Final Report Summary - EDIM (Ethnic Discrimination in Israeli Markets)
The study of discrimination against minorities in the marketplace has received great amounts of attention by economists in recent decades. There are two leading explanations for discriminatory behaviour in markets. The first focuses on 'taste-based discrimination', or personal prejudice, of economic agents who dislike associating with individuals of a given gender, race, or ethnicity. The second leading theory of discrimination focuses on 'statistical discrimination'. According to this theory, discriminatory behaviour is the result of differences across groups in aggregate characteristics. The decision-maker uses these differences to evaluate some outcome-relevant individual characteristics which are not easily observable.
Establishing that differential economic outcomes across groups are due to discrimination is a difficult challenge for empirical research using non-experimental data, as in many cases there are unobserved, potentially outcome-relevant factors that cannot be controlled for in the analysis. To overcome this challenge, economists have increasingly turned to quasi-experimental and experimental methodologies. Although experimental studies have often yielded highly convincing evidence of discrimination, it is often unclear how to interpret the results, as observed outcomes are consistent with competing explanations.
This project contributes to filling the void by using a novel two-stage experimental procedure to study discrimination towards Arabs both buyers and sellers in the Israeli online market for used cars.
The first stage of our analysis of discrimination against Arab buyers follows the correspondence study methodology and applies it in Israel's leading website for classified adverts. To uncover discriminatory behaviour by car sellers, we sent eight thousand pairs of e-mails from fictitious buyers, where each pair contains one e-mail from a buyer with a distinctly Arab (minority-group) first name and another e-mail from a buyer with a distinctly Jewish first name. This part of the study has two key features. First, we specifically target adverts posted by Jewish sellers. This allowed us to focus on the typical form of discrimination: that exhibited by members of the majority group against members of the minority group. Second, we randomly assign to buyers' e-mails suggested discounts. This feature facilitates direct monetisation of the extent of discrimination.
The second stage of the experimental procedure is novel to the literature on discrimination. It consists of a nominally independent telephone survey that collects socio-demographic information and elicits the attitudes towards Arabs of the Jewish sellers who were contacted in the first stage. Some of the attitude questions were aimed to capture prejudicial views while others were aimed to capture the type of views highlighted in models of statistical discrimination.
This two-stage procedure allowed us to identify, at the individual level, a direct relationship between explicit attitudes towards minorities and discriminatory behaviour.
The results of the first stage provide robust evidence that Jewish car sellers discriminate against the Arab buyer. The overall response rate to e-mails is 22 % higher for the Jewish than for the Arab buyer. We find that the extent of discrimination varies with monetary incentives: when an Arab buyer 'offers' the car's posted price, he obtains the same response rate as a Jewish buyer who 'requests' a five to ten percent discount relative to the posted price.
Survey results demonstrate that a large share of Jewish car sellers hold intolerant views towards Arabs. For example, more than half of the survey participants indicated that they do not want to have Arab Israeli neighbours. We also find that attitudes towards Arabs are significantly correlated with sellers' socio-demographic characteristics. Thus, for instance, unfavourable views of Arabs are positively correlated with religiosity and negatively correlated with education. The associations we find between attitudes and socio-demographic characteristics are consistent with existing literature and conventional wisdom, lending credibility to the research design.
By combining the results of the two stages of the experimental procedure, we demonstrate that discriminatory behaviour by Jewish sellers is associated not only with monetary incentives but also with explicit attitudes towards Arabs as captured in the survey. Crucially, variation in the response to only one of the attitude statements is significantly correlated with variation in discriminatory behaviour: Jewish car sellers who strongly disagree with the statement that 'Arabs in Israel are more likely to cheat than Jews' do not discriminate against the Arab buyer while others sellers do.
We argue that these results are consistent with the theory of statistical discrimination: behaviour of majority group members towards representative members of the majority and minority groups are determined by perceptions regarding aggregate differences across groups in transaction-relevant characteristics. A seller who is concerned about being cheated will naturally prefer to transact with a potential buyer who is ex-ante and believed to be more trustworthy.
Our analysis of discrimination against Arab sellers in the same market provides further support for the centrality of 'statistical' considerations in agents' decision-making. To establish whether and to what extent Arab sellers are being discriminated against in this market, we conducted a field experiment in which we posted a large number of fictitious adverts for used cars. To signal the ethnic identity of the sellers, we randomly assign to the name field in the adverts the same distinctly ethnic yet highly common Arab and Jewish first names used in the first experiment.
The main outcome measures for each advert are the number of incoming phone calls and the number of unique phone numbers associated with these calls. Results of the field experiment provide robust evidence of discrimination against Arab sellers. The extent of discrimination is large: on average, adverts posted by the Arab seller are associated with only half as many phone calls and unique phone numbers as adverts posted by the Jewish seller.
To investigate which of the two leading theories of discrimination can account for the results, we use the telephone numbers collected in the field experiment to contact prospective buyers and solicit their participation in a (nominally independent) survey similar to the one discussed above. The fact that Arabs and Jews have distinct accents enables us to identify the ethnicity of prospective buyers, even when they refuse to participate in the survey.
The evidence we obtained suggests that discrimination against Arab sellers is motivated by statistical rather than taste considerations. This conclusion is based on two results. First, we find that it is not only Jewish buyers who discriminate against Arab sellers: Arab buyers discriminate against sellers from their own ethnic group, although to a lesser extent than Jewish buyers do. This pattern is clearly inconsistent with taste-based discrimination playing a major role since this theory would predict that both Arab and Jewish buyers would prefer to transact with sellers from their own ethnic group. Secondly, survey results indicate that the discrimination exhibited by Jewish buyers is associated with the strength of their belief that Arabs are more likely than Jews to cheat but not with other beliefs, in particular those capturing prejudicial tastes.
The results of the project may have important policy implications. The relationship between Arabs and Jews is a hotly debated question in Israeli politics, mainly because Arab-Jewish relations are strongly influenced by the intensity of the Israeli-Palestinian conflict. Not much attention has been paid to date to the economic relationship between Israeli Arabs and Jews. Anti-Arab discrimination leads to economic segregation which reduces potentially large gains from trade between the two groups. By documenting and attempting to explain patterns of discrimination, this project may influence the formulation of policies aimed at reducing discrimination. Improved economic relations between Arabs and Jews hold the potential to spill over to the social and political spheres. Thus research projects like this one may contribute to improving Arab-Jewish relations.
The results of the project are likely to be relevant in other contexts as well since ethnic tensions characterise many countries around the world, including members of the EU. Following the terrorist attacks in the United States in September 2001, the invasion of Iraq in March of 2003, the Madrid train bombings in March 2004, the London bombings in July 2005, and other politically charged events, ethnic tensions have risen in many European countries. The rising tensions were accompanied by charges of heightened discrimination against minorities (especially Muslims). I believe that lessons learned from my research would be relevant not only for Israel but for Europe at large.
Establishing that differential economic outcomes across groups are due to discrimination is a difficult challenge for empirical research using non-experimental data, as in many cases there are unobserved, potentially outcome-relevant factors that cannot be controlled for in the analysis. To overcome this challenge, economists have increasingly turned to quasi-experimental and experimental methodologies. Although experimental studies have often yielded highly convincing evidence of discrimination, it is often unclear how to interpret the results, as observed outcomes are consistent with competing explanations.
This project contributes to filling the void by using a novel two-stage experimental procedure to study discrimination towards Arabs both buyers and sellers in the Israeli online market for used cars.
The first stage of our analysis of discrimination against Arab buyers follows the correspondence study methodology and applies it in Israel's leading website for classified adverts. To uncover discriminatory behaviour by car sellers, we sent eight thousand pairs of e-mails from fictitious buyers, where each pair contains one e-mail from a buyer with a distinctly Arab (minority-group) first name and another e-mail from a buyer with a distinctly Jewish first name. This part of the study has two key features. First, we specifically target adverts posted by Jewish sellers. This allowed us to focus on the typical form of discrimination: that exhibited by members of the majority group against members of the minority group. Second, we randomly assign to buyers' e-mails suggested discounts. This feature facilitates direct monetisation of the extent of discrimination.
The second stage of the experimental procedure is novel to the literature on discrimination. It consists of a nominally independent telephone survey that collects socio-demographic information and elicits the attitudes towards Arabs of the Jewish sellers who were contacted in the first stage. Some of the attitude questions were aimed to capture prejudicial views while others were aimed to capture the type of views highlighted in models of statistical discrimination.
This two-stage procedure allowed us to identify, at the individual level, a direct relationship between explicit attitudes towards minorities and discriminatory behaviour.
The results of the first stage provide robust evidence that Jewish car sellers discriminate against the Arab buyer. The overall response rate to e-mails is 22 % higher for the Jewish than for the Arab buyer. We find that the extent of discrimination varies with monetary incentives: when an Arab buyer 'offers' the car's posted price, he obtains the same response rate as a Jewish buyer who 'requests' a five to ten percent discount relative to the posted price.
Survey results demonstrate that a large share of Jewish car sellers hold intolerant views towards Arabs. For example, more than half of the survey participants indicated that they do not want to have Arab Israeli neighbours. We also find that attitudes towards Arabs are significantly correlated with sellers' socio-demographic characteristics. Thus, for instance, unfavourable views of Arabs are positively correlated with religiosity and negatively correlated with education. The associations we find between attitudes and socio-demographic characteristics are consistent with existing literature and conventional wisdom, lending credibility to the research design.
By combining the results of the two stages of the experimental procedure, we demonstrate that discriminatory behaviour by Jewish sellers is associated not only with monetary incentives but also with explicit attitudes towards Arabs as captured in the survey. Crucially, variation in the response to only one of the attitude statements is significantly correlated with variation in discriminatory behaviour: Jewish car sellers who strongly disagree with the statement that 'Arabs in Israel are more likely to cheat than Jews' do not discriminate against the Arab buyer while others sellers do.
We argue that these results are consistent with the theory of statistical discrimination: behaviour of majority group members towards representative members of the majority and minority groups are determined by perceptions regarding aggregate differences across groups in transaction-relevant characteristics. A seller who is concerned about being cheated will naturally prefer to transact with a potential buyer who is ex-ante and believed to be more trustworthy.
Our analysis of discrimination against Arab sellers in the same market provides further support for the centrality of 'statistical' considerations in agents' decision-making. To establish whether and to what extent Arab sellers are being discriminated against in this market, we conducted a field experiment in which we posted a large number of fictitious adverts for used cars. To signal the ethnic identity of the sellers, we randomly assign to the name field in the adverts the same distinctly ethnic yet highly common Arab and Jewish first names used in the first experiment.
The main outcome measures for each advert are the number of incoming phone calls and the number of unique phone numbers associated with these calls. Results of the field experiment provide robust evidence of discrimination against Arab sellers. The extent of discrimination is large: on average, adverts posted by the Arab seller are associated with only half as many phone calls and unique phone numbers as adverts posted by the Jewish seller.
To investigate which of the two leading theories of discrimination can account for the results, we use the telephone numbers collected in the field experiment to contact prospective buyers and solicit their participation in a (nominally independent) survey similar to the one discussed above. The fact that Arabs and Jews have distinct accents enables us to identify the ethnicity of prospective buyers, even when they refuse to participate in the survey.
The evidence we obtained suggests that discrimination against Arab sellers is motivated by statistical rather than taste considerations. This conclusion is based on two results. First, we find that it is not only Jewish buyers who discriminate against Arab sellers: Arab buyers discriminate against sellers from their own ethnic group, although to a lesser extent than Jewish buyers do. This pattern is clearly inconsistent with taste-based discrimination playing a major role since this theory would predict that both Arab and Jewish buyers would prefer to transact with sellers from their own ethnic group. Secondly, survey results indicate that the discrimination exhibited by Jewish buyers is associated with the strength of their belief that Arabs are more likely than Jews to cheat but not with other beliefs, in particular those capturing prejudicial tastes.
The results of the project may have important policy implications. The relationship between Arabs and Jews is a hotly debated question in Israeli politics, mainly because Arab-Jewish relations are strongly influenced by the intensity of the Israeli-Palestinian conflict. Not much attention has been paid to date to the economic relationship between Israeli Arabs and Jews. Anti-Arab discrimination leads to economic segregation which reduces potentially large gains from trade between the two groups. By documenting and attempting to explain patterns of discrimination, this project may influence the formulation of policies aimed at reducing discrimination. Improved economic relations between Arabs and Jews hold the potential to spill over to the social and political spheres. Thus research projects like this one may contribute to improving Arab-Jewish relations.
The results of the project are likely to be relevant in other contexts as well since ethnic tensions characterise many countries around the world, including members of the EU. Following the terrorist attacks in the United States in September 2001, the invasion of Iraq in March of 2003, the Madrid train bombings in March 2004, the London bombings in July 2005, and other politically charged events, ethnic tensions have risen in many European countries. The rising tensions were accompanied by charges of heightened discrimination against minorities (especially Muslims). I believe that lessons learned from my research would be relevant not only for Israel but for Europe at large.