SELECTION BIAS: Introduction

A standard method for evaluating social programs uses the outcomes of nonparticipants to estimate what participants would have experienced had they not participated. The difference between participant and nonparticipant outcomes is the estimated gross impact of a program reported in many evaluations read more.

The outcomes of nonparticipants may differ systematically from what the outcomes of participants would have been without the program, producing selection bias in estimated impacts. A variety of nonexperimental estimators adjust for this selection bias under different assumptions. Under certain conditions, randomized social experiments eliminate this bias.
Continue reading

TAX ARBITRAGE: Concluding remarks 3

Fourth, models that allow for tax arbitrage and asset trade have important implications for the econometric analysis of labor supply, and for applied work on income distribution. Studies that ignore the effect of tax arbitrage and asset trade on labor supply incentives will come up with biased estimates of important elasticities, and international comparisons of income inequality will exaggerate the redistributive achievements of high-tax countries like Sweden.
Continue reading

TAX ARBITRAGE: Concluding remarks 2

Also, all our models are single-period ones, and they do not account for the fact that there is an intertemporal dimension to many tax arbitrage strategies (think of the postponed taxation associated with pension plans). Developing models that explore the implications of such alternative mechanisms certainly seem like a worthwhile exercise read only.

We believe that research on how tax arbitrage interacts with labor supply decisions may shed light on a number of important questions concerning modem systems of income taxation. First, what makes people work, given the very high marginal tax rates that can be observed for some countries during some periods? The traditional answer has been that labor supply is rather inelastic. Our proposed answer is different. With the tax avoidance technologies that became increasingly available during the 1980s, those who care about incentives need not pay those high tax rates.
Continue reading

TAX ARBITRAGE: Concluding remarks

Income redistribution in the welfare state

A final observation is that our analysis casts strong doubt on the common use of tax return data in applied work on income distribution. According to several studies Sweden ranks, together with some other high-tax countries, as one of the most egalitarian countries in the western industrialized world (see e.g. Gottschalk and Smeeding (1997)). The Gini coefficient for reported disposable income is much smaller in Sweden than in the USA, a finding that often is attributed to the equalizing impact of progressive income taxes and transfers.

But for reasons discussed in section 5.1, the reliability of tax return data as an indicator of true economic income is likely to be a decreasing function of the tax rate. We would expect disposable income of the affluent, computed from the income tax returns, to underestimate actual disposable income. Because Swedish high-income earners are confronted with much higher tax rates than their U.S. counterparts, this bias is likely to be larger in the Swedish than in the U.S. data, thereby creating an impression of a much higher degree of equality in Sweden.
Continue reading

TAX ARBITRAGE: Implications for empirical research 2

Second, the standard procedure in the literature is to trace out individuals’ budget sets by changing the supply of hours, holding capital income constant.
However, in the presence of tax arbitrage it is inappropriate to treat (reported) capital income as an exogenous variable in the regression. In our models, individuals’ asset transactions are an integral part of the labor supply decision, and the position and slope of individuals’ budget sets depend on the simultaneous determination of hours and portfolio composition.
Continue reading

TAX ARBITRAGE: Implications for empirical research

The analysis of the previous sections suggests that there are few general results on the incidence and efficiency effects of tax arbitrage. The answers to questions such as “Who gains and who loses from introducing tax arbitrage?”, and “Does tax arbitrage lead to a Pareto improvement?”, depend on the exact specification of the model. However, more robust insights appear to emerge when we turn to the implications of tax arbitrage and asset trade for empirical research. All our models – irrespective of their precise assumptions – seem to suggest that empirical studies that fail to account for tax arbitrage may come up with estimates that are seriously biased. Let us see why.
Continue reading

TAX ARBITRAGE: Introducing limitations 5

Third, assume that the government chooses between a system where interest on housing is deductible from taxable income (as in (24)), and a system without deductions (i.e. the budget constraint reads c-wi- r(X – X) – T(w£)). Which system is preferable from a welfare point of view? Interest deductions stimulates the housing consumption of high-income earners, which makes the allocation of housing consumption more uneven across income groups than it would otherwise have been.

This could imply a welfare loss – but a loss that could be outweighed by an efficiency gain from lower marginal tax rates on the labor income of high-wage individuals. Thus, the common view that tax arbitrage is costly for society is not necessarily correct; depending on the curvature of the tax function, and on the form of the utility function, it could actually be welfare-enhancing. This was true in our simple model of costless paper transactions in section 3, but even in the present model, where deductions lead to a distorted consumption pattern, we can not rule out that arbitrage creates an efficiency gain. Whether this is the case in reality is of course an open question.
Continue reading