SUMMARY FINDINGS

Given the above background, context and headline analysis, we move to a more detailed investigation of labour market outcomes and the behaviour of migrants. The following section uses census data for the 1981, 1996, 2001 and 2006 years. The main requirement is to create models that can measure the cross-section profiles at each census, and the inter-census changes, where the census data are consistent. The models are also required to be able to create tables of the labour market outcomes of migrants with specified characteristics.

The characteristics of immigrants investigated using data from these four censuses relate to four main labour market outcomes: income levels, sources of income, labour force status and occupations.

In this section, we outline suggested propositions that motivate our investigation and define the terms and variables that are used in the discussion. Thereafter, we provide a summary of the findings of this investigation.

In particular, this investigation has two parts. One part makes observations from numerous cross-tabulations of the data. Of course, not all cross-tabulations can be covered in this report. However, a spreadsheet tool has been developed to enable desired comparisons to be easily generated and depicted. A selection of these comparisons is provided in sections 5-8, with a summary of findings provided in sub-section 4.4.

Another part undertakes a formal diagnosis of the migrant- and non-migrant-related influences determining labour market outcomes.^[8] Note that this diagnosis is by no means comprehensive. It is derived from tabulated data from each census, as opposed to unit record data. It is used to both complement and supplement the observations from the numerous data cross-tabulations. It also forms a useful platform for future comprehensive investigations of the various relationships. The appendix section 10 details this examination, with a summary provided in sub-section 4.5.

Suggested propositions

From the overall project objective, and noting the above background, the following propositions suggest themselves:

• Default proposition: Labour market outcomes for migrants to New Zealand are not significantly different from those for comparable New Zealand born people.
• Additional proposition A: Labour market outcomes for migrants to New Zealand are initially different from the comparable New Zealand born, but these differences become insignificant the longer the migrants remain in New Zealand.
• Additional proposition B: Labour market outcomes in New Zealand for migrant groups, according to their place of birth, are significantly different from each other. Consequently, labour market outcomes in New Zealand for some migrant groups are significantly different from the New Zealand born population.
• Additional proposition C: Labour market outcomes in New Zealand for some migrant groups, according to their place of birth, are significantly different from the New Zealand born population, but these differences become insignificant the longer the migrants remain in New Zealand.
• Additional proposition D: Labour market outcomes for migrants to New Zealand have been different from the New Zealand born population in the past, but these differences are progressively becoming less significant.

Diagnostic statistics/variables of interest

With the formulation of propositions, the next step in the research method is to establish variables or measures that can be investigated. Aligned to this task, it is necessary to note the distinction between 'outcomes' and 'characteristics'. For this purpose, the categories of labour market outcomes identified are:

• income
• occupation
• labour force status (for example, employer, self-employed, employed, unemployed, not in labour force)
• source of income (for example, wages, benefits, no income).

In turn, the categories of characteristics for labour market participants identified are:

• where born
• years since first arrival in New Zealand
• age
• highest qualification.

Within this characteristics and outcomes framework, variables and/or measures of interest can be identified. However, it should be noted that not all labour market outcomes observable from census data are investigated. Given the limitations of this project, we have focused our assessment on what we consider are the primary drivers of labour market outcomes. As a result, hours worked, industry of employment and job location are not assessed. As for characteristics, we have omitted ethnicity, language and household or family composition from our analysis. Thus, it has to be acknowledged that we have not undertaken a comprehensive investigation of differences in labour market outcomes - just the contribution of the factors listed.

It is pertinent to note the particular importance of the age characteristic. More specifically, the inclusion of full fee-paying international students in the data relating to the migrant population can result in misleading observations regarding labour market outcomes for particular population sub-groups. To redress this concern, we repeated our investigation for the population sub-group aged 25-54. This age group would limit the influence of full fee-paying international students on the data and so would provide a population group comparable to the New Zealand born 25-54 age group.

Further, more detailed analysis using micro-level data is going to be undertaken in a 2007-2012 Foundation for Research, Science and Technology-funded research programme carried out by Waikato University and Massey University.

1991 Census year data was not used because it does not contain information on the arrival date of the migrant population.

Definitions and classifications

1. Unless otherwise stated, population groups are restricted to those aged 15 and over.
2. Migrants are defined as those with non-New Zealand places of birth. Length of residence subsets of the migrant population are defined as:
   1. 'recent' migrants - those with less than 5 years residence in New Zealand
   2. 'intermediate' migrants - those with 5-14 years residence in New Zealand
   3. 'earlier' migrants - those with at least 15 years residence in New Zealand.
3. Region of birthplace sub-sets of the migrant population are defined as Australia, Pacific Islands, United Kingdom and Ireland, Europe and North America, Asia and Other.
4. Income is defined as annual, personal, pre-tax income from all sources.
5. Data for highest qualification were grouped as listed in Table 4.1.
6. Occupation data were grouped as listed in Table 4.2.

Table 4.1 Grouping of census highest qualification data

Table 4.2 Grouping of census occupation data

Census data cross-tabulations

According to the measures for income investigated, the observations clearly suggest migrant labour market outcomes are different from those of the New Zealand born. This observation holds, even after adjusting for the qualification characteristics of labour market participants.

However, this conclusion is not supported if the length of residence in New Zealand characteristic of each migrant group is captured. If adjustment for this effect is incorporated, the income measures of earlier migrants are not dissimilar to those of the comparable New Zealand born population group. Conversely, income measures for recent migrants clearly diverge from those of the New Zealand born.

As to sources of income, the observations again suggest migrant labour market outcomes are different from those of the New Zealand born. In general, the Asian and Pacific Islands born groups show lower proportions receiving wage and salary income and higher proportions reporting no source of income. Other migrant groups by birthplace report proportions similar to those for the New Zealand born.

After adjusting for age and years of residence characteristics of labour market participants, the outcomes for the Asian born group move closer to those for the New Zealand born population, as their length of residence in New Zealand increases. This leaves the conclusion that the labour market outcomes investigated by these income source measures are noticeably different for those born in the Pacific Islands, compared to the New Zealand born population.

Participation rate outcomes for migrant groups appear to be noticeably lower than those for the New Zealand born population. While this difference is not apparent for the intermediate migrant population group, the difference does reappear for the earlier migrant group.

The prevalence of self-employed is relatively high for migrant groups, compared to the New Zealand born population. The Europe and North American migrant group is a particularly noticeable feature in this observation. Furthermore, this migrant group also leads in the prevalence of employers. Other migrant groups, however, feature less so in this measure.

In terms of occupations, relatively higher proportions of earlier migrants are found to be in professional occupations. Conversely, the prevalence of migrants in trades or technical occupations is lower than for the New Zealand born population group.

Multivariate analysis

Isolating the impacts of the various characteristics on labour market outcomes is challenging. We also seek to identify the key explanatory influences through a series of simple multivariate regressions. In particular, we use census data to estimate equations to explain variations in the following measures of labour market outcomes.

Incomes

1. The percentage of population sub-groups that have income above the 70^th percentile of the national income distribution (termed 'high-income' for short-hand purposes).
2. The percentage of population sub-groups that have income below the 30^th percentile of the national income distribution (termed 'low-income' for short-hand purposes).

Workforce status

3. The labour force participation rate (%) of population sub-groups.
4. The unemployment rate (%) of population sub-groups.

Occupations

5. The percentage of population sub-groups that are employed in professional occupations.
6. The percentage of population sub-groups that are employed in associate professional or technician occupations.

Note further, attempts to estimate equations for the proportion in each sub-group that are employers and self-employed were unsuccessful.^[9] In addition, attempts to estimate an equation to explain the proportion of the population sub-groups employed in trades occupations also proved unsuccessful.

As noted earlier, only the 1996-2006 census data information was used here.

Methodology

Given the definitions stated in section 4.3, we identify 19 different groups in the population, i.e. six different non-New Zealand birthplaces, each with three sub-groups according to years in New Zealand, plus one New Zealand born group. Further, for each of these 19 groups, we have three census year observations. This provides us with a dataset with 57 observations as a basis for our investigation.

The size of each of these population sub-groups in 2006 is listed in Table 4.3. Remember that, throughout this discussion, the population sub-groups referred to are restricted to those aged 15 years and over. Data for those with unknown birthplace and/or years in New Zealand were omitted from the analysis. This related to some 185,000 people, or just under 6% of the 3.16 million in the total New Zealand population of those aged 15 years and over.

Table 4.3 The 2006 New Zealand population (aged 15 years and over)

For each of the six labour market outcomes investigated, we followed a similar process to construct an appropriate equation.

Our first step was to identify the presence, if any, of relationships excluding the migrant-specific characteristics. That is, we firstly established the influence (significance) of age composition and/or qualifications possessed by the sub-groups.

Thereafter, we introduced and tested the significance of census year identifiers. This was aimed at determining the importance of changes over the 1996-2006 period in the labour market outcomes across the different groups. At this stage of the process, we would have an appropriate explanation of the labour market outcomes, according to the non-migrant-related characteristics of each of the groups. Of course, such an equation would only successfully explain a proportion of the total variance in outcomes across each group.

The next step was the introduction of migrant-specific characteristics, i.e. birthplace (including New Zealand) and years in New Zealand. The significance or otherwise of identifiers for each of these characteristics enabled us to ascertain whether such migrant-related characteristics significantly improved our explanation of the labour market outcomes. That is, to what degree are labour market outcomes a function of migrant-related characteristics? And, conversely, to what degree are they a function of general population-wide factors?

Findings

This analysis found that migrant-related characteristics improved the explanation of labour market outcomes in only a selection of cases. Furthermore, in these cases, there was a relatively small degree of improvement in the explanations.

In most cases, the large majority of the variation in labour market outcomes across the various sub-groups of the population could be explained by the non-migrant-related characteristics, i.e. by differences in age composition and highest qualifications possessed. After controlling for these factors, the differences in labour market outcomes between and within migrant sub-groups (and compared to the New Zealand born sub-group) was relatively small.

Amongst the migrant-related factors that are significant, the main one was the years in New Zealand variable. In particular, the identifier for recent migrants^[10] was significant in indicating inferior labour market outcomes for this sub-group in some of the equations constructed. Notably, though, identifiers for intermediate and earlier migrants were insignificant. Recall that these categories are migrants with 5-15 years in New Zealand and more than 15 years in New Zealand respectively.

As to regions of origin, the identifiers for the separate birthplaces (including New Zealand) were also insignificant^[11] in most equations.

Propositions and conclusions

Note that, in sub-section 4.1, we proposed this default proposition:

Labour market outcomes for migrants to New Zealand are not significantly different from the New Zealand born population.

It is clear that both the observations from the cross-tabulations of data and the multivariate analysis do not support this proposition. However, the degree to which migrants' labour market outcomes are significantly different from the New Zealand born vary from measure to measure.

In particular, the large majority of the variation in labour market outcomes across the various sub-groups of the population can be explained by the non-migrant-related characteristics. Differences in age composition and highest qualifications possessed are the most important.

The migrant-related factor that remains important is the number of years in New Zealand. It is clear that there is a selection of inferior labour market outcomes for migrants with less than five years in New Zealand. In addition, the birthplace characteristic also appears as a factor for a few outcome measures, but the effect of birthplace is by no means uniform or widespread.

Consequently, we would conclude by arguing the most appropriate of the alternative propositions suggested is that labelled C:

Labour market outcomes in New Zealand for some migrant groups, according to their place of birth, are significantly different from the New Zealand born population, but these differences become smaller the longer the migrants remain in New Zealand.

---

^[8] Note that, due to difficulties in matching classifications, this formal diagnosis did not use the 1981 Census data. In particular, qualification and occupation classifications could not be satisfactorily matched, even at relatively highly aggregated levels.

^[9] That is, we were unable to obtain an equation where at least one of the explanatory variables available to us from our dataset was statistically significant.

^[10] That is, those with less than five years in New Zealand.

^[11] Or have relatively weak explanatory power.