Occupational Attainment and Immigrant Economic Progress in Australia*

I Introduction

Immigrants’ labour market outcomes are generally discussed around three key concepts—the less-than-perfect international transferability of the human capital skills they acquired in their country of origin, the positive selectivity of immigrants for labour market success, especially economic immigrants, and their relatively rapid economic progress in the destination country.¹ The less-than-perfect international transferability of human capital skills results in immigrants being at an economic disadvantage during their first year in the destination country. Immigrants’ rapid economic progress, particularly in the immediate post-arrival period, results in a narrowing of this gap and, especially for economic immigrants, can result in a ‘catch-up’ of their economic position compared to that of their native-born counterparts (Chiswick, 1978). This catch-up will occur when the effects of positive selection more than off-set the lingering effects of imperfect skill transferability and any discrimination against immigrants. However, after three decades of intensive research, understanding of the process of immigrant labour market adjustment that gives rise to these patterns is still far from complete.

In a recent study, using data on adult men in the USA, Chiswick and Miller (2007) argue that insights into the labour market adjustment of immigrants can be gained through estimation of earnings equations that take account of occupational status.² Equations that also include controls for occupation show the role that occupation has as an intermediary between immigrants’ human capital skills and their earnings. Nearly 60 per cent of immigrants’ earnings gains in the USA can be attributed to inter-occupational earnings differences, with just over 40 per cent to intra-occupational differences, in contrast to 55 or 45 per cent for native-born men in the USA.

The comparison of the Australian analyses in this paper with the Chiswick and Miller (2007) findings is particularly relevant given the difference in the primary focus in rationing immigrant visas, the applicant's skills in Australia and family reunification in the USA.

The structure of this paper is as follows. Section II reviews the data from the 2001 Australian Census of Population and Housing that is used in the statistical analyses, with a special emphasis on the information on occupation. It also outlines the specification of the estimating equation. Empirical results from the analysis of earnings are presented in Section III. Section IV provides information on the determinants of occupational attainment that assists in the explanation of the findings reported in Section III. A summary and conclusion are provided in Section V.

II Data and Earnings Equation

The data are from the 2001 Australian Census of Population and Housing 1 per cent sample of households (ABS, 2003). They include information on age, birthplace, educational attainment, marital status, current employment status, earnings and occupation, among other variables. The Expanded Confidentialised Unit Record Files available only through the Remote Access Data Laboratory (RADL) is used in this study.³

The information on occupation is coded according to the Australian Standard Classification of Occupations (ASCO), second edition (ABS, 1997). In the 1 per cent sample only 44 occupational categories are distinguished (See Appendix A). These are used in two forms in the analyses. First, the 44 occupational categories are aggregated into the nine ASCO Major Groups. Second, all of the 44 separate occupational categories are used as the basis of the empirical investigation.⁴

The analyses are restricted to men aged 20–64 who were employed on a full-time basis (i.e. they worked 35 or more hours per week) in the week before Census night and who reported positive weekly earnings. Appendix B contains definitions of all variables and a table of means and standard deviations.

The earnings function initially estimated takes the following form:

where the dependent variable is the natural logarithm of hourly earnings, Z_i is a vector of the individual and job-related characteristics that affect the earnings of individual i, ɛ_i is the error term, and β is a vector of parameters to be estimated. The variables considered in Z consist of years of schooling, labour market experience and its square, dichotomous variables for government employment, marital status, and birthplace (Australia or foreign born), and variables for duration of residence for immigrants and English language proficiency. Controlling for duration in Australia, the labour market experience variables measure the effect on earnings of pre-immigration labour market experience.

The estimates obtained from Equation 1 provide a benchmark set of results for the links between productivity related characteristics and earnings. Two extensions of Equation 1 are considered. The first of these involves augmentation with dichotomous variables for the Major Group occupations. Eight dichotomous variables are considered, with Managers and Administrators as the benchmark group. The second extension involves including dichotomous variables for each of the 44 occupations included in the Census classification.

Each of these extensions controls for the occupational earnings structure, albeit at different levels of detail. The coefficients on the variables for occupation provide information on the effect on earnings of employment in the particular occupation. This is a direct effect of occupation on earnings. With occupation held constant, the coefficients of the other explanatory variables reflect their effect on earnings within occupations. Hence, comparison of the estimates in the equation with the occupation variables with estimates from the benchmark Equation 1 provide information on the effect of these variables on earnings through intra- and inter-occupational change.

III Analysis for Earnings

(i) Earnings Functions

The results from these analyses are presented in Table 1 and summarised in Table 2. These tables have separate panels for the Australian born, immigrants from English-speaking developed countries, and immigrants from non-English-speaking countries.

Table 1. Estimates of Earnings Functions by Birthplace, Men Aged 20–64 Years, 2001 ^†

Variables	Australian born			Foreign born
				English-speaking countries			Non-English-speaking countries
	(i)‡	(ii)	(iii)	(i)‡	(ii)	(iii)	(i)‡	(ii)	(iii)
Constant	1.403 (58.13)	1.721 (49.28)	2.022 (36.97)	1.559 (24.91)	2.237 (28.24)	2.153 (15.12)	1.944 (30.68)	2.422 (31.12)	2.468 (22.30)
Education	0.088 (52.41)	0.067 (31.36)	0.052 (24.71)	0.081 (22.37)	0.050 (11.57)	0.046 (10.87)	0.058 (16.84)	0.031 (7.82)	0.020 (5.22)
Experience	0.028 (22.79)	0.027 (21.05)	0.026 (21.89)	0.029 (8.53)	0.025 (7.63)	0.026 (8.13)	0.008 (2.70)	0.011 (3.68)	0.014 (4.84)
(Experience)2/100	−0.047 (17.86)	−0.048 (18.45)	−0.044 (17.65)	−0.051 (7.49)	−0.046 (6.93)	−0.047 (7.26)	−0.015 (2.46)	−0.022 (3.71)	−0.027 (4.69)
Married	0.097 (12.11)	0.088 (11.13)	0.083 (10.83)	0.151 (7.05)	0.125 (6.08)	0.109 (5.37)	0.100 (4.97)	0.093 (4.82)	0.086 (4.69)
Government	0.150 (19.20)	0.135 (17.32)	0.116 (14.43)	0.106 (4.91)	0.091 (4.42)	0.109 (4.75)	0.162 (7.66)	0.154 (7.33)	0.145 (6.74)
Speaks English§
Very Well	NE	NE	NE	NE	NE	NE	−0.030 (1.35)	−0.021 (1.01)	−0.021 (1.04)
Well	NE	NE	NE	NE	NE	NE	−0.135 (5.50)	−0.086 (3.51)	−0.066 (2.79)
Not well	NE	NE	NE	NE	NE	NE	−0.226 (6.90)	−0.183 (5.60)	−0.152 (4.86)
Not at all	NE	NE	NE	NE	NE	NE	−0.421 (4.89)	−0.397 (4.84)	−0.312 (3.70)
Year of arrival§
1991–1995	NE	NE	NE	−0.040 (0.92)	−0.033 (0.81)	−0.030 (0.76)	0.011 (0.32)	0.025 (0.75)	0.012 (0.39)
1986–1990	NE	NE	NE	−0.055 (1.58)	−0.030 (0.92)	−0.026 (0.85)	0.031 (0.99)	0.039 (1.31)	0.030 (1.04)
Before 1986	NE	NE	NE	−0.085 (3.31)	−0.077 (3.25)	−0.069 (3.05)	0.107 (3.65)	0.100 (3.56)	0.083 (3.06)
Occupation	NI	INC	INC	NI	INC	INC	NI	INC	INC
	0.190	0.210	0.291	0.194	0.256	0.306	0.168	0.223	0.300
Sample size	20 709	20 709	20 709	3127	3127	3127	3752	3752	3752

• Notes: ^† Heteroscedasticity-consistent ‘t’ statistics in parentheses; ^‡ specification (i) is the benchmark model that does not contain information on occupation, specification (ii) contains dichotomous variables for the Major Group occupations, while specification (iii) contains dichotomous variables for the more detailed (44) Census occupational categories; ^§ The omitted category for the Speaks English variable is ‘Speaks only English’ and that for the Year of Arrival variable is ‘After 1995’; NE, variables not entered; NI, occupation not included; INC, occupation included. The dependent variable is the natural logarithm of hourly earnings.
• Source: 2001 Australian Census of Population and Housing, one per cent sample.

Table 2. Payoffs to Selected Characteristics from Analysis of Earnings, by Birthplace, Men Aged 20–64, 2001

Variable	Payoffs from earnings function			% Change
	Standard	Controlling for 9 occupations	Controlling for 44 occupations	With 9 occupations	With 44 occupations
A. Australian born
Educational attainment	8.8	6.7	5.2	−23.9	−40.9
Experience
10 years	1.86	1.74	1.72	−6.5	−7.5
20 years	0.92	0.78	0.84	−15.2	−8.7
B. Overseas born, English-speaking developed countries
Educational attainment	8.1	5.0	4.6	−38.3	−43.2
Pre-immigration experience
10 years	1.88	1.58	1.66	−16.0	−11.7
20 years	0.86	0.66	0.72	−23.3	−16.3
Migrated
Before 1986	−8.5	−7.7	−6.9	9.4	18.8
C. Overseas born, non-English-speaking countries
Educational attainment	5.8	3.1	2.0	−46.6	−65.5
Pre-immigration experience
10 years	0.5	0.66	0.86	32.0	72.0
20 years	0.2	0.22	0.32	10.0	60.0
Migrated
Before 1986	10.7	10.0	8.3	−6.5	−22.4
Speaks English
Well	−13.5	−8.6	−6.6	−36.3	−51.1
Not well	−22.6	−18.3	−15.2	−19.0	−32.7
Not at all	−42.1	−39.7	−31.2	−5.7	−25.9

• Source: Authors’ calculations based on Table 1.

The payoff to years of schooling declines by between 24 and 47 per cent when the occupational structure is taken into account at the major group level (nine occupational categories). The largest decline is for immigrants from non-English-speaking countries, and the smallest is for the Australian born. For immigrants from non-English-speaking countries, these results indicate that almost one-half of the increase in earnings associated with each extra year of schooling comes about through this extra schooling facilitating access to higher paying occupations.

There are further reductions in the payoff to schooling when the finer degree of detail on occupation is included in the estimating equation. Among the Australian born, the payoff to schooling falls by 41 per cent, from 8.8 to 5.2 per cent. The change in the payoff to schooling for the foreign born from English-speaking countries is similar: it falls from 8.1 to 4.6 per cent, a 43 per cent reduction. In other words, slightly more than 40 per cent of the increments in earnings associated with extra years of schooling for immigrants from English-speaking countries derives from inter-occupational earnings differences, and slightly less than 60 per cent derives from increases in earnings within occupations.

Among immigrants from non-English-speaking countries, however, the payoff to a year of schooling falls from 5.8 to 2.0 per cent when account is taken of employment in the 44 occupations. This is a 66 per cent reduction in the payoff to schooling. In other words, two-thirds of the payoff to schooling for immigrants from non-English-speaking countries is associated with access to higher paying occupations. Schooling is indicated here as being of far greater importance for earnings via occupational change for immigrants from non-English-speaking countries than it is for immigrants educated in Australia or other English-speaking countries.

There is minimal change to the payoff to labour market experience for those born in Australia, with the reduction ranging from 7 to 15 per cent. This implies that labour market experience has only a modest impact on occupational status for those born in Australia.

The control for occupation has a slightly greater impact on the payoff to pre-immigration experience for immigrants from English-speaking countries. This ranges from 12 to 23 per cent. Among immigrants from non-English-speaking countries, the pattern of effects is quite different, with the payoff to pre-immigration labour market experience rising once account is taken of occupation. Evaluated at 10 years, the payoff to pre-immigration labour market experience rises from 0.5 per cent per year in the benchmark model, to 0.66 per cent (a 32 per cent increase) following control for Major Group occupation. It rises further to 0.86 per cent (a 72 per cent increase over the benchmark model) when dichotomous variables for the 44 Census occupations are included in the model.

The immigrant duration variables (i.e. post-migration experience) have opposite patterns for the two groups of immigrants. Immigrants from non-English-speaking countries who arrived before 1986 are shown to have significantly greater earnings than the most recent arrivals (1996–2001). This earnings advantage falls by between 7 and 22 per cent once occupation is held constant. Immigrants from English-speaking countries, however, who arrived in the past 5 years are shown to have relatively high earnings, though their earnings position only differs significantly from the group who arrived before 1986.⁵

Finally, it is seen that there are pronounced changes to the earnings effects associated with English proficiency following the incorporation of information on occupation into the earnings equation. The changes range from 6 to 36 per cent when the information on Major Group occupation is used, and from 26 to 51 per cent when the information on all 44 Census occupational categories is used.

These changes in the estimated effects as the earnings equation is augmented with information on occupation follow the pattern found for the US labour market by Chiswick and Miller (2007). Comparison of Table 2 for Australia and Table 3 for the USA reveals that the changes in the effects of educational attainment on earnings following standardisation for occupation are broadly the same in Australia and the USA. Precise estimates of the differences in the relative magnitudes of the changes in Australia and the USA due to holding occupation constant are hard to assess, given the different definitions of the variables and the level of detail on occupation. It appears, however, that inter-occupational earnings differences are greater in the payoff to education in Australia than in the USA. This may follow from the more centralised system of wage determination, and perhaps greater union power, in Australia than in the USA, and as a result the more egalitarian distribution of earnings within each occupation (see Miller et al., 1995).

Table 3. Payoffs to Selected Characteristics from Analysis of Earnings, by Birthplace, Men Aged 25–64, 2000 US Census

Variable	Payoffs from earnings function			% Change
	Standard	Controlling for 23 occupations	Controlling for 509 occupations	With 23 occupations	With 509 occupations
A. US born
Educational attainment	10.6	8.2	5.8	−23	−45
Experience
10 years	2.16	2.16	2.10	0	−3
20 years	1.02	1.02	1.00	0	−2
B. Foreign born
Educational attainment	5.3	3.2	2.3	−40	−57
Pre-immigration experience
10 years	0.88	1.24	1.30	+41	+48
20 years	0.56	0.68	0.70	+21	+25
Years since migration
10 years	0.88	0.96	0.92	+9	+5
20 years	0.66	0.72	0.64	+9	−3
Speaks English
Very well	−8.0	−7.1	−5.7	−11.3	−28.8
Well	−26.1	−17.7	−13.4	−32.2	−48.7
Not well	−37.3	−26.9	−21.7	−27.9	−41.8
Not at all	−37.8	−30.0	−25.2	−20.6	−33.3

• Note: Only 11 per cent of the foreign born were from English-speaking developed countries.
• Source: Sourced from Chiswick and Miller (2007).

The impact of taking account of occupation in the earnings function on the payoff to experience is also broadly the same for immigrants from non-English-speaking countries in Australia (Table 2) and for immigrants in the USA (Table 3).

Similar changes are associated with limited English skills in the two labour markets. There is a reduction in the earnings disadvantage associated with limited English skills when occupation is held constant, or equivalently, part of the earnings advantage associated with better English language skills comes about through these skills facilitating the workers’ access to higher paying occupations. The changes in the partial effects of English-language proficiency are greater in the Australian labour market than is the case in the US labour market for those who report they speak English very well, but smaller for those with a lower level of proficiency.

Using the standard formula for analysing omitted variables bias, the changes in the estimated coefficients summarised in Table 2 are due to two sets of factors. First, there is the independent effect that occupation has on earnings in the augmented equation. Second, there are the correlations between the other explanatory variables (such as educational attainment) and occupation. The differences in the effects that controlling for occupation has on the partial effects of educational attainment, duration, pre-immigration labour market experience and the English proficiency variables for the Australian born and immigrants must therefore be due to both of these factors. If either one is zero, the effect of the explanatory variable is the same in the benchmark and the augmented equation.

(ii) Occupation Fixed Effects

Differences in the impact of occupation on earnings across the three birthplace groups can be assessed informally by plotting the fixed effects from the respective earnings equations. 1, 2, respectively, present the plot of the estimated occupational ‘fixed effects’ coefficients from the model for the Australian born against the ‘fixed effects’ coefficients for immigrants from English-speaking countries and non-English-speaking countries. The straight line AA in these figures is the simple regression of the coefficients on occupation for the Australian born on the coefficients for immigrants from English-speaking countries and non-English-speaking countries, respectively.⁶

It is clear from 1, 2 that relatively high-paying occupations for one birthplace group are generally also relatively high-paying occupations for the other birthplace groups. The correlation coefficient between the occupational fixed effects in Figure 1 is 0.716, while that for Figure 2 is 0.549.⁷ If the data points are weighted by the employment shares of the immigrant groups the correlation coefficients are 0.889 and 0.917, respectively. If the occupational employment shares of the Australian born are used then the correlation coefficients are somewhat lower, 0.796 and 0.840, respectively.

These comparisons of the occupational fixed effects suggest that they are so close that they are not likely to be the main contributor to the different pattern of results across birthplace groups in Table 2. Given these findings, the explanation for the differences in the estimates of the earnings equation between the Australian born and the foreign born when occupation is held constant needs to focus on the partial effects of the explanatory variables on occupational choice.

IV Occupational Attainment

This section presents estimates of a model of occupational attainment that assists in accounting for the pattern of effects reported in Table 2. A model in the tradition of the status attainment models of Nickell (1982) and Evans (1987) is employed. Hence, the analysis proposed is the estimation (using OLS) of a status attainment model:

where Occ_i is the mean occupational earnings of the Census occupational category (i.e. mean natural logarithm of hourly earnings in each of the 44 Census occupations) in which individual i works, X_i is a set of the individual's attributes that influences this occupational outcome, and ν_i 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.⁸

Table 4 contains two sets of OLS estimates of the status attainment model for each birthplace group. Specification (i) is based on mean occupational earnings for the nine major group occupations, while specification (ii) is for mean occupational earnings for the 44 Census occupational categories.

Table 4. Estimates of Model of Occupational Status, with the Mean Occupational Earnings as the Dependent Variable, by Birthplace, Men Aged 20–64, 2001 ^†

Variables	Australian born		Foreign born
			English-speaking countries		Non-English-speaking countries
	(i)‡	(ii)	(i)‡	(ii)	(i)‡	(ii)
Constant	2.245 (307.27)	2.054 (210.94)	2.269 (90.91)	2.202 (71.72)	2.412 (100.39)	2.299 (75.76)
Education	0.048 (97.25)	0.062 (94.10)	0.051 (38.67)	0.055 (33.85)	0.041 (35.14)	0.050 (34.49)
Experience	0.001 (3.59)	0.004 (7.58)	0.005 (3.13)	0.005 (2.86)	−0.005 (3.64)	−0.006 (3.81)
(Experience)2/100	0.003 (3.29)	−0.004 (3.39)	−0.006 (1.96)	−0.007 (2.13)	0.011 (4.74)	0.013 (4.58)
Married	0.017 (6.40)	0.022 (5.82)	0.036 (3.97)	0.054 (5.03)	0.010 (1.21)	0.017 (1.62)
Speaks English§
Very well	NE	NE	NE	NE	−0.013 (1.67)	−0.015 (1.44)
Well	NE	NE	NE	NE	−0.084 (9.33)	−0.105 (8.98)
Not well	NE	NE	NE	NE	−0.085 (7.55)	−0.123 (8.31)
Not at all	NE	NE	NE	NE	−0.032 (1.01)	−0.146 (3.71)
Year of arrival§
1991–1995	NE	NE	−0.011 (0.63)	−0.006 (0.32)	−0.016 (1.09)	0.000 (0.01)
1986–1990	NE	NE	−0.033 (2.41)	−0.029 (1.71)	−0.005 (0.36)	0.003 (0.19)
Before 1986	NE	NE	−0.009 (0.77)	−0.011 (0.83)	0.016 (1.22)	0.035 (2.25)
	0.321	0.278	0.281	0.253	0.330	0.321
Sample size	20 709	20 709	3127	3127	3752	3752

• Notes: ^† Heteroscedasticity-consistent ‘t’ statistics in parentheses; ^‡ Column (i) specification has the mean occupational earnings at the Major Group level as the dependent variable, column (ii) specification has the mean occupational earnings at the Census occupational classification as the dependent variable; ^§ The omitted category for the Speaks English variable is ‘Speaks only English’ and that for the Year of Arrival variable is ‘After 1995’; NE, Variables not entered.
• Source: Sourced from 2001 Australian Census of Population and Housing.

The estimates in Table 4 for the Australian born show that the main determinant of occupational status, as measured by the t-ratio and magnitude of the partial effect, is educational attainment. Each year of schooling is associated with entry into occupations having 5 (specification (i)) to 6 (specification (ii)) per cent higher earnings. In comparison, for men born in Australia, the effects on mean occupational earnings of labour market experience are very small, and sensitive to the level of aggregation of the occupational categories. Under either classification, the difference in mean occupational earnings between the least and most experienced workers is less than the effect of 2 years of schooling. Thus, among the native born, schooling is a far more important determinant of occupational attainment than is labour market experience. A similar pattern is found among the native born in the USA (Chiswick & Miller, 2007).

The results for immigrants from English-speaking countries are similar to the findings among the Australian born. While additional education apparently readily opens up access to higher-paying occupations, additional years of pre-immigration and post-immigration labour market experience are not associated with similar access.

Among the foreign born from non-English-speaking countries, years of education are also associated with higher mean occupational earnings, with the partial effect of 0.041–0.050 being around 1 percentage point less than that estimated for the other birthplace groups. This finding is consistent with the smaller partial effects of education on individual earnings among immigrants from non-English-speaking countries in Table 1. However, immigrants from non-English-speaking countries with moderate amounts of pre-immigration labour market experience have relatively low mean occupational earnings compared to those with even less pre-immigration experience. This is the same finding as in Chiswick and Miller (2007), on the basis of study of the US labour market. The finding is consistent with the increase in the payoff to labour market experience once occupation is held constant in the study of individual earnings (Tables 1 and 2). The effect of duration in Australia has a positive effect on mean occupational earnings, but this effect is also quite modest.

Finally, it is noted that proficiency in English is associated with substantial occupational advancement, though the estimated coefficients in Table 4 are only around one-half the magnitude of the effects found in the study of individual earnings (without the variables for occupation) in Table 1.

Findings similar to these are obtained when an ordered probit model is applied to the occupational data (ranked by mean earnings).⁹ In particular, the main determinant of membership in a higher-ranked occupation is educational attainment. The effects of labour market experience on occupational outcomes for immigrants from non-English-speaking countries is opposite those estimated for the Australian born and immigrants from English-speaking countries. English proficiency is a major determinant of the likelihood of being employed in a high-earnings occupation among immigrants from non-English-speaking countries.

V Conclusion

The analyses for Australia using the 2001 Census data on adult men in full-time employment reported in this paper show that when occupation is held constant in the earnings equation, there is a reduction of 41 per cent in the payoff to schooling for the Australian born, a similar reduction, of 43 per cent, in the payoff to schooling for immigrants from English-speaking developed countries, and an even greater reduction, by 66 per cent, in the payoff to schooling for immigrants from other countries. The latter have a much lower transferability of their skills to Australia.

At the same time, holding occupation constant is associated with a quite modest reduction, of less than 10 per cent, in the payoff to labour market experience for the Australian born. It is also associated with only a minor reduction, of around 16 per cent, in the payoff to pre-immigration experience for immigrants from English-speaking countries. However, controlling for occupation increases the payoff to pre-immigration labour market experience by 60–70 per cent for immigrants from non-English-speaking countries. These remarkable differences are due to a negative association between occupational status in Australia and pre-immigration labour market experience for many immigrants from non-English-speaking countries. This may arise if there is negative selectivity among those from low transferability countries who immigrate to Australia at an older age, which is after many years of work experience in their country of origin.

The comparison of the findings for the Australian labour market with the study by Chiswick and Miller (2007) for the US labour market revealed that inter-occupational earnings mobility is of greater importance in gaining a payoff to education in Australia than in the US. This is likely to be due to the more egalitarian distribution of earnings within occupations that is associated with the more centralised system of wage determination, and perhaps greater union power, in Australia than in the USA.

These findings suggest that attention needs to be focused on occupational outcomes at the time of labour market entry in the destination country. The different immigrant selection regimes of the US (emphasis on family reunion) and Australia (emphasis on skills) do not appear to matter in this regard. More fundamental labour market processes seem to be at work. The study of occupational attainment for subgroups of the population who may face different transitions (e.g. have access to well established networks or settle in areas with tight labour markets) may assist in understanding these processes.

Appendix A

Appendix B
Definitions of Variables

The variables used in the statistical analysis of the 2001 Australian Census of Population and Housing are defined below. The analyses are restricted to male full-time workers (i.e. working 35 hours or more per week) aged 20–64 years.