Income Underreporting and Tax Evasion in Italy: Estimates and Distributional Effects

• Author: Dino Rizzi,Michele Bernasconi,Andrea Albarea,Anna Marenzi
• DOI: http://doi.org/10.1111/roiw.12444
• Published date: 01 December 2020
• Date: 01 December 2020

© 2019 International Association for Research in Income and Wealth

904

INCOME UNDERREPORTING AND TAX EVASION IN ITALY:

ESTIMATES AND DISTRIBUTIONAL EFFECTS*

by AndreA AlbAreA, Michele bernAsconi, AnnA MArenzi* and dino rizzi

Department of Economics,Ca’ Foscari University of Venice

The paper estimates the extent of evasion of personal income tax (PIT) in Italy by integrating two

methods that the literature has previously applied separately. The consumption-based method intro-

duced by Pissarides and Weber (1989) is used to estimate misreporting of income in micro data col-

lected in the household IT-SILC survey. We adopt an econometric specification close in spirit to that

of Feldman and Slemrod (2007), which allows us to estimate income misreporting at different rates for

different income sources. The misreporting estimates are then used in the discrepancy method to correct

the incomes compared with administrative registered data. The comparison provides new estimates of

evasion of personal income tax by type of income, region and income class. The estimates are used to

improve microsimulation analyses of the distributional impact of tax evasion.

JEL Codes: C63, D31, H26

Keywords: household surveys, income misreporting, personal income tax, tax-benefit microsimulation,

tax evasion

1. introduction

Measuring tax evasion is often described as attempting to obtain “evidence on

the invisible” (Slemrod and Weber, 2012).1 Several approaches have been developed

to obtain evidence on tax evasion that depend on the purpose of the analysis and

on which effects of tax evasion one wants to measure.

Here we propose an approach that integrates two methods that the literature

has previously applied separately. Both methods adopt a microeconomic per-

spective. The analysis focuses on the personal income tax (PIT) in Italy (Irpef—

“imposta sui redditi delle persone fisiche”—and other local income taxes) and also

studies the distributional effects of this type of tax evasion.

Pissarides and Weber (1989) developed the first method, known as the con-

sumption-based approach. It uses micro-economic observations from consump-

tion-expenditure surveys to estimate the consumption function for certain classes

1Slemrod and Weber (2012) also provided a methodological review of various approaches, distin-

guishing between micro methods—or, as they are sometimes called, bottom-up or direct methods—and

approaches based on macro-economic aggregates, which are referred to as top-down or indirect meth-

ods (see also Schneider, 2005; Giovannini, 2011; Alm and Embaye, 2013; Schneider and Enste, 2013 for

a review of micro and macro studies in Italy).

Note: We would like to thank Roberto Casarin, Carlo Fiorio, Raffaello Seri, Francesca Zantomio

and, especially, an anonymous referee and the editor for several helpful suggestions.

*Correspondence to: Anna Marenzi, Department of Economics, Ca’ Foscari University of Venice,

Cannaregio, 873, Venezia 30121, Italy (anna.marenzi@unive.it).

Review of Income and Wealth

Series 66, Number 4, December 2020

DOI: 10.1111/roiw.12444

bs_bs_banner

Review of Income and Wealth, Series 66, Number 4, December 2020

905

© 2019 International Association for Research in Income and Wealth

of goods. After controlling for several household characteristics, the method uses

differences in consumption propensities estimated for various categories of income

earners to measure their tendency to misreport their incomes. In particular, the

method assumes that all categories report consumption expenditures accurately,

while incomes are reported correctly by only some categories (reference catego-

ries) of income earners. For example, Pissarides and Weber (1989) used employees

as the reference category, while the self-employed were estimated as substantially

underreporting. The method has since been applied to estimate misreporting rates

in various countries and for other income categories (studies include Besim and

Jenkins, 2005; Feldman and Slemrod, 2007; Hurst et al., 2014; Ekici and Besim,

2016; and several others quoted in Section 2). However, as far as we know, the

method has never been applied to Italy.

The consumption-based method can be used to estimate tax evasion further

assuming that people behave in the survey as they do in filing their official tax-

returns. This assumption can be criticized. A diverse hypothesis is behind a differ-

ent micro-economic method to measure evasion. The alternative method is based

on comparisons between the income distributions from the surveys and the income

distribution derived from data of official tax-return registers. Typically, these com-

parisons show that the distributions obtained from the surveys have higher incomes

than the distributions obtained from the registers, with the differences interpreted

as measures of evasion. For this reason, the procedure is also called the discrep-

ancy approach. Indeed, according to Feige (1990, p. 995), “the discrepancy

approach is feasible whenever independent means exist to estimate the same con-

ceptual entity. If one procedure for measuring a particular form of underground

activity is believed to be relatively free of biases induced by the activity, while

another is known to be affected by the activity, the discrepancy between the two

can be used to measure the net effect of the underground activity”.2 Clearly, the

method here assumes that people report their income truthfully in surveys, just as

they do with variables like consumption and expenditures, because they trust that

their data will not be disclosed to the tax authorities, eliminating the incentive to

lie.

The method is often combined with microsimulation analyses and the eva-

sion rates estimated by the discrepancy method are then employed to compare

the income distribution to counterfactual distributions simulated assuming full tax

compliance in order to determine the distributional impact of tax evasion. Analyses

carried out with this approach have been conducted to investigate tax evasion in

Italy (e.g. Marenzi, 1996; Cannari et al., 1997; Fiorio and D’Amuri, 2006; Baldini

et al., 2009) and other countries (e.g. Matsaganis et al., 2010; Figari et al., 2012).

Despite the intuition on which the discrepancy method is built, a large liter-

ature has identified various biases that affect people’s answers to surveys, which

in addition to the tendency to underreport income may include other forms of

measurement errors, e.g. due to inaccuracies, sampling errors, misclassifications

(Atkinson and Brandolini, 2001).

2The discrepancy method can be based on various measures, including in macro studies conducted

to obtain aggregate estimates of evasion via some gap that can be estimated (Alm, 2012). One of the

most famous application is for example in the so called currency method, based on the gap between

incomes and expenditures (Caridi and Passerini, 2001; Ahumada et al., 2007; Ardizzi et al., 2014).