Earnings and occupational attainment: immigrants and the native born



Download 413.1 Kb.
Page2/4
Date conversion24.05.2016
Size413.1 Kb.
1   2   3   4

Table 3

Payoffs to Educational Attainment and Labor Market Experience from Analysis of Earnings, Native Born Males, Age 25-64, 2000





Payoffs from Earnings Function

% Change



Variable



Standard

Controlling for 23 Occupations

Controlling for 509 Occupations


With 23 Occupations


With 509 Occupations

Educational Attainment

10.6

8.2

5.8

-23

-45


Labor Market Experience

- 10 years

- 20 years

2.16


1.02

2.16


1.02

2.10


1.00

0

0


-3

-2



Source: Authors’ calculations based on Table 1 results.
In the case of the foreign born, however, the payoff to pre-immigration experience (i.e., experience controlling for years since migration) actually increases compared to that obtained with the conventional model that eschews information on occupation. Payoffs evaluated at 10 and 20 years of pre-immigration experience are presented in Table 4. These reveal that, following standardization for occupation, the payoff to pre-immigration experience increases by around 40 percent at 10 years of experience, and by 20 percent at 20 years of experience. In other words, pre-immigration experience must be associated with immigrants being channelled in the destination into relatively low paying occupations. This appears to be a major result of the less-than-perfect international transferability of skills acquired on the job in the country of origin (see Chiswick, 1978).

When the relationship between earnings and years since migration is considered, it is apparent that controlling for the inter-occupational earnings structure has minimal impact on the estimates (see, in particular, the summary in Table 4). This is very similar to the finding in relation to labor market experience for the native born. In other words, the earnings growth that immigrants and the native born achieve as a result of U.S. labor market experience comes about through intra-occupation earnings mobility.


Table 4

Payoffs to Educational Attainment, Pre-Immigration Labor Market Experience and Duration of Residence from Analysis of Earnings, Foreign Born Males, Age 25-64, 2000





Payoffs from Earnings Function

% Change



Variable



Standard

Controlling for 23 Occupations

Controlling for 491 Occupations

With 23 Occupations

With 491 Occupations

Educational Attainment

5.3

3.2

2.3

-40

-57


Pre-Immigration Experience

- 10 years

- 20 years

0.88


0.56

1.24


0.68

1.30


0.70

+41


+21

+48


+25

Years Since Migration

- 10 years

- 20 years


0.88


0.66

0.96


0.72

0.92


0.64

+9

+9



+5

-3



Source: Authors’ calculations based on Table 2 results.
The estimated effects of English proficiency on earnings are also affected by the statistical control for occupation. For example, among the foreign born, the estimated effect in the standard earnings function of speaking a language other than English at home and speaking English “Very Well” is 8 percent lower earnings compared to English only speakers. This hardly changes (an increase of only one percentage point to 7 percent lower earnings) when account is taken of the major occupational groups. The changes in estimated impacts are more pronounced for the poorer English proficiency groups, with the largest change following standardization for occupation being the improvement from -0.373 to -0.269 in the estimated coefficient for the “English Not Well” variable. Thus, some of the earnings disadvantage of immigrants with limited English language proficiency is due to this deficiency placing them in lower earnings occupations.

The patterns of change to the remaining estimated coefficients in the earnings equation following the standardization for occupation are similar for the two birthplace groups. Hence, there is little change in the estimated elasticity of earnings with respect to weeks worked, a reduction of 13 to 16 percent in the estimated marriage premium, an increase of 14 to 20 percent in the earnings penalty associated with residence in the South, and a large fall (of between 33 to 40 percent) in the earnings penalty associated with a Black racial origin. The changes to the coefficients for residence in the metropolitan areas and veteran status are in the same direction for the native born and the foreign born, but the changes are much more pronounced for the foreign born.

The inclusion of more detailed information on occupation in column (iii) of Tables 1 and 2 reinforces these key results.3 Hence, when the inter-occupational earnings structure is held constant at this more detailed level, the payoff to schooling (achieved through intra-occupational earnings mobility) for the native born falls, to a little more than one-half of that reported in the absence of controls for occupation. For the foreign born, the payoff to schooling achieved through intra-occupational earnings mobility is only around two-fifths of that reported in the conventional earnings equation that combines both inter-occupational and intra-occupational earnings effects. The payoff to pre-immigration labor market experience among the foreign born increases, by between 50 and 70 percent of the payoffs reported for the conventional earnings function. The payoffs to labor market experience for the native born and to years of residence in the US for the foreign born are virtually unaffected by the degree of detail on occupation used in the analysis. The estimated effects on earnings of English proficiency for both birthplace groups are reduced even further with the more detailed occupational classification. These range between 30 and 50 percent less (in absolute value) in the model with the larger number of occupations than when occupation is not included in the earnings function.

These results present a conundrum. Labor market experience among native-born males aged 25 to 64 in the US labor market does not, from this study of earnings, appear to be associated with mobility to higher paying occupations. Yet the foreign born with labor market experience are being assigned to lower-ranked occupations. One possibility is that this is a reflection of a matching process in the labor market, where education and other, essentially pre-labor-market-entry, characteristics determine occupation. Provided their education is of a recent vintage, an immigrant will get access to an appropriate occupation. But if the education is of an older vintage, the immigrant gets assigned to a lesser job, possibly because it is more difficult for employers to assess how relevant the immigrant’s skills are to the current labor market, or because these generalized skills become more country-specific with longer work experience in the origin. It is noted that the same phenomenon appears to occur in relation to the earnings of women. Miller (1987), for example, estimated an ordered probit model of occupational attainment for married women (albeit with only six separate occupations) and reported that labor market experience was linked to employment in occupations with lower mean earnings. This was shown to be a result of a negative relationship between time spent out of the labor market and occupational prestige after returning to employment. Time spent out of the destination labor market, as measured by pre-immigration experience, may have a similar effect.

Clearly, occupation plays a key role in determining the earnings outcomes for both immigrants and the native born. It seems important, therefore, to attempt to understand this directly through analysis of models of occupational attainment, rather than inferring patterns from the extended earnings functions considered in Tables 1 and 2. These matters are addressed in the remainder of this paper.
III. MODELLING OCCUPATIONAL OUTCOMES

Studies of occupational outcomes have usually been based on a general model as follows:

(1)

where Y is a continuous variable which records the perceived “goodness of an occupation”, X is a set of attributes of workers thought likely to impact on occupational outcomes, and is a stochastic disturbance term.

Studies of this genre generally proceed in either of two main ways. In the first of these, typified by the work of Brown et al. (1980a)(1980b), Miller and Volker (1985), and Kidd and Shannon (1996), the likelihood of working in a number of occupational categories is examined using probability models. The basis of these models is the argument that Y, the perceived “goodness of an occupation”, is unobserved. Instead, what is usually analyzed is discrete data on occupation type. These data take the form of a variable , where if the individual works in the jth occupation. These studies then examine the determinants of the conditional probability that an individual i works in occupation h as:

(2)


When occupational outcomes have been examined from this discrete choice perspective, the main method of estimation has been the multinomial logit model. With this model the discrete response variable consists of a set of mutually exclusive and exhaustive occupation categories that can be ranked arbitrarily without any effect on the value of the estimated parameters. Under this approach, the conditional probability that individual i ends up in occupation h is given as (see Schmidt and Strauss, 1975; Brown et al., 1980; and Polachek, 1981):

(3)


The problem with this approach is that the large number of categories makes interpretation difficult.

An alternative probability model that has been used is the ordered probit model, which is to be preferred in this analysis. This model is similar to that of the unordered models in providing a prediction of the conditional probability that individual i ends up in occupation h, although it presumes that the occupational categories can be suitably ranked (see McKelvey and Zavoina, 1975; Miller and Volker, 1985). Accordingly, the j occupations are ranked from lowest to highest, using an underlying scale of measurement. These observed occupational categories are then linked to the unobserved underlying index of the “goodness” of the occupation (Y in equation (1)) as follows:


where the s are unknown threshold parameters separating the adjacent occupational categories. These can be estimated together with the in the index specified in equation (1). With the normal distribution, the conditional probability of individual i being in occupation h can be calculated as:


(5)
Various alternatives have been proposed for ranking occupations. For example, Miller and Volker (1985) use both status-attainment scores and income to establish rankings among occupations in their empirical analyzes and find marked differences between the estimated probabilities based on these two alternative rankings. However, they contend that no one ordering scheme is necessarily superior, and recommend analysis with several ordering variables.

The alternative approach to modelling occupational outcome is to focus on occupational attainment models, estimated using Ordinary Least Squares. Examples of this approach are Nickell (1982), Evans (1987), Evans and Kelley (1986) and Polachek (1981). Polachek (1981) characterizes occupations by their atrophy rates, that is, the rate at which earnings decreases with absence from the labor market. He uses the occupation atrophy rates as the dependent variables in a model of occupational choice. Evans (1987) and Evans and Kelley (1986) measure occupational attainment in Australia using a status attainment score, specifically the ANU2 occupational status scores. This scale is based on prestige ratings, and provides a link between census occupational classifications and popular ratings of the social standing of jobs (Jones, 1989). On the other hand, Nickell (1982) measures success using the average hourly earnings within each occupation. An argument in favor of the use of mean earnings over a status attainment scale is that changes in earnings are amenable to a clear, quantitative scale. An argument in favor of the use of a status attainment scale is that it is more encompassing than the simple monetary magnitude provided by mean occupational earnings. Duncan (1961), for example, discusses education as an indicator of social status and income as a measure of economic status, and the socioeconomic index as reflecting both of these. However, Nickell (1982) reports a correlation of 0.85 between his hourly income variable and a status ranking variable4, which suggests that similar findings should be derived from the study of these alternative measures of the goodness of occupations within a linear regression framework.5,

The relative merits of the probability and occupational attainment models cannot be evaluated formally (i.e., in a statistical sense). In this situation, an expedient way of proceeding is to tailor the method to the specific issue that is to be addressed. In this particular instance, the main research question is the role of the worker’s characteristics in getting them access to better jobs. From this perspective, following the status attainment literature has merit.6 Within this framework, the approach of Nickell (1982) is followed, and mean occupational earnings are used to rank occupations.

Hence, the analysis proposed is the estimation of a status attainment model:



(6)

where is the mean occupational earnings of the Census occupational category (around 500 specific occupations) in which individual i works, is a set of the individual’s attributes that influences this occupational outcome, and is a random error term. As a check on the robustness of the empirical findings, ordered probit models are also estimated using mean occupational earnings as the ranking instrument, but with the more limited number of occupations – 23 – provided by the Census major occupational groups.


IV. ESTIMATES OF MODELS OF OCCUPATIONAL ATTAINMENT

This section presents estimates of several variants of the model of occupational attainment set out in equation (6).



A. OLS and Ordered Probit Analysis of Occupational Attainment

Results are presented in Table 5 for the native born and in Table 6 for the foreign born. These results refer to a model of occupational attainment, estimated by Ordinary Least Squares, and using the unit level occupation data for about 500 occupations (columns (i) and (ii) of each table), and a probability model of occupational outcomes estimated using an ordered probit model, based on only the 23 major group occupations (columns (iii) and (iv) of each table). The estimated separation points (i.e., the s) for the ordered probit model are not listed: these are available from the authors upon request.

Two specifications of the estimating equation are presented under each approach. The first does not contain variables for English language proficiency, while the second does. While there are reasonably large numbers in the various English proficiency categories for the foreign born, this is not the case for the native born. Among the native born there are limited numbers in the English skills categories other than the “Very Well” group, and the meaning of reporting speaking English in the “Well”, “Not Well” and “Not at All” categories for the native born is not at all clear. Chiswick and Miller (1998), for example, argue that the native born who report that they are bilingual and speak English “Very Well” may have lesser proficiency in English than monolingual English speakers because speaking in childhood and/or as adults this other language competes with their obtaining full proficiency in English. Chiswick and Miller (1998) also argue that this small group of the native born may experience discrimination because of an accent or speech pattern related to their other language. Finally, it has been advanced that living and working in an ethnic concentration area because of their language deficiencies may also impact on their labor market outcome. Yet, it is unclear why there would be adult men born in the US who report their English-speaking proficiency as less than “Very Well”.

The pattern of effects in the OLS models of occupational status is remarkably the same as that in the ordered probit models. For brevity, only the OLS results will be discussed.7



Table 5

Estimates of Models of Occupational Attainment (Ranked by Mean Occupational Earnings), Native Born Males, Age 25-64, 2000

(Dependent Variable: Mean Occupational Natural Logarithm of Annual Earnings)


Variable


OLS-without English Variables

(i)


OLS-with English Variables

(ii)

Ordered Probit-without English Variables

(iii)

Ordered Probit-with English Variables

(iv)

Constant

8.754

(1223.98)



8.759

(1221.85)



-1.267

(73.58)


-1.262

(73.11)


Years of Education

0.083

(312.62)


0.083

(311.98)


0.209

(358.19)


0.209

(357.54)


Experience

0.001

(2.56)


0.001

(2.28)


-0.009

(15.74)


-0.009

(15.86)


Experience Squared/100

0.002

(4.15)


0.002

(4.31)


0.025

(21.50)


0.025

(21.56)


Log Weeks Worked

0.088

(62.26)


0.088

(62.02)


0.157

(45.97)


0.157

(45.85)


Married

0.103

(89.28))


0.102

(88.98)


0.179

(59.17)


0.179

(59.03)


South

0.026

(23.62)


0.026

(23.72)


0.051

(17.47)


0.051

(17.50)


Metropolitan Area

0.065

(24.16)


0.065

(24.36)


0.095

(13.44)


0.095

(13.53)


Veteran

-0.009

(7.50)


-0.009

(7.43)


-0.055

(16.22)


-0.055

(16.17)


Blacks

-0.115

(67.47)


-0.116

(67.91)


-0.275

(58.87)


-0.276

(59.03)


English Very Well

(a)

-0.023

(9.34)


(a)

-0.034

(5.31)


English Well

(a)

-0.030

(4.95)


(a)

-0.007

(0.44)


English Not Well/Not at All

(a)

0.015

(1.67)


(a)

0.048

(2.17)




0.271

0.271

-

-



-

-

127556.4

127589.8

Sample Size

533,906

533,906

533,906

533,906

Notes: ‘t’ statistics in parentheses.

(a) Variable not entered.



Source: 2000 US Census 1% PUMS.
1   2   3   4


The database is protected by copyright ©essaydocs.org 2016
send message

    Main page