Who decides who gets in? Diplomats, bureaucrats, and visa issuance

Abstract

Does resident diplomacy influence international outcomes? Theoretically, I argue that resident diplomats tend to adopt uniquely cooperative stances toward their hosts. I test this expectation using a natural experiment involving British visa issuance. Starting in 2007, the UK transferred visa decision making from local diplomatic posts to centralized hubs, located either at third country diplomatic posts or domestically. I study this rollout to credibly estimate the causal effects of visa adjudication by local posts. I find that resident diplomats implement a much more lenient visa policy—transferring adjudication to an outside hub reduces issuance by about fifteen percent. There is a robust difference between the behavior of local and third-country posts, showing that this cooperative effect of diplomacy is relationship-specific.

In October 2009, British Defence Secretary Bob Ainsworth and Home Secretary Alan Johnson traveled to Islamabad for a summit with the Pakistani leadership. Major issues loomed on the bilateral agenda including the escalating war in Afghanistan, counterterrorism, and the fate of hundreds of millions of pounds in foreign aid. But for the Pakistani leadership, another issue overshadowed all of these—the difficulties faced by Pakistani nationals in obtaining British visas (Morgan, 2009). In the months leading up to the visit, the refusal rate for Pakistani visa applicants had jumped from 39 percent to 57 percent, much to the dismay of applicants and the Pakistani government. This led the top British diplomat in Pakistan to describe visas as the “single most toxic” problem in the relationship (Vine, 2010, p. 17).

The dramatic increase in the visa refusal rate did not come from any deliberate change in immigration criteria. Instead, it was an unintended side effect of an administrative adjustment, whereby Britain transferred visa casework for Pakistan from its diplomatic posts in the country to a regional hub in Abu Dhabi. An internal investigation would later conclude that staff at this hub had enforced visa criteria excessively stringently against Pakistani as compared to Emirati applicants, driving up the refusal rate and sparking the diplomatic row (Vine, 2010). Changing the location of visa adjudication had a major impact on outcomes. As I will show below, this is part of a general pattern in which resident diplomats implement visa policy differently than other officials.

Scholarly interest in the effects of resident diplomacy has grown considerably in recent years. One of the core questions animating this work is whether resident diplomats produce different outcomes than other officials or institutional arrangements. Visa issuance provides an important window into this topic, given that it has long been one of the core functions associated with diplomatic missions and plays an important role in diverse areas of international interaction. Visa decisions directly shape global flows of people. In turn, this impacts the structure of global trade, investment, knowledge diffusion, and cultural exchange. All of this makes access to visas a major objective for many countries. In 2008, for example, 45 percent of Macedonians identified visa liberalization as their top foreign policy priority (Reactor: Research in Action, 2009).

Visa diplomacy is often asymmetric. Nationals of the most developed countries typically have little trouble obtaining visas, and their governments tend not to emphasize the issue. But, when visas are hard to obtain, easing restrictions becomes crucial. In 1994, for example, the Irish Foreign Minister announced that tourist visas for Irish citizens to attend the World Cup were a higher priority for his country than accession to the World Trade Organization (Callahan and Stuart Kennedy, 2017, p. 55). A decade later, the Polish government identified an easing of American visa requirements for Poles as the crucial item on “a list of favors that [it] expected the US to grant Poland as a payoff” for its participation in the Iraq War (Lubecki, 2005, p. 78).

As I will argue below, resident diplomats are uniquely attuned to these host nation interests. Using a natural experiment, I show that resident diplomats issue visas much more generously than other officials. In particular, I compare the visa practices of local diplomatic posts to those of both diplomatic posts based in a third country and home-based immigration bureaucrats. I find that local posts issue considerably more visas than either alternative. This result is robust across empirical specifications and substantively large. Some evidence suggests that third country hubs grant more visas than domestic ones, but this effect is not as large and is not consistent across empirical specifications.

These results have important implications for both diplomacy and international migration as well as clear policy relevance. From the perspective of diplomacy, they reinforce the conclusion of a growing body of literature showing that delegated diplomacy matters and that diplomats have a special connection with their host countries that tends to promote cooperative outcomes. In terms of migration, I show that substantial variation can be explained simply through changes in administrative arrangements even when de jure policies are unchanged. On the policy side, policymakers should be aware that the form and location of visa adjudication can have a large effect on outcomes.

1. Diplomacy and international outcomes

The study of diplomacy has experienced something of a renaissance in recent years (Trager, 2016). While much of this work has taken a unitary actor approach, interest is also growing in the specific role played by embassies and diplomats (Lindsey, 2017; Gertz, 2018; Malis, 2021; Schub, 2022). At the same time, an influential body of public commentary holds that delegated institutions are irrelevant and anachronistic in light of modern communications technology. Gone are the days when an ambassador could act autonomously without meaningful oversight from home, and technology now allows central authorities to bypass diplomatic posts if they like.

Potentially, diplomacy matters but diplomats do not. Implicitly, unitary actor theories of diplomacy take this position, focusing on some kind of alignment of interstate interests rather than the independent behavior of diplomats (Sartori, 2005; Trager, 2017). Diplomats might be mere functionaries with no independent effect. But, as the principal-agent literature shows, even if diplomats are merely agents and not policymakers, they may still have a great deal of influence.

At a correlational level, diplomatic exchange is strongly associated with more cooperative international outcomes, suggesting some role for diplomats. There is particularly extensive evidence for this when it comes to trade (see Moons and Bergeijk (2017) for a meta-analysis). Other studies have found similar results for treaty negotiations, dispute resolution, conflict and other areas (Plouffe and van der Sterren, 2016; Gray and Potter, 2020). Of course, the selection problem here is obvious—perhaps states have good relations because they have exchanged diplomats but perhaps they exchange diplomats because they have good relations.

Three recent studies have responded to this by using exogenous variation in diplomatic representation. Interestingly, these all rely on the same source—vacancies in American ambassadorships. American ambassadors often leave their posts for personnel reasons that are plausibly exogenous to the bilateral relationship. Gertz (2018) finds that these vacancies increase the risk of investor-state arbitration disputes. Malis (2021) finds they increase the risk of militarized disputes and reduce American exports. Ahmed and Slaski (2022) show that this is amplified when a career diplomat departs. An ambassadorial vacancy, however, is not equivalent to closing an embassy or transferring its functions. That is, these compare the effects of a permanent ambassador as opposed to a subordinate filling the role temporarily. In contrast, I aim to compare the behavior of a resident diplomatic mission to non-resident arrangements.

2. Visa diplomacy

I focus here on visas. Visas are a laboratory for studying the effect of resident diplomacy more broadly, but they are also important in their own right. Even single visa decisions can be significant (Stringer, 2004), but the principal impact is at the aggregate level. Visa restrictions substantially reduce bilateral travel and the circulation of people. This, in turn, constricts downstream flows of goods, money, and ideas (Czaika and Neumayer, 2017). Umana-Dajud (2019), for example, estimates that existing visa requirements reduce gross national welfare in developing countries by an average of 1.1 percent and by up to 5 percent for certain countries by functioning as a de facto trade barrier.

For the issuing country, visas present a tradeoff. Restrictions reduce travel, trade, and so on, but they also may be necessary to screen out undesired travelers. For the origin country (i.e., the country whose nationals are seeking visas), incentives are more straightforward—the economic and other benefits are present but the security costs are not—and countries usually want it to be easier for their own nationals to secure visas, at least when it comes to short-term visits. Motives are somewhat more ambiguous for long-term visas where the origin country might wish to limit or reduce outward migration and potential brain drain, but the overall direction of diplomatic pressure leans toward liberalization (Rosenblum, 2004).

While most research to date has focused on the formal rules, there is a second aspect to the strictness of a visa regime in practice—the behavior of the agents who grant visas. That is, the difficulty of obtaining a visa depends not only on central rules but also on the way that diplomats or other government officials use their discretion in applying these. The “street-level” behavior of visa officials thus has important implications for migration generally (Zampagni, 2016). With this in mind, I turn to a series of theoretical expectations about the way diplomatic posts will handle visas.

3. A theory of diplomatic preferences and visa administration

Analysts have long recognized that different foreign policy bureaucracies hold distinctive preferences and outlooks toward international affairs (Allison and Zelikow, 1999). Diplomats are often described as holding generally dovish, cooperative, or accommodating views along with a skillset that tends toward negotiation and compromise (Rosati and DeWitt, 2012, p. 186).

While diplomats may prefer cooperation, I argue that the typical case is more nuanced. We should not expect diplomats to adopt a monolithically accommodating position; rather, as the result of some combination of socialization, selection, skills and institutional incentives, we should expect diplomats to take an accommodating stance toward their host countries in particular rather than foreign countries in general. In terms of socialization, living and working in the host country may be able to favorably influence diplomats’ views and attitudes, a phenomenon commonly known as “clientitis” or “going native” (Rubin, 1987, p. 247). Selection pressures will often push in the same direction. Lindsey (2023) finds that countries tend to select diplomats who are relatively sympathetic toward their hosts. These forces are specific to the host country and do not necessarily translate toward a different overall attitude on cooperation.

Institutionally, resident diplomats will also tend to contextualize visa administration (or any other administrative task) within the broader diplomatic context, rather than treating it as an isolated, migration-control device. Migration is but one element of the broader interstate relationship. Immigration bureaucrats operating outside the diplomatic chain of command have little reason to concern themselves with broader political or economic considerations. On the other hand, diplomats—and especially senior embassy officials—have an incentive to treat visas as a diplomatic tool that can be used to pursue other bilateral objectives. An ambassador charged with promoting trade, for example, is likely to recognize that a more liberal stance on visas opens up trade opportunities. Similarly, a relaxed application of visa criteria might function as an easy side payment, and senior diplomats will recognize the public opinion costs of an overly harsh application of visa rules.

Within the British context, ambassadors and senior diplomats do not directly issue visas. But, when a diplomatic post issues visas, the entry clearance officers handling visas are supervised by the post leadership and “under the overall authority of heads of mission” (House of Commons Foreign Affairs Committee, 2009, p. 43). This reflects a general principle of organization by which the senior British diplomat at a post, acting as head of mission, exercises “overall control over the whole operation, whether or not that involves reporting back to the Foreign Office” (House of Commons Foreign Affairs Committee, 2011, p. 63). Thus, when a local diplomatic post issues visas, the ambassador has supervisory authority to set the overall tone and climate of the operation, but holds no such authority when visas are adjudicated elsewhere.

I expect, therefore, that local diplomatic posts will operate more leniently when they adjudicate visas given the channels described above. That is, entry clearance officers based in a given country are likely to have some affinity for the country and its people based on living and working among them. They are also responsible to diplomatic superiors who are likely to incentivize a more lenient approach. The empirical analysis below will allow me to test this expectation against two competing theories: the view that resident diplomacy does not matter, and the view that diplomats are more accommodating across the board (rather than specifically toward their hosts). Before testing these, I sketch the specifics of the British system and the policy shock that allows for the analysis here.

3.1 The British visa system

Within the British system, visa rules and policies are established centrally under the authority of the Home Secretary (Gower, 2014). Entry clearance officers, based at “Decision Making Centres” (DMCs) either at home or abroad apply these rules and issue or deny visas. Permanent entry clearance officers can enter the role through the Home Office or the Foreign Office, and individual officers rotate over time between foreign and domestic postings regardless of their background. That is, domestic DMCs are not specifically staffed by officers with a Home Office background nor are overseas DMCs specifically staffed by officers with a diplomatic service background (Home Office, 2017). Both domestic and foreign DMCs also sometimes engage entry clearance officers locally—often retired civil servants—as needed (Bolt, 2017). In sum, what varies based on DMC type is not the staff but rather the nature of their supervising authority and the socialization effects to which they are exposed in their environment.

Visa decision-making involves considerable discretion. The formal rules are relatively elastic, and many criteria are explicitly discretionary. DMCs also frequently bend the rules. A 2011 global review, for example, identified deviations from the governing rules in 33 percent of sampled visa decisions (Vine, 2011, p. 2). Individual entry clearance officers within a DMC are subject to tight supervision by post management. Front-line supervisors routinely review a large percentage of visa decisions and consider complaints or appeals.¹ But routine oversight does not extend above the post. There is no established internal process to review applications for visitor visas above the DMC level aside from spot checks as part of broader reviews, and these decisions are not subject to judicial review (Bolt, 2015b, p. 23). For longer-term visas, applicants have broader (and sometimes judicial) appeal rights. While this provides some constraint, entry clearance officers face no personal consequences if their decisions are overturned on appeal, and do not even receive notification when this occurs (Vine, 2010, p. 43). This setup provides tight control for local superiors but little day-to-day supervision from other sources. For overseas DMCs, this local accountability chain ends at the top diplomat at the post issuing the visa. For domestic DMCs, it runs to the senior Home Office civil servant overseeing the facility.

Prior to 2007, Britain operated a decentralized visa network. Most diplomatic and consular posts handled local applications independently. This arrangement was inefficient. Maintaining over 150 visa sections presented challenges in terms of staffing, technology, equipment, training and management. Visa demand in many countries is variable, and it was also difficult and expensive for posts to adjust staffing in response to fluctuating demand (National Audit Office, 2004, p. 17).

To reduce this inefficiency, Britain developed the “hub and spoke” model for visas. Visa application centers or “spokes,” mostly operated by commercial partners, now conduct the front-end of the visa process. They gather documents from applicants, collect biometric data, and perform customer service functions, regardless of where adjudication takes place (Bolt, 2019, p. 55). The “spokes” then forward compiled materials to a “Decision Making Centre” (DMC), which issues or denies the visa (Bolt, 2020, p. 13). Importantly for interpreting the analysis here, this front-end of the visa process does not vary depending on where the decision is ultimately made. That is, my empirical analysis looks only at changes in back-end decisionmaking.

With front-end functions handled at the “spokes,” adjudication could happen anywhere. In the second half of 2007, Britain began shuttering local DMCs and transferring decisions to centralized hubs (Bolt, 2020, p. 13). Initially, these hubs were located at overseas posts, and consolidation was largely regional (i.e., a single diplomatic mission would handle applications for a number of nearby countries). Subsequently, the Home Office began onshoring visa work to domestic facilities. By the end of 2019, this program had substantially consolidated visas, leaving only ten DMCs overseas (Bolt, 2020, p. 6). The initial hubbing transferred processing of all visa types from one DMC to another, but the later onshoring was segregated by visa type with long-term visas moving to onshore facilities separately from visitor visas (Bolt, 2020, p. 14).

DMC closures proceeded on a staggered basis, driven by administrative factors. Sir Peter Ricketts, the top civil servant in the Foreign and Commonwealth Office, described the closures to Parliament as “really an effort… to drive down the cost of the operation” (Public Accounts Committee, 2009). Of relevance to the research design, closures were not motivated by an effort to reduce issuance rates or otherwise change immigration policy, rather they proceeded on a haphazard basis designed to cut costs and streamline staffing (Bolt, 2020, p. 6). While hubs were usually placed in larger countries, these posts do not appear to have otherwise operated distinctively beforehand. For example, considering the 25 non-EU countries that would go on to become hubs as of 2011, the average approval rate before the closures began (79.7 percent) was essentially equal to the overall approval rate for visas (80.8 percent).

In short, closing a DMC or onshoring casework represents a change in the agents charged with implementing visa policy without a change in the governing rules. Given the nature of the rollout, it is possible to separately analyze three types of agency: entry clearance officers at a local post, at a third country hub, or at a domestic hub. Given the number of shifts, it is possible to test these separately.

For the sake of clarity, Table 1 provides a concrete illustration that shows how visitor visa applications submitted in Canada were adjudicated over time, including all three possible modalities. Initially, these visas were adjudicated at the British High Commission (i.e., embassy) in Ottawa by Ottawa-based entry clearance officers under the supervision of the British High Commissioner (i.e., ambassador) to Canada. After the closure of the Ottawa DMC, Canadian visa cases moved to a hub at the British Consulate in New York, under the supervision of the Consul General in New York. Finally, processing shifted to a domestic hub in Croydon under the supervision of the Home Office.

Table 1. Adjudication of Canadian visitor visas over time

Qualitative evidence shows how the location of adjudication can matter. A clear example comes from a detailed investigation of visa processing in Abu Dhabi conducted by the Independent Chief Inspector of Borders and Immigration in response to the dramatic decline in visa issuance rates for Pakistani applicants after casework shifted to Abu Dhabi described above. Senior diplomatic staff in Abu Dhabi described the drop as a positive development. When interviewed by inspectors, they argued that the reduction in Pakistani visa issuance represented “improved quality and consistency of decision making.” In contrast, the British High Commissioner in Pakistan characterized the same decline as a negative development, reflecting a low quality of work after casework shifted (Vine, 2010, pp. 13–17). This divergence in attitudes is precisely the one I predict theoretically above.

Inspectors ultimately concluded that the evidence mostly supported the latter view, and that the Abu Dhabi DMC favored Emirati applicants relative to Pakistani ones. At the time, both sets of applicants were subject to the same formal criteria, but the Abu Dhabi DMC applied these differentially to the two populations. Inspectors wrote:

Inconsistent approaches were taken in regard to the weight attached to evidence, depending on the nationality of the customer. We found that customers from [the UAE] who provided limited evidence to support their applications were granted entry clearance, whereas customers from Pakistan were being refused in similar circumstances (Vine, 2010, p. 26).

The report walks through a variety of illustrations. For example, an entry clearance officer granted a two year visa to an Emirati applicant who supplied no documentation other than a passport and a photograph. This officer applied discretion to ignore missing documents and concluded that the applicant “satisfied [requirements] on maintenance, accommodation, and intention,” despite the absence of any supporting documentation whatsoever. Inspectors found that this was a routine practice. In a random sample of successful applicants, inspectors found that 34 percent of visas had been issued despite insufficient supporting documentation. In interviews, entry clearance officers informed inspectors that they were applying a de facto visa waiver to Emirati applicants, based on a strong presumption in their favor (Vine, 2010, p. 25).

Applicants from Pakistan were not granted any similar benefit of the doubt in Abu Dhabi. Officers refused visas to Pakistani applicants for failing to provide the same kind of documentation that was routinely waived for their Emirati counterparts, and some applicants were subjected to demands that exceeded the guidelines (Vine, 2010, p. 26). Inspectors also found that entry clearance officers had often “disregarded positive evidence” with respect to Pakistani but not Emirati applicants—for example, refusing a Pakistani applicant for failing to demonstrate sufficient funds for a short-term visit to the UK despite a bank statement demonstrating a balance of over £34000 (Vine, 2010, pp. 22–23). The findings of discrimination against Pakistani applicants (i.e., demands in excess of the rules) generated substantial high-level interest and pressure for change. A subsequent re-inspection concluded that this practice had ended; however, inspectors found that the Abu Dhabi DMC continued to discriminate in favor of Emirati applicants—that is, granting them visas despite insufficient evidence under the rules (Vine, 2012).

3.2 Testable hypotheses

Given this substantive backdrop, it is possible to derive three hypotheses, corresponding to the three theoretical traditions discussed above.

H0 (no effects): The location of visa processing should have no impact on visa issuance.

While this is the null hypothesis, it also corresponds to the theoretical predictions of the unitary actor approach to diplomacy. If diplomatic missions have no independent effects, then the form and location of visa adjudication should not matter. Neither DMC closures nor onshoring should lead to changes in issuance.

H1 (relationship-specific effects): The closure of a local DMC reduces visa issuance for nationals of that country.

This hypothesis tests the theoretical expectation derived above—that diplomatic missions promote cooperation with their hosts in particular. If such a special relationship exists, then the transfer of visa work away from the relevant embassy or consulate to any other location should reduce issuance. Once local officials cease making visa decisions, issuance should drop regardless of who takes over.

H2 (generally cooperative diplomats): Onshoring visa work for a given country to a domestic facility reduces visa issuance for nationals of that country above and beyond any effects of closing the local DMC.

H2, on the other hand, tests the alternative theory sketched above—that diplomats have an accommodating or cooperative orientation in general. If diplomats behave this way overall, then any overseas post operating an entry clearance operation ought to take a relatively lenient approach to issuance. In contrast, home-based entry clearance officers operating outside diplomatic supervision should exhibit no such tendency. Thus, onshoring visa work should reduce issuance relative to any form of diplomatic adjudication, and there should be a significant difference between visa processing at third country hubs and visa processing at onshore hubs. That is, if there is a non-zero relationship-specific effect found for H1, then onshoring (as opposed to third country adjudication) should lead to an additional reduction in issuance above and beyond this relationship-specific effect.

In the empirical specifications below, I operationalize these hypotheses using two separate dummy variables: “DMC Closed?” and “Onshored?” For clarity, I show in Table 2 how these variables define the possible states of visa processing, and then explain how they test the hypotheses.

Table 2. Dummy variables and states

.

The fourth row of the table represents an impossible combination. Visas can be onshored only when they are not processed locally. Thus, in the regressions below, the two dummy variables define three states of the world with an excluded category of local adjudication (where both dummies are zero). In the fully specified regression, “DMC Closed” measures the difference between local adjudication and adjudication at a third country hub. H1 predicts a negative value for this coefficient. The variable “Onshored” measures the effect of onshoring relative to adjudication at a third country hub. H2 predicts a negative value for this coefficient. Note that the difference between local adjudication and onshore adjudication is measured by the sum of both. Finding a coefficient of zero on the “Onshored” variable would mean that onshore hubs are no different than third country hubs, but it would not imply that onshore hubs operate equivalently to local decision-making. If both “Onshored” and “DMC Closed” have negative coefficients, then their relative magnitudes would tell us about the relative importance of relationship-specific versus general cooperation effects in delegated diplomacy.

4. Research design and data

As my primary empirical specification, I adopt a staggered difference-in-differences design implemented with two-way fixed effects. Causal identification requires a common trends assumption—that there are not omitted, time-varying factors that differentially impact the treatment group and the control group after conditioning on the controls (Athey and Imbens, 2022). The canonical difference-in-differences design uses only two groups and two periods, and estimates causal effects using an interaction for treated units in the post-treatment period but cannot accommodate staggered timing. The two-way fixed effects design adopted here instead allows for multiple periods with staggered timing and estimates causal effects using an indicator for treatment status. The estimate recovered is a weighted average that is equivalent to the average treatment effect if treatment effects are homogenous. If treatment effects are heterogenous, however, the estimate may be biased. While I rely primarily on the standard estimator here, I also estimate models in the appendix using the Sun and Abraham (2021) method to estimate an average treatment effect given heterogeneity. This provides substantively equivalent results.

I test a variety of different fixed effects beyond the baseline. I start with just country effects, then move to the baseline two-way model with quarter and country effects. The quarter fixed effects control for changes in the overall orientation of visa policy over time—that is, general loosening or tightening of immigration restrictions. The country fixed effects control for the fact that visa issuance varies from country-to-country based on, for example, a country's level of development, its population, or the risk represented by its nationals. Next, I include fixed effects at the region-quarter level based on the administratively-defined regions of the world used in implementing British visa policy. These control for any kind of time-varying but regionally-specific shock to visa issuance. Finally, I use country-year fixed effects—these control for all variation affecting a particular country in a particular year and isolate the identifying variation to only the pre-closure and post-closure quarters for a given country in a given year.

In combination, the fixed effects ought to capture most likely confounders, but what they cannot control is an idiosyncratic shock that impacts issuance for a particular country and is specifically associated with the timing of closure. That is, in the most conservative specification with country-year fixed effects, the relevant confound would be some unobserved factor that causes both a drop in visa issuance in a specific quarter and the closure of a DMC in that same quarter. The necessary identifying assumption is that country-quarter shocks are not related to the timing of closures. Fortunately, in this context, it is possible to directly observe all changes in the explicit rules that target particular countries. The rules themselves are public, and changes are clearly noted. I exclude all cases involving a country-specific shock to the visa regime so as to avoid bias from these changes.² But, there could still be unobserved, implicit changes in the processing of particular visas or changes in conditions in a country impacting issuance. Conceivably, these could be related both to DMC closures and to issuance. Suppose, for example, that hostile relations with a given country prompted senior officials both to close down the local DMC in a given quarter and to issue informal, nonpublic directives to apply visa criteria more strictly. Such a relationship between closure timing and implicit changes in visa requirements could threaten the design.

The history and nature of the closure program support the assumption that the specific timing of DMC closures and onshoring were unrelated to such factors. Instead, the record clearly shows that the program was primarily driven by administrative factors. A systematic examination by the Independent Chief Inspector concluded that officials saw “the cost saving argument as overwhelming” in closing DMCs (Bolt, 2020, p. 9). Inspectors further found that there was no strategic or political direction for the program; rather, “until 2017 there was no effective oversight of the network consolidation process beyond that provided as part of everyday line management.” Closure timing depended, first and foremost, on the availability of sufficient capacity in place at hubs—a factor exogenous to any aspect of conditions in the affected country itself (Bolt, 2020, pp. 17–18). Section 6 provides further empirical validation for this.

For my dependent variable, I take total visa issuance from Home Office statistical releases that record visa outcomes at quarterly frequency aggregated by nationality starting in 2005. I end all analyses in the fourth quarter of 2019 prior to the substantial disruptions to global travel associated with the COVID-19 pandemic as well as any Brexit-related effects. Raw issuance numbers are highly skewed; few visas are issued to the nationals of small countries while many are issued to the nationals of the most populous countries. Thus, I use the natural logarithm of the number of visas issued by country quarter plus one as my dependent variable. The resulting coefficients are interpretable as a proportional reduction in visa issuance, which is theoretically appropriate.

The data on the two independent variables—DMC closures and and onshoring—come from the Independent Chief Inspector's report on network consolidation (Bolt, 2020) and an earlier Parliamentary report on the same topic (House of Commons Home Affairs Committee, 2012). During the relevant time period, nationals of the European Economic Area and Switzerland enjoyed freedom of movement into the United Kingdom without a visa, so I exclude these countries. I also exclude visas issued to individuals whose listed nationality does not correspond to a sovereign state (e.g., stateless persons). One important caveat is necessary—the shift of work from closing a DMC is rapid but not instantaneous. Hubs might begin processing some proportion of casework before final closure (Bolt, 2020, p. 66). Unfortunately, data is not available on where a particular application was adjudicated, so the analysis here is limited to the overall effect of DMC closures rather than a case-by-case measure based on where an application was processed. The gap in time between closures and shifts in work is generally minimal, so this should not have a large impact.

Before moving to the regression models, Figure 1 presents a preliminary visualization of the data. I plot the number of years before or after a DMC closure along the x-axis (zero corresponds to the quarter of the closure), limiting the dataset to those countries with at least three years of available data both before and after closure. The y-axis shows the average logged visa issuance by post, with one point for each quarterly average. The solid vertical line at zero indicates the quarter when the DMC closes, while the dotted horizontal lines show the average logged issuance in the pre-closure and post-closure periods.

Figure 1. Visa issuance before/after DMC closure.

In the three years preceding closure, logged issuance averages 6.92 (equivalent to 1,012 visas per country-quarter). In the three years after closure, this drops substantially to 6.76 (equivalent to 863 visas per country-quarter). As I will show below, this simple estimate without controls closely tracks the magnitude estimated in the regression specifications.

5. Results and discussion

Moving to the analysis itself, I perform a series of OLS regressions and adjust for the error structure of the data by computing standard errors with two-way clustering by country and quarter in each model—adjusting for serial correlation by country as well as any sort of cross-country correlation within periods (Cameron et al., 2011). In Table 3, I test hypothesis one (which predicts a negative coefficient on DMC closure) using data on overall visa issuance. I gradually test different fixed effects specifications across models.

Table 3. Regression of logged issuance by country-quarter

*p < 0.05, **p < 0.01, ***p < 0.001.

Note: Standard errors in parentheses with two-way clustering for quarter and nationality.

The results strongly support hypothesis one. Across all five models, there is a large and statistically significant reduction in visa issuance after a DMC closure. Given the log scale, the coefficient point estimates correspond to a 12 percent to 15 percent drop in issuance after the closure of a DMC depending on the specification. Aggregated across all decisions, this is a very substantial decrease—implying the issuance of about 150000 fewer British visas in 2019 alone as the result of the DMC closure program.³ In Table A1 in the appendix, I replicate the same specifications using the Sun-Abraham correction for possibly heterogenous treatment effects. The results are largely the same. Figure A1 in the appendix shows an event study plot constructed from those results, while Table A2 presents another robustness check using the issuance rate (i.e., visas issued as a percentage of decisions made) rather than logged total issuance, and Table A4 conducts the same analysis using the panel matching method developed by Imai et al. (2023).

As noted above, it is necessary to separate visas by type in order to study the impact of onshoring, which occurred separately for three broad categories: visitor and other temporary visas; student visas; and work, family, and settlement visas. In Table 4, I test both hypothesis one and hypothesis two using visa issuance split into these categories. The two independent variables of interest are DMC closures (where hypothesis one predicts a negative coefficient) and onshoring (where hypothesis two predicts a negative coefficient). In the interest of space, I present only the specifications using the basic two-way fixed effects (country and quarter) and the specification using the country and region-quarter effects in the table. The full set of alternative specifications, and a replication using the Sun-Abraham correction for possibly heterogenous treatment effects are shown in the appendix in Tables A5 and A6. Note that visitor requirements are waived for nationals of a number of countries, and I exclude these from the models where visitor visas are the dependent variable.

Table 4. Regression of logged issuance by country-quarter split by visa type

*p < 0.05, **p < 0.01, ***p < 0.001

Note: Standard errors in parentheses with two-way clustering for quarter and nationality.

Once again, the results strongly support hypothesis one. Across visa types and specifications, DMC closures lead to substantively large and statistically significant reductions in visa issuance. As shown in the appendix, this finding is very robust to alternative specifications for visitor visas though somewhat less robust for long-term visas. The results provide limited support for hypothesis two at most. In each specification, I estimate that onshoring does reduce visa issuance. But, with the exception of a single specification (see Model 2 of Table 8 in the appendix), this reduction is never statistically significant and the magnitude of the effect is generally much smaller than the effect of DMC closures. Variation across visa types is also notable. I find the largest and most robust impact of closures in the case of visitor visas. As noted above, DMCs have the greatest discretion when it comes to visitor visas, as these decisions generally cannot be appealed to a higher authority, and this may explain the difference.

In short, these findings reveal that closing a DMC (i.e., shifting away from local issuance) substantially and significantly reduces visa issuance, consistent with the theory that local diplomats pursue especially cooperative or accommodating polices. On the second major question—whether domestic hubs run by the Home Office operate differently than third-country hubs run by diplomatic posts—the results are ambiguous. There is suggestive evidence that onshoring to domestic hubs may reduce issuance above and beyond the impact of a local closure (i.e., onshore hubs may issue fewer visas than third-country hubs), but given the lack of statistical significance, it would be unwise to overinterpret this finding. With this in mind, I now turn to a further robustness check to validate the conclusion that DMC closures reduce issuance.

6. Placebo test: visa extensions

As noted above, the major potential threat to causal inference is the possibility that some kind of country and period specific shocks both cause DMC closures and lead to a tightening of visa issuance. While I have argued that the structure of the closure program makes such a relationship unlikely, a a second way to approach the issue is a placebo analysis using visa extensions. Requests to extend existing visas, unlike initial visas, have always been adjudicated in the UK and not at overseas posts. Thus, a DMC closure (or onshoring) has no impact on the way extensions are handled. On the other hand, extensions ought to be influenced by the same implicit factors that influence initial applications. If, for example, some non-public directive targeted nationals of a given country for enhanced scrutiny, this ought to affect both initial and extension applicants.

This provides an excellent opportunity for a placebo test. There is no plausible way that onshoring or DMC closures could change the number of granted extensions. If, therefore, such an effect is identified by the model, then it suggests that there is a problem in the inferential design. Once again, I use Home Office statistical releases to test for a change in visa extension grants. The available timeframe is, however, shorter as the data are only available from 2010 to 2019, covering about three million extension requests. In Table 5, I replicate the same models as shown in Table 3 using logged extensions plus one, rather than initial grants, as the dependent variable. I expect to find no effect.

Table 5. Regression of logged extension grants by country-quarter

*p < 0.05, **p < 0.01, ***p < 0.001

Note: Standard errors in parentheses with two-way clustering for quarter and nationality.

Across all five specifications, there is never a significant impact of DMC closures on extension grants. The relevant coefficient is positive in two models and negative in three models, consistent with the expected pattern if there is no underlying effect. In sum, the placebo results provide strong support for the validity of the design, indicating that the results are not driven by some kind of implicit changes in the stance of British visa policy toward the countries in question.

Another way to use the same data is as a control variable—that is, to incorporate granted extensions into the regression models a way to control for any kind of changes in the implicit stance of visa policy not captured by the fixed effects. Table A7 in the appendix shows this model in the various fixed effects specifications. I find a positive and significant relationship between extensions and initial issuance—that is, extensions do successfully proxy for some kind of changes in visa processing. After controlling for this, I continue to find a negative and significant effect of DMC closures (of comparable magnitude to the estimates above). This provides especially strong and robust support for the finding that DMC closures reduce issuance. That is, in combination with the fixed effects, this rules out nearly any plausible confound. It would be necessary to hypothesize some kind of confounding variable that is specific to the quarter of closure for a country and not the year (or else it would be controlled by the country-year effects), affects only the country in question and not its region (or else it would be controlled by the region-quarter effects), and only affects initial issuance but not extensions (or else it would be controlled by the extensions variable).

7. Discussion and conclusions

The empirical results here show that transferring responsibility for visa issuance away from local diplomatic posts substantially reduces issuance. This, in turn, affects many other forms of international cooperation that depend on travel or migration. While this is the first quantitative assessment of the program, stakeholders have qualitatively recognized the impact of these changes for migration and cooperation, and have spoken out against the network consolidation program based on its inhibitory effect on cooperation (All-Party Parliamentary Group, 2019).

From a policy perspective, national leaders should be aware that transferring visa responsibility away from local diplomatic posts is a de facto tightening of visa availability. The results here do not speak directly to the desirability of such a tightening from the perspective of the national interest. Some countries or leaders may see a tightening of visa criteria as desirable, while others will not. Turning to the theoretical side, the results provide strong evidence that diplomats promote cooperation with their host countries. There is limited evidence here that diplomats take a more cooperative or accommodating stance in general—the behavior of third country diplomatic hubs appears to be similar to that of onshore hubs. This is an important distinction, highlighting the significance of bilateral diplomatic ties in particular. Diplomats are not simply interchangeable agents of cooperation; their impact is relationship-specific. Diplomacy, as practiced by local missions, will be distinctive and uniquely cooperative. When diplomats administer other programs—such as supervising the distribution of foreign aid—they are likely to do so in a way that more closely reflects host-state interests.

It is not directly possible in this context to distinguish among the multiple channels potentially leading to the result—that is, to cleanly distinguish socialization, selection, incentives, and skills or knowledge. There are, however, clear reasons to be skeptical of the skill or local knowledge mechanisms. Most notably, the Independent Chief Inspector explicitly investigated the issue and found “little evidence to support the [lack of] local knowledge argument” (Bolt, 2020, p. 8). Immigration intelligence staff continue to engage locally, and supply local information, after DMC closures. Locally-engaged verification teams are also used to perform document verification or other specific tasks requiring detailed local knowledge (Bolt, 2017, pp. 13–14). The qualitative evidence from the Pakistani case described above also suggests a clear preference mechanism; that is, the divergent attitudes of the staff reflect a preference gap rather than a knowledge gap.