Ethnic discrimination on paper: uncovering realtors’ willingness to discriminate with mystery mails

ABSTRACT

Introducing a new method, we measure real estate agents’ intentions to act upon a discriminatory request on the housing market. Our results show that realtors are willing to discriminate on request and that the method of mystery mails allows scholars to measure and contextualize these discriminatory intentions in a detailed and elaborate way.

KEYWORDS:

• Ethnic discrimination
• housing market
• mystery mails
• experimental design

JEL classification:

• C93
• J15
• J71

I. Introduction

Research has previously shown persistent rental discrimination against ethnic minorities in both urban and rural areas through the use of audit or correspondence tests (Auspurg, Schneck, and Hinz 2019; Flage 2018). In these field experiments, two (fictitious) candidates – who only differ from each other in their ethnic origin – apply for real rental advertisements. Subsequently, it is assessed whether landlords and real estate agents treat these candidates unequally or not. Audit and correspondence tests thus measure discriminatory behaviour, evaluating whether landlords and realtors commit acts of discrimination in a real-world selection process (Gaddis 2018).

Even though correspondence tests are increasingly used in field experimental research, some limitations remain. First, actions prior to the unequal treatment are unmeasurable and hence not taken into account. Therefore, it is often hard to distinguish between the different drivers of discrimination. Second, correspondence tests lack the ability to identify later barriers that hinder ethnic minorities in finding a dwelling. For example, a realtor could initially respond positively to a request of a minority applicant to visit a dwelling, but at the same time impose higher barriers in later stages of the selection process (see e.g. (Hanson, Hawley, and Taylor 2011).

A potential alternative method to audit or correspondence tests is the mystery call, which measures the intention to act upon a discriminatory request rather than discriminatory behaviour itself (Verstraete and Verhaeghe 2020). Mystery calls are operationalized as follows: a realtor receives a call from a fictitious dwelling owner with the question of whether it is possible to rent out their dwelling through the realtor’s agency. Subsequently, the owner makes a discriminatory request – for example, by asking not to let the dwelling to foreigners. This method allows researchers to (i) measure to which extent the realtor intends to accommodate a discriminatory request, (ii) record the sometimes more subtle aspects of discrimination, and (iii) identify ethnic barriers in later stages of the rental process.

However, limitations regarding the mystery calls method exist as well. First, these calls are often performed by different actors. Even though these actors use standardized scripts to eliminate possible between-caller differences, there remains a potential for inter-rater bias (Verstraete and Verhaeghe 2020). For example, the volume of the caller’s voices could differ, or the realtor’s answers could be categorized in different ways depending on the callers’ interpretation. Second, in line with Heckman and Siegelman’s (1993) critique on audit studies, mystery calls are not double-blind, which raises the possibility that mystery callers alter their behaviour to actively scout for the agents’ willingness to discriminate. Third, performing mystery calls is time-consuming in itself and outsourcing these calls to an external company can be costly.

In light of the shortcomings of correspondence tests and mystery calls, we propose that researchers explore the possibility of using the mystery mails method as a novel method to evaluate discriminatory intentions. First, by standardizing the discriminatory request in a written (email) format, inter-rater bias is minimized and answers become comparable. Second, mystery mails can easily be sent out in large batches and are therefore cost-effective. Third, while we admit that realtors could behave in a more socially desirable manner because their answers are in writing, their responses will most likely be more elaborate and could therefore provide us with (i) a better insight in their legitimations and (ii) additional context to their intent – unlike the binary outcome of the classic correspondence test.

In this study, we measure the agent’s willingness to discriminate. This does not necessarily imply a personal distaste; the agent could merely be acting upon the discriminatory request of the client. In fact, the willingness could even be considered economically rational (though still morally undesirable). Only when willingness results in actual discriminatory behaviour, one could speak of owner-driven discrimination. Nonetheless, even then, it could be that the driver for this behaviour is purely economical. In other words, an otherwise non-discriminatory agent may, independent of their own preferences, accommodate a client’s discriminatory intention if this does not come at a cost, or if the cost is outweighed by the potential advantage of not missing out on a business opportunity.

II. Method

Our mystery mails experiment consisted of a scenario that was similar to that of Verstraete and Verhaeghe’s (2020) mystery calls experiment. This enabled us to compare the response rates and the willingness to accept the discriminatory request in our study to those reported in their article. In the email itself, a fictitious client first asked the realtor whether they could help out with some practicalities of renting out a house. The client then asked whether the agency could help with finding the right tenant and subsequently mentioned that they would rather not rent out to foreigners. Additionally, the client mentioned that ‘friends of ours’ had bad experiences with foreigners in the past. The emails were sent out from an email address that signalled a Flemish name. A second email was sent out when the realtor merely requested the fictitious client to call back or visit their office to further discuss their request. In this second email, the client asked whether the discriminating request could be discussed more comprehensively.

Emails were sent out to 200 realtors who rented out property in the cities of Gent and Antwerp (Belgium) in February of 2020.¹ Previous research shows that ethnic minority candidates face substantial discrimination in search for housing in these cities, with discrimination rates that fluctuate around 20% (Verhaeghe and Ghekiere 2020; Verhaeghe, Martiniello, and Ghekiere 2020). The realtors in this study were randomly selected through the popular real estate website immoweb.be. We received a total of 123 valid responses within a week, which equals to a response rate of 61,5%. A mere 20 cases were invalid because these realtors worked in different geographical areas. In the remaining 57 cases, the realtor did not respond to our email – these cases were excluded from the analysis.

III. Results

Agents willing to discriminate

As shown in Table 1, the first category consists of answers that show a direct intention to act upon the discriminatory request of the client. More concretely, in 5,6% of the answers, the client’s request to not rent out to foreigners is addressed by the realtor. This indicates a clear intention to act upon the request and to be selective on ethnic grounds when looking for possible rental candidates. The proportion found in this study (through mystery mails) is much lower than the proportions found in previous Belgian research using mystery calls (Van den Broeck and Heylen 2015; Verstraete and Verhaeghe 2020). Possibly because the risk of expressing the willingness to discriminate is higher when written on paper than answered over the phone. Most of these answers seem to have a rather cautious way of agreeing to the request, for example: ‘We obviously take your wishes into account. You are the one who decides who rents your place’.

Table 1. Categorization of the answers to the mystery mails (n = 123).

Type	Sub-type	Proportion
Willingness to discriminate	Direct	5.6%
	Indirect	5.6%
No willingness to discriminate	Direct	27.3%
	Indirect	20.0%
No response to discriminatory question		18.5%
Request to call back		12.8%
No property rental		10.2%

The second category, also 5.6% of our sample, consists of answers that show an indirect intention to act upon the discriminatory request of the client. These answers are indirect because the realtor proposes to filter the candidates in a later phase of the rental process. More concretely, they propose to be selective based on non-ethnic, subjective criteria which can nevertheless be used to exclude ethnic minorities. Examples of these criteria are personal experience, morality, language and appearance. However, the assessment of a candidate based on subjective measures could (un)consciously lead to discrimination in a later phase of the application process (Quillian 2006).

Agents not willing to discriminate

A third category, which sums up to 27.3% of the answers, comprises the explicit, direct refusals to act upon the discriminatory request. These mostly included a reference to anti-discrimination law or the realtor’s deontology that prohibits them from discriminating.

A final group of realtors would not directly address the discriminatory request of the client but would clear the pathway for discrimination by the client. These realtors express no willingness to discriminate but explain to the client that the eventual choice is theirs to make. By voicing this option, discrimination can still occur in later phases of the rental process, not by the broker but by the client. While, in this category, there is no intention by the realtor to act upon the client’s request, there is an indirect, facilitative nature to these answers. For example: ‘You as the owner always have the final choice of tenant. We cannot exclude anyone on visits but respect the owner’s choice’.

Finally, about one in six (18.5%) did not respond to the discriminatory request. This happened mostly in combination with an invitation to the realtor’s office. Similarly, about one in seven (12.8%) asked to call for further discussion and to comment on the request. We could have categorized these answers as an indirect willingness to discriminate, as the realtor would possibly discuss the discriminatory question on the phone or in person. However, the emails show no sign that later contact would result in a positive or negative answer to the request. Therefore, we did not categorize these answers as such. Finally, about one in ten did not rent out but was only interested in selling the dwelling.

IV. Discussion and conclusion

We constructed a new method to study discriminatory intentions, which we conducted among 200 real estate agents in Flanders, Belgium. Unlike classic correspondence tests, our research shows that the method of mystery mails allows scholars to measure and contextualize the intention to accommodate a discriminatory request in a more detailed way. In contrast to mystery calls, they allow us to conduct the research in a more standardized, less costly and more elaborate way. Our results provide evidence that more than one in ten realtors expresses their willingness to discriminate by email. However, because of the social desirability bias that is intrinsic to this method, we expect that this result is an underrepresentation of the willingness to discriminate and the actual discriminatory behaviour in later stages of the rental process. Furthermore, we could only speculate to what extent this willingness to discriminate results in actual behaviour. We include some recommendations for future research. First, future research should study if these direct or indirect intentions constitute accurate predictors of discriminatory behaviour in the later rental process, by linking mystery mails to correspondence tests. Second, this study only shows a small proportion of the issue, as the focus only lays on real estate agents and not on private property owners. Studies should uncover the private landlords’ intentions equally, for little is known about this group but they are at least of equal importance. Third, to get a clearer understanding of the mechanisms behind the intentions, future research should study the effect on intentions when potential income could be generated by differentiating between candidates based on ethnicity, by for example asking if the realtor would be willing to ask a higher price from the ethnic minority candidate.

Finally, the main limitation of the mystery mails method is the intrinsic bias towards non-discriminatory intentions. Because the answers to the mystery mails are in written form, our results possibly underestimate the level of intent to act upon the discriminatory request. For this reason, mystery mails cannot be used as a full substitute for mystery calls. However, the method does allow us to understand, compare and contextualize the various intentions expressed by the agents.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by SBO-FWO under Grant Research Foundation Flanders [S004119N].

Notes

¹ The average share of ethnic minorities in the neighbourhoods where the dwelling was located equalled to 16.3% and 19.1% – compared to the city average of 14.5% and 21.6% – for Ghent and Antwerp, respectively (Buurtmonitor 2020).