1
The labour battle between high skilled
first-and second- generation immigrants
Nina Timmer
10345906
June 2018
Faculty of Economics and Business
Field: Behavioural Economics and Game Theory
15 ECTS
i
Statement of Originality
This document is written by student: Katherina Maria Timmer who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible
ii
Abstract
This thesis presents a field experiment that aims to investigate possible labour market discrim-ination. To this end we compare callback rates of first- and second-generation immigrants with a Middle Eastern background. We apply to 94 advertisements from companies who are hiring graduates and we randomly assign one resume with either a first or a second generation immi-grant profile to each. The results of the experiment in the Netherlands show no indications of a difference in callback ratio’s between first and second generation immigrants. A hopeful note, the found callback rates are high for both generations as compared to other similar experiments, namely 57,44% for second-generation and 51,06% for first generation. We do find differences in the response rates for traineeships, which looks at whether or not an applicant received any re-sponse at all from the firm. This could suggests that there is a difference in discrimination levels between the generations.
Contents
Abstract . . . ii
1 Introduction 2 2 Previous methodologies and research 5 3 Experimental design 9 3.1 Market . . . 9 3.2 Application . . . 10 3.3 Name . . . 11 3.4 Procedure . . . 12 4 Results 13 4.1 Callback Rates . . . 13 4.2 Probit Estimation . . . 15 4.3 Response Rates . . . 16 5 Concluding remarks 18 A Appendix 21 B Appendix 25 Bibliography 26 1
Chapter 1
Introduction
Throughout the last decade, field experiments from all over the world show that ethnic labour discrimination in the hiring process is a common phenomenon (e.g., Riach and Rich (2002), Bertrand and Mullainathan (2004), Carlsson and Rooth (2007)). Their results show that ethnic minorities receive less callbacks and have a disadvantage when applying for a job. A large num-ber of these field experiments focus on first-generation immigrants or, more broadly, investigate the consequences of ethnic features within an individuals profile. Blommaert et al. (2013) con-clude that within the Dutch labour market, applicants with an Arabic name are perceived as less desirable than those seen as natives.
In the Netherlands, a large proportion of the population is second-generation immigrant: table 1.1 shows that 21,7% of the Dutch population are immigrants of whom more than 50% are non-Western. If the difference between second-generations immigrants opposed to natives or the first-generation immigrants is of importance in the labour market, this rises the question whether or not they can expect to have different chances in applications. In this study we focus on the difference between the various generations of immigrant workers. We test the
hypothe-Table 1.1: Population of the Netherlands
Total population First-generation Second-generation One immigrant Both parents immigrants immigrants parent immigrants Total population 16900726 1860977 1804344 1021904 782440 Native 13235405
Immigrant 3665321 1860977 1804344 1021904 782440 Non western 2038509 1113274 925235 307401 617834
c
Central Bureau of Statistics (CBS), Den Haag/Heerlen 17-05-2018
CHAPTER 1. INTRODUCTION 3
sis that the second-generation immigrants have a higher probability of receiving a callback for a job interview. The expectation that the second-generation immigrant receives more callbacks is based on data from the annual integration report of the Netherlands in 2016 (CBS, 2016). Here we see that second-generation immigrants perform better in the labour market than first generation immigrants. They work more hours, have a better income and their educational per-formance is higher compared to the first generation (CBS, 2010). But still, as several studies have shown, compared to natives within their generation, second-generation immigrants underper-form in the labour market (e.g., Behrenz et al. (2007), Liebig (2007), Algan et al. (2010)).
Performance in the labour market can be split up into income and either securing or holding a position, It is particularly in this latter part of the split where second-generation immigrants seem to be less successful. However, we can also pose a question with regards to the develop-ment of this performance over subsequent generations of immigrants within the labour market. Thus, an extension to the previous body of research can be to investigate the disparity in the level of discrimination against applicants in the hiring process. For this, first-generation immigrants are opposed to their offspring. In particular, we conduct an experiment about labour mar-ket discrimination towards high-skilled first- and second-generation immigrants with a Middle Eastern background in the Netherlands. We investigate the variance between immigrant gen-erations, in their rate of success of commencing a high profile professional career. To this end, we send nearly identical fictitious resumes to employers that post ads for starting positions, traineeships or internships.
Online services such as Google, LinkedIn and FaceBook complicate the process of sending out fake applications nowadays in comparison to experiments performed 14 years ago, such as the one of Bertrand and Mullainathan (2004). Because of these services, it is common for employers to screen applicants online in conjunction to the conventional process of scanning their CV’s. As such, we construct an online identity for our one fictitious applicant.
Perhaps surprisingly, we find no statistically significant difference in callback rates between first- and second-generation immigrants. This could imply that both generations essentially have the same probability of being invited to an interview for high-skill jobs. We do find some differences in the response rates, which looks at whether or not an applicant received any re-sponse at all from the firm. First-generation immigrants receive less frequently a rere-sponse for
CHAPTER 1. INTRODUCTION 4
a traineeship, may it be positive or negative, compared to second-generation applicants. This could suggest that there is a difference in discrimination levels between the generations, but not in the way we suspected.
This thesis continues as follows. Chapter 2 reviews the current body of research and vari-ous field experiments that investigate discrimination, and we particularly emphasize those that have made contributions regarding ethnic discrimination. We delve into the methodology and design of our field experiment in section 3. In section 4 the results are discussed and in the fifth one the drawn conclusions are covered.
Chapter 2
Previous methodologies and research
Over the last decade several approaches have been developed for conducting field research on the topic of discrimination. In particular for field research about discrimination within the labour force three methods can be found that are generally applied. Riach and Rich (2002) de-fine and discuss these methods in their paper. Two of these methods are techniques that involve actors playing the role of applicants, while the last one merely portrays the fictitious applicants on paper. The complication that arises in the design of field experiment involving actors re-volves around the problem that it is generally hard to control for all the forces affecting the behavior of an actor during an interview. As such it is harder to attribute the outcome of these interviews to only the race of the interviewee actors. For example, even after training actors to strictly behave within the confines of the experiment, it is still hard to ascertain that they will not adjust their behavior in favor of their desired outcome. Riach and Rich (2002) mention these as criticisms for the methods involving actors and are strongly in favor of the method that employs fictitious profiles on paper, because the latter method reduces the aforementioned potential for heterogeneity in unobservables.
Experiments utilizing fictitious applicants can be conducted in several ways for various pur-poses. Bertrand and Mullainathan (2004) performed a correspondence test in Boston and Chicago where they created high and low quality resumes, after which they randomly match each one to either an African-American, or a white name. Because of the randomization certain companies receive either a high quality African-American named resume or a low quality one and the same holds for white named resumes, as such they control for quality when determining the effects
CHAPTER 2. PREVIOUS METHODOLOGIES AND RESEARCH 6
of ethnicity. On the other hand in Sweden, Carlsson and Rooth (2007) conducted an experi-ment where they sent out resumes with identical skills, and each one randomly received either a Middle Eastern- or Swedish-sounding name. Both these correspondence tests show that eth-nic minorities receive significantly less callbacks for interviews.
Carlsson and Rooth (2007) also control for either high-, low- or middle-skilled occupations in their correspondence test, and they show that there is less discrimination within high-skill jobs. However, using demographic data from Sweden, Irastorza and Bevelander (2017) shows that in comparison to natives immigrants are underrepresented in the high-skill job segment. Furthermore Bovenkerk et al. (1995) and Goldberg et al. (1996) show that in the Netherlands and Germany discrimination within low- and middle-skill jobs is stronger than within the high-skill occupations. Both studies in the Netherlands and Germany use the same methodology. For the low- and middle-skill jobs they use actors who apply by phone and for the high-skill jobs they use a correspondence test. These studies however can be less reliable, because they are conducted a while ago and employ two different experiment designs. On the other hand Liebig (2007) also found that immigrants are less represented in high-skill professions using the available German demographic data. He concludes that second-generation immigrants are 63% less likely to be employed as a high-skilled professional and 58% more likely to work in lower skilled occupations.
Carlsson (2010) conducted an experiment where he used three nearly identical resumes. To every job ad he sent out one typical Swedish named resume and two typical Middle Eastern named ones. One of the Middle Eastern named resumes was accompanied by some background information reflecting that the fictitious individual was a first-generation immigrant, while the other one was identical to the Swedish cv. The range of ads he replied to in this experiment vary from that for shop sales assistants to school teachers, which require different skill sets. However he does not reply to ads that are either for or can lead to high profile or management positions. In this experiment he does not find a significant difference in the probability of a callback between the first- and second-generation immigrants.
Where Carlson’s research is focused on middle-class labour, Kaas and Manger (2012) center their study around high-skilled labour. In the experiment they looked at the difference in prob-ability of a callback for an internship in the business economics field. They generated fictitious
CHAPTER 2. PREVIOUS METHODOLOGIES AND RESEARCH 7
resumes for university students with German names and those with common names of Turkish immigrants in Germany. Both profiles were nearly the same, and some minor differences were included to ascertain that they were not identical. The Turkish named profile did not contain any additional sign of it being that of an immigrant other than the name. Both the profiles were sent out to all companies who had posted an ad for a university intern. The results of this exper-iment show that there is a difference in callback rate, depending on the size of the firm, where in all cases the second-generation immigrant is at a disadvantage.
In the annual integration report of the Netherlands we see that students with a non western migration background are more represented in studies associated with economics compared to natives. And the average inflow to higher education is nearly the same for natives and immi-grant students (CBS, 2016). So when graduating they have the same high level degree, but we established that they are under-represented in high-skilled professions. Together with these re-sults, Huijnk and Andriessen (2016) found that second-generation immigrants achieve better at school, but that does not lead to a better position on the labour market. One could hypothesize that this under representation in the high-skilled labour market, shown by Liebig (2007), could be a result of the lack of callbacks, early in the careers of the second-generation immigrants. Gault et al. (2000) performed an investigation into the relationship between career success and internships. They found that those who had an internship on their resume, obtained higher positions in a shorter time frame compared to workers who did not. Furthermore, Gault et al. (2000) also concluded that these workers also received higher salaries and experienced higher levels of overall job satisfaction than those without an internship on their resume. So if high-skilled second-generation immigrants have a smaller chance of getting an internship, they miss out on an important opportunity to start of their career.
To conclude there are some limitations to the use of correspondence tests. First, discrimi-nation can occur in every stage of the application. This is especially the case for traineeships, where the hiring process can often involve more than 6 rounds, and our data can only inform us about a possible difference in treatment for the first round. Secondly, as Rooth (2014) mentions, this method will only allow us to send resumes to firms that utilize standard search methods for their recruitment activities, which in turn could affect the external validity. Furthermore, the popularity of the correspondence tests could potentially aggravate this problem. If firms are
CHAPTER 2. PREVIOUS METHODOLOGIES AND RESEARCH 8
aware that this type of approach is frequently employed, they are encouraged to find applicants via alternative routes to avoid being part of an experiment. It could also happen that firms may no longer reply as soon as they suspect that they are part of an experiment. Rooth (2014) also states that with this method we only look at one side of the labour market, namely the demand side, but we do not provide any behavioral insights on the supply side. Finally correspondence tests do not tell us anything about wages or promotions after labour market entry, where dis-crimination could also occur.
This thesis will add to the current literature by providing a field experiment where we inves-tigate the difference in callback rates for first- and second-generation immigrants. In addition, given the aforementioned hypothesized effect of the early stages of a high-skill professional ca-reer, we specifically consider near graduation university students applying for an early career opportunity. In particular, we look at opportunities within the Dutch finance labour market where there is ample need for these types of students.
Chapter 3
Experimental design
To start this experiment first a LinkedIn profile is generated, to fill the profile a network of peers is created. This is done because there are vacancies where it is mandatory to provide LinkedIn information and sometimes recruiters search for this before replying to an application. An email address linked to this profile, as well as a simcard and an automatic personalized voicemail. A common student accommodation in Amsterdam is given as the home address, because provid-ing this is mandatory. However, companies don’t send out letters without first seekprovid-ing contact through email or by phone, and as such a dummy address will not result in data collection com-plications. The next step is creating a resume that is representative of students who are seeking to find a junior position, traineeship or an internship. For high skilled jobs it is essential to in-clude a cover letter with the application, so we draw one that is standardized and can be slightly alerted to the specifics of different vacancies. Here we make the trade off between a higher probability of a callback that is inherent to a perfectly customized letter and an adjustable stan-dardized one that eliminates potential heterogeneity.
3.1 Market
The experiment focuses on a specific segment of the Dutch labour market, in particular the market for traineeships, internships and junior positions for students in the final stage of their finance study. With the current trend in the Dutch labour market that most starting positions are designed as traineeships, the main focus is on the first two. The decision of applying for
CHAPTER 3. EXPERIMENTAL DESIGN 10
starting positions is based on the expectation of the importance of the early stages of a high-skill professional career. In addition, this decision also simplifies the design of the experiment, because higher level positions require a higher level of customization which in turn could result in additional biases. The restriction of only applying to vacancies within the finance segment of the market allows us to completely automate the application process by sending near identical covering letters and to promote homogeneity. For similar reason we limit our study to positions within the Netherlands. The jobs within finance can be segmented into those that are either in banking, consulting, or corporates but they are all are quite homogeneous. Most of the vacan-cies we reply to are posted by companies with more than 500 employees except those posted by M&A boutiques.
3.2 Application
For this experiment a first-generation applicant and second-generation applicant are generated. Both applicants are 23 year old master students. For this field experiment both applicants need to be of similar age. To let it be logical that two equally named and educated 23-old-applicants can continue as a first- and as a second-generation immigrant, there must be a reason to come to the Netherlands either about 15-20 years ago and the possibility of being born here in a family who moved here about 25-30 years ago. This will not be the obvious situation for first- and second-generation guest workers in the Netherlands from for example Maroc or Turkey. Most Turkish and Moroccan immigrants in the Netherlands came here during the 70’s (Prins et al., 1996). In line with this an applicant was created who came as political refugee and not as a guest worker. Iranian immigrants did come to the Netherlands in two waves. The first wave of immigrants came after the Islamic revolution in 1979 and a second wave is seen as around 2000 the requests for asylum were peaking (Dourleijn and Dagevos, 2011). These historical facts are in line with the story of our applicant, he could be either a first- or second-generation immigrant.
We only apply to vacancies that allow us to complete the application by only including a resume, covering letter and a grade list. This amount of information is sufficient for most of the Dutch vacancies, however there are those that also require a transcript of the high school grades and given the time constraints these are out of the scope of this thesis. Furthermore
CHAPTER 3. EXPERIMENTAL DESIGN 11
companies that require a video pitch or an assessment for the completion of an application are also not taken into account. This leaves out some large market participants that can be relevant for graduates, however it does not exclude any entire segment within the finance job market.
To distinguish between first- and second-generation immigrants we created two slightly dif-ferent profiles, to which we refer to as Tehran and Hilversum from here onwards. On one resume it is stated that he is born in Tehran and on the other one Hilversum is given as the place of birth (see Appendix A). Furthermore, the Tehran resume includes Farsi as a language whereas this is excluded in the Hilversum profile. The only difference in the cover letters between the profiles is that in the one of the Tehran profile it is mentioned that he is a refugee who came to the Nether-lands at the age of 8. In both of the resumes it is stated that he has worked as an intern, was a member a student committee and he has worked a few months for a non-profit student con-sultancy. This is a common collection of experiences for students in finance, because it allows them to pursue several paths afterwards.
3.3 Name
Both applicants have the same name, namely Mohamed Ahmadi. Because the name needs to be eligible for use in both profiles, the first name employed is the most common Middle East-ern name given in the Netherlands in 1995 (the birth year of this applicant)1. For the surname we used one of the most common ones in Iran and Afghanistan, which was also carried by 549 people in the Netherlands in 2007 2. With this name it is easy for recruiters to recognize the ethnic background of the applicant. Because in this experiment we do not use a baseline with a native name to measure the level of discrimination, the social status associated with the name is of less importance than in most previous field research. Because the purpose of this experiment is to uncover a possible variance in the opportunities encountered by the first- and second-generation immigrants, we only consider a male applicant. Considering a female pro-file is beyond the scope of this thesis, because demonstrating that a female applicant is in fact a 1The Meertens institute collects information about the Dutch languages and culture, and therefore also about the popularity of first names https://www.meertens.knaw.nl
2CBG provides information about the distribution of surnames in the Netherlands http://www.cbgfamilienamen.nl and Forebears show how many people carry a certain name in a certain country http://forebears.co.uk
CHAPTER 3. EXPERIMENTAL DESIGN 12
second-generation immigrant and not married into the name would result in a large difference between the profiles.
3.4 Procedure
In April and May of 2018 all vacancies that are open for finance students who are in their final stage of their studies were collected. For each advertisement, one of the two profiles is randomly assigned to that company, and as such only one resume is sent to each one. This randomization is done per batch, where a batch is defined as an even number of ads found in one week. From that batch, using STATA, a random sample with a size equal to half of the batch is taken without
replacement and assigned randomly to either the Tehran or the Hilversum profile. The other half of the batch then is assigned to the other profile. The total of all batches together gave us 94 ads. For all vacancies we send out a cv, cover letter and a grade list coming from either the personal email address of our fictitious candidate or these are included as attachments in the online form provided on the vacancy website. These are either in Dutch or in English, depend-ing on the language of the ad, but both provide exactly the same information. In our database we collect the type of contract (intern, trainee, junior), the size of the company and the dates. After applying for the vacancy, we register callbacks within the next 30 days, where a callback is defined as any action of a firm that signals positive interest in the applicant, including offers for interviews, direct job offers, going to the next round and leaving recruitment contact informa-tion in the voice mail. Standardized emails that confirm the applicainforma-tion are not considered as callbacks. If a positive response is received we reply and kindly withdraw the application due to personal circumstances. For traineeships the hiring process consists of several rounds, where the first round usually consists of a CV and motivation screening, and if the result thereof is pos-itive the candidate progresses to the next round. The next round generally involves an online capacity and personality assessment, and because of possible biases that could occur at that stage we decide to exclude this stage form the research. We are aware that discrimination could occur at every stage of the process, but we expect the pre-selection of the resume the have the largest impact from a discriminatory point of view.
Chapter 4
Results
The data for this experiment is gathered over a period of 2 months, and in total we have 94 observations equally distributed over the Tehran and Hilversum profile. We do not observe a significant difference in the callback rate between the Tehran and Hilversum profile, when look-ing at the total of the resumes that have been sent. But we do observe some differences in the manner that companies communicate with the candidates.
4.1 Callback Rates
The first set of results provide us with the callback rates, their differences and whether or not these are statistically significant. In table 4.1 the callback rate of both the Hilversum and Tehran profile is shown, and in the last column the difference between the two proportions is presented. Below the callback rates we provide the number of resumes that are sent out. The first row shows the results for the total of all the resumes sent out. We find that the percentage of callbacks for the Hilversum profiles is 57.44 percent and for the Tehran profiles it is 51.06 percent. This results in a difference of 6.38 percent due to the difference in place of birth, but this does not seem to be statistically significant as we cannot reject the hypothesis that the difference is equal to zero. As such, we cannot say that there is a difference in the callbacks between the Tehran and Hilversum profile at any reasonable confidence level.
Rows 2 up to 4 break down the sample into the different types of contracts mentioned in the ads. The total number of resumes sent to traineeship ads is roughly the same as the number of
CHAPTER 4. RESULTS 14
resumes sent to junior positions and internship openings together. This is due to the number of junior positions marketed as traineeship nowadays. Implying that we found more traineeship to apply for compared to the other type of contracts. For traineeships and junior positions we see a higher callback rate for the Hilversum profile, whereas for internships the Tehran profile received a higher percentage of callbacks. However, once more these differences are not signifi-cant.
Table 4.1: Callback rates
Hilversum Tehran Difference [obs] [obs] (p-value) All resumes 57.44 % 51.06% 6.38% [47] [47] (0.535) Traineeship 65.22% 42.31% 22.91% [23] [26] (0.109) Internship 52.94% 73.33% -20.39% [17] [15] (0.234) Junior 42,86% 33.33% 9.52% [7] [6] (0.725) Small size 80,00% 53.85% 26.15% [20] [13] (0.110) Large size 40.74% 50.00% -9.26% [27] [34] (0.471)
Note: Callback rates in this table indicate that a applicant
received positive news to go to a following step in the re-cruitment process. The table reports the callback rates for the applicant with Hilversum as place of birth (column 1), the one with Tehran as place of birth (column 2) and the difference (column 3) of these callback rates. In brackets in each cell we have the number of resumes sent in that cell. Column 4 reports the p-value for a test of proportion testing the null hypothesis that the callback rates are equal across ethnic groups.
Finally, rows 5 and 6 show the results segmented for different firm sizes. Small size refers to companies with less than 500 employees and large size to the ones with more than 500 employ-ees. The largest proportion of the resumes is sent to large companies, which could stem from the fact that they are more inclined to attract graduates through job advertisements, whereas smaller firms more often employ recruitment firms. To generate this data set we did not send resumes to recruitment firms, we only send applications directly to the companies of interest. This could explain why there are less small firms in the data set, because they more often use
CHAPTER 4. RESULTS 15
recruitment firms to find applicants compared to large companies. For both small and large companies the difference in the callback rate is not significant. Given these results we cannot say that the companies show preferential behaviour towards candidates originating from one of the birth places. But we can see in Appendix B that the study is underpowered, which could imply that the true effect is not discovered.
4.2 Probit Estimation
In table 4.2, we report the findings of a probit regression with the callback rate as the dependent variable. With this model we can look for the effects of various employer characteristics on the probability of a callback. We employ the following probit model:
P(callback= 1|Hilversum, . . . ,Internship) =
Φ(β0+ β1Hilversum+ β2Small+ β3Traineeship+ β4Internship)
WhereΦ(·) is the cumulative distribution function of the standard normal distribution. In column 1 we only regress the dummy Hilversum on the probability of getting a callback. The sign of the coefficient of the Hilversum dummy is positive but not significantly different from zero, so we cannot say that being born in Hilversum increases the likelihood of receiving a call-back. This effect is robust to adding two dummy’s for the type of contract (column 3) but not if we add two dummy’s for the size of the firm (column 2). The dummy variable for the place of birth stays small and not significant across all regressions. The size of the firm has a statisti-cally significant effect on the callback probability, where applications to small size companies increase the likelihood of a callback with approximately 25 percent. The type of contract being either a traineeship, internship or junior position does not seem to have a significant effect on the callback probability.
This all suggests that only the size of the firm has an impact on the callback rate. Place of birth and the type of contract do not seem to be predictive indicators for the callback rate of high-skill jobs.
CHAPTER 4. RESULTS 16
Table 4.2: Marginal effects Probit regression
Callback (1) (2) (3) (4) Constant 0.0267 -0.1397 -0.3807 -0.6025 (0.1838) (0.2021) (0.3824) (0.4099) Hilversum 0.0638 0.0330 0.0633 0.0342 (0.1031) (0.1065) (0.1037) (0.1066) Small 0.2337** 0.2497** (0.1047) (0.1156) Traineeship 0.1512 0.2120 (0.1547) (0.1569) Internship 0.2359 0.1867 (0.1547) (0.1614) Standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Note: The table reports in each column the results of a probit regression
where the dependent variable is the callback dummy. Reported in the table are the marginal effects on the likelihood of a callback. Standard errors corrected for clustering at the employment-ad level.
4.3 Response Rates
Because the previous results do not show us any type of differential attitude from the companies towards one of the profiles, we also provide a table that reports the type of response received by the two candidates. Discrimination can occur in different ways. In table 4.3 we report whether or not a profile has received any type of response. As mentioned before a response can be pos-itive or negative, and can be communicated either via email or by phone. It could be seen as discrimination if the Hilversum profile receives a response indicating that a firm is not inter-ested, while the Tehran profile is simply ignored. In row 1 of table 4.3 we see that 74.46 percent of the time the Hilversum profile was used to apply he received a response. For the Tehran pro-file this percentage is 68.09, but this difference is not statistically significant.
In row 2 we see that the Hilversum profile received more replies than the Tehran one, namely 22.07 percent more, and this difference is statistically significant at the 10% confidence level. For the other types of contracts, row 3 and 4, the difference is not statistically significant. Only for internships and large size companies the difference is negative, meaning that the Tehran profile received more responses that were either negative or positive. But these two differences are not
CHAPTER 4. RESULTS 17
statistically significant.
Table 4.3: Response Rate
Hilversum Tehran Difference [obs] [obs] (p-value) All resumes 74.46% 68.09 % 6.38% [47] [47] (0.494) Traineeship 91.30% 69.23% 22.07%* [23] [26] (0.056) Internship 64.70% 80.00% -15.29% [17] [15] (0.337) Junior 42.86% 33.33% 9.52% [7] [6] (0.725) Small size 85.00% 61.53% 23.46% [20] [13] (0.124) Large size 66.67% 70.59% -3.92% [27] [34] (0.743) *** p<0.01, ** p<0.05, * p<0.1
Note: Response rates in this table indicate that applicant
received any kind of news, positive or negative. The table reports the rates of receiving an answer for the applicant with Hilversum as place of birth (column 1), the one with Tehran as place of birth (column 2) and the difference (col-umn 3) of these response rates. In brackets in each cell we have the number of resumes sent in that cell. Column 4 reports the p-value for a test of proportion testing the null hypothesis that the response rates are equal across ethnic groups.
Summarizing we found that on average applicants that have Hilversum as their place of birth do not receive significantly more callbacks than those with Tehran as their place a of birth. We did find that in certain situations the resumes of the first-generation receives a larger number of responses that can be either negative or positive. We discuss the implications of these results in the next chapter.
Chapter 5
Concluding remarks
The research question of this experiment was aimed at investigating whether or not we can observe a difference in the probability of securing an interview, between first- and second-generation immigrants. To this end we conducted a field experiment where we sent out a fake profile with either Hilversum or Tehran as the place of birth of the applicant. Although we ex-pected to see more interview invitations for the second generation applicant, given that they are born and raised in the country that the advertised job is located at, the results do no show evidence of a difference in treatment when it comes to invitations. This means that the first-and second-generation applicants have the same chance of being invited for an interview.
From our estimated models we observe that the place of birth does not increase the likeli-hood of receiving a callback, and it could be argued that employers perceive them as equals. However, in nearly all the cases, except for large size companies and internships, we do notice the difference skewed towards favouring the Hilversum profile (table 4.1). One possible expla-nation for the difference in the case of large firms is that they have a diversity policy and aim to invite more foreign born applicants. The question remains whether these applicants make it to the other rounds in the process. But more research is required to assess the effect of a diversity policy on the recruitment behaviour of large firms.
In contradiction to the arguments that the profiles are being perceived as equal, it could be said that possibly, with a larger sample size, we might find significant differences, which would imply that our study is underpowered. The sample size bottleneck stems from the fact that the research is focused on a specific market, namely the finance market, where we did not have
CHAPTER 5. CONCLUDING REMARKS 19
access to more vacancies. We had to make a trade-off between sending application to job ads from multiple sectors, and thus obtaining more data at the cost of heterogeneity, and only one sector for the sake of homogeneity at the cost of a smaller set of job ads. In the high skilled labour market these profiles need to be tailor made to at least the sector, hence we expected a severe heterogeneity problem. Furthermore, because of time constraints we were unable to generate more online identities, and within this research we have seen that online presence was crucial for conducting the experiment. For several firms supplementing the application with a LinkedIn profile was mandatory and numerous company recruiters had viewed the profile be-fore responding to the application. For future research aimed at a population with the same demographic and socioeconomic characteristics, we advise to have a sample size of approxi-mately 1000 observations. Then the estimator found in table 4.2 will be significantly different from zero. The sample size increase could be achieved by increasing the number of near identi-cal profiles sent to a company, targeting a broader variety of job types within the finance market, or a combination of both.
We noticed that the behaviour of the firms towards second-generation immigrants is differ-ent than towards the first-generation immigrants in how applications are handled, and that the Tehran profile received significantly less responses be it positive or negative. This is the case for firms who were hiring trainees. Most companies that hire university graduates do this under a traineeship contract, so more and more young ambitious talents commence their careers as trainees. Given our findings this may suggest that it is harder for first-generation immigrants to start off their careers. Lets for example consider the situation that the Hilversum profile has received a response that he will not be invited to a following step in the recruitment process, while the Tehran profile has received no response at all. One could conjecture that the firm may have had an initial interest in the Hilversum profile, and declined to invite the candidate after consideration, whereas a lack of response to the Tehran profile may indicate an initial disregard of it. This initial disqualification can lower the chances of the Tehran profile in the beginning of his career and can be considered a result of discrimination.
On a brighter note, for both profiles we do find a high callback rate as compared to other similar experiments, where the lowest one we have seen is below 10 percent (Bertrand and Mul-lainathan, 2004) and a higher one is between 32 and 42 percent (Kaas and Manger, 2012). From
CHAPTER 5. CONCLUDING REMARKS 20
the data of the CBS, it can be drawn that since the last quarter of the 2017 onward the labour market in the Netherlands is relatively tense1. This labour market tightness implies that the demand for labour is above average and the available applicants are relatively low, which could be an explanation for the high callback rates in this research. It could also be the case that our profile is above average compared to other near graduates that apply, but in both cases these factors do not influence the validity of our observations for the sake of our research question. This is because both our profiles are sent out in the same economic environment and are equally strong. But it could be the case that the rate of discrimination increases for applicants with lower grade point averages and extracurricular experience on their resume. Recruiters could link these lower academic and professional achievements to their background, however such a conclusion is beyond the scope of this thesis and could be subject for future research.
From this research we cannot draw any conclusions with regards to the level of discrimina-tion as compared to natives, and given the results seen in previous literature we expect that the two profiles we used are in a disadvantageous position. From the data obtained through this research, we can not conclude that there is a difference in callback rates between different gen-erations of immigrants. We do observe that the second-generation immigrant received more responses, be it positive or negative. It could be argued then that differences in callback rates found in previous literature more likely stems from the ethnicity of the applicants rather than the place of birth or their first language. In the wisdom of knowing this, government and politics are obliged to take action and come up with some serious measurements to prevent this type of labour discrimination. In the end it does not matter in what family you were born, or what’s your given name: second-generation immigrants are as much Dutch as so-called natives. At least by their language, level of education and place of birth.
Appendix A
CV, cover letter and grades second
generation
Mohamed
Ahmadi
Curriculum Vitae
Willem Beukelsstraat 51 1097 CR Amsterdam 03-03-1995 in Tehran (+31) 6 15686478 ahmadimohamed318@gmail.com www.linkedin.com/in/mohamed-ahmadi
ㅡ
Education
UvA/MasterSeptember 2017 - Current, Amsterdam
Finance, Specialisation: Corporate Finance.
UvA/Bachelor
September 2014 - September 2017, Amsterdam
Economics and Business, Specialisation: Economics and Finance. 7/10
Alberdingk Thijm College/VWO
September 2007 - September 2014, Hilversum
Specialisation: Economics and society. 7.1/10
ㅡ
Experience
NIBC/M&A intern
January 2018- March 2018, The Hague
Valuing companies by means of different valuation methodologies and participating in day-to-day transaction executions.
FSA/Committee Member (Promotion) Beroependagen 2017
September 2016 - September 2017, Amsterdam
Together with six committee members I was responsible for the promotion and organisation of the Beroependagen 2017.
De Kleine Consultant/Consultant
December 2016 - June 2017, Amsterdam
De Kleine Consultant provides value-based strategic advice for (non-) profit organizations to help them achieve their ambitions.
ㅡ
Languages
Dutch: C2 German: B1 English: C2 Farshi: B1
ㅡ
Skills
Office software: Microsoft Office, Salesforce, LATEX Operating systems: Windows, macOS
Statistical packages: Stata, SPSS, Excel, R
Dear,
With this letter I would like to apply for a traineeship position at "company", and in this letter I will highlight my motivation for this and elaborate on why my profile is a good fit for the role as well as your firm.
My name is Mohamed Ahmadi, I am 23 years old and living in Amsterdam, I moved to the Netherlands from Iran when I was 8 years of age. I would be pleased to start my career at “company”. My interest in the financial world dates back to my childhood, where I came into contact with various financial concepts, ranging from leveraged acquisitions to equity research through my uncle who would hand me down his books and journals he read. This interest is reflected in my academic endeavours, and here I learned that the Corporate Finance world appeals the me the most, because of its dynamic and often transaction driven nature.
I know that “company” offers a high quality development and training programme, tailored to the competencies of each individual and where … plays a central role.The dynamics of having several clients appeals a lot to me, because I believe that it steepens my learning curve substantially. I apply for this position because I believe “company” to be a perfect learning school, where I can gain the professional skills necessary for my future
development. Furthermore, I'm confident that this opportunity will give me the challenges that I'm seeking, whilst working at one of the most prestigious companies in the Netherlands. I recognize myself in the profile described in your vacancy I believe that my academic and professional experience can be useful assets for the “company finance” department. I am conscious of the challenges a programme like this represents, however I consider that my predisposition, perseverance and background will allow me to successfully meet your expectations. If I were to describe myself, I would say that I am a positive thinking person who at the same time is patient and never loses control during an assignment. As part of my skill set, I also regard my great job satisfaction and my ability to meet tight deadlines under stressful situations. As for my analytical capabilities, these have been developing and put to practice during courses such as “Capita selecta finance” and “Applied Financial
Econometrics” as well as my internship at NIBC. In my spare time, I enjoy exercising and spending time with friends.
Thank you very much for this opportunity and for taking the time to consider my application. Sincerely,
Appendix B
Power analysis
.2 .4 .6 .8 1 Po w e r (1 -β ) 0 500 1000 1500 2000Total sample size (N)
Parameters: α = .05, δ = -.064, p1 = .57, p2 = .51
Pearson's χ2 test H0: p2 = p1 versus Ha: p2 < p1
Estimated power for a two-sample proportions test
Bibliography
Algan, Y., Dustmann, C., Glitz, A., and Manning, A. (2010). The economic situation of first and second-generation immigrants in France, Germany and the United Kingdom. The Economic Journal, 120(542).
Behrenz, L., Hammarstedt, M., and Månsson, J. (2007). Second-generation immigrants in the Swedish labour market. International Review of Applied Economics, 21(1):157–174.
Bertrand, M. and Mullainathan, S. (2004). Are Emily and Greg more employable than Lakisha and Jamal? A field experiment on labor market discrimination. American Economic Review, 94(4):991–1013.
Blommaert, L., Coenders, M., and Van Tubergen, F. (2013). Discrimination of Arabic-named applicants in the Netherlands: An internet-based field experiment examining different phases in online recruitment procedures. Social forces, 92(3):957–982.
Bovenkerk, F., Gras, M. J., Ramsoedh, D., Dankoor, M., and Havelaar, A. (1995). Discrimination against migrant workers and ethnic minorities in access to employment in the Netherlands. International Migration Papers, 4:1–39.
Carlsson, M. (2010). Experimental evidence of discrimination in the hiring of first-and second-generation immigrants. Labour, 24(3):263–278.
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.
CBS (2010). Jaarrapport integratie 2010. Den Haag: Centraal Bureau voor de Statistiek, pages 61–86.
BIBLIOGRAPHY 27
CBS (2016). Jaarrapport integratie 2016. Den Haag: Centraal Bureau voor de Statistiek, pages 42–53.
Dourleijn, E. and Dagevos, J. (2011). Vluchtelingengroepen in Nederland. Over de integratie van Afghaanse, Iraakse, Iraanse en Somalische migranten. Den Haag: Sociaal en Cultureel Planbureau.
Gault, J., Redington, J., and Schlager, T. (2000). Undergraduate business internships and career success: are they related? Journal of Marketing Education, 22(1):45–53.
Goldberg, A., Mourinho, D., and Kulke, U. (1996). Labour market discrimination against foreign workers in Germany. International Labour Office, Employment Department.
Huijnk, W. and Andriessen, I. (2016). Integratie in zicht. Den Haag: Sociaal en Cultureel Planbu-reau.
Irastorza, N. and Bevelander, P. (2017). The labour-market participation of highly skilled immi-grants in Sweden: An overview.
Kaas, L. and Manger, C. (2012). Ethnic discrimination in Germany’s labour market: a field ex-periment. German Economic Review, 13(1):1–20.
Liebig, T. (2007). The labour market integration of immigrants in Germany. OECD Social, Em-ployment and Migration Working Papers, (47):39–44.
Prins, K. S. et al. (1996). Van’gastarbeider’tot’Nederlander’: adaptatie van Marokkanen en Turken in Nederland. PhD thesis, Rijksuniversiteit Groningen Groningen.
Riach, P. A. and Rich, J. (2002). Field experiments of discrimination in the market place. The Economic Journal, 112(483):F480–F518.
Rooth, D. O. (2014). Correspondence testing studies: What can we learn about discrimination inhiring? IZA World of Labor, 58.
