The 1996 Census Housing Paper evaluates the data quality of the housing questions in the Census. The housing variables analysed in the paper include: Tenure Type, Rent/Housing Loan Repayment, Type of Landlord, Furnished/Unfurnished, Number of Bedrooms, Structure of Dwellings, Location of Dwellings and Non-Private Dwellings. The housing questions were completely restructured for the 1996 Census in an effort to improve data quality and to reduce respondent burden. Comparisons between 1991 and 1996 Census data has been undertaken. The Census data are also compared with data from the 1994 Australian Housing Survey and the 1995-96 Survey of Income and Housing Costs. The main conclusions of the analyses were as follows:
Census Working Paper 99/3
1996 CENSUS DATA QUALITY: HOUSING
- Overall the restructure of the housing questions was a success as the data quality for most topics improved substantially.
- An increased number of categories for Tenure Type generally provided a more accurate representation of tenure types.
- The increase in options for Type of Landlord did not increase data quality as desired, however it did provide users with a wider range of data.
- There was an increase in non-response rate for housing loan repayments (1.9 percentage points), and a decrease for rent (0.3 percentage points). This was a slight inconsistency, for which a couple of reasons have been considered, including the change to a write-in box, and privacy issues.
- Non-response rate increased for Structure of Dwelling which indicated potential difficulties experienced at the collection phase. This issue is currently being addressed in the field operations procedures for 2001.
- The introduction of a new classificatory variable 'Dwelling Location' proved to be reasonably successful, however collection procedures are being refined for the 2001 Census.
- Overall Type of Non-Private Dwelling was much better captured in 1996 compared with 1991.
- There was some improvement in the quality of data for 'relationship' for persons in Non-private dwellings. However modifications to the question design will be tested for the 2001 Census to further improve the data quality to the standard required by users.