Shedding light inside the black box of implementation: Tax crimes as a predicate crime for money laundering

Even perfect transposition of EU Directives does not necessarily translate into homogeneous rules or application of rules across the European Union. Europeanization literature focused on the formal transposition of EU Directives. Newer studies suggest looking into the black box of how this translates into law in action. The 4th Anti-Money Laundering Directive incorporated taxes as a predicate crime for money laundering. We analyze how and why this Directive has been implemented so differently across EU countries both in the books and in action through a novel dataset. We ﬁ nd that country characteristics can explain formal transposition patterns and in ﬂ uence the domestic adaptation of regulation as well as how practitioners, the second front line of implementation, use these rules in action. We ﬁ nd that corruption, government effectiveness, regulatory quality, tax morale, and tax administrative capacity are important factors to explain lingering differences in the books and in action among EU Member States.


Introduction
In 2012 the Financial Action Task Force (FATF), an intergovernmental organization that sets international standards to fight money laundering, recommended including tax crimes as a predicate offense 1 for money laundering. The European Parliament followed the FATF and in 2015 passed the 4th Anti-Money Laundering Directive 2 ("4th AMLD," Directive (EU) 2015/849) 3 that incorporates tax crimes 4 as a predicate crime for money laundering into EU regulation. EU Member States had until 2017 to transpose the 4th AMLD into their legislation. However, tax crimes are included without a concrete definition, leaving each jurisdiction to define or redefine what they consider a tax crime. This means discrepancies can arise in the practical implementation of said principle. 5 An example of the striking differences across countries is the following. We asked respondents from different countries to analyze the same case, inspired by a former Bayern Munich football player who evades 5 million euros in taxes through a Swiss bank account. Responses diverged so much that this superstar's situation would have been entirely different in neighboring countries, Belgium and the Netherlands. In Belgium, the former player would not face jail time, while in the Netherlands, he would face both tax crime and money laundering charges and go to jail for at least eight years. In Greece, this same case would have been only subject to administrative charges due to "bureaucratic reasons," as our respondent highlights. What factors can explain the cross-country divergencies of how such a case is treated differently after implementing the same EU Directive?
This paper seeks to contribute to the ongoing quest to understand how the Europeanization and domestication of policy are complementary forces (Thomann & Sager 2017) that interplay in the practical implementation of policy both in the books and in action. We choose the implementation of tax crimes as a predicate crime for money laundering as an example of a single issue tackled by a European Directive. Single issues can be a more relevant unit of analysis as Directives regulate diverse issues that can often be transposed into different national legislations (Toshkov 2010;Thomann 2015). This particular single issue is the result of European regional governance and global recommendations. The global influences are evident because, like the FATF, the EU did not incorporate a concrete definition of tax crimes. The decision to designate which tax offenses are defined as crimes was left to each member state, potentially leading to a divergent (practical) implementation of this principle and other rules of the 4th AMLD (Mitsilegas & Vavoula 2016;Unger 2017). The absence of a unified definition of tax crimes in the EU acquis communautaire results from a lack of consensus among Member States (Thirion & Scherrer 2017;Turksen & Abukari 2020). However, it is relevant to highlight that the parliament noted that "Agreeing on a definition of tax crimes is an important step in detecting those crimes" (Council of the European Union 2014, p. 11).
Directives by construction leave discretion to the Member States as to how regulations are transposed into legislation. As a result, traditional Europeanization research has focused on whether European Directives are or not transposed into domestic laws, focusing mainly on dates and rates of transposition and the legal compliance of minimum standards (conformance implementation) set by European Directives across different countries (Toshkov 2010;Treib 2014). However, this neglects the degree to which policies and recommendations can be modified and the domestic variations, customization, or domestication process (Bugdahn 2005) that can occur when implementing supranational regulation. Moreover, it neglects that while a Directive can be "perfectly" transposed into national legislation, it might still be practically implemented and applied differently and not function similarly in practice across countries. For example, although by 2005, there was a 98 percent success rate on the transposition of EU Directives, this number does not tell how these Directives were transposed (Mastenbroek 2005). Finally, as Thomann and Zhelyazkova (2017) point out, "studying legal compliance without considering adaptations of EU policy to domestic circumstances provides an incomplete picture of EU implementation" (p. 1270).
However, a more comprehensive analysis of the law in the books still misses how regulation is interpreted, used, and applied in practice (law in action). Bureaucrats are implementing EU Directives and therefore act as a second front line. This can create significant legal ambiguities (Dörrenbächer 2017). There has not been enough research on how individual actors implement EU policy on the ground (Thomann & Sager 2017). How the practical implementation of EU Directives works out across the EU Member States is thus considered a black box or black hole in the study of EU regulatory compliance (Mastenbroek 2005;Versluis 2007; Thomann & Sager 2017).
We study the legal adaptations or differences in defining tax crimes across the European Union after implementing tax crimes as a predicate crime for money laundering and how this principle is used in practice. We build a unique dataset 6 that includes legal and practical information for the 28 EU Member States. 7 For the law in action, we conducted a survey where we asked experts in the second front line of implementation (e.g., lawyers, public prosecutors, and tax inspectors) how they would proceed in the prosecution and investigation of three sample cases related to tax crimes and money laundering. Our methodology allows us to compare the handling of the same case across all EU Member States and understand its legal basis. Furthermore, it allows us to understand to what extent the directive has impacted the actors that it was meant to benefit the most, such as prosecutors and investigators. We use this survey as an alternative to the traditional comparative criminal law analysis that often results in hundreds of pages to answer only one single question 8 for few countries.
For the law in the books, opposite to a binary yes/no variable approach (Schwarz 2011), we collected legal information, such as maximum and minimum prison penalties and prescription times for both crimes, to capture more than whether a crime is punishable or not by law. Knowing more than whether the country abides by the minimum standards of the Directive allows us to go beyond the compliance versus noncompliance debate that masks significant implementation variance (Bondarouk & Liefferink 2017). We complement our data with semistructured, in-depth interviews with selected investigators and prosecutors.
Our findings contribute to the understanding of how and why EU policy can entail divergent national outcomes. In general, we find that when zooming in the implementation of a single issue, countries fall back in their old ways when they have the leeway to do so. As Thomann and Zhelyazkova (2017) suggest, the customization process can show how Member States try to regain control. By falling back, we mean that country characteristics do not explain just transposition patterns, as previously found in the literature (Falkner et al. 2005;Toshkov 2010). Instead, country characteristics might also explain the domestication and customization that regulation goes through when implemented in a countries' legislation and how the actors that are part of the second front line of implementation use and apply this in action.
We put forward and test two hypotheses that can explain said differences. First, basing ourselves on implementation theory that suggests that bureaucratic capacity, regulatory styles, and administrative traditions can explain transposition patterns and noncompliance (König & Luetgert 2009;Börzel et al. ;Toshkov 2010;Zhelyazkova et al. 2016), we argue that government capacity and quality can explain differences in the performance implementation of tax crimes as a predicate crime for money laundering. We operationalize government capacity and quality through country features such as corruption, government effectiveness, and regulatory quality. Second, we use implementation theory complemented by theory on preferences and attitudes of Member States and their actors and apply it directly to the tax realm. Thus, we argue that cross-country differences can be explained by tax-specific authorities' characteristics and domestic actors' beliefs and preferences (Mastenbroek & Kaeding 2006) regarding taxation. We operationalize this through a tax profile concept, a composite of variables such as tax morale and tax administrative capacity.
We find that less corrupt countries that rank better in terms of regulatory quality and have more effective governments are less punitive in the books and tend to prosecute a case only for tax crimes rather than for money laundering. Furthermore, we also find that countries with high tax morale and less secretive tax courts, and high administrative capacity, give more discretionary power to judges. Finally, we find that countries with low tax administrative capacity and harmful tax structures tend to limit the prosecution of tax crime cases.
The rest of the article is structured as follows. Section 2 provides a framework that merges the global and European governance of money laundering with the literature on Europeanization and domestication of European Directives and provides the underpinnings behind our choice of explanatory variables. Section 3 details the data collection, variable operationalization, and methodology. Section 4 contains the analysis and results, and Section 5 the conclusions, limitations, and suggestions for further research.

Global influence and European implementation of money laundering directives
Money laundering has become subject to supranational governance on a global and regional scale. On a global level, the most recognizable form of soft multilateral law on money laundering started when the G7 and other guest countries initiated the FATF in 1989. The FATF established international standards by issuing a set of Forty Recommendations. Countries that did not commit to these standards were pressured to comply through blacklists (Muller et al. 2007;Unger & Ferweda 2008;Schwarz 2011). Given the FATF Recommendations' soft law nature, individual jurisdictions have the flexibility to adapt their legal framework to comply with both the international standard and their own needs. This feature remains polemic as academics and practitioners have not reached a consensus on whether the flexibility hinders or fosters effective enforcement and convergence (Nance 2018). On a regional level, the European Union was not exempt from the pressure to regulate money laundering. Hence, its regime evolved through regional standard-setting parallel to the global standards of the FATF. As Mitsilegas and Vavoula (2016) point out, all EU anti-money laundering Directives have been justified as necessary to implement FATF recommendations. As a result, the expansive approach in terms of predicate crimes by the FATF is followed regionally.
Directives have been the main focus for Europeanization research as they incorporate the notion of discretion among jurisdictions during implementation. However, the study of how this discretion works has long been focused on formal transposition issues such as dates and delays. This does not capture that Member States have their own regulatory or administrative traditions and styles that influence how they formulate policy (Thomann & Zhelyazkova 2017).
The differences in implementation can be explained by two, often ignored, reasons. First, countries customize laws according to their domestic settings (Thomann 2015), the process by which policy is Europeanized is complemented by the domestic policy choices of each country that are tailor-made to their circumstance (Bugdahn 2005). Second, the individuals in each country in charge of the implementation of EU policy can interpret the law in the books differently. In this sense, actors that are practical implementers often become EU lawmakers, as EU implementation does not end when transposition is done given that rules and principles continue to be used in practice (Versluis 2007;Dörrenbächer & Mastenbroek 2019).
In the case of money laundering, the individuals in direct contact with the application of EU Directives rest on three pillars. The first pillar consists of administrative authorities (e.g., AML supervisors within tax authority or banking authority) who supervise and impose administrative fines on entities (e.g., banks, casinos, accountants, etc.) if they do not comply with the regulation. The second pillar consists of financial intelligence units (FIUs), which are in charge of collecting, analyzing, and disseminating the reporting done by the entities mentioned earlier. Finally, the third pillar consists of law enforcement agencies and the justice system in charge of the prosecution of individuals and entities that do not abide by AML regulation. These three pillars apply the implemented regulations individually and interact with each other (Kirschenbaum & Véron 2018).
Case studies have already suggested that even when the implementation of EU policy is considered compliant, it can still result in divergent national outcomes (Falkner et al. 2005;Versluis 2007). Based on this, we expect that the case of tax crimes as a predicate crime for money laundering will also diverge across jurisdictions, as national legislations adapt and interpret EU Directives differently. We present the reasons that can explain said divergences. In Section 3, we operationalize these variables to apply them to a taxation context.

Government quality and capacity
Implementation theorists have tried to understand what explains patterns in EU compliance. This includes whether Member States transpose directives on time or not (König & Luetgert 2009), explanations for noncompliance (Mbaye 2001), and the existence of differences between legal and practical implementation (Falkner et al. 2005). The varied explanations are most likely due to a lack of agreement on what compliance is and if it is transposition delays, infringements, or application records (Angelova et al. 2012). Yet, a common thread to this research is the finding that state power or government quality and capacity matter, especially when this is proxied through administrative capacity, bureaucratic capacity or efficiency, and government quality.
Toshkov, for example, confirms in a literature review of all quantitative research on EU law up until 2010 that there is strong evidence that administrative efficiency positively influences Member State compliance with EU law (Toshkov 2010). A potential explanation for this is that in the presence of low bureaucratic capacity, administrative actors recognize that their ability to comply with legislation is low, which results in a lack of incentives to ensure proper implementation of policy (Huber & McCarty 2004), hence in worse compliance. Besides, authors such as Berglund et al. (2006) and Mbaye (2001) argue that inefficient bureaucracies are more prone to private interests.
When it comes to the practical implementation of EU regulation, few studies try to understand what determines legal and practical compliance systematically. Falkner et al. (2005) state that the "enforcement systems of some member states are institutionally ill-equipped to assure practical compliance" (p. 243). Yet, to the best of our knowledge, Zhelyazkova et al. (2016) are the first to analyze legal and practical implementation across all Member States systematically. They find that practical implementation is mostly shaped by institutional capacity. In this line, government quality and capacity are even more critical for the practical implementation of national laws (Zhelyazkova et al. 2016).

Tax profile
Research on EU implementation has focused on a limited range of policy areas (Angelova et al. 2012). However, issues do matter. The saliency of an issue determines whether it is ignored, therefore, less salient matters tend not to be customized (Thomann 2019). Moreover, saliency can indicate the importance of a policy, and as Perkins and Neumayer (2007) suggest, the policy preferences of national governments matter. Hence the importance of including considerations specific to the policy field of study. In the case of tax crimes as a predicate crime for money laundering, salience might be relevant because whether tax offenses should be a predicate offense for money laundering has been "politically sensitive" (Bell 2003) and a "key area of dispute" (Levi 2002).
We can divide the literature on preferences and attitudes into two: those that focused on the preferences and attitudes and their influence in state-level decisions, and those that focused on how these preferences and beliefs affect the domestic institutions and actors who implement policy. We build upon both. When it comes to statelevel preferences, research has found that Member States with a strong preference for EU policy will comply faster than countries that oppose said policy (Toshkov 2010). Perkins and Neumayer (2007) found that when focusing on the implementation of EU environmental policy, Member States with more vested interests or more to lose have a worse implementation record. Yet, it is not only the topic that influences how policies are implemented, countries formulate and implement policies influenced by previous and long-lasting policies on an issue and by their regulatory tradition (Adam et al. 2017). We can add that the tradition of specific institutions, such as a country's legal system, can also affect compliance (Perkins & Neumayer 2007). This is especially relevant because, as Thomann (2019) points out, local administrations interpret or reinterpret the overarching norm to fit their national identities. Given the importance of saliency, we expect this to differ across different administrative institutions.
When it comes to traditions and preferences influencing actors, Mastenbroek and Kaeding (2006) highlight the need to focus on the "preferences and beliefs of domestic actors regarding a particular EU policy input" (p. 339), this is especially important since implementers on the ground are the second front line of implementation and those who interpret definitions from directives (Bondarouk & Mastenbroek 2018) and that legislators can "pass the buck" to practical implementers (Dörrenbächer & Mastenbroek 2019). Both the state and the preferences of practical implementers encounter each other through societal legitimacy or public support. As Zhelyazkova et al. (2016) say, "public support for EU policy and national institutional settings seem to shape the incentives of administrative actors to comply with domestic legal outputs and EU policy requirements" (p. 843).

Methodology and data
We base our research on three sources of information. The first is the tax crime and money laundering legislation the law in the books (Rossel et al. 2019). 9 Our second is data obtained from an expert survey of the so-called second-frontline implementers, which gives insight into how the national regulation is perceived, understood, and applied by those who interact with it. Both information sources were compiled into a single comprehensive database that includes law in the books and law in action for all 28 EU Member States. Finally, we use in-depth semi-structured interviews to provide context and accurately interpret our data.

The law in the books and law in action database
The field of comparative law has mainly relied on either detailed qualitative comparisons of a small sample, two to four countries (e.g., Hufnagel 2004), or on the extraction of concrete quantitative variables from broader samples (e.g. the datasets used for legal origins by La Porta et al. 2008). We combine these, as our database contains primary qualitative information from the original national law, translated into English, and quantitative data such as the maximum and the minimum number of years somebody could be sent to prison if he/she is convicted for tax crimes or money laundering.
To compare the law in the books, we gathered tax and money laundering laws from all 28 EU Member States. Taking the PANA committee's inquiry, we complemented and updated this information, completed, categorized, and found the articles in the laws that corresponded to what was reported. This was done based on primary sources such as the official gazettes. We used official English translations, when not available, we translated with the help of online legal dictionaries and local lawyers. 10 We revised our work with Thomson Reuters and International Bureau of Fiscal Direction legal databases' and sent our collected information to each country's relevant ministry requesting feedback. Table 1 presents a summarized version of the variables available in the dataset.
We capture two crucial areas of the law in the books through these variables: the punitiveness or jail time and the prescription time for a crime. Analyzing penalties is relevant for three reasons. First, variations between countries can cause forum shopping across EU Member States, suggesting criminals can focus their activities in countries with less severe (expected) sanctions (Bondt & Miettinen 2015). Criminals might not solely choose different locations but rather conjointly exploit the variations in legal systems to their advantage (Arnone & Borlini 2010). Second, penalties have a deterrence function (Carlsmith et al. 2002). Third, they reflect how "serious" a crime is considered (MacKinnell et al. 2010).
Current literature has already discussed that countries have different interpretations of what money laundering (Unger et al. 2014) and tax crimes are (Levi & Soudijn 2020) and how they are punished. Our contribution to the literature is that we do not only capture whether a crime is punishable or not by law through a binary yes/no variable (Schwarz 2011), but rather the extent to which it can be punished.
Jail time reflects the punitiveness that a country gives to a crime, prescription times (or statute of limitations) reflect the limits put by the law to the prosecution of crimes. The prescription period is the number of years that a crime can be prosecuted after being committed. Their relevance is twofold. On the one hand, prescription periods are crucial for cooperation among countries. 11 On the other hand, they represent a ticking-clock for prosecutors and investigators. By making tax crimes a predicate crime for money laundering, both crimes' tickingclocks have to some extent merged. This means that a tax crime that has already prescribed and hence is timebarred can still be prosecuted for money laundering if the prescription period for money laundering has not ended. This is, of course, subject to each country's legislation on how prescription periods of predicate offenses interact. For example, in the Netherlands and Italy, prosecution for money laundering is possible even if the underlying crime has prescribed.
We complement the law in the books with variables that reflect the law in action through an expert survey to understand how international recommendations translate into national regulation and how they are perceived, understood, and applied. For example, the maximum of years an individual can be sent to prison, according to the law in the books, is not the only key factor in analyzing the differences in the country's punitiveness or penal policy  Case Details 1 (individual) It has been proven that a football player, who is a taxpayer in your country, did not declare 10,000,000€ (or equivalent in your currency) of his/her personal income to tax authorities over the period of 2010-2015. This money is placed in a Swiss bank account through a company. It has been proven that this was done with the explicit intent to pay less tax. Please apply current regulation to answer.

(company)
A pizza company, that has its own legal personality and is a taxpayer in your country, has evaded 10,000,000 € (or equivalent in your currency) of taxes on profits by generating false invoices to reduce its profits by increasing its deductibles over the period of 2010-2015. Please apply current regulation to answer.

(crossborder)
For this case respondents get the case twice. The first for coding is referred to as A and second time as B, this is not visible for respondents.
A. Company X (a company with its own legal personality) has evaded 300,000 € in taxes in your country. Over the last year, this individual has spent this money investing in real estate in another jurisdiction within Europe B. Company X (a company with its own legal personality) has evaded 300,000 € in taxes in another jurisdiction within Europe. Over the last year, this individual has spent this money investing in real estate in your country (Krajewski 2014). It is also essential to know the likelihood of really having to go to prison or the practical enforcement of a sentence, whether served or not (Coffee Jr. 2007). Through our survey, we ask questions relative to the prosecution of tax crimes, money laundering, and tax crimes as a predicate crime for money laundering, the penalties or sentences attributed to both of these, and the overall challenges that prosecutors and investigators face. To make this tangible and comparable, we gave survey participants three cases: a tax crime committed by an individual, one committed by a company, and a cross-border case. They were inspired by real situations and did not divulge any private judicial outcomes. The questions were tested beforehand on two occasions by practitioners with different linguistic, professional, and legal (common and civil) backgrounds. The cases were intended to trigger participants into fringe situations, to see where and which problems would arise. Table 2 details the three cases. With our survey, we try to overcome the inherent complexity of comparative tax law by formulating specific tax crime cases and ask practitioners in each country how they would deal with these cases and based on which legal arguments. 12 We ask them to quote their national legal sources, which served as a double-check for our legal recompilation.
To obtain answers, we used nonprobabilistic sampling in the form of purposive expert and snowball sampling, which helped us target those who work directly or indirectly with the transposition or application of the 4th AML Directive. To reach possible participants, we contacted all prosecutorial departments in Europe and ministries of justice/finance, FIUs, tax administrations, and Europol, among others. We also asked these participants to share the survey with colleagues to harness the power of snowball sampling.
The survey was hosted through Qualtrics and was open from November 2018 until February 2020. We first asked participants to evaluate the inclusion of tax crimes as a predicate crime for money laundering in the 4th AMLD. Second, we requested respondents to analyze the three model cases. Respondents were asked questions such as how each case would be prosecuted in their country, based on which laws the case would be handled, whether there would be a jail sentence for the individuals involved, 13 and whether this sentence would be served or not. To minimize respondent fatigue, we limited the number of follow-up questions and asked mostly multiple-choice questions.
To have one response per country, we condensed the answers when there were multiple participants. 14 This was done in various steps: we took the most common answer for each question (mode). In case this was not possible, or there was a tie between answers, we used the answer of the most experienced respondent. In the few cases where two respondents had discerning answers and had the same years of experience, we chose the answer of the respondent with the highest rank, for example, we would pick the answer of a Head Prosecutor over that of an investigator. We received responses for all EU Member States, except Malta. 15 For 22 countries, we have at least two answers. On average, respondents were between 35 and 44 years old and predominantly male, had 11 years of experience with taxation, and had been in their current position between 5 and 10 years.

Interpreting the law in the books and practice: In-depth interviews
We conducted in-depth semi-structured interviews to provide context and accurately interpret our data. These interviews took place between September 2019 and January 2020. The focus was to obtain a mixed pallet of interviewees (Table 3).

Country characteristics from other databases
We operationalize our explanatory factorsgovernment quality and tax profilewith country characteristics from publicly available data sources. To measure government quality and capacity, we use the following indicators: Government Effectiveness (like Zhelyazkova et al. 2016), Corruption Perception Index, and Regulatory  Quality. All these indicators are World Bank Governance Indicators. 16 Furthermore, we add inequality measured through the GINI coefficient. We emphasize in our theory section the need to focus on field-specific characteristics. We select variables that can reflect the preferences and characteristics of tax and money laundering related authorities, practitioners, and the overall civil society. We call this the tax profile. We include tax morale or the nonpecuniary motivation to comply with taxes (Luttmer & Singhal 2014) as a measure of the societal opinion of taxation. The Tax-to-GDP 17 ratio measures a countries' tax revenue relative to the size of its economy, which also serves as a measure of the tax burden (Joumard 2002). Effective corporate tax rates reflect the average rate at which companies are effectively taxed instead of the statutory tax rate (Tørsløv et al. 2018), which can indicate the countries position on taxation.
We also incorporate the Financial Secrecy Index 18 that ranks countries and jurisdictions based on their contribution to secrecy in global financial flows on a 0 (lowest) to 1 (highest) scale (Cobham et al., 2015). We add 5 sub-indicators from the FSI: 1 Tax administration capacity: Reflects a country's tax administration capacity to collect and tax people and companies. It evaluates organizational capacity, informational data processing preconditions, and the availability of rules for targeted collection of intelligence. This indicator is crucial as it reflects administrative capacity, which has been found significant in the literature, specifically for the realm of taxes. 2 Promotion of tax evasion 19 : This indicator judges whether a jurisdiction facilitates tax avoidance and encourages tax competition. 3 Tax court secrecy: This indicator assesses the openness of a countries' judicial system regarding tax matters, intended to operationalize judicial culture, specifically for taxation. 4 Harmful tax structures: This indicator evaluates the availability of four harmful instruments and structures: large banknotes, bearer shares, series limited liability companies, and trusts with flee clauses. In the theory section, we delve into how long-lasting policy traditions can shape current policy decisions. For example, countries with a long history and tradition of financial secrecy, such as Switzerland, long refused to cooperate with any attempt to improve policy on international tax cooperation   (Emmenegger 2017). Hence the harmful tax structures indicator is meant to reflect the country's position in the tax realm. 5 Anti-money laundering: This indicator reflects to what extent the AML regime of the country fails to meet FATF recommendations. Table 4 provides an overview of the variables we use to operationalize our explanatory factors.

Shedding light inside the black box
We see stark differences in the law in action and law in the books. For the law in action, we illustrate this through the first case in which a football player fails to pay taxes for 5 million Euros in a Swiss bank account. We asked respondents if this case would be pursued as a criminal case or as an administrative case. 20 In most countries, this case would be subject to both administrative and criminal sanctions. However, in Slovenia, 21 it would only be subject to administrative sanctions. Moreover, in Greece's case, one respondent answered that the case would have been only subject to administrative charges due to "bureaucratic reasons." We then asked respondents if the case would be prosecuted as a tax offense, money laundering, or both. Interestingly no jurisdiction would prosecute this case only as a money laundering offense. Figure 1 illustrates how this differs across countries.
Once the type of prosecution that a country's authorities would pursuit was evident, we proceeded to ask questions on penalties. When it comes to jail time in Belgium, the former player would not serve any time, while    Table 2. ‡ For this question, N = 11 is explained by the fact that it could be answered only by respondents who had previously stated in Case 1 that the individual could be prosecuted for money laundering.
in the Netherlands, he would face both tax crime and money laundering charges and go to jail for at least eight years. Figure 2 illustrates the responses of all countries when asked if, in this case, an individual would serve jail time or not. The x-axis shows the maximum number of years for tax crimes, and different shades of grey represent whether the individual would actually serve time in prison. Respondents in Belgium, Estonia, France, Greece, Italy, Latvia, and Lithuania say that when it comes to applying the law in action, said individual would not serve the time in jail in their country. When shedding light on the implementation of tax crimes as a predicate crime for money laundering in the EU through the law in the books, we can illustrate two key points of contention: the maximum jail penalties established by law and the prescription period for these crimes.
The 4th AMLD gave the minimum indication that tax crimes were those that "are punishable by deprivation of liberty or a detention order for a maximum of more than one year," yet the maximum penalty for tax crimes across the EU can go anywhere between 6 months and up to 12 years according to the law in the books. 22 In Figure 3, we illustrate the maximum years of prison that can be given by law for money laundering and tax crimes, here it becomes clear that the practical implementation of EU law is very different across countries. The practical relevance of this is that by making tax crimes a predicate crime for money laundering, a criminal can be either sentenced for both crimes (summing the sentences) or just the highest sentence.
Equally interesting are the differences in the ticking-clocks given to prosecutors through prescription periods or statute of limitations in Figure 4. 23 These play a crucial role in cooperation among countries through the mutual assistance that jurisdictions can give each other with information exchange, notification of liabilities, and asset (unpaid tax) recovery. Throughout our interviews, this issue was highlighted by respondents. From the Netherlands: "Prescription is an important issue with tax evasion. Some countries have a low prescription, this means you can't prosecute because the information we get is always old." Our respondent from Portugal highlighted that previously if the crime had ended, you could not prosecute the crime. Our respondent from Hungary pointed at the fact that "When the statute of limitations has already passed, money laundering can still be used." Out of the 29 jurisdictions in the figure below, nine have higher prescription periods for money laundering than for tax crimes, 14 have the same prescription periods for both offenses, and six have higher prescription periods for taxes than for money laundering.

Explaining the black box
The question remains how we can explain the heterogeneity of tax crime and money laundering law in the books and in practice from Section 4.1. We find that countries fall back on their old ways or patterns. We discuss this for our two explanatory factors: government quality and capacity and the tax profile.

Government quality and capacity
Literature describes how government quality and capacity affect the implementation (transposition, practical application, and enforcement) of European Directives. Although the findings of this line of research are varied, the results in general point at factors such as government quality and corruption that affect overall compliance (Toshkov 2010). We are interested in whether these variables would affect the domestication of laws in the books and laws in action. We measure government capacity and quality with indicators such as corruption, regulatory quality and government effectiveness, GDP per capita, and the level of inequality. We find that these variables affect how the tax crime as a predicate crime for money laundering principle is transposed from the 4th AML Directive into national legislation, both in the books (Table 5) and in action ( Table 6).
The heterogeneity of the number of years that an individual can be sent to prison for both tax crime and money laundering is related to government capacity. Correlation results in Table 5 show that countries that are less punitive than their counterparts have a higher GDP per capita, rank better regarding perceived corruption, have higher regulatory quality and are more effective governments, which means that countries with higher government capacity have significantly lower minimum and maximum prison times for tax crimes and money laundering. This matches the notion that countries with high government capacity have high political legitimacy and are less punitive, as they tend to focus more on social policy, which in turn fosters penal moderation (Snacken, 2010). These results are, to some extent, mirrored when analyzing the law in action, although we find less significance.
When asked if they would prosecute the same case for money laundering and/or a tax crime, these countries were less likely to prosecute the case for both crimes, choosing more often for prosecuting only tax crimes and not money laundering.
The preference of rich developed countries for prosecuting tax offenses only as tax crimes and not for money laundering can be due to two factors. First, the inclusion of tax crimes as a predicate crime for money laundering is related to the need to use money laundering as a tool to fight tax evasion. This is further reinforced by the correlation between countries with high GDP per capita being more likely to agree with the statement that incorporating tax crimes as a predicate crime for money laundering was unnecessary. As with other predicate crimes such as corruption, where money laundering is more helpful for developing countries (Sharman & Chaikin 2009), our results signal that for countries with stable government quality and capacity, this tool is neither much used nor necessary.
Second, our results could indicate a specific line of interpretation of the ne bis in idem principle, where countries with higher regulatory quality go more often after one crime only, in this case, the underlying predicate crime. 24 We asked respondents whether an individual guilty of a tax crime was likely to serve his/her prison time. In seven countries, the individual would not have to serve his/her time in prison. Our analysis shows that this is correlated with a countries' inequality level. More unequal countries are less likely to have effective prison times. This answer is even more relevant considering that the case described to respondents was that of a football player evading 5 million euros, hinting that wealth plays an important factor in more unequal countries, especially when it comes to sentence enforcement. A practical example of how wealth and popularity influence sentencing or case ***P < 0.01; **P < 0.05; *P < 0.1. † The FSI and its sub-indicators range from 0 to 1, where 1 is highly secretive, so countries that fare poorly in an indicator get closer to 1. All correlations in this table are based on 28 observations.
treatment is the answer from a respondent from the Dutch National Police who stated that this individual would be both legally and practically prosecuted for a tax crime and money laundering because: "The purpose was to hide the money for the government, it's a huge amount, and a football player has a kind of exemplary function."

The tax profile
We find indications that the way tax crimes as a predicate crime for money laundering is implemented is influenced by what we call the national "tax profile." These characteristics define the dynamics of tax crimes and money laundering within a jurisdiction both in the books (Table 7) and in action (Table 8).
A key component of the law in the books is judges' discretionary power, broadly understood as the space that judges have to make their own decisions. We operationalize this as the leeway judges have regarding sentencing prison time. Sentencing is ultimately related to judiciary independence, and it is an area in which judges are limited by regulation (Asp 2013). The minimum and maximum prison times set these limits. We find that countries with high tax morale, less secretive tax courts, and more administrative capacity give higher discretionary power  to judges for sentencing in the case of money laundering. Although this relationship must be interpreted carefully, it can indicate that in countries with closed or secretive judicial systems, the law in the books tries to constrain the discretionary power of judges due to a lack of trust.
A second finding is about the role of prescription times. If discretionary sentencing power is a limit to judicial decision-making, prescription times are a limit or a ticking-clock for investigators and prosecutors. Furthermore, similarly to jail times, prescription times reflect the seriousness of a crime. The more serious the crime, the longer it takes to prescribe (Tak 2008). We find that countries with poor tax administrative capacity, and more harmful tax structures, also have lower prescription times either for tax crimes or money laundering. This can hint that countries that follow these profiles might not want tax crimes chased; prescription times can be used to limit the prosecution and sentencing of such tax crimes.
Regarding the correlations between our law in action variables and the tax profile, we find evidence of the "false-friend" effect whereby tax havens punish money laundering activities harshly (Masciandaro 2005;Schwarz 2011). This is especially evident when analyzing the outcomes of a case that included the cross-border factor. We asked respondents whether and how they would prosecute a fictitious case in which a tax crime was committed in a jurisdiction different from theirs, but the money was invested in their jurisdiction. We find that countries that have a secretive profile meaning that they rank high in the indicators of tax court secrecy, harmful tax structures, and tax administrative capacity, respond more often than they would punish and chase this as money laundering.
Our findings suggest that countries do not only "fall back on their old ways" or implement a principle in a path-dependent manner regarding the way they apply the law but also regarding the timing of adopting a principle. In our survey, we asked practitioners if their laws had already incorporated tax crimes as a predicate crime for money laundering, and if so, when. We find that countries that have more harmful tax structures and are more secretive also incorporated this principle later than other EU Member States.

Conclusion
There has been an increased awareness that even perfect transposition of EU Directives does not necessarily translate into homogeneous rules or homogeneous application of rules across the EU. Ambiguities in EU regulation leave space for second frontline implementers or those who work with the rules to apply and use them differently. In 2012 the FATF put forward the standard to consider tax crimes a predicate crime for money laundering. In 2015 the EU followed suit through the 4th Anti-Money Laundering Directive. By 2017, EU Member States had to transpose and implement the 4th AMLD. This paper analyzes how one concrete part of the 4th AMLD was implemented both in paper and in practice. Tax crimes were made a predicate crime for money laundering, so AML regulation could be used to combat tax evasion. However, tax crimes were not defined by the FATF or the 4th AML Directive, leaving leeway for countries to incorporate this term into their legislation as they saw fit.
Through a database that includes both the law in action and in the books, we find differences in (i) whether a case is prosecuted through an administrative or a criminal procedure; (ii) whether a case is prosecuted as a tax crime, a money laundering crime, or as both, and (iii) whether this prosecution leads to the same punitive consequences for those who commit the crime, both in the books and in practice. The differences are striking. For example, the same crime can lead to six-month imprisonment in Malta and six years in Sweden or Spain. Regarding the explanations for these differences, we find four key results.
First countries that have higher government quality, meaning that they are less corrupt and have better regulatory effectiveness, are less punitive than their counterparts. Second, these countries also fall back on their government quality and capacity when it comes to choosing how to prosecute crimes, leaning on administrative and tax prosecution rather than criminal and money laundering prosecution. The fact that the differences in punitiveness and type of prosecution are related to administrative characteristics can suggest that harmonization might be possible in the future. The creation of European-level instruments such as the European Anti-Fraud Office (OLAF) are then of added value. Third, countries with more "secretive" profiles limit their judicial structure more in two ways: judges get less discretionary power, and investigators and prosecutors get less time investigating and prosecuting the crimes. Fourth, countries that have a secretive profile fall back on money laundering prosecution, hinting at a "false-friend" effect.
The correlation results in this research must be interpreted with caution, considering that we relied on a small sample and at one point in time. This is why said results are accompanied by a qualitative narrative stemming from in-depth interviews and qualitative answers in our survey. Although our study does not provide a legal analysis of the treatment of tax crimes across the EU and rather seeks to find overarching patterns. The database we created can serve as a starting point for other types of analysis, such as longitudinal research on the impact of AML directives, when information on future legal changes is added. In addition, the dataset can be used by legal scholars to do qualitative legal (comparative) analysis.
Yet, we believe the results are relevant in at least two respects. First, we find that although regulation has gone global and regional, in practical terms, the nation-state still has control and a path-dependent implementation of international principles. Second, our analysis opens the door to questions on the future of global and regional regulation, given that national implementation is still so diverse.
The results of this study paint a picture where European countries seem to fall back on old national habits when it is their turn to domesticate global and regional policy both in the books and in action. This, on the one hand, can be positive for maintaining the variety of Europe by keeping alive the original EU motto In varietate concordia (United in diversity). On the other hand, too much diversity can hinder cooperation in dealing with global challenges such as transboundary crimes.

Acknowledgments
We thank Dr Jan van Koningsveld and two research assistants, Jason Batchelor and Francisca Vallejo, who contributed significantly to this dataset. This paper has been prepared in the Combatting Fiscal Fraud and Empowering Regulators (COFFERS) project. This project has received funding from the European Union's Horizon 2020 Research and Innovation Program under grant agreement No 727145.

Data Availability Statement
The law in the books can be found online at: https://doi.org/10.5281/zenodo.3476656 (Rossel, Unger, Batchelor, & Vallejo, 2019) Endnotes 1 Defined as "the underlying criminal offense that gave rise to the criminal proceeds which are the subject of a money laundering charge" (Bell 2003, p. 137). 2 So far, the EU approach towards money laundering has come in the form of Directives as they allow individual countries to decide how they transpose regulation into their legislation. Directives work more as guidance of minimum standards leading to minimum levels of harmonization. Unlike the two other European Policy instruments regulations and decisions that are directly applicable and result in maximum harmonization or identical legislative agreements (Versluis 2007;Kirschenbaum & Véron 2018). 3 There have been many changes in the tax and anti-money laundering legislation landscape since the issuance and implementation of the 4th AMLD in 2015 and 2017, respectively. The 5th AMLD was adopted by the European Parliament on 19 April 2018 and published in the Official Journal of the European Union 19 June 2018. The 5th AMLD is often called the 4th AMLD amendment, not only because it came so soon after it but because rather than putting forward new initiatives it modifies and deepens the 4th AMLD in three areas: ultimate beneficial ownership, use of prepaid credit cards and financial intelligence units. The 5th AMLD keeps the inclusion of tax crimes as a predicate offense for money laundering (EU, 2018/843) but still does not provide a concise definition of tax crimes. For an analysis of the 5th AMLD refer to Turksen and Abukari (2020). The 6th AMLD issued in 2018 must be transposed by December 2020. 4 "All offenses, including tax crimes relating to direct taxes and indirect taxes and as defined in the national law of the Member States, which are punishable by deprivation of liberty or a detention order for a maximum of more than one year or, as regards Member States that have a minimum threshold for offenses in their legal system, all offenses punishable by deprivation of liberty or a detention order for a minimum of more than six months (4th AMLD Art. 3(4)f)." 5 The lack of a concrete definition for tax crimes goes in line with the notion that directives work as guidance of minimum standards and with Art. 83 (1) and (2) of the Treaty on the Functioning of the European Union (TFEU) that allows the establishment of minimum rules concerning definitions of criminal offenses and their sanctions while keeping a balance between European interests and internal coherence of the own laws and sanctions of Member States (Richardson 2012;ECLAN & ECORYS 2015). And in line with Directive 2017/1371 on the fight against fraud to the Union's financial interests by means of criminal law, whereby the structure and functioning of the tax administrations is left to each Member State. 6 We thank Dr Jan van Koningsveld and two research assistants, Jason Batchelor and Francisca Vallejo, who contributed significantly to this dataset. The law in the books dataset can be found online at: https://doi.org/10.5281/zenodo.3476656 (Rossel, Unger, Batchelor, & Vallejo, 2019) 7 When this research was carried out the United Kingdom was still part of the European Union and hence is included in the analysis. 8 For example, the PhD of Hufnagel (2004), answers a question (whether a lawyer who defends a drug dealer is a money launderer) for Germany, Austria, Switzerland, and the United States in 293 pages. 9 The full dataset is available through http://www.coffers.eu/. It includes complete texts we use from each country's tax crime and money laundering law that we base the research on. 10 We thank: Alexandra Nagoyeva -Hungarian and Slovak; Linda Kunertova -Czech; Catalina Papari -Romanian; Tomas Balciunas -Lithuanian; Giovani Caroli -Italian, and Andoni Montes Nebreda -Spanish, for their contributions in translating.

11
The cooperation role of prescription times is related to the mutual assistance that jurisdictions can give each other in terms of exchange of information, notification of liabilities, and asset (unpaid tax) recovery. Although states must assist each other, the formal procedure of requesting said assistance is limited by prescription periods. 12 The complete survey is available upon request. 13 For Case 2 we ask this question twice to understand the differences in liability of natural and legal persons. Given that this is more pertinent to a legal analysis we do not use this information in our correlations in Section 4. The survey data is available upon request and can serve to further understand the workings of Articles 6 and 9 of Directive 2017/1371 (issued after the 4th AMLD) on liability of natural and legal persons. 14 Documentation on this process is available upon request. 15 Due to the snowball sampling we also got responses from Moldova; however, we do not use these in our analysis 16 Although most research has focused on bureaucratic quality indicators, some authors have used GDP and GDP per capita as a proxy for state power or state capacity (Börzel et al. 2007) finding no significant results. However, as Angelova et al. (2012) find 67 percent of the research focused on environmental and labor and social policy. This finding suggests that GDP variables might still influence other policy fields. 17 We have no clear indication of the expected sign of the tax to GDP ratio, because the relationship with tax evasion can be complicated: as low ratio can either indicate a low tax burden or a high tax burden with a high evasion rate. Yet the taxto-GDP ratio is still a relevant indicator, for example in Europe, Denmark has one of the highest tax-to-GDP ratios and very low evasion (Kleven 2014). Furthermore as Kleven et al. (2016) mention developing countries tend to have a low tax-to-GDP ratio which tends to increase as they develop. 18 For information on the Financial Secrecy index visit: fsi.taxjustice.net/en/ 19 In the 2020 edition of the FSI this variable is renamed as "avoids promoting tax evasion." 20 Note that we do not focus on the discussion on where such a case will and should be prosecuted. Money laundering is often committed across multiple jurisdictions. In such cases, offenders can be subject to the money laundering laws of several jurisdictions. This can lead to a jurisdictional conflict because two or more sovereign states have a right to assert criminal jurisdiction over the same crime. It can also lead to difficulty prosecuting non-resident defendants outside the boundaries of the state. (Amrani 2012, p. 149). This can lead to complex legal issues which we cannot discuss here. For a more detailed description of such issues, see Amrani (2012, especially chapter 6). 21 Also in the non-EU member state Liechtenstein. 22 Some considerations when interpreting this data as the maximum number of years an individual can be sent to jail can depend on factors such as the accumulation of sentences or aggravating factors. Hence some discretionary decisions have been made to facilitate the correlation analysis. For example, in the case of the UK this sentence could be up to a lifetime when the individual is charged under cheating the public revenue. However, this is a "judge-made" criminal offense. Hence we include the maximum penalty as established in the law for personal income tax evasion which is seven years. In the case of Poland, jail sentencing for tax crimes can be found in the Fiscal Criminal Code (Art. 54 and 56) under tax evasion and tax fraud and in the Criminal Code (Article 270a) related to false invoices for tax fraud. We take the Fiscal Criminal Code to make it comparable to other countries in the sample. In Hungary the maximum cumulative sentence according to Section 36 of the Criminal code is 25 years; however, Section 396 of the same code specifies 10 years as the maximum years for budget fraud, we take the latter into account. 23 It should be noted that as of the 5th AMLD (Directive 2017/1371) includes prescription periods in recital 22 "Member States should lay down rules concerning limitation periods necessary to enable them to counter illegal activities at the expense of the Union's financial interests. In cases of criminal offenses punishable by a maximum sanction of at least four years of imprisonment, the limitation period should be at least five years from the time when the criminal offense was committed. This should be without prejudice to those Member States which do not set limitation periods for investigation, prosecution and enforcement." 24 Maugeri (2018) discusses these issues in more detail.