Corporate Income Tax Reform: The Neglected Issue of Tax Incidence

Corporate Income Tax Reform: The Neglected Issue of Tax Incidence

III1111111111 ~II 200612275 Economic Analysis & Policy ©2006 Vol 35 Nos 1&2, March/Sept 2005 45 CORPORATE INCOME TAX REFORM: THE NEGLECTED ISSUE O...

671KB Sizes 0 Downloads 12 Views

III1111111111 ~II 200612275 Economic Analysis & Policy

©2006

Vol 35 Nos 1&2, March/Sept 2005

45

CORPORATE INCOME TAX REFORM: THE NEGLECTED ISSUE OF TAX INCIDENCE John Ablett and Neil Hart University of Western Sydney

The 'double taxation' of corporate income is often used as an argument in support of the integration of company and shareholder taxes, as occurred with the introduction of tax imputation in 1987 in Australia. These arguments are based, often implicitly, on the premise that the economic incidence of company taxes falls on shareholders receiving dividend income. However, a review of the available theoretical and empirical literature fails to provide an unambiguous answer to the corporate income tax incidence question. Empirical results presented in this paper suggest the existence of significant forward shifting of the tax on to consumers though higher prices in the case of Australian manufacturing corporations. A more informed discussion of tax reform must therefore consider more carefully the implications arising from the likely existence of significant forward shifting of the corporate income tax.

1. INTRODUCTION Calls for the integration of company and shareholder taxes are often predicated on the belief that the 'double taxation' of dividends inherent in the classical tax system is inequitable and may distort corporate financial decision-making. However, suchan argument depends on the existence of backward shifting of the tax to shareholders and overlooks the likely forward shifting of the tax on to consumers through higher prices. To the extent that forward shifting does occur, the corporate income tax's effects on income distribution come to resemble those of a sales tax that is regressive to income. In this paper, some preliminary results are presented that suggest that, in the case of Australian manufacturing corporations, forward income tax shifting was significant during the time period being considered. If accepted, these results would challenge much of the economic argument used in support of tax imputation systems, such as the one that has operated in Australia since July 1987. Discussion in the paper will be organised as follows. Corporate tax incidence issues, and its significance to the income tax integration debate, are outlined in section 2. In section 3 a review of some of the alternative approaches to the analysis of corporate tax incidence are reviewed. An empirical model used to investigate the possible presence of forward corporate income tax shifting is developed in Section 4. Preliminary results derived from the model are presented in Section 5. Some conclusions relating to corporation tax structure are suggested in the final section.

?opyright, of F,ulI Text rests with the original c~pyright owner and: except as pe,r~itted und,er the C?pyright Act 1968, copying this copyright material is prohibited without the permission of the owner or Its exclUSive licensee or agent or by way of a licence from Copyright Agency Limited. For information about such licences contact Copyright Agency Limited on (02) 93947600 (ph) or (02) 93947601 (fax)

46

Economic Analysis & Policy

©2006

Vol 35 Nos 1&2, March/Sept 2005

2. INCOME TAX INTEGRATION AND CORPORATE INCOME TAX INCIDENCE ISSUES The issue of 'double taxation' of corporate income distributed as dividends arises in the context of the 'classical' tax system where corporate income is taxed and dividends paid out of this income are taxed in the hands of shareholders without any relief for the tax paid by the corporation. Alternatively, corporate taxation could be fully integrated with the personal income tax system. In such a system, taxes would be collected from the corporation; however, this tax would be treated as pre-payment of shareholder's tax. Profits would be taxed only once at the shareholder's marginal tax rates, with corporations in effect being treated as if they were partnerships. Arguments against a fully integrated tax system often emphasise administrative difficulties. Instead, partial integration systems have been devised, whereby corporation and shareholder taxes are integrated with respect to profits distributed as dividends. Imputation tax systems now widely used allow shareholders an 'imputation credit' for tax paid at the corporation level. While not fully removing the 'distortions' claimed to be inherent in the 'classical' system, imfutation does provide a partial solution to the so-called 'double taxation' problem. However, before the policy responses outlined above can be accepted, the 'double taxation' issue needs to be more carefully examined. A useful way to proceed is to consider the principle of tax incidence that plays an important role in the economic analysis of the effects of taxation. The incidence of a tax can be defined in terms of its effect on the level and distribution of individuals' real income over a specified time period. The legal incidence refers to the immediate impact of a tax, while the economic incidence allows for the shifting of the tax to some other unit in the economic system. There are a number of possible mechanisms via which the corporate income tax may be shifted to various economic units. Importantly, there is the possibility of forward shifting onto consumers in the form of higher prices. Backward shifting onto shareholders, raw material suppliers and employees represent further shifting avenues. Imperfections in tax legislation leading to avoidance and evasion of tax liabilities and the ability of firms to partially absorb taxation through a rearrangement ofasset and liability structures further compI icate incidence analysis, particularly when the incidence question is considered in an international setting. All of these factors higWight the need to consider the effects of taxation not in terms of legal incidence, but rather. in the context of economic incidence where the ultimate effects of the tax on income distribution and decision making cannot be determined until tax shifting mechanisms have been taken into account. 'Double taxation' of distributed corporate income becomes more apparent than real if corporations are capable of significantly shifting the incidence of the tax onto other sectors in the economy. Differences in corporate and personal income tax scales complicate the issues, and it needs to be noted that income tax payable by corporations for a given financial year is assessed and paid during the following financial year, while personal income tax is collected at essentially the same time it is received.

6conomicAn'ilIysis & Policy

©2006

Vol 35 Nos 1&2, March/Sept 2005

47

3. CORPORATE INCOME TAX INCIDENCE ANALYSIS Recent literature on the analysis of corporate tax incidence has as its origins Harberger's (1962) theoretical contributions based on the neoclassical general equilibrium framework and Krzyzaniak and Musgrave's (1963; 1966) pioneering empirical investigation. However, despite this work, unambiguous incidence conclusions remain elusive. This situation is clearly observed in overall tax incidence studies, such as that in the Australian context by Harding and Warren (1999), where it is arbitrarily assumed that the economic incidence of the corporate income tax is 50 per cent on shareholders and 50 per cent on consumers. The lack of unambiguous conclusions, from both the theoretical and the empirical literature, is the major conclusion emerging from Whalley's (1997, p. 1) literature review; 'the question Df the incidence of the corporate tax is not well-defined, and should be approached with a certain amount of scepticism' . This ambiguity has not been clarified in the recent literature, with Auerbach's (2005) review indicating the extent to which theoretical analysis is unable to provide unambiguous or widely accepted answers to the corporate income tax incidence issue. 2 The review also highlights the scarcity of empirical investigations over the past few years. Some insights into how this situation has emerged can be gained by briefly looking at the alternative research methodologies that have been used to analyse the corporate income tax incidence issue. Harberger 's (1962) analysis ofthe incidence ofthe corporation tax was developed within a familiar 2-factor 2-industry general equilibrium framework divided into a corporate and a non-corporate sector. The corporation tax is treated as a partial factor tax on 'capital' employed in the corporate sector and has the affect of reducing the (after tax) rate of return on capital in the corporate sector. The subsequent flow Df ('perfectly mobile') capital from the corporate to the non-corporate sectors will increase the relative price and reduce the level of output in the corporate sector, increasing the gross rate of return remaining in the corporate sector and reducing the return on capital in the non-corporate sector (given the usual assumption of a declining marginal product of capital). Ultimately, the economic incidence of the tax, with respect to labour and capital (in both sectors), depends on the relative factor intensities in both sectors, the elasticity of substitution between capital and labour and the elasticity of substitution of demand for each sector's products. Harberger then proceeded to develop a numerical analysis of his model, assuming various 'plausible' elasticity and income share values, and concluded that capital (in both sectors) 'probably' bears close to the full burden of the corporate income tax in the US . Importantly, this much quoted result does not emerge directly from Harberger's theoretical model, but rather from his 'numerical simulations' of the model.3 2

Other reviews of aspects of the corporate income tax issue include Sorenson (1995) and Gravelle (1995).

3

This is sometimes overlooked by commentators who consider the special case in which consumption and production behaviour can be represented by Cobb-Douglas functions. Under such assumptions 'capital' must bear the full economic incidence of the tax as the flow of capital between sectors leaves factor shares unchanged; see McLure and Thirsk (1975) for an example of such as exposition of Harberger's model.

---------------------

48

Economic Analysis & Policy

©2006

Vol 35 Nos 1&2. March/Sept 2005

Harberger's neoclassical approach has since been modified and extended in two directions. Firstly attempts have been made to 'relax' some of the very strong assumptions to be found in the theoretical analysis. This has largely taken the form of adding 'market imperfections' which limit the extent to which factors of production are able to respond to market signals. 4 Not surprisingly, the indefinite incidence conclusions are amplified once these real world 'complications' are incorporated into the model. In an attempt to overcome the open-ended theoretical results emerging from the neoclassical modelling, attempts have been made to refine the numerical simulations developed initially by Harberger.The computational procedure introduced by Shoven and Whalley (1973) cleared the way for the construction of neoclassical applied general equilibrium tax incidence models. 5 Construction of these models generally takes the following pattern. Firstly functional forms for demand and production are chosen which are consistent with the theoretical parentage ofthe model (e.g. application ofWalras 'Law, profit maximisation, savings determined capital stock, marginal productivity distribution theory etc.). Next is the critical 'calibration' procedure whereby parameter values are determined. The economy under consideration is assumed to be in equilibrium, and the parameter values are chosen in such a way that the model is able to reproduce this data set as an equilibrium solution. Specification of exogenous elasticity values may also be required. Once the calibration procedure is completed, a fully specified model is available and policy changes can be considered by comparing the counterfactual with the original equilibria. Despite the increasing sophistication of applied general equilibrium models, the relevance to policy-makers of the results derived from such models should not be exaggerated. Even if the theoretical pedigree of this approach is accepted,6 many serious analytical weaknesses are to be found. These arise largely out of the calibration procedure referred to above. A key assumption is that the data collected for the initial observations,appropriately 'adjusted' ,representanequilibriumfortheeconomybeing considered. Critical elasticity and other parameter value choices proceed in a setting where data sources are often scarce and of questionable reliability, leading to doubts surrounding the confidence which can be placed on elasticity dependent outcomes. Similarly, even within the confines of the selected theoretical framework, various model specifications are possible. These differences are important, as the relationships 4

See for example Pami (1988), Bhatia (1988) and Gravelle and Kotlikoff (1989). The 'marketimperrections', which often entail wage rigidities,are discussed in McLure (I C£75). Atkinson and Stiglitz (1980) and Miyagiwa (1988). More recent examples are outlined in Fullerton and Metcalf (2002. pp. 1812-1815) and Auerbach (2005).

5

For examples of the usage of such models in the Australian context see Piggott and Whalley (1987) and Piggot (1983). Shoven and Whalley (1984) provide an extensive survey of applied general equilibrium tax incidence models.

6

Groenewegen (1984, pp. 113-119) presents a very useful summary of the theoretical objections to the neoclassical approach.

Economic Analysis & Policy

©2006

Vol 35 Nos 1&2, March/Sept 2005

49

between the key parameters and tax incidence are substantially transfonned once modifications such as increasing returns,intennediate goods and the mutual production of each product by corporate and non-corporate producers are introduced, and limited market imperfections are pennitted. Problems also existin selectingthe appropriate tax base and in the choice of closure of the system. However, because of the absence of hypothesis testing procedures, applied general equilibrium analysis does not propose a way of discriminating between alternative models. While the neoclassical modelling appeared to support the notion that the incidence of the corporate income tax fell predominately on 'capital', alternatively specified empirical investigations were commonly reporting significant tax shifting. Initially, following the pioneeringKrzyzaniakand Musgrave's (1963; 1966) contributions,this modelling was rather ad-hoc in nature, consisting of the estimation of unconstrained reduced fonn equations with the corporate rate of return expressed as a function of a number of independent variables including some measure of the effective corporate income tax. However, the tax shifting conclusions from this ad-hoc modelling was shown to be very sensitive to the choice and specification of the independent variables, which led Krzyzaniak and Musgrave (1968) to call for the development of a more complex approach involving a structural model in which price, wage and shifting behaviour were specified and all equations fully identified. The challenge identified by Krzyzaniak and Musgrave was taken up by Duransky (1972) and, in particular, by Sebold (1979). Sebold (1979) developed a simultaneous equation model, where the effective tax rate variable is included in a number of stochastic equations explaining manufacturing and materials prices, wages, and depreciation. His model suggested total shifting equal to 69%, accounted for largely by forward shifting in the fonn of higher prices which included the responses of manufacturers to simultaneous increases in materials prices and wages. While Sebold's methodology would appear to represent the most fully developed corporate income tax incidence modellingframework, there are significantoperational difficulties to be confronted. Such an approach places enonnous demands on available data sources and the increased complexity of the model amplifies the likelihood of specification ambiguities and errors. Consequently, the majority of empirical investigations have subsequently concentrated primarily on forward shifting of the corporate income tax through price increases,and have therefore been closely related to alternative models of short-period pricing behaviour of finns. In the traditional partial equilibrium representation of the profit maximising finn in the short period, significant forward shifting of the corporate income tax is not predicted because these taxes do not affect marginal costs. However, a noticeable feature of the modelling of forward corporate income tax shifting has been the application ofnon-profit maximising theories of finn behaviour. Most significantly, the mark-up pricing principle has been accorded the dominant role, a pricing principle attracting empirical support from a long history of investigations into industrial pricing behaviour. Intuitively, significant forward shifting of the corporate income tax would appear to be consistent with the mark-up pricing principle, with the tax liability treated as a component of direct or prime costs

50

Economic Analysis & Policy

©2006

Vol 35 Nos 1&2, March/Sept 2005

to which the mark-up is applied.' However, while forward shifting ofcorporation taxes may appear intuitively to be implied by the mark-up pricing principle, the existence and extent of such shifting requires empirical verification. The alternative ways forward shifting can be modelled are clearly illustrated in the approaches developed subsequent to the Krzyzaniak and Musgrave contribution. Gordon's (1967) influential empirical investigation was based on a model which assumes that firms determine prices on the basis of applying a constant mark-up over capacity average direct costs. On the basis of results derived from an estimating equation significantly modified because of data limitations, Gordon concluded that corporations suffered the full burden of the tax in the US during the period (1925-62). Adopting similar techniques, Davis (1972) found that UK data suggested the absence of significant tax shifting in a variety of alternatively specified models. Similarly the absence of significant forward tax shifting is consistent with Agapitos' (1979) UK estimating equations based on simple models of mark-up and target rate of return pricing behaviour which introduced the corporate tax directly into the price equation. Included in Coutts, Godley and Nordhaus' (1978) study of industrial pricing in the UK is an alternative method of testing for corporate tax shifting. Their test basically attempted to determine whether price is a fixed gross mark-up on normal costs, or whether the mark-up instead varies so as to maintain a fixed net mark-up. Their results suggested that the former was the case, implying the absence of significant tax shifting at least in the short-run. However, in an important investigation using a similar testing principle,Beath (1979) reached the oppositeconclusion. Beath postulated a relationship between mark-up of manufacturing prices over expected unit labour costs and the effective corporation tax rate. He concluded that over the period 1953-73, there was a decline in the mark-up of manufacturing prices and that this could be explained by a reduction in the effective corporate tax rate. On the basis of these results Beath put forward the hypothesis that corporate taxation is shifted forward in the short run in an attempt to maintain the net profit margin. As noted above empirical work on corporate tax incidence has been rarely attempted during recent decades. The only empirical modelling using Australian data was developed by Daly and Hart (1994); this adopted a similar methodology to that used in Beath's (1979). This investigation reached similar conclusions to that of Beath, with the available Australian data suggesting forward tax shifting in the case of manufacturing sector corporations for the period 1968-90. Firms appeared to respond to changes in the expected effective tax rates by altering prices so as to maintain a target net return. However the scope of the analysis was severely curtailed by data limitations, which compromised diagnostic testing within the model, and which also led to some uncertainty in interpreting the derived shifting conclusions. The study presented in the following section of this paper attempts to further develop 7

Examples of corporate income tax analysis that include the mark-up pricing hypothesis include Asimakopulos and Burbidge (1974), Laramie (1991) and Damania and Mair (1992).

Economic Analysis & Policy

©2006

Vol 35 Nos 1&2, March/Sept 2005

51

the Daly and Hart analysis through the construction of a much broader data base and with the application of more rigorous estimating procedures.

4. AN EMPIRICAL MODEL OF FORWARD CORPORATE INCOME TAX SHIFTING As in Beath (1979) and Daly and Hart (1994), the model used in this study starts with the following identity: V=PQ =:n:+ WQ

(1)

where V is the value added, P is the price index of value added or implicit deflator, :n: is gross profits, Wis an index of unit labour costs and Q is net output. The model assumes that firms have a target share,~, of net profits (:ltn) as a proportion of value added, giving :ltn

=(1 -

or E =

e):n:

= ~V

(2)

I~V

(3)

-e

where e is the effective rate of tax on corporate profits, i.e. the company tax liability divided by gross profits. Substituting (3) into (1), rearranging and dividing by Q yields

(4) If we transform equation (4) by taking natural logarithms, substituting

e*

= l/(I--e) ,rearranging and adding time subscripts, we obtain In P, - In WI =In

(_1_.) 1- ~e,

(5)

To arrive at a simpler empirical formulation for the price equation, we replace the right hand side of (5) by its first order Maclaurin series expansion, giving InP,-ln WI

=~e;

(6)

In the estimated equations described below, we incorporate Beath's (1979) assumption that the mark-up is determined with reference to expected labour costs, using a simple adaptive scheme, where the expected unit labour cost in a year t (denoted W,') is set equal to realised unit labour costs in the previous year (W,-I). In addition, some attempt is made to capture cyclical effects by inclusion of a capacity utilisation variable (CU) that can affect ~ in an additive way, so that ~

=~o + lCU,

(7)

52

Economic Analysis & Policy

©2006

Vol 35 Nos 1&2, March/Sept 2005

Rewriting equation (6) with these modifications gives the basic model as:

In P,-In W,_l

= 13 oe; + 13 le;CU/

(8)

The choice of a suitable capacity utilisation variable is difficult. In estimating the empirical models of this paper, we used Australian Bureau of Statistics data showing average weekly hours of overtime worked in manufacturing industry, as estimated from a survey of employers. It was felt that, given rigidities in the labour market, the level of overtime worked could capture short-term cyclical effects. Formulations Allowing for Specific Industry Subdivision Effects An important improvement over the time series methodology previously used by Daly and Hart (1994) is the construction and use in this paper of a pooled crosssectional and time series data set consisting of ten years of data (1989/90 to 1998/99) for each of the nine ANZIC manufacturing subdivisions. It is believed that this data set has allowed a more credible testing of the forward shifting hypothesis. In particular, it has allowed the estimation of subdivision specific effects. The nine ANZIC subdivisions considered are (abbreviations in brackets): • • • • • • • • •

Food, Beverages and Tobacco (FBT) Textile, Clothing, Footwear and Leather (TCF) Wood and Paper Products (WPP) Printing, Publishing and Recorded Media (PPR) Petroleum, Coal, Chemicals and Associated Products (PCC) Non-Metallic Mineral Products (NMM) Metal Products (MErr) Machinery and Equipment (MAC) Other Manufacturing (OTH)

The question arises as to how any subdivision specific effects are to be included in the model. Two possibilities suggest themselves. 8 Firstly, a fixed effects model can be considered, leading to the inclusion of subdivision specific constant terms. Although the basic model does not imply the existence of an intercept term, such a formulation allows for systematic fixed differences in the mark-up level across subdivisions. With the addition of a disturbance term (EJ, this results in equation (8) being modified to the following equation to be estimated: (i

In equation (9), subdivisions. 8

~o

and

~l

= 1,.....,9)

(9)

are assumed to be common coefficients across all

A third possible formulation would be a random effects (error components) model. However such a model implies the disturbance vectors of different cross-sectional units are uncorrelated, which is not supported by the available data.

Economic Analysis & Policy

©2006

Vol 35 Nos 1&2, March/Sept 2005

53

A second possibility, which is perhaps more in line with the basic model presented above, is to allow for subdivision specific coefficients of the tax rate variable eO, Also allowing for the possibility of a common constant, this yields an estimable equation of the form (i = 1,.....,9)

(10)

In both equations (9) and (10), the forward shifting hypothesis predicts positi ve coefficients on the tax rate variable. One would also expect ~I to be positive, since in economic upturns firms would find it easier to pass on increases in the effective tax rate they face.

The Data Set Manufacturing subdivision specific annual data (1989/90 to 1998/99) were obtained forall variables except the capacity utilisation proxy (average weekly overtime hours worked). Unpublished annual data were obtained from the Australian Bureau of Statistics for chain-volume gross value added (VA) and the implicit price deflator index for subdivision turnover. 9 The latter was used as a proxy for the price index of value added (P). Total labour cost (TLC) was obtained from the Australian Tax Statistics by adding together total wages and salaries, external labour costs and employer superannuation contributions. The calculation of the unit labour cost index (by manufacturing industry subdivision) can be explained by letting P" Q, and w, represent, respectively, the price index of value added, net output and unit wages in year t. Then, relative to a base year b, an index of unit labour costs in year t is given by: w wQ PQ W, = ----!. X 100 = _,-' x --1L....£. x 100

Wb

WbQb

PbQ,

(11)

By using TLC, in place of W,Q" and VA, in place of PbQ" our equation for estimating the index of unit labour costs becomes: _ TLC 1 VA b W,- TLC x VA x 100 b

(12)

,

The company tax Iiability was calculated by multi pIying the statutory corporate tax rate by company profits before income tax. Subsequently the company tax liability was divided by company profits before income tax, net interest paid and depreciation (gross profits) to obtain our measure of the effective tax rate on corporate profits (e).10 Over the period covered by the data set, the statutory corporate income . tax rate began at 39% (till 1992/93), fell to 33% (1993/94 and 1994/95), and then 9

Thanks are due to Paul Curran and John Ridley from ABS for their help in providing this data.

10

All data on manufacturing subdivision profits were obtained from published Australian Bureau of Statistics tables.

54

Economic Analysis & Policy

©2006

Vol 35 Nos 1&2, March/Sept 2005

increased again to 36%. It could be argued that in theory franked dividends should be taken into account when calculating the 'effective' corporate tax rate. However it is felt unlikely that corporate managers concern themselves with the amount of dividend imputation tax credits that are actually claimed by domestic shareholders, at least when they are determining mark-ups. Thus the assumption used here is simply that managers treat the nominal company tax paid as the effective company tax liability, even though the existence of dividend imputation means at least part of these taxes represents personal taxation withheld at the company level.

5. PRELIMINARY ESTIMATION RESULTS Preliminary pooled ordinary least squares estimation showed strong evidence of first order autocorrelation of the disturbance terms of equations (9) and (10). Furthermore, a Lagrange multiplier test of cross-section homoscedasticity (Green' 2000, p. 596) clearly indicated rejection of the null hypothesis, implying the presence of cross-section heteroscedasticity. The null hypothesis of a diagonal disturbance covariance matrix was also clearly rejected using the Breusch-Pagan (1980) Lagrange multiplier test. These test results suggest that ideally equations (9) and (10) should be estimated by full generalised least squares, i.e. allowing for first order autocorrelation, cross-section heteroscedasticity and cross-section correlation of disturbances. Unfortunately a lack of degrees of freedom, caused by the use of only 10 years of data, precluded estimation taking account of cross-section correlation of disturbances. This also necessitated the estimation of a common first order autoregressive (AR( 1» disturbance parameter, rather than cross-section specific AR(1) terms. It is hoped that in the future the availability of more years of data will make full GLS estimation possible, thus improving the efficiency of estimation. Afurtherestimation issue concerns the possible existence of unit roots in the data. The usual unit root tests on the individual manufacturing subdivision data generally led to non-rejection of the hypothesis of a single unit root. This was to be expected, since the availability of only 10 annual observations significantly limited the power of these tests. In the event, attempts to estimate models (9) and (10) using differenced data yielded qualitative results similar to those reported below (i.e. a statistically significant coefficient on the tax rate variable), but a generally poor fit. In view of the above considerations,Tables 1-4 below present results for models (9) and (10) using data in levels and taking account ofcross-section heteroscedasticity and anAR(1) coefficient that is common across all manufacturing subdivisions. An iterative procedure was used in estimation. lI

11

For the results given in Tables 1,3 and 4, the convergence criterion of the iterative estimation procedure was set such that iteration ceased once the change in all estimated coefficients was less than 0.001. For the model of Table 2, however, this criterion was not able to achieve convergence in 100 iterations; for this model a convergence criterion of 0.015 was required to obtain convergence in a reasonable number of iterations.

Economic Analysis & Policy

©2006

Vol 35 Nos 1&2. March/Sept 2005

TABLEt REGRESSION RESULTS FOR EQUATION (9) Model: In Pi; -In Wi;-l =Oi + 13oei; + 13 1ei;CU, + E;; Variable e* e*CU

AR(I) Fixed Effects (ai's): FBT TCF

WPP PPR PCC

NMM MEr

MAC OTH

(i = 1•..... ,9)

Coefficient

t statistic

P-Value

1.5958 -0.0489 0.6387

4.0638 -0.9892 6.7911

0.0001 0.3256 0.0000

-1.8947 -1.7307 -1.9429 -1.8621 -1.7732 -1.6351 -1.6829 -1.7525 -2.0598

Weighted Statistics: Adjusted R2 = 0.6789. SE of regression= 0.1453, DW = 2.56 Unweighted Statistics: Adjusted R 2 = 0.6705 •SE of regression = 0.1453, DW = 2.27

TABLE 2 REGRESSION RESULTS FOR EQUATION (9) CYCLICAL VARIABLE OMITTED Variable e*

AR(l) Fixed Effects(
WPP PPR PCC

NMM MEr

MAC OTH

Coefficient

t statistic

P-Value

1.2890 0.5914

3.4156 5.7287

0.0010 0.0000

-1.6298 -1.4720 -1.6925 -1.5578 -1.4988 -1.3630 -1.4244 -1.4751 -1.8182

Weighted Statistics: Adjusted R 2 = 0.6662. SE of regression 0.1400. DW = 2.45 Unweighted Statistics: Adjusted R2 =0.6701. SE of regression = 0.1454, DW =2.18

55

56

Economic Analysis & Policy

©2006

Vol 35 Nos 1&2, March/Sept 2005

TABLE 3 REGRESSION RESULTS FOR EQUATION (10) + ~ 0,1

+ ~ t eiPU, + Ei )

(i = 1,.....,9)

Model: In Pi" -In Wi) -1

=00

Variable

Coefficient

t statistic

P-Value

Constant (common) e*CU AR(I)

-1.7987 -0.0486 0.6369

-4.2317 -0.9969 6,7555

0.0001 0.3228 0.0000

Specific e* Variables: FBT TCF WPP PPR PeC NMM MEr MAC OTH

1.5209 1.6568 1.4682 1.5481 1.6136 1.7264 1.7005 1.6225 1.3432

3,9494 4,2061 3.8007 4.1902 4.1303 4.2780 4.3171 4.2125 3.2056

0.0002 0.0001 0.0003 0.0001 0.0001 0.0001 0.0001 0.0001 0.0022

Weighted Statistics: Adjusted R 2 =0.6906, SE of regression 0.1429, DW =2.57 Unweighted Statistics: Adjusted R2 =0.6814, SE of regression =0.1429, DW =2.28

TABLE 4 REGRESSION RESULTS FOR EQUATION (10) CYCLICAL VARIABLE OMITTED (i = 1,..... ~)

Variable Constant (common) AR(l) Specific e* Variables: FBT TCF WPP PPR PCC NMM

MET MAC OTH

Coefficient

t statistic

P-Value

-1.7925 0.5836

-4.4737 6.1488

0.0000 0.0000

1.4045 1.5453 1.3459 1.4517 1.5084 1.6191 1.5867 1.5165 1.2213

4.3229 4.6058 4.1026 4.6390 4.5521 4.7308 4.7573 4.6472 3.3798

0.0001 0.0000 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000 0.0013

Weighted Statistics: Adjusted R 2 =0.6946, SE of regression 0.1444, DW =2.56 Unweighted Statistics: Adjusted R 2 = 0.6742, SE of regression = 0.1444, DW = 2.20

Economic Analysis & Policy

©2006

Vol35 Nos 1&2. March/Sept 2005

57

The estimated coefficient of the cyclical variable in Tables 1 and 3 is not statistically significant and, moreover, it is not of the expected sign. One can speculate that average overtime hours does not adequately capture cyclical effects, rather than that cyclical effects are unimportant. The choice of a suitable cyclical variable thus remains an issue to be resolved. Tables 2 and 4 give results with the cyclical term omitted. Both Wald and likelihood ratio tests lead to easy rejection at conventional significance levels of the null hypothesis of no subdivision specific effects in both models (9) (fixed effects) and (10). In both models, the results reveal estimated coefficients of the effective corporate tax rate variable that are both positive and highly statistically significant. Hence the ~esults support the hypothesis of forward shifting of the corporate income tax, together with significant subdivision specific effects. The existence of the latter could be explained by, for example, differences in the level of competition, industry concentration and regulation across manufacturing subdivisions. A more competitive environment would make it more difficult for firms to shift forward changes in company taxation. It is also of interest to consider the degree of forward shifting of the corporate income tax implied by the estimation results. To do this we can conveniently start from a situation in which all variables are at their sample means and suppose there is a one percentage point increase in the effective corporate tax rate ('e'); this represents a 5.3533 per cent increase in 'e' over its mean value of 0.1868. In terms of the Table 2 estimates this would imply a 1.9888% increase in the ratio of the index of prices to the index of expected unit labour costs. Assuming the index of expected unit labour costs remain unchanged, this indeed would represent an increase in the mark-up over unit labour costs of 1.9888%. In other words about 37.15% of the increase in the effective corporate tax rate would be shifted forward.

6. CONCLUSIONS The hypothesis tested in the model presented in this paper is that variations in the gross mark-up on unit labour costs in part reflect the goal of maintaining a constant target net of tax profits share in value added. Subject to the limitations noted above, the preliminary empirical results derived support the hypothesis and suggest significant forward corporate income tax shifting inAustralian manufacturing companies within the dataset. Significant subdivision effects were also found to exist. The incidence of company taxes has considerable implications for the analysis and formulation of tax policy. Most directly, the incidence issue is central to debate about the role and nature of corporate income taxation, and its relationship with other forms of taxes. Arguments in support of the integration of company and shareholder taxes focus on the efficiency and equity implication arising from the 'double taxation' of dividends inherent in the 'classical' system of company taxation. The acceptance of such arguments played an important role in the introduction of an imputation tax system in Australia in 1987. However, the legitimacy of such arguments is called into question if, as the empirical findings in this paper suggest,

58

Economic Analysis & Policy

©2006

Vol 35 Nos 1&2, March/Sept 2005

there exists significant forward shifting of the tax onto consumers in the form of higher product prices. In this context, the corporate income tax resembles a general sales tax that is regressive as to income. A more informed tax reform debate requires a more serious consideration of the implications arising from the likely existence of forward shifting of corporate income taxes.

REFERENCES Agapitos, G. (1979), 'The Impact ofTaxes on Price Inflation in UK Manufacturing' , Bulletin ofEconomic Research, vol. 31, no. 11 (May), pp. 14-23. Asimakopulos, A. and Burbidge, J.B. (1974), 'The Short-period Incidence of Taxation', Economic Journal, vol. 84, no. 334, pp. 267-88. Atkinson, A.B. and Stiglitz, J.E. (1980), Lectures in Public Economics, McGrawHill, London. Auerbach,AJ. (2005), 'Who Bears the Corporate Tax? A review of what we know' , NBER Working Papers W11686. Beath, J. (1979), 'Target Profits, Cost Expectations and the Incidence of Corporate Income Tax' , Review of Economic Studies, vol. 46, no. 144, pp. 513-25. Bhatia, K.B. (1988), 'Tax Incidence in a Hierarchical Model', Journal of Public Economics, vol. 37, no. 2. Coutts, K., Godley, W. and Nordhaus, W. (1978), Industrial Pricing in the United Kingdom, Cambridge University Press, Cambridge. Daly, K. and Hart, N. (1994), 'Mark-up Pricing and the Forward Shifting of the Corporate Income Tax', The Australian Economic Review, July-September, pp.45-54. Damania,D. and MOOr,D. (1992), 'The Short-Period Incidence ofTaxation Revisited', Cambridge Journal of Economics, vol. 16, no. 2, pp. 195-206. Davis, J .M. (1972), 'The Short-Run Shifting of Company Taxation in the United Kingdom', Oxford Economic Papers, vol. 24, no. 2 (July), pp. 259-286. Duransky, R. (1972), 'The Short-Run Shifting of the Corporate Income Tax in the United States', Oxford Economic Papers, vol. 24, no. 3 (November), pp. 357-371. Gordon, RJ. (1967), 'The Incidence of the Corporation Income Tax in U.S. Manufacturing, 1925-62', American Economic Review, vol. 57, no. 4, pp.731-58. Fullerton, D. and Metcalf, G.E. (2002), 'Tax Incidence', in Auerbach, AJ. and Feldstei n, M. (eds) ,Handbook ofPublic Economics, Vol ume 4, Elsevier Science B.V, New York, chapter 26. Gravelle, J.G. and Kotlikoff, LJ. (1989), 'The Incidence and Efficiency Costs of CorporateTaxation When Corporate and Noncorporate Firms Produce the Same Good', Journal of Political Economy, vol. 92, no. 4. Gravelle, J. (1995), 'The Corporate Income Tax: Economic Issues and Policy Options', National Tax Journal, vol. 48, no. 2, pp. 267-277.

Economic Analysis & Policy

©2006

Vol35 Nos 1&2, March/Sept 2005

S9

Greene, W.H. (2000), &onometric Analysis, 4th edition, Prentice Hall, Upper Saddle River. Groenewegen, P.D. (1984), Public Finance in Australia, Prentice-Hall, Sydney. Harberger, A.C. (1962), 'The Incidence of the Corporation Income Tax', Journal of Political Economy, vol. 70, no. 3, pp. 215-40. Harding, A. and Warren, N. (1999), 'Who Pays the Tax Burden in Australia? Estimates for 1996-97', NATSEM Discussion Paper no. 39. Krzyzaniak, M. and Musgrave, R.A. (1963), The Shifting ofthe Corporation Tax: An Empirical Study of its Short-Run effect upon the Rate of Return, Johns Hopkins Press, Baltimore. Krzyzaniak,M. and Musgrave,R.A. (eds) (1966), Effects ofCorporate Income Tax, Wayne University Press, Detroit. Krzyzaniak, M. and Musgrave, R.A. (1968), 'Incidence of the Corporation Income Tax in U.S. Manufacturing: Comment', American Economic Review, vol 58, no. 5 (December), pp. 1358-1360. Laramie,AJ. (1991), 'Taxation and Kalecki 's Distribution Factors' ,Journal ofPost Keynesian Economics, vol. 13, no. 4, pp. 583-94. McLure, C.E. (1975), 'General Equilibrium Incidence Analysis: The Harberger Model after Ten Years' ,Journal ofPublic &onomics, vol. 4, no.2 (February), pp. 125-161. McLure, C.E. and Thirsk, W.R. (1975), A Simplified Exposition of the Harberger Model I: Tax Incidence', National Tax Journal, vol. 28, no. 1, pp. 1-27. Miyagiwa, K. (1988), 'Corporate Income Tax in the Presence of Sector-Specific Unemployment', Journal of Public Economics, vol. 37, no. 1. Parai ,A.K. (1988), 'The Incidence of Corporate Income Tax under Variable Returns to Scale', Public Finance, vol. 43, no. 3. Piggott, J. (1983), 'Some General Equilibrium Implications of Tax Substitutions: A Numerical Assessment for Australia', in J.G. Head (ed.), Taxation Issues in the 1980s, Australian Tax Research Foundation, Sydney. Piggott, J. and Whalley, J. (1987), 'Interpreting Net Fiscal Incidence Calculations', Review of Economics and Statistics, vol. 69, no. 4, pp. 685-94. Sebold, ED. (1979), 'The Short-Run Shifting of the Corporation Income Tax: A Simultaneous Equation Approach', Review of Economics and Statistics, vol. 61,no.3,pp.401-09. Shoven,J.B. & Whalley,J. (1973), 'General Equilibrium withTaxes: A Computational Procedure and an Existence Proof', Review of Economic Studies, vol. 60, October. Shoven,J.B. and Whalley,J. (1984), 'Applied General Equilibrium Models of Taxation and International Trade', Journal of Economic Literature, vol. 22, September. Sorenson, P.B. (1995), 'Changing Views of the Corporate Income Tax' ,National Tax Journal, vol. 48, no. 2, pp. 279-294.

60

Economic Analysis & Policy

©2006

Vol 35 Nos 1&2, March/Sept 2005

Whalley, J. (1997), 'The Incidence of the Corporate Tax Revisited', NBER Working Paper 97-7.