Location: United States

Would Electing More Women Make the U.S. Congress Less Polarized?

20240512 Would Electing More Women Image 03

With growing ideological polarization in the electorate and among U.S. Congress members, the view that electing more women would help solve partisan gridlocks has also grown especially popular. In this policy brief we review recent evidence on gender differences in cooperative behavior among legislators and argue that the prediction that a more female U.S. Congress would be less polarized does not find strong support in the data. While, in the past, Republican women have cooperated more with Democrats than their male colleagues we find evidence that this was due to higher ideological proximity between Republican women and Democrats rather than gender per se. Among Democrats, women actually appear to cooperate less with the opposite party than their male colleagues. Moreover, in recent years gender differences in ideology among Republicans have been narrowing, which also reduce gender differences in cooperation with the opposite party.

Gender Differences in Cooperative Behavior

Observers of U.S. politics have repeatedly reported increasing polarization in the U.S. electorate and Congress over the last decade, with growing concerns that the partisan gridlocks that have impaired Congress’ activities in the last two years will only grow after the 2024 elections. At the same time, it is widely believed that electing more women to the U.S. Congress would help reduce partisanship among legislators and promote cooperation across party lines. For instance, a report by the Center for American Women and Politics found that “collaboration by women across party lines is often fostered by participation in bipartisan, single-sex activities […] which can lead to policy collaboration” (Dittmar et al. 2017). These beliefs are rooted not only in anecdotal evidence but also in academic studies that, through laboratory experiments, have shown that women tend to cooperate more than men (cooperation is considered as working in a team to achieve a common good). However, this finding is not universal across settings and studies (Balliet et al. 2011), which suggests some caution in foreseeing fewer partisan gridlocks when more women are elected. Moreover, while laboratory experiments are a very important tool to discover patterns of human behavior in “ideal” conditions, testing for the robustness of experimental findings in real-world settings is a necessary step to draw definite implications for society-level outcomes.

What then is the research-based evidence on women’s willingness to cooperate with opposite parties as legislators?

Do Women in the U.S. Congress Cooperate More With the Opposite Party Than Men?

The proportion of women in Congress continues to be low, currently standing at 29 percent of the House of Representatives and 25 percent of the Senate. However, women’s representation has massively increased over time, especially since the 101st Congress, which was elected in 1989 (see Figure 1). This change has prompted researchers to investigate the effects of women’s different approaches to competitive and cooperative situations on the day-to-day working of Congress.

In examining the dynamics of legislative cooperation, contrasting viewpoints shed light on the role of gender in policymaking. Volden et al. (2013) find that women’s increased cooperativeness especially helps female lawmakers from minority parties who are able to sustain their bills throughout the legislative process, while more obstructive Congress members fail to find consensus. Offering an alternative explanation, Anzia and Berry (2011) show that female lawmakers indeed sponsor and co-sponsor more bills than male lawmakers but argue that this is due to only the best and most ambitious women entering Congress due to discrimination.

Figure 1. Women in Congress over time.

Source: Bagues et al. (2023), data from the Congressional Research Service.

This early work highlights the importance of studying gender differences in Congress overall and by party, while comparing women and men who have similar characteristics and are elected in comparable districts.

In a recent study, Gagliarducci and Paserman (2022) adopt several empirical strategies to assess the extent to which largely comparable women and men in Congress behave differently in terms of cooperativeness. Their measure of cooperation is the number of co-sponsors that women and men respectively attract on their bills, and what share of these co-sponsors that are from the opposite party. Each bill presented to the U.S. Congress has a main sponsor and can have an unlimited number of co-sponsors. These co-sponsors attract support for the bill and aid its passage through the necessary legislative steps. Gagliarducci and Paserman (2022) consider bills proposed to the U.S. Congress between 1988 and 2010 and find that among Democrats there is no significant gender gap in the number of co-sponsors recruited, but women-sponsored bills tend to have fewer co-sponsors from the opposite party. On the other hand, they establish robust evidence that Republican women recruit more co-sponsors and attract more bipartisan support on their bills than Republican men. They conclude that this pattern indicates that cooperation is mostly driven by a commonality of interest, rather than gender per se. This since during this period female Republican representatives were ideologically closer to Democrats than their male colleagues, whereas Democratic women were ideologically further away from Republicans. They proxy representatives’ ideology using information on the ideological leaning of voters in representatives’ constituency in the presidential elections. As the authors observe, these findings challenge the commonly held view that an increase in female representation in the US Congress would help solve partisan gridlock.

In a recent working paper (Bagues et al. 2023), we assess the replicability and reproducibility of these findings, given their practical relevance in the face of the upcoming 2024 Congress elections. Our work is part of a large effort promoted by the Institute for Replication to improve the credibility of social science by systematically reproducing and replicating research findings published in leading academic journals.

Using the same data and empirical strategies as in Gagliarducci and Paserman (2022), except for correcting for some data collection errors and proposing different assumptions on the empirical specifications, we virtually confirm all their original findings. Most importantly, we also extend the analysis to cover 2011-2020 to study gender differences in legislative cooperation in a context that differs in at least two relevant aspects. During this period the share of women in the House of Representatives became substantially larger and, moreover, within-party gender differences in ideology changed compared to previous decades. While Democratic female representatives are still less conservative that Democratic men, women became ideologically more similar to their male colleagues among Republicans. We reach this conclusion by proxying representatives’ ideology using information on the ideological leaning of voters in representatives’ constituency in the presidential elections, as in Gagliarducci and Paserman (2022).

Consistent with the hypothesis that gender differences in cooperation across parties are driven mainly by ideological distance, we observe that bills sponsored by female Democrats are less likely to have opposite party co-sponsors than bills sponsored by male Democrats. We also, do not observe any gender differences in bipartisan cooperative behavior among Republicans. Finally, we observe more robust evidence that during the last decade bills from both Republican and Democratic women attracted more sponsors than bills from their male colleagues.

In sum, the novel evidence from the 2011-2020 period strengthens the finding that cooperation with members of the other party is driven mainly by ideological proximity rather than gender per se.


We have reviewed the recent academic literature on gender differences in willingness to cooperate among legislators, considering the largely popular view that a more female U.S. Congress would be less polarized and thus face fewer partisan gridlocks. Such a view is particularly salient at a time of increased polarization in U.S. politics and growing representation of women in the U.S. Congress.

Overall, studies of the extent to which bills promoted by women and men in Congress attract co-sponsors from members of the opposite party invite caution in predicting fewer gridlocks from the election of more women. Women legislators do not appear to be inherently more willing to cooperate with the opposite party. Gender differences in cooperation noticed in the past seem to be mainly driven by Republican women being more likely to legislate with Democrats because of a higher degree of ideological proximity to the opposite party compared to their male colleagues. However, analysis of recent data also show that Republican women have become ideologically more aligned to their male colleague in the last decade. This suggests that as the share of women in Congress increases, their characteristics and ideological standing might also change, making it hard to predict patterns of future behavior based on the past.


  • Anzia, Sarah F., and Christopher R. Berry. (2011). The Jackie (and Jill) Robinson effect: Why do congresswomen outperform congressmen? American Journal of Political Science 55, no. 3. pp. 478-493.
  • Bagues, Manuel, Pamela Campa, and Giulian Etingin-Frati. (2022). Gender Differences in Cooperation in the US Congress? An Extension of Gagliarducci and Paserman. No. 75. I4R Discussion Paper Series, 2023.
  • Balliet, Daniel, Norman P. Li, Shane J. Macfarlan, and Mark Van Vugt. (2011). Sex differences in cooperation: a meta-analytic review of social dilemmas. Psychological Bulletin 137, no. 6. p. 881.
  • Dittmar, Kelly, Sanbonmatsu, Kira, Carroll, Susan, Walsh, Debbie, and Wineinger, Catherine. (2017). Representation Matters: Women in the U.S. Congress. Centre for American Women and Politics, Rutgers University.
  • Gagliarducci, Stefano, and M. Daniele Paserman. (2022). Gender differences in cooperative environments? Evidence from the US Congress. The Economic Journal 132, no. 641, pp. 218-257.
  • Volden, Craig, Alan E. Wiseman, and Dana E. Wittmer. (2013). When are women more effective lawmakers than men?. American Journal of Political Science 57, no. 2. pp.326-341.

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.

It’s a Bird, It’s a Plane, It’s Superman! Using Mass Media to Fight Intolerance

20240116 It’s a bird Image 01

Joseph F. Gomes, an assistant professor at the School of Economics, Louvain, will present and discuss his working paper titled ‘It’s a bird, it’s a plane, it’s Superman! Using Mass Media to Fight Intolerance.


This paper examines the impact of progressive radio programming on societal change during the early period of desegregation in post-World War II U.S. We investigate the influence of the popular radio show “The Adventures of Superman” on promoting tolerance and exposing the bigotry of the Ku Klux Klan (KKK) in 1946. Using state-of-the-art radio propagation models, we map the broadcast’s exposure and analyze its effects on various socioeconomic outcomes. We find that counties with higher exposure to the broadcast experienced a significant decrease in support for segregationist political candidates. Individuals potentially exposed to the Superman program during their childhood exhibited more progressive attitudes towards racial desegregation and African Americans later in life. These individuals were also more likely to be in interracial marriages and less likely to participate in the Vietnam war. Additionally, we uncover significant and progressive long-term effects of the radio coverage on county-level outcomes such as the presence of active KKK branches, civil rights organizations, and accessibility of non-discriminatory services for African Americans listed in the “Green Books.” These results underscore the potential of progressive radio programming as a catalyst for social change and contribute to our understanding of how media shapes societal attitudes and beliefs.

About the Speaker

Joseph Flavian Gomes is a (tenured) Assistant Professor in Economics at the Economics School of Louvain and Research Director at the Institute of Economic and Social Research (IRES) at UCLouvain. Furthermore, Professor Gomes is also a CEPR Research Affiliate (International Trade and Regional Economics program, Development Economics program), and an Invited Researcher at J-PAL. His primary research interests are in the fields of Development Economics and Political Economy.

Join the Seminar

Interested in attending the SITE Brown Bag seminar at SSE or online via Zoom? The link to the seminar will be distributed by invitation only. If you are interested in attending the seminar – please contact site@hhs.se. Follow the instructions below:

  • Type the subject box with “Brown bag seminar *INSERT SEMINAR TITLE*”
  • Indicate your affiliation and field of interest
  • Please also indicate if you want to attend in person or online

For registered applicants, a Zoom link will be provided prior to the event via email with further instructions.

Environmental Enforcement in the EU: Insights from Administrative Cases in the US

20230523 Environmental Enforcement Image 01

In March 2023, the European Parliament’s legal affairs committee voted unanimously in favor of a proposed update to the EU Directive on environmental crimes (Directive 2008/99/EC). The update seeks to step up enforcement of environmental legislation across Members States through criminal law aimed at severely punishing very serious environmental offenses. We argue that, while laudable in its goal of strengthening enforcement of environmental regulation at the EU level, the current effort might be insufficient since moderately serious offenses might remain largely unpunished. To address this shortcoming, we propose harmonizing administrative law as well. We consider additional benefits from relying on administrative law in terms of flexibility of punishment design, based on the US experience of using environmentally beneficial projects performed in affected areas as a form of punishment in administrative environmental settlements. We discuss evidence on the merits and potential limitations of the US approach based on Campa and Muehlenbachs (2022) and conclude that such an approach is worth considering in the EU context.

While the EU has set aggressive pollution reduction targets across its Member States (European Commission, 2021a), for example pledging to reduce deaths due to particulate matters to 55 percent of 2005 levels by 2030 (European Commission, 2023a), much work remains to be done. As documented in Lehne (2021), in 2020 all countries in Europe reported PM2.5 concentrations above the World Health Organization (WHO) guideline of 5mg/m3. Six countries, including three EU Member States (Italy, Croatia, and Poland) reported levels above the EU’s annual limit value of 25mg/m3. Further, Bulgaria, Poland, Portugal, Croatia, and Romania did not meet national targets for PM2.5 reduction (European Environment Agency, 2023). Main contributors to PM2.5 pollution are transportation and industrial activity, including energy production. High concentrations of these particles are known to increase physical and mental health risks (Persico, 2022; Persico et al., 2016), and risk of premature deaths (Fuller et al., 2022).

Environmental concerns across EU Member States are also not limited to air pollution. Across the EU, 28 percent of groundwater sources are affected by pollution from agriculture, 14 percent from industrial contamination, and 7.5 percent from mining waste (Kampa et al., 2021). The persistent pollution problems in the EU and their unequal distribution across regions despite growing EU-level environmental legislation underscores the importance of law enforcement. While all EU Member States are theoretically subject to the same overarching environmental standards and regulations, the enforcement of environmental laws differs widely across countries. To address this issue, the EU Commission (henceforth EC) has recently taken steps to further harmonize environmental enforcement across EU Member States.

In this brief we consider the EC’s proposal and argue that, while commendable in the goal of strengthening enforcement of environmental regulation at the EU level, it is also quite limited in terms of enforcement tools that it considers. Specifically, we discuss potential advantages of leveraging administrative law tools to enforce environmental regulation, whereas the EC approach is currently focused on criminal law. We consider the higher probability of prosecution and the enhanced flexibility in the type of penalties allowed by administrative enforcement actions. Finally, we discuss results from Campa and Muehlenbachs (2022), which studies the use of administrative penalties for environmental violations in the US and draws some lessons for environmental enforcement in other jurisdictions.

Strengthening Environmental Enforcement at the EU Level

While environmental regulation is a shared competence of the EU, enforcement has historically been left to national environmental authorities (European Parliament, 2016). In the face of a lack of institutional capacity at the national level, a result of this arrangement are generally low levels of environmental enforcement, widely heterogeneous across Member States (Mazur, 2011). EU institutions have tried unsuccessfully over time to address this challenge and harmonize enforcement across EU Member States. An early attempt was made in 2001, when the EU put in place minimum standards for environmental inspections that Member States carry out, though these were only non-binding guidelines, and Member States could not be sanctioned for flouting them (European Parliament, 2001). Mandatory standards were then introduced in 2008, with the EU Directive on environmental crimes (Directive 2008/99/EC), which forced national governments to apply criminal sanctions to those causing “substantial damage” to the environment. However, it has typically been difficult for the EC to sanction non-abiding Member States. Moreover, the obligation is limited to areas where the EU has competence and does not include minimum penalties.

In another attempt to step up their enforcement efforts, in 2016 the EC began publishing the annual Environmental Implementation Review, where each country is evaluated on its environmental affairs and enforcement (European Commission, 2023b). Although this does not improve the EC’s ability to efficiently sanction Member States, it does increase scrutiny and visibility. In 2021, the EC tabled a proposal to update the 2008 Directive on environmental crimes (European Commission, 2021b). The proposal acknowledged the insufficient number of environmental criminal cases successfully investigated and prosecuted as well as the large discrepancies in the transposition of the 2008 Directive across Member States. Against this background, the EC proposed to enlarge the scope of the 2008 Directive, establish minimum penalties, foster cross-border investigation and prosecution, and promote data collection and dissemination on criminal enforcement actions. In March 2023, the European Parliament’s legal affairs committee voted in support of the EC proposal, extending the list of offenses that would be criminally charged and increasing the size of the minimum penalties.

Environmental Enforcement, Administrative Law and “In-kind” Punishment

The efforts of EU institutions to improve and harmonize enforcement are exclusively focused on criminal law instruments. The EC’s 2021 proposal specifically links poor enforcement in Member States to their reliance on administrative law, which limits fines and thus allegedly reduces the deterrence value of enforcement actions. Indeed, sufficiently high fines are considered crucial to deter future violations (see, e.g., Aguzzoni et al., 2013). However, we argue that reliance on administrative law also has some advantages. In particular, we consider two potential benefits of administrative law based on existing studies, namely higher probability of case initiation and more flexibility in terms of penalty design.

Probability of Case Initiation

One of the shortcomings of the current enforcement framework highlighted by the EC is the very low number of environmental criminal cases that are ultimately prosecuted. Research on enforcement tends to link the low frequency of observed criminal cases to the high cost of criminal proceedings, especially relative to more informal administrative procedures (Faure and Svatikova, 2012). The cost dimension is especially relevant for cases that are moderately serious, but that nevertheless in aggregate contribute significantly to environmental degradation. The probability of catching violations is also relevant, together with the size of the penalty. A very large penalty for a criminal case that is highly unlikely to be prosecuted might be less deterring than a moderate penalty associated with very high probability of prosecution.

“In-kind” Penalties

Federal environmental regulations in the US are enforced through a combination of administrative and criminal law. The Environmental Protection Agency (EPA) initiates administrative cases or refers them to the Department of Justice when the gravity of the violation is large. Administrative cases result in settlements where the defendant can be ordered to pay a fine, which can vary from a few thousand to a few million dollars and which is determined according to various factors, such as the magnitude of environmental harm, the firm’s economic gain from violation, its violation history, and its ability to pay. Additionally, when a fine is established, defendants are given the opportunity to volunteer to pay for an environmentally beneficial project in the affected area. The EPA encourages these projects especially in areas subject to environmental justice concerns, namely those characterized by a large share of minority and low-income households.

Campa and Muehlenbachs (2022) study the implications of using these projects in environmental enforcement cases in the US. The study reveals a large preference among the public for this “in-kind” form of penalty versus traditional fines, based on a survey of US residents. Moreover, a randomized survey experiment reveals that these environmental projects elevate the profile of the firm among the public as compared to a firm that only pays a fine, even when the penalties stem from the same violation. Similarly, the stock-market response to the announcement of these projects is positive, whereas announcing a settlement with a large penalty causes a drop in the stock-market price of the defendant. In terms of implications for environmental justice, the data analysis shows that the whitest and richest communities are the most likely to receive these projects, but the second largest share goes to communities where there are highest concentrations of minorities and low-income households.

Overall, the study finds that punishing firms through environmental projects can be beneficial for political economy reasons, given the large preference for this enforcement tool among the public and likely among firms, since firms seem to benefit from undertaking the projects. Moreover, while the targeting of environmental justice communities in the US is not perfect, tweaking the US arrangement could guarantee that the projects predominantly benefit those communities most harmed by environmental violations.

For EU adoption of environmental projects enforcement, a caveat is that the perception of these projects might be different among the public in the EU. Nonetheless, large-scale surveys modelled on those presented in Campa and Muehlenbachs (2022) can help in understanding public views in different regions. Moreover, the paper emphasizes that on the one hand, by benefiting defendants, the environmental projects might ultimately be a more lenient punishment than fines, with implications for deterrence and future environmental quality. On the other hand, environmental quality might also improve as a direct effect of the projects being implemented and due to improved monitoring in affected communities (Dimitri et al., 2006). Overall, the study finds that future environmental quality might be more likely to improve following fines rather than environmental projects. However, it cautions the reader on data limitations that causes the result to not be conclusive enough and calls for further research.


The persistence of environmental problems in the EU, as well as the striking differences in pollution levels across EU Member States, underscores the need for more and better environmental regulation. However, even in the presence of comprehensive and strict environmental rules, the protection of the environment is still inadequate if a proper enforcement mechanism is not in place. As observed in OECD (2009), proper enforcement ensures deterrence. Successful deterrence provides the best protection for the environment, while reducing the resources necessary to administer laws by addressing non-compliance before it occurs. EU institutions have recently taken important steps to improve and further harmonize enforcement of environmental regulation across Member States, with proposed updates to the existing Directive on the matter scheduled for Member-State discussion in upcoming months.

Specifically, the EU is seeking to step up the use of criminal law to prosecute environmental offenses across Member States, with mandatory penalties and increased cross-border coordination. We argue that the focus on criminal law has some drawbacks, which could be addressed by also harmonizing administrative enforcement across EU Member States. Researchers have previously argued that reliance on administrative law might increase the likelihood that offenses are investigated and prosecuted. We also present evidence from the use of administrative law in the US, where defendants in environmental cases can settle to pay part of their penalty “in-kind”, i.e. by performing environmental projects in areas affected by the alleged violations. The evidence suggests that the use of these projects is worth considering in other jurisdictions, including the EU, because they might be preferred by the public and could help addressing environmental justice concerns. An important caveat is that their implications for environmental protection are not clear, and more research should address this important aspect. On the subject, the existing evidence on environmental enforcement in the US, such as that presented in Campa and Muehlenbachs (2022), is established thanks to the availability of rich data sources kept by the US’ EPA. The EC’s recent proposal to systematically collect and disseminate data on environmental crimes is thus particularly welcome and should not be overlooked in the upcoming negotiations with Member States on the final content of the proposed Directive.


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.

Spillover Effects from the Nordic Model of Prostitution Legislation

20230424 Spillover Effects Nordic Model of Prostitution Image 00

In recent years several European countries alongside Canada and Israel have adopted the so-called Nordic model of prostitution legislation to try and reduce the risk of sexual exploitation. While the reforms directly affect the regulation of the domestic sex market, their effects may also spill over to other outcomes in nearby areas and internationally – for example affecting sex tourism flowsMaking use of data on tourism flows and Google searches, a new study examines the causal effect from the implementation of the reform in four different countries on sex tourism in popular destinations. The findings indicate that domestic reforms increase sex tourism, calling for the design of policies to account also for these adverse effects.


Since 1999, when Sweden introduced the so-called Nordic model of prostitution legislation, similar legislation has been introduced in Canada, Iceland, Ireland, France, Norway, and most recently Israel. While the legislation design differs between countries (for an overview see Perrotta Berlin and Spagnolo, 2019), the common foundation is to effectively criminalize the purchase but not the selling of sexual services. The introduction of such reforms aims at battling human trafficking and reducing the risk of exploitation. While the effect from the asymmetric prostitution legislation has been found to increase rape incidence in Sweden (Ciacci, 2018), when it comes to the sex market the Nordic model is mainly thought to affect it in two contrasting ways. Firstly, it may suppress domestic supply, which could result in people travelling to destinations where prostitution is not criminalized. Secondly, it might affect the general view on prostitution (Kotsadam and Jakobsson, 2011), thus reducing domestic demand as well as international sex tourism.

Sex tourism is associated with human trafficking, child exploitation and increased spread of sexually transmitted diseases (Herold and Van Kerkwijk, 1992; Brooks and Heaslip, 2019; Newman et al. 2011). Despite this, few studies have explored the impact of prostitution laws on the practice – in part due to measurement difficulties.

This brief presents evidence from a forthcoming paper by Perrotta Berlin and Latour on sex tourism patterns following the implementation of the reform in four different countries.

Quantifying Sex Tourism

Perrotta Berlin and Latour use tourism patterns and Google searches to quantify sex tourism flows, in order to evaluate the effect from changes in prostitution legislation in Canada, France, Ireland and Norway. Specifically, they use data on the number of monthly tourist arrivals to Thailand and The Philippines, and weekly Google searches originating from the above-mentioned reform countries for popular sex-tourism and other tourism destinations, including attractions within cities. German tourism data and Google searches originating from France as well as Google searches originating in the US are used to estimate the effect on sex tourism to bordering countries (France to Germany and US to Canada, respectively). To evaluate the respective effects, they identify treated and control groups for each considered setting, and proceed to compare data between these groups before and after the reform (in line with the so-called difference-in-differences specification, as pioneered by Card and Krueger, 1994). In the following sections, each of these specifications and the subsequent results are discussed.

Evident Spillover Effects

Thailand and The Philippines

For Thailand and The Philippines, monthly data was available on tourist arrivals differentiated by country of origin from 2013 to 2020 and from 2008 to 2020, respectively. The underlying assumption is that, absent a prostitution legislation reform in the four considered countries (Canada, Ireland, France and Norway), the tourism flows from the country in question to Thailand and The Philippines would have remained the same over time. Thus, the change in the number of tourist inflow (out of which an unknown number are sex tourists) from the country in question – when compared to the number of tourists from other countries used as the control group – can be interpreted as a causal effect from the legislative reform on sex tourism.

The results show that, when compared to tourists arriving from other countries, the number of tourists arriving from one of the countries having recently implemented the Nordic model increased by 0.312 and 0.158 standard deviation points for The Philippines and Thailand respectively. Figure 1 below illustrates the results from an event study specification, in which the reform dates in the four different countries are aligned at 0, depicting how the increase is spread over the two years following the reform.

Figure 1. Number of tourists before and after the reform, The Philippines to the left and Thailand to the right.

Notes: The horizontal axis is the time variable. Time is normalized such that 0 is the month when the reform came into force. On the left panel the vertical axis is the number of tourist arrivals to The Philippines from reform countries in deviation from control countries. On the right panel the vertical axis is the number of tourist arrivals to Thailand from reform countries in deviation from control countries.

France-Germany Border

In Germany, the legislative status of prostitution is determined at the level of municipality. For the analysis, German municipalities where prostitution is to some extent legal were considered to form the treatment group and municipalities where it is illegal constituted the control group. The outcomes of interest were i) tourists travelling to German municipalities of interest, and ii) Google searches from France for the same municipalities.

The analysis shows an increase in foreign tourism to the treatment municipalities following the implementation of the Nordic model of prostitution legislation in France.  At the same time, no changes in domestic tourism was detected. The conclusion that the increase in foreign tourism is driven by an increase in French tourists, by which one could then argue the implemented reform to increase cross-border sex tourism, was validated by the analysis of French Google searches. In these data it can be seen that distant German municipalities where prostitution is legal become relatively more interesting in French Google searches after the reform compared to municipalities where prostitution is illegal.

Figure 2. Searches of German municipalities originated in France relative their distance from the French border.

Notes: The vertical axis is the weekly index of Google Trends for searches for municipalities in Germany originated in France. The horizontal axis is distance from the French border. The red line shows that the slope decreased, i.e. distance became more salient for municipalities with illegal prostitution after the reform.

Canada-US Border

Data on Google searches for Canadian municipalities from one year before to one year after the reform in Canada were considered for the analysis. Searches originate in different US states, which also differ in the extent to which purchase of sexual services is legally punishable. The length of imprisonment in each US state determines whether a state was considered treated – when the length of imprisonment equals or exceeds that in Canada following the reform – or control. Results show that after the introduction of the Canadian reform, Google searches for Canadian municipalities dropped, in particular, in US states with high punishments for purchase of sexual services – most likely those where sex tourism to Canada used to originate before the reform. The results from the event study is depicted in Figure 3 below.

Figure 3. Number of searches of Canadian cities before and after the reform, deseasoned.

Notes: The horizontal axis is the time variable. Time is normalized such that 0 is the month when the reform came into force. The vertical axis is the number of searches from US states with high punishments in deviation from control states.

Sex Tourism Destinations

Finally, Google Searches for sex tourism destinations were considered as the outcome variable with the underlying idea being that – in the absence of a legislative change in the four considered countries – the difference in number of searches for sex tourism vs tourism destinations would have been the same over time. Sex tourism destinations were defined in two alternative ways: first, a list of popular destinations was selected within countries where prostitution is legal; second, this list was augmented with information from websites that list popular destinations for sex tourism, regardless of the legal status of prostitution in that country.

The results from this analysis are less clear, varying with the definition of sex tourism destinations and with the country of origin. But by and large they showed, if anything, that the interest in sex tourism destination countries decreased after the reform. This might indicate a change in attitudes towards lower acceptance of sex trade in general in the countries where the reform was implemented.


Prostitution legislation reforms affect the domestic sex market and have potential cross-border and international spillover effects. One such impact from criminalizing the purchase of sexual services domestically is increased levels of sex tourism, which might in turn impose adverse effects on the destination countries.

Filling a research gap by studying the effect from introducing asymmetric prostitution laws on sex tourism, Perrotta Berlin and Latour find evidence suggesting that harsher domestic regulation, while potentially changing attitudes in the general population (as indicated by Google Searches) also, in specific cases, increases, the outflow of tourists to destinations with less stringent laws.

After the introduction of the Nordic model, Norway has imposed legislation prohibiting their citizens to purchase sexual services even in countries where it is legal and implemented awareness campaigns on the detrimental effects of sex tourism on local populations. Given that sex tourism is associated with human trafficking, child exploitation and increased spread of sexually transmitted diseases, the results call for other countries to follow suit with domestic prostitution legislation taking on a more global approach to achieve greater effectiveness.


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.

Carbon Tax Regressivity and Income Inequality

20210517 Carbon Tax Regressivity FREE Network Image 01

A common presumption in economics is that a carbon tax is regressive – that the tax disproportionately burdens low-income households. However, this presumption originates from early research on carbon taxes that used US data, and little is known about the factors that determine the level of regressivity of carbon taxation across countries. In this policy brief, I explore how differences in income inequality may determine the distribution of carbon tax burden across households in Europe. The results indicate that carbon taxation will be regressive in high-income countries with relatively high levels of inequality, but closer to proportional in middle- and low-income countries and in countries with low levels of income inequality.


Climate change is one of the main challenges facing us today. To reduce emissions of greenhouse gases, and thereby mitigate climate change, economists recommend the use of a carbon tax. The environmental and economic efficiency of carbon taxation is often highlighted, but the equity story is also of importance: who bears the burden of the tax?

How the burden from a carbon tax is shared across households is important since it affects the political acceptability of the tax. For instance, the “Yellow Vests” protests against the French carbon tax started due to concerns that the tax burden is disproportionately large on middle- and working-class households. Research in economics also shows that people prefer a progressive carbon tax (Brännlund and Persson, 2012).   

In this brief, I explore what we know about the distributional effects of carbon taxes and analyze the link between carbon tax regressivity and levels of income inequality in theory and in application to Sweden as well as other European countries.

Carbon Tax Burden Across Households

It is a common finding in the economics literature that carbon taxes are, or would be, regressive (Hassett et al., 2008; Grainger and Kolstad, 2010). However, most of the earlier literature is based on US data, and the US is unrepresentative of an average high-income country in terms of variables that are arguably important for carbon tax incidence. Compared to most countries in Europe, income in the US is high but unequally distributed, carbon dioxide emissions per capita are high, the gasoline tax rate is low, and the access to public transport is poor. If we want to understand the likely distributional effects of carbon taxes across Europe, we thus need to look beyond the US studies.

A recent study by Feindt et al. (2020) examines the consumer tax burden from a hypothetical EU-wide carbon tax. They find that the distributional effect at the EU-level is regressive, driven by the high carbon intensity of energy consumption in relatively low-income countries in Eastern Europe. At the national level, however, carbon taxation in Eastern European countries is slightly progressive due to car ownership and transport fuel being luxuries. Conversely, in high-income countries – where transport fuel is a necessity – carbon taxation is slightly regressive.

That the incidence of carbon and gasoline taxation varies across countries with different levels of income, has been found in numerous studies (Sterner, 2012; Sager, 2019). To understand the source of this variation, we need to identify the determinants of the incidence of carbon taxes.

The Role of Income Inequality

In a recent paper, I, together with Giles Atkinson at the London School of Economics, present a simple model where the variation in the carbon tax burden across countries and time can be determined by two parameters: the level of income inequality and the income elasticity of demand for the taxed goods (Andersson and Atkinson, 2020). The income elasticity specifies how the demand for a good, such as gasoline, responds to a change in income. If the budget share decreases as income increase, we refer to gasoline as a necessity. If the budget share increases with income, we refer to gasoline as a luxury good. Our model predicts that rising inequality increases the regressivity of a carbon tax on necessities. Similarly, we will see a more progressive incidence if inequality increases and the taxed good is a luxury.

To mitigate climate change, a carbon tax should be applied to goods responsible for the majority of greenhouse gas emissions: transport fuel, electricity, heating, and food. To estimate the distribution of carbon tax burden, we must then first establish if these goods are necessities or luxuries, respectively. Gasoline is typically found to be a luxury good in low-income countries but a necessity in high-income countries (Dahl, 2012). Food, in the aggregate, is consistently found to be a necessity. A carbon tax on food would, however, mainly increase the price of red meat – beef has a magnitude larger carbon footprint than all other food groups – and red meat is generally a luxury good, even in high-income countries (Gallet, 2010). Lastly, electricity and heating are necessities, with little variation across countries in the level of income elasticities.  A broad carbon tax would thus likely be regressive in high-income countries, but more proportional, maybe even progressive, in low-income countries. The overall effect in low-income countries depends on the relative budget shares of transport fuel and meat (luxuries) versus electricity and heating (necessities). A narrow carbon tax on transport fuel has a less ambiguous incidence: it will be regressive in high-income countries where the good is a necessity and proportional to progressive in low-income countries where the good is a luxury.  

The income elasticities of demand, however, only provide half of the picture. To understand the degree of regressivity from carbon taxation, we also need to take into account the level of income inequality in a country. Our model predicts that a carbon tax on necessities will be more regressive in countries with relatively high levels of inequality. And increases in inequality over time may turn a proportional tax incidence into a regressive one.

To test our model’s prediction, we analyze the distributional effects of the Swedish carbon tax on transport fuel and examine previous studies of gasoline tax incidence across high-income countries. 

Empirical Evidence from Sweden

The Swedish carbon tax was implemented in 1991 at $30 per ton of carbon dioxide and the rate was subsequently increased rather rapidly between 2000-2004. Today, in 2021, the rate is above $130 per ton; the world’s highest carbon tax rate imposed on households. The full tax rate is mainly applied to transport fuel, with around 90 percent of the revenue today coming from gasoline and diesel consumption.

 Figure 1. Carbon tax incidence and income inequality in Sweden

Sources: Andersson and Atkinson (2020). Gini coefficients are provided by Statistics Sweden.

Using household-level data on transport fuel expenditures and annual income between 1999-2012, we find that the Swedish carbon tax is increasingly regressive over time, which is highly correlated with an increase in income inequality. Figure 1 shows the strong linear correlation between the incidence of the tax and the level of inequality across our sample period. The progressivity of the tax is measured using the Suits index (Suits, 1977), a summary measure of tax incidence that spans from +1 to -1. Positive (negative) numbers indicate that the tax is overall progressive (regressive) and a proportional tax is given an index of zero. The level of income inequality, in turn, is summarized by the Gini coefficient (0-100), with higher numbers indicating higher levels of inequality.

In 1991, when the Swedish carbon tax was implemented, income inequality was relatively low, with a Gini of 20.8. If we extrapolate, the results presented in Figure 1 indicate that the tax incidence in 1991 was proportional to slightly progressive. Since the early 1990s, however, Sweden has experienced a rise in inequality. Today, the Gini is around 28 and the carbon tax incidence is rather regressive. This can be a potential concern if people start to perceive the distribution of the tax burden as unfair and call for reductions in the tax rate.

Empirical Evidence Across High-Income Countries

Figure 2 presents the results of our analysis of previous studies of gasoline tax incidence across high-income countries. Again, we find a strong correlation with inequality; the higher the level of inequality, the more regressive are gasoline taxes.  In the bottom-right corner, we locate the results from studies on gasoline tax incidence that have used US data. The level of inequality in the US has been persistently high, and the widespread assumption that gasoline and carbon taxation is regressive is thus based to a large part on studies of one highly unequal country. Looking across Europe we find that the tax incidence is more varied, with close to a proportional outcome in the (relatively equal) Nordic countries of Denmark and Sweden.

Figure 2. Gasoline tax incidence and income inequality in OECD countries

Sources: Andersson and Atkinson (2020). Gini coefficients are from the SWIID database (Solt, 2019).


A carbon tax is economists’ preferred instrument to tackle climate change, but its distributional effect may undermine the political acceptability of the tax. This brief shows that to understand the likely distributional effects of carbon taxation we need to take into account the type of goods that are taxed – necessities or luxuries – and the level and direction of income inequality. Carbon taxation will be closer to proportional in European countries with low levels of inequality, whereas in countries with relatively high levels of inequality the carbon tax incidence will be regressive on necessities and progressive for luxury goods.

This insight may explain why we first saw the introduction of carbon taxes in the Nordic countries. Finland, Sweden, Denmark, and Norway all implemented carbon taxes between 1990-1992, and income inequality was relatively, and historically, low in this region at the time. Policymakers in the Nordic countries thus didn’t need to worry about possibly regressive effects. Looking across Europe today, many of the countries that have relatively low levels of inequality have either already implemented carbon taxes or, due to the size of their economies, have a low share of global emissions. In countries that are responsible for a larger share of global emissions – such as, the UK, Germany, and France – inequality is relatively high, and they may find it to be politically more difficult to implement carbon pricing as the equity argument becomes more salient and provides opportunities for opponents to attack the tax.

To increase the political acceptability and perceived fairness of carbon pricing, policymakers in Europe should consider a policy design that offsets regressive effects by returning the revenue back to households, either by lump-sum transfers or by reducing tax rates on labor income.   


  • Andersson, Julius and Giles Atkinson. 2020. “The Distributional Effects of a Carbon Tax: The Role of Income Inequality.” Grantham Research Institute on Climate Change and the Environment Working Paper 349. London School of Economics.  
  • Brännlund, Runar and Lars Persson. 2012. “To tax, or not to tax: preferences for climate policy attributes.” Climate Policy 12 (6): 704-721.
  • Dahl, Carol A. 2012. “Measuring global gasoline and diesel price and income elasticities.” Energy Policy 41: 2-13.
  • Feindt, Simon, et al. 2020. “Understanding Regressivity: Challenges and Opportunities of European Carbon Pricing.” SSRN 3703833.
  • Gallet, Craig A. 2010. “The income elasticity of meat: a meta-analysis.” Australian Journal of Agricultural and Resource Economics 54(4): 477-490.
  • Grainger, Corbett A and Charles D Kolstad. 2010. “Who pays a price on carbon?” Environmental and Resource Economics 46(3): 359-376.  
  • Hassett, Kevin A, Aparna Mathur, and Gilbert E Metcalf. 2009. “The consumer burden of a carbon tax on gasoline.” American Enterprise Institute, Working Paper.
  • Sager, Lutz. 2019. “The global consumer incidence of carbon pricing: evidence from trade.” Grantham Research Institute on Climate Change and the Environment Working Paper 320. London School of Economics.  
  • Thomas, Sterner. 2012. Fuel taxes and the poor: the distributional effects of gasoline taxation and their implications for climate policy. Routledge.
  • Suits, Daniel B. 1977. “Measurement of tax progressivity.” American Economic Review 67(4): 747-752.

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.

Pay-for-Performance and Quality of Health Care: Lessons from the Medicare Reforms

Pay-for-Performance and Quality of Health Care Policy Brief Image

Health care attracts major attention in terms of hospital and physician reimbursement, owing to the large share of public expenditures and the presence of welfare issues demanding regulation. The focus of this policy brief is quality adjustments of prospective payments in the health sector. Using the data on the 2013 reform in Medicare, we show differential effects of value-based purchasing, where price setting is related to benchmark values of quality measures. The theoretical and empirical evidence indicates that unintended effects appear for acute-care U.S. hospitals at the best percentiles of quality. The findings provide insights into benchmarking within pay-for-performance schemes in health care.


The Russian national project “Health”, which was started by the federal government a decade ago and has expanded to regionally financed hospitals, is an example of a public remuneration scheme targeted at increasing health care efficiency. The project emphasized the role of the primary sector and raised salaries of general practitioners. A part of salaries was linked to patients’ assessment of the quality of health care. The reimbursement was seen as a means to stimulate higher quality.

However, cautiousness is required in introducing such payment mechanisms. Indeed, international experience shows that quality-related pay in health care may lead to heterogeneous effects across different groups of providers. A recent CEFIR working paper uses administrative panels of the U.S. hospitals to analyze the changes in quality owing to the introduction of the quality-pay.

The U.S. Health Care Sector

Pilots of pay-for-performance

In the early 2000s, numerous private and public programs linking quality and reimbursements in health care existed in the U.S., mostly at employer or state level (Ryan and Blustein, 2011; Damberg et al., 2009; Pearson et al., 2008). A nationwide pilot of quality-performance reimbursement started with the Hospital Quality Incentive Demonstration, where quality measures for five clinical conditions (heart failure, acute myocardial infarction, community-acquired pneumonia, coronary-artery bypass grafting, and hip and knee replacements) were accumulated from voluntarily participating hospitals. Some of these quality-reporting hospitals opted for the pay-for-performance project (initially established for 2003-2006, and later extended to 2007-2009). The project provided respectively 2% and 1% bonus payments for hospitals in the top and second top deciles of each quality measure (as of the end of the third year of the project). Hospitals in the bottom two deciles, on the other hand, were to receive 1-2% penalties (Kahn et al., 2006). Overall, the financial incentives helped improving the quality of the participating hospitals, but the improvement was inversely related to baseline performance (Lindenauer et al., 2007). Moreover, low-quality hospitals required most investment in quality increase; yet, they were not financially stimulated (Rosenthal et al., 2004).

The accumulation of the measures within the Hospital Quality Incentive was followed by the launch of the Surgical Care Improvement Project (SCIP) and Hospital Consumer Assessment of Healthcare Providers (HCAHPS). HCAHPS was the first national standardized survey with public reporting on various dimensions of patient experience of care. The measures of the clinical process of care domain are collected within the Hospital Inpatient Quality Reporting (IQR) program. These are measures for acute clinical conditions stemming from the Hospital Quality Incentive (i.e. acute myocardial infarction, heart failure, pneumonia), as well as measures from the Surgical Care Improvement Project and Healthcare Associated Infections.

The 2013 reform of Medicare

The success of the pilot project in the U.S. in terms of average enhancement of hospital quality has resulted in the nationwide introduction of these reimbursement policies. Namely, a value-based purchasing reform started at Medicare’s acute-care hospitals in the fiscal year of 2013. The reform decreased Medicare’s prospective payment to each hospital by a factor α and redistributes the accumulated fund. As a result of this rule, all hospitals performing below the mean value of the aggregate quality are financially punished, as their so-called adjustment coefficient is less than unity. At the same time, hospitals above the mean value are rewarded (See details in the Final Rule for 2013: Federal Register, Vol.76, No.88, May 6, 2011.)

The aggregate quality – called the total performance score – is a weighted sum of the scores of the measures in several domains: patient experience of care, clinical process of care, outcome of care, and efficiency. The scores on each measure are based on the hospital’s position against the nationwide distribution of all hospitals. In short, positive scores are given to hospitals above the median, and higher scores correspond to performance at the higher percentiles. The scores are a stepwise function, assigning flat values of points to subgroups within a given percentile range. Hospitals above the benchmark (the 95th percentile or the mean of the top decile) are not evaluated according to their improvement relative to the performance in the previous year.

If one assumes that hospitals are only maximizing profit, then such a linear payment schedule should stimulate quality increases across all spectrums of hospitals. However, the theoretical literature generally separates the hospital management, interested in profits, from the physicians who make decisions affecting the level of quality. In particular, physicians are treated as risk-averse agents, who have a decreasing marginal utility of money; that is, their valuation of monetary gains of a certain size decreases as their income increases. In such behavioral model (Besstremyannaya 2015, CEFIR/NES WP 218) physicians’ decisions about the quality of care is shaped by the trade-off between the potential losses they may incur if fired in case of hospital budget deficit and/or bankruptcy and their own costly effort to maintain and improve quality.

In this respect, the reform introduced two mechanisms: (1) it decreased the level of reward for low-quality hospitals and increased it for high-quality hospitals; and (2) it established a positive dependence of reward on quality. We show that the two forces compete, and the first one may outweigh the second for physicians at hospitals with high quality. Indeed, in these hospitals improved budget financing makes the bankruptcy, and probability of firing, less likely. As a result, physicians may be satisfied with a given sufficient level of a positive reward and not willing to exert any further efforts to raise the amount of this reward. Furthermore, physicians may even become de-stimulated. As a result, in these higher quality hospitals, the quality of care stabilizes or even goes down after the reform.

To sum up, we hypothesize that quality scores increase at the lowest tails of the nationwide distribution, while it may stay stable or fall among the highest quality hospitals. The sign of the mean/median effect is ambiguous.


Data on quality measures and hospital characteristics such as urban/rural location and ownership come from Hospital Compare. The panel covers the period from July 2007 to December 2013, and consists of 3,290 hospitals (12,701 observations). We exploit first-order serial correlation panel data models – longitudinal models where the value of the dependent variable in the previous period (lagged value) becomes one of the explanatory variables (see notations and definitions of analyzed measures in Tables 1-2.) The empirical part of the study evaluates the impact of the reform on changes of the quality scores of hospitals belonging to different percentiles of the nationwide distribution of each quality measure.

Table 1. Patient experience of care

Comp-1-ap Nurses always communicated well
Comp-2-ap Doctors always communicated well
Comp-3-ap Patients always received help as soon as they wanted
Comp-4-ap Pain was always well controlled
Comp-5-ap Staff always gave explanation about medicines
Clean-hsp-ap Room was always clean
Quiet-hsp-ap Hospital always quiet at night
Hsp-rating-910 Patients who gave hospital a rating of 9 or 10 (high)

Notes: Score on each measure is the percent of patients’ top-box responses to each question.

Table 2. Clinical process of care

AMI-8a Primary PCI received within 90 minutes of hospital arrival
HF-1 Discharge instructions (heart failure)
SCIP-Inf1 Prophylactic antibiotic received within 1 hour prior to surgical incision
SCIP-Inf3 Prophylactic antibiotics discontinued within 24 hours after surgery end time
SCIP-Inf4 Cardiac surgery patients with controlled 6 a.m. postoperative blood glucose
SCIP-VTE2 Surgery patients who received appropriate venous thromboembolism prophylaxis within 24 hours prior to surgery to 24 hours after surgery

Notes: Score on each measure is the percent of percent of cases with medical criteria satisfied.

The results of the estimates offer persuasive evidence for a non-rejection of our hypotheses: quality goes up at 1-5th deciles and falls at the 6-9th deciles (see Figures 1-2).

Figure 1. Mean change of scores owing to value-based purchasing across percentile groups of hospitals


It should be noted that the hypotheses concerning differential effects also rely on the fact that there is a certain population of hospitals to which each of the step-rates apply (Monrad Aas, 1995). Hence, the threshold and/or benchmark value in the national schedule may be worse than the value in a given hospital. Therefore, reimbursement with benchmarking becomes an additional cause of undesired effects.

Figure 2. Mean change of scores owing to value-based purchasing across percentile groups of hospitals



Our analysis confirms the presence of adverse effects of quality performance pay in health care. A remedy may be found in establishing benchmark at the value of the best performing hospital or employing ‘episode-based’ payment, which rewards a hospital for treating each patient case with corresponding criteria satisfied (Werner and Dudley, 2012; Rosenthal, 2008).

While the above results are based on the US data, they suggest that cautiousness is required in applying the pay-for-performance schemes to healthcare financing also in transition countries, and much attention should be paid to the potential adverse effects.


  • Besstremyannaya, Galina, 2015. “The adverse effects of incentives regulation in health care: a comparative analysis with the U.S. and Japanese hospital data” (2015) CEFIR/NES Working Papers, No.218, www.cefir.ru/papers/WP218.pdf
  • Damberg, Cheryl L, Raube, Kristiana, Teleki, Stephanie S and dela Cruz, Erin, 2009. ”Taking stock of pay-for-performance: a candid assessment from the front lines”, Health Affairs, Volume 28, pages 517-525.
  • Kahn, Charles N, Ault, Thomas, Isenstein, Howard, Potetz, Lisa and Van Gelder, Susan, 2006. “Snapshot of hospital quality reporting and pay-for-performance under Medicare”, Health Affairs, Volume 25, pages 148-162.
  • Lindenauer, Peter K, Remus, Denise, Roman, Sheila, Rothberg, Michael B, Benjamin, Evan M, Ma, Allen and Bratzler, Dale W, 2007. “Public reporting and pay for performance in hospital quality improvement”, New England Journal of Medicine, Volume 356, pages 486-496.
  • Monrad Aas, I., 1995. Incentives and financing methods, Health policy, Volume 34, pages 205-220.
  • Pearson, Steven D, Schneider, Eric C, Kleinman, Ken P, Coltin, Kathryn L and Singer, Janice A, 2008. “The impact of pay-for-performance on health care quality in Massachusetts, 2001-2003”, Health Affairs, Volume 27, pages 1167-1176.
  • Rosenthal, Meredith B, Fernandopulle, Rushika, Song, HyunSook Ryu and Landon, Bruce, 2004. “Paying for quality: providers’ incentives for quality improvement”, Health Affairs, Volume 23, pages 127-141.
  • Ryan, Andrew M and Blustein, Jan, 2011. “The effect of the MassHealth hospital pay-for-performance program on quality”, Health Services Research, Volume 46, pages 712-72.
  • Werner, Rachel M and Dudley, R Adams, 2012. “Medicare’s new hospital value-based purchasing program is likely to have only a small impact on hospital payments”, Health Affairs, Volume 31, Number 9, pages 1932-1940.




Why Do Scientists Move? The Mobility of Scientists

Image of a man walking inside the airport with his luggage representing a mobility of Scientists

This policy brief provides an overview of new evidence on the determinants of the mobility of scientists – high human capital workers who generate new ideas and expand the frontier of knowledge. New evidence from a large dataset of elite US life scientists shows that professional factors, including individual productivity and the quality of a scientist’s peer environment, matter for mobility. Strikingly, family structure also plays a significant role, with the likelihood of moving to decrease when a scientist’s children are in high school (14-17 years old). This suggests that even “star” scientists take into account more personal, family factors in their mobility decisions, likely due to the costs associated with disrupting their children’s social networks.

Workers often face an important decision during their career: should I move to a new city for a new job? Relocation is a complex decision that can involve numerous professional and personal factors, and some of these factors can constrain moves while others facilitate them. Relocating can, on one hand, mean significant transition costs in terms of uprooting one’s family and navigating a new city and workplace; on the other hand, it can open up new career opportunities and provide environments where one’s skills are put to better use.

Why should we care about whether and why workers move? Economic theory suggests that mobility is one channel through which worker productivity can be increased by improving the employer-employee “match”. Moreover, particularly for highly-skilled workers, the theory suggests that the mobility of workers can impact the productivity of their peers; if the human capital of the mobile worker “spills over” to their peers, then the peers left behind would experience a decline in productivity and those at the new destination would get a boost.

In light of this, understanding the mobility of scientists – high human capital workers who are generating new ideas and expanding the frontier of knowledge – is of particular importance when considering the potential role that mobility can play in increasing productivity and innovation, which are central to models of economic growth.

The Determinants of Mobility

While there is a growing literature trying to document how the mobility of scientists can impact their own productivity and the productivity of their peers (see e.g. Agrawal, McHale, & Oettl, 2014), a significant challenge is finding plausibly exogenous variation in both the timing and location choices of movers. In order to fully understand the impacts of mobility, we first need to know more about the determinant of mobility: why and when in their careers scientists move.

Several studies have examined the determinants of the mobility of scientists and inventors, but the literature has been hampered by lack of data that allow researchers to observe the relevant factors that may matter for mobility. These studies have tended to focus on professional factors for moves; especially how individual productivity measures, like a scientist’s number of publications, citations and patents, predict moves. Importantly, there has been less attention in this literature on the constraints to mobility, including more personal factors, like the role of children and family, and the quality of one’s peer environment.

The findings from these studies on the role of individual productivity for mobility is mixed with evidence pointing to a positive relationship (Zucker, Darby and Torero, 2002; Lenzi, 2009), and some evidence showing negative (Hoisl, 2007) and/or no effects (Crespi et al, 2007). One key professional factor that has remained underexplored in these studies is the quality of the peer environment, or how one’s colleagues can influence the choice of moving.

Moreover, very few studies have been able to examine non-professional factors such as the role of family and children. There is some evidence on family factors and inventor mobility from Sweden, where detailed data is widely available but within country mobility is relatively low (Ejermo and Ahlin, 2014). There is also some evidence from the sociology of science literature showing that children influence the scientific performance and mobility of scientists. Using data from the U.S. Census, Shauman and Xie (1996) find that children tend to constrain mobility; for women, children negatively impact mobility regardless of the age of the children, while for men, it is older, high school-age children that tend to constrain mobility. However, given that the study uses Census data, individual productivity measures are limited, which are needed to compare the effects for similarly accomplished scientists.

Evidence from Elite Life Scientists in the US

In Azoulay, Ganguli and Graff Zivin (2016) we examine the determinants of mobility of elite life scientists in the U.S. and for the first time, provide evidence on both professional and personal determinants of mobility. We use a unique panel dataset we compiled from the career histories of 10,004 elite life scientists to understand why and when scientists make decisions to move to new locations. We are able to observe the transitions scientists make between institutions, and focus on moves that are at least 50 miles apart (based on distance between the zip codes of the institutions) to increase the likelihood that a transition leads the scientists to change their place of residence.

The dataset includes individual productivity measures of publication counts and the U.S. National Institutes of Health (NIH) funding data. We also measure the quality of the peer environment at the scientists’ origin and destination locations using publication and funding counts of peers. We define peers as both collaborating and non-collaborating peers (those who are close in “idea space”), and those who are geographically close (less than 50 miles apart) and those who are distant (more than 50 miles apart). Finally, to examine the personal factors, for each scientist in our sample, we hand-collected information on their children, including each child’s year of birth.

Through regression analysis, we find that individual productivity is a positive predictor of moves, which is consistent with several other studies (Zucker, Darby and Torero, 2002; Coupé, Smeets & Warzynski, 2006; Lenzi, 2009; Ganguli, 2015). We also provide new evidence on additional professional factors that influence the propensity to move. For example, we find that obtaining recent NIH funding deters moves, perhaps as a result of transaction costs associated with transferring federally funded research between institutions. We also find that a scientists’ peer environment is a significant predictor of mobility, as scientists are less likely to move when the quality of the peer environment near their home institution is high and more likely to move when the quality of the peer environment at distant institutions is high.

Figure 1. Age of Children and Mobility: Age of Oldest Child


Figure 2. Age of Children and Mobility: Age of Youngest Child


Our most striking result is the role that family structure plays for mobility. We find a significant decrease in mobility when scientists’ children are of high school age. The likelihood of moving increases just before their oldest child enters high school, and again when their youngest child is beyond high school age. Figure 1 shows a notable spike in distant moves just before the oldest child in the household enters high school (11-14 years old), while Figure 2 shows a similar spike after the youngest child completes high school (18-20 years old). In both figures, the relationship between age of children and local (less than 50 miles) moves does not show a similar spike. This relationship between mobility and age of children persists in regressions that allow us to control for productivity measures and potential confounders.


This brief has discussed theory and evidence related to scientific mobility. New evidence from a large dataset of elite life scientists shows that while professional factors do matter for mobility, we also find that even “star” scientists take into account more personal, family factors in their mobility decisions, likely due to potential disruptions to the social networks of their children.

Given that there is still little evidence about what drives relocation decisions, it is important for further analysis of these issues, and our study raises several more questions for researchers to examine, many of which have important policy implications. For example, what is it about recent NIH grants that deter mobility – the terms of the grant contracts or costs of moving personnel and equipment? Regarding the family factors, we were unable to look at differences among female and male scientists, but an important question for further research is whether the age of children and other factors appear to affect women’s and men’s relocation decisions differently.


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.

Highly Educated Women No Longer Have Fewer Kids

This policy brief summarizes evidence that the cross-sectional relationship between fertility and women’s education in the U.S. has recently become U-shaped. The number of hours women work has concurrently increased with their education. The theory that the authors advance is that raising children and home-making require parents’ time, which could be substituted by services such as childcare and housekeeping. By substituting their own time for market services to raise children and run their households, highly educated women are able to have more children and work longer hours. The authors find that the change in the relative cost of childcare accounts for the emergence of this new pattern.

In 2012, the European Commission published a special report on “women in decision making positions”, suggesting legislation to achieve balanced representation of women and men on company boards. Some countries such as Norway, France, Italy, Belgium and the Netherlands have already taken legal measures in that direction. Trends in women’s education give hope that such goals may be achieved as women are increasingly occupying more prestigious and demanding careers. Indeed, in today’s world, women have surpassed men in higher education in most developed countries (Goldin et al 2006; Hazan and Zoabi 2015a).

What are the consequences of this important development for fertility? Historically, highly educated women have had fewer kids than less educated women (see, for example, Jones and Tertilt 2008). This relationship is deep rooted in the economic and sociological literature to the extent that many theories have already been proposed to explain this relationship. Leading explanations rely on the difficulty to combine children and career (Mincer, 1963; Galor and Weil, 1996) and the quantity-quality tradeoff (Becker and Lewis, 1973; Galor and Weil, 2000; Hazan and Zoabi 2006). The shift in women’s education coupled with more demanding careers for women means that if the cross-sectional relationship between women’s education and fertility is stable over time, then future fertility rates will continue to decline from their already historically low levels.

In Hazan and Zoabi (2015b) we find, however, that the cross-sectional relationship between women’s education and fertility has changed from monotonically declining until the 1990s to a U-shaped pattern during the 2000s. This change is due to a substantial increase in fertility among women with advanced degrees who increased their fertility by 0.7 children, or by more than 50%. This is illustrated in Figure 1, which plots the cross-sectional relationship between fertility and women’s education in 1980 and during the period 2001-2011.

Figure 1: Fertility Rates by Women’s Education, 1980 and the 2000s.


What can explain the rise of fertility among highly educated women during the period that saw the largest increase in the labor supply of highly educated women? We argue that the rise in college premium increased the demand for child-care and housekeeping services by highly educated women and a rise in the supply for such services by low educated women. This ‘marketization’ weakened the tradeoff between career and family life and enabled highly educated women to pursue demanding career without giving up on their desired family size.

To establish the relationship between the rise in the college premium and fertility of highly educated women, we use data from the March CPS for the period 1983-2012. We estimate the average hourly wage in the “child day-care services” industry and allow it to vary by state and year. In addition, we compute the hourly wage of all women in the 25-50 year-old age group who reported a positive salary income. Figure 2 presents the fitted values of the average of this variable for each of our five educational groups. The figure shows that childcare has become relatively more expensive for women with less than a college degree but relatively cheaper for women with a college or an advanced degree.

Figure 2: Linear Prediction of the Log of the Ratio of Average Wage in the Childcare Industry to Average Wage in the Five Educational Groups 1983–2012


To utilize the micro data we estimate regression models where the dependent variable is a binary variable that takes the value of one if a woman, living in a specific state in a specific year gave birth during that year and zero otherwise. Our main explanatory variable is the labor cost in the child daycare industry divided by the own wage of that woman. We show that there is a highly statistically significant and economically large negative correlation between this measure of childcare cost and the probability of giving a birth. In our empirical analysis we find that this change in the relative cost can account for about one-third of the increase in the fertility of highly educated women. We use a battery of tests to show that this correlation is not driven by selection of women into the labor market, by the endogeneity of wages, or by specific years over the last three decades.

Figure 3: 2000s Actual and Hypothetical Fertility under the 1980s prices of Childcare


Figure 3 uses the estimates from the regression models described above and shows a hypothetical fertility for 2001-2011 under the 1983-1985 relative childcare cost. The figure shows that the hypothetical fertility curve is obtained by a clockwise rotation of the actual fertility curve around the group of women that has some college education.

Direct evidence on paid childcare services is consistent with this view. Figure 4 shows the average weekly paid childcare hours by all women aged 25-50 in 1990 and 2008. The figure has two salient features. First, in each of these years, paid childcare is increasing with women’s education. Secondly, between 1990 and 2008, paid childcare hours have stagnated for women with up to some college education but have sharply increased for highly educated women.

Figure 4: Paid Childcare Weekly Hours for Women aged 25-50.


We then rule out potentially other explanations. What if the increase in labor supply stems from women who did not give birth during that year? To address this concern we shows that the cross-sectional relationship between education and usual hours worked for the sub-sample of women age 15-50 who gave birth during the reference period exhibit the same positive correlation. Another concern might be that it is in fact the spouses who respond to a birth by lowering their labor supply enabling their wives to work more. Find that men who are married to highly educated women work more than men who are married to women with lower levels of education. Interestingly, fathers to newborns work more than husbands who do not have a newborn at home, regardless of the education of their wives. More importantly, usual hours worked by fathers to newborns monotonically increased with their wives’ education. Thus, the spouses of highly educated women are not the ones substituting in childcare for their working wives.

Another concern our model may raise is that marriage rates differ across different educational groups. If married women have higher fertility rates and if more educated women have higher marriage rates, more educated women’s higher fertility rates may not be caused by the availability of relatively cheaper childcare and housekeeping services, but rather simply by their higher marriage rates. We find that the fraction of women with advanced degrees who are currently married is somewhat lower than that of women with college degree.

Figure 5: Number of birth per 1,000 White Women in the US in Age Groups 35-39, 40-44 and 45-49: Women with Advanced Degrees (2001-2011) and Historical Rates.


A final potential explanation might be related to recent advancements in Assisted Reproductive Technology (ART) that enable women to combine long years in school without scarifying parenthood. We address this possibility in three ways. First, we show that historical levels of fertility rates among women above age 35 were higher than the levels during the 2000s (see Figure 5). This stands in contrast to the argument that highly educated women were not able to have higher fertility rates in the past due to medical reasons. Secondly, we note that ART accounts for less than 1% of births occurred during the 2000s. Finally, fifteen U.S. states have infertility insurance laws that provide coverage to infertile individuals. We compare fertility patterns by women’s education in these states to the rest of the country and find no difference in fertility rates during the 2000s between the two groups of states.

The results of this study have several implications. For public policy, it highlights potential benefits from pro-immigration policies. Unskilled immigrants can potentially have positive effect on fertility via an increase in the supply of cheap home production substitutes. For many developed countries that are facing aging and shrinking population this may be something to consider. It also has consequences for economic growth. Given the strong correlation between parents’ education and kids’ education, an increase in the relative representation of kids coming from highly educated families means that the next generation is going to be relatively more educated. These are good news for economic growth.


  • Gary S. Becker and Gregg H. Lewis. On the interaction between the quantity and quality of children. Journal of Political Economy, 81:S279–S288, 1973.
  • Oded Galor and David N. Weil. The gender gap, fertility, and growth. American Economic Review 86(3): 374–387, 1996.
  • Oded Galor and David N. Weil. Population, technology, and growth: From Malthusian stagnation to the demographic transition and beyond. American Economic Review 90(4): 806–828, 2000.
  • Claudia Goldin, Lawrence Katz, and Ilyana Kuziemko. The homecoming of American college women: A reversal of the college gender gap. Journal of Economic Perspectives 20(4): 133–156, 2006.
  • Moshe Hazan and Hosny Zoabi. Does longevity cause growth? A theoretical critique. Journal of Economic Growth, 11 (4), 363-376, 2006.
  • Moshe Hazan and Hosny Zoabi. Sons or Daughters? Endogenous Sex Preferences and the Reversal of the Gender Educational Gap. Journal of Demographic Economic, Vol 81, pp: 179-201, 2015a.
  • Moshe Hazan and Hosny Zoabi. Do highly educated women choose smaller families? Economic Journal, 125(587):1191–1226, 2015b.
  • Larry E. Jones and Michele Tertilt. An economic history of fertility in the u.s.: 1826-1960. In Peter Rupert, editor, Frontiers of Family Economics, pages 165 – 230. Emerald, 2008.
  • Jacob Mincer. Market prices, opportunity costs, and income effects. In Carl F. Christ, editor, Measurement in economics: Studies in mathematical economics and econometrics in memory of Yehuda Grunfeld. Stanford University Press, pages 67-82, 1963.

Reputation for Quality and Entry in Procurement: Is there a Trade-Off?

20140224 Reputation for Quality and Entry Image 01

How much weight should be given to past performance indicators when selecting contractors? Does a large weight assigned to suppliers’ previous performance deter entry by new, innovative suppliers that have no track records for the very reason that they are new? If yes, how should we take this into account when designing procurements for firms or governments? This note describes recent research that sheds light on these questions crucial for every government and organization.


How should past performance be accounted for when selecting (public or private) contractors? On one hand, giving large weight to suppliers’ previous performance in assigning contracts may improve incentives in procurement; on the other hand, it may deter entry by new, innovative suppliers for the very reason that they are new. Are there ways to structure procurement rules and procedures to minimize or eliminate these costs? How can well performing suppliers be rewarded, but not at the expense of losing the most innovative start-ups, that could pose important positive externalities on the buyer and on society overall? These questions are important ones and every procurement manager, in the private and public sector, should know how to answer, or at least how to think about, them. Unfortunately, if one looks at the leading management and operations textbooks, or at public procurement textbooks, it is hard to find a line that could help in making these crucial decisions. The only procurement book that at least mentions these crucial questions, to our knowledge, is the Handbook of Procurement (2006; see Ch. 18, by Dellarocas et al.). However, even that handbook falls short in providing evidence-based or – more generally – research-based guidance for these questions. This is the case because there is practically no research dealing with these everyday problems. One recent exception is an experimental study recently undertaken by Butler, Carbone, Conzo and Spagnolo (2013) that will be discussed in depth in the reminder of this brief.

Public Policy Relevance  

Before getting into the results of this recent study, let me provide some background information that will give an idea of the relevance of these questions for public policy.

Public procurement currently accounts for between 15% and 20% of the GDP in developed countries (see http://cordis.europa.eu/fp7/ict/pcp/key_en.html).  In 2011, the total public procurement market in the EU – i.e. the purchases of goods, services and public works by governments and public utilities – reached a size of approximately €2,500 billion, corresponding to 19 percent of GDP (see e.g. http://ec.europa.eu/internal_market/publicprocurement/docs/modernising_rules/public-procurement-indicators-2011_en.pdf). As Table 1 below shows, despite year-to-year fluctuations, there has been an overall increase in procurement expenditures relative to 2007 levels, both absolutely and proportionately. The increasing shift in focus in EU innovation policies, from “push-based” mechanisms like R&D subsidies/tax breaks towards demand-led “pull” mechanisms, like Pre-Commercial Procurement, is likely to further increase the volume of this market.

Table 1: Total EU procurement expenditure on works, goods and services
In EUR billion











As a % of GDP











Source: DG MARKT. 2011. Public Procurement Indicators.

The enormous size of this market notwithstanding, we know relatively little about whether, when, and how buyers should use reputational indicators based on past performance in selecting among sellers, and whether the use of such indicators necessarily reduces the ability of new sellers—i.e., sellers with no history of past performance—to enter the market.

It is well known that reputational mechanisms that reward past performance are important governance tools that complement (and sometimes substitute for) contracts in private transactions (Calzolari and Spagnolo 2009). Private buyers, however, are typically only concerned about the price and quality of the good they buy. Regulators in charge of public procurement, instead, are usually also concerned that the public procurement process is transparent and open for obvious accountability reasons. The need to prevent favoritism and corruption has led lawmakers around the world to ensure that open and transparent auctions where bidders are treated equally—even when in some crucial dimensions they have very different track records—are used whenever possible.

The trend in the US

The costs of limiting discretion to ensure public buyers’ accountability – such as the possibly large cost of not allowing reputational forces to complement incomplete procurement contracts – were stressed by Kelman (1990), who pushed for a deep reform of the US procurement system when he was the head of public procurement during the Clinton administration. The reform was targeted at reducing the rigidity of procurement rules in the Federal Acquisition Regulations and allowing public buyers to adopt more flexible purchasing practices common in the private sector, including giving more weight to suppliers’ past performance. Since the Federal Acquisitions Streamlining Act of 1994, US federal departments and agencies are expected to record past contractors’ performance evaluations and share them through common platforms for use in future contractor selection.

However, the US Senate recently expressed the apparently widespread concern that past performance-based selection criteria could hinder new and small businesses’ ability to enter and compete effectively, leading to an intriguing, but inconclusive report by the General Accountability Office.

This is not to say that US regulators were not concerned with the ability of small and medium enterprises (SME) – sometimes the most innovative part of the economy – to enter public procurement markets. In the US, this long-held concern led to large programs like the Small Business Act, with its rules limiting the bundling of public demand in very large procurements and establishing the Small Business Agency, and the ‘set aside’ (procurement only open to SMEs) common in many types of procurement auctions. However, the worry that past-performance based selection may contribute to the exclusion of novel and smaller firms only arose in the last couple of years.

The (opposite) trend in the EU

The European Union has instead been moving precisely in the opposite direction. An important concern driving procurement regulation in Europe since the Treaty of Rome has been helping the process of common market integration by increasing cross-border procurement, i.e., the amount of goods and services each EU Member State buys from contractors based in other states. The EU Procurement Directives that coordinate public procurement regulation in the various European states have been limiting the use of past-performance information in the process of selecting among offers—a feature that came under broad attack during the 2011 consultation for the revision of the EU Directives (see Replies to the Consultation on the 2011 EU Green Book on Public Procurement regulation). The EU regulators appear to have been always convinced that using reputational indicators as a criteria for selecting contractors leads to manipulations in favour of local incumbents, at the expense of cross-border procurement and market integration.  (see e.g., http://ec.europa.eu/internal_market/publicprocurement/modernising_rules/consultations/index_en.htm).

Only very recently – now that the US is moving towards reconsidering the effects, and possibly limiting the use of past-performance indicators – has the EU started to move (again in the opposite direction to the US) towards leaving more space to these indicators, perhaps as a reaction to the comments they received in recent consultations.

Finally, to have an idea of the lack of research-based knowledge guiding policy in this field, note that in some cases EU regulation already acknowledges the crucial importance of past-performance based reputation for some types of procurement. For example, the European Research Council (ERC) provides funding to top researchers in Europe, who are selected through peer review, and the track record of the researchers is usually the main awarding criterion. ERC funding is distributed almost exclusively based on reputation criteria in order to support the best and the brightest. Other European instruments for the procurement of research, such as the FET-OPEN program, are based on a strictly enforced, completely anonymous evaluation instead, without obvious reasons justifying this opposite approach. On the dedicated homepage of these programs it states: “The anonymity policy applied to short proposals has changed and is strictly applied. The part B of a short STREP proposal may not include the name of any organization involved in the consortium nor any other information that could identify an applicant. Furthermore, strictly no bibliographic references are permitted.”

Reputation and Entry in Procurement

In a recent research paper (Butler, Carbone, Conzo and Spagnolo, 2013) we have been trying to fill at least part of this knowledge gap and offer some initial evidence-based guidelines for future policy.

The study

We build a simple model of repeated procurement with limited enforcement and potential entry and implement it in the laboratory. We focus on reputation as an incentive system to limit moral hazard in the quality dimension as well as on the effect of reputation on selection through entry. We assume that some costly-to-produce quality dimension of supply, although observable to the parties, is too costly to verify for a court to be governed through explicit contracting and is therefore left to reputational governance. We make the additional assumption that there is a potential entrant firm that is more efficient than all incumbent firms. In this context, we study how quality, price, entry and welfare change with the introduction of a simple and transparent reputational mechanism. This mechanism rewards an incumbent firm that chooses to provide (costly) high-quality production with a bid subsidy in the subsequent procurement auction, and may also award a bid subsidy (of varying size) to an entrant with no history of production.

Note that in the case of public procurement and of firms’ vendor rating systems, we are talking about reputational mechanisms based on public rules, known and accepted by suppliers. Formal mechanisms and rules give commitment power to the buyer and can be designed in many different ways. A common mistake is to assume that reputational mechanisms must be designed along the line of the eBay feedback system, in which new sellers start with “zero reputation”. However, a buyer with some commitment power concerning the rules for information aggregation and diffusion and for selecting suppliers may well award a positive rating to new entrants—e.g. the maximum possible rating, or the average rating in the market, putting entrants at less of a disadvantage—and ensure that this is taken into account by the scoring rule that selects the contractor, even if the contractor has never before interacted with the buyer. Indeed, private corporations often have vendor rating systems in which all suppliers start off with the same maximal reputational capital—a given number of points—and then lose points when performing poorly and are suspended for some time if their reputational capital falls below a certain low threshold. This type of vendor rating system creates an advantage for new suppliers, most likely stimulating rather than hindering entry, suggesting that it is possible to design a reputational mechanism in public procurement that simultaneously sustains quality and entry.

The results

First of all, the study shows that concerns about reputation-based selection hindering entry are justified: naively introducing a “standard” reputational mechanism in which only good past performance is rewarded with a bid subsidy in the following procurement auction increases quality provision, but it also significantly reduces entry.

In contrast to this first result, the study goes on to show that properly designed reputational mechanisms in which new entrants, with no history of past performance, are awarded a moderate or high reputation score—as is often done in the private sector—actually foster rather than hinder entry while, at the same time, delivering a substantial increase in high quality goods provision.

The third important result of this study is that the total cost to buyers (buyer’s transfer) does not increase when a reputational mechanism is introduced, even though (costly) quality provision increases. The introduction of bid subsidies for good past performance appears to benefit the buyer/taxpayer by increasing competition for incumbency, driving winning bids down sufficiently to offset the potential increase in procurement costs due to bid subsidies.


Considered together, the findings in Butler et al. (2013) suggest that there need not be a trade-off between reputation and entry in procurement, and that the debates both in the EU and the US are rather misplaced. The results suggest that the dual goals of providing incentives for quality provision and increasing entry and cross-border procurement – in the EU or elsewhere – are achievable through an appropriately designed reputational mechanism. Policy makers should therefore probably stop quarrelling about whether a generic past-performance based reputational mechanism should be introduced, and instead focus on how such a mechanism should be designed.


  • Butler Jeff, Carbone Enrica, Conzo Pierluigi and Giancarlo Spagnolo. 2013. “Reputation and Enty in Procurement.” CEPR Discussion Paper No. 9651.
  • Calzolari, Giacomo and Giancarlo Spagnolo. 2009. “Relational Contracts and Competitive Screening.”  CEPR Discussion paper No. 7434.
  • European Commission. 2011. Green Paper on the modernisation of EU public procurement policy Towards a more efficient European Procurement Market. Available at http://ec.europa.eu/internal_market/consultations/index_en.htm
  • European Commission. 2011. Green Paper on the modernisation of EU public procurement policy Towards a more efficient European Procurement Market. Synthesis of Replies.  Available at http://ec.europa.eu/internal_market/consultations/docs/2011/public_procurement/synthesis_document_en.pdf
  • Government Accountability Office. 2011. (GAO-12-102R, Oct. 18, 2011). Prior Experience and Past Performance as Evaluation Criteria in the Award of Federal Construction Contracts. Available at  http://www.gao.gov/products/GAO-12-102R
  • Kelman, Steven. 1990.  Procurement and Public Management: The Fear of Discretion and the Quality of Government Performance: American Enterprise Institute Press.
  • Yukins, Christoher. 2008. “Are IDIQS Inefficient? Sharing Lessons with European Framework Contracting.” Public Contract Law Journal,  37(3): 545-568.

Putting the “I” Back in Team: The Rise of International Teams in Science

20181217 Conference Image 01

In this policy brief, I discuss the increasing prevalence of international teams in the production of scientific knowledge.  I outline several potential factors that may explain these trends and discuss recent evidence from an original survey of coauthors on scientific papers regarding their collaboration behavior.  Finally, as a notable example of increased international collaboration, I discuss the increase in scientific collaboration between Russia and the US after the end of the Cold War.

The Increase in Collaboration and Internationalization of Teams

Teams are becoming more prevalent in science.  Both the share of papers produced by teams and the number of scientists working on scientific papers has increased in recent decades (Wuchty, Jones and Uzzi, 2007).  Economic theory suggests that scientific research is becoming increasingly collaborative since the frontier of scientific knowledge has become more complex and specialized so that more researchers are needed to combine their expertise to make advances (Jones, 2009).  Team members are also becoming more geographically dispersed: the share of papers resulting from international collaborations has increased, and within the US, scientists today are more likely to have coauthors located in a different city than before (Freeman, Ganguli and Murciano-Goroff, 2014).

These trends can be seen clearly in the graph below from the National Science Board’s Science and Engineering Indicators 2012.  It shows the share of both world papers and US papers from 1990-2010 that are coauthored, coauthored with domestic coauthors only, and coauthored with at least one international coauthor.  Collaboration in general and international collaboration have been increasing steadily since 1990 both in the world and in the US.  However, for the US, the share of domestic-only collaborations has plateaued, while it is increasing in the rest of the world.  In a recent Nature article, Adams (2013) shows that this trend similarly holds for other Western countries (United Kingdom, Germany, France, the Netherlands, Switzerland), while for emerging economies (China, India, South Korea, Brazil, Poland), domestic collaborations are also increasing.

Figure 1. World and US Trends in Scientific Collaboration, 1990-2010
Source: From National Science Board (2012)

Why has Science Become More International?

There are many potential reasons for the recent increases in international collaboration.  An important factor has likely been the spread of the scientific workforce and R&D activities throughout the world (Freeman, 2010). The growing number of science and engineering PhDs in developing countries, some of whom are international students and post-docs returning to their home countries has expanded the supply of potential collaborators around the world (Scellato, Franzoni, and Stephan, 2012).  Another factor is funding that has shifted scientific production towards international teams, as increased government and industry R&D spending in developing countries and grant policies by the European Union and other countries have supported international cooperation.

The lower cost of travel and communication in recent decades has also reduced the cost of collaborating with people in different locations.  For example, Agrawal and Goldfarb (2008) show how the expansion of Bitnet, the precursor to the Internet, led to increased collaboration between institutions within the US.  Finally, the location of scientific equipment and materials, such as the CERN Large Hadron Collider, telescopes, or climatological data available only in certain parts of the world, have increased international collaboration, and in some fields, has made international collaboration a necessity.

Survey Evidence on Scientific Collaborations

In a recent paper, my coauthors and I present the results of an original survey we conducted of scientists regarding collaboration (Freeman, Ganguli and Murciano-Goroff, 2014).  In August 2012 we conducted a web-based survey of the corresponding authors of scientific papers with at least one US coauthor published in 2004, 2007, and 2010 in the fields of Nanotechnology, Biotechnology, and Particle Physics.

We customized each survey to ask the corresponding author about the collaboration and individual team members.  The survey questions asked about how the team formed, how it communicated and interacted during the collaboration, the contribution of each coauthor, types of research funding, and the advantages and disadvantages of working with the team.  We received 3,925 responses, so that our response rate was approximately 20%.

The survey also asked the respondent which country each coauthor was “primarily based in during the research and writing” of the article. This gives us a more accurate measure of whether teams are international than can be typically gleaned from publication data, which are based on author affiliations at the time of publication.  Defining international teams from author affiliations alone can produce errors if affiliations change between the time the research was undertaken and the time of publication, or because some people have affiliations from more than one country.

Our analysis of the survey data uses the respondents’ information to define US collocated, US non-collocated and international teams. One of our key results is that face-to-face meetings continue to play an indispensible role in collaborations: most collaborators first met while working in the same institution.  Teams also reported that while carrying out the research, they communicated often through face-to-face meetings, even with coauthors from distant locations.

Figure 2 below displays how the corresponding author responded about how they first met their team members.  It shows that former colleagues play a very important role in the formation of international teams, followed by former students, conferences and institution visits, which equally contribute.  The graph also shows the similarity between international teams and US non-collocated teams in how coauthors met.  For other survey questions, our analysis also shows similarities between international teams and US non-collocated teams, suggesting that the salient issues are more about geography in general rather than necessarily about national borders.

Figure 2. How Coauthors First Met
Source: From Freeman, Ganguli and Murciano-Goroff (2014)


Another key finding from our survey is that the main reason for most collaborations, whether domestic or international, is to combine the specialized knowledge and skills of coauthors. We also asked the corresponding authors their views of the advantages and challenges of their collaboration.  The most often cited advantage for all types of collaborations was “Complementing our knowledge, expertise and capabilities” and “learning from each other”.  For the challenges, US non-collocated and international teams tended to agree more that there was “Insufficient time for communication”, “Problems coordinating with team members’ schedules”, and “Insufficient time to use a critical instrument, facility or infrastructure”, but international teams did not report these problems more often than US non-collocated teams. Where international teams differed is that these teams were the most likely to agree that their “research reached a wider audience”.

International Collaboration After the End of the USSR

A small but significant part of the increase in international collaboration since the 1990s can be attributed to the end of the Cold War.  In “Russian-American Scientific Collaboration” (Ganguli, 2012), I examine trends in international collaboration by Russian and US scientists since the end of the USSR.  Given the nature of the Cold War and restrictions on travel and communication with the West, I show that there was a dramatic increase in the number of publications with at least one Russian and a US coauthor from 1985 to 2005.

In addition to the lifting of travel and communication restrictions, there are several factors that contributed to the surge in collaborations between American and Russian scientists after the end of the USSR.  First, at the level of the Russian government, there was a switch to a more open and collaborative approach to science. Part of this effort included establishing international centers for research in Russia aimed at integrating Russia into the global science community. Another important factor facilitating collaborations with Western researchers were foreign grant programs. The large increase in the emigration of Russian scientists in the 1990s to the West also contributed to international collaboration.  After emigrating, many Russian scientists maintained close links to their colleagues in Russia, and coauthored papers with their former colleagues, which are counted as internationally coauthored publications.

While many of these factors have aided international cooperation after the end of the USSR, there have also been significant challenges that made cooperation difficult.  Some of these challenges in the early 1990s included the political instability, organizational turnover making long-term funding agreements difficult to implement, difficulty transferring funds due to the underdeveloped banking system, high taxation and customs duties, lack of effective intellectual property rights, poor infrastructure, lack of a shared language (both linguistic and cultural), and external regulations (see further discussion in OECD, 1994).  However, many of these challenges have now been overcome, leading to the continued increase in international collaboration between Russian and US scientists.

My analysis in Ganguli (2012) shows that the increase in Russian-American collaboration was more pronounced in some fields of science versus others, particularly in Physics.  Figure 3 shows that the bulk of the articles published with Russian and American coauthors were Physics articles, with a sharp increase occurring immediately after 1991.

Figure 3. Russia-United States Publications By Field, 1985-2005
Source: From Ganguli (2012)


While some of the differences across the fields can be attributed to the number of scientists active in these fields, there are also other potential contributing factors.  For example, it may be that there was greater emigration of scientists from certain fields abroad, and links between emigrants and those who remained in Russia persisted. Graham and Dezhina (2008: 24) suggest that over 50 percent of emigrants were physicists and mathematicians. Another reason may be that international collaboration was more important in some fields due to the knowledge or resources needed to conduct research during the economic crisis of the 1990s.  As Wagner Brahmakulam, Peterson, Staheli, and Wong (2002) point out, physics research received significant amounts of US government funding for international collaboration, partly because expensive equipment that is needed and through collaboration, countries could share costs.  Also, physicists from many countries often meet and work together at international research centers like CERN.  Moreover, in some fields, the US and Russian governments shared priorities in funding international cooperation, like biomedical and health sciences, energy, physics, while there were gaps in some areas where Russia devoted resources and the US did not, like chemistry (Wagner et al. 2002: 24).  Graham and Dezhina (2008: 141) also discuss how Western colleagues benefited from working with Russians especially in fields like zoology, botany and the earth sciences, since the Russian colleagues provided access to data from unique regions not available previously.

Support for International Teams?

This policy brief has discussed some reasons for the increase in international scientific collaboration and related empirical evidence, including insights from collaboration after the end of the USSR.  The growth in collaboration and the geographic dispersion of teams is likely to continue; the frontier of scientific knowledge will become more complex and specialized, so that an even greater numbers of researchers will be needed to combine their expertise, and they are likely to be spread across increasingly distant locations.

These trends raise many complex issues for policymakers.  For some countries, international collaboration may be the only way to sustain the science sector as the frontier of knowledge becomes more complex and resource-intensive.  For some, international collaborations may increase the emigration of home-grown talent to wealthier countries.  To what extent international collaboration should be supported, and how, will be important policy questions going forward. Typically, funding for international projects has been the main policy lever, and the Russian experience suggests that grant programs did play a critical role in that case.  As our survey evidence in Freeman, Ganguli and Murciano-Goroff (2014) suggests, face-to-face meetings are especially important in forming and sustaining international collaborations.  Thus, funding mechanisms that include provisions for research stays and face-to-face meetings may be the most effective means for fostering international collaborations.


  • Adams, J. (2013). “Collaborations: The Fourth Age of Research.” Nature, 497(7451), 557-560.
  • Agrawal A, Goldfarb A (2008). “Restructuring Research: Communication Costs and the
  • Democratization of University Innovation,” American Economic Review, 98(4):1578-1590.
  • Freeman, Richard B. (2010). “Globalization of Scientific And Engineering Talent: International Mobility of Students, Workers, and Ideas and The World Economy.” Economics Of Innovation And New Technology, Volume 19, issue 5, 201 pp. 393-406.
  • Freeman, Richard B., Ina Ganguli and Raviv Murciano-Goroff (2014).  “Why and Wherefore of Increased Scientific Collaboration,” NBER Working Paper No. 19819, Issued in January 2014.
  • Ganguli, Ina (2012).  “Russian-American Scientific Collaboration” in Y.P. Tretyakov (ed), Russian-Аmerican Links: Leaps Forward and Backward in Academic Cooperation. St. Petersburg, Russia: Nestor-Historia, pp. 120-135.
  • Graham, Loren and Irina Dezhina (2008).  Science in the New Russia: Crisis, Aid, Reform. Bloomington and Indianapolis: Indiana University Press, 2008.
  • Jones, Ben (2009). “The Burden of Knowledge and the ‘Death of the Renaissance Man’: Is
  • Innovation Getting Harder?” Review of Economic Studies, 76:283-317.
  • National Science Board (2012). Science and Engineering Indicators 2012. Arlington VA: National Science Foundation (NSB 12-01).
  • National Science Board (2006). Science and Engineering Indicators 2006. Arlington VA: National Science Foundation (NSB 06-01).
  • OECD (1994).  Science, Technology, and Innovation Policies. Federation of Russia. Paris: Organisation for Economic Co-operation and Development, 1994.
  • Scellato, G., Franzoni, C., & Stephan, P. (2012). “Mobile Scientists and International Networks,” NBER Working Paper 18613.
  • Wagner, Caroline, Irene Brahmakulam, D.J. Peterson, Linda Staheli, and Anny
  • Wong (2002).  U.S. Government Funding for Science and Technology Cooperation with Russia.
  • Santa Monica, CA: RAND Corporation, 2002.
  • Wuchty, S., Jones, B. F., & Uzzi, B. (2007). “The Increasing Dominance Of Teams In Production Of Knowledge.” Science, 316(5827), 1036-1039.