APPENDIX: MULTIVARIATE ANALYSIS

Relationships determining income

As noted earlier, we use the term 'high-income' for short-hand purposes as a label representing income above the 70^th percentile of the national income distribution. Similarly, we use the label 'low-income' for income below the 30^th percentile of the national income distribution.

10.1.1 Determining high-income earners

Table 10.1 lists the proportion of each of the 19 sub-groups that were high-income earners. This proportion ranges from a high of 47.6% for recent migrants born in the United Kingdom and Ireland to a low of 11.2% for recent migrants born in Asia. The proportion across all recent migrants was 24.2%, while the figure for all migrants (i.e. the total overseas born population) was 27.4%. The comparable number for the New Zealand born sub-group in 2006 was 31.0%.

Table 10.1 Percentage of high-income earners in each sub-group 2006

Birthplace	Years in New Zealand
	<5	5 to 15	>=15	Total
New Zealand				31.0
Total population				30.0
Australia	39.1	27.5	31.1	31.6
Pacific Islands	12.1	16.9	22.6	19.0
UK & Ireland	47.6	46.3	31.9	36.2
Europe & N. America	31.5	35.8	30.0	31.6
Asia	11.2	17.9	28.7	17.0
Other	32.7	37.1	40.9	35.7
Total overseas born	24.2	27.1	29.9	27.4

In explaining these variances, together with those in the 1996 and 2001 years, we found only one significant migrant-related characteristic.

Conversely, of most significance in determining the proportion of high-income earners was the highest qualification held within each population group. In particular, the proportions with a vocational or degree qualification as their highest qualification were most significant in this determination. As expected, both were positively related to the proportion of high-income earners. Figure 10.1 illustrates one of these relationships - that between high-income earners and the proportion with their highest qualification being a vocational qualification.

Figure 10.1 High-income earners and vocational qualifications

Data table for Figure 10.1

We report that estimated relationships including age-group variables were inferior to the qualification variables in determining the proportion of high-income earners. Including both qualification and age-group variables together also resulted in inferior relationships.

The two qualification variables (degree and vocational) together explained nearly 75% of the variation in the proportion of high-income earners.

In addition, the insignificance of the census year identifiers, when introduced, confirmed that the relationship determining high-income earners has not significantly altered over the 1996-2006 period.

The one migrant-related characteristic that did enter the equation was the identifier for recent migrants. We found that this group had a significantly lower proportion of high-income earners than for other groups. This conclusion held after controlling for the different qualifications. While significant, though, this variable contributed only another 3 percentage points to the power of the explanation.

On the other hand, both intermediate and earlier migrants had no significantly different proportions of high-income earners than the overall population. Furthermore, introducing specific places of birth also had no significant impact on the estimation of high-income earners in the population. This included the New Zealand born identifier. Hence, we conclude that the determination of high-income earners was not significantly related to birthplace over the 1996-2006 period.

Help displaying mathmatical formuale

10.1.2 Determining low-income earners

Table 10.2 lists the proportion of each of the 19 sub-groups in 2006 that were low-income earners. This proportion ranges from a low of 22.7% for earlier migrants born in the Other region to a high of 59.3% for recent migrants born in Asia. The proportion across all recent migrants was 44.7%, while the figure for all migrants (i.e. the total overseas born population) was 35.5%. The comparable number for the New Zealand born sub-group in 2006 was 28.0%.

Table 10.2 Percentage of low-income earners in each sub-group 2006

Birthplace	Years in New Zealand
	<5	5 to 15	>=15	Total
New Zealand				28.0
Total population				30.0
Australia	28.4	42.1	26.6	30.2
Pacific Islands	48.1	39.6	28.9	35.1
UK & Ireland	23.5	24.7	25.9	25.5
Europe & N. America	34.5	33.2	28.0	31.0
Asia	59.3	49.2	30.7	50.1
Other	35.9	35.0	22.7	33.5
Total overseas born	44.7	40.2	27.2	35.5

Similar to the findings for high-income earners, we found few significant migrant-related determining factors for low-income earners. In this case, the primary determining factors were those with only school qualifications and the proportion of the group aged 15-24 years old.

As expected, the signs on each of these coefficients were positive. This indicates that groups with only school qualifications would have a larger proportion of low-income earners in their population. In addition, groups with a larger proportion of youth (15-24 years old) will also have a larger proportion of low-income earners. This latter relationship is illustrated in Figure 10.2.

We note that the 15-24 year old age-group variable was a better predictor than the proportion with no qualifications in this relationship.

Figure 10.2 Low-income earners and those aged 15-24 years old

Data table for Figure 10.2

Notably, again, the insignificance of census year identifiers, when introduced, suggested that the determination of the proportion of low-income earners in a group has not altered over the 1996-2006 period.

The proportion of low-income earners can be nearly 58% explained by the qualification and age-related variables.

Moving to migrant-related characteristics, the identifiers that were significant when introduced individually were:

• recent migrants (with a positive coefficient)
• intermediate migrants (with a negative coefficient)
• Australia birthplace (with a negative coefficient)
• Pacific Islands birthplace (with a positive coefficient)
• Asia birthplace (with a positive coefficient).

However, when introduced together, the association between some of these factors led to inferior or insignificant estimates for some of the variables. We also further note that the New Zealand born identifier was insignificant in determining low-income earners.

Of the above list, only the identifiers for recent migrants and those born in Asia remained significant in the combined equation. Both the coefficients in this case were positive. This suggests that the higher the proportion of recent migrants, the higher the proportion of low-income earners in that group. Similarly, the higher the proportion of those born in Asia, the higher the proportion of low-income earners in that group.

Again, though, the significance of the migrant-specific characteristics is relatively low. These add 13 percentage points to the explanation of low-income earners - resulting in an overall explanation of 71%.

Hence, we conclude that the determination of low-income earners was predominantly related to age composition and qualifications over the 1996-2006 period. However, there were relatively small but significant differences associated with recent migrants and migrants born in Asia.

Relationships determining workforce status

The determinants of labour force participation and unemployment rates were investigated. We report that efforts to explain proportions in each group that are employers and proportions that are self-employed were unsuccessful.

10.2.1 Determining labour force participation

Table 10.3 lists the labour force participation rate in 2006 of each of the 19 sub-groups. We note the highest participation rate was the 79.3% for recent migrants born in the United Kingdom and Ireland, while the lowest was the 56.3% for recent migrants born in Asia. Across all recent migrants the participation rate was 65.7%, while the figure for all migrants (i.e. the total overseas born population) was 63.6%. The comparable number for the New Zealand born sub-group in 2006 was 70.6%.

Table 10.3 Labour force participation rate (%) for each sub-group 2006

Birthplace	Years in New Zealand
	<5	5 to 15	>=15	Total
New Zealand				70.6
Total population				68.5
Australia	78.8	74.6	68.7	71.3
Pacific Islands	62.6	65.0	66.5	64.9
UK & Ireland	79.3	79.0	55.8	62.2
Europe & N. America	72.9	75.3	55.5	63.9
Asia	56.3	60.8	67.1	59.6
Other	73.1	74.7	72.5	73.2
Total overseas born	65.7	68.2	60.6	63.6

As for the income determination exercise reported in the previous section, a majority of the variation in the labour force participation rate across different groups can be explained by qualification and age-related variables.

In particular, the proportion of the population group in the 25-54 year old age group has a positive impact on the labour force participation rate of that group. Similarly, there is a positive impact from the proportion possessing vocational qualifications as their highest qualification. Together, these two variables explain 53%.

Unsurprisingly, noting the economy-wide surge in participation rates over the last few years, the census year identifier for 2006, when introduced, was a significant factor. Including this variable enabled 59% of the variation in participation rates across the different groups to be explained.

The migrant-related characteristics that were also significant when introduced individually were:

• recent migrants (with a negative coefficient)
• intermediate migrants (with a positive coefficient)
• Australia birthplace (with a positive coefficient)
• United Kingdom and Ireland birthplace (with a negative coefficient).

However, when introduced together, the association between some of these factors led to inferior or insignificant estimates for some of the variables.

Consequently, retaining the recent migrant and Australia birthplace identifiers (as they remained significant and together produced a superior estimate) resulted in some 81% of the variation in labour force participation rates to be explained, that is, an additional 22 percentage points is explained by the introduction of the migrant-related characteristics.

At this stage, adding the New Zealand born identifier is also appropriate, with a significant positive coefficient. However, its significance is marginal, adding another 2 percentage points to the overall explanation.

Hence, we conclude that, after controlling for qualifications and age, certain migrant-related characteristics are significant in the determination of labour force participation rates. In particular, both Australia and New Zealand born sub-groups have significantly higher participation rates than other sub-groups. In contrast, recent migrants have significantly lower participation rates.

However, these migrant-related factors account for less than a quarter of the overall variation in participation rates over the 1996-2006 period. On the other hand, nearly 60% can be explained by the combination of qualification and age-related variables and the fact that 2006 participation rates are higher than in previous years across all groups.

10.2.2 Determining the unemployment rate

Table 10.4 lists the unemployment rate in 2006 of each of the 19 sub-groups. We note the lowest rate was the 2.3% for earlier migrants born in the United Kingdom and Ireland, while the highest was the 11.6% for recent migrants born in Asia. Across all recent migrants, the unemployment rate was 8.6%, while the figure for all migrants (i.e. the total overseas born population) was 6.0%. The comparable number for the New Zealand born sub-group in 2006 was 4.8%.

Table 10.4 Unemployment rate (%) for each sub-group 2006

Birthplace	Years in New Zealand
	<5	5 to 15	>=15	Total
New Zealand				4.8
Total population				5.1
Australia	5.2	7.5	4.4	5.3
Pacific Islands	10.9	8.8	6.1	7.8
UK & Ireland	4.5	3.2	2.3	3.0
Europe & N. America	7.0	5.1	2.7	4.7
Asia	11.6	8.2	4.1	8.9
Other	7.4	6.6	3.3	6.5
Total overseas born	8.6	6.8	3.6	6.0

The non-migrant-related variables successful in explaining the rate of unemployment within each population group were age-related, along with the identifier for the 2006 census year. The two age-related variables were the proportion aged 15-24 and the proportion aged 24-54. Both of these have positive coefficients. This reflects the fact that these age groups are more likely to be participating in the labour force and, so, more likely to be unemployed. The 2006 census year identifier has a negative coefficient. The significance of this variable and its sign is unsurprising given the reduction in the economy-wide unemployment rate over the latter-half of the 1996-2006 period.

These three variables together explain 45% of the variation in unemployment rates across the different groups.

As for the migrant-related variables, when introduced individually, we particularly note that New Zealand born identifier was insignificant in determining the unemployment rate. Further, the identifiers for intermediate and earlier migrants were also insignificant. Of the other migrant-related variables, those found to be significant were:

• recent migrants (with a positive coefficient)
• intermediate migrants (with a negative coefficient)
• Australia birthplace* (with a negative coefficient, but see note below)
• Pacific Islands birthplace* (with a positive coefficient, but see note below).

(* It should be noted that the introduction of the identifiers marked with an asterisk did significantly erode the robustness of the equation to make it unsatisfactory.)

Each of these variables added in the range of 7-11 percentage points to the explanatory power of the estimated equation. This took the total explained by the equations to 52% to 56%.

However, attempts to include all these variables (or sub-sets thereof) together into one equation were not successful, that is, this process resulted in the relationships between the variables being too difficult to separately identify. Consequently, the robustness of the estimated equation including these migrant-related variables was significantly eroded.^[16]

Thus, we conclude that there remain many factors that determine the rate of unemployment across various sub-groups of a population. It is clear that the relative age composition of a group is a significant factor in its unemployment rate. In addition, unemployment rates within the groups in 2006 are significantly different to those in earlier census years. Of the many other variables likely to influence groups' unemployment rates, there is a selection of migrant-related factors. However, the determination of unemployment rates appears complex, and isolating the effects of individual migrant factors (and other influences) was not successful with the limited range of variables investigated by this study.

Relationships determining occupations of sub-groups

The determinants of the proportions of each group employed as professionals, and associate professionals and technicians were investigated. However, we report that efforts to explain proportions in each group that are employed in trades occupations were unsuccessful.

Employment in professional, and associate professional and technician occupations recorded large changes over recent years. Indeed,

Table 10.5 confirms that the number of migrants in professional occupations grew by more than 50,000 over the past five years. That is, migrants accounted for more than 40% of the total change in this occupation group between 2001 and 2006.

This observation is even more striking when looking at the associate professional and technician category. As listed in Table 10.6, the number of migrants employed in these occupations rose by more than 22,600 in the last five years. As a result, migrants accounted for more than half of the total expansion in employment in this category over the 2001-2006 period.

For this category, we particularly note the difference between the 1996-2001 experience and that between 2001 and 2006. The contraction over the earlier five-year period resulted in mixed outcomes across the different birthplace groups. However, the dramatic expansion over the latter five-year period appears to have been spread over all sub-groups.

Determining professional occupations

Table 10.7 lists the proportion in 2006 of each of the 19 sub-groups that were employed in professional occupations. This rate was highest for recent migrants born in the United Kingdom and Ireland at 33.9% and lowest for recent migrants born in the Pacific Islands, at 7.4%. Across all recent migrants, 18.5% were employed in professional occupations, while the figure for all migrants (i.e. the total overseas born population) was 19.6%. The comparable number for the New Zealand born sub-group in 2006 was 18.8%.

Table 10.7 Percentage of each sub-group in professional occupations 2006

Birthplace	Years in New Zealand
	<5	5 to 15	>=15	Total
New Zealand				18.8
Total population				18.2
Australia	29.4	20.6	21.3	22.1
Pacific Islands	7.4	9.4	11.2	9.6
UK & Ireland	33.9	33.6	20.8	24.2
Europe & N. America	25.0	29.1	20.2	22.8
Asia	11.9	19.0	23.7	16.4
Other	21.6	29.9	32.8	26.4
Total overseas born	18.5	22.4	19.7	19.6

Of the non-migrant-related variables, the proportion of a group that have vocational or degree qualifications are most significant in determining the proportion of a group that are employed in professional occupations. Indeed, these two variables are successful in explaining nearly 81% of the variation in employment in professional occupations. As expected, both coefficients are positive, indicating the higher the proportion of a group that have such qualifications, then the higher the proportion employed in professional occupations. As an example, the relationship between proportions of a group with degree qualifications and proportions in professional occupations is illustrated in Figure 10.3.

Figure 10.3 Professional occupations and degree qualifications

Data table for Figure 10.3

We note that none of the census year identifiers were found to be significant. This implies that the determination of proportions in professional occupations has remained unchanged over the 1996-2006 period. Age-related factors were also not significant here.

As for migrant-related variables, all except one, on introduction, were found to be insignificant in adding further to the explanation. The one exception was, again, the identifier for recent migrants. The negative coefficient here means that, controlled for qualification, recent migrants have a lower proportion employed in professional occupations. The introduction of this variable adds a further 5% to the explanatory power of the equation.

We particularly note the insignificance of all the migrant birthplace identifiers, including New Zealand born, in determining the percentage of those in professional occupations. In addition, the intermediate and earlier migrant identifiers were also insignificant.

Hence, we conclude that the determination of the percentage of those in professional occupations was dominated by qualifications characteristics and not significantly related to birthplace, over the 1996-2006 period. We further note, the similarity of this conclusion (and the explanators included in the associated equation) with that stated earlier for high-income earners.

10.3.2 Determining associate professional and technicians occupations

Table 10.8 lists the proportion in 2006 of each of the 19 sub-groups that were employed in associate professional and technician occupations. This rate was highest for recent migrants born in Australia at 13.8%, while the lowest was the 6.1% for intermediate migrants born in Pacific Islands. Across all recent migrants, the rate was 9.7%, while the figure for all migrants (i.e. the total overseas born population) was 9.6%. The comparable number for the New Zealand born sub-group in 2006 was 13.1%.

Table 10.8 Percentage of each sub-group in associate professional and technician occupations 2006

Birthplace	Years in New Zealand
	<5	5 to 15	>=15	Total
New Zealand				13.1
Total population				11.7
Australia	13.8	12.1	12.0	12.2
Pacific Islands	6.4	6.1	6.8	6.4
UK & Ireland	14.3	14.6	9.9	11.1
Europe & N. America	12.6	14.9	11.3	12.1
Asia	7.1	8.0	8.7	7.6
Other	11.9	11.0	12.8	11.5
Total overseas born	9.7	10.0	9.6	9.6

Again, the non-migrant-related variables successful in explaining those employed in associate professional or technician occupations were vocational and degree qualifications. In this case, these two variables are successful in explaining nearly 71% of the variation in employment in professional occupations. As with the professional occupations, none of the census data year identifiers was found to be significant. Thus, the determination of proportions in associate professional and technician occupations has remained unchanged over the 1996-2006 period. Age-related factors were also not significant here.

As for the migrant-related variables, we found the New Zealand born identifier to be significant, with a positive coefficient. This implies that New Zealand born groups, after controlling for qualifications, were significantly more likely than other groups to be employed in associate professional and technician occupations. This identifier added 3 percentage points to the determination, taking the total explanatory power of the equation to 74%.

Of other migrant-related variables, when introduced individually, those found to be significant were:

• recent migrants (negative coefficient)
• intermediate migrants (positive coefficient)
• Australia birthplace* (positive coefficient, but see note below)
• Europe and North America birthplace (positive coefficient)
• Other birthplace (negative coefficient)

(* It should be noted that the introduction of the Australia birthplace identifier did significantly erode the robustness of the equation and make it unsatisfactory. Individually, each of these variables added in the range of a further 2-5% to the explanatory power of the estimated equation.)

Attempts to include all these variables (or sub-sets thereof) into the one equation were of varying success, that is, this resulted in the relationship between the variables to be difficult to separately identify. In several cases, the robustness of the estimated equations, including these migrant-related variables, were significantly eroded^[17] and thereby rendered unsatisfactory.

However, a satisfactory equation was determined. This retained the identifier for recent migrants, and those born in Europe and North America, as well as the Other birthplace region. Including these three additional migrant-related characteristics added 9 percentage points to the explanatory power of the estimated relationship. Consequently, this equation successfully explained 83% of the variation in the proportion employed in this occupation group.

Despite the relatively low level of significance of the migrant-related factors, the identification of three different birthplace identifiers^[18] in this equation suggests their significance should not be understated.

We conclude that, while the determination of those in associate professional and technician occupations was dominated by qualifications characteristics, migrant-related factors played an important role over the 1996-2006 period. After controlling for qualification factors, employment in these occupations is clearly related to birthplace. Those born in New Zealand, and Europe and North America are more likely to be employed in these occupations, while those born in the Other birthplace region are less likely. In addition, recent migrants are also significantly less likely to be employed in these occupations.

 

Preceding the above formal multivariate analysis, we provided observations from numerous cross-tabulations of the data (sections 5-8). As noted earlier, not all cross-tabulations can be covered in this report. However, the spreadsheet tool has been developed to enable desired comparisons to be easily generated and depicted. A selection of comparisons regarding income of migrants was provided in section 5. Sections 6-8 provided a selection covering source of income, labour market status and occupation respectively.

Additional details of multivariate analysis

Each of the estimated relationships took the following form:

`Y_(ijt)=a^mA_(ijt)^m+q^kQ_(ijt)^k+alpha^oT^(o)+beta^bB_(ijt)^b+delta^rR_(ijt)^r`

where

`A_(ijt)^m`= proportion of group ijt that is in age group m

`Q_(ijt)^k`= proportion of group ijt that possess qualifications of type k

`T_(ijt)^o`= identifier (dummy) for census year (i.e. T=1 if o=t, T=0 otherwise)

`B_(ijt)^r`= identifier (dummy) for birthplace characteristic (i.e. B=1 if b=i, B=0 otherwise)

`R_(ijt)^r`= identifier (dummy) for years in New Zealand characteristic (i.e. R=1 if r=j, R=0 otherwise)

`ijt` = group with birthplace i and years in New Zealand j for census year t

`i` = New Zealand, Australia, Pacific Islands, United Kingdom and Ireland, Europe and North America, Asia, Other

`j` = less than 5 years, 5-14 years, 15 or more years, if i≠New Zealand; all if i = New Zealand

`t` = 1996, 2001, 2006

and

`a`, `q`, `alpha`, `beta` and `delta` are coefficients

`Y_(ijt)`= dependent variable for population group ijt to be explained.

 

Within this construction, we have 19 different population groups (i.e. six different non-New Zealand birthplaces, each with three sub-groups according to years in New Zealand, and one New Zealand born group). For each of these 19 groups, we have three census year observations, thus making a dataset with 57 observations in all.

Our first step was to investigate the presence, if any, of relationships excluding any of the migrant-specific characteristics. We also exclude the census year dummy variables. In other words, we set α=β=δ=0, for all i, j and t.

Next, we introduced each census year dummy identifier to establish their significance.

Thereafter, we progressively introduced each migrant specific identifier individually, testing for their significance as well as the robustness of the equation.

Once the set of significant migrant characteristics were established, they were together introduced into the equation to enable associations and independence to be eliminated. Thus, the final relationship chosen was the one with the highest explanatory power subject to tests of significance and independence of residuals.

The dependent variables investigated within this construction were as follows:

Ytop3_ijt

percentage of population group ijt that has income above the 70th percentile of the national income distribution

Ybottom3_ijt

percentage of population group ijt that has income below the 30th percentile of the national income distribution

LFPR_ijt

labour force participation rate (%) of population group ijt

UNEM_ijt

unemployment rate (%) of population group ijt

Oprof_ijt

percentage of population group ijt that are employed in professional occupations

Oapftec_ijt

percentage of population group ijt that are employed in associate professional or technician occupations.

As noted in the text, attempts to estimate equations for the proportion in each sub-group that are employers and self-employed were unsuccessful. In addition, attempts to estimate an equation to explain the proportion of the population sub-groups employed in trades occupations also proved unsuccessful.

---

^[16] That is, the residuals did not pass the test of independence.

^[17] That is, the residuals did not pass the test of independence.

^[18] And further noting the significance of the other birthplace identifiers, although their independence could not be established.