M AGNUS C ARLSSON & D AN -O LOF R OOTH
2012-6
Can media attention about tests of ethnic discrimination change the employers’
behavior?
Abstract
Situation testing is used to investigate whether employers discriminate when hiring.
This study analyzes whether authorities can implement such situation tests to hinder ethnic discrimination and enforce anti-discrimination legislation more effectively by taking advantage of the opportunities provided by news media coverage. To this end, we use unique data that relates an exogenous shock, in the form of extensive media coverage of situation testing conducted in the labor market, to data on employers´ actual discriminatory behavior collected in two situation-testing field experiments. The media coverage, which luckily occurred in the middle of these experiments, and implementation of a difference-in-difference methodology make a causal interpretation of the results possible. Based on previous studies that have shown how the news media can affect economic outcomes, the hypothesis here is that such information, as the news media provided in this case, leads to a decline in the degree of discrimination in the labor market. However, the results reveal no sign of employers changing their hiring practices after the media coverage.
Contact information
Magnus Carlsson Linnaeus University
Centre for Labour market and Discrimination Studies SE- 391 82 Kalmar
magnus.carlsson@lnu.se Dan-Olof Rooth
Linnaeus University
Centre for Labour market and Discrimination Studies SE- 391 82 Kalmar
dan-olof.rooth@lnu.se
1. Introduction
In the spirit of the journalist Günter Wallraff's work, contemporary researchers have relied on using field experiments to test for the existence of ethnic discrimination in recruitment.
1Such real life, or situation, testing implies sending two qualitatively identical fictitious job applications to employers advertising for labor – the only difference being the name of the applicant with one having a majority, and the other a minority, sounding name. The degree of discrimination is then quantified by calculating the difference in the probability of being invited to a job interview for the two groups of applicants. The main advantage of this method is that, by construction, it guarantees that the researcher and the employer will observe the same characteristics of the applicant, and any difference in the probability of being invited to a job interview must be because of employers’ reaction to the name of the applicant when hiring.
The EU has recently shown renewed interest in using such tests to hinder ethnic discrimination in employment situations and to enforce anti-discrimination legislation.
For instance, in the UK, the Equality and Human Rights Commission has the right to conduct situation tests and even to prosecute firms.
2Other European countries, such as Sweden, have for long been reluctant to use situation tests as a tool for counteracting discrimination. However, Swedish authorities have recently reconsidered implementing this tool for combating ethnic discrimination. An important event for this turn-around occurred in 2005 when law students initiated lawsuits against restaurants and night clubs based on situation tests of ethnic discrimination.
3This initiative was successful in that some of these night clubs were later convicted.
The policy relevance of this paper is whether authorities can take advantage of opportunities provided by the news media to more effectively implement situation tests simply by informing the employers about the testing. Previous studies have established that what is reported in the news media can affect individual behavior and ultimately economic and political outcomes as well. For example, it has been shown that goods and service markets can be affected when the news media provides information about a product, voting behavior can be influenced by what information the news media report about politicians, policy decisions can be based on how well informed the public is by the media, and health behavior of individuals can change as a result of health campaigns in the media.
4In this study, we analyze whether extensive news media coverage of discrimination testing being conducted in the labor market leads to a lower degree of ethnic discrimination in hiring. Our strategy for identifying this effect is to utilize two random
1 A considerable number of situation tests have been conducted in Europe and the US during the preceding decades, see Riach and Rich (2002) for a summary of these experiments.
2 In the US, the Equal Employment Opportunity Commission has developed a testing methodology and a procedure on how to enforce civil rights.
3 These tests were conducted at night clubs in Stockholm, Gothenburg and Malmö.
4 Berger et al (2010) find that negative publicity about a product can have negative effects on sales, but positive effects in other cases, DellaVigna and Kaplan (2007) find that the Republicans in the U.S. gained votes in towns where the conservative Fox News channel was introduced in the cable market, Strömberg (2004) finds that politicians in the United States in the 1930s targeted voters in areas where a larger share of the population had radios, and Wen et al (1995) show that health information about cholesterol and saturated fat resulted in significant increases in consumption of corn, cottonseed, and soybean oils and decreased consumption of butter and lard.
and unique events which interrupted the data collection of two previous situation test experiments in hiring (Carlsson & Rooth, 2007, and Carlsson, 2010). About halfway into both of these experiments there was extensive news media coverage of situation testing being conducted in the Swedish labor market. For instance, both Dagens Nyheter, one of the largest daily newspaper in Sweden, as well as the teletext of Swedish Television, the Swedish public service television company, had it as their headline news. Hence, employers advertising for labor after the news media coverage would have been aware that these kinds of experiments were being conducted in the Swedish labor market. Since the news media coverage was unexpected, employers were either exposed or not exposed at random.
We hypothesize that news media attention lowers the measured degree of discrimination because it changes some employers´ behavior. There are at least three potential mechanisms through which this might occur. A first possibility is that firms simply attempt to avoid negative publicity in the form of being exposed as a discriminatory employer in the news media. The rationale for this might be that such publicity can have a deteriorating effect on the sales and the performance of a firm.
A second possibility is that, as a result of the news media attention, discriminatory employers fear that authorities monitor discrimination in the labor market and that there is a risk of a sanction associated with being caught.
5Finally, this kind of attention in the news media on situation testing and ethnic discrimination might act as a public agenda setter that changes some employers´
discriminatory behavior.
6For example, the picture painted in the media might make non-prejudiced employers more thorough in the recruitment process. That would potentially reduce the degree of statistical discrimination in the recruitment process and also the incidence of implicit discrimination – a phenomenon where a person’s attitudes and stereotypes operate in an unconscious mode, typically in situations where quick decisions are made (see Rooth, 2010). Also, putting ethnic discrimination on the public agenda could influence the employers´ attitudes toward minority workers in a positive direction.
However, our results reveal no sign of employers changing their hiring practices after the news media coverage of situation testing in the labor market. This suggests that none of the mechanisms described above are in operation. In other words, information about situation testing being conducted in the labor market appears irrelevant for the employers as regards their discriminatory behavior.
The next section describes the method used, while the results are presented in Section 3, followed by a discussion in Section 4.
5 The first case of media coverage was of Swedish authorities’ plans to implement situation testing to combat ethnic discrimination in the labor market.
6 See World Bank (2002a, 2002b) for examples where the media has acted as a public agenda setter and influenced outcomes in economic markets.
2. Methods
2.1 The field experiments
The data is taken from two previous field experiments measuring the extent of ethnic discrimination in Sweden: Experiment A (Carlsson & Rooth, 2007) and Experiment B (Carlsson, 2010). In these experiments, the job applications belonged to young native males with identical levels of occupational qualifications. The only difference was the names attached to the applications – either a typical Swedish sounding name or a typical Middle Eastern sounding name.
Experiment A was conducted between May 2005 and February 2006. When jobs in selected occupations (see below) – found on the Swedish Employment Agency’s webpage – were applied for, each employer received one application with a Swedish sounding name and one with a Middle Eastern sounding name. In total, 3,228 applications were sent to 1,614 employers. The callback rates for a job interview were 28.6 percent for applicants with a Swedish sounding name and 19.0 percent for applicants with a Middle Eastern sounding name, a statistically significant difference of 9.6 percentage points.
In the case of Experiment B, data was collected between August 2006 and April 2007, following a procedure similar to the one in Experiment A. In total 2,628 applications were sent to 1,324 employers who had advertised for labor on the Swedish Employment agency’s webpage. In this experiment the callback rates for an interview were 41.5 percent for applicants with a Swedish sounding name and 24.4 percent for applicants with a Middle Eastern sounding name, a statistically significant difference of 17.1 percentage points.
7Thirteen occupations were used: computer professionals, business sales, four categories of teachers, accountants, nursing, construction, restaurants, shop sales, motor-vehicle drivers and cleaners.
8Hence, both skilled and semi/unskilled occupations were represented. Although the complete labor market was not analyzed, we applied for all the vacant jobs found on the webpage under the chosen occupations.
2.2 The news media coverage
On one occasion in experiment A and on one occasion in Experiment B the news media reported on large-scale field experiments to measure employer discrimination.
7 The applicants in Experiment B had higher levels of qualifications compared to those in Experiment A.
It appears that this higher level of qualification helped native Swedes more in receiving a callback for interview than those with a Middle Eastern sounding name, which might explain the larger ethnic difference in callback rate found in Experiment B.
8 Experiment B did not include the occupations cleaners and teachers (language). The experiments were restricted to the two major cities of Sweden: Stockholm and Gothenburg.
Figure 1. The timing of the news media coverage.
Notes: This figure illustrates when the news media coverage occurred in Experiments A and B.
As can be seen from Figure 1, of the 1,614 participating firms in Experiment A, 744 were observed before and 870 after the news media coverage. The corresponding figures for the 1,314 firms in Experiment B are 592 before and 722 after the news media coverage.
Data collection
May 2005 – 15 September 2005 n=744
Data collection
August 2006 – 4 November 2006 n=592
Media coverage
16 September 2005 Media coverage
5 November 2006
Data collection
16 September 2005 – February 2006 n=870
Data collection
5 November 2006 – April 2007 n=722
Experiment A (n=1,614): Experiment B (n=1,314):
Table 1. The extent of news media coverage during the field experiments.
Notes: This table reports news media coverage as daily circulation in the case of newspapers and as the number of readers in the case of Text-TV. For newspapers marked with (*), the sum of the total circulation for these newspapers is given. Source: Svensk Dagspress 2007 and www.svt.se/texttv
Table 1 reports the extent of the news media coverage for Experiments A and B.
Regarding Experiment A, the headline on the front page of Text-TV
9on the 16
thof September, 2005 disclosed that the authorities intended to send fictitious job applications to employers in order to measure ethnic discrimination.
10Hence, this event occurred right in the middle of the data collection. Furthermore, between the 16
thand 22
ndof September three out of nine newspapers with the largest circulation in Sweden, including the largest daily newspaper, Dagens Nyheter, published similar articles as headline news, and so did at least fourteen other local newspapers.
119 The teletext of Swedish Television.
10 2.8 million daily readers of a total Swedish population of approximately 9 million.
11 See appendix B for some examples of this media coverage.
News media Circulation Experiment A Experiment B
Largest Daily Newspapers Circulated in the Major Cities of Sweden
Aftonbladet 416 500 Yes
Dagens Nyheter 347 100 Yes Yes
Expressen 326300* Yes
GT 326300* Yes
Kvällsposten 326300* Yes
Göteborgs-Posten 242 700
Svenska Dagbladet 193 500 Yes
Sydsvenskan 123 200 Yes
Dagens Industri 117 800 Yes Yes
Free Newspapers in Stockholm and Gothenburg
Stockholm City 297 700 Yes
Metro, Stockholm 280 800 Metro, Göteborg 102 100 Other News Media
Text-TV 2 800 000 Yes
Other Local News Media - Experiment A Other Local News Media - Experiment B
ST Press Lärarnas Tidning
Arbetarbladet Hotellrevyn
Vestmanlands Läns Tidning SR Sörmland
Uppsala Nya Tidning SVT.se Nyheter
Bohuslänningen Västerbottens-Kuriren
Gefle Dagblad Ljusnan Sundsvalls Tidning Borås Tidning Helsingborgs Dagblad Borlänge Tidning Falukuriren
Södra Dalarnes Tidning Nya Ludvika Tidning
Turning to Experiment B, on the 5
thof November 2006 the headline in Dagens Nyheter was “Employers discriminate against Arabic names”, referring to the results in Carlsson
& Rooth (2007). Altogether, at least seven of the nine largest newspapers published articles on the subject over the next few days, as did at least five local newspapers and other news media. In this case, too, the news media attention occurred in the middle of the data collection.
In summary, employers advertising for labor after the news media coverage would have been aware of the news media attention to situation testing of ethnic discrimination in the Swedish labor market. Indeed, we find this quite likely since a clear majority of about a hundred firms/recruiters, contacted as part of the Rooth (2010) study, had been aware of the news media attention. Also, since the first of these employers/recruiters were contacted at least four months after the news media attention, it had a long-run impact on their awareness. However, the reader should keep in mind that this information has been collected only for the small sample interviewed.
3. Results
This section analyzes the impact of the news media coverage, which occurred during Experiments A and B, using the difference-in-differences (DD) estimator. The DD estimate is easily calculated as the difference in the probability of applicants with a Middle Eastern sounding name being invited to an interview, before and after the news media coverage, minus the corresponding difference for applicants with a Swedish sounding name. In other words, it is the difference in the time effects across the two groups that constitutes the DD estimate. The DD estimate is obtained using the linear regression equation:
T M T M
y
0 1 2 3* (1)
In this linear probability model
12,
yequals one if the applicant was invited for a job interview, and zero otherwise,
Tequals zero if the job was applied for before the news media coverage and one if it was applied for after,
Mequals zero for applicants with a Swedish sounding name and one for applicants with a Middle Eastern sounding name, while
0,
1, and
2give the corresponding parameters. The parameter of the interaction term
T *M ,
3, provides the DD-estimate of interest. We would expect this estimate to be positive, that is, that the relative callback rate of the minority increased after the news media coverage.
12 The results are more or less identical when probit models are instead estimated using the dprobit command in Stata 11.
Table 2. The impact on the callback rate for interview.
Experiment A Experiment B
(1) (2) (3) (1) (2) (3)
After news media coverage 0.00
[0.02] -0.01
[0.02] -0.01
[0.02] -0.01
[0.03] -0.01
[0.03] -0,03 [0.03]
Middle Eastern name -0.10***
[0.02] -0.10***
[0.02] -0.11***
[0.02] -0.19***
[0.02] -0.20***
[0.02] -0.20***
[0.02]
After news media coverage *
Middle Eastern name 0.00
[0.02] 0.00
[0.02] 0.02
[0.02] 0.03
[0.03] 0.03
[0.03] 0.03 [0.03]
Occupational fixed effects No Yes No No Yes No
Firm characteristics controls No No Yes No No Yes
N 3,228 3,228 2,996 2,628 2,628 2,136
Notes: This table reports the effect on the callback rate for interview (from estimated Linear Probability Models). Model 1 reports the effect on the callback rate for applying after the news media coverage, having a Middle Eastern sounding name and the interaction effect of the two, which is the DD-estimator.
Models 2 and 3 are identical to Model 1 but with occupational fixed effects and firm characteristics added. *, **, and *** denote the ten, five and one percent significance levels, respectively. Reported standard errors (in brackets) are adjusted for clustering at the workplace.
Unconditional on any covariates (Model 1), there is no evidence of monitoring having an impact on the callback rate of minority applicants, see the third row of Table 2. The DD estimate is virtually equal to zero in both experiments and the result is not altered when occupation fixed effects (Model 2) or additional firm characteristics (male share in workforce, if male recruiter, and number of employees at the workplace, Model 3) are added to Model 1. Furthermore, the results in Table 2 show no evidence of a time trend in the callback rate for interview for the majority either (see the estimate of
1in the first row). Finally, irrespective of the news media coverage, having a typical Middle Eastern sounding name, as opposed to a typical Swedish sounding name, implies a ten and twenty percentage points lower probability of being invited for a job interview in Experiments A and B, respectively (see the estimate of
2in the second row).
Although there is no support for an effect of the news media coverage on the difference in callback rate for a job interview, there could nonetheless be an impact for specific occupations or at the regional level. Therefore, Model 1 is re-estimated on each occupation separately, for only low/high skilled occupations, and separately for the Stockholm and Gothenburg labor market areas, see Tables A1 and A2 in Appendix A.
The results of these estimations confirm the findings at the aggregated level.
Before turning to the discussion of the results, a sensitivity check of the DD model is
appropriate. Suppose, for example, that there was a negative time trend in the
probability of being invited to an interview for both applicants, but the slope of this
trend was steeper for the applicant with a Middle Eastern sounding name. Then the
news media coverage might have broken this divergence in the trends, but this might
not appear in the results above since the DD model assumes parallel time trends. We
have studied this issue to some extent by dividing the pre-media-coverage period into
two equal time periods and then performing a “fake” DD type of analysis without
finding any indications of diverging time trends in receiving a callback for interview for the two applicants.
134. Discussion
The callback rate for a job interview for applicants with a Middle Eastern sounding name did not increase, compared to applicants with a Swedish sounding name, after extensive news media coverage of situation testing of ethnic discrimination conducted in the labor market. In other words, there is no sign that employers changed their hiring practices as a result of the news media coverage.
So why is that? A possible explanation for the negligible impact might be that the news media coverage itself was perhaps too small and only lasted for, at most, one week.
However, it cannot be said that the media coverage was too small, since it included almost all major news media in Sweden during the week of action. Also, the fact that almost all employers in a subsample, who were interviewed four months after the media coverage, were aware of these tests makes it improbable that a longer duration of the media coverage would have produced a different result.
Given that the news media coverage was comprehensive enough, the absence of an effect must be explained by the fact that the provided information was not important for the employers´ discriminatory behavior. This explanation is related to the three mechanisms discussed earlier and none of them seemed to be in operation.
To start with, discriminatory employers do not seem to fear being exposed in the news media and the negative consequences that might be associated with it. Perhaps this is because no individual firms were hung out during the media coverage and the extent of ethnic discrimination was mainly discussed at an aggregated level.
Moreover, discriminatory employers apparently do not fear authorities’ monitoring of discrimination in the labor market as long as there is no risk of a sanction. This is despite the fact that such a sanction exists in the Swedish legislation on ethnic discrimination and was also enforced in the aforementioned juridical case where law students investigated if ethnicity was a factor that determined who was allowed to enter restaurants and night clubs. However, in contrast to these successful law students, The Equality Ombudsman
14lost all twelve cases of ethnic discrimination by employers pursued in the Swedish Labour Court from 2003 to 2009 (Dagens Nyheter, 2009).
Therefore, discriminatory employers might still regard the risk of a sanction as small.
Lastly, the media coverage of situation testing and ethnic discrimination did not set the public agenda such that discriminatory behavior among employers declined. It seems like the news media coverage was comprehensive enough to make the employers aware of situation testing being conducted, but not to change the public agenda.
13 These results are available upon request.
14 The Equality Ombudsman is a Swedish government agency that seeks to combat discrimination in compliance with the Discrimination Act.
Given the above discussion, we believe that we have explored relevant data, and that the
absence of a causal impact of the news media coverage suggests that it is difficult to
make an anti-discrimination measure, based on situation testing, more effective by
combining it with an information campaign in the media. There might, however, be
scope for the authorities to put the second mechanism into operation by strengthening
the discrimination legislation. As long as no employers are convicted of ethnic
discrimination, they would seem to have little to fear.
References
Berger, J., Sorensen, A., and Rasmussen, S. (2010) ”Positive Effects of Negative Publicity: When Negative Reviews Increase Sales” Marketing Science 29(5): 815-827.
Carlsson, M. and Rooth, D-O. (2007), “Evidence of ethnic discrimination in the Swedish labor market using experimental data”, Labour Economics 14(4): 716-729.
Carlsson, M. (2010) “Experimental evidence of discrimination in the hiring of first- and second-generation immigrants”, Labour 24(3): 263-278.
Dagens Nyheter, 2009-02-19
DellaVigna. S. and Kaplan, E. (2007) “The Fox News Effect: Media Bias and Voting”, The Quarterly Journal of Economics 122(3): 1187-1234.
Rooth, D-O, (2010) “Automatic associations and discrimination in hiring: Real world evidence”, Labour Economics 17(3): 523-534.
Strömberg, D. (2004) "Radio's Impact on Public Spending," The Quarterly Journal of Economics, 119(1): 189-221.
The Ombudsman against ethnic discrimination (2005) ”Diskrimineringstester som bevismedel”, Stockholm
Wen S ,C., Edna T, L. and Steven T, Y. (1995) "Information, Health Risk Beliefs, and the Demand for Fats and Oils," The Review of Economics and Statistics 77(3): 555-64.
World Bank 2002a, The Right to Tell: The Role of Mass Media in Economic Development. Washington DC: World Bank.
World Bank 2002b, Building Institutions for Markets: World Development Report.
Washington DC: World Bank.
Appendices
Appendix A
Table A1. Effect on the probability of an invitation to interview.
High Skilled
(1)
Low Skilled
(2) Stockholm
(3) Gothenburg
(4) Experiment A
After news media coverage 0.01
[0.03] -0.03
[0.02] 0.01
[0.02] -0.09*
[0.05]
Middle Eastern name -0.10***
[0.03] -0.10***
[0.02] -0.09***
[0.02] -0.19***
[0.05]
After news media coverage *
Middle Eastern name -0.01
[0.03] 0.01
[0.03] -0.01
[0.02] 0.08
[0.06]
Occupational fixed effects Yes Yes Yes Yes
N 1,584 1,644 2,432 796
Experiment B After news media coverage 0.02
[0.04] -0.02
[0.03] 0.00
[0.03] 0.00
[0.06]
Middle Eastern name -0.18***
[0.03] -0.20***
[0.03] -0.21***
[0.02] -0.13***
[0.05]
After news media coverage *
Middle Eastern name 0.02
[0.04] 0.03
[0.04] 0.04
[0.03] -0.03
[0.06]
Occupational fixed effects Yes Yes Yes Yes
N 1,260 1,368 2,124 504
Notes: This table reports the effect on the callback rate for interview (from estimated Linear Probability Models; the results are almost identical when Probit models are estimated). Each column, for Experiments A and B respectively, reports the effect on the callback rate for applying after the news media coverage, having a typical Middle Eastern name and the interaction effect of the two, which is the DD-estimator. The models are estimated separately for 1) high skilled occupations, 2) low skilled occupations, 3) Stockholm and 4) Gothenburg, respectively. *, **, and
*** denote the ten, five and one percent significance levels, respectively. Reported standard errors (in brackets) are adjusted for clustering at the workplace.
Table A2. Effect on the probability of an invitation to interview.
Teachers
Accountants
Business Sales
Assistants Computer
Professionals Construction Workers††
Motor Vehicle
Drivers Nurses Preschool
Teachers Restaurant Workers
Shop Sales Assistants
Upper Sec.
School Language† Math/
Science†††
Experiment A After news media
coverage -0.03
[0.04] 0.01
[0.05] 0.12
[0.08] -0.02
[0.07] -0.07
[0.07] 0.15**
[0.07] 0.03
[0.03] -0.01
[0.03] 0.05
[0.11] -0.25*
[0.14] 0.00 [0.17]
Middle Eastern name -0.05
[0.03] -0.07*
[0.04] 0.11
[0.09] -0.12**
[0.05] -0.07*
[0.04] -0.18***
[0.06] -0.05
[0.05] -0.09**
[0.04] -0.09
[0.07] -0.33**
[0.16] -0.14 [0.09]
After news media coverage * Middle Eastern name
0.04
[0.06] -0.01
[0.05] -0.16
[0.09] -0.08
[0.07] -0.08
[0.06] 0.03
[0.07] 0.00
[0.05] -0.02
[0.05] -0.14
[0.09] 0.17
[0.20] 0.04 [0.13]
N 372 556 212 156 300 368 280 400 128 120 84
Experiment B After news media
coverage 0.07
[0.07] 0.04
[0.06] -0.01
[0.12] 0.06
[0.24] -0.13
[0.10] 0.04
[0.08] 0.01
[0.07] -0.05
[0.06] -0.02
[0.04] -0.04 [0.15]
Middle Eastern name -0.21***
[0.06] -0.16***
[0.04] -0.24***
[0.07] -0.30***
[0.08] -0.24***
[0.07] -0.13**
[0.06] -0.10**
[0.05] -0.18**
[0.07] -0.15***
[0.05] -0.20*
[0.10]
After news media coverage * Middle Eastern name
0.07
[0.08] 0.01
[0.06] 0.02
[0.12] 0.09
[0.24] 0.02
[0.12] -0.07
[0.07] -0.01
[0.07] 0.03
[0.09] -0.01
[0.06] 0.08 [0.15]
N 276 516 156 96 156 300 372 300 300 96
Notes: This table is similar to Table A1 but divides data on the occupational level. The occupation Cleaners are not part of the table since cleaners were not included in Experiment B and for Experiment A no job application received a callback after the news media coverage. (†) The occupation Teachers (Language) was not included in Experiment B. (††) In the case of Construction Workers in Experiment A, there are only three observations before the news media coverage. (†††) In the case of Teachers (Math/Science) in Experiment B, there are only three observations after the news media coverage.
Appendix B
Picture B1. Example of news media attention during Experiment A.
Notes: On the 22nd of November 2005 “Sydsvenskan” and other news media disclosed that the authorities intended to send fictitious job applications to employers in order to measure ethnic discrimination.
Picture B2. Example of news media attention during Experiment B.
Notes: On the 5th of November 2006 “Dagens Nyheter” and other news media reported that “Employers discriminate against Arabic names”, referring to the results in Carlsson & Rooth (2007).
