural break test results are presented in panel II of
Table 3, and plotted in Fig. 3. According to the ADFn statistic, a
5.1. Testing for the PIH and structural break
structural shift occurred during the second quarter of 2008. This
break date roughly coincides with two events, the global financial
Taking consumption, wealth, and income as defined above,
crisis and the latest revision to the domestic pricing policy. In light
we begin by considering whether their long-run relationships
of the severity of the global financial crisis, we consider this to be
are stable over the estimation sample. To do so, we first
the dominant of the two forces.10
apply standard cointegration checks, including Johansen's
These results point toward a fundamental shift in consumers'
approach and the Engle and Granger residual-based test for
treatment of wealth before and after the structural break, with the
comparison. The results, reported in panel I of Table 3, indicate
sign of the coefficient switching. It appears that the role of the
one cointegrating vector, suggesting that a long-run relationship
financial crisis was to fundamentally alter the long-run expecta-
exists.
tions of consumers, but short-run adjustment behaviors remained
While the results suggest evidence of cointegration and the
unchanged. The long-run equation is therefore modified to reflect
potential to proceed to the estimation of the short-run Eq. (9), it is
this:
nonetheless important to recognize that, among other things,
our sample period incorporates the global financial crisis, which C pt ¼ α0 þ α1 Y t þα3 W t þ DBt ðα0b þα1b Y t þα3b W t Þ: ð11Þ
has been the source of fundamental shifts in economic activity
worldwide. Notwithstanding the existence of a cointegrating
relationship, a structural shift due to the global financial crisis is 9
More precisely, this means that enough observations must exist on both
a distinct possibility. To test the possibility of some significant subsamples, at the start and at the end, to allow the test procedure to produce
accurate results.
structural change in this long-run relationship, we adopt the 10
In the following discussion, we refer to this as a break due to the crisis, but
structural break test proposed by Gregory and Hansen (1996). we concede that it may, in part, also reflect structural change due to policy
They propose a simple model with regime shift that can be adjustment.

Please cite this article as: Zhang, D., et al., International oil shocks and household consumption in China. Energy Policy (2014), http://dx.
doi.org/10.1016/j.enpol.2014.08.034i
8 D. Zhang et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

-4.5

-5.0

-5.5

-6.0

-6.5

-7.0

2002 2004 2006 2008 2010

( Solid line = ADF statistic; Dashed line = Z statistic; Dotted line = 1% critical value )

Fig. 3. Gregory and Hansen (1996) test statistics. Notes: solid line ¼ ADF statistic; dashed line ¼Z statistic; dotted line ¼ 1% critical value.

Table 4
Regression results for aggregate consumption (these tables report only the optimal model specifications obtained using Akaike Information Criterion, AIC).

Baseline M0 Augmented M1 Augmented M2 Augmented M3 Augmented M4

Constant 2.5555nnn 2.6221nnn 2.4385nnn 2.5884nnn 2.3196nnn

(0.6070) (0.6473) (0.5973) (0.6305) (0.5802 )
ECTt  1  1.3866nnn  1.3219nnn  1.3343nnn  1.3790nnn  1.3546nnn
(0.2074) (0.2077) (0.2012) (0.2170) (0.2054)
ΔY t  1  0.3033  0.3518  0.3483  0.3081  0.3364
(0.2313) (0.2525) (0.2404) (0.2369) (0.2341 )
ΔW t  1 0.0304 0.0542 0.0469 0.0336 0.0379
(0.0575) (0.0623) (0.0587) (0.0299) (0.0678 )
oilt  0.0216nn
(0.0104)
þ  0.0051 0.0168
goilt  1
(0.0299) (0.0306 )

goilt  1  0.0345nn  0.0375nn
(0.0136) (0.0156 )

Additional model information

R2 0.4439 0.4671 0.4794 0.4441 0.4818
AIC for general model N/A 3.9974 3.9744 4.0420 4.0852
AIC for optimal model 3.8713 3.8694 3.8461 3.9117 3.8824

Notes: nnn, nn and n denote significance at the 1%, 5% and 10% levels respectively. Heteroscedasticity and autocorrelation corrected standards errors are used since the sample
size is relatively small and heteroscedasticity needs to be controlled. Standard errors are reported in parentheses. ECTt  1 is the error correction term (the residual from the
first stage regression in Eq. (5)), lagged by one period. R2 is the R-squared measure of goodness of fit. A value close to zero shows a poor fit, while a value close to one shows a
strong fit e.g. that the model describes the data very well.

Thus, the structural break is incorporated into the long-run model, were also considered but were not significant in any lag orders so
and the error-correction term that is then passed through to the are not reported.
short-run equation incorporates the structural instability expli- The results in Table 4 are not without merit, though they do
citly. By itself, this introduces no new statistical problems because arguably lack sufficient clarity over different types of consumption
the procedure helps to ensure stationarity of the error correction to accurately depict the true role of oil prices. We therefore
term, as required. The remaining part of this section presents the proceed to the disaggregated consumption category results.
results from the estimated short-run equations.
5.3. Regression results in categories

5.2. Short-run relationships between consumption, income, and We have found that aggregate consumption in Chinese house-
wealth, and the role of international oil shocks holds is negatively affected by international oil shocks. One natural
extension would be to try to understand how the shocks are
We turn next to estimation of the short-run total consumption transmitted—in other words, whether some consumption cate-
equations. First, we estimate the baseline model shown in Eq. (8) gories are affected more than others. When households meet a
without oil shocks included and then with oil-shocks and other new budget constraint, such as would happen with a change in the
factors i.e. Eq. (9). The residuals from these models, as per the cost of energy, they may have to rebalance the bundle of goods
long-run equations above, are tested for structural stability, but no they consume. Clearly, rising oil prices may reduce household
additional breaks are found in these short-run equations. consumption in energy-related categories. For example, one would
From the estimation results in Table 4, we can clearly observe a drive less when the price of fuel oil is higher and may therefore cut
role for oil shocks in shaping household consumption in China. In transportation and communication spending. However, it might
contrast to popular opinion, increasing oil prices do not have a not be plausible to reduce consumption of a necessity like food. It
significant impact, whereas falling prices increase household is therefore interesting to investigate the impact of international
consumption. We also find a clear tendency for error correction, oil price shocks on consumptions of different categories.
though after the correction to the long-run equilibrium, short-run Here we investigate the impact of oil shocks on the following
dynamics in income and wealth have no significant impact on categories of consumption expenditure: transportation and com-
changes in consumption decisions. Changes in short-term interest munication, food and clothing, medical expenditure, education

Please cite this article as: Zhang, D., et al., International oil shocks and household consumption in China. Energy Policy (2014), http://dx.
doi.org/10.1016/j.enpol.2014.08.034i
 D. Zhang et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 9

Table 5
Regression results for consumption categories (these tables report only the optimal model specifications obtained using Akaike Information Criterion, AIC).

Food and clothing Medical expenditure Educ. and ent. Transportation Other living expenses

M1 M2 M1 M1 M1 M2 M1 M2
Constant 2.1964nnn 3.2879nnn 3.1717nnn 1.8782 4.7546n 5.9188nn 1.7656 -1.4865
(0.4547) (0.7328) (1.0385) (1.3803) (2.6069) (2.4509) (1.4798) (2.8461)
ECTt  1  0.6189nnn  0.5112nn  0.3150  0.0415  0.7578  0.6220  2.8868n  2.8754n
(0.1975) (0.2354) (0.7721) (1.1624) (1.1600) (1.1803) (1.6464) (1.5963)
ΔY t  1  0.0717  0.0669  0.1540 0.3289 0.0378  0.0431 0.5450 0.7012
(0.1578) (0.1375) (0.2608) (0.4517) (1.0941) (1.1163) (0.6019) (0.5252)
ΔW t  1  0.0017  0.0101  0.2158  0.0178 0.3846 0.4529  0.0743  0.0477
(0.0857) (0.0752) (0.2201) (0.2226) (0.2727) (0.3039) (0.2427) (0.3100)
ΔC t  1  0.3642nn  0.6134nnn  0.6082nnn  0.4260nn  0.4780nnn
(0.1523) (0.1058) (0.0992) (0.1582) (0.1602)
ΔC t  2  0.1834n  0.1834n  0.2311nn  0.1695n
(0.0952) (0.0839) (0.0922) (0.0986)
d
goilt  1  0.0087  0.0346  0.1425nnn
(0.0164) (0.0443) (0.0410)
d
goilt  2  0.0175  0.0429n
(0.0171) (0.0240)
d
goilt  3 0.0082  0.0527n
(0.0160) (0.0267)
d
goilt  4  0.0361nn
(0.0172)
þ
goilt  1  0.0370  0.2854nnn 0.2198
(0.0341) (0.1001) (0.1474)
þ  0.0182  0.1326
goilt  2
(0.0280) (0.2164)
þ  0.0272 0.0926
goilt  3
(0.0434) (0.0916)
þ
goilt  4  0.0955nnn 0.2261n
(0.0306) (0.1315)

goilt  1  0.0834n
(0.0422)

Additional model information

R2 0.2094 0.2428 0.1058 0.1181 0.3614 0.3794 0.3270 0.4054
AIC for general model 4.0331 4.0331 5.5786 6.8669 6.6230 6.6036 6.9137 6.9768
AIC for optimal model 3.9625 3.9193 5.5645 6.7091 6.5385 6.5516 6.7762 6.8487

Notes: nnn, nn and n denote significance at the 1%, 5% and 10% levels respectively. Heteroscedasticity and autocorrelation corrected standard errors are in parentheses. ECTt  1
is the error correction term (the residual from the first stage regression in Eq. (5)), lagged by one period.

Table 6
The reactions to oil shocks across consumption categories.

Periods after shock Price rise Price fall

þ1 Quarter Transportation expenditure ↓ Transportation expenditure ↑

þ2 Quarter Medical expenditure ↓ Medical expenditure ↑
þ3 Quarter Medical expenditure ↓ Medical expenditure ↑
þ4 Quarter Living expenditure ↑
Food and clothing expenditure ↓

and entertainment, and other living expenditure. We use the same parsimony). A smaller model that offers the same amount of
general short-run equation in (9), applied to each of the consump- information is favored, and reflected by smaller values of AIC,
tion categories. Table 5 reports the regression results for each of hence the optimal model is that with the smaller AIC.
the categories, using M1 to denote model specifications in which The transportation and communication category picks up the
oil shocks are assumed to be symmetric and M2 for specifications highest and most rapid impact from oil shocks as it is the only
in which asymmetry is allowed.11 To control for possible auto- category to feel shocks after just one lag. M1 uses total oil returns
correlation, we include four lags on both the dependent variable only, whereas M2 decomposes oil price rises and oil price falls.
and oil shocks. From the most general specification, an optimal Taken together, the results from these two specifications provide
model specification is obtained using Akaike information criteria reasonably clear evidence of asymmetry insofar as increasing oil
(AIC). AIC works by selecting the model specification that best prices change consumption by a greater amount than do falling
balances the amount of information captured by a model, i.e. its prices. The impacts of oil shocks on the other consumption
ability to explain the data that it is intended to explain, against the categories, such as education and entertainment, is generally
size of the model used to generate that information (model much smaller, and their influence, when significant, takes at least
two quarters to manifest.
The consequences of an oil shock to the key consumption
11
In these specifications, the error correction term is taken from the total categories are summarized in Table 6. Column one defines the
consumption model, reflecting an overall disequilibrium adjustment effect. time period, running from the first period after the shock occurs

Please cite this article as: Zhang, D., et al., International oil shocks and household consumption in China. Energy Policy (2014), http://dx.
doi.org/10.1016/j.enpol.2014.08.034i
10 D. Zhang et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎

through the fourth subsequent period (þfour quarters) when all oil shocks to pass through to domestic activity more quickly.
effects are realized. The second and third columns, respectively, Arguably, the changes in domestic regulation have been cautiously
show which consumption items are affected and in which time applied, taking steps toward oil price liberalization very carefully.
periods given (a) a price rise or (b) a price fall. In the period Ultimately, we consider that the path of price regulation has
immediately following a shock, the only consumption category been in the right direction, and that it should be maintained.
affected is transportation, irrespective of whether the shock is Taking a slightly more general perspective on the economy, Lin
positive of negative. In the second quarter, medical expenditure and Mou (2008) argue that the industrial sector of the economy in
decreases.12 In the third period, additional reductions in medical particular may benefit from oil shocks; here, we additionally
expenditure imply there may have been some further health argue that household consumption may also benefit, or at least
benefits, greater even than in the previous period. It is only in not suffer. Since these effects imply reduced oil consumption,
the fourth period that other consumption transfer patterns are transportation-related emissions should also fall. We do, however,
revealed, with expenditures on food decreasing and expenditures acknowledge the full complexity of any economic system, and that
on other living expenses increasing at the same time. But these our analysis and discussion are ultimately based on a partial
final effects manifest only when prices rise, not when prices fall. assessment of oil shocks and consumption.
In sum, a positive oil shock reduces transportation expenditure,
reduces medical expenditure, and causes a change in the balance
of food expenditure and consumption related to other living
expenses. In the literature review section it was generally found 6. Conclusions
that oil shocks negatively affect economic behavior, but the results
here suggest a largely positive reaction to oil shocks. To reconcile In this paper, we have attempted to answer a simple question: Do
this it should be recalled that previous literature looks at international oil price shocks pass through to domestic consumption
economy-wide effects in most cases, the lens offered here is more by the household sector in China? In this regard we demonstrate not
clearly focused on the relatively under-studied aspect of consumer only that oil price shocks influence consumption, but that the nature
expenditure. Since it looks at a different part of the economy it of these impacts differ across various different types of consumption
should not be expected that the same conclusions are to appear. e.g. transport versus medical expenditure. This is in many ways
The results here do not so much offer a different conclusion for the intuitive, since different consumption categories have differing levels
same area, rather they offer a conclusion for a different area, and of connectivity to oil prices. Nonetheless empirical quantification of
moreover the results are both plausible and intuitive. such effects at an aggregate level has to date been somewhat limited
within the known literature.
5.4. Further discussion and policy implications Ultimately, our results point towards the conclusion that lifting
the domestic oil price policy that is currently in place in China, is a
We have, to this point, illustrated that oil shocks are relevant in very serious option that domestic policy makers should at the very
household consumption choices, but in a generally asymmetric least entertain. To be more specific, given our results there is a
manner. Here we consider further the implications of our results compelling argument (notwithstanding wider general equilibrium
for related policy—in particular, possible revisions to existing effects that we do not model) that price regulations should be
domestic price regulation. removed since the regulation appears to offer no great benefits,
Some have argued that exposure to shocks from international but does (i) create a costly regulation process, (ii) hinder progress
oil markets would not necessarily be a bad thing, as it may force towards integration with international oil markets and (iii) dimin-
industry to be more responsible in cost management and create ish the co-benefits of health and environmental protection that
further benefits for industrial structural reform (Lin and Mou, would result from a higher (and/or more flexible) price of oil.
2008). However, since 1993, China has been a net importer of oil, Our results have their boundaries. While on the one hand there is
and the pricing policies discussed above have played an important an argument for removing price regulation, on the other hand there
role in ensuring stable economic growth and protection for exists a counter argument which potentially over-emphasizes the
domestic Chinese oil producers. importance of the partial nature of our analysis, which concentrates
The results indicate that household consumption is not on a reasonably narrow area of the economy. While acknowledged,
adversely affected by rising oil prices; rather, our results (consis- this is not considered a limiting feature of the current work: though
tent with economic theory) imply an adjustment of household it must be conceded that other areas of the economy, for example
consumption patterns among alternative goods and services in the producers/firms nor the role of the government, are explicitly
light of the new household budget constraint. Conversely, falling characterized, the apparent gap in the extant literature motivates the
oil prices seem to stimulate overall consumption, which appears to need for any empirical insights on the residential sector, partial or
be a positive effect for the economy. This suggests that allowing otherwise. However, this line of objective criticism does give rise
international oil price shocks to pass through to households (e.g., to some directions for future study: who exactly are the players
through international price liberalization), is not necessarily some- impacted by oil shocks (consumers, firms, governments, oil refiners
thing to be afraid of. and producers etc.); what is the balance of benefits and costs that
Considering the history of price regulation discussed earlier in impact each of these players; from an economists perspective, what
the paper, it is clear that domestic prices have become increasingly are the welfare implications—who benefits more the consumer or
market oriented, particularly from 1998 onward. For example, the the producer, and is this as intended? Better understanding the
bundle of prices pegged against was revised in 2001 and again in welfare implications of oil shocks would seems like a particularly
2008 to reflect the true cost of importing from international oil worthy direction for future study. Doing this properly would, in our
markets. Moreover, the window size over which oil prices are opinion, benefit most from complementary research to this study,
pegged has been shortened over the years, allowing international that takes a more careful look into the micro-level foundations of the
topics discussed here. Explorations in this direction could usefully
12
include analysis from micro-level firm data as well as consumer data,
A speculative explanation for this could be that fewer cars are traveling on
the roads (the first-period effect), resulting in reduced pollutants and fewer
seeking among other things to add deeper clarity to the substitution
instances of respiratory disease and/or other ailments. However, our data provide mechanisms across alternative consumption categories in response
no way to confirm this. to a sudden energy price shift.

Please cite this article as: Zhang, D., et al., International oil shocks and household consumption in China. Energy Policy (2014), http://dx.
doi.org/10.1016/j.enpol.2014.08.034i
 D. Zhang et al. / Energy Policy ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 11

References Hamilton, J., 2003. What is an oil shock? J. Econom. 113 (2), 363–398.
Hamilton, J.D., 2009. Causes and consequences of the oil shock of 2007–08.
Brook. Pap. Econ. Act. 40 (1), 215–283.
Bernanke, B., Gertler, M., Watson, M., 1997. Systematic monetary policy and the Hamilton, J.D., Herrera, A.M., 2004. Oil shocks and aggregate macroeconomic
effects of oil shocks. Brook. Pap. Econ. Act. 28 (1), 91–157. behavior: the role of monetary policy. J. Money Credit Bank. 36 (2), 265–286.
Bhattacharyya, S.C., 2011. Energy Economics: Concepts, Issues, Markets and Huang, Y., Guo, F., 2007. The role of oil price shocks on China's real exchange rate.
Governance. Springer-Verlag, London, UK.
China Econ. Rev. 18 (4), 403–416.
Broadstock, D.C., Cao, H., Zhang, D.Y., 2012. Oil shocks and their impact on energy
Kilian, L., 2008. The economic effects of energy price shocks. J. Econ. Lit. 46 (4),
related stocks in China. Energy Econ. 34 (6), 1888–1895.
871–909.
Broadstock, D.C., Hunt, L.C., Sorrell, S.,October 2007. Elasticity of substitution
Kilian, L., Vigfusson, R.J., 2011. Are the responses of the US economy asymmetric in
studies. UK Energy Research Centre (UKERC) Review of Evidence for the
energy price increases and decreases? Quant. Econ. 2 (2), 419–453.
Rebound Effect. Technical Report 3. Ref: UKERC/WP/TPA/2007/011.
Lee, K., Ni, S., 2002. On the dynamic effects of oil shocks: a study using industry
Broadstock, D.C., Wang, R., Zhang, D., 2014. Direct and indirect oil shocks and their
level data. J. Monet. Econ. 49, 823–852.
impacts upon energy related stocks. Econ. Sys. 38, 451–467.
Lee, K., Ni, S., Ratti, R., 1995. Oil shocks and the macroeconomy: the role of price
Brown, S.P.A., Yucel, M.K., 2002. Energy prices and aggregate economic activity: an
variability. Energy J. 16, 39–56.
interpretative survey. Q. Rev. Econ. Finance 42, 193–208.
Li, X., Ma, N., 2011. Oil Prices. Peking University, China center for energy develop-
Campbell, J.Y., Mankiw, G., 1989. Consumption, income, and interest rates:
ment. CCED working paper series no. 20110303 (in Chinese).
reinterpreting the time series evidence, NBER Working Paper no. 2436.
Lin, B.Q., Mou, D.G., 2008. The impact of energy price increases on macro-economy:
National Bureau of Economic Research, Cambridge, MA.
China statistical yearbook, 2011. National Bureau of Statistics, China Statistics Press, an analysis based on CGE method. Econ. Res. J. 11, 88–101 (in Chinese).
Beijing, China. Mehra, Y.P., 2001. The wealth effect in empirical life-cycle aggregate consumption
Cong, R.G., Wei, Y.M., Jiao, J.L., Fan, Y., 2008. Relationships between oil price shocks equations. Fed. Reserve Bank Richmond Econ. Q. 87 (2), 45–68.
and stock market: an empirical analysis from China. Energy Policy 36 (9), Mehra, Y.P., Peterson, J.D., 2005. Oil prices and consumer spending. Fed. Reserve
3544–3553. Bank Richmond Econ. Q. 91 (3), 53–72.
Dargay, J.M., Gately, D., 1995. The imperfect price-reversibility of non-transport oil Mork, K.A., 1989. Oil and the macroeconomy when prices go up and down: an
demand in the OECD. Energy Econ. 17, 59–71. extension of Hamilton's results. J. Political Econ. 91, 740–744.
Du, L.M., He, Y.N., Wei, C., 2010. The relationship between oil price shocks and Mork, K.A., Olsen, O., Mysen, H.T., 1994. Macroeconomic responses to oil price
China's macro-economy: an empirical analysis. Energy Policy 38 (8), increases and decreases in seven OECD countries. Energy J. 15, 19–35.
4142–4151. Odusami, B.O., 2010. To consume or not: how oil prices affect the comovement of
Edelstein, P., Kilian, L., 2009. How sensitive are consumer expenditures to retail consumption and aggregate wealth. Energy Econ. 32, 857–867.
energy prices? J. Monet. Econ. 56 (6), 766–779. Ou, B.L., Zhang, X., Wang, S.Y., 2012. How does China's macro-economy response to
Fan, Y., Jiao, J.L., Liang, Q.M., Han, Z.Y., Wei, Y.M., 2007. The impact of rising the world crude oil price shock: a structural dynamic factor model approach.
international crude oil price on China's economy: an empirical analysis with Comput. Ind. Eng. 63 (3), 634–640.
CGE model. Int. J. Glob. Energy Issues 27 (4), 404–424. Palumbo, M., Rudd, J., Whelan, K., 2006. On the relationships between real
Faria, J.R., Mollick, V.A., Albuquerque, P.H., Len-Ledesma, M.A., 2009. The effect of oil consumption, income, and wealth. J. Bus. Econ. Stat. 24 (1), 1–11.
price on China's exports. China Econ. Rev. 20 (4), 793–805. Su, X., Zhou, W., Nakagami, K.I., Ren, H., Mu, H., 2012. Capital stock-labor-energy
Gregory, A.W., Hansen, B.E., 1996. Tests for cointegration in models with regime substitution and production efficiency study for China. Energy Econ. 34 (4),
and trend shifts. Oxf. Bull. Econ. Stat. 58 (3), 555–560. 1208–1213.
Hall, R.E., 1978. Stochastic implications of the life cycle-permanent income Tang, W.Q., Wu, L.B., Zhang, Z.X., 2010. Oil price shocks and their short- and long-
hypothesis: theory and evidence. J. Political Econ. 86, 971–988. term effects on the Chinese economy. Energy Econ. 32, 3–14.
Hamilton, J., 1983. Oil and the macroeconomy since World War II. J. Political Econ. Wang, Y.S., 2013. Oil price effects on personal consumption expenditures. Energy
91 (2), 228–248. Econ. 36, 198–204.

Please cite this article as: Zhang, D., et al., International oil shocks and household consumption in China. Energy Policy (2014), http://dx.
doi.org/10.1016/j.enpol.2014.08.034i