Tag: asymmetric punishment

The Nordic Model of Prostitution Legislation: Health, Violence and Spillover Effects

20190422 The Nordic Model of Prostitution Image 01

An emerging literature is studying, with the help of new types of data and clever identification strategies, the effects of different legislative measures regulating the market for sexual services. The primary target of such measures are arguably the participants in the market, prostitutes and their clients, and law and order concerns in their immediate vicinity. In a new research project, we mean to shift the spotlight on potential broader spillovers from these policies, both to other outcomes and other countries. In their presence, we cannot understand the full impact of a law change if we limit our analysis to the prostitution market in that country alone. We focus on a particular model of prostitution legislation, first adopted in Sweden in 1999 and known since as the Nordic model.

The Nordic model

The debate on prostitution legislation shares clear similarities with the standard arguments put forward for or against alcohol prohibition or drug liberalization. The criminalization of an activity is most likely shrinking the corresponding market, because it increases the cost of participation. It also functions as a signal of what a society deems acceptable or not, and coordinates behavior to potentially change social norms. At the same time, however, it pushes the remaining market into the darkness, where criminal activity potentially increases. In the specific case of the prostitution market, what is particularly feared is an increased risk of violence and general worsening of conditions for the potentially fewer sex workers.

When, in 1999, Sweden enacted the first asymmetric criminalization of prostitution, whereby buyers but not sellers of sexual services are punished, a third way between criminalization and legalization seemed to appear. This legislation would still give a clear signal on societal values, but at the same time protect the, in large part female and in large part exploited, sex workers. The model proved very successful in deterring street prostitution, and, under the catchy name of the “Nordic model”, has subsequently been adopted by Norway, Iceland, Canada, and France. It is currently under consideration in further countries as well.

This is where most reports and policy evaluations stop. In a new project at SITE, involving an international research cooperation, we propose to investigate the impacts of this legislation beyond the participants in the prostitution market. Specifically, we encompass other outcomes such as gender-based violence, health outcomes and online behavior, both within Sweden and other countries that implemented the reform, but also, most importantly, across their borders. The idea is that law changes in one country may also affect the demand and supply of prostitution in other countries, especially but not exclusively those bordering the country that enacts the law change. Two possible channels for such cross-border effects are sex tourism and human trafficking.

This brief summarizes the preliminary evidence we collected so far.


The focus on the role of policies is a recent but rapidly growing addition to the economic literature on prostitution. The risk of violence, both for the participants and within the neighboring geographic areas, is a natural area of concern for policy in relation to the sex market, and to criminal activities in general. To improve on cross-country comparisons and draw causal links from policies to outcomes, the most robust contributions in this area focus on natural experiments. Cunningham and Shah (2018) study an unintentional, and therefore unexpected and temporary, decriminalization of indoor prostitution in Rhode Island, and find that reported rape offences fall by 30%. Cunningham and coauthors (2019) also look at the geographic expansion of the erotic services section of Craigslist, a popular advertisements website, before online solicitation was banned in 2018. The possibility to use online platforms for their work, by allowing prostitutes to keep mostly indoors, and screen their potential clients to a larger extent, appears to have been very beneficial: the study finds lower female homicide rates by 10-17% when and where the service was available. Ciacci and Sviatschi (2018) find that the opening in a neighborhood of indoor prostitution establishments decreases sex crime by 7-13%, with no effect on other types of crime, arguing that the reduction is mostly driven by potential sex offenders resorting to the establishments, instead, to satisfy their needs. What is common to these studies is the finding that allowing the sex market to exist in some form is beneficial for outsiders, while indoor prostitution is safer for the sex workers themselves.

Preliminary findings from our project (Berlin et al., 2019 a) are consistent with this. We base our strategy on a comparison, within Sweden, between counties that are above or below average in terms of representation of women among police force and elected officials (we refer to them as treated and control counties, respectively). Both these indicators have been found in previous studies to drive greater reporting and lower incidence of crimes against women (Iyer et al., 2012; Miller and Segal, 2018). Looking at population-wide rates of violence against women in Sweden, we observe an increase in assaults committed by acquaintances indoors by about 10% and an increase in rapes indoors by more than 20% in treated as compared to control counties. Since the reform is argued to have eliminated street prostitution, and pushed the remaining sex trade indoors, violence against prostitutes will be counted in the indoor assaults statistic. However, in treated counties, where we observe the increase in violent crimes against women, we at the same time find fewer convictions for buying sex. We argue therefore that the increase in assaults we observe is not likely in the context of the sex market, but rather indicates increased violence against non-prostitutes from frustrated former customers, in other words a negative externality of deterring prostitution. In order to distinguish whether this increase is only in reported or actually committed crimes, we look at hospitalizations of women for injuries that are related to sexual interactions. If we think that seeking hospital care is less sensitive than reporting a violent man to the police, the series of hospitalizations should be closer to the true violence than the convictions. Although numbers are small and differences not significant, hospitalizations spike up in treated counties directly after the reform, as Figure 1 shows. All in all, our preliminary evidence from Sweden suggests that intimate partner violence and violence on women in general might have increased as a consequence of the “Nordic model”.

Figure 1. Hospitalizations of women

Source: Hospitalizations of women for injuries related to sex, from Berlin et al. (2019 a).

Other outcomes

Besides violence, health outcomes are also a policy relevant objective with the regulation of prostitution. Indicators such as the spread of sexually transmitted infections serve the double purpose of giving a rough indication of the changes in the size of the sexual market while at the same time enabling inference on the work environment and general living conditions for prostitutes. In a companion paper, which is underway, we examine these statistics for Sweden and Norway, in terms of within country changes but also with a mind to capture potential cross-border spillovers between the two countries.

Cross-border spillovers

In another working paper (Berlin et al., 2019 b) we study the reform enacted in France on April 13th, 2016, which removed the punishment for solicitation of prostitution (previously set to two months imprisonment plus a fine) and introduced instead a range of fines for the purchasing of sexual services, thereby, pushing the punishment to the side of the buyer. In order to study the cross-border effect of this change, we focus on the German Bundesländer bordering France: Baden-Württemberg, Saarland and Rheinland-Pfalz. The national law in Germany generally allows prostitution, but gives federal states the right to regulate it on a more detailed level. This generates variation at the level of the Gemeinde, the administrative division corresponding roughly to a municipality. The idea behind our analysis is to compare municipalities where prostitution is at least in part allowed with municipalities where it is banned (we refer to them as treated and control municipalities, respectively). Our preliminary results show that foreign tourism to cities where prostitution is at least partly legal increased after the reform more than to those completely overlapping with a Sperrbezirk, i.e. an area in which prostitution is banned. However, so does domestic tourism. This might be seen as a threat to our interpretation, since we can’t connect this increase directly to the French reform, unless we can show that there is a dynamic adjustment of the supply of sexual services, which also attracts domestic flows. We can’t isolate tourism from France in this data, so we go a step further by looking at online behavior.

Google searches

A key contribution of this project is to gather new data that haven’t been analyzed to date in the existing literature. In particular, we collected detailed data on Google searches originating in France using as keywords different German cities. The idea is to capture potential deviations of search trends over time driven by prostitute customers who after the legislative change find it more attractive to look for sexual services across the German border. Preliminary findings show that after the policy change there is a larger increase in search activity for cities closer to the French border relative to cities further away. While searches are generally downward trending over time, the trend is slowed after the French reform, and this effect is stronger the closer a city is to the border, although intermittently significant. Figure 2 reports the differential increase in searches (with 95% confidence intervals) as related to the distance from the border. The negative relationship between size of the impact and distance to the border is consistent when controlling for city and time fixed effects. However, further analysis is needed in order to validate the results and control for confounding factors.

Figure 2. Google searches for German cities before and after the French reform

Source: Google Search data on searches originating in France for cities closer to VS farther from the German border than the indicated distance (in km).

We are currently repeating the same exercise at the French borders with Belgium and Spain, with searches originating in Norway around the time of the Norwegian reform (2009), and at the US-Canada border around the time of the Canadian reform (2014).


When adopting a version of the Nordic model in 2014, the Canadian Department of Justice stated that the “overall objectives [of the reform] are to:

  • Protect those who sell their own sexual services;
  • Protect communities, and especially children, from the harms caused by prostitution; and
  • Reduce the demand for prostitution and its incidence.”

Research seems to show that restrictions on the sexual services market, rather than the sex trade itself, have substantial negative impacts on communities and sex workers. Nevertheless, it is understandable that legislators in many countries, sharing similar concerns and expectations as expressed by the Canadian DoJ, find it unattractive to legalize prostitution. What our project points to, then, is that when considering various forms of criminalization, it is crucial to understand how best to pursue each of these objectives. Taking into account side-effects, or spillovers, such as the ones we highlight above, might reveal the need for complementary policies, in order to avoid unexpected and counterproductive consequences.


Disclaimer: Opinions expressed in policy briefs and other publications are those of the authors; they do not necessarily reflect those of the FREE Network and its research institutes.

Leniency, Asymmetric Punishment and Corruption: Evidence from China

20150928 FREE Policy Brief Leniency, Asymmetric Punishment Image 01

Since coming into office two years ago, Chinese President Xi Jinping has carried out a sweeping, highly publicized anticorruption campaign. Skeptics are debating whether the campaign is biased towards Mr. Xi’s rivals, and even possibly related to the current economic slowdown. What is less debated is the next stage of Mr. Xi’s anti-corruption strategy, which is going to alter the legal statutes. Amendment IX, proposed in October 2014, includes heavier penalties, but two important tools in the fight of corruption – one-sided leniency and asymmetric punishment – became more limited and discretional. We argue that studying a 1997 reform and its effects can shed some light onto why the Chinese leadership seems dissatisfied with the current legislation and the likely effects of the proposed changes.

What We Know about Leniency

In our context, leniency can be defined as the concession of reduced sanctions (or full immunity) to wrongdoers that cooperate by self-reporting and providing information against former partners in crime. Formal and informal exchanges of leniency against information and collaboration are normal features of law enforcement in most countries. Policies of this kind have been extensively and quite successfully used to fight the Italian and American mafias, drug dealing and other organized crimes, and have become the main instrument to fight collusion in antitrust since the US reform in 1993 (see Spagnolo, 2008).

For crimes in which multiple offenders cooperate, one-sided leniency conditional on being the first to self-report can be a very powerful tool of law enforcement: by playing the partners in crime against each other, it may elicit information, greatly facilitate prosecution and generate deterrence at a very low cost. A conspicuous scientific literature with theoretical, experimental and empirical contributions shows the great potential of these policies, when properly designed and administered, for deterring collusive crimes (Miller 2009; Spagnolo 2008; Bigoni et al. 2012, 2015). On the other hand, Buccirossi and Spagnolo (2006) show specifically for the case of corruption that, when poorly designed or administered, these same policies may become ineffective or even counterproductive.

Asymmetric Punishment

A related way of using leniency towards one party (to play it against the other) in the fight against corruption has been at the center of a recent intense policy debate after the popular note “Why, for a Class of Bribes, the Act of Giving a Bribe Should Be Treated as Legal”, by Kaushik Basu (2011). Then chief economist of the Indian government and now of the World Bank, Basu advocated asymmetric depenalization of bribe giving, which can be thought of as a form of unconditional, one-sided leniency. More precisely, the note proposed to legalize bribe giving in the form of harassment bribes (also called extortionary, or discharge-of-duty bribes) paid to obtain something one is entitled to, while strengthening sanctions against bribe taking. As with other forms of leniency, the idea is to create a conflict of interests between the partners in crime by increasing the temptation for one party to betray and report the illegal act, leading to a severe punishment of the other.

In the debate sparked by this note many different arguments have been put forward, both against it and in favor of it. Dufwenberg and Spagnolo (2015) discuss formally some of the issues raised by critics of the proposal, while Abbink et al. (2014) provide (mixed) experimental evidence on its effectiveness. Later, a blogpost by a Chinese law scholar, Li (2012), attracted our attention to the case of China, where asymmetric punishment (bribe-giver impunity) has been in place since 1997. She argued, probably reflecting the political debate in the country rather than based on factual evidence, that the system had not been successful. We felt this claim granted a deeper investigation into the details of the Chinese legal reform and the changes it introduced, and of course a careful inspection of the data to back it.

A Study in Red

In a new working paper, Perrotta Berlin and Spagnolo (2015), we set out to understand the evolution of the anti-corruption legislation in China over the last decades, and then to evaluate the effects of the policy changes occurring in 1997. Two new elements were given the strongest legal status in 1997: leniency for wrongdoers that self-reported and cooperated with investigators, and asymmetric punishment (no charge for bribe givers) for bribes paid to obtain something one was entitled to. Concurrently, penalties were decreased, in particular for bribe-takers.

To understand the likely effects of this policy change we would ideally look at correspondent changes in corrupt transactions. Data on the prevalence of bribery, however, are notoriously hard to come by because of the secretive nature of this activity. Instead, we use several data sources which capture on the one hand actual corruption cases tried in courts, and on the other hand surveys of corruption perceptions. In particular, we have collected the number of arrests and public prosecutions on the counts of corruption and bribery from the Procuratorates’ Yearly Reports for each Chinese province since 1986.

It is not straightforward to infer changes in total corruption, which is unobserved, from changes in discovered cases tried in court. The data on prosecutions mix together corruption and anticorruption activities, as they fail to distinguish occurrence of the criminal activity from detection. A policy that deters crimes but at the same time increases the fraction of those that are successfully prosecuted will have an ambiguous effect on the number of prosecutions. We adapt for this purpose the testable predictions developed by Miller (2009): he models the occurrence of criminal activity (cartel formation, in this case) and derives predictions for how changes in the rate of occurrence and the rate of detection affect the time series of detection.

The preliminary evidence we have so far points to a substantial and stable reduction in the number of major corruption cases around the 1997 reform, a result consistent with a positive deterrence effect of the 1997 reform. The evidence is suggestive, and some alternative interpretations of the patterns in the data, shown in the plot below, cannot be excluded at the moment. While a peak-and-slump pattern as in Miller (2009) would have been much stronger evidence supporting the success of the reform at deterring corruption, we cannot exclude that the drop in prosecutions is simply due to a general worsening in detection. Although we deem this unlikely in the light of the general political climate of the time, we need more and better data to support our interpretation. Still, claims that the reform did not have an effect appear not supported by the data.

Figure 1. Change in Corruption Prosecutions before and after law reform in 1997

MariaGiancaPicSource: Perrotta-Berlin and Spagnolo (2015).

More to be done

A case study analysis is under way to corroborate and help the interpretation of these preliminary findings. We will analyze in depth a stratified random sample of prosecution case files between 1980 and 2010. Given that we sample a given number of cases, in this part of the analysis we cannot gain any insight about the incidence of bribery in general. We can instead observe the impact of the legislative reform on specific details of the corrupt behavior, and the mechanisms through which this behavior occurs or is deterred. In particular, we will be able to distinguish between cases of extortionary (harassment) bribes and bribes paid to obtain illegitimate benefits. Moreover, this will allow us to shed light on whether and how leniency and asymmetric punishment were applied in practice. The details of the case files might even allow us to gain insight into how the bribe-size and the value of corrupt deals evolved through the reform and even the selection into bureaucracy.


One-sided leniency, conditional on reporting an act first, or unconditional, as when bribe giving is depenalized, may be powerful corruption deterrence instruments if well designed and implemented in the right environment, but may also have negative effects. It has been argued that these instruments have been ineffective in China, after they were reformed in 1997, however, without data supporting the claim. Part of the reason lies in the difficulty to obtain good data on corruption. Another obstacle is the subtlety of interpreting them when they relate only to detected and convicted cases, rather than to the whole population of corruption cases.

We cannot solve completely the issue of data quality, as we also need to rely on official reports of counts of corruption cases. However limited, the exercise performed on aggregated data clearly shows that the 1997 Criminal Law reform did have an effect, consistent with increased corruption deterrence. To further support this finding we will collect and analyze micro-data from a randomized sample of these cases. This will allow us to isolate at a higher level of detail the changes in criminal behavior, reporting behavior and prosecution activity, and link them to the details of the legal reform to highlight the mechanisms at work.

China is home to a sixth of humanity, and currently undergoing a massive crackdown on corruption. Whatever we can learn about the effectiveness of their past and present anti-corruption policies is likely to have considerable welfare effects. Moreover, the 1997 reform was the object of a policy debate, and comments on its effectiveness came without data to support them. We believe our effort to use data to shed light on what this reform actually changed will be a valuable input to further research and policy discussion on this important topic.


  • Abbink, K., U. Dasgupta, L. Gangadharan, and T. Jain. “Let-ting the Briber Go Free: An Experiment on MitigatingHarassment Bribes.” Journal of Public Economics, 111,2014, 17–28.
  • Basu, K. “Why, for a Class of Bribes, the Act of Giv-ing a Bribe Should Be Treated as Legal.” WorkingPaper 172011 DEA, Ministry of Finance, Governmentof India, 2011
  • Bigoni, M., S.-O. Fridolfsson, C. LeCoq, and G. Spagnolo.“Fines, Leniency and Rewards in Antitrust.” RANDJournal of Economics, 43, 2012a, 368–90.
  • Bigoni, M., S.-O. Fridolfsson, C. LeCoq, and G. Spagnolo.. “Trust and Deterrence.”. Journal of Law, Economics, and Organization (2015)
  • Buccirossi, P., and G. Spagnolo. “Leniency Policies and Ille-gal Transactions.” Journal of Public Economics, 90,2006, 1281–97.
  • Buccirossi, P., Marvão, C. M. P., & Spagnolo, G. (2015). Leniency and Damages. Available at SSRN 2566774.
  • Dufwenberg, M. and Spagnolo, G., Legalizing Bribe Giving (April 2015). Economic Inquiry, Vol. 53, Issue 2, pp. 836-853, 2015.
  • Li, X. Guest post: bribery and the limits of game theory – the lessons from China. http://blogs.ft.com/beyond-brics/2012/05/01/guest-post-bribery-and-the-limits-of-game-theory-the-lessons-from-china/, 2012. Accessed: 2015-05-20.
  • Miller, N. H. Strategic leniency and cartel enforcement. The American Economic Review, pages 750–768, 2009.
  • Perrotta Berlin, M. and G. Spagnolo, Leniency, Asymmetric Punishment and Corruption: Evidence from China, SITE Working Paper, 2015 (forthcoming)