https://orcid.org/0000-0002-6947-0663
Abstract
Despite representing a crucial day-to-day diplomatic tool, travel by heads of government and state (HOGS) has remained an under-investigated topic in international relations, inhibiting our ability to better understand how these visits change foreign aid, interstate conflict, diplomatic affinities, and more. Here, we fill that gap by introducing the first global dataset on the foreign visits of state leaders, the Country and Organizational Leader Travel (COLT) dataset, which allows us to present descriptive analysis and assess the monadic and dyadic drivers of foreign travel by HOGS. We find evidence consistent with previous literature explaining the motives of leader travel: development, trade, conflict, institutional co-membership, and regime type. In addition, we show a potential further application of the dataset, presenting original results on the relation between diplomatic visits and international trade. Overall, these data represent a unique indicator of international interaction that cuts across levels of analysis.
A pesar de que los viajes de los jefes de Gobierno y de Estado (HOGS, por sus siglas en inglés) representan una herramienta diplomática crucial en el día a día, estos siguen siendo un tema poco investigado en el campo de las Relaciones Internacionales, lo que inhibe nuestra capacidad para comprender mejor el efecto que tienen estas visitas sobre la ayuda exterior, los conflictos interestatales y las afinidades diplomáticas, entre otros. En este artículo, llenamos ese vacío gracias a la presentación del primer conjunto de datos global sobre las visitas al extranjero de los líderes estatales: el conjunto de datos de viajes de líderes nacionales y organizacionales (COLT, por sus siglas en inglés). El COLT nos permite presentar un análisis descriptivo y evaluar los impulsores monádicos y diádicos de los viajes al extranjero que llevan a cabo los jefes de Gobierno y de Estado. Encontramos evidencia consistente con la literatura previa que explica los motivos de los viajes de los líderes: desarrollo, comercio, conflicto, comembresía institucional y tipo de régimen. Además, mostramos una posible aplicación posterior que podría tener el conjunto de datos, mediante la presentación de resultados originales con respecto a la relación entre las visitas diplomáticas y el comercio internacional. En general, estos datos representan un indicador único de la interacción internacional que comprende todos los niveles de análisis.
Malgré leur importance cruciale en tant qu'outil diplomatique au quotidien, les déplacements des chefs d’État et du gouvernement (CEG) restent sous-étudiés en relations internationales, ce qui nous empêche de mieux comprendre, entre autres, l'incidence de ces visites sur l'aide étrangère, les conflits interétatiques et les affinités diplomatiques. Nous pallions ici cette lacune en présentant le premier ensemble de données mondial sur les visites de dirigeants à l’étranger, l'ensemble de données Country and Organizational Leader Travel (COLT), qui nous permet de présenter une analyse descriptive et d’évaluer les facteurs monadiques et dyadiques des déplacements à l’étranger de CEG. Nous trouvons des éléments probants qui viennent confirmer la littérature antérieure qui explique les motifs de déplacements des dirigeants : développement, commerce, conflit, appartenance commune à une institution et type de régime. En outre, nous montrons qu'il existe peut-être une autre application pour l'ensemble de données, en présentant des résultats originaux quant à la relation entre les visites diplomatiques et le commerce international. Dans l'ensemble, ces données représentent un indicateur unique des interactions internationales aux différents niveaux d'analyse.
Introduction
US President Joseph Biden’s diplomatic visit to Ukraine in 2023 marked an unprecedented event in modern diplomacy. As pointed out by his National Security Advisor Jake Sullivan, it was the first instance of an American president visiting the capital of a nation at war without the US military being in control of the site’s critical infrastructure (Beaumont and Borger 2023). Underscoring the visit’s significance to Ukraine’s adversary, Russia responded by deploying a MiG-31 fighter jet to fly over Kyiv shortly after Biden’s departure, signaling the Kremlin’s negative reception of this diplomatic development. Such visits by political leaders have served as a cornerstone of international diplomacy since antiquity, often yielding considerable impacts on alliances, conflict resolution, trade, and national prestige.
Recent literature has shed light on the multifaceted impact of heads of government and state (HOGS) travel. Visits can exacerbate or mitigate international tensions (Fuchs and Klann 2013; McManus 2018), while also serving to solidify existing partnerships and cultivate new ones (Hoshiro 2020). Scholars have identified a range of factors influencing such trips: personal motivations, domestic political considerations, and broader systemic dynamics (Lebovic and Saunders 2016; Koliev and Lundgren 2021). Studies have also demonstrated that these visits can significantly shape international trade patterns, reinforce interdependence, and influence other key elements of the global landscape (Nitsch 2007; Goldsmith and Horiuchi 2009).
Despite the demonstrated importance of leader visits in the global system, research on their motives and effects has predominantly focused on a few, primarily Western or powerful countries. For instance, Lebovic and Saunders (2016) analyzed US presidential and secretary of state travel from 1946 to 2012, while Nitsch (2007) examined US, French, and German leader travel from 1948 to 2003. Hoshiro (2020) tracked diplomatic visits to Japan from 1969 to 2015, and Kastner and Saunders (2012) analyzed data on Chinese presidents’ and premiers’ travel from 1998 to 2008. While these studies offer valuable insights, their geographic limitations restrict our ability to draw broader conclusions.
We address this gap by introducing the Country Organization Leader Travel (COLT) dataset, a resource that tracks bilateral and multilateral HOGS trips from 1990 to 2023 for 212 countries and territories. These data account for 78,641 foreign trips allowing for the first generalizable findings on the causes and effects of global leader travel. The COLT data and codebook are included with this article’s replication materials and are freely available to use. After elaborating the data collection process used to produce COLT, we demonstrate their utility through a series of statistical tests. First, we provide descriptive statistics related to trends in HOGS travel across time. Second, we test the drivers of monadic and dyadic travel abroad. Finally, we evaluate the relationship between HOGS travel and trade, a relationship established in the literature that we test with global data for the first time.
Our findings reinforce many explanations of what drives leader travel, showing that levels of development, material capabilities, economic and institutional interdependences, conflict, distance, and homophily all matter (Russett, Oneal, and Davis 1998). More specifically, material flows matter: states that receive large amounts of foreign aid have high levels of economic dependence or that receive large inflows of arms trade are also more likely to receive visits (Lebovic and Saunders 2016). Institutional co-membership also plays a role in shaping state visits in various ways, with multilateral treaties, United Nations (UN) voting patterns, and international organization membership all shaping which states are likely to be visited. Furthermore, states that are more economically developed are more likely to both send and receive visits (Kastner and Saunders 2012). States with high levels of internal conflict are also less likely to see incoming visits though more likely to visit other states. These findings also generally support previous results related to case studies (e.g., Lebovic and Saunders 2016 on the United States, Hoshiro 2020 on Japan).
Our findings on the relationship between HOGS travel and trade complicate results presented elsewhere. Existing literature shows that HOGS travel abroad drives increased trade between pairs of states, a finding that has been explored in various studies (Nitsch 2007; Fuchs and Klann 2013; Moons and van Bergeijk 2017). However, using a global dataset, we find that the relationship is much more endogenous than literature shows: increased international trade also drives an increased likelihood of state visits. By replicating and extending previous work, we introduce a wrinkle into our picture of this dynamic relationship, complicating our understanding of the role of leaders, structure, and economic incentives.
Beyond offering general analytical confirmation of previous theories of diplomatic exchange in the literature, our novel and open-access dataset presents unique opportunities to advance research on timely issues such as national prestige, public diplomacy, weaponized interdependence, and interstate conflict. We note that these data will appeal to researchers with general and country-specific interests in leader travel, permitting investigations across various levels and in countries whose diplomatic activities are often overlooked. Similarly, scholars who focus on status could certainly employ outgoing and incoming diplomatic visits as a reliable proxy for, respectively, status seeking and recognition. Finally, leader travel can be explored and connected with many other important international political phenomena such as militarized disputes or humanitarian aid.
The article proceeds by (1) describing the COLT data collection process with examples; (2) analyzing the drivers of HOGS travel; (3) evaluating the relationship between HOGS travel and trade; and (4) discussion/conclusion.1
The COLT Dataset
The unit of analysis used in the COLT dataset is leader-trip and the database includes information on 78,641 trips for 212 political units between 1990 and 2023. The data are global in scope, tracking bilateral, multilateral, and personal visits abroad by priority leaders from all countries and territories. The codebook provides information on the seventy-nine indicators that our data capture for each trip, which can be categorized into four groups. First, we identify the characteristics of the trip itself, such as the destination, dates of visit, and whether the trip involved various events such as leader bilateral meetings or participation in a multilateral summit. Second, we include various indicators identifying the subject traveling abroad, such as their name, title, and other contextual details.2 Third, we introduce additional details and notes for each trip. Finally, we include a confidence rating to account for uncertainty introduced by source validity (Ackerman and Pinson 2016). To ensure further reliability of the data, each group of leader visits were put through a series of additional vetting steps, including double-blind coding, cross-validation with extant leader travel data produced by other researchers, and following similar data gathering methods to those used by others (Moyer, Turner, and Meisel 2021; Moyer et al. 2024).
These data can be explored at the event level (e.g., leader-trip) or aggregated to the monad (e.g., country-year) or dyad (e.g., country-pair-year). For the analysis of drivers, we aggregate trips to both monad- and dyad-year counts. But for the analysis exploring the drivers of bilateral trade, we focus on country counts of bilateral meetings between heads of state. Other units of analysis can be derived from the data, although the presence or absence of a specific meeting between two HOGS or groups of people is characterized by much greater uncertainty than the count of trips. Below is a list of examples of indicators that can be derived from COLT:
Bilateral “event”: A HOGS travels to a country and does not attend a multilateral event.
Multilateral “event”: A HOGS travels to a country and attends a multilateral event.
Bilateral “meeting”: A HOGS meets another HOGS face-to-face. This can happen either during a bilateral event with a host HOGS or in a multilateral event with a host-country HOGS or on the sideline of a multilateral event with a nonhost HOGS.
“Trip”: A HOGS travels to a country, regardless of whether the purpose is for a multilateral event, bilateral event, or other reasons.
Various combinations of the above: For example, only sideline meetings with nonhost HOGS at multilateral events or only bilateral meetings with a host HOGS.
To gather these data, coders begin by identifying the priority leader for each political unit (Gleditsch and Ward 1999), either the head of government (HOG) or the head of state (HOS), depending on the constitutional powers vested in the executive.3 Data on diplomatic visits and meetings are collected using primary sources where available, including (1) publicly available records of travel, including those produced on government-sponsored and international governmental organization websites; (2) content aggregation databases; and (3) online sources such as Getty Images and official social media accounts.4 To provide more context on data creation processes, we present four cases below.
Example 1: In June 2023, Indian Prime Minister Narendra Modi made a diplomatic trip to the United States. While visiting, Prime Minister Modi participated in several activities, including participating in the International Day of Yoga at the United Nations headquarters in New York, addressing the United States Congress, and participating in bilateral talks with President Joe Biden. Prime Minister Modi was also present at a White House state dinner, met with American CEOs, and signed several Memorandums of Understanding (MoUs) related to the transportation sector, semiconductor supply, and the manufacture of jet engines. Given the information provided, a meeting with the host country’s HOGS, US President Joe Biden, was captured in the MetHostHOGS variable, the address to Congress was captured under the PublicAddress variable, Modi’s meetings with CEOs was captured under the BusinessLeaderOrForum variable, and the signed MoUs were captured under the SignedAgreement variable. Using the confidence rating framework, Prime Minister Narendra Modi’s trip was coded with a level 1 confidence, indicating that there was no ambiguity in the trip details and that they were reported from reputable sources.
Example 2: US President Donald Trump traveled briefly to North Korea in June 2019. During his visit, President Trump participated in bilateral talks with North Korean Supreme Leader Kim Jong-un. This meeting with the host country’s HOGS was captured in the MetHostHOGS variable. Although this trip did not result in signed agreements or a public address, it was a significant event in international diplomacy. Using the confidence rating framework, President Donald Trump’s trip to North Korea was coded with a level 1 confidence rating, indicating that all variables were captured and there was no ambiguity in the trip details.
Example 3: Mauritania’s government website described President Mohamed Cheikh El Ghazouani’s trip to the United States in December 2022. While visiting the United States, Ghazouani attended and spoke at the US–Africa Leaders’ Summit, received a distinguished leadership award, and met with several CEOs on the sidelines of the summit. It was also reported that Ghazouani met with the International Monetary Fund’s (IMF) Managing Director and World Bank’s President. Given the information provided, attendance at the US–Africa Leaders’ Summit was captured in the AttendedMultilatEvent variable, the address at the summit was captured under PublicAddress, the distinguished leadership award was captured under the CulturalSiteOrCeremony variable, Ghazouani’s meeting with CEOs was captured under the BusinessLeaderOrForum variable, and the meetings with the IMF Managing Director and World Bank President were captured under the MetIGOLeader variable. Using the confidence rating framework, President Mohamed Cheikh El Ghazouani’s trip to the United States was coded with a level 1 confidence rating, indicating that all variables were captured and there was no ambiguity in the trip details. This high confidence rating is due in part to the extensive media coverage of this trip and the available resources from Mauritania.
Example 4: Seychelles President Wavel Ramkalawan traveled to the United Arab Emirates in December 2021. Although Seychelles has an official government website, limited trip details were released; it was reported that President Ramkalawan would be in the United Arab Emirates and when he would be visiting, but more information was not available. Given the limited information, we were only able to cover so much information within our data for this trip. We included the TripStartDate and TripEndDate variables along with the HOGS traveling and his destination, but the remaining variables regarding his activities abroad were reported as No. Using the confidence rating framework, President Wavel Ramkalawan’s trip to the United Arab Emirates was coded with a level 4, indicating that only basic information such as the date and location of the trip were available.
The data collection process involved systematically cleaning and vetting all trips through a tiered review process. After an original coder identifies and codes a trip, those data are provided to a second coder to check for incorrectly coded visits and missing variables. This first check includes investigating visits longer than 9 days, identifying blank variables, correcting misspellings, reviewing visits where all variables are marked as “no,” and verifying source links to ensure accurate and comprehensive interpretation of the trip details. Next, a third coder, referred to as the “vetter,” further interrogates the coded data by cross-referencing visits with official government websites and verified social media pages. If discrepancies are identified, the “vetter” makes corrections.
Upon completing the vetting process, the “vetter” provides feedback to the original coder and submits the vetted travel file to the project lead, an experienced research aide responsible for overseeing all data from a small sample of countries. The project lead also provides an additional quality review of the data based on sources provided by the primary coder. Afterward, the project manager—a full-time staff researcher assigned solely to the COLT project—conducts a similar review as the project lead, selecting a randomized sample of trips to double-code and vetting nontraveling leaders to confirm their lack of travel in each time period. Finally, a full-time senior researcher attached to the project conducts a final review of the data for cleanliness and accuracy.
We performed an initial test of the accuracy of COLT data against previously published travel data. Figure 1 compares COLT with Chinese, Turkish, and US leader travel (Kastner and Saunders 2012; Lebovic and Saunders 2016; Wang and Stone 2023; Balci and Pulat 2024). Overall, the correlations between these data sources and COLT are very high, ranging from 0.820 to 0.968. Differences in visits appear to be driven by different coding criteria and source material access. For example, Wang and Stone (2023) and Lebovic and Saunders (2016) exclude travel solely for a multilateral meeting, which COLT includes. Further, Kastner and Saunders (2012)combine both Chinese president and premier visits into a single unified count of travel, making it difficult to directly compare with COLT, which focuses only on China's general secretary as the priority leader (who since 1993 has also been the president).
Comparison between COLT and other published leader travel data. The Kastner and Saunders comparison is ranked by country from highest to lowest values for COLT data.
Balci and Pulat’s (2024) data are the most comparable to ours, although COLT only tracks either Turkey’s president or prime minister, depending on who was the priority leader at the time (1990–2014 president; 2014–2023 prime minister). Across the 1990s, Balci and Pulat reported more trips than were found in the COLT data, which we suspect is a product of their greater access to local, nontranslated sources from a time when the Internet was at its infancy. From 2000 onward, after the advent of more globalized media coverage and Internet expansion, the COLT data are very similar to Balci and Pulat’s.
We find that the COLT data possess a high degree of external validity with other travel data. However, we note that data collection has limitations and requires revision. We welcome users of the COLT dataset to provide feedback and flag errors via our GitHub issue tracker—we will then incorporate these revisions into future updates.5 Meanwhile, we caution users of the COLT data against drawing strong conclusions regarding non-English-speaking countries prior to the late 1990s, when non-English language Internet pages were drastically under-represented relative to native speakers (Graddol 2006) and Internet-based research and web-page archiving were relatively uncommon (Jensen 2011, 47). It may be that the leader trips are likelier to be missing for these countries in these early years of the COLT data, meaning that they would not be missing at random and as such there would be no statistical means to correct for this potential bias (Tsikriktsis 2005). From the mid-1990s onward, and especially since the dawn of the social media age, roughly defined as 2004 onward (Kaplan and Haenlein 2010), we expect online information used to code COLT data to be far more readily available.
Descriptive Analysis
From 1990 to 2023, the average leader traveled abroad 10.9 times (see Table 1). Bilateral events accounted for the most annual travel with 6.2 trips annually per leader, while multilateral events accounted for 4.7 trips a year per leader. Globally, the average number of HOGS trips has increased by 5.2 percent per year (normalized, calculated by dividing trip counts by the total number of countries). Unsurprisingly, the 2020 COVID pandemic significantly reduced overall leader travel, reducing the total global trips by 65.8 percent relative to 2019. By 2022, global travel counts had rebounded from the pandemic, with the annual count of HOGS trips (2,706) climbing back to just below the 2019 total count (2,779). By 2023, overall HOGS trips grew to a post-1990 high of 3,188 total.
Summary statistics for HOGS travel at the dyad-year and country-year
| Unit type | Mean | SD | Min | Max | |
|---|---|---|---|---|---|
| Dyad-tear | All | 0.054 | 0.324 | 0 | 33 |
| Dyad-year | Bilateral | 0.031 | 0.22 | 0 | 28 |
| Dyad-year | Multilateral | 0.023 | 0.2 | 0 | 22 |
| Country-year in trips | All | 10.884 | 21.642 | 0 | 289 |
| Country-year in trips | Bilateral | 6.201 | 10.36 | 0 | 111 |
| Country-year in trips | Multilateral | 4.684 | 13.879 | 0 | 234 |
| Country-year out trips | All | 10.903 | 9.106 | 0 | 115 |
| Country-year out trips | Bilateral | 6.222 | 6.365 | 0 | 107 |
| Country-year out trips | Multilateral | 4.682 | 4.257 | 0 | 28 |
| Unit type | Mean | SD | Min | Max | |
|---|---|---|---|---|---|
| Dyad-tear | All | 0.054 | 0.324 | 0 | 33 |
| Dyad-year | Bilateral | 0.031 | 0.22 | 0 | 28 |
| Dyad-year | Multilateral | 0.023 | 0.2 | 0 | 22 |
| Country-year in trips | All | 10.884 | 21.642 | 0 | 289 |
| Country-year in trips | Bilateral | 6.201 | 10.36 | 0 | 111 |
| Country-year in trips | Multilateral | 4.684 | 13.879 | 0 | 234 |
| Country-year out trips | All | 10.903 | 9.106 | 0 | 115 |
| Country-year out trips | Bilateral | 6.222 | 6.365 | 0 | 107 |
| Country-year out trips | Multilateral | 4.682 | 4.257 | 0 | 28 |
Note: Statistics are for the years 1990–2023, with country values representing pooled totals.
Summary statistics for HOGS travel at the dyad-year and country-year
| Unit type | Mean | SD | Min | Max | |
|---|---|---|---|---|---|
| Dyad-tear | All | 0.054 | 0.324 | 0 | 33 |
| Dyad-year | Bilateral | 0.031 | 0.22 | 0 | 28 |
| Dyad-year | Multilateral | 0.023 | 0.2 | 0 | 22 |
| Country-year in trips | All | 10.884 | 21.642 | 0 | 289 |
| Country-year in trips | Bilateral | 6.201 | 10.36 | 0 | 111 |
| Country-year in trips | Multilateral | 4.684 | 13.879 | 0 | 234 |
| Country-year out trips | All | 10.903 | 9.106 | 0 | 115 |
| Country-year out trips | Bilateral | 6.222 | 6.365 | 0 | 107 |
| Country-year out trips | Multilateral | 4.682 | 4.257 | 0 | 28 |
| Unit type | Mean | SD | Min | Max | |
|---|---|---|---|---|---|
| Dyad-tear | All | 0.054 | 0.324 | 0 | 33 |
| Dyad-year | Bilateral | 0.031 | 0.22 | 0 | 28 |
| Dyad-year | Multilateral | 0.023 | 0.2 | 0 | 22 |
| Country-year in trips | All | 10.884 | 21.642 | 0 | 289 |
| Country-year in trips | Bilateral | 6.201 | 10.36 | 0 | 111 |
| Country-year in trips | Multilateral | 4.684 | 13.879 | 0 | 234 |
| Country-year out trips | All | 10.903 | 9.106 | 0 | 115 |
| Country-year out trips | Bilateral | 6.222 | 6.365 | 0 | 107 |
| Country-year out trips | Multilateral | 4.682 | 4.257 | 0 | 28 |
Note: Statistics are for the years 1990–2023, with country values representing pooled totals.
European countries receive the largest number of average annual in-visits, followed by states from (in descending order) the Americas, Asia-Pacific, and Africa. The prominence of the Americas in received visits is biased by travel to the United States, which is compounded by annual global multilateral summits such as the UN General Assembly in New York City. However, the growing importance of multilateral organizations in Asia and Africa through the Association of Southeast Asian Nations (ASEAN) and the African Union (AU) has led to an increase of HOGS multilateral-event-related visits in recent years. The region with the fastest growth in normalized visits is Africa, with an average annual trip growth of 4.5 percent, followed by Europe at 4.3 percent, and the Asia-Pacific region at 3.7 percent.
Figure 2 shows the global distribution of countries ranked by the number of in-visits compared with out-visits. There are a handful of countries that receive a disproportionately large share of HOGS visits and relatively few countries that are disproportionately large travelers. Countries receiving the largest number of in-visits are most often relatively powerful states, as measured by Beckley's (2018) measure of “net” resources (GDP times GDP per capita, both at purchasing power parity (PPP)) or the Diplomacy, Military, and Economy (DiME) index (Moyer, Meisel, and Matthews 2023), along with those that are members of diplomatically engaged alliances, such as the European Union and NATO. The top 10 territories for cumulative visits received from 1990 through 2023 are (in order from greatest to fewest): the United States, Belgium, France, the United Kingdom, Germany, Russia, China, Italy, Egypt, and Saudi Arabia. The top ten territories that traveled most frequently from 1990 through 2023 are (also ranked greatest to fewest): Palestine, France, Germany, Italy, Jordan, Spain, Estonia, the United Kingdom, Senegal, and Poland. We make a special note that Palestine HOGS travel skews the distribution of out-visits significantly, making 1,460 visits compared with the second largest traveler, France, having 1,039 out-visits.
Total count of COLT visits into a country and total count of visits made to other countries, ranked from greatest to least (left to right) with each series independently ranked.
The net difference between in-visits and out-visits for trips is mapped in Figure 3. Countries shaded in blue are greater net receivers of foreign leader visits, while those in red are net senders. The top countries by average annual net count of in-visits versus out-visits are (highest to lowest) the United States, Belgium, France, the United Kingdom, Russia, China, Saudi Arabia, Germany, the United Arab Emirates, and Switzerland. The bottom ten are (ranked from lowest to highest): Palestine, Estonia, Luxembourg, Guinea-Bissau, Slovenia, Serbia, Central African Republic, Benin, Croatia, and Mali.
The country-level annual net in-visits minus out-visits averaged from 1990 to 2023 with higher values indicating more net in-visits.
Travel flows normalized by regime type show that that the average democratic country travels more often than the average autocratic country, as measured by the Varieties of Democracy (V-Dem) Institute’s polyarchy index (Kasuya and Mori 2019; Coppedge et al. 2024).6 COLT data show that democracies receive slightly more visits on average per year (13.79) than they send outward (13.08), while autocracies travel slightly more frequently per year on average (9.31) than they receive visiting leaders (9.23). However, there is a significant variation at the country level with several nondemocratic countries receiving more in-visits than out-visits. The plurality of trips in the COLT data is from HOGS in democratic states visiting other democracies (34,049). Autocrat-to-autocracy travel as a category was less than half as common (12,026) and autocracy-to-democracy travel outnumbered democracy-to-autocracy travel (10,180 and 9,619 trips, respectively).7
Drivers of Travel
Beyond descriptive statistics, we are also able to use the COLT data to assess common theories underpinning theories of diplomatic exchange in the international system. Figures 4 and 5 show findings from analysis conducted to better understand the drivers/correlates of monadic and dyadic travel counts. We used Poisson pseudo-maximum likelihood (PPML) regressions with multiway fixed effects (Correia, Guimarães, and Zylkin 2020) and country-clustered, heteroscedasticity- and autocorrelation-consistent standard errors (Freedman 2006). We find that a variety of established theoretical explanations for state behavior correlate with HOGS travel abroad.
Dyadic model results.
Monadic model results.
Foreign visits by HOGS are very costly by their nature (personnel, logistics, etc.), so decisions about where to go often involve considering trade (Koliev and Lundgren 2021). This inevitably entails a political decision whereby some partners are favored over others based on different factors. Starting from this baseline and following previous studies on the drivers of leader travel (e.g., Lebovic and Saunders 2016; Koliev and Lundgren 2021), we identify various logics that explain decision-making about diplomatic travel.
First, we focus on the logic of material interests, a correlate previous literature has highlighted as important in explaining the likelihood of both incoming and outgoing visits that may drive trade-offs in decision-making (Koliev and Lundgren 2021). In particular, leaders should seek major powers and/or economically interdependent states as their preferred partners (e.g., Lebovic and Saunders 2016). We therefore expect to find that countries prioritize visits to partners with greater wealth, where trade and arms transfer relationships are stronger, and where relations centered on foreign aid are established (e.g., Balci and Pulat 2024). In addition, conflict dynamics influence global diplomacy, as states are involved, for material and status-related factors to be involved in conflict-resolution processes (Ward 2020; Malis and Smith 2021).
Our findings show that economic interests are an important condition for leader travel, although the relationship with GDP per capita is statistically insignificant, perhaps because of multicollinearity with the national power measure used here (Beckley 2018). In the monadic model, HOGS from countries that are fossil fuel exporters both travel and receive more visits. Foreign aid is a significant factor in the dyadic model, as donor countries travel more often to recipient countries and vice versa. As previously identified by Malis and Smith (2021), arms trade between states is a significant predictor of leader visits; countries that receive weapons are more likely to visit their suppliers. Finally, countries experiencing an interstate or intrastate conflict are likely to travel more than those at peace. Our results show an indeterminate relationship with respect to in-visits, suggesting that states with lingering low-level conflict (such as India) are still recipients of visits. One explanation may be that sustained conflicts provide enough time for hosts to improve security conditions to the point that HOGS can safely insert themselves for critical diplomatic efforts (Ostrander and Rider 2019).
Second, we focus on a homophily logic. As already demonstrated (e.g., Ridgeway and Cornell 2006; Duque 2018), leaders can build their diplomatic agenda primarily with partners that share some values and characteristics. Scholars showed that such processes and factors influence alliances, leaders’ networks and trade as well (Maoz 2012; Sheafer et al. 2014). We also assess whether political regimes increase the likelihood of visits as well as vote agreement in the UN General Assembly. Our findings suggest that voting agreement between states in the UN General Assembly correlates significantly with bilateral leader travel. This signals that policy affinity could be encouraging deeper personal exchanges between the leaders of these systems, influencing their shared global values and voting tendencies in intergovernmental organizations such as the United Nations. Our results also show that shared regime type matters, with leaders from similar types of regimes (more/less democratic/autocratic) typically traveling to and receiving visits from states with similar levels of democratic governance.
A final set of potential drivers—those representing structural factors—also show a strong correlation with leader travel. Similar to the findings of Ostrander and Rider (2019), both the routineness and reciprocity of visits matter (Pouliot 2008; Hopf 2010). Separate models excluding the country and dyad fixed effects to include geographical distance found that increasing distance between states substantially and significantly corresponds with fewer numbers of trips, an unsurprising result given that greater distance also increases the costs associated with travel.
Travel and Trade
We provide an additional demonstration of the utility of our data by using them to assess the statistical relationship between HOGS travel and patterns of international trade. Specifically, we examined how bilateral HOGS travel (with a face-to-face meeting between the traveling and host HOGSs) may lead to shifts in a bilateral relationship (Wang and Stone 2023).
International trade is a key feature of most bilateral interstate relationships (Reuveny and Kang 1996; Kleinberg and Fordham 2010). Trade can be used to influence other states, in part through trade dependencies (Farrell and Newman 2019). Changes to trade flows are not only a significant dynamic in international relations but also directly influence consequential domestic political and economic developments that relate to state stability (Hainmueller and Hiscox 2006; Hiscox 2006). Previous works have established a method of quantitatively modeling bilateral trade, which we can leverage to incorporate HOGS visits to assess the impact of leader travel on this important outcome (Pollins 1989; Egger 2008).
To analyze this relationship, we rely on Anderson's (1979) gravity-based theory of trade, which argues that the economic size of two countries and the physical distance between them can explain and estimate a significant amount of their bilateral trade patterns. Although political economies and commercial interests have long been integrated into considerations of diplomatic exchange (Lee and Hudson 2004), the study of how HOGS travel affects these outcomes has focused particularly on China and the United States. While these studies found that an increase in diplomatic travel is a leading indicator of increased trade, their limited geographic coverage suggests that a global dataset could test the robustness of this finding (Thompson 1981; Nitsch 2007; Goldstein 2008; Zhang, van Witteloostuijn, and Elhorst 2011; Kastner and Saunders 2012; Fuchs and Klann 2013; Fuchs 2016).
Again, we estimate our models using PPML, which has been shown to provide robust estimations of bilateral trade relationships (Yotov et al. 2016). This model is described in equation (1), where i and j denote the countries of the traveling leader and the leader being visited, respectively; t the year; π time-invariant dyad fixed effects; χ year fixed effects; X1 through Xk a suite of time-varying monadic and dyadic factors; β1 through βk their respective relationships with trade; and ε the idiosyncratic error.
We also include lagged travel in our analyses for both the traveling leader’s country and the country they are visiting. The former is intended to control for the routineness of visits from Country A to Country B and the latter for reciprocity, where a prior trip from B to A may make a trip from A to B more likely. The results of this analysis are presented in Table 2, including a full sample of country-pairs, country-pairs that include at least one major power, and a sample that excludes major powers (as defined by the Correlates of War Project 2017).
Regression of bilateral goods trade and bilateral HOGS meetings on selected factors
| All dyads | Dyads with ≥1 major power | Dyads without a major power | ||||
|---|---|---|---|---|---|---|
| DV = bilateral goods trade | DV = trips, A to B with HOGS meeting | DV = bilateral goods trade | DV = trips, A to B with HOGS meeting | DV = bilateral goods trade | DV = trips, A to B with HOGS meeting | |
| Independent variables | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) |
| GDP at MER, A (1-year lag) | 0.000** (0.000) | −0.000** (0.000) | 0.000** (0.000) | −0.000** (0.000) | 0.000* (0.000) | −0.000 (0.000) |
| GDP at MER, B (1-year lag) | 0.000** (0.000) | −0.000** (0.000) | 0.000** (0.000) | −0.000** (0.000) | 0.000* (0.000) | −0.000** (0.000) |
| Power share, A (1-year lag) | −0.028 (0.069) | 0.058 (0.229) | 0.006 (0.098) | 0.679 (0.489) | −0.021 (0.052) | −0.133 (0.261) |
| Embassy, A in B (1-year lag) | 0.156** (0.021) | 0.139* (0.062) | 0.080* (0.037) | 0.187 (0.201) | 0.166** (0.021) | 0.123 (0.065) |
| Embassy, B in A (1-year lag) | 0.146** (0.021) | 0.157* (0.061) | 0.070 (0.028) | −0.051 (0.178) | 0.163** (0.021) | 0.165* (0.064) |
| Alliance (1-year lag) | 0.042 (0.024) | −0.042 (0.079) | 0.020 (0.028) | −0.106 (0.103) | 0.160** (0.024) | 0.095 (0.125) |
| Arms transfers, A to B | −0.000** (0.000) | 0.000 (0.000) | −0.000** (0.000) | −0.000 (0.000) | 0.000 (0.000) | 0.001 (0.001) |
| Regional trade agreement (1-year lag) | 0.161** (0.027) | −0.035 (0.092) | 0.287** (0.065) | 0.525** (0.171) | 0.063** (0.018) | −0.285** (0.107) |
| Polyarchy, A (1-year lag) | −0.352** (0.039) | 0.692** (0.146) | −0.340** (0.066) | −0.369 (0.311) | −0.321** (0.032) | 0.945** (0.172) |
| Polyarchy affinity (1-year lag) | −0.176** (0.029) | 0.564** (0.124) | −0.287** (0.050) | 1.470** (0.269) | −0.093** (0.026) | 0.375** (0.141) |
| UN General Assembly voting coincidence (1-year lag) | −0.163** (0.050) | 0.340 (0.221) | −0.114* (0.053) | −0.205 (0.329) | 0.494** (0.067) | 1.005** (0.312) |
| Conflict, A (1-year lag) | −0.031** (0.008) | 0.018 (0.036) | −0.028** (0.010) | 0.009 (0.061) | −0.024* (0.011) | 0.026 (0.044) |
| Conflict, B (1-year lag) | −0.029** (0.008) | 0.003 (0.034) | −0.026* (0.011) | 0.118* (0.057) | −0.025* (0.011) | −0.064 (0.043) |
| Foreign direct investment stocks, A in B (1-year lag) | −0.000** (0.000) | 0.000** (0.000) | −0.000** (0.000) | 0.000** (0.000) | −0.000 (0.000) | 0.000** (0.000) |
| Total goods trade, A with B (1-year lag) | 0.002** (0.000) | 0.003* (0.001) | 0.001** (0.000) | 0.001 (0.002) | 0.019** (0.001) | 0.018** (0.006) |
| Trips, A to B with HOGS meeting (1-year lag) | 0.013** (0.004) | −0.149** (0.019) | 0.012** (0.004) | −0.109** (0.026) | −0.010 (0.006) | −0.203** (0.027) |
| Trips, B to A with HOGS meeting (1-year lag) | 0.012** (0.004) | 0.229** (0.019) | 0.012** (0.004) | 0.129** (0.029) | −0.011 (0.006) | 0.278** (0.024) |
| Constant | 23.731** (0.081) | −2.777** (0.248) | 24.277** (0.116) | −2.023** (0.441) | 21.484** (0.074) | −3.466** (0.339) |
| Dyad fixed effects | Yes | Yes | Yes | Yes | Yes | Yes |
| Year fixed effects | Yes | Yes | Yes | Yes | Yes | Yes |
| Sample size | 478,382 | 98,584 | 40,390 | 22,090 | 437,992 | 76,494 |
| Pseudo R-squared | 0.991 | 0.136 | 0.990 | 0.173 | 0.982 | 0.116 |
| All dyads | Dyads with ≥1 major power | Dyads without a major power | ||||
|---|---|---|---|---|---|---|
| DV = bilateral goods trade | DV = trips, A to B with HOGS meeting | DV = bilateral goods trade | DV = trips, A to B with HOGS meeting | DV = bilateral goods trade | DV = trips, A to B with HOGS meeting | |
| Independent variables | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) |
| GDP at MER, A (1-year lag) | 0.000** (0.000) | −0.000** (0.000) | 0.000** (0.000) | −0.000** (0.000) | 0.000* (0.000) | −0.000 (0.000) |
| GDP at MER, B (1-year lag) | 0.000** (0.000) | −0.000** (0.000) | 0.000** (0.000) | −0.000** (0.000) | 0.000* (0.000) | −0.000** (0.000) |
| Power share, A (1-year lag) | −0.028 (0.069) | 0.058 (0.229) | 0.006 (0.098) | 0.679 (0.489) | −0.021 (0.052) | −0.133 (0.261) |
| Embassy, A in B (1-year lag) | 0.156** (0.021) | 0.139* (0.062) | 0.080* (0.037) | 0.187 (0.201) | 0.166** (0.021) | 0.123 (0.065) |
| Embassy, B in A (1-year lag) | 0.146** (0.021) | 0.157* (0.061) | 0.070 (0.028) | −0.051 (0.178) | 0.163** (0.021) | 0.165* (0.064) |
| Alliance (1-year lag) | 0.042 (0.024) | −0.042 (0.079) | 0.020 (0.028) | −0.106 (0.103) | 0.160** (0.024) | 0.095 (0.125) |
| Arms transfers, A to B | −0.000** (0.000) | 0.000 (0.000) | −0.000** (0.000) | −0.000 (0.000) | 0.000 (0.000) | 0.001 (0.001) |
| Regional trade agreement (1-year lag) | 0.161** (0.027) | −0.035 (0.092) | 0.287** (0.065) | 0.525** (0.171) | 0.063** (0.018) | −0.285** (0.107) |
| Polyarchy, A (1-year lag) | −0.352** (0.039) | 0.692** (0.146) | −0.340** (0.066) | −0.369 (0.311) | −0.321** (0.032) | 0.945** (0.172) |
| Polyarchy affinity (1-year lag) | −0.176** (0.029) | 0.564** (0.124) | −0.287** (0.050) | 1.470** (0.269) | −0.093** (0.026) | 0.375** (0.141) |
| UN General Assembly voting coincidence (1-year lag) | −0.163** (0.050) | 0.340 (0.221) | −0.114* (0.053) | −0.205 (0.329) | 0.494** (0.067) | 1.005** (0.312) |
| Conflict, A (1-year lag) | −0.031** (0.008) | 0.018 (0.036) | −0.028** (0.010) | 0.009 (0.061) | −0.024* (0.011) | 0.026 (0.044) |
| Conflict, B (1-year lag) | −0.029** (0.008) | 0.003 (0.034) | −0.026* (0.011) | 0.118* (0.057) | −0.025* (0.011) | −0.064 (0.043) |
| Foreign direct investment stocks, A in B (1-year lag) | −0.000** (0.000) | 0.000** (0.000) | −0.000** (0.000) | 0.000** (0.000) | −0.000 (0.000) | 0.000** (0.000) |
| Total goods trade, A with B (1-year lag) | 0.002** (0.000) | 0.003* (0.001) | 0.001** (0.000) | 0.001 (0.002) | 0.019** (0.001) | 0.018** (0.006) |
| Trips, A to B with HOGS meeting (1-year lag) | 0.013** (0.004) | −0.149** (0.019) | 0.012** (0.004) | −0.109** (0.026) | −0.010 (0.006) | −0.203** (0.027) |
| Trips, B to A with HOGS meeting (1-year lag) | 0.012** (0.004) | 0.229** (0.019) | 0.012** (0.004) | 0.129** (0.029) | −0.011 (0.006) | 0.278** (0.024) |
| Constant | 23.731** (0.081) | −2.777** (0.248) | 24.277** (0.116) | −2.023** (0.441) | 21.484** (0.074) | −3.466** (0.339) |
| Dyad fixed effects | Yes | Yes | Yes | Yes | Yes | Yes |
| Year fixed effects | Yes | Yes | Yes | Yes | Yes | Yes |
| Sample size | 478,382 | 98,584 | 40,390 | 22,090 | 437,992 | 76,494 |
| Pseudo R-squared | 0.991 | 0.136 | 0.990 | 0.173 | 0.982 | 0.116 |
Source: DV = Dependent Variable GDP = Gross Domestic Product MER = Market Exchange Rates SE = Standard Error UN = United Nations. * p < 0.05 (95% confidence); ** p < 0.01 (99% confidence).
Regression of bilateral goods trade and bilateral HOGS meetings on selected factors
| All dyads | Dyads with ≥1 major power | Dyads without a major power | ||||
|---|---|---|---|---|---|---|
| DV = bilateral goods trade | DV = trips, A to B with HOGS meeting | DV = bilateral goods trade | DV = trips, A to B with HOGS meeting | DV = bilateral goods trade | DV = trips, A to B with HOGS meeting | |
| Independent variables | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) |
| GDP at MER, A (1-year lag) | 0.000** (0.000) | −0.000** (0.000) | 0.000** (0.000) | −0.000** (0.000) | 0.000* (0.000) | −0.000 (0.000) |
| GDP at MER, B (1-year lag) | 0.000** (0.000) | −0.000** (0.000) | 0.000** (0.000) | −0.000** (0.000) | 0.000* (0.000) | −0.000** (0.000) |
| Power share, A (1-year lag) | −0.028 (0.069) | 0.058 (0.229) | 0.006 (0.098) | 0.679 (0.489) | −0.021 (0.052) | −0.133 (0.261) |
| Embassy, A in B (1-year lag) | 0.156** (0.021) | 0.139* (0.062) | 0.080* (0.037) | 0.187 (0.201) | 0.166** (0.021) | 0.123 (0.065) |
| Embassy, B in A (1-year lag) | 0.146** (0.021) | 0.157* (0.061) | 0.070 (0.028) | −0.051 (0.178) | 0.163** (0.021) | 0.165* (0.064) |
| Alliance (1-year lag) | 0.042 (0.024) | −0.042 (0.079) | 0.020 (0.028) | −0.106 (0.103) | 0.160** (0.024) | 0.095 (0.125) |
| Arms transfers, A to B | −0.000** (0.000) | 0.000 (0.000) | −0.000** (0.000) | −0.000 (0.000) | 0.000 (0.000) | 0.001 (0.001) |
| Regional trade agreement (1-year lag) | 0.161** (0.027) | −0.035 (0.092) | 0.287** (0.065) | 0.525** (0.171) | 0.063** (0.018) | −0.285** (0.107) |
| Polyarchy, A (1-year lag) | −0.352** (0.039) | 0.692** (0.146) | −0.340** (0.066) | −0.369 (0.311) | −0.321** (0.032) | 0.945** (0.172) |
| Polyarchy affinity (1-year lag) | −0.176** (0.029) | 0.564** (0.124) | −0.287** (0.050) | 1.470** (0.269) | −0.093** (0.026) | 0.375** (0.141) |
| UN General Assembly voting coincidence (1-year lag) | −0.163** (0.050) | 0.340 (0.221) | −0.114* (0.053) | −0.205 (0.329) | 0.494** (0.067) | 1.005** (0.312) |
| Conflict, A (1-year lag) | −0.031** (0.008) | 0.018 (0.036) | −0.028** (0.010) | 0.009 (0.061) | −0.024* (0.011) | 0.026 (0.044) |
| Conflict, B (1-year lag) | −0.029** (0.008) | 0.003 (0.034) | −0.026* (0.011) | 0.118* (0.057) | −0.025* (0.011) | −0.064 (0.043) |
| Foreign direct investment stocks, A in B (1-year lag) | −0.000** (0.000) | 0.000** (0.000) | −0.000** (0.000) | 0.000** (0.000) | −0.000 (0.000) | 0.000** (0.000) |
| Total goods trade, A with B (1-year lag) | 0.002** (0.000) | 0.003* (0.001) | 0.001** (0.000) | 0.001 (0.002) | 0.019** (0.001) | 0.018** (0.006) |
| Trips, A to B with HOGS meeting (1-year lag) | 0.013** (0.004) | −0.149** (0.019) | 0.012** (0.004) | −0.109** (0.026) | −0.010 (0.006) | −0.203** (0.027) |
| Trips, B to A with HOGS meeting (1-year lag) | 0.012** (0.004) | 0.229** (0.019) | 0.012** (0.004) | 0.129** (0.029) | −0.011 (0.006) | 0.278** (0.024) |
| Constant | 23.731** (0.081) | −2.777** (0.248) | 24.277** (0.116) | −2.023** (0.441) | 21.484** (0.074) | −3.466** (0.339) |
| Dyad fixed effects | Yes | Yes | Yes | Yes | Yes | Yes |
| Year fixed effects | Yes | Yes | Yes | Yes | Yes | Yes |
| Sample size | 478,382 | 98,584 | 40,390 | 22,090 | 437,992 | 76,494 |
| Pseudo R-squared | 0.991 | 0.136 | 0.990 | 0.173 | 0.982 | 0.116 |
| All dyads | Dyads with ≥1 major power | Dyads without a major power | ||||
|---|---|---|---|---|---|---|
| DV = bilateral goods trade | DV = trips, A to B with HOGS meeting | DV = bilateral goods trade | DV = trips, A to B with HOGS meeting | DV = bilateral goods trade | DV = trips, A to B with HOGS meeting | |
| Independent variables | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) | Coefficient (SE) |
| GDP at MER, A (1-year lag) | 0.000** (0.000) | −0.000** (0.000) | 0.000** (0.000) | −0.000** (0.000) | 0.000* (0.000) | −0.000 (0.000) |
| GDP at MER, B (1-year lag) | 0.000** (0.000) | −0.000** (0.000) | 0.000** (0.000) | −0.000** (0.000) | 0.000* (0.000) | −0.000** (0.000) |
| Power share, A (1-year lag) | −0.028 (0.069) | 0.058 (0.229) | 0.006 (0.098) | 0.679 (0.489) | −0.021 (0.052) | −0.133 (0.261) |
| Embassy, A in B (1-year lag) | 0.156** (0.021) | 0.139* (0.062) | 0.080* (0.037) | 0.187 (0.201) | 0.166** (0.021) | 0.123 (0.065) |
| Embassy, B in A (1-year lag) | 0.146** (0.021) | 0.157* (0.061) | 0.070 (0.028) | −0.051 (0.178) | 0.163** (0.021) | 0.165* (0.064) |
| Alliance (1-year lag) | 0.042 (0.024) | −0.042 (0.079) | 0.020 (0.028) | −0.106 (0.103) | 0.160** (0.024) | 0.095 (0.125) |
| Arms transfers, A to B | −0.000** (0.000) | 0.000 (0.000) | −0.000** (0.000) | −0.000 (0.000) | 0.000 (0.000) | 0.001 (0.001) |
| Regional trade agreement (1-year lag) | 0.161** (0.027) | −0.035 (0.092) | 0.287** (0.065) | 0.525** (0.171) | 0.063** (0.018) | −0.285** (0.107) |
| Polyarchy, A (1-year lag) | −0.352** (0.039) | 0.692** (0.146) | −0.340** (0.066) | −0.369 (0.311) | −0.321** (0.032) | 0.945** (0.172) |
| Polyarchy affinity (1-year lag) | −0.176** (0.029) | 0.564** (0.124) | −0.287** (0.050) | 1.470** (0.269) | −0.093** (0.026) | 0.375** (0.141) |
| UN General Assembly voting coincidence (1-year lag) | −0.163** (0.050) | 0.340 (0.221) | −0.114* (0.053) | −0.205 (0.329) | 0.494** (0.067) | 1.005** (0.312) |
| Conflict, A (1-year lag) | −0.031** (0.008) | 0.018 (0.036) | −0.028** (0.010) | 0.009 (0.061) | −0.024* (0.011) | 0.026 (0.044) |
| Conflict, B (1-year lag) | −0.029** (0.008) | 0.003 (0.034) | −0.026* (0.011) | 0.118* (0.057) | −0.025* (0.011) | −0.064 (0.043) |
| Foreign direct investment stocks, A in B (1-year lag) | −0.000** (0.000) | 0.000** (0.000) | −0.000** (0.000) | 0.000** (0.000) | −0.000 (0.000) | 0.000** (0.000) |
| Total goods trade, A with B (1-year lag) | 0.002** (0.000) | 0.003* (0.001) | 0.001** (0.000) | 0.001 (0.002) | 0.019** (0.001) | 0.018** (0.006) |
| Trips, A to B with HOGS meeting (1-year lag) | 0.013** (0.004) | −0.149** (0.019) | 0.012** (0.004) | −0.109** (0.026) | −0.010 (0.006) | −0.203** (0.027) |
| Trips, B to A with HOGS meeting (1-year lag) | 0.012** (0.004) | 0.229** (0.019) | 0.012** (0.004) | 0.129** (0.029) | −0.011 (0.006) | 0.278** (0.024) |
| Constant | 23.731** (0.081) | −2.777** (0.248) | 24.277** (0.116) | −2.023** (0.441) | 21.484** (0.074) | −3.466** (0.339) |
| Dyad fixed effects | Yes | Yes | Yes | Yes | Yes | Yes |
| Year fixed effects | Yes | Yes | Yes | Yes | Yes | Yes |
| Sample size | 478,382 | 98,584 | 40,390 | 22,090 | 437,992 | 76,494 |
| Pseudo R-squared | 0.991 | 0.136 | 0.990 | 0.173 | 0.982 | 0.116 |
Source: DV = Dependent Variable GDP = Gross Domestic Product MER = Market Exchange Rates SE = Standard Error UN = United Nations. * p < 0.05 (95% confidence); ** p < 0.01 (99% confidence).
We find that bilateral HOGS travel with a face-to-face meeting is a leading indicator of increased trade between dyads. Using a 1-year lag, we find that a one-trip increase corresponds with an increase in goods trade between two countries by 1.2–1.3 percent.8 When compared with other drivers of trade identified in the literature, such as the GDP of the two trading countries or the distance between them (see Head and Mayer 2014 for a review of drivers), we find leader travel to be an important but auxiliary factor with respect to increased trade. For example, the presence of an embassy corresponds with an increase in bilateral goods trade of 15.8–16.9 percent, relative to a country-pair without diplomatic exchange holding all else equal. A regional trade agreement corresponds with a 17.5 percent increase in trade. We note, however, that this correlational analysis does not establish a causal relationship, but as Donnelly (2023, 64–69) argues, we must think of such interactions in terms of “systemic causation” rather than “independent variables as causes.”
Despite previous findings suggesting that leader travel drives trade (Nitsch 2007), our exploration of this and the reversed relationship—trade as a driver of travel—illustrates that neither hypothesis passes a causality test (Lopez and Weber 2017). In other words, there is a correlation between travel and trade in outward and inward directions when lagging whichever component is the independent variable. As such, the causal relationship does not clearly flow in only one direction or the other.
Our findings generally hold when examining dyads with and without major powers. One notable difference is the sign, size, and significance of the coefficients for routine and reciprocal visits with meetings [“Trips, A to B with HOGS meeting (1-year lag)” and “Trips, B to A with HOGS meeting (1-year lag)”] when analyzing trade between country-pairs without a major power. Rather than demonstrating a highly statistically significant, positive relationship with bilateral trade, routine and reciprocal visits between non-major powers demonstrated a negative, albeit statistically insignificant, relationship with bilateral trade. This difference in statistical relationships—as with future differences that can now be uncovered using the COLT dataset—merits further exploration in future research.
Conclusion
The use of data in international relations is not a panacea for building knowledge about our complex global system. But data, when developed transparently and produced for external review and iteration, can act as a “model” to help us understand and evaluate key dynamics in the international system. International relations is characterized by deep social interaction, constructed patterns of engagement, and an ideational superstructure that connects to a “rump material” core (Wendt 1999, 110). As such, data can only help us go so far.
Data building in the social sciences should be understood in the way George Box understood statistics, as models that are wrong (said another way, they do not accurately reflect the reality of the world as an open system) but that still can be useful (Box 1976; Luhmann 1995; Donnelly 2023). A model is a representation of reality; the best models are those that are transparent, designed to be interrogated, and those used to build an understanding of how our complex world has changed or remained the same, and what humans might do to make the future better. Data in the social sciences are useful when they advance these outcomes. Data that are black boxes—tools that are meant to be used but not understood—can be dangerous. The COLT data series is meant to be interrogated and used—we hope that making this resource public will allow others to highlight gaps and areas where we can improve our coding (see, e.g., our GitHub resource referenced earlier in this article). With this in mind, we believe that this dataset can help us better understand how actors are shaped and shoved by forces in the international system and how they in turn make global politics what it is and what it might be (Avant 2024).
Funder Information
Funding for this project was provided by the US Government. The results and views expressed are those of the authors alone and do not represent the views of the US Government.
Author Biography
Jonathan D. Moyer is an Associate Professor at the Josef Korbel School of International Studies, University of Denver, USA and Director of the Frederick S. Pardee Institute for International Futures and is corresponding author ([email protected]).
Collin J. Meisel is the Associate Director of Geopolitical Analysis at the Pardee Institute for International Futures, University of Denver, USA, a senior fellow with The Hague Centre for Strategic Studies a nonresident fellow with The Henry L. Stimson Center, and a term member of the Council on Foreign Relations.
Adam Szymanski-Burgos is a Statistical Analyst with the Colorado Department of Labor and Employment Labor Market Information team. He formerly worked as a Research Associate with the Frederick S. Pardee Institute for International Futures at the University of Denver, USA.
Andrew C. Scott was a Senior Research Associate at the Pardee Center for International Futures, USA, who led the original scaling-up of the COLT data project. His interests are in intrastate relations and global diplomacy.
Matteo C.M. Casiraghi is an Assistant Professor of International Relations at the University of Groningen, Netherlands, and a former Marie Curie Global Fellow. His research focuses on international norms, public opinion on war, and global diplomacy.
Alexandra Kurkul formerly served as Project Manager for the Country and Organization Leader Travel Project and a Research Associate at the Frederick S. Pardee Institute for International Futures, University of Denver, USA.
Marianne Hughes is Deputy Project Manager of the Research and Analysis Program, providing contract support for the US Trade and Development Agency. She formerly served as Deputy Project Manager at the Frederick S. Pardee Institute for International Futures, University of Denver, USA.
Whitney Kettlun is a Senior Data Analyst at S&P Global. She has a background in international studies, with research interests in international development and data-driven approaches to global development challenges.
Kylie X. McKee is the Project Manager for the Country and Organization Leader Travel (COLT) Project and a Research Associate at the Frederick S. Pardee Institute for International Futures, University of Denver, USA.
Austin S. Matthews is an Assistant Professor of Political Science at East Carolina University, USA. His research focuses on authoritarian politics, civil–military regimes, and political violence.
Notes
Authors’ note: The authors would like to thank the hundreds of undergraduate and graduate students who contributed to this dataset. They would also like to thank the university and institute administrators who make this work possible. The GBAT database estimates that 27 percent of the references in this paper are from female scholars. The data for this project as well as replication files can be found at https://dataverse.harvard.edu/dataverse/isq.
Footnotes
Replication files, data, and codebook can be found here: https://dataverse.harvard.edu/dataverse/isq.
For example, we identify whether a “visit” is the HOGS residing in exile, which would technically qualify in our set. We use these flags to allow researchers to omit trips like these that would not fit their intention.
The priority leader for each country-year in our sample (1) serves as the top representative of the country in the international arena, (2) is primarily responsible for implementing foreign policy, and (3) is not primarily ceremonial. For countries with a separate HOG and HOS, the priority leader was identified using the Archigos dataset (Goemans, Gleditsch, and Chiozza 2009) except when in contradiction to criteria 1–3.
Databases included ProQuest, Nexis Uni, Foreign Broadcast Information Service Daily Reports, and World News Connection. Additionally, Getty Images was used to confirm travel and meetings from 2005; Twitter/X for travel from 2010.
Here we define democracy using the V-Dem Institute's polyarchy score using a threshold of 0.42, as recommended by Kasuya and Mori (2019).
For 4,888 trips, either the country of the traveling leader or host country did not have a polyarchy score in that year, not allowing us to make comparisons in these cases based on regime type.
PPML models produce coefficients that are defined by elasticities (the percent change in Y divided by the percent change in X) for log-transformed independent variables (e.g., natural log of GDP) and semi-elasticities (the percent change in Y divided by the unit change in X) for linear independent variables [e.g., Trips, A to B with HOGS meeting (1-year lag)]. Thus, with respect to coefficients that function as elasticities, the PPML model coefficients can be interpreted as percentages (e.g., 0.984 equals 0.984 percent). With respect to semi-elasticities, the interpretation is more complex. The relationship is equal to (eβ–1) × 100 percent, with e representing Euler's number and β the linear independent variable's coefficient.
