Contemporary public policy towards prostitution regulation on a global scale

Ðàçìåùåíî íà http://www.allbest.ru/

FEDERAL STATE AUTONOMOUS EDUCATIONAL INSTITUTION

FOR HIGHER PROFESSIONAL EDUCATION

NATIONAL RESEARCH UNIVERSITY HIGHER SCHOOL OF ECONOMICS

Saint Petersburg School of Social Sciences and Area Studies

Master's thesis

Contemporary public policy towards prostitution regulation on a global scale

Maxim Raschupkin

Saint Petersburg, 2020



Abstract

This thesis examines different classifications of prostitution policies and compares them to contemporary prostitution policies around the world. It examines 4 classifications commonly used in prostitution policy research and points out lack of their empirical verification, especially outside of select country cases. This quantitative study relies on average-linkage hierarchical clustering method using Hamming distance measure and applies it to a novel dataset based on national legislative documents and other sources of data about prostitution policy, that contains 205 instances of different prostitution regimes. While a number of subtypes consistent with existing research and discourse (such as the Nordic Model) can be established based on the data, my overall findings indicate that both policy research and policy discussion needs to move from using simple classifications based on laws to broader classifications that reflect intent and implementation of the studied policies, such as Petra Östergren's 2017 classification. The conclusion provides potential avenues for further research and application of the dataset.



Acknowledgements

I would firstly like to sincerely thank my supervisor Alexander Sungurov from the Saint-Petersburg School of Social Sciences and Area Studies - Higher School of Economics for his timely involvement and support of my research. Secondly I would like my consultant and long-time supervisor Tatiana Barandova from NWIP RANEPA for priceless inspiration and support throughout the years, this thesis would not have been possible without her involvement. I would also like to additionally thank professors Galina Selivanova, Olessia Kolstova, Sergei Koltsov and Boris Sokolov who helped me in strengthening my methodological choices and the overall quality of my research.

Finally, I would like to thank my friends and relatives who have been supporting me since the beginning of this research.

Maxim Raschupkin



Table of contents

Introduction

1. Thus, the research question of this paper is as follows: Which types do prostitution regulation regimes break into?

1.1 Objectives

1.2 Definitions

2. Literature review

2.1 Pajumets' classification

2.2 Mossman's classification

2.3 McCarthy's classification

2.4 Decker's classification

2.5 Östergren's classification

2.6 Hypotheses.

3. Empirical chapter

3.1 Methodolody

3.2 Data gathering

3.3 Data Overview

3.4 Data analysis

Conclusion

Bibliography

Appendix



Introduction

Statement of the research problem

Prostitution policy process has been extremely turbulent for the last two decades, specifically in but not limited to Europe. It sporadically enters and leaves public attention as underground conflicts between various interest groups surface in legislative bouts - France criminalizes purchase of sex, Amnesty International calls for complete decriminalization of the sex industry, lawyers of escort agency owners claim that Canadian client criminalization (enacted in 2014) unconstitutionally endangers prostitutes, and so on. This policy issue has an unusual breadth, as not only the methods of achieving a socially desirable goal are contested, but the character of that goal itself: is prostitution moral, can it be accepted in a modern society, should it be considered an issue at all? While different countries have their own sets of path-dependent options, some global patterns seem to emerge - for example, a neo-abolitionist movement led by radical feminists and conservatives For an example of their interconnections, see Emily St Denny,”The gradual transformation of a weak but enduring regime: contemporary French prostitution policy in transition (1946-2016)”, in Modern & Contemporary France, 25:3 (2017), 299-314 For an analysis of different feminist perspectives towards prostitution, see Katie Beran, “Revisiting the Prostitution Debate: Uniting Liberal and Radical Feminism in Pursuit of Policy Reform”, in Law and Inequality 30 (2012), pp. 19-56 that seeks to combat prostitution by punishing clients. What remains unclear is the policy landscape left in the wake of these conflicts.

The concept of prostitution model (or prostitution regime) is a core one in both advocacy and research. Journalists, activists and politicians use these models as shorthands in discussing and drafting new laws, researches use them to compare impacts of laws in different countries, and their research is then used by interest groups, models included. They are extremely important, because they allow to easily frame a policy in a certain way, its actual outcomes notwithstanding, and this framing is hard to challenge. A case in point: Canada, where a “decriminalized” framework was stricken down by the Supreme Court as effectively making legal sex work impossible, only to be replaced by a different “decriminalized” framework For a case study of Canada's prostitution policy, see Lauren Sampson, “The Obscenities of this Country”: Canada v. Bedford and the Reform of Canadian Prostitution Laws , in Duke Journal of Gender Law & Policy 22 (2014) 137-172 that, as we've mentioned earlier, is challenged again on the same basis. However, despite their importance, these models never become an object of study themselves. As we later see, categorizations created are rarely justified besides citing “common practice” and using few select countries as examples (i.e. “Swedish model” or “New Zealand model”). Different categorizations disagree on status of prostitution in particular countries. As Petra Östergren notes:

“To use several categorical concepts is not problematic in itself, but a review of prostitution policy studies reveals that scholars define, combine and apply them in such diverse ways that it makes it difficult to employ them as analytical categories and classification devices”.

All of that comes on top of existing methodological challenges to prostitution research, which have been consistently outlined in literature Cecilia Benoit, Michaela Smith, Mikael Jansson, Benoit, C., Smith, M., Jansson, M. et al““The Prostitution Problem”: Claims, Evidence, and Policy Outcomes. “in Archives of Sexual Behavior, 2018; Elaine Mossman, “International Approaches to Decriminalising or Legalising Prostitution” in Crime and Justice Research Centre, University of Wellington, 2007. Those include “hard-to-reach” status of people involved in prostitution (including a lack of a sampling frame and stigma on acknowledgement of one's status as a prostitute), various sampling biases inherent in convenience sampling (which is the type that researchers often have to resort to in view of the hidden status of the sector), widespread and multidirectional ideological biases and abundance of low-validity research. When all that is combined with a necessity to either create a new typology or voluntarily choose one from a multitude of existing ones (and thus define what the alternatives for comparison are, what is typical and what is an aberration), reliable comparative research becomes unreachable. This is the gap in current knowledge that this study addresses.



1. Thus, the research question of this paper is as follows: Which types do prostitution regulation regimes break into?

1.1 Objectives

To identify most common prostitution policy classifications used in modern research, their origins, bases and shortcomings

To gather all available information about prostitution regimes in various states

To create a prostitution categorization that would most accurately reflect the data

1.2 Definitions

Before I proceed to discuss the state of the research, three definitional issues need to be discussed: prostitution itself, accurate term to refer to it and finer distinctions between typical and specific models.

While a certain degree of theoretical debate exists on what constitutes prostitution and where does the border between it and other sex-related activities lie (see Green 2016 for a detailed review), its general broad definition usually looks like “provision of sexual services in exchange for money or other benefits”, while referring to a narrower part of such activities that involve commercial transaction. As this paper focuses on prostitution policies, rather than prostitution itself, I can avoid this debate entirely by defining prostitution as whatever activities local laws (or legal practice in their absence) treat as prostitution. While convenient, this tactic relies on an assumption that different states use the term “prostitution” to refer to the same set of activities - an assumption that is generally uncontroversial, but should be noted nevertheless.

There is a tendency in modern research to refer to prostitution as “sex work”, which is motivated by allegedly neutral or “non-judgmental” nature of the term. However, this term has been criticized as implicitly normalizing prostitution via equating it to other kinds of work, while neglecting problems of violence and coercion that are inherent to the industry, as well as erasing the boundary between people actually selling sex and third parties such as pimps/madams/managers. Partially agreeing with this criticism, in this paper we'll keep using “prostitution” and “prostitutes” as the middle ground between two conflicting frameworks (“sex workers” vs “people involved in prostitution”). While in some social contexts these terms can be seen as pejorative and judgmental, they have modern academic precedent and should be reclaimed as such. However, aforementioned frameworks can be referred to when describing policy models that expressly use them (especially when applied by the author of a classification).

For the sake of clarity there needs to be a distinction between general policy types that I am working with and actual policies applied in specific countries. While a number of mutually exchangeable terms is used for both, in this paper I use “prostitution model” or “prostitution type” to describe the former (e.g. “The legalization model”), “prostitution regimes” or “prostitution policies” to describe the latter (e.g. “The French prostitution model”), and “prostitution (policy) classifications” to refer to theoretical systems that use multiple prostitution models to describe variety of prostitution laws in different countries (e.g. “Mossman's classification”).



2. Literature review

In this chapter I provide a broad overview of literature in prostitution policy studies. As you will see, modern literature in this field includes a number of different prostitution policy classifications, with some overarching tendencies (such as separation between models that prohibit and condone prostitution) and a lot of idiosyncrasies. As no classification is used universally, I had to resort to chronologically describing a number of them used in the literature (in detail, because many significant distinctions lie in details), and infer from them the general state of the field. I also take note of countries cited as having a certain model, as it might be useful for spotting differences between classifications (some differences inevitably arise due to policy changes that happened between publications of the papers, these are ignored).

2.1 Pajumets' classification

The first classification was created by Marion Pajumets in 2004 as part of the project “Prostitution - a social problem? The views on prostitution's nature, causes and effects in the Baltic states and north-western Russia” Marion Pajumets, “Prostitution - A social problem? : The views on prostitution's nature, causes and effects in the Baltic states and north-western Russia.” Equality Department of Estonian Ministry of Social Affairs, 2004; For more examples of this and similar classifications, see Bernstein, 2007; Carline, 2009; Dewey & Kelly, 2011. Like most of the researchers we'll discuss, she does not claim to create a classification, but rather to reflect an obvious structure of both theoretical field and legal systems. Four models are identified:

Criminalization/prohibitionism

Regulationism/legalization/classic order/French order

Neoregulationism

abolitionism/decriminalization/toleration

neoabolitionism/decriminalization combined with human right approach

The first model is described as treating prostitution as either a crime or a social disease and expecting the state to “uphold the morality of the people in the country”. Examples cited are China, Iceland, Japan, most of the US and Canada. The second model claims to be morally neutral towards “the world's oldest business” and focuses on licensing brothels and prostitutes based on various health and safety regulations. “Neoregulationist” subtype is mentioned, which differs by banning brothels/procuring while maintaining a similar general direction. Examples cited are Germany, Austria, the Netherlands, Greece, Switzerland and Turkey, with Latvia being an example of neoregulationism. The third model treats prostitution as free business association between consenting adults and does not seeks to license or regulate it. However, it also identifies brothels/procuring as inherently non-free (trafficking/slavery) and bans them, while maintaining a laissez-faire attitude towards prostitution in general (difference with neoregulationism is on the issue of licensing individual prostitutes). Examples cited Great Britain, Estonia, France, Ireland, Italy, Spain and Norway. The fourth model is identified as a form of abolitionism that emphasizes women's rights and protection of human dignity, seeing prostituted women as “victims of male-centered social manipulation” and seeing formal consent to prostitution as irrelevant. The only example cited in Sweden, with Finland mentioned as discussing its adoption.

This is a relatively simple classification, establishing what it seems as four “clear types”. Notably, it focuses as much (or more) on values and intentions inherent in these models, as on actual policy frameworks. While not necessarily true for all applications of them (and not necessarily relevant to this research in particular), this is an important dimension often overlooked in other classifications.

2.2 Mossman's classification

The second classification was created by Dr. Elaine Mossman as part of 2007 policy review for the Ministry of Justice of New Zealand Elaine Mossman, “International Approaches to Decriminalising or Legalising Prostitution” in Crime and Justice Research Centre, University of Wellington, 2007. It establishes three main models:

Criminalization

Prohibitonist

Abolitionist

Legalization

Decriminalization

Criminalization is referred to as a policy model where “it is not legally possible to engage in prostitution, because prostitution or its associated activities would be contravening some law, regardless of the level of tolerance existing”. Two subtypes of criminalization are established: prohibitonist and abolitionist. While prohibitionist criminalization outlaws prostitution entirely (examples cited are most of the US and the Middle East), abolitionist prohibition allows the sale of sex itself, but banning all related activities, thus making it effectively impossible to sale sex without breaking some kind of law. Abolitionist prohibition refers both to classical abolitionist models of England and Canada, and to Swedish neo-abolitionism. Legalization is described as the model with government-controlled prostitution, which is only legal under specific conditions, with different justifications but the same key indicators: prostitution-specific controls and conditions, such as licensing, registration and health checks. Examples cited are Netherlands, Germany, Iceland, Switzerland, Austria, Denmark, Greece, Turkey, Senegal, the USA state of Nevada, and many Australian states (Victoria, Queensland, ACT and Northern Territory). Decriminalization is described as repealing all laws against prostitution, using general (non-specific) norms to regulate it, while still outlawing coercion and child prostitution. It is noted that this model seeks to avoid two-tier system (legal and illegal prostitution) that is created by legalizations. Examples cited are New South Wales (Australia) and New Zealand, with both of them retaining elements of legalization, such as brothel licensing in New Zealand. The paper also cited “unregulated regimes” as a common phenomenon in Central and Eastern Europe and Central Asia, but the term likely refers to lack of enforcement of legal norms, rather than to their absence.

This classification has a sharper focus on policy distinctions, while still citing intentions and moral stances as an integral part of the models. It stands in a stark contradiction to earlier classifications in seeing all abolitionist regimes as criminalized ones. While on surface this distinction might seem insignificant, it means that the author disagrees with others on a quite fundamental question “Is prostitution legal in a significant number of states?”. It also means that the author sees these systems as more closely related to prohibitionist regimes, while it can easily be argued that abolitionist regimes are a form of decriminalization model with more severe restrictions on what can be considered “free and acceptable exchange”. While the author attempts to draw sharp distinctions between the models, it only makes the difficulty of this task more apparent. It is not entirely clear where the borders between decriminalization, legalization and abolitionist criminalization lie, or whether these borders exist at all.

2.3 McCarthy's classification

The third classification has been created by Bill McCarthy and others in 2012 Bill McCarthy, Cecilia Benoit, Mikael Jansson, Kat Kolar,B., Benoit,C., Jansson M., & Kolar K. (2012). “Regulating Sex Work: Heterogeneity in Legal Strategies” in Annual Review of Law and Social Science, 8(1) (2012), 255-271.. They use in their analysis three broad categories:

Full criminalization (proabolition)

Partial decriminalization (partial regulation)

Full decriminalization (regulation/antiabolition)

Full criminalization is described, broadly, as a model that criminalizes any or both of the main parties (buyers and/or sellers of sex), it includes both prohibitionist and neo-abolitionist models from aforementioned classifications, but does not include abolition. Partial decriminalization includes various “nuanced approaches” that criminalize prostitution in some forms or places and legalize it in others. These regimes include both “legalized” regimes such as Netherlands and “abolitionist” regimes such as pre-2014 Canada. Full decriminalization refers to regimes that minimize their involvement in the sex industry and employ limited regulation (often regional or municipal). These regimes include not only regimes considered “decriminalized” by Mossman, but some “legalized” regimes, such as Germany, as well.

This classification is deliberately broad: it correctly identifies issues with traditional models, acknowledges that real prostitution policies often do not fit established classifications (even this one, to a lesser extent) and aims to provide a general overview of existing policies that would encourage discussion and further research. While this approach can be useful as a part of broader discourse, its own utility is very limited: not only a country's position in this classification can tell us almost nothing about its policies (which is arguably a prerequisite for broader comparative research), but here said position can be determined differently, as criminalization of buyers and sellers is treated almost entirely separately by the authors, and the line between partial and full decriminalization is based more on an attitude towards prostitution, rather than on policy distinction.

2.4 Decker's classification

The fourth classification has been created by Michele R. Decker Michele R. Decker and others, “Human rights violations against sex workers: Burden and effect on HIV” in Lancet, 385(9963) (2015), 186-199. and others as part of their paper on human rights violations against sex workers. It establishes 4 models:

Full criminalization

Partial criminalization

Legalization

Decriminalization

Full criminalization prohibits “all of the following: the selling of sex, the purchasing of sex or earning money from someone's sex work (i.e. as a manager, as sex workers' working together cooperatively, as support staff, or as a landlord renting a home to a sex worker)” (Ibid.). Partial criminalization prohibits either one or two of the aforementioned activities and includes “the Swedish approach”. Legalization allows sex work under specified condition, most often with strict regulation and often through criminal law. Decriminalization allows sex work under normal occupational health and safety regulations.

On a superficial level, this classification is superficially similar to Mossman's: it sees abolitionist regimes as criminalized, has a policy-oriented definition focus and arrives to broadly similar conclusions. However, the criminalization models of this classification are defined radically differently, with full criminalization being almost non-existent, and most regimes elsewhere defined as prohibitionist, abolitionist or neoabolitionist becoming a singular unified category. Choosing one over the other as a basis of inter-model comparative research would lead to vastly different sets of countries in each of the four categories, despite a lot of similarities between these classifications.

2.5 Östergren's classification

The fifth classification has been created by Petra Östergren and others as part of the European DemandAT project. It establishes three broad “ideal types” of prostitution policy:

Repressive

Restrictive

Integrative

Repressive policy type aims to eradicate prostitution altogether in order to protect society and/or people selling sex from harm. Criminal law is used to discourage the sale and/or the purchase of sexual services, as well as to discourage or prohibit third-party involvement. It typically involves campaigns of deterrence of the targeted activities, and may encourage sex workers to exit the sector, with the dominant discourse being wholesale condemnation of sex work. The author, while acknowledging variations within the type (e.g. radical feminist approach to that model would treat those who sell sex as plaintiffs rather than criminals), notes that all repressive policies force the whole sex sector to operate illegally. It is also noted that this model can be more symbolic than pragmatic, valuing “sending a signal” over measurable outcomes.

Restrictive policy type aims to restrict prostitution in order to protect society and/or people selling sex from harm. It has the same general attitudes towards it, but has a degree of tolerance to prostitution, often pragmatic ones. Criminal legislation is mostly aimed at third parties, combined with criminal and/or administrative regulation of who, when, where an how can legally sell and/or buy sexual services.

Integrative policy type aims “to integrate the sex work sector into societal, legal and institutional framework in order to protect those selling sex from harm” (Ibid.). It is a rights-based approach that is based on a multifaceted understanding of commercial sex. Its policies mostly treat sex work as any other service sector, with specific protections for sex workers as an occupational category and a focus on collaboration between all actors in an effort to prevent and report crimes.

Östergren's classification is one of the latest attempts to address the same issues with prostitution policy classifications that have been outlined in this paper, and tries to be as complex as necessary to include all nuances in ideology, policies and impacts while remaining relatively simple in terms of actual types. The author also includes a qualitative assessment protocol of evaluating a policy's intentions, instruments, implementation and impact (the 4i's). While this classification exhibits an unparalleled amount of high-quality effort towards creating a precise, unambiguous and practically applicable classification, it is reliant on the previous discourse that it tries to part with, and often prioritizes overall intent over policies themselves, with no specific country completely matching its types. It is reliant on each researcher's ability to analyze all aspects of a prostitution policy, including intentions of its architects, and to arrive to a similar replicable conclusion. This classification also has a prescriptive focus - with Integrative policy type being here more as an aim than a description of any actual contemporary policy. All of this means that even this classification would benefit from an empiric grounding.

While this is not a complete list of commonly used classifications for a more complete discussion, see Cecilia Benoit, Michaela Smith, Mikael Jansson, Benoit, C., Smith, M., Jansson, M. et al““The Prostitution Problem”: Claims, Evidence, and Policy Outcomes. “in Archives of Sexual Behavior, 2018., it gives a representative view of how researchers perceive this policy area. It can be seen that, aside from a general tendency towards more nuanced classifications, there is little agreement on anything: number of models, their boundaries, distinctiveness and applicability for specific cases. The same policy regime (e.g. England) can be described, in order of aforementioned classifications, as decriminalized, abolitionist (criminalized), partially regulated, partially criminalized and restrictive. Not only are these terms in themselves imprecise and create additional problems in a field already ridden with methodological problems, they provide few clear reasons to choose one over the other. This indicates the need for a data-driven approach to classification that would be able to compare them in representing the actual policy regimes and either confirm some to be better than the others or indicate a new, integrated classification.

2.6 Hypotheses.

As I am deviating from established classifications (if any of them can be considered such), my reliance on their predictive power is be necessarily low. The most that can be expected before conducting data analysis is a rough estimate of how the field will look like after the analysis is concluded, rather than any specific features of any policy type. Based on the aforementioned classifications, I can expect several hypothetical patterns to appear in the data:

H1. 4 defined clusters with relatively little overlap, as established in I, with “abolitionist” countries being a cluster of their own;

H2. 3 defined clusters with relatively little overlap, as established in II and IV, with “abolitionist” countries belonging to the same group as other criminalized regimes;

H3. 2-3 loose clusters with few clear boundaries, as established in III and V.



3. Empirical chapter

3.1 Methodolody

As this paper is concerned with large scale cross-national comparisons of different prostitution policies, the unit of this study is a prostitution regimes - i.e. territories governed but separate sets of prostitution legislation. In most of the cases this means states. However, some countries - notably, Great Britain and Australia - have separate prostitution regimes in its regions, and for the purposes of this analysis countries with such policy subdivision are divided into separate units that reflect that. At the same time, the ability of regional and municipal governments in many countries to modify general prostitution policy (as in Germany or the United States) that does not lead to differentiation in policy models, is acknowledged but does not cause a country to be broken up in the dataset. I acknowledge that most prostitution policies have regional variance that exists on a continuous scale and believe that this approach is the best way to reflect that given the constraints of this analysis.

As I am faced with an abundance of data (in terms of existing policies, if not their impact) and a clearly defined lack of strong theoretical structures and the aim of partitioning the objects into homogenous groups Stephen C. Johnson, “Hierarchical clustering schemes.” Psychometrika iss. 3 (1967) pp 241-254., the best quantitative method that would allow us to detect patterns in large data frames is cluster analysis, and it is used in this paper. Cluster analysis is a generic designation for a large group of techniques that can be used to create a classification. Such procedures results in empirically clusters or groups of strongly similar objects Dalson Filho and others, “Clustering Analysis for Political Scientists” in Applied Mathematics, Vol. 5 No. 15 2014.. Among several general types of clustering methods, hierarchical cluster analysis in particular has an advantage of not requiring a pre-set number of clusters for the analysis, which would not be feasible given the research goals of this paper. Hierarchical clustering creates a dendrogram based on dissimilarities between all each pairs of observations, starting with the least dissimilar. This also allows us to see a complete picture of the field, including sub-divisions and low-level relations between different prostitution regimes. While more advanced clustering methods, such as co-clustering, exist, their relatively low level of application in political science and lack of well-developed software suggests that something more reliable and familiar should be utilized, as long as the chosen method has all necessary properties. While hierarchical clustering requires prior methodological choices and its results still require interpretation in its relation to the theoretical classifications, its result is as close to giving us an inductive view of the current policy field. The data that used is legal norms codified primarily in binary format (“Is it legal to sell sex/buy sex/operate a brothel”), the collection methods vary from browsing official legislation, where available, to mentions in other research, to consulting regional experts, to broader sources including other broad overviews of prostitution policies. The scarcity of data even on such a broad level means that in some cases the exact information on certain legal norms will not be possible to obtain - sometimes due to practical limitations, sometimes due to the fact that the issue underlying the specific norm was not meaningfully discussed in that country. Because highlighting such situations as N/A would significantly compromise the dataset and because in most situations the status of a particular norm can be reasonably inferred from the known information on the country's broader prostitution policy, each variable has a procedure for that, outlined below. I use the Orange Data Mining as the software. Hamming distance is chosen as the distance measure as it is best suited for dealing with non-numeric and binary data (which all of clustering variables in this paper are). Hamming distance is based on the number of positions in a string of numbers that are different. Jaccard dissimilarity measure was considered as an alternative, but rejected because it eliminates 0-0 pairings of binary variables in its calculations - it is meant to eliminate features, traits or units that are absent within the comparison, but I cannot expect all variables to align in that way, given the complexity of many aspects of prostitution policy. It means that eliminating a certain type of pairing influences the dataset in a way that is does not improve the degree to which it deflect the policy field, leaving Hamming distance measure as the best option. Average linkage was chosen as the method of clustering - it is one of the most popular ones and the default setting of the software. The main alternative is complete linkage, which is based on the distance between farthest pairs of observations (and thus also known as the method of “most dissimilar”), After a clustering solution is reach I validate its results, randomly separating the dataset and clustering the sample datasets separately to compare it with the result of the full clustering. While this clustering method may give additional weight to outliers, it might be a beneficial trait for this research, as this paper is focused on picking out distinct policy types, which may be defined by outliers. policy prostitution protocol

3.2 Data gathering

The aforementioned methodology requires the creation of a dataset that would reflect the breadth of existing prostitution policy regimes and their distinct measures. These parameters do not make distinctions between levels of offences - while in some countries selling sex is a criminal offence and in others it is a “misdemeanor” or an “administrative offence”, the effect of outlawing a practice is broadly similar and the differences are not significant enough on this level of analysis. “Direct criminalization” is often used as a qualifier because many of these actions are done by the same people, meaning that criminalization of public soliciting or living off income of prostitution can be considered as indirectly criminalizing selling sex itself, but the premises of this research necessitate the distinction.

First and foremost, every prostitution policy begins with legal status of both sides - the prostitute and the client. This gives us the selling sex variable, where 1 means that it is not directly criminalized and 0 means it is directly criminalized, and buying sex variable, where 1 means that it is not directly criminalized and 0 means it is directly criminalized. In many countries, the legality of buying sex (in comparison to having non-commercial sex) is dependent on the status of the other side, e.g. its age or degree of outside coercion. This gives us two more parameters - buying sex from an underage person and buying sex from a trafficking victim (or a person otherwise subjected to force); for both of these parameters, 1 means that it is not directly criminalized, 0 means that it is specifically criminalized. If no clear information is found about such norms, it is considered not criminalized (1). While the latter parameter depends on country's anti-trafficking policy, it is a sector that in policy discussions is often adjacent or even conflated with prostitution policy Emily Shuckman “Anti-Trafficking Policies in Asia and the Russian Federation, a Comparative Perspective”, Demokratizatsiya, 14(1) (2006), 85-102., and the presence of these measures is additional evidence of that.

Secondly, prostitution policy has to concern itself with broader structure of the prostitution industry - its organization, promotion and involvement of third parties. The status of brothels gives us the Owning or managing a brothel parameter, where 1 means that it is not directly criminalized, 0 means it is directly criminalized. If no clear information is found about such norms, it is considered not criminalized (1). While owning and managing a brothel are two similar activities, the prohibition of any of them effectively outlaws brothels, which is the aim of both of them, making the distinction unwarranted for this analysis. Outside of brothels, organizational efforts by third parties are also often criminalized. Pimping is a colloquialism that describes a number of third-party activities, such as procurement, recruitment and organization of people involved in prostitution. While norms restricting these activities use different wording (and are often vague), they target the same actor - a non-prostitute third party involved in organizing the work, with or without a brothel. 1 means that this set of actions is not directly criminalized, 0 means that it is directly criminalized. If no clear information is found about such norms, it is considered not criminalized (1). Additionally, Living off income of prostitution, especially for a third party, can be criminalized. 1 means that it is not criminalized for any party, 0 means that it is directly criminalized. If no clear information is found about such norms, it is considered not criminalized (1). Some countries establish licensing systems for brothels or individuals prostitutes and make obtaining such a license a prerequisite for legal work. 1 means that such system is present, 0 means that such system is absent or is not a prerequisite to legal work, i.e. not directly enforced. . If no clear information is found about such system, it is considered absent (0).

Tertiary measures related to prostitution are numerous, but I found four that are important to highlight. Some countries establish Exit programs that give people involved for prostitution an opportunity to leave the industry and move into a different profession. In this parameter, 1 means that such programs are included into the official policy, 0 means that they are not included. If no clear information is found about such programs, they are considered not included (0). Presence of prostitution, in public areas, its proposal and negotiation is another concern that, it is occasionally restricted with Public soliciting laws for one or the other party. In this parameter 1 means that it is not criminalized for either party, 0 means that it is criminalized for one or both parties. If no clear information is found about such norms, it is considered not criminalized (1). Finally, in a number of countries regional or municipal governments have a significant degree of authority over the specifics of prostitution policy, including setting up designated zones or “red light districts” (making certain or all forms of prostitution illegal outside of those) or changing the status of prostitution entirely. While the specifics and complexity of such distinctions is outside of the scope of this research, existence of such systems might significantly alter the policy of a country, e.g. giving it de-facto legal status in spite of federal prohibitions. In Local differentation parameter 1 means that such authority is present, 0 means that such authority is absent. If no information about this differentiation is found, it is considered absent. All sub-units of countries that maintain separate prostitution regimes in countries that maintain separate regimes for its states or territories (as discussed at the beginning of the Methodology section) marked as having such a system (1). Another important parameter related to prostitution policy is immigration restriction against people who are suspected of planning to sell sex - their visa applications can be rejected, they can be refused entry at the border, sometimes this area is specifically excluded from work permits even if it's legal for permanent residents Nora Butler Burke, “Double Punishment: Immigration Penality and Migrant Trans Women Who Sell Sex” in Red Light Labour, ed. Elya M. Durisin et al. (Vancouver: UBC Press, 2018). . However, very few countries have explicit legal norms related to that, which means that properly quantifying this area means breaching the divide between policy and its application that was established as part of this research (a standard which would be impossible to apply consistently across different variables given on the scale of this research and the scarcity of detailed information about most countries). Additionally, the available information in that area is often limited to individual complaints, especially in non-Western countries, which additionally worsens the reliability of the data that would potentially be gathered. Given these reasons, no parameter related to immigration policy was created in the dataset, leaving us with the following 11 parameters:

Selling sex;

Buying sex;

Buying (underage);

Buying (trafficking);

Brothels;

Licensing;

Living off income;

Pimping;

Exit programs;

Public soliciting;

Local differentiation.

3.3 Data Overview

The resulting dataset contains 205 units from 198 countries (see Annex 1). Countries separated into sub-units include Great Britain (England and Wales, Scotland, Northern Ireland), China (China, Hong Kong, Macao) and Federated States of Micronesia (Yap, Kosrae, Chhuk and Pohnpei). The United States of America were not separated, because their prostitution policy is largely uniform over all 50 states aside from a number of counties within Nevada, putting the US in line with other countries where local authorities allow limited tolerance zones. Mexico and Indonesia, on the other hand, exhibited significant degree of local differentiation between their states and should, ideally, have been separated. However, due to the lack of detailed data about prostitution policies in 32 states of Mexico and 34 provinces of Indonesia they were treated as singular countries with existing local differentiation, with their federal legislation assessed via standard procedures.

While prostitution legislation of many countries was available online in the languages I had access to, a number of secondary sources was helpful to locate information on less accessible countries. These include The Map of Sex Work Law by Institute of Development Studies of the University of Sussex, The Global Mapping of Sex Work Laws by Global Network of Sex Work Projects and Trafficking in Persons (TIP) Report by the U.S. Department of State. While these sources applied different methodologies and classifications (and TIP report in particular treats prostitution legislation as a peripheral subject), the use of cross-referencing and double-checking with available primary sources made them useful supporting sources.

While the resulting dataset may involve a statistically inevitable minimum of erroneous assignments, especially given the availability of detailed policy information within the Global South, I believe that its proportion does not influence quality or robustness of the data analysis.

3.4 Data analysis

Applying aforementioned methodological decisions to the dataset, I have created the hierarchical dendrogram that can be found in Annex 2. The following groupings could be distinguished from the dendrogram by breaking it into increasing number of clusters by descending dissimilarity:

2 broadest clusters that can be very vaguely described as “suppressing” vs “regulating” prostitution (no relation to any aforementioned classification)

3 clusters, with 4 units (Dominican Republic, Philippines, Luxembourg and Nigeria) separating from the “suppressing” cluster. These countries can be characterized by both buying and selling sex being legal (with exception of buying sex from minors), criminalized pimping , brothels and living off prostitution income while keeping public soliciting legal. These countries also do not have any licensing systems.

4 clusters, with 9 units (Afghanistan, Georgia, Haiti, Liberia, Maldives, Nepal, Oman, San Marino, São Tomé and Príncipe) separating from the “regulating” cluster. The main distinguishing trait of this cluster is the ban on selling sex in combination with uneven but lax restrictions on other areas (1 country prohibits brothels, 2 countries criminalize living off prostitution income etc.) and a lack of licensing.

5 clusters, with 8 units (Sierra Leone, Slovenia, El Salvador, Colombia, Kosrae (FSM), Queensland (AU), Italy and Macao (C)) separating from the “suppressing” cluster. their main similarities being lack of bans on buying or selling sex, having both types of additional client restrictions, not allowing pimping or public soliciting and not having any licensing systems.

6 clusters, with 5 units (Canada, Sweden, Northern Ireland, France, Norway) breaking off from the “suppressing” cluster. These regimes are characterized by banning buying sex but not its selling, severe restrictions on organized prostitution (no brothels, no pimping, no living off prostitution income) and the existence of exit programs.

7 clusters, with 5 units (Tunisia, Aruba, Ecuador, Northern Territory (AU) and Victoria (AU) breaking off from the “regulating”cluster. These regimes are characterized by a combination of non-criminalized pimping and living off prostitution income, bans on public soliciting, a lack of additional client restrictions and the presence of some form of mandatory licensing system.

8 clusters, with 2 units (Sierra Leone and Slovenia) breaking off from the 5th cluster. These countries are distinguished from their main cluster by restrictions on buying sex from victims of trafficking and a lack of local differentiation.

9 countries, with 2 units (Germany and the Netherlands) breaking of from the “regulating” clusters, their main distinguishing feature being the combination of a licensing system and an exit program.

10 clusters, with the “suppressing” cluster breaking roughly in half, the main feature of the division being the legality of selling sex.

In this analysis, I mostly focus on the 6-cluster interpretation of the dendrogram, although the possibility of a different interpretation is inherent in our methodology. So, to reiterate, these six clusters are:

9 regimes that criminalize selling sex while leaving most of other areas of the industry non-criminalized. The closest theoretical model to that group is “prohibitionist” criminalization mentioned in Pajumets's and Mossman's classifitcation. The notable artifact of this cluster is its broader grouping with cluster 2 despite lack of any legal avenue for prostitution's existence. The most reasonable explanation for that is that their lack of restrictions on organized prostitution (since these countries rely on the criminalization of selling sex to criminalize the entire industry) puts them closer to cluster 2 than to other prohibitionist regimes, that typically criminalize third parties' involvement.

30 regimes that decriminalize selling and (all) buying of sex , generally have low restrictions on third parties (pimping, brothels and living off prostitution income), with many having some kind of a mandatory licensing system. Notable exceptions include Tunisia and Grenada - selling sex is generally illegal there, but there are licensing systems, which put them close to the countries where selling sex is technically legal, but you need to go through licensing to do that. The closest theoretical model to this cluster is Östergren's “Restrictive” model, since both “decriminalization” and “legalization” used by Decker and Mossman leave some countries of this cluster uncovered and there is no significant divide between the two.

4 regimes that decriminalize buying and selling sex (while putting additional restrictions on the former), leave public soliciting legal while universally criminalizing third parties' involvement (brothels, pimping and living off prostitution income). This cluster fits Mossman's “Decriminalization”, McCarthy's “Partical decriminalization” and Decker's “Legalization”. These countries clearly allow for individual sex work to be practiced while trying to criminalize areas that are more likely to involve violence or abuse.

8 regimes that decriminalize both buying and selling sex, leave living off prostitution income legal while criminalizing third parties' involvement (all countries in this cluster ban either pimping, brothels or both). Similar theoretical models to cluster 3 can be applied here, although there seems to be less tolerance to prostitution's public presence.

5 regimes that decriminalize buying sex but prohibit everything else (additional restrictions on buying are absent, but the general buying ban covers these actions anyway) while introducing exit program. The closest theoretical model to that is Pajumets's “Neoabolitionism” and Mossman's “abolitionist criminalization”. This is the cluster of the “Nordic Model” - all of these regimes publicly subscribed to that approach during the adoption of their current policies. It is worth noting however that 3 other countries known to self-describe as belonging to that model - Ireland, Iceland and Israel - did not end up in that cluster. It seems that the absence of exit programs brought these regimes closer to cluster 6, and a mere combination of criminalizing buying and criminalizing selling sex is not sufficient to be distinguished from them.

149 regimes (the largest cluster by far) whose most consistent features include bans on brothels and public soliciting, criminalization of either buying, selling of both, a lack of additional buying restrictions, exit programs or licensing systems. The overwhelming majority of countries that criminalize selling sex ended up here, with only 3 such countries (Tunisia, Grenada, Philllipines) belonging to clusters other than this and cluster 1. However, another common subtype is a type of regime that prohibits everything except for selling and (all) buying, common both in the Global South (especially the British Commonwealth) and in Europe (this is the divide that appears in the aforementioned 10-cluster interpretation). The only theoretical model that adequately describes this cluster is Östergren's “Repressive” approach. McCarthy's classification was another classification that could potentially handle it, with “Full criminalization” and “Partial criminalization” covering the sub-clusters that appear in the 10-cluster interpretation. However, both of these models - “Partial criminalization” in particular - can also be applied to most of the other clusters.

While this interpretation give us a lot of interesting information, its validity needs to be assessed before I make any conclusive statements. To do that, I randomly sampled 70% of the dataset (144 units) and ran the same clustering procedure, repeating this process a number of times (see Annex 3 for example dendgrograms, with the clustering cutoff set at the degree of dissimilarity set at the level similar to the main interpretation). While clusters similar in characteristics in composition to some of the smaller clusters (most notably cluster 1) seem to appear in numerous iterations, the only two clusters of the main interpretation that are consistent in all versions are clusters 2 and 6 of the main interpretation, with different groups of regimes breaking off with differing degrees of consistency. This, along with an overall low degree of dissimilarity between clusters (the distance value of 0.31 being the cutoff level for the 6-cluster interpretation) doesn't allow us to conclude that the 6-cluster interpretation (or any other with the same degree of specificity) represents consistent separate models of prostitution policy that exist in reality.