Expenditure
In every wave, HILDA collects housing expenditures (rent and mortgage repayments) in the Household Questionnaire. The household expenditure on groceries, food and meals eaten outside were collected in the Household Questionnaire for wave 1, 3, 4, and 5. Household expenditure on a wide range of goods and services were first collected in the wave 5 Self-Completion Questionnaire. The list of items collected was expanded to include consumer durables from wave 6.
While the person in the household responsible for the household bills was asked to complete the household-level expenditure questions in the SCQ, sometimes more than one person in a household provided answers. The variables with the prefix _hx average the responses across all individuals who provided a response to these expenditure questions (the responses from dependent students who stated they are not responsible for the household bills are excluded).26
Imputation Method
The expenditure variables were imputed for the first time in Release 8 and were extended to include the imputation of child care costs for Release 9. A summary of the steps in the imputation process is provided below:
- Step 1 – Create a longitudinal household identifier. For variables imputed at the household-level, households are linked longitudinally if they had common membership.27 Deaths and births, for the purposes of expenditure imputation, are counted as a membership change.
- Step 2 – Identify lumpy expenditure items. Some items (such as cars, white goods, etc) would not be purchased each year, so need to be treated differently in the imputation process.
- Step 3 – Carryover zeros. The population carryover method is used to determine zero and non-zero expenditure flags for non-lumpy expenditure items prior to any other imputation. Lumpy expenditure items were excluded from this step.
- Step 4 – Nearest Neighbour Regression imputation of zeros. The predicted values from a regression model are used to identify a donor from which to flag zero or non-zero imputes for the recipient. This is essentially a filter process to decide whether the case has the expense or not.
- Step 5 – Nearest Neighbour Regression imputation of non-zero amounts. The predicted values from a regression model are used to identify a donor from which the reported value is taken as the imputed value for the recipient. The models and donor pools are restricted to cases with non-zero amounts. For households without any expenditure data reported in the SCQ, a single donor for all expenditure variables collected in the SCQ was used.
- Step 6 – Little and Su imputation. This method incorporates (via a multiplicative model) the trend across waves (column effect), the recipient's departure from the trend (row effect), and a residual effect donated from another case with complete expenditure information for that component (residual effect). Only cases that have been enumerated in more than one wave, longitudinally linked, and have at least one wave of non-zero data available can be imputed via this method. For the lumpy expenditure items, the donors selected had to have the same zero pattern for the non-missing waves as the recipients. Wherever possible, the Little and Su imputation replaces the Nearest Neighbour Regression imputation. The zero or non-zero determination from steps 3 and 4 is observed.
Imputation classes were used for some variables to ensure the donors and recipients match on a small number of characteristics (typically equivalised household disposable income bands and the age group of the highest income earner were used).
A full description of the imputation process for the expenditure variables is provided by Sun (2010). Appendix 2 provides an extract from Hayes and Watson (2009) which details the Nearest Neighbour Regression method, the Little and Su method and the Population Carryover method.
Table 4.22 shows the percentage of missing cases that were imputed by each imputation method.28 Ideally all the records should be imputed by a longitudinal imputation method, such as the Little and Su method or the Carryover method. The households which cannot be linked between waves were imputed by the Nearest Neighbour Regression method regardless of their situation. For the housing expenditure variables (rent payment, mortgage repayment and second mortgage repayment), which have been collected in all waves so far, the majority of cases were imputed by the Little and Su method. For the expenditure items collected from wave 6 onwards where we fewer waves of data available, more than half of the cases were imputed by the Nearest Neighbour Regression method.
Table 4.22: Percentage of missing cases imputed by imputation method (expenditure), waves 1 to 9
| Imputation Method | Wave | ||||||||
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | |
| Household and work-related child care expenditure variables (collected in all waves Household Questionnaire) | |||||||||
| Nearest Neighbour | 45.6 | 13.6 | 22.2 | 22.4 | 14.4 | 25.6 | 16.3 | 21.8 | 41.2 |
| Little & Su | 46.6 | 70.9 | 71.4 | 73.5 | 77.5 | 69.9 | 75.0 | 74.6 | 49.3 |
| Carryover | 7.8 | 15.5 | 6.5 | 4.1 | 8.1 | 4.5 | 8.7 | 3.6 | 9.6 |
| Non-work-related child care expenditure variables (collected in all waves Household Questionnaire) | |||||||||
| Nearest Neighbour | – | 66.7 | 20.0 | 44.4 | 81.8 | 75.0 | 66.7 | 76.2 | 85.2 |
| Little & Su | – | 19.4 | 40.0 | 44.4 | 18.2 | 25.0 | 28.6 | 14.3 | 8.2 |
| Carryover | – | 13.9 | 40.0 | 11.1 | 0.0 | 0.0 | 4.8 | 9.5 | 6.6 |
| Weekly household expenditure variables (collected in wave 1, 3, 4, and 5 Household Questionnaire) | |||||||||
| Nearest Neighbour | 56.2 | – | 27.6 | 23.3 | 35.6 | – | – | – | – |
| Little & Su | 42.8 | – | 65.0 | 70.2 | 56.6 | – | – | – | – |
| Carryover | 1.0 | – | 7.4 | 6.5 | 7.8 | – | – | – | – |
| Annualised household expenditure variables (collected in the Self-Completion Questionnaire from wave 5) | |||||||||
| Nearest Neighbour | – | – | – | – | 62.3 | 41.4 | 38.4 | 40.0 | 53.9 |
| Little & Su | – | – | – | – | 27.8 | 41.7 | 45.7 | 43.7 | 35.8 |
| Carryover | – | – | – | – | 9.9 | 17.0 | 15.9 | 16.3 | 10.3 |
| Annualised household expenditure variables (collected in the Self-Completion Questionnaire from wave 6) | |||||||||
| Nearest Neighbour | – | – | – | – | – | 69.3 | 57.0 | 58.3 | 66.9 |
| Little & Su | – | – | – | – | – | 25.2 | 34.7 | 33.8 | 27.4 |
| Carryover | – | – | – | – | – | 5.5 | 8.3 | 7.9 | 5.6 |
Imputed Household Expenditure Variables
All expenditure imputation was undertaken at the household level. Both the pre- and post-imputed variables are available in the datasets, along with an imputation flag. Table 4.23 provides an overview of the pre- and post-imputed expenditure variables and the waves in which they are available.
Table 4.23: Imputed household expenditure variables
| Wave | Pre–imputed1 | Post–imputed | flag | |
| Usual payments/repayments per month (Collected in the HQ) | ||||
| Rent | 1 - 9 | _hsrnt | _hsrni | _hsrnfg |
| First mortgage | 1 - 9 | _hsmg | _hsmgi | _hsmgfg |
| Second mortgage | 1 - 9 | _hssl | _hssli | _hsslfg |
| Weekly household expenditure (Collected in the HQ) | ||||
| All groceries | 1, 3 - 5 | _xpgroc | _xpgroci | _xpgrocf |
| Groceries for food and drink | 1, 3 - 5 | _xpfood | _xpfoodi | _xpfoodf |
| Meals eaten outside | 1, 3 - 5 | _xposml | _xposmli | _xposmlf |
| Annualized household expenditure (Collected in the SCQ1) | ||||
| Groceries | 5 - 9 | _hxygroc | _hxygrci | _hxygrcf |
| Alcohol | 5 - 9 | _hxyalc | _hxyalci | _hxyalcf |
| Cigarettes and tobacco | 5 - 9 | _hxycig | _hxycigi | _hxycigf |
| Public transport and taxis | 5 - 9 | _hxypubt | _hxypbti | _hxypbtf |
| Meals eaten out | 5 - 9 | _hxymeal | _hxymli | _hxymlf |
| Leisure activities | 5 | _hxyhsge | _hxyhsgi | _hxyhsgf |
| Motor vehicle fuel | 5 - 9 | _hxymvf | _hxymvfi | _hxymvff |
| Men’s clothing and footwear | 6 - 9 | _hxymcf | _hxymcfi | _hxymcff |
| Women’s clothing and footwear | 6 - 9 | _hxywcf | _hxywcfi | _hxywcff |
| Children’s clothing and footwear | 6 - 9 | _hxyccf | _hxyccfi | _hxyccff |
| Clothing and footwear | 5 | _hxyclth | _hxyclti | _hxycltf |
| Telephone rent and calls | 5 | _hxytel | _hxytli | _hxytlf |
| Telephone rent and calls, internet charges | 6 - 9 | _hxyteli | _hxytlii | _hxytlif |
| Holidays and holiday travel costs | 5 - 9 | _hxyhol | _hxyholi | _hxyholf |
| Private health insurance | 5 - 9 | _hxyphi | _hxyphii | _hxyphif |
| Other insurances | 6 - 9 | _hxyoi | _hxyoii | _hxyoif |
| Fees paid to health practitioners | 6 - 9 | _hxyhltp | _hxyhlpi | _hxyhlpf |
| Medicines, prescriptions and pharmaceuticals | 6 - 9 | _hxyphrm | _hxyphmi | _hxyphmf |
| Health care | 5 | _hxyhlth | _hxyhthi | _hxyhthf |
| Electricity bills | 5 | _hxyelec | _hxyelei | _hxyelef |
| Gas bills | 5 | _hxygas | _hxygasi | _hxygasf |
| Other heating fuel | 5 | _hxyohf | _hxyohfi | _hxyohff |
| Electricity, gas bills and other heating fuel | 6 - 9 | _hxyutil | _hxyutli | _hxyutlf |
| Repairs, renovation and maintenance to home | 5 - 9 | _hxyhmrn | _hxyhmri | _hxyhmrf |
| Motor vehicle repairs and maintenance | 5 - 9 | _hxymvr | _hxymvri | _hxymvrf |
| Education fees | 5 - 9 | _hxyeduc | _hxyedci | _hxyedcf |
| Buying brand new vehicles | 6 - 9 | _hxyncar | _hxyncri | _hxyncrf |
| Buying used vehicles | 6 - 9 | _hxyucar | _hxyucri | _hxyucrf |
| Computers and related services | 6 - 9 | _hxycomp | _hxycmpi | _hxycmpf |
| Audio visual equipment | 6 - 9 | _hxytvav | _hxytvi | _hxytvf |
| Household appliance | 6 - 9 | _hxywg | _hxywgi | _hxywgf |
| Furniture | 6 - 9 | _hxyfurn | _hxyfrni | _hxyfrnf |
| 1 | The household-level responses provided by each person in the household responsible for household expenditure are provided in equivalent variables to the pre-imputed household expenditure variables from the SCQ (_hx is replaced by _xp to give variables _xpgroc to _xpyfurn). Most users will use the _hx variables. Back to where you were |
Endnotes:
| 26 | For each of the _hx pre-imputed variables listed in Table 4.23, corresponding _px variables are provided, which are the derived annualised response for each person who provided a response to these questions. Most users will use the _hx variables. Back to where you were |
| 27 | _hxylink is an indicator variable for whether a household was linked to another household in the next wave for the purposes of imputing expenditure. Back to where you were |
| 28 | For the proportion of cases which are missing, see Table 6.8. Back to where you were |