Rumors and Refugees: How Government-Created Information Vacuums Undermine Effective Crisis Management

Although more than 800,000 displaced people arrived in Greece by sea in 2015, fewer than 5 percent applied for asylum in this first country of arrival. Instead, they either traveled northward informally or remained in Greece in legal limbo. The resultant chaotic conditions deprived many refugees of the benefits of asylum and formal relocation procedures, and also reduced the Greek government’s popularity among natives. We argue that governments, regional and international organizations, and aid groups can undermine compliance with their own policies by mishandling information dissemination. Common crisismanagement tools—such as frequent policy changes, information dissemination limits, and ad-hoc policy implementation— can easily backfire. Information mismanagement can lead people to develop deep distrust in government and aid organizations, and instead turn to informal brokers like smugglers. To assess our theory, we draw on over 80 discussions with migrants and refugees in Greece, on 25 semistructured interviews with aid workers and government officials, and on weekly rumor correction newsletters produced by the nongovernmental organization Internews. We conclude that governments must prioritize effective communication and policy transparency, especially in crisis contexts.


Introduction
On May 24, 2016, hundreds of Greek riot police reached Idomeni to dismantle the crowded, informal refugee camp (Labropoulou and Hunt 2016). Rumors regarding the imminent opening of the Greek-Macedonian border, fueled by smugglers and believed by migrants who wished them to be true, encouraged thousands of camp residents to stay for months despite harsh conditions. Although Greek, Macedonian, and European Union (EU) leaders announced that they closed the border indefinitely, migrants deemed this information not credible. Riot police had to physically escort migrants to buses to disperse the crowds. 1 The incident at Idomeni was not isolated. The Greek government faced a broad-ranging governance crisis in the first years of the Syrian refugee crisis. While more than 800,000 displaced people arrived to Greece by sea in 2015, fewer than 5 percent applied for asylum in this country (Eurostat 2016;UNHCR 2016). According to the EU's Dublin regulations, asylum seekers must first file paperwork in Greece, even if they plan to reunite with family elsewhere in Europe. Brief policy "openings," family ties, economic conditions, and language barriers influenced refugees' decisions to move onward from Greece (Riddle and Buckley 1998, 238;Moore and Shellman 2007, 814; Robinson and Segrott 2002, 4-5). However, these factors cannot fully explain why so many refugees opted to remain in Greece informally or why refugees used smugglers rather than legal routes for onward travel.
Refugees' aversion to legal pathways has been particularly puzzling given that opportunities for formal relocation exist, especially for Syrians. The EU emergency relocation scheme aims to relocate over 160,000 people from Greece and Italy to other EU states (Council Decision 2015/1523Council Decision 2015/1601. 2 Syrians' applications for protection have had a 99 percent success rate in Greece; moreover, as of December 2016, there were more places for relocation available than relocation requests filed (Greek Asylum Service 2016, 6;Konstantinou, Georgopoulou, and Drakopoulo 2016, 8). If Syrians knew and believed these facts, we should not have observed so many turn to informal routes.
Similarly, limited state capacity cannot explain Greece's crisis mismanagement. Although the recent debt crisis weakened Greek institutions, the government had moderate bureaucratic capacity and extensive EU resources at its disposal. Moreover, Greek politicians did not strategically benefit from sustaining chaotic conditions and restricting refugees' access to services. Low asylum application rates made it difficult for the Greek government to distinguish between refugees (who they must protect) and undocumented migrants (who they can legally deport). 3 This in turn slowed the asylum process, stalled the deportation of undocumented migrants, and trapped tens of thousands in legal limbo.
We argue that rumors and misinformation play a critical role in explaining this governance crisis. By pursuing policies that created information vacuums, the Greek government and aid organizations inadvertently allowed rumors to flourish. Although rumors and misinformation have wider reach today than at any other point in history, their effect on policy compliance and governments' enforcement capacity receives limited scholarly attention (Berinsky 2017, 3;DiFonzo and Bordia 2007, 205-27). Prior scholarship on information management focuses on authoritarian states and on active government efforts to mislead enemies during war. We significantly broaden this literature by identifying a range of government and international organization policies that create information vacuums. Whereas government propaganda aims to spread misinformation, the policies we study can unintentionally create a space for rumors. Informal brokers, such as smugglers, easily exploit this space, encouraging policy noncompliance and undermining crisis management strategies.
Rumors are commonly defined as "claims of fact about people, groups, events, and institutions that have not been shown to be true, but that move from one person to another and hence have credibility not because direct evidence is known to support them, but because other people seem to believe them" (Sunstein 2009, 6). We follow the literature in defining rumors as propositions that are neither verified nor verifiable, but in wide circulation. While all types of rumors-true and false-can contribute to distrust and instability, we focus on false rumors to demonstrate how they directly lead displaced persons to suboptimal decisionmaking.
In crisis situations, false rumors proliferate and create destructive feedback cycles that encourage noncompliance with a broad range of government policies. Drawing from the regime transition literature, we argue that these feedback cycles can rapidly transform refugees' perceptions of state legitimacy by creating "tipping points" that dramatically shift public opinion (Lamberson and Page 2012, 3-5;Watts and Dodds 2007, 442-47). Indeed, scholars link the catalyzing effect of rumors to regime collapse (Coyne and Leeson 2009, 3;Kuran 1997, 256;Kuran 1998, 628) as well as to conflict onset (Greenhill and Oppenheim 2017, 673;Horowitz and Varshney 2003, 2-5).
Ultimately, successful crisis management-and policy implementation broadly-depends on how governments and relevant stakeholders disseminate information. Our theory of information mismanagement does not directly compete with existing explanations that emphasize strategic enforcement, bureaucratic capacity, and push-pull factors. Instead, we argue that information mismanagement is a critical mediating factor that explains how exactly bureaucratic limitations or refugees' preferences produce unanticipated and unfavorable outcomes. 3 International law distinguishes between refugees, who flee persecution and are entitled to protection, and undocumented migrants, who lack these rights. Throughout the article, we use the term refugee loosely to include a variety of displaced persons who might not fit the UNHCR definition, but might qualify for other forms of subsidiary protection under EU or national law, following many scholars who focus on de facto protection needs. For example, while Syrians' protection claims are almost always granted across the EU, they are increasingly given only subsidiary protection. Unlike asylum claims, subsidiary protection claims do not require the applicant to prove individualized persecution. However, subsidiary protection also has fewer benefits in many cases, notably a shorter residency period and more limited family reunification rights.
To develop our theory, we draw on 25 semistructured interviews with government employees and aid workers, more than 80 interviews with refugees living in Greek camps and detention centers, and on the nongovernmental organization (NGO) Internews weekly rumor correction newsletters, which are in turn based on thousands of interviews. The Greek case provides multiple leverage points. First, Greece serves as Europe's main entry point for migrants and refugees. Although initially considered a transit country, closed land borders indicate that Greece will host tens of thousands of asylum seekers for the long term, making questions about rumors and policy compliance critical. Moreover, a variety of successive regional policy changes, including the EU-Turkey deal and border closures, provide shifts in existing bureaucratic obstacles, push-pull factors, and other key contextual factors. These variations create opportunities to identify, isolate, and understand the role that policy transparency and rumors play in compliance.
The article proceeds as follows: first, we outline existing explanations, highlighting how strategic logic, capacity limitations, and the existing rumor literatures fail to explain the governance crisis in Greece. Next, we develop our theoretical argument. We then discuss the methods that we use to assess our claims. We describe our findings and continue with a counterfactual section, exploring what would change in a world with more complete information. We conclude with policy implications and consider risks that come with transparency as well.

Existing explanations for policy implementation failures
The literature traditionally attributes policy implementation failures to selective law enforcement and limited state capacity; push-pull factors, it contends, prove particularly influential in the migration context. Scholars who examine the role of governments in rumor propagation focus primarily on active propaganda dissemination. Unlike propaganda efforts, which are often led by authoritarian governments and amplified in times of war, the policies we study are implemented by a broad range of governments in a broad range of circumstances. Below, we briefly explain how our theory builds on and contributes to each of these literatures.

Strategic Law Enforcement
Strategic law enforcement theories focus on officials' rentseeking and vote-seeking behavior. Politicians may frequently reward friends by providing legal exemptions and punish enemies via targeted penalties (Gans-Morse 2012, 264-65;Holland 2016, 235;O'Donnell 1998, 117). Strategic enforcement can occur because of significant resource constraints: governments need to assemble police, judges, and bureaucrats in order to enforce the law (Holland 2015, 359). Often, politicians enforce policies strategically to benefit the rich and well-connected, or to favor co-ethnics of the ruling party. Sometimes, strategic enforcement even allows politicians to benefit less privileged constituencies, such as unlicensed street vendors and undocumented immigrants (Holland 2015, 357-58).
Scholars have used strategic enforcement theories to explain aspects of Greek public policy, notably widespread tax evasion and illegal construction in exchange for money or votes (Skouras and Christodoulakis 2015, 537-38). In prior migration waves, Greek governments prioritized co-ethnics due to their potential voting capacity. However, in this crisis, politicians could not gain any votes from Syrians or Afghans, since they lack Greek citizenship. While Greek officials could have, in theory, intentionally created chaotic conditions to deter migration waves or push refugees onward, this would have been costly to the Greek population. Moreover, March 2016 land and border closures substantially curtailed outflows, rendering such a strategy inoperable. Indeed, pollsters have attributed the governing Syriza's party's significant decline in popularity to its mishandling of the refugee crisis (Kathimerini 2016).

Bureaucratic Capacity
Low bureaucratic capacity may also explain poor policy enforcement. States with weak institutions and limited resources lack the capacity to control street-level bureaucrats, who often apply policy arbitrarily (Saltsman 2014, 469;Hansson, Ghazinour, and Wimelius 2015, 103;Post, Agnihotri, and Hyun 2018, 3). For example, when weak states lack the resources to hire translators, they inadvertently delay refugees' asylum applications, lengthening their stay while limiting their access to critical services (Bohmer and Shuman 2007, 68;Burchett and Matheson 2010, 86). Bureaucratic obstacles, then, may push refugees to move onward prior to securing legal protection in hope of acquiring it more quickly elsewhere.
Greek officials face numerous bureaucratic obstacles to effective refugee management in the country. Currently, the country faces an economic crisis, and the government has limited administrative experience. Syriza came to power as a protest party in 2015, having won less than 5 percent of the vote prior to the austerity crisis. Indeed, when economic recessions discredit the established political order, new and inexperienced governments must often manage major crises. That said, when a European country with substantial resources at its disposal mishandles a refugee influx and refuses additional guidance and administrative support from the United Nations High Commissioner for Refugees (UNHCR), 4 the bureaucratic capacity explanation raises more questions than it answers. If capacity issues explained refugee noncompliance, we should not see institutionally weaker and relatively resource-poor governments managing larger refugee crises. Poor Middle Eastern countries like Jordan and Lebanon currently host roughly 656,400 and 1,000,000 Syrian refugees, respectively-over ten times the number in Greece (UNHCR 2016). 5 While small, contained outbreaks of violence have occurred, both countries have avoided the instability reported in Greece. Notably, shared language and culture bolster the stability of Jordan and Lebanon, whereas Greece has faced serious complications in communication.
In the following sections, we develop a theory to explain how bureaucratic limitations contribute to crisis mismanagement via the politics of information. We suggest that the relationship between bureaucratic capacity and crisis mismanagement is not monotonic: low-capacity states do not necessarily mishandle crises more than high-capacity states. Since moderate-capacity countries tend to place greater faith in their own bureaucratic procedures, low-capacity 4 Interviews 8, 9, 12, 13, 14, 18, 19, and 20 states may rely more on international bodies running their crisis operations. In moderate-capacity Greece, this form of self-reliance increased information mismanagement at the height of the refugee crisis.

Push-pull Factors
Extant literature demonstrates that push-pull factors influence refugees' decisions to move onward and settle in another country (Kunz 1988, 23). Factors that can "push" refugees to leave a particular host country range from insufficient legal protections, to a lack of basic rights, to restricted or nonexistent public services (Betts 2009, 55). Targeted violence, arbitrary arrest, and police harassment can also lead refugees to travel to safer host countries (Brewer and Yükseker 2006, 13). Conversely, factors that "pull" or entice refugees to travel beyond a particular host country include the presence of family members, friends, or co-ethnics in another country, anticipated cultural acceptance, and linguistic knowledge (Riddle and Buckley 1998, 238;Collyer 2005, 713;Saltsman 2014, 462). Similarly, economic conditions and educational prospects, cost of living, and employment opportunities can lead refugees to move onward (Lindley and Van Hear 2007;Peters 2015, 7-9). Scholars have also identified the geographic position of the destination country as well as policy "openings" that facilitate travel, including porous borders, lax entry controls, and liberal visa regulations, as important pull factors (Boyd 1989, 646;Brewer and Yükseker 2006, 10). Refugees' preference for Germany or Sweden as their final destination is unsurprising. However, push-pull factors cannot explain how refugees decide to move onward, namely their decision to forego the formal application process (the legal route to northern Europe), and entrust their lives to smugglers. In particular, these factors cannot explain why Syrians, who enjoy streamlined asylum procedures and have opportunities for relocation, opt for informal pathways. Push-pull factors also cannot explain why refugees maintain an informal status in Greece, foregoing access to free services from the Greek government.
Similarly, policy openings may provide temporary legal incentives for onward movement. However, they cannot explain why so many asylum seekers have opted for informal pathways rather than legal routes. Figure 1 below presents a timeline of European countries' migration policy changes from 2015 to 2017. At the end of August 2015, German Chancellor Merkel announced that Germany would not enforce the Dublin regulation for Syrians already on German soil. Although this prompted some refugees to eschew legal pathways to try to reach Germany, the German government quickly closed this policy opening. Three months later, Germany announced that it would enforce the Dublin regulation for new entrants, and start returning refugees to their country of first arrival. Moreover, only Germany and the Czech Republic suspended the Dublin regulations for Syrians at any point. Other desired destination countries, including Sweden, France, and the United Kingdom, did not. Meanwhile, other countries pursued restrictive policies, such as border closures along the Balkan route 6 -drastically increasing the difficulty (and danger) of informal travel (Dernbach 2015). If asylum seekers interpreted Germany's brief opening as a permanent guarantee, it reinforces our argument that rumors can undermine policy implementation by encouraging noncompliance. 6 Macedonia and Hungary, among other Southern and Central European countries, closed their borders and detained asylum seekers in response to the 2015 migrant influx. Push-pull factors and bureaucratic obstacles influence refugees' choice of final destination. However, depending on how they develop and communicate policies, governments encourage refugees to engage in behaviors that undermine crisis management efforts. Rumors and misinformation about policy exacerbate the effects of bureaucratic obstacles and push-pull factors, leading refugees to act in ways that undermine host states' asylum policies. We contribute to this literature by demonstrating that information mismanagement can lead refugees to act in counterproductive ways.

Active Rumor Dissemination
Prior work on information management focuses on how governments actively propagate rumors in wartime and in authoritarian contexts. Extant literature dates at least as far back as WWII, when Allied and Axis powers established propaganda committees and incorporated rumors into formal military strategy to raise civilian morale and strengthen national security (Nasi and Sweatland 2015, 31). During the Cold War and post-Cold War era, this cycle of securitybased propaganda flourished. For example, the FBI's counterintelligence program solicited journalists to produce fake news discrediting communists (Jeffreys-Jones 2007, 149-74). Currently, the Russian government sows rumors against domestic opposition and the United States through the St. Petersburg Internet Research Agency and other related firms. Using more subtle types of rumor propagation, the Chinese Communist Party intentionally creates uncertainty in the country by enforcing incoherent and inconsistent censorship policy using a decentralized bureaucracy (Stern andHassid 2012, 1236). Countries that face domestic terrorist threats also frequently propagate rumors (Bernardi et al. 2012, 10). While the Internet age has amplified the spread of false information, governments have propagated rumors as a long-standing security practice (Berinsky 2017, 3).
Our theory differs from existing work in three distinct ways. First, intent sharply differentiates government propaganda from government-created information vacuums; while the policies we study unintentionally create misinformation and chaos as second-order effects, propaganda policies actively aim to generate disorder and confusion. Second, we focus on different methods: while government propaganda utilizes false messaging, we emphasize that frequent policy changes, information restrictions, and arbitrary implementation can inadvertently create information vacu-ums. Third, we broaden the scope of inquiry from authoritarian governments and wartime conditions to democratic governments in a broad range of circumstances. We contend that rumors can proliferate even without tools in place to intentionally promote them. Governments' incompetence, inexperience, or inattention to information dissemination can shape refugees' perceptions and behaviors as substantially as malicious misinformation and censorship.

Theory
We argue that governments do not need to actively deceive to produce governance crises or impede policy implementation. Even when governments pursue policies aimed at stabilizing crises, these choices can backfire, producing an information vacuum and an environment rife with mistrust. In the case of refugee influxes, when governments fail to provide accurate, consistent, and timely information, they exacerbate refugee communities' reliance on rumors and create negative feedback cycles that substantially weaken compliance.

How Governments Fuel Rumors, Often Unintentionally
Xenophobic governments may intentionally spread misinformation about migrants and refugees to win votes and deter refugee flows (Ivarsflaten 2005, 21). For example, Hungary's ruling right-wing Fidesz party set up a referendum on the EU's resettlement scheme, using xenophobic material that implicated migrants in recent terrorist attacks and in sexual harassment (Thorpe 2016). Both far-right and mainstream European leaders often call refugees "migrants" and suggest that they migrate to steal locals' jobs and abuse welfare benefits rather than seek protection, blurring the distinctions between these categories. Just as governments at war develop the "friend/foe" binary, xenophobic governments sharpen in-group and out-group distinctions in their political narratives. 7 We argue that well-intentioned governments fuel rumors through three common practices: frequent policy shifts, restricted information dissemination, and inconsistent policy implementation. In these cases, government policies have unanticipated second-order effects that exacerbate the spread of misinformation. We observe this phenomenon in Greece, where the ruling far-left Syriza party has employed refugee-friendly language, even though it places them at odds with more xenophobic public opinion (Linos 2003, 320-22). The Syriza Party's current position manifesto calls on progressive forces to resist racism and respect refugees' fundamental rights (Syriza 2016). The Greek electorate finds Syriza's refugee policy deeply unpopular-only 20 percent support the government in this area (Kathimerini 2016). While Syriza is progressive in ideology, it did not accommodate refugee hardships in practice, as subsequent sections clarify. 8 Crisis environments sharpen the need for effective communication, because misinformation spreads most easily in these high-anxiety environments. At the same time, crises place extensive demands on governments, so effective communication can easily become an afterthought. We argue that, due to this inattention, many common responses to crises fuel rumors.
First, governments responding to crises often change policies rapidly to manage on-the-ground developments. The humanitarian community normally lauds rapid crisis response as a hallmark of effective disaster management, especially when a government recognizes a policy's ineffectiveness and recalibrates (Rosenthal and Kouzmin 1997, 299-300). However, rapid policy shifts can impede communication flows and reduce individuals' trust in official information. Second, governments sometimes actively restrict information about their policies to control refugee movement and ensure national security. For example, the Greek government and UNHCR restrict information about how long asylum procedures take to deter refugees from leaving the country through smugglers. The Greek government's Asylum Center, the body charged with processing asylum claims, also actively limits such information, as does its EU counterpart-the European Asylum Support Office (EASO). While in theory these forms of information management expand government control, they also push asylum seekers to seek information from smugglers and rely on rumors to make decisions. Third, government officials, particularly street-level bureaucrats, may decide to enforce rules that deviate from official policies out of selfinterest or because the policy appears ineffective (Maynard-Moody and Musheno 2000, 329). Governments may have good reasons to pursue informal policies, particularly because they often have short-term advantages. However, as the crisis extends to the medium term, informal deviations from official policy increase perceptions of arbitrariness and discrimination.

From Uncertainty to Noncompliance and Ethnic Tensions
While rumors spread in many communities across diverse contexts, refugee crises offer particularly fertile ground for misinformation. Refugees, fleeing government persecution in their home countries, often start with low levels of trust in government institutions, and sometimes assume ulterior motives behind the international community's actions (Pearlman 2016, 25;Sperl 2002, 150;Carlson, Jakli, and Linos 2018, 13). Low trust may also extend to NGOs and IGOs due to their close cooperation with governments. For example, refugees in Piraeus and Scaramangas reported low trust in UNHCR because its workers would not provide information about the asylum process. Generally, refugees grouped government and UNHCR actions together. 9 Moreover, the high-anxiety context of migration, especially migration triggered by persecution, renders refugees vulnerable to information processing errors. Individuals that face anxiety and threats will likely engage in motivated reasoning and believe information that supports their desires rather than fact (Jost et al. 2003, 340-41). In experiments, highly anxious individuals repeat rumors more "eagerly" than less anxious ones (Oh, Kwon, and Rao 2010, 231). In this high-anxiety, low-information, and low-trust environment, displaced persons actively seek out additional information from unofficial sources and sources they trust to inform their decisions, including family, friends, Facebook groups, and, critically, smugglers (Allport and Postman 1947;De Feyter 2015, 150). The ethnic politics literature indicates that migrants place greater trust in co-ethnics (Chandra 2007, 36;Habyarimana et al. 2007, 709).
Migrants' trust in smugglers stems in part from smugglers' willingness to provide information when needed, their shared linguistic and ethnic backgrounds with clients, and shared migration journeys. Although smugglers may sometimes provide accurate information, they often provide false or biased information because they have the incentive to influence asylum seekers to use their services to leave the country. Using social media sites, text messages, and phone calls, 10 asylum seekers can more easily and consistently access smugglers than government officials and aid workers. Ease of consistent access leads many asylum seekers to perceive smugglers as a more reliable information source. Smugglers also customize their services to provide refugees flexibility in ways that governments cannot. For example, while refugees applying for relocation cannot choose their destination country, smugglers guarantee that they will travel with refugees until they reach their desired destination. 11 Poorly designed official communication policies create the space for informal brokers to offer alternative narratives. Migrants act on these narratives in case they are true. Through these feedback cycles, false rumors encourage noncompliance and may even spark widespread violence. Figure 2 visualizes the relationship between government policies, uncertainty, and noncompliance.
Asylum seekers' engagement with misinformation encourages a wide range of noncompliant behaviors. First, marginalized individuals use rumors to determine whether they should access certain government services (DeClerque et al. 1986, 83;Freedman 1991;Rosnow 1991, 484). Migrants also use rumors to anticipate host state actions, leading them to either avoid or circumvent the state (Ordóñez 2015). Additionally, rumors help establish and perpetuate informal economic practices and moral hierarchies among migrants (Harney 2013, 227).
Rumors also influence interactions between migrant groups of different ethnicities. When migrants possess little information about other ethnic groups, they are more likely to circulate negative stereotypes (Gardner 2012, 25-26). Moreover, we find that nontransparent and inconsistent policy implementation blurs the distinction between corruption and order and can lead all groups, even advantaged ones, to perceive disadvantage (Young 2016, 68). This is in line with decades of behavioral research; researchers have quickly generated in-group and out-group conflict in laboratory experiments by assigning subjects to arbitrarily defined 10 Interviews 9 and 20. 11 Interview 9; discussions with aid workers and refugees in Piraeus and Scaramangas.

Figure 2. Origins and effects of rumors on migrant behavior
teams (Rabbie and Horwitz 1969, 269;Turner, Brown, and Tajfel 1979, 187). In the same way, governments can spark novel, violent conflicts between ethnic groups with limited previous contact, including those we see between Syrians and Afghans.

Methodology
Identifying rumors and their effects on compliance poses significant methodological challenges. Rumors often concern sensitive information that government officials, aid organizations, and migrants themselves wish to mask. Additionally, asylum seekers frequently move between host communities seeking work or safe living conditions, making them hard to reach.
To overcome these challenges, we designed a qualitative research program that combines ethnographic and interview-based research with data collected by aid organizations that track rumors within Greek refugee communities. Below, we describe the advantages and limitations of our research methodology and briefly discuss the application to Greece.

Case Selection
We focus on Greece because, as the first transit point to Europe for hundreds of thousands asylum seekers, it plays a central role in the regional refugee crisis. While focusing on a single country limits generalizability, the significant variation over time, space, and refugee nationality within Greece allows us to better understand how rumors develop and spread (Linos and Carlson 2017, 220-31; Linos 2015, 477-80).

Semistructured Interviews and Participant Observation
During the summer of 2016, we conducted 25 semistructured interviews with government employees in health and asylum agencies and aid workers that ranged from field officers to heads of mission. Additionally, we observed and interviewed eighty refugees in their camp surroundings for six days in Piraeus, an informal camp; seven days in Scaramangas, a formal camp; and two days in Moria, a detention center on Lesvos. Although we used snowball sampling, we addressed this technique's limitations by having multiple "seeds," or starting points, from disconnected networks. Due to language constraints, we spoke with Arabicspeaking asylum seekers, including Syrians, Iraqis, and Kurds. We collected information on Afghan and Pakistani migrants through interactions with refugees and aid workers and through the Internews database, described below. Table 1 summarizes this work; Appendix A provides more detail.

Internews' Rumor Trackers
Due to the prevalence of misinformation during the refugee crisis, NGOs tried to facilitate information access. One prominent NGO, Internews, sends refugee liaison officers to formal camps, detention centers, and informal settlements to collect rumors and ask refugees about their experiences. In 2015, Internews established a website called News that Moves to aggregate the highest frequency rumors and produce weekly newsletters to debunk them. In the findings section, we draw primarily on rumors from the weekly newsletters, ensuring that we select common rumors. We also draw on Internews' full rumor database, which mentions refugees' ethnicities, to analyze inter-ethnic tensions. Their database contains 7,000 rumors collected across 16 formal and informal Greek camps from 2015 to 2017. This comprehensive dataset captures refugee perceptions of major events, particularly local, national, and regional policy changes. Even though Internews' rumor examples suffer from diverse types of selection bias, their data collection effort represents a significant addition to the rumor literature, which often relies on a scholar's rumor collection from a single site or within a more limited time frame.

Findings
Governments facing crises often design policies that address short-term problems, thinking they can quickly revise policies in response to changing conditions on the ground. Below, we focus on common actions that governments take: frequently changing policies, restricting information, and implementing informal deviations from stated policy. We provide examples of the Greek government implementing each type of policy, and explain government and aid organization logic in adopting these policies. We also document unfortunate policy side effects and show how these policies sparked rumors, decreased compliance, and fomented ethnic tensions, ultimately weakening the Greek government's crisis management capacity. Before outlining these policies, we briefly introduce official EU asylum procedures. The Common European Asylum System typically requires refugees to apply for asylum in their country of first arrival, regardless of whether they intend to relocate. The country of first arrival assesses asylum claims; those denied asylum can appeal, but otherwise face deportation.

Government Policies: Rapid Policy Shifts, Information Restriction, and Arbitrary Implementation
In this section, we detail three mechanisms the Greek government employed to manage the refugee crisis: fastchanging policies, information restrictions, and ad-hoc policy adjustments. We highlight that the Greek government did not intend to spread rumors among refugees. Rather, their efforts had second-order effects, unintentionally facilitating the spread of rumors and causing widespread instability and noncompliance. First, the Greek government changed asylum policies rapidly, attempting to remove bottlenecks that obstructed asylum seekers from initiating their applications. These policy changes had unanticipated consequences: they contributed to the spread of rumors by confusing asylum seekers and eroding their trust in the government. Between March and July 2016 the government implemented four major shifts in the asylum process. Prior to the Greek-Macedonian border closure in March 2016, the government focused on facilitating refugees' onward movement to Europe. Police stations on the Greek islands recorded asylum-seekers' basic information and provided them a registration paper, referred to as a kharti, allowing them limited residency and service access. 12 After the border closed, the kharti became a significant obstacle for asylum seekers, particularly Afghans, whose month-long legal residency passed quickly. Greek officials were more likely to detain individuals with expired khartia and exclude them from living in formal refugee camps. 13 The Asylum Center had no clear policy regarding whether asylum seekers with expired khartia could apply for asylum, 14 creating cases where Greek officials arrested and deported expired khartia holders who attempted to apply for asylum. 15 Recognizing the kharti's limitations, the Greek government established a preregistration process in which individuals called the Asylum Center via Skype to initiate their application. 16 This change, however, created new obstacles. First, calling required Internet access, which the Greek government provided inconsistently across formal camps. Second, since the Asylum Center operated the hotline only several hours per week, thousands called simultaneously, creating a massive bottleneck. 17 Lastly, Asylum Center officials lacked guidance for conducting the interviews, because no legal provision outlined Skype as a registration tool, 18 allowing officials to arbitrarily decide who qualified and prioritize high-publicity cases. 19 To remove these obstacles and gain a clearer picture of the volume and location of refugees in the country, in June 2016 the Greek government rolled out an in-person preregistration process, sending mobile registration units directly to formal camps. 20 This new policy, however, created additional obstacles. A mobile unit would give refugees the date of their preregistration appointment written on a wristband, confusing refugees between this date and the date of their asylum interview. 21 Language obstacles, such as Greek officials writing the appointment time in English, exacerbated confusion for non-English speakers. After preregistration, the Asylum Center provided refugees with SIM cards, planning to text them the date of their first asylum appointment. This created difficulties for refugees without phones and those who obtained new SIM cards monthly. 22 Following a mass gathering of asylum seekers during preregistration at Scaramangas camp in July 2016, the Asylum Center suspended in-person preregistration due to security concerns. 23 In total, this process preregistered 20,100 of the near 66,400 asylum seekers recorded in Greece before the government shut it down (Hellenic Republic Ministry of Interior and UNHCR 2016). The Asylum Center reopened the Skype hotline in August 2016. By this last policy change, refugees' confidence in the system had collapsed, creating an environment ripe for rumors. Figure 3 below outlines the cycle of policy changes. 12 The Greek government did not systematically store this information, motivating the pre-registration exercise. Discussions with aid workers in Piraeus and Scaramangas and interviews 3, 4, 7, 9, and 19. 13 Interviews 9, 16, 20, 23, and 24. 14 Interviews 6, 20, and 23. 15   The Greek government improvised these registration processes, despite the fact that UNHCR had established sophisticated procedures to conduct asylum registration over the course of previous crises. 24 The Greek government refused UNHCR support because early miscommunication between Greek and UNHCR officials soured their relationship. At the beginning of the crisis, UNHCR workers reportedly did not treat government officials with adequate respect; they assumed that Greek institutions could not manage the crisis, akin to failing institutions in developing countries hosting refugee communities. 25 In response, the Greek government largely cut UNHCR from key crisis policy development and coordination. 26 Second, the Greek government restricted information about asylum procedures to try to stabilize refugees' movement in the country and reduce their use of smugglers. UN-HCR and Greek officials frequently obfuscated the length of the asylum application process to discourage refugees from leaving through smugglers. 27 During the in-person preregistration process, up until a mobile unit entered a camp, asylum seekers had little to no information about who conducted preregistration, when it would start in a given camp, and the rights that they would receive once preregistered. 28 Government and UNHCR officials hoped that, by restricting this information, they would prevent refugees from traveling en masse to the camp where preregistration was underway. In the absence of such critical information, asylum seekers' confusion about legal options to leave the country increased and they turned to rumors from informal information providers.
Additionally, the Greek government limited the information it provided aid organizations, in part due to its inexperience and in part to control the (sometimes false) information spread by the aid and volunteer groups that rushed into Greece to assist refugees. 29 Sometimes, this form of information restriction had serious consequences. For example, the Greek government generally gave aid organizations less than a twenty-four-hour notice before they would close one camp and transfer refugees to another. 30 One NGO worker said government officials notified them at midnight to provide toilets for a camp opening the next morning. Such short notice prevented aid organizations from ensuring that 24 For example, since 2010, UNHCR has implemented a biometric identity management system, which it uses to coordinate food distribution, vaccination, and other needs for over 4 million displaced people in Africa and the Middle East (Nonnecke 2017 newly opened camps met basic humanitarian standards. 31 Moreover, by rapidly moving refugees, the government disincentivized aid organizations from investing in camp infrastructure. Greek officials also restricted refugees' access to information in camps. 32 For example, when the government moved residents of Piraeus port E1 to port E1.5 (about one mile away), they gave refugees a three-hour notice. This quick move particularly burdened refugee families who moved more slowly than single migrants. Since families did not want their children in the blinding sun, fights broke out over shaded tent locations. 33 Moreover, the government restricted aid organizations' and legal aid workers' access to camps and detention centers, hoping to prevent unverified or illegitimate volunteer groups and aid organizations from spreading misinformation. 34 However, this prevented refugees from accessing critical legal information, particularly about filing appeals. 35 For example, EASO's pervasive barring of lawyers from the Moria detention center led local police to inform an EASO representative that he would have to allow attorneys access. 36 Government officials also limited information about the rights refugees received during the asylum process. Even after preregistration, refugees' asylum application remained pending until their asylum appointment, which generally occurred six months later. 37 Often, asylum seekers did not know that preregistration was separate from the asylum process and that they could only exercise certain rights after their first asylum interview. 38 The Asylum Center left the communication of information about rights access to camp managers, who themselves had little knowledge of refugees' rights. 39 Third, in attempts to more quickly process asylum applications, Greek government officials implemented ad-hoc policies to distinguish between those who had valid asylum claims and those who did not. Even prior to the influx of asylum applications that occurred after the Macedonia border closure, Greek government officials used asylum seekers' nationality to determine rights access. 40 For those arriving after the implementation of the EU-Turkey deal on March 20, the Greek government put in place expedited asylum procedures, meaning they would decide whether they could safely return applicants to Turkey, or whether they had a sufficient asylum claim to stay, in a 31 Ibid. 32 Interviews 13,15,16,18,20,21,22,and 25. 33   Increased confusion about the rights received during asylum application process "After having an appointment, you have the right to demand an apartment."-August 5, 2016 When you register via Skype, you immediately get housing and money card."-August 12, 2016 Decreased trust in government officials "The relocation program is a lie. They use it to calm us down and slowly get rid of us by sending us to camps in nowhere, where we will be forgotten, unlike here with all the media."-April 22, 2016 "The Wifi in the camp is really weak and that is on purpose, because they do not want people to be in contact with outside world (including smugglers)."-July 15, 2016 "They take us to the mainland, keep us there for a while, then deport us to Turkey"-April 1, 2016 matter of days. 41 If the government accepted an applicant's asylum claim, they remained in island detention centers until they provided paperwork necessary to travel to the mainland to complete the application process. 42 While instituting a fast-track process seemed sensible, officials' actual implementation of EU law on the ground differed drastically. Challenges with the implementation of EU directives are widespread across issue areas (Linos 2007, 547-50), but are especially acute in this field. Government officials prioritized Syrian applications under the blanket assumption that they had valid asylum claims. Conversely, the Greek government often stalled Afghans' and Pakistanis' applications because government officials assumed they lacked valid asylum claims. Human rights experts have criticized these procedures, and UNHCR suspended many of its activities in Greece following these policy changes (UNHCR 2016b).
Greek government officials also relied on arbitrary policy implementation to control asylum seekers' movement, particularly to shift them from informal settlements to formal camps. For example, Piraeus Port police officers allowed only migrants with a special stamped card to move in and out of the port, regardless of legal status. Many did not obtain the stamp, meaning they could leave the port, but not return. This policy created particular problems for asylum applicants and those with health problems that had to leave the port for doctor's appointments and asylum interviews. 43 Those without the stamped card would find dangerous ways of returning to Piraeus, often resorting to jumping off the freeway overpass near the port to gain entry. 44 While the Greek government pursued the policies described above to increase their control of the crisis and deter asylum seekers from using informal pathways to move onward, the policies backfired by encouraging the very behavior the government sought to restrain. We explain how in the next section. 41 Interviews 18,20,23,and 25. 42 Interviews 16,18,20,23,and 25; discussions with refugees in Moria detention center. 43 Discussions with volunteers and refugees in Piraeus. 44 Ibid.

Effects on Refugee Behavior: Reduced Interactions with Authorities, Mass Mobilizations, and Emerging Ethnic Tensions
In this section, we first present evidence that asylum seekers became more distrustful of government and aid organization officials in response to the government's changing policies and insufficient information provision. These policies made refugees increasingly reliant on informal sources of information, leading them to withdraw from government officials and mobilize to demand their rights. Next, we demonstrate that the Greek government's arbitrary policy implementation heightened ethnic tension between asylum seekers, escalating policy noncompliance and outbreaks of violence. The Greek government's rapid policy changes and selective provision of information heightened uncertainty and anxiety among refugee communities and reinforced distrust in UNHCR officials as well. Almost every single refugee family we spoke with expressed high levels of distrust in Greek government and UNHCR officials, citing inconsistent information, incorrect information, or its complete absence. 45 Refugees only reported trusting aid workers who could answer their questions and provide consistent, concrete information about the asylum process; these were few and far between. 46 Distrusting official sources, refugees sought alternatives, namely smugglers, other asylum seekers, social media sites, and websites. Several refugees in Scaramangas openly discussed their conversations with smugglers, saying that they provided critical information on when they could easily cross the border.
Relying on these unofficial sources ultimately bolstered refugees' perceptions that they could viably leave the country through smuggling. In Table 2, we draw from News that Moves newsletters to illustrate examples of false rumors that emerged among asylum seekers after the Greek government pursued these policies.
While officials believed that rapid policy adaptation helped correct mistakes, these policies influenced refugees' behavior in unanticipated ways. First, this distrust led refugees to reduce their communication with government and UNHCR officials, exacerbating noncompliance. During interviews with refugees in Scaramangas and Piraeus, roughly 36 of the 60 refugees asked for information about 45 Discussions with refugees in Scaramangas, Piraeus, and Moria. 46 Ibid.

Government policy Impact on asylum seekers Subsequent rumors
Restricted refugee movement Decreased trust in government officials "The police said that the stamps they are putting on our khartis are for exit/entry, but we believe that these stamps represent our acceptance to be moved to other camps. They are lying to us but we can do nothing since they put policemen on all gates who require to see the stamps to let us in."-April 22, 2016 Arbitrary application of refugee laws Increased willingness to use smugglers "If you make it through one of the neighbor countries, even if you are detained, the change of relocation to better countries like Germany is higher."-July 8, 2016 "You can go to Canada from France. You talk to the mafia and in 5-6 days they'll take you to Canada by boat for 10,000 euros."-June 17, 2016 Decreased trust in government officials and aid organizations "Greece is blackmailing the EU regarding its debts. If the debts are not cancelled then Greece will give Schengen visas to all refugees."-May 20, 2016 "Warehouses are full of tuna and sardines for instance. When we ask for one they say 'there is none.' Maybe they are selling them."-May 13, 2016 Increased ethnic discrimination "We did the preregistration and our next appointment is on the 16th of August, but we heard that all Iraqis will be rejected."-July 29, 2016 "They say Afghans are the third most vulnerable on the list. They will separate Afghans and spread Syrians to other EU countries. Afghans will be stuck here."-May 6, 2016 preregistration. When we recommended speaking with UN-HCR or Greek government officials, most refugees said they did not want to approach them. Reducing their interactions with officials likewise affected how refugees accessed other services. For example, refugees believed primary healthcare services were unavailable even though aid workers asserted that doctors were there during working hours. 47 Distrust and lack of interactions with government and UNHCR officials led migrants to disregard government demands, in turn weakening Greek officials' ability to maintain security. For example, on July 27, 2016, a group of 1,500 Pakistani men gathered at Scaramangas camp, where the mobile unit was preregistering refugees, and demanded that the government preregister them as well. The government had only allowed Pakistanis to preregister through Skype, assuming they lacked valid asylum claims. 48 This mass gathering led to the suspension of the preregistration exercise due to safety concerns. UNHCR suspended its service programming in the camp thereafter. 49 This mobilization barred the Asylum Center from registering tens of thousands of migrants and prevented the government from ascertaining how many refugees resided in the country and how many planned to legally move onward.
Lastly, mistrust indirectly influenced the rate at which migrants attempted to leave the country informally. Indeed, migrants generally perceived information from smugglers as more credible not because of its truthfulness, but because of its accessibility, consistency, and concreteness. 50 Moreover, smugglers created rudimentary customer service procedures that offered informal passage between countries. 51 For example, smugglers who facilitated travel by plane often bought migrants clothes and haircuts to make them appear more European, increasing the probability of successful passage. By creating these services, smugglers seemingly reduced risk. 52 However, smugglers scammed, trafficked, or 47 Discussions with aid workers at Scaramangas. 48 Ibid. 49 Ibid. 50 Interviews 6, 9, and 16; discussions with refugees in Scaramangas. 51 Interviews 9, 16, 21, and 22; discussions with refugees and volunteers in Piraeus and Scaramangas. 52 Ibid. otherwise exploited refugee "customers" on these passages. Despite these risks, smugglers had the incentive to spread (mis)information about the risks of informal movement for economic gain. For example, an aid organization official reported that some smugglers imitated the asylum application process to make it seem as if they were affiliated, requiring that refugees apply through Skype and attend interviews. 53 Through this intentional misinformation, they convinced asylum seekers to move onward informally. Furthermore, Greek government officials' arbitrary implementation of asylum policies led asylum seekers to believe that government officials were biased towards particular ethnic groups. Formal EU and Greek policies created important advantages for Syrian nationals in the speedy processing of asylum claims. However, rumors about the discriminatory treatment of particular ethnic groups extended to other policy areas, such as food distribution, medical care, and police enforcement. For example, many refugees believed government officials targeted particular ethnic groups with harassment, beatings, and arrests. 54 In Table 3, we draw from the News that Moves weekly newsletters to provide examples of false rumors that emerged in response to the government's restriction of refugee movement and arbitrary application of asylum policy.
These beliefs influenced asylum seekers' behavior in three ways. First, perceived discrimination incentivized noncompliance. Afghan migrants resisted the government's efforts to move them to formal camps, making it difficult to control their movement throughout the country. Greek officials often segregated camps by ethnicity, meaning that they sent Afghans and Pakistanis to specific camps while sending Syrians and Iraqis to others; this exacerbated perceptions of discrimination. 55 Since the conditions and services available in each camp varied dramatically, many believed Arab asylum seekers received preferential treatment. 56 Rather than remain in camps in which they were registered, Afghans and Pakistanis returned to and resided in informal camps or 53  illegally entered majority-Arab formal camps, believing that majority-Arab camp had a higher quality of life than the majority-Afghan and Pakistani camps. 57 Second, people who anticipated discrimination were less likely to access critical services and make rights claims. For example, in Piraeus, Afghans generally refused to provide their name, phone numbers, and other personal information to Greek doctors. 58 Afghan migrants feared that, since they did not have a sufficient asylum claim in the eyes of the government, doctors would share this information with the police, who would deport them. 59 As a result, Afghan migrants only accessed primary healthcare for emergencies. Moreover, because Afghans refused to provide personal information, doctors were unable to arrange for them to receive services only provided in hospitals. 60 Migrants with serious, chronic illnesses would forego medicine and care to maintain anonymity; for them, refusing access to maintain anonymity secured their freedom of movement. 61 This disengagement undermined the government's ability to monitor and track health issues within refugee communities. 62 Third, perceived government discrimination increased ethnic tensions between asylum seekers, leading them to actively discriminate against, and in some cases attack, refugees of other ethnicities. For example, volunteers in Scaramangas stated that an incoming group of Iraqi Yazidis demanded to live within the Yazidi "block" in the camp, fearing that Syrians would attack them if they were dispersed. When transferring from one camp to another, many would refuse to move if they could not live within their ethnic "neighborhood," citing concerns about safety. Across Piraeus and Moria Detention Center, refugees would organize their tents by ethnic group. The ethnic segregation in camps created separate networks through which refugees and migrants transmitted rumors. In Moria, Syrian women would refuse to travel outside their block. This limited their access to information about available services. In Scaramangas, we observed Syrian refugees refusing to tell Iraqis information about services, saying that only Syrians could access them.
Ethnicity-based rumors also exacerbated perceptions of discrimination. Rumors about Syrian discrimination against Afghan migrants, and vice versa, were pervasive within each ethnic community. For example, Afghans felt that the shortage of Farsi translators and relative abundance of Arabic translators restricted their access to services. 63 The translator gap lead Afghan migrants to believe that Syrians had stronger connections with volunteers and could secure better treatment.
Even though many EU and Greek government policies privileged Syrians, misinformation spread through ethnically segregated networks led many Syrians to believe that Greek officials discriminated against them relative to other ethnicities. Discussions with Syrian refugees in Moria indicated that they believed Afghan and Pakistani families received higher quality and more secure housing, as well as greater access to available food. This perceived discrimination led many Syrian refugees to form and spread nega-57 Discussions with refugees living in and volunteers working at Piraeus Port and Scaramangas Camp. Carlson also observed many Afghan and Pakistani families objecting to transfers to camps with co-ethnics, believing that camp conditions were particularly bad. 58 Table 4.

Counterfactual analysis
This section explains how an environment rife with rumors might differ from a world of full information. First, in a world with full information, we would expect to see greater use of legal routes compared to informal routes. We would expect favored groups, such as Syrians, for whom legal procedures were almost guaranteed success, to use legal routes. We observed low initial application and appeals rates for asylum in Greece across all ethnic groups, as many refugees believed the process was a money-making scam.
Only 51,091 people applied for protection in Greece in 2016, even though over 173,000 displaced people arrived in Greece by sea that year, adding to the 800,000 plus people who had arrived in 2015 (Konstantinou et al. 2016, 8, 13). Although the Greek Asylum Service did not process most asylum applications in a timely manner, applicants receive important protections and rights guarantees the moment they file an application. Moreover, even though the Asylum Service rejected many processed applications at the first instance, refugees rarely appealed these decisions (Konstantinou et al. 2016, 9). If displaced persons knew that for 2015 and much of 2016, Greek judges granted approximately one-half of all appealed cases, we would expect many more asylum seekers to file appeals (Eurostat 2017 application rates for asylum and other forms of protection among favored groups such as Syrian nationals. In a counterfactual world of full information, we would still expect use of smugglers among less privileged groups, notably Afghans and Pakistanis. Groups who expected that the government would reject their asylum claims often eschewed legal processes, either staying in Greece without papers or traveling northward with smugglers. However, in a world with better information, we would also expect the government to more effectively differentiate between undocumented migrants and refugees and thus more quickly grant asylum. As such, we would expect more individuals who could document individualized threats of persecution to file for asylum and more individuals to file appeals against negative initial decisions.
In addition, relative to a world rife with rumors, we expect that a world with greater information about asylum policies would lead to major differences in refugees' interactions with government and aid workers. Specifically, we would expect greater trust in government and aid organization policies, whether these policies concern much-needed benefits or compliance with unfavorable developments like border closures. Indicatively, a Syrian refugee living in Athens stated, "Everyone just gives fake promises so as to get rid of us. . . . It's better if they give honest answers and not manipulative ones . . . we don't trust any organization anymore." 67 In a world of information, we should not see migrants refusing to provide doctors personal information because they believe that doctors would share it with police. Rather, we should see increased trust in formal institutions and greater access of services. At the same time, greater trust in government would also imply greater adherence with unfavorable government announcements. For example, asylum seekers would perceive EU and host government announcements about camp and border closures as credible, and refugeesespecially favored groups-would look to find alternative housing rather than squatting for months in unfavorable conditions.
Finally, we expect refugees to more correctly attribute blame for unfavorable policies and events on policymakers rather than on one another, reducing ethnic conflict. Whereas ethnic conflict among earlier waves of Balkan refugees has been explained through ancient hatreds, in Greece, violent tensions developed between groups without extensive prior contact-such as Pakistanis and Syrians. More accurate information could help reduce these newfound hostilities, because refugees could more closely tie policy decisions to government officials. With more accurate attribution, fewer people would circulate and act on rumors blaming other ethnic groups for unfavorable policies. We do not imagine that accurate policy communication would fully satisfy refugees; Afghans and Pakistanis would still understand that they require more documentation and time to successfully obtain asylum compared to Syrians. However, with better communication, all ethnic groups would turn to aid workers for basic needs more frequently. Moreover, more Syrians would perceive their advantaged status rather than understand themselves as disadvantaged vis-a-vis other ethnicities.

Conclusions and implications
Rumors are a critical, yet understudied aspect of refugee crises. In this article, we demonstrated how government and aid organization policies unintentionally exacerbated anxi-67 Drawn from the News that Moves dataset, collected on 11/10/16. ety among asylum seekers, creating an information vacuum. We showed how frequent policy changes, limited communication, and ad-hoc policy implementation increased asylum seekers' distrust in government officials and ethnic outgroups. In this information vacuum, refugees actively sought smugglers and other informal sources for information. In turn, refugees acted on rumors in ways that undermined migration policy and stymied the Greek government's crisis management efforts.
While our study speaks directly to migration and refugee policy, we anticipate that our theoretical model extends to other crises contexts, such as natural disasters and epidemics, which also involve high anxiety and uncertainty (Allport and Postman 1947;Grein et al. 2000;Oh et al. 2010). For example, Liberia, a developing country affected by the Ebola crisis, acutely felt the pervasive, destructive effect of rumors. Government agencies and aid organizations initially provided limited information about the Ebola outbreak; in this information vacuum, Liberians increasingly believed that healthcare workers were spreading the virus through vaccines. Ultimately, more Liberians died from not accessing primary healthcare services than from the virus itself (Kamradt-Scott 2016). Similarly, after the 2015 Nepalese earthquake, thousands swarmed the streets of urban areas, believing viral misinformation on WhatsApp and Facebook that warned of a more devastating earthquake ahead. This made crisis management extremely difficult for authorities (Express News Service 2015).
We also expect that our theoretical model explains some acute governance failures in non-crisis contexts. Developing countries often have complex, disorganized, or corrupt government bureaucracies, making them likely to mishandle information. When developing countries provide contradictory or insufficient information about policy changes or public services, it exacerbates already low levels of trust in bureaucracies. We anticipate that residents will rely on rumors, particularly information provided by informal brokers, rather than navigate such bureaucracies. We also expect that the accessibility of informal brokers weakens residents' relationships with officials, and that over time they stop relying on government information to make critical decisions about services. For example, poorly publicized government housing programs allow slumlords to promote squatter settlements. Feedback cycles cement spaces for informal actors to operate, creating long-term governance problems.
Even in highly developed countries, inconsistent government messaging has deleterious effects. In the United States, the Trump administration announced plans to cut the Deferred Action for Child Arrivals (DACA) program in 2017'. Fearing that Department of Homeland Security would use their personal information to identify and deport them, DACA recipients reportedly disengaged from the government, relying instead on private actors in informal sectors for information and assistance (Mark 2017). For example, some DACA recipients stated that they will set aside career ambitions, preferring to work in jobs like construction, where they "can get paid under the table" (Gonzales 2017, 1).
Our findings therefore matter for information management, dissemination, and implementation across a wide range of contexts. Currently, a wealth of important human rights documents, including the 1951 Refugee Convention, offer little by way of conceptualizing the right to information despite the fact that refugees often consider the ability to access information more critical than food or shelter (Gillespie et al. 2016, 11).
In the past decade, national governments slowly adopted information transparency laws in response to widespread criticism. However, international organizations still lack comparable information transparency policies. Recently, a UN special rapporteur contacted dozens of international organizations for details on their information transparency policies. Only a handful of organizations-primarily financial institutions-responded (Kaye 2017, 10). The UN Secretariat, which lacks transparency standards across agencies and employs ad-hoc standards for access-to-information requests, did not respond (Kingsbury and Casini 2009;Kaye 2017, 4). Moreover, international organizations lack external accountability mechanisms, because news media do not subject them to the same "journalistic microscope" as national governments. Although crisis situations make international organizations' efforts at transparency uniquely pertinent, institutional leaders often assert a false trade-off between policy transparency and easing human suffering as quickly as possible (Kingsbury and Casini 2009, 10). In crisis contexts, we anticipate that rumors can seriously reduce beneficiary engagement with international organization programming.
We argue that governments and aid organizations need to redesign and expand migration policies in particular, and crisis policies more generally, in consideration of the critical need for information. We anticipate that governments and aid organizations must do more than simply disclose policies; they must incorporate transparency into policy design at an earlier stage. 68 Governments and aid organizations maintain greater legitimacy when they provide clear, consistent, and timely information and signal transparency and trustworthiness to affected populations. Governance institutions must increase trust in policy and reduce reliance on rumors to bolster policy compliance and in turn boost their own management capacities.

Supplementary Data
Appendix A, presenting supplemental information on the interviews and fieldwork that forms the basis of this article, can be found at https://www.melissaannecarlson.com and at the International Studies Quarterly data archive.