The Labour Market Adjustment of Immigrants in New Zealand Report
3. Data and Sample Characteristics
This paper uses unit record data from the 1997-2007 New Zealand Income Survey (NZIS). This is a departure from previous studies of immigrant adaptation in New Zealand, which have invariably used data from the five-yearly Census of Population and Dwellings. While there are certain advantages to using Census data, in particular the availability of large samples of immigrants and detailed country of birth information, there are two important limitations. First, since the Census only provides five-yearly snapshots of the populations, it requires strong assumptions to separately identify the impact of additional years in New Zealand on labour market outcomes from general macroeconomic and ageing effects. Second, the Census does not collect any information on hourly wage rates and thus these previous studies have been unable to examine wage adaptation.
Since 1997, the NZIS has been carried out by Statistics New Zealand (SNZ) each June quarter as a supplement to the Household Labour Force Survey (HLFS). Taken together, the two surveys collect data on household structure, the socio-demographic characteristics of household members, and labour force activity in the reference week and recent incomes for individuals at least 15 years old. The HLFS has a sample size of approximately 15,000 households and 28,000 adults. About 85% of these respondents also complete the NZIS. Sampling weights are calculated by SNZ to increase the representativeness of the HLFS, and are used in all analyses in this paper.
The HLFS collects information on how many years each individual has lived in New Zealand and aggregated country of birth. We restrict our analysis throughout to individuals aged 25-59 to exclude students and individuals nearing retirement. This provides a sample of nearly 185,650 observations. We drop a further 610 observations who are foreign-born and missing years in New Zealand and 865 observations who are missing other key covariates. For our descriptive statistics, we classify individuals as being either New Zealand-born, a recent migrant or an earlier migrant. Recent migrants are all individuals who have lived in New Zealand for less than 5 years and earlier migrants are all other individuals born in a foreign country. We also stratify all of our analysis by gender, given the large differences in labour market outcomes between men and women, particular for immigrants.
We examine four labour market outcomes throughout this paper. The first is employment, defined as whether an individual worked any hours in the last week for pay, was away from work but receiving accident compensation, or worked any unpaid hours for a family business. The second is the (log) real hourly wage rate for all workers, which is calculated by dividing the sum of actual income from wage/salary employment in the last week and actual self-employment income in the last year divided by 52, by actual total hours work in the last week. Because of dropping imputed records and the suppression of outliers, this measure is missing for roughly one-quarter of the employed population (as well as for all the non-employed). The implications of this are discussed when presenting the results.
The third labour market outcome is annual total income measured in brackets in the final survey question which reads, "I am going to read out a list of (thirteen) income groups, and I'd like you to tell me which of these groups covers your total income from all the kinds of income we have talked about. This is before tax and is for the 12 months ending today. But don't include irregular lump sum payments." These brackets are then assigned a continuous value by SNZ using distributional information for total income as measured in the separate Household Economic Survey. While there are obvious disadvantages to examining this outcome, it is the only annual measure of income in the NZIS and is the same question that is used in the Census, which allows us to directly compare our results to those in previous papers. The measure is also dropped for the roughly fifteen percent of the population with imputed NZIS records, but is available for non-working individuals.
Our final labour market outcome is a constructed continuous measure of occupational rank, as in Chiswick et al. (2005). We have access to information on each employed worker's current occupation at the two-digit NZSCO90 classification group level, which records twenty-six different occupations. For each of these occupations, we calculate the average real wage of New Zealand-born workers over the entire sample period, separately by gender. We then assign these values to each New Zealand-born and immigrant worker based on their gender and occupation. This method ranks occupations in a continuous metric that has the same explicit ordering for immigrants and the New Zealand-born and can be examined using the same framework that is used to look at the other labour market outcomes. This measure is available for individuals with imputed records in the NZIS since the occupational information comes from the HLFS, but is unavailable for people who are not currently employed.
Table 1 presents the demographic characteristics of the three nativity groups (recent migrants, earlier migrants, New Zealand-born) stratified by gender. Our analysis sample consists of 68,526 New Zealand-born men, 4,461 male recent migrants, 13,313 male earlier migrants, 77,659 New Zealand-born women, 5,188 female recent migrants, and 15,015 female earlier migrants. Immigrants increased from 18 percent of the overall population in 1997 to 25 percent of the overall population in 2007. As in most countries, recent migrants are younger than the non-immigrant population. But, unlike the United States where most immigrants are low-skilled, in New Zealand, recent migrants are more highly qualified than the New Zealand-born, with 41 percent of male recent migrants and 36 percent of female recent migrants having university degrees compared with only 14 percent of the New Zealand-born men and 13 percent of New Zealand-born women. This is reflected throughout the qualification distribution, with fewer migrants having no qualifications compared to the New Zealand-born. This is not surprising given that, as discussed above, New Zealand operates a structured immigration system that focuses mainly on higher-skilled migrants.
There are also notable differences in other characteristics. Unsurprisingly, the ethnic distribution of migrants differs a great deal from that of the New Zealand-born. Only 41 (38) percent of male (female) recent migrants and 56 (53) percent of male (female) earlier migrants classify themselves as European compared with 89 (88) percent of New Zealand-born males (females). In fact, almost the entire non-European and non-Māori population is foreign-born (and hence we do not control for ethnicity when examining differences in outcomes between migrants and the New Zealand-born in a regression framework). Immigrants are more likely to be married than the New Zealand-born and recent immigrants are less likely to be divorced/separated/widowed. Interestingly, earlier migrants are as likely or more likely than the New Zealand-born to be in this category. Similarly, immigrants are more likely to live in a household classified as 'couple with children' than the New Zealand-born. There are large differences in settlement location of migrants compared to the New Zealand-born. For example, 95 percent of recent migrants and 92 percent of earlier migrants live in urban areas compared with 84 percent of the New Zealand-born.
Table 1 also presents the labour market outcomes for the three nativity groups stratified by gender. Employment rates are much lower among recent migrants compared to both earlier migrants and the New Zealand-born, confirming earlier NZ findings by Winkelmann and Winkelmann (1998a), and Boyd (2006) . For example, only 78 percent of male recent migrants and 54 percent of female recent migrants are employed compared with 86 percent of male earlier migrants, 68 percent of female recent migrants, 89 percent of New Zealand-born males and 73 percent of New Zealand-born females. Wage variation across the nativity groups is much smaller, with male recent migrants having an average wage of $23 per hour in 2007 dollars compared with $24 per hour for male earlier migrants and New Zealand-born and female recent migrants having an average wage of $19 per hour versus $21 per hour for female earlier migrants and New Zealand-born. Male immigrants work in occupations than pay on, average, $1 more per hour than the occupations in which New Zealand-born males are working, while female immigrants work, on average, in the same occupations as New Zealand-born females.
However, it is worth nothing that, based on differences in qualifications, we might expect migrants, to have higher wages and be working in higher paid occupations than the New Zealand-born, and this is why a regression analysis is needed to make a proper comparison. The large differences in employment rates, together with possible differences in hours of work, translate to large differences in annual incomes between recent migrants and the other nativity groups. For example, the average recent male migrant earns 40 thousand dollars per annum, while the average earlier male migrant earns 47 thousand per annum, and the average New Zealand-born male earns 48 thousand per annum. The same figures for women are 21, 27 and 28 thousand dollars, respectively.
Finally, Table 1 presents information on immigrant-specific characteristics. On average, earlier migrants have lived in New Zealand for 20 years and were aged 23 when they arrived. Among this group, 32 percent of men and 30 percent of women arrived prior to age eighteen, and thus are likely to have done some of their formal education in New Zealand. Among recent migrants, the average age is 35. The difference in the average arrival age between earlier and recent migrants is partially mechanical since recent migrants who were less than 21 years-olds at arrival are excluded from our sample since the lower age cut-off is 25. In our empirical analyses, we group the immigrant population into six arrival cohorts: before 1958; 1958-67; 1967-78; 1978-87; 1988-97; 1998-2007 to control for differences in the quality of migrants coming to New Zealand over time.
The source region distribution of recent immigrants differs from that of earlier migrants in a way that reflects the movement away from traditional source country preferences in 1987. For example, 36 (32) percent of male (female) earlier migrants were born in the United Kingdom compared with only 21 (18) percent of male (female) recent migrants. Similarly, 22 (23) percent of male (female) earlier migrants were born in the Pacific Islands versus only 11 (11) percent of male (female) recent migrants. Conversely, recent migrants are much more likely to have been born in Asian countries, with 28 (30) percent of male (female) recent migrants born in Asia versus only 14 (16) percent of male (female) earlier migrants.
Table 2 presents the same characteristics stratified by gender and region of birth (ie. New Zealand-born, Australia, United Kingdom, Pacific, Asia, Other). Pooling recent and earlier immigrants, the average age of immigrants is quite similar to that of the New Zealand-born, except for immigrants born in the United Kingdom, who are on average 3 years older than New Zealanders, and immigrants born in Asia, who are on average 2 years younger than New Zealanders. On the other hand, there is a large variation in the qualification distribution for migrants from different sources countries. Only 8 (6) percent of male (female) migrants from the Pacific Islands have university degrees versus 49 (38) percent of male (female) migrants from Asia. These differences are largely related to the different immigration categories under which individuals from different countries are migrating (mainly family versus skilled migration). The changing mix of source-countries over time is also clearly evident in the average years since arrival, which is only 8 years for immigrants from Asian countries, and 22 years for immigrants from the United Kingdom. Asian immigrant men and women first arrived at older ages than did other immigrant groups, with an average of 31 years of age, compared with a range of 22 to 25 years of age for immigrants from the United Kingdom, Pacific Islands and Australia.
Outcomes also vary across different groups of migrants defined by region of birth or qualifications. Asian immigrants have the lowest employment rate of all the region-of-birth groups shown, with only 73 percent of men and 52 percent of women being employed, compared with a maximum of 91 percent for United Kingdom men and 75 percent for United Kingdom women. Immigrants from two of the regions, Asia and the Pacific, earn hourly wage rates that are on average lower than those for the New Zealand-born. For Pacific immigrants, some of this difference is associated with their lower qualifications levels, whereas this is not so for Asian immigrants, whose higher qualifications would be expected to lead to a wage premium. Similarly, while Pacific Islanders are found to work in lower paying occupations than both other immigrants and New Zealanders, Asians work, on average, in higher (for men) or similar (for women) paying occupations as the New Zealand-born. Real annual income differences reflect the employment and wage variation, and also capture differences in hours of work over the year. In accordance with the comparatively low employment and wage rates for Asian and Pacific immigrants, these groups have substantially lower mean annual incomes.
 Unfortunately, neither the Census nor the NZIS/HLFS collect immigrant specific data, such as citizenship status or visa category upon entry to New Zealand.
 Wage and income data are imputed for all HLFS sample members who fail to complete the NZIS. Individuals with imputed data are dropped when examining wage rates and annual incomes because, as discussed in Hirsch and Schumacher (2004), including imputed data leads to biased estimates of mean differences between groups when the attribute being studied (here, migration status) is not a criterion used in the imputation procedure.
 There are eight possible choices which were the most common immigrant countries in 1986 when the HLFS was started. These can be aggregated up to four meaningful groups, Australia, United Kingdom, Pacific Islands, and Asia, and a residual category for all other foreign-born individuals. Based on figures from the 2006 Census, the rough breakdown of the residual category is 40% non-United Kingdom Europe, 40% Africa and the Middle East (mainly South Africa) and 20% Americas (mainly the United States and Canada).
 Individuals reporting real wages less than $4 or greater than $150 are recoded to missing along with all individuals with imputed data. These thresholds are approximately the real youth minimum wage at the start of our sample period and the 99.5 percentile of the wage distribution. This mainly has the effect of dropping individuals with negative self-employment income and thus negative wages and a few observation with unrealistically high wage rates (ie. over $1000 per hour). This recoding effects 4-5% of workers in each gender and migrant group. Overall, for men, 9-10 percent of workers are either missing wage data or have wages that are outside the valid range and a further 17-19 percent have imputed data. For women, 10-13 percent of workers are either missing wage data or have wages that are outside the valid range and a further 12-14 percent have imputed data. There is little difference in the percentage of workers with valid wage data across migrant groups; for men, 74% of employed New Zealand-born, 73% of employed earlier migrants and 72% of employed recent migrants have valid wage data while for women the numbers are 75%, 74% and 76%, respectively.
 Because the NZIS only asks how many years each individual has lived in New Zealand and not their year of first arrival, immigrants who have not lived continuously in New Zealand since first arriving will be assigned to a more recent arrival cohort than their true arrival cohort.