Risk Perception in a Multi-Hazard Environment

World Development Vol. xx, pp. xxx–xxx, 2017 0305-750X/Ó 2017 Elsevier Ltd. All rights reserved. www.elsevier.com/locate/worlddev

http://dx.doi.org/10.1016/j.worlddev.2017.04.002

Risk Perception in a Multi-Hazard Environment KIRA A. SULLIVAN-WILEY a,b and ANNE G. SHORT GIANOTTI c,* a Brown University, USA b University of Antwerp, Belgium c Boston University, USA Summary. — Environmental disasters cause enormous losses of life and property every year, a threat that is recognized and addressed in both the Sendai Framework for Disaster Risk Reduction and the 2015 Sustainable Development Goals. Organizations from both the risk reduction and development fields are working to design programs that build risk understanding and risk perception to encourage protective action in communities that are often at risk from multiple, overlapping threats. We know little, however, about how individuals perceive and prioritize multiple hazards at once and how this relates to their adoption of protective action strategies in the developing world. Our work addresses environmental hazard risk perception in a multi-hazard context in eastern Uganda, with particular attention paid to the role that risk reduction and development organizations (RDOs) play in shaping risk perceptions, as well as their potential to influence protective action. To better understand risk prioritization, we used survey data from farming households to generate four indices reflecting several components of risk perception and to predict holistic risk perception through multivariate regression analysis. Our study finds that the factors shaping smallholder risk perception vary among hazards within the study population and that characteristics of both hazards and individuals are important. The regression analysis also reveals a surprising relationship between risk perception, self-efficacy, and protective action. Our findings suggest that risk reduction and development programs can play an important role in affecting both risk perception and the capacity of smallholders to respond to environmental threats. Our work adds to the growing body of literature on how people perceive and respond to risk in a multi-hazard environment, a context increasingly common in a changing world. Improved understanding of how RDO programs in the developing world are engaging with and influencing risk mitigation in the multi-hazard environments is fundamental for reducing vulnerability. Ó 2017 Elsevier Ltd. All rights reserved. Key words — risk perception, development organizations, DRR, multi-hazard, Africa, Uganda

1. INTRODUCTION

inconsistencies in perceived responsibility for protection and trust in protective agencies, and the perception of limited self-efficacy (i.e., the capacity to undertake protective actions) have each been found to act as intermediaries to prevent the understanding of risk from translating into action (Wachinger, Renn, Begg, & Kuhlicke, 2013). The decision to take, or not take, action can in turn influence risk perception (Brewer, Weinstein, Cuite, & Herrington, 2004). These challenges in translating risk perception to action may be especially critical in multi-hazard environments where people are vulnerable to multiple, overlapping threats, with which they have limited resources to cope. Examining risk perception in a multi-hazard environment, and the role of RDOs in shaping those risk perceptions, is important to better reflect the reality of vulnerable individuals and to allow us to tease out the influence of particular hazard characteristics versus individual characteristics on risk perception. Yet we know little about how individuals perceive and prioritize multiple hazards at once and how this relates to their use of the protective actions that are frequently particular to an individual threat (Doss, McPeak, & Barrett, 2008). While a large body of work has examined who perceives risk and why, this work has focused on single hazards in isolation

Globally, environmental disasters result in the death of tens to hundreds of thousands of people (IFRC, 2014) and the loss of US$250 billion to US$300 billion every year (UNISDR, 2015). In addition to the threat of an individual hazard event, there is increasing awareness that hazards are often found in combination with other threats, both environmental and social and that these threats can interact to exacerbate each other in a multi-hazard landscape (Cutter, Mitchell, & Scott, 2000; O’Brien et al., 2012; UNISDR, 2015). High population growth rates exacerbate threats in multi-hazard environments (Huppert & Sparks, 2006) and the threat of climate change, an additional uncertainty overlaying existing vulnerabilities, further complicates the meteorological component of hazards (IPCC, 2014). The international community has recognized the interconnectedness of these threats in the adoption of the Sustainable Development Goals and the Sendai Framework for Disaster Risk Reduction in 2015. Both risk reduction and development organizations (hereafter referred to collectively as RDOs) are making substantial efforts to encourage vulnerable populations to adopt protective actions, designing programs that aim to build risk understanding and risk perception (Shaw & Izumi, 2014; Thomalla, Downing, SpangerSiegfried, Han, & Rockstro¨m, 2006). In order to take protective actions against a hazard, people must have some understanding of the risk associated with that hazard and the capacity to act on their concern (Lindell & Perry, 2012). While higher levels of risk perception would be expected to lead to higher rates of protective action, this relationship is not always straightforward. In a phenomenon termed the ‘‘risk perception paradox”, elevated risk perception is not always linked to protective action. A lack of motivation,

* This research would not have been possible without the people who shared their time and experiences. We are grateful to the interview and survey respondents who informed this research. The authors gratefully acknowledge the Boston University Moorman-Simon Civic Fellowship and Graduate Research Abroad Fellowship for providing financial support for this research. They would also like to acknowledge the logistical assistance of the field and translation teams in both the Bududa and Manafwa districts of Uganda. Final revision accepted: April 3, 2017. 1

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

2

WORLD DEVELOPMENT

(e.g., Lo´pez-Marrero, 2010 flooding risk perception in Puerto Rico; Nathan, 2008 landslides in Bolivia; Gaillard, 2008 volcanoes in the Philippines; review articles such as Gallina, Torresan, Critto, Sperotto, Glade, & Marcomini, 2016 similarly highlight studies focused on single hazards). Further, none of these studies directly addresses the role of RDOs in affecting risk perceptions in a multi-hazards context. We address this gap in knowledge through a study of environmental hazard risk perception in a multi-hazard context in the Bugisu region of eastern Uganda. We pay specific attention to the role of risk and development oriented organizations in shaping risk perceptions and their potential to influence protective action. Understanding the factors that shape risk perception and the implications for those on changing action in a multi-hazard environment is important to inform RDOs in their work to reduce risks to the most vulnerable populations. This paper begins with an overview of the literature on risk perception of environmental hazards and an introduction to our study area of the Bugisu region of eastern Uganda. We then present results of hazard ranking and regression analyses for risk perception that show a disconnect between RDO and local prioritizations and perceptions, the difference in factors shaping perception of different hazards, and the role that RDOs may play in shaping perception. Finally, we discuss the theoretical and policy implications of our study, as well as areas of future research based on this work. 2. BACKGROUND & THEORETICAL FRAMEWORK Risk perception is a key component in encouraging protective action in the context of natural hazards (Lindell & Perry, 2012; Wachinger et al., 2013). Risk perception contrasts with ‘‘real risk”, or the statistical likelihood of fatality from the hazard, through its reference to a person or population’s interpretation of the hazard and its risk (Sjo¨berg, 2000). There are three issues implicit in perceived, as opposed to real, risk. First is that, while distinct from real risk, the notion of probability still exists in perceived risk, but instead of reflecting a calculated statistical probability, perceived risk reflects perceived likelihood, which frequently differs from statistical probability in meaningful ways through biases such as the availability heuristic (Siegrist & Gutscher, 2006; Tversky & Kahneman, 1974). Secondly, perceived risk comprises uncertainty in event outcomes and the severity of those outcomes for the individual or group interpreting the risk; even the same physical outcome of a hazard can represent different danger to different people depending on their preferences and coping capacities. Finally, there is the social construction of risk that relates to the level of risk society is willing to accept in exchange for the social benefits associated with its cause, a relationship that is influenced by perceptions of to whom the responsibility for risk mitigation falls (Bronfman, Lo´pez Va´zquez, & Dorantes, 2009; Kasperson et al., 1988). Much early work in the field focused on assessing the differences between perceived and real risk (e.g., Lichtenstein, Slovic, Fischhoff, Layman, & Combs, 1978; Slovic, Fischhoff, & Lichtenstein, 1979, 1980), while later work began investigating the implications of these differences for risk management and risk communication (e.g., Boholm, 1998; Renn, 1999; Wachinger et al., 2013). A large and mature body of research investigates how people perceive risks associated with technological hazards (e.g., nuclear power, genetically modified organisms). This body of work shows that risk perception varies with respect to the characteristics of the individual perceiver as well as the

characteristics of the hazard itself (Fischhoff, Slovic, Lichtenstein, Read, & Combs, 1978; Slovic, 1986; Slovic et al., 1979; Wachinger et al., 2013). Early research identified differences in how expert and non-expert communities perceive risk. While experts generally equate risk with fatality frequency (annual death rates associated with a given hazard), non-experts include factors like catastrophic potential and sensationalism into their risk calculus (Lichtenstein et al., 1978; Slovic et al., 1979). In addition, non-experts tend to rate concern about risks more highly when the hazard is uncontrollable, catastrophic, involuntary, not equitable in its impacts, and not well-understood (Boholm, 1998; Slovic, 1986). Like technological hazards, the most essential components of environmental hazard risk perception are generally considered to be the perceived probability (likelihood) and the severity of the consequences of the hazard (Lindell & Perry, 2012). These, however, are insufficient to account for variability in risk perceptions (Table 1). The characteristics of the individual also affect risk perception. Characteristics related to social vulnerability are associated with higher risk perceptions of hazards, a relationship likely to reflect individual self-efficacy, or one’s perceived ability to affect change (mitigate risk) through protective action (Bandura, 1995; Bickerstaff, 2004; Martin, Bender, & Raish, 2007). Gender, age, and educational attainment are often (though not consistently) found to be mediating factors in risk perception. Women have been found to perceive greater risk than do men, older adults to perceive greater risk than young, and less educated to perceive greater risk than more educated (Flynn, Slovic, & Mertz, 1994; Gyekye & Salminen, 2009; Mayhorn, 2005; Siegrist, 2000; Terpstra & Lindell, 2013; Wachinger et al., 2013). Those living in poverty (Cutter, 1981; Nyland, 1993; Sjo¨berg, Kolarova, Rucai, Bernstro¨m, & Flygelholm, 1996), those with children in the household (Turner, Nigg, & Paz, 1986); people who are divorced or unemployed (Boholm, 1998), and other characteristics like cultural identity (Rohrmann, 1994) have also been shown to be associated with elevated risk perception. Other studies of perceptions of individual environmental risks, however, have found weak or non-existent trends with respect to some or all of these socioeconomic characteristics (Burningham, Fielding, & Thrush, 2008; Plapp & Werner, 2006). Familiarity and experience can also affect perception of risk. In the context of environmental hazards, direct personal experience has consistently been shown to be positively associated with risk perception (Grothmann & Reusswig, 2006; Heitz, Spaeter, Auzet, & Glatron, 2009; Miceli, Sotgiu, & Settanni, 2008; Plapp & Werner, 2006; Siegrist & Gutscher, 2006; Terpstra, 2011). The recency, frequency, and severity of this experience can affect the strength of its relationship to risk perception (Lindell & Perry, 2012). Those who have experienced mild forms of a hazard, for example, tend to underestimate subsequent danger, with an attitude that Mileti and O’Brien (1992) describe as ‘‘normalization bias”, whereby people interpret the mild impacts of the early experience as the norm and believe that future severe impacts can also be avoided. This can be seen in the example of non-experts consideration of familiar actions such as driving in a motor vehicle less risky than less familiar actions that are statistically less likely to result in fatalities (Slovic et al., 1979). Baan and Klijn (2004) found that those most experienced with floods were among those least concerned by them, but in this case the effect was mediated through a sense of preparation on the part of the perceiver. Risk perception is also influenced by communication about risks from external expert sources in complex ways (Fischhoff,

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

RISK PERCEPTION IN A MULTI-HAZARD ENVIRONMENT

3

Table 1. Factors of importance in shaping risk perception Factor categories

Factors

Selected references

Hazard characteristics

Likelihood of occurrence, severity of consequences (catastrophic potential), sensationalism, controllability, voluntariness, equitability of impacts, scientific understanding Self-efficacy characteristics (gender, age, education, children, employment, cultural identity, personal experience)

Bickerstaff (2004), Boholm (1998), Lichtenstein et al. (1978), Lindell and Perry (2012), Slovic et al. (1979), Slovic (1986) Grothmann and Reusswig (2006), Lindell and Perry (2012), Siegrist and Gutscher (2006), Sjo¨berg et al. (1996), Wachinger et al. (2013) Bickerstaff (2004), Earle (2010), Fischhoff (1995), Peters et al. (1997), Renn and Levine (1991), Siegrist and Cvetkovich (2000), Wachinger et al. (2013)

Individual characteristics

Trust in communicating institutions

Similarity heuristic, knowledge and expertise, openness and honesty, and concern and care; important under conditions of uncertainty

Figure 1. Uganda shaded elevation map with Bududa and Manafwa districts highlighted and surveyed villages shown.

1995; Renn, 2009). Much of the work on risk communication emphasizes that, beyond accuracy and relevance of message content, the trust between the non-expert and the expert is important (Fischhoff, 1995; Renn & Levine, 1991; Wachinger et al., 2013). Paton (2008) notes that trust is an especially important component of communication when people are dealing with decisions under conditions of uncertainty, a condition satisfied in all cases of hazard risk. In these cases, trust is used as a proxy in place of complete information, allowing a simplified message to be believed and taken up by the individual without all underlying complexity needing to be understood. In addition to reducing uncertainty, external experts can provide information to people who lack direct experience with a particular hazard. Through risk communication, the expert provides indirect experience to the non-expert, but only when trust between the parties is present (Earle, 2010; Siegrist & Cvetkovich, 2000; Wachinger et al., 2013). Trust is influenced by characteristics of the expert, as perceived by the non-expert. A large body of work shows trust as determined by knowledge and expertise, openness and honesty, and concern and care (Fisher, 2013; Kasperson, 1986; Peters,

Covello, & McCallum, 1997; Renn & Levine, 1991). However, this list of characteristics has received some criticism (Cvetkovich & Lofstedt, 2013). Earle (2010) suggests that trust is primarily influenced by perceived morality, interpreted through the ‘‘similarity heuristic”, which reflects shared values and priorities of the expert and non-expert. This trust can be undermined by perceived differences in values and priorities, which trigger the non-expert to perceive expert bias. In some cases, value congruence is more important than even transparency and other factors such as competence in generating trust between parties (Earle & Siegrist, 2006; Pirson & Malhotra, 2008). While this body of research highlights the many factors that influence individual risk perception and the role that trusted organizations can play (Table 1), we lack an understanding of how these factors interact in a multi-hazard context. Risk perception studies that do evaluate the perception of multiple risks are rare and focus on industrialized countries (Lindell & Hwang, 2008; Perry & Lindell, 2008). Lin, Shaw, and Ho (2008) is a notable exception in their focus on landslide and flood risk in Taiwan. They found that while perceived

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

4

WORLD DEVELOPMENT

likelihood and negative impacts of floods were higher than those of landslides, the greater dread of landslides was associated with greater willingness to invest in mitigation activities (Lin et al., 2008). Their work highlights the differences in risk perception and risk response in a single population, and demonstrates the importance of understanding how people perceive one hazard in the context of others to which they are exposed. Given the frequent intersection of multiple hazards with social components of vulnerability in the developing context, a better understanding of how people perceive multiple threats may help to inform our understanding of how they respond to and cope with these threats. 3. STUDY REGION AND METHODS

ards directly and/or to engage in other development activities that can indirectly reduce hazard risk. Some risk reduction agencies, like the Red Cross Society and the Ugandan Department of Natural Resources and the Environment, address environmental hazards through training programs to reduce risk levels through disaster preparedness and improved land management training. Others, like Technoserve, Send-ACow, and the Uganda national agricultural advisory service (NAADS) address risk indirectly through livelihoods development and income generation programs. The confluence of multiple hazards (Shi & Kasperson, 2015), social vulnerability (UNDP, 2013), and the presence of a number of risk reduction and development organizations makes the Bugisu region a compelling site to investigate the factors at play in shaping risk perception in a multi-hazard environment.

(a) Study area

(b) Methods

Our study area includes 10 small communities in the Bududa and Manafwa Districts, Eastern Province, Uganda (Figure 1). These districts border the main commercial and transport hub of Mbale, which connects to other major cities in Uganda by tarmac road. People in the region are ethnically Bagisu and speak Lugisu dialects and, to varying extents, English. Most have lived in the Mount Elgon region for generations and have close cultural ties to the region. The population of Bududa is approximately 210,000 (UBOS, 2016). Though Manafwa’s population is 68% larger, it is slightly less rural (UBOS, 2016). The population of Bududa is also growing more quickly, and this rate has increased since 1991, while the growth rate in Manafwa is lower and is decreasing (Table 2). Vulnerability in the region stems from poverty and population pressures as well as multiple environmental hazards. The region is characterized by steep slopes, river valleys, and is dominated by loose, volcanic soils. Rainfall is distributed across two rainy seasons with average annual rainfalls of 1000–1800 mm (Manafwa) and 1400–2200 mm (Bududa) (NEMA, 2010). Available land is scarce and land scarcity has pushed cultivation onto the steepest slopes of Mount Elgon, the extinct volcano that defines the region. Together these factors result in a landscape at risk for chronic soil loss due to erosion, severe landslides, mudslides, and flooding along the Manafwa River, which runs through both districts. In 2010, the leading disaster relief organization in the region, the Uganda Red Cross Society, responded to 11 landslides in the Bugisu region and in a period of about 15 years 98 landslides occurred in Bududa alone (Claessens, Knapen, Kitutu, Poesen, & Deckers, 2007), with one catastrophic landslide in 2010 killing well over 300 people and displacing many more (URCS, 2010). Landslides and related hazards have increased in recent years (Mugagga, Kakembo, & Buyinza, 2012) and represent only a few of the hazards facing the study region. Other hazards include drought, deforestation, and the proliferation of pests and diseases. Numerous non-governmental, private sector, and government organizations are active in these districts. These RDOs aim to reduce smallholder vulnerability to environmental haz-

To investigate the role of RDO programs and other factors on risk perception we used a mixed methods approach that combines data collected through semi-structured interviews with RDO personnel, surveys of farming households across ten villages in the Bugisu region, and focus groups with individuals from the study villages. (i) Interviews with RDO development agents We conducted 40 semi-structured interviews with development agents at RDOs to identify the range of RDO programs involved in risk communication and to characterize the ways in which they perceive and communicate those risks. Organizations and agents within those organizations were identified using snowball sampling, wherein each interviewed agent was asked to identify other organizations operating in the Bugisu region in the field of agricultural development, disaster relief, or risk reduction that: (1) have a consistent local presence in Bududa or Manafwa districts, (2) implement programs intended to reduce farmer risk from any hazard, and (3) target either land management or environmental education as a key node of program design. Once an organization was identified, an interview was set up with either a field worker (for organizations with staff large enough to support separate coordinator and field positions) or a coordinator or chairperson for organizations with more limited staff where these agents served multiple roles. We focused interview efforts on individuals who were familiar with the actual implementation of organization programs in the field, at the point of contact with farmers. Interviews took place over the course of two field seasons in 2012 and 2013 and continued until no new organizations were named, resulting in 40 interviews and coverage of 12 nongovernment non-profit organizations, 6 community-based farmer’s and women’s groups, and 7 government offices/officers (Appenidx 1) The interviews were semi-structured and elicited information on:
The agent’s perception of the principal risks and hazards faced by the local population, inclusive of any challenge that threatened well-being in any way,

Table 2. Demographic and socioeconomic information for Bududa and Manafwa Districts, Uganda (UBOS, 2016)

Bududa Manafwa

Population (2014)

Average household size

Population density (per sq km)

Literacy rate (18 + years) (%)

% rural households

Population growth rate (2002–14) (comparison to 1991–2002)

210, 173 353,825

5.7 4.8

662 661

70.2 66.0

93.4 85.5

4.5 (higher) 2.5 (lower)

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

RISK PERCEPTION IN A MULTI-HAZARD ENVIRONMENT


The benefits and rewards that farmers accrue from their land and environment, and for which they make land management decisions,
The design process, implementation characteristics, beneficiary selection methods, and evaluation techniques for programs of the organization, including intended targets of program outreach, and
The relationships of that organization with others in the field. Interview data were transcribed and coded using NVivo 10 qualitative analysis software (QSR International Pty Ltd. Version 10, 2012). These data were used to determine the perception of risk and appropriate management within the RDO community and to understand RDO practices in this context. Information from these interviews also informed the development of the household survey, especially with reference to the hazards and management strategies queried. To select study villages for the household survey (Section 3 (ii)), we used the interviews to identify villages that had recently been targeted by program implementation by relevant RDO organizations. Since we were interested in the role of outside organizations in shaping local perceptions, we eliminated the community-based organizations from further quantitative analysis, and identified seven organizations that represented a cross-section of active RDOs in the region. The focal organizations identified were the Uganda Red Cross Society (the only strictly risk reduction organization), the Bududa District Environment Office (the only government organization that was running a single-village agricultural development and risk reduction project), and five nongovernment organizations (Spark Microgrants, Send-A-Cow, Coffee-A-Cup, Mbale Farmers Association, and Technoserve). We identified a single village that had recently been targeted for program implementation by each of above organizations with the exception of the Red Cross, for which two villages were identified, one in Bududa and another in Manafwa.

5

villages—five in Bududa and five in Manafwa. Eight of the villages (the ‘‘program” villages) were selected because they were identified as the targets of programs administered by one of the regional development and disaster risk reduction organizations (RDOs). The remaining two ‘‘control” villages (one in each district) were selected because they have similar social and environmental conditions but have not been a part of any development or disaster risk reduction program beyond outreach by NAADS, the government extension program that operates in all villages across Uganda. In each village, surveyed households were randomly selected by copying every fifth name from a comprehensive list of headof-household names obtained from the local elected chairperson until the correct number of households was identified. This method is the standard used by the Uganda Red Cross Society in village-level vulnerability and capacity surveys. The average number of households in each of nine villages was 80 (range 41–118), with one outlier village containing 358 households. In control villages and villages where organizations aim to reach the entire village, 55 households were selected in this manner to be approached for participation in the survey. In villages in which organizations target only a sub-population of the village, surveyed households included 10 non-random households identified as beneficiaries by development agents and a random selection of 45 additional households to ensure that the sample included at least 10 beneficiary households. Only heads-of-households or their spouses responded to the survey. Surveyed households in Bududa and Manafwa (n = 426) averaged 0.7 ha with a mean household size of just over 6 people, with 15% of households reporting household sizes of 10 or more (see Table 3 for more detailed information). Surveyed households were an average of 3.5 km from the nearest market and the mean annual incomes of surveyed households was 650 USD (values reported in Ugandan shillings and converted to USD based on the exchange rate of mid-September 2013). We surveyed an average of 54% of households in each village. Of these, 75% reported male heads-of-household, though only 49% of survey respondents were male.

(ii) Household survey A household survey collected data on individual risk perceptions, the use of risk management and other best management strategies, engagement with RDO programs, and a variety of demographic and socioeconomic variables. The survey was pre-tested with three smallholder farmers in each district, with minor modifications made as necessary to survey questions to ensure clarity, and administered in ten

(iii) Focus groups To gain insight into risk perception from the level of the community, one focus group was held in each study village. Focus group participants included a randomly selected subset of survey respondents who had indicated a willingness to participate in additional discussion beyond the survey. 79% of

Table 3. Demographic and economic means and standard deviations of the surveyed population of ten target villages in each of the two districts Village

RDO active in village

HH in sample (%)

Bududa Bunasaba Buwabusera Bushibuya Bunamutunyi Bunamalishe

Environment Office Mbale Farmers Assn Coffee A Cup Coop. (control) Red Cross I

51 83 100 40 88

0.73 0.48 0.86 0.63 0.46

± ± ± ± ±

0.63 0.37 0.68 0.63 0.23

6.4 5.7 6.7 4.8 5.3

± ± ± ± ±

3.1 2.7 3.3 2.8 2.5

354 ± 501 528 ± 1315 381 ± 436 233 ± 345 301 ± 343

3.74 2.32 5.40 3.57 9.26

± ± ± ± ±

1.93 0.81 4.72 1.58 3.17

91.6 72.7 75.4 67.4 80

100 44.9 87.3 51.1 75

Manafwa Shiruku Bumwangu Buwangota Bunokomola Silumbusa

Send-A-Cow Technoserve Spark Microgrants Red Cross II (control)

34 11 41 43 50

0.60 0.84 1.00 0.68 0.63

± ± ± ± ±

0.53 0.74 0.94 0.48 0.33

7.4 6.4 6.4 6.1 6.3

± ± ± ± ±

3.7 4.0 2.7 3.4 2.2

709 ± 933 774 ± 846 1513 ± 1544 1019 ± 893 949 ± 874

0.56 1.06 3.37 3.42 1.50

± ± ± ± ±

0.30 0.61 1.45 2.31 0.68

54.8 75 81.4 64.1 81.6

53.1 35 95.3 78 31.6

n/a

54

0.70 ± 0.62 6.1 ± 3.1

3.50 ± 3.26

74.6

65.9

Total *

Farm size (ha)

HH size (total)

Annual income Distance to HOH male HH with (USD)* nearest market (km) (%) uphill land (%)

652 ± 959

Exchange rate of 1 USD = 2588 Ugandan shillings based on daily rate from mid-September 2013.

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

6

WORLD DEVELOPMENT

surveyed farmers volunteered to participate in further research, and t-tests reveal that the 21% that did not volunteer had significantly lower incomes (p = 0.02), smaller farms (p < 0.001), and smaller household sizes with fewer children (p < 0.001). They did not differ in their gender, nor their general risk profiles, though volunteers did tend to have higher perceptions of flood risk. These tendencies indicate that the focus group results may underrepresent the voices of farmers from the smallest households and farms, though these farmers are represented equally in the regression and indexing results. During focus group discussions, information was elicited about which hazards were currently or historically experienced in the village. (iv) Data analysis Data from the survey, interviews, and focus groups were analyzed in multiple ways. To assess multiple dimensions of risk across the survey population and to compare the prioritization of concern between farmers and RDO development agents, we constructed several indices: the risk perception index (RPI), which represents individual risk perception, and the incidence (I), importance (P), and severity (S) index of each hazard, which reflect the ranking of hazards across households and study villages. The indices are based on two sets of survey responses; the first is a ranking of the three hazards of most concern and the second a rating of each of twelve hazards along three risk components (Table 5). We then utilized a series of regression analyses to examine the variables that are predictive of an individual farmers’ perception of risk, using the RPI as the dependent variable of analysis. The approximately normal distribution of the risk perception index (RPI) data for each hazard makes ANOVA and ordinary least squares (OLS) regression analyses appropriate. The significance results presented here from the OLS analysis are also tested for and are robust to heteroscedasticity. 4. RESULTS AND INTERPRETATION Surveyed farmers and RDO interview respondents collectively identified eight environmental hazards (landslide, soil erosion, flood, drought, hailstorm, pests and diseases, climate

change, and deforestation) and four social issues (corruption, market prices, the sale of counterfeit seeds, and overpopulation) as local concerns The results focus on three issues: (i) the perceptions and prioritizations of environmental and social issues by RDOs and the lay population, (ii) the individual perception of risk by smallholder farmers, and (iii) the importance of predictor factors in shaping farmer risk perceptions for a subset of four environmental hazards using regression analysis. (a) Relative importance of hazards to RDOs and farmers (i) RDO agent responses and results RDO development agents interviewed for this study expressed awareness of a wide array of challenges facing the local (primarily small-scale agricultural) population. Interview respondents were asked to speak to the issues of most concern in the region. In contrast to farmer rankings (described below), RDOs emphasized landslides as the most important environmental hazard and many social issues like overpopulation and a lack of knowledge over other environmental hazards (Table 4). When asked to link these concerns to farmer behavior and to speak to what farmers can and should do to address them, a number of solutions were posited. While land management strategies (e.g., terracing, contouring, planting trees, planting crops across rather than down the slope, rerouting water via trenches) were frequently mentioned and are the focus of RDO programs, the principal solution cited for long-term alleviation of the stresses of poverty and food insecurity was outmigration, either to another less densely populated agricultural area, or to the city. For example, one agent commented ‘‘. . .our 30-year vision is to move the 70% [rural population] to be with the 30% [urban population] where people live an urban life with a farm elsewhere” (Bugisu Production Officer, 2012). Another explained ‘‘You can’t talk about population in these areas, though, people say there is no problem, but it is for sure an issue. The land gets smaller and fragmented and therefore they have to take more and more land. . . and productivity is only going down because of over-cultivation and no replenishing of minerals or nutrients being added to the soil” (RDO agent, 2012). The focus of the interview respondents was on

Table 4. List of hazards and issues named by RDO interview respondents, separated by listed frequency. More than three-quarters of respondents Landslides Overpopulation Heavy rains Lack of knowledge

More than one-quarter of respondents Worsening soil quality/low yields Insufficient markets for agricultural products Human diseases Poverty Deforestation Land degradation Corruption (primarily political) Presence of middlemen who pay low prices Unreliable transportation infrastructure

More than half of respondents Hilly terrain Floods Animal pests and diseases Overcultivation of land Soil erosion Climate change Fewer than one-quarter of respondents Drought Hailstorms Water runoff Water contamination Overuse of fertilizers Windstorms Water shortage Changes in the river Hunger Theft Fake seeds Insufficient improved seeds No insurance available

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

RISK PERCEPTION IN A MULTI-HAZARD ENVIRONMENT

7

Table 5. Data collected directly or derived from the household survey and the focus groups and used in regression analysis Risk Perception Index (RPI)

Single index value to represent three component variables (perceived likelihood of experience, perceived severity of outcomes, and holistic concern about the issue) collected in the survey for each of 8 environmental hazards and 4 social issues (Environmental: landslide, soil erosion, flood, drought, hailstorm, pests and diseases, climate change, and deforestation; Social: corruption, market prices, the sale of counterfeit seeds, and overpopulation) Derived for each individual and each hazard based on responses on a 5-point scale from 0 to 4 (0 = no likelihood, not severe, and no concern; 3 = definite likelihood; 4 = extremely severe, and extreme concern). See Section 4(b) for further explanation.

RPI i;j ¼ Mean RPI deviance



  Li;j 2  S i;j þ C i;j 3

Derived measure of general environmental risk perception of an individual relative to the average perception of the sample population.

Pg

RPI deviancei;h ¼

j¼a ðRPI i;j

 RPI j Þ

7

Derived for each individual (i) for each of 8 environmental hazards (h) as the average difference between that individual’s RPI value and the mean population RPI value for all other environmental hazards (a-g) (e.g., in the OLS regression for landslide RPI, the mean RPI deviance for the individual would take into account only RPI values for the remaining 7 hazards of soil erosion, floods, drought, hailstorms, climate change, deforestation, and pests & diseases) Demographic variables

Engagement with organizations Hazard experience

Income (from any goods or services sold; does not include the value of goods both produced and consumed by the household; continuous; log adjusted, in 2013 USD); Income from coffee (binary; 0 = does not grow and sell coffee, 1 = grows and sells coffee) Farm acreage (continuous; in hectares); Respondent sex (binary; 0 = female, 1 = male); Children in household (binary; 0 = no school-aged children in household, 1 = presence of school-aged children in household); Uphill land (binary; 0 = no portion of cultivated land is on the slopes of the mountain; 1 = some portion of cultivated land is on the slopes of the mountain); Fragmented land (binary; 0 = all owned land is contiguous; 1 = land is in at least two pieces) Categorical variable comparing (1) those with no engagement to (2) those who engaged with an organization other than the Red Cross and (3) those who engaged with the Red Cross. The Red Cross is separated in this variable as it is the only organization focusing purely on risk reduction. Occurrence of hazard in village in the past (binary; 0 = hazard was not experienced; 1 = hazard was experienced) Derived from focus group responses; experience of a hazard in the village was used as a proxy for experience of that hazard by individuals living in the village

the belief that the root of all these challenges lay in overpopulation leading to land fragmentation and unsustainable land management (e.g., denuding of the landscape, draining of wetlands, a lack of fallows, and cultivation in riparian zones and in landslide areas). A multitude of other problems stemmed from these root issues, including environmental hazards as well as negative well-being outcomes like food insecurity and pervasive poverty. The environmental hazards identified by RDOs were primarily limited to those for which some mitigation through land management was possible (e.g. soil erosion named by over half of RDO interview respondents, flooding by twothirds, and landslides by all respondents). Landslides were the only hazard noted by representatives of every RDO. RDO agents rarely mentioned hazards for which there is little land management mitigation possible. For example, only one interviewee mentioned hailstorms and few RDOs engage in or promote any activities (e.g., crop insurance programs) that could mitigate hailstorm risk. A plurality of RDO agents (two-thirds) also note that climate change, pervasive deforestation, and environmental degradation expose farmers to additional risk, and that these combine with insufficient knowledge (noted by three-quarters of respondents) and/or an unwillingness to change their behavior to limit the potential for farmers to mitigate their risks in the near-term future.

(ii) Population hazard indices for small-scale farmers The indices used to analyze the rankings of hazards across the study population—incidence (I), importance (P), and severity (S)—are based on methods from Smith, Barrett, and Box (2000), Tschakert (2007) and Lo´pez-Marrero and Yarnal (2010). Each provides a slightly different viewpoint on how a hazard is perceived in aggregate by a given population and is calculated as a single metric from the individual responses in the sample. The incidence index (I) represents the proportion of participants in the study sample that named a specific hazard in their ranking of the top three hazards of concern. This is based on but differs slightly from the rankings used in Tschakert (2007) as the survey in this study constrained the total number of ranked hazards to a maximum of three. Values range from 0 (ranked by no one) to 1 (ranked by everyone) and are reported as percentages. Across the survey population, the hazards and issues with the highest incidence were environmental rather than social (Table 6). More than 50% of respondents named hailstorms in their top three hazards of most concern, followed by soil erosion (49%) and pests and diseases (46%). Rankings generally reflected our expectations across most villages based on geographical features. Concerns about soil erosion and hailstorms were ubiquitous while flooding and landslides were more localized.

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

8

WORLD DEVELOPMENT

Table 6. Incidence and importance index (I and P) values for all hazards ranked in the top three by farmers. Index values reported in aggregate for the study sample Hazard/issue

n

Hailstorm Soil erosion Pests & diseases Flooding Drought Landslide Market Prices Heavy rains Corruption Fake seeds Overpopulation Deforestation Climate change Hunger Windstorms Theft Other

233 207 192 134 105 103 47 39 38 18 16 16 14 10 6 2 12

Incidence index (I) (%) Importance index (P) 55.3 49.2 46.3 31.7 24.8 24.6 11.1 9.2 9.0 4.3 4.0 3.8 3.3 2.4 1.4 0.5 3.1

0.582 0.594 0.378 0.470 0.490 0.675 0.202 0.551 0.474 0.194 0.281 0.594 0.286 0.150 0.250 0.750 0.458

Market price fluctuation is the only social issue with an incidence index value of more than 10%. Climate change, deforestation, and overpopulation, issues frequently discussed as issues of concern by RDO personnel, ranked among the lowest for farmers, with incidence index values of only 3–4%. The importance index (P) represents how highly each issue was ranked by respondents, to get some measure of its relative position to other ranked issues in the population. The importance index was calculated for each issue by: P i;j

ðr  1Þ ¼1 ðn  1Þ

ð1Þ

where P i;j is the importance value for a given hazard j for a given individual i, r is the rank of that hazard as it relates to n, the total number of hazards named by the individual. The index ranges from 0 (lowest importance) to 1 (highest importance). A mean value of P was calculated for the subset of participants who identified a particular hazard

Figure 2. Plot of the incidence index (I) against the importance index (P) for each environmental hazard or social issue ranked by a survey respondent. Percentage of RDO representatives listing the issue is indicated by the shading of each point. A lack of knowledge, over-cultivation, and hilly terrain were ranked by more than two-thirds of experts, but were never ranked by individual farmers and thus do not appear on this chart.

(Lo´pez-Marrero and Yarnal, 2010) (Figure 2). While the incidence index (I) presented above indicates the proportion of the population very concerned about the hazard, the importance index (P) indicates the relative ranking of that hazard by those who ranked it. For example, a hazard ranked first, but by only one respondent would have a high importance (P) rating, but a low incidence (I) rating, while something that was ranked by many, but always in third place would have a high incidence, but low importance rating. Landslides 1 are the most important hazard for farmers as measured by the importance index, with a P rating of 0.675, demonstrating that, while they are not of ubiquitous concern across the population (an incidence index value of only 25%), they are an important concern where they are ranked (Figure 2). Of the remaining hazards ranked by at least 20% of the sample, soil erosion and hailstorms are the most important, with 0.594 and 0.582 importance index values respectively. Both hailstorms and soil erosion are ubiquitous throughout the study districts. Finally, the severity index (V) provides information on how dangerous people perceive these issues to be. It was calculated as the mean value of perceived severity for a given hazard for the subset of participants who ranked that particular hazard in their top three. V values range from 0 (least severe) to 4 (most severe). This index value varied little among issues, from a low of 2.30 ± 0.81(drought) to 2.87 ± 0.52 (hailstorms). When calculated for all respondents, including but not limited to those that ranked the issue, variation increased, ranging from 1.90 ± 1.33 (landslides) to 2.83 ± 0.55 (the sale of counterfeit seeds). This indicates that perceived severity is closely tied to ranking decisions. (iii) Comparison of RDOs and farmers While there were significant similarities in the range of hazards named by farmers and RDO agents, these two groups focused on different sub-sets of hazards, especially with respect to landslides, hailstorms, and the treatment of environmental as opposed to social issues. Given the high level of concern about landslides by local RDOs, it is notable that only 25% of respondents ranked this hazard in their top three. Even in Bushibuya, the entirety of which is located in a steep slope prone to mass movement events, concern about landslides was not ubiquitous. Fifteen percent of respondents in Bushibuya did not rank landslides in their top three concerns. These differences are even more pronounced between villages. In three villages (Shiruku, Bumwangu, and Sirumbusa), there were no respondents who ranked landslides as the primary hazard of concern and, in two of these (Shiruku and Sirumbusa), landslides were never listed. While this latter result reflects our expectations, given the relatively flat topography of these two villages, the low ranking overall is surprising and contrasts to the picture painted by RDOs. 2 Similarly, only one RDO agent mentioned hailstorms during the interview, though hailstorms are a key concern of farmers, having the highest incidence index value of all hazards (Figure 2). While this is consistent with the tendency of RDOs to focus on the hazards for which they offer a solution, it marks a clear difference in the ways in which farmers and development agents are conceptualizing risk in the landscape. Other environmental hazards were more consistent across the two groups. Both farmers and RDOs emphasized the risks and challenges associated with pests and diseases, soil erosion, and flooding. The partial overlap between the two groups in the case of environmental issues disappears almost completely when comparing RDOs and farmers on social issues. Market price

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

RISK PERCEPTION IN A MULTI-HAZARD ENVIRONMENT

fluctuation is the only social issue with an incidence index value of more than 10%, though it was named by 44% of RDO agents. Similarly, climate change, deforestation, and overpopulation issues were frequently named as issues of concern by RDO personnel, but ranked among the lowest for farmers, with incidence index values of only 3–4%. (b) Risk perception index for individuals Based on a more complex index created by Leiserowitz (2006) as a holistic measure of risk perception, the risk perception index (RPI) simplifies the multiple measures of risk perception into a single metric and is derived for each individual based on his or her perceived likelihood, perceived severity, and holistic concern for each hazard as

  Li;j 2 ð2Þ  S i;j þ C i;j RPI i;j ¼ 3 Here RPI i;j is the risk perception index rating for respondent i for hazard j calculated as a simple mean of stated holistic concern C and perceived ‘‘real risk” from that hazard, calculated as expected losses: perceived severity S, multiplied by perceived likelihood L. RPI is continuous, varies from 0 (no risk) to 4 (extreme risk), and represents a single metric describing an individual’s perceived risk for a given hazard. RPI ranges vary by hazard (Figure 3), with mean RPI values ranged from 2.00 (drought) to 2.48 (land fragmentation from overpopulation). One-way analysis of variance (ANOVA) tests indicate that these differences are significant across villages at the p < 0.05 level for all hazards except for perceived risk from deforestation. (c) Factors influencing risk perception for selected hazards A series of ordinary least squares regressions were performed to predict farmers’ risk perception (RPI). Variables included socioeconomic variables, geographic factors, respondent RPI deviation (indicating the general level of worry of individual respondents), respondent experience to the hazard either directly or indirectly, and respondent’s use of protective action (Table 5). No single set of independent variables consistently explains the Risk Perception Index (RPI) for all hazards across the

9

population. Though RPI ratings did vary across hazards and among people, this variability was small compared with the magnitude of variability in the independent variables. We highlight the results from a subset of four regression analyses to discuss these results, focusing on two of the hazards of most concern to farmers (soil erosion and hailstorms), and two hazards of most concern to development and risk reduction organizations in the region (landslides and floods). Across all four of our target hazards, the models were significant at the p < 0.001 level and reasonable adjusted R-squared values ranging from a low of 0.2975 for floods to 0.5276 for landslides (Table 7). No single set of independent variables was significant across all four hazards. Counter to results from other studies, the presence of school-aged children in the home (Lindell & Perry, 2000) is not significant for any hazard, while gender (Lindell & Perry, 2012; Wachinger et al., 2013) and other household characteristics are significant for some hazards but not others. Similar to other work, our study shows that hazard experience is significantly related to heightened risk perception for some hazards, when experienced indirectly through others in the same village (experience: landslides and marginally significant for floods) (Wachinger et al., 2013). Soil erosion, which is ubiquitous across the region, is an exception to this tendency, while interaction with RDOs has a marginally significant and slightly dampening effect on risk perception of hailstorms. Though hazard experience was an important predictor, the most powerful predictor of RPI across all four hazards was the tendency for an individual to perceive more or less risk from environmental hazards in general, relative to other individuals (RPI deviance). Those who worried more for one hazard also tended to do so for others. Both the significance (p < 0.001 for all hazards) and magnitude of this effect were strong, with a 1-point increase in RPI deviation corresponding to an effect more than twice as strong as the effect of experience for both landslides and flooding. For all hazards, RPI deviation reflected the largest magnitude effect behind the constant, and in the case of landslides its effect was greater than the constant. There is significant baseline concern across our study sample for each hazard, indicated by the regression constant. This is consistent with the focus group data indicating that these hazards are ubiquitous across the region. For most hazards,

Figure 3. Box plots for RPI of all environmental hazards queried in the household survey, with boxes containing the central 50% of respondents, circles marking the median, and whiskers extending to 1.5 the interquartile range of the nearest quartile (Tukey, 1977). Across all hazards, the 75th percentile RPI value of 2.625 reflects likelihood, severity, and concern all rated of 3 out of 4, the most common rating.

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

10

WORLD DEVELOPMENT

Table 7. Ordinary least squares (OLS) regression results for four target hazard RPI values. All models are very significant (p < 0.001). In bold are those variables that were significant (p < 0.05). In bold are those variables that were significant (p < 0.05). Experience was not included as a variable for soil erosion or hailstorms because all villages reported these hazards as occurring in their communities. Only the most salient protective land management action was assessed for each landslide, soil erosion, and flooding, but none was assessed for hailstorms as the authors are aware of no protective action advocated for this hazard Landslide RPI

Soil erosion RPI

Flood RPI

Hailstorm RPI

187 0.0000 0.5276

193 0.0000 0.4575

193 0.0000 0.3953

205 0.0000 0.2975

Number of obs p-Value (model) Adj R-squared variable

p-value

coefficient

0.022

0.518

0.048

0.305

0.000 0.646 0.593 0.006 0.062 0.534 0.394

0.278 0.035 0.004 0.043 0.002 0.008 0.237

0.010 0.830 0.971 0.664 0.983 0.671 0.005

0.008 0.535 0.224 0.566 0.356 0.023 0.097

0.956 0.014 0.090 0.000 0.011 0.380 0.403

0.010 0.200 0.034 0.036 0.004 0.003 0.028

0.906 0.110 0.634 0.626 0.965 0.865 0.670

1.126

0.000

0.902

0.000

0.879

0.000

0.505

0.000

Engagement (baseline: none) Other RDO Red Cross Experience

0.031 0.005 0.699

0.834 0.981 0.000

0.079 0.010 .

0.453 0.949 .

0.122 0.329 0.315

0.390 0.110 0.066

0.145 0.187 .

0.069 0.113 .

Protective action Contour Trench constant

0.238 . 0.824

0.054 . 0.004

. 0.095 2.274

. 0.291 0.000

. 0.210 1.609

. 0.092 0.000

. . 2.862

. . 0.000

Income (log-adjusted, 2013 USD) Income from coffee Non-farm income Acreage (ha) Hilly land Fragmentation Children in household Gender (0 = female, 1 = male) RPI deviation

coefficient

p-value

coefficient

0.040

0.413

0.572 0.102 0.072 0.378 0.272 0.017 0.099

baseline concern was both highly significant (p < 0.001) and had the highest coefficient value, ranging from 1.609 for floods to 2.862 for hailstorms. For landslides, baseline concern was still significant (p = 0.004), but its coefficient (0.824) is less than the magnitude of the effect of RPI deviation (1.126). The adoption of protective action measures is also interesting. For all hazards for which a protective action was available, the use of contour hedgerows for landslides and trenches for soil erosion and flooding, adoption and RPI are positively correlated indicating that those who perceived more risk were also those more likely to take protective action. The relationship ranged from insignificant (p = 0.291) in the case of trenches for soil erosion, to marginally significant in the case of trenches for flooding (p = 0.092) and contours for landslides (p = 0.054), but in each case with a positive association. For landslides, in addition to the three broadly significant factors, the only additional factors of significance are the cultivation of land on the sloped hillside and having income from coffee (Table 7). The relationship between slope cultivation and landslide RPI is significant (p = 0.006) and is also positive (0.378). Those who grew and sold coffee also perceived higher risk of landslide (p < 0.001), an effect comparable in magnitude to landslide experience in determining risk perception. In the case of soil erosion and only soil erosion among the hazards evaluated does gender plays a role in risk perception. Women perceive greater risk of soil erosion than do men (p = 0.005). Though the magnitude of this effect is much less than the relative effects of both baseline risk perception and the individual effect of RPI deviation, men perceive less risk than women in the case of soil erosion. For no other hazard was gender even marginally significant. As with landslides,

p-value

coefficient 0.008

****

p-value 0.774

perceived risk of soil erosion was also positively associated with coffee incomes (p = 0.010), with a similar magnitude of effect as that of gender on risk perception. Those who have invested in the cultivation of coffee are more concerned than others about soil erosion and landslides. For flooding, beyond the effects of baseline concern, RPI deviation, and experience, the presence of off-farm income and certain farm characteristics also play a role in risk perception. Having non-farm income through family members or non-agricultural employment has an attenuating effect on flood risk perception. Those who were employed in paid labor activities such as teaching, transport, and business ownership in addition to farming showed a lower perceived risk of flooding than their neighbors who had no off-farm income (p = 0.014). Farm characteristics are also significant contributors to flooding risk perception. Those who farmed on the hillier slopes perceived lower flooding risk (coefficient = 0.566, p < 0.001), while those with fragmented land perceived higher flooding risk (coefficient = 0.356; p = 0.011). Land fragmentation is very common in Bududa and Manafwa and frequently represents farmer use of more marginal lands for food production. As mentioned above, those who have experienced a flood (p = 0.066) also perceive more risk from floods. In the case of hailstorms, baseline risk perception and RPI deviation are the only significant factors in individual risk perception. No additional individual, household, or farm characteristics play a role in the extent to which individuals perceive hailstorm risk. Interestingly, interaction with RDOs has a marginally significant and negative relationship with hailstorm risk perception (coefficient = 0.145; p = 0.069 for RDOs other than the Red Cross). While this effect is slight in

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

RISK PERCEPTION IN A MULTI-HAZARD ENVIRONMENT

magnitude compared to baseline concern and RPI deviation, it is nonetheless interesting in that the direction of the relationship (lower risk perception) is different for hailstorms than for the other three hazards (heightened risk perception with RDO engagement) (Table 7). In fact, using only RPI deviation and RDO interaction as independent variables, the adjusted R2 value for hailstorm RPI increases to 0.3067, indicating that socio-economic factors contribute little to the development of the perceived risk of hailstorms. 5. DISCUSSION Smallholder farmers face multiple hazards. Understanding those hazards and translating that knowledge into protective action is essential for successful vulnerability reduction. Yet, risk perception in a multi-hazard environment is poorly understood. Our findings demonstrate that the factors shaping smallholder risk perception vary among hazards within the same study population and show that characteristics of both hazards and individuals shape risk perception. The regression analysis reveals an unexpected relationship between risk perception, self-efficacy, and protective action. Our findings further suggest that RDOs can play an important role in shaping risk perception, but also illuminate differences in how RDOs and smallholder farmers perceive and prioritize environmental and social concerns. Together, these findings point to the need for additional research about the mechanisms through which RDOs most successfully work with individuals in a multi-hazard environment. (a) Hazard and individual characteristics both shape risk perception Previous research on the predictors of risk perception for individual environmental hazards has found conflicting results regarding the importance of socio-economic and other individual characteristics (Wachinger et al., 2013). However, it is unclear if these inconsistencies are due to the characteristics of the hazards and/or differences among study populations. Our examination of risk perception of multiple hazards for a single population shows that the characteristics of the hazards themselves are likely to explain some of this variation. Our regression analysis demonstrates that risk perception drivers vary by hazard and are shaped by both hazard and smallholder characteristics. Factors such as off-farm income and gender, generally associated with capacity, socioeconomic status, and self-efficacy, are important only for some hazards. Experience at the village level, too, was a consistently important factor, suggesting geographic contributors or the role of social networks in shaping risk perception. Social networks have been shown to play a role in risk through the importance of social capital and mutual aid in mitigating risk (Adger, Hughes, Folke, Carpenter, & Rockstro¨m, 2005; Nakagawa & Shaw, 2004), and affecting sub-groups of a population differently (Yazdanpanah, Monfared, & Hochrainer-Stigler, 2013). Off-farm income is negatively associated with risk perception of flooding, while gender-differentiated risk perception is only a factor when considering the chronic stressor of soil erosion, but not for more acute hazards like landslides or flooding (Table 7). Similarly, the cultivation of coffee is associated with heightened risk perception of landslides and soil erosion. Coffee is the most common export crop in Bugisu, and one of the few crops grown by Bagisu farmers for sale outside local markets (Mugagga et al., 2012). While investment in coffee production is one of the foremost income-generating

11

activities (IGAs) pursued by the Bagisu, our results indicate it may also increase farmers’ risk perception. Experience of landslides and flooding increases risk perception, the two hazards of our study that are spatially heterogeneous. This builds on other work that has demonstrated the importance of experience in heightening risk perception (Grothmann & Reusswig, 2006; Siegrist & Gutscher, 2006; Terpstra, 2011) even in the case of indirect experience (Wachinger et al., 2013). Measures of experience are moot in other cases, like hailstorms and soil erosion, which are ubiquitous throughout the region. In a multi-hazard environment, chronic and ubiquitous hazards must not be forgotten in the shadow of catastrophic, but less broadly experienced, hazards. Some smallholders tend to perceive higher risks than others, regardless of hazard. Though our study did not investigate why some smallholders are more prone to worry than others, we suggest three likely explanations for this tendency toward generally elevated risk perception. First, that some people, regardless of underlying susceptibility to hazards, are more likely to perceive greater risk than are their neighbors, and that this tendency is not based solely on gender or socioeconomic status, but rather reflects individual risk aversion (Wossen, Berger, & Di Falco, 2015). Secondly, that some people, as a result of where they live or farm, are at greater risk from a great number of hazards compared to counterparts who are geographically susceptible to fewer (i.e., there is spatial heterogeneity of hazards within a multi-hazard environment). Lastly, it could reflect a compounding effect of multiple hazards, whereby initial susceptibility to one hazard may lead to increased vulnerability to other hazards, either because of compounded environmental susceptibility or double exposure to environmental and social risk (Cutter & Finch, 2008; O’Brien & Leichenko, 2000). This overlapping susceptibility to multiple risks is reflected in the significant baseline concern for all hazards in our sample, as reflected in the regression constant (Table 7) and in the focus group data indicating that many hazards are present in each village, though the set of hazards varies. (b) Self-efficacy and the ability to take protective action The regression analysis also illuminates an unexpected relationship between risk perception, self-efficacy, and protective action, with implications for the role of RDOs. Previous research has found that the perception of limited self-efficacy and associated low levels of protective action are linked to elevated risk perception (Bickerstaff, 2004; Martin et al., 2007; Wachinger et al., 2013). In contrast, our regression results demonstrate that the use of protective measures, where applicable, is positively associated with risk perception: those that perceive more risk are more likely to have adopted protective measures. This relationship is marginally significant for both landslides and flooding, even taking into account factors usually associated with vulnerability and low self-efficacy expectancy. If low self-efficacy were driving heightened risk perceptions, we would expect to see a negative relationship between risk perception and protective action because those who were most fearful would also be those who felt unable to act. The positive association between these two, however, indicates that heightened risk perception is not primarily driven by feelings of helplessness. While capacity metrics are still related to risk perception, as mentioned above, they are not preventing people from taking protective action. This result may indicate success on the part of RDOs, which frequently provide both information and the means to act (through trainings, material resources, or both). The majority

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

12

WORLD DEVELOPMENT

of smallholders who have adopted contour hedgerows (59%) and trenches (54%), credit RDOs with having provided the information necessary to implement the protective action and 72% of RDO beneficiaries received a farming tool (hoe, machete, or bucket), while 64% received seeds or seedlings. Most interactions involved the transfer of both information and material benefit. Furthermore, farmers report high levels of satisfaction with the efficacy of contour hedgerows (89% say they work) and trenches (90%). By providing knowledge and material benefits in combination, RDOs may be demonstrating their contribution to increasing the self-efficacy expectancy of stallholders as well as their understanding of hazards so that they can take appropriate protective action. This is discussed in more depth in (Sullivan-Wiley, 2016). (c) The role of RDOs Our analysis reveals the inconsistent influence of RDO activity in shaping the perceived risk associated with multiple hazards by smallholder farmers in Bugisu, as well as the contrast between RDOs and farmers in their risk concerns. Understanding the sources of these inconsistencies in both the regression results and risk prioritizations, whether stemming from the characteristics of the hazards or characteristics of RDO activity, can inform the design of more successful risk reduction policy and programs. Our regression results indicate that the relationship between RDO engagement and risk perception is not consistent across hazards. RDO engagement was only a significant predictor of hailstorm risk perception and, in this case, had a negative influence indicating that those who had received training or material goods from an RDO perceived lower hailstorm risk. While this result may indicate that engagement eases concern, it is inconsistent with the direction of RDO influence on perception of the other hazards, which is either positive or neutral (though not statistically significant). While these results demonstrate that RDOs may play a role in shaping risk perception, they raise questions regarding the nature of influence. Our work also illustrates differences in the perception of risks between the smallholder and RDO communities. RDOs generally focus training efforts on hazards that are within their capacity to affect (e.g., soil erosion through improved land management; a knowledge gap through training), while those hazards less readily addressed by land management, like hailstorms, are the most common concern ranked by smallholders. While these differences are not unexpected given that RDOs face a distinct set of incentive structures related to funding and evaluation metrics, the different emphasis may influence the relationship between RDOs and smallholder farmers. Though our research was not designed to tease out the longterm implications of these differences for the relationship between RDOs and the communities they work with, our data led us to posit two competing hypotheses regarding the potential impact of shared prioritization on long-term trust. First, there may be no relationship; the difference in priorities may be readily understood and accepted by both groups. This hypothesis is supported by the high rates of RDO goodwill reported by smallholders. The alternative hypothesis contends that the difference in priorities has the potential to weaken the trust at the foundation of the RDO-smallholder relationship, making RDO success more difficult over time. The importance of trust in facilitating success has been much discussed in the literature, and within this, the importance of the similarity heuristic (Bickerstaff, 2004; Boholm, 1998; Earle, 2010). This hypothesis is also supported by anecdotal evidence of

dissatisfaction and perceptions of a lack of understanding on the part of RDOs for what smallholders really need. One community member in Bududa, when asked about the help received from a local RDO after a recent mudslide, said ‘‘They don’t give us what we need. After a disaster we get food, but really the problem is education and the people cannot afford school fees [for their children]. We need help for education, for the next generation”. This is especially interesting in light of the emphasis RDO agents place on education as a long-term solution during interviews. The similarity in underlying understanding between the RDO and the farmer in this case may be undermined by the RDO field activities that do not reflect this understanding. Beyond the dangers it may pose to trust, inappropriate or misplaced relief may even result in direct negative side-effects for some intended beneficiaries, as was observed of government ex-post disaster relief in Iran, which exacerbated existing community inequalities (Yazdanpanah et al., 2013). Further research that includes long-term and/or ethnographic studies that assess how trust relationships evolve over time in multi-hazard landscapes would provide insight into if and how differences in prioritizations and field program actions impact trust. An improved understanding of how messaging and communication work to build that trust would provide practical insights to inform program design in vulnerable communities. 6. CONCLUSION Reducing vulnerability to multiple hazards and subsequently promoting economic development requires smallholders to perceive multiple hazards simultaneously and have both the will and the capacity to act (Wachinger et al., 2013). Though multi-hazard environments are widespread in the developing world and are likely to become more prevalent in the wake of climate change (IPCC, 2014; O’Brien et al., 2012; UNISDR, 2015), there are few studies that simultaneously examine overlapping risks in a multi-hazard environment. Our study of smallholder farmers in the Bugisu region of Uganda extends past work on risk perception, the risk perception paradox, and the role of RDOs in risk management by examining risk perception in a multi-hazard environment. Our findings clearly show that the factors that shape risk perception are specific to particular hazards, suggest that heightened risk perception can sometimes reflect greater understanding and motivation for protective action rather than just helplessness, and highlight the challenges associated with RDO involvement in the complex relationship between risk perception and protective action in a multi-hazard environment. Farmers in the Bugisu region are most concerned about the hazards like hailstorms, but their perception of the risk associated with ubiquitous stressors like soil erosion is also high. Smallholders in our study also showed that heightened risk perception and actually taking action were positively associated in some cases, indicating a potential role for the relationship between farmers and the RDOs that provide training and context for protective actions. Further research is required to improve risk management in a multi-hazard and development context. A better understanding of how farmers balance and prioritize protective action is essential, especially when the recommended protective actions are as varied as the sources of risk. This research should compare protective action adoption across multiple hazards within a single population, and the ways in which engagement with RDOs may be influencing these decisions and prioritizations.

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

RISK PERCEPTION IN A MULTI-HAZARD ENVIRONMENT

In addition, the nature of trust in long-term RDO-farmer relationships and the relative importance of the similarity heuristic in facilitating or eroding trust should be addressed. This research should examine how cycles of engagement, risk perception, action, and risk outcomes develop over time and

13

the influence this development has on the trust between RDOs and beneficiaries. Improved understanding of how RDO programs in the developing world are engaging with and influencing risk mitigation in the multi-hazard environments is fundamental for achieving the goal of reduced vulnerability.

NOTES 1. The importance index value (P) for Theft was greater than that of Landslides, but only two respondents ranked this threat, so that while it is of great importance to a very few, it is not discussed further in the context of the other hazards of broader concern.

2. The eight remaining villages contain terrain hilly enough for landslide risk. Between 45% and 100% of households within these villages have a portion of their land on the slope, indicating that they retain at least some exposure to a mass movement event.

REFERENCES Adger, W. N., Hughes, T. P., Folke, C., Carpenter, S. R., & Rockstro¨m, J. (2005). Social-ecological resilience to coastal disasters. Science, 309 (5737), 1036–1039. Baan, P. J. A., & Klijn, F. (2004). Flood risk perception and implications for flood risk management in the Netherlands. International Journal of River Basin Management, 2(2), 113–122. http://dx.doi.org/10.1080/ 15715124.2004.9635226. Bandura, A. (1995). Self-efficacy mechanism in psychobiologic functioning. In R. Schwarz (Ed.), Self-efficacy in changing societies (pp. 355–394). Cambridge University Press. Bickerstaff, K. (2004). Risk perception research: Socio-cultural perspectives on the public experience of air pollution. Environment International, 30(6), 827–840. http://dx.doi.org/10.1016/j.envint.2003.12.001. Boholm, A. (1998). Comparative studies of risk perception: A review of twenty years of research. Journal of Risk Research, 1(2), 135–163. Brewer, N. T., Weinstein, N. D., Cuite, C. L., & Herrington, J. E. Jr, (2004). Risk perceptions and their relation to risk behavior. Annals of Behavioral Medicine, 27(2), 125–130. Bronfman, N. C., Lo´pez Va´zquez, E., & Dorantes, G. (2009). An empirical study for the direct and indirect links between trust in regulatory institutions and acceptability of hazards. Safety Science, 47 (5), 686–692. http://dx.doi.org/10.1016/j.ssci.2008.09.006. Burningham, K., Fielding, J., & Thrush, D. (2008). ‘It’ll never happen to me’: Understanding public awareness of local flood risk. Disasters, 32 (2), 216–238. Claessens, L., Knapen, A., Kitutu, M. G., Poesen, J., & Deckers, J. A. (2007). Modelling landslide hazard, soil redistribution and sediment yield of landslides on the Ugandan footslopes of Mount Elgon. Geomorphology, 90(1–2), 23–35. http://dx.doi.org/10.1016/j.geomorph.2007.01.007. Cutter, S. C. (1981). Community concern for pollution social and environmental influences. Environment and Behavior, 13(1), 105–124. Cutter, S. L., & Finch, C. (2008). Temporal and spatial changes in social vulnerability to natural hazards. Proceedings of the National Academy of Sciences of the United States of America, 105(7), 2301–2306. http:// dx.doi.org/10.1073/pnas.0710375105. Cutter, S. L., Mitchell, J. T., & Scott, M. S. (2000). Revealing the vulnerability of people and places: A case study of Georgetown County, South Carolina. Annals of the association of American Geographers, 90(4), 713–737. Cvetkovich, G., & Lofstedt, R. E. (2013). Social trust and the management of risk. Routledge. Doss, C., McPeak, J., & Barrett, C. B. (2008). Interpersonal, intertemporal and spatial variation in risk perceptions: Evidence from East Africa. World Development, 36(8), 1453–1468. Earle, T. C. (2010). Trust in risk management: A model-based review of empirical research. Risk Analysis : An Official Publication of the Society for Risk Analysis, 30(4), 541–574. http://dx.doi.org/10.1111/j.15396924.2010.01398.x. Earle, T. C., & Siegrist, M. (2006). Morality information, performance information, and the distinction between trust and confidence. Journal of Applied Social Psychology, 36(2), 383–416. Fischhoff, B. (1995). Risk perception and communication unplugged: Twenty years of process. Risk Analysis, 15(2), 137–145.

Fischhoff, B., Slovic, P., Lichtenstein, S., Read, S., & Combs, B. (1978). How safe is safe enough? A psychometric study of attitudes towards technological risks and benefits. Policy Sciences, 9, 127–152. Fisher, R. (2013). ‘A gentleman’s handshake’: The role of social capital and trust in transforming information into usable knowledge. Journal of Rural studies, 31, 13–22. Flynn, J., Slovic, P., & Mertz, C. K. (1994). Gender, race, and perception of environmental health risks. Risk Analysis, 14, 1101–1108. http://dx. doi.org/10.1111/j.1539-6924.1994.tb00082.x. Gaillard, J.-C. (2008). Alternative paradigms of volcanic risk perception: The case of Mt. Pinatubo in the Philippines. Journal of Volcanology and Geothermal Research, 172(3–4), 315–328. http://dx.doi.org/ 10.1016/j.jvolgeores.2007.12.036. Gallina, V., Torresan, S., Critto, A., Sperotto, A., Glade, T., & Marcomini, A. (2016). A review of multi-risk methodologies for natural hazards: Consequences and challenges for a climate change impact assessment. Journal of Environmental Management, 168, 123–132. http://dx.doi.org/10.1016/j.jenvman.2015.11.011. Grothmann, T., & Reusswig, F. (2006). People at risk of flooding: Why some residents take precautionary action while others do not. Natural Hazards, 38(1–2), 101–120. http://dx.doi.org/10.1007/s11069-0058604-6. Gyekye, S. A., & Salminen, S. (2009). Educational status and organizational safety climate: Does educational attainment influence workers’ perceptions of workplace safety?. Safety Science, 47(1), 20–28. Heitz, C., Spaeter, S., Auzet, A.-V., & Glatron, S. (2009). Local stakeholders’ perception of muddy flood risk and implications for management approaches: A case study in Alsace (France). Land Use Policy, 26(2), 443–451. http://dx.doi.org/10.1016/j.landusepol.2008.05.008. Huppert, H. E., & Sparks, R. S. J. (2006). Extreme natural hazards: Population growth, globalization and environmental change. Philosophical Transactions of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, 364(1845), 1875–1888. International Federation of the Red Cross and Red Crescent Societies (IFRC). World Disasters Report IFRC, Geneva, 2014. IPCC (2014). Climate change 2014: Synthesis report. In Core Writing Team, R. K. Pachauri, & L. A. Meyer (Eds.), Contribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Geneva, Switzerland: IPCC, 151 pp.. Kasperson, R. E. (1986). Six propositions on public participation and their relevance for risk communication. Risk Analysis, 6(3), 275–281. Kasperson, R. E., Renn, O., Slovic, P., Brown, H. S., Emel, J., Goble, R., ... Ratick, S. (1988). The social amplification of risk: A conceptual framework. Risk Analysis, 8(2), 177–187. Leiserowitz, A. (2006). Climate change risk perception and policy preferences: The role of affect, imagery, and values. Climatic Change, 77, 45–72. http://dx.doi.org/10.1007/s10584-006-9059-9. Lichtenstein, S., Slovic, P., Fischhoff, B., Layman, M., & Combs, B. (1978). Judged frequency of lethal events. Journal of Experimental Psychology: Human Learning and Memory, 4(6), 551. Lin, S. Y., Shaw, D. G., & Ho, M. C. (2008). Why are flood and landslide victims less willing to take mitigation measures than the public?. Natural Hazards, 44(2), 305–314.

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

14

WORLD DEVELOPMENT

Lindell, M. K., & Hwang, S. N. (2008). Households’ perceived personal risk and responses in a multihazard environment. Risk Analysis, 28(2), 539–556. http://dx.doi.org/10.1111/j.1539-6924.2008.01032.x. Lindell, M. K., & Perry, R. W. (2000). Household adjustment to earthquake hazard: A review of research. Environment and behavior, 32(4), 461–501. Lindell, M. K., & Perry, R. W. (2012). The protective action decision model: Theoretical modifications and additional evidence. Risk Analysis, 32(4), 616–632. http://dx.doi.org/10.1111/j.1539-6924.2011.01647. x. Lo´pez-Marrero, T. (2010). An integrative approach to study and promote natural hazards adaptive capacity: A case study of two flood-prone communities in Puerto Rico. Geographical Journal, 176(2), 150–163. http://dx.doi.org/10.1111/j.1475-4959.2010.00353.x. Lo´pez-Marrero, T., & Yarnal, B. (2010). Putting adaptive capacity into the context of people’s lives: A case study of two flood-prone communities in Puerto Rico. Natural Hazards, 52, 277–297. http:// dx.doi.org/10.1007/s11069-009-9370-7. Martin, I. M., Bender, H., & Raish, C. (2007). What motivates individuals to protect themselves from risks: The case of wildland fires. Risk Analysis, 27(4), 887–900. http://dx.doi.org/10.1111/j.15396924.2007.00930.x. Mayhorn, C. B. (2005). Cognitive aging and the processing of hazard information and disaster warnings. Natural Hazards Review, 6(4), 165–170. Miceli, R., Sotgiu, I., & Settanni, M. (2008). Disaster preparedness and perception of flood risk: A study in an alpine valley in Italy. Journal of Environmental Psychology, 28(2), 164–173. http://dx.doi.org/10.1016/ j.jenvp.2007.10.006. Mileti, D. S., & O’Brien, P. W. (1992). Warnings during disaster: Normalizing communicated risk. Social Problems, 39(1), 40–57. Mugagga, F., Kakembo, V., & Buyinza, M. (2012). Land use changes on the slopes of Mount Elgon and the implications for the occurrence of landslides. Catena, 90, 39–46. Nakagawa, Y., & Shaw, R. (2004). Social capital: A missing link to disaster recovery. International Journal of Mass Emergencies and Disasters, 22(1), 5–34. Nathan, F. (2008). Risk perception, risk management and vulnerability to landslides in the hill slopes in the city of La Paz, Bolivia. A preliminary statement. Disasters, 32(3), 337–357. http://dx.doi.org/10.1111/j.14677717.2008.01043.x. NEMA (2010). State of the Environment Report for Uganda 2010-, Retrieved from http://www.nemaug.org/reports/national_state_report_2010.pdf. Nyland, L. G. (1993). Risk perception in Brazil and Sweden. RHIZIKON: Risk Research Report No. 15. Center for Risk Research, Stockholm School of Economics. O’Brien, K. L., & Leichenko, R. M. (2000). Double exposure: Assessing the impacts of climate change within the context of economic globalization. Global Environmental Change, 10(3), 221–232. http:// dx.doi.org/10.1016/S0959-3780(00)00021-2. O’Brien, K., Pelling, M., Patwardhan, A., Hallegatte, S., Maskrey, A., Oki, T., ... Viguie´, V. (2012). Toward a sustainable and resilient future. In C. B. Field, V. Barros, T. F. Stocker, D. Qin, D. J. Dokken, & K. L. Ebi (Eds.), Managing the risks of extreme events and disasters to advance climate change adaptation. A special report of working groups I and II of the Intergovernmental Panel on Climate Change (IPCC) (pp. 437–486). Cambridge University Press. Paton, D. (2008). Risk communication and natural hazard mitigation: How trust influences its effectiveness. International Journal of Global Environmental Issues, 8(1–2), 2–16. Perry, R. W., & Lindell, M. K. (2008). Volcanic risk perception and adjustment in a multi-hazard environment. Journal of Volcanology and Geothermal Research, 172(3), 170–178. Peters, R. G., Covello, V. T., & McCallum, D. B. (1997). The determinants of trust and credibility in environmental risk communication : An empirical study. Risk Analysis, 17(1), 43–54. Pirson, M., & Malhotra, D. (2008). Unconventional insights for managing stakeholder trust. Sloan Management Review, 49, 43–50. Plapp, T., & Werner, U. (2006). Understanding risk perception from natural hazards: Examples from Germany. Risk, 21, 101–108. Renn, O. (1999). A model for an analytic–deliberative process in risk management. Environmental Science & Technology, 33(18), 3049–3055. Renn, O. (2009). Risk communication: Insights and requirements for designing successful communication programs on health and environ-

mental hazards. In R. L. Heath, & H. Dan O’ Hair (Eds.), Handbook of Risk and Crisis Communication (pp. 80–98). London: Taylor and Francis. Renn, O., & Levine, D. (1991). Credibility and trust in risk communication (pp. 175–217). Netherlands: Springer. Rohrmann, B. (1994). Risk perception of different societal groups: Australian findings and crossnational comparisons. Australian Journal of Psychology, 46(3), 150–163. Shaw, R., & Izumi, T. (Eds.) (2014). Civil society organization and disaster risk reduction: The Asian Dilemma. Springer Science Business. Shi, P., & Kasperson, R. (Eds.) (2015). World atlas of natural disaster risk. Springer. Siegrist, M. (2000). The influence of trust and perceptions of risk and benefits on the acceptance of gene technology. Risk Analysis, 1220(2), 195–203. Siegrist, M., & Cvetkovich, G. (2000). Perception of hazards: The role of social trust and knowledge. Risk Analysis, 20(5), 713–719, Retrieved fromhttp://www.ncbi.nlm.nih.gov/pubmed/11110217. Siegrist, M., & Gutscher, H. (2006). Flooding risks: A comparison of lay people’s perceptions and expert’s assessments in Switzerland. Risk Analysis, 26(4), 971–979. http://dx.doi.org/10.1111/j.15396924.2006.00792.x. Sjo¨berg, L. (2000). Factors in risk perception. Risk Analysis, 20(1), 1–12. Sjo¨berg, L., Kolarova, D., Rucai, A.-A., Bernstro¨m, M.-L., & Flygelholm, H. (1996). Risk perception and media reports in Bulgaria and Romania. Stockholm: Center for Risk Research. Slovic, P. (1986). Perception of risk. Science, 236, 280–285. Slovic, P., Fischhoff, B., & Lichtenstein, S. (1979). Rating the risks. Environment, 21(3), 14–39. Slovic, P., Fischhoff, B., & Lichtenstein, S. (1980). Facts and fears: Understanding perceived risk. In Societal risk assessment (pp. 181–216). US: Springer. Smith, K., Barrett, C. B., & Box, P. W. (2000). Participatory risk mapping for targeting research and assistance: With an example from East African pastoralists. World Development, 28(11), 1945–1959. Sullivan-Wiley K., (2016). Risk reduction and development in a multihazard landscape: A case study of eastern Uganda (Order No. 10193193), Available from ProQuest Dissertations & Theses Global. (1847054412). Retrieved from https://search.proquest.com/docview/ 1847054412?accountid=9758 Terpstra, T. (2011). Emotions, trust, and perceived risk: Affective and cognitive routes to flood preparedness behavior. Risk Analysis, 31(10), 1658–1675. http://dx.doi.org/10.1111/j.1539-6924.2011.01616.x. Terpstra, T., & Lindell, M. K. (2013). Citizens’ perceptions of flood hazard adjustments: An application of the protective action decision model. Environment and Behavior, 45(8), 993–1018. http://dx.doi.org/ 10.1177/0013916512452427. Thomalla, F., Downing, T., Spanger-Siegfried, E., Han, G., & Rockstro¨m, J. (2006). Reducing hazard vulnerability: Towards a common approach between disaster risk reduction and climate adaptation. Disasters, 30(1), 39–48. http://dx.doi.org/10.1111/j.14679523.2006.00305.x. Tschakert, P. (2007). Views from the vulnerable: Understanding climatic and other stressors in the Sahel. Global Environmental Change, 17(3–4), 381–396. http://dx.doi.org/10.1016/j.gloenvcha.2006.11.008. Tukey, J. (1977). Exploratory data analysis. Reading, MA: AddisonWesley. Turner, R. H., Nigg, J. M., & Paz, D. H. (1986). Waiting for disaster: Earthquake watch in California. University of California Press. Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science (New York, N.Y.), 185(4157), 1124–1131. http://dx.doi.org/10.1126/science.185.4157.1124. UBOS (Uganda Bureau of Statistics) (2016). National population and housing census 2014: Main report and Eastern Region census sub-county report. Uganda: Kampala, Retrieved from http://www.ubos.org/2016/ 03/24/census-2014-final-results/. UNDP (United Nations Development Programme) (2013). Human Development Report 2013 Human Progress in a Diverse World. Human Development, 5 pages. Retrieved from https://www.hdr.undp. org/sites/default/files/country. UNISDR (United Nations Office for Disaster Risk Reduction) (2015). Making Development Sustainable: The Future of Disaster Risk Management, Global Assessment Report on Disaster Risk Reduction. Geneva, Switzerland: United Nations Office for Disaster Risk Reduction (UNISDR).

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002

RISK PERCEPTION IN A MULTI-HAZARD ENVIRONMENT URCS (Uganda Red Cross Society) (2010). 2010 Annual Report of the Disaster Management Department. Uganda Red Cross Society. Wachinger, G., Renn, O., Begg, C., & Kuhlicke, C. (2013). The risk perception paradox-implications for governance and communication of natural hazards. Risk Analysis, 33, 1049–1065. http://dx.doi.org/ 10.1111/j.1539-6924.2012.01942.x. Wossen, T., Berger, T., & Di Falco, S. (2015). Social capital, risk preference and adoption of improved farm land management practices in Ethiopia. Agricultural Economics, 46, 81–97. http://dx.doi.org/ 10.1111/agec.12142.

15

Yazdanpanah, M., Monfared, N., & Hochrainer-Stigler, S. (2013). Interrelated effects due to droughts for rural populations: A qualitative field study for farmers in Iran. International Journal of Mass Emergencies & Disasters, 31(2).

APPENDIX A. SUPPLEMENTARY DATA Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.worlddev.2017.04.002.

Available online at www.sciencedirect.com

ScienceDirect

Please cite this article in press as: Sullivan-Wiley, K. A., & Short Gianotti, A. G. Risk Perception in a Multi-Hazard Environment, World Development (2017), http://dx.doi.org/10.1016/j.worlddev.2017.04.002