The results presented in this study are the work of the authors, not Statistics New Zealand (Statistics NZ); they are not official statistics. This article describes the construction of a unique longitudinal data set that enables the study of the dynamics of family incomes in New Zealand. The most comprehensive longitudinal survey in New Zealand is the Survey of Family, Income and Employment (SoFIE).
Therefore, the linking of data sets involved an extensive deterministic and probabilistic matching exercise.5 The structure of the IDI can be described as a 'central backbone' through which a series of administrative and survey data sets are linked at the individual level. The IDI backbone aims to include all people who have ever been resident in New Zealand. Individuals can be linked across different datasets using these unique identifiers, which change and are reassigned in each 'refresh': the reload archive used for the current exercise is IDI_Clean_20200120.
As a result of the Census transformation strategy adopted by the New Zealand Government in 2012, a number of studies investigating the feasibility of providing census-type information from administrative data sources have been undertaken by Statistics NZ.9 The potential for such transformation has previously been explored by the Nordic countries, where the census in Denmark (since 1981), Finland (since 1990) and Norway and Sweden (since 2011) has been replaced by administrative data.10. 11 As mentioned earlier, the IDI backbone aims to include all people who have ever been 'resident' in New Zealand.
The Unit of Analysis
Finally, the constructed families in Gath and Bycroft (2018) are compared to the 2013 census, while family information used here includes the 2013 census along with the two other stand-alone censuses in 2001 and 2006.
Constructing Families
Parent-Child Relationships
In order for the reference date to match the income data, a reference year is based on the tax year: this runs from 1 April each year to 31 March the following year. To address this concern, other complementary data sources such as MSD, WfF and MBIE are used to improve coverage. It is important to emphasize that family information on people who migrated to New Zealand before 1997, or those who do not require a visa to live and work in New Zealand, such as Australian migrants, is not available in the MBIE data sources (and possibly not be available in the DIA data if none of their children were born in New Zealand).
These include unemployment benefit (UB), household benefit (DPB), widow's benefit (BB), emergency maintenance allowance (EMA), independent youth, orphans and dependent children, sickness benefit (SB) ) and disability benefit (IB). Since this source retains the dates of inclusion in receipt of child benefit, the reference date is set again to capture those parental relationships that exist in the relevant tax year. As with the MSD data source, the reference date is set to capture parental relationships recorded in each tax year. This contains information on all migrants who have ever made an application to Immigration New Zealand since 1997.
Third, the tax year limitation is applied to the decision date based on when the final decision on the visa application is made. Parent-child relationships extracted from these data sources are then combined using IDI's unique personal identifiers. Data sources are prioritized according to their coverage of the parental links provided and are checked in order.
Therefore, if a parental link is available in the DIA-Births for a particular individual, it will be registered as final. Otherwise, the availability of links in the WfF, migration data from MBIE and MSD is then checked. The corresponding parent identifiers (parent 1 and parent 2) may appear in reverse order when using different data sources.
While it is possible to infer the partnership between parents from the previously described data sources, the status of the partnership may change over time.
Partnerships
The next data source from which partnership information can be gathered is the Benefit Dynamics Datasets (BDD) provided by MSD. These datasets contain information on all people who have received a social welfare benefit during working age since 1993. Given that the information on time is given for each record, the links of partners are preserved if they are together at any time in a given period. tax year.
As with other datasets, partnership information is collected for those individuals who are in a relationship in a given tax year. Previous administrative data sources used to obtain information about the partnership focused primarily on local residents. However, the NZ population is made up of many individuals who have migrated from overseas at some point.
In particular, the administrative data sources previously examined may be representative of the total national population, but do not include individuals (and their partners, if any) who have recently migrated to New Zealand. That's partly because these individuals (and their families) are less likely to come into contact with government agencies. It is also possible that no matching record can be found in the last census performed.
An important caveat, when using this data to derive partnership information, is that two applicants with the same application number are not necessarily partners. To avoid any wrong inference, partnerships are only inferred for those partners who have children together. Partnerships information collected from all these data sources is then combined using the IDI unique person identifiers.
An individual may appear multiple times in the combined data set, in which case only the latest indicated partner from any source in each year is retained.
Additional Relationship Information from Census Data
Therefore, the other censuses that could potentially be used to create linked datasets are 2001 and 2006, neither of which are available in the IDI environment.15 Access to these two independent datasets was provided to the project by Statistics NZ. Link variables are required to manually link records from each of these inventories to the backbone. To overcome this problem, two shortened versions of these datasets, including a date of birth variable, were subsequently requested and provided to the project by Statistics NZ.
In order to have a stand-alone census with date of birth instead of age, the most recent versions of these datasets (short versions) were first linked to the longer versions originally provided. This added the dates of birth, which are available in abbreviated versions, to the existing stand-alone census data sets. The next step was to add two key linking variables, namely date of birth and gender, to the residential address.
Finally, these individual census data sets are linked to the IDI backbone using core linkage variables. Therefore, the final step is to identify partners that are both linked to the IDI backbone. 20 The existence of name and day of birth could potentially improve the linkage significantly, but these are not provided for confidentiality reasons.
The 2013 census provides full coverage of partners in the NZ population at the time of the census. This is mainly because at the individual level this census corresponds to the IDI with a fairly high rate of agreement. The overall linkage rate of the 2013 census with ordinary residents to the IDI was 92.4 percent.
Given the fact that the relationship information from the previous censuses (2001 and 2006) is limited to those partners where both are linked to the IDI backbone, the relationship information from the 2013 census is split into two files.
Collecting Relationship Information from All Sources
For simplicity, this table shows cases where the family structure remains the same over the period of study. Unlike the previous table, a family type in this table can change from one period to the next depending on changes in family composition. The next segment refers to a single adult living with a dependent child until 2016, when a new adult member, a partner of the main adult member, is added to the family (Partner_snz_uid = 69810).
A year later, in 2017, one of the children leaves the family and forms a new family. To do this, individual-level income data is linked to the household-level datasets using the individual-level identifiers (Snz_uid). By creating a unique identifier at the family level, any changes in family composition over time can be tracked.
In cases where the adult member or members of the family remain unchanged over time, the same identifier is always assigned to the family. In cases where there is only one adult member in the family (single and single with dependent children), the individual identifier is assigned as a family level identifier, so the two identifiers are the same. Family income is then calculated by adding together the income of the adult members of the family.
The family level identifier assigned to this single family is identical to the individual level identifier (Snz_uid . = Fam_id = 105185). The taxable income for the family is also equal to the income of the single adult member of this family. Because this family consists of two adult members, the individual-level identifiers are sorted first for the adult members so that the family-level identifier begins with the smaller number.
By combining the income between the adult members of this family, the taxable income for the family in 2000 is calculated as LLLL2000 + XXXX2000 (the taxable income corresponding to the main individual and partner is LLLL2000 and XXXX2000, respectively). The construction of the identifier at the family level for more complex cases is shown in Table 5. The composition of the family corresponding to this individual changes in 2016, when a couple-type family is formed.
Conclusions
Nazila Alinaghi is a Research Fellow in Public Finance at Victoria Business School, Victoria University of Wellington, New Zealand. John Creedy is Professor of Public Finance at the Victoria Business School, Victoria University of Wellington, New Zealand. Norman Gemmell is Professor of Public Finance at the Victoria Business School, Victoria University of Wellington, New Zealand.
Working Papers in Public Finance
