Do international parliaments matter? An empirical analysis of influences on foreign policy and civil rights

International parliamentary institutions (IPIs), which give parliamentarians regular opportunities to communicate with their foreign counterparts, have become a common feature in global governance. Recent research has shed light on why IPIs are created, on the similarities and differences in their institutional design, and on the reasons that lead members of national parliaments to engage with them. By contrast, there is little systematic empirical research on whether and how IPIs affect global politics. This article addresses this question by assessing their ability to influence states in relation to the position they take on issues of global concern and to how they treat their own citizens. The study identifies several mechanisms of IPI influence, leading to the hypothesis that more frequent opportunities for parliamentarians to interact with their foreign counterparts within IPIs leads in time to greater similarity in the foreign policy positions expressed by their governments and affects how those governments protect the civil rights of their citizens. A statistical analysis spanning multiple international organizations, member states, and decades indicates that IPIs offer a distinct contribution to convergence in foreign policy. By contrast, participation in IPIs is not robustly associated with civil rights protections. The finding that IPIs can be consequential on which policies governments promote internationally even though such institutions typically lack substantial authority may be encouraging for advocates of further international parliamentarization and specifically the creation of a United Nations parliamentary assembly.


Introduction
Each year, thousands of parliamentarians and legislators from all parts of the world meet and communicate with their foreign counterparts. When they do so, they occupy a peculiar and underexplored space in world politics. Parliamentarians are distinct from representatives of the executive branch, such as heads of government, foreign ministers or diplomats, because typically only executive actors are entitled to formally represent the state in international relations (IR). But, they are also distinct from the non-governmental organizations (NGOs) and transnational activists that have long attracted the attention of IR scholars who emphasize that global politics is not limited to interactions between states. This is because parliaments are public bodies with constitutionally assigned functions in the national governance system, even if their ability to exercise those functions is sometimes limited in practice. To use the language of English School theory (Buzan, 2004), parliamentarians involved in cross-border activities roam the grey zone between the 'international society' of states and the 'world society' of transnational actors.
The phenomenon of parliamentarians stepping out of their home state is not new. A key rationale for it was stated already in 1789, by Jeremy Bentham (2002Bentham ( [1789: 250): Were the French and English legislature to interchange a few Members, there could not be a more powerful means of wearing away those national antipathies and jealousies which as far as they prevail are so disgraceful and so detrimental to both countries. Roy (1962Roy ( [1831: 503) suggested that political and commercial matters affecting any two people with constitutional governments should be decided by an annual 'Congress' composed of an equal number of parliamentarians from each country, 'amicably and justly to the satisfaction of both', so that 'profound peace and friendly feelings might be preserved between them from generation to generation'. In 1889, 55 French, 28 British and 5 Italian members of parliament, and 1 representative each from the parliaments of Belgium, Denmark, Hungary, Liberia, Spain and the United States, met in Paris and founded the Inter-Parliamentary Conference, which soon became the Inter-Parliamentary Union (IPU). However, it is after World War II and especially after the end of the Cold War that bodies bringing together parliamentarians from various countries have proliferated: while only 3 such bodies existed in 1939, there were 40 by 1990, and over 100 were created after that year, including sub-institutions (Kissling, 2014).
By contrast, we do not know much about the effects of IPIs. With the notable exception of the European Parliament, there is little systematic research on whether and how IPIs make a difference for domestic and international politics, and specifically whether they influence the behaviour of governments. At least in their public statements, parliamentarians taking an active role in IPIs tend to credit such institutions with significant achievements. By contrast, the detractors of IPIs dismiss them as enablers of mere 'parliamentary tourism' (e.g. Douzinas, 2017). Even a sympathetic observer who advocates a larger role of IPIs in global governance suspects that 'many existing regional "parliaments" or "assemblies" are quite ineffective -the kind of entities that spread skepticism about international law or institutions of any kind' (Slaughter, 2004: 106). The scarcity of evidence about impact should be troubling for policy makers and scholars alike, and this article aims to fill this research gap using a quantitative approach. To our knowledge, this is the first study to do so.
The article proceeds as follows. We first discuss what IPIs are expected to achieve according to their proponents and the relevant literature, to identify theoretically grounded causal mechanisms and develop testable hypotheses. We argue that a key expectation of practitioners and scholars is that IPIs affect how state agents relate to each other and to their own citizens. Because IPIs are still relatively unfamiliar to IR scholars, we use this section also to provide examples of IPIs and their activities. Then, we present in detail our strategy for assessing the argument that involvement in IPIs promotes the diffusion of international and domestic policies. More specifically, we test the hypothesis that more frequent opportunities for parliamentarians to interact with their foreign counterparts within IPIs leads in time to greater similarity in the foreign policy positions expressed by their governments and affects how those governments protect the civil rights of their citizens. To parse the role of IPIs from the role of international bodies to which they are often associated, we apply an empirical strategy based on pairing intergovernmental organizations (IGOs) that have an IPI with IGOs that do not, using propensity score matching (PSM). Our statistical analysis suggests that IPIs promote convergence in foreign policy, whereas the evidence that they affect civil rights protections is much weaker.

The influence of IPIs: who, how and what
An IPI is an international institution that is of a parliamentary nature, whether legislative or consultative, and has three or more member states, of which the parliamentarians are either selected from national legislatures in a manner that they determine or popularly elected by the electorates of the member states, and that is a regular forum for multilateral deliberations on an established basis, either attached to an international organization or itself constituting one. (Cutler, 2001: 209, bullet points omitted) It is common to distinguish between two main categories of IPIs: free-standing 'international parliamentary associations' that have no formal links with IGOs; and 'international parliamentary organs', which are formally attached to an IGO and possess various degrees of authority vis-à-vis its other organs. In turn, international parliamentary organs differ in relation to various dimensions of institutional design, such as the selection modes of IPI members, the distribution of seats among member states, and their oversight, legislative and budgetary powers (Cofelice, 2018;Habegger, 2005;Winzen and Rocabert, 2021).
How should we think about the effects of IPIs on world politics? There are three aspects to consider: first, the targets of influence, that is, the actors whose behaviour an IPI is expected to affect; second, the channels of influence, that is, how the outputs of the IPI reach the targets and third, the expected outcome, that is, what kind of behaviour the targets are expected to change. These dimensions will be considered in turn.

Whom can IPIs influence and how?
International institutions can be expected to influence primarily the beliefs, aims and identities of individuals who interact with and within them. In the case of secretariats and bureaucracies, the most immediate potential targets of socialization processes are the members of the transnational civil service (Murdoch et al., 2019). In the case of IPIs, socialization processes involve most immediately the participating parliamentarians (Scully, 2005). While these are important aspects, practitioners and observers of IPIs seem to agree that, ultimately, their crucial contribution to global politics lies -or at least should lie -in influencing the behaviour of states, and more specifically national executive actors. Executive actors comprise governing politicians, high-level bureaucrats, military leaders, diplomats, members of security services, law enforcement officers and other actors that control the key levers of the state apparatus. This control gives executive actors access to the main resources that affect the course of domestic and international politics, and thus it is unlikely that IPIs can be said to have a significant impact if they make no difference for what those national executive actors do. For this reason, this study focusses on how IPIs may influence the latter.
Stressing the importance of influence on executive actors does not mean to suggest that such influence must be direct. As Jančić notes (2015: 241-242), '[t]hough the ultimate goal of international parliamentarism is to influence intergovernmental decisionmaking, this is an inherently incremental process that is premised on a much more informal capacity to mobilize political support and steer policy development'. We expect the influence of IPIs to be generally indirect and potentially acting through at least three distinct channels. The first channel goes through the intergovernmental organs of IGOs. IPIs can influence the policies adopted by those intergovernmental organs, either through formal powers of legislative co-decision (as the European Parliament) or through weaker consultation and information processes. In turn, the policies adopted by IGOs can influence the behaviour of national executives.
The second channel goes through national parliaments. The members of most IPIs are also members of national parliaments and, typically, they are formally or informally expected to promote compliance with the IPI's resolutions and recommendations within their own national parliament (Habegger, 2005: 180-182). Alongside the social pressure this entails, interactions within IPIs can lead to genuine persuasion among participants. In turn, preference shifts among IPI-participating parliamentarians can lead to a shift in the balance of preferences in their home parliaments, for instance as a result of the former promoting parliamentary alliances in favour of a policy change. In a last step, shifting preferences in the national parliament can then potentially change the behaviour of executives, either through formal mechanisms, such as legislation that limits executive discretion, or informal pressure exercised through parliamentary questions and debates.
The third channel goes through civil society actors and sectors of public opinion (Pufleau, 2017). Policy positions expressed by and through IPIs can strengthen likeminded civil society groups, for instance by widening support among the relevant sections of public opinion, and this in turn can lead to pressure on national executives to change their behaviour.

What can IPIs influence?
The preceding discussion has referred generically to influences on the behaviour of executive actors, but what kind of behaviour can be influenced exactly? Three potential outcomes are regularly emphasized by both practitioners and observers of IPIs: a reduction of conflicts between states, a change in how states treat their own citizens and an increase of the public accountability of international governance. These will be discussed in turn.
The arguably oldest expectation is that IPIs will help prevent and resolve conflicts between states (Wagner, 2019). This aim underlies Bentham's view quoted in the introduction and led to the creation of the IPU in the 19th century. By promoting trust and mutual understanding among members of different national parliaments, IPIs can help identify mutually acceptable solutions to international controversies. The aim of preventing diplomatic and militarized disputes has been prominent for most of the past 150 years. For instance, in his speech on the occasion of receiving the Nobel Peace Prize in 1903, Randal Cremer, a founder of the IPU, maintained that during the time when there was considerable friction and danger of war between Italy and France, the appeals made by the [IPU] to the members of the Italian Parliament produced an excellent effect and were largely instrumental in preventing strife between those two nations. (Cited by Wagner, 2019: 94) IPIs are expected to contribute both to conflict prevention and conflict mitigation. An example of the conflict prevention function is provided by the Parliamentary Assembly of NATO (NATO-PA): Since its formation in 1955, the traditional principal role of NATO-PA has been concerned with information and education through its meetings and seminars with alliance parliamentarians, in order to ensure a sufficient level of expertise among legislators to enable them to contribute to their national debates, and to ask pertinent questions relating to defence . . .. At a more general level, however, NATO-PA activities have been aimed at inducing a sense of 'weness' amongst European and North American parliamentarians. In that particular role the NATO-PA has fulfilled an important integrative role in the transatlantic relationship by improving understanding and mutual confidence and reinforcing a sense of partnership. (Flockhart, 2004: 371; see also Šabič, 2017) An example of the conflict mitigation function is the role played by the ECOWAS Parliament in defusing tensions between Liberia and its neighbouring countries during Liberia's 1999-2003 civil war. ECOWAS Parliament members conducted missions in the affected countries, engaged with national parliaments and civil society groups, and mediated between the main decision-makers in the conflict. According to some observers, '[s]uch visits turned out to be crucial in preventing the escalation of crisis and were eventually decisive in resolving the conflicts which would have had further devastating human rights consequences in the troubled areas and the entire sub-region' (Nwankwo, 2017: 249).
While security cooperation remains a major concern of IPIs, in recent decades the aim of IPIs extended to the promotion of international cooperation in other areas, such as economic relations and environmental protection. One example is the role of the ASEAN Inter-Parliamentary Assembly (AIPA) in promoting parliamentary support for trade liberalization under the auspices of ASEAN's Economic Community pillar, against the backdrop of opposition to trade liberalization voiced in the legislatures of Indonesia, the Philippines and Thailand (Rüland, 2013).
While the discussion so far has focussed on relationships between states, the second expectation is that parliamentarians' involvement in IPIs can shape how national executive actors behave towards their own citizens. IPIs are seen to be able to influence the way political power is exercised domestically and the extent to which it protects the rights of individuals. In this article, we focus on 'civil rights', which encompasses equality (including gender and race equality) before the law and various personal freedoms (Marshall, 1950). But the argument can be extended to other categories of rights. Most directly, IPIs can affect the level of civil rights protections in the participating states by scrutinizing them and putting them under pressure to improve. More indirectly, interactions within IPIs expose members of parliaments to different ways of organizing statesociety relations than what they know from their own country, and this may alter their beliefs of what is legitimate. In other words, like other international institutions IPIs are social environments that enable and promote socialization processes (Johnston, 2001;Zürn and Checkel, 2005). The expectation is usually that beliefs will shift in the direction of giving greater importance to civil rights vis-à-vis discretionary state authority. Parliamentarians involved in IPIs then become agents of diffusion in their own national parliaments and domestic political environments.
To illustrate what IPIs can do in this domain we can consider the parliamentary dimension of the Council of Europe and the Southern African Development Community (SADC). The core aims of the Council of Europe are the promotion of human rights, democracy and the rule of law in Europe. In addition to its main legal instrument, the Convention for the Protection of Human Rights and Fundamental Freedoms of 1950, the Council of Europe promotes human rights through other international conventions, such as the Convention on Preventing and Combating Violence against Women and Domestic Violence. Over a third of all Council of Europe conventions have been initiated by its IPI, the Parliamentary Assembly of the Council of Europe (PACE) (Habegger, 2005: 155). Moreover, PACE plays a significant role in shaping the content of these conventions, not least as a result of its right to be consulted by the Council of Europe's intergovernmental body, the Council of Ministers (Habegger, 2010). But PACE does not aim to influence government policies only through the intergovernmental channel: the national-parliamentary channel is relevant too. It takes an active role in the implementation of the European Convention on Human Rights and specifically in cases involving delays in execution and problems of noncompliance with judgements of the European Court of Human Rights.
Debates and reports serve to inform national delegations to the Parliamentary Assembly about problems concerning the execution of judgments in their countries. When they return home, members are therefore able to raise the relevant issues in their own parliaments. They have mechanisms and possibilities to promote, or indeed initiate, the adoption of new legislation or changes in existing legislation and put pressure on the relevant authorities to amend practices which are not compatible with the Convention as interpreted by the Strasbourg Court. (Drzemczewski, 2010: 177) For instance, in the early 2000s PACE (2005) repeatedly put Italy under the spotlight for major delays in implementing Court decisions, with a particular emphasis on overcrowding and mistreatment in prisons and the excessive length of judicial proceedings. In January 2006, the Italian parliament responded to the Assembly's pressure by adopting a bill that had been submitted by the parliamentarian and chairman of the Italian delegation to PACE, Claudio Azzolini (PACE, 2008). The so-called Azzolini Law created a more robust procedure for the supervision of the implementation of judgements by the Italian government and parliament, specifically by giving direct responsibility to the Prime Minister and his Office for compliance with Court rulings (Sileoni, 2011).
The combination of intergovernmental and national-parliamentary channels of influence can be seen also in the activities of the SADC Parliamentary Forum in relation to gender equality and violence against women. In 1997, the leaders of SADC member states had signed a non-binding Gender and Development Declaration that stated the goal of promoting gender equality in political, economic and social domains, and to act against violence against women. In 2005, an audit report commissioned by the SADC Parliamentary Forum and the SADC Gender Unit showed substantial gaps in the implementation of those commitments, and the two bodies recommended that the SADC heads of state upgrade their commitments by adopting a legally binding protocol on gender equality with specific targets. After negotiations among governments and the involvement of civil society groups, the SADC Protocol on Gender and Development was adopted in 2008 (van Eerdewijk and van de Sand, 2014). The SADC Parliamentary Forum complemented its efforts to deepen intergovernmental commitments with activities aimed at national parliaments. The Forum played a leading role by implementing gender quotas in its own decision-making bodies, by promoting women's caucuses in several parliaments of SADC member states and in the SADC PF itself, by preparing a Member of Parliament's (MP) guide on gender and development and by training parliamentarians in gender advocacy and analysis skills (Kandawasvika-Nhundu, 2004;Karuuombe, 2008;Sadie, 2005).
The third expectation is that IPIs affect the behaviour of executive actors because they expose the latter to greater scrutiny. By requiring the decision-making bodies of IGOs to be transparent about their decisions and to justify them in a public forum, IPIs can influence the content of those decisions before they are finalized, and this in turn may affect the policies that governments implement domestically. Lipps (2021) shows that national parliamentarians are more likely to attend IPIs when they have more reasons to mitigate the traditional information asymmetry between executives and legislatures in relation to international issues and negotiations. Establishing whether and how exposure to an international layer of parliamentary scrutiny and co-decision causes shifts in intergovernmental policy outcomes raises significant complexities, and so we set aside this third expectation for the time being. The remainder of this article develops an empirical test of the first and the second expectations.
Based on the preceding discussion, we conjecture that IPIs are channels of diffusion between the states that send parliamentarians to IPIs. Specifically, we consider two hypotheses, one relating on their influence on state-to-state relationships and one relating to state-citizens relationships: Hypothesis 1 (foreign policy convergence): Interaction of parliamentarians within IPIs makes the foreign policy positions of their states more similar.

Hypothesis 2 (civil rights influence): Interaction of parliamentarians within IPIs affects the civil rights practices of their states.
Instead of considering individual IPIs, we study the sum of opportunities that parliamentarians have for interacting with their foreign counterparts within IPIs. Our assumption is that the more frequent and numerous such opportunities are, the greater the chances that parliamentarians will do things that can lead -after a suitable time period -to greater similarity in the foreign policy positions expressed by their governments and in how much those governments protect the civil rights of their citizens. This assumption leads us to the research design presented in the next section.
While the focus on this study is the effect of IPIs on the practices of executives, it is important to bear in mind that this is not the only potentially important effect. IPIs could matter also because they shape the attitudes of politically active groups and, ultimately, ordinary citizens. Specifically, IPIs can increase the legitimacy of international governance in the perception of such groups and the interested public. This legitimation function is often discussed as a key raison d'être of IPIs (e.g. Habegger, 2010;Jančić, 2015). Empirical research has shown that the creation of IPIs resulted from 'a legitimation strategy that governments employ strategically when the IO's core constituencies perceive the organization to be failing on this count' (Schimmelfennig et al., 2020: 41; see also Cofelice, 2018;Grigorescu, 2015; and, on the uses of democracy as legitimizing principle, Dingwerth et al., 2019Dingwerth et al., , 2020. This raises the question: does this legitimation strategy work? Answering this question requires looking beyond the effect that IPIs may have on executives and consider other groups. We present some suggestive evidence on this at the end of the 'Additional analyses' section.

Research design
To assess the conjecture that participation in IPIs influences the foreign policy positions and civil rights practices of participating states, we draw on research designs used in prior research on whether and how IGOs promote diffusion across member states.
With regard to Hypothesis 1 -on convergence of foreign policy preferences -we follow the approach pioneered by Bearce and Bondanella (2007). They assessed the socialization effect of IGOs by estimating whether dyads of states sharing membership in a larger number of IGOs were subsequently more likely to vote in a similar way in the United Nations General Assembly (UNGA). UNGA voting behaviour is widely regarded as a reasonably good proxy of foreign policy preferences on a wide range of issues, and scholars often use the degree of voting similarity between pairs of states as independent variable in estimations of the likelihood of conflict between those states (Gartzke, 1998(Gartzke, , 2000Oneal and Russett, 1999;Reed et al., 2008;Sweeney, 2003). If participating in the same IPIs makes foreign policy preferences more similar, this can indirectly prevent conflicts between them or mitigate their severity. To identify foreign policy preferences we use the dataset assembled by Michael Bailey, Anton Strezhnev and Erik Voeten, who applied item response theory (IRT) statistical models to UNGA roll call votes to derive state ideal points in relation to a single dimension that can be interpreted as degrees of support for and contestation of the Western-led international order (Bailey et al., 2017). Our dependent variable for the assessment of Hypothesis 1 is the distance between the ideal points of any two pairs of states (ideal point distance). In addition, we also estimate models using a measure of UNGA voting similarity between any two states (affinity), which was developed by Gartzke (1998) and used by Bearce and Bondanella (2007).
The unit of observation in the assessment of Hypothesis 1 is the dyad-year. For each observation, the independent variable is the number of IPIs that have parliamentarians from both countries in that year, based on the assumption that a higher number provides parliamentarians with more frequent opportunities for interacting with their foreign counterparts. We call this variable joint IPI memberships. Hypothesis 1 is operationalized as the expectation that a higher number of joint IPI memberships will be associated with a lower ideal point distance after a period of time. Based on constructivist arguments on the time scale of international socialization (Zürn and Checkel, 2005: 1065-1066), Bearce and Bondanella focus on 5-year lag between membership and position changes and we do the same here.
Hypothesis 2 states that the extent to which states recognize and protect the civil rights of their citizens is affected by the civil rights practices of other states, and that the influence is stronger among states whose parliamentarians have more numerous opportunities to interact with each other through IPIs. As dependent variable for this part of the analysis, we use the 'Equality before the law and individual liberty index' provided by the Varieties of Democracy project (V-Dem), which captures well the concept of civil rights as we use it here (Marshall, 1950). The index is constructed by taking the point estimates from a Bayesian factor analysis model of country expert assessments of the extent to which a country has rigorous and impartial public administration, transparent laws with predictable enforcement, access to justice for men and women, property rights for men and women, freedom from torture, freedom from political killings, freedom from forced labour for men and women, freedom of religion, freedom of foreign movement and freedom of domestic movement for men and women (Coppedge et al., 2020;Pemstein et al., 2020). We label this variable civil rights.
To capture the diffusion of civil rights practices through IPIs, we follow the approach developed by Greenhill (2015). He created a diffusion variable that is a weighted average of the level of human rights practices in all other countries, where each country's weight increases with the number of shared IGO memberships it has with the focal country. Conceiving the IGO network as a bipartite network yielded a country-level variable that Greenhill labels 'IGO Context'. We adopt this approach in our analysis of IPI participation. For each IPI in which a state was involved in a given year, we calculated the average civil rights score of the other states that were involved in the same year, which yielded a civil rights value for each IPI. By averaging the civil rights values of all IPIs in which a state was involved, we obtain the state's yearly value on a variable that we call IPI context. This is the independent variable in our assessment of Hypothesis 2.
For the same reason noted in relation to foreign policy convergence, we consider a 5-year lag between explanatory and dependent variable to be appropriate to the causal mechanisms discussed earlier in the article. The anecdotal evidence presented earlier supports this time frame. PACE delegates urged governments to implement judgements of the European Court of Human Rights in a resolution adopted in 2000, they formally asked the Italian delegation at PACE (and seven others) 'to prevail upon their respective governments to implement the unexecuted decisions' in February 2003, and the Italian parliament passed legislative reforms to promote compliance in January 2006 (PACE, 2004(PACE, , 2008. For most cases, 4-6 years is a plausible interval to be expected between exposure of parliamentarians to their foreign counterparts and changes in the behaviour of the state.

Sample and matching strategy
Earlier, we mentioned a commonly made distinction between two main types of IPIs: free-standing international parliamentary associations that have no formal links with IGOs, and international parliamentary organs, which are formally attached to an IGO and possess various degrees of authority vis-à-vis its other organs. In the remainder of this article, we limit ourselves to studying the impact of international parliamentary organs, for two reasons. The first reason is theoretical: the causal mechanisms producing impact may differ substantially between the two types and it seems preferable to conduct separate analyses. While the vast majority of IPIs has at most consultative powers in relation to IGOs (Cofelice, 2018;Schimmelfennig et al., 2020) and any effect is likely to stem from socialization mechanisms, including free-standing associations alongside formal organs would increase the heterogeneity of the possible channels of influence. The second reason is practical: we can infer the membership in international parliamentary organs from information about the membership in the IGO to which they are constitutionally linked, which we obtained from the International Governmental Organizations Dataset of the Correlates of War (COW) project (Pevehouse et al., 2020). By contrast, attendance in international parliamentary associations typically depends on the initiative of individual members or speakers of parliaments, and so national representation in meetings is likely to change frequently over time. Determining which countries were represented in each single session of a wide range of international parliamentary associations would be exceedingly expensive in terms of research time and resources, and the present study can help other researchers decide whether such an investment would be worthwhile.
The fact that international parliamentary organs are part of an IGO poses a significant challenge to attempts to determine their distinctive impact on foreign policy preferences and civil rights practices. In brief, the challenge is this: if we find that participation in the same IPIs is associated with convergence between the participating states, how can we know that the effect is due at least in part to the existence of the IPI and cannot be attributed entirely to the fact that there is an IGO of which the IPI is a component? In other words, how can we know that IPIs have any effect beyond that of the intergovernmental bodies of their respective IGOs? We address this challenge through a matching strategy. For each IGO with an IPI, we identify a closely matched IGO without an IPI, and then we compare the effect of diffusion through the set of IGOs with IPIs to the effect of diffusion through an equal number of matched IGOs without IPIs. Finding that the former effect is significantly larger than the latter can be evidence that IPIs make a difference over and beyond the other components of the IGOs in which they are embedded.
We employed PSM for this purpose (Rosenbaum and Rubin, 1983). While PSM is usually employed to achieve balance in the covariates used in the estimation of the dependent variable, we used it for a different purpose here: to identify a set of IGOs without IPI that can function as comparator to the set of IGOs with IPIs. The first step was to select a suitable dataset of IGOs that may or may not have IPIs. We favoured the Measure of International Authority (MIA) dataset over the COW International Governmental Organizations dataset because the more stringent inclusion criteria used for the former yields a more homogeneous set for comparison. The MIA dataset includes only those IGOs in the COW dataset that have 'standing in international politics', which is operationalized as 'having a distinct physical location or website, a formal structure (i.e., a legislative body, executive, and bureaucracy), at least thirty permanent staff, a written constitution or convention, and a decision body that meets at least once a year' (Hooghe et al., 2017: 16). Since IGOs with parliamentary organs form a near-perfect subset of IGOs with standing, it is appropriate to compare IGOs with and without IPIs within this more restrictive category. We follow the coding of Schimmelfennig et al. (2020) in establishing which IGOs in the MIA dataset had an IPI and which ones did not in any given year since 1950. Based on our hypotheses, we regard an IGO has having an IPI from the year of the first session of the IPI rather than from the year when its creation was decided.
The second step consists of the selection of variables for matching. We used five: the proportion of IGO members that are democracies, whether the IGO has a general purpose (as opposed to a specialist policy focus), whether it deals with international security, whether it has an economic focus and whether it imposes human rights constraints on member states. The first two variables have been shown to affect the propensity of having an IPI in previous research: Rocabert et al. (2019) found that IGOs are more likely to have an IPI when they have a general purpose and when they have a higher proportion of members that are democracies. Data are from Hooghe et al. (2017) and Rocabert et al. (2019), respectively. The other three variables have not been examined as potential determinants of IPI creation, but they may affect the likelihood that an IGO will be involved in policy issues discussed in the UNGA and civil rights protection and hence deserve attention as potential confounders. Not only IGOs with a general purpose, but also those with a focus on security may be tasked with coordinating the foreign policies of its member states, which makes it useful to match IGOs also based on whether they are securityfocussed. We consider also whether an IGO has an economic focus, since convergence may be motivated by the prospect of gaining or losing material benefits. Finally, IGOs that have an explicit human rights focus combined with compliance mechanisms may be more likely to influence civil rights protections. We use the data collected by Hafner-Burton et al. (2015) to construct a dummy variable that takes a value of 1 if the IGO has a human rights focus and a score of at least 4 on a 0-10 scale that measures the level of precision, delegation, and obligation of human rights commitments entailed by membership, otherwise a value of 0. Our coding for general purpose, security, economic and sovereignty-constraining human rights focus is presented in the supplemental material (Section 1).
As a third step, we calculated propensity scores for the presence of an IPI in relation to our matching variables, and matched the IPI-IGOs to IGOs from the non-IPI pool, so that the difference in propensity scores between these is minimized. To fulfil the common support criterion, we removed treatment observations whose propensity score is higher than the maximum or less than the minimum propensity score of the non-IPI group (Leuven and Sianesi, 2003). The matches are also restricted to the same year and are re-matched, which means that a match for a given IPI-IGO does not necessarily remain the same over the period of investigation, but is potentially replaced from 1 year to another by a more suitable match -a shift arising either from changing composition of the matching pool or changes in the values of the IOs matching variables over time. A list of the treated observations and their corresponding matches for each year can be found in the supplemental material (Section 1) for an illustrative selection of years and fully in the replication data.
This exercise yielded two additional independent variables to be used alongside our main explanatory variables (joint IPI membership and IPI context). These are joint IPI membership, which counts the number of matched IGOs without parliamentary organs in which two countries have joint membership; and non-IPI context, which is constructed in the same way as IPI context except that it is based on matched IGOs without parliamentary organs.

Control variables
Ideal point distance and civil rights are likely to be affected by a range of factors beyond interactions within IPIs. In all models, we control for the dependent variable lagged by 1 year to account for its temporal dependence and ensure that we estimate the effect of IPIs on the change rather than the absolute level of the dependent variable. We follow Bearce and Bondanella (2007) in the selection of control variables for the analysis of convergence of foreign policy preferences. The analysis controls for: extra-Igo contact, defined as the lower number of diplomatic missions maintained by the two states (Bayer, 2006); domestic political difference, as the absolute difference between the higher and the lower Polity 2 scores (Marshall et al., 2014);dyadic trade, capturing the lower of the two bilateral trade/gross domestic product (GDP) ratios in the dyad (Barbieri et al., 2009;Barbieri and Keshk, 2016);relative economic development as the ratio between the log of the higher and the lower GDP per capita levels of the dyad members (Gleditsch, 2002); relative economic size, measuring the log of the larger dyad member's GDP relative to that of the smaller one (Gleditsch, 2002); relative military power as the log of the higher military capability of the two dyad members divided by the lower one (Greig and Enterline, 2017;Singer et al., 1972); joint military alliance in form of a binary variable taking the value 1 when a military alliance (including ententes, neutrality pacts and defence pacts) is present between dyad members and otherwise 0 (Gibler, 2009); the dichotomous variable Cold War taking the value of 1 for dyad-years before 1991 and otherwise 0; geographical distance, capturing the distance in miles between the capitals of the two states in the dyad (Gleditsch, 2013) and colonial relationship in form of a binary variable taking the value of 1 if such relationship is present and otherwise (Hensel, 2014).
In the analysis of civil rights, we control for the following variables. Civil rights violations may be more likely if a state is involved in internal or external armed conflict (Reiter, 2001). As an indicator of large-scale political violence, we use the variable actotal from the Major Episodes of Political Violence database, which measures the intensity of both interstate and intrastate violence (Marshall, 2019). We include gDP per capita logged to account for the level of economic development and the expectations of modernization theory (Bilson, 1982). We control for the size of the population, to reflect the argument that there are economies of scale in the production of repression (Bilson, 1982). GDP per capita and population data are from Bolt and van Zanden (2020) and from the National Material Capabilities (v6.0) dataset (Singer et al., 1972), respectively. To account for the argument that free and fair elections with broad suffrage lead to governments that are more respectful of civil rights (De Mesquita et al., 2005), we control for V-Dem's electoral democracy index, which combines indices measuring elected officials, suffrage, clean elections, freedom of association and freedom of expression. To account for transnational economic, social and political connections that could affect the protection of civil rights in a country, we control for trade (import plus exports) as a percentage of GDP (Barbieri et al., 2009;Barbieri and Keshk, 2016;Feenstra et al., 2015), for a measure of centrality in the global network of international NGOs (Paxton et al., 2015) and for the number of IGOs of which a country is member (Pevehouse et al., 2020).
The supplemental material (Section 2) presents descriptive statistics. The analyses span the years 1950-2010.

Model specification
To estimate effects on ideal point distance, we apply ordinary least squares (OLS) models to country-dyad-years with standard errors clustered on the dyad level with and without dyadic fixed effects. Fixed-effects models are estimated with the reghdfe package (Correia, 2016). We specify the model as follows where IDEAL POINT DISTANCE ij t , captures how dissimilar the foreign policy positions of states i and j are at time t, IDEAL POINT DISTANCE ij t , −1 is the lagged dependent variable, JOINT IPI MEMBERSHIPS ij t , −5 counts the number of IGOs with an IPI in which the states i and j have joint membership at time t−5, JOINT NON-IPI MEMBERSHIPS ij t , −5 counts the number of IGOs without an IPI in which the states i and j have joint membership at time t−5; ′ X ij t , is the vector of the control variables for states i and j for time t, and υ ij t , is the error term. To determine effects on civil rights, we estimate OLS models with and without country fixed effects. We specify the model as follows where CIVIL RIGHTS i t , measures the level of civil rights protection by state i at time t, CIVIL RIGHTS is the lagged dependent variable IPI CONTEXT − −5 captures state i's exposure to other states' civil rights practices through IGOs with an IPI at time t−5, NON-IPI CONTEXT i t , −5 captures state i's exposure to civil rights practices through IGOs without an IPI at time t−5, ′ X i t , is the vector of the control variables for state i at time t − 1, and υ i t , is the error term. Table 1 presents the findings on foreign policy convergence. The negative sign of the coefficients for joint IPI memberships indicates that an increase in the number of shared memberships in IGOs with IPIs of two states is associated with more similar foreign policy positions 5 years later in the models with and without dyad fixed effects. By contrast, the association between joint memberships in the matched IGOs without an IPI and vote choice similarity is positive in the models with no dyad fixed effects, but not statistically significant; the association becomes negative if dyad fixed effects are added, but the coefficient is smaller than the IPI coefficient and the confidence intervals do not overlap. 1 These findings indicate that IPIs provide a distinctive contribution to convergence in foreign policy positions, supporting Hypothesis 1. Table 2 presents the findings on the protection of civil rights. The models show that the IGO context variable referring to IGOs with IPIs is positively associated with a change in civil rights 5 years later, with the association being statistically significant at the 95% level in the model without country fixed effects but only at the 90% level in the model with country fixed effects. The association of non-IPI context with civil rights is also positive but fails to reach conventional levels of statistical significance. As expected, the coefficient of IPI context is larger than the coefficient of non-IPI context, and a seemingly unrelated estimation test (Weesie, 1999) indicates that the difference is statistically significant at the 95% level in the model with country fixed effects but insignificant in the model without fixed effects. Taken together, these results do not give us high levels of confidence that IGOs with IPIs influence civil rights more than matched non-IPI IGOs do. Considering also that the alternative matching approach described in the next section does not point towards an effect of IPIs, we conclude that there is insufficient support for Hypothesis 2. .790*** (.003) Joint IPI memberships ( .216*** (.007) Observations 366,080 366,078 R 2 /adj. R 2 .880 .886

Additional analyses
We conducted several additional analyses to deepen our understanding of IPI effects and to identify promising avenues for further research. This section summarizes those analyses and the supplemental material provides detailed descriptions and results. First, we used exact matching as an alternative to PSM for the purpose of matching IGOs with IPIs with IGOs without IPIs. Results for IPI variables based on this alternative matching procedure confirm the conclusion based on PSM: interaction within IPIs affects foreign policy positions but not civil rights practice (supplemental material, Section 3).
Second, as an alternative to ideal point distance we estimated models using a measure of UNGA voting similarity between any two states (affinity), which was developed by Gartzke (1998) and used by Bearce and Bondanella (2007). As Bailey et al. (2017) note, changes in affinity reflect changes in the UNGA agenda as well as changes of state preferences. Results confirm those based on ideal point distance: joint IPI memberships are associated with increased affinity, the association is stronger than for joint non-IPI memberships and the confidence intervals of the two variables do not overlap (supplemental material, Section 4). Third, we checked whether the findings are robust to the inclusion of a dummy for European countries, which have the most IPI memberships, and we controlled for possible differences across time periods by adding year fixed effects to our models. These checks support the findings reported earlier (supplemental material, Section 5).
Fourth, we considered the time scale of the effect of IPIs on the dependent variable with which it has a significant and consistent association, ideal point distance. As noted earlier, we focus on a 5-year time lag between the interaction with foreign counterparts within IPIs and the adjustment of government behaviour, as this seems most consistent with the constructivist literature on international socialization. To examine the time scale question more directly, we relaxed our assumptions about the time lag and considered all lags at yearly intervals between 1 and 7 years. As detailed in the supplemental material (Section 6), joint IPI memberships has a statistically significant effect on ideal point distance at each of the seven lags, with the coefficient remaining fairly stable over time and consistently exceeding joint non-IPI memberships.
Fifth, we checked whether our results change if we take into account differences in the institutional design of IPIs (Cofelice, 2018;Winzen and Rocabert, 2021). This exercise is made possible by the new data on design features included in the International Parliamentary Institutions Dataset (Schimmelfennig et al., 2020). We modified joint IPI memberships to give more weight to IPIs that can provide more opportunities for interaction and discussion among members of parliament from different countries, because of the following features: delegates are organized in ideological rather than national groupings, IPI rules stipulate that it should meet for 1 week every year or more, and the IPI has autonomy from outside actors in relation to their agenda, meetings and internal organization. We found that such institutional design features augment the impact of IPIs on foreign policy convergence (supplemental material, Section 7). We also examined whether having formal rights of participation in decision-making of the associated IGO affects IPIs' impact. We found that such rights increase the distance between the foreign policy positions of member states (supplemental material, Section 7). We can speculate that IPIs with authority encourage parliamentarians to mirror clashes occurring between national governments, and in some cases even exacerbate them, but we have to leave a thorough examination of this possibility to future research.
Sixth, we considered whether the effect of involvement in IPIs varies depending on characteristics of the 'receiving' state. The earlier discussion suggested that members of legislatures can affect government practices by formally challenging executive officials and by exercising persuasion and social influence. A mixture of both processes is probably at play in most cases, but it would be desirable to gauge their relative importance across the sample of countries. We conduct an initial analysis of this question by examining whether participation in IPIs matters more when members of parliament are generally more active in challenging the behaviour of executive actors in their respective countries. A positive answer would be evidence of the importance of the adversarial process, whereas a negative answer would suggest that persuasion and social influence in the legislative-executive relationship are the exclusive or at least predominant mechanisms at play. To assess these contrasting expectations, we added to our main models interaction terms between the IPI variables and each of three variables from the V-Dem dataset that relate to legislative constraints on the executive (Coppedge et al., 2020;Pemstein et al., 2020). They, respectively, capture whether the legislature routinely questions executive branch officials, the likelihood that a legislative body would conduct an investigation that would result in a decision or report that is unfavourable to the executive if the latter were engaged in unconstitutional, illegal or unethical activity, and whether opposition parties are able to exercise oversight and investigatory functions against the wishes of the governing party or coalition. We found that all three interaction terms are negative and statistically significant (supplemental material, Section 8). This suggests that the ability of joint IPI memberships to reduce ideal point distance is amplified when members of parliament routinely challenge executive actors and hold them to account. It is, however, notable that the sign and statistical significance of joint IPI memberships in the models, including the interaction term, suggests that joint IPI memberships are still associated with lower ideal point distance even when parliament does not perform those functions.
Seventh, we probed potential causal mechanisms further by considering the argument that officials of countries that experienced a recent transition from one regime type to another are particularly sensitive to socialization effects, because they are 'novices' whose social identity and interests are relatively malleable (Johnston, 2001;Simmons, 2009). We tested this possibility by interacting joint IPI memberships with a variable indicating whether the country has experienced a regime transition during the 10 years prior to the year when its ideal point is observed. A regime transition is defined as a change between any of the four regime types identified by the Regimes of the World project based on V-Dem data: closed autocracy, electoral autocracy, electoral democracy and liberal democracy (Lührmann et al., 2018). While the interaction term has the expected sign, it reaches conventional levels of statistical significance only in the model without dyad fixed effects and not also in the model with dyad fixed effects, which precludes us from drawing a definite conclusion on this issue (supplemental material, Section 8).
Finally, we addressed the question asked at the end of the second section above: given that IPIs were often created to protect or enhance the legitimacy of IGOs, has this strategy worked in practice? The literature on the politicization of international governance suggests that emerging and established political parties use contestation of IGOs to gain electoral support among voters (Ecker-Ehrhardt, 2014;Hooghe et al., 2019;Zürn et al., 2012). If the strategic legitimation strategy is successful, IPIs should be able to shield IGOs from attacks by political parties, at least to some extent, because it becomes more difficult to portray IGOs as undemocratic when they have IPIs and/or because IPIs give a wider range of parties -including opposition parties -opportunities to participate in the workings of IGOs. To provide an initial assessment of this hypothesis, we build on Ecker-Ehrhardt's (2014) analysis of position statements on international governance to be found in party manifestos from 26 countries between 1970 and 2008. If creating IPIs were to support the legitimacy of IGOs in domestic political arenas, we should observe a difference between the reaction of political parties to their country's membership in IPI-less IGOs and how they react to IGOs with IPIs: the latter should be associated with less anti-IGO contestation than the former. The analysis reported in the supplemental material (Section 9) is consistent with this expectation: controlling for a range of factors that can shape party positions on IGOs, we find that a higher number of memberships in IGOs without IPIs is associated with political parties expressing increased levels of opposition to international governance in their manifestos; by contrast, memberships in IGOs with IPIs do not have a statistically significant effect on party positions on average. IPIs appear to 'protect' international governance from negative party statements. Moreover, the analysis suggests that membership in IGOs with IPIs increase the salience of international governance in party manifestos, whereas IGOs without IPIs have the opposite effect. This can be interpreted as a sign that involvement in IPIs encourages parties to openly address international governance as an important matter for public debate and scrutiny, which arguably has positive implications for democratic accountability. We present these findings as initial and suggestive evidence, which is consistent with the view that IPIs have at least some of the legitimation effects that their creators hoped for. While this is only a first take on a complex issue, we hope that it will inspire further research.

Conclusion
As noted earlier in this article, the thousands of parliamentarians who travel each year to meet their fellow parliamentarians around the world are exposed to the suspicion of being mere 'parliamentary tourists'. To our knowledge, we provide the first systematic quantitative assessment of the impact that such activities have on state behaviour. The evidence we collected suggests that the foreign policy positions of governments tend to become more similar when their parliamentarians participate in a larger number of IPIs. We also found some suggestive evidence that differences between IPIs and between national parliaments matter: diffusion through IPIs is stronger when the latter provide more opportunities to parliamentarians to interact across national divides, and when domestic parliaments can routinely challenge executive actors and hold them accountable. By contrast, we did not find robust evidence that governments adjust their protection of civil rights in response to the practices of states with which they are connected through IPIs.
We conclude by pointing at some directions for further investigation and some broader implications of our research. First, we selected two important areas of executive action -foreign policy and civil rights -but in principle IPIs could operate as channels of diffusion in other areas, notably social rights, such as access to education, health care and housing, and rights pertaining to the protection of environmental quality and cultural goods. Future research could examine whether government policies relating to such rights are subject to peer influence conveyed through interactions between parliamentarians. Second, our findings suggest that the domestic political context matters for IPI influence, but we have considered only a limited range of domestic-level moderating variables. Future research could develop and test hypotheses on other internationaldomestic interactions. Third, our analysis remained at the macro level, but future research could study more closely what parliamentarians do after participating in IPIs, how the experience may lead them to shift priorities and/or positions, and how they may promote those positions and priorities in the domestic arena. We hope that our research will encourage other researchers to regard IPI influence as a worthwhile area of study.
In conclusion, we would like to highlight a theoretical and a practical implication of the research agenda pursued in this article. The former stems from what earlier we described with the language of English School theory: parliamentarians involved in IPIs move across a grey zone between the 'international society' of states and the 'world society' of transnational actors (Buzan, 2004). IR theorists have developed a rich understanding of the range of mechanisms that can produce international diffusion through the interaction between government actors (e.g. Johnston, 2001). Together with sociologists, they have also developed an analytical toolbox for understanding the impact of transnationally active civil society actors (e.g. Boli and Thomas, 1997;Keck and Sikkink, 1998). By contrast, the analysis of the mechanisms through which interactions between parliamentarians affect the domestic and international policies of governments is much less developed. As discussed in this study, they may consist of a mix of mechanisms that are familiar from studies of intergovernmental and transnational politics, or they could include mechanisms that are specific to the 'in-between' nature of parliamentary practices. We highlight this as a promising set of questions.
The second implication worth highlighting relates to policy. From the moment of the creation of the United Nations in 1945, politicians, activists and scholars have called for a parliamentary body to be added to its structure -a United Nations parliamentary assembly (UNPA) (Archibugi, 2008;Cabrera, 2018;European Parliament, 2018;Global Greens, 2008;Leinen and Bummel, 2018;Mendlovitz andWalker, 2003 see also De Wilde et al., 2016;Dingwerth, 2014;Ghassim et al., 2022). Sceptical assessments of this proposal suggest that it may face a dilemma: an UNPA with weak powers may be feasible, but it would fail to have real impact; conversely, an UNPA with strong powers might make a difference if it existed, but it is for all practical purposes unachievable given the current realities of world politics (Miller, 2010;Nye, 2002). The research presented in this article is relevant to this debate. With a few notable exceptions, the IPIs covered in this study have very limited legislative powers (Cofelice, 2018;Schimmelfennig et al., 2020). However, the evidence we presented suggests that this does not condemn them to irrelevance in relation to the positions promoted by governments on global issues. This finding may be encouraging for those who regard an UNPA to be both achievable and consequential.
Nevertheless, we urge caution before using our results to advocate institutional change in global governance, for two reasons. First, the identification of causal effects in this political domain is challenging and ideally the conclusions would be confirmed using additional analytical strategies before influencing policy making. Second, we found evidence of convergence between governments, but convergence is not necessarily towards substantive positions that many advocates of a global parliament would welcome. On one hand, major international organizations increasingly articulate commitments to liberal and democratic norms (Dingwerth et al., 2020;Tallberg et al., 2020). On the other hand, research on 'authoritarian diffusion' shows that international institutions can also spread antidemocratic norms and practices, especially when powerful autocracies are involved (Bank and Weyland, 2019;Bromley et al., 2020;Costa Buranelli, 2020;Hackenesch and Bader, 2020). Moreover, higher democratic accountability can lead members of international assemblies to endorse positions that are less aligned with cosmopolitan values (De Wilde et al., 2016). Further research on the positions and practices that IPIs prompted governments to adopt would be advisable before concluding that their contribution to international democracy is unambiguously positive.