The Proposed GST: Impact on the Distribution of Income

Current Issues Brief 7 1998-99

David Richardson
Economics, Commerce and Industrial Relations Group
30 March 1999

Contents

Introduction

Distribution Effects of Indirect Tax Changes

Inflation Impact

Exempting Food

Conclusions

Endnotes

Introduction

All participants in the current tax debate would acknowledge that a pure Goods and Services Tax (GST) would be regressive-that is, it would involve a proportionately higher burden on lower income earners than higher income earners. A study carried out by the International Monetary Fund put it:

For that reason the Government has designed a compensation package to offset the regressive impact of the GST. The purpose of this paper is to examine the burden likely to be imposed by the indirect tax package, in particular the proposed GST offset by the various indirect tax measures proposed in the Government's tax package.(2) The use of the word 'burden' in reference to taxation follows standard practice among economists discussing taxation arrangements and is not meant to convey the impression that a particular tax or tax arrangement is 'good' or 'bad.'

The compensation package itself is not examined here, nor are the changes to direct taxes that will affect the distribution of income. Compensation arrangements are set out in the documentation the Government took to the 1998 election.(3) However, to judge the adequacy of the compensation arrangements requires the type of analysis of distributional impacts that are presented below.

We start by looking at some general aspects of the impact of the GST on income distribution. A major effect of the GST on income distribution will come from the fact that, on average, low income earners spend more than they earn while high income earners spend less than they earn. The only data on income and expenditures for different income groups comes from the household expenditure survey (HES) undertaken by the Australian Bureau of Statistics (ABS). This particular data source is said to have problems. The Department of the Treasury warns that we should not make direct comparisons between income and consumption using these data. Those data problems are examined below. The approach here is to do the exercise with the household expenditure data and then consider how useful the exercise is in light of the problems with the data. It is also worth reflecting on the fact that while the HES has been said to have problems for many years, no-one has come up with anything better. Neither the Department of the Treasury nor the ABS have commissioned other surveys that might have overcome the deficiencies in the HES. With these qualifications in mind the distributional issues can be addressed.

Distribution Effects of Indirect Tax Changes

As a first approach to this question, the estimated price effects due to the indirect tax package are added to total goods and services expenditures to determine the impact on different income groups. Table 1 presents the results of that exercise using the Government's figure for the price impact on final consumption expenditures of 2.2 per cent.(4) The actual inflation estimate is discussed further below.

Table 1: Household Expenditure Survey: Distributional impact of the GST/indirect tax package

Table 1 shows that the GST impact is equivalent to a burden of 4.4 per cent of income for the bottom 20 per cent of households. That compares with 1.4 per cent of income for the top 20 per cent of households. The major differences in the burdens likely to be experienced by the bottom and top quintiles are due to the enormous differences in the HES's estimates of the savings of the two groups. The HES has the bottom quintile spending 200 per cent of their income while the top quintile spends 61.8 per cent of their income.

There are some reasons for thinking that these figures overstate the difference between spending and income for the lower income group. Some of the data problems can be understood by considering the following example. Consider a family whose breadwinner experiences a sudden bout of unemployment that lasts six weeks. Before they experience the unemployment the family is a middle income household, perhaps in the third quintile. With the unemployment, income goes right down to the first quintile. However, that family is likely to want to try to roughly maintain its consumption standards in the expectation/hope that the unemployment will prove temporary. If now during that spell of unemployment we sample that family in a household income and expenditure survey we will find they are earning near the bottom of the income distribution but spending near the middle of the spending distribution. Lower income groups contain people who are normally in a higher income group and spending according to the norms of higher income groups. The effect of those type of households is to bias the results towards overstating the dissavings of lower income groups

There are also a number of statistical problems raised by Treasury.(5) The ABS adjusts for purchases and sales of housing and other durables by treating them as positive and negative expenditures respectively. Over a large enough sample that should give a good estimate of net expenditures on a flow basis. The more the data are disaggregated the more the estimates are subject to large statistical errors. However, in this paper the total survey is only split into five income quintiles as in the published version of the HES. Hence this paper should not be subject to the sort of criticism that might be levelled at studies that rely on more disaggregated data made available by the ABS. A more serious problem for our purposes is the charge that the pensions and benefits identified by the HES do not add up to a total consistent with Department of Social Security figures.(6)

In principle the under-reporting of pensions and allowances can be adjusted for when using the HES. For example, the HES reports all households, on average, received pensions and allowances of $94.02 per week. That implies total Social Security and other income support payments of $32.351 billion in 1993-94. According to the Budget Papers, total personal benefit payments were $41.084 billion in that year.(7) This is a 21 per cent under-reporting. Given the stigma of reporting social security payments in a survey context that seems a reasonably good success rate. Nevertheless, we should be able to make some useful adjustments to the data to compensate for this known and quantifiable under-reporting. Applying a 21 per cent adjustment alone would mean that dissavings (dissavings is the opposite of 'savings' and simply refers to people spending more than they earn) rates for the bottom quintile come down from 100 per cent of income to 75.2 per cent. It should be possible to make other similar adjustments to the data to correct for other known biases arising from the survey.

In the Draft White Paper prepared before the 1985 tax summit, the Department of the Treasury used its own guesstimates, low income groups were assumed to spend 10 per cent more than their income while high income earners saved 15 per cent of their disposable income.(8) In the discussion below this set of assumptions are referred to as the '1985 assumptions'. In the context of the present debate the Department of the Treasury does not appear to have allowed for the different savings rates and the implications they might have for the distribution of the burdens of the GST. The present position put by Treasury simply says:

• Set the saving rate to zero.
• Use the population-wide saving rate implied by the National Accounts data.(9)

The implication is that because the HES data have problems, and despite the importance of the issues it raises, we should ignore the high dissavings rates among lower income groups and dismiss distributional issues by also ignoring any differences among income groups.

If we use the 1985 assumptions in Table 1 above the GST impost is 2.4 per cent of income for the bottom 20 per cent and 1.4 per cent for the top 20 per cent. That represents a substantial difference in the burden of the price effects of the tax package. The package is still regressive but with a much more moderate impact on the lowest income group.

There may be other biases against low income groups. The above discussion has assumed that it is appropriate to use the same estimated price effect for both low and high income groups. In the 1985 Draft White Paper Treasury found that low income groups would, on average, experience higher price increases than the average expected from a GST. It specifically mentioned food and clothing as examples where the lower income groups spend a disproportionate share of their income and where the GST would have a disproportionate effect. As a result, using the aggregate consumer price index (CPI) understates the impact on lower income groups. As an example, Treasury presented estimates suggesting that a pensioner would experience a CPI increase 1.6 percentage points higher than a single person.

Economics Professor, John Nevile from the University of NSW, estimates that when all corrections are made for differences in consumption patterns between those in the bottom 20 per cent of the income distribution, the inflationary impact goes from the Government estimate of 1.9 per cent to between 3.5 and 4 per cent.(10)

Inflation Impact

In the previous section the Government's estimate of a 2.2 per cent increase in the weighted average price of final consumption expenditures was used as our measure of the inflationary impact. However, that is not exactly the same thing as the CPI. Elsewhere the Government estimates that the CPI will increase by 1.9 per cent as a result of the measures it proposes in the package. That estimate ignores the inflationary effects of the additional taxation on tobacco products on the grounds that 'the impact of the GST on tobacco prices should not, for public health reasons, be offset by income tax cuts and/or increases in social security payments'.(11) It is unlikely the Australian Statistician would take the same view. According to the Government documentation, the planned increases in tobacco taxes would increase private final consumption expenditures by 0.27 per cent and would most likely have virtually the same effect on the CPI. That confirms the use of 2.2 per cent in the calculations used here.

At 2.2 per cent the Government estimate of the price impact on consumption is relatively modest. That is because the 10 per cent GST is offset by reductions in other indirect taxes. The Government assumes any reduction in business costs is passed on in full to the consumer. Of course, by saying that changes are passed on in full, it would be inconsistent to suggest that business might gain from cheaper inputs. Business could only be better off if they keep some of the cost reductions to boost their own incomes. Hence it cannot be said that there is both full pass on to consumers and that business would be better off. However, the effect of a less than complete passing through of the tax reductions can be examined.

Colin Hargraves of the Economic Modelling Bureau of Australia looked at the effect of a 50 per cent passing on of the business cost reductions. He found the CPI would increase by 5.95 percentage points as a result of the indirect tax changes.(12) That would translate into a burden equal to 6.6 per cent of the income of the low income group and 3.7 per cent of the income of the high income group under the 1985 assumptions. Of course there are likely to be regional differences. Business in the regions of Australia is less competitive than Australian business generally. Therefore it is possible for the price effect on lower income groups in regional Australia to look much more like Hargraves' figures than the Government's figures.

The official estimate of the inflation impact has also been questioned by various people who gave evidence to the Senate Select Committee on a New Tax System. As is often the case, the assumptions behind the estimates are very important. It is assumed that wage increases do not follow the higher prices caused by the GST. The package states that:

The other main assumptions, as discussed above, are that all changes to indirect taxation-the GST less cuts in other indirect taxes-are exactly passed on to purchasers and, ultimately, to consumers. The changes are assumed to be passed on in full, no more and no less. Reinforcing this the Government has announced an intention to use the Australian Competition and Consumer Commission to 'take action under its Act against unfair business practices that adjust prices in a manner that is not consistent with changes in tax rates'.(14) It is worth recalling, however, that the old Prices Surveillance Authority (PSA) closely examined an analogous case, the extent to which the 1988 appreciation of the Australian dollar was passed through into consumer prices. It was forced to conclude that 'the full benefits of the 1988 appreciation of the Australian dollar have not been transmitted to consumer prices'.(15) In particular, instead of passing on the benefits of appreciation to consumers in full, the lower import prices were also reflected in increased profit margins. Unfortunately the PSA was not able to quantify the extent to which cost reductions were passed through to consumers.

Finally it needs to be stressed that the Government CPI estimate is the long run estimate of the impact on the CPI.(16) The Senate heard evidence to the effect that the first year impact effect of the package would be greater than the long run effect. Treasury officials said the CPI impact in the first year could be 3.1 percentage points.(17)

Exempting Food

The proposed GST has raised a good deal of debate about the merits or otherwise of taxing the 'necessities of life', especially food. Most countries that impose a GST (or Value Added Tax) give favourable treatment to food. Food is either subject to a lower GST or zero rated as in Canada, Ireland and the UK.(18) In this section the distributional implications of exempting food are examined. It is assumed here that the proponents of exempting food have in mind zero rating food. That is, a seller of tax-exempt food would nevertheless be able to claim GST credits on the GST paid on any inputs into the production of food. Taxing the 'necessities of life' also raises ethical and moral issues for many commentators. Those are of course beyond the scope of the present paper.

Food is defined here as the household expenditure items 151-237 in the ABS household expenditure survey.(19) It should be noted that if indeed there was a decision to exempt food then the actual definition of food would be a policy question. However, the approach here is to examine the effect of including items 151 to 237. Excluded are 'meals out and take-away food' and 'alcoholic beverages.' The food group obviously includes all the basics such as milk, bread and meat. Exempting food from the GST would produce a reduction in the price of food as the effect of abolishing other indirect taxes on inputs into food production flows through to the final price of food. According to Government estimates, the price of food under the indirect tax reform package would increase by 4.4 per cent.(20) Deducting the 10 per cent GST means that the other indirect tax reforms, on their own, would reduce food prices by 5.6 per cent. The effect of exempting food is given in Table 2.

The table also illustrates how the regressivity of the GST would change if food were exempt and if the GST rate was increased to make good the loss in revenue. As a consequence of not imposing the GST on food, the new GST rate would have to be 11.96 on our estimates.(21) This means the price increase expected on all non-food items would be 4.16 per cent.

Table 2 GST burdens with Food Exempt

Recall that with food included, as shown in Table 1, the GST burden on the lowest income group was 4.40 per cent of income compared with 1.36 per cent for the highest income group. By excluding food and no other changes the burden on the lowest quintile is more than halved down to 2.04 per cent and the burden on the top group declines from 1.36 to 0.88 per cent. Hence while the burden would be substantially reduced on all income earners, exempting food would still leave us with a regressive tax change. The final row of Table 2 estimates the effect of exempting food but increasing the GST rate on non-food items to compensate for the exemption for food. Under that scenario the burden on the low income quintile is even greater than in the original case. The burden of the changes is now 5.37 per cent on the bottom quintile. The change is still regressive and the impact on low incomes is even larger when food is exempted but rates increased on non-food items. It seems, according to the HES data, that the bottom quintile spends so much on all items that exempting even such a major item as food still leaves us with a very regressive change when the net indirect tax burden is increased.

Unfortunately these conclusions depend critically on the validity of the HES data. Given the above criticisms of the HES there is a strong possibility that the conclusions in the above paragraph may be upset by better data. Accordingly the above calculations are repeated for the bottom and top quintiles using the assumptions that the bottom spend 10 per cent more than they receive and the top spend 15 per cent less than their after tax income. In the absence of any other information it is assumed that the HES gets right the amount spent on food by the bottom and top quintiles. In this way the affect of exempting food under these assumptions can be examined.

Table 3: GST Burdens: when bottom quintile dissaves 10 per cent and top quintile saves 15 per cent of after-tax income

This table illustrates some important results. The GST including food remains regressive in that it imposes a higher proportionate burden on the bottom quintile. The GST without food (second case) imposes a trivial burden on the lowest quintile, less than 0.1 per cent of income while imposing a burden of almost one per cent on the highest quintile. This case is actually progressive in that a higher proportionate burden is imposed on the highest quintile. By exempting food but increasing the GST on non-food items (third case) the result is also progressive albeit to a lesser extent.

The above conclusions must remain tentative, not least because they contradict the earlier results. The results in Table 2 based on HES data show that the GST is highly regressive even when food is excluded. The reason is that according to the HES the lower income groups spend such a high proportion of their income on all goods and services that exempting even major items of expenditure by low-income groups still leaves a heavily regressive tax.

It is thought that the HES exaggerates the dissavings of lower income groups, but no one knows by how much. Hence there is no way of knowing whether the true position of lower income groups is closer to the results reported in Table 2 or those of Table 3.

Conclusions

Any conclusions must be tentative. The GST and related measures would have a significant impact on the real income of different income groups. However, the only data that allows modelling of the effects is the household expenditure survey which gives detailed expenditure estimates according to income groups. Using those data shows that the bottom 20 per cent of the population would experience a substantial burden equivalent to 4.40 per cent of their income while the top quintile would experience a more modest burden at 1.36 per cent of their income. Other quintiles fall in-between these two limiting cases with the proportionate burden falling as income increases. However, we know there are problems with the data and some mistrust their use given the apparently unrealistic dissavings among lower income groups. One argument which could be seen to undermine the Government's position could be that Treasury appears not to have adjusted for the problems in the data, preferring instead to ignore distributional complications in looking at the impact of the indirect tax package.

In exercises such as this, attention is mainly on lower income groups. This paper has concentrated on the bottom quintile. We can state our results fairly briefly if we confine our attention mainly to that group. First, it seems that, using the HES, the proposed indirect tax changes will impose a much higher proportionate burden on the bottom quintile, a 2.2 per cent price effect becomes a 4.4 per cent of income burden on the low income group. The burden on the bottom quintile is more than halved if food is exempted. However, the burden is even higher if the GST is increased on non-food items to pay for the exemption of food.

The HES has problems, as perhaps do most surveys. In the absence of any agreed means of compensating for the HES deficiencies, one approach is to repeat the assumptions used in the 1985 Draft White Paper on tax reform. That assumed the bottom group spent 10 per cent more than it earned while the top quintile saved 15 per cent of after tax income. Under those assumptions the GST package is still quite regressive. However, if we exempt food the package becomes progressive, the proportionate burden increases on higher income groups and falls to a trivial amount for the bottom quintile. If the non-food GST rate is increased to make good the revenue lost through the exemption of food, then the GST is again regressive, though much less so than under the HES assumptions.

Endnotes

2. A. A. Tait, 'VAT policy issues: Structure, regressivity, inflation, and exports,' in A.A. Tait (ed), Value-Added Tax: Administrative and Policy Issues, International Monetary Fund, Washington DC, October 1991, p. 5.

3. Tax Reform-Not a New Tax: a New Tax System, August 1998. The indirect tax changes referred to are set out and discussed in chapter two, which is headed 'Security: Reforms to indirect tax and State finances'.

4. ibid.

5. That is distinct from the more widely publicised estimate of a 1.9 per cent increase in the CPI (adjusted to remove the effect of increased tobacco prices).

6. M. Carnahan, 'Does demand create poor quality supply: A critique of alternative distributional analyses', Economic Roundup (Treasury), Spring 1998, pp 37-50.

7. ibid.

8. 1994-95 Budget Paper No. 1, Budget Statements 1995-96, Canberra, 9 May 1995.

9. Reform of the Australian Tax System: Draft White Paper, Canberra, AGPS, June 1985.

10. M. Carnahan, op. cit., p. 49.

11. J. W. Nevile, 'The Coalition's tax package: A failure on Mr Howard's own criteria', in J. Disney, J. Nevile, J. Quiggin and J. Smith, eds, Perspective on the GST Package, Australia Institute Discussion Paper, no. 19, September 1998, pp. 11-20.

12. Reform of the Australian Tax System: Draft White Paper, op. cit., p. 162
13. C. Hargraves, 'Some comments on the tax plan modelling', at website http://www.anu.edu.au/emba/GSTpaper.html (last visited 18 September 1998).

14. ibid, p. 157

15. ibid, p. 85.

16. Prices Surveillance Authority, Inquiry into Effects of Exchange Rate Appreciation on Prices of Consumer Goods, Report no. 21, 20 May 1989, p. xvii.

17. See for example, C. Murphy, 'The long-term economic effects of the Government's tax plan - Modelled using MM303', September 1998.

18. Report of the Senate Select Committee on a New Tax System, at website http://wopablue/senate/committee/gst/report1/chapter2.htm

19. N. Warren, 'Food: Staple of life or staple of the GST?' paper given to ATAX Conference, 'The GST law: Key issues for business and administration,' Sydney, 16 December 1998.

20. Australian Bureau of Statistics, Detailed Expenditure Items: 1993-94 Household Expenditure Survey, Australia, Cat No 6535.0, 29 February 1996.

21. Tax Reform-Not a New Tax: a New Tax System, August 1998, p. 158.

22. Treasury estimates the cost of exempting food at $4.454 billion in 2000-01 compared with total GST receipts (including food) of $27.20 billion. John Quiggin's estimate is 12.0 to 12.5 per cent. See transcripts of Senate Select Committee on a New Tax System, Hansard, 2 February 1999 and 3 February 1999 respectively.