Using Behavioral Insights to Improve Disaster Preparedness, Early Warning and Response Mechanisms in Haiti

USING BEHAVIORAL INSIGHTS TO IMPROVE DISASTER PREPAREDNESS, EARLY WARNING AND RESPONSE MECHANISMS IN HAITI Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized Public Disclosure Authorized CONTENTS 1 Introduction 6 2 Methodology and description of data 9 3 Background and Context 13 4 Disaster Risk Management in Haiti 20 4.1 Operating model 20 4.2 Roles and key actors 22 4.3 Preparedness process 22 4.4 Communication channels 23 4.5 Evacuation process 23 5 Key Insights on Barriers to Evacuation 25 5.1 Obstacle 1: EWS Messages Do Not Arrive 25 5.2 Obstacle 2: People Do Not Understand EWS Messages 27 5.3 Obstacle 3: People Struggle to Internalize Risk 29 5.4 Obstacle 4: People Lack Access to Resources and Shelters to Evacuate 30 5.5 Obstacle 5: People Prefer Not to Evacuate 30 6 Ideas to Nudge People to Evacuate to A Safe Place on Time 34 7 Conclusion 39 Bibliography 41 Appendix 1 – Field Study Methodology and Diagnostic Activities 43 Appendix 2 – Hurricane classification 45 ACKNOWLEDGEMENTS T his report was prepared by a World Bank team composed of Jime- na Llopis (Behavioral Scientist, Poverty & Equity GP), Emilie Perge (Economist, Poverty & Equity GP), Zeina Afif (Senior Social Scientist, Poverty & Equity GP), Claudia Soto Orozco (Disaster Risk Management Spe- cialist, Social, Urban Rural and Resilience GP), Lace Padilla (Consultant, Pov- erty & Equity GP), Jessica Hsu (Consultant, Poverty & Equity GP) with inputs from Cecile Lorillou (Consultant, Social, Urban Rural and Resilience GP); Ali Alwahti (Consultant, Social, Urban Rural and Resilience GP), Roland Brad- shaw (Senior Disaster Risk Management Specialist, Social, Urban Rural and Resilience GP), and Giovanni Michele Toglia (Consultant, Social, Urban Rural and Resilience GP) with overall guidance from Oscar Calvo-Gonzalez (Prac- tice Manager, Poverty & Equity GP), Raju Singh (Program Leader, LCC8C), and Pierre-Xavier Bonneau (Program Leader, LCC8C). The team would like to thank Oscar Anil Ishizawa (Senior Disaster Risk Management Specialist, Social, Urban Rural and Resilience GP) for his comments on an earlier draft. The present research, including qualitative data collection was fi- nanced by the EU-funded ACP-EU Natural Disaster Risk Reduction Program, managed by the Global Facility for Disaster Reduction and Recovery. This report has benefited from discussion with government officials, develop- ment partners and members of the community, including representatives from The Civil Protection Directorate (DPC) within the Ministry of Interior and Local Authorities (MICT), the Ministry of Education (MNEFP), the munic- ipalities of Paillant and les Cayes; the United Nations Development Program (UNDP), the United Nations Office for the Coordination of Humanitairian Affairs (OCHA), the International Organization for Migration (IOM), and the United States Agency for International Development (USAID). The qualitative data collection was performed by the team with assis- tance from Jessica Hsu, Donald Antoine, James-son Vamblain, Manouchka Justin, and Reginald Milfort. The team would like to thank key informants and participants in the focus-group discussions and semi-structured inter- views in Les Cayes and Paillant in the South Department. The opinions, in- terpretations, and conclusions expressed herein do not reflect the views of the World Bank, its Board of Executive Directors, or the Governments they represent. Photo: UPC-DPC Given the high exposure in Haiti to natural hazards, ensuring that proper disaster preparedness and re- sponse mechanisms are well established and managed can save lives. Systems such as Municipal Civil Protec- tion Committees ( Comités Communaux de Protection Ci- vile – CCPCs) have played an important role in reducing casualties from natural disasters in recent years. Howev- er, structural and behavioral barriers have limited desired evacuation behaviors in Haiti. This report leverages behav- ioral science to identify key entry points to improve early warning systems (EWS) and enable people to evacuate to safe locations before adverse natural events, including hurricanes. The behavioral approach allows for the detec- tion of barriers related to psychological and social factors that go beyond access and cost. The objective of this report is to understand, through a behavioral approach, the structural and behavioral barriers that limit evacuation decisions in Haiti . Our goal is to identify key entry points to improve EWS and enable people to evacuate to safe locations in anticipation of a hur- ricane. This type of approach is based on the understanding that people think automatically, socially, and with mental models (World Bank, 2015). The framing of a problem, the context in which decision making takes place, and details of the design of an intervention play such an essential role in determining behavior that not accounting for them can result in an ineffec- tive intervention. Thus, looking for both structural and behavioral barriers means to examine a wider set of influences and pay attention to the social, psychological, as well as economic factors that affect what people think and do. 1 Introduction 7 Informed through desk research and primary qualitative data (inter- views and focus groups), this report maps the decisions and actions that influence individuals in Haiti and identifies the barriers impeding people from following EWS and seeking safe shelters. They are: 1. Warning dissemination and communication: EWS messages do not always reach the population due to limited resources and funding. When messages reach the population, people often do not understand them given that the language is unclear and lacks information about the desired behavior. Furthermore, people may distrust the messenger. 2. Preparedness and response capacity: Messengers may lack ap- propriate training to make people evacuate and contingency plans may not be available or up-to-date. Moreover, people lack prepara- tion experiences. The population lacks access to resources needed to evacuate and has difficulty accessing shelters. Often the experi- ences of emergency evacuation shelters are dissatisfactory. Some perceive shelters as unsafe as structures do not always abide by building codes/norms, resulting in not everyone preferring to evac- uate to a shelter. 3. Risk knowledge and internalization: People struggle to internal- ize the level of risk. They either underestimate the risk level due to state of denial or fatalistic belief or they experience myopia, pre- ferring to stay with their possessions (short-term gains) instead of evacuating to a shelter (long-term gains). In a complex context such as Haiti, potential solutions range from simplifying communication messages and channels, training messengers to deliver concise and persuasive messages, designing clear messages to counter fatalistic beliefs, and using loss aversion framing to nudge people to internalize risk and evacuate. To keep shelters preserved and promote codes of conduct, a community recognized figure could be set up as a shel- ter manager. The following document is organized as follows. Section two describes the methodology while section three outlines background and context. Sec- tion four describes the functioning of Haiti’s disaster risk management sys- tem. A fifth section explains barriers to safe evacuation. Section sixth pro- vides key ideas for the design of the intervention(s). Finally, section seven provides concluding thoughts. Detailed information on hurricane classifi- cation and diagnostic activities conducted can be found in the appendices. 1. Introduction Photo: UPC-DPC 2 Methodology and description of data The evidence in this diagnostic is based on qualitative research: desk review, key-informant interviews, and qualitative fieldwork. The diagnostic started with an ex- tensive desk review of existing reports and research papers on (i) EWS in the Haitian context, (ii) empirical research on evacuation behavior, and (iii) case studies in develop- ing countries. The reports helped inform the background section as well as our in-depth understanding of how the current disaster risk management system works in Haiti. Through a meta-analysis comprising various research pa- pers (Huang, Michael K., & Prater, 2016), we learned that, throughout the world, people are more likely to evacuate when they have an accurate understanding of the severity of a storm. Severity has numerous facets including inten- sity, damage, landfall, and rapid onset. Moreover, people use rules of thumb to help them determine their person- al risk, including if their neighbors are evacuating and if businesses in their community are closing. Some studies found that women with children are more likely to evac- uate. Several case studies of evacuation experiences in Marikina, Philippines, Mumbai, India, Uganda, and rural Indonesia show that risk perception about the likelihood and severity of a hazard is one of the key factors that pre- dicts risk preparedness. To complement past research USING BEHAVIORAL INSIGHTS TO IMPROVE DISASTER PREPAREDNESS, EARLY WARNING AND RESPONSE MECHANISMS IN HAITI 10 findings and to better understand the complex nature of the issues surrounding disaster preparedness, EWS and response mechanisms in Haiti, the team interviewed key informants such as the Civil Protection Directorate ( Direc- tion de la Protection Civil – DPC), national counterparts working with the DPC, international partners working on disaster risk management (DRM) in Haiti (UNDP, USAID, OCHA), and NGOs. A total of nine interviews were conduct- ed, all in Port-au-Prince and its surrounding vicinity. Qualitative fieldwork was later conducted to capture individual experience, choices, perceptions, and attitudes towards evacuation in Haiti. Fieldwork aimed at understanding barriers to EWS and what limits people’s ability to evacuate to safe locations during a hurricane. The meth- ods chosen for this study were six Focus Group Discussions (FGDs) and 16 Semi-Structured Interviews (SSIs) with a purposely selected range of ac- tors from three categories: (i) CCPCs, (ii) community leaders, and (iii) the general population. Community leaders included teachers, Red Cross staff/ volunteers, pastors, women’s group leaders, departmental delegates, and beyond (see Appendix 1 for more details). The general population included men and women that did and did not evacuate in the past. The diversity of interviewees is essential in qualitative research to gather more detailed information on behaviors and barriers to safe evacuation. Local consultants transcribed audio recordings of FGDs and interviews in Creole. Data was coded in Excel. Sites and respondents were intentionally selected to answer our research questions best. Site selection was done through a two-stage pro- cess. We first looked at communal sections that were located within a risk area and who had a relatively easy access to a shelter. The second stage controlled for past performance (good or bad) and location (rural or urban). The DRM World Bank (WB) team provided a list of five communal sections that met our preferred criteria and logistical requirements. They helped se- lect Paillant (pop. 16,000) and Les Cayes (pop. 86,780), municipalities lo- cated in Nippes and Sud departments, respectively. Paillant was selected as the rural site and positive deviant, 1 where the team visited the Centre d’Etudes Classiques Theophile . 2 Les Cayes (in Sud) was selected as the ur- ban site and the team visited Lycée P Guerrier shelter because its access to 1 Paillant is known as a good practice in terms of preparedness and CCPC performance. A positive deviant is an approach to behavioral and social change based on the observation that in any community there are people whose uncommon but successful behaviors or strategies enable them to find better solutions to a problem than their peers, despite facing similar challenges and having no extra resources or knowledge than their peers (Wikipedia, 2018). 2 Initially the team planned to visit Ecole Note Dame de la Victoire shelter, but the school was not accessible since it was in session during the visit. 11 schools allowed us to speak to teachers who commonly communicate EWS to students. Analyzing qualitative data provide in-depth details about expe- riences, attitudes, and perceptions. Qualitative data analysis allows re- searchers to uncover unexpected themes during analysis (Fusch and Ness 2015, Tracy 2010). Although qualitative data analysis does not prioritize the quantity of statements over richness and detail, a rule of saturation point was followed. According to this rule, the researcher will explore data until no new themes can be identified (Fusch and Ness 2015). Also, the sample is not and does not pretend to be representative of the whole population. The objective was not to test causal links or to generalize findings, but to capture views and experiences of people and the ways they express them. The recruitment strategy followed best-practices and was success- ful in recruiting desired profiles. Mobilization was conducted one week in advance with the help of the Departmental Technical Coordinator from the DPC and the Board of Directors of the Communal Section ( Conseil d’Ad- ministration de la Section Communale – CASEC). Once in the field, the team further mobilized participants by going door to door distributing flyers, ac- companied by a community member. As is common in qualitative research, our study overcame some challenges in its implementation. The initial dates for fieldwork were pushed back due to country-wide protests from October 15 to October 22, 2018. Furthermore, in Paillant, the CASEC had misunderstood some of the instructions, so many of the members of the CCPCs FGD on the first day were also community leaders, such as pastors, teachers, or public health representatives. Therefore, the team was unable to get their perspectives from the other roles that they play in the community, but only as CCPCs vol- unteers. The field team worked with the CASEC to recruit additional com- munity leaders after the initial CCPCs FGD. In Les Cayes, the member des- ignated to assist in mobilization only mobilized community leaders he was close to and would not assist in the mobilization of the general population, introducing a potential bias in the data. The field team made its best effort to further mobilize the general population in a short-time frame to reduce this bias. 2. Methodology and description of data Photo: UPC-DPC Haiti is highly exposed to natural hazards. Over 93 per- cent of its surface and more than 96 percent of the pop- ulation are exposed to two or more hazards, primarily hurricanes, floods, earthquakes, and landslides, but also tsunami and drought (World Bank, 2018). Haiti is highly subject to earthquakes, created by the interaction of the Caribbean and North American tectonic plates. Other sec- ondary hazards impacting Haiti include landslides, tor- rential debris flows, soil liquefaction, and tsunamis. The Grand-Sud , which includes the departments of Grand’Anse, Nippes, and Sud are more vulnerable to hurricane risks, and the Nord and Nord Ouest Departments are highly ex- posed to the risks of floods and tsunami. The human and economic impact of disasters has been extremely severe, due to Haiti’s exposure to hazards, the high vulnerability of its infrastructure, unplanned urban expansion, and institutional fragility. Between 1961 and 2012, the country experienced more than 180 disasters causing the death of more than 240,000 people (World Bank, 2018), 3 includ- ing the death of about 220,000 people after the 2010 earthquake (World Bank, 2015). Regarding economic impacts, the 2010 earthquake destroyed the equivalent of 120 percent of GDP and Hurricane Matthew resulted in estimated damages and losses equivalent of around 32 percent of GDP. 3 An event is considered a disaster by the EM-DAT database if (i) it caused at least 10 deaths, (ii) affected at least 100 people, (iii) caused an emergency declaration or (iv) led to a call for international assistance. 3 Background and Context USING BEHAVIORAL INSIGHTS TO IMPROVE DISASTER PREPAREDNESS, EARLY WARNING AND RESPONSE MECHANISMS IN HAITI 14 Additionally, the potential future maximum losses that could be caused by hurricanes and earthquakes occurring within a 250-year return period are estimated at US$1.6 billion (13.3 percent of 2016 GDP) and US$2.41 billion (27.5 percent of 2016 GDP), respectively (World Bank, 2018). Climate change will increase the frequency of, and impacts from, extreme weather events. In 2017, the Maplecroft index ranked Haiti as amongst the top three most vulnerable countries in the world to climate change (Maplecroft, 2018). With forecasted increases in temperature from 1.2ºC to 2.3ºC by 2100, Haiti will likely experience a rise in the frequency and intensity of hurricanes while more than doubling the length of the dry season (World Bank, 2018). The increased frequency of hurricanes is already evident in historical data (Figure 1). FIGURE 1. HURRICANES IN HAITI (1950-2018) Source : National Hurricane Center https://www.nhc.noaa.gov/aboutsshws.php Additionally, Haiti has not been able to reduce the vulnerability of its in- frastructure (public and private buildings) and residential buildings through the application and enforcement of building regulations. Available data in- dicates that built-up areas in Haiti are particularly vulnerable. Most Haitians live in self-produced housing, built without the appropriate technical ex- pertise, where 51 percent are exposed to flood events and 60 percent are concentrated in high seismic hazard zones (World Bank, 2017). The vulnera- bility of infrastructure was evidenced by the impact of Hurricane Matthew; 59 percent of the hurricane’s damages and losses were in infrastructure and residential buildings and 31 percent were in the housing sector alone. 15 In addition, the frequency of natural disasters makes it harder for Haiti to recover fully. In 2018, just two years after Hurricane Mathew, a 5.9 earth- quake shook Haiti and killed 11 people (The New York Times, 2018). Hurricane Mathew (2016) is considered the most recent devastat- ing hurricane that has made landfall in Haiti. Hurricane Matthew made landfall in Haiti’s western peninsula as a category 4 and caused a loss equal to roughly 32 percent of GDP, about 600 deaths, and 2 million people af- fected, particularly in the southern regions of Haiti (Government of Haiti; United Nations; European Union; Inter-American Development Bank and World Bank, 2017) (Map 1). Although the death toll was high, deaths associ- ated with Hurricane Matthew were significantly lower than 2004’s Hurricane Jeanne (approximately 3,000 deaths) and 2008’s Hurricanes Fay, Gustav, Hanna, and Ike (approximately 800 deaths) (World Bank , 2018). The most affected provinces by Hurricane Matthew – Grand’Anse, Nippes, and Sud – are home to nearly 1.6 million Haitians (14.5% of the country’s population, estimated at 10.9 million in 2015). MAP 1. SPATIAL DISTRIBUTION AND IMPACT OF HURRICANE MATHEW 1. Speed (mph), per commune 2. Estimation of infrastructure damage (USD) per commune Source: National Hurricane Center/University College London/ CNIGS Spatial Team, October 2016 Source: Haiti Rapid Post-Disaster Buildings Economic Loss Assessment D-RAS, World Bank A report by the Haiti Disaster Risk Management and Reconstruction Project (DRMRP) links the reduction in loss of lives to the establishment of CCPCs (World Bank , 2018). A CCPCs is a municipal level committee formed of volunteers from the community who coordinate first emergency respons- es, warn the population about a storm, and help people find shelters as part of the EWS (Box 1). 3. Background and Context USING BEHAVIORAL INSIGHTS TO IMPROVE DISASTER PREPAREDNESS, EARLY WARNING AND RESPONSE MECHANISMS IN HAITI 16 BOX 1. MUNICIPAL CIVIL PROTECTION COMMITTEES (CCPCs) Source: DPC Municipal Civil Protection Committees ( Comités Communaux de Protection Civile - CCPCs) are responsible for evacuating populations at risk and have been instrumental in saving lives. The role of CCPCs is crucial in the National Risk and Disaster Management System ( Système National de Gestion des Risques et des Désastres – SNGRD) as they are the closest to the population and can go from door-to-door to explain EWS, communicate the alert, and organize their community for evacuation when a disaster is about to happen. CCPCs are composed of volunteers and have successfully engaged in Haiti’s most adverse natural events, including FGHI (Hurricane Fay and Tropical Storms Gustav, Hannah and Ike) in 2008, the January 2010 earthquake, Hurricane Tomas in 2010, and Hurricane Matthew in 2016. In rural areas with little access to traditional communication channels (phones, radios, newspapers), volunteers can reach people that are isolated. Structure of CCPCs CCPCs are directed by the DPC, through the Departmental Civil Protection Committees ( Comités Départementaux de Protection Civile - CDPC). They work in all stages of the SNGRD: preparedness, response, and recovery. There is one CCPC per municipality, totaling 140 CCPCs formed by 3, 100 volunteers serving the entire population of Haiti. CCPCs’ responsibil- ities include disseminating early warnings, evacuating populations to shelters/safe havens, conducting search and rescue, providing first aid in the aftermath of a disaster, and conduct- ing preliminary human losses and material damage evaluations. Once a year, they receive a set of standardized trainings from the DPC on how to perform their duties and cover the following topics: internal management; emergency and response planning, including early warning, evacuation and protection of people; emergency communication; shelter manage- ment; and risk mapping. Beyond the yearly training, CCPCs are active in their municipality all year long, as they organize simulation exercises, conduct information campaigns for the community on preparedness-related aspects, and prepare and update the municipal’s con- tingency plan in support to the municipality. Since CCPCs are dependent on volunteers, their ability to operate is vulnerable to high volunteer turnover. CCPCs volunteers are not entitled to benefits, insurance, or reimbursement for expenses. 17 Assessment of CCPCs CCPC capacity has significantly improved since 2013 thanks to the standardized training pro- vided by DPC, with support from various partners. CCPC capacity is evaluated by DPC and each CCPCs is assigned a score from 1 to 4 (excellent to weak), depending on organizational and functional factors combined in a performance index. The index includes factors related to the institutions represented in the committee, number and nature of specialized training received (internal management; emergency and response planning, including early warn- ing, evacuation, and protection of people; emergency communication; shelter management; and risk mapping); whether they have developed and maintained an emergency and contin- gency plan; number of years the CCPC has been active; the level of implication of the mayor in the committee; and the level of coordination with the Departmental DPC (Map 2). MAP 2. CCPCS’ RANKING (1-4) IN 2016 AND 2018 Source: World Bank analysis based on data from the Disaster Risk Management and Reconstruction Project (P126346). Note. CCPC of Category 1 are deemed “Excellent”, those of Category 2, 3 and 4 are evaluated as “Good”, “Average” and “Weak” respectively. While the decrease in death tolls from hurricanes is promising, Haiti can still improve its resilience to shocks and reduce its exposure. Several structural barriers inhibit the government, local population, and NGOs from making progress on improving resilience to natural disasters. Buildings are poorly constructed and building codes are sparsely enforced, making Hai- ti’s building infrastructure vulnerable to hurricane force winds (World Bank, 2018). The EWS is weak and inconsistently implemented and, according to the Thematic Committee Evacuation and Management of Emergency Shelters ( Comité Thématique Evacuation et Gestion des Abris Provisoires – CTEGAP), the number of emergency evacuation shelters is insufficient to meet the needs of Haiti’s population, particularly in rural areas. 3. Background and Context USING BEHAVIORAL INSIGHTS TO IMPROVE DISASTER PREPAREDNESS, EARLY WARNING AND RESPONSE MECHANISMS IN HAITI 18 A thorough study on shelter coverage for the population in areas highly exposed to flood hazard in the Sud, Grande Anse, and Nippes Departments (Grand Sud) revealed the need for the expansion and im- provement of the shelter network. There are currently 314 emergency shelters in the Grand Sud covering 1.6 million people, in particular 447,589 living in areas highly exposed floods (100-year return period). The multi-cri- teria methodology developed by the DPC to assess shelter needs and prior- itize interventions combines GIS data and field information that considers: (i) flood risk 4 ; (ii) accessibility (3 km radius or about a 45 minute walk); (iii) number of people at risk with regards to existing shelter coverage; and (iv) existing shelter building characteristics, such as size, land availability, and level of destruction post-Matthew; among others. According to study on shelter coverage already conducted by DPC for the Grand Sud, additional shelter capacity is very high for all three departments. Additionally, while structural vulnerability data is not systematically available for shelters, experience from Hurricane Matthew and field evidence collected preliminarily by the DPC and the National Center of Geo-spatial Information ( Centre National de l’information Géo-Spatiale – CNIGS) demon- strated the poor structural and functional state of most shelters, which are not built to withstand a 100-year return period event. At the same time, be- haviors associated with disaster preparedness and evacuation can prevent Haiti from improving its resilience to shocks (World Bank, 2019). As a result, the DPC is working to strengthen and expand its emergency shelter network with the support from the World Bank. 4 A 100-year return period flood event was used as it would be associated with floods caused by a high category hurricane. 19 Photo: UPC-DPC 4.1 Operating model Haiti has a complex structure to organize its National Disaster Risk Management System. Within the Ministry of Interior and Local Authorities, the DPC oversees the design and implementation of the National Disaster Risk Management System ( Système National de Gestion des Risques et des D ésastres – SNGRD) and coordinates the response actions to disasters and risk management (Figure 2). The highest unit within the SNGRD is the Na- tional Committee for Disaster Risk Management ( Comité National de Gestion des Risques et des Désastres – CNGRD), which brings together the highest authorities of the state (including most ministries) and the civil society (Red Cross). The CNGRD oversees the planification, organization and coordina- tion of all actions aimed at the reduction of risks and the response to di- sasters. The Permanent Secretariat is in charge of coordinating all techni- cal actions that are to be implemented during a disaster by the Emergency Operation Center ( Centre d’opérations d’urgence – COU). The SNGRD has thematic committees to establish plans to manage risks and disasters, a consultative committee from civil society, and an international committee composed of international and non-governmental organizations support- ing DRM actions. 4 Disaster Risk Management in Haiti 21 FIGURE 2. SNGRD ORGANIZATIONAL PLAN Risk Reduction in Sectors Sectorial Committes Territorial level risk reduction actions Emergency Prepardness, Response and Recovery Emergency Prepardness, Response and Recovery Committe Emergency Operations Center National (COUN), Department (COUD), Municipal (COUC) and local levels (COUL) National Committee for Disaster Risk Management (CNGRD) Permanent Secretary for DRM Comprised of all sectorial focal points Chaired by Prime Minister. It is comprised by the all the ministers and the Haitian Red Cross President Source: Adapted from the Ministère de L’Interieur et des Collectivités Territoriales, « Plan National de Gestion des Risques et des Désastres 2019-2030 » page 43. Volunteers within CCPCs select and propose to the local mayors which buildings could be used as emergency evacuation shelters. The CCPC negotiates the conditions of use for the structures and ensures that these shelters meet the following criteria set by the DPC under its Shelter Management Guide “G uide de Gestion des Abris d’Evacuation ” ( Secretariat Permanent de Gestion des Risques et des Desastres , 2013). 1. Emergency evacuation shelters must not be in a flood zone, 2. They must be close to vulnerable people and basic social services, including hospitals or health centers, 3. They must be accessible to people with disabilities, and 4. They must be secure. Of the almost 1,500 emergency public shelters in the country identified by the DPC (comprised of schools, municipal buildings, and other public buildings, e.g., courts and city halls, community centers, as well as some private buildings) more than 90 percent are schools. When emergency evac- 4. Disaster Risk Management in Haiti USING BEHAVIORAL INSIGHTS TO IMPROVE DISASTER PREPAREDNESS, EARLY WARNING AND RESPONSE MECHANISMS IN HAITI 22 uation shelters are multifunctional centers such as auditoriums, churches, or gymnasiums, they must be able to stop their activity in time to be used as an evacuation location. 5 4.2 Roles and key actors When a climate event is about to happen, the EWS conveys informa- tion through multiple steps and agents (Figure 3). When a hurricane is approaching, forecasts are interpreted by experts at the Haiti Hydromete- orological Unit (UHM). During an emergency, the Prime Minister activates a state of emergency alert and the Office of the Secretary becomes a Na- tional Emergency Operation Center ( Centre d’Opérations d’Urgence Nation- al – COUN ). COUN notifies departmental delegates, who then activate the Departmental Emergency Operation Center ( Centre d’Opération d’Urgence Départemental – COUD). The COUD notifies the mayors, who then activate the Communal Emergency Operation Center ( Centre d’Opération d’Urgence Communal – COUC), where, in theory, mayors relay this information to CCPC. Finally, the latter disseminates weather related information to the population and evacuate populations to shelters as needed. FIGURE 3. HIGH-LEVEL OVERVIEW OF THE EWS STRUCTURE IN HAITI COUN (activated by Prime Minister) COUC (activated by mayors) COUD (activated by Dep. delegates) CCPC General Population 4.3 Preparedness process Simulation exercises occur at the departmental level but are insuffi- cient to respond to the needs . Every year, one to three departments par- ticipate in a simulation organized by the DPC. Participants spend two to three days simulating the mechanisms that must be implemented during a hurricane and the entire evacuation process, including evacuation man- agement and operations. However, these simulations are insufficient given Haiti’s exposure to natural hazards. There is also no standard debriefing after the simulation. ‘Lessons learned’ are not always documented and im- provements are not suggested for future simulations in a systematic way. 6 5 Basic facility minimum requirements include, a toilet for 50 people, a shower for 30 people, 2m per person (standing), 3m 5 per person (lying down), and sufficient water. 6 Key informant interviews, October 2018. 23 4.4 Communication channels Various communication channels are used to disseminate emergency messages to the general population: 789 CHANNEL PROCESS Radio EWS messages are broadcast in the national radio of Haiti, as well as in commercial and community radios in both FM and AM. SMS AGERCA 7 and CONATEL 8 send SMS to cellphones and through social media. Announcements in public spaces Community leaders are encouraged to disseminate EWS messages in schools, churches, and markets. Megaphones Volunteers drive around the community with megaphones or sirens and communicate information about the storm. Door-to-door visits In high-risk areas 9 , volunteers visit people in their homes and tell them about the approaching storm, instructing them evacuate quickly and prior to the hazard. Flag system In some remote areas where standard forms of communication are not possible, a flag system has been designed to communicate storm warning. Flags of three colors (red, yellow, and green) are raised to indicate the intensity of a storm. Locations that the flags are placed include markets, mountains, and major intersections. 4.5 Evacuation process Ideally, the evacuation process begins long before the storm, hurricane or flood hits the communities. Communities and households are encour- aged to have an evacuation plan so that families know what actions to take, what to prepare, and where to go in case of an event. Once the EWS information is received, people are encouraged to travel to a family shelter (i.e. a more solid house of a friend, neighbor, family member, etc.). In case households do not have a family shelter, they are asked to seek refuge in an emergency evacuation shelter prior to the arrival of the hurricane. People are supposed to stay in the evacuation location for no more than 72 hours after the storm passes at which time they should return to their homes. 7 AGERCA is the institution that represents the private sector, and technically also civil society in the secretariat that manages disasters in the country (SNGRD). 8 CONATEL is the state institution that is responsible for regulatory issues around communication in the country. 9 There is not a precise definition of high-risk areas but are rather defined as areas susceptible to at least one or more vulnerabilities. 4. Disaster Risk Management in Haiti USING BEHAVIORAL INSIGHTS TO IMPROVE DISASTER PREPAREDNESS, EARLY WARNING AND RESPONSE MECHANISMS IN HAITI 24 Photo: UNDP Research has shown that many disaster-related fatalities are preventable if people evacuate to a safer place in a timely fashion. The main problem we are trying to ad- dress in Haiti is that often people do not evacuate in a timely manner when there is an approaching hurri- cane . Five main obstacles and related barriers limit peo- ple from early evacuation. We analyze behavioral and structural barriers given that both can be tackled with a behavioral approach. Findings in this section have also been identified in behavioral science research looking at risk perception (Milch, Broad, Meyer, & Robert, 2018) and (Kunreuther & Meyer, 2017). 5.1 Obstacle 1: EWS Messages Do Not Arrive On paper, Haiti’s communication plan works. But, in practice, no reliable and effective methods communicate information about an approaching storm to the entire population. As one participant mentioned “…I did not know Hurricane Matthew was coming….” (FGD Woman, Paillant ). The team found inefficiencies at each level of the communication pipeline, especially for those in rural regions who are generally more vulnerable. 5 Key Insights on Barriers to Evacuation USING BEHAVIORAL INSIGHTS TO IMPROVE DISASTER PREPAREDNESS, EARLY WARNING AND RESPONSE MECHANISMS IN HAITI 26 Barrier 1. Limited resources and funding at all levels • The DPC has limited operational budget, which limits its ability to fulfill roles such as training, comprehensively coordinating, com- municating, accompanying, responding, and assisting in recovery. • Volunteers have limited access to the necessary equipment or re- sources to perform their role. Often, they borrow private vehicles or pay out-of-pocket for transportation costs. Without regular access to motorcycles, they cannot quickly and effectively access areas with more challenging terrain. • The population lacks functional devices used to circulate EWS alert messages. Most people have access to their own or a peer’s phone or radio, but it is not guaranteed that the network or the equipment functions. Many do not have the funds to purchase batteries or phone minutes and, as a volunteer explained, “ We can do a lot with phones, but frequently during a catastrophe the networks of Natcom and Digicel stop working. If there was another way of communicat- ing, perhaps by satellite phone, it would be very helpful” (FGD, CCPC, Les Cayes). Barrier 2. Lack of standardization on how EWS information is transmitted • High level communication does not always happen efficiently be- tween sectoral ministries, affecting coordination on all levels. • There may be confusion about which CCPC is responsible for which areas. As a man explained: “Every time a hurricane may come, Civil Protection passes on a motorcycle and just arrives at the intersection. When they arrive at that last intersection, they finish speaking and sounding the siren and then do not go anywhere else” (FGD, Men, Paillant). • A lack of standardization makes it is unclear what areas have re- ceived warnings and which still need to be informed. Prior to Hur- ricane Matthew, some households were visited multiple times by volunteers while others received no visits at all. • The means in which information is disseminated varies by area. In many instances, it is transmitted by international NGOs, bypassing the DPC. Additionally, NGOs are not always present and often leave. • Information circulated through multiple communication channels, especially in the urban areas, can become confusing. 27 Barrier 3. Inexperienced messengers • CCPCs are composed of unpaid volunteers. This causes high turn- over, that needs to be trained on a regular basis. New members have no experience when they start. Mayors may not always under- stand their role in disaster management. They lack training and an understanding of the role of civil protection. This poses a challenge as it is the mayor’s responsibility to activate COUC and work closely with the CCPC. As the General Director of Les Cayes said: “ Me per- sonally, this is not a role I can play. I am only a part of the Mayor/ Municipal office in Les Cayes. Let me take a look because I can’t say personally but it is civil protection who is responsible” (Interview, Mayor, Les Cayes ). 5.2 Obstacle 2: People Do Not Understand EWS Messages When the information arrives, messages are not always adapted in ways the average person in Haiti understands and are rarely validated by trustworthy sources. As a result, people do not always know where to go or what to do or feel urgency in evacuating. Barrier 4. Unclear messaging • Messages are not always adapted to local educational and literacy levels. For example, the Saffir-Simpson Hurricane Wind Scale rang- ing from category 1-5 (Appendix 2) is not informative to the gener- al population in Haiti. As the General Director posed: “It should be easier for people to understand the different categories. It should be presented in a way that the people understand, in a language that they understand. Perhaps, the words heavier, lighter, more violent” (Interview, Mayor, Les Cayes). • Although some seem to be familiar with the flag system, the pop- ulation is not fully aware of what the three colors indicate or have not they seen flags even if they know where they are supposedly located. Barrier 5. Distrust in the messenger • There is a general distrust in the State and its representatives. Past neglect leads the population to perceive them as prioritizing their own interests and pockets. As a delegate explained: “If you wait for the State, the State will never come“ (Interview, Delegate, Les Cayes) . • Though volunteers within CCPC are members of the community, they may be seen as representatives of the State or NGOs. The pop- 5. Key Insights on Barriers to Evacuation USING BEHAVIORAL INSIGHTS TO IMPROVE DISASTER PREPAREDNESS, EARLY WARNING AND RESPONSE MECHANISMS IN HAITI 28 ulation does not fully understand the term volunteer. As one volun- teer explained: “People may think you are not speaking the truth and it is a false alarm.” (FGD, CCPC, Les Cayes). • The likelihood and impact of a disaster cannot be predicted with certainty due to changing storm trajectories. As a result, false alarms may be given to people that end up not experiencing the damage. This carries the risk of the “cry wolf” effect (Kunreuther & Meyer, 2017), negatively affecting the credibility of the messenger. Barrier 6. Lack of prior hurricane experience or simulations • Although Hurricane Matthew is now etched in people’s memories, especially in the Southwestern Peninsula, people had little expe- rience with storms prior to its landfall. As a teacher mentioned: “There were many who understood the warnings, but others…let me tell you, it is a people who need to see to believe. For me, I had experienced Hurricane Allen and did not want to relive that experi- ence” (Interview, Teacher, Les Cayes). Some people suffer amnesia, a tendency to forget too quickly the lessons of past disasters (Kun- reuther & Meyer, 2017). • Population may not give enough importance to preparedness ac- tivities such as yearly simulation exercises due to present bias: pre- paredness activities have an immediate cost with benefits unclear or in the future. • Although yearly simulation exercises are conducted, these simu- lations are insufficient to fully train all actors involved in EWS and the entire population, exacerbated by the lack of disaster prepared- ness activities in schools. Barrier 7. Little, wrong, or misappropriate information about best behaviors in a disaster scenario • Messaging generally informed people to move animals away from ravines, without explanation as to why and what the consequences might be if animals were left by the ravine. • Messaging said to stay away from windows while most of the popu- lation does not have glass windows. • People were observed taking their animals from the mountains and tying them by rivers, despite the risk of rising waters or swelling rivers. • The population does not know that they live in high-risk zones or what defines a high-risk zone. The DPC does not have a precise defi- nition, which may contribute to further confusion. 29 5.3 Obstacle 3: People Struggle to Internalize Risk Even when information arrives and is understood, people do not realize or accept their actual level of risk. Many negate the information received (state of denial) just because they do not want to believe it. Others feel there is little to nothing they can do to prepare to save themselves (fatalistic belief) so prefer to stay with their possessions instead of saving their lives. Barrier 8. Underestimation of risk level • State of denial: People tend to underestimate the probability of negative events due to lack of historic references or experiences. Religion is also very present in Haitian society. Culturally speaking, many neither believe something until they have seen it nor think God would let such a disaster befall them. From the experience of a woman: “I heard about Matthew on the radio and when I told people, they told me I was lying because God would not let such a thing hap- pen to us and everyone said it was in God’s hands. They said let’s pray and they started to pray…” (FGD, Women, Les Cayes). • Fatalistic belief: There is frequent resignation to the possible shocks related to natural disasters due to the everyday struggles that most Haitians face. Many believe God will protect them which may stem from both faith and the socioeconomic conditions giving them little option but to put faith in God. As a woman put it: “You make the preparations that you are able to if you have the means. If you do not have the means, you are the mercy of God’s will because if you do not have the means, you may not stay alive” (FGD, women, Paillant). Barrier 9. Temporal and spatial myopia • When faced with an evacuation decision, the population focuses on possessions rather than their lives, experiencing myopia - the tendency to make decisions that prioritize short-term gains when appraising immediate costs at the expense of long-term ones. Vol- unteers noted that they had to remind people that the most im- portant thing was to save their life, and that everything else comes after. This is most likely due to a combination of socioeconomic conditions and perceived threat the hazard poses. • Emergency evacuation shelters do not allow livestock, requiring that people leave them behind. Livestock (goats, donkeys, horses) is the households’ main productive and financial asset. People cit- ed fearing looters immediately after the hurricane, especially if they 5. Key Insights on Barriers to Evacuation USING BEHAVIORAL INSIGHTS TO IMPROVE DISASTER PREPAREDNESS, EARLY WARNING AND RESPONSE MECHANISMS IN HAITI 30 did not have means to secure their house (i.e. sturdy door or strong lock). Little to no police patrol the streets that might deter thieves. Therefore, the preference is to remain in the familiarity of and pro- tect one’s home. 5.4 Obstacle 4: People Lack Access to Resources and Shelters to Evacuate Even when people internalize risk levels, structural challenges prevent them from evacuating. Many people do not have a shelter close to their homes, making it impossible to heed warnings. In rural areas especially, there is a widespread lack of emergency evacuation shelters that can withstand hur- ricane force winds. People may lack access to transportation or simply can- not gain access to emergency evacuation shelters. Barrier 10. Lack of resources needed to evacuate • Quality of and limited access to vehicles and roads limit one’s abil- ity to evacuate. • The locations of vulnerable people (people with reduced mobility, children, elderly) are not mapped and therefore unknown in ad- vance to volunteers. Barrier 11. Lack of access to shelters • Some shelters are closed before a disaster occurs. Because most shelters are schools, it is the responsibility of the Ministry of Educa- tion in collaboration with Ministry of Interior and Local Authorities to ensure schools are opened once the alert is given by the DPC. In some instances, people have been forced to search for the person holding the keys to the school. 5.5 Obstacle 5: People Prefer Not to Evacuate Given that people are not forced to evacuate, many prefer staying in their homes with hopes they can wait out the storm, thinking the shelter might be more harmful than staying (omission bias). Participants mentioned pre- ferring to seek shelter in a friends’ or family member’s house instead of an emergency evacuation shelter. 31 Barrier 12. Unsatisfactory experiences in emergency evacuation shelters • Unmanaged . CCPC volunteers or members of civil society who have had adequate training are meant to manage and set the rules at a shelter. But, rules and codes of conduct are not guaranteed. Their limited presence and oversight may create spaces that are unsafe, frequently overcrowded, and with hygiene challenges. • Unsafe . The DPC has established clear guidance to guarantee safe- ty in shelters ( Secretariat Permanent de Gestion des Risques et des Desastres , 2013), but a lack of resources means oversight cannot al- ways be guaranteed. Emergency evacuation shelters have no sepa- rate areas for women and children. Young men often drink alcohol, smoke, and listen to loud music in the shelter. There is no privacy and people are responsible for their own security and the safety of their belongings. Volunteers cited cases of sexual aggression, es- pecially in urban areas: “There are sometimes incidences of sexual aggression and rape that go unreported in the shelters” (FGD, CCPC, Les Cayes) . Many go unreported due to stigma, and because the perpetrator can sometimes be someone the victim knows. In other cases, men benefited from humanitarian aid distributions and used the aid to coerce (transactional sex). There are issues of theft. • Lacking resources. Many people do not see why they should go to emergency evacuation shelters given the lack of resources such as food, water, first aid supplies or care, beds, and sheets. There is no checklist of what should be stored in a shelter in preparation for a disaster. Approximately 90 percent of shelters do not have appro- priate WASH (water, and sanitation/hygiene) infrastructure includ- ing toilets, latrines and running water. Most emergency evacuation shelters do not have a power source or lighting, flashlights, clean- ing or cooking supplies. When supply distributions do happen, they seem to be infrequent and subpar. Barrier 13. Shelters do not abide by building codes and standards • Emergency evacuation shelters in the municipalities most heavily affected by Hurricane Ma