VAT Fraud in Intra-EU Trade

VAT Fraud in Intra-EU Trade Katerina Gradeva Goethe University Frankfurt ∗ August 7, 2014 Abstract After the introduction of the EU Single Market i...
Author: Norah Walsh
6 downloads 2 Views 339KB Size
VAT Fraud in Intra-EU Trade Katerina Gradeva Goethe University Frankfurt



August 7, 2014

Abstract After the introduction of the EU Single Market in 1993, VAT evasion in intraCommunity trade has become a serious issue. Recently it has gained considerable attention from policy-makers. This study addresses empirically the issue of VAT evasion in cross-border trade in the context of intra-EU trade flows from the EU-15 to seven of the new Eastern European EU member states (Czech Republic, Estonia, Hungary, Latvia, Lithuania, Slovak Republic and Slovenia). It estimates the responsiveness of the trade gap to changes in the VAT rate for the time period of 2004-2009 at the six-digit product level, defining the trade gap as the difference between the value of exports to the seven new EU member states reported by the EU-15 and the value of imports declared in the Eastern European countries of the same product flow. This is the first study to investigate the link between the trade gap and the level of the VAT rate. The empirical evidence shows that the trade gap is positively correlated with the VAT rate in three of the seven Eastern European countries. Correcting for outliers and restricting the sample to large trade flows leads to a positive and significant correlation in five countries. In the latter cases a one-percentage-point increase in the VAT rate is associated with a 0.6% up to around 3% increase in the trade gap. Keywords International Trade · VAT Evasion · European Union JEL Codes H26 · K42



Department of Applied Econometrics and International Economic Policy, Goethe University. RuW Postbox 46, Gr¨ uneburgplatz 1, D-60323 Frankfurt am Main. E-mail: [email protected]

1

1

Introduction

Value added tax (VAT) is one of the most important sources of government revenue. Thus, a pronounced evasion might cause substantial shortage in the provision of public services and hinder fair competition since it entails a higher tax burden on the honest taxpayers. In addition, tax fraud is a threat to the proper functioning of the European Union (EU) internal market (Borselli, 2008). According to current estimates of the European Commission, 18% of the total VAT tax base in the EU is evaded (European Commission, 2013). While VAT evasion within one country has always been an issue for the national tax authorities, the case of VAT fraud in intra-EU trade flows has gained importance since the introduction of the EU Single Market. The present study aims to contribute to the growing literature on the problem of VAT evasion in intra-EU trade. The topic is currently extensively debated within the EU, with the EU Commission taking a leading role and initiating action plans and additional procedures and instruments to combat this type of fraud (EU Commission, 2012; Borselli, 2008). The subject is of high relevance also with respect to the recent Eastern EU enlargements and the potentially growing number of member states in the future which leads to an increase in the volume of intra-EU trade and thus, to a bigger scope of VAT evasion. In general, VAT evasion might occur in every international trade flow. Each importer has to pay the applied customs duties and VAT on the imported goods upon their arrival at the destination border. Hence, even in the absence of tariffs, which is the case with trade flows between countries that have a free trade agreement (FTA) or are members of a customs union (such as the EU), importers have an incentive to misreport their trade in order to evade VAT. The EU presents a specific case because, since the introduction of the EU Single Market, it is particularly vulnerable to VAT fraud in trade flows across its member countries. Beginning in 1993, the EU Single Market has led to the complete abolishment of all customs procedures at the borders across EU member states (free movement of goods) so that the VAT on intra-EU traded goods could not be collected anymore at the border before entering the destination country. Instead, the so-called “transitional” VAT regime, incorporating the “destination principle” in taxation, was established and is currently still applied (Baldwin, 2007; Keen and Smith, 2006). The transitional VAT regime implies that intra-EU trade is VAT-free; the importer receives the goods VAT-free from the exporter, charges the VAT in the importing country when selling the products to the customers and remits the collected VAT to the national tax authorities. Thus, the transitional regime breaks the VAT chain at a very vulnerable stage; all intra-EU trade flows are transported VAT-free, which leaves potential for massive VAT fraud (Baldwin, 2007; Keen and Smith, 2006). A prominent case of VAT evasion in intra-EU trade flows is the so-called “Missing Trader Intra-Community” (MTIC) fraud, which has gained increased attention from policy-makers and tax authorities in the recent years. MTIC fraud presents the extreme case, where the importing firm does not just underreport the correct value of its imports but rather disappears completely without transferring any VAT to the tax authorities, therefore the “missing trader”. This study addresses empirically the issue of VAT evasion in cross-border trade based on the case of intra-EU trade flows between the EU-151 and seven of the new EU member 1 Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden, United Kingdon (UK).

2

states (The Czech Republic, Estonia, Hungary, Latvia, Lithuania, The Slovak Republic and Slovenia, hereafter referred to as EU-7) for the time period of 2004-2009. The trade gap (or “missing” trade) is defined as the difference in each product-specific trade flow between the declared export value in the country of origin (EU-15) and the reported import value in the destination country (EU-7). I will then estimate the relationship between the trade gap and the level of the VAT rate in the importing (EU-7) country at the six-digit product level. Thus, the empirical analysis will consider all levels of underreporting of imports and not only the extreme case of MTIC fraud. In all EU-7 countries a standard VAT rate is applied to the majority of products; however, a reduced VAT rate exists for certain goods. The implementation of a reduced VAT rate, its level as well as the standard VAT rate vary across six-digit products and over the specified time period of 2004-2009. Currently, the standard VAT rate in the countries included in the sample ranges up to 27%. This is a substantial amount of taxes and thus potentially an incentive to underreport imports in order not to remit the collected VAT fully to the tax authorities. Hence, the estimation strategy relies on the variation of the VAT rates across products over time to identify the responsiveness of the trade gap to changes in the VAT rate. Using the same empirical approach, various empirical studies have shown that higher tariff rates are associated with a higher magnitude of missing trade, meaning that higher customs duties are correlated with more misreporting of imports (Fisman and Wei, 2004; Javorcik and Narciso, 2008, 2013; Mishra, Subramanian, and Topalova, 2008; Jean and Mitaritonna, 2010; Gradeva, 2014). Fisman and Wei (2004) estimate the responsiveness of the trade gap to a one-percentage-point increase in the tariff rate to be around 3%, whereas the other studies report a lower magnitude of the correlation between trade gap and tariff rate. Javorcik and Narciso (2008) investigate trade flows from Germany to ten Eastern European countries (including six of the EU-7 countries) and report a positive correlation between the trade gap and the tariff rate; a one-percentage point increase in the customs duty is associated with 0.4% to 1.7% increase in the trade gap. Gradeva (2014) examines the issue of tariff evasion in trade flows from the EU-15 to the ten Eastern European countries in the context of their EU accession in 2004 and 2007. The results of the analysis point towards a positive relationship between the trade gap and the tariff rate in the majority of countries, which is more pronounced when restricting the sample to manufacturing products. However, the potential correlation between the trade gap and the VAT rate in the importing country has so far been neglected in research. With a general trend towards lower tariff rates and higher number of FTAs, VAT rates might play a key role in explaining the presence of a trade gap. The present paper aims to fill this gap. This is the first study, which examines whether higher VAT rates in the importing country are associated with higher underreporting of imports. The pure existence of a discrepancy between the export and import data of the same product-specific trade flow cannot be, however, interpreted as a sign of evasion because of possible measurement error in the trade data due to timing, exchange rate or incorrect reporting issues. Hence, only a systematic correlation between the trade gap and the VAT rate will be interpreted as evidence for VAT evasion. Also, it is a standard procedure in international trade statistics to report imports including the cost of freight and insurance additional to the value of the product. So, in the absence of evasion and measurement issues, the declared value of 3

exports should be lower than the import value. The results show that a positive correlation between the level of the VAT rate and the trade gap exists in three of the seven countries in the sample, namely in Estonia, the Slovak Republic and Slovenia. The relationship ranges from a one-percentage point increase in the VAT rate associated with 1% increase in the trade gap in Slovenia to more than 3% in Estonia and the Slovak Republic. The correlation is robust to correction for outliers in the sample. Restricting the sample only to trade flows with a value which is higher than 50,000 US dollars and correcting for outliers as ways to avoid measurement error in the data leads to a significant relationship between the VAT rate and the trade gap, in addition to the other three countries, also in the Czech Republic and Hungary. The study by Allingham and Sandmo (1972) is one of the first theoretical contributions on the topic of tax evasion. According to their model, increasing the probability of detecting evasion and higher penalty rate lead to less tax evasion. However, the model cannot predict unambiguously the effect of changes in the tax rate on tax evasion; the relationship depends on the risk aversion of taxpayers. Assuming decreasing absolute risk aversion of the taxpayer, tax evasion becomes more profitable at the margin when the tax rate increases (substitution effect) but a higher tax rate reduces simultaneously the taxpayer’s wealth (income effect). Yitzhaki (1974) shows that when the penalty rate depends on the evaded tax, and not on the undeclared income as in the original model by Allingham and Sandmo (1972), a higher tax rate leads to less evaded income (assuming decreasing absolute risk aversion). Andreoni, Erard, and Feinstein (1998) discuss the puzzling fact that taxpayers are by far more honest and actual tax compliance is higher than the Allingham-Sandmo model predicts (evidence based on US tax data). The authors present evidence that tax moral, perception of fairness and satisfaction with the provision of public goods might explain the behavior of taxpayers. Andreoni, Erard, and Feinstein (1998) and Slemrod and Yitzhaki (2002) review the various augmentations of the original Allingham-Sandmo model and conclude that the empirical evidence is rather scarce and controversial, especially for other countries than the USA. In a more recent overview Slemrod (2007) discusses in particular the topic of VAT evasion, which does not apply to the USA (no VAT implemented) but is of high relevance to other developed countries as the EU member states. Several studies and national tax authorities estimate the magnitude of VAT fraud using a “top-down” approach based on national accounts data and input-output tables (Reckon LLP, 2009; Woon Nam, Parsche, and Schaden, 2001; Gebauer, Nam, and Parsche, 2005; Christie and Holzner, 2006; Agha and Haughton, 1996; HM Revenue & Customs, 2006, 2013). They apply a similar approach to estimate the “VAT gap”, the difference between a theoretical VAT liability and the total VAT receipts. The theoretical VAT liability is calculated on the basis of the categories of expenditure which contribute to the VAT base. The use of this top-down approach incorporates various assumptions in order to estimate the theoretical VAT liability. These include estimating the proportion of intermediate consumption on which VAT is not recoverable (especially regarding the financial, education and rental activities sectors), the size of adjustments needed for small businesses for which special tax schemes apply, and the correct matching of VAT rates to the product groups, which sometimes might be influenced by the judgement of the authors themselves (Reckon LLP, 2009). However, very few of the studies reveal 4

in detail the exact assumptions in the methodological part. So, although all of them employ a similar top-down approach measuring the VAT gap, their results would not be necessarily comparable. It should be noted here that besides fraud the VAT gap might be evoked also by legal tax avoidance, unpaid VAT due to insolvencies, incompleteneess or inaccuracy of the national accounts data (Reckon LLP, 2009). In contrast to all studies that calculate the VAT gap, this paper examines only one possible source of VAT fraud, namely VAT evasion in intra-EU trade flows. Measuring the VAT gap with a similar top-down approach as previously described, Agha and Haughton (1996) estimate the responsiveness of the VAT gap to the average standard VAT rate for a cross-section of 17 OECD countries for 1987. The results show that a one-percentage-point increase in the average VAT rate is associated with a decrease in the tax compliance rate by 2.7 percentage points. The complexity of the applied VAT scheme matters also for the evasion rate; an additional VAT rate lowers the compliance rate by 7.0 percentage points. This substantial effect should, however, be interpreted with caution since the number of VAT rates per country is limited and the empirical specification of the study does not account for the size of the tax base of each VAT rate. Christie and Holzner (2006) come to a similar conclusion when estimating the correlation between the VAT gap and the average weighted VAT rate for all current EU member states2 and Turkey for the time period from 2000 to 2003 in an empirical framework based on the Allingham-Sandmo model. Although the relationship is much lower, in their study a higher average VAT rate is also associated with a lower tax compliance rate. In addition, the authors conclude that the quality of the legal system as well as the satisfaction with public services matter for the VAT compliance rate. Reckon LLP (2009) estimates the VAT gap development in the EU member states (except Cyprus) for the period of 2000 to 2006 on behalf of the EU Commission (DG Taxation and Customs Union). The average VAT gap for the EU in this time period ranges from roughly 91 to 113 EUR billion, which represents 12% to 14% of the theoretical VAT liability. The results of the study suggest that the VAT gap has decreased steadily in Slovenia (from 16% in 2000 to 4% in 2006) whereas the other EU-7 countries do not report a clear trend. The VAT gap in 2006 is measured to amount to 28% for the Slovak Republic, 23% for Hungary, 22% for Latvia and Lithuania and 18% for the Czech Republic. The estimate for Estonia has increased from 12% in 2000 to 21% in 2004 and decreased afterwards to 8% in 2006. The descriptive statistics of the VAT gap in the study by Reckon LLP (2009) do not show a direct connection to the development of the trade gap used in this study. While according to Reckon LLP (2009) Estonia and Slovenia have the lowest VAT gaps in 2005 and 2006 among the EU-7 countries, these are two of the countries with the highest trade gaps and a robust positive correlation between the trade gap and the VAT rate. The empirical analysis of the study by Reckon LLP (2009) follows Christie and Holzner (2006) and examines the determinants of the VAT gap over the pooled sample of all EU member states (except for Cyprus and Malta) for the time period of 2000-2006. The study presents mixed evidence about the relationship between the VAT gap and the standard VAT rate. The results contradict the findings of Christie and Holzner (2006): the positive and significant correlation between the VAT 2

Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, Netherlands, Poland, Portugal, Romania, Slovak Republic, Slovenia, Spain, Sweden, UK.

5

gap and the VAT rate turns to be negative once controlling for possible autocorrelation and heteroskedasticity. Thus, a higher VAT standard rate is associated with a smaller VAT gap, which is the predicted effect in the model by Yitzhaki (1974). The study provides some evidence that the VAT gap decreases with a longer EU membership. Woon Nam, Parsche, and Schaden (2001) calculate the VAT gap rates for ten of the EU member states for the years 1994, 1995 and 1996 (due to data issues for France, Italy and the UK, the estimates for these three countries are for the years 1991, 1992 and 1993). The average ratio of VAT non-compliance over the respective three years ranges from 2.4% in the Netherlands to 34.5% in Italy. For Italy, Greece, Spain and Belgium the share of the VAT gap of the hypothetical VAT tax base amounts to around 20% or more while in the Netherlands, the UK, Denmark and Germany the average share is less than 5%. In a follow-up study Gebauer, Nam, and Parsche (2005) estimate the VAT evasion ratio for Germany for each year between 1994 and 2001. Although not constantly increasing, the VAT non-compliance ratio is 9% higher in 2001 than in 1994 and amounts to 9.6% of the total VAT liability. The UK tax authority, HM Revenue & Customs (HMRC), is one of the most active national tax authorities in the fight against VAT evasion. For more than a decade it has been issuing a report on “Measuring Indirect Tax Losses” (HM Revenue & Customs, 2006). HMRC uses a top-down approach to estimate the share of the VAT gap out of the theoretical VAT liability. Overall, for the financial years between 2001-2002 and 20112012, the VAT gap ranges between 10% and 16% of the total hypothetical liability, being stable around 10.5% in the last two periods, 2010-2011 and 2011-2012 (HM Revenue & Customs, 2006, 2013). In a joint study of the HMRC and the Office for National Statistics Ruffles, Tily, Caplan, and Tudor (2003) investigate the extent of the MTIC fraud in the UK and its impact on trade statistics and balance of payments. However, the authors do not explain the methodology used to calculate the amount of MTIC fraud, referring to confidentiality issues. A massive MTIC fraud would bias the official trade statistics with the other EU member states because the import data and the trade deficit would be understated. Ruffles, Tily, Caplan, and Tudor (2003) apply their adjustment method only to certain transactions involving products that present an interest for fraudsters (mobile phones and computer components). The authors calculate the necessary adjustments for the effect of MTIC fraud on UK imports from the rest of the EU for the years 1999-2002 and show that the impact of MTIC fraud has increased within the four-year period: MTIC fraud amounted to 0.7% of the total imports from the EU in 1999 and around 4.0% in 2002. Ruffles, Tily, Caplan, and Tudor (2003) argue that MTIC fraud is partly responsible for the increasing asymmetries in the trade figures between EU member states. After 1993, when the EU Single Market was introduced and the opportunity for massive MTIC fraud arose, the discrepancy between the UK import data and the export figures of the other EU member countries to the UK has continuously increased. The present study builds on this finding since it examines the extent to which the level of the VAT rate of a given product can explain the existence of a mismatch in the mirror trade statistics. Several studies offer empirical evidence on the topic of tax morale and tax evasion in the Eastern European EU countries based on survey data (Torgler, 2004, 2011; Torgler, Demir, Macintyre, and Schaffner, 2008; Hanousek and Palda, 2008; G¨erxhani, 2004). 6

High perception of corruption, high number of evaders and a malfunctioning tax administration might lower the public trust in state officials and with it, the tax morale in the society, which in turn leads to more tax evasion (Torgler, 2004, 2011; Torgler, Demir, Macintyre, and Schaffner, 2008). The empirical evidence from the EVS shows that during the accession period and the first years as EU members there has been some improvement in the tax morale in some countries (Gradeva, 2014). However, this phenomenon is not observed in all countries. Corruption and a high acceptance of bribes still remain important issues to tackle in these states. Hereafter, the study is structured as follows: The next section describes in more detail the specific features of the EU Single Market with regard to VAT collection and explains the reasons for the increased probability of VAT fraud in intra-EU trade flows after 1993. Section 3 reviews the data used in the empirical analysis and possible data quality issues. Section 4 presents the empirical strategy of the study. Section 5 discusses the results and the last section offers conclusions.

2

The EU Single Market and VAT fraud schemes

Before the introduction of the EU Single Market in 1993 the EU states had enjoyed full national VAT authority, which should have changed afterwards. However, the member countries could not agree upon a common VAT rate and a uniform VAT regime for the whole EU. Hence, since the differences across the national legislation were too big, a “transitional” VAT regime has been adopted within the EU and is still implemented until present day. Even though the VAT rates are not fully harmonized, the common EU VAT system sets the general framework, which has to be applied in all member states in order to ensure “neutrality in competition, such that within the territory of each Member State similar goods and services bear the same tax burden,...” (The Council of the European Union, 2006). With respect to the national VAT rates, the EU members and the EU Commission have agreed upon a minimum standard VAT rate of 15% and a minimum reduced VAT rate of 5% as well as the product categories for which a reduced VAT rate or exemptions from VAT can apply (The Council of the European Union, 2006). The product categories, which can be taxed with a reduced VAT rate, include broadly defined foodstuffs, pharmaceutical products, medical equipment and books. Importers had presumably, already in the pre-1993 system, incentives to underreport their imports in order to evade VAT (tariffs had already been abolished on intra-EU trade flows). However, after the introduction of the EU Single Market some of the key features of the VAT system have changed so that VAT evasion has become a bigger issue in intra-Community trade. After 1993, all border controls had been abolished so the VAT on imported goods could no longer be collected before entering the destination country. Instead, now the importing firm receives the VAT on the imported products after selling them to its customers and has the obligation to remit the collected VAT to the national tax authorities. In addition, the “destination” principle, as in most international trade flows, applies in business-to-business relations. This means that exports are VAT-free so that the importing firm collects from its customers the VAT on the whole production chain and not only on the own value added. The main argument for the destination principle is that there are no incentives for delocation of industries to low-tax countries

7

because all supplies are taxed with the same VAT rate, independent of whether they are produced in the home country or imported (Baldwin, 2007). However, there are several major objections against the transitional VAT regime. First, the implementation of different VAT regulations for acquisitions from national or foreign suppliers leads to an asymmetry and high compliance costs. Second, the transitional regime breaks the VAT chain and leaves opportunities for massive VAT fraud. Third, the transitional regime sets incentives for price arbitrage to cross-border shoppers; however, this issue is regarded as one of lesser importance (Baldwin, 2007). VAT fraud can take place through different channels: underreported sales (“off the books” sales), no firm registration to the tax authorities (“ghost” firms), misclassification of products in the case when a firm sells several goods taxed at different VAT rates, false claims for credit based on overstated VAT paid on inputs and imported products which are not brought into tax (Keen and Smith, 2006). The latter mechanism of VAT fraud is of most interest for this study. Two simplified examples show the differences in the VAT collection between domestic production and a production chain which involve importing goods from other EU members and the incentives for VAT fraud in the latter case. Figure 1 presents a very simplified example of domestic production and sale of cell phones in Slovenia, where currently a 20% VAT standard rate is applied. A supplier sells inputs for cell phones with a value of 800,000 EUR to firm A and charges 160,000 EUR VAT (20% of the value of the inputs). Thus, firm A pays in total 960,000 EUR (inputs’ value plus VAT) to its supplier. Using the inputs Firm A produces cell phones and sells them to firm B for 1 million EUR, collecting additionally from company B VAT at the amount of 200,000 EUR. The net VAT liability of firm A to the Slovenian tax authorities is 40,000 EUR (200,000 EUR - 160,000 EUR) since each firm has to pay VAT only on its own value added. Firm B buys the cell phones from firm A for 1.2 million EUR, including VAT, and resells them to the final customer for 1.3 million EUR plus 260,000 EUR VAT. So, company B has to transfer 60,000 EUR to the tax authorities. That is the difference between the collected VAT on the sold products and the VAT that the company has paid already for the products. In this case, the maximum VAT that company B can evade are the 60,000 EUR VAT due (this would be the extreme case where firm B does not report any disposals and sells all products “off the books”).

8

Figure 1: VAT liability on domestically produced and sold products (the country of Slovenia as an example).

9

Figure 2 shows the case where the cell phones are not produced domestically but imported from another EU member state, in this case Austria, where the standard VAT rate is currently also 20%. The first step in the production is similar to the previous example, only it takes place abroad. The major difference is the disposal of the cell phones from firm A to firm B. Because firm A sells the cell phones now to a company abroad, it does not charge any VAT upon the 1 million EUR valued products. Instead, firm A sells the cell phones VAT-free to company B and receives from the Austrian tax authorities the VAT that it has already paid on its inputs (160,000 EUR). Firm B in Slovenia pays now only 1 million EUR for the cell phones (without the 200,000 EUR VAT as in Figure 1 ) and resells them again for 1.3 million EUR plus 260,000 EUR VAT to the final customers. In this case, firm B owes the Slovenian tax authority 260,000 EUR since it has not made any VAT payments in previous production stages which can be deducted (the imported products are VAT-free). Thus, the maximum VAT that firm B can now evade is the VAT on the whole production chain in the amount of 260,000 EUR, which is substantially higher than in the case shown in Figure 1.

Figure 2: VAT liability on foreign produced and domestically sold products.

The second example shows the increased opportunity of evading VAT in intra-EU flows after the introduction of the EU Single Market when the importing firm is itself responsible for remitting the collected VAT to the tax authorities. The widely discussed MTIC fraud is similar to a situation as in Figure 2 but instead of just underreporting its imports firm B goes bankrupt (the missing trader) before transferring any VAT to the tax authorities. This would be the extreme case where firm B evades completely the VAT on the whole production chain (Baldwin, 2007; Keen and Smith, 2006; Borselli, 2008). 10

Another fraud scheme that has recently gained increasing attention among policy-makers is the so-called “carousel fraud” (Baldwin, 2007; Keen and Smith, 2006; Borselli, 2008). The carousel fraud includes a MTIC situation, where the importing firm (Firm B in Figure 2) sells the goods to another firm, firm C, in Slovenia and goes bankrupt. Firm C, not necessarily involved in the fraud scheme, buys the products from firm B and resells them to firm D in Slovenia. Firm D exports the products back to Austia or another EU country, thus collecting the VAT that the missing trader (Firm B) should have remitted to the Slovenian tax authorities. The products can be then circulated around several times, including also exporting to a non-EU member country which makes it more difficult to be tracked and re-exporting the goods back to the EU, therefore “carousel fraud”. In most cases all firms are involved in the criminal acitivity but there can be also “honest” companies as “buffers” (for example firm C) in order to complicate the fraud scheme. Preferred products for these types of frauds are goods with high value but low weight like computer chips or cell phones (Baldwin, 2007). This paper considers in the empirical analysis the trade gap, the difference between the export and import value of the same product-specific trade flow. MTIC or carousel fraud schemes might be the reasons for the existence of the trade gap but the asymmetry in the mirror trade statistics might be caused also by only partly evading VAT through underreporting of imports. Thus, it will not be possible to differentiate in the empirical analysis across the different types of VAT evasion.

3

Data description

The UN COMTRADE database of the United Nations Statistical Division, provided by the WITS database, offers the information on bilateral trade flows between the EU-7 and the EU-15 countries for the time period of 2004-2009. The UN COMTRADE database contains two figures belonging to the same trade flow: the value of a product that is declared in each EU-15 country to be exported to each of the EU-7 countries (hereafter referred to as “exports”) and the reported import value of the same product in the EU-7 countries (hereafter referred to as “imports”). Thus, each trade flow is registered twice, once when leaving the country of origin and secondly when arriving to the destination country. The data are collected by the national agencies and then transmitted to the United Nations Statistical Division, which converts the values into US dollars, if necessary, to make the figures comparable across countries. Apart from the costs of freight and insurance, there should be no other difference between the export and import value of the same trade flow. In international trade statistics, exports are reported as f.o.b. (free on board) and imports as c.i.f. (including the costs of insurance and freight). Thus, in the absence of measurement error or evasion the import value should be higher than the corresponding export value. Trade flows in the sample are declared according to the Harmonized Commodity Description and Coding System (HS) at the six-digit product level which is the highest available disaggregated level. The trade data are classified according to the initial HS product classification, HS 1988/92. One issue in the trade data for Hungary concerns the implementation of a threshold of 1,000 US dollars per product category for a given year and partner country below which no imports are reported. Therefore, all exports to Hungary with a value below 1,000 US dollars will be excluded from the empirical analysis in order not to bias the results.

11

As previously mentioned, since the introduction of the EU Single Market the intra-EU trade data are collected not on the basis of customs declarations but through the Intrastat system (European Council and European Parlament, 2004; European Commission, 2004). Each firm, which is importing to/exporting from another company within the EU, has to declare its trade within the Intrastat system and the information is then submitted directly to the national statistical offices. The latter process the data and transmit it to other statistical databases. Small firms are the only exception from the regulation since they do not have to fill out the Instrastat declarations in order to have lower administrative burden. The threshold for defining the firms which do not have to comply with the general rule is set each year and for each country in a way that the collected Intrastat declarations cover 97% of all exports and 95% of all imports (European Council and European Parlament, 2004). The national statistical offices gather information on the firms which remain below the threshold based on their VAT declarations and estimates from previous years. There is no available database that offers information on the VAT rates at the sixdigit product level for the EU member countries. The data in the present study are gathered with the help of the national VAT legislation of the EU-7 countries, which has been in force in the time period of 2004-2009 (Czech Republic Government, 2004; Estonia Government, 2010; Hungary Government, 2004, 2009, 2006; Latvia Government, 2010; Lithuania Government, 2006, 2009; Slovak Republic Government, 2008; Slovenia Government, 2002; EU Commission, 2014). The public availability of a translated version of the VAT laws is the reason why the sample does not include all Eastern European EU members.3 Some of the VAT laws define in a very detailed way the product codes which are taxed with a reduced VAT rate (for example the Slovak Republic and Hungary). For the majority of the countries only the product description is available which has to be matched with the correct product codes. The matching is done as rigorously as possible and equally for all countries, however, there is still potential for some errors in the VAT data. Product groups, where it is not possible to identify whether all goods included in a certain six-digit product category are taxed with the same VAT rate, are defined as “mixed” product codes and will be not considered in the empirical analysis. An example for such a “mixed” product is the six-digit product code 853180 “Other electronic sound or visual signalling apparatus”, which in the case of the Slovak Republic is taxed with the standard VAT rate except when the final customers are persons with hearing and/or visual disability, then the reduced VAT rate applies. The VAT rates applied in the EU-7 countries in the sample period 2004-2009 are shown in Table 1. In the years when the VAT rate has changed during the year, the VAT rate for the respective year is calculated as the weighted average of the two different rates according to the number of months they have been applied. The implementation of a reduced VAT rate is decided by the national government. The column “Reduced rate 1” of Table 1 reveals that all countries in the sample have applied at least one reduced rate during the time period of 2004-2009, with Hungary and Lithuania being the exceptions where there has been also a second reduced rate (“Reduced rate 2”). There has been 3

VAT data for Poland and Romania are missing. Bulgaria is not part of the sample because the standard VAT rate of 20% has not been changed during the time period and no reduced rate has been implemented.

12

at least one change in the VAT rates in each country with the only exception of Slovenia where the VAT rates have remained the same. In the majority of countries (Czech Republic, Estonia, Hungary and Latvia) both the standard and reduced VAT rates have changed during the sample period, whereas in Lithuania only the standard rate and in the Slovak Republic only the reduced rate vary in the years between 2004 and 2009. Overall, the VAT rates show rather little variation, which should be considered when analyzing the results of the empirical analysis. Table 1: VAT rates between 2004 and 2009. Country Czech Republic Czech Republic Czech Republic Estonia Estonia Estonia Hungary Hungary Hungary Hungary Latvia Latvia Latvia Lithuania Lithuania Lithuania Slovak Republic Slovak Republic Slovenia

Standard rate

Reduced rate 1

Reduced rate 2

Implementation period

22 19 19 18 18 20 25 20 20 25 18 18 21 18 19 21 19 19 20

5 5 9 5 9 9 5 5 5 5 9 5 10 5 5 5 10 8.5

15 15 18 9 9 9 -

01.01.2004-30.04.2004 01.05.2004-31.12.2007 01.01.2008-31.12.2009 01.01.2004-31.12.2008 01.01.2009-30.06.2009 01.07.2009-31.12.2009 01.01.2004-31.12.2005 01.01.2006-31.08.2006 01.09.2006-30.06.2009 01.07.2009-31.12.2009 01.01.2004-30.04.2004 01.05.2004-31.12.2008 01.01.2009-31.12.2009 01.01.2004-31.12.2008 01.01.2009-31.08.2009 01.09.2009-31.12.2009 01.01.2004-31.12.2006 01.01.2007-31.12.2009 01.01.2004-31.12.2009

Applied VAT rates in the EU-7 countries in the sample period 2004-2009. Source: EU Commission (2014)

4

Estimation strategy

Following the specification of Fisman and Wei (2004), the trade gap is defined for each of the EU-7 countries separately as:

T radeGapcpt = lnExportscpt − lnImportscpt

(1)

where c stands for the partner country (EU-15), p for a six-digit HS product code and t for year (2004-2009). Thus, the trade gap is equal to the difference in the log-value of exports and the log-value of imports of the same product-specific trade flow. According to Equation 1, the trade gap is calculated only for trade flows for which the export and the import values are non-missing and non-zero. Equation 1 implies that in the absence 13

of evasion and measurement errors the trade gap should be negative since imports are declared including the costs of freight and insurance additionally to the export value. Therefore, the higher the costs of insurance and freight are, the higher the trade gap should be in absolute terms. Gradeva (2014) discusses in detail the reasons, other than evasion, for the existence of a discrepancy between the export and import value of the same trade flow and the case of asymmetries in mirror trade statistics within the EU in particular. Incorrect specification of the exporting and importing firm, respectively of the countries of origin and destination, is one of the most common mistakes in trade documents, especially in the cases of transit trade or trade involving export-processing zones (De Wulf, 1981; Yeats, 1995). Another explanation for the trade gap is a mismatch in the timing information for exports which are dispatched at the end of a certain year but the imports arrive in the beginning of the following year. Additionally, differences in the export and import values might arise due to different currencies and the fluctuations of their exchange rates to the US dollar. Ruffles, Tily, Caplan, and Tudor (2003), Loschky (2006), Koufen (2001) and Krockow (2007) consider the specific sources of the existing asymmetries in intra-EU trade flows. The authors see the main reasons in the incorrect identification of the product code and of the origin and destination country. Incorrect time mapping and exchange rate fluctuations also play a role. A specific issue of the implementation of the Intrastat system is the exclusion for small firms. Since the thresholds for submitting regular Intrastat declarations vary across EU members and time, it is possible to have only one of the exporting and importing firms falling into the category below the threshold. Also, the national statistical offices of the EU members do not apply the same method for estimating the trade flows of firms below the reporting threshold which might lead to differences in the results. Ruffles, Tily, Caplan, and Tudor (2003) and Loschky (2006) examine the development of the trade asymmetries for the UK and Germany respectively, particularly after the introduction of the EU Single Market and the change in the reporting system. While Loschky (2006) sees the German bilateral trade asymmetries with Italy, the UK and the Czech Republic descreasing after the replacement of the customs declarations with the Intrastat system, Ruffles, Tily, Caplan, and Tudor (2003) present evidence that the UK missing trade with the other EU member states has been increasing since 1993. Loschky (2006) argues that the improvements in the trade data are caused by less corruption at the customs since the importers no longer have any interaction with the customs officials and submit their Intrastat declarations directly to the national statistical office. Another reason is a better coordination across the national statistical offices. On the other hand, Ruffles, Tily, Caplan, and Tudor (2003) attribute the higher UK trade gap with the other EU members to the increased opportunities for VAT evasion in intra-EU trade flows, particularly in the form of MTIC or carousel fraud. To conclude, one cannot interpret the pure existence of the trade gap as a proof of VAT evasion in the trade flows between the EU-7 and the EU-15 countries because of the various possible data quality issues or of a discrepancy in the export and import data due to a random measurement error. Thus, analyzing the development of the trade gap on the basis of descriptive statistics alone has a limited informative value with respect 14

to VAT evasion. It is not possible to disentangle the “real” trade gap from the noise in the data. Therefore, the empirical analysis in this study is concentrated on the question whether there is a systematic correlation between the level of the VAT rate and the trade gap. Assuming that the measurement error is random with respect to the VAT rate, such a systematic correlation may not exist in the cases where measurement error causes the existence of a trade gap. Defining the trade gap as Equation 1 does, the main specification in the empirical analysis, which is estimated for each EU-7 country separately, would be as follows :

lnExportscpt − lnImportscpt = T radeGapcpt = β0 + β1 V ATpt + αt + γc + δp + cpt (2) where c stands for the partner country (EU-15), p for each six-digit HS product code and t for year (2004-2009). Equation 2 estimates the responsiveness of the trade gap to changes in the VAT rate for trade flows between the EU-7 and the EU-15 states for each EU-7 country separately and tests whether there is a systematic and significant correlation between changes in the VAT rate and the trade gap. The estimated coefficient on the VAT rate should be positive and significant if a higher VAT rate is associated with more missing trade. The model exploits the variation of the VAT rate across products (standard vs. reduced VAT rate) and time within each EU-7 country. The VAT rate is the same for all products within the same product code, independently of the country of origin so the VAT rate does not vary over the partner countries. The preferred specification of the model includes year, partner country and product fixed effects. The year fixed effects control for particular changes in a certain year that might influence the trade gap such as a reform in the administration or changes in the reporting. The partner and product fixed effects account for any partner country or product characteristics which are systematically correlated with the trade gap. Because the VAT rate does not change regularly and there is little variation at the six-digit product level, the product fixed effects, which are included, will be at the four-digit level. Another possible specification might include partner country-year fixed effects in order to control for specific time trends of the partner countries. However, the results of this specification are similar to including separate year and partner country fixed effects, so they will be not presented. Equation 2 assumes that the VAT rate changes exogeneously with respect to the trade gap. The level of the VAT rate is a government decision, thus endogenously set, and part of the tax system in a country. However, the VAT revenues are generated mostly through the domestic market and to a lesser extent through international trade. Therefore, it is unlikely that when considering a change in the VAT legislation the government would take into account the trade gap at the six-digit product level.

5 5.1

Results Descriptive statistics

The descriptive statistics of the trade gap for the products subject to a standard or reduced VAT rate are shown in Table 2 (the table includes only the product codes for which 15

the VAT rate can be identified, thus excluding the mixed categories). These observations present the sample used in the empirical analysis afterwards. The mean and the median trade gap for the products with a standard VAT rate are negative only for Latvia and Lithuania. Considering the products with a reduced VAT rate, the median trade gap is negative for four countries, namely Estonia, Hungary, Lithuania and the Slovak Republic. As Table 2 demonstrates, the trade gap (mean and median) is substantially higher for Estonia, the Slovak Republic and Slovenia compared to the other countries in the sample. In the majority of countries the trade gap is on average smaller for the products taxed with a reduced VAT rate. This observation supports the hypothesis that a higher VAT rate is associated with a higher trade gap. The last column of Table 2 shows the difference in the mean trade gaps and it is, indeed, positive and significant for Estonia, Hungary, the Slovak Republic and Slovenia, meaning that the trade gap of the products taxed with the standard VAT rate is significantly higher on average than the one of the product categories taxed with the reduced VAT rate. The only exception is Lithuania, where the difference in the mean values goes in the opposite direction. Table 3 shows the descriptive statistics of the trade gap per EU-7 country for the two types of product codes, those with identified VAT rate and the mixed categories, which are excluded from the empirical analysis. Considering the total number of observations, it is obvious that the mixed categories are only a small share of all product codes. Similar to the descriptive statistics of the trade gap for products taxed with the standard VAT rate (Table 2) are the mean and the median of the trade gap in the sample without the mixed product groups negative only for Latvia and Lithuania. The highest mean trade gap for both types of product groups is again reported for Estonia, the Slovak Republic and Slovenia; the median values are lower but still high, especially for Estonia and the Slovak Republic. The mixed categories offer potentially more opportunities to evade VAT since they include both products with a standard and reduced VAT rate but with the same six-digit product code. The last column presents the difference in the means of the trade gap between the two types of product categories. If the mixed product codes are more prone to VAT evasion, this would mean that the difference in the means should be negative. The evidence is rather mixed; the difference is negative and highly significant in the cases of Estonia and the Slovak Republic (the countries with the highest trade gaps) but positive and significant for the Czech Republic and Hungary. It does not seem that there is any clear pattern along this differentiation of product groups, at least on the basis of the descriptive statistics.

5.2

Baseline results

Table 4 presents the baseline results of estimating Equation 2. The first panel of the table includes only year fixed effects, the second one adds also partner country fixed effects and the third panel contains all three types of fixed effects (year, partner country and product at the four-digit level). The estimates respond to the inclusion of the partner country fixed effect, which points towards a systematic correlation between partner country characteristics and the trade gap. Therefore, the first panel will not be reviewed further and the discussion will concentrate on the second panel. The point estimates for Estonia and the Slovak Republic show that a one-percentage-point increase in the VAT rate is associated with more than 3% increase in the trade gap, for the Slovak Republic 16

17

0.045 0.238 0.050 -0.013 -0.056 0.131 0.092

Czech Republic 0.136 Estonia 0.383 Hungary 0.060 Latvia -0.033 Lithuania -0.153 Slovak Republic 0.311 Slovenia 0.249

1.956 1.987 1.476 1.666 1.650 2.208 1.925

Standard Deviation 135,894 92,791 118,061 91,562 104,826 102,497 102,605

Observations 0.147 -0.045 -0.053 0.006 0.017 0.068 0.182

Mean (2) 0.019 -0.007 -0.013 0.021 -0.041 -0.002 0.056

Median 1.943 2.115 1.361 1.807 1.648 2.418 1.882

19,234 1,818 5,790 1,601 2,283 829 13,272

-0.009 0.428*** 0.113*** -0.039 -0.170*** 0.243*** 0.067***

Standard Observations Difference Deviation Mean(1) - Mean(2)

Reduced rate

Trade gap by products taxed with a standard and reduced VAT rate (only product with identified VAT rates considered). Mean and median tariff values are calculated for each country and year. Test for equality of means: *** p < 0.01, ** p < 0.05, * p < 0.1.

Median

Mean (1)

Country

Standard rate

Table 2: Descriptive statistics.

18

0.042 0.232 0.046 -0.012 -0.055 0.131 0.088

Czech Republic 0.137 Estonia 0.375 Hungary 0.055 Latvia -0.032 Lithuania -0.150 Slovak Republic 0.309 Slovenia 0.242

1.955 1.990 1.471 1.669 1.650 2.210 1.920

Standard Deviation 155,128 94,609 123,588 93,163 107,109 103,326 115,877

Observations 0.075 0.467 -0.072 -0.069 -0.140 0.833 0.273

Mean (2) 0.039 0.311 -0.017 -0.025 -0.064 0.530 0.128

Median 1.878 2.025 1.380 1.752 1.647 2.259 2.011

6,201 4,533 5,439 4,193 6,733 149 6,182

0.062*** -0.092*** 0.127*** 0.037 -0.010 -0.524*** -0.031

Standard Observations Difference Deviation Mean(1) - Mean(2)

Mixed=1

Trade gap by product categories; product codes with identified VAT rates are presented in the first four columns (Mixed=0), those with mixed VAT rates in columns 6-9 (Mixed=1). Mean and median tariff values are calculated for each country and year. Test for equality of means: *** p < 0.01, ** p < 0.05, * p < 0.1.

Median

Mean (1)

Country

Mixed=0

Table 3: Descriptive statistics (“mixed” vs. “non-mixed” product categories).

the increase is almost 4%. For Slovenia, the response in the trade gap is of smaller magnitude, but also highly significant. The results presented in Table 4 are consistent with the descriptive statistics from Table 2, which show that Estonia, the Slovak Republic and Slovenia are the countries with the highest trade gaps and thus potentially suffer most from VAT evasion. Lithuania is an exception, where the correlation between the VAT rate and the trade gap is negative and significant, meaning that an increase in the VAT rate is associated with a decrease in the trade gap. Table 4: Trade gap and VAT rate. (1)

(2)

(3)

(4)

(5)

(6)

(7)

Year FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

-0.001 (0.002)

0.034*** (0.009)

0.004 (0.003)

-0.003 (0.007)

-0.013*** (0.004)

0.042*** (0.012)

0.006** (0.003)

Observations Adj. R2

155,128 0.001

94,609 0.002

123,588 0.007

93,163 0.004

107,109 0.001

103,326 0.004

115,877 0.000

Year FE and Partner country FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

0.002 (0.002)

0.033*** (0.009)

0.003 (0.003)

-0.001 (0.007)

-0.012*** (0.004)

0.038*** (0.012)

0.009*** (0.003)

Observations Adj. R2

155,128 0.017

94,609 0.015

123,588 0.016

93,163 0.012

107,109 0.010

103,326 0.023

115,877 0.023

Year FE, Partner country FE and Product FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

-0.004 (0.010)

0.033** (0.016)

-0.004 (0.003)

-0.019 (0.033)

-0.006 (0.008)

0.038*** (0.013)

-0.017 (0.018)

Observations Adj. R2

155,128 0.076

94,609 0.068

123,588 0.049

93,163 0.035

107,109 0.034

103,326 0.095

115,877 0.093

Note: *** p < 0.01, ** p < 0.05, * p < 0.1. First panel: All regressions include year fixed effects and a constant. Second panel: All regressions include year and partner country fixed effects and a constant. Third panel: All regressions include year, partner country and four-digit product fixed effects, and a constant. Standard errors, clustered at the six-digit product level, in parentheses.

As previously discussed, the theoretical and empirical literature on the topic of tax evasion show that the relationship between the level of tax rate and the rate of evasion is ambigious. Agha and Haughton (1996) estimate a decrease of 2.7 percentage points 19

in the VAT compliance rate as a response to a one-percentage-point increase in the VAT rate and Christie and Holzner (2006) measure a decrease of 0.3 percentage points. However, Yitzhaki (1974) and Reckon LLP (2009) provide theoretical and empirical evidence respectively that an increase in the tax rate is associated with a decrease in the VAT non-compliance rate. The findings of Table 4 are therefore in line with the previous studies since they show that across the EU-7 countries the direction of the correlation between the trade gap and the VAT rate is also ambigious. In addition, the magnitude of the effect of the VAT rate on the trade gap varies substantially across the states; a one-percentage-point increase in the VAT rate is associated with 0.9% to 3.8% increase in the trade gap. Thus, the results reveal the importance of country heterogeneity in the relationship between the VAT rate and the trade gap. The last panel of Table 4 presents the results when four-digit product fixed effects are included. As expected, the significance level of the estimates is lower since the variation of the VAT rate over time is limited at the six-digit product level. For Estonia and the Slovak Republic the correlation between the VAT rate and the trade gap remains in the same magnitude as before. For Slovenia and Lithuania the relationship is no longer significant. A possible explanation for Slovenia is the fact that the VAT rates do not vary over time in the sample period so the variation in the VAT variable comes only from differences in the VAT rates across products. Thus, once controlling for product fixed effects at the four-digit level, the variation of the VAT rate is very limited which leads most likely to the change in the significance level.

5.3

Robustness checks

Equation 2 is also estimated weighted by the mean value of exports of each partner country-product code pair controlling for the relation between the missing trade and the export value, which is assumed to be the “true” value of the trade flow. Table 5 presents the results. Overall, the outcomes are very similar to the baseline results in Table 4. The only difference is the result for Hungary, for which the correlation between the VAT rate and the trade gap is now significant at the 5%-level except when product fixed effects are included. The estimate for Hungary is similar in its magnitude to the outcome for Slovenia, a one-percentage-point increase in the VAT rate is associated with around 1% increase in the trade gap, and thus much lower than in the cases of Estonia and the Slovak Republic. Estonia and the Slovak Republic remain the only countries, for which the coefficient on the VAT rate is still significant when adding product fixed effects. Instead of a variable, capturing the level of the VAT rate, Equation 2 might be estimated including a dummy variable, which indicates whether the product code belongs to the goods for which a standard VAT rate applies. Compared to a VAT variable in levels, the dummy variable shows the difference in the trade gap between the products which are taxed with the standard VAT rate and those taxed with the reduced rate independent of the exact level of the VAT rates. Table 6 shows the results. The findings are consistent with the outcomes of Table 4. Products subject to the standard VAT rate have a significantly higher trade gap than the goods taxed with the reduced VAT rate in Estonia, the Slovak Republic and Slovenia. The inverse relationship applies again for Lithuania. 20

Table 5: Trade gap and VAT rate, weighted by exports. (1)

(2)

(3)

(4)

(5)

(6)

(7)

Year FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

0.000 (0.002)

0.035*** (0.009)

0.007** (0.003)

-0.004 (0.007)

-0.011** (0.004)

0.041*** (0.013)

0.007** (0.003)

Observations Adj. R2

155,116 0.001

94,606 0.002

123,588 0.007

93,154 0.004

107,106 0.001

103,309 0.005

115,863 0.000

Year FE and Partner country FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

0.002 (0.002)

0.034*** (0.009)

0.006** (0.003)

-0.001 (0.007)

-0.010** (0.004)

0.038*** (0.013)

0.010*** (0.003)

Observations Adj. R2

155,116 0.018

94,606 0.015

123,588 0.017

93,154 0.012

107,106 0.009

103,309 0.025

115,863 0.024

Year FE, Partner country FE and Product FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

-0.000 (0.009)

0.037** (0.018)

-0.001 (0.003)

-0.018 (0.037)

-0.006 (0.008)

0.035*** (0.013)

-0.018 (0.019)

Observations Adj. R2

155,116 0.083

94,606 0.073

123,588 0.050

93,154 0.037

107,106 0.034

103,309 0.101

115,863 0.101

Note: *** p < 0.01, ** p < 0.05, * p < 0.1. First panel: All regressions include year fixed effects and a constant. Second panel: All regressions include year and partner country fixed effects and a constant. Third panel: All regressions include year, partner country and four-digit product fixed effects, and a constant. Standard errors, clustered at the six-digit product level, in parentheses.

21

Table 6: Trade gap and VAT rate, dummy variable for the standard VAT rate. (1)

(2)

(3)

(4)

(5)

(6)

(7)

Year FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

Standard V AT

-0.009 (0.031)

0.427*** (0.113)

0.037 (0.027)

-0.040 (0.086)

-0.174*** (0.058)

0.377*** (0.109)

0.068** (0.034)

Observations Adj. R2

155,128 0.001

94,609 0.002

123,588 0.007

93,163 0.004

107,109 0.001

103,326 0.004

115,877 0.000

Year FE and Partner country FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

Standard V AT

0.022 (0.030)

0.411*** (0.111)

0.027 (0.027)

-0.014 (0.085)

-0.163*** (0.058)

0.346*** (0.109)

0.103*** (0.034)

Observations Adj. R2

155,128 0.017

94,609 0.015

123,588 0.016

93,163 0.012

107,109 0.010

103,326 0.023

115,877 0.023

Year FE, Partner country FE and Product FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

Standard V AT

-0.299 (0.399)

0.449** (0.207)

-0.053* (0.029)

-0.239 (0.426)

-0.107 (0.111)

0.340*** (0.117)

-0.194 (0.203)

Observations Adj. R2

155,128 0.076

94,609 0.068

123,588 0.049

93,163 0.035

107,109 0.034

103,326 0.095

115,877 0.093

Note: *** p < 0.01, ** p < 0.05, * p < 0.1. First panel: All regressions include year fixed effects and a constant. Second panel: All regressions include year and partner country fixed effects and a constant. Third panel: All regressions include year, partner country and four-digit product fixed effects, and a constant. Standard errors, clustered at the six-digit product level, in parentheses.

5.4

Controlling for outliers

In all previous tables the estimates for Hungary are restricted only to trade flows with a value higher than 1,000 US dollars. Table 7 presents the results if this sample restriction is applied to all countries. The estimates are very similar in magnitude and significance to the baseline results. Table 8 shows the results of estimating Equation 2 only for trade flows with a higher value than 50,000 US dollars. The probability of measurement error is likely to be higher for trade flows of a smaller value and also the statistical offices control potentially more strictly trade flows with a higher value. Excluding all trade flows below 22

the threshold of 50,000 US dollars leads to dropping out a substantial share of observations, more than 50% of the observations in the sample are left out. Table 8 shows that there is a positive and significant correlation between the VAT rate and the trade gap in five of the seven countries in the sample (Czech Republic, Estonia, Hungary, Slovak Republic and Slovenia). The significant relationship remains also when partner country fixed effects are added. Similar to the previous results, the magnitude is the highest for Estonia and the Slovak Republic, although lower than in Table 4. Again, when adding four-digit product fixed effects the significance level decreases, only the estimates for Estonia and Hungary remain significant. Since the variation in the VAT rates is rather low over time, this result is not surprising. Table 7: Trade gap and VAT rate (Trade flows with value higher than 1,000 USD). (1)

(2)

(3)

(4)

(5)

(6)

(7)

Year FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

0.001 (0.002)

0.030*** (0.008)

0.004 (0.003)

-0.005 (0.006)

-0.009** (0.004)

0.033*** (0.012)

0.009*** (0.002)

Observations Adj. R2

137,015 0.001

78,160 0.002

123,588 0.007

77,519 0.001

91,167 0.001

87,823 0.006

98,513 0.001

Year FE and Partner country FE included Czech

Estonia

Hungary

Latvia

Lithuania

Slovak

Slovenia

V AT

0.002 (0.002)

0.028*** (0.008)

0.003 (0.003)

-0.004 (0.006)

-0.008** (0.004)

0.029** (0.012)

0.011*** (0.002)

Observations Adj. R2

137,015 0.019

78,160 0.019

123,588 0.016

77,519 0.009

91,167 0.010

87,823 0.027

98,513 0.025

Year FE, Partner country FE and Product FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

-0.012 (0.008)

0.042*** (0.013)

-0.004 (0.003)

-0.015 (0.025)

-0.006 (0.008)

0.024** (0.011)

-0.016 (0.016)

Observations Adj. R2

137,015 0.086

78,160 0.081

123,588 0.049

77,519 0.037

91,167 0.038

87,823 0.094

98,513 0.104

Note: *** p < 0.01, ** p < 0.05, * p < 0.1. First panel: All regressions include year fixed effects and a constant. Second panel: All regressions include year and partner country fixed effects and a constant. Third panel: All regressions include year, partner country and four-digit product fixed effects, and a constant. Standard errors, clustered at the six-digit product level, in parentheses.

In order to control for outliers, Equation 2 is also estimated dropping the top and 23

Table 8: Trade gap and VAT rate, trade flows with a value higher than 50,000USD. (1)

(2)

(3)

(4)

(5)

(6)

(7)

Lithuania

Slovak Republic

Slovenia

0.003 (0.003)

0.027** (0.011)

0.009*** (0.002)

35,096 0.000

41,184 0.005

45,566 0.002

Year FE included

V AT

Observations Adj. R2

Czech Republic

Estonia

Hungary

0.005*** (0.002)

0.026*** (0.007)

0.008*** 0.002 (0.002) (0.006)

76,483 0.002

28,297 0.005

67,317 0.007

Latvia

26,766 0.001

Year FE and Partner country FE included

V AT

Observations Adj. R2

Czech Republic

Estonia

Hungary

0.006*** (0.002)

0.026*** (0.007)

0.007*** 0.003 (0.002) (0.006)

76,483 0.024

28,297 0.020

67,317 0.018

Latvia

26,766 0.009

Lithuania

Slovak Republic

Slovenia

0.003 (0.003)

0.023** (0.011)

0.010*** (0.002)

35,096 0.010

41,184 0.027

45,566 0.032

Year FE, Partner country FE and Product FE included Czech Republic

Estonia

Hungary

V AT

0.003 (0.005)

0.038*** (0.007)

0.008*** 0.001 (0.003) (0.017)

Observations Adj. R2

76,483 0.100

28,297 0.092

67,317 0.059

Latvia

26,766 0.045

Lithuania

Slovak Republic

Slovenia

-0.007 (0.006)

0.001 (0.009)

0.004 (0.016)

35,096 0.054

41,184 0.104

45,566 0.122

Note: *** p < 0.01, ** p < 0.05, * p < 0.1. First panel: All regressions include year fixed effects and a constant. Second panel: All regressions include year and partner country fixed effects and a constant. Third panel: All regressions include year, partner country and four-digit product fixed effects, and a constant. Standard errors, clustered at the six-digit product level, in parentheses.

24

bottom 1% or 10% of trade gap observations. Table 9 and Table 10 show the results without the 1% or 10% outliers of the trade gap, respectively. There is no difference between the findings of Table 9 and the baseline results, except that dropping the 1% outliers in the trade gap observations leads to a slightly lower magnitude of the significant coefficients. The results of Table 10 are very similar to the ones of Table 8 implementing the threshold of 50,000 US dollars, except that the share of the dropped observations is much lower. The coefficient on the VAT rate is significant and positive for the Czech Republic, Estonia, the Slovak Republic and Slovenia when year and partner country fixed effects are included (for Hungary only in the first panel). In the specification with product fixed effects, the relationship between the VAT rate and the trade gap is significant for Estonia, Hungary and the Slovak Republic. The magnitude of the significant coefficients is again lower, particularly for Estonia and the Slovak Republic, although they still remain the highest among the other countries in the sample. Some product groups might be regarded as outlier cases because additional taxes such as an excise tax or confidential restrictions apply for them. Excluding HS24 (tobacco), HS26 (ores), HS27 (mineral fuels and oils), HS84 (nuclear reactors), HS88 (aircraft and spacecraft), HS89 (ships, boats and floating structures) and HS93 (arms and ammunition) based on these concerns does not change the baseline results. Table 11 summarizes the results of the empirical analysis displaying the sign and the significance level of the coefficient on the VAT rate variable for each EU-7 country estimated in each of the empirical specifications with year and partner country fixed effects included. The estimates are robust to changes in the specification and controlling for outliers for three countries, namely Estonia, the Slovak Republic and Slovenia, and show a positive and highly significant correlation between the VAT rate and the trade gap. For the Czech Republic and Hungary the relationship is also positive, however, only in two regressions it is significant. For the Czech Republic, an increase in the VAT rate is associated with an increase in the trade gap only when the outliers in the trade gap data under the broader definitions (either excluding trade flows below 50,000 USD or 10% of the top and bottom trade gap observations) are left out from the sample. The results for Latvia are mostly negative, though never significant. Lithuania presents an outlier case since in the majority of the specifications the coefficient on the VAT rate points towards a negative and siginicant correlation between the VAT rate and the trade gap. Similar to the results of the empirical studies on the VAT gap and to the estimations on the relationship between the trade gap and the tariff rate the correlation between the trade gap and the VAT rate is not straightforward. Even though the majority of the results in this study point towards a positive relationship between the trade gap and the VAT rate, the empirical analysis shows that the inverse relationship might also apply. Despite the evidence that VAT evasion in intra-Community trade is prevalent in many of the EU-7 countries, the direction and the magnitude of the effect of the VAT rate on the trade gap seem to be dependent on country characteristics.

25

Table 9: Trade gap and VAT rate, controlling for outliers (without top and bottom 1% of trade gap observations). (1)

(2)

(3)

(4)

(5)

(6)

(7)

Year FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

0.001 (0.002)

0.031*** (0.008)

0.001 (0.003)

-0.005 (0.006)

-0.009** (0.004)

0.034*** (0.011)

0.006** (0.003)

Observations Adj. R2

152,024 0.001

92,715 0.001

126,287 0.006

91,299 0.004

104,965 0.001

101,258 0.005

113,559 0.000

Year FE and Partner country FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

0.002 (0.002)

0.030*** (0.008)

0.001 (0.003)

-0.003 (0.006)

-0.009** (0.004)

0.031*** (0.011)

0.008*** (0.003)

Observations Adj. R2

152,024 0.018

92,715 0.015

126,287 0.011

91,299 0.013

104,965 0.010

101,258 0.025

113,559 0.024

Year FE, Partner country FE and Product FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

-0.009 (0.008)

0.032*** (0.012)

-0.004 (0.003)

-0.013 (0.026)

-0.006 (0.008)

0.025*** (0.010)

-0.009 (0.014)

Observations Adj. R2

152,024 0.079

92,715 0.068

121,949 0.048

91,299 0.035

104,965 0.032

101,258 0.096

113,559 0.098

Note: *** p < 0.01, ** p < 0.05, * p < 0.1. First panel: All regressions include year fixed effects and a constant. Second panel: All regressions include year and partner country fixed effects and a constant. Third panel: All regressions include year, partner country and four-digit product fixed effects, and a constant. Standard errors, clustered at the six-digit product level, in parentheses.

26

Table 10: Trade gap and VAT rate, controlling for outliers (without top and bottom 10% of trade gap observations). (1)

(2)

(3)

(4)

(5)

(6)

(7)

Year FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

0.004*** (0.001)

0.015*** (0.004)

0.003* (0.001)

-0.003 (0.003)

0.001 (0.002)

0.022*** (0.008)

0.006*** (0.001)

Observations Adj. R2

124,102 0.001

75,687 0.001

98,870 0.012

74,529 0.004

85,687 0.002

82,660 0.006

92,701 0.001

Year FE and Partner country FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

0.005*** (0.001)

0.015*** (0.004)

0.002 (0.001)

-0.003 (0.003)

0.001 (0.002)

0.020** (0.008)

0.007*** (0.001)

Observations Adj. R2

124,102 0.017

75,687 0.013

98,870 0.020

74,529 0.013

85,687 0.011

82,660 0.022

92,701 0.020

Year FE, Partner country FE and Product FE included Czech Republic

Estonia

Hungary

Latvia

Lithuania

Slovak Republic

Slovenia

V AT

-0.005 (0.004)

0.031*** (0.007)

0.003* (0.002)

-0.007 (0.011)

-0.002 (0.003)

0.017*** (0.006)

-0.009 (0.006)

Observations Adj. R2

124,102 0.074

75,687 0.068

98,870 0.042

74,529 0.027

85,687 0.028

82,660 0.082

92,701 0.084

Note: *** p < 0.01, ** p < 0.05, * p < 0.1. First panel: All regressions include year fixed effects and a constant. Second panel: All regressions include year and partner country fixed effects and a constant. Third panel: All regressions include year, partner country and four-digit product fixed effects, and a constant. Standard errors, clustered at the six-digit product level, in parentheses.

27

28

+ + + + +*** + +***

+*** +*** +*** +*** +*** +*** +***

(4)

+ +** + + +*** + +

+ -

Hungary Latvia

(3)

-*** -** -*** -** + -** +

Lithuania

(5)

(7)

+*** +*** +*** +** +** +*** +**

+*** +*** +*** +*** +*** +*** +***

Slovak Slovenia Republic

(6)

Note: *** p < 0.01, ** p < 0.05, * p < 0.1. Summary of the sign and the significance level of the coefficient on V AT based on all previous estimations including year and partner country fixed effects.

Baseline Weighted by exports VAT dummy variable Trade above 1,000 USD Trade above 50,000 USD 1% outliers of trade gap dropped 10% outliers of trade gap dropped

(2)

Czech Estonia Republic

(1)

Table 11: Summary of the results with year and partner country fixed effects.

6

Conclusion

Recent estimations and studies have suggested that VAT fraud has been increasing since the introduction of the EU Single Market in 1993. Intra-EU trade flows, in particular, have become more vulnerable to VAT fraud schemes since the change in the procedure of VAT collection in intra-Community trade. Policy-makers and the national tax administration are aware of this issue and the threat that it imposes on the functioning of the Single Market. However, until now, the empirical evidence has remained very scarce with regard to the relationship between the level of the VAT rate and the asymmetries in the bilateral trade flows between EU members. This is the first study to use an empirical approach employed in the literature on tariff evasion and to estimate the responsiveness of the missing trade between EU member states to changes in the VAT rate. The study shows that there is a robust positive correlation between the level of the VAT rate and the trade gap for Estonia, the Slovak Republic and Slovenia. The relationship is strongly pronounced for Estonia and the Slovak Republic. Depending on the specification, a one-percentage-point increase in the VAT rate in these two countries is associated with more than 3% increase in the trade gap. When restricting the sample to trade flows with a higher value than 50,000 US dollars or dropping the top and bottom 10% outliers in the trade gap data, the correlation is positive and significant in five countries (Czech Republic, Estonia, Hungary, Slovak Republic and Slovenia), potentially due to better quality of the trade data and more correct estimates. This systematic correlation between the level of the VAT rate and the trade gap suggests that VAT evasion in intra-EU trade flows is a serious issue to tackle in many of the EU-7 countries. The empirical evidence of the study indicates that the VAT rate, which has been so far neglected in the literature on the missing trade, is a relevant determinant of the mismatch in mirror trade statistics between the EU-15 and the EU-7 countries. In addition, the results of the study provide evidence that the correlation between the VAT rate and the trade gap varies substantially across the EU-7 countries so it is important to take into account the specific institutional and political environment of each country. This finding points towards the fact how difficult it might be to overcome the differences in the national VAT laws and to agree upon a common VAT legislation for all EU members in the future. However, it is potentially because of these differences in the national VAT legislation that the VAT fraud schemes exist to such an extent in intraCommunity trade flows. Better coordination and exchange of data across the national tax authorities might be effective, at least as a first step, to limit the scope of VAT evasion. Better availability of detailed data on VAT rates and their implementation would allow for more extensive future research on the topic of VAT fraud.

29

References Agha, A., and J. Haughton (1996): “Designing Vat Systems: Some Efficiency Considerations,” The Review of Economics and Statistics, 78(2), 303–308. Allingham, M. G., and A. Sandmo (1972): “Income Tax Evasion: A Theoretical Analysis,” Journal of Public Economics, 1(3-4), 323–338. Andreoni, J., B. Erard, and J. Feinstein (1998): “Tax Compliance,” Journal of Economic Literature, 36(2), 818–860. Baldwin, R. E. (2007): “EU VAT Fraud,” http://www.voxeu.org/article/ vat-fraud-part-1, [Online; accessed 02-May-2013]. Borselli, F. (2008): “Pragmatic Policies to Tackle VAT Fraud in the European Union,” International VAT Monitor, (5), 333–342. Christie, E., and M. Holzner (2006): “What Explains Tax Evasion? An Empirical Assessment based on European Data,” WIIW Working Paper 40. Czech Republic Government (2004): “Value Added Tax Act,” http://www. czechlegislation.com/en/235-2004-sb, [Online; accessed 02-May-2013]. De Wulf, L. (1981): “Statistical Analysis of Under- and Overinvoicing of Imports,” Journal of Development Economics, 8(3), 303–323. Estonia Government (2010): “Value Added Tax Act,” http://www.legaltext.ee/ et/andmebaas/paraframe.asp?loc=text&lk=et&sk=en&dok=X70060K6.htm&query= k\%C3\%A4ibemaks&tyyp=X&ptyyp=RT&pg=1&fr=no, [Online; accessed 02-May-2013]. EU Commission (2012): “An Action Plan to Strengthen the Fight against Tax Fraud and Tax Evasion,” http://ec.europa.eu/taxation_customs/resources/ documents/taxation/tax_fraud_evasion/com_2012_722_en.pdf, [Online; accessed 02-May-2013]. (2014): “VAT Rates Applied in the Member States of the European Union,” http://ec.europa.eu/taxation_customs/resources/documents/ taxation/vat/how_vat_works/rates/vat_rates_en.pdf, [Online; accessed 14January-2014]. European Commission (2004): “Commission Regulation No 1982/2004 implementing Regulation (EC) No 638/200 of the European Parliament and of the Council on Community statistics relating to the trading of goods between Member States and repealing Commission Regulations (EC) No 1901/2000 and (EEC) No 3590/92,” http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L: 2004:343:0003:0019:EN:PDF, [Online; accessed 02-May-2013]. (2013): “The Missing Part,” http://ec.europa.eu/taxation_customs/ taxation/tax_fraud_evasion/missing-part_en.htm, [Online; accessed 14-January2014].

30

European Council and European Parlament (2004): “Regulation No 638/2004 of the European Parliament and of the Council on Community statistics relating to the trading of goods between Member States and repealing Council Regulation (EEC) No 3330/91,” http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L: 2004:102:0001:0008:EN:PDF, [Online; accessed 02-May-2013]. Fisman, R., and S.-J. Wei (2004): “Tax Rates and Tax Evasion: Evidence from “Missing Imports”’ in China,” Journal of Political Economy, 112(2), 471–496. Gebauer, A., C. W. Nam, and R. Parsche (2005): “VAT Evasion and Its Consequence for Macroeconomic Clearing in the EU,” FinanzArchiv, 61(4), 462–487. ¨rxhani, K. (2004): “Tax Evasion in Transition: Outcome of an Institutional Clash? Ge Testing Feige’s Conjecture in Albania,” European Economic Review, 48(4), 729–745. Gradeva, K. (2014): “Trade Liberalization and Evasion of Import Tariffs: Evidence from the Eastern EU Enlargement,” mimeo. Hanousek, J., and F. Palda (2008): “Tax Evasion Dynamics in the Czech Republic: First Evidence of an Evasional Kuznets Curve,” CERGE-EI Working Paper, Center for Economic Research and Graduate Education, Charles University 360. HM Revenue & Customs (2006): “Measuring Indirect Tax Losses - 2006,” Discussion paper. (2013): “Measuring Tax Gaps, 2013 Edition,” Discussion paper. Hungary Government (2004): “Act on Value Added Tax,” http://www.lexadin. nl/wlg/legis/nofr/eur/arch/hun/lxxiv1992.pdf, [Online; accessed 02-May-2013]. (2006): “Act on Value Added Tax,” http://www.ecovis.hu/angol/value_ added_tax.pdf, [Online; accessed 02-May-2013]. (2009): “Act on Value Added Tax,” http://www.icnl.org/research/ library/files/Hungary/vatact_ENG.pdf, [Online; accessed 02-May-2013]. Javorcik, B. S., and G. Narciso (2008): “Differentiated Products and Evasion of Import Tariffs,” Journal of International Economics, 76(2), 208–222. Javorcik, B. S., and G. Narciso (2013): “WTO Accession and Tariff Evasion,” mimeo. Jean, S., and C. Mitaritonna (2010): “Determinants and Pervasiveness of the Evasion of Customs Duties,” AgFoodTrade Working Paper 2010-05. Keen, M., and S. Smith (2006): “VAT Fraud and Evasion: What Do We Know, and What Can be Done?,” National Tax Journal, 59(4), 861–887. Koufen, S. (2001): “Spiegelvergleich der deutschen und spanischen Intrahandelsergebnisse,” Wirtschaft und Statistik, (1), 46–50. Krockow, A. (2007): “Vereinfachung der Intrahandelsstatistik,” Wirtschaft und Statistik, (7), 670–678. 31

Latvia Government (2010): “Law on Value Added Tax,” http://www.fm.gov.lv/ en/s/taxes/, [Online; accessed 02-May-2013]. Lithuania Government (2006): “Law on Value Added Tax,” http://www3.lrs.lt/ pls/inter3/dokpaieska.showdoc_l?p_id=282353, [Online; accessed 02-May-2013]. (2009): “Law on Value Added Tax,” https://www.vmi.lt/cms/documents/ 10162/7977078/The+Law+on+VAT/c5ec1bf1-564a-42d9-a5c6-e3771ad611b7, [Online; accessed 02-May-2013]. Loschky, A. (2006): “Asymmetrien in der Außenhandelsstatistik,” Wirtschaft und Statistik, (3), 257–263. Mishra, P., A. Subramanian, and P. Topalova (2008): “Tariffs, Enforcement, and Customs Evasion: Evidence from India,” Journal of Public Economics, 92(10-11), 1907–1925. Reckon LLP (2009): “Study to Quantify and Analyse the VAT Gap in the EU-25 Member States,” Discussion paper. Ruffles, D., G. Tily, D. Caplan, and S. Tudor (2003): “VAT Missing Trader Intra-Community Fraud: The Effect on Balance of Payments Statistics and UK National Accounts,” Economic Trends, (597), 58–70. Slemrod, J. (2007): “Cheating Ourselves: The Economics of Tax Evasion,” The Journal of Economic Perspectives, 21(1), 25–48. Slemrod, J., and S. Yitzhaki (2002): “Tax Avoidance, Evasion, and Administration,” in Handbook of Public Economics, ed. by A. J. Auerbach, and M. S. Feldstein, vol. 3, pp. 1423–1470. Amsterdam: Elsevier B.V. Slovak Republic Government (2008): “Act on Value Added Tax,” http://www. finance.gov.sk/en/Default.aspx?CatID=34, [Online; accessed 02-May-2013]. Slovenia Government (2002): “Value Added Tax Act,” http://www.mf.gov.si/ fileadmin/mf.gov.si/pageuploads/Davki_in_carine/Sprejeti_predpisi/ Zakon_o_davku_na_dodano_vrednost/VALUE_ADDED_TAX_ACT__ZDDV-1__-_ UNOFFICIAL_TRANSLATION_1_.pdf, [Online; accessed 02-May-2013]. The Council of the European Union (2006): “Council Directive 2006/112/EC on the Common System of Value Added Tax,” http://eur-lex.europa.eu/ LexUriServ/LexUriServ.do?uri=OJ:L:2006:347:0001:0118:en:PDF, [Online; accessed 02-May-2013]. Torgler, B. (2004): “Tax Morale, Trust and Corruption: Empirical Evidence from Transition Countries,” CREMA Working Paper, Center for Research in Economics, Management and the Arts 2004-05. (2011): “Tax Morale, Eastern Europe and European Enlargement,” Communist and Post-Communist Studies, 45(1-2), 11–25.

32

Torgler, B., I. C. Demir, A. Macintyre, and M. Schaffner (2008): “Causes and Consequences of Tax Morale: An Empirical Investigation,” Economic Analysis & Policy, 38(2), 313–339. Woon Nam, C., R. Parsche, and B. Schaden (2001): “Measurement of Value Added Tax Evasion in Selected EU Countries on the Basis of National Accounts Data,” CESifo Working Paper 431. Yeats, A. J. (1995): “Are Partner-Country Statistics Useful for Estimating “Missing” Trade Data,” World Bank Policy Research Working Paper 1501. Yitzhaki, S. (1974): “A Note on Income Tax Evasion: A Theoretical Analysis,” Journal of Public Economics, 3(2), 201–202.

33