Abstract
It has been widely documented that immigrants receive lower earnings than the majority of the population in most Western countries. Previous research has pointed to various forms of discrimination that affect immigrants’ wage rates. The authors discuss whether the source of this wage inequality can be found at the occupational level. In this article, the authors argue that occupational closures reduce within-occupation wage inequality. To test their expectations, the authors use Norwegian register data that span all employees. The results align with their expectations, as both occupational licensing and union density strongly reduce immigrant-majority earnings’ inequality. However, neither certifications nor credentialization reduces the immigrant-majority earnings gap.
In most Western labor markets, ‘most groups with non-European origins continue to experience major ethnic penalties’ (Heath & Cheung, 2007, p. 686), but the size of the minorities’ disadvantage varies considerably across occupational groups in Norway (Drange, 2013, 2016; Karlsen, 2012) and the United Kingdom. (Elliott & Lindley, 2008). How and why these wage differentials vary with labor market settings is not well-documented, however, and this article investigates how variation in occupational wage returns between the majority and minority groups in Norway correlates with different mechanisms of occupational closure. A central question is whether institutions of occupational closure may serve as a shelter against wage discrimination for immigrants who have gained access to closed occupations. We study these questions by analyzing individual-level data from Norwegian administrative registers for the entire population of Norway.
We make three contributions in this article. First, we expand on Drange (2013), who suggested that the closed professions provide shelter against discrimination, and examine if occupational closure can actually reduce inequality, even if we take selection into account. We combine discrimination theory with closure theory to understand better variations in ethnic wage differences across different institutional settings and theorize that occupational closure mechanisms suppress the mechanisms leading to discriminatory wage setting. In addition, we pinpoint which closure mechanisms have such an inequality reducing effect. Finally, we examine our expectations using a unique data set of the entire Norwegian working population purposely compiled for studying wage inequality across the entire occupational spectrum, containing a fine-grained classification of occupations according to the four different closure devices: credentialization, licensure, unionization, and certification (Norwegian Occupational Regulations Dataset; Alecu & Drange, 2016). We can therefore give a fuller account of where wage differences are generated between ethnic minorities and the majority population.
Why Do Immigrants Have Lower Earnings?
We can divide the literature explaining wage differences between the ethnic majority and minority groups into two main perspectives (which, in fact, are often combined). The starting point of the human capital perspective is that wage differences are due to differences in productivity and individual-level skills, such as education, experience, and other job-related skills (Becker, 2009). Research indicates that immigrants have lower productivity than the majority, on average, and explains this with reference to country-specific human capital. Education, work experience, and skills cannot be transferred without incurring loss (Chiswick & Miller, 2002). Over time, immigrants are able to transfer more of their human capital and increase their earnings (Brekke & Mastekaasa, 2008; Chiswick & Miller, 2002, 2008, 2009) and labor market position (Duvander, 2001).
The other main perspective refers to different kinds of discrimination to explain why immigrants have lower earnings than the majority of the population (Banton, 1994). One kind of employer discrimination is ‘taste for discrimination’ (Arrow, 1974; Banton, 1994; Becker, 2009), which implies that the employer is willing to pay for hiring certain candidates rather than others. If the employer chooses to employ an ethnic Norwegian applicant rather than a better qualified, and presumably more productive, minority candidate, this kind of discrimination will be present. Another related example of discrimination is if the employer prefers a majority applicant because of other people’s prejudice, so-called unprejudiced discrimination (see Merton, 1949). Overt discrimination is illegal in Norway and many other countries, however, and is not a seen as a prevalent cause of earning differentials.
A more common, but weaker, form of discrimination occurs when employers only prefer majority applicants in situations where the majority and minority applicants have comparable qualifications and productivity. Statistical discrimination occurs whenever an individual is judged by the average characteristics of the group or groups to which he or she belongs, rather than on his or her own personal characteristics (Thurow, 1975, pp. 171–172). This form of discrimination exists where an employer prefers one category of applicants (e.g., ethnic majority to minority) because it is expected that, on average, they have more of the desirable characteristics (e.g., higher productivity). There is some empirical evidence of statistical discrimination of immigrants in Scandinavian countries (both in Sweden, Tomaskovic-Devey, Hällsten, & Avent-Holt, 2015; and in Norway, Drange, 2013; Wiborg, 2006).
Status characteristics theory claims that social categories such as ethnicity, gender, and social class are hierarchically ordered and that these characteristics and their ordering become the basis for social closure (or opportunity hoarding). The mechanism leading to statistical discrimination is information deficiency, whereas status-based discrimination occurs because of cognitive biases (Ridgeway, 2014; Tomaskovic-Devey, 1993).
Finally, according to relational inequality theory (RIT; Avent-Holt & Tomaskovic-Devey, 2014; Tilly, 1998; Tomaskovic-Devey, 2014), inequality is generated by labor market positions and exaggerated by the mapping of nominal status characteristics onto these positions within an organizational hierarchy. The theory ties status characteristics to wage inequality, as the strength of an individual’s claims-making and whether those claims are perceived as legitimate will be influenced by the social categories to which he or she belongs (Avent-Holt & Tomaskovic-Devey, 2014).
In a recent article, Tomaskovic-Devey et al. (2015) investigate their theoretical expectations to workplace-generated earnings inequality among non-Western immigrants and the majority population in Sweden and find considerable support. They find that workplace characteristics, such as share of minority in management positions and minority–majority job segregation, affect the immigrant–native wage gap. Their analyses also show that the native–immigrant wage gap is smaller in blue-collar occupations than in white-collar occupations, a finding they attribute to collective wage bargaining in occupations of the former kind. Their findings show that the presence of collective wage agreements mitigates inequality among non-Westerners and Swedes in the workplace. Moreover, the difference in earnings between immigrants and majority workers is almost fully explained by occupational segregation. This indicates that opportunity hoarding, that is, closure, is an important mechanism behind earnings inequality.
A central assumption in RIT is that wages are set in the individual workplace and that ‘[…] status distinctions, such as gender or race or immigrant status, operate more powerfully when wage setting is more strongly influenced by local decision making […]’ (Avent-Holt & Tomaskovic-Devey, 2014, p. 392). Although not always so, it is generally the case that people receive their pay from the organization where they work. 1 How much latitude the individual employer has in deciding the size of this payment varies, however. In some industries and for some workers, wages are set through centralized wage agreements. For others, for example, physicians, both what they can charge their patients and the reimbursement they receive from the State is set through fixed rates. There is of course still some leeway for discrimination (e.g., in hiring and promotion, or in which physician patients choose to consult), but we assume it to be considerably smaller than in situations where wage is set locally. In many closed occupations, local decision-making plays a negligible role and, somewhat paradoxically perhaps, we expect that effective closure mechanisms may benefit minority employees who have managed to get a foot in the door.
Occupational Closure
In this article, we propose that the institutions of occupational closure curb the discriminatory mechanisms described earlier and that closed occupations may serve as a shelter against wage discrimination for the immigrants who have gained access. From the preceding pages, we assume that to have such sheltering effects, the occupational closure devices have to overshadow employers’ or customers’ inclinations to discriminate, either by affecting supply and demand or by signaling high quality of service. Finally, occupational closure may affect macro/mesoprocesses in ways that remove wage setting from the local organizational level, and thus leave less room for discriminatory practices. Our argument is that these mechanisms are effective means of reducing discrimination due to information deficiency, which is the basis for statistical discrimination, and restricting the employers’ opportunity to set wages locally, which is the basis for wage inequality in RIT. We will discuss how the mechanisms leading to higher wages in ‘closed’ occupations may also affect the native–immigrant wage gap.
Closure theories assume that wage differences are due to groups’ abilities to close off their domain from competition by various means of discrimination (Avent-Holt & Tomaskovic-Devey, 2014). The starting point of closure theories is the assumption that an elite ‘closes off opportunities to another group of outsiders beneath it’ (Murphy, 1988, p. 8). Neo-Weberian closure theorists have developed this outline into a theory on social stratification through the means of property (Murphy, 1988) and credentials (Collins, 1979; Parkin, 1979), but it is also used to discuss closure based on ethnicity and other status characteristics (Tomaskovic-Devey, 1993). Immigrant origin may be used as a basis for exclusion, but education and credentials are more commonly used (Murphy, 1988) and so often, in fact, that exclusion based on credentials seems ‘necessary and inevitable because they appear to be closely connected with efficiency’ (Murphy, 1988, p. 3). By requiring formal educational credentials for access to certain positions, employers or professional groups control access to privileged positions. Such educational requirements for particular positions may become institutionalized and even sanctioned by the State, as is the case with licensed professions and trades. These kinds of closure are examples of what we here label occupational closure.
The basic argument of occupational closure is that some occupations pay more than others because they are able to establish and maintain institutional barriers to access and raise demand for their services (Bol & Weeden, 2015; Weeden, 2002). This produces a wage surplus that would not accrue under free market conditions, creating a monopoly rent (Sørensen, 2000; Weeden & Grusky, 2014). The mechanisms that raise the rewards of occupational members are restrictions on the labor supply, enhancing diffuse demand, channeling demand, and signaling a particular quality of service (Kleiner, 2006; Weeden, 2002). According to Weeden (2002), the strongest and most resilient forms of occupational closure combine mechanisms that restrict supply and increase demand. Weeden (2002) identifies licensure, educational credentialing, unionization, and voluntary certification as institutionalized strategies of occupational closure.
We will in turn discuss how licensure, unionization, credentialization, and certification can effectively reduce earning differentials. All four closure devices are perceived as legitimate sources of pay inequality, while discrimination is not. The forms of closure are not mutually exclusive, but a central point in our argument is that the strength of the closure strategy is relevant to its ability to reduce pay inequalities between immigrants and majority colleagues. The closure devices draw on different social and legal foundations; the strongest devices are licensure and union representation, which are enforced by law or by-laws. Credentialization and certification constitute closure by social convention and are therefore weaker.
Licensure
A licensed occupation is regulated by law, making unauthorized practice illegal. Hence, it is a true monopoly. According to Weeden (2002), licensure raises mean wages by three mechanisms: First, it reduces the supply of labor, as prospective practitioners have to acquire a license to practice the occupation. In many instances, the occupation can regulate the influx of new practitioners by placing restrictions on the number of training spots and by defining competence demands. Second, licensure channels demand to the licensed occupations, as other occupations cannot legally supply the services that are licensed. Third, licensure signals a higher quality of service by removing the least competent service suppliers (Angrist & Guryan, 2008; Kleiner & Krueger, 2010). The consumer need not to be concerned with the existence of Akerlof’s (1970) ‘lemons’ in the market, at least not in theory.
State regulation of licensed occupations and the way these privileges are legitimized vis-à-vis the public give reason to expect that closure mechanisms can have an effect on the intraoccupation wage inequalities. First, strategies that limit the supply of qualified practitioners reduce the risk of overeducation and effectively stop the employer from buying alternative, cheaper labor. Second, because licensure signals quality of service by raising barriers to access, it can reduce the information asymmetry in the market. This especially applies to professions where licensure is granted to practitioners based on examinations in theoretical and practical skills. By issuing licenses, the authorities and professional organizations ‘guarantee’ that the license holder’s competence is up to standard. This signal can replace other signals that employers and clients make use of, such as immigrant and ethnic minority origin. Participation in government supported occupational retraining programs decreased the risk of occupational degradation among former Soviet Union immigrants to Israel, apparently because their credentials received greater legitimacy when awarded from a state agency (Lerner & Menaheim, 2003, p. 25).
The establishment of a minimum skill requirement can also have an effect on wage setting. Not knowing the productivity level of the prospective employee, the employer can differentiate entry wages. However, licenses document that the skill set meets the occupational standard, which can lead to less acceptance of wage differentiation by the professional associations and the occupations’ members. Research shows that licensure increases the cost of low-quality service and reduces the cost of high-quality service (Kleiner, 2006). Last, the occupations, and particularly the professions, try to equalize wages internally and reduce competition among practitioners (Freidson, 2001; Murphy, 1988; Parkin, 1979).
The effect of occupational regulation on immigrants and ethnic minorities’ wages has only been investigated to a limited degree in previous research, but this research supports an understanding of licensed occupations as open to ethnic minority groups and immigrants. Freeman (1981) found that licensure reduced the occupational attainment of Blacks in America prior to but not after 1970. More recently, Chavez and Redbird (2015, p. 316) documented that licensure increased immigrants’ access by providing a pathway to entry, but not for recent immigrants and adult immigrants. Similarly, Alecu and Drange (2018) show ethnic minorities educated in Norway have equal access to licensed occupations and that cross-national reciprocity agreements 2 facilitate the inclusion of foreign-educated immigrants.
Furthermore, the earnings gap between the majority population and immigrants was significantly smaller in licensed occupations in Canada (Gomez, Gunderson, Huang, & Zhang, 2015) and for the Black minority in the United States (Redbird, 2017). Redbird (2017) finds that licensure results in more equal earnings profiles for underprivileged groups. This suggests that licensure has an equalizing effect that works independent of a supply restriction. Her argument is that access to unlicensed occupations may not be as open and equally accessible for all as theory will have it, but rather dependent on different forms of discrimination, or on informal criteria such as social networks and nepotism (Redbird, 2017). In this perspective, the introduction of a license may level the playing field through more meritocratic entry and result in ‘a net gain for historically excluded workers’ (Redbird, 2017, p. 19). In this article, we pursue this understanding of licensure, by examining the immigrant-majority earnings gap across occupational settings.
The research of Moore, Pearce, and Wilson (1981), who investigate the earnings of women in regulated occupations in the United States, is also relevant in this regard. They find that licensure both raises and compresses the wage levels of women, adjusted for human capital characteristics. Their analyses show that human capital characteristics are only of limited relevance for explaining wage differences within licensed occupations. This suggests that employers and clients perceive licensed workers as more homogeneous than other workers and are therefore less attentive to quality differences. The stricter selection may also reduce actual quality differences. It might also reflect the ability of a regulated occupation to establish a uniform pricing system (Moore et al., 1981, p. 378). As mentioned previously, occupational status becomes the defining trait of the incumbents, not gender or ethnicity. One of the arguments of the ‘professional shelter’ hypothesis put forward here is that ethnicity may lose relevance as a status marker for ethnic minority persons with professional titles and status group memberships that such titles entail, especially among the prestigious ‘elite’ professions. Collins (1979, p. 72) underscores the fact that education may replace ethnicity as a status culture by calling education ‘pseudoethnicity.’ Professional kinship can thus trump ethnic kinship. We expect this mechanism to be strongest in the licensed professions.
Unionization
Unionization raises mean wages and compresses the wage dispersion, as shown by a multitude of studies in several countries (Askildsen & Nilsen, 2002; Barth, Raaum, & Naylor, 2000; Kleiner & Krueger, 2010; Rosenfeld, 2014; Wallerstein, 1999). Hence, unions will decrease the wage differentials within the unions between immigrants and the majority (Metcalf, Hansen, & Charlwood, 2001; Rosenfeld, 2014; Tomaskovic-Devey et al., 2015). According to Weeden (2002), unionization works through the mechanism of restricting supply but does not increase demand for a service. Unions do not monopolize an occupation as licensure does, but negotiating wages and working conditions is arguably the main task of a union, and through unions, employees can impose collective wage demands on employers. In a company with collective agreements, there is no alternative cheaper labor to the unionized workers. A strong union, in terms of high membership rates, constitutes a bigger threat to the company and is therefore more likely to gain acceptance for wage demands. Unions also engage in ‘opportunity hoarding’ to ensure that profits are more evenly distributed. Previous research has shown that unions are primarily a collective good, as union density and not individual membership results in a wage premium, conditional on union coverage (Barth et al., 2000). For that reason, the effect of union density is likely found at the occupational level, not the individual level.
Centralized wage bargaining systems will tend to reduce within-occupation wage inequality because negotiations take place between unions and employers’ peak associations with reference to occupational skill level, not individual skill level. Thus, employees are treated equally with respect to credentials, experience, and job level (Wallerstein, 1999). The Norwegian system of wage bargaining is highly centralized in comparison with other countries (Barth, Moene, & Wallerstein, 2003). Approximately 58% of Norwegian employees are covered by collective wage agreements in the private sector, and 100% in the public sector (Nergaard & Stokke, 2010). Norway has no minimum wage, but firms that are part of an employers’ confederation are liable to follow the collective wage agreements. Under the Scandinavian collective bargaining regime, settled agreements obtain ‘force of law’ in relation to wages and working conditions specified in the agreement, and neither party can terminate a settlement (Scheuer, 1997). Because an agreement is settled between the union and employers’ associations, employees in a company belonging to different unions can be covered by different agreements. The strength of collective agreements thus formally regulates the relationship between employers and employees at the company level, and often also at the occupational level.
The role of unions is to unite workers and show solidarity in wage setting. In theory then, in occupations with a high degree of unionization, pay levels should be equal, also for immigrants. Previous research has shown that unions decrease the earnings inequality between minority ethnic and majority ethnic workers, mainly by compressing the wage structure in the unionized sector (Metcalf et al., 2001). The ‘protectionist’ theory of unions suggests that immigrants will seek to become union members because their chances of receiving fair treatment are greater in a unionized occupation or company (Rosenfeld, 2014). Unions stand for solidarity and wage equity from transparent criteria, usually education, occupation, and experience (Avent-Holt & Tomaskovic-Devey, 2010; Høgsnes, 1989; Metcalf et al., 2001; Rosenfeld, 2014). Unions should thereby reduce the extent of discriminatory wage setting against immigrants. Because employers are bound by the collective wage agreement, unfair wage setting can be overruled in unionized companies. However, fewer immigrants are union members (Nergaard, Barth, & Dale-Olsen, 2015).
Credentialization
Credentialization implies that educational credentials are used as a source of closure. Occupational positions are reserved for those with the right amount and type of education (Collins, 1979). From this, it is clear that this strategy is mostly found in professional and associate professional occupations, although some craft and trade occupations are also strongly selective on education. Credentialization raises mean wages by placing restrictions on supply and signaling quality of service, but it does not channel demand to the occupation in the same way that licensure does (Weeden, 2002).
In contrast to licensure and unionization, this closure device does not have any legal/formal support. While restrictions on supply can be beneficial to immigrants, it is more uncertain whether this will be as beneficial as under licensure because the employer has alternatives. Moreover, a licensed occupation has a stronger hold on supply because the qualification of prospective members partially takes place in the occupational practice. In many credentialized occupations, qualification and practice take place in separate domains, as the education is not directed at one occupation in particular. In the words of Murphy (1988), the organization is ‘atomistic’ (p. 186). However, educational credentials should also reduce the information asymmetry between employers and employees in the same manner as licensure does. Lastly, because the privileges of credentialized occupations stem from ‘quality of service,’ systematic underpayment of immigrants could ultimately undermine this basis.
Certification
Certification mimics licensure by sending a signal to employers and clients that the occupation itself considers the complexity of the task to be so advanced that they recommend the use of certified practitioners. Certification can increase mean wages through the mechanism of channeling demand or signaling quality of service (Weeden, 2002). Voluntary certification often originates in the occupation itself, professional associations, or trade organizations. Because certifications are voluntary, the occupation has no means of restricting supply, as employers and clients are free to use uncertified practitioners. The value of the certification is determined by whether or not gatekeepers consider it to be important.
Certification can reduce wage inequality between immigrants and the majority by adding an alternative source of information on productivity, thus reducing the value of immigrant status as a signal. Increased demand for a service can be beneficial, but without a simultaneous restriction on supply, it is questionable whether this will increase relevant employment for immigrants. Because certification is voluntary, it is even possible that it has an inequality-increasing effect. If certificates raise wages, and if natives are more likely to hold certificates (e.g., because certificates are expensive to attain), then certificates could have an inequality-increasing effect, regardless of whether they affect supply.
Preliminary Recapitulation
We have discussed how the closure devices leading to higher wages in ‘closed’ occupations may also affect the native–immigrant wage gap. To have the sheltering effects we predict, the occupational closure devices have to overshadow employers’ or customers’ inclinations to discriminate, either by affecting supply and demand or by signaling high quality of service. This discussion is illustrated in Figure 1, which shows how the different closure strategies map onto these two main dimensions of occupational closure.
The closure devices.
In the figure, we see that certification is only high on the signaling quality dimension, whereas unionization is high mainly on the supply/demand dimension. Licensure is high on both. Credentialization is most difficult to map onto the two dimensions because it is what we label a social closure device without any legal support. It will be a question of degree to a larger extent than the other closure devices and will probably have greater variability among occupations.
The main contribution of this research is to establish a theoretical argument for the hypothesis that closed occupations may serve as a shelter against wage discrimination for the immigrants who have gained access. Because of this, we have devoted less attention to the effect sizes found in previous studies. The effects of different closure devices certainly vary a great deal across countries and occupations but also over time and with the applied methodology (see, e.g., Redbird, 2017 for licensure). It is therefore likely that the ‘shelter’-effect of occupational closure also varies a great deal between closure devices, occupations, and countries. Here, we will examine this hypothesis in the Norwegian labor market, which, as will be discussed, we find particularly well suited for these analyses.
Are Immigrants in Closed Occupations More Selective?
The discussion so far is based on the assumption that the institutional arrangements associated with the four closure devices under study impact on the wage gap between the majority and the ethnic minority groups. However, the wage gap might also be affected by a stricter selection among the immigrants who enter closed occupations. This would reduce the occupation-internal variation in human capital and productivity. If that is the case, the mechanisms assumed in human capital theory may reduce the occupation-internal wage differences.
Licensed occupations have clearly defined entry requirements relating to some form of education or training, and possibly also citizenship or registration fees to which prospective workers must oblige. To foresee these entry barriers might prove a greater obstacle to immigrants who likely have less economic and cultural resources to invest in education or training. Similarly, for credentialized occupations, high credentialization implies that a high share of the workers share educational credentials, and such selection processes through the educational system may result in workers with similar productivity. Selection might also be an issue for occupations with high union density, as employers might be more discriminating when hiring for positions where the union has negotiated wages and benefits, especially for low-skilled occupations where the price of labor is relatively high (Barth et al., 2003).
To separate between selection effects and ‘true’ closure effects would require within-occupation variation stemming from a change in the degree of closure. In the 6 years under study, only one occupation goes from being unlicensed to licensed. The change in the other closure devices is also not sufficiently large to facilitate the use of this identification strategy. However, we can estimate the wage premium associated with occupational closure as individuals change occupations. For instance, if someone switches occupation from juridical deliberation and planning, which do not require licensure, to becoming a lawyer, we can observe the wage increase associated with moving into a licensed occupation. The interaction between ethnic minority origin and the closure devices will reveal whether there is an additional wage premium for the minority groups.
The Norwegian Context
We find the Norwegian labor market particularly well suited for an examination of our sheltering hypothesis. The reason for this is the country’s particular industrial relations (i.e., its high degree of centralized wage bargaining, even among professional white-collar workers), and the strong egalitarian tradition in favor of small income differences and a relatively compressed wage structure (Barth et al., 2003). Norwegian culture has been described as putting strong emphasis on equality defined as sameness (Gullestad, 1992). Even compared with the other Scandinavian countries, Norwegians consider their own country to be more egalitarian (Hjellbrekke & Korsnes, 2014, p. 56). We thus test our argument in a conservative environment, meaning that effects are possibly even greater in less egalitarian societies.
Norway has a long history of immigration from other Scandinavian countries and European countries in general, while immigration from non-Western countries is much more recent. The era of immigration from non-Western countries began with labor migrants from Turkey, Pakistan, and India in the 1970s, and with refugees from Vietnam and Iran in the late 1970s and 1980s. Since 1975, immigration has continued through family reunification and refugees/asylum seekers, as Norway no longer accepts regular labor migration from non-Western countries. In the 1990s and 2000s, Norway has increasingly received refugees from Iraq and African countries, especially Somalia. The Norwegian-born children of immigrants are still very young. In 2010, only 16% were 20 years or older (Horgen, 2011).
Immigration from European and Western countries is mainly driven by targeted labor markets and educational mobility, with the exception of refugees from the Balkan war in the 1990s. Since the European Union (EU) enlargement in 2004, a large number of labor migrants from Poland have established themselves in Norway. Immigrants from Poland have become the largest immigrant group in Norway, but Swedes, Danes, Germans, Bosnians, British, and Russians make up a substantial share of European immigrants to Norway (Daugstad, 2008).
Bratsberg, Raaum, and Røed (2010, 2014) have studied labor market outcomes over the life cycle of the non-Western immigrants who came to Norway in the 1970s and 1980s and find a pattern where immigrants are assimilated early on but that their employment rates drop sooner and at a larger rate than the majority. Immigrant employment is more sensitive to business cycle fluctuations, and economic recessions have more permanent effects on their employment. The children of these immigrants, however, are more similar to the majority, both in education and in earnings (Bratsberg et al., 2014). Hermansen (2016) compares educational attainment and earnings in the Norwegian majority population with that of minority groups and their descendants and finds support for an optimistic scenario in which immigrants’ offspring enjoy at least the same upward socioeconomic mobility as the majority peers with the same socioeconomic origin. Even though the immigrants have lower earnings and educational attainment, their children do considerably better, and Hermansen finds a convergence in the minority–majority gap.
In the period studied here, 2007 to 2012, the Norwegian economy had high employment levels and low unemployment levels. The employment rates in the majority population in 2012 were 82% to 77% among men and women aged 20 to 66 years. The corresponding rates were lower in almost all immigrant groups, varying from 75% (among immigrants from Western European countries) to 46% (immigrants from Africa). Only immigrants from the Nordic countries have higher employment rates compared with the majority population (Statistisk Sentralbyrå, Table 09837, 2018). The unemployment rates in the majority population fluctuated between 1.2% in 2008 and 2.5% in 2010. 3 Among immigrants from Western countries, the unemployment rates were 1.7% in 2007 and 6.5% in 2010. The lowest unemployment rate for non-Western immigrants was 5.9% in 2007 and the highest was 9.9% in 2010 (Statistisk Sentralbyrå, Table 07115, 2018). The unemployment rate is higher at the end of the period, but the relative difference between the majority and immigrant groups show stability.
There are probably several reasons for the observed variation in labor market integration by country of origin. One reason is selection, as labor migrants generally integrate more easily into European labor markets. A second reason is the short residency of some groups. A third reason is the institutional structure in the countries from which people emigrate (e.g., the educational system; Blom & Henriksen, 2008; Daugstad, 2008).
Data and Variables
We analyzed individual-level data derived from Norwegian public administrative registers that contain complete individual histories of education, labor market status, occupation, wage, and demographic information for the entire working population. The data cover the years 2007 to 2012, which give a strong empirical basis for estimating wage effects. The data consist of all persons born from 1945 onward. Information about the sources of the administrative data is available in online Appendix A.
Specification of the Sample
The individual-level data comprise all individuals with resident status in Norway in 2007 to 2012 who were employed for at least 1 week and for a minimum of 4 hours during the calendar year and had a gross yearly wage above the national insurance basic rate (1G = $9,600 in 2012). Immigrant workers with temporary residence of less than 6 months are not included. It is mainly residents of EU countries who have temporary access to the Norwegian labor market, as, since 1975, Norway no longer accepts regular labor migration from non-EU countries. Moreover, we have restricted our sample to individuals aged 25 to 67 years. This limitation is imposed with regard to the second-generation immigrants, a large share of whom are employed part-time and in unskilled occupations. People aged 25 years and older are likely to have a permanent labor market attachment. Sixty-seven years is the standard retirement age in Norway. 4
Information on education is lacking for some immigrants in the registers, and they are for that reason excluded from the analyses. It is important to adjust the analyses for educational attainment, as the wage level within an occupation may vary depending on the formal skill level of employees. We are interested in estimating the earning differentials, net of individual skill. Hence, it would exaggerate the earning differentials if immigrants had lower education levels than majority colleagues. Immigrants who have their foreign education approved by the Norwegian Agency for Quality Assurance in Education are included. The share of immigrants who complete their highest level of education in Norway varies with country of origin, and it is naturally higher among those immigrant groups with long residency (e.g., Vietnam, 63%) and lowest among recently arrived immigrants (e.g., Lithuania, 3%; Steinkellner, 2015). We compare the immigrant groups with a 30% random sample from the majority population, as the use of all observations is computably inefficient. The results are unaffected by this random sampling.
Specification of the Occupational Level
The measure for occupation is central to this study. The classification structure for occupations is Standard for Occupational Classification (Standard for yrkesklassifisering [STYRK]), which is a Norwegian adaption of the Isco-88 code. The occupational classification is structured along two dimensions: skill level and skill specialization. Employers have reported occupations since 2003. However, public sector employers started using the standardized occupational codes at a later stage. The data cover most employees, except state employees at universities, hospitals, ministries, and directorates (Villund, 2014). Instead, employers in the public sector report occupational codes from collective wage agreements, and because these codes are mainly used for wage setting, they are less detailed than the STYRK standard. For some employees, these codes are sufficiently detailed to allow imputation. For instance, the titles professor or assistant professor, senior physician, or nurse correspond to STYRK codes. In total, 7% of occupations are imputed from exact matching of job title or matching based on job title, education, and industry. After imputation, information on occupation is missing only for 7% of employees. They are mainly consultants, officials, and administrative staff in public sector employment. It is unlikely that this group of employees will change the general pattern of earnings equality or inequality found in the analyses.
The structure consists of 356 occupation unit groups. We have excluded the armed forces and legislators, as well as skilled agricultural and fishery workers. 5 This leaves 328 occupational groups to be included in the final analyses.
Measurement of Closure at the Occupational Level
We measure four closure devices in our analyses, namely licensure, certification, unionization, and credentialization. The occupational-level closure variables are aggregated from the register data or derived from a data set collected especially for this research that contains information about licensure and certifications coded from Norwegian laws and regulations (see Alecu & Drange, 2016).
An occupation is classified as licensed if the right to practice it is regulated by the authorities, either by law or by regulations to a law. We mapped the extent of licensure using the database on Norwegian laws and regulations (www.lovdata.no). The inclusion criterion is that licenses are granted to actual persons and not legal persons. An occupation can be fully licensed (e.g., medicine) or partly licensed (e.g., accounting, where only chartered accountants need a license). The licensure measure was coded based on occupational titles within occupational unit groups and aggregated to the occupational unit group level.
Because certificates are not part of the educational system, we are unable to identify certificate holders in the registers. That is not problematic because the theory suggests that the existence of certificates raises an occupation’s esteem, which implies that all occupational incumbents reap the benefits of this arrangement (Weeden, 2002). The data on certification were coded based on extensive searches. We searched the Internet for the combination of the occupational title and the key word certific* (Norwegian: sertifi*) for each occupational unit/title in the STYRK standard. This approach has been used in previous studies (Carter, 2005; Weeden, 2002). To have an impact, a certification must be known to prospective certificate holders and the public. Information should therefore be readily available. The inclusion criteria were that the certificate was (a) issued by an association, union, private, or public educational institution situated in Norway, (b) awarded to an individual practitioner, not a business, and (c) not related to a specific method, product, or health and work safety. If a certificate was found, the occupational title was coded 1 and 0 otherwise. Examples of certified occupations are appraisers and coaches. To reduce bias, two independent coders based on the prescribed rules coded the prevalence of certifications. The correlations between the two coders were between 85% and 95% for the crafts and trade occupations, and the associate professional and technician occupations, respectively.
Unionization is used as a proxy for centralized, collective wage bargaining. The measurement of unionization is derived from the register data, which contains information on union membership. We have aggregated the share of union members per occupation for all occupations.
Lastly, the measure of credentialization is a continuous variable that measures the strength of educational segregation in the labor market. The measure is developed by DiPrete, Bol, Ciocca, and van de Werfhorst (2017, Appendix A), and it combines information about education level and field of study and how this relates to labor market positions. More precisely, it measures the extent to which incumbents of an occupation have similar or diverse educational credentials. Occupations whose incumbents have similar levels and types of education have a higher score for this variable than occupations with a wide range of educational levels. Thus, the higher the score, the more closed the occupation is. Because a few, highly segregated occupations might have a high score, we take the natural logarithm to reduce the influence of outliers.
Because of aggregation from individual-level data, the variables measuring licensure, certification, and unionization vary between zero and one. The closure effect is therefore best understood as average treatment effects for those occupations subjected to closure. Moreover, because there is limited variation in the closure devices within occupations across time, we measure all closure devices as occupational averages; hence, these measures are fixed for the occupations.
Operationalization of Wage
The dependent variable is yearly gross wages earned from employment. The variable contains benefits received in place of wages, such as parental benefits, sickness benefits, and unemployment insurance benefits. The dependent variable is consumer price index adjusted to 2013 level. The data do not link wages and employment relationships, so annual wages can potentially result from multiple employment relationships. This does not affect the results (cf. the chapter on robustness checks). The data identify the employment relationship that contributes most to the individual’s annual earnings, and we use this employment relationship in the analyses.
Operationalization of Immigrant Origin
The definitions of immigrants and ethnic minorities in this study are based on their own and their parents’ country of birth, and not on citizenship. Nevertheless, most ethnic minorities with legal and permanent resident status have denizenship that gives them the same social, economic, and civil rights as citizens (Brochmann & Hagelund, 2010).
The variable for immigrant origin distinguishes between first- and second-generation immigrants from Western Europe, North America, and Oceania, on one hand, and immigrants from Eastern Europe, Asia, Africa, and Latin America, on the other. Our hypotheses do not distinguish between immigrants depending on country of origin, and neither should they. The non-Western category undoubtedly masks a great deal of heterogeneity, and if occupational closure is indeed beneficial to those ‘protected’ by it, this should apply to all country groups. To check consistency of findings at an aggregated level, we have rerun the analyses using a multitude of 19 different countries and country groups, in addition to the Western and non-Western dichotomy. 6 The results from these subgroup analyses are largely the same as in the analyses with the less fine-grained minority classification. These results are available in Table 4 in the online appendix.
Control Variables
At the individual level, we control for education, labor market experience, age and age squared, geographical location of the workplace, public or private sector employment, gender, and family status, and we include an interaction term between gender and family status, as marriage and children are known to have opposite effects on men and women. We use a detailed control for education, with 30 combinations of field and level. Those with a generic high school diploma constitute the reference category. We include time since graduation and time since graduation squared as proxy measures of labor market experience. We also adjust the estimates for the number of days employed and average weekly work hours, as both measures affect the wage level of employees. Our focus here is on closure mechanisms and not on human capital, but we still control thoroughly for individual-level human capital in the analyses to reduce the likelihood that any observed wage differences are due to differences in education and experience. At the occupational level, we control for the average share of females in the occupation, as female-dominated occupations tend to have lower wage levels.
Methods
Multilevel Regressions
We investigate the development of wage inequalities between and within occupations dependent on incumbents’ country origins. We use multilevel models with a three-level structure, that is, individuals nested within occupation-years, nested within occupations, to estimate occupational wage differences. Because individuals in the same occupation tend to have similar wage levels, the usual ordinary least squares assumption of identical and independently distributed standard errors is violated. According to Aguinis, Gottfredson, and Culpepper (2013), intraclass correlations of > 0.1 indicate clustering. The intraclass correlation of the null model is 0.35 at the occupational level; hence, a multilevel specification is appropriate. We specify a random intercept model, allowing the occupations to have different intercepts, but restraining the slopes of the estimated coefficients to be fixed across occupations. For the closure devices, this is predicated on closure not having any within-occupations variance.
Cross-level interactions between the Level 1 variables of ethnic origin and the Level 3 variables measuring the average strength of the closure devices are required to test the hypotheses that closure strategies moderate wage inequality. Positive and significant coefficients will support our hypotheses, as we expect to find lower levels of ethnic wage inequality in those occupations compared with occupations without restrictions on access.
Testing the hypotheses requires cross-level interactions with each of the four closure devices. The analytical design is to test the interactions in separate models. Although this could lead to bias of the interaction effects due to correlation between the devices, testing all four interactions simultaneously will clutter the interpretation of the effects. Moreover, the statistical power is usually lower for cross-level interactions; hence, testing several effects simultaneously can reduce the likelihood of finding nonzero results, that is, type B error (Aguinis et al., 2013, p. 1515). However, we have modeled all interactions simultaneously to check the robustness of the separate models. The simultaneous estimates do not differ substantially from those of the separate models (see online Appendix Table 3).
Equation 1 specifies the multilevel regression model:
The dependent variable ln(wage)itj is the logarithm of annual wage observed for individuals (i), nested within occupation-years (t), nested within occupations (j). β0 is the intercept. The vector xitj represents the individual characteristics estimated by β1. The coefficient β2 gives the estimate for immigrant origin. The measure for the occupational closure devices is averaged over occupation-years and is fixed for all occupation-years. The vector
Fixed-Effect Regressions
We address the question on selection into ‘closed’ occupations with individual-level fixed-effects regression. This model uses only within-individual change, which means that it removes all time-invariant individual characteristics, both observed and unobserved characteristics. This statistics do not estimate mean occupational wage levels and cannot show whether the ethnic wage gap in an occupation decreases with licensure. Instead, these models do provide evidence for whether the wage premium associated with licensure is higher for immigrants while keeping the unobserved skill-levels constant. Hence, if closed occupations produce more similar earnings profiles because these occupations recruit high-ability immigrants, we would not expect to observe a higher wage premium for immigrants moving into closed occupations.
Descriptive Statistics on the Share of Immigrants Covered by Occupational Closure and the Share of Immigrants and Children of Immigrants Across the Occupational Structure.
Analyses of variance reveal significant differences between all groups.
Analyses of variance reveal significant differences between all groups, except between natives and first-generation Western immigrants.
Analyses of variance reveal significant differences between all groups, except between natives and first-generation Western immigrants and between first- and second-generation non-Western immigrants.
The χ2 test is significant at the p ≤ .001 level.
Equation 2 specifies the fixed-effects model:
The dependent variable ln(wage)it is the logarithm of annual wage, observed over individuals (i) and time (t). The intercept is β0. The coefficient β1 gives the estimate of the vector of time-varying individual effects, xit. The coefficient β2 is the wage effect for individual’s movement into more or less closed occupations, and β3 gives the additional wage effect of occupational closure that is associated with immigrant origin. The model has two error terms, μi is the unobserved individual error and ɛit is an idiosyncratic error.
Descriptive Results
Table 1 summarizes the closure devices and presents the extent to which the majority population and the immigrant groups are found in occupations covered by the different closure devices. The lower part presents the distribution of the majority population and the immigrant groups across the major occupational groups (the first digit of the ISCO-88 classification).
The first row in Table 1 shows the share of immigrants in occupations with licensure requirements. Because licensure is necessary for all employees in a licensed occupation, this share is equal to the share of immigrants with a license. Norwegians have the highest share of licensed employees, followed in descending order by first-generation Western immigrants, second-generation non-Western immigrants, second-generation Western immigrants, and, finally, first-generation non-Western immigrants.
The second and third rows of Table 1 show the mean level of certification and unionization, respectively, in the majority and the minority groups. These variables are measured at the occupational level and do not reflect the share of immigrants who are union members or certified. There is less variation in the share of employees in occupations where certification is available. Second-generation immigrants are more concentrated in certified occupations than the three remaining groups.
The credentialization measure in the fourth column shows the concentration of the education to occupation linkage at the occupational level. This measure is not interpretable by its mean value (DiPrete et al., 2017). The measure is stronger, that is, further from zero, if there is a high correspondence between field of education and occupation, and weak otherwise.
Generally, first-generation Western immigrants and the majority have larger proportions in occupations covered by the closure devices, than the non-Western immigrants do. The patterns revealed are in line with our expectations, given that Western immigrants are more highly selected on skills and occupations than the non-Western immigrants are.
The correlations between the different closure mechanisms are low to moderate. The correlation between licensure and union density is 0.55, the correlation between licensure and credentialization is 0.60, and the correlation between credentialization and union density is 0.61. The correlation between licensure and certification is −0.17, which reflect that licensed occupations do not tend to be certified. Importantly, this means that licensure, credentialization, and union density partly cover the same immigrant and majority groups, whereas certification targets a different segment of the working population.
Descriptive Statistics and Mean Scores for All the Individual and Occupational-Level Control Variables.
Three-Level Model, Maximum Likelihood Estimation; 30% Random Sample of the Majority Population.
p ≤ .05. **p ≤ .01. ***p ≤ .001.
Two-Level Generalized Least Squares Model With Individual Fixed Effects (Full Sample).
p ≤ .05. **p ≤ .001.
Results
The results from the regression analyses are presented in Table 3. We present only the coefficients of interest, but the models including control variables are available in online Appendix B, Table 3.
Model 1 (column 1) shows the results from a hierarchical regression nested at the occupational and occupation-year level with all explanatory and control variables included. The model shows that first- and second-generation non-Western immigrants earn approximately 8.1% 7 and 4.2% less, respectively, than the majority population. The wages of Western immigrants are 1.8% lower. Because we control for individual and occupational-level characteristics, this result is not explained by majority Norwegians, non-Western, and Western immigrants being distributed across different occupations. It shows that, even after adjusting for this, the non-Westerners earn less on average than the other groups.
Turning to the closure devices, the analysis in Model 1 shows that licensure, certification, and unionization significantly raise mean wage levels. Credentialization does not have any significant effect on earnings.
Models 2 to 5 (columns 2 to 5) introduce interaction terms between the immigrant dummy variables and the closure devices licensure, unionization, certification, and credentialization, respectively. To keep it clear and simple, we have estimated four interaction models between the closure devices and the immigrant origin dummies, that is, one model for each device, while controlling for the other devices. The changes in log likelihood between the model without interaction terms (Model 1) and models with interaction terms (Models 2 to 5) show that Models 2 to 5 provide a better fit to the data compared with Model 1. 8
Model 2 shows that the earning differentials between non-Western immigrants and the majority in unlicensed occupations are larger than the average earnings gap estimated in Model 1, adjusted for controls. In nonlicensed occupations, the mean wage of first- and second-generation non-Westerners is approximately 10.3% and 4.4%, respectively, below the majority. The interaction terms between immigrant origin and licensure show that first- and second-generation non-Westerners in licensed occupations earn an average premium of 11.5% and 1.5%, respectively, on top of the licensure premium. Hence, there is no earnings gap between the majority and first-generation non-Westerners in licensed occupations, and the difference between the majority and second-generation non-Westerners is significantly reduced. Interestingly, the interaction term between first-generation Western immigrants and licensure is zero.
Model 3 estimates interactions between immigrant origin and union density. In line with our expectations, the mean wage gaps between the majority and first- and second-generation non-Western immigrants are large and significant in nonunionized occupations. The interaction term between first-generation non-Westerners and union density is positive and significant (21.3%), however, and in occupations with complete organization, non-Westerners and the majority receive equal mean wages. In other words, the mean wage differentials between non-Western immigrants and the majority increase as unionization decreases, and the estimated earnings gap as unionization goes to zero is larger than the average earnings gap estimated in Model 1.
Model 4 estimates interactions between immigrant origin and certification. The wage gap between the majority and first- and second-generation non-Western immigrants is considerable, and the interaction term between first-generation non-Westerners and certification is negative and significant. Thus, certification does not, on average, equalize non-Westerners’ earnings to that of the majority.
Model 5 shows the interactions between credentialization and immigrant origin for log mean wages. The interaction terms between credentialization and immigrant origin are generally small, although significant. The wage returns to credentialization reduces the wage gap but does not equalize earnings in more strongly linked occupations.
The results from Models 2 to 5 are graphically displayed in Figure 2. The x axis shows the wage gap. For the three closure devices licensure, union density, and certification, we simply estimate the difference between people in occupations that are fully covered by licensure, union density, or certification and people in occupations without them. For the credentialization device, we compare log linkages equal to 0 and 1, where the latter indicates stronger credentialization.
Immigrants’ wages compared with majority wages.
The upper left graph of Figure 2 shows the wage gap between immigrants and the majority in occupations that are not licensed and contrasts this with a scenario where occupations are fully licensed, for instance, medical doctors and nurses. The figure clearly shows that the earnings gap between the majority and non-Westerners is reduced in licensed occupations.
The upper right graph in Figure 2 shows the wage difference in occupations with no union density compared with occupations with full union density. First-generation and second-generation immigrants of non-Western origin have significantly lower earnings than the majority in occupations where there is no union density. In the contrasting case of complete union density, the mean wages of the majority and first-generation immigrants are equal. Union density raises the wages of second-generation immigrants but does not fully eradicate the earnings gap to majority colleagues.
The lower left graph of Figure 2 presents the predicted marginal effects of certification. As can be seen, certification does not reduce the wage differentials between the majority and non-Westerners.
The lower right graph of Figure 2 shows the credentialization effects. The wage differentials between the majority and the immigrant groups are somewhat reduced in more strongly credentialized occupations.
Fixed-Effects Estimates
To address the selection issue of immigrants into closed occupations, we have specified an individual fixed-effects model over the years 2007 to 2012 to account for the possibility that equalization of wages is due to more productive immigrants being recruited to the licensed occupations and occupations with high union density. This estimation procedure shows wage returns as individuals move into or out of occupations encircled by closure institutions, or returns to changes in the strength of these intuitions over time. The results are presented in Table 4 (see online Appendix B for full table).
The results show a wage premium to licensure of 1% for the majority, but the wage premium for non-Western immigrants is approximately 6.2%. First- and second-generation Western immigrants also receive an additional wage premium when transferring to a licensed occupation. If an employee of majority origin switches from an occupation with no union density to full union density, this would result in a wage premium of 8.1%, and first-generation non-Western and Western immigrants would receive an additional premium of 8.7% and 2.2%, respectively. A one-unit increase in the credentialization measure is associated with a wage increase of 0.3% and an additional increase of 2.5% and 1.4% for the first- and second-generation non-Western immigrants. Both the direction and magnitude of these results are in line with the results from the multilevel models in Table 3. The exception between Tables 3 and 4 is that first-generation non-Western immigrants also receive a wage premium when moving into certified occupations of 1%.
Based on these models, we can once again conclude that the results reveal a higher wage premium in licensed and unionized occupations, with an additional wage premium to immigrants as they move into occupations where these institutions encompass more employees.
Robustness Checks
We have performed four different robustness checks, and the results are available in online Appendix B Table 5, Table 6a to 6e, Table 7a and 7b, and Table 8, respectively. First, because the non-Western category can conceal internal diversity, we reestimated the model for 19 different countries and country groups (see Note 6). The interaction terms between country origin and both licensure and unionization are positive for all country clusters, and the effect is significant for most of these interactions. The results for certification and credentialization are more mixed.
Second, we also ran models for all the major occupational groups, except for Group 4 ‘Clerks’ and Group 9 ‘Elementary occupations,’ to check that the findings are consistent across the occupational hierarchy. 9 The results from these analyses also verify that licensure and unionization reduce the wage differences between majority and non-Western employees in most major occupational groups.
Third, we ran models for men and women separately to investigate whether the closure devices had different effects for the two groups. The results were highly similar, which suggest that the sheltering effect of occupational closure does not depend on gender.
Finally, we ran models for full-time (>29 hours a week), full-year (>329 days) employees to ensure that the results are not biased by, for example, unmeasured job change or multiple employment relationships. The results from this limited sample were concurrent with the results for all employees.
A possible limitation of this research is that we do not control for human capital beyond education and potential labor market experience. The wage gap between the non-Western immigrants and the Norwegian majority might be augmented by unobserved skill differentials in these two populations, and different selection mechanisms may reduce the ethnic pay gap by reducing the unmeasured differences between natives and minority groups within, for example, the licensed occupations. The individual fixed-effects estimates remedy this deficit and strengthen the conclusion that labor market regulations such as licensure and union density curbs wage discrimination against workers of non-Western origin.
Discussion
This study shows that occupation matters to earnings inequality between the ethnic majority and minority employees, and our analyses have found the largest wage differences in unlicensed occupations with low union density.
In general, immigrants’ earnings are substantially below the earnings of the majority, even in similar occupations and after adjusting for human capital controls. Interestingly, and in line with our predictions, the results show that immigrants benefit more from entering licensed and unionized occupations.
Our ‘labor market shelter’ hypothesis stipulated that we would observe a reduction in wage differentials between the majority and non-Westerners in occupations with licensure requirements and in occupations with strong union representation. The analyses revealed that immigrants receive a higher wage premium from being in licensed and unionized occupations compared with majority colleagues. The consequence is that licensure and unionization equalizes immigrant and majority mean log wages, and these results are uniform and positive, and consistent across the occupational hierarchy. Moreover, the individual fixed-effects regressions show that first-generation non-Western employees receive a larger wage premium than majority colleagues when moving into occupations associated with stronger licensure or unionization. Thus, the observed wage convergence is not due to selection effects. The findings support our hypothesis of licensed and unionized occupations being a ‘labor market shelter’ for non-Westerners.
Finally, our analyses showed a limited negative effect of certification and a small, positive effect of credentialization on the mean wage differentials between non-Westerners and the majority. Although the interaction effects were statistically significant, these closure devices did not converge the earnings gap between the majority and immigrants. The low estimate for credentialization is interesting in light of the moderate correlation this device has with both licensure and union density because the correlation coefficients imply that these devices partly cover the same occupations. Yet, licensure and union density reduce the ethnic pay gap, whereas certification does not. Certification has a low correlation with the other devices and therefore mostly compare wage outcomes for individuals in different occupations.
Overall, the results support our argument, which is that occupational closure reduces within-occupation wage inequality and that the strongest effect is from the closure devices that have statutory authority, namely licensure and unionization.
Conclusion
In Figure 1, we illustrated how different closure strategies map onto the two main dimensions of occupational closure, manipulating supply/demand and signaling quality of service. We anticipated that both supply restrictions and signaling of quality would contribute to decreasing ethnic wage differentials in regulated occupations. Whether it is the ‘limitation of supply mechanism’ or the ‘signaling mechanism’ that leads to lower wage differentials in licensed occupations is difficult to disentangle. However, comparisons with the results for certification and unionization can give a hint. Although licensure is high on both the supply/demand and the signaling dimension, unionization is only high on the former while certification only on the latter. Unionization has a considerable sheltering effect, which we do not find for certification. Certification is supposed to signal quality in the same way as licensure without placing restrictions on supply (Humphris, Kleiner, & Koumenta, 2010; Kleiner, 2006). This suggests that the important equalizing mechanism is restriction of supply, assuming that the quality signals from licenses and certificates are of equal strength. However, it is doubtful whether certification has equal status to licenses among consumers, and, for that reason, is perhaps not in demand among employers. Another difference between certification and licensure is that whereas licensure is the same for all members of the occupation, certifications make it possible to distinguish between occupational members (not everyone in the occupation have the certificate). If certificates raise income, and if natives are more likely to hold certificates (e.g., because certificates are expensive to attain), then certificates should have an inequality-increasing effect, regardless of whether they affect supply.
Credentialization and certification classify as selective closure devices because legal/formal authority does not back these devices to the same extent as licensure and unionization. The main reason why the two latter devices have a stronger equalizing effect is that they effectively limit the employers and clients’ latitude in wage setting and hiring. The employer has no choice but to hire a licensed optician, and, considering that they are short in supply, the employer has a strong incentive to provide decent working conditions and wage levels.
Moreover, in the case of immigrants, the government ensures that the education of foreigners in licensed occupations is comparable with Norwegian education. For unlicensed occupations, it is up to the employer to verify that the education is comparable with a Norwegian degree. For foreigners, then, a license is a much more reliable signal than a university degree.
This article makes three contributions to the research literature. The main contribution is that the article expands on previous studies that have investigated the closure hypotheses about intraoccupational wage inequality in selective occupations (Drange, 2013, 2016; Karlsen, 2012). Occupational closure can actually reduce inequality, even if we consider selection.
Second, we investigate the context dependency of discrimination theories in the framework of occupational closure theory and pinpoint what closure mechanisms that have such an inequality reducing effect. The main difference between the occupational closure theory and discrimination theory is whether it is the occupation or the workplace that is the main arena for wage setting. A central tenet of RIT is that categorical status characteristics serve as the basis for opportunity hoarding and claims-making, and the strength of these claims depends on the social position of the status group. The relatively weaker bargaining position of immigrants explains their lower earnings (Tomaskovic-Devey et al., 2015). Tomaskovic-Devey et al. (2015) argue that the workplace is the origin of these differences and cite the increase in wage differentials in the most segregated workplaces in support of the theory. We emphasize that occupational-level institutions can substantially mediate immigrant/majority wage inequality. The mean wage level is more equal in occupations with strong institutional barriers because employers’ opportunity and incentive to discriminate in wage setting is lower when the occupation has control of supply of labor through regulations and through the occupation’s organizations participating in centralized wage negotiations.
A final contribution of this article is the use of Norwegian register data on the entire labor force coupled with data on occupational closure devices. So far, the theory of occupational closure has only been tested using survey data from the United States, Germany, and Great Britain (Bol & Weeden, 2015; Humphris et al., 2010; Kleiner & Krueger, 2010; Weeden, 2002). A major strength of using register data is that they cover the entire workforce so that even numerically small groups are sufficiently represented in our data. Another strength is that we estimate the effects for 19 country groups (in the online appendix). In our opinion, this is a more conservative test of the hypotheses because the category non-Western immigrants conceals much heterogeneity. To our knowledge, this is the first article to investigate the consequences of a wide range of occupational regulations on within-occupation wage differentials for the occupational structure, and our findings support those of previous studies looking at single occupations.
It is beyond the scope of this article to investigate whether occupational regulation affects other aspects of working life, such as job mobility, unemployment, and skill enhancement. Nor is it within the scope of this article to test the underlying mechanisms of occupational closure, that is, restricting supply and increasing demand and quality of service. In terms of enhancing demand and signaling higher quality of service, it is necessary to consider the clients and employers’ attitudes and preferences with regard to the different groups of suppliers. Furthermore, we would expect the same findings to apply to women and other groups experiencing wage discrimination. Exploring these questions is an avenue for future research.
We find the Norwegian context particularly well suited for these analyses because of the country’s industrial relations (i.e., its high degree of centralized wage bargaining, even among professional white-collar workers) and the strong egalitarian tradition in favor of small income differences and a relatively compressed wage structure (Barth et al., 2003). We thus test our argument in a conservative environment, meaning that effects are possibly even greater in less egalitarian societies. An avenue for further research is to replicate this study in less egalitarian and regulated economies.
Regarding policy implications, the solution to the majority–minority wage gap is not what one (to paraphrase Wilensky) might call “the licensing of everyone,” which of course is not feasible. It is also probable that employers with a preference for discrimination still have considerable leeway in decisions about hiring and promotion, even though other forms of within-occupational discrimination may be somewhat hampered by closure mechanisms. However, more centrally set wages seem to reduce ethnic wage differences, which may serve as a warning against attempts to abolish collective wage bargaining.
This research has implications for closure theory. So far, closure theorists have focused on exclusion and usurpation strategies as part of the general stratification processes in society. While it is acknowledged that closure may lead to greater equality among incumbents (Murphy, 1988; Weeden, 2002), this side of closure theory remains, both theoretically and empirically, highly underdeveloped. Little attention has been paid to how these groups are internally structured with regard to (in)equality among incumbents, and particularly among people with similar education or vocational qualifications. To grasp the link between closure and wage inequality requires an investigation of how closure affects internal distributions of wages (Bol & Weeden, 2015, p. 367). This study provides conclusive evidence to the hypothesis that the mechanisms specified in the theory on occupational closure moderates ‘demand-side’ explanations of labor market inequality. Frank Parkin (1979) has previously theorized this, but empirical investigations have been scarce.
Occupational closure, then, seems to be part of the explanation of the labor market disadvantage experienced by people of immigrant origin, but at the same time, such closure mechanisms serve as shelters for immigrants who have managed to get a foot in the door. Closure theory has mainly focused on the ability of closure mechanisms to raise mean wages (Bol & Weeden, 2015; Weeden, 2002). This does not necessarily imply equalization of wages among occupational incumbents (Kleiner & Krueger, 2010; Wallerstein, 1999). The results of our analyses indicate, however, that the mechanisms that raise mean wages also lead to more similar wage levels among immigrants and the majority population, even if we consider selection.
Supplemental Material
Online appendix -Supplemental material for The Sheltering Effect of Occupational Closure? Consequences for Ethnic Minorities’ Earnings
Supplemental material, Online appendix for The Sheltering Effect of Occupational Closure? Consequences for Ethnic Minorities’ Earnings by Ida Drange and Håvard Helland in Work and Occupations
Footnotes
Author Note
This article is the part of the project ‘Income Inequality in Professional and Vocational Occupations’ financed by the Norwegian Research Council, grant number 237039. This article has been presented at the Transitions in Youth conference in Brno 2015.
Acknowledgments
The authors would like to thank the participants for their helpful suggestions. They would also like to thank two anonymous reviewers, Thijs Bol, Kim Weeden, Jan O. Jonsson, Herman van de Werfhorst, and their colleagues at the Centre for the Study of Professions for their comments to an earlier version of this article. Any errors or omissions are the responsibility of the authors.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Supplemental Material
Supplemental Material for this article is available online.
Notes
Author Biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
