The Australian Economic Review, vol. 32, no. 4, pp.

32748

Changes in Indirect Taxes in Australia: A Dynamic General Equilibrium Analysis

Peter B. Dixon and Maureen T. Rimmer* Centre of Policy Studies Monash University

Abstract Over the last thirty years, the effects of indirect taxation changes have been analysed using comparative static general equilibrium models. We use a new method to analyse current changes in Australias indirect taxes: dynamic computable general equilibrium modelling. Comparative static methods compare the situation in a given year (usually unspecified) with and without a policy change. The dynamic method shows the effects of a policy change through time. Comparative static methods are usually restricted to estimates of long-run changes in allocative efficiency. The dynamic method provides information not only on efficiency but also on adjustment processes, including variations in employment. With our dynamic method, the effects of policy changes are analysed as deviations from explicit forecasts. We find that these forecasts are important for the policy results. For Australias current set of indirect tax changes, our main conclusions are (i) the short-run employment effects depend critically on the wage response; (ii) merchandise exporters benefit but tourism is harmed; and (iii) the long-run welfare effects are likely to be negative, reflecting a decline in the terms of trade and increased compliance costs.

1.

Introduction

* We thank Brian Parmenter for valuable suggestions made during the preparation of this paper.

In August 1998 the Treasurer, Peter Costello, circulated Tax Reform: Not a New Tax, a New Tax System hereafter referred to as ANTS. This is a description of Government plans for changes in Australias tax system. It involves five types of tax changes: increases in taxes on consumption and cuts in taxes on labour income, on capital income, on intermediate inputs and on inputs to capital creation. We analyse the effects of these tax changes using MONASH. This is one of three general equilibrium models of the Australian economy which were prominent in discussions of the ANTS package during 1998 and 1999. Unlike the other two models, ORANI-G1 and MM3032 which are comparative static, MONASH is dynamic. It produces time paths of policy effects, not just long-run equilibrium effects. These time paths are deviations from an explicit set of base-case forecasts. In this and earlier applications of MONASH we have found that the base-case forecasts matter for policy results. For example, in this paper our base-case forecasts involve an increase in world prices of service exports (tourism and education) relative to those for commodity exports. Thus we find that the ANTS package, which favours commodity exports relative to service exports, produces a negative impact on Australias terms of trade. This is missed by comparative static models with no explicit background forecasts. Readers are not expected to be acquainted with the details of MONASH. Consequently

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research 1999 Published by Blackwell Publishers Ltd, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA

328

The Australian Economic Review

December 1999

we spend considerable effort explaining the main MONASH results in terms of familiar economic mechanisms. Our explanatory tool is a back-of-the-envelope theoretical model which includes the five types of tax changes in the ANTS package. The paper is organised as follows. In Section 2 we review the Treasurys application in ANTS of the PRISMOD model. Section 3 contains our central MONASH simulation of ANTS. In this simulation we assume real aftertax wage bargaining and no additional compliance costs. In Section 4 we continue to ignore compliance costs but assume real before-tax wage bargaining. Compliance costs are introduced in Section 5. Then in Section 6 we extend the earlier analysis to take account of the modifications negotiated between the Government and the Democrats in June 1999. Concluding remarks are in Section 7. 2. The Treasurys Analysis

A central part of ANTS is the application of the PRISMOD inputoutput model to estimate the effects of the proposed tax changes on producer, consumer and investment prices. In essence, PRISMOD takes the form: pg = pg*A + tg + w*L pc = pg + tc and: pi = pg + ti (3) (1) (2)

vate investment goods used by business; and ti is the row vector of taxes collected per unit of sales of each commodity to private business capital creators. In applying PRISMOD, the Treasury introduced changes in tg, ti and tc to reflect the proposed reductions in taxes on producers and investors and the proposed increases in taxes on consumers. It then used (1) to (3) to compute the resulting impacts on pg, pc and pi. From these computations, the Treasury concluded that the proposed tax changes would, on average, reduce basic prices (pg) by 3.2 per cent, increase consumer prices (pc) by 1.9 per cent3 and reduce the cost of private investment goods used by business by 7 per cent. The first of these results can be understood as follows. For a typical Australian industry, intermediate inputs (including used up capital) represent about 75 per cent of costs. Taxes on these inputs are about 2 per cent of costs and labour is about 23 per cent of costs. In the Treasurys calculations, the nominal wage (w) is held constant and intermediate-input taxes are reduced by about 40 per cent, from 2 per cent of costs to 1.2 per cent.4 Assuming that prices are initially unity, a one-industry version of the Treasurys equation (1) gives: pg = pg*0.75 0.008 generating: pg = 0.032 (5) (4)

In these equations pg is the row vector of basic prices or costs of commodities; A is the input output coefficient matrix showing the use of each good i per unit of output of each good j (i can be used up directly in js production or it may be embedded in capital used up in js production); tg is the row vector of taxes collected on the inputs to each industry per unit of output; L is the row vector of labour input to each industry per unit of output; w is the wage rate; pc is the row vector of prices of commodities to consumers; tc is the row vector of taxes collected per unit of sales of each commodity to consumers; pi is the row vector of prices of pri

In obtaining the second result, it is apparent that the Treasury assumed that the proposed changes in the tax mix would increase rates of consumer tax by, on average, about 5.2 percentage points. This gives: pc = pg + tc = 0.032 + 0.052 = 0.020 (7) (6)

The current rate of consumer taxes is about 7 per cent (that is, tc = 0.07). With basic prices on unity, (7) implies an increase in consumer prices of 1.9 per cent (= 2.0/1.07).

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

Dixon & Rimmer: Changes in Indirect Taxes in Australia In obtaining the third result the Treasury assumed that the proposed changes in the tax mix would reduce rates of tax on private investment goods used by business by, on average, about 4 percentage points. This gives: pi = pg + ti = 0.032 0.04 = 0.072 (9) (8)

329

that is a reduction in the price of investment goods of about 7 per cent. The Treasury and the Government are inclined to portray the 3.2 per cent reduction in business costs (pg) as a major advantage to business arising from the proposed tax changes (see, for example, ANTS, p. 24). To us, this advantage is an illusion. It comes about because business is assumed to reduce its selling prices by an average of 3.2 per cent. Thus there is a general deflation of 3.2 per cent in the costs of business inputs and the value of business outputs. General changes in the price level are not normally an important determinant of business prosperity. At some points in ANTS it is argued that cost reductions will give Australian producers a competitive advantage in international markets. This is also an illusion. We would expect that general changes in the business cost/ price level would be offset by exchange rate movements. In fact, in their PRISMOD calculations, Treasury assumes an offsetting loss in competitiveness via an appreciation of the Australian dollar of approximately 3.2 per cent.5 What is important to individual industries is the movement in their costs relative to the movements in the costs of other industries. For example, an export-oriented industry gaining a 5.2 per cent cost reduction benefits in a situation of an economy-wide cost reduction of 3.2 per cent. This is because currency appreciation offsets only the economy-wide cost reduction leaving the individual industry with a 2 per cent competitive advantage in international markets. Thus, on the basis of the PRISMOD results in ANTS (p. 167) we would expect the Iron ore industry to gain from the tax changes (their costs are projected to decline by 5.8 per cent) whereas we expect the Sheep industry to lose

(their costs are projected to decline by only 2.6 per cent). Similarly, an import-competing industry experiencing less than a 3.2 per cent cost reduction loses in a situation of an economywide cost reduction of 3.2 per cent. This is because the appreciation causes only a 3.2 per cent reduction in the prices of competing imported products. Thus, on the basis of the PRISMOD results (p. 167) we would expect the Clothing and Footwear industries to be adversely affected by the proposed changes in the tax mix. Both are projected to have unit-cost reductions of only 2.9 per cent. For consumer products there is an additional complication. The effect of the tax package on industries producing such products depends not only on relative cost reductions but also on relative changes in consumer taxes. If tax changes in combination with cost reductions lead to an increase in the price to households of commodity i relative to the price to households of commodities in general (the CPI), then we can expect producers of commodity i to suffer from a shift in consumer purchases away from their product. Thus, for example we would expect the Beer industry to lose from the proposed tax changes. According to the PRISMOD calculations (p. 170), the consumer price of beer products will increase by 3.3 per cent, compared with an overall increase in consumer prices of only 1.9 per cent. Similar complications could apply to producers of investment goods. However, in MONASH we assume that there is little priceinduced substitution between different investment goods. This is a point of contrast with the MM303 model. In that model, it is possible to substitute between cars, buildings and other investment goods in the creation of units of capital. To the extent that substitution is possible, MONASH understates the benefits of the tax package to industries producing investment goods (for example, vehicles) on which tax rates will be reduced relative to the tax rates on other investment goods. Similarly MONASH overstates the benefits to industries producing investment goods on which there will be relative increases in tax rates. From the point of view of the macro economy, the most important relative price change

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

330

The Australian Economic Review

December 1999

emerging from the PRISMOD calculations is for labour compared with consumer goods. As already mentioned, the Treasury assumes no change in the nominal wage rate (w) in a situation in which the CPI rises by 1.9 per cent. Treasury relies on workers accepting the income tax cuts offered in the tax package as compensation for price rises. The assumption that workers will accept reductions in real before-tax wage rates in return for income tax cuts is also made in the central MONASH simulation, reported in Section 3. In Section 4 we use the MONASH model to work out the macroeconomic implications of the Governments tax proposals under the alternative labour market assumption that workers resist declines in w relative to the CPI. At the micro (commodity) level, economic theory suggests that the tax changes should be designed to increase the consumer prices of commodities which are currently lightly taxed relative to the prices of commodities which are currently heavily taxed. This requires reducing high tax rates and increasing low tax rates. A glance at the PRISMOD results for consumer prices (ANTS, pp. 1702) shows mixed outcomes on this issue. In accordance with theory, the tax changes are projected to increase the prices of some commodities (for example, Electricity) which are currently lightly taxed and to reduce the prices of some (for example, Cars) which are currently heavily taxed. On the other hand, there are some lightly taxed items (for example, Health services and Education) for which the tax changes are likely to reduce consumer prices and some heavily taxed commodities (for example, Beer and Tobacco) for which prices are likely to rise. A useful formula for making a preliminary back-of-the-envelope assessment of whether

the proposed tax changes will alter relative consumer prices in a welfare-enhancing manner is:6 Welfare = iWi*(R Ri)*(ti tave) (10)

where Welfare is the change in consumer welfare measured by the percentage increase in household budgets which would have the same effect on consumer utility as the proposed tax changes; Ri is the tax rate currently applying to good i (calculated as tax collected per dollar of the tax-inclusive price); R is an average of the Ris; Wi is the share of consumer spending currently devoted to good i; ti is 100*Ri /(1 Ri) with Ri being the proposed change in the tax rate applying to the tax-inclusive price of i; and tave is the expenditure-share weighted average of the tis. This formula projects gains from increases in the relative taxation (ti > tave) of lightly taxed items (R Ri > 0) this is the electricity example; gains from reductions in the relative taxation (ti < tave) of heavily taxed items (R Ri < 0)this is the car example; losses from reductions in the relative taxation (ti < tave) of lightly taxed items (R Ri > 0)this is the health and education example; and losses from increases in the relative taxation (ti > tave) of heavily taxed items (R Ri < 0)this is the beer and tobacco example. In Table 1, we have applied (10) with data from the MONASH model on Wi, Ri and R and data deduced from ANTS on ti. We obtained: Welfare = 0.017 (11)

that is, we calculated that the proposed tax changes will alter the pattern of consumer

Table 1 Back-of-the-Envelope Welfare Calculation Commodity Motor vehicles Electricity Tobacco Health All other Total Wi 0.026 0.012 0.011 0.063 0.888 1.000 R Ri 0.161 0.128 0.713 0.062 ti tave 10.51 4.39 11.09 3.51 Wi*(R Ri)*(ti tave) 0.045 0.007 0.089 0.014 0.068 0.017

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

Dixon & Rimmer: Changes in Indirect Taxes in Australia spending in a manner which increases consumer welfare by 0.017 per cent. In 1998 prices, this is equivalent to an annual gain to consumers of about $50 million. Because (10) captures the effects of only part of the proposed tax changes (it ignores the effects of changes in relative producer taxes) and because it involves some crude assumptions (for example, CobbDouglas preferences and no externalities), not much quantitative significance should be attached to result (11). For example, if it were established that the present taxing of Tobacco is optimal on the grounds of externalities then Tobacco could be left out of the calculation in Table 1 giving: Welfare = 0.106 (12)

331

Nevertheless, the calculations in Table 1 contain a message about orders of magnitude: a detailed investigation using a comprehensive general equilibrium model is likely to show that the proposed tax changes will lead to a negligible change in the economic efficiency with which households allocate their budgets across commodities. 3. The Effects of the Proposed Change in the Tax Mix: The Central Monash Simulation

rately from the published PRISMOD results for pg, pc and pi. To deduce the ANTS assumptions concerning changes in consumer taxes (tc) and changes in taxes on investment (ti), we used versions of (6) and (8).7 Deducing the changes in taxes applying to business inputs (tg) was more complicated. We started by configuring MONASH as an inputoutput model in the style of PRISMOD. We then conducted a simulation imposing our best guess of tg, based on the qualitative information in ANTS on which taxes are to be removed and quantitative information on these taxes in the MONASH model. In this simulation, we adopted Treasury assumptions: no change in wage rates; exchange rate appreciation of 3.5 per cent;8 and no change in rental prices of capital relative to asset prices. Next, we compared the results from this simulation for changes in basic prices (pg) with those published in ANTS. In the few cases in which there were significant discrepancies, we finetuned our guesses for tg. In this way, we arrived at a set of tax changes (tc, ti and tg) which we think accurately reflect those used in ANTS. It is these tax changes that are used in the MONASH simulations described below. 3.2 Key Assumptions

This section describes our central MONASH simulation of the effects of changing Australias tax system as proposed in ANTS. Subsection 3.1 describes how we ensured that the tax changes assumed in the MONASH simulation are in line with those assumed in ANTS. Subsection 3.2 sets out the key assumptions underlying the central simulation. Subsection 3.3 presents the results and explains them by the use of back-of-the-envelope algebra. 3.1 Deducing the Tax Changes ANTS contains comprehensive qualitative information on the Governments tax plan but does not provide detailed quantitative information on the proposed changes in the indirect tax rates (tg, tc and ti, see (1) to (3)). Fortunately, it is possible to deduce these changes quite accu

3.2.1 Labour Market In the central MONASH simulation (reported in this section), we assume that workers are concerned with the real after-tax wage rate, that is, the wage rate less income taxes, deflated by the CPI. If the labour market tightens, then we assume that the real after-tax wage rate increases in response to increased worker demands. More technically, we assume that the deviation in the after-tax real wage rate from its base-case forecast level increases in proportion to the deviation in employment from its basecase forecast level. The coefficient of proportionality is chosen so that the employment effects of a shock to the economy are largely eliminated after 5 years. In other words, after about 5 years, the benefits or costs of a shock, such as tax reform, are realised almost entirely

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

332

The Australian Economic Review

December 1999

as an increase or decrease in real after-tax wage rates. This labour market assumption is consistent with conventional macroeconomic modelling in which the NAIRU is exogenous. Further explanation of our labour market assumption is given in Appendix 2. 3.2.2 Public Expenditure In the central simulation and all subsequent simulations, we assume that the change in the tax mix makes no difference to the path of real public consumption. 3.2.3 Consumption, Investment, Ownership of Capital and Measurement of Welfare In each year of our tax-mix simulations, we assume that Australians save sufficient to finance the same quantity of investment as in the basecase forecasts. Together with our assumption of no change in the path of real public consumption, this means that aggregate real consumption diverges from its base-case forecast level by an amount reflecting the divergence in real income available to Australians. In other words, we assume that the benefit or cost in year t from the change in the tax mix is absorbed in that year entirely as a change in real household consumption. This is consistent with a zero marginal rate of real national saving. Marginal rates of saving in the Australian economy are low but not zero. Consequently, our household consumption assumption leads to a small overestimation of the immediate consumption effects of income changes. Against this, our assumption has two important simplifying advantages. First, it means in our model that it is easy to keep track of foreign/ domestic ownership of units of capital. Extra units created as a result of tax changes are entirely foreign owned. Similarly, if the tax changes lead to a reduction in the capital stock, then there is a corresponding reduction in the quantity of foreign-owned capital. Consequently, in our policy simulation, all of the variation in after-tax capital income associated with variations in the capital stock is excluded in the calculation of the change in income available to Australians. The second simplify

ing advantage is that compensating and equivalent variation calculations based on the divergences in the paths of the volumes of consumption of each commodity provide a valid indicator of the welfare effects of the tax changes under consideration. This is because in our policy simulations the domestic population undertakes no extra investment, owns no extra capital and incurs no extra debt. 3.2.4 Rates of Return on Capital In simulations of the effects of changes in policy variables, MONASH allows for short-run divergences in after-tax rates of return on industry capital stocks from their levels in the base-case forecasts. Short-run increases/decreases in rates of return cause increases/decreases in investment and capital stocks, thereby gradually eroding the initial divergences in after-tax rates of return. 3.2.5 Production Technologies MONASH contains variables describing primary-factor and intermediate-input-saving technical change in current production; inputsaving technical change in capital creation; and input-saving technical change in the provision of margin services. In our policy (tax-mix) simulations, we assumed that all technology variables have the same values as in the base-case forecasts, that is, we assume that the change in the tax mix has no effect on technology. 3.3 Results

3.3.1 Back-of-the-Envelope Model Figures 1 to 11 show macro and industry effects from our central simulation. In explaining these figures, we use a back-of-the-envelope (bote) model which includes the five types of tax changes in the ANTS package. The bote model sharply simplifies MONASH by assuming that the economy produces one good (grain) and imports one good (vehicles). Grain production is via a constantreturns-to-scale production function of capital and labour inputs. Grain and vehicles are both

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

Dixon & Rimmer: Changes in Indirect Taxes in Australia consumption and investment goods. Units of consumption and investment are formed as Cobb-Douglas functions of grain and vehicles leading to Cobb-Douglas unit-cost functions. Finally, we assume that the costs per unit of employing capital and labour equal the values to the employer of their marginal products. Under these assumptions we have: Pc = P g gc * P v vc *Tc Pi = P g gi * P v vi *Ti W*Tw = (Pg/Tg)*M1 Q*Tq = (Pg/Tg)*Mk WrealA = W/Pc WrealB = W*Tw /Pc and:

333 (21)

M1(K/L) = WrealB*Tg*Tc*(Pv /Pg)vc and that: Mk(K/L) = *Tq*Tg*Ti*(Pv /Pg)vi

(22)

(13) (14) (15) (16) (17) (18)

In (20) to (22), we emphasise that marginal products are functions of K/L. M1 is a positive function of K/L and Mk is a negative function of K/L. In terms of the bote model we can think of ANTS as having the following effects. It reduces Tg from 1.036 to 1.022. Taxes on intermediate inputs to production currently are about 2 per cent of production costs for most industries. This translates to about 3.6 per cent of capital and labour costs. Consequently in our bote model, which has no intermediate inputs, we represent existing taxes on these inputs as a 3.6 per cent tax on output. The tax package involves the removal of about 40 per cent of the taxes on intermediate inputs. We represent this as a movement in Tg from 1.036 to 1.022. It reduces Ti from 1.037 to 1.013. Taxes on inputs to capital creation are about 3.7 per cent of investment costs for most industries.9 The tax package involves the removal of most of the existing taxes but the imposition of new taxes on housing construction. We calculate that after the changes to the tax mix have been implemented, the average rate of tax on sales of goods and services to capital creation will be 1.3 per cent. Thus in the bote model we assume that Ti moves from 1.037 to 1.013. It reduces Tw from 1.250 to 1.215. Taxes on wages are about 25 per cent of after-tax wage income. The tax package involves a 14 per cent reduction in this rate. We represent this as a movement in Tw from 1.250 to 1.215. It reduces Tq from 1.250 to 1.233. Taxes on capital are about 25 per cent of after-tax capital income. The tax package involves a 14 per cent reduction in this rate. However, we estimate that about half of the reduction in

= Q/Pi

(19)

where Pg and Pv are the basic price of grain and the c.i.f. price of vehicles; Pc and Pi are the purchasers prices of a unit of consumption and a unit of investment; Tc, Ti and Tg are the powers (one plus rates) of the taxes applying to consumption, investment and production; Q and W are after-tax factor payments, the rental rate and the wage rate; Tw and Tq are the powers (one plus rates) of the income taxes applying to labour and capital income; M1 and Mk are the marginal products of labour and capital; WrealA and WrealB are the after-tax real wage rate and the before-tax real wage rate; is the rate of return on capital calculated as the after-tax rental or after-tax user price of capital divided by the cost or asset price of a unit of capital; and the s are positive parameters reflecting the shares of grains and vehicles in consumption and investment, such that gc + vc = 1 and gi + vi = 1. From these equations we find that: M1(K/L) = WrealA*Tw*Tg*Tc*(Pv /Pg)vc or equivalently:

(20)

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

334

The Australian Economic Review

December 1999

the rate of tax on capital income will be offset by dividend imputation. Thus, we represent the movement in the capital tax rate by a shift in Tq from 1.250 to 1.233. It increases Tc from 1.070 to 1.105. Taxes on consumption currently average about 7 per cent. After the imposition of the GST, our MONASH calculations indicate that this will become about 10.5 per cent. Thus in the bote model we assume that Tc moves from 1.070 to 1.105. 3.3.2 The MONASH Results for the Central Simulation Explained by the Bote Model With the change in the tax mix Tw*Tg*Tc declines by 1.0 per cent (from 1.250*1.036*1.070 to 1.215*1.022*1.105). Under the central labour market assumption, WrealA is sticky in the short run. In the absence of a change in the terms of trade (a movement in Pv /Pg), the bote model (see (20)) indicates that the change in the tax mix will cause a short-run reduction in the marginal product of labour. Consequently K/L will decrease. Because K moves slowly, there must be a shortrun increase in L. This is confirmed in Figure 1 where we see that employment moves above the base-case forecast.

In implementing ANTS, the Government planned to make a net reduction in tax revenue of $6 billion. This accounts for about half of the planned reduction in income taxes. Without the reduction in tax revenue, Tw would decline from 1.250 to 1.233 rather than to 1.215, and Tw*Tg*Tc would increase by about 0.5 per cent. This would generate a short-term decline in L. Thus we conclude that the short-term employment gain associated with the package is entirely the result of fiscal stimulation. Looking now at (22), we ask what is the short-run impact of the change in the tax mix on the rate of return ()? With a decrease in K/L, Mk rises. Tq*Tg*Ti falls by 4.9 per cent (from 1.250*1.036*1.037 to 1.233*1.022*1.013). Again, ignoring changes in Pv /Pg, we see that must rise. Thus in our MONASH simulation we expect to see a short-run increase in investment followed by an upward movement in capital. This is confirmed in Figure 2. The short-run increase in employment leads to increased demands for after-tax real wages, generating increases in WrealA (Figure 3). Thus, after initially moving below its forecast path, M1 moves up towards this path (see (20)). This means that after its initial fall, K/L must rise (see Figure 1). This forces Mk to fall back towards forecast generating a reduction in , thereby slowing the rise in K. With growth in K

Figure 1 Central: Capital and Labour Inputs (per cent deviation from base-case forecasts) Per cent 1 0.8 0.6 0.4 0.2 L 0 -0.2 -0.4 -0.6 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year K K/L

Figure 2 Central: Investment and Capital (per cent deviation from base-case forecasts) Per cent 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 -0.5 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year K Investment

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

Dixon & Rimmer: Changes in Indirect Taxes in Australia

Figure 3 Central: Employment and Real Wage Rates (per cent deviation from base-case forecasts) Per cent 1.5 1 0.5 0 -0.5 -1 -1.5 -2 -2.5 -3 -3.5 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year Real wage before tax Real wage after tax

335

being choked off, further rises in WrealA (necessitating increases in K/L) must be achieved by reductions in L. Thus, as can be seen in Figure 1, L falls back towards forecast. In the very long run, the deviation in is zero. In the absence of terms of trade effects, the reduction in Tq*Tg*Ti leaves the K/L ratio permanently increased (Figure 1). Thus M1 is permanently raised. This together with the reduction in Tw*Tg*Tc leaves WrealA permanently increased with L approximately on its forecast path. 3.3.3 The Terms of Trade Throughout the explanation so far we have ignored movements in the terms of trade. As can be seen in Figure 4, the terms of trade initially moves above forecast and then moves below forecast. In the context of the bote model this reinforces the short-run decrease in M1 (an increase in the terms of trade lowers Pv /Pg). Similarly, it reinforces the short-run increase in and the consequent increases in I and K. In the long run, the decline in the terms of trade reinforces the upward movement in M1. This accelerates the return of L to its forecast path; the increase in Mk; the decline in R; and the choking off of the upward movement in K. What explains the movements in the terms of trade in Figure 4? In MONASH we treat Aus

tralia as a small country on the import side, that is we treat c.i.f. import prices in foreign currency as exogenous. On the other hand, we recognise that Australia has considerable shares of world markets for several relatively homogeneous agricultural and mineral products, and that Australia exports distinctive varieties of manufactured goods and tourist and education services. Thus we assume that expansions of any of Australias exports reduce their world prices and generate a decline in Australias terms of trade. Consequently the deviation path of the terms of trade is closely connected with the deviation path in aggregate exports. As can be seen from Figure 4 and will be explained shortly, the change in the tax mix initially reduces exports but eventually leads to an increase in exports. This is consistent with the initial upward movement in the terms of trade and the eventual downward movement. A reinforcing long-run negative effect on the terms of trade is changes in the composition of exports. As is apparent from Figure 5, the change in the tax mix favours exports of primary and manufactured goods relative to exports of services. This is discussed below. In our base case we are forecasting decreases in the world prices of primary and manufactured goods relative to the world prices of services. This is consistent with historical trends. An
Figure 4 Central: Exports, Imports and the Terms of Trade (per cent deviation from base-case forecasts) Per cent 2 1 0 -1 -2 Export volume -3 -4 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year Terms of trade Import volume

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

336

The Australian Economic Review

Figure 5 Central: Export Volumes (per cent deviation from base-case forecasts)

December 1999

Per cent 6 4 2 0 -2 -4 -6 -8 -10 -12 -14 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year Tourism exports Foreign student numbers Non-traditional exports Total exports Traditional exports

implication is that a change in the composition of Australias exports away from services will be terms-of-trade reducing. 3.3.4 Imports, Exports and the Composition of Exports The first step in understanding the results for the trade quantities is to look at investment. In the short run, investment jumps strongly to facilitate the upward movement in K. In the longer term the deviation in K stops growing, allowing investment to fall back towards its forecast path (Figure 2). When investment is strong, the balance of trade moves towards deficit with an associated strengthening of the real exchange rate (Figure 6) and consequent export contraction. As the investment deviation declines, the balance of trade moves towards surplus, with an associated weakening of the real exchange rate and consequent export expansion. A surprising result is the occurrence of a positive long-run deviation in aggregate exports in combination with a positive long-run deviation in the real exchange rate (Figures 4 and 6). The change in the tax mix is cost reducing for major export industries such as Coal, Iron ore and Non-ferrous metal ores relative to major import-competing industries such as Motor vehicles, Aircraft, Electronic equipment

and Other machinery. This allows long-run export expansion despite long-run real appreciation. It also explains the long-run increase in import volumes (Figure 4). The change in the composition of exports (Figure 5) referred to above reflects the planned GST treatment of exports of goods compared with the planned treatment for the main service exports (tourism and Australianbased education). Goods exports will not be subject to GST. For these exports there will be a reduction in $A costs largely but not completely offset by exchange rate appreciation. On the other hand, foreign tourists and students will find that most of their purchases in Australia are subject to GST. For tourists, we estimate that the foreign-currency price of their Australian visit will rise by about 3.6 per cent.10 (This takes account of the appreciation of the Australian dollar; increases in the GSTinclusive $A prices of hotels and other tourist services; and of the fact that in-Australia costs are only about 60 per cent of tourist expenditures.) For foreign students we estimate that the foreign-currency price of their Australian studies will rise by about 3 per cent. (This takes account of the appreciation of the Australian dollar; increases in the GST-inclusive $A prices of accommodation, food and other

Figure 6 Central: The Real Exchange Rate and Investment (per cent deviation from base-case forecasts) Per cent 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year Investment Real exchange rate

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

Dixon & Rimmer: Changes in Indirect Taxes in Australia student requirements; and of the fact that education fees will be GST free.) 3.3.5 Real GDP, Real Consumption and Welfare Figure 7 compares the deviation results for capital and labour with that for real GDP. At first glance, the GDP path seems too low, especially in the long run, relative to the deviation paths for the factor inputs. With capital and labour shares in GDP being about 0.3 and 0.7, we might expect the GDP deviation in 2008 to be about 0.31 per cent (= 0.3*0.90 + 0.7*0.05). Instead it is only 0.15 per cent. In earlier applications of MONASH11 we have provided a detailed decomposition of GDP results. This work shows that deviations in real GDP depend not only on deviations in quantity variables (such as capital, labour and tax-carrying commodity flows) but also on differences between forecast and policy simulations in the composition of GDP. If the policy under consideration increases the shares in GDP of slow-growing quantity variables relative to the shares of fast-growing quantity variables then on this account real GDP growth is reduced in policy relative to forecast. In our tax-mix simulations, the tax shares in GDP of various slow-growing consumption items (for

337

example, Beer and Tobacco) are increased in policy relative to forecast. On the other hand, the tax shares of various fast-growing consumption items (for example, Motor vehicles) are reduced in policy relative to forecast. This has the effect of reducing GDP growth in policy relative to forecast, generating an everexpanding negative effect on the policy level of GDP relative to the forecast level. A similar phenomenon affects the measurement of real consumption. The path of real consumption in policy deviates from the forecast path not only because of deviations in the quantities consumed of each commodity but also because of deviations in budget shares. By increasing their prices relative to those other consumption goods, the change in the tax mix increases the budget shares of alcohol and tobacco. (Price elasticities of demand for these commodities are low.) Simultaneously, by lowering their relative prices the change in the tax mix decreases the budget shares of vehicles and appliances. Overall the budget shares of slow-growing consumption items are increased by the tax changes while those of fast-growing items are reduced. This produces the everexpanding negative long-run consumption deviation shown in Figure 8.

Figure 8 Central: Real Consumption and Welfare (per cent deviation from base-case forecasts) Figure 7 Central: Real GDP and Factor Inputs (per cent deviation from base-case forecasts) Per cent 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 -0.1 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year Real GDP L 0.2 0 -0.2 Real consumption -0.4 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year Note: (a) Welfare is the Laspeyre cost difference as a per cent of consumption. Welfare (lcd)a K 0.8 0.6 0.4 Per cent 1.2 1

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

338

The Australian Economic Review

December 1999

From the point of view of economic welfare, movements in real GDP and real consumption associated with share effects are of little significance. In measuring the welfare effects of the changes in the tax mix, we rely on calculations of Laspeyre and Paasche cost differences (see Ng 1983, p. 89). The Laspeyre cost difference (lcd) is the amount of money that could be taken away from households in the policy situation, leaving them with just sufficient money to buy the base-case forecast bundle of commodities, and the Paasche cost difference (pcd) is the amount of money that must be given to households in the forecast situation so that they can just buy the policy bundle of commodities. These cost differences can be expressed conveniently as percentages of the household budget and calculated according to: lcdT = and: pcdT =

BSH Ti *diff T ( consr i )

p p i

(23)

BSH Ti *diff T ( consr i )

f f i

(24)

p f where BSH Ti and BSH Ti are the shares in year T of the household budget accounted for by good i in the policy and forecast situations; and p f diff T (consri) and diff T (consri) are the differences in year T between the levels of consumption of good i in the policy and forecast situations expressed as percentages of the levels of consumption of good i in the two situations. Under our assumptions explained in Subsection 3.2, differences in consumption quantities provide a legitimate basis for calculating welfare effects. However (23) is a lower bound on welfare changes. This is because consumers faced with the policy prices and with income sufficient to buy the base-case forecast bundle could achieve at least the base-case level of welfare, implying that the percentage reduction in their policy income could be at least lcd before their welfare is driven below its forecast level. On the other hand, (24) is an upper bound because consumers faced with the forecast prices and with income sufficient to buy the

policy bundle could achieve at least the policy level of welfare, implying that the percentage increase in their forecast income could be no more than pcd before their welfare is pushed above its policy level. As illustrated in Figure 9, in the present application of MONASH there is little difference between the paths of lcd and pcd. Either is an adequate indicator of welfare. We choose to work with lcd.12 The lcd path in Figure 8 indicates a small long-run welfare loss from the tax package. This is a result of offsetting factors. On the one hand, welfare is reduced by the decline in the terms of trade. At the end of the forecast period, exports and consumption are 27 and 60 per cent of GDP. This suggests that the long-run decline in the terms of trade of 0.5 per cent (Figure 4) would reduce real consumption by about 0.22 per cent (= 0.5*0.27/0.60). On the other hand, the expansion in the capital stock carries a welfare gain. As explained in Subsection 3.2 we assume that the extra capital generated as a result of changing the tax mix is owned by foreigners. Nevertheless there are benefits to Australians. These arise from extra tax collections on foreign-owned capital. The extra collections are worth about 0.11 per cent of consumption, calculated as the percentage deviation in capital (0.9) times the capital share in GDP (0.3) times the tax rate on capital income

Figure 9 Central: Welfare Measures (per cent of consumption expenditure) Per cent 1.2 1 0.8 0.6 0.4 0.2 0 -0.2 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year Laspeyre cost difference (lcd) Paasche cost difference (pcd)

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

Dixon & Rimmer: Changes in Indirect Taxes in Australia (0.25) divided by the share of consumption in GDP (0.6). Together the terms-of-trade and the capital tax effects suggest a long-run welfare loss of about 0.11 per cent of consumption. Figure 8 indicates a smaller loss (0.03 per cent

339

of consumption), leaving an unexplained gain of 0.08 per cent of consumption. This is of the order of magnitude expected for the efficiency gains associated with the reorganisation of household expenditures (see Section 2).

Figure 10 Central: Output of Main Winners (per cent deviation from base-case forecasts) Per cent 7 6 5 4 3 2 1 0 -1 2000

2001 I12 Iron ore

2002

2003

2004

2005

2006

2007

2008 Year

I30 Wool processing I95 Water transport I68 Motor vehicles Figure 11 Central: Output of Main Losers (per cent deviation from base-case forecasts)

I17 Mining services I50 Basic chemicals I73 Electronic equipment

I64 Non-ferrous products I13 Non-ferrous ores

Per cent 2 1 0 -1 -2 -3 -4 -5 -6 2000

2001 I71 Aircraft I27 Beer

2002

2003

2004

2005

2006

2007

2008 Year

I110 Entertainment I39 Footwear

I87 Residential building I112 Personal services

I96 Air transport I111 Hotels

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

340 3.3.6 Industry Results

The Australian Economic Review

December 1999

Figure 10 shows the deviations in the output paths of the industries which are the main winners from the change in the tax mix. These are industries for which the output deviation in 2008 is more than 3 per cent. They fall into three groups. The first contains traditional export industries and related industries (Iron ore, Non-ferrous ores, Non-ferrous products, Wool processing, Water transport and Mining services). Output in these industries is stimulated by tax-related cost reductions which are only partially offset by currency appreciation. The second group is export-oriented manufacturing industries (Electronic equipment and Basic chemicals). The relatively strong stimulation of these industries reflects relatively strong growth in non-traditional exports as a whole. The third group consists of one industry, Motor vehicles. Under the change in the tax mix, this industry will benefit strongly from reductions in sales taxes. Figure 11 shows the deviations in the output paths of the industries which are the main losers from the change in the tax mix. These are industries for which the output deviation in 2008 is below 1.40 per cent. Hotels, Air transport, Personal services, Entertainment and Aircraft appear among the losers because of their connection with both domestic and foreign tourism. As already explained, the change in the tax mix will increase the costs to foreign tourists of their Australian visits. MONASH also recognises that the change in the tax mix will reduce the costs to Australians of holidaying overseas relative to holidaying in Australia. Residential building is a loser because of the proposed GST treatment of the sale of new houses. Beer and Footwear lose because the Government proposes a sharp increase in the tax rate on these products causing their consumer prices to rise relative to the CPI. 4. The Effects of the Proposed Change in the Tax Mix with Before-Tax Wage Bargaining

the real after-tax (AT) wage rate. An alternative assumption is that workers are concerned with the real before-tax (BT) wage rate. Under this assumption, the deviation in the BT real wage rate from its base-case forecast level increases in proportion to the deviation in employment from its base-case forecast level. Results generated by MONASH for the effects of the Governments proposed tax changes under the BT assumption are shown in Figure 12. Under the AT assumption, wage demands by workers can be damped by reductions in income taxes. This is not the case under the BT assumption. Consequently, the short-run macro effects of the proposed change in the tax mix (which involves cuts in income taxes) are more favourable with the AT assumption than with the BT assumption. The adverse short-run macro repercussions of the change in the tax mix under the BT assumption can be understood in terms of equation (21). Under the tax package Tg*Tc will rise by about 2 per cent (from 1.036*1.070 to 1.022*1.105). With sticky before-tax real wages (WrealB), there will be a tendency for Ml to rise, necessitating a short-run fall in L. As can be seen in Figure 12 the short-run fall in employment reaches 1.3 per cent, about 120 000 jobs.
Figure 12 Before-Tax: Capital and Labour Inputs and Welfare (per cent deviation from base-case forecasts) Per cent 1 K 0.5 0 Welfare (lcd) -0.5 -1 -1.5 -2 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year L

In the simulation discussed in Subsection 3.3, we assumed that workers are concerned with

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

Dixon & Rimmer: Changes in Indirect Taxes in Australia The long-run effects under either wage assumption are quite similar. This is because under both assumptions, employment in the long run is driven close to its base-case forecast level. Relative to the AT results, the BT results show a delayed movement to the long run. Under the BT assumption, the short-run increases in K and I are strongly damped, reflecting negative initial movements in output and employment. Thus, under the BT assumption the capital stock takes longer to achieve its final deviation from the base-case forecasts than it takes under the AT assumption. Notice in Figure 12 that K is still rising in 2008 whereas in Figure 1 it has stabilised. 5. Central Simulation with Compliance Costs

341

pliance costs on 1.6 million firms whereas the wholesale sales tax (which the GST replaces) involves only 75 000 firms. We recognise the extra compliance costs in this simulation by imposing a reduction in labour productivity of 0.5 per cent. This is equivalent to assuming an annual wastage of labour worth about $1.5 billion. As can be seen from Figure 13, the recognition of compliance costs largely eliminates the short-run welfare gains which were apparent in the central simulation (Figure 8) and generates a long-run welfare loss of about 0.6 per cent of consumption. 6. Government/Democrats Package with Compliance Costs

In Figure 13, we return to the AT wage assumption. However this time we make an allowance for compliance costs. These were not recognised in the simulations described in the previous two sections. Binh (1999) draws on international experience with GSTs and establishes that the GST will impose on taxpayers and on state and federal tax offices additional clerical costs of between $1 billion and $2 billion a year. Binhs main argument is that the GST will impose comFigure 13 Central with and without Compliance Costs: Welfare (lcd) (per cent of consumption expenditure) Per cent 1.2 1 0.8 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year Central with compliance costs Central without compliance costs

To ensure passage of the GST legislation through the parliament, the Government was forced by the Democrats in June 1999 to agree to a series of modifications of the ANTS package. In Figure 14 we compare welfare results from Section 5 for the ANTS package with results for the Government/Democrats package. In simulating the Government/Democrats package we made similar assumptions to those in Section 5: real after-tax wage bargaining and compliance costs of $1.5 billion a year. In introducing the Democrats modifications we recognise the following:
Figure 14 Central and Government/ Democrats with Compliance Costs: Welfare (lcd) (per cent of consumption expenditure) Per cent 1.2 1 0.8 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 2000 2001 2002 2003 2004 2005 2006 2007 2008 Year Government/Democrats with compliance costs Central with compliance costs

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

342

The Australian Economic Review

December 1999

The Government/Democrats package will make $32 billion of expenditure on food GST-free, costing revenue $3.2 billion a year. The Government/Democrats package provides extra transfer payments to pensioners and low income families of about $0.8 billion a year. The cut in income tax will be reduced from the 14 per cent promised in the ANTS package to 12.9 per cent, a revenue saving of $1.1 billion a year. Relative to the ANTS package, the Government/Democrats package involves an increase in taxes on diesel fuel usage of about $0.7 billion a year. (For usage in the rail industry the Government/Democrats package provides for a cut in the costs of diesel fuel of about 25 per cent. However, for other users, the Government/Democrats package imposes, on average, an increase in the price of diesel fuel of about 4.8 per cent.) Relative to the ANTS package, the Government/Democrats package involves an increase in state taxes of about $1.1 billion in 2001 and $2.2 billion in later years achieved by deferral of planned abolition of the debits tax and various other state taxes. After 2001, these five sets of modifications recognised in our Government/Democrats simulation are budget neutral. In addition to these modifications the Democrats have persuaded the Government to allocate extra environmental expenditures worth about $1 billion a year. We have left these out of our simulation because we think that this extra expenditure will be offset by cuts or reduced rates of growth in other government expenditures. As can be seen from Figure 14, the Government/Democrats package lowers welfare relative to the ANTS package. This reflects the retention of various distorting state taxes. In a more complete representation of the Government/Democrats package we would expect a greater negative welfare effect relative to

the ANTS package than is shown by the comparison in Figure 14. In the Government/Democrats simulation we have not allowed for extra compliance costs associated with the partial exemption of food from the GST or with the imposition of different taxes on different uses of diesel fuel. Nor have we allowed for distortions in consumer decisions between the purchase of food carrying a GST and food not carrying a GST. 7. Concluding Remarks

7.1 Implications of the MONASH Central Simulation Under favourable assumptions (no additional compliance costs and real after-tax wage bargaining, the central simulation in Section 3), the ANTS package has a small short-run stimulatory effect on employment. This is dissipated in the long term by increases in real wage rates. Investment and the capital stock increase in the short term, leading in the long term to a permanent increase in the capital/labour ratio. The package has positive effects on some industries, particularly those trade-oriented industries receiving greater-than-average cost reductions from the cuts in taxes on intermediate inputs. On the other hand, tourism-related industries will be quite adversely affected. The overall long-run welfare effect of the package in the central simulation is small and slightly negative. The main reason for the negative outcome is a projected decline in the terms of trade. Our conclusion from the central simulation that the long-run welfare benefit from ANTS is small is supported by Murphys (1999) simulation using MM303. This shows a long-run annual welfare gain of about 0.12 per cent of GDP. We suspect that the MM303 results are slightly more optimistic than the central MONASH results for two reasons. First, MM303 projects a decline in the terms of trade only about half as great as that projected by MONASH. We think this is because Murphys comparative static analysis misses the dynamic effects on the terms of trade of the ANTSinduced compositional switch in exports away

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

Dixon & Rimmer: Changes in Indirect Taxes in Australia from services and towards merchandise. Second, MM303 projects a considerably greater increase in the long-run capital/labour ratio than is projected by MONASH. Increases in Australias capital stock are welfare enhancing via their effects on tax collections from foreign investors. The MM303 long-run increase in the capital stock is larger than that in MONASH because (i) MM303 assumes that the changes in the tax mix cause no deviation in rates of return and (ii) MM303 ignores differences between industries in their composition of capital. Assumption (i) is appropriate only for the very long run. Over the period of policy interest (to 2008, for example) we assume in MONASH that adjustment of rates of return is incomplete. Thus in the central MONASH simulation of the ANTS package, rates of return in 2008 in most industries are still above their base-case forecast levels. Under assumption (ii), MM303 implies that the ANTS-induced percentage reduction in the cost of creating a unit of capital is the same in each industry. MONASH on the other hand shows considerable variation across industries in the effects of ANTS on capital costs. For example, consider the Road transport and Beer industries. In the MONASH database, vehicles are an important capital item for Road transport but not for Beer. With the ANTS package causing a sharp decline in the price of vehicles, MONASH shows a sharp de-

343

cline in the cost of a unit of capital to Road transport but not to Beer. Because Australias Beer industry is experiencing slow growth and has durable capital, its investment/capital ratio (I/K) is low. The Road transport industry has the opposite characteristics, giving it a high I/K ratio. In these circumstances, MM303s assumption (ii) can lead to a sharp overestimate of the increase in aggregate capital associated with a given reduction in the costs of investment. This is illustrated by a hypothetical example in Table 2. At first glance, the ORANI-G results of Johnson et al. (1998, Table 3.9) appear to be more favourable with respect to welfare than either the MONASH or MM303 results. ANTS (p. 156) cites the ORANI-G results as support for the proposition that the Governments proposed tax changes will produce substantial economic growth benefits. ORANI-G shows a long-run GDP gain of 3.75 per cent (associated mainly with a sharp increase in the capital stock) and a consumption gain of more than 1 per cent. However these measures should not be considered as indicators of the welfare effect of the ANTS package. First, Johnson et al. completed their work before the ANTS package was announced and they simulated a broader based GST and a cleaner set of reforms than those in ANTS. Second, unlike the MONASH and MM303 calculations, the ORANIG calculations do not account for the increases

Table 2 Capital Stimulation under MM303 and MONASH Assumptions: An Examplea Beer Capital Investment Per cent change in cost of capital MONASH MM303 Per cent change in capital stock MONASH MM303 0 3 6 3 1.5 3 0 6 12 6 6 6 3 0.4 Transport 1 0.4 Economy average

Note: (a) In this example we assume that MM303 and MONASH agree that capital stimulation in each industry is 0.5 times the reduction in capital costs and that the overall reduction in capital costs (an investment-weighted average of the industry reductions) is 6 per cent. We assume that all of the reduction in capital costs is concentrated in the transport industry. Failure to recognise this (MM303s assumption (ii)) leads in our example to a 100 per cent overestimate (3 per cent rather than 1.5 per cent) of the increase in the economys capital stock (a capital-weighted average of the industry increases). The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

344

The Australian Economic Review

December 1999

in foreign liabilities or domestic saving required to finance policy-induced increases in Australias capital stock. In effect ORANI-G overestimates the benefits of increases in capital by treating these increases as if they are gifts from a foreign benefactor. In common with MM303, ORANI-G shows a much stronger stimulation of aggregate capital than does MONASH. ORANI-G does not suffer from MM303s assumption (ii). However Johnson et al. (1998) adopt MM303s assumption (i). In addition, we think that the failure of Johnson et al. to account for foreign liabilities or domestic saving leads them to overestimate the long-run consumption effects of capital-cost-reducing changes in indirect taxes. Because consumption expenditure is capital intensive, this leads to an overestimate in ORANI-G of the increase in the economys capital stock. 7.2 Downside Risks We think that the MONASH central simulation is an optimistic representation of the likely outcome from the changes in the indirect tax system which will be taking place over the next few years. In Section 4 we showed that the ANTS package combined with before-tax wage bargaining will reduce employment in the short run. Some commentators, for example Murphy (1999, p. 273), have claimed that the assumption of before-tax wage bargaining is implausible. However, we think that this possibility should not be dismissed lightly. In their submission to the Senate inquiry into the new tax system, the Australian Council of Trade Unions (1999) explicitly rejected cuts in income taxes as a rationale for reduced wage demands. They argued that the income tax cuts offered in the ANTS package should be viewed as compensation for bracket creep experienced since 1993 and not as compensation for a GST-induced jump in the CPI. Thus they indicated that they will be seeking to maintain before-tax real wage rates. In Section 5 we included in the MONASH simulations conservative estimates of compliance costs. These left the long-run welfare effect of the ANTS package as clearly negative

and substantially eliminated the short-run employment benefits which were generated in the central simulation (with after-tax wage bargaining). In Section 6 we included the Democrats modifications to the ANTS package. These have a small negative effect on overall economic welfare. 7.3 The GST: A Failure in Policy Formulation The Governments tax package (ANTS) was based on inadequate economic analysis. It appears that the only modelling done by the Government was the Treasurys PRISMOD calculations of the effects of the package on commodity prices and on the real incomes of different types of families. The Treasurys modelling ignored 60 years of progress in economics and used a theoretical framework first applied by Wassily Leontief in the 1930s, see Leontief (1953). It seems that the Treasury gave the Government no worthwhile quantitative advice on the likely effects of the package on employment, industry structure, the trade accounts, overall economic efficiency or any other vital macroeconomic variable. Nevertheless, the Government became committed to the package, making strong but unsupported claims about its beneficial effects. The Senate inquiry into the package elicited detailed analysis from Australias top economic modellers. Among their conclusions are that the package will: not increase employment in the long run; harm Australias exports of tourism and educational services; reduce the terms of trade; and have a negligible long-run effect on Australias overall economic welfare. In ANTS (p. 8), the Government claimed that:
The current tax system is ineffective. It provides a crumbling base from which to derive

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

Dixon & Rimmer: Changes in Indirect Taxes in Australia

the necessary revenue to fund essential government services . . .

345

changes in tax rates. Then, with Cobb-Douglas preferences the households demand functions and indirect utility function (U) are given by: Xi = WiY/Ti and: U = kY T i
i W i

No quantitative evidence was used to support this claim. Contrary to the Governments claim, separate forecasts by the Centre of Policy Studies and Access Economics for the next 10 years suggest that the present system of indirect taxes would produce revenue growth within a percentage point of that of GDP. These forecasts are discussed in Richardson (1999). In formulating ANTS, the Government was loath to seek advice from outside the Treasury. It was worried that the mere act of commissioning research on consumer taxes would have lead to premature and unhelpful reactions from the media and from lobby groups. However the in-house approach had a considerable cost. It left the Government (i) formulating ANTS on the basis of narrow analysis; (ii) unable to change fundamentally its policy in light of broader independent analysis which subsequently became available as a result of the Senate inquiries; and (iii) in a position where it was compelled to accept unsatisfactory changes to its policy imposed by the Democrats. If the Government had formed its policy with the Senate evidence at hand, it is difficult to believe that it would have chosen a GST as the central component of tax reform. In the area of microeconomic reform, the Government can refer potential policy changes to the Productivity Commission. The Commission collects information and opinions in transparent processes involving draft reports, public hearings and final reports. The Government need not commit to a policy change until all the relevant data and analyses are available. It is unfortunate that a similar approach was not adopted with respect to ANTS. First version received January 1999; final version accepted September 1999 (Eds). Appendix 1: Back-of-the-Envelope Calculation of Welfare Effects of Changes in Indirect Taxes: Derivation of Equation (10) We assume that the producer prices of all goods are one and that they are unaffected by

(A1)

(A2)

where Xi is the consumption of good i; Ti is the power of the tax on good i (1 + the rate applied to the producer price) and also the consumer price of i; Wi is the exponent on the ith good in the direct utility function and also the share of good i in household expenditure; k is a parameter; and Y is household income given by: Y=F+

(Ti 1)Xi G
i

(A3)

The first term on the right-hand side of (A3) is factor income which we assume is unaffected by tax changes; the second term is tax revenue which we assume is returned to households as a lump sum grant from the Government; and the third term can be thought of as a payment by households to an agent outside the economy. In making a back-of-the-envelope assessment of the welfare effects of a set of tax changes we compute the compensating variation as the change in G required to preserve the initial value of utility. This can be done conveniently (but only approximately) by working with percentage changes. (A1) to (A3) can be written in percentage change form as: xi = y ti u=y and: y= (A4) (A5)

Wi ti
i

Wi ti + Wi Ri xi
i i

(A6)

where Ri is the tax rate calculated on the taxinclusive price, that is Ri equals (Ti 1)/Ti ; is the change in G expressed as a percentage of Y;

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

346

The Australian Economic Review

December 1999

and the lower case variables x, y, t and u are percentage changes in the corresponding upper case variables. Assuming that u equals zero, we find from (A4) to (A6) that:

= Wi Ri(ti tave)
i

(A7)

where tave is the average percentage change in the powers of the taxes, that is tave is iWi ti. Finally we obtain equation (10) by recognising that (A7) can be rewritten as:

gate employment and long-run changes in real wages. In the simulation reported in Subsection 3.3, the real wage variable is after-tax wages deflated by the CPI. In Section 4 we discuss a simulation in which the real wage variable is the before-tax wage rate deflated by the CPI. In either simulation, where W refers to the relevant real wage concept, our algebraic specification is: Wt Et Wt 1 ------------- 1 = ------------------- 1 + ----------- 1 W t, old W t 1, old E t, old (A8)

= Wi (R Ri)(ti tave)
i

(A7) In this equation, old indicates a base-case forecast value. Wt, old and Et, old are the real wage rate and the level of employment in year t in the base-case forecasts. Wt and Et are the real wage rate and the level of employment in year t in the policy simulations, and is a positive parameter. Under (A8), the real wage rate in the policy simulations will continue to move further

where R is any convenient average of the Ris. Appendix 2: Labour Market Specification We assume that real wages are sticky in the short run and flexible in the long run. Thus policy shocks generate short-run changes in aggre-

Figure A1 Operation of the EmploymentWage Specification in a Steady-State W/Wold D'

W/Wold W3/Wold W2/Wold W1/Wold 1 S3 S

2

S3 S2 S S

D'

S D 1 E3/Eold E2/Eold E1/Eold

E/Eold

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

Dixon & Rimmer: Changes in Indirect Taxes in Australia above the real wage rate in the base-case forecasts whenever employment in the policy simulations is above that in the forecasts. We set the value of so that the effect on aggregate employment of a policy change in year t will be largely eliminated by year t + 5. That is, we assume that employment gains/losses from policy changes are a short-run phenomenon with the economy tending in the long run to an exogenously given natural rate of unemployment. The operation of the employmentwage specification is illustrated in Figure A1 for a steady-state case in which technology, consumer tastes, foreign prices and capital availability are unchanged from year to year. In this steady state, the demand curve for labour in each year t is DD and the supply curve is SS. In each year employment is Eold and the real wage rate is Wold, that is, the employmentwage combination is at point I. Now assume that there is a policy change in year 1 which causes the demand curve for labour to shift up to D'D', where it remains for all future years. The supply curve for year 1 is the initial supply curve SS. The policy-simulation levels for employment and the real wage rate in year 1 are E1 and W1. In year 2 there is a vertical upward shift in the supply curve reflecting the gap between W1 and Wold. In our diagram employment and the real wage rate in year 2 are E2 and W2. Eventually the supply curve for labour stops moving when W reaches W. At this stage employment has returned to Eold. Endnotes 1. See Johnson et al. (1998). 2. See Murphy (1999). 3. This excludes the prices of new houses and of tobacco. 4. In the package, wholesale taxes on intermediate inputs are removed but existing taxes on energy products are substantially retained. The Government has judged that it would be difficult to impose the goods and services tax (GST) on financial services and the services of housing. Instead the Government intends to tax

347

inputs to financial services and to home construction. 5. ANTS (p. 161) mentions that the exchange rate will be moderately stronger over time. A Treasury official told us that this means an appreciation of between 3 and 3.5 per cent. 6. This formula is valid under the assumptions that consumer taxes are the only distortions in the economy and that consumer preferences are Cobb-Douglas, see Appendix 1. 7. We modified these equations to allow for wholesale, retail, transport and other margins. 8. This was our best guess of the Treasury assumption. As indicated in endnote 5, the Treasury confirmed only a range for its exchange rate assumption. 9. Here we are referring to all industries. In Section 2, where we mentioned a 7 per cent reduction in the cost of investment goods, we were referring to private investment goods used by business, that is we excluded housing and investment by government. 10. As mentioned in endnote 5, Treasury assumed appreciation of between 3 and 3.5 per cent. MONASH implies a smaller appreciation, between 1 and 2 per cent, for most of the simulation period. Consequently we project a smaller increase in the foreign-currency price of visits to Australia than would be obtained on the basis of Treasury calculations. The difference between the MONASH and Treasury exchange rate projections arises because the Treasury ignores terms-of-trade deterioration. 11. See for example Dixon, Malakellis and Rimmer (1997). 12. An alternative approach, recommended by Creedy (1998) for example, is to calculate compensating and equivalent variations. This involves the use of the utility function. Where adequate, we prefer lcd and pcd. These measures avoid the utility function and are immediately interpretable.

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research

348 References

The Australian Economic Review

December 1999

Australian Council of Trade Unions 1999, Submission no. 383 to the Senate Select Committee on a New Tax System, Parliament of the Commonwealth of Australia, Canberra. Binh, T.-N. 1999, Assessing the revenue and simplification impacts of the governments tax reform, Journal of Australian Taxation, forthcoming. Costello, P. 1998, Tax Reform: Not a New Tax, a New Tax System, AGPS, Canberra. Creedy, J. 1998, The welfare effects of indirect taxes in Australia, Working Paper no. 6/ 98, Melbourne Institute of Applied Economic and Social Research, University of Melbourne. Dixon, P. B., Malakellis, M. & Rimmer, M. T. 1997, The Australian Automotive Industry from 198687 to 200910: Analysis Using the MONASH Model, a Report to the Industry Commission, Centre of Policy Studies, Monash University.

Johnson, D., Freebairn, J., Creedy, J., Scutella, R., Cowling, S. & Harding, G. 1998, Evaluation of Tax Reform with Producer and Consumer Responses, Tax Reform: Equity and Efficiency Report no. 3, Melbourne Institute of Applied Economic and Social Research, University of Melbourne. Leontief, W. 1953, The inputoutput approach in economic analysis, in Input Output Relations: Proceedings of a Conference on Inter-Industrial Relations, ed. The Netherlands Economic Institute, H.E. Stenfert Kroese N.V., Leiden. Murphy, C. 1999, Modelling a New Tax System (ANTS)Comparing MONASH and MM303, in Senate Select Committee on a New Tax System, First Report, Parliament of the Commonwealth of Australia, Canberra. Ng, Y.-K. 1983, Welfare Economics: Introduction and Development of Basic Concepts, rev. edn, Macmillan, London. Richardson, C. 1999, Is the indirect tax system broken?, Business Council of Australia Papers, vol. 1, pp. 4651.

The University of Melbourne, Melbourne Institute of Applied Economic and Social Research