Ethnic discrimination in contacts with public authorities: a correspondence test among Swedish municipalities

ABSTRACT

We present a field experiment conducted in order to explore the existence of ethnic discrimination in contact with public authorities. Two fictitious parents, one with a Swedish-sounding name and one with an Arabic-sounding name, sent email inquiries to all Swedish municipalities asking for information about preschool admission for their children. Results show that the parents were treated differently by the municipalities since the individual with the Swedish-sounding name received significantly more responses that answered the question in the inquiry than the individual with the Arabic-sounding name. Also, the individual with the Swedish-sounding name received more warm answers than the individual with the Arabic-sounding name in the sense that the answer from the municipality started with a personal salutation. We conclude that ethnic discrimination is prevalent in public sector contacts and that this discrimination has implications for the integration of immigrants and their children.

KEYWORDS:

• Discrimination
• subtle discrimination
• field experiment
• authorities

JEL CLASSIFICATION:

• J15
• C93
• H83

I. Introduction

Today, there is a relatively large body of research that has used field experiments in order to detect ethnic and other forms of discrimination in different markets (Bertrand and Duflo 2017; Riach and Rich, 2002). While much attention has been paid to the existence of ethnic discrimination in labour and housing markets, less attention has been paid to the extent to which ethnic discrimination exists in contacts with public authorities. In the US, Giuletti, Tonin, and Vlassopoulos (2019) carried out an email correspondence study with queries to about 19,000 local public service providers and documented discrimination against black email senders. In Sweden, Adman and Jansson (2017) conducted a field experiment and found that municipality responses were significantly less friendlier and welcoming to individuals with Arabic-sounding names than to individuals with Swedish-sounding names who sent inquiries regarding preschool admission of children.

Since previous field experiments has shown that individuals with Arabic-sounding names are discriminated against in the labour as well as in the housing market in Sweden, the country is for sure a suitable ground for a test of ethnic discrimination by public authorities (Ahmed, Andersson, and Hammarstedt 2010; Ahmed and Hammarstedt 2008; Bursell 2014; Carlsson and Rooth 2007). We therefore conduct a field experiment of ethnic discrimination in contact with Swedish public authorities. The experiment was possible to implement by the fact that the Swedish municipalities according to the Education Act are responsible for arranging preschool for children from the age of one year (Swedish Code of Statutes 2010, 800).

Our experimental set up is closely related to Adman and Jansson (2017). We test the extent to which public authorities treat people with Arabic-sounding names differently than people with Swedish-sounding names by sending fictive inquiries to all Swedish municipalities regarding preschool admission of children. Besides replicating the study by Adman and Jansson (2017), we improve their work in different ways. First, Adman and Jansson (2017) included both male and female applicants in their study. Since there are only 290 municipalities in Sweden, their experiment resulted in low statistical power. By using only male applicants we increase the statistical power in our experiment. Second, the names in Adman and Jansson (2017) where chosen arbitrarily. Since one may overstate the magnitude of the discrimination by using Arabic-sounding names that are difficult to pronounce, we use Swedish- and Arabic-sounding names that belong to the 100 most common first names and surnames in Sweden.

II. Method

We used a randomized correspondence test to examine whether municipalities in Sweden treat people of different ethnic backgrounds unequally in a situation where they are asked to provide information about a simple and usual matter. Each municipality received an inquiry from one out of two fictitious and ethnically distinct parents regarding preschool admission of their child. The ethnicity of the parent was conveyed through the name, using either a Swedish-sounding or an Arabic-sounding male name. The names were randomly assigned to each inquiry that was made to municipalities. The experiment was conducted during Spring 2019. We sent one inquiry to all 290 municipalities in Sweden through email in a single day. The email addresses to the municipalities were collected from their official websites. Three of the municipalities’ email addresses, however, malfunctioned during the day of the experiment. These municipalities were, therefore, excluded, and as a result, the final data set subject to analysis consisted of 287 municipalities.¹

Municipality responses were coded in terms of four outcome variables: Contact, Explanation, Wordcount, and Salutation. Contact was a binary variable indicating whether a parent received a response from a municipality, regardless of the content of the response. Contact included any type of response, i.e., even autogenerated replies. Explanation was a binary variable indicating whether a parent received a response from a municipality that answered the parent’s question in a satisfactory way. Wordcount was a continuous variable that simply measured the total number of words in a municipality response that contained a satisfactory explanation. Finally, Salutation was a binary variable indicating whether a municipality response containing a satisfactory explanation was initiated with a salutation in combination with the parent’s name (e.g., Dear The father’s name).

The parent that made the inquiry to the municipality was the father in the family, and a five-year-old boy in the family needed a preschool admission. The father asked the municipality about how he can find information about available preschools in the municipality and how he can place his son in the municipality’s preschool que. An email account was created for the fictitious father through which we sent his inquiry to the municipalities. The inquiry consisted of a brief email message expressing the father’s need for information (English translation):

“Hi, we are a family with a five-year-old son who needs to be admitted to a preschool. How do we get more information about this, and how do we put him in the que for preschool? Kindly, [The father’s name]”

The name of the father was randomly assigned to each inquiry. We chose two distinctive given male names with clear ethnic connotations: Mattias Hansson, a typical Swedish-sounding male name, and Mohamed Hassan, a typical Arabic-sounding male name.

III. Results

Table 1 presents the percentage of preschool inquiries that resulted in a response from the municipalities. The parent with a Swedish-sounding name achieved a contact with the municipalities in almost all cases except for one, i.e., in 99 percent of the time. Similarly, the parent with an Arabic-sounding name achieved a contact in 98 percent of the time. The small difference in contact rates is not statistically significant, χ²(1, N = 287) = 1.06, p = .304.

Table 1. Differences in municipality response Rates.

	Swedish-sounding name	Arabic-sounding name	Hypothesis Testing
Contact	99.3%	97.9%	χ^2(1, N = 287) = 1.06,
	(144)	(139)	p = 0.304
Explanation	97.2%	91.5%	χ^2(1, N = 287) = 4.42,
	(141)	(130)	p = 0.036
Number of	145	142
municipalities

Note: Contact is simply a response from the municipalities, regardless of the content. Explanation is a response from the municipalities that contained satisfactory answers to the inquirers’ questions. The number of municipalities is given in parentheses.

Next, we count the number responses that contained a satisfying answer to the parents’ questions (either directly in a first reply or in a later one). Table 1 shows that the parent with a Swedish-sounding name received a response with a satisfying answer in 97 percent of the cases. The parent with an Arabic-sounding name received a response with a satisfying explanation in less than 92 percent of the cases. Hence, the parent with a Swedish-sounding name received 6 percent more satisfying responses than the parent with an Arabic-sounding name. This difference is statistically significant, χ²(1, N = 287) = 4.42, p < .05.

Table 2 provides results for two outcome variables, Wordcount and Salutation. These measures are examined for those municipalities that actually provided an acceptable explanation to the parents’ inquiries (i.e., the 271 municipalities that appeared under Explanation in Table 1). Wordcount gives the mean number of words used by municipalities in their explanations to our parents. Salutation is the percentage of municipalities that started their letter with a personal salutation, i.e., a salutation in combination with the parent’s name.

Table 2. Differences in municipality response quality.

	Swedish-sounding name	Arabic-sounding name	Hypothesis Testing
Wordcount	59.0	54.5	t(269) = 1.02,
	(36.96)	(36.65)	p = 0.310
Salutation	48.9%	23.8%	χ^2(1, N = 271) = 18.29,
	(69)	(31)	p < 0.001
Number of municipalities	141	130

Note: Wordcount is the mean number of words in letters, received by inquirers. Salutation is the percentage of municipality response letters that started with a salutation in combination with an inquirer’s name. Standard deviations and actual number of cases are given in parentheses for Wordcount and Salutation, respectively. The number of municipalities in this table consists of those municipalities that provided the inquirers an explanation (see Table 1).

Table 2 shows that the mean number of words used by municipalities in their explanations was 59 and 55 when the parent had a Swedish-sounding and Arabic-sounding name, respectively. The small difference in the wordcount of municipalities’ explanation is not statistically significant, t(269) = 1.02, p = .310.

Finally, we look at named salutations in the municipalities’ responses. Almost half of those municipalities that responded to the parent with a Swedish-sounding name, with a satisfying explanation, initiated their letter with a salutation in combination with the parent’s name. Less than one-fourth of the municipalities that responded to the parent with an Arabic-sounding name, with a satisfying explanation, did the same. Thus, we find a statistically significant difference in personal salutations, χ²(1, N = 271) = 18.29, p < 0.001. While both inquirers received the same amount of information from the municipalities, there was a considerable difference in treatment as regards amiability.²

IV. Conclusions

We have conducted a field experiment of ethnic discrimination against individuals with Arabic-sounding names in contacts with public authorities. Our results reveal that ethnic discrimination exists and that it has different dimensions. We conclude that discrimination in contacts with public authorities, in different ways, is an obstacle for certain ethnic groups. As regards limited access to preschool admission, this may have a negative impact on the parent’s labour supply. Research has also shown that preschool attendance is positively related to an individual’s years of schooling and highest educational degree completed later in life. Furthermore, it is also positively related to an individual’s employment and earnings as an adult (Dietrichson, Kristiansen, and Nielsen 2018) and to health (Aalto et al. 2019). Finally, research indicates that preschool attendance seem to be an efficient way to improve language knowledge among children of immigrants (Drange 2018). Thus, limited access to preschool attendance may be an obstacle for the integration of immigrants and their children in different ways.

Acknowledgments

Financial support from the Swedish Research Council, the Marianne and Marcus Wallenberg Foundation and Jan Wallander och Tom Hedelius stiftelse are gratefully acknowledged.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the Jan Wallanders och Tom Hedelius Stiftelse samt Tore Browaldhs Stiftelse;Vetenskapsrådet [2018-03487].

Notes

¹ The final anonymized data are openly available in Zenodo at https://doi.org/10.5281/zenodo.2613067.

² We have also conducted a number of regression analyses with Contact, Explanation, Wordcount and Salutation and the dependent variables. The results from the regressions are in line with the results in Tables 1 and 2.