Ageat Arrival, Economic Assimilation, and the Occupational-Workplace Sorting ofChildhood Immigrants: Evidence from Longitudinal Employer-Employee Data

Are SkeieHermansen, University of Oslo, Email: a.s.hermansen@sosgeo.uio.no

Backgroundand the Contribution of this Study

Age at arrival clearly demarcates the life cycle stageat which immigrants start life in a new country and is a key determinant of socioeconomicassimilation. Prior research on the earnings gradientsby age at arrival among childhood immigrants has addressed processes ofeducational and occupational sorting (Wangand Wang, 2011, Bacolod and Rangel, 2017). While access to linkedemployer-employee data has made a focus on workplace variation in inequality-generatingprocesses possible, these types of data has not yet been exploited for thestudy of childhood immigration.

This study investigates the relation between age atarrival and adult labor market productivity. First, I presentthe first evidence on distributional variation in the effects of age at arrivalon adult earnings using sibling fixed-effects models.  Second, I present thefirst evidence using longitudinal employer-employee data to study age-at-arrivalgradients in adult earnings and sorting processes within the labor market.

Dataand Methods

Norwegian registries enable the linkage of parents tochildren as well as employees to their employers. I consider all children with twoforeign-born parents in the birth cohorts 1960-1989 who were either born inNorway or arrived at age 18 or earlier, observed annually from age 25 andonwards. The final analytic sample includes about 45,000 persons and about202,000 person-years.

            Age at arrival is measured as thedifference between year of immigration and year of birth. I focus on annualobservations on each person’s natural logarithm of annaul earnings. Completededucation is measured using information on the child’s highest level ofeducational qualification at age 25, and is entered in the models as a set of eightdummies. Occupations are measured using a four-digit code for occupationaltitles. Establishments are defined as distinct establishments with distinctlocations defined by the postal address of the workplace. Occupation-establishmentunits refer to unique combinations of observations holding the sameoccupational title within the same establishment.            Istart comparing estimates in models that compare childhood immigrants arrivingat different ages from the same origin country (i.e., country-of-origin fixedeffects models) and childhood immigrants from the same family of origin (i.e.,sibling fixed effects) The sibling fixed-effects models enablesme to disentangle the effect of age at arrival from all fixed family-levelconditions and endowments shared by siblings (Hermansen,2017).In addition to OLS regressions, I use unconditionalquantile regression (UQR) (Firpo,Fortin and Lemieux, 2009) to address whether the  exhibit varyacross the earnings distribution.

            Then,I use the OLS regressions with country-of-origin fixed effects to explore thesemediating channels by first adjusting for own education and then subsequentlyby including fixed effects for occupation, establishment, andoccupation-establishment units. By making comparisons at different levels inthe data, these fixed-effects models are informative on whether theage-at-arrival gradients in (average) earnings reflect the differential sortingprocesses by age at arrival or whether earnings differ by age at arrival evenwhen they occupy similar positions.

PreliminaryResults and Future Directions

Figure 1compares distributional variation in the age-at-arrival effect on log annualearnings between Norwegian-born children of immigrants (as indicated by solidline at zero) and childhood immigrant arriving at different ages (dummies forarrival ages in two-year intervals) throughout childhood and adolescence withUQR models. While there is some variation across different quantiles and thetwo model specifications, the overall pattern in both panels is that age at arrivalhas a long-run negative impact on adult earnings and that this effect can beobserved at different levels throughout the earnings distribution.

            Figure2 compares the estimates from the baseline age-at-arrival earnings gradients inthe country-of-origin fixed-effects models with estimates where I addadjustments for education and fixed effects for occupations, establishments,and occupation-establishment units. These models show is that while adjustingfor education accounts for a part of the age-at-arrival gradient, the mainmediator of this relationship is differences in the sorting of childhoodimmigrants into different occupations. Moreover, sorting by establishments andoccupation-establishment units only adds a relatively little in terms of accountingfor the overall relationship between age at arrival and adult earnings. 

References

Bacolod, M. andRangel, M. A. (2017). Economic Assimilation and Skill Acquisition: EvidenceFrom the Occupational Sorting of Childhood Immigrants, Demography, 54,pp. 571-602.

Firpo, S., Fortin, N.M. and Lemieux, T. (2009). Unconditional Quantile Regressions, Econometrica,77, pp. 953-973.

Hermansen, A. S.(2017). Age at Arrival and Life Chances Among Childhood Immigrants, Demography,54, pp. 201-229.

Wang, C. and Wang, L. (2011). Language Skills andthe Earnings Distribution Among Child Immigrants, Industrial Relations, 50,pp. 297-322.