How should evidence be used in the study of tax design? What is the appropriate balance between theory and empirics?
(pages 39-45 of printed journal)
The Mirrlees Review was an attempt to build a foundation for tax reform based on the large body of economic theory, empirical evidence and practical experience of tax policy built up over the last three decades. It was inspired by the Meade Report (1978) with the idea of reviewing tax design from first principles for modern open economies – particularly the UK. The Review was published in two volumes, Dimensions of Tax Design (Mirrlees et al, 2010), summarising expert evidence across a wide range of aspects of tax reform, and Tax by Design (Mirrlees et al, 2011), outlining the conclusions and recommendations.
One aim of the Review was to bridge the gap between academic research and policy practise – much in the spirit of Dick Downing, after whom this lecture is named. For consistency and coherence, this lecture focuses on the taxation of earnings, although the Review itself concerned all aspects of the tax system.
There were three broad principles adopted by the Review:
- System: Consider the tax system as a whole, defining the marginal tax rate as sum of all additional taxes paid when income increases by £1.
- Neutrality: Do not discriminate (unnecessarily) between similar activities.
- Progressivity: More tax from the better off.
The current structure of UK taxes and benefits neither reflects these principles nor works as a coherent system. Welfare benefits, personal taxes and corporate taxes are disjointed. Neutrality is not observed – savings taxes are inconsistent; the corporate tax system favours debt over equity; and there are different tax rates on carbon and failure to price congestion properly. Progressivity is not achieved efficiently – and work incentives are damaged unnecessarily.
The broad proposals in Tax by Design reflect these principles. A single integrated welfare benefit and the alignment of tax-rates across personal and corporate earnings are proposed, with all taxes and benefits being considered, including value-added, congestion and carbon taxes (not examined in this lecture). The Review authors aimed to achieve progressivity through the direct tax and benefit system, recognising constraints imposed by responses to incentives.
Tax by Design is loosely organised under five related headings:
(i) Key margins of adjustment – particularly establishing empirical facts about key aspects of behaviour on which taxes could impact.
(ii) Measurement of effective tax rates – considering the tax system as a whole, including the implicit tax rates in the benefit and tax-credit systems.
(iii) The importance of information and complexity – relating to the incentives implicit in the tax and benefit system acting on individuals, households and firms, including the stigmas and hassles endured by those accessing the system.
(iv) The size of responses - the core of any rigorous empirical analysis, concerning the robust measurement of the causal impact of tax reforms. The Review used a mix of (quasi-) experimental and structural approaches with the experimental approaches acting as a 'reality check' on the structural model.
(v) Implications from theory for tax design – linking empirical relationships and design mechanisms from economic theory to determine efficiency costs, overall optimality and improvements to tax design.
There are three main ingredients to any optimal tax analysis: accurately measuring response elasticities, establishing the distribution of income, and deciding social welfare weights. The first two are positive and emerge from evidence-based analysis. The last is normative and something about which reasonable people may differ. The aim was to draw broad conclusions using evidence on the first two empirical aspects while making fairly weak assumptions on social welfare weights – basically, that they are declining in some measure of equivalised income.
Earnings taxation, the focus here, is ideal for examining the role of evidence in tax design. There are substantial empirical results on labour-supply responses to tax reform for individuals and families. This research highlights the need to distinguish between the intensive and extensive margins of labour supply – respectively, between the decisions of whether to work or not and how much to work. Research has also shown clear differences in responses by age, gender and family composition. Additionally, tax-return information provides evidence on taxable income elasticities relevant to the design of earnings taxation. This evidence naturally supplements and extends work on the responses of employment and hours of work to tax reform.
That is not to say the taxation of earnings should stand separately from the design of the rest of the tax system. Indeed, the taxation of earnings will need to adjust for the level of redistribution and work incentives induced by other tax-reform elements. Nonetheless, we can derive overall directions for the taxation of earnings, while the precise tax-rate schedule will depend on the shape of the whole tax system.
Recent labour market history demonstrates three central trends which also point to the three key margins where responses to tax reform are most likely to occur: declining male employment, especially among older men; a strong rise in employment and total hours of work for women; and a decline in employment among those in their late teens and early twenties, reflecting increasing educational attainments.
Although both extensive and intensive margins of labour-supply matter in explaining the broad changes in total hours over the last three decades in the UK, France and the US, in recent work with Guy Laroque and Antoine Bozio I have shown that they matter in different ways for different age and demographic groups. This empirical evidence was highly influential in the Review.
For men, variations in the extensive margin of labour supply occur mainly at the start and finish of their working lives. Figure 1 provides a broad view of (pre-recession) employment rates by age for the UK, France and the US in 2007. The strikingly similar average rates for men aged 30-54 in these three economies suggest that differences in employment are concentrated at early and later points in working lives, indicating the importance of the extensive margin at these life-cycle points.
Figure 2 demonstrates that differences in hours, conditional on employment, matter for men across their working life. Although it is unlikely that tax and benefit systems alone explain all these differences, it would seem unwise to play down the intensive margin.
For women, Figures 3 and 4 show that hours and employment itself, both vary across working lives. As with men, average employment rates in 2007 were surprisingly close at ages between the late 20s and early 50s. Again it is at the early and later periods of working lives that extensive-margin choices become important.
Figure 1: Male employment in France, the UK and the US, 2007
Figure 2: Male total hours worked in France, the UK and the US, 2007
Figure 3: Female employment in France, the UK and the US, 2007
Figure 4: Female total hours worked in France, the UK and the US, 2007
For women with younger children it is not usually just an employment decision that is important, but also whether to work part- or full-time, influenced by specific elements of the tax and benefit system. In the UK there still remains a dip in hours of work around child-bearing ages. Indeed, underneath these broad figures, there are also important variations at the extensive margin for mothers with pre-school children and lower levels of education.
Optimising the balance between the extensive and intensive margins is crucial to understanding responses to reform. Overall, those with school-age children or aged 55-70 are most responsive at the extensive margin. To understand how taxes and benefits might affect labour supply choices, we need to measure the effective work incentives implicit in the tax and benefit system.
Perhaps the main (perceived) defects in current welfare/benefit systems are that participation tax rates at the bottom remain very high. In the UK for example, effective marginal tax rates are well over 80 per cent for some low-income working families, mainly due to interaction between taxes, means-tested benefits and tax credits. But high implicit tax rates at low incomes can be optimal for welfare functions that prioritise redistribution. The key to good tax policy design is to avoid high work-disincentives where responses to taxes are likely to be high while maintaining a desired level of redistribution.
Consider a typical budget constraint for a low earning family. A complete analysis of the effective tax rate will combine the implicit tax rates in the benefit, tax-credit and income tax systems. Figure 5 provides such a case study for a single mother in the UK, assuming all applicable benefits are accessed.
Figure 5: The interaction between taxes and benefits in the UK
The tax-credit system is an increasingly important part of the effective tax system facing low-income earning families in many countries. Taken together with income support and other benefits, low-income earners in the UK can face a complex rate schedule with relatively high effective tax rates. Indeed, families in receipt of other benefits gain less from the working tax credit (WTC) than otherwise equivalent families not receiving these benefits.
In this respect, there are two summary measures that are usefully documented in Tax by Design: the effective marginal tax rate (EMTR) – that is the proportion of a small increase in earnings taken in tax and withdrawn benefits; and the participation tax rates (PTR) – the incentive to be in paid work at all, defined by the proportion of total earnings taken in tax and withdrawn benefits. The overall distribution of these tax rates by income and family type in the UK is presented in Figures 6 and 7.
Importantly, it is the PTR that is relevant for the employment margin, and the EMTR for the effort margin. The EMTRs and the PTRs can be negative as well as positive, but they are typically positive and often high at lower incomes. The effective tax rates in these Figures indicate the relatively strong redistribution towards low-income families with children in the current UK tax system. Indeed, the more the tax system targets low incomes, the higher EMTRs on low earnings are likely to be.
Not surprisingly, tax schedules can easily possess the feature of high EMTRs at low earnings. It is simply the result of means-testing which is the flip-side of targeted redistribution. Whether it is optimal or not will depend on the responsiveness of labour supply to these implicit tax rates, on the distribution of income and on the desire to redistribute to low-income families of a particular composition.
Figure 6: Effective marginal tax rates in the UK
Figure 7: Participation tax rates in the UK
The EMTRs and PTRs in the last section are just (local) averages at each gross earnings level. In the UK and elsewhere, multiple benefits with an array of overlapping means tests can create EMTRs of over 90 per cent. This degree of complexity can leave some individuals unwilling and/or unable to access all the benefits and tax credits to which they are eligible.
A useful way to formalise some of the issues surrounding information and complexity in earnings taxation is to allow eligible individuals not to 'take up' certain benefits and tax credits. This reflects the idea that individuals may not understand the system, or may find the stigma or hassle costs of accessing benefits or tax credits too high to be worthwhile.
Typically, take-up is an increasing function of the eligible amount of benefit or tax credit but full take-up is rarely achieved. Rates of take-up can be quite low, especially for families eligible for minimal benefits. These observations provide some insight into the importance of modelling take-up decisions when engaging in tax and benefit reform.
Although for many workers, employment and hours margins are the key measures of their labour supply, for others it is the level of effort for any hour of work that they can use to respond to tax incentives. For yet others there will be exemptions and deductions which will allow them to change their taxable incomes with little change in their overall earnings. Accordingly, the next section uses the impact of taxation on taxable income to examine tax-rate reform for top earnings.
There are three dominant empirical approaches to the measurement of responses: the 'experimental' approach using randomised control trials, the 'quasi-experimental' approach using historic reforms and the 'structural' approach based on a formal optimisation model of individual and family choices. There are many comprehensive reviews of quasi-experimental approaches. There are also some influential control-trial experiments on labour supply.
It is difficult to envisage a full-fledged tax reform analysis that does not draw on a structural model. Policy simulation, and understanding the impact of particular rate structures, requires a model of decision-making and of the budget constraint. However, to fully specify the choice problem and the budget constraint inevitably requires assumptions based on weaker empirical foundations.
Deriving convincing response effects from structural models is the key to reliable tax-design analysis. At a minimum, quasi-experimental analyses of policy reforms should be used to validate structural models. Inevitably, the more comprehensive and robust the empirical evidence, the better we can address the tax design problem.
Overall, the evidence suggests a key role for labour-supply responses at the extensive and intensive margins. Both matter but differ by gender, age, education and family composition. There are larger extensive responses by families with school-age children and older workers in 'pre-retirement' years.
Once individuals are allowed to respond to changes in a tax schedule by deciding whether or not to work, as well as how hard to work, then the optimal tax schedule can change dramatically. Optimal marginal tax rates can be lower (perhaps even negative) for those with low earnings capacities. Hence, in contrast to a purely intensive model, acknowledging responses at the extensive margin of labour supply can imply that earnings subsidies or work-contingent credits (such as the earned income tax credit or the WTC) should be part of an optimal tax system. A key lesson from recent tax-design research is that a 'large' extensive elasticity can 'turn around' the impact of declining social weights, implying both higher transfers to low-wage workers than to the unemployed and a role for earned-income tax credits.
Finally, the past three decades have seen expanding earnings inequalities and a change in the nature of earnings risks. The redistributive element of the earnings tax and benefit system acts, in part, as insurance against earnings risks. As the nature of these risks changes, and as underlying inequalities grow, inequality and work incentives become harder to balance. Designing an efficient structure for the earnings tax and benefit system – such that it achieves desirable distributional objectives – becomes ever more salient.
We assume that the government seeks to maximise social welfare subject to revenue constraint. As an illustration consider the tax design for low-income single mothers, suppose we want to redistribute '£R' to this group. Our aim will be to recover an optimal tax/credit schedule in terms of earnings. There are two related approaches. The first is to use the Diamond-Saez approximation in terms of extensive and intensive elasticities at different earnings. Second, a 'complete' Mirrlees optimal tax computation requiring a complete specification of choices and constraints.
The first approach uses usefully intuitive expressions for the optimal tax-rate schedule in terms of the extensive and intensive elasticities, simple summary measures of the distribution of earnings and the social welfare weights. This approach has been spearheaded by Emmanual Saez at Berkeley and Raj Chetty at Harvard. The simple formulae though are only approximate and assume away income effects. Income effects are likely to be important for some individuals, especially parents with young children. However as a guide to setting earnings tax rates they are extremely informative.
The second approach is appropriately labelled the structural micro-econometric approach to tax design. Effectively a stochastic mechanism design problem, the optimal tax model is the labour supply model. Consequently all of the assumptions concerning behaviour are also required for this analysis. The distribution of earnings, fixed costs of work, childcare, demographic differences and unobserved heterogeneity (described in the previous section) all influence choice of tax-rate schedule
The objective is to find robust tax-rate schedules for fairly general social-welfare weights. Given the structural parameter estimates, we can solve optimal schedules. In recent work with Andrew Shephard at Princeton we found that marginal rates are broadly increasing in earnings for all groups. This research, used in the Review, also recommended a shift out of work-support towards families with younger children, pointing to an optimal tax schedule with tagging according to age of children. These findings imply pure tax credits but only at low earnings and for those with school-age children. The analysis also suggested placing hours bonuses at full-time hours. Typically, part-time hours rules were sub-optimal.
At the top of the income distribution we focus particularly on the tax base. The greater the opportunities for exemptions and deductions and associating remuneration with lower-tax jurisdictions, the more difficult it is to raise revenue from top earners. High tax rates on narrow tax bases generate incentives to shift businesses to corporate forms, increase charitable giving, and orientate compensation packages towards tax-preferred capital gains. Consequently, we require a more general elasticity measure that captures these other phenomena. The taxable income elasticity does just that.
The responsiveness of taxable income to the tax rate will be affected by the tax base. The broader the tax base, the larger the elasticity is expected to be. Given the need to capture all these margins, and with effort difficult to quantify, the behavioural effect will require a different kind of measurement from that used to gauge hours and employment responses.
Figure 8: The Pareto distribution and the taxable income distribution at the top
Consider an 'optimal' top tax rate and suppose the welfare weight on top bracket incomes is negligible. The optimal rate – the Laffer rate – will be the revenue-maximising rate. When e is the taxable income elasticity and a is the Pareto parameter, the revenue-maximising rate is given by:
This assumes that the top of the taxable-income distribution is well-approximated by a Pareto distribution. Figure 8 shows this to be the case for the UK, where a is estimated at 1.67.
To estimate e reliably is harder – a difference-in-difference methodology is typically used. Applying this approach to past changes in tax rates among the top 1 per cent in the UK, using the 2–5 per cent group as a control, suggests a preferred estimate of e of 0.46 with a standard error of 0.13.
Exploring various formulations of the differences-in-differences specification for the UK, the estimate of e remains in the 0.35–0.55 range with a central around 0.45, but is clearly quite fragile. An estimate of this magnitude would suggest the optimal top 1 per cent bracket rate of around 57 per cent, close to the current top rate.
This analysis suggests little room for any further raising of the top rate in the UK without changes to the tax base for earned income itself.
In developing its recommendations in Tax by Design, the authors of the Review drew on empirical evidence across a wide range of fields. Here, as an illustration, I have focussed on recommendations for the reform of earnings taxation. Overall, the evidence points to a key role for labour-supply responses at the extensive and intensive margins. Both matter but differ by gender, age, education and family composition. Labour-supply responses for families with children vary by age of the youngest child. We also found different responses for older workers in 'pre-retirement' years.
The implications for tax reform are quite radical. They point to lower marginal rates at the bottom of the earnings distribution. They also point to less-aggressive means testing for some key groups and tax credits better targeted to lower incomes and to families where labour supply is most responsive. We particularly stress reforms for lone parents, married parents, and older workers pre-retirement, and 'tagging' tax rates by age of (youngest) child for mothers/parents and also at pre-retirement ages.
In summary, the reform agenda for earnings taxation can be interpreted through a lifetime view of taxation, implying a 'lifecycle' rearrangement of tax incentives and welfare payments to match elasticities and early-years investments – effectively redistributing across the lifecycle, distinguishing by age of (youngest) child for mothers/parents and at pre-retirement ages. The simulation results reported in Tax by Design suggest significant employment and earnings increases from such a reform package.
For top incomes, tax reforms are better directed towards base-broadening to address anti-avoidance and revenue shifting. Accordingly, in Tax by Design we recommend an alignment of personal and corporate tax rates to equalise tax treatments of income derived from employment, self-employment and running small companie
Mirrlees, J, Adam, S, Besley, T, Blundell, R, Bond, S, Chote, R, Gammie, M, Johnson, P, Myles, G and Poterba, J (eds) 2010, Dimensions of Tax Design: The Mirrlees Review, Oxford University Press for Institute for Fiscal Studies, Oxford.
Mirrlees, J, Adam, S, Besley, T, Blundell, R, Bond, S, Chote, R, Gammie, M, Johnson, P, Myles, G and Poterba, J (eds) 2011, Tax by Design: The Mirrlees Review, Oxford University Press for Institute for Fiscal Studies, Oxford.
A condensed version of the 2011 Downing Lecture presented at the University of Melbourne on 22 July 2011.