Australian Family Tax Reform and the Targeting Fallacy

Authors

  • Patricia Apps,

    1. Law School, The University of Sydney
    Search for more papers by this author
  • Ray Rees

    1. Center for Economic Studies, University of Munich
    Search for more papers by this author
    • *

      We are grateful to the editor and two referees for helpful comments on an earlier draft of this article. The description for the contents page: Reforms to Australia's income tax and targeted family benefit system have shifted the burden of taxation unfairly and inefficiently. The authors propose returning to progressive individual taxation and universal family payments.


Abstract

Over recent decades, Australia's highly progressive, individual-based taxation of families has been replaced by a system that tends towards joint taxation with an inverted U-shaped rate scale. The reform has been implemented by introducing family-income-targeted child payments (now Family Tax Benefit Part A) and by lowering tax rates on higher incomes. The new system has shifted the burden of taxation to two-earner families on low and average wages and, in particular, to working married mothers as second earners. For reasons of fairness and efficiency, we propose returning to more progressive individual taxation and universal family payments.

1. Introduction

In the early 1980s support for dependent children was provided in the form of universal family allowances funded by a highly progressive rate scale on personal incomes. Since then we have seen:

  • • the introduction of income tests for child support payments based on family income.1 As a consequence, Australia now has a family tax system that closely approximates one of joint taxation under a rate scale that exhibits an inverted U-shaped profile—the highest marginal tax rates apply across average incomes and to the income of the second earner in the family. The latter is a characteristic feature of joint taxation.
  • • significant tax cuts at high income levels, together with the expansion of the Low Income Tax Offset (LITO). These reforms have contributed to replacing the strongly progressive rate scale applying to personal income with one that is less progressive in general and no longer strictly progressive over certain ranges. The withdrawal of the LITO at 4 cents in the dollar above $30 000 creates a new 34 cents rate at that point, which then falls to 30 cents once it is fully withdrawn.

The introduction of joint-income-tested family payments and the new rate scale on personal incomes has had the effect of funding tax reductions on top incomes and transfers to families at the bottom of the income distribution by raising taxes on the population of earners between these two groups, in a way that bears particularly heavily on the two-earner family and working married mothers. The set of policy measures to achieve this outcome—the Personal Income Tax scale in combination with the LITO, the Medicare Levy and Family Tax Benefit (FTB) system—have resulted in what appears to be a very complex system. However, the changes amount simply to a change in the rate structure and base of the family tax system, which could have been made directly and openly. Such a degree of simplicity and transparency was, however, apparently not the intention. Similar developments have occurred in other countries, including especially the United States and the United Kingdom.2

In this article, we review the main elements of these changes and their rationalisations and isolate their effects. Income-tested family payments are often supported by the argument that universal payments are necessarily more ‘costly’. It is clear from optimal tax theory, however, that the idea of achieving a ‘cost saving’ by targeting reflects a misunderstanding of the trade-off between efficiency and equity in tax design. Nevertheless, the idea has been influential in promoting reforms that have now completely eliminated universality. For this reason, Section 2 explains at some length the error in the logic of the cost-saving argument for targeting family payments, drawing on the theory of optimal taxation.

We also know from optimal tax theory that the merits of a particular tax system can be assessed only on the basis of reliable estimates of behavioural responses—changes in labour supply in response to the incentives created by the tax system—and information on household living standards.3 It is always possible to construct a model that supports a particular direction of reform, or a particular ideology, by specifying an appropriate set of assumptions, but these may or may not be supported by the data, and it is important to check this. It is therefore essential to be familiar with the data. Section 3 develops this point in some depth. Section 4 presents an empirical analysis of the effects of the Personal Income Tax scale, LITO, Medicare Levy and FTB system on the structure of tax rates, and reports results showing the high taxation of women as second earners in low- and average-wage families. Section 5 contains concluding comments.

2. Tax Rates and Targeting

2.1 The ‘Cost’ of Universal Family Payments

The view that universal family payments are ‘more costly’ than income-tested payments appears in the Australian Treasury's 2008 Consultation Paper Australia's Future Tax System. In discussing Family Tax Benefit Part A (FTB-A), which is withdrawn on joint income, and Family Tax Benefit Part B (FTB-B), which is withdrawn on the second earner's income, the Treasury argues:

If the primary purpose of the payments is considered to be part compensation for the direct and indirect costs of having children, it could be argued that they should not be income-tested. While this would lower effective marginal tax rates (EMTRs) … and possibly improve workforce incentives, it would be more costly. On the other hand, if the payments are directed at reducing child poverty, arguably they could be more tightly targeted.

This statement implies that universal family benefits are ‘more costly’ than income-tested benefits in terms of net government expenditure. Superficially this seems to be self evident. Giving all households a fixed transfer is clearly more ‘expensive’ in this sense than giving only some households that transfer, while giving all other households either a smaller transfer or none at all. However, the argument fails to understand that any tax system that gives a transfer, and then withdraws it at so many cents in the dollar, is equivalent to a system with a given universal payment and a particular structure of marginal tax rates. It is not the ‘universality’ of the payment, but the value of the payment and the structure of marginal tax rates that is the relevant basis for evaluating the tax system. In this sense, it is irrelevant whether or not a tax system gives universal benefits to families who ‘do not need’ them. A given structure of marginal tax rates withdraws these universal benefits in a specific way. Given the level of the tax revenue requirement, all tax structures that meet this requirement are equivalent in terms of ‘cost’, in the sense implied by the quotation.

What matters, as is made clear by the theory of optimal income taxation,4 is the way in which a particular tax structure trades off fairness of the distribution of tax burdens across households against deadweight welfare losses arising from its effects on work incentives. One tax structure is in this sense more costly than another if, for a given degree of income redistribution, it generates a larger welfare loss because of its greater adverse effects on work incentives. In this sense, FTB-A is very probably more costly than a system of universal payments financed by an individual-based progressive income tax. We develop this point at some length in this article.

In a piecewise linear tax system of the type that exists in Australia and almost all other countries, the amount of tax an individual pays on a given gross income can be written as a function of just two tax parameters: a lump sum and a marginal rate. This is an important point in light of the often confused discussion of ‘churn’ (see Australian Treasury 2008, p. 113), which fails to recognise that targeting is simply a policy instrument for changing the ‘menu’ of lump sums and marginal rates. We therefore show this diagrammatically for the simplest case of a linear income tax on a single-person household. We go on to present a numerical example for a progressive marginal rate schedule. Finally, we show the effect of targeting on joint income in a two-person household.

2.2 A Single Person Household and Targeting

Figure 1a illustrates the structure of a linear income tax. The figure plots net income, x, as a function of gross income, y, for a set of households with differing income levels, 0 ≤ y ≤ yn. The universal transfer is denoted by a. The household's net income after tax and transfer is given by the function:

image

which is represented by the line aL. The slope of the line is (1 − t), with t the marginal tax rate (MTR). A household with y = 0 is completely dependent on a.

Figure 1.

Tax Systems and Targeting

Now consider what happens when the government decides to withdraw the universal transfer, a, on incomes above the level ya at a rate, r, and to use the revenue thus ‘saved’ to cut the tax rate from t to t while keeping total tax revenue the same. Income above ya is taxed at the rate t′ + r until an income level is reached at which the additional tax paid, r(y − ya), exactly equals a. Denoting this income level by yb, we have yb = ya + a/r. The MTR is no longer constant across incomes but has an inverted U-shaped profile—the higher rate t′+ r applies over the ‘middle’ income range and the lower rate, t′, over the bottom and top income ranges.

This is illustrated in Figure 1b. Households with incomes in the range (0, ya) receive a and pay the tax rate t. Those with incomes in the range (ya, yb) receive a, and pay t on their incomes up to ya and t′+ r on their incomes y − ya. Finally, households in the range (ya, yn) receive a, and pay t on their incomes up to ya, t′+ r on their incomes yb − ya, and t on their incomes y − yb. The linear tax represented by the line aL has been replaced by the piecewise linear tax resulting in a budget constraint corresponding to the kinked line abcd. The line aL in the figure represents a linear income tax with the same transfer but the lower marginal tax rate, t′ < t, so that along it x = a + (1 − t′)y. It obviously generates less tax revenue than aL.

The new non-convex piecewise linear tax system can be described by the function

image
image
image

This policy is equivalent to introducing a ‘menu’ of three different linear income tax systems, each defined by a lump sum5 and marginal rate as follows:

image
image
image

where:

image
image
image

and t1= t, t2= t′+ r, t3= t.

Households with an income below ye are better off (the budget line abe lies above aL); those with an income in the interval (ye, yf) are worse off (the budget line ecf lies below aL); and those with incomes above yf are better off (the budget line fd lies above aL) by an amount that rises with their income level.6 It is always possible to find a tax-rate schedule such as abcd, which, by choosing a lower basic tax rate and withdrawing the universal benefit at an appropriate rate from an appropriate threshold, will shift the tax burden to the middle while raising the same tax revenue.

It is straightforward to show that targeting can lead to a similar distributional outcome when the pre-reform income tax is convex piecewise linear, that is, when MTRs are progressively increasing as in the case of the Australian Personal Income Tax scale. We illustrate this case with a numerical example.

Consider a hypothetical economy in which average annual earnings rise from $20 000 in quintile 1 (the bottom 20 per cent of the distribution) to $200 000 in quintile 5 (the top 20 per cent), and each individual receives a cash transfer of $20 000 funded by a progressive rate scale, as shown in Panel 1 of Table 1. The table also reports the implicit lump-sum payments (the counterparts to a1, a2 and a3 in the above example). Assume there are no behavioural effects (that is, no change in labour supplies) and that the distribution can be represented by the average in each quintile.

Table 1.  Marginal Tax Rates, Cash Transfers and Lump Sums ($pa)
Income $pa 20 00040 00060 00080 000200 000
  1. Note: MTR, marginal tax rate.

1. Pre-reformMTR%0.025.025.025.050.0
Cash transfer20 00020 00020 00020 00020 000
Lump sum20 00025 00025 00025 00050 000
2. Reported reformMTR%012.512.512.525.0
Cash transfer20 00015 00010 00050000
3. True reformMTR%0.037.537.537.525.0
Cash transfer20 00020 00020 00020 00020 000
Lump sum20 00027 50027 50027 50050 000

Now suppose the government thinks that it can reduce ‘costs’ by withdrawing the cash transfer of $20 000 at a rate of 25 cents in the dollar above a threshold income of $20 000. This halves the government's tax revenue requirement. Transfers can now be financed by MTRs that are half those of the pre-reform scale. Panel 2 of Table 1 shows the new MTR scale and distribution of transfers that can be reported by a government that wants to claim that it has cut taxes and reduced ‘middle class welfare’. Panel 3 shows the true reform. The progressive rate scale has been replaced by an inverted U-shaped MTR scale, with the universal cash transfer left in place. The overall effect of the reform is to shift the tax burden to the middle, as reflected by the average tax rates (ATRs) in Table 2. ATRs rise in quintiles 2 to 4 and fall in quintile 5. The representative individual in quintile 5 gains by $15 000.

Table 2.  Average Tax Rates
Income $pa 20 00040 00060 00080 000200 000
  1. Note: ATR, average tax rate.

Pre-reformATR%−100.0−37.5−16.7−6.325.0
Post reformATR%−100.0−31.3 −8.3 3.117.5

It is clearly mistaken to say that the targeted payment is ‘less costly’. If both systems have the same administrative costs,7 the only relevant criterion on which to judge the change is in terms of the effects on fairness of the income distribution and the change in work incentives, as shown in optimal tax theory. In general, the redistribution from middle to top would be regarded as regressive, that is, socially costly. Therefore, the beneficial incentives from reducing the top marginal tax rate must be sufficiently higher than the negative effect on work incentives of the higher tax rates across the middle to at least offset the social cost of the growth in inequality of the net income distribution. Thus, support for a reform that shifts the tax burden from the top to a lower segment of the income distribution must be based on empirical evidence on wage elasticities. It needs to be shown that labour-supply elasticities (or at least earned-income elasticities)8 at the top are sufficiently larger than those lower down that the reform achieves net efficiency gains that at least offset the welfare loss from the rise in inequality.

This is recognised by Brewer, Saez and Shephard (2008) in their paper prepared for the Report of the UK Commission on Reforming the Tax System for the 21st Century, chaired by Sir James Mirrlees (the ‘Mirrlees Review’).9 The authors first argue that the relevant elasticities are not those of labour supplies with respect to the net of tax wage rates, but rather those of gross earnings with respect to the income tax rate. The reason is that a change in the tax rate on (declared) earned income can cause responses such as tax evasion and avoidance, changed effort intensity and productivity (an efficiency wage effect), differences in readiness to accept promotion and increased levels of responsibility, switching between taxed and untaxed forms of remuneration, and so on. The paper then uses data on the top 1 per cent income share and the changes in the top rate of income tax to argue that the earnings elasticity of this group is so large that an increase in the top tax rate would actually reduce total tax revenue, and therefore they propose no change to this tax rate.10 The basis for this argument is the steadily rising trend in the share of income going to the top 1 per cent over the period 1978–2002, with 1978 representing a turning point from a previously falling trend over the period 1962–77. In 1979 the top rate of income tax fell from around 80 per cent to 60 per cent, and in 1988 there was a further fall to just over 40 per cent, though outside these years the tax rate was roughly constant.

The key question of course is the extent to which the rising income share was due to the tax rate changes—hours elasticities were as usual close to zero, the standard empirical result. Brewer, Saez and Shephard's (2008) paper shows that the income share of the next 4 per cent of top earners also increased steadily over this period, even though marginal effective tax rates on this group were mainly constant; however, the paper suggests that this may have been because of the incentive effects of the prospect of achieving promotion to the top 1 per cent. On the other hand, the paper also recognises (on p. 16) the argument that the change in high incomes was not entirely a result of the cuts in the marginal effective tax rate, and may have been caused by other reforms that were enacted by the Thatcher administration that were favourable to high incomes.

Important examples of such reforms were the wave of privatisations of publicly owned firms, which hugely increased managerial salaries in the industries concerned, and the deregulation and liberalisation of the banking and finance sectors, which had very marked effects on income levels in those sectors. Nevertheless, the paper appears to accept, after using a number of estimation methods, an earnings elasticity of about 0.46 for the top 1 per cent of income earners, with a value of 0.25 for everyone else. The decision of the UK Government in April 2009 to raise the top income tax rate on incomes above £150 000 to 50 per cent, which is expected to raise an additional £6 billion of tax revenue in a full year, should allow further testing of these assumptions.

2.3 Two-Person Households and Joint-Income Targeting

We now illustrate the effects of targeting on the basis of joint income, by considering an economy in which a household can switch from being single-earner to two-earner—in other words, a household can switch ‘type’ by changing the labour supply of the female partner as second earner. For the purpose of exposition, we construct a numerical example for an economy in which male labour supply is fixed and there is an equal split between the following two types:

  • • single-earner household: the male partner works full time in the market and the female works full time at home providing child care and related services.
  • • two-earner household: both partners work full time in the market and buy in substitute services for child care and home production.

We assume the income of the male partner as primary earner rises from $20 000 in quintile 1 to $200 000 in quintile 5, as in the preceding example, while the second income rises from $20 000 in quintile 1 to only $100 000 in quintile 5. A progressive, individual-based income tax funds a $20 000 cash transfer to each household. The MTRs and cash transfer under this ‘pre-reform’ tax system are shown in Panel 1 of Tables 3a and 3b for a single- and two-earner household, respectively. Note that, because of the increase in the tax base with the entry of married mothers into the workforce, the universal cash payment of $20 000 can now be financed by lower MTRs. Note also that married mothers with incomes above the tax-free threshold, by switching from untaxed home production to taxed market work, partly subsidise the transfer payment to the single-earner household.

Table 3a.  Single-Earner Household: Marginal Tax Rates and Cash Transfers ($pa)
Primary income $pa 20 00040 00060 00080 000200 000
  1. Note: MTR, marginal tax rate.

1. Pre-reformMTR%0.020.020.020.040.0
Cash transfer20 00020 00020 00020 00020 000
2. Reported reformMTR%07.07.07.014.0
Cash transfer20 00015 00010 00050000.0
3. True reformMTR%032.032.032.014.0
Cash transfer20 00020 00020 00020 00020 000
Table 3b.  Two-Earner Household: Marginal Tax Rates on Second Income and Cash Transfers ($pa)
Second income $pa 20 00040 00060 00080 000100 000
  1. Note: MTR, marginal tax rate.

1. Pre-reformMTR% 2nd income0.020.020.020.020.0
Cash transfer20 00020 00020 00020 00020 000
2. Reported reformMTR% 2nd income07.07.07.07.0
Cash transfer15 0005000000
3. True reformMTR% 2nd income25.032.07.07.07.0
Cash transfer20 00020 00020 00020 00020 000

Again, suppose that the government thinks that it can reduce ‘costs’ by withdrawing the cash transfer of $20 000 at a rate of 25 cents in the dollar above a threshold joint income of $20 000; that is, it applies the withdrawal rate to joint income. The effect is dramatic. Assuming no behavioural effects, the government can claim a ‘cost’ saving of 65 per cent. Applying a proportional reduction to the pre-reform marginal rates, the government can report a new rate scale of only 7 cents in the dollar on incomes up to $100 000 and 14 cents in the dollar on incomes above $100 000, as shown in Panel 2 of Tables 3a and 3b.

The cash transfer is fully withdrawn at a primary income of $100 000 for the single-earner household, but at a primary income of only $50 000 for the two-earner household. The much greater loss for the low- and average-wage two-earner family can be concealed by reporting the reform by household income. This is a common practice that allows, for example, the two-earner family in quintile 2 of primary income to be regarded as just as well off as a single-earner family in quintile 4 in which only one partner needs to work full time in the market to earn the same household income.

The large shift in the tax burden towards the middle, and especially towards low- and average-wage two-earner families, is indicated by the pre- and post-reform distribution of ATRs in Tables 4a and 4b. The additional tax revenue from these families funds tax cuts in quintile 5 that reduce ATRs from 18 per cent to 9.8 per cent for the single-earner household, and from 17.33 per cent to 8.4 per cent for the two-earner household.

Table 4a.  Single-Earner Household: Average Tax Rates
Primary income $pa 20 00040 00060 00080 000200 000
  1. Note: ATR, average tax rate.

Pre-reformATR%−100.00−40.00−20.00−10.0018.00
Post-reformATR%−100.00−34.00−12.00  1.00 9.80
Table 4b.  Two-Earner Household: Average Tax Rates
Second income $pa 20 00040 00060 00080 000100 000
  1. Note: ATR, average tax rate; hh, household.

Pre-reformATR% 2nd income  0.0  4.0013.3315.0016.00
ATR% hh income−50.00−15.00−3.33 6.4017.33
Post-reformATR% 2nd income 25.00 28.5021.3311.50 5.60
ATR% hh income−12.50 −2.75 4.67 5.25 8.40

3. Optimal Income Taxation of Couples: Do the Facts Matter?

The question we now consider is whether the direction of reform described by the examples in the preceding section can be supported on the basis of any set of plausible assumptions. Our introductory comments imply that it cannot. In this section we present data on female labour supply and the allocation of time to child care (Subsection 3.1) and on earnings distributions (Subsection 3.2) to support this view. Subsection 3.3 provides a brief overview of developments in optimal tax theory that are relevant to the taxation of families.

3.1 Female Labour Supply and Parental Child Care

According to the data for couples in the Household, Income and Labour Dynamics in Australia Survey, Wave 5, 2005 (HILDA 2005), 92 per cent of males and 73 per cent of females of prime working age, defined as 25 to 59 years, are employed—a difference of less than 20 percentage points. However, almost all prime aged married males—85 per cent—work full time while only 34 per cent of prime aged married females are in full-time work. The result is that married women work about half the hours of married men in this prime age category. These figures also reveal a high degree of heterogeneity in female hours. While male labour supply shows relatively little variation, with almost all men working full time, females are distributed more evenly between zero hours and full time work.

In empirical work on labour supply, heterogeneity in female hours is found to be strongly associated with the presence of children, as we would expect. However, a fact that is often missed is that significant heterogeneity emerges with the arrival of the first child and thereafter tends to persist, which means we need to search for other explanations. To see the significance of taking a family life-cycle approach, it is useful to compare the labour supplies of couples across four broad family phases: pre-child phase; youngest child 0–4 (pre-school) phase; youngest child 5–17 phase; and a post-child phase (no child under 18 years present). Figure 2 presents histograms of male and female hours of market work in these phases based on HILDA 2005 data for ‘usual weekly hours of work’ of partners aged from 25 to 59 years.

Figure 2.

Labour Supplies across Family Life-Cycle Phases

In the first phase, the profiles closely match—partners of prime working age tend to work full time and for the same hours.11 In the 0–4 child phase there is very little change in male hours, but female hours fall dramatically. About 40 per cent of mothers in this phase work less than 5 hours per week and less than 20 per cent remain in full-time employment. In the 5–17 child phase, 24 per cent continue to work less than 5 hours, and full-time employment rises to 31 per cent. In the post-child phase, the proportion of females reporting working less than 5 hours per week rises to 33 per cent, and about 36 per cent work full time. Thus, in the post-child phase female labour supply remains well below its pre-child level, indicating a high degree of ‘persistence’ in the labour-supply decision made in the pre-school phase.12

From this life-cycle organisation of the data, it is clear that variation in female labour supply cannot be attributed simply to demographic variables. More plausibly, it can be argued that the observed changes in female labour supply are driven by the economics of investment in the care and education of children, much of which is directly influenced by government policy, and by the gender wage gap.

The argument is straightforward. In phase 1 there is a low demand for home-produced goods and services because there are few of the kinds of goods and services couples in this phase consume for which there are not good, affordable market substitutes, and so there is a low demand for domestic labour in this phase. Put simply, there is nothing much to do in the home, and so it would make no sense for either partner to specialise in household production, or for singles who have not yet had children to do so.

The arrival of children creates a very large demand for their care and for investment in their education. Although the government has taken over much of the role of investing in the education of children once they reach school age, it has largely neglected those under school age. The result is that market child care can be very costly, primarily because of the long-term failure of government to invest in the sector's infrastructure in line with that of other education sectors, in an imperfect capital market.13 At the same time, the income of the second earner is subject to a high average tax rate, as shown in the sections to follow. These policies undermine the capacity of a second earner to finance child care, especially when her future wage is uncertain and she faces an imperfect capital market in which the borrowing rate is above the lending rate.

Child care can be provided by some combination of parental time and services bought in from the market. The opportunity cost of parental child care is determined by the present value of the current and future net market income foregone. The higher the effective tax rate on the second earner and the more costly and difficult it is to access market child care, the more of it will be provided at home, other things being equal. The demand for child care then implies a large induced demand for household production and introduces a fundamental change in the work choices of couples, which will reflect the relative costs of each partner's time. Furthermore, withdrawal from the labour market by the female as the lower wage partner in phase 2 can lead to a lower wage because of loss of human capital and career possibilities.14 Under these conditions, it is not surprising to find a large gender gap in market hours and a high degree of heterogeneity in female hours.

The substitution of home child care for market work after the first child is evident from time-use data. Table 5 reports average weekly female hours allocated to market work, home child care and domestic work (laundry, cooking, cleaning etc.) in the child 0–4 and child 5–17 phases.15 Of a total of 62 hours spent on work at home in the 0–4 phase, more than 35 hours are on child care.

Table 5.  Female Market, Domestic and Child-Care Hours Per Week
Child 0–4 phaseMarket hours16.9
Domestic hours26.2
Child-care hours35.7
Child 5–17 phaseMarket hours21.0
Domestic hours24.0
Child-care hours20.2

These data have implications for the modelling approach needed for estimation of the parameters of behavioural responses to policy changes. Most importantly, it makes no sense to specify a model that ignores home child care as a substitute for female labour supply and bought-in child care, especially in an economy with a poorly developed child-care sector. These data indicate that it is this substitution that drives female wage elasticities.

The data also serve to highlight the fact that a household in which one partner specialises in home production pays for child care through the opportunity cost of her time, and her contribution to home production is remunerated implicitly through intra-household exchange of home output for market goods funded by the earnings of the partner in market work. Ideally, therefore, the household should be modelled as a small economy engaged in production and intra-household exchange, with heterogeneity reflecting varying degrees of production and exchange across households.

It is also important to note that neither an income tax nor a consumption tax can be applied to a tax base that includes household production, and so households with the same wage rates and demographic characteristics will pay different amounts of tax, depending on the second earner's choices between market and domestic work. Those who substitute household production for market work avoid paying tax on the implicit income derived from domestic labour, and they avoid consumption taxes on the output. In respect of the latter, it is important to recognise that a shift from income towards consumption taxation does not represent a solution to the problem of taxing couples. Unfortunately, the literature on indirect taxation can often give misleading advice because it ignores the policy implications of the untaxed status of household production and heterogeneity in female labour supply.

3.2 Household Income: An Unreliable Indicator of Household Welfare

If families with the same wage rates and demographic characteristics were observed to make the same time-allocation decisions, then, all else being equal, we could reasonably expect to find a strong correlation between household income and family welfare within a demographic group. Under these conditions, progressive joint taxation would not necessarily be unfair in terms of the distribution of tax burdens across households. It would, of course, widen the net-of-tax gender wage gap and could therefore be expected to disadvantage women in general by widening inequality within the family. However, it would not discriminate on the basis of employment status because, at given wage rates, all would be the same type.

However, with heterogeneity in female labour supply across households with the same wage rates and demographic characteristics, this is no longer the case. Furthermore, the problem of errors in a welfare ranking defined on household incomes becomes especially serious when, as the analysis to follow will show, the profile of male earnings, and therefore also of primary earnings, for full-time work is relatively flat across the middle of the distribution and then rises sharply towards the top.

The analysis is based on data for a sample of 1608 ‘in-work’ couple income units with a dependent child drawn from the Australian Bureau of Statistics (ABS) 2005–06 Survey of Income and Housing (SIH06). The sample is selected on the criterion that at least one parent is employed. Families with negative incomes from investments or unincorporated enterprises, or with incomes below $15 000 per annum are also excluded. All incomes are indexed to the 2008–09 financial year.

The sample is split into three household types:

  • SE: single-earner household;

  • PT: two-earner household with second earner employed part time; and

  • FT: two-earner household with second earner employed full time.

Table 6 reports the distribution of types by income of the primary earner, defined as the partner with the higher private income.16 The male is the primary earner in more than 88 per cent of records. Although there is a somewhat stronger representation of the two-earner household in the middle quintiles, the main point to note is that all three types are distributed relatively smoothly across the distribution of primary income.

Table 6.  Household Type by Primary Income ($pa)
Quintile12345All
  1. Note: SE, single-earner household; PT, two-earner household with second earner employed part time; FT, two-earner household with second earner employed full time.

Primary income $pa34 81550 19763 10579 902150 62076 233
SE%40.131.326.728.340.532.9
PT%37.937.444.240.435.639.0
FT%22.031.332.131.424.028.0

Figures 3a and 3b show graphically the distribution of hours and earnings of each of the three household types, by quintiles of primary income. Because the male is the primary earner in the vast majority of cases, the distribution of primary- and second-earner hours reflects gender differences in labour supply and the high degree of heterogeneity in the labour supply of married mothers at every level of primary income, and across families of the same size. The number of dependent children is 1.93, 1.88 and 1.83 in the SE, PT and FT household type, respectively. While there is more variation with respect to age of youngest child, there are clearly many families with the same demographic characteristics making very different time-use choices.

Figure 3.

Labour Supplies and Earnings by Primary Income and Household Type

The crucial feature of the earnings profiles is the relatively flat segment across the middle quintiles. This means that the position of a family in a ranking defined on household income will be very sensitive to the earnings, and therefore to the labour supply, of the second earner because it will take only a small increase in her earnings to shift a family from a low percentile of family income to a significantly higher point in the distribution. This is illustrated in Table 7. The table shows the quintile distributions of the three household types by household income. In contrast to the ranking by primary income, the majority of single-earner families are in the lower quintiles because household income omits the family's implicit income from household production.

Table 7.  Household Type by Household Income ($pa)
Quintile12345All
  1. Note: SE, single-earner household; PT, two-earner household with second earner employed part time; FT, two-earner household with second earner employed full time.

Household income $pa42 89767 21986 397108 410196 060100 632
SE%60.335.922.725.220.832.9
PT%27.040.142.141.836.239.0
FT%12.717.134.942.941.828.0

The upper limit of quintile 1 is $55 956 and the lower limit of quintile 4 is $96 369. A single-earner family with an income of, say, $50 000 will be located in quintile 1. If the family switches ‘type’, with the second partner working full time for the same income, the family will be re-ranked from quintile 1 to quintile 4. If the household has a preschool child, much of the second net income may be spent on child care. Clearly, such a household cannot be said to have the same standard of living as another in which only one parent needs to work full time to earn $100 000 while the other works full time at home. To argue to the contrary, it is necessary to assume that home child care makes no contribution to family welfare or that market child care is costless.

3.3 Optimal Tax Theory

Contributions to the theory of optimal taxation have provided important insights into the structure of the tax design problem. However, the literature has been slow to develop models that are consistent with the data on two-parent families in an economy in which female labour supply and the allocation of time to home child care are significant and heterogeneous.

Much of the literature, and certainly that part of it which is taught in standard textbooks at the undergraduate and graduate levels, is based on the model of a household as a single individual who divides time between market work and leisure—the direct consumption of one's own time—to earn an income that is then spent on market goods. Household production is missing. The models also assume that market wage rates reflect innate abilities, thereby implying, for example, that the gender wage gap reflects a gap in innate ability. The models in Apps (1981, 1982) provide an early critique of these assumptions, presenting instead an analysis in which women are ‘crowded’ into household production by labour market discrimination. The household is modelled as a small economy engaged in production and intra-household exchange, where the terms of exchange are set by the ‘outside’ female wage. In this kind of model a joint tax system reduces the outside net wage of the female partner and, in turn, makes the terms of intra-household exchange less favourable to her.

One of the most widely cited contributions that extends the analysis to a two-person household is the linear tax model of Boskin and Sheshinski (1983). The authors derive the result that, optimally, women should be taxed at a lower rate than men. This builds on the observation, dating back to Munnell (1980) and Rosen (1977), that because women have higher compensated labour-supply elasticities, standard Ramsey arguments would imply, other things equal, lower tax rates.

Drawing on Ramsey principles does not, however, provide a conclusive argument. The optimal tax rate in a linear tax model depends not only on the efficiency effects of taxation, but also on distributional effects, and it is a priori possible that the tax rate on women could optimally be higher, despite the higher elasticities, if this tax rate were a sufficiently better instrument for redistribution than that of men. This depends on the across-household covariance between the marginal social utility of income and gross income of men and women, respectively. Boskin and Sheshinski use a model calibrated with parameter values that are representative of empirical estimates to derive the result that, when distributional effects are taken into account, the optimal tax rate on women is indeed below that on men. But this is just an example. There has been little further empirical work done to test its robustness.

More recently, Apps and Rees (1999, 2007) show that, both in the tax reform and optimal linear tax cases, the Boskin and Sheshinski result, hailed as the ‘conventional wisdom’ in this area, can be put on a firmer foundation. Given that the ratio of female to male income falls as we move through the income distribution, as indicated in the data presented earlier, and that the correlation between male and female wage rates across households is sufficiently strong (positive assortative matching), which is also evident from the data, the male tax rate is always a better instrument for income redistribution from higher wage to lower wage households. This then reinforces the effect of higher labour supply elasticities in supporting a lower tax rate on women.

It is almost a trivial result that male and female tax rates should differ. Equalising their marginal tax rates, as is done in a joint taxation or income splitting system such as those in the United States and Germany, amounts to imposing a constraint on the optimal tax problem that cannot increase, and in general will reduce, the optimised value of social welfare. Less trivial is the argument that women should be taxed at lower rates than men with the same gross income. It is often objected that ‘gender-based taxation’ does not exist and could not be introduced. This is not true. Australia has gender-based taxation in the form of FTB-B. The problem is, however, that the discrimination goes in the wrong direction—FTB-B raises the tax on the second earner, typically the female partner, instead of lowering it.

More powerful, however, is the argument that labour supply elasticities are on average higher for women because of the greater proportion of women who earn lower wages and also because of the higher degree of non-participation among women. Heckman (1993) argues that as the female wage distribution and distribution of working hours become more like those of men (assuming of course that there are changes in tax and child-care policies that allow it to do so), their labour supply elasticities will converge toward those of men.

An answer to both these arguments is a tax system that is not gender-based but rather based on progressive individual taxation, where primary and secondary earners are taxed individually but on the same progressive rate schedules. This would necessarily imply that tax rates on second earners would be below those on primary earners when their incomes are sufficiently lower as to place them in a lower tax bracket. It would also mean that ATRs would be positively related to the gap in partners’ earnings—a single-earner family would face a higher ATR on household income than a two-earner family with the same joint income. The difference would at least partly compensate for the higher level of untaxed home production in the single-earner family. Progressive individual taxation therefore has the potential to achieve a higher degree of horizontal equity, as well as a higher degree of vertical equity, for a given efficiency cost.

In the case of non-linear taxation in the tradition of Mirrlees (1971), the papers by Apps and Rees (2006), Brett (2007), Cremer, Lozachmeur and Pestieau (2007), Kleven, Kreiner and Saez (2009) and Schroyen (2003) consider the problem of extending the Mirrlees analysis to the case of two-earner households. General results are hard to find, essentially because of the complexity of the two-dimensional screening problem that arises when the productivity of each household member is the household's private information. Even in the relatively simple case of two wage types for primary and secondary earners, which is the approach adopted by Apps and Rees, Brett, and Schroyen, the multiplicity of potentially binding incentive compatibility constraints gives rise to a wide range of possible solutions. The main general result of these analyses is that the tax rates on men and women will vary with their productivity type, and so individual taxation is still in general optimal, but the tax rate on a given individual of one of the two types will also depend on the type of his or her partner. In this sense, the tax unit consists of both the individual and the couple.17

The approach adopted by Kleven, Kreiner and Saez, in simplifying the model to make the optimal tax problem analytically more tractable, is very different. It is worth discussing this model at some length because it is cited as providing authoritative support for income-tested benefits in Institute for Fiscal Studies papers for the Mirrlees Review on reforming the UK tax system.18

Key assumptions underlying the model are:

  • • There is a continuum of primary earners with a given innate productivity level. Associated with each primary earner type is a continuum of second earner types, all of whom face the same market wage rate with no variation in their innate productivity; they are all of the same ‘wage type’.
  • • Whereas primary earners choose a level of labour supply on a continuum between zero and full employment, second earners are restricted to choosing either to work full time or not at all. There are no part time jobs for second earners. Thus they have a pure participation decision.
  • • There is a fixed utility cost of working, with second earners distributed according to its value on [0,∞).
  • • The characteristics of primary and second earners, respectively market productivity and fixed utility cost of working, are independently distributed; there is no correlation of types across households.19

Each of these assumptions is counterfactual, as the data presented earlier show. There is variation in the market wage of working women; women choose market labour supplies right across the distribution from zero to full time; and there is significant positive correlation across couples between male and female wage rates.20 Indeed it is men who tend to work either full time or not at all, although, as the data presented earlier show, working full time is consistent with quite a wide range of weekly hours.21 Finally, given that for many families a major cost of going out to work is child care, this can be expected to vary with hours worked, rather than being a fixed utility cost—why else would so many women work part time? It is of course always necessary to make simplifying assumptions for modelling purposes. The resulting model ought, however, to bear some resemblance to the essential aspects of the real economy for which it is purporting to prescribe an optimal tax system.

The main result of the analysis is what the paper calls ‘negative jointness’, which can be expressed in either of two equivalent ways: the optimal marginal tax rate on the primary earner income in the two-earner household is at each value of primary income below the corresponding marginal tax rate in the single earner household; or, the implicit participation tax rate on the second earner, arising out of the increase in tax on the primary earner's income when the second earner goes out to work, falls as the income of the primary earner rises.

In an attempt to show the plausibility of this result, the paper by Kleven, Kreiner and Saez (2009) points out that most OECD countries, including those that have moved to individual tax filing, also operate family-based means-tested welfare programs with transfers being phased out with joint family income. This combination creates negative jointness. Under such a system, a woman married to a husband with a low income, and in the benefit-phase-out stage of the system, faces higher marginal and participation tax rates than a woman married to a husband with a high income, with a joint income beyond the phase-out range.22

Note, however, that a further characteristic of these systems is that, over the phase-out range of joint incomes, at each given level of primary income, the marginal tax rates on the primary earners in each of the two types of household must be equal, holding constant the income of the second earner. In a fully joint taxation system, as in the United States and Germany, the marginal tax rate on a primary earner will in general be higher if his wife works than if she does not, because her income tends to push him into a higher tax bracket.

This implies that the Kleven, Kreiner and Saez interpretation is very problematic. In the model, as just pointed out, there are two equivalent definitions of negative jointness: that the marginal tax rate of the primary earner in the two-earner household is lower than that of a primary earner with the same income in a single-earner household; and that the second earner's implicit participation tax rate is falling with primary earner income. In actual tax systems, as just noted, the first of these does not in general hold. The general point is that these two equivalent properties of the optimal tax functions in the model of the paper are not equivalent properties of the real-world tax systems, which simply reflects the fact that real-world tax systems are essentially different to that analysed in the paper. It is therefore not possible to justify some aspect of the real tax system as being ‘optimal’ simply because it is a characteristic of the optimal solution of the model in the paper.23 This shows a lack of understanding of the theory of second best.

In spite of all this, the paper by Brewer, Saez and Shephard misleadingly asserts:

Hence, the results in Kleven et al. suggest that the broad way in which tax and benefit systems of many OECD countries treat the incomes of a couple, including that of the UK, are consistent with optimal tax results.

[2008, p. 31]

4. Australian Family Income Tax and Targeting

We now turn to the structure of marginal and average rates under the Australian family income tax system, resulting from four key policy instruments:24

  • • Personal Income Tax;
  • • LITO;
  • • Medicare Levy; and
  • • FTB-A and FTB-B.

We first identify the MTR rate scale of the Personal Income Tax when combined with the LITO and then go on to show how tax rates and the tax base change when FTB-A, FTB-B and the Medicare Levy are included. Results are presented for the 2008–09 financial year.

4.1 Personal Income Tax and the Low Income Tax Offset

The left-hand side of Table 8 lists the formal MTR scale of the Personal Income Tax. The right-hand side lists the true rate scale when the LITO is included.25 The formal rate scale of the Personal Income Tax is a strictly progressive, piecewise-linear income tax. When the LITO is included this is no longer the case. The LITO raises the zero-rated threshold from $6000 to $14 000. It also raises the MTR on incomes from $34 001 to $60 000 by 4 cents in the dollar. This is followed by a lower rate of 30 cents in the dollar up to $80 000, as shown.

Table 8.  Marginal Tax Rates on Personal Income, 2008–09
Taxable incomeMTR (PIT)Taxable incomeMTR (PIT+LITO)
  1. Note: MTR, marginal tax rate; PIT, Personal Income Tax; LITO, Low Income Tax Offset.

$0–$60000.00$0–$14 0000.00
$6001–$34 0000.15$14 001–$30 0000.15
$34 001–$80 0000.30$30 001–$34 0000.19
$80 001–$180 0000.40$34 001–$60 0000.34
$180 000+0.45$60 001–$80 0000.30
 $80 001–$180 0000.40
 $180 000+0.45

The standard argument for the LITO is also that of ‘cost’ saving—it is said to be a less costly use of taxpayers’ money for assisting those on very low incomes than an increase in the zero-rated threshold, which provides a benefit to all taxpayers above the threshold. The argument again reflects a misunderstanding of the relevant criteria for evaluating a tax system, as discussed in Section 2. In this case, however, it is somewhat extraordinary that the idea could find any degree of acceptance, because successive budgets simultaneously reduced tax rates at high income levels.

The LITO is in fact an entirely redundant policy instrument that serves only to reduce the transparency of the ongoing shift towards an effective Personal Income Tax scale with higher rates across average incomes and lower rates at the top. Although the changes in each successive year are relatively small, they are always in the same direction. In the 2008–09 Budget, tax cuts provided individuals with incomes at or above $180 000 a gain of $2600. This reduced the ATR on $180 000 by 1.44 percentage points. At $60 000, the threshold income at which the LITO is fully withdrawn, the gain was only $600, a 1 percentage point reduction in the ATR at this income level. This fell to a 0.8 percentage point reduction at $75 000. The cumulative effect of changes of this kind from 2006–07 to the Government's ‘aspirational’ Personal Income Tax scale and the LITO for 2013–14 is a disproportionate shift in the tax burden towards the ‘middle’, in effect denying individuals on around average earnings an equi-proportional rate of compensation for the failure to index tax bands.26

In discussing the Personal Income Tax scale and the LITO, the Australian Treasury (2009, p. 81) notes:

Australia has a progressive personal income tax system. The personal rate scale has four personal income tax rates, as well as a zero rate of tax below the tax free threshold. In addition, other elements such as the low income tax offset (LITO) alter the effective rate of taxation. A progressive income tax could be achieved with a tax-free threshold and a single rate of tax above this point. While this would be less progressive than the current system, it would be simpler and could potentially provide better participation incentives.

From this comment, we can infer that the direction of change in the rate scale on personal income over the past decade is no accident. It reflects the policy aim of reducing taxes at high income levels. Unlike Brewer, Saez and Shephard (2008), who search for empirical evidence of high gross earnings elasticities, the rationalisation here is ‘simplicity’. However, a government seriously concerned with reducing complexity would begin with a revenue-neutral reform that combined a more progressive Personal Income Tax rate scale with the elimination of the LITO and the Medicare Levy, and would then move towards making FTBs universal. It is also worth noting that a less progressive income tax implies a shift in the tax burden towards women because they predominate at lower earnings levels. Therefore, unless it is the intention of government to hold the earnings of women below the zero-rated threshold, or slightly above it in poorly remunerated part-time work, a less progressive income tax can be expected to reduce participation incentives.

4.2 Family Tax Benefits and the Medicare Levy

Because FTB-A and the Medicare Levy exemption are withdrawn on joint income, tax rates on a partner's income depend on the income of the spouse. We therefore present results separately for the single-earner family and then show what happens when the second partner goes out to work. As rates also vary with demographic characteristics, we report results for the case of a family with two children under 13 years, one of whom is under 5 years.

4.2.1 Single-Earner Family

Table 9 shows the impact of the Medicare Levy27 and the FTB system for the single-earner household. At $34 571, the lower income limit of the Medicare Levy exemption for the two-child family, the MTR rises to 10 cents in the dollar. A MTR of 55.5 cents in the dollar applies across a middle income band of $42 600 to $60 000 followed by 51.5 cents in the dollar up to an income of $69 423, because of the withdrawal of FTB-A above the ‘base rate’ at a rate of 20 cents in the dollar.28 The progressive rate scale of the Personal Income Tax has now been replaced by one that exhibits an inverted U-shape up to an income of $80 000. At $98 112 the MTR rises to 71.5 cents in the dollar because of the withdrawal of the base rate at 30 cents in the dollar. At $150 001 there is a MTR spike, denoted by *, because of the withdrawal of FTB-B at that income.29 ATRs (calculated at upper income thresholds) are progressive. We have a negative income tax system for the single-earner, two-child family up to about $54 000.

Table 9.  Marginal and Average Tax Rates: Single-Earner Family
Taxable incomePIT+LITO+ML+FTBs
MTRATR
  1. Note: PIT, Personal Income Tax; LITO, Low Income Tax Offset; ML, Medicare Levy; FTB, Family Tax Benefit; MTR, marginal tax rate; ATR, average tax rate.

$0–$14 0000.00−0.926
$14 001–$30 0000.15−0.352
$30 001–$34 0000.19−0.288
$34 001–$34 5710.34−0.278
$34 572–$40 6720.44−0.170
$40 673–$42 5590.355−0.147
$42 600–$60 0000.5550.057
$60 001–$69 4230.5150.119
$69 424–$80 0000.3150.145
$80 001–$98 1120.4150.195
$98 113–$111 0820.7150.256
$111 083–$150 0000.4150.297
$150 001*0.322
$150 002–$180 0000.4150.337
$180 000+0.465

4.2.2 Two-Earner Family

Table 10 presents MTRs and ATRs on a second income for a primary income of $50 000. The highest rates apply across low to average female earnings. The MTR on the first dollar of the second income is 21.5 cents because of the withdrawal of FTB-A and the Medicare Levy exemption. The rate rises to 41.5 cents at $4526 with the withdrawal of FTB-B. It then rises to 56.5 cents in the dollar at $14 000, the zero-rated threshold of the Personal Income Tax plus LITO. Again, at an income of about $23 000, the second earner loses more than 40 per cent of her earnings. At $48 112 the base rate of FTB-A begins to be withdrawn and so her MTR rises to 65.5 cents in the dollar, and her ATR begins to rise again.

Table 10.  Primary Income of $50 000: Tax Rates on Second Income
Taxable incomeTax rates on second earnings
MTRATR
  1. Note: MTR, marginal tax rate; ATR, average tax rate.

$0–$45260.2150.215
$4527–$14 0000.4150.35
$14 001–$19 4230.5650.41
$19 424–$22 9950.3650.403
$22 996–$30 0000.1650.348
$30 001–$34 0000.2050.331
$34 001–$48 1120.3550.338
$48 113–$50 0000.6550.35

Figure 4 plots the ATR profiles faced by the second earner for four levels of primary income: $30 000, $40 000, $50 000 and $60 000 pa. The graph shows that, in all four cases, the second earner reaches high ATRs at relatively low income levels. For families with more than two dependent children, the results will be more extreme because the withdrawal rates will apply across wider bands of income.

Figure 4.

Average Tax Rates on Second Income, 2008–09

Taxes on second earners and their families at the levels indicated, together with a lack of access to affordable, high-quality child care, can be expected to have strong negative effects on female labour supply, not only during the child-rearing years, but throughout the entire life cycle, as indicated by the data on labour supplies reported for life-cycle phases in Subsection 3.1.

4.3 Graphical Exposition of the Australian Family Tax System

The effect of targeting family payments on joint income can be illustrated graphically. The following analysis shows how rates change when a household switches ‘type’, as defined in Section 2. For the purpose of simplification, we compare rates for the two types: the single-earner household (SE), in which the male partner works full time in the market and the female works full time at home; and the two-earner household (FT), in which both partners work full time.30 Partners in the FT household are assumed to earn the same income.

Figure 5a plots the marginal tax rates on partners’ incomes as a result of the Personal Income Tax and the LITO, as annual primary income rises. Figure 5b depicts the resulting average tax rates. Because the tax base is individual income, the primary earner in each household and the second earner in the FT household all face the same MTRs and ATRs at any given income level.

Figure 5.

Marginal and Average Tax Rates by Primary Income

Figures 5c and 5d show the effect of the FTB system and the Medicare Levy. The new MTR profiles exhibit two striking features. First, the MTR profile of the second earner (MTR2) is now strongly to the left of that of the single-earner household (MTR SE).31 This is a characteristic feature of joint taxation. Second, marginal rates in the ‘middle’ are much higher—there is a much stronger tendency towards an inverted U-shaped rate scale up to the lower income threshold for the withdrawal of the base rate of FTB-A The consequences of both are very high ATRs on the income of the second earner (ATR2) at relatively low primary and second income levels.32 Overall, the higher ATR on the second earnings raises the ATR of the FT household (ATR FT) to well above that of the SE household (ATR SE). At any given level of primary income, the two-earner family now no longer pays twice as much tax, but more than twice as much tax as the single-earner with the same primary income.

4.4 Impact on Working Families

We examine the impact of this type of tax system using the ABS SIH06 data sample of 1608 ‘in-work’ families described in Subsection 3.2. Results are reported for the three household types: single-earner (SE), two-earner with the second in part-time work (PT), and two-earner with both partners in full time work (FT). As indicated in Table 6, the three types tend to be fairly evenly distributed across quintiles.

Table 11 reports ATRs on the household income of each type and on the second income of the PT and FT households, by quintiles of primary income, for the 2008–09 financial year. The overall data means in the final column show that, on average, primary earners in the SE household pay $8150 in tax. The PT household's tax is close to double that figure, at $14 492, because, on average, the second earner pays $6600 on an income of only $22 911. The tax burden on the FT household is $20 613, with the second earner contributing $12 613. While all second earners face high ATRs, the highest rates appear in quintile 1, a result broadly consistent with the effect of targeting on joint income illustrated in Table 4b.

Table 11.  Tax Burdens by Primary Income, 2008–09 $pa
Primary income quintile12345All
  1. Note: SE, single-earner household; PT, two-earner household with the second in part-time work; FT, two-earner household with both partners in full time work; ATR, average tax rate; ATR2, average tax rate on the income of the second earner.

SE
 Net tax $pa–8 737–1 4153 1219 16933 7188 150
 ATR%–22.4–2.54.415.522.99.5
PT
 Second earnings $pa18 09021 54622 78122 82629 72722 911
 Tax on second earnings $pa5 9976 2676 1836 2488 4566 600
 ATR2%33.229.127.127.428.428.8
FT
 Second earnings $pa24 53231 96938 91445 91965 94141 342
 Tax on second earnings $pa7 5999 33210 68013 83120 52712 311
 ATR2%31.029.229.927.431.129.8

ATRs on the second income at the levels indicated mean that, on average, a married mother who decides to go out to work will lose about one-third of her income in taxes and reduced FTBs. She will also contribute more to GST revenue, because her additional income will be spent at least partly on GST-rated goods and services bought as substitutes for those she could produce herself by working full time at home. Moreover, if she decides not to go out to work, and if all mothers make the same decision, tax revenue from families could fall by almost 50 per cent. Over time, the revenue from all couples in the post-child phase can be expected to fall, given the evidence on persistence of the female labour decision made in the pre-school phase of the life cycle.

5. Conclusions

In this article, we have shown how Australia's progressive individual income tax has been transformed into a system with strong elements of joint taxation and a rate scale that is no longer progressive. The overall effect of the transformation has been to shift the tax burden from the top of the income distribution to the lower and middle income ranges, and in particular to working married women in these income ranges. An explanation for this direction of reform is an ideological position, prevalent since the 1980s, that exaggerates the mobility of top income earners and the effects on their incentives to earn income of tax rates that place on them a fairer share of the overall tax burden. The drive to reduce taxes at the top, and the need to deal with in-work poverty of families at the bottom of the income distribution, has resulted in a set of changes to the Australian tax system that cannot be justified on the grounds of either fairness of the distribution of tax burdens or the effects on work incentives and economic efficiency.

The central policy proposal of this article is the reintroduction of a progressive individual-based income tax combined with universal family payments, together with the elimination of unnecessary policy instruments, such as the LITO and the Medicare Levy, which serve only to reduce transparency of the true rate scale. The article sets out at some length the merits of a progressive individual income tax in a modern economy. While the majority of families now have two-earners, there is a high degree of heterogeneity in the labour supply of the second earner. It is therefore essential to evaluate the effects of tax reform within the framework of a model that recognises the multi-person household as a small economy engaged, to varying degrees, in intra-household production (predominantly of child care) and exchange. The relevance of the single-person household model, in which a representative individual makes a choice between work and leisure in a perfectly competitive labour market, and inter-temporal consumption and saving decisions in a perfect capital market, has long since passed.

Section 3 explained at some length the superiority of a progressive individual income tax over joint taxation, or indirect taxation. Neither an income tax nor a consumption tax can be applied to a tax base that includes household production. Households with the same wage rates and demographic characteristics will therefore pay different amounts of tax, depending on the second earner's choices between market and domestic work. Under these conditions, individual taxation is a less constrained policy instrument for redistribution. Progressive individual taxation, in addition to being consistent with the Ramsey rule for efficiency, has merits in terms of horizontal and vertical equity, because the tax a household pays will be negatively correlated with the dispersion in partners’ incomes, and therefore positively correlated with the allocation of time to home production. Joint taxation has the opposite effect. Indirect taxation offers no solution because individual consumptions within the household are not observed. The base for indirect taxation is inevitably limited to joint consumption or some essentially arbitrary assumption about unobserved consumption shares within the household. Australia therefore needs to move to a tax system centred more heavily on a well-designed progressive individual income tax, together with reforms that address the widespread problem of tax avoidance.

Endnotes

  • 1

    1983 saw the first step in the process, with the introduction of the Family Income Supplement withdrawn on joint income, which has since evolved into Family Tax Benefit Part A, excluding the base rate. A series of subsequent reforms have completely eliminated universality.

  • 2

    See Chapter 6 of Apps and Rees (2009) for a comparative analysis of the family income tax systems of these countries.

  • 3

    Note that the focus of this article is the family tax benefit system, which makes payments on the basis of observable demographic characteristics. The relevant second-best setting for other payments may differ.

  • 4

    The modern foundations of which were laid by James Mirrlees (1971) and Eytan Sheshinski (1972).

  • 5

    Note that the implicit lump sum in the second tax bracket is in fact higher than the previously universal benefit, a.

  • 6

    Put simply, whether a consumer is better or worse off depends on the exact equilibrium positions before and after the tax reform, which in turn depend on preferences. For more exact analysis see Apps and Rees (2009, Subsections 6.3 and 7.4).

  • 7

    There is also the misconception that a system of universal benefits creates higher administration costs for the tax system, as if the uniform payment is physically made to everyone and withdrawn portions then have to be physically returned to the tax authority. If the uniform payment and withdrawal rates are incorporated into the tax system, no such costs arise. Indeed, as Brewer, Saez and Shephard (2008) make clear, targeted benefit or means-tested systems can be far more costly and complex in administrative terms, and less effective in getting benefits to people who need them, than a simple uniform payment with a given structure of marginal tax rates that determine how it is withdrawn.

  • 8

    We discuss the point of this distinction below.

  • 9

    This review and the work that has contributed to it are likely to figure prominently in future discussions of tax policy.

  • 10

    However, if they really take this seriously, they should recommend a cut in the tax rate.

  • 11

    There are fewer records in this phase than in later phases. However, when young singles without children, who are essentially in the same life-cycle phase, are included, we obtain a similar result. Almost all men and all women not in higher education work full time prior to having children.

  • 12

    This is consistent with the results of panel data studies for the United States. See, for example, Shaw (1994).

  • 13

    In a perfect capital market, children would be able to borrow to pay for their consumption and investment in their human capital, and they would repay the debt during their working years. For further discussion, see Apps and Rees (2001, 2003).

  • 14

    This effect offers an explanation for the persistence of female labour supply decisions made in the preschool phase. An extensive literature on work-related human capital accumulation includes the contributions of Altug and Miller (1998) and, more recently, Imai and Keane (2004).

  • 15

    The analysis is based on data from the ABS 2006 Time Use Survey.

  • 16

    Private income is defined by the ABS as income from all non-government sources, such as wages and salaries, profits, investment incomes and superannuation.

  • 17

    Cremer, Lozachmeur and Pestieau also show, in a very general model, that equalisation of tax rates on primary and second earners is almost never optimal.

  • 18

    See in particular Brewer, Saez and Shephard (2008).

  • 19

    Strictly, this assumption is made only for the proof of Proposition 3 in the paper, but this, the ‘negative jointness’ result, is the result of main interest.

  • 20

    The paper presents simulations based on UK data, which claim to show that their conclusions are robust to this assumption, but for reasons given below we doubt the usefulness of these simulations.

  • 21

    Furthermore, it is hard to believe that all non-participation by men is purely a matter of choice.

  • 22

    Note that in the model there is no explicit marginal tax rate on the second earner's income.

  • 23

    Furthermore, the empirical evidence suggests that a wife's income has no statistically significant effect on the pre-tax earnings of the husband. In the Kleven, Kreiner and Saez model, because the husband's marginal tax rate in the two-earner household is lower, he is predicted to have a higher effort supply and level of earnings than in the single-earner household.

  • 24

    The Child Care Benefit is not included. This is unlikely to alter significantly the overall results, given that available unit record data on government direct and indirect benefits for child care indicate that they tend to be distributed independently of employment status.

  • 25

    The 2008–09 LITO is $1200 and is withdrawn at a rate of 4 cents in the dollar above a lower income threshold of $30 000.

  • 26

    For a detailed analysis, see Apps (2009).

  • 27

    The Medicare Levy is a flat rate tax of 1.5 per cent above a joint-income targeted exemption. When the family's income reaches the lower income threshold for the exemption, it is withdrawn at a rate of 10 cents in the dollar. For a family with two dependent children, the threshold is $34 571. There is also a surcharge at a higher income level for those without the required level of private health cover, which we omit here. The private health insurance tax rebate is also omitted.

  • 28

    FTB-A provides a cash transfer of $4631.83 per child under 13 years. The ‘maximum rate’ is withdrawn up to the ‘base rate’ on family income above $42 559 at a rate of 20 cents in the dollar. The base rate is $1945.45 and is withdrawn at a rate of 30 cents in the dollar on joint income above $98 112.

  • 29

    FTB-B is a cash transfer of $3693.80 for a child under 5 years and is withdrawn on the second earner's income above $4526.0; it is fully withdrawn at $22 995. Since 1 July 2008, it has been limited to families in which the primary earner has an adjusted taxable income of $150 000 or less.

  • 30

    For a comparison of tax rates on primary and second incomes that includes households with the second earner in part-time employment, see Apps and Rees (2009, ch. 6).

  • 31

    In a model in which male and female labour supplies are variable, both partners face the same MTR under joint taxation. This will not however apply in the Australian context because of FTB-B, which selectively raises the rate on the second earner.

  • 32

    The ATR on the income of the second earner is calculated as the additional household tax due to her labour supply decision.

Ancillary