CHAPTER 2

ECONOMIC MODELLING

Macro Economic Effects

2.1 The Select Committee on a new Tax System was established to examine the impact of the Government's proposed New Tax System. The measures in that proposal are expected by the Government to deliver substantial long-term benefits in the operation of the economy, such as higher economic growth, better export outcomes (supply costs reduced by 3.5 per cent), a reduction in industry costs (by an average 3.2 per cent), reduced private investment costs (nearly 7 per cent) and costs of government reduced by about $1 billion per year. 1

2.2 The first report of the Select Committee concerned itself with the technical assumptions underlying the Government's modelling of the proposals for tax reform. The Committee took evidence from several of Australia's best-known economic modellers about the economic theories, assumptions, calculations, projections, estimates and modelling which produced those proposals. 2

2.3 The Committee spoke also to Professor Ken Wallis, the Director of the ESRC Macroeconomic Modelling Bureau at the University of Warwick in England. Professor Wallis gave the Committee a private briefing on the principal issues and difficulties involved in large scale economic modelling.

2.4 To achieve a balanced view of the Treasury's economic modelling results, the Committee commissioned three modelling projects to examine the likely economic outcomes of the introduction of the Government's proposed tax package.

2.5 Professor Peter Dixon of the Centre of Policy Studies at Monash University was asked to analyse the likely outcome of the introduction of the new package, using his MONASH model. Mr Chris Murphy of Econtech was engaged to review Professor Dixon's results using his Murphy model.

2.6 The third project was an introductory paper on Microsimulation modelling prepared by Professor Ann Harding of NATSEM at the University of Canberra and Professor Neil Warren of ATAX at the University of NSW.

2.7 Professor Dixon's central simulation produced findings that:

the long-run resource allocation gains flowing from the proposed tax changes will be negligible;
the package will harm tourism and education exports and benefit most traditional exporters, e.g iron ore;
the effects on consumer-good industries will be mixed;
employment will be stimulated in the short-run by about 30,000 jobs due to the net fiscal stimulus from the reduced budget surplus, not from the GST;
investment will be increased, especially in the short-run; and
the package will produce a long-run increase in the capital stock in Australia, but little change in economic welfare.

2.8 In addition to the central simulation, Professor Dixon conducted six sensitivity simulations. The findings of these additional simulations strengthened the findings from earlier assessments that the proposed tax changes will have little effect on Australia's long-run macro-economic performance.

2.9 The findings revealed two short-run risks: that the package will cause job losses in the short-run if wage earners refuse to allow before-tax wage rates to fall relative to the CPI or if increases in indirect taxes are passed on more quickly than reductions. MONASH simulations also produced no evidence to support the Treasury position that a major tax mix change is necessary because the present tax regime will raise insufficient revenue to meet Australia's future needs.

2.10 In his review of the MONASH modelling results, Chris Murphy commented that this MM303 model and the MONASH model are not strictly comparable. MM303 provides long-run estimates while for this project MONASH gives progressive estimates down to the year 2007-8.

2.11 There is a difference also in the complexity of the models, with MM303 identifying 303 different goods and services and MONASH about 100. Murphy argued that this extra detail helps to give a better estimate of the gains achieved by a broad based consumption tax like a GST. 3

2.12 The predicted CPI effects from the two models showed 2.2 per cent for MONASH and 0.9 per cent for MM303. Murphy, however, claimed that the MONASH number was derived from Treasury price effects recalculated in the MONASH model. MM303's number arises from the modelling calculations and reflects the estimated long-run effects. 4

2.13 Another interesting point of comparison is the predicted wage and employment positions. MONASH indicated a sustainable gain in real after-tax wages of 1 per cent, MM303 predicted 1.8 per cent. Murphy broadly agreed with the MONASH central simulation that there would be a short-term gain in employment of 30,000 jobs in 2000-1. He also agreed that the jobs gain would largely disappear by 2007-8, as wages catch up. In the long-run national employment is fixed in both models as a modelling assumption. 5

2.14 Murphy disagreed with one of the MONASH wage assumption scenarios in which workers ignore income tax cuts and obtain CPI equivalent wage rises. He considered this implausible, especially in view of the trend to deregulated labour markets. Professor Dixon of MONASH considered this scenario to be plausible. He estimated that in this case the ANTS package could cost 100,000 jobs in the short term. 6

2.15 In considering industry production effects the two models produced broadly equivalent results. Because both models hold total employment at a fixed level, job shifting between industries must occur in the modelling results. In a comparison of the results, Murphy argued that neither model revealed ANTS as a significant force in job shifting. He argued that the results reflected a response to the evening out of the tax burden between industries. Professor Dixon 7 described the GST as "job destroying". 8

2.16 In assessing overall welfare gain, the MM303 model estimated an annual welfare gain of around $600 million per year, while MONASH estimated a loss of $30 million. 9

2.17 Murphy stated that the difference lay in the greater complexity and flexibility of MM303. The MONASH model he said did not independently model the existing and new tax systems because it lacked sufficient detail to do so. He also claimed that the process of estimating tax rate changes limited the results:
That means it is not possible directly to model the change from the old to the new tax system using Monash. Instead, the Monash modellers have relied on Treasury's estimates of the price changes due to ANTS and work backwards to try and figure out what adjustments they need to make to the summary tax measures in their model to generate the same kinds of price changes. This is a limitation of the modelling, but it does not mean the modelling is invalid. What it means is that the modelling of the change to the tax system with Monash is somewhat approximate. 10

2.18 Professor Dixon said that even $600 million is only a small number when taken in terms of the total Australian consumption of goods and services (0.2 per cent). He commented that if different (slightly stronger) terms of trade effects are assumed, even that amount disappears. He stressed repeatedly that the main point revealed by the modelling projects is that the overall economic efficiency impact of ANTS is small - either plus or minus but small. 11

2.19 The MONASH assessment of the results if food were excluded from the GST indicated a loss to revenue of $2.5 billion (based on a narrow definition of food) and a consequent reduction in the generosity of income tax cuts. There would also be marginally better short-term employment results. Overall this change would make the economy marginally smaller than it otherwise would have been. 12

2.20 On the question of making food GST-free, Murphy used a wider definition of food, excluding all but restaurants and takeaways. His results showed a revenue reduction of $4 billion and a reduction of 80 per cent in the revenue available to fund tax cuts. He estimated that the long-run change in the CPI would move from an increase of 0.9 per cent to a reduction of 0.5 per cent. 13

2.21 Chapter 5 discusses the effects of making food GST-free in more detail.

NATSEM report

2.22 The National Centre for Social and Economic Modelling (NATSEM) at the University of Canberra, was contracted to carry out an examination of the distributional impact of the Government's proposed New Tax System. The study was carried out by Professor Ann Harding of NATSEM and Professor Neil Warren of the University of New South Wales and was released by the Committee at a public hearing on 8 April 1999. 14 The report is attached at Appendix IV.

2.23 Evidence provided by Professors Warren and Harding at this hearing was that:

there are a number of "losers"; 15
compensation should be increased by more than 50% for pensioners and beneficiaries; 16
the analysis of Mr Geoff Carmody and Mr Mitch Hooke that pensioners' compensation will erode to zero within a few years is correct. 17

2.24 In addition, Professor Warren indicated that a positive distributional result in the NATSEM report for a particular household group did not indicate that all of the people within that particular group were better off:

2.25 During the project NATSEM modelled ten different scenarios. The first of these attempted to replicate Treasury's methodology in preparing the ANTS document. The second also closely followed the original but with specific changes to some major parameters. In almost all instances, the results obtained in these cases were very close to Treasury's results.

2.26 In the other eight scenarios, at the Committee's request other feasible assumptions were introduced to test their effect on the overall distributional outcome. Changes of particular interest to the Committee were: the possible exclusion of food from the GST; the effect of changes on tobacco and the price of new homes; and the effects produced when cost changes resulting from the abolition of WST are passed on into consumer prices at differing rates. The project also examined the effects of seeking to balance these changes by various means, including increasing the GST rate, the provision of increased compensation, and/or a reduction in the scope of the income tax cuts.

2.27 The modelling was carried out on the STINMOD-STATAX microsimulation model, which according to NATSEM bears a close resemblance to the PRISMOD model used by the Treasury. It should be noted that full details of the PRISMOD model have not been released into the public domain by Treasury.

2.28 In presenting their findings, the two authors warned that economists' models involved in these calculations are highly complex attempts to represent mathematically the even more complex economy. Consequently, the results produced are attempts to indicate the likely impact of tax changes - they do not provide absolutely precise measures of outcomes, but show the likely direction and approximate magnitude of changes:
What we would emphasise is that you should not treat any of these numbers as other than broadly indicative. I do not think you can say, for example, that a household of this particular type will be better off by exactly $2.10 a week, which is what could be implied from a first reading of the tables. The science of the modelling is not accurate down to that sort of detail. What you are looking for is where there appears to be possible potential problem areas, in particular with regard to the adequacy of the compensation. 19

The Scenarios

Option 1. Option 1 is the ANTS cameos replicated, using the same methodology and assumptions as adopted by the Treasury;

Option 2. The impact in July 2000 of the Government's base package with 100% pass through of both indirect tax cuts and increased costs in the first year, assuming: (a) a single CPI price effect but including price effects on new housing and tobacco, and (b) disposable income equals consumption;

Option 3. Option 2 but with: (a) cameo specific price effects, inclusive of housing and tobacco price effects, and (b) hypothetical dissaving ratios for different income groups, which vary within the cameos;

Option 4. Option 3 but with 70% pass through of indirect tax cuts in the first year, but with 100% of increased costs passed through in one year;

Option 5. Option 3 with Food GST-Free but with the cost of this change funded through a progressive adjustment to the personal income-tax rate schedule;

Option 6. Option 3 with Food GST-Free and with the cost of this change funded through increasing the GST rate to 12 per cent;

Option 7. Option 3 but with the cost of increased compensation funded through adjustments to the personal income-tax rates and cameo specific price effects exclusive of housing and tobacco price effects.

2.29 Three additional options were modelled to demonstrate the effects of further variations to some of the above scenarios:

Option 3b: Same as Option 3, but excluding the price effect on tobacco and the price effect on new houses and with the compensation rate for pensions and allowances increased to 3.5 per cent;
Option 4b: Same as Option 4, but excluding the price effect on tobacco and the price effect on new houses and with the compensation rate for pensions and allowances increased to 4.2 per cent; and
Option 5b: Same as Option 5, but excluding the price effect on tobacco and the price effect on new houses and with the compensation rate for pensions and allowances reduced to 2.3 per cent and greater tax cuts.

Indicated Results
2.30 Option 1: This option used the same key distributional assumptions about the CPI effects of the tax package, as were used in the Treasury's modelling. The results reflected this, with the overall price effect estimated at 2 per cent, compared with the Treasury's estimate of 1.9 per cent. This is the price effect for 2001-2 but, as in the cameos in ANTS, it has been assumed to apply from July 2000 in the distributional analysis tables.

2.31 The results of the 29 distributional tables for various household types are little different to the Treasury results. Some variations are inevitable because of the more recent projections of average weekly earnings and CPI that are now available. Although nearly all of the estimates are very close to the Treasury results, the estimated net gains overall are marginally reduced by the slightly higher CPI effect in NATSEM's results.

2.32 Option 2: Inclusion of the CPI effect of changes in the prices of new homes and tobacco, are estimated to increase the price effect from 2 per cent to 2.7 per cent. NATSEM noted that this figure could be a little high because they were unable to model the positive effects of the New Home Owners Scheme or the probable appreciation in the value of existing homes resulting from it.

2.33 When run against the 29 cameo household units, with the increase in pensions and allowances still set at 3.4 per cent (as in Option 1), the results show that for those fully dependant on social security, their position is not as good as under Option 1 - the estimate is that they would be 0.7 per cent better off after introduction of the tax package. Overall, however, while there are no losers under Option 2, as in Option 1, the actual spending patterns of these households are not considered.

2.34 Options 3 and 3b: These options vary two of the controversial assumptions included in the Treasury modelling. One is reliance on the CPI as a gauge of total price effect and the other is the assumption that there is no dissaving. The two options differ in their treatment of the new homes and tobacco price effects - these are included in Option 3 and excluded in Option 3b.

2.35 The data on consumption patterns used in these Options are drawn from the HES data set. NATSEM while regarding this as the appropriate data set pointed to the fact that it can only provide a broad indication of effects. The HES sample size was not large enough to be fully disaggregated and NATSEM has used averages for each population sub-group. If each sub-group had been divided into quintiles, for example, some would have become too small to be statistically valid:
In essence, the Committee faced a trade-off here between having analysis done by quintiles and having the diversity of the 29 income groups that you selected. So, for example, for a single income couple with one child aged between five to 12 years, there were only 28 such single income unit households in the household expenditure survey, so it would not have been possible to take that group and break them down into quintiles because there would be only three or four observations in each quintile and the results would be statistically completely invalid. So ot was a trade-off between doing it by quintiles and income levels and having this diversity of household types. You cannot have both. 20

2.36 In applying household specific price effects therefore, the study assumes the same consumption pattern within sub-groups but allows differences in the patterns between groups.

2.37 NATSEM reported that there was no time for an examination of the extent of dissaving among various Australian households to be investigated. The authors have therefore adopted an arbitrary measure of 10 per cent as the assumed maximum rate of dissaving by lower income households.

2.38 The overall CPI effect for Option 3 is the same as Option 2, i.e. 2.7 per cent. However, when the different spending patterns for various household types are taken into account, the CPI results range from a high point of 3.8 per cent (age pensioner couples) to a low point of 1.7 per cent (sole parent, 1 child under 5 years).

2.39 If the assumptions are varied to allow for imputed dissaving, the results change substantially. In the case of single age pensioners, for example:

at the general CPI effect of 2.7 per cent and no dissaving - a gain of 0.7 per cent or about $1.36 in disposable income;
at the same CPI effect but with 10 per cent dissaving (based on Treasury estimates made in the 1985 Draft White Paper on Tax Reform) - the gain turns to a loss of about $1.39 per week;
at the same CPI effect but with dissaving at 5 per cent - the loss is reduced to about $1.02 per week.
These estimated outcomes illustrate the fact that the results for low income earners particularly, are very sensitive to the assumptions applied to price effects and dissaving.

2.40 Treasury's estimate of the CPI effect for age pensioners was 2.4 per cent, more than 25 per cent higher than the general effect of 1.9 per cent. The STATAX model indicated an average effect on pensioners of 3 per cent, compared to 2 per cent generally. The results from STATAX indicated then an additional burden on pensioners which is about double the Treasury estimate.

2.41 In summary for Option 3, on average only aged and disability-support pensioner couples fully dependent on social security are estimated to experience a decline in their incomes if the ANTS package were introduced using Option 3 assumptions and the benefits claimed in ANTS are generally shown to be overstated for most low income groups. If the price effects from new home prices and tobacco are removed then these groups would become marginal winners.

2.42 For Option 3b the results indicate that maximum rate age pensioners and disability support pensioner couples would be very marginally better off under the tax package using the assumptions of Option 3b. However, in evidence before the Committee Professor Harding cautioned:

2.43 Options 4 and 4b: These Options seek to examine the price effects if some businesses delayed the passing-on of price reductions from lower indirect tax rates. The CPI effect in this case would be substantially higher than for the earlier Options - NATSEM estimates 3.6 per cent, or 0.9 per cent higher than Options 2 and 3.

2.44 The question then arises, what would the Government do about the level of pension/compensation increases? If it were left at 3.4 per cent (Option 4), the result would be price rises which in many cases exceeded 3.4 per cent and there would, consequently, be losses.

2.45 If the Government responds with a further pension increase, however, (Option 4b) the results become less clear-cut and more difficult to assess properly. The Harding/Warren paper does not attempt a summary of these results. The paper asserts, however, that the important issue is that in year 2 the CPI can be expected to be reduced further than the assumption in ANTS indicates, because of the flow through of the remaining price reductions.

2.46 Options 5 and 6 (Food excluded): In Option 5 the revenue shortfall from the exclusion of food is made up by reducing the generosity of income tax cuts. In Option 6 the same effect is achieved by increasing the GST rate. In both cases restaurants and takeaways were not included in the definition of food and were retained in the GST tax base.

2.47 If food were removed from the GST base, the CPI effect (measured with new houses and tobacco included) would be reduced substantially. The estimated CPI impact of ANTS (2.7 per cent) would be cut to 1.6 per cent - at the cost of $5 billion in revenue in 2001-2.

2.48 Low income households would be the main beneficiaries from the exclusion of food from the GST. The price index for these households would fall at nearly twice the average rate. Although low income households gain proportionately more, high income households also benefit.

2.49 NATSEM used two methods to model the recovery of the lost revenue on zero rated food products:
" the first was to increase tax rates. The proposed method was to increase the tax rates on incomes between $20,001 and $50,000 by 3.5 per cent and between $50,001 and $75,000 by 3 per cent.
" the alternative method was to increase the GST rate from 10 per cent to 12 per cent. The difficulty with this alternative is that it would increase the CPI effect also, because of the weighting of products in the CPI scale.

2.50 Comparing Option 5 with Option 3 shows gains for lower income households generally. Both CPI and household-specific price effects are lower than for Option 3. Overall, the effect of Option 5 is more progressive than Option 3, it transfers money from higher to lower income groups.

2.51 The major winners are age pensioners and disability support couples - those who would be expected to have losses under Option 3. These groups become net winners under Option 5; about 1.4 per cent for those fully dependant upon social security and without private income.

2.52 Similarly, under Option 6 age pensioners are relatively big winners, because they spend such a large proportion of their income on food. Despite this, however, Option 5 still delivers more assistance to age pensioners than Option 6.

2.53 For other household types, there is little difference between the distributional outcomes of Options 3 and 6. Overall, about half the groups modelled are losers on this change and the other half are winners. Single people are mainly losers.

2.54 Option 5, on the other hand, delivers a much better result to lower income families with children than Option 3. Its effect is much more progressive. Option 6 is unsatisfactory for helping families with children, it delivers much better results for age pensioners.

2.55 The general result for options 3, 5& 6 is to:

assist single income families with children more than two income families with children;
assist families with children more than families without children;
Options 3 and 6 have their benefits skewed to higher income families; and
Option 5 has its gains more evenly spread.

2.56 Option 7 (Food remains in but Compensation increased): This Option has been added to illustrate the relative merits from an equity viewpoint, of the food out/tax up approach and the food in/ compensation up approach.

2.57 This Option is properly assessed against Options 3b and 5b:

3b is the ANTS package but with household-specific price effects and dissaving allowed for and a 3.5 per cent pension increase;
5b is food out and pensions increased by 2.3 per cent; reduced tax cuts keep this Option revenue neutral;
7 is food in, pensions increased by 6 per cent and reduced tax cuts;
in all three cases the effects of new house prices and tobacco have been excluded.

2.58 The comparative results are relatively clear:

compared to 3b, Option 7 gives higher gains to lower income groups and lower gains to higher income groups;
compared to 5b, Option 7 also delivers higher gains to the lower income groups;
Option 7 allows more generous tax cuts to middle and higher income families than Option 5b, because it is cheaper to target lower income groups through increased compensation than to exclude food from the GST;
for middle income families there is little difference between the packages.

2.59 The NATSEM study was not able to model any increased compliance costs which might arise from the exclusion of food from the GST base. The authors argue, however, that the exclusion of food is likely to increase the burden on households, reduce GST revenue (net of administrative costs) and result in higher prices for goods and services than currently expected.

2.60 As a final comment on Option 7, the authors stated that the modelled outcomes made it clear that the effects of excluding food can be replicated by leaving food in the GST base while at the same time increasing and more accurately targetting compensation to lower income groups.

2.61 NATSEM'S Conclusions:

distributional impacts are sensitive to the modelling assumptions adopted;
assumptions made about price effects and the extent of dissaving, have a major effect on projected gains and losses from the tax package;
there are hidden losers from the tax package, predominantly low income earners;
making food GST-free and increasing the GST tax rate (Option 6) is not an attractive option;
the distributional outcome of making food GST-free and reducing tax cuts can be reasonably replicated by leaving food in the GST base and increasing and more accurately targetting compensation.

2.62 However, in response to Treasury's effective confirmation that compensation to pensioners will erode over time, Professor Harding stated that the question of the exclusion of food should be re-examined:

Footnotes

1 ANTS, pp 155-61.

2 Most of this section is drawn from the Select Committee's First Report, February 1999, Appendix VI and Appendix VII.

3 Submission No. 57.

4 Evidence, p.872.

5 Evidence, p.745.
6 The Government's Tax Package: Further Analysis Based on the MONASH Model, PB Dixon & MT Rimmer, 25 January 1999, p.ii.
7 Evidence, p.764.
8 Evidence, p.529.
9 Evidence, p.875

10 Evidence, p.869.
11 Evidence, pp 884, 900, 903.
12 Select Committee's First Report, February 1999, pp 197-8.
13 Select Committee's First Report, February 1999, p. 293.
14 This chapter is a summary of the NATSEM report: Distributional Impact of Possible Tax Reform Packages, by Neil Warren, Ann Harding, Martin Robinson, Simon Lambert and Gillian Beer, National Centre for Social and Economic Modelling, University of Canberra, 1 April 1999.
15 Evidence, pp 2346, 2353 and 2405.
16 NATSEM Report, p.17.
17 Evidence, p.2332.
18 Evidence, p.2405.
19 Evidence, p.2319.
20 Evidence, p.2335.
21 Evidence, p.2347.
22 Evidence, p.2402.