Ingroup and Outgroup Biases in the Marketplace:
A Field Experiment during the World Cup
Fernanda L. Lopez de Leony Sang-Hyun Kimz
March 2015
Abstract
This paper estimates the e¤ects of group identity on discrimination in
the marketplace. We conducted an audit study in two electronics markets in
São Paulo, Brazil during the 2014 Brazil World Cup (WC), and recorded the
behaviour of 557 sellers in 922 market interactions. To identify discrimination
patterns and the rationale behind them, we manipulated the salience of
buyers’group membership by making them wear t-shirts of national football
teams, and exploit the outcomes of the WC matches, which arguably
strengthened sellers’ group identity. We detect ingroup market favouritism
(i.e. lower prices) towards buyers wearing the Brazil t-shirt when Brazil had
won a match in the very recent past, and discrimination (i.e. higher prices
We are grateful to Edward Cartwright, David Hugh-Jones, Joo Young Jeon, Charles Noussair, Ernesto Reuben, Judy Rich, Judit Temesvary, seminar participants at University of East
Anglia and Yonsei University, and audience at 3rd Antigua Experimental Economics Conference
for helpful comments. We thank Leonardo Rosa, Silvio Doria, Pamela Felix, Solange Goncalves,
Ilaria Masiero, Robert McDonell, Erika Medina, Mayra Nagase, Paula Thais, Ghitz Sqalli, Arthur
Souto, Fabiana Souza, Francisco Urdinez, Murilo Zacareli and other students at Universidade de
São Paulo and Fundação Getulio Vargas for excellent research assistance. We thank Daniel Monte
and Angelica da Silva Almeida for helping with the advertisement for recruitment. Sang-Hyun
Kim gratefully acknowledges …nancial support from National Research Foundation of Korea Grant
funded by the Korean Government (NRF-2014S1A5A2A03065638).
y
School of Economics. University of Kent. E-mail: [email protected]
z
School of Economics. University of East Anglia. E-mail: [email protected]
1
and lower willingness to trade) against buyers wearing a t-shirt of another
winning team. Our analysis rejects the explanation that these di¤erential
treatments were always motivated by economic pro…ts. Instead, the results
indicate taste-based discrimination (Becker, 1957) and shed light on the ways
in which ingroup and outgroup biases occur in market outcomes.
JEL Classi…cation: C93, D71, J15
Keywords:
ingroup and outgroup discrimination, bargaining in the
marketplace
1
Introduction
Group identity is of key importance in understanding social behaviour. This
is not surprising as we tend to like people who are similar to us (or those who are
in the same group) more than those who are di¤erent from us (or from another
group). The consequences of these preferences have been documented in games in
the laboratory and in the …eld (e.g. Chen and Li 2009, Eckel and Grossman 2005,
Goette et al. 2006) as well as in real-life situations, such as charitable giving (Fong
and Luttmer 2009) and judicial choices (Shayo and Zussman 2010).
However, despite the substantial body of literature demonstrating the
importance of group preferences in non-market interactions, the evidence that such
biases play a role in the marketplace has thus far been restricted to the lab (Ball
et al. 2001 and Li et al. 2011). In contrast, recent evidence from the …eld suggests
that the discrimination in the market is largely motivated by economic pro…ts, not
by a distaste against people from other groups (Castillo et al. 2013, Ewens et al.
2014, Gneezy et al. 2012, List 2004, Zussman 2013).
The contrast made by these two strands of study mirrors a larger debate within
the literature that has developed around two explanations for the existence of
discrimination in the market (for reviews, see Gneezy et al. 2012, Guryan and
Charles 2013 and Rich 2014). The …rst one, put forward by Gary Becker in his
work "The Economics of Discrimination" (1957), simply regards discrimination as an
outcome of agents’preferences and prejudices (consistently with the group preference
hypothesis). The most recognised competing theory, proposed by Phelps (1972), is
2
a statistical discrimination model where purely rational agents treat individuals
di¤erently due to incomplete information. According to this theory, it is rational to
discriminate between buyers from di¤erent groups because these memberships (e.g.
nationality) signal unknown pro…t-related characteristics of the buyers (e.g. income
or willingness to pay).
This paper sheds new light on this discussion by proposing a novel experiment
to understand whether and how ingroup and outgroup preferences a¤ect bargaining
outcomes in real markets. We conducted an audit study in informal electronics
markets in São Paulo, Brazil during the 2014 Brazil World Cup championship
(WC). In these markets, the prices are very often determined by bargaining. We
employed Brazilian (ingroup) and non-Brazilian (outgroup) research assistants to
negotiate with local sellers, following a …xed bargaining script similar to the one in
Castillo et al. (2013). They enquired about relatively inexpensive and homogeneous
electronic devices (Playstation game, headphones or memory stick). In total, we
recorded the behaviour of 557 sellers in 922 market interactions before and during
the championship. We analyse the prices o¤ered by sellers, the likelihood of a seller
accepting the buyer’s price and measures of sellers’willingness to trade.
Although comparing the prices charged to buyers of di¤erent nationalities (e.g.
Brazilians and non-Brazilians) could provide evidence of discrimination based on
group membership, it would be di¢ cult to identify the rationale behind such
discrimination (economic pro…t or personal prejudice). To identify discrimination
motivated by sellers’ preferences, we manipulated the salience of buyers’ group
membership –based on the support for Brazil or another team –by making them
wear t-shirts of national football teams. Brazilian buyers wore the Brazil t-shirt, and
non-Brazilian buyers wore a t-shirt of one of the following teams: Chile, Croatia,
Mexico or Spain. We randomized the group of sellers to whom the buyers’ team
t-shirts were visible (the treatment) or hidden by a jacket (the control) during the
negotiation. This allowed us to quantify discrimination in a very controlled way by
making within-buyer comparisons.
The 2014 WC was an ideal place to investigate group discrimination in the
market, as it provided a great opportunity to prime buyers’identity to sellers. Brazil
3
is very well-known for its passion about football, which peaked during the WC as
the country hosted the championship.1 This made the connection between football
t-shirts and the categorization of buyers into ingroup or outgroup natural to the
market environment.
Our test relies upon how the e¤ect of the football t-shirts varied over the course of
the WC. We make use of the fact that the outcomes of matches and the rivalry status
against Brazil induced exogenous ‡uctuations in sellers’emotion towards buyers with
speci…c t-shirts. In this regard, we follow Edmans et al. (2007), who conclude that
the news of their own country’s defeat in the WC a¤ected investors’ mood and
thus …nancial market outcomes. We conjectured that a victory for Brazil would
intensify the feeling of national pride among sellers and hence ingroup bias. We
test this hypothesis by looking at the impact of the Brazil t-shirt on the bargaining
outcomes before and after Brazil won a match. We also test for outgroup bias by
examining whether non-Brazilian buyers were more discriminated against when they
wearing a t-shirt of an imminent rival of Brazil (e.g. when buyers wore the Croatia
t-shirt one day before the Brazil vs. Croatia game) than in another situation (e.g.
the Croatia t-shirt in another situation).2
In between- and within-buyer comparisons, we …nd no notable impacts of team
t-shirts on market outcomes, either when worn by Brazilians or by non-Brazilians
buyers; however, we do detect statistically and economically signi…cant changes in
the e¤ects of t-shirts in a di¤erence-in-di¤erences analysis. When Brazil had just
won a WC game, sellers charged signi…cantly lower prices to and were more likely to
accept o¤ers from Brazilian buyers wearing the Brazil team t-shirt. We test whether
income-related consideration was the main motive of this di¤erential treatment, and
reject this explanation.3
1
2
"World Cup Survival Guide", New York Times, 2014.
Observing reactions to shocks has as advantage of being more revealing than post-experiment
surveys where respondents might be answering in a politically correct manner. Surveys are also
used to identify reasons for discrimination (Oreopoulos 2011, Zussman 2013).
3
This does not mean that we prove the absence of statistical discrimination, but rather that
the particular discrimination pattern observed in our data cannot be explained by statistical
discrimination theory.
4
We also …nd evidence of outgroup discrimination, but in an unexpected way. In
particular, sellers did not signi…cantly discriminate against buyers who were wearing
the t-shirt of Brazil’s rival, but they did discriminate against those who were wearing
the t-shirt of another team which had recently enjoyed a victory (against another
team than Brazil). After ruling out some possible explanations, we conclude that
the most plausible account for this outgroup discrimination is that it was motivated
by envy towards another successful group. That is, reminding the sellers of another
team’s victory (by the t-shirt) probably made the Brazilian sellers feel less victorious.
As mentioned above, our results add to a surprisingly scarce literature on
marketplace discrimination based on group preferences. The closest studies to this
paper are Zussman (2013) and Li et al. (2011). Di¤erently from this current article,
they focus on investigating discrimination based on the likelihood of transaction
among agents from di¤erent groups, rather than bargaining outcomes.4 They …nd
mixed results.
Zussman (2013) investigated discrimination in an online market experiment in
Israel. He sent enquiring emails to sellers, manipulating buyers’names to be Jewish
or Arabic. He found that sellers were more likely to reply to buyers with a Jewish
name (Arabs are the minority and thus more likely to be outgroup members). To
explain the source of discrimination (statistical or taste-based), he correlated the
discriminatory behaviour with attitudes in a post-experiment survey. The only
signi…cant association was found for the statement "The Arabs in Israel are more
likely to cheat than Jews", from which Zussman concluded that the discrimination
is based on a transaction-related characteristic (i.e. Arabs in Israel are believed to
be less trustworthy) and statistical reasoning.
On the other hand, Li et al.
(2011) did …nd e¤ects of ingroup bias on
discrimination in a laboratory market experiment. They assigned subjects to the role
of seller or buyer and the membership to groups, and made them interact repeatedly
4
The results in Zussman (2013) and Li et al. (2011) speak more to the homophily literature.
‘Homophily’refers to the empirical fact that people tend to interact with similar others. Our paper
does not address this issue because, as will become clear later, matching of a buyer and a seller was
not a choice for either of them. For a careful discussion on homophily, see Currarini and Mengel
(2013).
5
in a small market (of three buyers and three sellers). They found that sellers were
more likely to make o¤ers to ingroup buyers, and that buyers were more likely to
accept o¤ers from ingroup sellers.
This paper proceeds as follows. In Section 2, we state our hypotheses and explain
our experimental design in more detail. Section 3 explains the markets and the
experimental procedure. The results are presented and discussed in Section 4. Then,
we conclude.
2
Identi…cation of Ingroup and Outgroup Biases
in the Marketplace and Hypotheses
Following a widely used approach in social psychology and recently in
economics as well (see, e.g. Benjamin et al. 2010, Chen et al. 2014, Perdue et
al. 1990 and Shih et al. 1999), we quantify ingroup and outgroup biases by priming
buyers’group identity to sellers. The priming technique involves the activation of
social representations by exposing subjects to some information reminding them of
their or other’s social identity and the application of these activated representations
in social behaviour (Bargh 2014 and Molden 2014). In a classical study, Perdue et
al. (1990) conducted experiments mixing the pronouns "us" and "them" (referring
to ingroup and outgroup status) with other syllables so as to understand how these
pronouns a¤ect the way information is processed. Chen et al. (2014) conducted an
experiment among Asians and Caucasians, exploiting their natural identities (their
ethnicity or their school a¢ liation) to identify e¤ects of ingroup and outgroup bias
on coordination and cooperation games. They primed their ethnic identity using a
pre-experiment questionnaire on family and cultural background.
By making buyers wear t-shirts of national football teams, we highlighted buyers’
association with sellers: that buyers are rooting for Brazil ("us") or for another
national group ("them").5 Given their high likelihood of playing against Brazil,
5
Researches in the social psychology of sports document that individuals report more positive
evaluations of other ingroup fans than of rival outgroup fans (Wann and Grieve 2005, Wann et
al. 2001), and show more pro-social behavior towards fans of the same team (Platow et al. 1999).
6
four teams –Chile, Croatia, Mexico and Spain –were chosen to be the outgroups.
As will be detailed in Section 3, we randomized whether our buyers were to wear
a football t-shirt (the treatment) or a jacket hiding the t-shirt (the control). To
control for income-related factors, we provided jackets that were priced similarly to
the o¢ cial team t-shirts which were worth approximately 100 US dollars. Moreover,
buyers followed a …xed script of negotiation. Because our design allows withinbuyer comparisons on how they were treated in the market, we avoid the most
important criticism related to audit studies (Heckman 1998, Guryan and Charles
2013), ensuring that the treatment and control buyers are matched on all relevant
attributes.
During the WC, football team t-shirts were a very popular item in Brazil, and
wearing one was a way to demonstrate one’s national pride and support for the
country. Stores predicted a 500% increase in the demand for the o¢ cial t-shirt of
the Brazilian national team as well as a large increase in the demand for the t-shirts
of non-Brazilian teams due to the large in‡ux of tourists.6
For these reasons, we believe that the football t-shirts did prime the sense of
group identity of sellers and hence stimulate their ingroup and outgroup biases
as well. An alternative interpretation of a team t-shirt (as opposed to a national
symbol) is that it reveals the buyer’s taste for football. But during the 2014 WC,
everyone in Brazil was a football fan.7
Thus, we hypothesize that Brazilian buyers will be more favourably treated when
wearing the Brazil t-shirt and that non-Brazilian buyers will be more discriminated
against when wearing a non-Brazil team t-shirt. However, these simple hypotheses
Others studied the e¤ects of priming national symbols. For example, Hassin et al. (2007) conducted
experiments exposing subjects to their national ‡ag, and …nd that it increased the "sense of unity"
among the participants, drawing them to the political center.
6
About the prediction, see http://jcrs.uol.com.br/mob/noticia.php?codn=164715. Also, during
the WC there was an estimate of one million tourists in Brazil.
7
As described in a New York Times article: "Across the country, schools public and private have
scheduled midyear vacations so they could start before the tournament’s inaugural game. Banks
are set to close early, slashing their hours almost in half. Hospitals must stay open, but doctors
have been routinely rescheduling nonemergency appointments so they have no commitments when
Brazil is playing." ("Making Holidays of Brazil’s World Cup Games", 2014.)
7
do not lead us to a de…nitive conclusion regarding the mechanism behind the
discrimination because the nationality demonstrated by the t-shirt may also signal
the buyer’s willingness to pay. To isolate the group biases from income confounders
(in line with statistical discrimination), we compare the e¤ect of team t-shirts before
and after Brazil’s matches (in a di¤erence-in-di¤erences approach). Since we visited
the markets over a relatively short period of time (from May 30 to July 7, 2014),
most factors a¤ecting sellers’ perception of buyers’ economic status (e.g., GDP of
the country and the cost of visiting Brazil) stayed constant during the period.
Because the most signi…cant event during the period was the WC football games,
which greatly stimulated nationalistic emotions, it would be safe to assume that
any detected changes in the e¤ect of a t-shirt over the WC are attributable to the
outcomes of the games making sellers’sense of national identity more or less salient.
Although nationality, which in our case determines group membership, is mostly
determined by birth and thus is di¢ cult to change, the salience of national identity
(or nationalism) is time-varying and depends on social context. For example, in
Benabou and Tirole (2011), one’s expressed identity is a reminder of his/her true but
easily forgettable preference. On the other hand, in Akerlof and Kranton (2010), by
choosing one’s identity, an individual chooses to follow social norms associated with
the identity. Similarly, we assume that national identity and associated preference or
social norms can be reminded or strengthened by the sports event and the t-shirts.
The following are the hypotheses:
H1 Ingroup bias: The e¤ect of the Brazil t-shirt on market favouritism will
increase just after Brazil wins a match; the Brazil t-shirt will be more e¤ective in
making sellers charge lower prices and accept buyers’o¤er.
H2 Outgroup bias: The e¤ect of a non-Brazil t-shirt on discrimination will
increase when the respective team is about to play against Brazil; the non-Brazil
t-shirt will be more e¤ective in making sellers charge higher prices and not accept
buyers’o¤er.
To test these hypotheses, we collected data on nine days, before and during the
WC. We visited the markets one day before and two days after the following three
matches: Brazil vs. Croatia (June 12), Brazil vs. Mexico (June 17) and Brazil vs.
8
Chile (June 28). We also collected data two weeks before the …rst game of the WC
(May 30 and 31) and one day before the Brazil vs. Germany match (July 8). The
dates of our visits, the set of t-shirts worn on each visit, and the score for each match
are shown in Figure 1. On all nine days, we had Brazilian buyers wearing the o¢ cial
Brazil t-shirt, and on some days they also wore cheap non-o¢ cial t-shirts (before
and after Croatia and Chile games).8
F igure1
To test the rivalry and outgroup bias hypothesis, we made non-Brazilians buyers
wear the t-shit of Brazil’s competing team before and after the …rst three matches.
To conduct our test for ingroup bias, we observe sellers’reaction to the Brazil t-shirt
just after Brazil won a match (after the Croatia and Chile games).9 To disentangle
an ingroup bias from some favoritism towards a “winner”, we also had non-Brazilian
buyers wearing a t-shirt of a team that had just won (e.g. Chile on June 14) or had
just been defeated (Spain on June 20 and Mexico on June 30) against a third team
that was not Brazil.
3
Data
3.1
Markets, Buyers and Products
The experiment was conducted in two main electronics markets in São Paulo.
The …rst market, Mercado Santa E…ginia, has more than 1,300 stores on seven
streets (Santa I…gênia, Aurora, Vitória, dos Andradas, Timbiras, General Osorio
and Gusmoes).10 The second one, Mercado Paulista, consists of two malls located
at Avenida Paulista, which are in a more cosmopolitan area of the city. The …rst
mall (Market Paulista) has 146 stores, and the second (Stad Center) 204 stores.
These markets play an important role in the commerce of São Paulo, and their
8
9
The rationale for this additional treatment will be explained in Section 4.
Since our ingroup bias test relies on a victory for Brazil, and given our …nancial constraints,
we did not collect data for any of the matches in which Brazil was defeated.
10
http://www.comprasnasantai…genia.com.br/historia_da_santai…genia.asp
9
prices are very often determined by bargaining.
Our research assistants were sent to the market in pairs. Each person was
assigned a role as a buyer or a buyer’s friend. There were two types of pairs: a
Brazilian pair consisting of a Brazilian buyer and a Brazilian friend and a nonBrazilian pair consisting of a non-Brazilian buyer and a Brazilian friend. The
experimental manipulation was only on the buyer, and he/she was the one enquiring
about prices and conducting the negotiation. The role of the friend was to observe
the buyer and protect him/her from any possible threat or violence since, many
times, non-Brazilian buyers were wearing the shirts of a team playing against Brazil.
Also, due to the possibility of pickpocketing and other crimes, it is very unusual that
a tourist would go to Mercado Santa E…ginia on his/her own. It is more natural for
a local to bring his/her foreign friend to this market to …nd a bargain.
In total, we had 41 pairs, combinations of 22 Brazilians and 13 non-Brazilians.11
Brazilians were undergraduate and graduate students at Universidade de São Paulo
or at Fundação Getulio Vargas. Non-Brazilian participants were exchange students
at the same schools, who had been in Brazil for at least one semester. All of
them spoke some Portuguese. They were recruited from an internship program
in these institutions or via advertisements on Facebook pages of the students’
organizations.12
Buyers enquired for homogeneous electronic products to minimize any potential
confounding factors related to product characteristics. The products were: Sony
headphones XZ 300 and XZ 100, Sandisk memory card 32 GB and 8 GB, and 2014
FIFA World Cup game for Playstation. Pictures of the products are included in the
Appendix. In selecting these products, we follow two main criteria: the product had
to be available in most of the stores (this was noticed during the pilot) and be a
11
Non-Brazilians were from the following countries: Argentina, Canada, Colombia, England,
France, Italy, United States and Venezuela.
12
During the recruitment phase, all participants responded a questionnaire in which they were
asked whether they would feel uncomfortable in shopping wearing speci…c team t-shirts or shopping
with someone who is wearing these shirts. We did not allocate anyone to a condition in which
he/she would feel uncomfortable. To avoid experimenter biases, research assistants were informed
neither of the research questions nor the expected results of this study.
10
product that a tourist could be interested in buying during a short trip abroad for
the WC.
3.2
Outcomes
Each pair was in the market for about 2 hours each day. They were assigned a
route, and visited some stores on the route, such as those of whose addresses ended
with an odd number.13 To increase the chances of collecting bargaining outcome
data, the pair was instructed to look for the selected products in the store window
and to enquire for whichever of those products was most likely to be available.
In each store, the pair followed a …xed script very similar to the one proposed by
Castillo et al.(2013),14 starting with the buyer asking "Do you sell product X?" In
the case that the product was available (77%), the buyer asked "How much does
product X cost if I pay in cash?"
After being informed of the product’s availability, the buyer asked "How much
does product X cost if I pay in cash?" After the seller gave the quote, the buyer
insisted on the pre-determined maximum acceptable price.15 Buyers repeated this
response twice in case the seller made any countero¤er. Then, the buyer terminated
the negotiation, the buyer’friend asked for the store’s business card, and the pair
left the store. In the case that the seller accepted the o¤er, the buyer said he/she
would make a phone call to con…rm the purchase or get some cash from an ATM.16
In the case that the product was not available (23%), we recorded whether the seller
13
We never sent two pairs of buyers to the same route in the same day, and a pair of buyer has
never repeated the same route.
14
Castillo et al. (2013) investigate Peruvian taxi drivers’patterns of discrimination regarding the
gender of passengers. The authors made their "taxi customers" negotiate with taxi drivers, using
a …xed-o¤er bargaining script that we follow closely in this study. Castillo et al. …nd evidence of
discrimination against male passengers, which they conclude is driven by statistical discrimination.
15
We established a low but reasonable maximum acceptable price for each product by choosing
the third lowest price in an online Brazilian market (http://www.buscape.com.br/eletronicos.html).
More speci…cally, the prices that we used were: 115 reais for Sony headphones XZ 300, 78 reais for
Sony headphones XZ 100, 52 reais for Sandisk memory card 32 GB, 18 reais for Sandisk memory
card 8 GB, and 140 reais for FIFA World Cup game.
16
It is a common practice to enquire about prices and not to buy.
11
o¤ered a similar product in his own store or if he recommended looking for the
product in a speci…c store in the neighborhood.
In addition to memorizing this information, each participant was asked to report
on a number of details about the store, the seller and how the buyer was treated
during the visit. Immediately after leaving the store, they made a phone call to one
of our interviewers who read the questions and recorded the outcomes. The buyer
and the friend were instructed to speak on the phone separately from each other in
order to avoid a¤ecting each other’s answers. Since the t-shirt could have a¤ected
the buyer’s attitude, we asked the buyer’s friend to evaluate the buyer’s level of
con…dence during the negotiation. In the Appendix, we present regression results
showing that the t-shirt and the interaction between the t-shirt and the WC outcome
are very largely uncorrelated with the buyers’attitude. Our main regression results
do not change if we include this assessment as an explanatory variable.
An obvious concern was to keep the experiment anonymous to sellers. It was rare
that a buyer visited the same store more than once (only 11 cases or 2.44% of the
total interactions), and only 21.7% (10.4%) of the sellers were asked for the same
product more than once (twice) over the whole experiment. We believe that the
phone calls were not suspicious as it is common that people discuss before making
a purchase.
3.3
Randomization
In Mercado Santa E…ginia, each pair was assigned to a street route that had 6
to 8 blocks, and they visited 2 to 4 stores on each block. Every time they were about
to move on to the next block, the buyer made a phone call, and the interviewer rolled
a dice to determine the treatment allocation: whether to wear a jacket to hide the
team t-shirt or show the t-shirt by not wearing the jacket. (We chose to conduct
the randomization at the level of the block, instead of at the level of the store, to
avoid contamination or the situation where a seller sees the buyer’s shirt in the
street, when they are assigned to the "jacket condition").17 In Mercado Paulista, we
17
On some days, Brazilian buyers had to wear a cheap uno¢ cial Brazil t-shirt as well as the
expensive o¢ cial Brazil t-shirt. On these occasions, in the …rst half of a day, they would wear one
12
follow a similar procedure. We explored the internal logistics of the malls to identify
clusters of stores that were separated by walls limiting sellers’vision. Randomization
was conducted at this level.
In total, we collected 922 interactions of buyers with 557 sellers. Table 1 shows
the number of negotiations by market, pair and treatment. In the Appendix, we
show evidence that visits were randomly assigned across stores. Mostly, we do
not …nd statistically signi…cant correlations between treatment conditions and store
characteristics or enquired product.
T able1
4
Results
We begin by documenting how the outcomes – sellers’ proposed …rst price,
second price and the likelihood of a seller accepting the buyer’s price – varied
depending on a buyer’s nationality and the t-shirt he/she was wearing.
The
…rst price quote is probably more sensitive to sellers’ prejudice as no information
regarding the buyer’s willingness to pay has been revealed. The probability of
accepting the buyers’o¤er, on the other hand, should disclose more about the sellers’
willingness to trade with a speci…c type of buyers. Although buyers visited stores
922 times, price quotes were given only in 697 cases due to the availability of the
products.
The main dependent variables in our analysis are the …rst and second price quotes
given by sellers in their logarithmic form and an indicator for whether the buyer’s
asked price was accepted by the seller. In the odd columns in Table 2, we present
results from estimating Equation (1).
Y =
+
BrazilianBuyer +
1
+
2
+
3
+
4
+
5
+"
(1)
where Y is an outcome variable, BrazilianBuyer is an indicator of buyers’
of the two Brazil shirts, change their clothes during a short break, and continue the experiment in
the second half. (The shirt order was randomized.)
13
Brazilian nationality and " is a random term. We included product ( 1 ), street and
market-…xed e¤ects ( 2 ). To account for demand e¤ects, we also added dummies for
day ( 3 ) and ( 4 ) are time of the visit …xed e¤ects (seven indicators for a onehour block, from 10-11 am to 4-5 pm). In the regression in Table 2, we have
buyer-pair random e¤ects (as we intend to show the e¤ect of buyers’ nationality
on discrimination), represented by
5.
(For the remaining Tables, we include buyer-
pair …xed e¤ects, instead.) The numbers in parenthesis in the Tables are robust
standard errors clustered at the store address level.18
T able2
Table 2 shows that non-Brazilians were given initial price quotes about 6.3%
higher than Brazilian buyers [column (1)]. In 103 interactions (20.7%), sellers stated
prices below the maximum acceptable price, and in 42 cases (6%) they refused to
make any countero¤er. In the remaining 553 cases (73.3%), sellers made second price
o¤ers. For these quotes, the price di¤erence across buyer’s nationality declined, but
did not disappear. On average, foreign buyers were o¤ered 5.4% higher second prices
in comparison to Brazilian buyers [column (3)]. This might be because non-Brazilian
buyers are perceived to have higher income and willingness to pay, or this might be
driven by some ingroup bias. We do not …nd a statistically signi…cant correlation (pvalue>0.19) between the likelihood of sellers’acceptance and the buyer’s nationality
[column (5)].
In columns (2), (4) and (6) in Table 2, we investigate the e¤ect of national identity
further by introducing the t-shirt treatment (…ve t-shirt dummies and excluding the
“jacket” category). Table 2 shows only a few statistically signi…cant correlations
between the market outcomes and the t-shirts. In regressions not included in the
paper, we tested whether the mismatch between buyers’mother language and the
football shirt (e.g. a Canadian buyer wearing a t-shirt of a Spanish-speaking country
like Chile) a¤ected the bargaining outcomes. Arguably, a mismatch is a stronger
indication of support for a team competing against Brazil as it involves an active
18
All the results in the paper are also visible in speci…cations, including only product and day
of the visit …xed e¤ects. We do not show these estimates for the sake of space.
14
choice. But again, we found no e¤ect.
In summary, we do not …nd supporting evidence that Brazilian buyers were
more favourably treated when wearing the Brazil t-shirt nor that non-Brazilian
buyers were discriminated against more severely when wearing a non-Brazil team
t-shirt. This may be because the t-shirt treatment was not strong enough to remind
sellers of their national identity or because there were other confounding issues in
the treatment as articulated in the previous section. Equally, it may be because the
e¤ectiveness of the treatment varied over the WC.
Thus, we further analyse the pattern of the discrimination by utilizing a
di¤erence-in-di¤erences method. In particular, we test the hypothesis that the e¤ect
of wearing a Brazil t-shirt increases when Brazil has won a game in the recent past
(H1). We estimate Equation (2), and focus only on the sample of Brazilian buyers.19
Y = +
1 Braziltshirt+
2 Braziltshirt
Brazilwon+
1+ 2+ 3+ 4+ 6+
where Braziltshirt is an indicator for whether the Brazil t-shirt was visible to
the seller, Brazilwon is an indicator for the dates following a Brazil victory (June
14 and June 30),
6
are buyer pair …xed e¤ects and
present the results for coe¢ cients
1
and
is the error. In Table 3, we
2.
T able3
The estimates in Table 3 indicate that the Brazil t-shirt e¤ect in determining
market favouritism increases just after Brazil wins a match. The …rst price given to
buyers wearing the Brazil t-shirt was about 14.5% lower than that charged to the
same buyers wearing the same t-shirt in another occasion. However, this di¤erence
is not statistically signi…cant [column (1)]. On the other hand, sellers turned out
to be signi…cantly more generous in their second price quote: the di¤erence in the
price was approximately 30% as shown in column (3) to a Brazil t-shirt buyer after
19
In Tables 3, 4 and 6, we present results separately for Brazilian and non-Brazilian buyers. All
…ndings also hold when we estimate the e¤ects using the whole sample. These are presented in
Table A3 in the Appendix.
15
a victory compared to the same buyer wearing a Brazil t-shirt at some other point.
Sellers also became 33.5% more likely to accept these buyers’price o¤ers when Brazil
had recently won.
A concern in interpreting this di¤erential treatment as a consequence of ingroup
bias is that the Brazil victory might also result in an increase in the number of buyers
wearing Brazil t-shirts, in general. Consider the following scenario: sellers give price
quotes according to their rational expectation of the buyers’income, whereas many
buyers emotionally responding to Brazil’s victory purchase the expensive o¢ cial tshirt after a winning match. If this were the case, the team t-shirt would carry less
information about the buyer’s wealth after the victory than before. This might be
the reason why sellers charge lower prices to buyers wearing the football t-shirt just
after a Brazil win. To account for this possibility, we introduced an extra treatment.
In addition to the o¢ cial t-shirt worth about 100 US dollars (229 reais), our buyers
also wore a cheap uno¢ cial t-shirt worth approximately 7 US dollars (15 reais).
Because the cheap t-shirt was not signaling the high income of the buyer in the
…rst place, the above argument does not apply to this t-shirt. So, we estimate a
variation of (2) decomposing the football t-shirts into these two categories: o¢ cial
and uno¢ cial. If income inferences were driving the di¤erential treatment in Table
3, only the coe¢ cient of (o¢ cial t-shirt X victory), not that of (uno¢ cial t-shirt X
victory), should have been statistically di¤erent from zero.
The results refute this scenario. Column (4) in Table 3 shows that both buyers
–with the o¢ cial and the uno¢ cial t-shirt –were charged a lower second price after
Brazil had won a game. In addition, column (6) shows that wearing a Brazil t-shirt
after a Brazil victory had a signi…cant impact on the probability of sellers accepting
the buyer’s insisted price. This is mostly driven by buyers wearing the inexpensive
Brazil t-shirt. Compared with other occasions, after a Brazil victory, sellers were
24.6% more likely to accept an o¤er of a buyer wearing the o¢ cial t-shirt and 51%
more likely to accept the o¤er from a buyer wearing the uno¢ cial t-shirt.
We test whether this market favouritism is due to a "winner favouritism" rather
than an ingroup bias, by turning our attention to how outgroup buyers were treated.
In Table 4, we present results replicating the analysis in Table 3, but restricting the
16
sample to non-Brazilian buyers. We …nd that, in contrast to results in Table 3, the
supporters of a winning team other than Brazil (Chile) were charged signi…cantly
higher prices by Brazilian sellers than when wearing the same t-shirt but on another
occasion. Column (2) shows that the prices initially o¤ered to buyers wearing the
victorious Chilean t-shirt were 21% higher than the prices given to buyers with a
non-victorious Chilean t-shirt. We do not …nd statistically signi…cant discrimination
in the second price. On the other hand, sellers were 56% less likely to agree to trade
at the o¤ered price [column (6)]. The asymmetric e¤ects detected in Tables 3 and 4
reject the hypothesis that the sellers might just favour the supporters of a winning
team because they wanted to identify themselves with the successful group to enjoy
the feeling of winning or because they believed that supporters of a victorious team
are tougher negotiators. The results so far raise a new question: why did Brazilian
sellers discriminate against supporters of a victorious non-Brazilian team?
T able4
One possible explanation for the results in Table 4 is that sellers discriminated
against these buyers to balance out the loss of pro…ts incurred by charging lower
prices to victorious Brazil t-shirt buyers (e.g. because victorious buyers are generally
happier, they might accept paying higher prices). To examine this possibility, we
explore the cases in which sellers did not have the enquired product. Recall that
we recorded whether sellers o¤ered another similar product in their own store. The
results are described in Table 5 for the entire sample [column (1)], the sample of
Brazilian pairs [column (2)] and the sample of non-Brazilian buyers [column (3)].
Column (3) shows that sellers were 46% less likely to o¤er a similar product to
buyers wearing winning non-Brazil team shirt, indicating that to some extent they
were willing to sacri…ce earnings in order to shun these buyers. We do not …nd
similar evidence for Brazilian buyers [column (2)]. We do not detect a statistically
signi…cant e¤ect of a victorious t-shirt in any sample when the dependent variable is
an indicator of whether the seller recommended the buyer look for another speci…c
store [columns (4) to (6)].
17
T able5
The di¤erence-in-di¤erences analysis so far shows that the observed
discriminations are taste-driven by ingroup and outgroup bias.
Discrimination
against supporters of a non-Brazilian winning team (Table 4) might arise because
the very existence of another winner made Brazil one of the many winners, making
Brazil’s victory less enjoyable. Another possibility is that Brazilian sellers might
dislike the supporters of another winner since the winner became a future rival.
To test the hypothesis of outgroup rivalry bias (H2), we …rst investigate whether
the impact of wearing the t-shirt of Brazil’s rival team just before the match
(e.g. the e¤ect of wearing a Croatian T-shirt one day before Brazil v. Croatia)
di¤ers from the e¤ect on other occasions. Table 6 presents the regression results.
Surprisingly, we …nd no statistically signi…cant impact of being a supporter of the
rival team on the market discrimination. We also check whether sellers discriminated
against supporters of a "recently defeated team", but again we …nd no signi…cant
pattern [columns (2), (4), and (6)]. The results reject hypothesis 2, and therefore
suggest that Brazilian sellers discriminated against buyers wearing the t-shirt of
non-Brazilian winners because their victories diminished the impact of Brazil’s.
T able6
5
Conclusion
In this paper, we investigate whether group identity determines bargaining
outcomes in the marketplace and under which circumstances such biases manifest.
The results of our audit study during the WC show that non-Brazilians are charged
higher prices than Brazilian buyers, which is perhaps not surprising given the
common perception that tourists are generally overcharged, due to, for example,
their lack of experience in the local marketplace; however, one cannot distinguish
whether some of this discrimination is due to group prejudices.
In a more controlled analysis, we …nd that ingroup and outgroup identities do
18
play a non-trivial role in market discrimination. First, we …nd that the Brazil tshirt was instrumental in determining bargaining outcomes when Brazil had just
won a match. We argue that the motive behind the discrimination was sellers’
preference towards ingroup buyers, which was primed by the t-shirt and the recent
victory. The sympathy towards buyers from the ingroup arose in the case of a group
success, and it manifested in the form of supply-side discrimination. This …nding is
strong evidence of ingroup bias given that the sellers in this experiment are in the
informal sector, are mostly self-employed and work with small pro…t margins, and
thus presumably cannot a¤ord to let their feelings a¤ect their economic pro…ts. Our
results resonate with the …nding that people are more willing to donate to those
in the same racial group (Fong and Luttmer 2009) in the sense that sellers became
more generous towards buyers that were rooting for the same football team.
In contrast, we …nd that sellers discriminated against non-Brazilian buyers (i.e.
charged higher prices and were less willing to trade) when they wore a t-shirt of
another winning team (Chile, in our case). Our tests reject the hypothesis that
this has to do with rivalry against Brazil, or that this is a way to counterbalance
the loss of pro…t from giving better deals to Brazil winning t-shirt buyers. The
most plausible account is, instead, related more to the comparison of achievements.
(Note that we observe Brazil and Chile winning t-shirts in the same day). That is,
another team’s victory than their own might make sellers feel less victorious, which
in turn makes them more likely to treat the buyers wearing the winning Chile t-shirt
unfavourably. This explanation is in line with previous studies that demonstrate a
negative causal e¤ect of neighbours/co-workers’ success on individuals’ happiness
(Card et al. 2012, Luttmer 2005). It is also in line with theories in social psychology
on negative feelings towards outgroup individuals when they threaten the ingroup
(i.e. the Chilean victory is a threat to the association of "victory" with Brazilian
group status).
Our …nal remark relates to the external validity of this study. Although the WC
provided a nice opportunity to identify ingroup and outgroup biases, the mental
state during the WC ("hot state") might di¤er substantially from that in everyday
life, when people might feel less nationalistic; however, there are many important
19
situations that are similar to our experimental environment. For example, during
election time, people support causes, take sides, and often express their opinions
through the clothes they wear. In such a case, the expressed political preference of
buyers may lead them be treated di¤erentially in the market.
References
[1] Akerlof, G. A. and R. E. Kranton. 2010. Identity Economics, Princeton
University Press.
[2] Ball, S., C. Eckel, P. Grossman, and W. Zame, 2001. "Status in markets".
Quarterly Journal of Economics 155 (1): 61–181.
[3] Bargh, J. A. 2014. The Historical Origins of Priming as the Preparation of
Behavioral Responses: Unconscious Carryover and Contextual In‡uences of
Real-World Importance. Social Cognition: Vol. 32, Understanding Priming
E¤ects in Social Psychology, pp. 209-224.
[4] Becker, G., 1971. The Economics of Discrimination, 2nd ed. University of
Chicago Press.
[5] Benabou, R. and J. Tirole .2011. "Identity, Morals, and Taboos: Beliefs as
Assets." Quarterly Journal of Economics 126: 805-855.
[6] Benjamin, D. J., J. Choi, and A. J. Strickland. 2010. "Social Identity and
Preferences." American Economic Review, 100(4): 1913-28.
[7] Card, D., A. Mas, E. Moretti, and E. Saez. 2012. "Inequality at Work: The
E¤ect of Peer Salaries on Job Satisfaction." American Economic Review,
102(6): 2981-3003.
[8] Castillo, M., R. Petrie, M. Torero and L. Vesterlund .2013. “Gender Di¤erences
in Bargaining Outcomes: A Field Experiment on Discrimination.” Journal
of Public Economics 99: 35-48.
[9] Chen Y, SX Li, TX Liu, M Shih. 2014. "Which hat to wear? Impact of natural
identities on coordination and cooperation." Games and Economic Behavior
84: 58–86
20
[10] Chen, Y. and SX Li. 2009. “Group Identity and Social Preferences.”American
Economic Review, 99(1): 431–57.
[11] Currarini, S. and F. Mengel, 2013. “Identity, Homophily and In-Group Bias,”
FEEM Working Paper No. 37.
[12] Eckel, C. and P. Grossman. 2005. “Managing Diversity by Creating Team
Identity.”Journal of Economic Behavior and Organization, 58(3): 371-392.
[13] Edmans, A., D. Garcia. and Ø Norli. 2007. “Sports Sentiment and Stock
Returns.”The Journal of Finance, 62: 1967–1998.
[14] Ewens, M., B. Tomlin. and L-C. Wang. 2014. ‘Statistical discrimination or
prejudice? A large sample …eld experiment’, Review of Economics and
Statistics, 96 (1): 119-134.
[15] Fong, C.M., and E. F. P. Luttmer. 2009. "What Determines Giving to Hurricane
Katrina Victims?
Experimental Evidence on Racial Group Loyalty."
American Economic Journal: Applied Economics, 1(2): 64-87.
[16] Gneezy, U., J. List and M. Price. 2012. “Towards an understanding of why
people discriminate: evidence from a series of natural …eld experiments.”
NBER Working Paper No. 17855.
[17] Goette, L., D. Hu¤man, and S. Meier. 2006. "The Impact of Group
Membership on Cooperation and Norm Enforcement: Evidence Using
Random Assignment to Real Social Groups." American Economic Review,
96(2): 212-216.
[18] Guryan, J. and Charles, K. K. 2013. "Taste-based or Statistical Discrimination:
The Economics of Discrimination Returns to its Roots." The Economic
Journal, 123: 417–432.
[19] Hassin, R. R., M. J. Ferguson, D. Shidlovski and L. Gross. 2007."Subliminal
exposure to national ‡ags a¤ects political thought and behavior".
Proceedings of the National Academy of Sciences, USA, 104: 19757–19761.
[20] Heckman, J.J. 1998. ‘Detecting discrimination’, Journal of Economic
Perspectives, 12(2): 101–16.
21
[21] Ifcher, J. and H. Zarghamee. 2011. “Happiness and Time Preference: The
e¤ect of Positive A¤ect in a Random-Assignment Experiment,”American
Economic Review, 101(4): 3109-3129.
[22] Li, S.X., Dogan, K. and E. Haruvy 2011. "Group identity in markets."
International Journal of Industrial Organization 29: 104-115.
[23] List, J.A. 2004. "The nature and extent of discrimination in the marketplace:
evidence from the …eld." Quarterly Journal of Economics 119(1): 49–89.
[24] Molden. D. C. 2014. Understanding Priming E¤ects in Social Psychology: What
is “Social Priming” and How does it Occur? Social Cognition: Vol. 32,
Understanding Priming E¤ects in Social Psychology, pp. 1-11.
[25] "World
Cup
Survival
Guide",
New
York
Times 12/6/2014, http://www.nytimes.com/2014/06/13/opinion/cohenworld-cup-survival-guide.html?_r=0
[26] "Making Holidays of Brazil’s World Cup Games", New York Times
9/6/2014, http://www.nytimes.com/2014/06/10/sports/worldcup/makingholidays-of-brazil-world-cup-games.html
[27] Oreopoulos, P. 2011. "Why Do Skilled Immigrants Struggle in the Labor
Market? A Field Experiment with Thirteen Thousand Resumes." American
Economic Journal: Economic Policy, 3(4): 148-71.
[28] Perdue, C. W., J. F. Dovidio, M. B. Gurtman, and R. B. Tyler. 1990. “Us and
them: Social categorization and the process of ingroup bias,” Journal of
Personality and Social Psychology, 59: 475–486.
[29] Phelps, E., 1972. "The statistical theory of racism and sexism." American
Economic Review 62: 659–661.
[30] Platow, M. J.; Durante, M.; Williams, N.; Garrett, M.; Walshe, J.; Cincotta,
S.; Lianos, G.; Barutchu, A. 1999. The contribution of sport fan social
identity to the production of prosocial behavior. Group Dynamics: Theory,
Research, and Practice, 3(2): 161-169.
[31] Rich, J., “What Do Field Experiments of Discrimination in Markets Tell Us?
22
A Meta Analysis of Studies Conducted Since 2000.” IZA Discussion Paper
No. 8584.
[32] Rick, S. and G. Loewenstein. 2008. “The Role of Emotion in Economic
Behavior,” in Lewis, M., Haviland-Jones, J. M., & Barrett, L. F. (Eds.).
Handbook of Emotions, 3rd Edition. New York: Guilford.
[33] Shayo, M. and A. Zussman. 2010. “Judicial In-group Bias in the Shadow of.
Terrorism.”Quarterly Journal of Economics. 126: 1147-1484
[34] Shih, M., T. L. Pittinsky, A. Trahan and N. Ambady. 1999. “Stereotype
Susceptibility: Identity Salience and Shifts in Quantitative Performance,”
Psychological Science, 10 (1): 81–84.
[35] Wann, D. L and F.rederick G Grieve. 2005. "Biased Evaluations of In-Group
and Out-Group Spectator Behavior at Sporting Events: The Importance of
Team Identi…cation and Threats to Social Identity", The Journal of Social
Psychology, 145 (5): 531-546.
[36] Wann, D. L., M. J. Melnick, G. W. Russell and D.G. Pease. 2001. Sport fans:
The psychology and social impact of spectators. New York: Routledge.
[37] Zussman, A. 2013. “Ethnic Discrimination: Lessons from the Israeli Online
Market for Used Cars.”The Economic Journal, 123:433-468.
23
Figure 1 - Timeline
Matches:
(Scores in parenthesis)
Dates of data collection
and t-shirts:
June 17
Brazil (0) vs. Mexico (0)
June 12
Brazil (3) vs. Croatia (1)
June 11
Brazil (official
/non-official)
Croatia
June 14
May 30, 31
Brazil (official)
Chile
Spain
June 16
Brazil (official)
Mexico
Brazil (official
/non-official)
Chile
Croatia
June 28
Brazil (3) vs. Chile (2)
July 8
Brazil (1) vs. Germany (7)
June 27
July 7
Brazil (official
Brazil (official)
/non-official)
Chile
June 30
June 20
Brazil (official
Brazil (official)
/non-official)
Mexico
Chile
Spain
Mexico
Table 1. Number of Interactions by Market, Buyer Nationality and Treatment
Brazilian Pair
Jacket
Brazil National Shirt
Official Shirt
Non-official Shirt
Mercado Santa Efiginia
374
145
229
151
77
Mercado Paulista
96
43
53
43
10
Total
470 (50.9%)
198 (21%)
272 (29.5%)
194 (21.04%)
87 (9.4%)
Non-Brazilian Pair
Jacket
National Official Shirt
Chile
Croatia
Mexico
Spain
Total
312
175
137
76
25
51
22
686 (74.4%)
140
74
66
12
22
18
21
236 (25.6%)
452 (49.02%)
249 (27%)
203 (22%)
88 (9.5%)
47 (5.1%)
69 (7.5%)
43 (4.7%)
922
Table 2. Effects of Nationality and Team T-Shirts on Market Outcomes
dependent variable:
First Price
(1)
(2)
Second Price
(3)
(4)
Accepted Offer
(5)
(6)
Nationality
Brazilian
-0.0632**
(0.0144)
-0.0683**
(0.0188)
-0.054**
(0.023)
-0.077*
(0.0412)
0.0400
(0.0309)
0.0706
(0.0509)
t-Shirt
0.0068
(0.0152)
-0.0357
(0.0309)
0.0075
(0.0451)
-0.0145
(0.0208)
0.0748*
(0.0434)
Brazil
Chile
Croatia
Mexico
Spain
Sample (pair)
Num. of Obs.
All
697
All
697
0.0257
(0.0327)
-0.0272
(0.0358)
0.0165
(0.0502)
-0.0434*
(0.0258)
0.0136
(0.0593)
All
553
All
553
-0.0338
(0.0559)
-0.0178
(0.0721)
-0.0964
(0.0808)
0.1421**
(0.0639)
0.0088
(0.0854)
All
697
Notes: All specifications include buyer-pair random effects, indicators for product, day, time of the store visit and
stores’ addresses. Robust standard errors clustered at the store address level are in parenthesis.
Price variables are transformed to their logarithm form.
* Significant at the 10% level, ** Significant at the 5% level.
All
697
Table 3. Effect of a Victorious Brazil Team T-shirt on Market Outcomes
dependent variable:
First Price
(1)
(2)
Second Price
(3)
(4)
Accepted Offer
(5)
(6)
Brazil t-Shirt × Victory
Any Brazil t-shirt
-0.1449
(0.1006)
-0.3008**
(0.1272)
-0.0235
(0.0275)
-0.0223
(0.0342)
Official t-Shirt
Unofficial t-Shirt
0.3354**
(0.1083)
-0.1074**
(0.0401)
-0.0915*
(0.0498)
0.2459
(0.1403)
0.5103**
(0.0993)
Brazil t-Shirt
Any Brazil t-shirt
0.1404
(0.0972)
Official t-Shirt
Unofficial t-Shirt
Sample (pair)
Num. of Obs.
Brazilian
362
0.2033
(0.1356)
0.0235
(0.0179)
0.0098
(0.039)
Brazilian
362
Brazilian
276
-0.1029**
(0.0432)
0.0681*
(0.0381)
0.0040
(0.0605)
Brazilian
276
Notes: All specifications include buyer-pair fixed effects, indicators for product, day, time of the store visit and
stores’ addresses. Robust standard errors clustered at the store address level are in parenthesis.
Price variables are transformed to their logarithm form.
* Significant at the 10% level, ** Significant at the 5% level.
Brazilian
362
-0.0699
(0.0432)
-0.2078**
(0.0622)
Brazilian
362
Table 4. Effects of a Victorious Non-Brazilian Team T-shirt on Market Outcomes
dependent variable:
t-Shirt × Victory
First Price
(1)
(2)
0.1195**
0.2107**
(0.0504)
(0.0708)
Second Price
(3)
(4)
0.1040
0.1901
(0.0902)
(0.1170)
-0.0167
(0.0225)
-0.0162
(0.0308)
Accepted Offer
(5)
(6)
-0.4257**
-0.5590**
(0.1926)
(0.209)
t-Shirt
Any Non-Brazil team
Chile
Croatia
Mexico
Spain
Sample (pair)
Num. of Obs.
-0.1051*
(0.0578)
0.0387
(0.0578)
-0.0342
(0.0329)
0.045
(0.0297)
non-Brazilian non-Brazilian
333
333
0.0721
(0.0445)
-0.0865
(0.0817)
0.0573
(0.0402)
-0.0495
(0.0464)
0.0140
(0.0436)
non-Brazilian non-Brazilian
276
276
0.1738
(0.1294)
-0.1362
(0.1227)
0.1698
(0.0901)
0.0245
(0.0810)
non-Brazilian non-Brazilian
333
333
Notes: All specifications include buyer-pair fixed effects, indicators for product, day, time of the store visit and
stores’ addresses. Robust standard errors clustered at the store address level are in parenthesis.
Price variables are transformed to their logarithm form.
* Significant at the 10% level, ** Significant at the 5% level.
Table 5. Effects of a Victorious Team T-shirt on Non-Market Outcomes
dependent variable:
t-Shirt × Victory
Offered Similar Product
in Own Store
(1)
-0.1470
(0.0908)
(2)
-0.1019
(0.0963)
0.1034
(0.0502)**
0.2795
(0.1348)**
0.0766
(0.1377)
0.1396
(0.1435)
0.0067
(0.2304)
-0.2508
(0.3139)
All
211
-0.0031
(0.0548)
-0.1469
(0.1002)
Gave a recommendation of another store
(3)
-0.4678
(0.2270)**
(4)
0.0149
(0.1734)
(5)
-0.0708
(0.3094)
-0.0176
(0.0687)
-0.2693
(0.1173)**
0.1412
(0.1998)
0.1557
(0.2702)
-0.0043
(0.2586)
-0.2875
(0.3992)
non-Brazilian
107
-0.0096
(0.0444)
-0.2554
(0.1375)*
0.0766
(0.1377)
0.3581
(0.1498)**
0.1182
(0.1496)
0.5470
(0.2874)**
All
211
(6)
0.2396
(0.2230)
t-Shirt
Brazil (Official)
Brazil (Unofficial)
Chile
Croatia
Mexico
Spain
Sample (pair)
Num. of Obs.
Brazilian
104
Brazilian
104
-0.2781
(0.2224)
0.4051
(0.1268)**
0.0418
(0.1796)
0.4699
(0.2732)*
non-Brazilian
107
Notes: All specifications include buyer-pair fixed effects, indicators for day, time of the store visit and stores’ addresses.
Robust standard errors clustered at the store address level are in parenthesis.
The observations refer to cases in which the enquired product was not available.
* Significant at the 10% level, ** Significant at the 5% level.
Table 6. Effects of a Rival and Defeated Team T-shirt on Market Outcomes
dependent variable:
t-Shirt × Rival
t-Shirt × Defeated
First Price
(1)
-0.0183
(0.0434)
(2)
-0.0120
(0.0366)
Second Price
(3)
(4)
-0.0026
(0.0625)
0.0162
(0.0358)
Accepted Offer
(5)
(6)
0.2040
(0.1845)
-0.0710
(0.1570)
-0.0481
-0.0537
(0.0703)
(0.0537)
0.0525
0.0438
(0.0433)
(0.0396)
-0.0415
-0.0462
(0.0534)
(0.0415)
0.0226
0.0118
(0.0420)
(0.0455)
non-Brazilian non-Brazilian
276
276
-0.0061
0.0648
(0.1413)
(0.1507)
-0.2385
-0.0914
(0.1825)
(0.1545)
0.0504
0.1598
(0.1499)
(0.0879)*
-0.0029
0.0512
(0.0806)
(0.1522)
non-Brazilian non-Brazilian
333
333
t-Shirt
Chile
Croatia
Mexico
Spain
Sample (pair)
Num. of Obs.
-0.0517
-0.0527
(0.0488)
(0.0435)
0.0434
0.0377
(0.0568)
(0.0566)
-0.0172
-0.0241
(0.0335)
(0.0307)
0.0532
0.0605
(0.0298)*
(0.0414)
non-Brazilian non-Brazilian
333
333
Notes: All specifications include buyer-pair fixed effects, indicators for product, day, time of the store visit and
stores’ addresses. Robust standard errors clustered at the store address level are in parenthesis.
Price variables are transformed to their logarithm form.
* Significant at the 10% level.
dependent variable:
mean
Panel A
Brazil (Official) t-shirt
Table A1- Randomization Check: Effects of T-shirts on stores and enquired product
Enquired Product
seller is wearing
seller is male
high demand store Sony headphone FIFA World Cup Sandisk memory
Brazil t-shirt
XZ 300
game
card 8 GB
0.035
0.665
0.151
0.289
0.231
0.308
(1)
(2)
(3)
(4)
(5)
(6)
Sony headphone
XZ 100
0.067
(7)
Sandisk memory
card 32 GB
0.110
(8)
0.0078
(0.0171)
0.0756
(0.0500)
0.0096
(0.0370)
0.0733
(0.0430)*
0.0110
(0.0480)
-0.0478
(0.0446)
-0.0171
(0.0231)
-0.0194
(0.0241)
Brazil (Unofficial) t-shirt
0.0338
(0.0296)
0.1124
(0.0656)*
0.0659
(0.0531)
0.0202
(0.0499)
0.0313
(0.0585)
-0.0083
(0.0504)
-0.0023
(0.0361)
-0.0408
(0.0392)
Chile t-shirt
0.0111
(0.0348)
0.0248
(0.0748)
0.0134
(0.0584)
0.0472
(0.0559)
0.0604
(0.0597)
-0.0676
(0.0601)
-0.0241
(0.0440)
-0.0160
(0.0523)
Croatia t-shirt
-0.0482
(0.0478)
-0.0008
(0.0943)
-0.0801
(0.0795)
-0.0833
(0.0925)
0.0581
(0.0868)
0.0316
(0.1034)
-0.0023
(0.0084)
-0.0042
(0.0104)
Mexico t-shirt
-0.0246
(0.0318)
0.0512
(0.0869)
0.0432
(0.0452)
-0.0113
(0.0723)
0.0002
(0.0768)
0.0079
(0.0767)
-0.0145
(0.0394)
0.0176
(0.0479)
Spain t-shirt
0.0198
(0.0338)
-0.0588
(0.0934)
-0.1689
(0.0848)
0.0937
(0.1066)
-0.1737
(0.0842)**
0.0800
(0.1062)
-0.0002
(0.0101)
0.0002
(0.0116)
0.0156
(0.0161)
0.0831
(0.0453)*
0.0260
(0.0339)
0.0582
(0.0384)
0.0156
(0.0421)
-0.0355
(0.0390)
-0.0130
(0.0219)
-0.0252
(0.0227)
-0.0087
(0.0196)
ALL
904
0.0190
(0.0458)
ALL
905
-0.0319
(0.0350)
ALL
904
0.0095
(0.0396)
ALL
907
0.0098
(0.0403)
ALL
907
-0.0041
(0.0427)
ALL
907
-0.0125
(0.0195)
ALL
907
-0.0027
(0.0235)
ALL
907
Panel B
Brazil t-shirt
t-shirt ofother team than Brazil
Sample (pair)
Num. of Obs.
Note: All specifications include buyer-pair and product fixed effects, indicators for day, time of the store visit and stores’ addresses.
High demand store refer to store for which the pair had to wait to be attended because there were other customers in the store.
Robust standard errors clustered at the store address level are in parenthesis.
* Significant at the 10% level, ** Significant at the 5% level.
dependent variable:
t-Shirt×Victory
t-Shirt×Rival
t-Shirt × Defeated
t-Shirt
Brazil (Official)
Brazil (Unofficial)
Table A2- Robustness Check
Friend' Buyer Assessment on Buyers' Confidence in his/her interaction with
Seller
(1)
(2)
(3)
(4)
(5)
-0.2412
0.0469
0.0750
(0.1586)
(0.2409)
(0.2283)
0.0395
0.0665
(0.1115)
(0.1186)
0.0001
0.0362
(0.0731)
(0.0785)
-0.0480
(0.0486)
-0.0375
(0.0749)
0.0261
(0.0750)
0.0117
(0.0622)
0.0528
(0.0672)
-0.2595
(0.0669)**
Chile
Croatia
Mexico
Spain
Sample (pair)
Num. of Obs.
Brazilian
467
0.0256
(0.0971)
-0.0112
(0.0610)
0.0354
(0.0873)
-0.2592
(0.0678)**
0.0371
(0.0790)
0.0096
(0.0799)
0.0541
(0.0590)
-0.2578
(0.0654)**
-0.0053
(0.0921)
-0.0383
(0.0868)
0.0137
(0.0816)
-0.2843
(0.0655)**
non-Brazilian non-Brazilian non-Brazilian non-Brazilian
439
439
439
439
NoteS: The dependent variable is the buyer’ friend perception on how the buyer felt/behaved during the interaction with
the sellers, in a scale 1 to 5, where 1 correspond to very uncomfortable and 5 relates to very comfortable.
All specifications include buyer-pair fixed effects, indicators for product, day, time of the store visit and
stores’ addresses. Robust standard errors clustered at the store address level are in parenthesis.
** Significant at the 5% level.
Table A3. Effect of Team T-shirt on Market Outcomes
dependent variable:
Brazil t-Shirt × Victory
First Price
(1)
(2)
-0.0197
-0.0225
(0.0271)
(0.0267)
Chile t-Shirt × Victory
Croatia t-shirt
Mexico t-shirt
Spain t-shirt
Sample (pair)
Num. of Obs.
0.1619
0.2248
(0.0895)* (0.0906)**
0.0187
(0.0621)
0.0365
(0.0546)
non-Brazil t-Shirt × Defeated
Chile t-shirt
Second Price
Accepted Offer
(4)
(5)
(6)
(7)
(8)
-0.1034
-0.1062
-0.1046
0.2588
0.2678
(0.0312)** (0.0315)* (0.0317)** (0.1190)* (0.1184)**
0.1440
0.1655
(0.0597)** (0.0757)**
non-Brazil t-Shirt × Rival
Brazil t-shirt
(3)
-0.0205
(0.0262)
0.0089
(0.0118)
-0.0737
(0.0387)*
0.04219
(0.0565)
-0.0321
(0.0307)
0.0882
(0.0251)**
All
696
0.0104
0.0098
(0.0119)
(0.0119)
-0.1078
-0.1288
(0.0514)** (0.0690)*
0.0461
0.0209
(0.0566)
(0.0595)
-0.0352
-0.0512
(0.0311)
(0.0444)
0.0854
0.0581
(0.0242)** (0.0537)
All
All
696
696
-0.4466
(0.1396)**
0.0909
(0.0828)
0.0981
(0.0625)
0.0489
(0.0298)
-0.0646
(0.0535)
0.0192
(0.0405)
-0.0668
(0.0540)
0.0482
(0.0358)
All
552
0.0513
0.0501
-0.0871
(0.0300) (0.0303) (0.0460)*
-0.0957
-0.1618
0.0895
(0.0721) (0.0874)* (0.1111)
0.0250
-0.0682
-0.1221
(0.0400) (0.0589) (0.1076)
-0.0702
-0.1375
0.1385
(0.0553) (0.0551)** (0.0751)*
0.0447
-0.0322
-0.0769
(0.0350) (0.0666) (0.0846)
All
All
All
552
552
696
Notes: All specifications include buyer-pair and product fixed effects, indicators for day, time of the store visit and stores’ addresses.
Robust standard errors clustered at the store address level are in parenthesis.
Price variables are transformed to their logarithm form.
* Significant at the 10% level, ** Significant at the 5% level.
(9)
0.2575
(0.1305)*
-0.4610
(0.1109)**
0.0648
(0.1708)
-0.0570
(0.1448)
-0.0920
(0.0454)**
0.1955
(0.1087)*
-0.1344
(0.1068)
0.1482
(0.0736)**
-0.0681
(0.0836)
All
696
-0.0888
(0.0462)*
0.2016
(0.1149)
-0.1504
(0.1871)
0.1287
(0.1221)
-0.0311
(0.1526)
All
696
Appendix – Enquired Products