Corporate Tax Evasion in a Globalized Enivronment

Teepe, Alexander

Corporate Tax Evasion in a Globalized Enivronment

BWL - Rechnungswesen, Bilanzierung, Steuern

Zusammenfassung

Inhaltsangabe:Abstract:
There are only few results of empirical research about corporate tax evasion and avoidance in comparison to individual tax evasion. One reason is a lack of exact data. Moreover, due to the nature of tax non-compliance it is not easy to calculate its volume. There is set of empirical questions on which we can make progress, such as the determinants of corporate evasion, for example the impact of penalties on corporate tax directors in large enterprises or the size of a company and the owner structure. It is material for policymakers to know which characteristic of enterprises let us assume that a firm evades more or less and which circumstances enhance corporate tax evasion.
We suggest that the developments in a more globalized world may change incentives for non-compliance of enterprises and change the options to evade taxes, in particular for import or export firms. The growing importance of transfer prices for tax directors of multinational enterprises may emphasize this assumption. But not only large multinational enterprises have options to evade taxes. Also smaller companies controlled by private owners can misstate import and export invoices in order to shift profits into low-tax countries or tax haven countries or build up capital in the informal economy.
We analyze the existing basic theory and provide a critical discussion to it. And we survey the existing empirical literature about corporate tax evasion. Furthermore, we present the results of a corporate tax survey in China and Germany. By this survey related to import and export firms of all size, evidence is provided that there are remarkable incentives to evade taxes by manipulation of commercial invoices.
We find that this type of tax evasion by importers and exporters is basically determined by the same effects as corporate tax compliance in general but there are in addition specific effects on compliance, such as effects by trading countries of the firms or by the supply chain. Finally, this paper presents various indications which imply the need of an international framework to analyze corporate tax compliance.

Inhaltsverzeichnis:Table of Contents:
Figures3
Tables4
ABSTRACT8
1.INTRODUCTION8
1.1DEFINITION CORPORATE TAX EVASION13
2.LITERATURE REVIEW OVER PREVIOUS STUDIES OF CORPORATE TAX NON-COMPLIANCE14
2.1THEORETICAL MODELS OF CORPORATE TAX EVASION14
2.1.1The TAG-model14
2.1.2Critical discussion of the TAG model18
2.1.3The TAG-model and […]

Leseprobe

Inhaltsverzeichnis

Alexander Teepe

Corporate Tax Evasion in a Globalized Enivronment

ISBN: 978-3-8366-0156-6

Druck Diplomica® GmbH, Hamburg, 2007

Zugl. Universität Mannheim, Mannheim, Deutschland, Diplomarbeit, 2006

Dieses Werk ist urheberrechtlich geschützt. Die dadurch begründeten Rechte,

insbesondere die der Übersetzung, des Nachdrucks, des Vortrags, der Entnahme von

Abbildungen und Tabellen, der Funksendung, der Mikroverfilmung oder der

Vervielfältigung auf anderen Wegen und der Speicherung in Datenverarbeitungsanlagen,

bleiben, auch bei nur auszugsweiser Verwertung, vorbehalten. Eine Vervielfältigung

dieses Werkes oder von Teilen dieses Werkes ist auch im Einzelfall nur in den Grenzen

der gesetzlichen Bestimmungen des Urheberrechtsgesetzes der Bundesrepublik

Deutschland in der jeweils geltenden Fassung zulässig. Sie ist grundsätzlich

vergütungspflichtig. Zuwiderhandlungen unterliegen den Strafbestimmungen des

Urheberrechtes.

Die Wiedergabe von Gebrauchsnamen, Handelsnamen, Warenbezeichnungen usw. in

diesem Werk berechtigt auch ohne besondere Kennzeichnung nicht zu der Annahme,

dass solche Namen im Sinne der Warenzeichen- und Markenschutz-Gesetzgebung als frei

zu betrachten wären und daher von jedermann benutzt werden dürften.

Die Informationen in diesem Werk wurden mit Sorgfalt erarbeitet. Dennoch können

Fehler nicht vollständig ausgeschlossen werden, und die Diplomarbeiten Agentur, die

Autoren oder Übersetzer übernehmen keine juristische Verantwortung oder irgendeine

Haftung für evtl. verbliebene fehlerhafte Angaben und deren Folgen.

http://www.diplom.de, Hamburg 2007

Printed in Germany

Table of Contents

Figures...3

Tables...4

ABSTRACT...8

1. INTRODUCTION ...8

1.1.

EFINITION CORPORATE TAX EVASION

...13

2. LITERATURE REVIEW OVER PREVIOUS STUDIES OF CORPORATE TAX NON-COMPLIANCE

...14

2.1.

HEORETICAL MODELS OF CORPORATE TAX EVASION

...14

2.1.1. The TAG-model...14

2.1.2. Critical discussion of the TAG model ...18

2.1.3. The TAG-model and endogenous probability of detection...19

2.1.4. Strategic models...20

2.1.5. Modeling the firm ...30

2.1.6. Critical Discussion of the "simple firm model" and prospects to modification...32

2.1.7. Sensibility of tax evasion models for changes of assumptions ...42

2.2.

REVIOUS EMPIRICAL STUDIES OF CORPORATE TAX NON

COMPLIANCE

ETERMINANTS OF COMPLIANCE

...44

2.2.1. Empirical Example: "What causes Firms to hide Output? The determinants of

Informality" (Dabla-Norris et al.) ...47

2.2.2. Empirical Example: "Keeping it Off the Books: An empirical Investigation into the

Characteristics of Firms that Engage in Tax Non-compliance" (Tedds) ...59

2.2.3. Empirical Example: "Tax Holidays and Tax Non-compliance: An Empirical Study of

Corporate Tax Audits in China´s Developing Economy" (Chan and Mo)...65

3. EMPIRICAL STUDY THROUGH A TAX SURVEY: TAX-COMPLIANCE OF IMPORT AND

EXPORT FIRMS IN CHINA AND GERMANY...71

3.1.

ATA COLLECTION

...75

3.2.

MPIRICAL

NALYSIS

HINA

...78

3.2.1. Dependent Variables ...78

3.2.2. Explanatory Variables...81

3.2.3 Hypotheses development for `underreporting of sales'...83

3.2.4 Hypotheses development for understatement and overstatement of invoices ...87

3.2.5. Descriptive Statistics ...89

3.2.6. Empirical Framework...95

3.2.7 Regression Results for `underreporting of sales' ...100

3.2.8 Regression Results for understatement and overstatement of invoices...105

3.3.

MPIRICAL

NALYSIS

ERMANY

...118

3.3.1 Dependent Variables ...120

3.3.2 Explanatory Variables...122

3.3.3 Hypotheses development for `underreporting of sales'...124

3.3.4 Hypotheses development for understatement and overstatement of invoices ...125

3.3.5 Descriptive Statistics ...133

3.3.6 Empirical Framework ...136

3.2.7 Regression Results for `underreporting of sales' ...137

3.2.8 Regression Results for understatement and overstatement of invoices...139

4. SUMMARY OF THEORETICAL AND EMPIRICAL FINDINGS BY THE LITERATURE REVIEW

AND OWN ANALYSIS WITHIN THIS PAPER ...146

5. CONCLUSION ...150

6. REFERENCES...228

Appendix 1: Survey China...233

Appendix 2: Survey Germany...239

Appendix 3: Example of used databases for potential participants German Survey...243

Appendix 4: Example of used database for potential participants German Survey

"Bundessteuerberaterkammer (BStBK)" ...244

Appendix 5: numerical example for optimal concealing rate according to optimality

condition in section 2.1.6...245

Erklärung...246

Figures

Figure 1:

tax moral in some OECD countries...10

Figure 2:

Potential indifference curve and budget set taxpayer in the ,,TAG-model"

...16

Figure 3: possible actions by taxpayers and IRS...22

Figure 4:

best response functions taxpayer and tax authority in an interactive model...26

Figure 5:

best response functions taxpayer and tax authority in an interactive model (2)

...27

Figure 6:

descriptive statistics for firm size, financial statements to shareholders and hiding output; data

source: WBES ...55

Figure 7:

correlation matrix for firm size, financial statements to shareholders and hiding output; data

source: WBES ...56

Figure 8:

ordered logit regression, dependent variable: percentage of hidden output,

data source: WBES...56

Figure 9:

correlation matrix for firm size, financial statements to shareholders and hiding output; data

source: WBES ...57

Figure 10: summary determinants of corporate tax compliance; source: own drawing...58

Figure 11: extract WTO international trade statistics 2005...72

Figure 12: extract 2, WTO international trade statistics 2005...74

Figure 13: Summary Dependent Variables Chinese Survey Version...81

Figure 14: Summary Explanatory Variables Chinese Survey Version...82

Figure 15: Firm size Chinese Sample...89

Figure 16: Industries Chinese Sample...89

Figure 17: means of understatement and overstatement variables for all Chinese

respondents...91

Figure 18: means of understatement and overstatement variables for respondents which are either import

or export oriented...92

Figure 19: means of understatement and overstatement variables of import invoices - respondents which

are import oriented among other orientation ...92

Figure 20: means of understatement and overstatement variables of export invoices - respondents which

are export oriented among other orientation ...93

Figure 21: means of understatement - all invoice misstatement types 

all

respondents...93

Figure 22: means of understatement - all invoice misstatement types 

only exporting or importing respondents...94

Figure 23: correlation matrix for sum_ou_sta and related variables all respondents...110

Figure 24: correlation matrix for sum_ou_sta and related variables only exporters and importers...110

Figure 25: mean for underreporting of sales and for profit shifting, each associated with the development

of capital controls...111

Figure 26: mean for fraction of respondents answering profit shifting happens, depending on the obstacle

`my competitors avoid sale taxes'...112

Figure 27: Summary Dependent Variables German Survey Version...122

Figure 28: Summary Explanatory Variables German Survey Version...123

Figure 29: Typical Asia loop: source: www.hapag-lloyd.com...131

Figure 30: Typical Asia loop: source: www.hapag-lloyd.com...131

Figure 31: means of understatement and overstatement variables for all German

respondents...135

Figure 32: means of understatement and overstatement variables for German respondents, importing and

exporting respondents only...135

Figure 33: descriptive statistics understatement and overstatement

variables...140

Figure 34: percentage of respondents estimating firms redirect shipments to evade taxes, depending on

Switzerland is a trading country ...141

Figure 35: percentage of respondents estimating firms misstate invoices to achieve capital is wired to third

party tax haven country (`wiretype'), depending on China is a trading country...143

Figure 36: summary of important effects on corporate tax noncompliance...147

Figure 37: summary of important effects on commercial invoice misstatements... 149

Tables

Table 1:

Determinants of Informality: Basic Specification in Dabla-Norris et al. (2005) ...158

Table 2:

Determinants of Informality: Firm Size and Obstacles in Dabla-Norris et al. (2005) ...159

Table 3:

Determinants of Informality: Institutions in Dabla-Norris et al. (2005) ...160

Table 4:

Definition of the Dependent Variable for the Interval Regression Model in Tedds (2005) ...161

Table 5:

Data Summary for independent Variables (source WBES -data) in Tedds (2005) ...162

Table 6:

Estimation Results in Tedds (2005) ...163

Table 7-10: Tax Holidays in China, Chan and Mo (2000) ...169

Table 10a: Descriptive Statistics China (including all firms: import, export and domestic) ...172

Table 10b: Descriptive Statistics China (including only firms which import or export) ...174

Table 11: Summarized Results for Model 1-3, for ordered logit regression - dependent variable is `not

reported sales' ... 176

Table 12: Model 1 alone - ordered logit regression - dependent variable is `not reported sales'... 177

Table 13:

Model 2 alone - ordered logit regression - dependent variable is `not reported sales'...178

Table 14:

Model 3 alone - ordered logit regression - dependent variable is ``not reported sales'...179

Table 15:

Model 1 - logit regression - dependent variable is `ou_sta' (=any kind of over or under

statement of commercial invoices), respondents: all ...180

Table 16:

Model 2 - logit regression - dependent variable is `ou_sta', respondents: all...180

Table 17: Model 3 - logit regression - dependent variable is `ou_sta', respondents: all...181

Table 18:

Model 4 - logit regression - dependent variable is `ou_sta', respondents: all...281

Table 19: Model 5 - logit regression - dependent variable is `ou_sta', respondents: all...182

Table 20:

Model 6 - logit regressions - dependent variable is `ou_sta', respondents: all...182

Table 21: Model 7 - logit regression - dependent variable is `ou_sta',

respondents: only importers and exporters ...183

Table 22:

Model 8 - logit regression - dependent variable is `ou_sta',

respondents: only importers and exporters...183

Table 23:

logit regression - dependent variable is `wire_type' (=commercial invoice over or

understatement to wire capital into tax haven countries), respondents: all...184

Table 24:

logit regression - dependent variable is `wire_type' (=commercial invoice over or

understatement to wire capital into tax haven countries), respondents: only importers and

exporters...184

Table 25:

model 1 - logit regression - dependent variable is `profout_ty' (any commercial invoice over or

understatement to underreport profit and shift profits in low-tax countries),

respondents:

all...185

Table 26: model 2 - logit regression - dependent variable is `profout_ty' (=any commercial invoice over or

understatement to underreport profit and shift profits in low-tax countries), respondents:

all...185

Table 27:

model 3 - logit regression - dependent variable is `profout_ty', respondents: all...186

Table 28:

model 4 logit regression - dependent variable is `profout_ty', respondents: only exporting and

importing firms...186

Table 29:

model 5 logit regression - dependent variable is `profout_ty', respondents: only exporting and

importing firms...187

Table 30:

model 1 - logit regression - dependent variable is `othprofd_ty' (=commercial invoice over or

understatement to help trade partners in other countries to understate their profits and to shift

their profits in low-tax countries), respondents: all...187

Table 31:

model 2 - logit regression - dependent variable is `othprofd_ty',

respondents:

all...188

Table 32:

model 3 - logit regression - dependent variable is `othprofd_ty', respondents: only exporting

and importing firms...188

Table 33:

correlation matrix China - respondents: all...189

Table 34a: Descriptive Statistics Germany (including all orientation: import, export and domestic) ...193

Table 34b: Descriptive Statistics Germany (including only respondents which are import- or export

oriented)...201

Table 35:

Summarized Results for Model 1-3 - logit regression - dependent variable is `percsa_1' ...203

Table 35a: Model 1 alone - logit regression - dependent variable is `percsa_1' ...204

Table 35b: Model 2 alone - logit regression - dependent variable is `percsa_1' ...204

Table 35c: Model 3 alone - logit regression - dependent variable is `percsa_1' ...205

Table 36:

logit regression model 1 - dependent variable is `ou_sta' (=any kind of over or under

statement of commercial invoices), respondents: all...206

Table 37:

logit regression model 2 - dependent variable is `ou_sta', respondents: all...206

Table 38:

logit regression model 2 - dependent variable is `ou_sta',

Respondents: all, but Swiss respondents excluded...207

Table 39:

logit regression model 2 - dependent variable is `ou_sta', respondents: only respondents which

are manager or firm owner (consultants excluded) ...207

Table 40:

logit regression model 2 - dependent variable is `ou_sta', respondents: only import and export

oriented respondents...208

Table 41:

logit regression model 1 - dependent variable is `routtype' (=over or under statement of

commercial invoices to shift profits to other countries through which the shipments are

redirected), respondents: all...208

Table 42:

logit regression - model 2 - dependent variable is `routtype', respondents: all...209

Table 43:

logit regression - model 2 - dependent variable is `routtype', respondents: all, but Swiss

respondents excluded...209

Table 44:

logit regression - model 2 - dependent variable is `routtype', respondents: only respondents who

are managers or firm owners (consultants excluded) ...210

Table 45:

logit regression - model 2 - dependent variable is `routtype', respondents: only import and

export oriented respondents...210

Table 46:

correlation matrix Germany all type of invoice misstatement and country variables

respondents:

all...211

Table 47:

correlation matrix Germany all type of invoice misstatement and country variables

respondents: only respondents who are managers or firm owners (consultants excluded) ...212

Table 48:

correlation matrix Germany all type of invoice misstatement and country variables

Respondents: only import and export oriented respondents...213

Table 49:

Size of the Shadow Economy: Europe and the USA versus Asia...214

Table 50:

China Top Trading Partners in 2005...215

Table 51:

Germany - Top Trading Partners in 2005...215

Table 52:

Word Corporate Tax Rates 2004...216

Table 53:

correlation matrix Germany all variables all respondents...218

Abstract

There are only few results of empirical research about corporate tax evasion and avoidance in

comparison to individual tax evasion. One reason is a lack of exact data. Moreover, due to the

nature of tax non-compliance it is not easy to calculate its volume. There is set of empirical

questions on which we can make progress, such as the determinants of corporate evasion, for

example the impact of penalties on corporate tax directors in large enterprises or the size of a

company and the owner structure. It is material for policymakers to know which characteristic

of enterprises let us assume that a firm evades more or less and which circumstances enhance

corporate tax evasion.

We suggest that the developments in a more globalized world may change incentives for non-

compliance of enterprises and change the options to evade taxes, in particular for import or

export firms. The growing importance of transfer prices for tax directors of multinational

enterprises may emphasize this assumption. But not only large multinational enterprises have

options to evade taxes. Also smaller companies controlled by private owners can misstate

import and export invoices in order to shift profits into low-tax countries or tax haven countries

or build up capital in the informal economy.

We analyze the existing basic theory and provide a critical discussion to it. And we survey the

existing empirical literature about corporate tax evasion. Furthermore, we present the results of

a corporate tax survey in China and Germany. By this survey related to import and export firms

of all size, evidence is provided that there are remarkable incentives to evade taxes by

manipulation of commercial invoices. We find that this type of tax evasion by importers and

exporters is basically determined by the same effects as corporate tax compliance in general but

there are in addition specific effects on compliance, such as effects by trading countries of the

firms or by the supply chain. Finally, this paper presents various indications which imply the

need of an international framework to analyze corporate tax compliance.

1. Introduction

"A crime is an act committed or omitted in violation of a law forbidding or commanding it and

for which punishment is imposed upon conviction."

Tax evasion is a violation of law and it is

punished if detected, in many countries and depending on the amount of evasion even punished

by imprisonment. Thus, tax evasion should certainly fall into the subset of crimes. But a crime

in its more limited sense is confined to felony and the effort to evade taxes is not in all

circumstances a felony. The definition of tax evasion with respect to crime theory may be

URL: http://www.thefreedictionary.com/crime

controversial; all the more if the importance of it is played down if it is denoted by words like

tax avoidance or tax optimisation. Indeed, such grading makes sense but the classification

becomes even more difficult when ethical arguments should be included. We may leave the

question of classification to others but nevertheless we will find that the large effort by some

individuals and corporations to minimize tax liability causes remarkable efficiency loss in the

economy. For that reason tax evasion is a field of interest which is a highly relevant part of

economics.

In detail, tax evasion affects the efficiency of the tax system by creating misallocations in

resource use, as agents alter their behavior, and evasion incurs costs, namely "noncompliance

costs" to cheat on taxes. In addition, tax-noncompliance has equity effects since it changes the

distribution of income in unpredictable ways. Moreover, tax evasion causes the governments

of countries to expand their resources to reduce its magnitude. And in case of individuals or

firms evade taxes they also affect the tax rates of those who are compliant taxpayers and

increase their tax rates. So finally by recognizing the existence of tax evasion, understanding

the real impact of taxation becomes more complicated.

When enterprises decide to underreport profit, in general this will, like for individuals,

generate distortions because of the effort needed to avoid detection. In particular, referring for

example to the shadow economy, it may affect the productivity level badly: This is suggested

for example by Johnson (2000) who states this happens when enterprises bear restrictions to

make use of supporting institutions as for example courts, due to operating underground (or

partly underground) while sales or output respectively is not reported.

The first theoretical and empirical examinations are originated in the United States and

detailed empirical results are available. Thus, as first impression, we refer to the volume of

the U.S. tax gap. It is described and estimated comprehensively by the International Revenue

Service (IRS) in the U.S. The International Revenue Service estimated on February 2006 the

updated overall gross tax gap for Tax Year 2001 at $345 billion. (The IRS has made tax gap

estimates for tax year 2001, but not later)

. The overall tax gap is defined by the IRS as the

difference between what taxpayers should have paid and what they actually have paid on a

timely basis. The overall tax gap has three components: Underreporting, nonfiling and

underpayment. For the tax gap due to underreporting more numbers are available and it is

estimated at $148.8 billion for individual underreporting and at $29.9 billion for corporations

(in 2001). The resulting underreporting rate (calculated as underreported tax divided by

This number is drawn from the article ,,IRS Updates Tax Gap Estimates" from the homepage

of the International Revenue Service

receipts plus underreported tax) is 17.4% for individuals and 13.8% for corporations

respectively.

Regarding the compliance behavior in general we can read in the 2004 articles

of the IRS that in the US "[...] one in five taxpayers now believes in the US that it's

acceptable to cheat on their taxes. Their ranks are growing, especially among younger

taxpayers. [...]"

However, the analysis of this paper mainly refers to Germany and China. For Germany the

overall tax gap (i.e. for individuals and corporations) is estimated in the year 2002 at 75 billion

(not including tax avoidance and not prosecuted cases)

while 450 billion are voluntarily paid.

In the official reports of the "Bundesministerium der Finanzen" no differences were showed

between overall tax gap and tax gap due to underreporting, so the comparison to the

underreporting rate in the US is not possible. Referring to the "world value survey of 2002"

(also published in the monthly report of the "Bundesministerium für Finanzen" (BMF), refer to

figure 1) we find differences in the tax moral, which is in Germany better than the OECD

average and the same survey reports that 57% of the German respondents answered that tax

evasion would not be acceptable for them at all.

Figure 1 - tax moral in some OECD countries

these numbers are drawn from Slemrod (2004)

refer to IRS Oversight Board Annual Report of 2004

refer to: ,,Bundesministerium der Finanzen, Monatsbericht 12.2002",

URL: www.bundesfinanzministerium.de

source : monthly report march 2005 ,,Bundesministerium für Finanzen", Germany

In addition, the monthly report of the BMF provides some insight about determinants of tax

moral. For example the relationship between increasing central organization of a state and tax

moral is negative, an increasing total burden of taxes and social contributions for the citizens

has a negative impact on the tax moral. If citizens expect that a large proportion of others evade

taxes their tax morale is lower.

In a more recent source about the problem of the corporate

German tax gap, we find the German ministry of finance reports German enterprises dislocated

65 billions of profits to abroad in 2005. We will get back to this fact and the source

interpretation in detail in section 3.3. For China, which is the second country of particular

interest beside Germany in this paper, there are fewer sources available about the compliance

gap. Referring to Dong (1998) alone foreign invested enterprises were found during audits to

evade 506 million of taxes in 1995.

These facts show the general importance of offering an economic perspective on the issue of tax

reporting behavior. Despite absolute figures about the evasion gap are challenging we will in

the following sections mainly care about the characteristics of the corporate evaders and the

determinants on compliance and we will analyze the problem theoretically and empirically.

Thereby we will describe why it may be suggested that the existing corporate tax evasion theory

is not continued completely in the existing empirical approaches, we will furthermore compare

theoretical results with the findings of some appropriate empirical papers. A further purpose of

this paper is to focus upon the compliance behavior of firms in connection with their

international relationships and international trade. We use Chinese and German survey data for

this approach and will compare the findings between the two countries and we will analyze,

whether certain evasion technologies of importers and exporters are triggered by the same

determinants as corporate profit underreporting in general.

The tax survey results, examined for this paper, reveal that the Chinese survey respondents

estimate in their field of business an average of 12.5% of sales are kept off the books.

Furthermore 62.9% of all Chinese respondents estimate that in their field of business

falsifications of commercial invoices of any type and amount are applied by importers and

exporters in international trade.

The German survey data suggests that 5% of total corporate sales are not reported and 50 % of

the respondents estimate that German importers and exporters apply any misstatements of

commercial invoices to reduce taxable profits. The great part of the German and Chinese tax

survey respondents are importers or exporters themselves.

,,BMF" monthly report of march, 2005

The rest of this paper proceeds as follows: A brief definition of what corporate tax evasion

denotes is described in Section 1.1. A review of the most relevant tax noncompliance theory for

both individuals and corporations, but with a focus on the theory for corporations is presented in

section 2.1. It starts in section 2.1.1. with the `Taxpayer as a Gambler model' which is the

seminal theory by Allingham and Sandmo (1972), then we present a critical discussion of this

model following Cowell (2002) in section 2.1.2. and afterwards derive the extension of this

simple model to the assumption of endogenous detection probability in 2.1.3 which goes back

to Yitzhaki (1974) and we follow Slemrod and Yitzhaki (2000) in the presentation. In section

2.1.4. we present the interactive model by Graetz, Reinganum and Wilde (1986) which is a

crucial extension as it includes the law enforcement agency into the compliance problem of the

individual taxpayer. In section 2.1.5. we present the basic firm model close to Cowell (2002),

however, the key reference for this basic approach is by Virmani (1989). Afterwards in section

2.1.6. we first discuss the basic firm model by presenting further insight of Cowell (2002) and

second discuss in detail the assumption and the difference of endogenous detection probability

on the one hand and endogenous audit probability on the other hand. Then we extend the basic

firms model to endogenous detection probability and fixed unit cost of concealment, as we

believe this reflects the decision problem of the firm most realistically. Finally in 2.1.7 we

survey shortly various prominent models of corporate tax compliance which address more

special problems than the basic approach does, as for example the monopoly market structure.

Thereby we summarize the popular assumptions which are crucial for a compliance model of

the firm and which affect the outcome and we explain the separability problem of output

decision and evasion decision.

In section 2.2. we first introduce the potential problems of analyzing tax evasion empirically

and subsequently three selected empirical papers are described in detail. That is (1) the

empirical testing by WBES-data

of a sophisticated model which elaborates the decision of a

firm to go informal, by Dabla-Norris and Gradstein (2005), (2) the cross country analysis of

Tedds (2005) which presents many determinants on corporate tax evasion and evidence by

WBES data and (3) the empirical example of Corporate tax compliance of Foreign Invested

Enterprises in China during Tax Holiday periods by Chan and Mo (2000). We will compare

these papers and we add for each a critical discussion and analyze to which extent existing

compliance theory is continued in the empirical approach.

The WBES dataset will be explained in detail in section 2.2.1.

In section 3 we work out an empirical analysis of the tax compliance of importers and exporters

in China and Germany by using own survey data. The data source is demonstrated in section

3.1.1. In section 3.2.1 to 3.2.8. the application of several dependent variables and explanatory

variables, which should describe tax evasion technology of firms in international trade, is

introduced and the statistical procedure and the statistical results are presented which result by

the Chinese data. In section 3.3.1. to 3.2.8. basically the same modus operandi is referred to the

German survey data, however, with different results and insight since we found the German

firms have other incentives due to another environment which needs slightly different variables

to describe the problem.

Section 4 provides a summary of all kind of material effects on corporate compliance found

within this paper. Section 5 concludes.

1.1. Definition corporate tax evasion

The amount of corporate tax evasion can be defined as the corporate tax that is legally owed

but not reported and by that not paid. As a result there exists a tax gap as difference between

legally owed tax and actually declared and paid tax. The reason for that are different kinds of

tax evasion technologies potentially applied by firms. The IRS for example explains the

following, for tax evasion in general (individual and corporate), as main components of the

tax gap: Nonfiling, underreporting and underpayment. Referring to the IRS website we can

find as explanation: ,,Nonfiling occurs when taxpayers who are required to file a return do

not do so on time. Underreporting of tax occurs when taxpayers either understate their

income or overstate their deductions, exemptions and credits on timely filed returns.

Underpayment occurs when taxpayers file their return but fail to remit the amount due by the

payment due date".

Since the U.S. - according to the IRS - underreporting of income tax, employment taxes and

other taxes represents about 80 percent of the tax gap, it makes sense to focus mainly on the

determinants of underreporting of income. That is actually what the empirical and theoretical

models do. The same is practical for the analysis of corporations, which understate not

income but profit. In the empirical part we will present various examples for evasion methods,

however, to decrease the reported profits of enterprises, basically only understatement of

revenue and overstatement of costs is what potential corporate evaders can do.

refer to the following URL for the definitions by the IRS

http://www.irs.gov/newsroom/article/0,,id=137246,00.html

2. Literature Review of previous studies of corporate tax non-compliance

2.1. Theoretical models of corporate tax evasion

2.1.1. The TAG-model

In the seminal theoretical literature about tax evasion

the decision to evade or not and how

much is presented as decision under uncertainty there is a tradeoff between a chance of a gain

if the evasion is undetected and a risk of loss if the evasion is detected and penalized. Thus the

compliance behavior is determined by the probability of detection and punishment, the penalty

structure and the risk aversion of the potential evader. Taxpayers are hence modeled as rational

calculators who are not honest or dishonest but just maximize expected utility. Before we focus

on differences about the demand for tax evasion by individuals or corporate taxpayers this very

basic theoretical model should be described. We follow basically the insight of the paper by

Cowell (2002) which explains comprehensively the TAG model (tax-payer-as-gamblermodel).

Cowell outlines that the TAG model is not strategic and it is atemporal, because neither the

reaction of the tax authority to the evader's signals nor vice-versa is modeled, which is not

realistic of course, but helps to explain the basic idea of the evasion decision. As described

above the taxpayer is performing a choice under risk and knows all relevant information as the

taxes that are liable be paid and the penalty for getting caught and he knows that the tax

authority has no information about his true liability unless the tax authority spends time and

incurs costs to find it out. In the paper of Cowell the following variables are used which are

taken over here: The taxpayer has an income y which is liable to a proportional income tax at

rate t. After the evasion decision was done the taxpayer is either not audited and receives a

consumption level c´ or he is audited and (if he evades taxes) is punished and if so the

consumption is c´´. If the amount of evasion is at Zero it follows c´=c´´; otherwise c´´<c´. In

this model it is furthermore assumed that the probability of detection, which is denoted by p,

(hence the probability of being undetected is 1 - p) is fixed no matter which proportion of

income is concealed and the tax on concealed income is subject to a surcharge at a rate s (s must

be positive). Based on these assumptions for each dollar of evaded tax one can think of a rate of

return r which is the sum of the value s with the detection-probability p and the value 1 with

probability 1 - p. So the rate of return to evasion is r = 1 p ps.

Only if all y is declared the

The seminal paper is by Allingham and Sandmo (1972)

it follows that the return on evasion is here independent of the proportional tax rate t. This is

already an extension introduced by Yitzhaki (1974), because in the original formulation of

Allingham and Sandmo the expected utility function was specified as

t]y

([1

t]y

([1

consumption level would be at [1 - t] y. But if income is concealed the resulting consumption is

given by a random variable:

c = [ 1 t ] y + r e t

(1)

where the not reported income is defined as e. We need to imply preferences of the taxpayers

over the consumption levels c´ and c´´ and assume the utility function of the "a-type" taxpayer's

preferences ( "a" includes all personal attributes as risk aversion or innate honesty ) is written as

the following expected utility function:

[1 p] u

(c´) + p u

(c´´)

(2)

and u

is an increasing and concave function. Consequently the marginal utility from increasing

consumption is increasing but at a decreasing rate. So everyone is assumed either to be risk

averse or risk neutral in this model to let it work. In addition, the slope of the indifference curve

of the tax payer in the neighborhood of perfectly honest behavior is at [1 - p] / p, that is on the

certainty line where c´=c´´. Making use of (1) it follows that the transformation rate of c´ and

c´´ is s, so the condition for maximizing utility with respect to concealed income if the tax-

payer conceals some but not all of his income is as following:

´´)

(

´)

(

Condition (3) has the interpretation "Marginal rate of substitution = proportional penalty (s)".

The difference is that based on the original formulation the penalty with probability p is based

on the understated income and not based on understated tax (but penalty for caught evaders on

base of the tax understatement reflects the practice in many countries

more realistically). The

expected payoff of each understated $ becomes then (1-p) t - ps instead of (1-p) - ps as stated

earlier. It is an important distinction which basis we choose for the penalty surcharge s

depending on which formulation is choose, increases in t would either proportionally increase

the reward of hiding income, but not proportionally increase the penalty ( making evasion

more attractive for increased t) or with the formulation as described above according to the

paper of Cowell, t only has an income effect in the utility function but does not change the

rate of return from hiding income (see also Slemrod, Joel and Yitzhaki, Shlomo (2000))

(3)

Figure 2 - Potential indifference curve and budget set taxpayer in

the ,,TAG-model"

There are two special cases in which the individual is completely honest, which results when

replacing "=" by "

" in (3) or it completely specializes in evasion when replacing "=" by " ".

The optimal evasion e* can be derived by the latter function and is determined by the tax-

enforcement parameters (p, s, t) and the individual characteristics (y, a). In figure 2 the utility

maximizing problem is showed graphically: The region O, A, B, C is the budget set of the tax

payer in this example and represents the tax-payer's income c´=[ 1 t ] y + e t if he is not

caught with probability (1-p) and the corresponding c´´=[ 1 t ] y - e t s if he is caught

occurring with probability p. Consequently the slope of the AB is s. Point A is the condition

where c´=c´´ and the taxpayer reports all income, thus e=0. Point B is the case where the

taxpayer conceals all income (e=y and c´=y and c´´=y[1-t(1+s]) . It follows that whenever the

slope of the indifference curve at A is lower than -s some income will be concealed.

It should be added that the fiscal systems of most countries imply values of E(r) greater than

zero and in the range of 0.99 to 0.75 (see e.g. Skinner and Slemrod (1985)) and people should

based on estimations of the degree of risk aversions in other situations evade much more than

they actually do, hence the results of the basic TAG-model don't fit with reality. For an

source: own drawing

According to Cowell (2002) the following deductions can be made from the TAG model: If r

>0 and the individual is an expected utility maximiser we can conclude it will always conceal

some income. In addition, when rewriting (3) as

set

)

]

([

)

]

([

we can find easily the equilibrium e* is decreasing as the probability of detection p or the

surcharge s is increasing, since u

is a concave function. Moreover, in order to receive explicit

results we assume that absolute risk aversion is a decreasing function of the consumption level

. This allows us to conclude 4 basic results listed below according to Cowell (2002):

1.) if the rate of return to evasion is positive everyone evades tax;

2.) people with higher risk-aversion tend to evade less;

3.) people with higher personal income tend to evade more;

4.) Increasing any of the tax-enforcement parameters p, s, t will reduce the amount of

concealed income.

Some doubt follows immediately about two conclusions. First, it is hardly imaginable that all

taxpayers will evade and secondly that compliance should be increasing with nominal tax rate t.

The review of empirical literature in section 2.2. will show whether this doubt is also justifiable

for corporations. However, the model helps us to know which parameters are essential in an

extension with a calibration of the parameters from empirical and experimental studies see

Bernasconi (1998) "Tax evasion and orders of risk aversion"

Absolute risk aversion is defined as

)

(

)

(

We should add to the this list of basic conclusions about the special role of the t with respect

to income and substitution effect: As described e.g. in Slemrod and Yitzhaki(2000) an increase

in t has always an income effect and since we assumed above decreasing absolute risk aversion

for an increasing consumption level, the income decline makes a smaller e* optimal, which is

the basis for the 4

basic result and works since t does not influence the tax evasion gamble

itself but results only in an income effect affecting the utility function. However we should

know that, if and only if the penalty is related to evaded income instead of evaded tax (which

we assume is not the standard case in most tax systems) there is also a substitution effect

because an increase in t would also increase the relative price of consumption in the state in

which the evader is caught and by that increases evasion.

(4)

empirical model specification though of course for corporate compliance analysis much more

parameters must be analyzed, such as firm size, country, industry and other possibilities or

obstacles to evade. This will be described later in the empirical analysis.

2.1.2. Critical discussion of the TAG model

There is plenty literature including critical discussion and extensions of the TAG model

. A

fundamental problem is the assumption that the tax payer is rationally optimizing expected

utility.

In general it is problematic to assume people behave rational, i.e. behave like economic

theory with concepts like utility maximization predicts - experimental research could already

show it is not necessarily true that individuals behave rational.

Moreover, also social questions

play a role - as Cowell outlines, the assumptions in the TAG model does not include state

dependent utility as feelings of shame or on the other hand feelings of enjoyment when evasion

is successful. In addition, Cowell describes that individuals may do systematic mistakes when

maximizing utility such as having misperceptions about the real probability of an audit. Cowell

suggests among other ways alternatively describing the model with the prospect theory as

alternative concept to the traditional utility maximizing theory the main argument is that the

latter can include that individuals rather use reference points from which they measure

outcomes instead of calculating absolute values in their value function.

Another interesting aspect not accounted for in the standard model is that the tax paid is used

more or less (depending on many factors as e.g. the efficiency of the public authorities) for the

production of public goods, thus, the cost of evasion is not only the risk of detection and

punishment but also indirectly the reduction of public goods. But we must notice a typical free-

rider problem exists as tax evaders are not excluded from enjoying public goods.

Another critical assumption is the fixed income and also the fixed detection probability in the

TAG-model. For the latter we will later go into details, for the former extended models include

for example the question of labour supply to make the income endogenous because it is

assumed that individuals make their decision how much to evade and how much to work

Andreoni, J., Erard, B and Feinstein, J.(1998) "Tax compliance", Cowell, Frank A.(2002)

"Stocks and Carrots", Slemrod, Joel and Yitzhaki, Shlomo(2000) "Tax avoidance, Evasion

and Administration"

refer for example to Gintis (1998) examining the role of the individual as `economic actor'

refer again to Cowell, Frank A. (2002) "Stocks and Carrots" for a large list of literature about

the prospect theory

for more arguments addressing the linkage of tax evasion and public goods and for literature

describing also welfare functions with respect to tax evasion refer to Myles (2001)

simultaneously

. Furthermore, it seems to be necessary to model the interaction process

between the evader and the tax-authority. This will be examined later in section 2.1.4.

The following approach deals with the TAG model with endogenous probability of detection:

2.1.3. The TAG-model and endogenous probability of detection

Since we stated earlier that the TAG-model overestimates the amount of evasion in the basic

formulation there might be another limit to the amount of tax evasion other than the risk

aversion of the taxpayer. A further problem is that the TAG-model is not effective in case of

risk neutral tax payers because then it results in an either or prediction in which either all or no

income is reported which is of course not conform to real observations. Both problems can be

solved by introducing an endogenous detection probability p, with p increasing when evaded

income is increasing. Slemrod and Yitzhaki (2000) specify the expected income (instead of

expected utility because we define the model alternatively for a risk neutral taxpayer now) as

following, while p[e] is a function for detection probability depending on the amount of tax

evasion e :

t))

t]y

p[e]([1

t]y

([1

p[e])

((

(5)

we assume that

and apply the first order condition on equation (5) to compute

the optimal e = e*. It follows that the condition for a maximum of expected income of the

potential evader is

1 p ps = p´ (1+s) e

(6)

The left handed side of (6) is the rate of return on each dollar of further evasion (actually the

same definition as already in 2.1.1) which increases expected income unless the rate of return is

zero or negative. This gain is offset by the right handed side by the fact that p increases in e (as

p' is positive). The right handed side increases as the amount of evasion e increases. For the

optimal amount of evasion e* both sides of equation (6) must be equal. Consequently as before

if the rate of return to evasion is positive everyone evades tax. Furthermore, in this model an

increase of the responsiveness of p to e increases compliance. The tax rate has no impact as it is

refer to Slemrod and Yitzhaki (2000) for a presentation of the related model

We go on while keeping the notation according to the symbols introduced in 2.1.1. conform

to the paper of Cowell (2002) instead of to the referenced paper of Slemrod and Yitzhaki

(2000) to let the equations be consistent and like in 2.1.1. for `y, e, p, s and t'.

not part of the condition for the maximum (because there is no income effect in this approach

with a risk neutral taxpayer) and in addition this model does not need any assumptions about the

risk aversion or absolute risk aversion. The equilibrium instead is determined by the

specification of the p[e] function. As in the standard TAG-model an increase of the penalty rate

s increases compliance. This concept with the endogenous detection probability can be also

applied to the riskaverse tax payer and it follows that in this case the gain in expected utility

is offset by both the increased risk bearing of the risk averse and the increased detection

probability due to

Furthermore, as Slemrod and Yitzhaki notice, the problem of the either or prediction of the

introductory model is, alternatively to the concept of an endogenous detection probability

function, also eliminated if different sources of income are subject to different detection

probabilities p, this is relevant for example if we think of employee labor who face a higher p

for employee income than for "moonlighting" income and by this simple concept they report

possibly all employee income but none of the "moonlighting" income and thus report a fraction

of total income.

2.1.4. Strategic models

The strategic approach is modeled in a game in which both the tax-payer and the tax-authority

maximize their revenue which is tax revenue for the tax-authority (government respectively)

and utility for the taxpayer and the outcome is an equilibrium in which both parties make their

best response to each other. It seems to be quite necessary to include both sides of the

compliance problem; not only the taxpayer but also the law enforcement agency and we also

should test whether former results are still stable then. In addition, this allows including more

parameters which affect the decisions of the enforcement agency as for example audit costs. A

basic approach is by Graetz, Reinganum and Wilde (1986). Graetz et al. argued, when they set

up their model, that a disadvantage of former theory was that the tax authority and their policy

were treated as exogenous parameters but in contrast the reaction of the tax policy should be

considered also and the interrelationships of taxpayers and authority in the tax collection

process should be modelled as well. They introduce the IRS as strategic actor into a game

theory approach. The Interactive theory according to the paper of Graetz, Reinganum and

Wilde should be introduced following: The law enforcement agency doing the audit is called

IRS following as it makes sense for the U.S.; we can think of any other authority performing

the audits. The IRS conducts audits at randomly selected groups of taxpayers. This is done by

compare to the explanation in Slemrod and Yitzhaki and further literature specialised into this

question is Usher (1986), Kaplow(1990) and Mayshar(1991)

a program known in the U.S. as TCMP (Taxpayer Compliance Measurement Program). The

IRS is assumed to be a rational maximizer of net tax revenue

which is risk neutral. The audit

selection process uses the information gained from the taxpayer i.e. reported income. After

the taxpayer reports income he will be audited or not. If not the declared income is the basis

for his actual tax liability. If the taxpayer is audited and misreports income he has to pay the

applicable tax and a fine in addition.

The taxpayer is assumed to be risk averse and utility

maximising. In order to keep the model as clear as possible the following simplifying

assumptions are made by Graetz et al.: There are two types of taxpayers: (1) strategic

noncompliers who strategically weigh their expected gain and loss from underreporting

and

(2) "habitual compliers" who just report honestly and do not play the "audit lottery"

. The

proportion of strategic taxpayers in the population is called , it follows the proportion of

habitual compliers is 1 . Income can be just high or low, denoted by I

and I

for high and

low income levels respectively, where I

< I

The IRS only observes whether high or low

income is reported but cannot directly observe income. The tax payments owed are denoted

by T

and T

for high and low incomes respectively. In addition it is naturally assumed that T

and T

. And eventually it is T

. The fine for discovered evaders is F 0.

Income levels for both kind of taxpayers, tax rates and fines should be exogenous in the

model which is a simplification.

The cost for a single audit is denoted by c 0 and T

+ F -

> c. Otherwise an audit would never be profitable for the tax authority, even not if it could

observe different income levels. Furthermore, it is assumed that taxpayers can always pay the

fine: T

+ F I

and T

+ F I

. Taxpayers who report honestly are never fined. In the basic

version - which is the only one introduced here - the taxpayers have no cost from the audit

itself if they are audited. The probability that a taxpayer has a high income should be q and 0

< q < 1. The game tree including the possibilities for both parties is showed following:

gross revenue plus fines and minus audit costs

such as assumed for the utility maximising taxpayer in the standard TAG-model above who

adjusts the amount of understatement depending on his degree of risk aversion and the other

relevant parameters

the reason may be they are persons who have no opportunity to evade tax as normal wage

earners or they just not evade due to moral reasons

refer to a detailed discussion about exogenous tax and fines in Graetz et al. However it makes

sense to treat fines and taxes exogenous because they are decided by the tax policy but not by

the tax authority.

Figure 3: possible actions by taxpayers and IRS

It shows first at the top of the game tree both type of taxpayers, the strategic and the habitual

complier. Both can have low or high incomes determined by the nature (step2) and both can

decide to report high or low incomes (step3) (but the habitual complier always reports only true

income in this step (right side)). Then the IRS decides to audit or not to audit (step4 and last

step in the game tree). Since there are no audit costs for the taxpayer we assume low incomes

will always report truthfully

. The dotted line areas show `information sets'. This means the

agent whose turn it is to make a decision cannot distinguish between the information's within

the same information set. This implies in this game tree that if high incomes are reported by the

taxpayer the IRS does not know which of the three situations is true for a report of high income:

(1) truthfully reported high income of strategic taxpayers, (2) not truthfully reported high

income of strategic taxpayers with low income or (3) high income correctly reported by habitual

compliers. However, there is no reason for the IRS to audit high income reports because it is

clear that it never pays to do so. More relevant is the second information set; we mark each

point of it with a green circle in the above game tree. Here the IRS must decide whether to audit

and does not know whether the report of low income they observe is of (1) a strategic taxpayer

who reports low income but has high income or (2) of a strategic taxpayer who reports honestly

If audit cost for the taxpayer would exist it would be necessary to consider that a higher

reported income may decrease the probability of the audit and the taxpayer may decide to

report higher income for this reason

source: Graetz, Reinganum and Wilde (1986), p. 10

low income because he has low income or (3) of a habitual taxpayer with low income. So it

follows that it is most interesting to consider the case what happens when strategic taxpayers

have high income: it is assumed they report low income in this situation with some probability,

denoted by , where 0 1. The IRS must evaluate their decision to make an audit or not. To

do so they need to compute the distribution of honest and dishonest taxpayers among low

income reports. The probability that a high income belongs to a low income report is denoted by

µ(). By making use of Bayes´s rule it follows:

µ() = q/[ q + 1-q ]

(7)

(7) is the probability that an observed report of low income is not true: q is the probability

that a taxpayer is a strategic taxpayer and has high income and reports low income, this is

related to the probability that low income is reported at all (represented by the right term in

brackets), the latter can happen if a taxpayer has low income (probability 1-q) or if he cheats

(probability q). As it makes no sense to audit taxpayers who report high incomes, thus the

strategy of the IRS is described by the probability that the IRS audits reports of low income, this

probability is given by and 0 1. The expected net revenue to the IRS can be written as

following:

)

(

]

)(

(

)

(

[

)

(

(8)

Hence, the net revenue depends on the audit probability () and the probability that high income

strategic taxpayers report low income () (but instead of using here we make use of the above

defined µ). The left term of equation (8) describes the expected value from an audit and the

right term the certain revenue T

. Expected value and certain value are multiplied by the audit

probability and (1- ) respectively. The expected utility function of the strategic taxpayer with

high income, who decides about the probability to report low income, is

)

(

)

(

)]

(

)

(

)

(

[

)

(

ßu

(9)

In equation (9) the first term is the expected utility of reporting untruthfully low instead of high

income and the second term the certain utility received if high income is reported. Also for this

model we expect as for the TAG model that utility is increasing in income but at a decreasing

rate (i.e. we assume the taxpayer is risk averse). The IRS is assumed to be risk neutral for this

reason we wrote above that `expected net value' is maximized. Equation (8) and (9) determine

best response functions for both agents which each depend on the strategy of the other agent. A

solution for this problem is a Nash equilibrium, which is a pair of strategies that are best

responses to each other. At a Nash equilibrium both the taxpayer and the IRS have no incentive

to change their strategies. The best response requirements to satisfy a Nash equilibrium are

written formally as following: For the IRS the best response strategy for a given is

)

(

))

(

)

(

And similarly for the potential noncompliers for a given the best response is

)

(

)

(

)

(

. It follows that a Nash equilibrium is a pair of strategies (*,

*) such that

(

ß*

(

and

To find the optimal to maximize the expected value of the IRS we need to derive the marginal

benefit of auditing a taxpayer who reports low income:

]

(

)

(

(10)

By (10) the marginal benefit of auditing a taxpayer who reports low income increases with the

probability that a taxpayer has high income but reports low income (denoted by µ as explained

earlier above) and by the tax owed by high income taxpayers (which is T

)

and by the fine (F).

Moreover, we know that the conditional probability of a high income for a given report of low

income (µ) is increasing with , q and , thus the marginal benefit of auditing a low income

report taxpayer is also increasing with these variables. On the other hand the marginal benefit

decreases with tax owed by low-income taxpayers (T

) and with the audit costs c which follows

also clearly from equation (10). Notice, there is no in the right handed term of equation (10)

consequently the marginal benefit is independent of the likelihood of auditing itself, it follows

that the benefits of auditing are linear in the audit probability. To know whether it pays to audit

we must just refer to the sign of (10), if for example µ is such that the marginal benefit of

auditing is positive, it always pays to audit, if marginal benefit is negative, it never pays to

audit. Thus, there is a critical µ so that (10) equals to zero which is

)

Consequently we also can write for the best response of the IRS:

)

(

)

(

]

[0,

)

(

)

(

where

)

(

is as defined in equation (7). In addition, since µ depends on , we may write the

response of the IRS depending on a critical as following:

]

[0,

)

(

(11)

and we obtain from equation (7) and (10) that

)

(

)

(

. (12)

It follows from (12) that

>0 is always valid unless the fine is infinite, which in turn is not

possible as we defined earlier that F I

- T

thus we know that the trigger value

is strictly

greater than zero and it follows that it is never a best response of the IRS to audit whenever the

probability of non-compliance is greater then zero. But contrariwise it is possible that it never

pays to audit, that is when

> 1 which results for example when c, the costs of an audit, are

high enough.

We continue with the response of the taxpayer who chooses a probability of non-compliance

depending on in order to maximise is expected utility as we explained above. Similar to the

expected net income of the IRS we derive the marginal benefit of non-compliance for the

taxpayer by computing the first order condition of the utility function (refer to equation (9)):

)]

(

)

(

[

)]

(

)

(

)[

(

)

(

(13)

This marginal gain decreases with the probability of audit (), the fine (F) and the tax owed by

the low-income taxpayers (T

). But an increase of T

has a more complicated interpretation as it

increases the gain from noncompliance but also the amount of loss from non-compliance in case

of detection, and it follows that for a risk averse taxpayer, the effect is ambiguous and depends

on the degree of risk aversion. That is because the nature of a utility function in case of a risk

averse taxpayer implies the utility of the expected value is higher than the expected utility.

Furthermore the marginal benefit is independent from the likelihood of underreporting the

income itself, thus the marginal benefit is increasing linear with . It follows immediately that

for a positive marginal benefit the taxpayer will never comply and for a negative marginal

benefit he will always comply and if the marginal benefit is zero he is indifferent - we can find a

critical so that (13) is zero:

)]

(

)

(

)]

(

)

(

[

(14)

then, since we know that the marginal benefit is linear in , we can write the best response of

the taxpayer depending on as

]

[0,

)

(

(15)

where

]

[0,

)

(

means that the taxpayer is indifferent to any probability of non-compliance.

Moreover 0 <

< 1 which is clear by equation (14) as

can only be 1 or larger if the

numerator of the right handed side of (14) would be larger then the denominator which is not

possible because of

)

(

)

(

and

cannot be 0 or smaller unless the

denominator of the right hand side of (14) becomes infinite which is not possible. Thus, the

trigger value of the probability of audit

strictly lies between 0 and 1 and only if this value is

sufficiently high the taxpayer will be compliant. At the same time it results of course that it is

certainly not necessary to audit every reported low income to achieve compliance. We will

show below graphically which equilibriums are possible by drawing the best response functions

for the IRS and the strategic taxpayer. We derived above that only for a certain trigger value of

the taxpayer is indifferent and his best response function can take values between 0 and 1,

otherwise it can be only 0 or 1 and equivalently for the IRS depending on the trigger value of .

However, it is possible that

is larger than 1 while for we have always 0 <

< 1. Thus, two

types of equilibrium are possible, for the first type when

> 1 there is no auditing as it never

pays to audit and all strategic taxpayers underreport - this is showed in figure 4

Figure 4 best response functions when

> 1

The equilibrium in figure 4 is (*, *) = (1, 0). The red dotted line denotes the best response

function for the IRS which is a zero audit probability for all possible in the range of 0 and 1.

The blue dotted line denotes the best response for the strategic taxpayer. His optimal probability

Source: Graetz, Reinganum

and Wilde (1986), p. 15

of underreporting is zero as long as is big enough, then for the critical denoted by

he is

indifferent (the horizontal blue line) and when is smaller than

underreporting pays and the

strategic taxpayer will underreport all income. Is it realistic that

exceeds 1 and it never pays

to audit? We can derive from equation (12) this situation actually occurs when the costs of

audit are high enough or the proportion of strategic taxpayers among all taxpayers is low

enough.

If auditing sometimes pays, that is

< 1, the situation becomes as sowed in figure 5. This

equilibrium is denoted by (*, *) = (

). So it follows that some taxpayers with high

income report low income and some low reports are audited.

Figure 5 best response functions when

< 1

By assuming for the following paragraph that the latter equilibrium (*, *) = (

) results,

i.e.

< 1, we apply comparative statics and analyze the determinants of 4 important

endogenous variables: (1) the probability of underreporting

, (2) the probability of audit

Source: Graetz, Reinganum

and Wilde (1986), p. 16

(3) the probability that any randomly choose taxpayer is a noncomplier, denoted by aggregate

noncompliance P

(refer to the introduction of the variables to verify)) and (4) the

probability that a randomly drawn report will be audited (denoted by aggregate probability of

audit P

)

(

)

. The material determinants of the latter 4 endogenous variables

affecting the resulting equilibrium are fines, audit cost, distribution of income, tax rates and

percentage of strategic noncompliers.

An increase of the fine results in a decrease of all 4 variables. The likelihood of non-compliance

decreases (refer to equation (12) to verify) and by that the aggregate non-compliance P

decreases. But also less auditing results and

decreases (refer to equation (14)) and by that P

decreases. In equilibrium terms the marginal benefit of auditing increases for an increase of F

and the critical must be smaller to let the IRS be indifferent, but also the marginal costs of

noncompliance increase for an increase in the fine rate, so the critical must be smaller to be

back in equilibrium (Refer to the green arrow in figure 5 indicating the direction of the new

equilibrium and ). What happens in other words is that the strategic taxpayers increase their

compliance rate because of the increased benefit of compliance due to the higher fine which in

turn allows the IRS to decreases the audit probability because of the increased marginal benefit

of auditing in the new equilibrium. The result is conform to the result in the TAG-model in

which an increased penalty surcharge increases compliance.

The introduction of the audit costs is an important extension of this theoretical part as in the

theory described in section 2.1-2.3 the impact of the audit costs was not modelled. The impact

of an exogenous increase of the audit costs in this model is an increase in both individual

noncompliance (

) and aggregate non-compliance (P

The probability of audit on the other

hand stays unaffected but the aggregate probability of audit is increased. This can be easily

verified when we argue again by the trigger value of . When the audit costs increase the

marginal benefit of auditing decreases (refer to equation (10)). Thus

must increase to offset

this change to be back to the equilibrium in which the IRS is indifferent. But

must not

change at all as the trigger value of does not depend on c. (The change of the equilibrium is

indicated by a black arrow in figure 5) But nevertheless the aggregate audit probability of

course increases because of the increase of the low-income reports for the increased probability

of non-compliance (refer to the above definition of P

If the share of high income in the distribution of income (q is the fraction of high incomes) is

increasing, this will increase the incentive to audit. This can be compensated, to keep the

which is correct because

)

(

is the probability that a report is a low income report

equilibrium, by a decrease of the probability of non-compliance (

) (the effect is indicated by

an orange arrow in figure 5). The type of argument is similar as for the explanation of the

increase of the audit costs but goes into the opposite direction and refers also to the change of

marginal benefit of auditing as the latter is increasing and not decreasing as it was for an

increase in c. Thus again

is unaffected (there is no q in the definition of

) but P

again is

affected and decreases due to the increased

. P

also decreases as it is linear in

To consider tax rates in this model we need to refer to which tax-rate is increased which is not

as simple as in the TAG-model which in turn includes only a proportional tax-rate on reported

income. The first possible case is an increase of taxes on high income while the tax on low

incomes stays the same, i.e. an increase of the progressivity of the tax system. The first effect is

that it pays more to underreport and the second effect is that marginal gain to auditing increases.

The second effect dominates and by that for an increased benefit of auditing the equilibrium

must be smaller so that equation (12) holds. Consequently individual and aggregate

compliance is increased. The effect on the equilibrium audit rate is ambiguous and depends on

the risk preferences of the taxpayers. Increased tax on low incomes has the opposite effect and

decreases individual and aggregate compliance.

Finally the effect of a change of the percentage of strategic noncompliers should be analyzed.

Graetz et al. mention as two possible reasons for a changed proportion of strategic actors a

change in the taxpayer's attitudes or alternatively a change of the structure of penalties and IRS

enforcement policies. An increase in the fraction of strategic noncompliers decreases

because

of more potential non-compliance increase the marginal benefit of auditing. However, the

aggregate compliance remains unchanged as the decreased rate of underreporting is offset by

the increase proportion of strategic noncompliers. Moreover, the audit rate is unaffected because

marginal benefit of underreporting does not change (which can be verified from equation (14)),

hence, also the aggregate probability of auditing is unaffected. It follows that the decreased

noncompliance rate is the only effect of an increased proportion of strategic noncompliers. In

addition as the audit rate does not change and the aggregate compliance rate does not change the

expected tax revenue is also unchanged which leads to the surprising result that an exogenous

increase of the proportion of strategic taxpayers does not change tax revenue

. Finally, the

analysis of comparative statics demonstrated, this model is quite powerful while including the

refer to Graetz et al. pp 20-21 for a more formal deduction for the tax revenue

tax authority and by that important parameters as for example audit costs are added to the

model

. Thus this model of Graetz et al. was an important development after the TAG-model.

2.1.5. Modeling the firm

The previous part provided an overview about individual taxpayers' compliance models the

following will concentrate on corporations. An introduction standard model for the behavior of

the firm will be described following close to Cowell (2002)

. This standard model successfully

demonstrates - as we will see later - at least the basic results of the existing empirical literature

about firms. Moreover, it is the theoretical starting point for the interpretation of the statistical

results of the survey in China and Germany examined in Section 3. The big difference of the

corporate model to the simple TAG model is the introduction of a decision about an output

level of the firm and in addition the introduction of an effort resulting from concealment

varying depending on how much profit is concealed. The firm is assumed to have constant

average and marginal cost and it is producing only 1 product. The output is subject to a

proportional tax rate t. The demand function in the market is written as x(P). Here - in the

simple version - the firm faces perfect competition and consequently the price P is constant in

the market. The proportion of concealed profit is denoted by and consequently the reported

profit is the proportion 1- . The unit costs of concealing are G() with

)

(

and

)

(

The detection probability p and penalty rate on evaded tax s are assumed to be exogenous as in

the TAG-model. The expected tax rate per unit of output is

]]

[

ßp

(16)

where r := 1 p ps, as before in the TAG model and the expected profit is:

[

]

x(P)

]

[

]

][

[

)

(

caught

4 3

4 2

caught

not

sß

ßG

(17)

which can be simplified to

[

]

x(P)

)

(

(18)

For further extensions of this model, for example for proportional tax rates and fines and

taxpayers audit costs and budget constraints refer to Graetz et al. pp 21-27

However the key reference for the model is by Gahvari (1993) and Virmani (1989)

where

)

( ß

is the average cost of concealment per unit of output and t is as in (16) the

expected tax per unit of output. So for a maximum of profit the first-order condition follows

based on (18):

)

(

dß

]]

[

)

(

(19)

Only for

]]

[

(20)

there results an interior solution for , otherwise would have the simple solution =0 since a

negative makes no sense. The following condition leads to the equivalent restriction for an

interior solution: t <t for 0<

1, so that the expected tax rate for any within 0< 1 can

only be smaller than the tax rate, otherwise the rate of return to evasion would not be positive

and the firm would be completely honest. This reminds us of the rate of return in the TAG-

model which was also required to be positive for an interior solution. Finally the market

equilibrium for a competitive market is

(21)

implying that expected profits are zero in the perfectly competitive market as the expected value

of the sum of constant average cost the tax and the average effort for concealment under the

optimal decision about must result in the constant market price P. Of course actual profits

vary for the interior solution depending on whether the firm is audited or not.

This model actually is a `double optimization problem' since the firm makes an output decision

and a decision about the optimal concealment , but, in this model the concealment decision is

independent of the output decision because the output has no impact on the profit per unit if the

model is stated as above. We sum up the conclusions below:

- The firm will conceal output up to the point where marginal cost of concealment are

equal to marginal reduction of the expected tax rate

- The firm will always conceal some output if t < t implying that the rate of return to

evasion is positive

- Reported sales decrease as the tax rate increases (in contrast to the TAG-model for

individuals in 2.1.1)

- An increase in tax increases the price but by less than the amount of the tax since some

of the tax increase is absorbed in increased evasion (because the increased tax leads to

an increased optimal )

- An increased probability of detection p or an increased surcharge s will decrease

concealment and will raise expected tax and the market price.

These basic results hold for the noncompetitive firms as well however, the equilibrium rule is

different (see Marrelli and Martina, 1988). According to Cowell the most important empirical

implications related to this theoretical part are as following:

- A firm's compliance is positively associated with being publicly traded and with

belonging to a highly regulated industry

- Having low profits relative to the industry median is correlated with higher corporate tax

evasion

We will review and extend these results in the proceeding analysis.

2.1.6. Critical Discussion of the "simple firm model" and prospects to modification

Extensions of the firm model apply to the Monopoly

and Oligopoly

case and in addition

refer to the impact on the equilibrium in the market and on the market structure and on welfare

implications of tax evasion. This paper in its main part sticks rather to how we can explain the

firm's evasion decision without referring to the resulting market equilibrium in particular. We

pick up below first the assessment by Cowell (2002) about the above firm model and then go on

with a survey of the insight of other recent literature addressing the theory of the firm's evasion

problem. Finally, we will proceed with an approach for a modification of the standard firm

model for endogenous detection probability.

See for example Marrelli and Martina (1988) for the noncompetitive industry

For example Goerke and Runkel (2005)

Cowell argues that in the above model 3 features stand out when comparing it to the TAG-

model for the individual taxpayer: That is (1) the nature of the taxpayer (2) the assumption

about risk preferences and (3) the determinants of responsiveness to economic incentives.

Cowell explains these items in detail starting with the taxpayer's nature:

The challenge is to model the entities of corporations since we cannot assume them to be

exogenously given as we maybe can apply it for individuals. Firms behave in response to

economic incentives mainly the tax system. They can change their shape and management. So

this includes an important difference to the household or individual. The idea linked to this

problem is that before we can model the tax system and its enforcement mechanism efficiently

it is essential to model the firm's behavior to enforcement policy and other incentives.

Furthermore, the risk aversion was an essential component of the TAG model but not of the

firm's model in which we assumed the firm to be risk neutral. Cowell outlines the neutrality

assumption is important to enable clear predictions from the theory.

The responsiveness of firms to incentives is one of the most researched questions in empirical

examinations. Cowell underlines the material difference between the TAG model and the firms

model in that respect: In the TAG-model one would explain different outcomes for the evasion

decision by differences in risk preferences or by different groups which either face different

penalty or detection probabilities. But in case of the firm, as it is modeled above, the

equilibrium is determined by the marginal concealment cost and the rate of return to evasion.

This approach is justified by the assumption that the function g in the model can include those

incentives firms face which of course can lead to different equilibriums. It is assumed that the

function g() is determined by the following:

- The nature of the product, because the output can be concealed harder or easier

depending on the type of product which is sold. For example output from physical

goods can be harder concealed than services. These circumstances result in more or less

opportunities to evade tax across different industries. Another intuitive example is that

self-employed can conceal much easier than employed.

- The size and organizational structure of the firm matters. Firms with a more complex

organization are assumed to have higher concealment costs and in such big firms the

But similar results will hold if the firm is risk averse it is straight forward that the risk averse

firm will ceteris paribus evade less than it's risk neutral counterpart

`problem' is that the more are involved in evasion activities the greater is the `security

problem'.

- A particular role of the reputation should lead to less evasion. Thus firms with a brand

name as assurance of their quality have more to lose in case illegal activity is proved,

the latter is again included in the g() as this can be interpreted as higher concealment

costs.

- As last impact on the concealment cost Cowell argues that the degree of concentration

of the industry has two effects: First, an auditor can easier police an industry with a large

number of firms, that is because in such an industries a firm deviating from the norm in

terms of reporting would be easier to spot and thereby such firms have higher

concealment costs. The second effect counteracting the first one is that a large number

of similar firms will encourage the spread of appropriate technology of concealment for

this industry.

It is furthermore essential if we want to understand the special problems of corporate tax

evasion to refer not alone to the complex organizational structure of the firm with respect to the

concealment costs (as described above based on Cowell) but also work out more detailed the

implications for tax evasion theory from separations of the ownership and control in a firm. The

recent literature about this problem is by Chen and Chu (2005)

and Crocker and Slemrod

(2003). The main result is that it is necessary to model the evasion decision in the context of the

contractual relationship between the manager and the owner. Crocker and Slemrod address the

fact that corporate tax officers will either have explicit compensation contracts in which their

success at tax minimization is rewarded, or will have this success reflected implicitly via

performance review procedures. Crocker and Slemrod found there is a major difference

between the case that the tax officer is penalized and the case that the corporation is

penalized. It is intuitive that penalties assessed on the tax officer are more effective (policy-)

tools than penalties on the corporation as a whole in the same amount because the former

Chen and Chu (2005) discuss the principal agent problem - relevant for larger firms - which

occurs when both the firm owner and the manager are involved in the evasion process in a

corporation. It is explained by Chen and Chu in detail why even for a positive rate of return

on evasion and a risk neutral owner (or principal in the notion of the principal agent relation)

tax evasion may be not profitable because the gain from evasion is offset by a loss due to a

efficiency loss within the contractual relationship between the owner and the manager. This

result of course supports the general idea that smaller firms, where the decision is only made

by an owner without such an efficiency loss, evade more.

complicate the conflict between the tax-manager and the shareholder and make evasion for

them together less profitable.

These recent findings by Chen and Chu and by Crocker and Slemrod give reason to imply that

it may be necessary to apply different theory for small and large corporations. Small firms are

more similar to individuals as they are likely to be risk averse and for them owner and tax

manager are the same actors, but in contrast large corporations and multi national enterprises

(MNE) imply completely different conditions as for example other possibilities to shift profits

by transfer pricing policies or other tax audit frequencies. And tax officers of an MNE or at

least of large firms follow their incentives based on their contracts. Thus, it can be questioned

whether it is possible at all to bring large and small firms together in one standard model such

as in section 2.1.5.

A further critical assumption in the introduction model of the firm's compliance similar as for

the seminal version of the TAG-model by Allingham and Sandmo which was described earlier

is the exogenous detection probability p. To find out about the question whether the

probability p is exogenous or endogenous it is among other assumptions crucial what we

assume about how the tax authority decides about their audit rule and about which firms should

be audited. That is because audit probability clearly influences detection probability positively.

The existing literature already suggests quite a number of determinants of the audit probability

but it should be always linked to something observable as reported profit or income, reported

cost or reported revenue.

This is reasonable because, as Lee (1998)

argues, the audit

probability cannot be based on a kind of statement deviation because any deviation will be

known after but not before the audit so it must be based on variables observable without audit. It

is mostly assumed in the standard model that the probability is a function of reported income.

In addition, the audit probability can depend on socioeconomic data

which however is not

directly relevant for the firm case. If the model should be valid for firms in all countries it would

For the complete elaboration and mathematical part refer to Crocker and Slemrod (2003)

refer to Marrelli (1984) or Marrelli and Martina (1988) or Lee (1998)

also refer to the assessment in Lee (1998)

for example Allingham and Sandmo (1972) or Reinganum and Wilde (1985)

for example in Germany, according to the monthly report of December 2002 of the

"Bundesministerium der Finanzen", tax authorities are recommended to group individuals

(which is an indication for the individual model but not relevant for our approach referring to

the firms model) by different ,,risk classes" depending on firstly how much tax they are liable

to pay and secondly on socioeconomic variables as example type of income, sector and

individual differences which are assumed to be correlated with tax moral

anyway be hardly possible to include all audit rules in one model.

But, we argue that an

endogenous detection probability in the firms compliance model does not necessarily need the

reason of endogenous audit probability we can alternatively assume detection probability is

influenced directly by the amount of evasion as we assumed for the TAG model with

endogenous detection probability

. The argument for this direct relationship of detection

probability and proportion of concealed income or output respectively is that detection

probability can rise by increased risk of detection during an audit (even if efforts to hide the

evasion increase), because if underreported profit is relatively high in comparison to a

benchmark (related to firm size and industry) known by the auditor or in comparison to the

output, the effort of the auditor to detect a reason for this finding can be assumed to increase.

Furthermore, if the output is understated and at the same time physical assets or employees

occur to be disproportionately much relative to the output reported, the auditor should also

become more skeptics.

All such circumstances may be assumed to increase efforts by the

auditor to detect evasion. In other words we assume it would be the easier to detect

characteristics which imply evasion, the more the fraction of concealed output would be.

Since this paper also focuses on China; an accountant in China was interviewed for this

paper relating to this question of endogenous audit probability. This accountant is an

accredited Chinese auditor (her name is Ms. Zhuo Shuqin and she is a senior registered

accountant in Xiamen, China (- for such sources the typical problem is that in particular

persons in China who are interviewed about confidential tax issues do not confirm to quote

full address, company and other information)) and we asked whether the probability of an

audit for a firm in china is fixed or changes based on the behavior of the firm (for example for

higher or lower reported income) or whether it is related to the kind of industry or similar

characteristics. Her answer was: "In China most firms needs to do Audit papers annually by

some authorized Accountant Company and then presented them to government bureau for

file. Some firms though few may not need to do audit every year under special permission.

The probability of audit doesn't change on base of the revenue a firm reports or on base of

costs or profit which is reported. In China since we have so many companies, big and small,

the government don't do the entire tax audit by themselves meanwhile, actually they are

entrusting accountant companies to carry out the tax audit and file out report to them. They

will then check by chance (maybe something about 10%) those tax audit reports and if they

find something inconsistent or suspicious, they will carry out tax audit themselves and if they

find false audit, then the accountant company as well as the firm they audit will be punished

seriously! An annual tax audit as well as other certificated inspection regulated by

government is a must for most companies in China"

though there no further specification was done why p should be increasing in e

which in turn is a similar argument as in Cowell (2002), however, he referred this to the

increasing concealment costs if firm behaviour is deviating from the norm

German tax authorities actually check according to the below article for deviations from the

norm and in particular for manipulations of values, i.e. over or understatement of figures. For

example they apply Benfords' rule which says that the probability that the first digit of a

number is a one is around 30%, the probability that it is a two is around 17 %, ..., and for

`nine' the probability is only 4.5%. Auditors detect misstatements of figures by comparing the

actual distribution with the expected distribution according to this law (refer to URL:

http://www.selbstanzeige.de/kanzleischwerpunkte/steuerrecht.html

and

http://www.daviddarling.info/encyclopedia/B/Benfords_law.html

)

Hence, we argue that a higher evasion relative to total real output decreases ceteris paribus

possibilities to conceal the hidden output successfully and thereby the detection probability

increases.

But, consequently the former argument is with respect to the result of the model very similar to

the earlier assumption that concealment cost increase for a higher evasion and by this reason

cost of evasion outweigh the gain of evasion. Finally, we may argue it basically makes no

difference for the result of the model whether the gain of evasion is offset by higher

concealment cost or increasing detection probability, but nevertheless it makes a difference for

the correct model specification

and thus it should be worthwhile to analyze which of both is

appropriate to describe reality

Since we suggest that detection probability increases due to a variable risk situation during the

audit and depending on how much output is concealed, we should add that this assumption can

only hold if firms internalize that they will be audited at all during the operating time of the firm

and by that internalize the increased risk. We assume it is definitely realistic that most firms are

audited during their lifetime and the decision makers know that

- of course such assumption

could never be made for an individual but for the firm it should be realistic. At least for large

enterprises it is common knowledge that they are subject to yearly audits. In addition, we will

report later based on Chinese and German survey data that in both countries firms of different

size (also smaller firms) expect an audit frequency average of around every 2 years in both

countries. Irrespective of whether the detection probability is increasing by either an increased

only for certain theoretical assumptions it makes a difference with respect to the separability

of the output decision and evasion decision as we will describe in detail earlier

Nevertheless with respect to empirical testing it can be expected that the basic firms model

as introduced in 2.1.5. - with fixed audit probability - already fits good with reality since with

an appropriate specification of the g() function the relating empirical model should work out

satisfying results because of a broad definition of the concealment cost and all characteristics

which can be found to influence the firms compliance can be explained by such cost

differences. Most empirical models are based on the research on determinants of the amount

of evasion, as for example firm size, sector, multinational or not, publicly traded and so on

(see for example Hanlon, Mills and Slemrod(2005)), these determinants can simplified be

interpreted as obstacles or opportunities for hiding output - the larger the obstacle the higher

the cost of concealment (in the broader sense ) - this idea is conform to the standard model in

which the cost of evasion g is the material effect which determines noncompliance (if p,s and

t are given exogenously)

in case of Germany for example refer to URL:

http://zeit.advogarant.de/index.jsp?Navi=instbpbetpruefI&ID=984564

, according to this

article German enterprises with more than 6 million yearly turnover are audited yearly, while

for smaller enterprises the audit frequency is much less and at around every 12 years, thus for

smaller enterprises the theory about endogenous audit probability is more relevant as

according to this article the report matters with respect to the audit probability, but for middle

and large enterprises it is more relevant to refer to the detection probability during an audit

(for Germany). This underlines once more the relevance of a different theory for small and

large enterprises.

audit probability for a higher fraction of concealed output or rather by an increased probability

of `success' of the auditor during an audit or by both it follows as suggestion for an improved

model to let p rise by the proportion of hidden output of the firm. However, we will explicitely

assume following that the endogenous detection probability results only by an increased

probability of detection during an audit. And we assume the audit is expected by the firm owner

or tax manager to take place and cannot be avoided anyway by any reported values to the tax

authority.

A further argument to use endogenous detection probability as concept in the firms model,

occurs when referring to the decision making of the firm: Is it really correct that firms - when

deciding about which output to hide check whether the rate of return is positive while

computing with a fixed p, s and t and then if so increasing the to the optimal level depending

on the cost of concealment? We argue instead that firms are rather affected by upcoming

opportunities and that they decide to use these options to evade or not, depending on how they

influence the detection probability during the audit. In other words a "nice option" will mean a

low detection probability which affects the compliance decision.

We assume if firms make use

of evasion opportunities to hide profit, in case they exist, they furthermore are likely to know at

the same time that the more output is concealed the more they raise the chance of detection and

will at some fraction of hidden output stop to increase further. In addition, also by using the

argument of `evasion opportunities' it would be possible - similar to the model with

concealment costs only to model different characteristics of the firm, since some can evade

without having a high risk of detection based on any characteristic as for example the industry,

while other firms are different in that respect which must be included in the function of

detection probability.

Moreover, if the endogenous detection probability is included into the model this allows for

changing the characteristic of the cost of concealment function. This may become necessary

when we question whether it is realistic that the function of the unit cost of concealment G() is

an increasing convex function. To challenge this suggestion reconsider the assessment in

Previous models already included variable audit probability for Monopoly and Duopoly

markets (Marrelli, 1984 and Marrelli and Martina, 1988 and Lee, 1998). The key literature for

the competitive industry is by Virmani (1989) and by Cremer and Ghavari (1994). Cremer

and Ghavari dealed with fixed audit probability, Virmani included variable audit probability

but determined by reported output and not referring to the situation during an audit as we will

alternatively assume in this section to keep the separability of output decision and evasion

decision.

that firms may make their decisions rather depending on upcoming opportunities to evade,

was also indicated during one interview we did with a respondent of the German survey of the

later section 3 of this paper

Cowell (2002) about what the G() function should explain according to Cowell it includes

the role of reputation (costly reputation loss), size and organization of the firm (cost of

complexity), nature of the product or industry (some outputs are easier to hide) and degree of

concentration (with effect on the visibility of concealment activity). For the reputation loss, the

firm organization, the nature of the product and the degree of concentration one could

alternatively think of a one time shift in G() based on these parameters or firm characteristics

instead of increasing unit costs by an increase of . (The size has a special role referring to the

formerly mentioned principal agent problem but the latter does not affect concealment cost in

the sense as used here). For example it is not intuitive that for a product which is particularly

hard to be concealed, this effect should disproportional increase for an increase in . By contrast

there may be also some fixed concealment cost if we think of a one time installment of a

complex evasion technology or if we think of the handling of evasion procedures. It is hard to

imagine that if an evasion technology is once introduced to evade 10 % of output it adds to the

unit concealment cost to evade 20% instead when the paperwork is the same but values change.

Even if more concealing is harder to be hided, also for fixed unit cost of concealment the total

cost of concealing would increase, but only linear in . If we allow for an endogenous detection

probability we can rather offset the return to evasion (if positive) by an increase of p which in

turn increases by . Consequently the model would work out traceable results as well for

constant unit cost of concealment (not depending on ).

Hence, we assume below fixed unit costs of concealment with respect to , but alternatively

depending on firm characteristics. We assume that various characteristics which are known to

have an impact on compliance, affect the unit cost of concealment. This should be summed up

in a function of unit concealment cost denoted by G(d). The constant d should include all such

determinants and various firm characteristics as size, age, owner structure, contractual

relationships and complexity of organization within the firm, industry, country, variables of the

business environment as regulatory cost or governmental services and business obstacles as

regulation, organized crime, government corruption

. This implies, when holding d constant,

the unit cost of concealment are fixed. By modifying the previous function (17), the expected

profits of the risk neutral competitive firm should be:

[

]

x(P)

]

)[

(

]

)][

(

[

)

(

caught

4 3

4 2

caught

not

sß

ßG

(22)

in the following empirical sections all these variables are described in detail

the variables P, m, , s, t and x follow the same definition as in (17) but the specification of G

changes and p depends now on ( p' > 0). We can write for the profit per unit of output

psßt

pßt

ßt

ßG

)

(

(23)

where p:=p(). The first order condition for a maximum is:

)

(

)

(

sß

)

(

)

(

(24)

For the optimal concealing rate the left hand side of (24) (which is the return to each dollar of

evasion) must equal the right hand side which is the costs per unit of evasion divided by t plus

the increase in p for the optimal multiplied by the penalty surcharge and . The right hand side

is comparable to the individual model for an endogenous p (see section 2.1.2.) but with costs of

concealment and t included here. Since we assume '

p >0 and s>0 we can see by writing

)

(

)

(

(24')

that the optimal can only be positive if the return to evasion minus fixed cost per unit of

concealment divided by t is positive (for p() and s always larger than zero), otherwise the firm

will report all income. So only if the fixed unit concealment cost G(d) are low enough the firm

will evade taxes. As in the introduction model an increase in t or a decrease in the penalty rate s

will increase noncompliance. For an interior solution of , when the amount of evasion will be

offset, will be determined by the specification of p(). This modified model has some

advantages: The interior solution for is only allowed for sufficiently low fixed unit cost of

concealment even if the rate of return to evasion is positive. So we can explain why some firms

evade no taxes even if the return to evasion - by the basic model - is positive. Whether an

interior solution results, now depends on the characteristics of the firm which should be

included in the G(d) function. Thus, G(d) can include all mentioned characteristics which were

assumed in the introductory model to rise the unit concealment cost. But, we assume they affect

the fixed unit concealment independent of while the increasing risk of detection for each

further unit of concealing offsets the gain from further evasion. Furthermore, this model is still

robust for fix unit costs of concealing which would, by contrast, not be possible in the earlier

presented model with fixed detection probability. Finally all other basic results regarding the

impact of tax rate and penalty rate of the former model above still hold. We present a numerical

example in Appendix 5 to verify the optimality condition in equation (24') yields reasonable

results.

The separability of the output decision and the tax evasion decision only holds if we assume that

the probability of detection is directly related to the understated fraction of output by the

argument that the risk of detection during an audit is increasing by the fraction of concealment

(p depends on but does not depend on X(P)). But in contrast, if the detection probability

would depend on the audit probability and audit probability would depend on reported output as

in Virmani (1989) or if the extent of tax evasion exerts a direct negative effect on gross profits

(Kreutzer and Lee, 1988, Virmani, 1989) separability would not hold. However, as stated above

this problem should not exist.

A further aspect is worthwhile to be considered which may play an increasing role for the firms

decision in a globalized environment: We may include in some cases the firms decision maker

can not only decide about the fraction of concealed output but can also compute first the

expected profit based on an optimal and afterwards depending on this result make a choice

about the country location of the firm or the head office of the firm. Determinants of this

decision are the result of the optimization regarding , the possibilities to change the location of

the business, the costs of such relocation and the amount of tax savings based on the potential

reduction of the tax liabilities. In such a model the set of actions by the firm would be extended.

It can be noncompliant or compliant within the same country but furthermore it can compare

this to the net profit in which it is compliant (or noncompliant) in another country with different

(less) tax liabilities. Since a location change has many other complicated determinants it is not

possible to model it easily within a tax evasion decision of the firm, however, it makes a

material difference for the resulting equilibrium to include this possibility of a country change.

It is known that low tax rates attract foreign direct investment (FDI)

. We may consider this

fact when referring to the tax evasion problem: Let us assume certain incentives exist for firms

in a country to make a country change, based on local and foreign tax rate differences and based

on a certain level of detection probability and penalty surcharge. And for these determinants

firms decide about a certain level of hiding output and afterwards decide about the country

change. Under these circumstances we consider a ceteris paribus increase of the detection

probability p or of the penalty surcharge s. The amount of firms changing the location would

refer for example to Altshuler, Grubert and Newlon (1998)

Details

Seiten
Erscheinungsform: Originalausgabe
Jahr: 2006
ISBN (eBook): 9783956361869
ISBN (Paperback): 9783836601566
Dateigröße: 7.4 MB
Sprache: Englisch
Institution / Hochschule: Universität Mannheim – Volkswirtschaftslehre
Erscheinungsdatum: 2007 (Februar)
Note: 1,3
Schlagworte: steuerrecht steuervermeidung verrechnungspreise globalisierung china

Autor

Alexander Teepe (Autor:in)

Corporate Tax Evasion in a Globalized Enivronment

Zusammenfassung

Leseprobe

Inhaltsverzeichnis

Details

Autor

Alexander Teepe (Autor:in)