Essay Two: A Longitudinal Analysis on the Incidence of Over- education among Immigrants and its Impacts on Earnings
This essay uses longitudinal analyses based on the Household, Income and Labour Dynamics in Australia (HILDA) Survey to investigate the extent of matching between education and occupation and resulting effects on earnings among immigrants in Australia. The panel approach based on nine years of longitudinal data addresses individual heterogeneity effects that are important to over-education analysis, and thereby extends the international literature. Correlated random effects (CRE) logit with Mundlak (1978) correction results suggest that both ESB (English speaking background) and NESB (Non-English speaking background) immigrants have high incidence rates of over-education. Age at migration and year of arrival have significant effects on the incidence of over-education among NESB immigrants, on the other hand, this appears to have no effects among ESB immigrants. Longitudinal analyses show an assimilation effect among both NESB and ESB immigrants, with ESB immigrants experiencing faster assimilation rates than NESB immigrants. Over-education has been shown to slow down assimilation for NESB immigrants. Pre-migration education obtained abroad is valued in Australia for ESB immigrants; although this is not the case for NESB immigrants. Current Australian immigration policy favours skilled migrants. However, if their skills are not fully used in their jobs, the under-utilisation of skills impedes their assimilation process.
Keywords: over-education, years since migration, age on arrival, year of arrival, country of qualification, earnings
It is a commonplace occurrence to hear of immigrants being employed in occupations that are below the level of their educational attainment; such as those from professional occupations driving taxis or working as kitchen hands. What is the extent of this phenomenon across host countries? What are the determinants of this disadvantageous situation among immigrants? How can immigrants’ skills be used to full advantage? A topic of significant debate among researchers and policy makers has been immigrants’ adjustment, assimilation, and success in their new labour market.
This study uses longitudinal data to examine labour market outcomes for immigrants in the Australian labour market.
During a two-year period (2005-2006), about 48,865 skilled migrants, 45,290 family migrants and 14,140 humanitarian migrants arrived in Australia. The number of skilled migrant visas issued in 1998-99 was 35,000, which increased to 97,340 in 2005-06. Of these, 17% permanent arrivals came from the United Kingdom and 11% came from New Zealand.
“Skilled visa holders were the most likely to be employed after arriving in Australia. Humanitarian visa holders were the least likely to be employed. However, the longer an immigrant remained in Australia, no matter what their visa class, the more likely they were to be in employment.” (DIMA 2007)
The evidence shows that Australian immigration policy has placed greater emphasis on skill based immigration because skilled immigrants are more employable and more productive than their unskilled counterparts. Thus, they are therefore likely to increase Australia’s productive capacity. However, if immigrants cannot work in occupations that fully utilise their skills, this productivity gain is reduced. The cause of “the unrecognised skills of immigrants” is the mismatching of educational attainment and the educational requirements for migrants prospective occupations in the host country, generally referred to as over-education. When compared to native-born, immigrants are more likely to be over-educated and to suffer an earnings loss and therefore explicit individual earnings disadvantage (See for example, Chiswick and Miller, 2008). Moreover, a potential loss to the economy as well as a significant burden on new arrivals may be caused (Ferrer & Riddell, 2008).
Over-education8 is defined as the extent of someone’s actual education exceeding the educational requirement to perform his or her job. Because the HILDA data does not provide any questions on over-education, workers’ self-reports (SR) are not applicable. Thus, the required years of education to do a job for a particular occupation can be defined by using a cross-wave Mode measure; this measures the number years of education required to undertake a position of employment; the number varies between waves. The amount of education that most commonly occurs within an occupational category is calculated for each wave. The required years of education for all nine waves, are derived by combining the Mode education of all waves; next, the years of over-education and years of under-education are obtained by comparing the actual years of education with the required years of education.
By employing the procedure described, it was found that the incidence of over-education differed considerably between the native born population and the immigrants. In particular, immigrants were shown to have a higher probability of being over-educated than natives9 . Non-English-Speaking background (NESB) immigrants were found to suffer especially from extremely high levels of educational mismatch. For example, the incidence of over-education ranged from 24 per cent to 28 per cent among natives. However, among English-Speaking background (ESB) immigrants it was 3 to 10 per cent higher, ranging from 28 per cent to 36 per cent; the incidence of over-education was 36 per cent to 48 per cent among NESB immigrants. This is 17 per cent to 21 per cent higher than for the native born population, depending upon the specific year of assessment. It was also found that ESB immigrants earn a premium wage and NESB immigrants suffer loss of earnings when compared to natives.
A number of questions arise from these findings. Why is it that immigrants have a higher incidence of over-education than natives? What are the determinants of educational mismatch? What is the relationship between earnings and over-education? Does over-education have a negative effect on earnings? Why do NESB immigrants earn less than natives? Can NESB immigrants reduce their earnings disadvantage with years since migration? These questions have motivated the research reported in this paper.
To date, immigrants’ over-education is under-researched in Australia. This study makes the following contributions to the international literature: It investigates the determinants of over-education among immigrants in Australia, and the extent of the impact of over-education on earnings after accounting for individual heterogeneity. I use the correlated random effects (CRE) logit model, and fixed effects earnings models to address endogeneity and individual heterogeneity. To the best of our knowledge, it is the first examination of the determinants of over-education and its impact on earnings among immigrants using longitudinal techniques based on panel data. In addition, specific sub-group effects, such as, age at migration, year of arrival and country of qualification effects are examined among ESB and NESB immigrants respectively. This study also examines new evidence based on panel data on the transferability of experience and education abroad for immigrants.
The over-education of immigrants is examined from the following perspectives:
Country of origin and language proficiency
In the study of immigrants ‘assimilation, country of origin is of importance. Immigrants from different countries have differing assimilation rates in the host country. Immigrants from a background that is similar to that of the host country are more likely to have similar incidence rates of over-education due to the higher transferability of human capital. However, those from a non-English speaking background may find it more difficult to settle down, which could produce serious over-education rates. The over-education rate of immigrants may not converge with the rate of natives, even after a lengthy period of residence.
As English is the main language in Australia, the English proficiency of immigrants may help them to obtain education-occupation matched jobs. Compared to NESB immigrants, in the host country, ESB immigrants would expect to face similar labour market conditions to those of their country of origin. Their prior migration experience and education may be portable to the host countries. As a result, relative to NESB immigrants, ESB immigrants may adapt to new environments quickly, and be more likely to find a matched job.
In this study, an immigrant is defined as a person who was born overseas. Base on the English proficiency and country of origin, I differentiate immigrants between ESB immigrants and NESB immigrants. People born overseas are asked whether English is the first language they learned to speak as child10. If English was the first language learned, the immigrant is defined as an ESB immigrant, otherwise, as a NESB immigrant. Thus, the sample is divided into three subsamples: Natives, ESB immigrants and NESB immigrants.
Chiswick and Miller (2009b) provided evidence of strong positive relationships between English speaking proficiency and occupational attainment.
Transferability of human capital
Human capital acquired both abroad and domestically may have a variety of effects on the rates of over-education. Since transferability of human capital is limited, education and experience obtained abroad are discounted in the host country (Friedberg, 2000).
Immigrants generally demonstrate high rates of over-education due to the imperfect transferability of human capital in the host country. Thus, the over-education rates of immigrants signify education-occupation matching difficulties in the host countries’ labour market, and they reflect an important dimension of immigrants assimilation (Friedberg, 2000).
Therefore, to analyse the impact of over-education on transferability of human capital, I distinguish human capital between human capital obtained abroad and that obtained domestically. I examine the impacts of both experience and education acquired abroad on the incidence of over-education.
Furthermore, based on the country where qualifications are obtained, two types of qualification are defined among immigrants: Qualifications obtained domestically (in Australia) and abroad (overseas qualifications). The differing incidence of the rates of over-education among these two groups may reflect the transferability of human capital.
With time, gaining local experience or investing in local education may help immigrants to improve educational and job matches, reduce the rates of over-education, and decrease the earnings penalty.
Age at migration
Migrating as a child or as an adult may give rise to differing effects on the incidence of over-education. Young immigrants are more likely than adults to adapt to their new country of residence and to achieve qualifications in the host country. Thus, they behave similarly to a member of the local population even though they may still face a certain amount of discrimination.
Year of arrival
Immigrants ‘quality’ affect them to allocate at matched position, in particular, immigration policy has favoured skilled immigrants in recent years. Thus, this implies that recent cohort ‘quality’ is increasing compared to the earlier cohort. However, the earlier entrants have acquired the domestic experience and are expected to less likely to be over-educated than the recent entrants.
The following questions are addressed in this study:
To what extent are immigrants and natives over-educated? Does the incidence of over-education among immigrants vary by country of origin, English proficiency, age on arrival and year of arrival?
Are there differing impacts of over-education on earnings between sub-groups based on country of origin and English proficiency?
Are there differing impacts of over-education on earnings between sub-groups based on age on arrival, year of arrival and country of qualification?
To estimate the effects of over-education on immigrants’ assimilation effects, I examine the following hypotheses.
1. NESB Immigrants are more likely to be over-educated in relation to ESB immigrants at the time of arrival.
2. As time passes, by gaining local experience or investing in local education, the over-education rates of immigrants converge to the rates of the native born population, and the immigrant earnings differential relative to that of the native born decreases. Therefore, the coefficients of YSM (years since migrating to Australia) are predicted to be negative when examining the incidence of with over-education, and positive with earnings.
3. Immigrants’ experience and education are divided into pre-migration experience and pre-migration education, and post-migration experience and post-migration education. ESB immigrants are predicted to have pre-migration human capital transferable to the host country and can thereby enhance the match of their education and occupation.
4. Younger labour market entrants are less likely to be over-educated compared to the older entrants because they are likely to gain more education and experience in host country than older entrants.
5. Over-education is more likely among the recent labour market entrants compared to the earlier entrants.
The remainder of this essay is organised as follows. Section 3.2 provides an overview of recent immigrants’ over-education literature, and it identifies the main factors affecting immigrants mismatch and labour market outcome in the host country. Section 3.3 develops the econometric framework. Section 3.4 outlines the data and variables. The results are presented in Section 3.5 to Section 3.6, followed by a summary in Section 3.7.
Review of the literature
A number of studies have examined over-education among immigrants in different countries, and reviews of the literature are presented in Table 3.1. Regardless of host country and official language, these studies have shown that immigrants have a high incidence rate of over-education, ranging, from 16 per cent (Kler, 2007) in Australia to 96 per cent (Aringa and Pagani, 2010) in Italy. And immigrants suffer an earnings loss from education-occupation mismatches (Chiswick and Miller, 2006; Kler, 2007; Green, Kler and Leeves, 2007; Lindley, 2009; Wald and Fang, 2008).
To date few studies have been conducted on immigrant assimilation in the Australian labour market. Based on the 2001 Census of Population and Housing, Chiswick and Miller (2006) reported that NESB immigrants have a lower rate of return to schooling accompanied by over-education and under-education. The payoff to years of schooling for Australian-born males is 8.8 percent. For ESB immigrants and NESB immigrants, it is 8 percent and 5.9 percent, respectively. However, there is the same payoff to required years of schooling of 15.2 percent for these three groups. The earning effects of over-education (under-education) is 5 to 6 (-3 to -4) percent for the Australian-born and ESB immigrants, and it is about 3 (-1) percent for NESB immigrants.
Based on longitudinal data for immigrants to Australia (LSIA), Green, Kler and Leeves (2007) examined the determinants of employment and over-education. They also studied the return to required schooling and surplus schooling by two cohorts among male immigrants aged 15-64. They found that immigrants, even those with skill-assessed visas are more vulnerable to over-education than natives. NESB immigrants are more likely to be over-educated, with the incidence of over-education between 32% and 49%. NESB immigrants also have lower returns to required and surplus education than do natives. Tighter welfare and support policies11 for immigrants may increase the employment at the expense of under-utilising their skills. However, their sample is limited to recent immigrants in their sample (arriving in 1993, 1995, 1999, and 2000). The analysis employed OLS estimation.
Using the same LSIA dataset with the addition of the inclusion of both genders, Kler (2007) examined the effects of over-education among tertiary educated immigrants. The evidence is in line with Green, Kler and Leeves (2007). The incidence of over-education is similar between ESB immigrants and natives, and is higher among Asian NESB immigrants. The rate of over-education is around 16% for ESB immigrants. Among Asian immigrants, approximately 50% are over-educated. Among other NESB immigrants, the rate of over-education is close to 40%. The payoff to over-education is much smaller than the payoff to required education. There is no significant effect of over-education on earnings among Asian immigrants.
Green, Kler and Leeves (2007) and Kler (2007) used a bivariate probit model to examine the incidence of over-education, and an augmented human capital earnings model (Frenette,2004) to examine earning effects in the Australian labour market. They focused on the effects of visa category and labour market conditions.
This study extends Green, Kler and Leeves’ (2007) work and it contributes to the Australian literature as follows. I extend the analysis to panel data, and I employ a correlated random effects (CRE) logit model with Mundlak (1978) correction to examine the incidence of over-education by focusing on the effects from years since migration, age at migration and year of arrival. The endogeneity due to the correlation between explanatory variables and error terms is addressed by Mundlak correction. I also employ both panel fixed effects (FE) and random effects (RE) models to examine the effects of over-education on earnings from years since migration and transferability of human capital by country of origin, age on arrival, year of arrival and qualification type respectively. The latter aspect of my study on the effects of transferability of human capital on over-education and earnings and the panel feature of the analysis extend the international literature.
In Spain, the effects of years since migration effects have been examined by Fernández and Ortega (2008). They used data from the Spanish Labour Force Survey for the period 1996-2006, and showed that compared to the rates for the native-born, immigrants experience initially higher participation and unemployment rates, and have a higher incidence of over-education and temporary contracts. Over a five-year period, immigrants’ participation rate was shown to be reduced to that of those who are native-born and unemployment rates to levels even lower than those of the native-born. The incidence of over-education and temporary contracts however remained constant.
Moreover, the portability of immigrants’ human capital into the Spanish job market has been studied by Sanroma, Ramos, and Simon (2008). They suggested that geographic origin has an influence on transferability of human capital. Immigrants from countries that are highly developed, or have a similar culture or language to that of the host countries, have higher transferability levels. This indicates their human capital acquired from abroad is portable to the host countries. The researchers’ empirical results were consistent with those of previous studies (Friedberg, 2000); schooling acquired abroad has a significant effect on earnings in the host country, whereas, seemingly, experience gained elsewhere has no such effect.
Similar evidence is also found in the study of the Italian labour market by Aringa and Pagani (2010). Based on data from the Italian Labour Force Survey for the years between 2005 and 2007, Aringa and Pagani found that foreigners arriving in Italy are much more likely to be over-educated than are the natives, and that work experience acquired in countries of origin is not valued in the Italian labour market. Furthermore, experience acquired in Italy did not help to improve their education-occupation match. The researchers suggested that foreigners struggle to catch up with natives even if they adapt their skills to the host countries.
Age at arrival is expected to have a negative effect on immigrant earnings. This was shown by Friedberg (1992), who found that there was an 11.6 per cent earnings disadvantage between an immigrant who arrived in the United States at age 30 and a comparable immigrant who had migrated at age 10.
Reference to Table 3.1 here. Table 3.1 provides a succinct summary of the studies reviewed in this section.
A longitudinal analysis is applied in this study to address the potential problem of “omitted unobservable bias” from cross-sectional analysis, which is important to examine both the incidence and potential earnings penalty to over-education. Therefore, both the determinants of over-education and the impact of over-education on earnings are examined with panel techniques.
In order to obtain the estimates for comparison between the Australian-born and immigrants, two samples are examined in this study. One sample consists of the Australian-born (natives) and ESB immigrants, and the other sample is natives and NESB immigrants.
This approach allows me to examine results for each immigrant group compared to the same base category of the Australian-born.
Part 1: Determinants of over-education
I apply the correlated random effects logit model to examine the likelihood of over-education with panel data. In this model a number of important variables, such as immigrant status, are time-invariant. A conditional logit (or fixed effects logit) model which was also considered, sacrifices time-invariant but potentially important information on any individual who presents no change in dependent variables by eliminating time-invariant variables. However, this model benefits from controlling for the endogeneity from individual effects. The random effects logit model, is in comparison able to estimate the coefficient of time invariant variables whilst also allowing for dynamic adjustment. Thus, based on these considerations, I have chosen the random effects logit model to examine the determinants of over-education.12
A potential problem arises from the biases occurring in the correlation between explanatory variables and error terms in random effects models. I address this problem by using the Mundlak (1978) correction.
Mundlak’s approach is used to control for endogeneity effects due to unobserved individual effects. It is considered as a compromise between the fixed and random effects models. It also provides a test for adjustment for endogeneity as an alternative to the Hausman test–If the coefficient on group mean δ is non-zero, that suggests that individual effects are not to be ignored (Greene, 2010).
It is noted that coefficients δ will differ between panels of different lengths T and they are specific to the particular sample. The estimates of approximate the fixed effects estimators, as shown by Wooldridge (2009).
In this study, I employed both a random effects logit model and a correlated random effects logit model with Mundlak (1978) correction to estimate the determinants of over-education. As noted earlier, I consider effects for natives and ESB immigrants, and among natives and NESB immigrants, respectively. The random effects logit model is applied as a benchmark. The endogeneity issue due to the individual effects is corrected by the correlated random effects logit model with Mundlak correction. If the results from these two models are significantly different, then endogeneity is addressed by the correlated random effects logit model.
I examine the hypothesis that the incidence of over-education for immigrants may decrease with their duration of stay (YSM) in Australia. This less-examined hypothesis has important implications for understanding the labour market assimilation of immigrants in earnings models.
By this logit model setup, the natural log of the odds ratio of over-education is explained by a quadric function of years since migration (YSM) with other explanatory variables. The observed variable takes the value of 1 if worker i is over-educated and is defined as 0 otherwise. Zit denotes a set of personal or job characteristics of individual i at time period t; denotes actual years of education obtained by individual i at time t. Mi is a dummy variable, and it takes the value of 1 if individual i is an immigrant, 0 otherwise. The coefficient of Mi, δ2, measures the initial over-education gap of immigrants upon arrival relative to comparable natives. denotes the number years of residence since migrating to the host country. The coefficient of , δ6, measures the way in which the over-education gap varies as immigrants spend time in the host country. The over-education rates of immigrants are expected to signify their levels of assimilation. Therefore, the coefficient of is predicted to be negative. 7, the coefficient of 2 examines the rate of over-education in a linear or quadric style over time. A quadratic form was chosen to examine the non-liner relationship between the rate of over-education and years since migration. The rate of over-education is expected to be decreasing with increased years of residence because immigrants are more likely to find a better education-occupation match after gaining Australian experience or education. And the relationship may be flatter when years since migration reach a certain point. This non-liner relationship is examined by a quadratic form. However, the result shows that the coefficient on the quadratic term is insignificant and the coefficient on YSM is negatively significant, which justifies the use of a linear relationship between rate of over-education and YSM.
∑ =1[ ̅ ] represents the Mundlak adjustments (where is m is the number of explanatory variables).
The unobservable individual specific µi as a function of individual means, that is = ∑ =1[ ̅ ] + , where ∼ N(0, 2).
It assumes zero correlation between the means of time varying explanatory variables and . And , denotes the disturbance terms, which are assumed to be independent and identically distributed (iid).
To further examine the effects of age on arrival and year of arrival on the probability of being over-educated, I replace and 2 with age on arrival and year of arrival dummy variables, respectively, in order to avoid an over-specification problem.
Part 2: Impacts of over-education on earnings
Unobserved heterogeneity, such as unobserved ability, motivation or work efforts influence earnings, and also are correlated with observed education and skills. If these unobserved individual effects, ui, are correlated with explanatory variables, cross-sectional analysis would result in omitted unobservable biases. Longitudinal data captures the same individual over time. Thus, unobservable individual effects are eliminated by using a panel fixed effects model. Thus, estimation results from fixed effects models are consistent. However, this model cannot evaluate the time-invariant explanatory variables because they are removed by within-group transformation. In contrast, a random effects Generalised Least Squares (GLS) model assumes that ui is uncorrelated with explanatory variables in which GLS uses the optimal combination of within-group and between-group variations. If individual effects do not matter, then the GLS estimator is equal to the ordinary least squares (OLS) estimator. A Hausman test is used to identify whether the random effects GLS estimator is biased.
Two specifications are employed to examine earnings effects. One specification is to examine the effect of over-education on earnings through years since migration (YSM). Thus, assimilation effects for immigrants are found by the significance of YSM. The other specification is to examine the impacts of over-education on earnings from both pre-migration and post-migration human capital perspectives. By doing so, the transferability of immigrants’ human capital by their country of origin is evaluated. These two specifications are expressed by Models 2 and 3 respectively.
In this section, the analysis focuses on the link between over-education and earnings. The following questions are of interest in the empirical analysis. How does over-education impact, directly or indirectly, on earnings via years since migration and migration status? Is the impact of over-education on earnings affected by unobserved heterogeneity, such as, personal ability or variable quality or under-valuation of immigrant qualifications?
The standard Over-education, Required-education, Under-education (ORU) earnings model (as originally proposed by Duncan and Hoffman (1981)13) is widely used in ‘over-education’ empirical research. Based on the standard ORU earnings model, the extended earnings’ model is applied into this study for issues of interest.
The ORU earnings model decomposes actual years of education (Sa) into required years of education (Sr), years of over-education (So), and years of under-education (Su). Thus Sa = Sr + So – Su, where So= Sa – Sr for the over-educated (i.e. if Sa > Sr), and 0 otherwise. Similarly, Su= Sr – Sa for the under-educated if (i.e. Sr > Sa), and 0 otherwise.
Then the log of earnings in the ORU model can be written as: (3.7) =α1+ + + + 1 1+ε
is the natural logarithm of earnings, 1 is a vector of a variety of other control variables that generally includes personal characteristics and job characteristics, , , are, respectively, the years of required education, over-education, and under-education. α1is the intercept term, and ε is an error term.
Equation (3.7) estimates βr, βo, βu continuously, and βr, βo, βu are the rates of returns to required education, over-education and under-education respectively.
Prior literature on ‘over-education’ has consistently found that βr > βo and βo >0, such that the return of over-education is lower than the return to required education; and the return to over-education is positive (Cohn, 1992; Groot, 1996; Rumberger, 1987; Sicherman, 199114). In contrast, they also found that βu < βr and βu <0, which means the return to under-education is lower than the return to required education; and that it is a negative return (Hartog, 2000).
In panel data settings the ORU model is expressed as follows:
The extended ORU earnings model is built by adding interaction terms to Equation (3.8) to examine the impacts of educational mismatch, years since migration and migrant status on the return to over-education, after controlling for the individual effects. By doing so, I can examine the earnings gap between immigrants and natives via educational mismatch. These results reveal an added and less-studied explanation for the existing earnings disadvantage for immigrants in the Australian labour market.
TABLE OF CONTENTS
TABLE OF CONTENTS
LIST OF TABLES
LIST OF FIGURES
2. ESSAY ONE: THE INCIDENCE OF OVER-EDUCATION AND ITS EARNINGS EFFECTS
2.2 Measures of over-education
2.3 An overview of the empirical literature
2.4 Data and variables
2.5 Incidence of over-education
2.6 Impacts of over-education on earnings
3. ESSAY TWO: A LONGITUDINAL ANALYSIS ON THE INCIDENCE OF OVER-EDUCATION AMONG IMMIGRANTS AND ITS IMPACTS ON EARNINGS
3.2 Review of the literature
3.3 Econometric framework
3.4 Data and variables
3.5 Determinants of over-education results
3.6 Impact of over-education on earnings results
4. ESSAY THREE: DYNAMIC EFFECTS OF OVER-EDUCATION AND OVER-SKILLING
4.2 Literature review
4.4 Analytical framework
4.5 Data and variables
4.6 Estimations and empirical results
GET THE COMPLETE PROJECT