The Effects of Recognition of Foreign Education for Newly Arrived Immigrants

We analyze the effects of formal recognition of foreign higher education on employment probabilities and earnings for newly arrived immigrants in Sweden. Prior research has found that immigrants have lower returns on education if it was acquired in the country of origin than if it was acquired in the host country. One reason for this is that foreign credentials work poorly as productivity signals and riskaverse employers avoid employees with credentials they do not fully understand. A formal recognition statement can help overcome this problem by providing credible information about the foreign education, thus reducing uncertainty. Data consists of immigrants who, within the first ten years of residence in Sweden, had their foreign degree formally recognized during 2007–2011. Using fixed effects regressions, we estimate the treatment effect of official recognition to be 4.4 percentage points higher probability of being employed, and 13.9 log points higher wage for those with employment. We also find considerable treatment effect heterogeneity across subcategories of immigrants from different regions of origin, with different reasons for immigration and who obtained recognition during different economic conditions. Our conclusions are that the mechanism of employer uncertainty is real, and that recognition does reduce it. But as the signal of foreign education becomes better, other mechanisms such as human capital transferability problems and quality differences, and the ability to use foreign human capital, become more salient, leading to heterogeneous effects.


Introduction
Labour market integration of highly skilled immigrants is an interesting case for both researchers and policy makers.The prediction of country-specific human capital theory that education is not perfectly transferable across national contexts is a particularly intricate problem for immigrants who are high educated (Chiswick and Miller, 2008).At the same time, highly educated immigrants are coveted in receiving countries because of their possible contribution to the labour market, growth rate, etc.Thus, there is a common interest in understanding how to overcome the transferability problem in order to more quickly integrate highly educated immigrants.One such attempt is policies directed towards formal recognition of foreign educational credentials, a form of translation of foreign education into the language of the host society.However, there has been little evaluation of the effectiveness of this recognition.
The objective of this study is to analyze the effects of recognition of foreign higher education on employment probabilities and earnings for newly arrived immigrants in Sweden.Our contribution is two-fold.First, it is-to our knowledge-the first study that explicitly and comprehensively analyses the effects of recognition on labour market outcomes.Previous research on the effects of formal recognition of foreign education has not been able to provide a general account of its treatment effects, since it has been limited to subpopulations of immigrants and contexts in which recognition is reserved for specific immigrant categories (Kler, 2006;Green, Kler and Leeves, 2007;Kogan, 2012).In this study, we draw on a larger data set including all immigrants in Sweden who have obtained a recognition statement for their tertiary degree for a non-regulated profession between 2007 and 2011.In the Swedish system, recognition of educational qualifications is both voluntary and available regardless of origin or reason for immigration.Thus, our study is less bound by the limitations previous research had to face.
Second, we contribute to the discussion on why immigrants have lower returns to origin-country education by testing if a part of the problem is employer uncertainty about foreign credentials.There are plentiful studies on labour market integration that have concluded that foreign education is lower valued, but considerably fewer have been able to isolate specific mechanisms to answer why it is so.Recognition is an interesting case because it neither affects educational origin or quality, it can only work by reducing employer uncertainty about foreign credentials.Because our data is limited to education for non-regulated professions, we also avoid the effect of immigrants with recognition of foreign degrees getting formal access to otherwise unattainable occupations.i.e. overcoming closure.Thus, by evaluating the policy, we also test a key component of why immigrants get lower returns to their foreign human capital.

The Swedish Context
Swedish integration policy is directed towards a fast transition into the labour market, and has become increasingly so in the last years (Larsson, 2015: p. 311).Sweden consistently ranks among the highest countries on the Migrant Integration Policy Index, including labour market mobility measures and immediate access to the labour market for non-labour market immigrants (Huddleston et al., 2015).Sweden also has a low prevalence of occupational regulation (Koumenta et al., 2014), giving immigrants formal access to a substantial proportion of the labour market upon arrival.Regarding labour market results, the employment rate of immigrants in Sweden is close to the OECD average, but Sweden has at the same time the highest employment rate gap between immigrants and natives (OECD, 2014).
Within the labour market focus of integration policy, recognition of origin-country education has become an important policy tool for the integration of highly educated immigrants.Recognition of academic higher education for non-regulated professions is handled by the Swedish Council for Higher Education. 1 The recognition procedure ends with the issuing of a recognition statement, a formal document that includes information about where the tertiary degree was obtained, what level it corresponds to in the Swedish education system, and what field of study the degree is in.What is required is that the applicant's education has resulted in a tertiary degree by an accepted provider, e.g. a university, and that it is documented, e.g. in the form of a diploma.The Swedish Council for Higher Education controls the authenticity of the documentation, the accreditation of the education provider, and asserts the level of the education (Petersson and Rislund, 2012).The process of recognition for non-regulated professions described here does not include a judgement of the education provider's quality or a test of the applicant's skills.Thus, a recognition statement attests to the authenticitiy, level, and content of a foreign education, not to its quality or the individual's actual skills.
Between 2007 and 2011, 16,927 recognition statements were issued, including for native born with a foreign education.About 74 per cent of those who applied for recognition obtained a statement with an average processing time of 3-5 months (Petersson and Rislund, 2012).The most frequent reasons for not obtaining recognition was that the documentation was insufficient, or that the education was not academic or incomplete (ibid.).Those in the last category were able to use their foreign education to apply to a Swedish university instead and acquire Swedish credentials through studies.The most common fields of study for immigrants who have received a recognition statement were business administration, engineering, education, and social science (Holmvall and Rislund, 2014).
The Swedish recognition system is unique in several ways.To apply for and gain recognition of a foreign degree is voluntary and free of charge.The availability is not tied to any specific immigration category, and immigrants from countries outside EU/EES 2 can obtain a recognition statement before they receive a residence permit.A recognition statement is not needed when applying for higher education in Sweden.Recognition for immigrants is, to all intents and purposes, purely a labour market integration tool.

Country-Specific Human Capital and Return on Foreign Education
In the theoretical framework of country-specific human capital, immigrants fare worse on host countries' labour markets because of a deficit in the host country's knowledge and experience and because the human capital that immigrants bring with them from the country of origin is valued less in the host country's labour market (Chiswick, 1978;Chiswick and Miller, 2009).Previous research has consistently found an association between the origin of human capital and labour market outcomes for immigrants, as well as a so-called integration gap between immigrants and natives with comparable overall levels of human capital (for the Swedish case : Duvander, 2001;le Grand and Szulkin, 2002;Hammarstedt and Shukur, 2006;Rosholm, Scott and Husted, 2006).Studies on immigrants' return on education that compare origin-and host-country schooling have shown that foreign education is less valued, especially if it is of non-Western origin (Bratsberg and Ragan, 2002;Zeng and Xie, 2004;Buzdugan and Halli, 2009;Kanas and van Tubergen, 2009;De Vroome and Van Tubergen, 2010;Hardoy and Schøne, 2011;Kaushal, 2011;Nielsen, 2011).
Several explanations have been proposed as causes behind the lower value placed on foreign education.From the human capital perspective, this has been attributed to problems with transferability across national contexts, or educational quality differences between different countries (Friedberg, 2000).Transferability problems lead to a mismatch between origin-country schooling and host-country labour market demand because only part of knowledge acquired in education in one country is general enough to transcend different national contexts.Educational quality differences lead to skill heterogeneity between immigrants and natives as well as between different immigrant categories, with the same level of education.If employers reward skill and not formal credentials, a lower return to foreign education with lower quality is expected.Variation in return to education among immigrants from different countries of origin has sometimes been interpreted as an effect of variation in transferability or quality (e.g.Friedberg, 2000;Bratsberg and Terrell, 2002;Kanas and van Tubergen, 2009;Chiswick and Miller, 2010).But these interpretations are not unproblematic.The same regional variation, with human capital from non-Western countries yielding lowest returns, is interpreted as support for both explanations, and countries' educational content and quality can be correlated with other factors that affect labour market outcomes, e.g.immigrant selection and discrimination.Other research designs have been used to better isolate the two different explanations from each other, with positive results.Kanas and van Tubergen (2014) used institutional variation within Belgium to test the mismatch explanation for Moroccan immigrants, and Prokic-Breuer and McManus (2016) included results on the PIAAC survey to differentiate formal qualifications from skill, to test the skill heterogeneity explanation for immigrants with different origin.
Another proposed explanation of why foreign education has lower returns is that risk-averse employers avoid potential employees with educational credentials that they do not fully understand (Chiswick and Miller, 2009).Because employers do not know the true productivity of an applicant, they use screening based on productivity signals, e.g.education, in hiring and wage decisions (Spence, 1973;Stiglitz, 1975).If employers do not understand what particular skills or level of productivity a foreign education corresponds to, they may, being risk-averse, be reluctant to employ applicants with foreign degrees, even if they hold no specific beliefs about other countries' educational systems.In other words, foreign credentials do not function properly as signals because their signalling value is unknown to employers.Lancee and Bol (2017) found using PIAAC data a substantial wage penalty for immigrants with a non-Western foreign degree after controlling for different forms of skill.They attribute the penalty to the weaker signals that specifically non-Western degrees have, as employers are less likely to be familiar with them.Using a factorial survey, Damelang and Abraham (2016) showed that foreign vocational certificates have some value for German employers, but are less valued than German certificates.These empirical results are in support of the explanation that one reason for why immigrants fare worse in the labour market is employer uncertanty about the signals of foreign degrees.
Note that the uncertainty explanation should be held separate from theories of discrimination against individuals (Phelps, 1972;Arrow, 1973).The mechanism at work in the former case is risk aversion, while in the latter it is stereotypical imputation.The distinction is important in the case of formal recognition of qualifications.A recognition statement does not change the origin of the individual.Thus, if employers hold stereotypical, downgrading beliefs about individuals from a country of origin, recognition should have no effect.
Finally, previous research has also shown that immigrants' returns to human capital, and labour market integration in general, depends on their ability to use their foreign skills in their host country.Individuals with better potential to gain host country specific human capital and access information about relevant job openings can better utilize their origin-country human capital (Chiswick and Miller, 2003;Kanas et al., 2012;Lancee, 2016), suggesting a positive interaction effect between the two forms of human capital.And immigrants are more susceptible to the overall economic performance of the host country's labour market compared to natives (Arai and Vilhelmsson, 2004;Kogan, 2004;Gustafsson and Zheng, 2006).The ability the get a hold on the labour market, to be able to use one's human capital, is therefore not only conditional on individual traits but also on the economic circumstances.

Hypothesized Effects
The main hypothesis of this study is that the mechanism of employer uncertainty is real, and that recognition will reduce it.We do not claim a recognition statement will reduce all employer uncertainty regarding individual skill, that is improbable and against signalling theory which we build on, but it will aid the foreign credential to better function as a signal, leading to better labour market outcomes.If the opposite case is true-if the lower return on immigrants' origin-country education can only be explained with mismatch or skill heterogeneity between immigrants and natives-then recognition will have no effect because gaining recognition of their qualifications affects neither the content of a degree nor the skill level of an individual.Additionally, as signalling is connected to information deficits, the mechanism will primarily work through new employments, and not within existing employee-employer relationships.Here, we are somewhat helped by the fact that newly arrived have a weak connection to the labour market, meeting (potential) new employers more often than the general public: H1: Recognition will have positive effects on labour market outcomes for newly arrived immigrants.
When employer uncertainty decreases, other mechanisms increase in relative importance.Human capital that is recognised becomes more salient and is therefore also more subjected to the processes of differing returns depending on employers' appreciation of its worth and of the individual's ability to use it.Stated simply, there is an interaction effect between recognition and other mechanisms described above that decide the returns to foreign human capital.
First, when employers understand the content of an immigrant's foreign education better, the mechanisms of transferability problems and (perceived) educational quality will have a relatively larger effect on labour market outcomes.When provided with a recognition statementwhich attests to the authenticity, level and content of the applicant's tertiary degree, but not to its quality or the skill of the individual-employers can have heterogeneous responses depending on their appreciation of that country's educational system regarding subject matter or quality. 3As noted in the theoretical section, it is difficult to separate between the human capital mechanisms of mismatch and quality differences empirically, but they follow the same pattern regarding origin: H2a: Immigrants from Western countries will experience larger effects of recognition compared to the rest of the world.
Second, there is a connection between the ability for immigrants to use their foreign human capital and the returns to it.Compared to other immigrant categories, family migrants have by definition some form of social network upon arrival.This gives them better opportunities to learn the Swedish language and culture, and get information about the labour market, making it easier to better use a recognition statement: H2b: Family migrants will experience larger effects of recognition compared to other reasons for immigration.
Third, if immigrants are highly susceptible to economic conditions, it follows that the labour market must be in a good enough state for recognition to have an effect on outcomes.This is because there must be enough job opportunities for newly arrived immigrants for the recognition statement to be useful to begin with: H2c: The effects of recognition will be larger in good economic years compared to bad economic years.

Research on Recognition
We have found three previous studies that in some way measure the effects of educational recognition.Kler (2006) and Green, Kler and Leeves (2007) studied inter alia qualification assessment of qualifications in Australia, which is mandatory for some visa categories and optional for others.The effect of assessment, net of the effects of visa category, was negative on employment, zero or negative on overeducation, and zero on earnings.Kogan (2012) analyzed how recognition affected employment probabilities of immigrants who were either ethnic Germans or Jewish quota refugees from the former Soviet Union in Germany.The results were that immigrants who succeeded in gaining recognition found a qualified job faster during the first three years after immigration, and had more qualified current jobs, compared to those who applied for recognition but were unsuccessful.Unfortunately, the design of Kogan (2012) did not identify when recognition was issued, which makes it hard to establish causality.
Our study differs from the previous studies in several ways.First, recognition is not tied to specific visa categories in Sweden and we can identify the timing of recognition.This allows us to make stronger causal claims of its effects.Second, our data is large enough to make comparisons between different subcategories of immigrants without a major loss of statistical power (the three studies all had samples below 2000).Third, we only analyse so-called de-facto recognition, i.e. for nonregulated professions.Our estimated treatment effects are therefore not artificially amplified by a shift in legal eligibility to access regulated professions, i.e. immigrants overcoming occupational closure.

Data
Information about individuals with recognition 4 was collected by the Swedish Council for Higher Education and matched to administrative population-level registries containing information about migration (STATIV) and labour market outcomes (LISA).Our data consists of immigrants who gained recognition between 2007 and 2011, and for whom we measure labour market outcomes from 2005, or, if later, the year of immigration, to 2012.Because recognition is a voluntary procedure, individuals were self-selected into treatment.A total of 14,878 immigrants gained recognition between 2007 and 2011.We restricted the sample to include only those who obtained recognition during the first 10 years in Sweden in order to focus on the newly arrived and to limit the potential number of individuals who studied abroad after immigration.In addition, we restricted the sample to individuals who were of working age, 20-60 years old, and lived in Sweden in all the measured years.This produced an unbalanced panel of 12,792 individuals who were followed between 2 and 8 years (5.5 years on average).

Analytical Strategy
The analyses are conducted using fixed effects models.The strength of fixed effects is that it allows us to control for all stable unobserved individual characteristics that can vary within the sample and affect labour market outcomes, such as skill or motivation.Our data only includes those with recognition, i.e. no control group.While this can be perceived as a limitation, we do not believe that a control group would help in estimating a causal effect of recognition.Finding a matched group can only be done on observable characteristics that exist in registry data, with the notion that registry variables will validly approximate unobservable traits.But because our sample contains newly arrived immigrants, relevant registry variables that can accurately approximate skill or motivation are all but non-existent, e.g.employment history.Furthermore, since recognition is voluntary, self-selection into treatment, the defining difference between the treated and a control group, is itself correlated with important unobservable traits affecting labour market outcomes.The sample consists of individuals who brought their credentials with them, found out about the possibility of recognition in Sweden, were motivated enough to apply for recognition, and so on.Any matched control group on observables would differ too much on unobservables.It is probable that labour market outcomes would therefore differ between the groups even in the absence of recognition, and that the effect of recognition would differ across the groups, bringing bias into the estimates (Winship and Morgan, 1999).That is, we expect the sample to both have a better labour market trajectory in the absence of recognition and be able to make better use, i.e. obtain higher effects, of recognition than a control group matched on variables existing in our data.
Individuals in our sample obtained recognition at different points in time.We use this variation in timing to estimate the treatment effect, a technique applied by Arai and Thoursie (2009) to study effect of name changes on labour market outcomes for immigrants in Sweden.The most important overall time effect for newly arrived immigrants is not calendar year but year since migration.During the first years, immigrants acquire the new language and other country-specific human capital, get access to domestic networks, and learn about the labour market in Sweden-all of which translates into an upward trajectory in labour market outcomes.Not controlling for this trajectory would overestimate the effects of recognition.The fixed effects model used here therefore contains years since migration (YSM) dummies as measurement of time.
Estimating the dependent variable y for individual i at YSM t, our model is as follows.
where a i is individual fixed effects, YSM t years since migration time period dummies, d the treatment effect of the Recognition it treatment, b estimate for X it time varying control, and it the residual.Under this specification, the control group changes for each t and consists of those who did not obtain a recognition statement at each YSM, i.e. of those who obtain it at an earlier or later t.The treatment effect d measure the average change in y for all years after recognition compared to all years before recognition, holding a, YSM and X constant.

Variables
We use two separate dependent variables; whether the individual was employed during a year, and earnings for that corresponding year.Employment is coded dichotomously 0/1, and defined as having earnings above a cut-off of 0 SEK.Earnings are measured as the natural logarithm above that cut-off.To conduct a sensitivity test of the results, we also ran our main analyses with higher cut-offs for employment and earnings in incremental steps of 50' SEK (Angrist and Pischke, 2009: p. 101;Ha ¨llsten, 2012).Earnings are measured by wage income during a year (in 2010 prices), excluding income from self-employment.This better captures the theoretical mechanisms behind the effect of recognition, i.e. that employers in some way react to a recognition statement among their job applicants or employees. 5 Recognition is a dichotomous 0/1 variable, with the value 0 representing the time up to the year in which an individual obtained the recognition statement, and 1 from the first year after and onwards.It takes some time to start using a recognition statement in job search, and additional time for employers to go through applications, interview candidates and come to a hiring decision.Setting the variable to 1 already in the year that the recognition statement was issued creates a risk of capturing a lag effect, underestimating the treatment.
In an ideal situation, all behavioural changes that accompany individuals' applications for recognition, e.g.job search intensity or differences in use of the recognition statement across sectors or occupations, would have been included in our models to estimate the true treatment effect of recognition.Due to registry data limitations, we include 21 county of residence dummies as control variables to approximate such behaviour.Labour market opportunities are unevenly distributed across Sweden, and individuals can move to a better regional labour market to improve their situation and make better use of the recognition statement, which could bias the treatment effect if uncontrolled for (Arai and Thoursie, 2009).
Estimations of effect heterogeneity, hypotheses H2ac, are done by augmenting the original model with interaction terms between the treatment of recognition and each subcategory, and estimating average marginal effects (AME) for the interactions.The subcategories are region of origin, reason for settlement and year of recognition.As a robustness check, we also divided the sample according to the subcategories and run separate models.The difference between the procedures is that the interaction model assumes equal labour market returns to YSM and county for all subcategories, but uses the full statistical power of the sample.The interaction procedure also controls for the correlation between the subcategories, e.g. that refugees tend to come from a specific region etc.
Gender is included as subcategory to be a control variable, because it has a high correlation with the other subcategories and need to be held constant in the analyses.Gender will be reported in the regression outputs but not commented on.

Descriptive Statistics
The average time from immigration to recognition was 2.1 years (sd ¼ 1.8).Immigrants who have obtained recognition lived in all of Sweden's 21 counties, with the county including Stockholm being the most common and somewhat overrepresented at the time of recognition (36 per cent in sample, compared to 22 per cent for the entire Swedish population).Table 1 displays the distribution of subcategories within the sample.The table also provides an illustrative comparison group in the form of all immigrants who came to Sweden between the years 2006 and 2010, the period in which the vast majority (82 per cent) of the sample immigrated.The comparison group serves to highlight how individuals with recognition differ on observables from the overall immigrant population in comparable cohorts.
The subcategories divided by region of origin are EU15þ (including EES countries, North America and Oceania), Other Europe (including Russia), Africa & Asia (predominantly the Middle East), and Latin America.Compared to all those who immigrated between 2006 and 2010, immigrants from Other Europe and Africa & Asia are overrepresented, while EU15þ is underrepresented, among those with recognition.Reason for immigration is divided into four subcategories.Work/study, Family (including family to refugees), Refugees and Other/missing (comprised mostly of immigrants from Nordic and EU countries, who thus do not need an official reason for immigration to Sweden).The Work/studies and Other/missing subcategories are underrepresented among those with recognition.The number of immigrants who obtained recognition increased during the sample period.This is not due to an increasing number of newly arrived immigrants, but rather to the fact that the policy instrument of recognition got a higher priority and visibility during the same time.Two thirds of those with recognition are females, even though they were a minority among comparable immigrants.
A final important difference between the sample and the comparison, not visible in Table 1, is level of education.Everyone in the sample is highly educated, compared to 30 per cent of the adult individuals who immigrated during this time (Statistics Sweden, 2014a).Registry information about the education level of newly arrived is self-reporteted through surveys (ibid.), and thus not validated by credentials.If the 30 per cent estimate is taken at face value, acquiring recognition is not common.There are several reasons for this.First, not all higher education is academic, completed with a degree or for a non-regulated profession, i.e. within the scope of recognition from the Swedish Council for Higher Education.Second, not all immigrants bring their credentials with them.Third, not everyone knows about the availability of recognition, feels that it is a good labour market strategy or is motivated enough to apply.Comparing the education levels before and after recognition, we find no evidence of any general downgrading during the recognition procedure (see Table A1 in the Appendix).
There are important correlations between region of origin, reason for immigration and gender in the sample (see Table A2 in the Appendix).In short, refugees almost exclusively come from Africa & Asia, women are overrepresented among migrants from Other Europe, women tend to be family migrants, and men are overrepresented among refugees.
The average employment rate for the whole sample and for all years is 49 per cent, and the mean yearly wage for those with employment is 150' SEK. 6,7 Figure 1 shows the distribution of employment and wage for the first 10 years after immigration.Both the employment rate and the yearly wage show the expected curvilinear form with a faster growth rate during the first years in Sweden and a slower rate in later years, demonstrating the importance of controlling for YSM in the fixed effects regressions as discussed above.Compared to immigrants in general with similar YSM and education level (Statistics Sweden, 2014b), the sample has a somewhat better employment rate trajectory.

Effects of Recognition
Table 2 displays the results of fixed effects regressions with treatment effects of recognition on employment and wage.Recognition increases the probability of being employed on average by 4.4 percentage points, and increases wage among the employed on average by 13.9 log points. 8Both sets of YSM and county dummies have significant effects on employment and wage when included.Comparing models with and without county dummies (the first and third column to the second and fourth, respectively) it is clear that while moving to different counties has an effect on employment and wage, it did not alter the effects of recognition, i.e. moving is on average not related to the timing of recognition.
We can assume that the intent of recognition, from the perspectives of highly educated immigrants and policy makers alike, is not only to ease transition into the labour market in general, but also to facilitate attainment of more qualified occupations.The data does not have a good measure of occupations, so it is not possible to get a direct estimate of the effect of recognition on job qualification.Instead, we use different cut-offs for the definition of being employed, and to measure wage only above those cut-offs, both as a test of sensitivity of the results, and in order to study whether there is an effect of recognition on the probability of obtaining more gainful employment.We interpret higher cut-offs as an approximation of higher qualification requirements for the job and of a more stable employment, i.e. of betterpaying work and longer working hours.The results of using different cut-offs are displayed in Figure 2. The cut-off 0 on the X-axis displays the same results as in the second and fourth column in Table 2, the treatment effect of recognition on employment is defined as any income above 0 SEK and wages for those with employment per the same definition.50' SEK shows the corresponding effects if employment is defined as having an income above 50' SEK and wages only for those now defined as employed, and so on for different cut-offs.
The effects of recognition on employment probability are steady, around 5 percentage points up until a cut-off of 100' SEK, where it starts to decline and became nonsignificant from 300' SEK onwards.The effect on wage declines immediately with higher cut-offs and becomes non-significant from 250' SEK onwards. 9For the 100' SEK cut-off, close to the value suggested by Eriksson et al. (2007) as a measure of a substantial entry into employment in Sweden, the treatment effect of recognition is 5.3 percentage points on employment and 3.8 log points on wage.For the 250' cut-off, close to the average wage in Sweden, recognition has an effect of 1.4 percentage points on employment but no significant effect on wage.In sum, the results show that recognition did help newly arrived highly educated immigrants to transition into employment and to earn higher wages, but it did so more in the lower range of the income distribution.

Effect Heterogeneity
The final step of the analysis is to estimate the treatment effect heterogeneity.Table 3 displays AME from fixed  Note: Employment defined as earnings above cut-off 0 SEK in a year.Log wage only for years with earnings above cut-off.Robust standard errors clustered on individuals in parentheses.F-test of joint significance for full set of dummies for YSM and County, respectively.*P < 0.05; **P < 0.01; ***P < 0.001.effects models in which the subcategories are interacted with the recognition treatment.The table includes both regressions using the 0 SEK and the 100' SEK cut-off for defining employment, in order to study whether there are treatment differences when defining employment as any income or as substantial entry into the Swedish labour market (see Table A3 in the Appendix for the procedure with separate regression models for each subcategory).
Categorized by region of origin, the general pattern is that Other Europe, and to some extent EU15þ (for the 100' SEK cut-off), show the overall largest effects of recognition, while Africa & Asia and Latin America show the smallest.Looking at employment, the effect for Other Europe is 8.7 and 10.3 percentage points for the two cut-offs, respectively.For Africa & Asia, the corresponding figures are 3.1 and 2.3 percentage points.Both EU15þ and Latin America display an interesting pattern with higher and more significant effects at the higher cut-off.Here, recognition seems to work closer to policy intent by primarily facilitating transition into better work.The separate regressions by each subcategory in Table A3 show the same general pattern, with EU15þ and Other Europe having higher effects than Africa & Asia and Latin America.As expected, when the recognition statement reduces employer uncertainty, other mechanisms become relatively more important in determining labour market success.Hence, immigrants from regions associated with human capital transferability problems and (perceived) quality issues experience an adverse effect to the usefulness of the recognition statement.But somewhat contrary to expectations is that the higher effects are not only evident for Western countries (EU15þ), but also Other Europe, H2a is thus only partly confirmed.
The category Work/study has significant negative effects of recognition at the 0 SEK cut-off and no effects at the 100' SEK cut-off.Labour migrants are a special case.Having an employment upon arrival, their motivation to obtain recognition is likely to coincide with a search for a new job at first employment's end (as mentioned in the section on the Swedish context, no recognition is done at the time of immigration to Sweden).The effect captures a short-term association of going from employment to unemployment, but it does not indicate going from better to worse employment, which would have been the case if the estimates at the 100' SEK cutoffs had been significantly negative as well.In Table 3, family migrants as well as refugees experience a positive effect of recognition, on both employment and wage and for both cut-offs.In contrast to this, the separate models for each subcategory of Table A3 show that only family migrants benefit from recognition.The difference stems from two reasons.First, family migrants have a higher return on YSM, i.e. a better labour market trajectory (not shown); second, refugees came almost exclusively from Africa & Asia (Table A2 in the Appendix).Thus, the fact that family migrants acquire more Sweden-specific human capital through work, and that refugees come from a region that is associated with low returns to foreign human capital implies that family migrants can use their recognition statement better, which gives rise to heterogenous effects.Since these differences are held constant in Table 3, but not in Table A3, the tables show different results. 10The category Other/missing, comprised mostly of de facto work, study, and family migrants from the Nordic and EU countries, had effects of recognition at the higher cut- off, close to the pattern of EU15þ.Overall, with family migrants experiences larger effects than refugees, and larger or equal effects as the Other/missing category, H2b is confirmed.Finally, the subcategory of Year of recognition shows how effect heterogeneity is correlated with the state of the economy and the labour market.The lowest effects on both employment and wage, regardless of cut-off, were for those who obtained recognition in 2009, right after the financial crisis.For the 0 SEK cut-off, the effect of recognition on employment is 3.2 percentage points for those who obtained recognition in 2009, while it is almost the double for those who got recognition 2007 or 2011 (Table A3 shows a similar pattern).H2c is confirmed.

Concluding Discussion
We have studied the effects of recognition of higher education for newly arrived immigrants in Sweden, a policy tool with the goal to improve labour market integration.We aim was to contribute to the literature by both evaluating the policy of recognition, and by testing whether employer uncertainty is a reason for why foreign human capital is valued less in host countries' labour markets.
Regarding the policy, using fixed effects regression models that utilize the variation in timing of recognition, we estimated the treatment effects to be 4.4 percentage points higher probability of employment and 13.9 higher wage among the employed.We also estimated the effects at different cut-offs for defining employment and for different subcategories.Our first conclusion is that recognition does improve labour market integration (H1), but that the policy intervention has its limitations.First, the effect is concentrated at the lower end of the income distribution.While this might not be surprising, as it is easier for newly arrived immigrants to find lowpaying jobs, it must be considered as policy intent that recognition would primarily facilitate a transition to higher qualified occupations for its highly educated recipients.Unfortunately, as there is no comparable research from other countries yet, we cannot relate effect size to make a sound judgment in whether the effect is large or small, or whether the distribution is expected or abnormal.Second, the analyses of effect heterogeneity revealed that the effect of recognition varies with different subcategories.Immigrants from EU15þ and Other Europe (H2a), family migrants (H2b), and those who acquired recognition during good economic years (H2c) experience larger effects.These results point to that origin-country human capital is positively associated with destination-country human and social capital acquisition, implying a complementary relationship between capital of different origin.
Based on the regression results we can also say something about why immigrants get lower returns to their foreign education.Our focus on only de facto recognition in a national context where it is both available to all immigrants and voluntary gave us a good possibility to isolate the mechanism of employer uncertainty.The effects of recognition here are not driven by a change in formal to work in specific professions (i.e.closure) or by a state-regulated selection process in which only some can acquire recognition, thus increasing its value by artificial scarcity.Because recognition affects employment and wage, our second conclusion is that employer uncertainty is an existing mechanism that explains a part of why foreign human capital is valued less in host countries' labour markets.But, as shown by the analyses of effect heterogeneity, when employer uncertainty decreases, the importance of other mechanisms that hinder labour market integration increase.Our results indicate that the answer to why immigrants get lower returns to foreign education lies not in either human capital or signalling explanations, but in both and that the mechanisms are intertwined.Both immigrants with education from countries that have either human capital transferability problems or (perceived) lower educational quality and immigrants that have lesser possibilities to use their foreign human capital benefit less from a better signal of their foreign education.
As discussed in the section on analytical strategy, we believe that the sample is positively selected on motivation as an unobserved individual characteristic.If employers generally interpret a recognition statement as a signal of such higher motivation, the results could be driven by this signal rather than by reduced educational uncertainty.In this study, we cannot clearly differentiate between these two effects and future studies must further try to find new ways to isolate the mechanism of uncertainty.
We have focused on tertiary education for nonregulated professions in a single country, and we believe that the effect of recognition will vary with education level, recognition procedure and in different contexts.Recognition of general education that is lower than tertiary will likely have a smaller effect because the human capital is itself of lesser value and of more limited usefulness.In contrast, recognition of vocational education should have higher effects, because the tighter connection between education and occupation implies a stronger signalling effect once employer uncertainty is reduced.Recognition for regulated professions-where the recognition procedure includes documentation of experience, tests, additional courses, or training-will have larger effects because it will both grant access to an otherwise inaccessible profession, and provide employers with information about the skill level.Countries that have a higher level of occupational regulation than Sweden and a tighter link between education and occupation will therefore likely have higher average effects of recognition for the labour market as a whole (Bol and Van De Werfhorst, 2011;Kogan, 2016;Bru ¨cker et al., 2018).
We have studied the short-term effects of recognition.Future research should-besides studying different national contexts, education levels, and professionsstudy the long-term effects as well.Possible avenues are distributed fixed effects11 to analyze whether the effect of recognition is most pronounced in the short or long term, as well as the inclusion of a control group to see if those with recognition have cumulative advantages or if those without recognition catch up on labour market outcomes in the long run.

Figure 1 .
Figure 1.Employment rate and wage for sample during first 10 years in Sweden Note: Employment defined as earnings above cut-off 0 SEK for a given year.Mean wage (in 2010 prices) only for those with employment.Different N for each YSM due to sample restrictions.

Figure 2 .
Figure 2. Effects of recognition with different cut-offs Note: Separate FE models for each dependent variable and cut-off point.All models held constant for YSM time dummies and county with robust standard errors clustered on individuals.Different N for Log wage at all data points.Effect for employment significant (P < 0.05) up until 250' SEK.Effect for Log wage significant (P < 0.05) up until 200' SEK.

Table 2 .
Effects of recognition on employment and log wage

Table 3 .
Effects (AME) of recognition on employment and log wage for different subcategories Average marginal effects.Employment defined as earnings above cut-off in a year.Log wage only for years with earnings above cut-off.Robust standard errors clustered on individuals in parentheses.F-test of joint significance for full set of dummies for YSM and County, respectively. Note: