APPENDIX 5 AGGREGATE IMPACT OF THE 2003–04 CHANGES TO SIH ON INCOME MEASURES
IMPACT ON INCOME BY SOURCE
The changes in methodology between 2002-03 and 2003-04 outlined in Section 4.2 have impacted on the comparison of the 2003-04 results with those for earlier cycles. While not all impacts can be quantified, the potential significance of the impacts on various sources of income are discussed below.
For wages and salaries, no obvious impacts were detected. Average wages and salaries in the 2003-04 results are 4.8% higher than in 2002-03, in line with the increase in average total weekly earnings reported in ABS business surveys. For selected distributional measures of gross wage and salary income (the Gini and the quintile income shares) the distributions in the two years are very similar.
For government pensions and allowances, no obvious impacts were detected. For 2003-04, the coverage of survey reported benefits compared to the benefits and allowances paid by government was slightly above the longer term average in cycles from the mid to late 1990s, but within one standard error of that average. Therefore, while a benefit benchmark had been introduced for the 1999-00 and 2000-01 cycles (when coverage fell significantly), no benchmark was used in either 2002-03 or 2003-04.
For investment income, the change in 2003-04 to ask about current income, rather than imputing the income on a 'no change' assumption from reported income for the previous financial year, has been significant. In the 2002-03 results, the imputed total current investment income estimate was $16.2 billion. This simple imputation methodology, which had been used since the mid 1990s as the practical approximation to measuring current investment income, did not always result in year to year movements that were consistent with the related property income series in the household income account of the Australian System of National Accounts. This was particularly so for the current income imputed estimates for 2002-03. In 2003-04, respondents reported investment income amounts earned in 2002-03 at $19.8 billion, and current income in 2003-04 at $22.3 billion. The year on year movement between the current and previous year investment incomes reported in 2003-04 is broadly in line with the related national accounts series. The difference between the imputed amount for 2002-03 and the subsequently reported amount for that year contributes about $9 to the increase between the results for 2002-03 and 2003-04 in average gross weekly household incomes.
The change in methodology to capture reported current income was expected to produce a one-off break in the level of the household income series. It is also expected to provide a significant improvement in the future investment income and total income series. However, from the testing that has been undertaken, it is not obvious that the change in methodology has significantly affected income distribution measures. For the Gini and the quintile income shares, the change in gross household income distributions by excluding investment income is very similar for 2002-03 and 2003-04. Increasing the 2002-03 imputed investment income amounts by the ratio of reported to imputed incomes results in very little change to the selected income distribution measures for 2002-03.
As for investment income, for income from own unincorporated business (business income) there was a change in 2003-04 to ask about current income, rather than imputing the income on a 'no change' assumption from reported income for the previous financial year. In the 2002-03 results, the imputed total current business income estimate was $33.2 billion. This did not reflect the decline in the related national accounts series for mixed income of households (adjusted to deduct depreciation and interest payments). In 2003-04, respondents reported business income amounts earned in 2002-03 at $28.0 billion, and current income in 2003-04 at $31.2 billion. The year on year movement between the current and previous year business incomes reported in 2003-04 is broadly in line with the related national accounts series. The difference between the imputed amount for 2002-03 and the subsequently reported amount for that year contributes about -$13 to the change between the results for 2002-03 and 2003-04 in average gross weekly household incomes.
Together the improved methodology for both investment income and own unincorporated business income largely offset each other in mean income terms when comparing 2003-04 to 2002-03.
Income from superannuation, annuities or allocated pensions (other than government benefits such as the age pension) is higher in 2003-04 than in 2002-03. These amounts have been recorded in previous survey cycles as current income amounts, and no explicit change in methodology affects the reporting of these values. However, it is possible that changes in non-response have impacted on the series, or that the reporting of superannuation assets in conjunction with income has improved the quality of reporting. The increase in gross weekly household incomes from superannuation etc. of about $7 between 2002-03 and 2003-04 is statistically significant (the difference is about 3 standard errors). It is also likely that the reported superannuation assets are an underestimate of the total value of these assets and it may be that the superannuation income series, although higher than previously reported, is still a somewhat conservative measure.
IMPACT ON SUMMARY MEASURES
Inspection of the Gini coefficient and other income distribution measures draws attention to the question of whether all these changes have introduced significant discontinuities into the time series of the summary measures.
Table 5.1 Inequality measures(a)
Share of lowest
Share of second and
third income deciles
Share of third
Share of highest
|(a) See Section 1.6 and Appendix 3 for a description of these measures. |
As can be seen from the inequality measures shown in Table 5.1, there appears to have been a large decline in measured inequality in 2003-04. Some of this decline is likely to be a result of the methodological changes introduced in 2003-04 and some a result of factors that may have caused a genuine decrease in inequality.
Table 5.2 illustrates the impacts of those factors for which the summary measures can be re-estimated for 2003-04. These are further discussed below.
Table 5.2 Impact of inequality measures
|Initial estimate |
|1. Based on whole sample |
|2. Based on HES sample |
|3. Based on SIH only sample |
|Impact of changes in survey |
|4. Revert to old methodology for current period business and investment income |
|Impact of real world changes |
|5. Remove one-off payments to families and carers (otherwise as for line 4) |
|6. Impose personal tax regime of 2002-03 (otherwise as for line 4) |
IMPACT OF SURVEY CHANGES
New data and methodology for current period business and investment income
The one factor which allows precise quantification is the introduction of reported current year income for own unincorporated business income and investment income. It increases mean household income by $16.54 (or 1.5%) per week. The impact on the income distribution measures can be seen by comparing lines 1 and 4 of Table 5.2, with the Gini coefficient increasing from 0.294 to 0.297 and P90/P10 increasing from 3.70 to 3.72. However, P80/P20 and the income share of the second and third equivalised disposable household income deciles do not change.
While reported current business and investment income may contain an overly optimistic assessment from some respondents (especially those interviewed early in the financial year), the results of the new methodology appear significantly more credible than the estimates based on the old methodology. In particular, it reduces the number of households with gross household income below $20 per week by 55,000 (69%). Therefore it is believed that the new methodology is significantly superior.
But clearly the introduction of the new methodology leads to a series break. It is not known whether the difference between the two methodologies would be of the magnitude shown here for all years, or whether it would vary substantially with changes in the business cycle.
Integrating SIH and HES, and SIH no longer being run using the outgoing sample of the MPS
Lines 2 and 3 of Table 5.2 decompose line 1 into results based on the HES sample only and results from the SIH only sample. The differences in the Gini coefficient and the income shares going to the third quintile and the highest quintile are statistically significant at about the 90% confidence level. This implies that there is a high probability that including HES with the SIH is having a noticeable impact on the income inequality measures.
If there is a substantial impact on the measures from combining the HES and the SIH, there is also likely to be a substantial impact from moving from an MPS-based survey to a survey of dwellings not recently included in an ABS household survey. Differences in non-response rates also indicate that this is likely. Non-response in the SIH only sample of 2003-04 is significantly less than for the 2002-03 SIH and for the 2003-04 HES sample. However, the high non-response in the HES sample is not necessarily for the same reasons as the non-response experienced in the previous MPS-based income surveys, and therefore the non-response impact is likely to be different between the two.
Changes in imputation and editing practices
The major identifiable element of this aspect is the loss of capacity to impute using MPS information. Such imputation was particularly important for lone person and single parent households. But there is no obvious way to quantify the impact.
IMPACT OF REAL WORLD CHANGES
One-off payments to families and carers
The one-off payments to families and carers has had a substantial impact on income inequality. These payments were modelled in SIH. A comparison of lines 4 and 5 in Table 5.2 shows the impact. Without the one-off payments, the Gini coefficient would be 0.007 higher, P90/P10 would be 0.009 higher, etc.
In 2003-04, marginal rate thresholds were increased for personal income tax, and there were other changes to other parts of the tax system such as the aged person's rebate. A comparison of lines 4 and 6 in Table 5.2 shows the results of retaining the 2002-03 tax rates for 2003-04 income. There is not a significant impact on income inequality.
There are substantial changes in the income inequality measures between 2002-03 and 2003-04.
The Gini coefficient declined from 0.309 in 2002-03 to 0.294 in 2003-04, a drop of 0.015. The previous largest annual movement was 0.011, between 1996-97 and 1997-98. The P90/P10 ratio declined by 0.30, with the previous largest annual movement being 0.11. The other measures had changes of similar magnitude.
Adjusting the income estimates to retain constant methodology for current year business and investment income reduces the difference in the Gini coefficient to 0.012 and the difference in P90/P10 to 0.28, still greater than any historic change.
On their own, the one-off payments to families and carers have resulted in the Gini coefficient being 0.005 lower than it would otherwise have been, and P90/P10 being 0.09 higher.
These two factors do not entirely explain the change in inequality. Therefore it is difficult to assess the changes in income distribution over time. However, it appears that there has been no significant change in income inequality from the mid 1990s to 2003-04.