Estimating the revenue loss from VAT tax avoidance using fiscal audits

Estimation des montants manquants de versements de TVA: exploitation des données du contrôle fiscal

Simon Quantin and Cécile Welter-Médée
Additional contact information
Cécile Welter-Médée: INSEE - Institut national de la statistique et des études économiques (INSEE)

Working Papers from HAL

Abstract: Estimating the revenue loss from tax avoidance is an important issue, but a difficult one. Extrapolating it from tax audits to all taxable companies requires taking into account the process which led to the selection of the firms to be audited. However, this selection process is particularly complex and not formalized: it is the result of an audit program concerning all sectors of activity, and of in-depth expertise work carried out by the tax auditors based on many private informations. This results in a selection bias that is significant, which should prevent from any simplistic extrapolation. This article is focused on the estimation of the French Value Added Tax (VAT) gap on firms subject to this tax. This exercise is based on the fiscal audits conducted by the fiscal administration. We adopt a two-step methodology that relies on a partition of VAT-taxable firms based on the administration in charge of the control, in order to take into account the structural difference in their organization. The first step aims at estimating a weighting for each controlled firm and is inspired by reweighting methods to correct for the non-response bias. Groups of firms are formed based on a similar estimated probability of being controlled, then each firm in the same group is assigned the same weighting, in this case the inverse of the proportion of firms actually controlled in the group. In a second step and based on these weightings, we propose to extrapolate the amounts of revenue loss from VAT avoidance applying estimators by the ratio and/or the average, applied to sets of homogeneous firms, in the sens of the probability to fraud. Overall, the different estimates obtained from the controlled 2012 fiscal exercises do not seem to depend very much on the methodological variations adopted. They stand between 20 and 25 billions euros.

Keywords: Fiscal Audits Classification; Tax Gap; Estimating Machine Learning; Machine Learning; biais de sélection; contrôles fiscaux; Selection Bias; écart de TVA (search for similar items in EconPapers)
Date: 2022-07
Note: View the original document on HAL open archive server: https://insee.hal.science/hal-05374863v1
References: Add references at CitEc
Citations:

Downloads: (external link)
https://insee.hal.science/hal-05374863v1/document (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:hal:wpaper:hal-05374863

Access Statistics for this paper

More papers in Working Papers from HAL
Bibliographic data for series maintained by CCSD ().