• No results found

Chapter 3. Data and methodology

3.1 Data sources

My analysis draws primarily on the Survey of Dynamics and Motivations for Migration in New Zealand, undertaken by Statistics New Zealand in 2007 and is referred to hereafter as the DMM survey or simply, the survey. The survey was run in conjunction with Statistics New Zealand‟s long running quarterly Household Labour Force Survey (HLFS). In addition to these surveys I also utilise data from the Statistics New Zealand‟s 2006 Census of Population and Dwellings (census) and Urban and Rural Profile classification, as well as the New Zealand Ministry of Health and Otago University‟s New Zealand Deprivation Index (deprivation index). I also utilise data on New Zealand‟s LLM catchments, which is derived from census commuting data.

Together, these data allow me to investigate not just the characteristics of the mover, but also the characteristics of their move, where they move from and where they move to. I introduce each in turn8.

The 2007 DMM Survey and HLFS

The DMM survey was appended as a supplement to the Household Labour Force Survey (HLFS) in the March 2007 quarter. As a supplement, the survey gains access to a pre-existing statistically representative sample of New Zealand households.

The HLFS is a nationwide, quarterly survey, comprising of about 15,000 statistically representative households, equating to around 30,000 individuals over the age of 15 and is used to generate some of New Zealand‟s key economic indicators, such as the unemployment rate and labour force participation rate (Statistics New Zealand, 2009c).

The HLFS collects information about the activities of households and individuals during a particular week and includes data on the characteristics of individuals, such as their

8 They are also provided in Appendices 3, 4 and 5

age, gender, ethnicity and nationality, as well as the location of the household (Statistics New Zealand, 2009e, Statistics New Zealand, 2009c, Statistics New Zealand, 2009b, Statistics New Zealand, 2007c). I use these data, where applicable, to capture the socio-economic characteristics of movers and where they live.

Households which are selected to participate in the HLFS are interviewed quarterly for up to two years and as such respond to a total of up to eight interviews.

The first interview is conducted face-to-face and subsequent interviews are conducted by telephone. Respondents to the HLFS survey were asked to participate in the DMM survey over the period 7 January to 7 April 2007, provided they were living in occupied private dwellings. Of the 88.7 per cent of respondents who responded to the HLFS in the March 2007 quarter, 87.8 per cent also responded to the DMM supplement, leading to an overall weighted response rate of 77.9 per cent, slightly below the target response rate. This 77.9 per cent of respondents resulted in a total sample size of 23,465 individual responses from 13,841 households.

Of these 23,465 sampled individuals, 24 per cent had moved within New Zealand over the two years prior to being interviewed. Those who had moved more than once in the two year period were only asked about the characteristics of their most recent move. Of those whose most recent move was within New Zealand, 4937 respondents were able to be subsequently geocoded using their street address. They represent 87% of all respondents whose most recent move was within New Zealand.

The DMM survey was explicitly designed to “investigate what motivates some people to move from one house to another, from one part of New Zealand to another, or to and from New Zealand, and what motivates people to stay where they are” (Statistics New Zealand, 2007c: p. 2). The survey was undertaken in order to provide a greater degree of information about internal migration in New Zealand than was previously available. Unlike other sources of information such as the census, the DMM survey was primarily focused on understanding the drivers and motivations for internal migration, as well as why some people move and others do not.

Respondents to the DMM survey who had moved within the previous two years were asked a range of questions relating to their move. In particular they were asked where they had moved from, their reasons for moving away from their previous location and their reasons for moving to their new location. Most importantly for this

inquiry, they were asked about their satisfaction with their move and how that satisfaction, in a number of domains, had changed relative to before their move. The DMM survey is therefore unique in that it asks a number of questions of the mover not yet addressed by migration researchers, within New Zealand and internationally. For example, the reason individuals give for moving is separated into the reasons for moving away from their place of origin and also the reasons for moving to their destination area. Individuals were also asked whether their income changed over the period and whether it was a result of their move.

Geographical framework

The survey fits within the wider Statistics New Zealand geographical framework, allowing the inclusion of a wide range of location specific data. The origin and destination locations of respondents to the survey can be calculated at the level of the meshblock and area unit9 (Statistics New Zealand, 2011), as well as to clusters of area units which I later define as LLMs. The presence of a meshblock identifier allows for the inclusion of additional statistics to be appended if needed. Examples include population density and neighbourhood deprivation data derived from the 2006 Census.

The meshblock coding also enables the application of GIS techniques to compute other measures such as the distance moved. The distance of a move is approximated by taking the Euclidean distance between the centroids of the origin and destination meshblocks, or if unavailable, area units10. Using ArcGIS, I undertake the following estimation. The centroid of each meshblock and area unit was calculated as (x , y), where the x and y values of the centroids are the midpoint on the plane of the maximum extents of each East/West, North/South dimension:

(3.1) (x , y) = ([max(x) - min(x)]/2 , [(max(y) – min(y)]/2).

9 The smallest geographic unit utilised in this project is the meshblock, which represent Statistics New Zealand‟s smallest spatial area. Meshblocks contain an average of approximately 60 individuals in rural areas and 110 individuals in urban areas, with urban meshblocks being roughly the size of a city block while in rural areas they may be considerably larger. Meshblocks cover the whole of New Zealand and are derived from population and household data from the 2006 Census. There are 41,376 meshblocks in total. Statistics New Zealand‟s second smallest statistical area is the area unit, which comprises of a number of complete meshblocks that combine to “be a single geographic entity with a unique name”

(Statistics New Zealand, 2011). In urban areas, area units usually contain between 3,000 and 5,000 individuals.

10 This approach of using centroids is also utilised by Goodyear and Ralphs (2009) in their modelling of journey to work data.

In other words, the x value is the midpoint between the most eastern value and most western value and the y value is the midpoint between the most northern value and the most southern value. Figure 3.1 illustrates using area units in Wellington City. The distance of the move between the centroids of the origin and destination area is then calculated using Euclidean distance:

(3.2) √( ) ( )

Figure 3.1: Calculating distance of moving using Euclidean distance between meshblock centroids, 2006 New Zealand Census.

While convenient, the use of centroids in this manner introduces a degree of error to the length of each move. The size of the error is dependent on a number of factors. The distance of the household from the centroid of the measured meshblock at both their origin and destination location influences the size of the error. For example, because of their larger meshblocks, the error is much more significant in rural areas than it is in urban areas. In irregular shaped meshblocks, as shown by the light shaded polygon in Figure 3.1, it is possible for the centroid (also shaded lightly) to be located outside the meshblock.

The size of the error is a proportion of the length of the move and is therefore greatest for short distance moves. In situations where an individual moves within a meshblock, the distance recorded is equal to zero, as there is no change in meshblock centroid. Because the natural logarithm of zero does not exist, when calculating ln(distance) moves within meshblocks are given a small nominal value of 100m.

Being a simple straight line calculation, Euclidean distance does not take into account the physical barriers such as the Cook Strait, which separates the North Island from the South Island, or non-linear configurations in transportation networks. I assume throughout that, for moves between meshblocks, the error associated with the estimated distances are randomly distributed about a mean of zero and not related to any of the arguments introduced.

New Zealand Census

Also incorporated in the dataset are data from Statistics New Zealand‟s 2006 Census of Population and Dwellings (Census), which was held on 7 March 200611. The 7th of March 2006 fell roughly halfway through the two year period that movers were asked to recall in the DMM survey. The timing was deliberate and as a result, the data from the Census reflects a timely snapshot of the demographic, social and economic landscape of New Zealand at the time of most moves. Use of census data takes two forms: that utilised directly by me, and that used by other organisations to generate the New Zealand Deprivation Index and the Urban and Rural Profile.

Full 1996, 2001 and 2006 Census datasets were available throughout the exploratory and analytical process. In my final analysis, little census data was directly utilised, primarily because many of the factors of interest were included indirectly as components of the New Zealand Deprivation Index and the Urban and Rural Profile.

New Zealand Deprivation Index

The census is used to calculate the New Zealand Deprivation Index 2006, referred to here simply as the deprivation index. The deprivation index is calculated by the New Zealand Ministry of Health in conjunction with Otago School of Medicine Wellington following each census. It measures the relative level of socio-economic deprivation of an area and is used in funding formulas for a range of social services. A range of variables is used, including the individual and household income, home ownership, family support, employment, qualifications and transport accessibility within a location, as outlined in Table 3.1.

11 New Zealand undertakes a census every five years, provides a snapshot of New Zealand on that day and the census is “used for policy-setting and implementation, research, planning and other decision-making” (Statistics New Zealand, 2006: p. 1).

Table 3.1: Description of the nine variables used to construct NZDep.

Deprivation domain Census variables (in order of decreasing weight) Income Aged 1-64 years receiving a means-tested benefit

Income Living in households with equivalised* income below an income threshold

Owned home Not living in own home

Support Aged under 65 living in a single-parent family Employment Aged 18-64 years and unemployed

Qualifications Ages 18-64 years and without any qualifications

Living space Living in households below an equivalised* bedroom occupancy threshold Communication With no access to a telephone

Transport With no access to a car

*Equivalisation: methods used to control for household composition

Source: (White et al., 2008: p. 9)

The geographic framework of the deprivation index is built using grouped meshblock data. Statistics New Zealand‟s 41,376 meshblocks are aggregated into 23,786 NZDep2006 small areas, each comprising of at least 100 people. Thus, every meshblock has an NZDep value but may share a value with a group of one or more geographically connected meshblocks. The NZDep values are calculated as follows:

“The NZDep2006 continuous score is a weighted sum of the nine variables created using a principal components analysis. This statistical method identifies weighted sums of variables that progressively account for the overall variation in the data.

“The NZDep2006 index is the first principal component scaled to have a mean of 1000 index points and standard deviation of 100 index points. The index is the weighted sum of the variables that accounts for the most variation. Each variable in the sum is a proportion of people in a small area. Each proportion is standardised in eight age–gender groups (0–17 years, 18–39 years, 40–64 years, 65 years and over, for each gender) to the New Zealand population structure. This equivalises the small areas, so that some areas cannot be considered more deprived than others simply because their populations have different age structures.” (White et al., 2008: pp. 9-10)

As a result, the deprivation index provides a standardised measure of the degree of socio-economic deprivation of small areas, interpreted by some as a neighbourhood. I have used this data to evaluate how the post-move satisfaction of movers varies when they change from “neighbourhoods” in different deprivation categories.

Local Labour Market Areas

The census also contains information about residents‟ travel to work and it is these data that has been used to partition the country into LLMs. From dwelling and workplace addresses, commuting flows have been generated for use in a number of applications (Goodyear and Ralphs, 2009). Using travel-to-work data from the 1991

Census, Papps and Newell (2002) quantified New Zealand‟s functional labour market catchments, following the algorithm developed by Coombes et al. (1986) and also used by (Casado-Díaz, 2000). The algorithm partitions the country based on the dominant commuting flows into and out of centres of employment.

One notable benefit of this approach is that all locations are included, regardless of whether they are a metropolitan area or rural. A further benefit from an analytical perspective is that these LLMs are non-overlapping, although this does

“ignore[s] the competition between labour catchments that occurs in reality” (Papps and Newell, 2002: p. 9). My thesis uses an updated dataset using 2006 Census data, kindly provided by J. Newell.

In summary, the use of several sources of data for the study of post-move satisfaction has enabled me to derive information about the individual, their move, and the spatial environments that they occupy. As a result, I have been able to consider a broader range of factors, such as income change, deprivation and distance, in this single study than reported elsewhere in the post-move satisfaction literature. In the following section I introduce my dependent variables.