ECONOMICS LABOUR TAXES AND WORK HOURS IN AUSTRALIA By Anton Hallam and Ernst Juerg Weber The University of Western Australia DISCUSSION PAPER 07.09 Labour Taxes and Work Hours in Australia Anton Hallam* Ernst Juerg Weber Business School University of Western Australia Crawley WA 6009 March 2007 * The authors thank Edward Prescott for his helpful advice on the availability of labour market data. The first author acknowledges the financial support of the C.A. Vargovic memorial fund and the UWA Business School. Email of corresponding author: [email protected] Abstract In the 1970s, work hours in Europe were similar to work hours in America, but today Europeans work less than Americans. Prescott (2004) attributes the decline in European work hours to an increase in the effective marginal tax rate on labour income. The Australian labour market experience confirms that the taxation of labour income is an important determinant of the decision to work. In Australia taxes and work hours did not change much in the long-run, but Australian work hours rebounded after a temporary increase in taxes in the 1980s. The resilience of Australian work hours suggests that a return to the tax rates of the 1970s would restore the European labour supply. 1 1. Introduction As recently as in the 1970s, work hours in Europe exceeded work hours in America, yet today Europeans spend less time at work than Americans. The reversal of European and American work hours in the last quarter of the 20th century provides an opportunity to study the effect of labour market institutions on national labour supplies. Institutional factors that are commonly thought to influence the decision to work include the tax system, unemployment benefits, the degree of unionisation and other labour market institutions. Table 1 illustrates the dramatic shift that occurred in weekly work hours in Europe and America. In the early 1970s, work hours per person aged 15 to 65 years averaged 24.4 hours in France and 24.6 hours in Germany, whereas Americans worked 23.5 hours. By the mid-1990s, average work hours had fallen to 17.5 hours in France and 19.3 hours in Germany, whereas American work hours had risen to 25.9 hours. This amounts to a decrease in labour supply of 28 percent in France and 22 percent in Germany, and an increase of 10 percent in America. Between the 1970s and the 1990s, weekly work hours fell in all G-7 countries, except in America and Canada. Prescott (2004) developed a labour market model that highlights the relationship between weekly work hours and taxes on labour income. He concludes that “virtually all of the large differences between the U.S. labour supply and those of Germany and France are due to the differences in the tax system.” Surprisingly, national differences in labour market institutions, including the system of unemployment benefits and the degree of unionisation, matter little. The distortionary effect of the high European taxes on labour income produces a substantial welfare loss. Prescott (2004) estimates 2 that if France were to reduce its taxes on labour income to the U.S. level, the welfare of French workers as measured by lifetime consumption equivalents would increase by 19 percent. Table 1. Work Hours per Week for the G-7 Countries Country 1970-74 1993-96 Canada 22.2 22.9 France 24.4 17.5 Germany 24.6 19.3 Italy 19.2 16.5 Japan 29.8 27.0 United Kingdom 25.9 22.8 United States 23.5 25.9 Note: Weekly work hours per person aged 15 to 65 years. Source: Prescott (2004). Davis and Henrekson (2005) conduct a cross country analysis of the effect of taxes on work effort. They concur with Prescott “that tax rate differences among rich countries explain much of the international variation in work activity outcomes.” Still, the findings of Prescott (2004) and Davis and Henrekson (2005) remain controversial. Olovsson (2004) shows that work hours in Sweden exceed work hours in France and Germany, although Sweden has the highest tax rate in the world. Nickell (2004) reckons that unemployment accounts for about half of the difference in labour supply between Europe and the United States. Nickell, Nunziata and Ochel (2005) consider the effect of the tax system, unemployment benefits, employment protection, the system of wage determination and barriers to labour mobility on unemployment. 3 Alesina, Glaeser and Sacerdote (2006) also reject Prescott’s claim that taxes explain most of the low work hours in Europe. They argue that a regression of work hours on taxes that omits the degree of unionisation overestimates the effect of taxes because tax rates are positively correlated with the degree of unionisation. Once the degree of unionisation and employment protection is included among the explanatory variables, the tax effect becomes insignificant. Rogerson (2006), however, finds no relationship between work hours and the degree of unionisation and employment protection. He also points out that country and time dummy variables account for most of the explanatory power in the regressions conducted by Alesina, Glaeser and Sacerdote (2006). Ljungqvist and Sargent (2007) hold that generous government transfers, which are conditional on recipients not working, account for low European work hours. Prescott (2007) rejects their model because, unlike his own, it does not provide “a quantitative general equilibrium analysis ….. that is restricted to be consistent with the national account statistics.” Yet, Prescott recognises that Ljungqvist and Sargent’s approach sheds light on the implications of labour indivisibility. Employed Europeans work similar hours per day and week as Americans. In Europe average weekly work hours of the working age population are low because people take longer vacations, more sick days and there are more public holidays. In addition, Europeans start to work later in life and they retire earlier than Americans. In this paper, the quantitative general equilibrium model of Prescott (2004) is used to analyse the relationship between taxes on labour income and work hours in Australia. Section II presents Prescott’s labour market model. Section III details how Australian national accounts data were transformed to fit the labour market model and 4 how the effective marginal tax rate on labour income was calculated. Section IV compares actual and predicted work hours in Australia with work hours in the G-7 countries. As in the United States, in Australia taxes on labour income and work hours did not change much in the long-run, but a short-lived increase in Australian taxes temporarily reduced work hours in the 1980s. Section V considers whether other factors besides labour taxes affected work hours. Section VI concludes with some remarks on taxation and labour supply in Australia. 2. Labour Market Model Prescott (2004) developed a labour market model that explains how a person allocates time between work and leisure. The distinction between work and leisure depends upon whether an activity is taxed or not. Market work is subject to taxation, whereas leisure comprises all tax-free non-market work, in particular home work and work in the shadow economy, together with ordinary leisure activities. The available time for work and leisure is 100 hours per week, the remaining time being used for sleep and other necessities of life. In period t, the preferences of the representative worker are: logc +αlog(100−h ) (1) t t c is consumption, h represents weekly work hours, and (100-h) measures leisure time. The parameter α is a weight that determines the subjective value of leisure. The worker allocates time between market work and leisure subject to the budget constraint: 5 (1+τ )c +(1+τ )x =(1−τ )wh +(1−τ )(r −δ)k +δk +T (2) c t x t h t t k t t t t x denotes gross investment, w the real wage, k the capital stock and δ the depreciation rate. The taxes are: τ consumption tax rate, τ investment tax rate, τ c x h marginal labour tax rate and τ capital income tax rate. The budget equation states that k the sum of wage income, capital income, depreciation allowances and government transfers T must equal expenses for consumption and gross investment, with all items t being adjusted by the pertinent tax rate. The government uses taxes to finance public services, which, except for military expenses, are assumed to be perfect substitutes for private consumption. Any excess of taxes over expenses for public services is returned to households as lump-sum transfer payments. Output y is produced with a Cobb-Douglas technology: t θ 1−θ y = Ak h (3) t t t t Assuming workers are paid the marginal product, the parameter θ is the capital share in income. A is a productivity parameter. Since the Cobb-Douglas production t function has constant returns to scale, the size of the productive unit does not matter; it may be a single worker or a firm. The effective marginal tax rate on labour income captures the combined effect of labour and consumption taxes on the work decision. Differentiating the budget equation with regard to consumption and labour income yields: 1−τ Δc = h Δ(wh ) (4) t t t 1+τ c 6 In this expression the fraction (1−τ )/(1+τ ) represents the effective marginal h c increase in labour income, unencumbered by labour and consumption taxes. Setting (1−τ)= (1−τ )/(1+τ ), the effective marginal tax rate τ is: h c τ +τ τ= h c (5) 1+τ c The first optimum condition requires that the marginal rate of substitution between consumption and leisure is equal to the after-tax real wage, using the effective marginal tax rate on labour income. α/(1−h ) t =(1−τ)w (6) t t 1/c t The marginal rate of substitution, which is shown on the left-hand side of the optimum condition, can be derived from equation 1. The second optimum condition states that the real wage must equal the marginal product of labour. w =(1−θ)y /h (7) t t t Substituting equation 7 into equation 6 yields the key equilibrium relationship between work hours, the effective marginal tax rate on labour income and the consumption-income ratio. 1−θ h = (8) t c α 1−θ+ t y 1−τ t t This equilibrium relationship contains two endogenous variables, work hours h and the consumption-income ratio c/y, whose values depend on the effective marginal tax rate τ and all other factors that determine the path of the economy in a dynamic 7 macroeconomic model. The equation splits the effect of a change in the tax rate into a present-time substitution effect and an intertemporal substitution effect. An increase in τ lowers the incentive to work because the after-tax wage falls, reducing the relative price of leisure in the first optimum condition. If the increase in τ is expected to be temporary, the c/y ratio rises because people work less and, maintaining con- sumption, save less as long as the higher taxes persist. Consequently, a temporary increase in τ reduces h because (1-τ) falls and the intertemporal substitution effect increases the c/y ratio. A permanent tax increase cannot be avoided by postponing work effort. As there is no intertemporal substitution effect, a permanent tax increase reduces h only by lowering (1-τ), without affecting the c/y ratio. Since extra tax revenues are returned to households through lump-sum transfer payments, there are no wealth effects in this analysis. Prescott (2004) uses equation 8 to measure the effect of taxes on work hours. 3. Australian Taxes Prescott (2004) calculates the effective marginal tax rate on labour income for the G-7 countries, using United Nations System of National Accounts data (SNA). The following calculations for Australia follow Prescott’s method as far as possible.1 Since SNA data are unavailable for Australia after 1998, OECD data have been used for the period from 2001 to 2003. 1 The Appendix to the electronic version of this paper includes spreadsheets with the Australian data and calculations. 8