Essays on Wealth Taxation, Aviodance and Evasion among the Rich

Abstract

This PhD thesis contributes to the ongoing debate about wealth taxation by providing empirical evidence on this matter. On the one hand, it studies behavioural responses to taxes that levy broad stocks of wealth - i.e. the inheritance tax and the annual net wealth tax -, paying special attention to tax avoidance and focusing on the upper part of the wealth distribution. On the other hand, it analyses tax evasion and wealth disclosed through voluntary disclosure programs.

More specifically, after a brief introduction, the second chapter of this Thesis studies the effect of the inheritance tax both on the apportionment of estates and on the reporting and assessment of inherited assets. It uses the universe of inheritance tax returns of Catalan tax residents from 2008 to 2015 and exploits two tax reforms that occurred during 2010 and 2011 in Catalonia. It first exploits the introduction of a deduction for heirs older than 74 to study the effect on estates’ apportionment and uses this age cut-off to instrument inheritance tax rates. Results indicate that spouses are more likely to inherit the entire estate when there is no need to minimize tax payments. Second, it exploits a natural experiment resulting from the quasi-repeal of the inheritance tax for bequests given to close relatives to study changes in reported inheritances. It implements a difference-in-differences strategy that compares estates mostly inherited by close heirs, and hence, affected by the quasi-repeal of the tax, to estates mostly inherited by distant heirs not affected by this reform. Results indicate that reported estates increased up to 40% due to this tax cut. This response is primarily explained by the rise in real estate assessments and, to a lesser extent, by the reporting of assets that otherwise would have been evaded, such as cash, antiques, jewellery, etc.

The third chapter presents a study which focuses on the net wealth tax. It studies how taxpayers reacted to the reintroduction of the Spanish Net Wealth Tax in 2011, focusing not only on wealth accumulation, but also on potential avoidance strategies related primarily to exemptions and the existence of a limit on tax liability. It uses a panel of tax return micro-data from the universe of Catalan wealth taxpayers between 2011 and 2015, which approximately accounts for the wealthiest 1% of income tax filers. As there are no data for the period when the wealth tax was not being imposed, it takes advantage of the unexpected reintroduction of the tax by the Catalan Government at the end of 2011. This serves as the control year. The variation in treatment exposure, measured through the average tax rates, is then used to identify the effects of the wealth tax. Results show that the taxpayers' response to the reintroduction of the wealth tax was significant and they reflect avoidance rather than real responses. Facing higher wealth taxes did encourage taxpayers to change their asset and income composition to take advantage of exemptions (mostly business-related) and the limit on tax liability. Overall, these avoidance responses are large in terms of revenues and increasing over time.

Leaving aside these avoidance strategies, the fourth chapter of this Thesis takes a step forward and studies wealth evasion among the rich. This chapter exploits a tax amnesty implemented by the Spanish government in 2012. Through belated wealth tax returns submitted after the voluntary period - by the end of the amnesty program -, it identifies taxpayers voluntarily disclosing hidden wealth and quantifies the levels of evasion. Then it estimates the probability of a taxpayer being evader given the values reported in wealth tax returns filed during the voluntary period. For this matter, it frames tax evasion detection as a binary classification problem and trains and evaluates multiple classifiers commonly used in supervised machine learning methods. The accuracy rates indicate that the relatively little information available from wealth tax returns already allows to distinguish evaders from (presumably) non-evaders. Nonetheless, the provision of additional taxpayers' information might help to achieve a better detectability. Finally, the fifth and last chapter concludes.