I
nthis chapter, I summarize the dissertation’s contributions to the study of how group-based biases shape economic and political interactions. I begin by identifying the key insights to be drawn from the five research articles. This first part of the chapter is organized into two sections, corresponding to the two tracks of my empirical work. The first section outlines the main results from the research articles on how discrimination shapes access to the labour market, when, and for which groups. The second section presents the main results from the second track concerning how discrimination shapes political representation. Note that this chapter only summarizes the core results from the five research articles while more detailed results and analyses can be found in the individual research articles. I then discuss the limitations and broader implications of the main findings in the research articles. Finally, I point to avenues for future theoretical and empirical work on discrimination.Chapter 4: Core results, limitations and implications
I
nthis chapter, I summarize the dissertation’s contributions to the study of how group-based biases shape economic and political interactions. I begin by identifying the key insights to be drawn from the five research articles. This first part of the chapter is organized into two sections, corresponding to the two tracks of my empirical work. The first section outlines the main results from the research articles on how discrimination shapes access to the labour market, when, and for which groups. The second section presents the main results from the second track concerning how discrimination shapes political representation. Note that this chapter only summarizes the core results from the five research articles while more detailed results and analyses can be found in the individual research articles. I then discuss the limitations and broader implications of the main findings in the research articles. Finally, I point to avenues for future theoretical and empirical work on discrimination.Chapter Four: Results
Core results
In the introduction of the dissertation, I formulated the following research question:How do group-based biases shape economic and political interactions between salient social groups? The next two sections briefly outline the main findings that speak to this question from each empirical track.
Note that since the chapter draws on the research articles, there are overlaps between the articles and this chapter.
Group-based biases significantly shape access to the labour market
As discussed in the previous chapters, the literature on labour market discrimination generally indicates that discrimination against immigrant-origin minorities is persistent and widespread. Ar-ticles A,Intersections, and B,Alike but different, engage with this literature to identify the causal effects of group categories on hiring practices, and to consider how and under what conditions some group cues result in disparities, whereas others do not. The articles rely on field-experimental designs in which the ascriptive traits of fictitious job applicants were randomly assigned before sending the applications to a variety of employers advertising for a range of positions – a design that cleanly sidesteps issues of unmeasured confounding factors and allows for an assessment of real-world behaviour.
Collectively, the findings from the articles indicate that discrimination is a serious barrier and plays a role in shaping contemporary labor market disparities. Both research articles demonstrate that origin minorities face labor-market barriers: job applicants with an immigrant-origin name receive significantly fewer invitations to job interviews compared to applicants with an ethnic majority name. This difference appears to be unrelated to discrimination based on so-cioeconomic characteristics. It is worth mentioning that the applicants had acquired educational merits, work experience and were written in flawless Danish. In addition to being the first studies to explore discrimination against immigrant-origin job applicants in the labour market in Denmark, they contribute to the literature more generally by advancing our understanding of the empirical patterns in several ways, which I outline in the following sections.
Chapter Four: Results
Discrimination is hierarchical
In Chapter Two, I proposed that we should expect group-based biases to increase as individuals appear more socially distant and when multiple out-group categories coincide. Articles A and B includes a variety of group configurations in order to test this proposition. That is, along the random assignment of ethnic affiliation, I manipulated gender, socioeconomic status, information about the applicants’ spare time interests, and whether the applicants were depicted wearing a headscarf in a CV photo.
Figure 1 summarizes the results from articles A and B across the main treatment groups.1 The figure depicts the average treatment effects (ATE) of having an immigrant-origin name ranging from 4.7 percentage points to 16.9 percentage points depending on the applicants’ other characteristics.
The figure reports five difference-in-means estimates using invitations to job interviews as the outcome variable. The first estimate depicts a difference between immigrant-origin minority and majority applicants of 5.1 percentage points for female applicants. This is a precision-weighted estimate derived from the average treatment effect of 5.7 percentage points in article A (second estimate) and an ATE of 4.7 percentage points in article B (third estimate). The fourth estimate (11.5 percentage points) indicates the ATE from comparing callbacks to immigrant-origin applicants wearing a headscarf with majority female candidates. The fifth and final estimate reveals a difference of 16.9 percentage points between male immigrant-origin minority applicants and male majority applicants.
Overall, discrimination does not appear to be homogeneous across out-groups, but rather de-pends on the specific configuration of group categories. Male immigrant-origin minorities are subject to significantly more discrimination than female immigrant-origin minorities, albeit when immigrant-origin females wear a headscarf, the effect of having a minority name is amplified. The results are consistent with the notion that stereotypes about ethnic minorities are mainly ascribed to males (Eagly and Kite 1987), findings from survey experiments that indicate that perceptions
1 While one should generally be cautious when comparing results from correspondence experiments (results are sensitive to context, design, job types etc.), the experimental procedures in the two studies are almost identical; this is why they can be integrated into a combined analysis.
Chapter Four: Results
Discrimination is hierarchical
In Chapter Two, I proposed that we should expect group-based biases to increase as individuals appear more socially distant and when multiple out-group categories coincide. Articles A and B includes a variety of group configurations in order to test this proposition. That is, along the random assignment of ethnic affiliation, I manipulated gender, socioeconomic status, information about the applicants’ spare time interests, and whether the applicants were depicted wearing a headscarf in a CV photo.
Figure 1 summarizes the results from articles A and B across the main treatment groups.1 The figure depicts the average treatment effects (ATE) of having an immigrant-origin name ranging from 4.7 percentage points to 16.9 percentage points depending on the applicants’ other characteristics.
The figure reports five difference-in-means estimates using invitations to job interviews as the outcome variable. The first estimate depicts a difference between immigrant-origin minority and majority applicants of 5.1 percentage points for female applicants. This is a precision-weighted estimate derived from the average treatment effect of 5.7 percentage points in article A (second estimate) and an ATE of 4.7 percentage points in article B (third estimate). The fourth estimate (11.5 percentage points) indicates the ATE from comparing callbacks to immigrant-origin applicants wearing a headscarf with majority female candidates. The fifth and final estimate reveals a difference of 16.9 percentage points between male immigrant-origin minority applicants and male majority applicants.
Overall, discrimination does not appear to be homogeneous across out-groups, but rather de-pends on the specific configuration of group categories. Male immigrant-origin minorities are subject to significantly more discrimination than female immigrant-origin minorities, albeit when immigrant-origin females wear a headscarf, the effect of having a minority name is amplified. The results are consistent with the notion that stereotypes about ethnic minorities are mainly ascribed to males (Eagly and Kite 1987), findings from survey experiments that indicate that perceptions
1 While one should generally be cautious when comparing results from correspondence experiments (results are sensitive to context, design, job types etc.), the experimental procedures in the two studies are almost identical; this is why they can be integrated into a combined analysis.
Chapter Four: Results
of cultural dissimilarity have negative effects on attitudes (Hainmueller and Hopkins 2015; Ostfeld 2017) and that perceptions of cultural dissimilarity with ethnic minorities are reinforced when sub-jects are introduced to a male and/or Muslim individual (as opposed to a female and/or non-Muslim individual) (Canan and Foroutan 2016; Adida, Laitin, and Valfort 2016). Notably, among majority applicants there is no differential treatment across gender which indicates that the gendered as-pect of discrimination among immigrant minorities is not driven by a general preference for hiring women. On a more general level, the results are consistent with the notion that some group cate-gories and specific combinations of catecate-gories can motivate different behaviours. The findings also underline an important methodological point, namely the importance of carefully considering proxy markers for minority group membership. Researchers examining discrimination against out-groups should attend to the fact that some characteristics strengthen the risk of discrimination.
Chapter Four: Results
Figure 1. Differences in callbacks between groups (articles (A) and (B))
Note: The Figure plots the average effect of having an immigrant-origin minority name as compared to a majority name on invitations to job interviews when applicants are female (estimates 1-3), wears a headscarf (estimate 4) and are males (estimate 5). The first estimate is a precision-weighted pooled estimate derived from articles A (2018) and B (2019). 90% and 95% CI are included. N = 2150.
Discrimination is consistent across sectors
During the process of collecting data, I read more than a thousand job ads. Frequently included in these ads was the statement that ’all applicants, unconditioned on their religious affiliation, ethnic-ity, gender or sexual orientation, are encouraged to apply.’ In fact, this proposal is integrated in all public sector job ads. Nonetheless, as shown in Figure 2, discrimination is consistent across private and public sector jobs. A precision-weighted estimate that combines the results from articles A and B indicates that public and private employers exhibit biases against immigrant-origin minority applicants to roughly the same extent. This finding contradicts a belief, commonly held, that public sector organizations are more egalitarian and discriminate less than do private workplaces (see for example Midtbøen (2015)). It also raises questions regarding the effectiveness of organizational procedures such as formalized hiring procedures and the use of job templates – typical features of public sector hiring practices – as mechanisms for combatting discrimination in hiring as is often suggested in the literature (Pager and Shepherd 2008; Moulton 1990).
Chapter Four: Results
Figure 1. Differences in callbacks between groups (articles (A) and (B))
Note: The Figure plots the average effect of having an immigrant-origin minority name as compared to a majority name on invitations to job interviews when applicants are female (estimates 1-3), wears a headscarf (estimate 4) and are males (estimate 5). The first estimate is a precision-weighted pooled estimate derived from articles A (2018) and B (2019). 90% and 95% CI are included. N = 2150.
Discrimination is consistent across sectors
During the process of collecting data, I read more than a thousand job ads. Frequently included in these ads was the statement that ’all applicants, unconditioned on their religious affiliation, ethnic-ity, gender or sexual orientation, are encouraged to apply.’ In fact, this proposal is integrated in all public sector job ads. Nonetheless, as shown in Figure 2, discrimination is consistent across private and public sector jobs. A precision-weighted estimate that combines the results from articles A and B indicates that public and private employers exhibit biases against immigrant-origin minority applicants to roughly the same extent. This finding contradicts a belief, commonly held, that public sector organizations are more egalitarian and discriminate less than do private workplaces (see for example Midtbøen (2015)). It also raises questions regarding the effectiveness of organizational procedures such as formalized hiring procedures and the use of job templates – typical features of public sector hiring practices – as mechanisms for combatting discrimination in hiring as is often suggested in the literature (Pager and Shepherd 2008; Moulton 1990).
Chapter Four: Results
Figure 2. Average treatment effects across sectors in articles (A) and (B)
Note: The Figure plots the average treatment effect of having an immigrant-origin minority name as compared to a majority name on invitations to job interviews across public and private sector jobs. The first estimate is a precision-weighted pooled estimate derived from all treatment groups in articles A (2018) and B (2019). 90% and 95% CI are included.N = 2395.
Immutability of ethnic discrimination
Departing from the findings of discrimination, I explore what options, if any, immigrant-origin minorities have to avoid discrimination. In Chapter Two, I suggested that if individuating cognitive processes that counter stereotypes are activated, discrimination should be curtailed. To explore this notion, in research article B, I randomly assigned information about applicants’ adherence to cultural norms (i.e. civil engagement, membership of local sports club, a relationship with a native Dane) and whether the immigrant-origin minority applicant wore a headscarf. Thus, the research article raises the question of whether discrimination is moderated by the extent to which a minority individual appears culturally deviant. The results in article B indicate that immigrant-origin minorities cannot affect their employment chances by providing implicit information that signals cultural adherence. While an explicit and visual signal (wearing a headscarf) increases discrimination significantly, applicants gain nothing by signalling that they are culturally proximate.
A potential explanation for this null-finding is, that employers rely on a ’lexicographic search’
whereby they stop reading once they see the applicants’ name or picture and therefore fail to see all credentials (Bertrand and Mullainathan 2004). Nonetheless, this finding demonstrates that the
Chapter Four: Results
effects of fixed traits like ethnicity and gender are difficult to circumvent, and that immigrant-origin applicants have few tools at their disposal to escape discrimination.
Group-based biases and political representation
Articles C and D contribute to longstanding debates in political science regarding the factors shap-ing and the implications of descriptive representation. A basic principle of democracy is the claim that citizens’ preferences should count equally in the realm of politics. However, as outlined in the following sections the research articles indicate that group-based biases significantly hamper immigrant-origin minorities access to political representation and influence in the context of Dan-ish local municipality councils.
Politicians exhibit ethnocentric behaviour – even when they face strong electoral incentives As discussed in Chapter Two, a U.S-based body of literature indicates that politicians are more likely to answer constituents who share their ascriptive traits. Article C,Who is responsive, engages with this literature by exploring how Danish local incumbents respond to constituent requests asking for help with locating their polling station for the upcoming election. As depicted in Figure 3, immigrant-origin constituents are significantly less likely (17.2 percentage points) to receive an answer from ethnic majority incumbents, whereas the reverse pattern is true for immigrant-origin incumbents responding to ethnic majority constituents (19.8 percentage points). This also manifests in responses of lower quality to requests from out-group constituents.2 Overall, this finding aligns with evidence from the American context (Mendez and Grose 2018; Butler 2014; Gell-Redman et al.
2018; Butler and Broockman 2011). The strong pattern of ethnocentrism in responsiveness among legislators in Danish local councils suggests that ethnocentric responsiveness is a more general phenomenon that extends to settings with much less ethnicized politics.
As previously discussed, recent work on representation points to two mechanisms that could explain why representatives are less responsive to immigrant-origin constituents (i.e. personal
pref-2 Specifically, when the requester holds an immigrant-origin minority name, the replies are shorter, less timely, significantly less friendly, less likely to provide an answer to the question, and less likely to include an invitation to follow up.
Chapter Four: Results
effects of fixed traits like ethnicity and gender are difficult to circumvent, and that immigrant-origin applicants have few tools at their disposal to escape discrimination.
Group-based biases and political representation
Articles C and D contribute to longstanding debates in political science regarding the factors shap-ing and the implications of descriptive representation. A basic principle of democracy is the claim that citizens’ preferences should count equally in the realm of politics. However, as outlined in the following sections the research articles indicate that group-based biases significantly hamper immigrant-origin minorities access to political representation and influence in the context of Dan-ish local municipality councils.
Politicians exhibit ethnocentric behaviour – even when they face strong electoral incentives As discussed in Chapter Two, a U.S-based body of literature indicates that politicians are more likely to answer constituents who share their ascriptive traits. Article C,Who is responsive, engages with this literature by exploring how Danish local incumbents respond to constituent requests asking for help with locating their polling station for the upcoming election. As depicted in Figure 3, immigrant-origin constituents are significantly less likely (17.2 percentage points) to receive an answer from ethnic majority incumbents, whereas the reverse pattern is true for immigrant-origin incumbents responding to ethnic majority constituents (19.8 percentage points). This also manifests in responses of lower quality to requests from out-group constituents.2 Overall, this finding aligns with evidence from the American context (Mendez and Grose 2018; Butler 2014; Gell-Redman et al.
2018; Butler and Broockman 2011). The strong pattern of ethnocentrism in responsiveness among legislators in Danish local councils suggests that ethnocentric responsiveness is a more general phenomenon that extends to settings with much less ethnicized politics.
As previously discussed, recent work on representation points to two mechanisms that could explain why representatives are less responsive to immigrant-origin constituents (i.e. personal
pref-2 Specifically, when the requester holds an immigrant-origin minority name, the replies are shorter, less timely, significantly less friendly, less likely to provide an answer to the question, and less likely to include an invitation to follow up.
Chapter Four: Results
erences and strategic motivations). Departing from this literature, we ask how, if at all, such tendencies can be curbed. More specifically, we scrutinize two different mechanisms for obtain-ing equal responsiveness among legislators. First, we test how majority legislators’ responsiveness varied with increasingly strong electoral incentives by randomly assigning legislators an explicit expression of intent to vote in that legislator’s favour. Moreover, we merge individual legislator responses with auxiliary data on their electoral performance (seat-winning margin) and their de-cision to run for re-election in an upcoming race. Overall, the results demonstrate that in-group favouritism in constituency service occurs even when there are clear strategic incentives for incum-bents to respond to voters. Figure 4 shows that when an explicit statement of intention to vote for the legislator is included in the request, the likelihood of responsiveness increases; however, the differential treatment is not mitigated. If anything, the in-group favouritism increases slightly when given an explicit vote cue as indicated by a negative but statistically insignificant interaction term. The article demonstrates that this is true for other types of electoral incentives as well.
This finding speaks directly against the notion, that if elections are sufficiently competitive it will animate office-seeking legislators to be responsive to their constituents independent of their ethnic background.
Figure 3. Effects of the immigrant-origin alias on responsiveness among ethnic majority and minority incumbents
Note: The figure plots the average treatment effect of signing a request to incumbents with an immigrant-origin name compared to an ethnic majority name. The outcome is receiving an answer. Treatment effects are provided for ethnic majority and ethnic minority politicians. N = 2395.
Chapter Four: Results
Figure 4. Responsiveness across constituents’ alias and inclusion of explicit voting cue
Note: The figure plots the average response rate to voters with a majority or a minority alias conditional on whether a personal voting cue was included in the requests. Finally, the difference-in-differences between these groups is reported as percentage points.N = 2395.
Incumbents’ political preferences vis-´a-vis immigration policies predict their personal biases The third contribution of article C is an examination of the strategies by which minority voters can identify politicians who are more responsive to them. Specifically, we consider politicians’
political preferences on questions that relate to integration and immigration policies by merging the experimental data with politicians’ stated preferences, gauged via a voting advice application.
Hence, we were able to examine whether politicians that explicitly favour a liberal immigratrion policy are also less likely to exhibit biases against immigrant-origin constituents.
Our findings provide compelling evidence that immigrant-origin constituents are able to iden-tify politicians that are less discriminatory by paying attention to their policy preferences. We first demonstrate that immigrant-origin minority constituents are significantly more likely to be provided an answer when they contact incumbents from parties that advocate liberal immigration and integration policies. We then show that even among candidates within the same party,
indi-Chapter Four: Results
Figure 4. Responsiveness across constituents’ alias and inclusion of explicit voting cue
Note: The figure plots the average response rate to voters with a majority or a minority alias conditional on whether a personal voting cue was included in the requests. Finally, the difference-in-differences between these groups is reported as percentage points.N = 2395.
Incumbents’ political preferences vis-´a-vis immigration policies predict their personal biases The third contribution of article C is an examination of the strategies by which minority voters can identify politicians who are more responsive to them. Specifically, we consider politicians’
political preferences on questions that relate to integration and immigration policies by merging the experimental data with politicians’ stated preferences, gauged via a voting advice application.
Hence, we were able to examine whether politicians that explicitly favour a liberal immigratrion policy are also less likely to exhibit biases against immigrant-origin constituents.
Our findings provide compelling evidence that immigrant-origin constituents are able to iden-tify politicians that are less discriminatory by paying attention to their policy preferences. We first demonstrate that immigrant-origin minority constituents are significantly more likely to be provided an answer when they contact incumbents from parties that advocate liberal immigration and integration policies. We then show that even among candidates within the same party,