Reprisals Remembered: German-Greek Conflict and
Car Sales during the Euro Crisis∗
Vasiliki Fouka
Hans-Joachim Voth
First draft: December 2012
This draft: May 2013
Abstract
During the debt crisis after 2010, car sales in Greece contracted sharply. This paper
examines how much German car sales declined in periods of German-Greek conflict.
It shows that German car sales fell sharply in areas affected by massacres during
World War II – especially in periods of public conflict between the German and the
Greek government. The German government was widely blamed for the harsh austerity measured imposed on Greece as a condition for several bailout packages. This
led to public outrage and a rapid cooling of relations between the two countries. We
conclude that cultural aversion was a key determinant of purchasing behavior, and
reflected the memories of past conflict in a time-varying fashion. We compile a new
index of public acrimony between Germany and Greece based on newspaper reports
and internet search terms. In addition, we use historical maps on the destruction
of villages and the mass killings of Greek civilians by the German occupying forces
between 1941 and 1944. During months of open conflict between German and Greek
politicians, sales of German cars fell markedly more than sales of those produced in
other countries. This is especially true in areas affected by German reprisals during World War II: areas where German troops committed massacres and destroyed
entire villages curtail their purchases of German cars to a much greater extent in
conflict months than other parts of Greece.
Keywords: Consumer behavior, cultural aversion, political conflict, boycott, car sales,
Euro crisis, German-Greek relations.
∗
We would like to thank Jordi Gali, Luigi Pascali, Giacomo Ponzetto, as well as seminar participants
at CREI for their helpful advice. The Hellenic Statistical Authority kindly provided car registration
data.
1
1
Introduction
When German Chancellor Angela Merkel visited Athens in 2012, she was greeted by
angry protests. Demonstrators carried placards depicting her in Nazi uniform, denouncing the rise of a “Fourth Reich”. Since 2010, there has been repeated and severe
conflict between the German and the Greek governments. Ever since it was discovered
that Greece had forged debt figures, forcing it into a humiliating EU bailout, German
politicians in particular were blamed for harsh austerity measures. Adding fuel to the
flames of discontent was the fact that politicians in Northern Europe – and especially
in Germany – repeatedly made disparaging remarks about the country, and the popular press printed incendiary headlines and insulting images. Greek consumer groups
quickly began callling for a boycott of German goods. Strikingly, memories of World
War II played an important role in Greek condemnations of German attitudes and policies: German institutions were defaced with swastikas; populist politicians demanded
German reparations for war crimes; and newspapers both foreign and Greek recounted
German massacres after 1941.
In this paper, we examine how these instances of public conflict affected consumer
behavior. In particular, we test if areas that suffered German war crimes during the
occupation saw sharper changes in purchasing patterns. To trace the impact of cultural aversion – and the way it interacts with location-specific memories – we focus
on car purchases. Cars are often seen as an archetypal German product.1 Examining
car purchases constitutes a demanding test – they represent a major (and rare) investment. We hypothesize that public conflict and calls for a boycott of German products
reduced sales of German automobiles, and that these reductions were greater where the
Wehrmacht had destroyed entire villages and committed massacres.
1
In the 1980s, Audi discovered that it was not perceived as a German car producer in some of its
core markets, such as the UK. It began an advertising campaign featuring a highly idiomatic – and
near-inscrutable – German motto (“Vorsprung durch Technik”) to drive home the point of its national
origin.
2
We first compile an index of German-Greek political clashes based on newspaper
reports in Greece and on the frequency of internet search terms. These indicators show
an explosion of conflict after 2010, with eight separate periods of particularly intense
public conflict. Sales of German-made cars suffered marked declines during political
clashes. We combine this time-series information with detailed data on the location
of German massacres during the occupation, 1941-44. Following an upsurge in partisan activity, the German occupying forces adopted a policy of harsh reprisal measures.
These involved burning entire villages, and killing the entire civilian population (Mazower, 1995). To measure the severity of these attacks at the local level, we use lists
drawn up by the Greek government designating localities as “martyred towns”. These
are based on a set of criteria including the percentage of homes destroyed, as well as the
loss of human life. The locations of these martyred towns are then mapped onto the
50 prefectures into which Greece is divided for administrative purposes. Car registration data are collected and disseminated at the level of the prefecture. We find strong
evidence that in areas of Greece affected by Wehrmacht reprisals during World War
II, sales of German-made automobiles fell more sharply than elsewhere during conflict
months. This suggests that consumer behavior – the purchase of big-ticket items like
cars – responds strongly to general public sentiment; where local memories of earlier
German wrong-doing could easily be activated, purchases were curtailed sharply.
Boycotts are frequently used to articulate political views. Evidence on their effectiveness is mixed. More than half of top brands in the US were targeted by a call for a
boycott in the period 1980-2000. There is only limited evidence that consumer behavior
is directly influenced by calls for political action. For example, boycotts of French wine
after the country’s failure to support the US in Iraq were probably ineffective Ashenfelter et al. (2007);2 . Teoh et al. (1999) examine the effect of the South African Boycott on
2
Chavis and Leslie (2006) earlier concluded that French wine sales in the US suffered after the start
of the Iraq war.
3
firm valuation, and find that it had no clear impact. In general, the stock prices of firms
in calls for boycotts typically do not react (Koku et al., 1997). John and Klein (2003)
argue that The main counter-examples include a decline in tourist visits by Americans
to France after 2003 (Michaels and Zhi, 2010); lower French car sales in China during
the 2008 Olympics (Hong et al., 2011); and suggestive evidence that French-sounding
products saw their sales slump in the aftermath of the Iraq war (Pandya and Venkatesan, 2012). In a closely related paper, (Fisman et al., 2012) examine the stock market
value of Chinese and Japanese firms after a cooling of Sino-Japanese relations in 2005
and 2010. This followed the introduction of new Japanese textbooks that downplayed
events during the Japanese invasion of China in the 1930s. Stock prices fell more for
firms that had a higher sales exposure to the other side, especially in industries with
large public sector involvement. They find no evidence of systematic changes in stock
prices of firms in consumer goods sectors.
Our paper also contributes to an emerging literature on the importance of cultural
factors in economic and social behavior. Countries that fought numerous wars in the
past continue to trade less with each other to the present day, and they engage in less
FDI Guiso et al. (2009). Fertility behavior of immigrants’ children is still influenced by
their parents’ country of origin (Fernández and Fogli, 2006), language characteristics
are associated with savings behavior (Chen 2012), and inherited trust can influence
national growth rates (Algan and Cahuc, 2010). Many attitudes persist over long
periods: Italian cities that were self-governing in the Middle Ages are richer and more
civic-minded today (Guiso et al., 2007), areas of Africa affected by 19C slave-hunts
have lower trust in the present (Nunn and Wantchekon, 2011), and German cities that
persecuted their Jews during the Black Death were markedly more anti-Semitic in the
1920s and 1930s (Voigtländer and Voth, 2012).
At the same time, culture is not only persistent, it can also change quickly: Attitudes towards pre-marital sex have been transformed in the last century (FernándezVillaverde et al., 2011); Islam changed from an open and tolerant religion to a rela-
4
tively intolerant one (Chaney, 2008); Franco-German conflict in the last 200 years was
repeated and seemingly deeply rooted in cultural differences (Mann 1916), but has vanished in the last 50 years. One of the key challenges for cultural economics is to analyse
the conditions for persistence, and the context in which contemporary attitudes are no
longer influenced by the past.
Relative to the existing literature, we make the following contribution: First, we
are among the first to show that actual consumer behavior is modified in response to
time-varying cultural aversion,3 and not simply stock-market valuations of large firms
mainly selling to the public sector (Fisman et al., 2012). Second, we show that crosssectional differences in memories of past conflict and “reasons to hate” matter in periods
of general conflict. Third, this paper demonstrates that it is not only small consumer
expenditures, such as purchases of shampoo or coffee, but purchases of big-ticket items
(like cars) that are affected by political conflict.
We proceed as follows. Section II presents the history and background of GermanGreek conflict since 1941 and describes our data. Section III presents the main results,
and section IV shows robustness checks and extensions. Section V concludes.
2
Historical Background and Data Description
In this section, we briefly summarize the history of German-Greek conflict during World
War II, as well as the period of crisis after the outbreak of the sovereign debt crisis in
2010. Next, we summarize how our data on car sales and on German reprisal measures
is constructed.
3
As in Michaels and Zhi (2010), Hong et al. (2011), and Pandya and Venkatesan (2012)
5
2.1
German retribution measures in Greece during WWII
Following a military campaign that lasted less than a month, Greece was occupied
by Axis forces in May 1941. The country was divided into three occupation zones.
The largest one was administered by Italy. Germany occupied a smaller part of the
territories, but controlled crucial locations including Athens, Thessaloniki and Crete.
Bulgaria administered a relatively small part of the country close to its own borders
(Figure 1). From the beginning, the civilian population suffered under the harsh measures of expropriation and plunder that followed the occupation. The German armed
forces requisitioned foodstuffs on a vast scale, leading to a major famine during the
winter 1941-1942. An estimated 300,000 people died, and the period still survives in
Greek collective memory (Hionidou, 2006).
Throughout Eastern Europe, the German armed forces targeted the civilian population in a bid to deal with partisan attacks. Shooting of potentially uninvolved civilians
in areas of armed resistance was first authorized in April 1941 in Yugoslavia (Mazower,
1995). Following the capture of Crete - involving heavy losses by the Wehrmacht in the
face of determined local and Allied resistance – reprisal measures were also used there
(Nessou, 2009). General Student, the temporary commander of Crete after the German
invasion of the island, explicitly instructed his forces to “leave aside all formalities and
deliberately dispense with special courts”, since these were not fit for “murderers and
beasts”. The town of Kondomari in Crete was the first to witness a mass execution of
civilians by the Germans on Greek soil: 19 people were shot in June 2, 1941, in retaliation for the death of a German military official in the town’s vicinity (Meyer, 2002).
Both mass shootings and the burning down of villages became standard practice. Until
1944, an estimated 2-3,000 Greek civilians were executed by the German armed forces
on Crete alone, and 1,600 (out of a total of 6,500) towns and villages were destroyed
(Nessou, 2009, p. 204).
After the defeat of Rommel in El Alamein and the Italian armistice with the allied forces, the Italian-occupied zone of Greece was handed over to the Germans in
6
September 1943. The Italian occupying forces had been notably lax in their attempts
to subdue local partisan groups (“andartes”). Following the German take-over, conflict
between guerrilla groups (mostly the Communist-led ELAS) and the Wehrmacht intensified. An upsurge of terror tactics employed by the occupying forces followed. Partisan
attacks were often followed by indiscriminate shootings of civilians and the destruction
of every village in a certain radius from the attack. For example, the town of Mousiotitsa in the northwestern part of Greece had 153 of its inhabitants killed, including
women and children, on July 25th 1943. Another 15 other localities in the area were
destroyed by the Germans (Nessou, 2009). The massacre was part of a mopping-up
operation in response to the killing of a German officer in the town of Zita. Similarly,
the entire male population of the town of Kalavryta in the Peloponnese was shot, along
with inhabitants of several neighboring towns, for a total number of 696 dead, after
guerrillas abducted and killed soldiers of the 117 Jaeger Division in October 1943. One
of the last massacres of civilians before the end of the occupation occurred in Distomo,
near Delphi. In total, 218 people, including infants, were killed by the Waffen-SS on
June 10th 1944. Post-war reports of the Ministry of Reconstruction estimate the total
number of dead in Greece due to reprisal measures at 30,000 (Doxiadis, 1947).
Memories of Nazi massacres during the occupation are far from erased in Greece
today. Family members of the victims of Distomo have tried to claim reparations,
taking their case to the German courts and to the International Court of Human Rights.
Despite the fact that the Constitutional Court in Germany dismissed the case in 2003,
it was recently revived when an Italian court awarded the descendents of the victims
property belonging to a German NGO in Italy. The case reached the International
Court in 2012 in the middle of the Greek sovereign crisis, and featured prominently in
the Greek press.4 .
4
“The government in the Hague for Distomo”, Kathimerini, 13 January 2011, http://news.
kathimerini.gr/4dcgi/_w_articles_politics_2_13/01/2011_428531
7
Data on towns that suffered reprisal measures by the Wehrmacht during the German
occupation of Greece come from Presidential Decrees no. 2130 (1993), 399 (1998), 99
(2000), 40 (2004) and 140 (2005). These decrees designate a number of municipalities
and communes throughout the country as “martyred towns”. Localities that fall in this
category were determined - by a committee created in 1997 by the Ministry of Internal
Affairs and Public Administration - to have suffered material and human losses in the
period 1941-1944 that fullfill one of the following criteria:
1. Complete destruction of housing stock by arson, bombings or explosions.
2. Loss of 10% of the period’s total population by individual or mass executions, as
well as by other causes, e.g. blind shootings of civilians.
3. Destruction of housing stock that approaches 80% of the total and loss of population that approaches 10% of the total, also taking into account the absolute
magnitudes of the losses.
This list of locations includes a total of 72 towns, from which we exclude the following: Doxato, Drama, Choristi (under Bulgarian occupation and destroyed by the
Bulgarians) and Domeniko, Tsaritsani, Nea Agchialos (destroyed by the Italians). Figure 2 depicts the regional distribution of affected localities. All places on the list of
martyred towns suffered due to German reprisals; they were not destroyed by bombing
during the war or during the invasion. 54 out of 72 witnessed mass executions of civilians; the rest were burnt to the ground in retaliation for an insurgency attack against
German armed forces in the vicinity(Nessou, 2009). Since data on car registrations,
our main dependent variable, are not available at a level of aggregation finer than the
prefecture, we construct a prefecture-level index of exposure to German reprisals, in the
form of the share of the prefecture’s total population in 1940 that lived in “martyred”
localities.
8
2.2
German-Greek relations during the Greek crisis
The Greek sovereign debt crisis began to unfold in late 2009, when revised budget
deficit figures revealed the country’s dire financial situation. This lead to successive
downgrades of its credit rating. Eventually, with debt markets all but closed to the
Greek state, an EU bailout became inevitable. From the beginning, the German government was sceptical of a financial rescue for Greece.5 It finally agreed to the bailout
in exchange for harsh austerity measures. From the onset of the crisis, Greek public
opinion saw Germany as the instigator of foreign-imposed austerity. The reaction was
immediate and intense: In February 2010, the Greek Consumers Association called
for a boycott on German products - explicitly highlighting the importance of cars and instructed consumers on how to identify German manufacturers by the products’
barcode.
Animosity was further aggravated by incendiary articles in the popular press. German newspapers routinely portrayed Greeks as notorious and lazy cheaters living it up
at the expense of the German taxpayer.6 A German weekly featured Aphrodite making
a rude gesture on the cover page; German populist politicians suggested that Greece
should sell some of its islands to repay its debts.7 As the Greek economy contracted and
unemployment surged amid severe austerity measures, anti-German feelings in Greece
deepened. Greek politicians openly referred to the German special envoy as a “military
commander”. In early 2012, Greek president Karolos Papoulias publicly complained
that the entire country was being insulted by the German finance minister Wolfgang
Schaeuble. During the 2012 visit of German chancellor Angela Merkel to Athens, thousands of people demonstrated in the streets of Athens.
Much of the criticism of German policy in Greece after 2010 used references to the
5
“German “no” to facilitating the repayment of the 110 billion euros”, Kathimerini, 13 October
2010
6
“Die Griechendland-Pleite”, Focus Magazine, Nr. 8, 2010.
7
“Verkauft doch eure Inseln, ihr Pleite-Griechen”, Bild, 27 October 2010
9
events of World War II, and employed Nazi-era symbols to protest against the way
Greece was being treated. Mentions of war crimes and unpaid German reparations
became much more frequent in the press. Populist politician fed the frenzy: Former
foreign minister Stavros Dimas, addressing the Greek parliament in March 2011, stated
that Greece never waived its right to claim reparations and the repayment of the loan
that Germany forced on it during Nazi occupation.8 Protesters often carried placards of
Angela Merkel in Nazi uniform; popular newspapers would print swastikas surrounded
by the stars of the European union to symbolize that EU policy was similarly harsh as
Nazi occupation.
An article by the English Daily Telegraph illustrates the way in which past conflict
suddenly came to matter for Greeks after the start of the debt crisis. In the issue of
February 11, 2012, the Telegraph profiled the life of Eleftherios Basdekis, who spent his
“entire life beneath a German cloud”. A survivor of the Distomo massacre, he eventually
build a successful trucking business, which went bankrupt after the start of the crisis.
The article also cited a mother from Distomo saying that she “hated Germany”, that
Angela Merkel was “a monster”, and that the Germans “killed Distomo; they stole our
gold; they belittle Greece.” A bar owner is quoted as saying “five years ago, no one had
any problem with Germany. But now people are getting upset. The Germans say we
are lazy, which is not fair”.
As the Telegraph article illustrates, hatred of Germans suddenly resurfaced after the
outbreak of the debt crisis. In addition, Greeks from towns destroyed after 1941 often
interpreted recent acrimony in the light of earlier war crimes. Our hypothesis is that
the persistence of collective memories of the German occupation is stronger in areas
of Greece that actually fell victim to German atrocities and that the revival of these
memories during specific conflict events manifests itself through consumer decisions.
8
“The issue of German reparations is open but...”, Kathimerini, 28 March 2012, http://www.
kathimerini.com.cy/index.php?pageaction=kat&modid=1&artid=83628&show=Y
10
In order to identify months of heightened conflict in German-Greek relations during
the euro crisis, we use information from Lexis-Nexis and Google Insights. Figure 3
shows the frequency of the co-ocurrence of the words “anti-German” and “Greece”
in international news media. For the years before 2009, the word pair is virtually
inexistent. Thereafter, the frequency count increases sharply, reaching peaks in 2010
and 2011.
To obtain a high-frequency measure of perceived German-Greek conflict within
Greece, we use data on web searches from Google Insight. Here, we use the number of terms related to the crisis in Greece and the particular role that Germany played
in it. For a given search term, the frequency index provided by Google Insights is a
normalization of the share of total searches represented by the term in a given time and
region.9 We use this index to construct a measure of public interest in German-Greek
related issues, based on the following searches - in Greek and from computers whose
IP identifies them as located in Greece - for the following terms: “Germans”, “German
reparations”, “Distomo”, “Merkel”, “Schäuble”, “Rösler”, “troika”, “firings”, “cuts”,
“haircut”, “measures”, “memorandum”.
These search terms were selected on the basis of their frequency of appearance in the
headlines of Greek newspapers and their general association to issues that posed a source
for tension in German-Greek relations during the period 2008-2012. It is surprising that
the vague term “Germans” first appears with a non-zero value in the Google index in
February 2010, exactly the month when the first austerity measures were announced
and the consumer boycott started. The volume of Google hits on the names of German
politicians, like chancellor Angela Merkel and Minister of Finance Wolfgang Schäuble
jumps discretely in early 2011. In the case of Vice-Chancellor Philipp Rösler it takes on
non-zero values in two instances: His visit to Athens in October 2011 and in March 2012,
9
Only terms with hits above a certain threshold are considered for the construction of the index.
As a result, the index often takes on the value 0 when the search volume for a term is low.
11
when the Greek Ministry for Development issued an official announcement accusing him
of “systematically undermining” Greece’s effort to recover.10 The term “memorandum”
refers to the various memoranda signed between Greece and the EU/IMF, with which
Greece agreed to economic policy conditionality in return for financial help. Though not
strictly a term referring to German-Greek relations (like the terms “troika”, “firings”,
“cuts”, “haircut” and “measures”) it is related to the general public annoyance caused
by the demands for austerity made of Greece during the crisis, which were perceived as
primarily German demands, pushed by the German government and voiced by German
politicians.
For each of the above terms we exctract a monthly search index from Google for
the period 2008-2012. The value of the index is practically zero until the second half
of 2009. We compute deviations from the mean for each of these series and take first
differences. We then sum up the resulting series and end up with an index for the
evolution of growth of the popularity of the combined search terms.
We identify conflict months based on turning points of the above series. The criterion
for identifying a month as a turning point is for it to be a local maximum in the quarter
(yt > yt−1 and yt > yt+1 ) and to be larger that one standard deviation. The turning
points resulting from this procedure are depicted in Figure 4. February 2010 is the
obvious first big event, followed by a series of turning points coinciding with major
episodes in the Greek debt crisis. Several of the event-months are characteristically
German-focused, like October 2010, when Germany led the EU summit that decided on
Greece’s bailout, by insisting on harsh conditionality and even suggesting that indebted
countries be stripped of their voting rights in the future.11 The largest peak in the series
is February 2012, a month especially conducive to agitated Greek nationalist feeling,
which saw both the dismissal of Greek claims for German war crime reparations in
10
“Scatter bomb of Chrisochoidis against Rösler”, To Vima, 1 March 2012, http://www.tovima.
gr/politics/article/?aid=446212
11
“EU Agrees to Merkel’s Controversial Euro Reforms”, Spiegel, 29 October 2010
12
the International Court of the Hague and the press statement of the Greek president
that he cannot accept insults to his country from Mr. Schäuble.12 In this month, the
Google search index for the term “Schäuble” takes on its maximum value. Table 2
offers a summary of the main events coinciding with our identified months of conflict.
2.3
Passenger vehicle registrations
In Greece, the Ministry of Transport and Communications collects data on registrations
of new passenger vehicles. These are disseminated by the Hellenic Statistical Authority
(HelStat). We have access to monthly information on the number of new passenger
vehicles registered in each prefecture for the period from January 2008 to October
2012, by manufacturing plant. The raw data does not contain information on the
brand of a registered vehicle. HelStat provides a correspondence list that allows the
matching of production plants to car manufacturers. This correspondence does not
always distinguish between different car brands produced by the same manufacturer.
This is especially true in the case of the Daimler group, producer of both Smart and
Mercedes vehicles, and for the Fiat group, which also produces Alfa Romeo and Lancia.
Despite this issue, we are able to distinguish German from non-German brands in our
sample; the former include Volkswagen, Opel, Audi, BMW, Porsche and the brands
of the Daimler group.13 For our purposes, a car’s “nationality” is not determined by
ultimate ownership of the company, but the place of manufacture of (most) cars. In
this sense, we count Seat as a Spanish car maker despite the fact that it is owned by
Volkswagen.
Summary statistics for the monthly sales of the brands in our sample are given in
Table 1. Toyota is the brand with the highest average sales number, followed by Opel
12
“Greek president attacks German minister’s insults”, Reuters, 15 February 2012, http://www.
reuters.com/article/2012/02/15/us-greece-germany-idUSTRE81E1VK20120215
13
Data on vehicle registrations are available from January 2004 on, but we are unable to distinguish
German brands in the earlier sample, due to the fact that Daimler was also owner and producer of
Chryslers.
13
and Volkswagen. At the opposite end of the spectrum are small luxury car makers such
as Ferrari and Maserati, with average sales of only one car per month. To compare like
with like, our sample does not include small manufacturers with less than 10 vehicle
sales in the total period 2008-2012.
Aggregate car sales slumped after the start of the financial crisis. Annual unit sales
had totalled close to 180,000 before 2007. By 2011, as the Greek economy contracted
at a rapid pace, car sales had fallen to barely 60,000 per annum, a decline by twothirds within four years. Analysing sales trends of cars in Greece is complicated by the
fact that German car manufacturers performed strongly over the last decade. Worldwide, the share of German brands has been rising. This partly reflects the recovery
of Volkswagen sales and the significant decline in Toyota’s market share.14 Figure 5
compares the share of German cars in the Greek car market with that in the European
market overall. The broad trends are very similar.
Figure 6 separately depicts the evolution of total car registrations in the period 20082012 in prefectures with and without a history of German reprisals. When excluding
Attica, the most populous prefecture and home to the capital Athens, affected and
non-affected prefectures show a high degree of co-movement in terms of total vehicles
registered. Total registrations are on a downward trend, particularly steep after mid2009, which seems to stabilize after 2011.
To deal with the fact that many German cars are luxury products – which may have
suffered greater declines in sales as a result of the crisis – we will perform part of our
empirical analysis for the “Volkswagen category” only – a group of brands that focuses
on the production of small family cars (or compact vehicles). This category includes the
following brands: Volkswagen, Opel, Citroen, Ford, Honda, Hyundai, Nissan, Peugeot,
Renault, Seat, Skoda, Toyota.15
14
“VW conquers the world”, The Economist, 7 July 2012, http://www.economist.com/node/
21558269
15
Including other brands to this category (Daewoo, Daihatsu, Isuzu, Kia, Mitsubishi, Subary, Audi)
14
2.4
Control variables and balancedness
Control variables are taken from the 2001 Greek Census and include population, share
employed in agriculture, share employed in industry, share with higher education, share
with secondary education, and unemployment rate. The 2001 controls do not account
for the differential economic reaction of prefectures during the crisis. Unfortunately,
variables like the yearly unemployment rate are not available at the prefecture level for
the years 2011-2012, in which most of the conflict events take place. To mitigate the
crisis effect and to control for prefecture-specific time-varying unobservable factors , we
extend our empirical specification to include prefecture-specific fixed effects (prefectureyear interactions).
Table 3 reports descriptive statistics for the main prefecture-level controls used in
the empirical analysis. Prefectures are generally balanced with respect to observables.
With the exception of a higher share of population with secondary education, the
economic structure and current and historic population does not significantly differ by
reprisal status.
3
Empirical analysis
In this section, we present our main empirical result - the dramatic decline in German
car sales in Greek prefectures affected by German massacres relative to sales in other
areas during the German-Greek crisis after 2010. This result holds under OLS and
when we employ geographical matching. In addition, we show that the results for an
IV exercise that instruments for the prevalence of German reprisals.
does not significantly alter the results.
15
3.1
Baseline results
We begin with a simple example, comparing three neighboring areas. The prefecture of
Achaia, home to the martyred town of Kalavryta, saw 2% of its 1940 population killed
or made homeless by German retribution measures. During months of German-Greek
conflict in the period 2008-2012, it showed an average drop of 1.5 percentage points
in the share of German cars registrations in the Volkswagen category. In contrast,
the neighboring prefectures of Korinthia and Aitoloakarnania - not affected by German
reprisals – experienced increases of 7.6 and 9.6 percentage points respectively.
How general are these differences? As a first step, we perform a difference-indifferences tabulation of the effect of a conflict month on German brand sales. Table
4 presents the basic result. In an average month of German-Greek conflict, prefectures
that witnessed German massacres after 1941 experienced a decline in the share of
German cars of 3.2 percentage points. In contrast, prefectures that had seen no reprisals
experienced a small increase in the share of German cars sold – a plus of 1.1 percentage
points. The overall treatment effect is 4.3 percentage points – a large change in the
market share of German cars within a single month, equivalent to more than a 12%
reduction relative to the pre-crisis level in prefectures with reprisals.
In Figure 7, we plot the density of the difference-in-differences in “normal” months
and months following a month of conflict. We take the monthly average of log(cars)
in reprisal and non-reprisal prefectures, for German and non-German brands in the
Volkswagen category and compute:16
[yg − yng ]r − [yg − yng ]nr
where {g, ng} stand for German and non-German brands and {r, nr} for ‘reprisal’
and ‘non-reprisal’ prefectures respectively. The dotted line represents the density of
16
To account for seasonality, we express these figures as the difference from their value 12 months
earlier.
16
the Diff-in-Diff in months without conflict, and the solid line shows the distribution
during months of public antagonism. There is a clear shift of the distribution to the
left, indicating a relative drop in the sales of German brands in prefectures affected by
reprisals during months of conflict.
More formally, we examine the differential effect of an event month on the market
share of German brands by estimating the following specification:
yjt = α + λt + β1 Ct−1 + β2 Dj + γCt−1 ∗ Dj + Xj δ + jt
(1)
where yjt is the share of vehicles of German manufacturers registered in prefecture
j at time t, λt are year fixed effects, Ct−1 is a dummy for a conflict month,17 Dj is
the share of the prefecture’s 1940 population that lived in towns affected by German
reprisals, and Xj is a vector of prefecture controls. The empirical model amounts to
a difference-in-differences strategy, comparing the share of German brands between
prefectures with and without a past history of German reprisals, in months following a
month of conflict relative to months without a conflict event. The only difference from
classical DID, is that the treatment variable Dj is not a dummy, but a continuous index
proxying for exposure to reprisals. We are interested in the sign and magnitude of the
interaction coefficient γ.
The baseline estimate of (1) is reported in the first column of Table 5, panel A. The
share of German brands seems to increase after months of tension in German-Greek
relations. This is true in all prefectures and could be attributed to the fact that all
car sales react negatively to event months, since these reflect general developments in
the Greek crisis and capture more than just a surge in anti-German feeling. The result
might not be so surprising considering that the sample in Panel A of Table 5 includes
luxury vehicles. If sales of these fall less as a response to crisis events, the German
17
Because of delays between purchasing decisions and car registration, we lag this variable by one
period.
17
share of total sales could increase, as most German cars fall in the luxury category.
Despite the above, there is always a statistically significant difference between
reprisal and non-reprisal prefectures. Prefectures with a higher share of population
affected by reprisals register relatively more German cars in normal times, but not so
in months following a month of conflict. The magnitude of the interaction coefficient γ
implies that a conflict month is correlated with a German car share that is lower by 1.14
percentage points in reprisal prefectures relative to non-reprisal ones. This difference
corresponds to approximately 4.6% of the average market share of German brands in
the period 2008-2012, an economically significant drop, particularly for a durable good
such as cars.
In columns (2) and (3) we include prefecture fixed effects and prefecture-specific
time trends respectively. The interaction coefficient remains significant and increases
in magnitude. In columns (4)-(6), we restrict the sample to the years after the onset of
the Greek sovereign default crisis, in which all our episodes of German-Greek tension
take place. The direction of the results remains unchanged and the coefficients become
slightly larger.
Panel B of Table 5 repeats the exercise for cars in the Volkswagen category. This
again shows a significant negative interaction effect between the share of the population
in reprisal prefectures and months of conflict. In contrast to the above panel, there is
no significantly positive effect of a conflict month on the German market share in nonreprisal prefectures. As before, the results are not affected by controlling for prefecture
or fixed-effects or prefecture-year interactions.
Instead of imposing various assumptions underlying the OLS procedure such as
linearity, we also use geographical matching to examine the effect of German-Greek
conflict. Every prefecture that experienced German massacres is matched to the two
nearest prefectures that did not suffer reprisals. We then take the difference in the share
of German cars in total sales between the affected and the unaffected provinces. By
matching on geographical proximity, we hope to account for unobserved heterogeneity.
18
3.2
Robustness
The German share of registered vehicles is an informative measure, but its movements
are affected by both the denominator and the numerator. In order to capture sales of
German vehicles directly - without imposing the expectation that all other countries’
vehicle sales should move in lockstep - we use a triple-difference specification of the
form:
log(yijt ) =α + λt + β1 Ct−1 + β2 Dj + β3 Gi + γ1 Ct−1 ∗ Dj + γ2 Ct−1 ∗ Gi
+ γ3 Dj ∗ Gi + δ(Ct−1 ∗ Dj ∗ Gi ) + Xj π + ijt
where yijt is the number of vehicles of brand i registered at prefecture j in month
t,18 Dj is the share of the prefectures 1940 population that lived in towns affected by
German reprisals and Gi is a dummy for a German brand. The coefficient δ measures the
effect of a conflict month on the gap between German and non-German cars registered
in reprisal locations relative to non-reprisal ones.
With this specification, we are able to control for unobservable factors that are
constant over time and influence car sales differentially in prefectures with a different
share of population affected by reprisals, as well as for factors that have a differential
influence on German vs non-German car sales. We continue to report specifications
controlling for prefecture-specific fixed effects; any confounding factor would have to be
a brand-prefecture-specific unobservable that changes on months of conflict.
Results are reported in Table 6, for all brands and for the restricted sample of the
Volkswagen category. In both cases, German brands register significantly more cars
on average. They also seem to be less elastic to conflict episodes: following an event
18
In practice, we use log(yijt + 1) as a dependent variable, in order to deal with the problem of zeros
in many prefecture-month-brand cells.
19
month, all car registrations fall, but German brand ones less so. This accounts for the
increase in the German market share after conflict months that was reported in Table
5. What is more important, a higher share of population affected by reprisals leads to a
smaller reaction of car sales to conflict events. Furthermore, this reaction, or decrease
in total car registrations in reprisal prefectures, seems to be accounted for to a larger
extent by a drop in German cars.
According to the triple interaction coefficient estimated in column (1) and accounting for average registrations by prefecture reprisal status, an event month signifies an
average reduction of 335 cars in reprisal prefectures, 85 of which, or approximately 25%,
are German cars. In non-reprisal prefectures only 18% of the total drop - or 179 out of
968 cars - is represented by German brands. The triple difference coefficient decreases
slightly when looking at the years 2010-2012 and more so when restricting the sample
to the Volkswagen category, but it remains always significant and negative.
Despite the fact that conflict episodes are not associated with particular months,
there remains the possibility that event months capture the seasonality of car sales. To
account for any seasonal registration pattern and as a robustness check, we replace the
dependent variable with the residuals of a regression of the log of registered vehicles at
prefecture j in month t on month, prefecture and brand dummies and their interactions.
In this way we allow for a seasonality pattern that varies by prefecture and brand,
though results remain unaffected if we control for seasonality common to all brands
and prefectures or all brands within a prefecture. Results are reported in Table 7 - the
coefficient of interest δ remains practically unaffected.
4
Discussion
The idea that the (collective) recall of memories can act as a powerful determinant of
behavior has a long lineage in psychology, history, and economics. The Roman historian
Tacitus famously argued that the “the principal office of history [is] this: to preserve
the memory of virtuous actions, and to prevent evil words and deeds by instilling the
20
fear of an infamous reputation with posterity.” History as collective memory, in other
words, serves a utilitarian function – to make people more virtuous. The organised
recollection of past events can have other functions as well. An influential school of
thought in cultural history traces its roots to the work of Pierre Nora, who analysed how
lieux de mémoire (“sites of memory”) acquire their significance. Sites of memory in this
approach include physical sites of memory, such as statues, archives, museums, palaces,
plus rituals and commemorative practices (such as feasts and celebrations) and physical
objects (such as sacred texts). Lieux de mémoire bring the past into everyday life by
recalling historical events.19 . As such, sites of memory are by definition ”constructed”,
reflecting the interests of those who perpetuated the celebration of a particular past
event.
In economics, much of the focus has been on limitations of individual memory as a
limiting factor of optimizing behavior. Becker (1993) considered it as such in his Nobel
lecture. Recent work uses imperfect recall as an explanation for various phenomena in
behavioral economics, including over- and underreactions to news.
Theories of limited rationality offer one way to explain our findings. Faced with
sudden conflict, individuals typically search for modes of interpretation that reduce
cognitive dissonance. For example, with limited cognitive resources, not all historical
examples and precedents will be present in everyones consciousness all the time it
takes specific events for them to “come to mind” (Gennaioli and Shleifer, 2010). In the
work by Gennaioli and Shleifer, dormant beliefs and preferences can be activated by
salient events. Once activated, the “mental matrix”, determined by the sum total of
attitudes and beliefs transmitted by parents and friends, contains seemingly convincing
explanations that can be used to shed light on puzzling events: Greeks who feel their
country is suddenly being punished unfairly by austerity-minded German politicians
19
“A lieu de mémoire is any significant entity, whether material or non-material in nature, which by
dint of human will or the work of time has become a symbolic element of the memorial heritage of any
community” (Nora, 1996)
21
remember tales of German atrocities during World War II and related beliefs about the
“German national character”. Such memories can be activated more readily in areas
where ancestors had first-hand experience of massacres such as geographical areas most
exposed to German reprisals.
5
Conclusions
Countries that often went to war with each other in the past still trade less. How does
a history of conflict affect consumer behavior? One possibility is that general cultural
differences make it more likely that countries go to war, and that they trade less with
each other.20 The second possibility is that current events are interpreted through the
lens of past experiences, making insults and transgression seem all the more outrageous.
Our study shows that boycotts can be effective, and that consumer behavior reacts particularly sharply in those areas where contemporary conflict sparks memories
of earlier misdeeds. We examine the case of Greece after the outbreak of the debt crisis
in 2010. Forced to borrow from EU partners, the country had to implement severe
austerity measures which led to massive unrest. Many of the policies implemented in
exchange for the EU bailout packages were blamed on German policies. Public spats
between German and Greek politicians deepened the impression of deeply-rooted antagonisms. The Greek public, when protesting, freely used Nazi-era symbols to express
its outrage about German demands for more spending cuts and the perceived unfairness
of the conditions imposed on Greece.
These events affected consumer behavior in Greece. German car sales suffered in
months of conflict, but in a differential way. Most affected were areas where German
occupying forces during World War II had committed massacres. Prefectures where no
major war crimes had been committed saw much smaller declines in car sales. This
20
Spolaore and Wacziarg (2009) argue that genetically close populations fight each other more often.
This would contradict our intuition if genetic differences translated directly into cultural differences.
22
strongly suggests that public conflict matters for economic behavior in part when it
revives latent hatreds, reflecting an earlier history of conflict. It also helps to rationalize
why empirical evidence on the effectiveness of boycotts has been mixed. Large effects
should mainly be expected where contemporary events trigger memories with earlier
events – as was the case with French automobile sales in China (reviving memories of the
humiliation of China at the hands of Western powers) and with Sino-Japanese conflict
(recalling the war crimes committed by Japan’s armed forces during the invasion of
China in the 1930s).
23
References
Algan, Y. and Cahuc, P. (2010). Inherited Trust and Growth. American Economic
Review, 100 (5), 2060–92.
Ashenfelter, O., Ciccarella, S. and Shatz, H. J. (2007). French Wine and the
U.S. Boycott of 2003: Does Politics Really Affect Commerce? Working Paper 13258,
NBER.
Becker, G. S. (1993). Nobel Lecture: The Economic Way of Looking at Behavior.
Journal of Political Economy, pp. 385–409.
Chaney, E. (2008). Tolerance, Religious Competition and the Rise and Fall of Muslim
Science. mimeo, Harvard University.
Chavis, L. and Leslie, P. (2006). Consumer Boycotts: The Impact of the Iraq War
on French Wine sales in the U.S. Working Paper 11981, NBER.
Doxiadis, C. (1947). Thysies tis Ellados: Aitimata kai Epanorthoseis ston B’ Pagosmio Polemo. Report 19, Hellenic Ministry of Reconstruction, Athens.
Fernández, R. and Fogli, A. (2006). Fertility: The Role of Culture and Family
Experience. Journal of the European Economic Association, 4 (2-3), 552–561.
Fernández-Villaverde, J., Greenwood, J. and Guner, N. (2011). From Shame
to Game in One Hundred Years: The Rise in Premarital Sex and its Destigmitization.
Discussion Paper 8667, CEPR.
Fisman, R., Hamao, Y. and Yongxiang, W. (2012). The Impact of Cultural Aversion on Economic Exchange: Evidence from Shocks to Sino-Japanese Relations, working Paper.
Gennaioli, N. and Shleifer, A. (2010). What Comes to Mind. The Quarterly Journal of Economics, 125 (4), 1399–1433.
Guiso, L., Sapienza, P. and Zingales, L. (2007). Social Capital as Good Culture.
Working Paper 13712, NBER.
—, — and — (2009). Cultural Biases in Economic Exchange? The Quarterly Journal
of Economics, 124 (3), 1095–1131.
Hionidou, V. (2006). Famine and Death in Occupied Greece, 1941-1944. Cambridge:
Cambridge University Press.
Hong, C., Hu, W.-M., Prieger, J. E. and Zhu, D. (2011). French Automobiles
and the Chinese Boycotts of 2008: Politics Really Does Affect Commerce. The B.E.
Journal of Economic Analysis & Policy, 11 (1), 26.
John, A. and Klein, J. (2003). The Boycott Puzzle: Consumer Motivations for
Purchase Sacrifice. Management Science, 49 (9), 1196–1209.
24
Koku, P. S., Akhigbe, A. and Springer, T. M. (1997). The Financial Impact of
Boycotts and Threats of Boycott. Journal of Business Research, 40 (1), 15–20.
Mazower, M. (1995). Inside Hitler’s Greece: The Experience of Occupation, 19411944. New Haven and London: Yale University Press.
Meyer, H. F. (2002). Von Wien nach Kalavryta. Die blutige Spur der 117. JägerDivision durch Serbien und Griechenland. Mannheim: Peleus.
Michaels, G. and Zhi, X. (2010). Freedom Fries. American Economic Journal: Applied Economics, 2 (3), 256–81.
Nessou, A. (2009). Deutsche Besatzungspolitik und Verbrechen gegen die Zivilbevölkerung - eine Beurteilung nach dem Volkerrecht. Göttingen: Vandenhoeck &
Ruprecht.
Nora, P. (1996). From Lieux de Mémoire to Realms of Memory. In P. Nora and
L. D.Kritzman (eds.), Realms of Memory: Rethinking the French Past. Vol. 1: Conflicts and Divisions., XV-XXIV, New York and Chichester: Columbia University
Press.
Nunn, N. and Wantchekon, L. (2011). The Slave Trade and the Origins of Mistrust
in Africa. American Economic Review, 101 (7), 3221–52.
Pandya, S. and Venkatesan, R. (2012). French Roast: International Conflict and
Consumer Boycotts – Evidence from Supermarket Scanner Data.
Teoh, S. H., Welch, I. and Wazzan, C. P. (1999). The Effect of Socially Activist
Investment Policies on the Financial Markets: Evidence from the South African
Boycott. The Journal of Business, 72 (1), 35–89.
Voigtländer, N. and Voth, H.-J. (2012). Persecution Perpetuated: The Medieval
Origins of Anti-Semitic Violence in Nazi Germany. The Quarterly Journal of Economics, 127 (3), 1339–1392.
25
A
Tables and Figures
Figure 1: Occupation zones of Greece mapped to modern-day prefectures
Source: 512 Field Survey Company, Royal Engineers, 1943.
26
Figure 2: Map of towns that suffered reprisals by German occupying forces, 1941-1944
Notes: “Martyred towns” as characterized by presidential decrees no. 399 (1998), 99 (2000), 40 (2004)
and 140 (2005). Population data from 1940 Greek Census.
27
30
20
0
10
Number of articles
40
50
Figure 3: Number of international news articles referring to German-Greek conflict
Jul 2007
Jan 2009
Jul 2010
Jan 2012
Jul 2013
Date
Notes: Number of international newspaper articles in English that mention the words “anti-German”
and “Greece”. Source: Lexis Nexis.
28
100
0
−200
−100
Google search index
200
300
Figure 4: Evolution of popularity of Google search terms related to German-Greek
relations, searched from the Greek web
Jan 2009
Jan 2010
Jan 2011
Jan 2012
Jan 2013
Date
Notes: The blue line is the sum of first differences of the demeaned time series of the Google Trends
search index for the following search terms - in Greek: “Germans”, “German reparations”, “distomo”,
“merkel”, “schäuble”, “rösler”, “troika”, “firings”, “cuts”, “haircut”, “measures”, “memorandum”.
The vertical reference lines indicate the turning points in the series that we identify as event months.
29
Figure 5: Share of German brands, Greece vs Western Europe
Volkswagen category
.35
.3
.3
.2
.2
.25
.25
Share German brands
.35
.4
.4
All brands
2008
2009
2010
2011
2012
2008
Greece
2009
2010
2011
2012
Western Europe
Notes: Western Europe includes Austria, Belgium, Denmark, Finland, France, Greece, Germany,
Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, Sweden, the UK, Iceland, Norway and
Switzerland.
Source: Hel.Stat. and Association Auxiliaire de l’Automobile (AAA).
30
15000
10000
5000
0
Total car registrations
20000
Figure 6: Evolution of total car registrations
Jan 2008
Jan 2009
Jan 2010
Jan 2011
Jan 2012
Date
Non−reprisal prefectures excluding Attica
Reprisal prefectures
Non−reprisal prefectures including Attica
Notes: The figure depicts the total number of passenger vehicle registrations in Greece for the period
2008-2012. Reprisal prefectures are those that have at least one “martyred” town. The dashed blue line
is the time series of car registrations in the non-reprisal prefectures, excluding Attica, the prefecture
in which Athens is situated.
31
0
Density
2
4
Figure 7: Kernel density of difference in differences
−.4
−.2
0
.2
Difference in differences
Months after event month
Other months
Notes: The graph depicts the kernel density of the difference in differences of German vs non-German
car sales in the Volkswagen category in reprisal vs non-reprisal prefectures: [yg − yng ]r − [yg − yng ]nr .
yi∈{g,ng} is the average of log(cars + 1) for German and non-German cars in reprisal and non-reprisal
prefectures, expressed as the difference from its value in the previous year, in order to account for
seasonal sales patterns. The solid line is the average density in months right after an event month.
The dashed line is the average density in all other months. Reprisal prefectures are those that have at
least one “martyred” town.
32
Figure 8: Marginal effect of a conflict month on the share of German brand sales
Notes: Estimated marginal effect and 95% confidence interval from OLS regressions in Table 5,
Column (1) of Panel B.
33
Table 1: Average montly car sales in Greece
Manufacturer
Mean
St. dev.
Min
Max
AUDI
BMW
BENTLEY
CHANG’AN
CHEVROLET
CHRYSLER
CITROEN
DACIA
DAIHATSU
DAIMLER
FERRARI
FIATa
FORD
GENERAL MOTORS
HONDA
HYUNDAI
JAGUAR
JIANGLING
KIA
LADA
LAMBORGHINI
LAND ROVER
LOTUS
MASERATI
MAZDA
MITSUBISHI
NISSAN
OPEL
PEUGEOT
PORSCHE
RENAULT
SAAB
SEAT
SHUANGHUAN AUTO
SKODA
SSANGYONG
SUBARU
SUZUKI
TOYOTA
VOLKSWAGEN
VOLVO
389
455
1
3
230
91
493
41
243
645
1
953
897
25
281
841
7
0
332
10
0
15
0
1
285
221
685
1202
504
19
215
36
423
4
511
10
71
695
1452
1182
136
277
318
1
3
188
94
296
30
232
472
2
488
506
32
232
625
11
0
285
13
1
20
1
1
311
169
444
613
335
19
136
40
308
5
294
13
66
532
888
622
73
99
90
0
0
5
0
87
0
4
51
0
269
216
0
36
133
0
0
33
0
0
0
0
0
2
25
86
384
97
0
65
0
45
0
138
0
0
91
284
220
30
1543
1334
5
15
698
361
1358
99
783
1785
7
2513
2087
129
1025
2524
48
2
1636
52
3
79
2
4
1081
675
1998
2806
1319
67
723
194
1227
18
1439
57
251
1896
3909
2436
319
a
Includes Alfa Romeo and Lancia.
Source: Hel.Stat. Data for the period January 2008 to August 2012.
34
Table 2: Chronology of Greek crisis and turning point identification
Date
Event description
February 2010
Deal with EU/ECB/IMF on bailout and first austerity package
Cover of Focus magazine with title “Cheaters in the Euro-family” displays Aphrodite of Milos making rude gesture
Greek Consumer Association calls consumers to boycott German products
May 2010
Greece announces series of austerity measures
National strikes and large-scale protests in the center of Athens
October 2010
Germany refuses time extension for repayment of Greek loans
Brussels EU summit sees acceptance of German-engineered new bailout
mechanism
Merkel-Sarkozy suggestion that indebted countries are stripped of voting
rights causes angry responses from Greek politicians
November 2010
Troika arrives to Greece in order to ensure implementation of measures
voted in May 2010
January 2011
Case of German reparations for WWII crimes on trial in den Haag
New wave of measures focused on labor market reforms starts in view of
troika’s visit to Greece
May 2011
Troika pushes company privatizations and painful labor reforms
Discussions for new round of austerity measures (Midterm plan)
Talk of Greece leaving the eurozone in the foreign press, initiated by an
article in Spiegel.
Merkel comment on “lazy Southerners” at political rally attracts large
attention in Greek press
September 2011
Eurogroup meeting in Brussels pressures Greece to go through with reforms
Greek government implements new measures including firings and pension cuts
German Finance minister says it is the Greeks’ decision whether they
want to leave the euro, while FDP members suggest a Greek orderly
default
October 2011
New austerity package is voted amidst severe rioting
50% “haircut” of Greek debt takes place
February 2012
Parliament approves new austerity plan
International court rules in favor of Germany in trial regarding WWII
reparations
Greek President Carolos Papoulias declares “I cannot accept Mr Schaeuble insulting my country”
May 2012
Month of Greek national elections
German ministers remind Greece that measures have to be carried
through irrespective of government outcome, if the country wants to remain in the Eurozone
35
Table 3: Summary statistics for Greek prefectures
Variable
All
Nonreprisal
Reprisal
Difference
Population
420542
(1084615)
481133
(1385816)
353823
(442978)
127309
(270851)
Share employed in agriculture
(0.11)
0.26
(0.11)
0.28
(0.10)
0.24
(0.0300)
0.0326
Share employed in industry
0.23
(0.08)
0.21
(0.05)
0.23
(0.07)
-0.0189
(0.0179)
Share higher education
0.11
(0.03)
0.11
(0.02)
0.11
(0.02)
-0.00745
(0.00671)
Share secondary education
0.18
(0.04)
0.17
(0.03)
0.19
(0.03)
-0.0212
(0.00854)**
Unemployment rate
0.12
(0.03)
0.12
(0.02)
0.13
(0.03)
-0.00705
(0.00880)
Population in 1940
146868
(177369)
1.06
(2.17)
1.08
(2.42)
147034
(223671)
146637
(83881)
2.58
(2.57)
2.63
(3.21)
397
(45389.1)
51
30
21
Percentage of 1940 towns
affected
Percentage of 1940 population
in affected towns
N
Source: 2001 and 1940 Greek Census.
36
Table 4: Simple comparison of differences in the average share of German cars in the
VW category
Share of German brands
Conflict months
Other months
Difference
(1)
(2)
(3)
Reprisal prefectures
0.312
0.345
-0.033
Non-reprisal prefectures
0.311
0.300
0.011
Difference
0.002
0.044
-0.042
Sample restricted to 2010-2012. Reprisal prefectures defined as having a share of 1940
population affected by reprisals in the 75th percentile of the distribution of affected
prefectures. The Volkswagen category includes the following brands: Volkswagen,
Opel, Citroen, Ford, Honda, Hyundai, Nissan, Peugeot, Renault, Seat, Skoda, Toyota.
37
Table 5: OLS estimates using shares of German cars
(1)
(2)
(3)
(4)
(5)
(6)
Panel A: All brands
Share pop. affected
0.00541
(0.00422)
0.00546
(0.00546)
Lagged conflict month
0.0181∗∗∗
(0.00622)
0.0182∗∗∗
(0.00628)
0.0192∗∗∗
(0.00684)
0.0181∗∗∗
(0.00620)
0.0186∗∗∗
(0.00638)
0.0192∗∗∗
(0.00686)
Lagged conflict month*
Share pop. affected
-0.00436∗∗∗
(0.00105)
-0.00440∗∗∗
(0.00107)
-0.00458∗∗∗
(0.00159)
-0.00441∗∗∗
(0.00150)
-0.00449∗∗∗
(0.00152)
-0.00458∗∗∗
(0.00160)
Mean share pop. affected in reprisal prefectures
0.0263
Mean share of German cars
0.251
0.251
0.251
0.264
0.264
0.264
Observations
2847
2847
2847
1623
1623
1623
0.0531
0.298
0.381
0.0376
0.274
0.335
R-squared
Panel B: Volkswagen category
Share pop. affected
0.00833
(0.00515)
Lagged conflict month
0.0100
(0.00907)
0.0103
(0.00905)
0.0139
(0.00930)
0.0124
(0.00904)
0.0130
(0.00905)
0.0139
(0.00933)
Lagged conflict month*
Share pop. affected
-0.00374∗∗
(0.00158)
-0.00382∗∗
(0.00159)
-0.00611∗∗∗
(0.00186)
-0.00587∗∗∗
(0.00182)
-0.00597∗∗∗
(0.00182)
-0.00611∗∗∗
(0.00186)
0.278
0.278
0.307
0.307
0.307
Mean share pop. affected in reprisal prefectures
0.0263
Mean share of German cars
0.278
0.0105
(0.00699)
Observations
2837
2837
2837
1615
1615
1615
R-squared
0.101
0.298
0.381
0.0545
0.276
0.343
Controls
Yes
No
No
Yes
No
No
Prefecture fixed effects
No
Yes
Yes
No
Yes
Yes
Prefecture-year fixed effects
No
No
Yes
No
No
Yes
Restricted to 2010-2012
No
No
No
Yes
Yes
Yes
Significance levels: *** p< 0.01, ** p< 0.05, * p< 0.1. Years 2008-2012. Dependent variable is share of German brand passenger vehicles registered in
a prefecture in a month. The Volkswagen category includes the following brands: Volkswagen, Opel, Citroen, Ford, Honda, Hyundai, Nissan, Peugeot,
Renault, Seat, Skoda, Toyota. All regressions include year fixed effects. Prefecture controls include population, share employed in agriculture, share
employed in industry, share with higher education, share with secondary education and unemployment rate, all in 2001. Standard errors are clustered
at the prefecture level.
38
Table 6: Robustness: Triple differences
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
-0.0502∗∗∗
(0.00522)
-0.0502∗∗∗
(0.00522)
-0.0502∗∗∗
(0.00522)
-0.0502∗∗∗
(0.00523)
0.151∗∗∗
(0.0426)
0.156∗∗∗
(0.0426)
0.156∗∗∗
(0.0427)
Panel A: All brands
Share pop. affected
-0.00955
(0.0147)
-0.00498
(0.0121)
-0.0508∗∗∗
(0.00797)
-0.0508∗∗∗
(0.00798)
-0.0451∗∗∗
(0.00475)
German brand
0.176∗∗∗
(0.0433)
0.183∗∗∗
(0.0433)
0.183∗∗∗
(0.0434)
Lagged conflict month*
Share pop. affected
0.00706
(0.00542)
0.00706
(0.00542)
0.00241∗∗
(0.00103)
0.00241∗∗
(0.00103)
0.00197
(0.00135)
0.00197
(0.00135)
0.00197
(0.00135)
0.00197
(0.00135)
Lagged conflict month*
German brand
0.0186∗
(0.0104)
0.0186∗
(0.0104)
0.0155
(0.00996)
0.0155
(0.00996)
0.0423∗∗∗
(0.00843)
0.0423∗∗∗
(0.00844)
0.0423∗∗∗
(0.00844)
0.0423∗∗∗
(0.00844)
German brand*
Share pop. affected
-0.00129
(0.00818)
0.000315
(0.00870)
0.000373
(0.00874)
0.00340
(0.00601)
-0.00310
(0.00799)
-0.00181
(0.00849)
-0.00181
(0.00850)
0.000469
(0.00694)
Lagged conflict month*German brand* -0.00797∗∗∗
Share pop. affected
(0.00250)
-0.00797∗∗∗
(0.00250)
-0.00838∗∗∗
(0.00246)
-0.00838∗∗∗
(0.00246)
-0.00620∗∗∗
(0.00171)
-0.00620∗∗∗
(0.00171)
-0.00620∗∗∗
(0.00171)
-0.00620∗∗∗
(0.00171)
119000
119000
119000
119000
68000
68000
68000
68000
0.194
0.223
0.229
0.626
0.178
0.203
0.205
0.582
-0.0997∗∗∗
(0.0116)
-0.0997∗∗∗
(0.0116)
-0.0997∗∗∗
(0.0116)
-0.0997∗∗∗
(0.0116)
0.639∗∗∗
(0.0471)
0.639∗∗∗
(0.0471)
0.639∗∗∗
(0.0473)
Lagged conflict month
Observations
R-squared
-0.0451∗∗∗
(0.00475)
Panel B: Volkswagen category
Share pop. affected
-0.0251
(0.0289)
Lagged conflict month
-0.120∗∗∗
(0.0159)
-0.120∗∗∗
(0.0159)
-0.111∗∗∗
(0.0117)
German brand
0.571∗∗∗
(0.0433)
0.571∗∗∗
(0.0433)
0.571∗∗∗
(0.0435)
Lagged conflict month*
Share pop. affected
0.0154∗
(0.00810)
0.0154∗
(0.00810)
0.00699∗∗∗
(0.00208)
0.00699∗∗∗
(0.00208)
0.00660∗∗∗
(0.00199)
0.00660∗∗∗
(0.00199)
0.00660∗∗∗
(0.00200)
0.00660∗∗∗
(0.00200)
Lagged conflict month*
German brand
0.0778∗∗∗
(0.0266)
0.0778∗∗∗
(0.0267)
0.0778∗∗∗
(0.0267)
0.0778∗∗∗
(0.0267)
0.00974
(0.0264)
0.00974
(0.0264)
0.00974
(0.0265)
0.00974
(0.0265)
German brand*
Share pop. affected
0.0115
(0.0112)
0.0115
(0.0112)
0.0115
(0.0113)
0.0115
(0.0113)
0.00912
(0.0126)
0.00912
(0.0126)
0.00912
(0.0127)
0.00912
(0.0127)
-0.0146∗∗∗
(0.00412)
-0.0146∗∗∗
(0.00413)
-0.0146∗∗∗
(0.00414)
-0.0146∗∗∗
(0.00414)
-0.0123∗∗
(0.00463)
-0.0123∗∗
(0.00464)
-0.0123∗∗
(0.00465)
-0.0123∗∗
(0.00465)
Observations
34272
34272
34272
34272
19584
19584
19584
19584
R-squared
0.504
0.604
0.616
0.696
0.488
0.574
0.581
0.661
Lagged conflict month*German brand*
Share pop. affected
-0.0168
(0.0249)
-0.111∗∗∗
(0.0117)
Controls
Yes
No
No
No
Yes
No
No
No
Prefecture fixed effects
No
Yes
Yes
Yes
No
Yes
Yes
Yes
Prefecture-year fixed effects
No
No
Yes
Yes
No
No
Yes
Yes
Brand fixed effects
No
No
No
Yes
No
No
No
Yes
Restricted to 2010-2012
No
No
No
No
Yes
Yes
Yes
Yes
Significance levels: *** p< 0.01, ** p< 0.05, * p< 0.1. Years 2008-2012. Dependent variable is log(cars) and observations are brand-prefecture-month cells. The Volkswagen
category includes the following brands: Volkswagen, Opel, Citroen, Ford, Honda, Hyundai, Nissan, Peugeot, Renault, Seat, Skoda, Toyota. All regressions include year fixed
effects. Prefecture controls include population, share employed in agriculture, share employed in industry, share with higher education, share with secondary education and
unemployment rate, all in 2001. Standard errors clustered at the prefecture level.
39
Table 7: Robustness: Removing seasonality
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
-0.00903∗∗∗
(0.00323)
-0.00903∗∗∗
(0.00324)
-0.00903∗∗∗
(0.00324)
-0.00903∗∗∗
(0.00324)
-0.0268∗∗∗
(0.00764)
-0.0280∗∗∗
(0.00761)
-0.0280∗∗∗
(0.00762)
Panel A: All brands
Share pop. affected
-0.000921
(0.000744)
Lagged conflict month
-0.00970
(0.00686)
-0.00970
(0.00687)
-0.00404
(0.00299)
German brand
-0.00216
(0.00147)
-0.00216
(0.00147)
-0.00171
(0.00141)
Lagged conflict month*
Share pop. affected
0.00645
(0.00521)
0.00645
(0.00521)
0.00182∗∗
(0.000881)
0.00182∗∗
(0.000881)
0.00138
(0.000906)
0.00138
(0.000906)
0.00138
(0.000907)
0.00138
(0.000907)
Lagged conflict month*
German brand
0.0151
(0.0103)
0.0151
(0.0103)
0.0120
(0.00990)
0.0120
(0.00990)
0.0382∗∗∗
(0.00852)
0.0382∗∗∗
(0.00852)
0.0382∗∗∗
(0.00853)
0.0382∗∗∗
(0.00853)
German brand*
Share pop. affected
0.00115∗∗∗
(0.000335)
0.00115∗∗∗
(0.000335)
0.00121∗∗∗
(0.000328)
0.00121∗∗∗
(0.000328)
-0.000594
(0.00230)
-0.000939
(0.00226)
-0.000939
(0.00226)
-0.00170
(0.00170)
Lagged conflict month*German brand*
Share pop. affected
-0.00804∗∗∗
(0.00234)
-0.00804∗∗∗
(0.00234)
-0.00848∗∗∗
(0.00230)
-0.00848∗∗∗
(0.00230)
-0.00633∗∗∗
(0.00172)
-0.00633∗∗∗
(0.00172)
-0.00633∗∗∗
(0.00172)
-0.00633∗∗∗
(0.00172)
119000
119000
119000
119000
68000
68000
68000
68000
0.200
0.200
0.223
0.223
0.0866
0.0944
0.104
0.223
-0.0202∗∗
(0.00912)
-0.0202∗∗
(0.00913)
-0.0202∗∗
(0.00916)
-0.0202∗∗
(0.00916)
0.0500∗∗∗
(0.0120)
0.0500∗∗∗
(0.0120)
0.0500∗∗∗
(0.0120)
Observations
R-squared
0.00365
(0.00240)
-0.00404
(0.00299)
Panel B: Volkswagen category
Share pop. affected
-0.00195∗
(0.00113)
Lagged conflict month
-0.0400∗∗∗
(0.0134)
-0.0400∗∗∗
(0.0134)
-0.0309∗∗∗
(0.00883)
German brand
-0.0138∗∗∗
(0.00381)
-0.0138∗∗∗
(0.00381)
-0.0138∗∗∗
(0.00382)
Lagged conflict month*
Share pop. affected
0.0137∗
(0.00792)
0.0137∗
(0.00792)
0.00530∗
(0.00276)
0.00530∗
(0.00276)
0.00491∗
(0.00287)
0.00491∗
(0.00287)
0.00491∗
(0.00288)
0.00491∗
(0.00288)
Lagged conflict month*
German brand
0.0966∗∗∗
(0.0266)
0.0966∗∗∗
(0.0267)
0.0966∗∗∗
(0.0267)
0.0966∗∗∗
(0.0267)
0.0328
(0.0264)
0.0328
(0.0264)
0.0328
(0.0265)
0.0328
(0.0265)
German brand*
Share pop. affected
0.00209∗∗∗
(0.000589)
0.00209∗∗∗
(0.000590)
0.00209∗∗∗
(0.000591)
0.00209∗∗∗
(0.000591)
-0.000255
(0.00252)
-0.000255
(0.00252)
-0.000255
(0.00253)
-0.000255
(0.00253)
Lagged conflict month*German brand*
Share pop. affected
-0.0146∗∗∗
(0.00412)
-0.0146∗∗∗
(0.00413)
-0.0146∗∗∗
(0.00414)
-0.0146∗∗∗
(0.00414)
-0.0123∗∗
(0.00463)
-0.0123∗∗
(0.00464)
-0.0123∗∗
(0.00465)
-0.0123∗∗
(0.00465)
Observations
34272
34272
34272
34272
19584
19584
19584
19584
R-squared
0.359
0.359
0.397
0.397
0.204
0.216
0.241
0.255
0.00639
(0.00398)
-0.0309∗∗∗
(0.00883)
Controls
Yes
No
No
No
Yes
No
No
No
Prefecture fixed effects
No
Yes
Yes
Yes
No
Yes
Yes
Yes
Prefecture-year fixed effects
No
No
Yes
Yes
No
No
Yes
Yes
Brand fixed effects
No
No
No
Yes
No
No
No
Yes
Restricted to 2010-2012
No
No
No
No
Yes
Yes
Yes
Yes
Significance levels: *** p< 0.01, ** p< 0.05, * p< 0.1. The dependent variable is the residual of a regression of log(carsijt ) on month, prefecture and brand dummies and their
interactions. The Volkswagen category includes the following brands: Volkswagen, Opel, Citroen, Ford, Honda, Hyundai, Nissan, Peugeot, Renault, Seat, Skoda, Toyota. All
regressions include year fixed effects. Prefecture controls include population, share employed in agriculture, share employed in industry, share with higher education, share
with secondary education and unemployment rate, all in 2001. Standard errors clustered at the prefecture level.
40
Table 8: Robustness: Excluding Athens and Crete
Excluding Athens
(1)
(2)
Excluding Crete
(3)
(4)
(5)
(6)
Panel A: All brands
Share pop. affected
0.00801∗
(0.00404)
0.0158∗∗∗
(0.00184)
Lagged conflict month
0.0185∗∗∗
(0.00640)
0.0186∗∗∗
(0.00643)
0.0196∗∗∗
(0.00700)
0.0166∗∗
(0.00633)
0.0168∗∗
(0.00637)
0.0191∗∗∗
(0.00691)
Lagged conflict month*
Share pop. affected
-0.00447∗∗∗
(0.00108)
-0.00450∗∗∗
(0.00109)
-0.00464∗∗∗
(0.00160)
-0.00501∗∗∗
(0.00159)
-0.00506∗∗∗
(0.00163)
-0.00726∗∗∗
(0.00143)
Observations
2791
2791
2791
2623
2623
2623
R-squared
0.101
0.293
0.375
0.128
0.295
0.379
Panel B: Volkswagen category
Share pop. affected
0.0108∗∗
(0.00474)
0.0201∗∗∗
(0.00204)
Lagged conflict month
0.0109
(0.00924)
0.0112
(0.00923)
0.0148
(0.00949)
0.00718
(0.00918)
0.00759
(0.00915)
0.0125
(0.00932)
Lagged conflict month*
Share pop. affected
-0.00395∗∗
(0.00160)
-0.00403∗∗
(0.00161)
-0.00625∗∗∗
(0.00186)
-0.00363∗
(0.00208)
-0.00374∗
(0.00212)
-0.00910∗∗∗
(0.00198)
Observations
2781
2781
2781
2613
2613
2613
R-squared
0.135
0.299
0.381
0.160
0.295
0.381
Controls
Yes
No
No
Yes
No
No
Prefecture fixed effects
No
Yes
Yes
No
Yes
Yes
Prefecture-year fixed effects
No
No
Yes
No
No
Yes
Significance levels: *** p< 0.01, ** p< 0.05, * p< 0.1. Years 2008-2012. Dependent variable is share of German brand
passenger vehicles registered in a prefecture in a month. The Volkswagen category includes the following brands: Volkswagen,
Opel, Citroen, Ford, Honda, Hyundai, Nissan, Peugeot, Renault, Seat, Skoda, Toyota. All regressions include year fixed effects.
Prefecture controls include population, share employed in agriculture, share employed in industry, share with higher education,
share with secondary education and unemployment rate, all in 2001. Standard errors are clustered at the prefecture level.
41