Department of Economics and International Business Working

Department of Economics and International Business
Working Paper No. 16-08
August 2016
GAINERS AND LOSERS OF POLITICAL
INSTABILITY: EVIDENCE FROM THE
ANTI-JAPANESE DEMONSTRATION IN
CHINA
Zijun Luo
Department of Economics and International Business
Sam Houston State University
237C Smith-Hutson Business Building
Huntsville, TX 77341
[email protected]
Yonghong Zhou∗
Department of Economics
Jinan Univerisity, China
601 Huangpudadao West, Guangzhou,
Guangdong 510632, P. R. China.
[email protected].
∗ Corresponding
author
Abstract
This paper quantifies the Chinese consumers’ boycott of Japanese cars that immediately
followed the anti-Japanese demonstration in September 2012.We decompose the total boycott
effect into cancel effect and transfer effect. We find that the temporary cancellation of orders
by potential buyers accounts for more than 90% of the total decline in Japanese car sales. Such
results indicate that Japanese car makers lost these customers only for the short-run. European,
Korean, American, and Chinese cars became the dominant substitutes for lost Japanese sales;
Chinese brands benefited the least. This paper provides evidence of both negative and positive
impacts of political conflicts for different market participants and includes analysis of welfare
implications.
Keywords: China; Japan; Boycott; Automobile; Political Conflict
JEL Codes: O11, F51, L62
Acknowledgments: The authors benefited greatly from discussions with Le Wang. This paper
was presented at the Chinese Economists Society 2015 North America Conference, the 2016 CEC
Workshop, and seminars at Fudan University and Peking University. We thank conference and
seminar participants, especially Zhao Chen, Wei Huang, Lixing Li, Pinghan Liang, Tianyang Xi,
Yiqing Xu, and Shilin Zheng for their helpful comments. Zhou acknowledges financial support
from the National Natural Science Foundation of China [71103075, 71203077, 71273116] and
the Fundamental Research Funds for the Central Universities of Jinan University [12JNYH002,
12JNKY001, 15JNQM001].
1
Introduction
The influences of politics on economic performance have long been studied by economists
(Alesina and Rodrik, 1994; Acemoglu and Robinson, 2012, 2013; Mitra and Ray, 2014). These
influences are magnified in the age of globalization. The effects of bilateral political relations on
bilateral economic relations, such as trade and foreign direct investment (FDI), can be considered
from both supply and demand sides. On the supply side, the emergence of bilateral political
instability may cause a decline in trade and retraction of FDI. On the demand side, consumers
may boycott products of foreign brands regardless of whether they are imported or domestically
made.
Focusing on the China-Japan1 relationship, this paper studies how bilateral political instability
between the two countries affect the Chinese economy from the demand side. Specifically, we use
the anti-Japanese demonstrations of 2012 as a quasi-experiment to assess the impact of politics on
the automobile market in China. In a market economy such as modern China, substitutes, both of
domestic and foreign brands, are readily available. The worsening of political and hence economic
relationships with a particular country may benefit the competitors. To show such benefits, we
quantify and decompose the boycott of Japanese cars as a result of the demonstrations, into cancel
and transfer effects. The cancel effect captures declines in consumption of the whole automobile
market by potential Japanese car buyers, whether temporary or permanent, while the transfer effect
refers to potential consumers who buy non-Japanese brands of cars; these actions resulted in a
decline in Japanese brands sold, but an increase in the purchase of brands from other countries. To
the best of our knowledge, such decomposition has not been done in similar studies.
In the literature, there is much research on political instability or unpleasant recent histories
between countries and how this influences their relationship in trade and FDI, and will eventually
affect outputs of the countries involved. An extreme case of political instability is war. With data
from 1950–2000, Martin et al. (2008) report that intra-state military conflicts had a sizable impact
on bilateral trade. But higher levels of trade may not lead to more peaceful relationships. Similar
1
Some literature use “Sino-Japanese” instead.
2
results are demonstrated by Gowa and Mansfield (1993) for 1905–1985, Glick and Taylor (2010)
for 1870–1997, and Berger et al. (2013) for the Cold War period.
Studies of less severe events such as terrorism, sanctions, and other types of bilateral political
instability are also abundant. For example, Abadie and Gardeazabal (2003) find that after the
outbreak of terrorism in the late 1960’s, per capita GDP in the Basque Country in northern Spain
declined about 10 percentage points relative to a synthetic controlled region without terrorism.
Chavis and Leslie (2009) study the boycott of French wines by the United States after the French
government’s opposition to being involved in Iraq in early 2003. The authors show that the boycott
resulted in 26% lower weekly sales at its peak, and 13% total lower sales over the six month period
that the boycott lasted. Based on their estimates, Fershtman and Gandal (1998) demonstrate that
the ending of the Arab economic sanctions on Israel led to a per buyer gain of approximately
$2,343 in the automobile markets, which can be interpreted as a “peace dividend”. The empirical
investigation by Fuchs and Klann (2013) show that meeting with the Dalai Lama led to a reduction
in a trading partner’s exports to China. Their paper shows the important role politics can play in
economic relations with China.2
Three recent papers have focused specifically on the China-Japan relationship. Che et al. (2015)
study the relationship between the regional civilian casualties caused by the Japanese’ invasion
from 1937 to 1945 and the sizes of inward FDI and bilateral trade in 2001. According to their
results, there was a long-term negative impact from the conflict between these two countries. Using
the “Textbook event” in 2005 and the “Senkaku Event” in 2010 as two experiments,3 Fisman et al.
(2014) provide evidence that stock markets react to adverse shocks of China-Japan relations. They
2
But some researchers find no or insignificant influence of political instability on the economies. Notable ones are
Ashenfelter et al. (2007), Davis and Meunier (2011), and Li and Sacko (2002). Ashenfelter et al. (2007) report no effect
for United State’s 2003 boycott of French wine, as opposed to information from Chavis and Leslie (2009). Davis and
Meunier (2011) point out that variation in the number of negative events did not change bilateral economic activities
for the United States and Japan in their relationships with 152 countries with the exception of French-American
(Ashenfelter et al., 2007; Chavis and Leslie, 2009) and China-Japan (Che et al., 2015; Fisman et al., 2014, and the
current study). Even for these two special cases, they claim that there was little short-term economic impact of political
tensions. Li and Sacko (2002) attempt to give an explanation to the complex and uncertain effects of political stability
by introducing expectation and the connection between ex ante and ex post effects in their study.
3
Our data also cover the period of the “Senkaku Event” in 2010 that arisen from disputes over the Senkaku Islands,
known as Diaoyu Islands in China. However, we find no evidence of negative impact resulted from the “Senkaku
Event” in the Chinese automobile market.
3
show that stocks with high exposures of the other countries suffered relative decline during the two
events. The paper that is most relevant to our study is that of Heilmann (2015), who investigates
the 2012 episode of the Chinese consumer boycott of Japanese products, especially automobiles,
as a result of the Senkaku conflict. He reports a, average 2.7% trade disruption yearly due to the
event.
Unlike these studies that used aggregate level data (Heilmann, 2015) or data of multiple
industries and regions (Che et al., 2015; Fisman et al., 2014), we focused on the automobile
industry. There are two advantages in using the automobile industry.4 First, sales of cars increased
rapidly in the last two decades in China due to fast economic growth. As a result, the automobile
industry provides one of the best examples for the study of consumer behaviors in China, the focus
of the current study. Second, Japanese cars are known for their high quality and reliability and
are among the top choices of Chinese buyers. Since Japanese cars were intentionally damaged by
passers-by during the anti-Japanese demonstration, the demonstration caused negative impacts on
the Japanese automobile market. This latter reason helps with our identification.
Our empirical investigation of the Chinese automobile sales also confirmed the presence of the
boycott effect following the anti-Japanese demonstration on September 2012. When the decline
of Japanese sales is decomposed into cancel and transfer effects, we show that more than 90% of
the decline in Japanese car sales was due to the cancel effect. The majority of potential customers
stopped buying Japanese cars during the boycott, but they had not switched to buying cars of
non-Japanese brands. This loss in potential buyers is usually short-term, for consumers may come
back once the boycott is appeased. In terms of gainers, our analyses show that cars of non-Japanese
brands all benefited from the boycott, but Chinese brands profited the least.
The rest of the paper is organized as follows. The next section provides background information
on the anti-Japanese demonstration in 2012 and also describes our data. Section 3 details our
empirical methodology in identifying and decomposing the boycott effects. Section 4 presents our
4
Studies of the automobile industry are abundant in the economics literature. See, for example, Bandeen (1957);
Sheahan (1960); Hess (1977); Berkovec (1985); Bresnahan (1987); Cooper and Haltiwanger (1993); Ries (1993);
Berry et al. (1995), and Park (2003), covering a variety of topics. In addition, Depner and Bathelt (2005), Deng and
Ma (2010), Luong (2013) and Hu et al. (2014) provide analyses of various aspects of the Chinese automobile industry.
4
empirical results. The last section concludes.
2
Background and Data
The growth in income of Chinese families since the Reform and Opening-up has resulted in
the automobile industry’s bpp,omg sa;es. Partly because the Chinese automobile industry was
underdeveloped during the early years of the reform, the large market attracted many foreign
automobile companies to engage in joint venture with Chinese manufacturers and to introduce
their brands into the Chinese market. In the past decades, sales of foreign-brand cars have soared
due to their high quality and reliability over domestic brands. Popular origins of foreign-brand cars
include America, France, Germany, Japan, and Korea. In our analysis, we follow Hu et al. (2014)
and classify origins into five categories: Chinese, Japanese, Korean, American, and European.
Our car sales data were mainly collected from the monthly reports of passenger cars from
the China Association of Automobile Manufacturers (CAAM) for years 2008 to 2014. The data
include only domestic made cars sold all over China and hence exclude imports and exports. That
is, foreign-brand cars in this paper refer to those produced by joint venture or FDI which bear
the logos of foreign brands–all of them were made in China.5 Japanese brands in our data include
Honda, Isuzu, Mazda, Mitsubishi, Nissan, Suzuki, and Toyota.
We only include in our data three types of passenger cars according to their sizes and purposes:
basic passenger cars such as midsize sedans, sport-utility vehicles, and multi-purpose vehicles.6
Figure 1 shows shares of sales by countries of origin in selected years. Japanese cars accounted
5
Our analysis would be richer and more convincing if import, and possibly export, data were available. However,
we believe that the most important component of our results, namely the decomposition of the two effects, still has
significant merit with the current data set. As argued by Deng and Ma (2010), Chinese car buyers make their decisions
in two stages. During the first stage, they choose between domestic-produced versus imported cars. And in the second
stage, they decide on a specific brand and model. Hence, our analysis is confined to the second stage of the decision
making.
6
CAAM also reports the sales of Jiaocha cars, which are designed for both cargo and passenger purposes. This type
of car is specially designed for the Chinese market, especially in the rural areas. Since they are not general passenger
cars, and there is no foreign branded cars in this category, we exclude them from our sample. In the data, Jiaocha and
imported cars together account for about 15% (2013) to 25% (2009) of the Chinese new car market. Thus, it is no less
representative by restricting our attention to only the three types of passenger cars in our analysis.
5
for 26.34% of the Chinese market in 2008, which is greater than any other foreign origins but
behind the number posted by domestic Chinese brands. Although the sales share has declined,
it still accounted for 16.74% of the Chinese market in 2014, behind only Europe among foreign
origins.
(Insert Figure 1 about here)
It is important to point out that all sales numbers are reported at the model level and the
complete hierarchy of our data is manufacturer–brand–model. For example, Volkswagen (VW)
has two joint-venture manufacturers in China: FAW-VW and Shanghai-VW. These are two distinct
brands and their models are considered distinct models, although some may share the same
model name or same specification. Meanwhile, FAW-VW and Shanghai-VW also produce Chinese
domestic cars, which are also regarded as different brands from the VW’s brand. In Table 1, we
report the number of companies, brands, and models of each country of origin along with the
summary statistics grouped by country. In the table, negative sale numbers are due to adjustments
of orders and returns made by the manufacturers.
(Insert Table 1 about here)
In our data, there are 73 manufacturers (companies) in total; 29 of these companies make
foreign-brand cars as well as their own domestic brands. None of these 29 companies is associated
with more than one foreign company but could manufacture multiple brands for the same foreign
company as in the cases of Volkswagen, Audi, and Skoda. The total number of brands, as defined
by “company-brand”, is 98. The total number of models is 601. That is, each brand offers about
6 models on average. It is worth noting that car models may be named differently in the Chinese
market as compared to other markets. For example, Nissan Tiida in China is equivalent to Nissan
Versa in the United States. There are also China-only models such as Honda Ciimo produced by
Dongfeng-Honda that are the inaugural product of the eighth generation Civic.
Politically, an unstable relationship between China and Japan las loomed in recent years. Fueled
by the unhappy history between the two countries (Che et al., 2015), a sovereignty dispute over
6
the Diaoyu Islands, known as the Senkaku Islands in Japan, ignited large scale anti-Japanese
demonstrations and protests all over China in September of 2012. Boycotting products of Japanese
brands, especially automobiles, was at the center stage of the protests. It should be noted that the
boycotts were primarily spontaneous and sometimes led by non-governmental consumer groups.
The 2012 anti-Japanese demonstration resulted in a huge decline in Japanese car sales. Figure 2
shows the movement of sales and sales shares of Japanese cars in China. In the figure, the vertical
line at September of 2012 indicates the beginning of the boycott of Japanese products in China. In
the two immediate months of the boycott, the sales share of Japanese cars dropped sharply from
20.94% (August) to 8.42% (October). Although it recovered to 16.80% in December, it was still
below the level before the boycott; the effect of the boycott was significant and lasting.
(Insert Figure 2 about here)
Before our full econometric analysis, we presented two preliminary analyses by aggregating
our data into Japanese versus non-Japanese car sales. First, we ran
Saleict = α + βBoycottt × Japanc + θyear + θmonth + θcompany + θtype + εict ,
(1)
where Saleict denotes sales of car model i from origin c at time t. Boycottt is a dummy variable
that equals to 1 for months within a certain period starting September 2012 and 0 otherwise. We
considered three possible lengths when constructing Boycottt : one month (September 2012 only),
three months (September 2012 to November 2012), and six months (September 2012 to February
2013). Japanc is a dummy variable that equals to 1 for Japanese cars and 0 otherwise. Furthermore,
θyear , θmonth , θcompany , and θtype are, respectively, year, month, company, and type fixed effects
controlling for factors such as monthly macroeconomic fluctuation, annual updates of models and
marketing strategies, company characteristics, and type fixed effects. The results are reported in
Table 2.
(Insert Table 2 about here)
7
Table 2 confirms the boycott effect on Japanese cars after the 2012 anti-Japanese demonstration
in China for all length considered: one, three, and six months post demonstration. The average
monthly drops in sales per model ranged from the highest of 2041 cars from the three-month
estimation to the lowest of 1598 cars from the one-month estimation. With 55 Japanese models in
the Chinese market,7 the monthly decline was greater than 87,000 of Japanese cars with the lowest
estimate. Put differently, for the four months (September to December) that the effect lasted in
2012, the total decline amounted to approximately 350,000 Japanese cars.
Second, we estimated an autoregressive moving average (ARMA) model to provide
preliminary results for the decomposition. With the monthly total sales data of Japanese versus
non-Japanese cars, we predicted sales numbers after September 2012 with observations before the
boycott. By comparing the actual sales and predicted sales, we get an idea about cancel effect (CE)
and transfer effect (TE). Figure 3 shows the predicted sales from ARMA(1,1)8 along the actual
sales for Japanese cars. The predicted sales were in general greater than the actual sales after
September 2012, indicating sales of Japanese cars fell below their normal levels. Similarly, Figure
4 shows the predicted sales from ARMA(1,2) along the actual sales for non-Japanese cars. In this
case, the predicted sales were in general smaller than the actual sales, indicating above normal
sales levels for non-Japanese cars.
(Insert Figures 3 and 4 about here)
The above analysis obtains the rough estimates of TE by exploiting only the time series
property of Japanese versus non-Japanese cars. Such estimations have not taken into account
the substitutability between Japanese and non-Japanese cars. It is important to consider the
substitutability in our analysis of boycott. Assuming away the substitution may result in either
overestimation or underestimation of either CE or TE. In the next section, we present a novel
7
The number of models, 55, is only for the month of September 2012. Because of introduction of new models and
exits of old models during our whole sample period, the numbers of models here are not equal to the numbers we have
reported in Table 1.
8
Tables A-1 and A-2 in the appendix give results of the relevant unit root tests, as well as AIC and BIC values for
the selections of optimal degrees and lags.
8
method to decompose the total effect that takes into consideration the substitutability between
Japanese and non-Japanese cars.
3
Methodology
Our empirical analysis included two steps. First, we decomposed the boycott effects into CE
and TE. CE refers to the portion of customers who exited the market, possibly temporarily, due to
the boycott. For example, they might be loyal followers of Japanese cars but feared that owning
a Japanese car would have an adverse effect. On the other hand, TE refers to potential customers
who switched to brands of non-Japanese origins. Intuitively, loss in customers due to CE may
come back later, for instance, when the demonstration has ended. But loss in customers due to TE
indicates an absolute decline in market shares for Japanese cars.
After the decomposition, we quantified the benefits among non-Japanese brands by looking at
which country of origin exhibited the biggest increase in sales as a result of the boycott.
3.1
Decomposition of the Boycott Effect
Figure 5 shows the basic principle of the decomposition, where T denotes the policy time
(September 2012), with T − 1 and T + 1 denoting pre-treatment and post-treatment, respectively.
The control group refers to non-Japanese cars while the treatment group is composed of Japanese
cars. When Chinese consumers boycott Japanese cars, some of them cancel their buying plan
(shown as BB), while others transfer to purchase other brands (shown as AA). Other things being
equal, TE is identified by an increase in sales of non-Japanese cars while CE is identified by the
decline in sales of Japanese cars net of TE.
(Insert Figure 5 about here)
In this research, the event (boycott) had an impact on both the treatment (Japanese-cars) and
control (non-Japanese cars) groups, with opposite results. We begin the decomposition of the two
9
effects with the following month-specific model:
Saleict = α0 + α1 Japanc +
TX
+11
j=T
(βj Japanc × Mt=j + γj Mt=j ) + α2 Japanc × Postt≥T +12
(2)
+ α3 Postt≥T +12 + θyear + θmonth + θcompany + θtype + trendict + εict ,
where i, c, and t index model, country and time. Japanc is equal to 1 for Japan and 0 for all
other countries. T denotes September 2012 which is the beginning of the boycott. Mt=j = 1 for
t = j indicate the month-specific treatments within a year of the anti-Japanese demonstration,
and 0 otherwise, while Postt≥T +12 = 1 for all months more than one year post-demonstration
and 0 otherwise. The reason for grouping all months after one year is because Japanese car sales
recovered in 2013. We therefore focus our attention to the shorter term effects of the boycott.
Similar to (1), θyear , θmonth , θcompany , and θtype are year, month, company, and type fixed effects.
εict denotes the error term.
For the variable trendict in Equation (2), we consider two different types of time trends: a
common trend for all car models and unique trends for Japanese versus non-Japanese models. The
common trend approach is similar to the difference-in-differences (DID) estimation. However,
with this approach, the coefficient of βj was overestimated.9 We therefore preferred regression
results with unique trends, in which we added “(1 − Japan) × Trend” as additional controls to
distinguish between Japanese and non-Japanese time trends.
In Equation (2), coefficients βj , ∀ j ∈ [T, T + 11], measure the conditional average treatment
9
The basic fixed effect regression with common trend is
Yit = αi + λt + ρ̂Dit + εit ,
while the regression with unique trends is
Yit = αi + λ¯t + λ̃t Dit + ρ̃Dit + εit ,
where αi and λt indicate the group fixed effect and the time fixed effect and Dit is a dummy variable that equals 1
for the treatment group after policy and 0 otherwise. In the second regression, λ¯t denotes the collection of trends of
the non-treatment groups while λt Dit is the unique trend of the treatment group. With the common trend regression,
ρ̂ gives the estimate of ATE. However, if the unique trend model is the true model, then ρ̂ would be overestimated
because its value is equal to λ̃t + ρ̃ from the regression with unique trend.
10
effect of month t:
βt ≡ Et (∆SaleJapan−other | Πt ) ≡ Et (∆SaleJapan | Πt ) − Et (∆Saleother | Πt ),
where ∆Salek denotes differences of sales between month t and the average pre-demonstration
level for k ∈ {Japan, other, Japan − other}. Πt denotes the information set, including the year,
month, company, and type fixed effects discussed earlier. According to the principle illustrated in
Figure 5, we can further rewrite the average monthly treatment effects into
transfer Et (∆SaleJapan−other | Πt ) =Et (∆Salecancel
Πt ) − Et (∆Saleother | Πt )
Japan Πt ) + Et (∆SaleJapan
transfer = Et ( ∆Salecancel
Π
)
+2
E
(∆Sale
Πt )
t
t
Japan
|
|
{zJapan
}
{z
}
cancel effect
transfer effect
=CE + 2TE.
(3)
The last step in Equation (3) uses the fact that
Πt ) = −Et (∆Saleother | Πt ).
Et ( ∆Saletransfer
Japan
(4)
To further separate CE and TE and express them as percentages of total changes in sales, we
need information on the substitutability between sales of Japanese cars and other cars. Let δ denote
the marginal rate of substitution (MRS), which is conditional on the information set Λt ; we then
have
d(ln Saleother,t ) δ≡E
Λt ,
d(ln SaleJapan,t ) which may be rewritten into a discrete form as10
Saleother,t0
∆Saleother δ=E
Λ /
.
∆SaleJapan SaleJapan,t0
10
(5)
For a fully defined Equation (5), we still need to specify t0 , a reference time period, for our decomposition. We
will delay the discussion to Section 4.1 when empirical results are presented.
11
The MRS, as conventionally defined, measures the substitutability between sales of Japanese
versus other cars, based on the information set. We envisioned the information set to contain
not only observable market conditions, but also unobservable attitudes of the Chinese customers
toward foreign countries especially Japan. There are three main reasons for the set up of δ in
Equation (5). First, consumer preferences are stable. It is especially important that consumers’
attitudes toward Japan are stable, which is arguably true because of the long unhappy history
between the two countries. This is also why Equation (5) does not have a time dimension. Second,
non-Japanese and Japanese cars, especially those of similar sizes, are close substitutes, if not
perfect substitutes, in the sense that a family may purchase only one car at a time. As a result, the
utility function of consumers is a linear combination of consumptions of all cars with a consumer
buying the car that gives the highest utility. Last, although price information is unavailable in our
data, the prices of new cars in China are stable and can be anticipated almost perfectly. This is
especially the case because the used car market in China is underdeveloped and most consumers
choose to buy new cars from dealership or flagship stores.
In order to estimate the value of MRS between Japanese and non-Japanese brands, we
constructed a panel of one-to-one correspondence by matching Japanese brands with non-Japanese
brands with replacement. For example, if we matched Honda and Toyota with Audi, Ford,
and Volvo, there would be six pairs: Honda-Audi, Honda-Ford, Honda-Volvo, Toyota-Audi,
Toyota-Ford, and Toyota-Volvo.
The simplest specification to obtain δ is to run the regression
ln Saleother,t = α + δ ln SaleJapan,t ,
where Saleother,t and SaleJapan,t denote, respectively, sales of the paired non-Japanese and Japanese
brands. However, such specification has not taken into account factors contained in the information
12
set Λt . Hence, we prefer the augmented version of the above equation:
ln Saleother,t = α + δ ln SaleJapan,t
+ β2 (Boycottt × ln SaleJapan,t ) + β3 ln Saletotal,t
(6)
+ θmonth + θyear + εct .
In Equation (6), δ is the value of MRS that we are estimating. In addition β2 captures the
deviation in consumption of Japanese cars during the boycott, with the dummy variable Boycottt
equals to 1 for two different durations: in the first six months or one year after September 2012.
β3 measures changes in sales that resulted from overall market conditions at time t as captured by
ln Saletotal,t . Lastly, the year and month fixed effects, θmonth and θyear , are included.
Combining Equations (3), (4) and (5), we decompose the boycott effect into CE and TE as

Saleother,t
0
1+δ Sale


Japan,t0

CE = Et (∆SaleJapan−other | Πt )

Saleother,t

0

1−δ Sale

Japan,t0
Saleother,t

0

−δ Sale


Japan,t0


TE = Et (∆SaleJapan−other | Πt ) 1−δ Saleother,t0
SaleJapan,t
.
(7)
0
From Equation (7), we can also infer the sizes of CE and TE as percentages of the total effect,
Saleother,t0
Saleother,t0
which are 1 + δ SaleJapan,t
and −δ SaleJapan,t
, respectively.
0
3.2
0
Quantifying the Benefits
Another part of our analysis is the comparative analysis of changes in sales among
non-Japanese cars. Once the TE is isolated, we want to further decompose it into cars of different
countries of origin. This is accomplished by applying Equation (2) by country of origin rather
than in a pooled estimation. Comparing δ’s from these regressions of different countries give the
comparative results we are looking for and we can then quantify the benefits to non-Japanese
brands posed by the anti-Japanese demonstration in 2012 in China.
13
4
Empirical Results
4.1
Decomposition of the Boycott Effect
The estimation results of the monthly boycott effects based on Equation (2) are reported in
Table 3. Column (1) and (2) are baseline regressions with common and unique trends, respectively.
In column (1), all coefficients are negative and statistically significant, while only coefficients of
the first six months are statistically significant in column (2). We believe the regression with unique
trends, i.e., column (2) in Table 3, gives the estimates that are less biased. According to column
(2), the boycott phased out after six months. During the first six months of the boycott, the monthly
average decline in sales of Japanese cars per model ranged from the lowest of 1037 in December
2012 to the highest of 3208 in October 2012.
(Insert Table 3 about here)
In columns (3) and (4) of Table 3, we included the GAIN Index in the regression, which is the
monthly average price of different car models weighted by sales.11 During the boycott, Japanese car
manufacturers responded by lowering prices. Hence, a price effect wass to be expected. Comparing
the results with and without GAIN in the regression, the estimates without GAIN are about 10-20%
larger in absolute values. It is shown in column (4) that, during the first six months of the boycott,
the monthly average decline in sales of Japanese cars per model ranged from the lowest of 848 in
December 2012 to the highest of 3012 in October 2012.
Table 4 shows the values of MRS, δ, obtained from the estimation of Equation (6). In the table,
column (1) includes only observations before the boycott, while column (2) is based on the full
sample. Columns (3) and (4) further control for different length of the boycott. The coefficients
of ln SaleJapan are all negative and statistically significant at 1% level, showing that Japanese and
11
The GAIN Index is compiled by Shanghai ISE Management Consulting Co., Ltd., and is provided by Wind Info
(http://www.wind.com.cn/), a leading provider of financial data in China. The GAIN Index is a measure of the average
price in the whole automobile market rather than of individual brand or model. The GAIN Index starts in January
2011, which is why columns (3) and (4) have a smaller number of observations than columns (1) and (2) in Table 3.
14
non-Japanese cars are indeed substitutes. For samples before the boycott, the coefficient was -0.01
while the estimate from the full sample was -0.02.
(Insert Table 4 about here)
In Table 5, we report results of the decomposition. These results are based on column (2) of
Table 3 and δ = −0.02 from Table 4. In the table, two different initial values of Saleother,t0 and
SaleJapan,t0 are considered. The first two columns use the mean of sales of pre-treatment (t ≤
August 2012) while the last two columns use the sales of August 2012 (t = August 2012). In both
sets of initial values (t ≤ August 2012), we find TE accounts for less than 10% of the total sales
change of Japanese cars.
(Insert Table 5 about here)
The result that CE accounts for more than 90% of the total drops in sales of Japanese-brand
cars implies that most would-be consumers chose to cancel or delay their purchases during the
boycott. This further implies that potential buyers of Japanese brands have strong brand loyalty.
On the other hand, those 7% potential buyers who walked away from Japanese cars due to the
boycott will benefit cars of non-Japanese origins. Japanese brands have lost these customers, as a
result of the boycott, until they will buy yet another car. This represents long-term loss in sales and
market shares.
4.2
Gainers of the Boycott
While it is straightforward to understand that sales of Japanese cars suffered during the boycott,
an important question remains unanswered: who benefited? The estimation results based on the
econometric model outlined in Section 3.1 are reported in Table 6. In the table, we used each
country other than Japan as a control group, and we only considered the boycott effects, with
unique trend, during the first six months.12 All non-Japanese brand cars benefited from the boycott,
12
Estimation results including the GAIN Index are reported Table A-3 in the Appendix.
15
although no country appeared to benefit the most during the whole six month period. European cars
were the top sellers in the first two months and absorbed the largest portion in the declining sales
of Japanese cars. Both Korean and American cars then caught up in sales. Korean cars stayed
strong until the end of 2012. In January 2013, all four origins considered are strong substitutes of
Japanese-brand cars. At the end of the six month period we considered, Chinese brands were the
clear winners. As shown in Table 3, the boycott effect was phased out after six months.
(Insert Table 6 about here)
Chinese cars did not benefit the most during the boycott of Japanese cars because, in general,
Chinese domestic-brand cars and foreign-brand cars are quite different in price, design, and quality.
When Chinese buyers decide to purchase a car, they often first decide on domestic versus foreign
brands (Deng and Ma, 2010). Only if they decide to buy a foreign-brand car, do they need to make
a choice among Japanese, Korean, American, or European cars. Therefore, when they stop buying
Japanese cars during the boycott, they turn to other foreign brands rather than purchasing domestic
cars.
5
Conclusion
In the literature, quantitative evidence about the impact of bilateral political instability on the
economy is mixed. In this paper, we used the anti-Japanese demonstration and the subsequent
boycott of Japanese products in China in September 2012 as a quasi experiment to shed light on
the topic. With automobile sales data, we confirmed the presence of a boycott effect and further
decomposed the boycott effect into cancel effect and transfer effect, where the former refers to the
cancellation of orders by potential buyers while the latter refers to consumers who switch to cars
from other countries of origin. We noted that the cancel effect accounted for more than 90% of the
decline in Japanese car sales in the first six months after the boycott. With the majority of decline
attributed to the cancel effect, Japanese car makers lost these customers only for the short-run.
These customers would most likely come back and buy Japanese cars once the anti-Japanese
16
emotion is pacified. Among different countries of origin, there was no top seller during the first
six months of the boycott. European, Korean, American, and Chinese cars took turns to be the
dominant substitutes of lost Japanese sales.
It is worth noting that all car manufacturers in our data were Chinese firms. As a result, the
absolute decline in car sales during the boycott, which lasted for at least six months according to
our estimation, hurt Chinese workers and their welfare and was likely to have minimum effect on
Japanese car makers and their Japanese branches. The boycott had a negative effect on the Chinese
economy.
In fact, the 2012 boycott was not the first time such boycotts had occurred. An earlier
anti-Japanese demonstration, also caused by issues related to the Diaoyu Islands, took place in
September 2010, however, we did not observe any obvious decline in sales of Japanese cars during
September 2010. The reason of this difference is unclear to us and should warrant efforts in future
research.
References
Abadie, A. and J. Gardeazabal (2003). The Economic Costs of Conflict: A Case Study of the
Basque Country. American Economic Review 93(1), 113–132.
Acemoglu, D. and J. A. Robinson (2012). Why Nations Fail: The Origins of Power, Prosperity,
and Poverty. New York: Crown Business.
Acemoglu, D. and J. A. Robinson (2013). Economics versus Politics: Pitfalls of Policy Advice.
Journal of Economic Perspectives 27(2), 173–192.
Alesina, A. and D. Rodrik (1994). Distributive Politics and Economic Growth. Quarterly Journal
of Economics 109(2), 465–490.
Ashenfelter, O., S. Ciccarella, and H. Shatz (2007). French Wine and the U.S. Boycott of 2003:
Does Politics Really Affect Commerce? Journal of Wine Economics 2(1), 55–74.
Bandeen, R. A. (1957). Automobile Consumption, 1940-1950. Econometrica 25(2), 239–248.
Berger, D., W. Easterly, N. Null, and S. Satyanath (2013). Commercial Imperialism? Political
Influence and Trade During the Cold War. American Economic Review 103(2), 863–896.
Berkovec, J. (1985). New Car Sales and Used Car Stocks: A Model of the Automobile Market.
Rand Journal of Economics 16(2), 195–214.
17
Berry, S., J. Levinsohn, and A. Pakes (1995).
Econometrica 63(4), 841–890.
Automobile Prices in Market Equilibrium.
Bresnahan, T. F. (1987). Competition and Collusion in the American Automobile Industry: The
1955 Price War. Journal of Industrial Economics 35(4), 457–482.
Chavis, L. and P. Leslie (2009). Consumer Boycotts: The Impact of the Iraq War on French Wine
Sales in the U.S. Quantitative Marketing and Economics 7(1), 37–67.
Che, Y., J. Du, Y. Lu, and Z. Tao (2015). Once an Enemy, Forever an Enemy? The Long-run
Impact of the Japanese Invasion of China from 1937 to 1945 on Trade and Investment. Journal
of International Economics 96(1), 182–198.
Cooper, R. and J. Haltiwanger (1993). Automobiles and the National Industrial Recovery Act:
Evidence on Industry Complementarities. Quarterly Journal of Economics 108(4), 1043–1071.
Davis, C. L. and S. Meunier (2011). Business as usual? Economic Responses to Political Tensions.
American Journal of Political Science 55(3), 628–646.
Deng, H. and A. C. Ma (2010). Market Structure and Pricing Strategy of China’s Automobile
Industry. Journal of Industrial Economics 58(4), 818–845.
Depner, H. and H. Bathelt (2005). Exporting the German Model: The Establishment of a New
Automobile Industry Cluster in Shanghai. Economic Geography 81(1), 53–81.
Fershtman, C. and N. Gandal (1998). The Effect of the Arab Boycott on Israel: The Automobile
Market. RAND Journal of Economics 29(1), 193–214.
Fisman, R., Y. Hamao, and Y. Wang (2014). Nationalism and Economic Exchange: Evidence from
Shocks to Sino-Japanese Relations. Review of Financial Studies 27, 2626–2660.
Fuchs, A. and N.-H. Klann (2013). Paying a Visit: The Dalai Lama Effect on International Trade.
Journal of International Economics 91(1), 164–177.
Glick, R. and A. M. Taylor (2010). Collateral Damage: Trade Disruption and the Economic Impact
of War. Review of Economics and Statistics 92(1), 102–127.
Gowa, J. and E. D. Mansfield (1993). Power Politics and International Trade. American Political
Science Review 87(2), 408–420.
Heilmann, K. (2015). Does Political Conflict Hurt Trade? Evidence from Consumer Boycotts.
Journal of International Economics, forthcoming.
Hess, A. C. (1977). A Comparison of Automobile Demand Equations. Econometrica 45(3),
683–702.
Hu, W.-M., J. Xiao, and X. Zhou (2014). Collusion or Competition? Interfirm Relationships in the
Chinese Auto Industry. Journal of Industrial Economics 62(1), 1–40.
18
Li, Q. and D. Sacko (2002). The (Ir)Relevance of Militarized Interstate Disputes for International
Trade. International Studies Quarterly 46(1), 11–43.
Luong, T. A. (2013). Does Learning by Exporting Happen? Evidence from the Automobile
Industry in China. Review of Development Economics 17(3), 461–473.
Martin, P., T. Mayer, and M. Thoenig (2008). Make Trade Not War? Review of Economic
Studies 75, 865–900.
Mitra, A. and D. Ray (2014). Implications of an Economic Theory of Conflict: Hindu-Muslim
Violence in India. Journal of Political Economy 122(4), 719–765.
Park, B.-G. (2003). Politics of Scale and the Globalization of the South Korean Automobile
Industry. Economic Geography 79(2), 173–194.
Ries, J. C. (1993). Windfall Profits and Vertical Relationships: Who Gained in the Japanese Auto
Industry from VERs? Journal of Industrial Economics 41(3), 259–276.
Sheahan, J. (1960). Government Competition and the Performance of the French Automobile
Industry. Journal of Industrial Economics 8(3), 197–215.
19
2008
2010
Japan
26.34%
China
Japan
23.14%
China
Korea
Korea
Europe America
Europe
2012
America
2014
Japan
16.74%
Japan
18.69%
China
China
Korea
Korea
America
America
Europe
Europe
Graphs by year
Data source: Monthly reports of The China Association of
Automobile Manufactures (CAAM).
Note: Only domestically made basic passenger cars, sport-utility
vehicles (SUV), and multi-purpose vehicles (MPV) are included.
Imported and exported cars are excluded.
Figure 1: The Brand Origins of Car Sales in China: 2009-2014
20
Dec 16.80% 209.39
0
Oct 8.42% 93.55
100
200
300
Sales of Japanese cars (Thousand)
30.00
Percent of sales of Japanese cars (%)
10.00
20.00
Aug 20.94% 223.81
2008-01
2012-09
2014-12
Time
Data source: Monthly reports of The China Association of
Automobile Manufactures
Note: Only domestically made basic passenger cars, sport-utility
vehicles (SUV), and multi-purpose vehicles (MPV) are included.
Imported and exported cars are excluded.
Figure 2: Sales and Market Share of Japanese Cars (Monthly)
21
400000
100000
200000
Sale
300000
Actual
Fitted
2008-01
2012-09
2014-12
Time
Note: Out-of-sample predictions, shown as dashed line, from by
regressing the ARMA(1,1) with time trend on subsample from
2009-01 to 2012-08 are shown.
Figure 3: The Predicted and Actual Sales Numbers of Japanese Cars
22
1.5e+06
500000
Sale
1.0e+06
Actual
Fitted
2008-01
2012-09
2014-12
Time
Note: The ARMA(1,2) with time trend is regressed here to get the
out-of-sample predictions with subsample from 2009-01 to
2012-08, shown as dash line.
Figure 4: The Actual and Fitted Sales of non-Japanese Cars
23
Sales
AA: Transferring
Control
AA: Transferring
BB: Canceling
Treatment
T-1
T
T+1
Time
Figure 5: The Principle of Decomposition
24
Table 1: Descriptive Statistics
Country
Japan
China
Korea
America
Europe
Total
No. of
Company
14
57
2
4
9
73
No. of
Brand
15
61
2
7
13
98
No. of
Model
89
367
31
39
75
601
Mean
S.D.
Min.
Max.
Obs.
4010
2094
5143
5497
5519
3387
4417
3980
5486
6626
6356
5038
-107
-771
0
-28
-248
-771
29184
81153
27054
35603
46447
81153
4394
13793
1609
2079
3702
25577
Note: Negative sales are due to adjustments of orders and returns made by the companies. In the table, all foreign
companies refer to Chinese-foreign joint ventures while Chinese companies refer to those wholly owned by
Chinese legal persons. However, Chinese brands (62) and models (382) include those manufactured by
Chinese-foreign joint ventures that bear Chinese domestic brand names.
25
Table 2: Evidence of Boycott Effect
Sales of Japanese Cars
1 month
(1)
-1598.38∗∗∗
(277.01)
(2)
3 months
-2041.36∗∗∗
(285.93)
6 months
Adjusted R2
Observations
(3)
0.2223
25577
0.2230
25577
-1711.11∗∗∗
(241.03)
0.2234
25577
Note: OLS results of Equation (1) are shown. Standard errors clustered at the brand level are in parentheses.
∗
p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
26
Table 3: Monthly Estimates of Boycott Effects
Sales
September 2012
October 2012
November 2012
December 2012
January 2013
February 2013
March 2013
April 2013
May 2013
June 2013
July 2013
August 2013
Year
Month
Company
Type
GAIN
Trend
Adjusted R2
Observations
(1)
-2295.28∗∗∗
(386.66)
-3473.80∗∗∗
(495.04)
-2419.11∗∗∗
(379.61)
-1322.61∗∗∗
(411.26)
-2292.18∗∗∗
(406.97)
-2197.92∗∗∗
(443.10)
-1377.62∗∗
(537.67)
-1116.27∗∗∗
(402.64)
-1144.27∗∗∗
(353.35)
-1151.38∗∗∗
(362.61)
-1166.69∗∗
(486.16)
-1804.21∗∗∗
(449.85)
Yes
Yes
Yes
Yes
(2)
-2038.93∗∗∗
(416.14)
-3207.77∗∗∗
(535.59)
-2143.37∗∗∗
(441.45)
-1037.25∗
(544.69)
-1997.18∗∗∗
(335.13)
-1893.21∗∗∗
(539.17)
-1063.28∗
(630.79)
-792.28
(496.49)
-810.64∗
(437.41)
-807.77∗
(445.24)
-813.54
(528.12)
-1441.20∗∗∗
(529.96)
Yes
Yes
Yes
Yes
Common trend
0.2260
25577
Unique trend
0.2261
25577
(3)
-2004.75∗∗∗
(373.54)
-3177.05∗∗∗
(489.04)
-2141.75∗∗∗
(382.06)
-1042.03∗∗
(471.92)
-1995.19∗∗∗
(322.21)
-1896.62∗∗∗
(456.71)
-1068.39∗
(559.66)
-803.30∗
(416.97)
-831.52∗∗
(350.97)
-836.57∗∗
(358.39)
-853.07∗
(469.86)
-1503.70∗∗∗
(443.86)
Yes
Yes
Yes
Yes
Yes
Common trend
0.2475
16571
(4)
-1854.76∗∗∗
(382.62)
-3012.46∗∗∗
(480.69)
-1962.51∗∗∗
(367.38)
-848.24∗∗
(388.76)
-1786.84∗∗∗
(405.90)
-1673.68∗∗∗
(460.87)
-830.86∗∗
(418.05)
-551.18∗
(295.07)
-564.83∗
(303.16)
-555.26
(386.48)
-557.26
(425.72)
-1193.26∗∗∗
(401.49)
Yes
Yes
Yes
Yes
Yes
Unique trend
0.2475
16571
Note: Non-Japanese cars are used as control group in this table. In regression with unique trend, we added
“(1 − Japan) × Trend” as control variable. Standard errors clustered at the brand level are in parentheses.
∗
p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
27
Table 4: Estimates of Marginal Rate of Substitution (MRS)
ln Saleother
ln SaleJapan
ln Salerest
Boycott6 month × ln SaleJapan
Boycott1 year × ln SaleJapan
Month
Year
Sample
Adjusted R2
Observations
(1)
-0.01∗∗∗
(0.00)
Yes
Yes
Yes
Before Boycott
0.0285
43039
(2)
-0.02∗∗∗
(0.00)
Yes
Yes
Yes
Full
0.0410
68665
Note: Robust standard errors are in parentheses. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01.
28
(3)
-0.02∗∗∗
(0.00)
Yes
Yes
Yes
Yes
Full
0.0411
68665
(4)
-0.02∗∗∗
(0.00)
Yes
Yes
Yes
Yes
Full
0.0420
68665
Table 5: Decomposition of Monthly Boycott Effects
δ = −0.02
Salet0 = Mean of (Salet≤Aug. 2012 )
Month
September 2012
October 2012
November 2012
December 2012
January 2013
February 2013
CE (93.30%)
-1783
-2805
-1874
-907
-1746
-1655
TE (6.70%)
-128
-201
-135
-65
-125
-119
29
Salet0 = Salet=Aug. 2012
CE (92.45%)
-1753
-2757
-1842
-892
-1717
-1627
TE (7.55%)
-143
-225
-150
-73
-140
-133
Table 6: Monthly Estimates of Boycott Effects with Unique Trend
September 2012
October 2012
November 2012
December 2012
January 2013
February 2013
Adjusted R2
Observations
China
(1)
-1827.03∗∗∗
(414.13)
-2992.37∗∗∗
(529.21)
-1917.68∗∗∗
(447.61)
-1261.70∗∗
(513.06)
-1592.77∗∗∗
(319.26)
-2003.83∗∗∗
(549.94)
0.2048
18187
Korea
(2)
-2360.45∗∗∗
(629.09)
-3394.78∗∗∗
(599.68)
-2627.36∗∗∗
(467.64)
-1698.83∗∗∗
(574.37)
-2291.41∗∗∗
(337.45)
-1336.48∗∗
(538.60)
0.1703
6003
America
(3)
-2402.25∗∗∗
(559.26)
-2893.91∗∗∗
(908.19)
-2589.97∗∗∗
(742.55)
-410.26
(772.63)
-2668.19∗
(1333.68)
-1496.31
(930.05)
0.1939
6473
Europe
(4)
-2428.33∗∗∗
(752.30)
-4159.93∗∗∗
(835.57)
-2378.91∗∗∗
(747.24)
-19.59
(1267.55)
-2472.24∗∗∗
(785.27)
-1368.96∗∗
(639.39)
0.2662
8096
Note: The titles of the columns, namely China, Korea, America and Europe, refer to the use of that country’s car
sales as control. Clustering standard errors at brand level are in parentheses. All regressions include year, month,
company, and type fixed effects. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
30
Appendix
Table A-1: Unit Root Tests
Series
Intercept
Trend
Lags
Z(t)
SalesJapan,t
Yes
Yes
1
Residual of ARMA(1,1), Japanese cars
Yes
No
1
Salesother,t
Yes
Yes
1
Residual of ARMA(1,2), non-Japanese cars
Yes
No
1
-5.14
(0.00)
-5.91
(0.00)
-6.89
(0.00)
-8.66
(0.00)
Critical Value
Stationary?
1%
-4.08
5%
-3.47
10%
-3.16
-3.53
-2.90
-2.59
Yes
-4.08
-3.47
-3.16
Yes
-3.53
-2.90
-2.59
Yes
Yes
31
Note: MacKinnon approximates p-value for Z(t) are reported in the parentheses of column Z(t).
Table A-2: Selection of Optimal Degrees and Lags
ARMR
(1,1)
(1,2)
(2,1)
(2,2)
Japanese cars
AIC
1041.70∗
1042.08
1042.17
1042.31
Note: Note: * denotes the minimum of AIC or BIC.
non-Japanese cars
BIC
1050.62∗
1051.00
1051.09
1051.23
AIC
1134.93
1133.95∗
1134.26
1135.66
BIC
1143.85
1142.87∗
1143.19
1144.58
Table A-3: Monthly Estimates of Boycott Effects with unique trend and GAIN Index (Starts
January 2011)
September 2012
October 2012
November 2012
December 2012
January 2013
February 2013
GAIN
Adjusted R2
Observations
China
(1)
-1642.46∗∗∗
(396.67)
-2791.21∗∗∗
(486.67)
-1707.26∗∗∗
(408.45)
-1035.26∗∗
(426.45)
-1342.19∗∗∗
(354.76)
-1725.01∗∗∗
(499.79)
Yes
0.2097
11932
Korea
(2)
-2162.97∗∗∗
(576.74)
-3179.76∗∗∗
(534.18)
-2370.54∗∗∗
(399.22)
-1421.05∗∗∗
(442.08)
-2002.87∗∗∗
(329.71)
-1028.63∗
(523.49)
Yes
0.1513
3938
America
(3)
-2024.45∗∗∗
(557.72)
-2465.71∗∗
(919.07)
-2138.68∗∗
(796.57)
77.22
(777.57)
-2074.48
(1211.45)
-875.47
(748.39)
Yes
0.1855
4140
Europe
(4)
-2236.86∗∗∗
(650.57)
-3983.90∗∗∗
(749.73)
-2234.50∗∗∗
(620.42)
110.85
(1069.46)
-2357.17∗∗∗
(718.58)
-1267.90∗∗
(593.99)
Yes
0.2748
5159
Note: The titles of the columns, namely China, Korea, America and Europe, refer to the use of that country’s car
sales as control. Clustering standard errors at brand level are in parentheses. All regressions include year, month,
company, and type fixed effects. ∗ p < 0.1, ∗∗ p < 0.05, ∗∗∗ p < 0.01
32