Whose tastes matter? Discrimination against immigrants in the Japanese labour market

ABSTRACT

The existence and mechanisms of ethnic discrimination in hiring decisions have been extensively studied, but the following two gaps remain: non-Western societies and taste-based discrimination. In this contribution, we investigate whether ethnic discrimination and its mechanisms are similarly observable in the Japanese labour market. Additionally, to reflect Becker’s initial idea, we decompose the sources of taste-based discrimination into employers’ preferences and societal preferences (i.e. perceived co-workers’ and customers’ preferences) and determine which and how preferences drive ethnic discrimination. We conducted factorial survey experiments via web surveys in Japan targeting HR professionals/employers working in almost all industries. The results show that immigrants are severely disadvantaged in hiring. Decomposing taste-based discrimination shows that only employers’ preferences contribute to discrimination, whereas societal preferences do not influence discrimination. However, the magnitude of the effects of employers’ preferences on discrimination depends on societal preferences as follows: hiring decisions are less dependent on employer attitudes towards socially preferred groups than those towards socially non-preferred groups. This study suggests that ethnic discrimination is observable even outside of Western societies and that taste-based discrimination is a more complicated process than expected in previous studies.

KEYWORDS:

• Ethnic discrimination
• labour market
• Japan

Acknowledgement

We would like to thank Sho Fujihara for helpful comments. This work was supported by Japan Society for the Promotion of Science (19K23254, 21K01932). We take full responsibility for all infelicities and errors of judgment or interpretation.

Disclosure statement

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

Notes

1 We do not claim that our study is the first study to investigate ethnic discrimination in the labour market outside of North America and European societies. Important predecessors include experimental studies conducted in China (Maurer-Fazio 2012), Hong Kong (Li and Liu 2021), Malaysia (Lee and Khalid 2016), Georgia (Asali, Pignatti, and Skhirtladze 2018) and Russia (Bessudnov and Shcherbak 2020). However, these studies did not investigate mechanisms. Baron (2020) conducted factorial survey experiments in Japan, but his respondents were non-professionals.

2 In addition to these measurements, some studies have regarded the remaining level of unexplained discrimination after adjusting for applicants’ human capital as evidence of taste-based discrimination (e.g. Kaas and Manger 2012). Previous studies have also tested potential contacts between foreign workers and customers and co-workers (e.g. Laouenan 2017).

3 However, notably, a method using incentivized resume rating has been newly developed to avoid deception (Kessler, Low, and Sullivan 2019).

4 However, see Pager and Quillian (2005).

5 We should note that as the institutions we belong to do not require ethical approval, we did not obtain such approval. Yet, we believe that the disadvantages the participants experienced is minimal, because, unlike standard audit studies, our experimental process did not contain deception. Furthermore, we explained to the respondents before they participated in the survey that they could stop taking the survey anytime and that their data would be anonymised.

6 We calculated the distribution of firms in the population based on the latest available data at the time of the survey (Statistics Bureau 2014). Using the quota sampling method, we set a goal for the sample size in each category (i.e. the 16 industries in this study); once the goal was achieved, we stopped the recruitment of respondents from that category.

7 Among the respondents, 48 did not answer the occupation question. Including these respondents in the model does not change the results.

8 Students graduate from high school when they are 18 years old; thus, employers may infer that both first- and second-generation immigrants have graduated from Japanese schools. Experience in Japanese schools is thus controlled for between first- and second-generation immigrants.

9 Because a factorial survey can be considered as a type of conjoint analysis, we made these arguments based on previous theoretical research about conjoint analysis (Bansak et al. 2021b). Conjoint experiment (and theoretically, a factorial survey) estimates causal effects as an average marginal component effect (AMCE) by estimating a particular attribute over the distribution of other attributes.

10 Compared to other methods that deal with clustered data (e.g. OLS with cluster-robust standard errors), multilevel analysis is preferable to study the cross-level interactions because it explicitly partitions the variance into between-cluster and within-cluster effects (Auspurg and Hinz 2014). We also estimated random coefficient models and ordered logistic multilevel models. These different specifications provide substantially similar results.

11 Because attitudes are interrelated we included all four variables simultaneously, while to reduce multicollinearity, we conducted a grand-mean for all the attitudinal variables (Paccagnella 2006; Iacobucci et al. 2016). Because the results did not change, when testing societal preferences, we used the original measurement. Some may wonder why we used interactions instead of simply including attitudinal variables. Including these variables without making interactions capture the effects of positive attitudes towards one of the immigrant groups on evaluations of hiring prospect regardless of the applicants’ migrant background. The results indicate employers’ strictness of the evaluations in general, but are not relevant to different evaluations between Japanese and non-Japanese applicants.

12 We originally conducted a representative survey in November 2021 to capture the hierarchy and heterogeneity of attitudes towards the four groups. This survey used stratified two-stage random sampling in 48 cities to target individuals aged from 18 to 81 who lived in Japan (for the detail and descriptive results, see the report by Research project for internationalisation and citizens’ political participation (2022)). The questionnaires were sent by postal mail to 8,640 respondents and we obtained 3,081 responses, for a 37.2% response rate. We asked the participants about their attitudes towards American, Japanese Brazilian, South Koreans and Chinese immigrants in Japan. The results are consistent with the findings of Igarashi (2015) that Japanese respondents have consensual hierarchical preferences for these four groups. The attitudes towards Chinese immigrants are the most negative (3.090; ranging from 1 to 5, with higher scores being more negative), followed by those towards South Koreans (2.953), Japanese Brazilians (2.455) and Americans (2.292). Interestingly, males tend to have more negative attitudes towards South Koreans than females; however, these gender differences are only significant for attitudes towards South Koreans and not towards other groups (although the order of preferences does not change). These gender differences may partly explain why we obtained the results that South Koreans are the most negatively viewed, because our data mostly consisted of male respondents.

Additional information

Funding

This work was supported by JSPS: [Grant Number 19K23254,21K01932].