pte

alia

Global Environmental Change 23 (2013) 12491257

mpo

Volu

by s

seh

resi

ciat

ere

ins

Contents lists available at ScienceDirect

Global Environm

l se1. Introduction

Economic losses from extreme weathers are rising due toclimate change (Michel-Kerjan and Kunreuther, 2011; The WorldBank, 2010; Warner et al., 2009). Average temperatures areprojected to increase and rainfall patterns to change. Consequent-ly, major ooding events are likely to become more intense andfrequent in the decades to come. This is expected to createenormous costs to communities, in the forms of rescue operations,loss of human life, asset damage, and business disruption(McDonald, 2010). Governments at all levels are at pains to paythe damage bill. This motivates the search for protective measuresagainst the risk of massive economic losses.

Flood insurance can supplement ofcial disaster relief schemesand provide a foundation for economic resilience. It can spread therisk of ooding across time and space and therefore reduce theuncertainties associated with climate change impacts. Well-designed insurance arrangements can protect communitiesagainst insured damage created by oods and provide economicincentives for voluntary efforts on risk mitigation. Flood insurancethus plays an important role in climate adaptation and hasattracted renewed interests (Botzen et al., 2010; Linnerooth-Bayer

et al., 2009; Penning-Rowsell and Pardoe, 2012; The World Bank,2010; Warner et al., 2009).

Lack of interest in insuring against natural hazards is one of thebarriers to climate adaptation at the household level. Someresidents of ood-prone areas are reluctant to voluntarilypurchase residential ood insurance cover even when it isaffordable and available (Handmer and Smith, 1989). Advancesin social sciences have identied a complex suite of social-cognitive factors responsible for the failures to insure (Baumannand Sims, 1978; Botzen et al., 2009; Botzen and van den Bergh,2009; Kunreuther, 1996, 2006; Kunreuther and Slovic, 1978;Laska, 1990; Zaleskiewicz et al., 2002), and more generally, thefailures to undertake adaptive behavioural adjustments (Adger,2003; Alexander et al., 2012; Grothmann and Patt, 2005; Hulme,2009; Raymond and Robinson, 2013; Wolf et al., 2010, 2013).Differential perceptions of risk have been cited as a key factorcontributing to these failures.

However, evidence on the linkage between risk perceptionand behaviour is far from consistent. The standard assumption isa simple positive relationship between perceived risk and thewillingness to purchase ood insurance cover (Botzen and vanden Bergh, 2012; Kunreuther, 1996; Warner et al., 2009). Asargued by Kunreuther (1996, p. 176), if the risk is perceived to berelatively high, then there is increased interest in purchasing apolicy. In reality, many individuals perceive the probability of anatural hazard causing damage to their home as being

Flood insurance

Social amplication of risk

Climate adaptation

Natural hazards

insuring decision indirectly through shaping perception of social norms. This implies that adaptive

behaviour is not necessarily a function of risk perception, but an outcome of its impacts upon the ways in

which the individuals situate themselves in their social circles or the society. There is a feedback process

in which individual perceptions of risk manifest as both a cause and effect, shaping and being shaped by

the socio-cultural context.

2013 Elsevier Ltd. All rights reserved.

* Tel.: +61 7 55527419; fax: +61 7 55528244.

E-mail address: alex.lo@grifth.edu.au

0959-3780/$ see front matter 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.gloenvcha.2013.07.019The role of social norms in climate adaperception and ood insurance purchas

Alex Y. Lo *

Grifth School of Environment, Grifth University, Gold Coast, Queensland 4222, Austr

A R T I C L E I N F O

Article history:

Received 4 February 2013

Received in revised form 13 July 2013

Accepted 21 July 2013

Keywords:

Risk perception

Social norms

A B S T R A C T

Flood insurance plays an i

of catastrophic ooding.

perception that is shaped

factors. It is based on a hou

of 501 randomly selected

insurance cover was asso

perceived norms and risk w

a mediating role between

jo ur n al h o mep ag e: www .eation: Mediating risk

rtant role in climate adaptation by recovering insured losses in the event

ntary adoption of ood insurance has been seen as a function of risk

ocial norms. This paper attempts to clarify the relationship between these

old survey conducted in the eastern cities of Australia and involving a total

dents. Results of a path analysis show that the likelihood of having ood

ed with perceived social norms, but not perceived ood risk. In addition,

statistically related to each other. It is concluded that social norms played

uring decision and risk perception. Risk perception might inuence the

ental Change

vier . co m / loc ate /g lo envc h a

A.Y. Lo / Global Environmental Change 23 (2013) 124912571250sufciently low lower than actuarial levels. Systematicmisperception or under-estimation of ood risks is thus listedas the primary reason for non-insurance (Kunreuther, 1996,2006).

Although the standard assumption has found empiricalsupport, it also comes with a fair amount of counter-evidence.For example Baumann and Sims (1978) and Laska (1990) nd noobservable relationship between risk perception and oodinsurance purchase. Hung (2009) even shows that these twovariables are negatively related to each other. Based on a review of16 relevant empirical studies, Bubeck et al. (2012) conclude thatthe explanatory power of risk perception has been overstated.Findings reported in the present paper also contradict the standardassumption. A point of departure for understanding the mixedndings is that risk perception is related to coping behaviour insome complex ways, precluding the use of over-simplisticdichotomous descriptors such as positive and negative. Thepresent paper suggests that the linkage becomes discernible onlywhen specic mediating factors are taken into account.

Past research has identied a number of factors contributing tothe coping responses of the individuals, other than risk perceptionand personal socio-economic characteristics. These include socialnorms (Frank et al., 2011; Grothmann and Patt, 2005; Hu et al.,2006; Lo et al., 2012; Nelson et al., 2007), which are regarded as akey driver of the decision to purchase ood insurance (Kunreuther,2006; Kunreuther et al., 2009). The dominant view holds that socialnorms inuence perception of climate risk (Grothmann and Patt,2005; Renn, 2011; Swim et al., 2010). Social referrals, expectationsand pressures inuence the individuals judgment on what is rightor true, and can therefore determine the ways in which riskinformation is processed and anchored, moderate or constrain thesense of impact, and consequently drive or prevent changes inbehaviour. Yet, recent research ndings indicate that the relation-ships between these variables are not linear. Frank et al. (2011), forexample, conclude that perception of climate risk is insufcient tomotivate adaptive responses. According to Norgaard (2011) andWolf et al. (2010), closer in-group social relations may even lead todenial or underestimation of risk. Thus, incorporating social factorsinto existing conceptual frameworks raise more questions thanoffering answers to the mixed observations regarding oodinsurance purchase and climate adaptation generally. Currentknowledge about the dynamics between risk perception, socialnorms and risk-related behaviour is far from complete.

Renn (2011) recently suggests that research into these issues,particularly in relation to climate change risk, could benet from aconceptual framework that features the reexive process of riskexperience. It is known as the social amplication of riskframework (Kasperson et al., 2003, 1988; Renn, 2011; Rennet al., 1992). The Social Amplication of Risk Framework isproposed as an interpretative framework for understandingexperiences of risk and their behavioural and broader societalimplications. It is based on the assumption that individuals processrisk information either by amplifying signals that appearfrightening or by attenuating those that are less threatening. Thisprocess is driven by or highly sensitive to social parameters, suchas social norms, which are creatively described as a socialamplier of risk. The functioning and transmission power ofsuch ampliers crucially inuence the formation of risk percep-tions. They operate through multiple feedback mechanisms andcomplicate the ways in which risk perceptions impact upon humanactions. By placing social-cognitive factors at centre, the frame-work could offer explanations as to why risk-related behavioursappear insensitive to risk perceptions in some cases.

A source of confusion, however, is that the standard assumptionbetween perception and behaviour is often taken for granted.Indirect pathways linking them to each other have received littleattention in the discussions of the Social Amplication of RiskFramework and other cognate theoretical accounts (e.g. Groth-mann and Patt, 2005). This paper argues that one of thesealternative pathways involves social norms as a mediatoroperating in between. It proves to be helpful for understandingthe observation reported here that perceived risk is not a keypredictor of ood insurance purchase. This paper seeks to clarifythe interrelationships between perceived risk, perceived socialnorms, and behavioural engagements with a focus on thevoluntary purchase of residential ood insurance. Findings of aquantitative analysis ascertain the effects of perceived risk andsocial norms on the likelihood of having ood insurance cover. Theresearch has broader implications for understanding the role ofsocial inuence in enhancing the capacity for coping with probableeconomic impacts of natural disasters on households.

This paper is organised as follow. The next section brieydiscusses the role of social norms in behavioural adaptation tonatural catastrophes, with a focus on ood insurance purchase. It isfollowed by a further elaboration on the conceptual problemsaddressed by this research. The inquiry is supported by a primarydataset collected in a questionnaire survey conducted in Queens-land, Australia. Research methods are introduced in the sectionthat follows. Research ndings are then presented, followed by adiscussion on conceptual implications.

2. Social norms: a source of behavioural distortions?

Norms that evolve from social interactions between individualsconstrain and guide their responses to known hazards. There isample evidence supporting the claim that peoples responses toprojected impacts of climate change are strongly inuenced bywhat they hear from other members of their social networks(Grothmann and Patt, 2005; Moser, 2007; Norgaard, 2011; Swimet al., 2010). These studies, however, do not specically focus ondecisions to insure against the rising risk of ooding due to climatechange. In the literature the role of social norms in inuencingthese decisions remains unclear and contested.

The psychometric tradition of risk analysis offers the dominanttheory of decision making for purchasing natural disasterinsurance (Kunreuther, 1996, 2006; Kunreuther and Slovic,1978; Slovic et al., 2000). Adherents to this approach suggest thatfailures to take out ood insurance stem from systematic bias ininformation processing and decision making on the part of theindividual. These cognitive failures include underestimation ofprobabilities and myopia, leading to misjudgments on riskexposure and future benets from risk-mitigating investments(Kunreuther, 2006). Conformity to social norms and socialinterdependencies are another example of such distortive factors(Kunreuther, 2006; Kunreuther et al., 2009).

An earlier report, dating back to 1970s, has indicated thatdiscussions with friends, neighbours, and family members couldincrease the likelihood of purchasing natural disaster insurance(Kunreuther, 1978). Based on their earlier research, Kunreutherand Michel-Kerjan (2009) argue that when homeowners hear thatother people have insured against ood risks, they becomemotivated to follow suit even without changing their beliefsabout the risks they face or knowing about the cost of coverage.Social norms may also result in premature cancellation ofinsurance policies after some years of coverage without makingany claim of insured damage (Kunreuther and Michel-Kerjan,2009, p. 126). Homeowners observe and tend to follow the actionsof their neighbours when deciding how much to spend onmitigating the ood risks they face.

Concerns have been raised about the normative role of socialnorms and interdependencies. Kuran (1995, p. 19) believes thatsocial norms are social artifacts that mask individual true

A.Y. Lo / Global Environmental Change 23 (2013) 12491257 1251preferences and result in undesirable social outcomes. In a similarvein, Kunreuther et al. (2009) argue that social interdependenciesare likely to impede selection of effective risk mitigation optionsand exacerbate the cognitive errors the individuals encounter. ToKunreuther (2006, p. 214), this is a regrettable outcome: a widerange of problems come under this rubric. The psychometricapproach stresses the pathological aspects of social normativeeffects.

Social norms are viewed in a more positive light in the study ofadaptive institutions. Operating social norms and networksgenerate social capital and give impetus to collective action(Ostrom, 2000). Networked relationships among members ofcommunity, built upon norms of trust and reciprocity, enable thesharing of knowledge, risk and resources, and can support recoveryfrom natural disasters and the resulting economic shocks throughmutual aids. Social norms and networks are generally deemed tobe conducive to adapting communities to climate change andreducing their vulnerability (Adger, 2003; Nelson et al., 2007;Pelling and High, 2005), despite some counter-evidence (Wolfet al., 2010). Afrmative empirical evidences exist in the literatureof ood risk management (Wong and Zhao, 2001), but very fewpertain to the purchase of ood insurance specically.

There is a shortage of evidence on how social norms might helpcommunities adapt to climate change by promoting voluntaryadoption of ood insurance. While the virtue of social norms isportrayed as a source of behavioural distortions under thepsychometric approach, it is considered to be pivotal to enhancingadaptive capacity under the social network approach. This meansthat it has mixed impacts on the capacity for climate adaptation.According to the Social Amplication of Risk Framework, the meritsof these claims depend on the ability of social networks and thenorms they engender to select the right aspects of risk and impactinformation, effectively transmit relevant risk signals to affectedgroups or individuals, and induce proper responses. In theseprocesses the persuasive forces of social norms play a mediating role.

A mediator relates one factor to another one which wouldotherwise lack observable linkage. But this does not seem to be akey premise of those conceptual frameworks that afrm the role ofsocial norms in climate adaptation, including the Social Ampli-cation of Risk Framework. This limitation is discussed in thefollowing section.

3. Relationships between risk perception, social norms, andbehaviour

3.1. Feedback mechanisms

The interaction between social norms and risk perceptionscomplicates the formulation of coping responses by the individu-als. Psychologists suggest that this is a dynamic and continuousprocess, in which individuals responses and their societal impactsfeedback and eventually become a determinant of risk perceptionand other related parameters (Swim et al., 2010). The SocialAmplication of Risk Framework explicitly recognises thesecomplexities.

Social amplication of risk is based on the idea that eventspertaining to hazards interact with psychological, social, institu-tional, and cultural processes in ways that can heighten orattenuate individual and social perceptions of risk and shape riskbehaviour (Renn et al., 1992, p. 139; see also Kasperson et al.,1988). Risk signals are subject to subjective transformations asthey run through social and cultural channels and individualcognitive processes. Such transformations can increase or decreasethe volume of risk information received, and may lead toreinterpretation of the risk events (Kasperson et al., 2003). Socialamplication thus denotes both intensication and attenuation ofrisk signals when they are transmitted through social stations,such as news media and social networks, and individual stations,such as the use of intuitive heuristics and attention lters by theindividual (Kasperson et al., 1988; Renn, 2011). This process maylead to overreactions of target audiences as well as their failures asrespond.

As a core element of the Social Amplication of Risk Framework,social norms and interactions operate within informal personalnetworks, which are one of the two major communicationnetworks that facilitate the ow of risk information and accountfor the societal processing and reconstruction of risks (Kaspersonet al., 1988, p. 185). The metaphor of amplier creatively illustratesthat the magnitude of outcoming risk perceptions depends on thefunctioning and the transmission power of various societal andindividual processors, such as social norms. The holistic scope ofthe Social Amplication of Risk Framework permits multiplefeedback mechanisms by which incoming risk signals, socialampliers, and output responses interact in a dynamic fashion (asschematically depicted in Kasperson et al., 2003, p. 14 and Renn,2011, p. 157).

The Social Amplication of Risk Framework could shed somelight on the empirically ambiguous role of risk perception ininuencing behaviour. The multiple feedback mechanisms allowalternative explanations for the ways in which risk perceptionsinduce behavioural adjustments. Although the idea is compelling,in practice there has been an emphasis on the impacts of socialnorms on risk perception, but less on the opposite way. This raisesquestions about the explanatory power of the Social Amplicationof Risk Framework as well as other cognate theoretical accountswith such a focus.

3.2. Mediation by what?

Although the metaphor of social amplier implies mediation bysocial norms, messages from the eld are confusing. For example,Renn et al. (1992, p. 151) nd that social stations, represented bynews media coverage in their study, create impacts on behaviourindirectly by shaping risk perceptions. In keeping with his earlierobservation, Renn (2011, p. 156) describes the causal relationshipin this way:

Social interactions can heighten or attenuate perceptions ofrisk. By shaping perceptions of risk, they also inuence riskbehaviour.

This suggests that risk perceptions mediate between socialinteractions and behavioural changes, rather than social interactionsbeing a mediator itself. Likewise, other empirical reports that adoptthe Social Amplication of Risk Framework tend to focus upon howsocietal processes affect risk perceptions (Pidgeon et al., 2003). Littleattention has been paid to the opposite route, although this is clearlypart of the Social Amplication of Risk Framework.

Similarly, other cognate theoretical accounts tend to assign amediator role for risk perception. For instance Wolf et al. (2010)conclude that social networks could impede proper copingresponses by perpetuating misperception of risk. Grothmannand Patt (2005) also put risk perception at the centre. In their socio-cognitive model, social discourse, which includes the factor ofsocial inuences, has no direct conceptual linkage to adaptiveintention, which is a precursor of adaptive behaviour. There is onlyan indirect linkage drawn through the mediator of risk perception(Grothmann and Patt, 2005, p. 205; Patt and Schroter, 2008, p.465). Likewise, Lindell and Hwang (2008) portray perceived risk asa mediating variable linking hazard adjustment and other factorswhich would otherwise be weakly related to each other.

A problem with this treatment is that the immediate causalrelationship between risk perception and behavioural response is

A.Y. Lo / Global Environmental Change 23 (2013) 124912571252taken for granted. Certainly there are other factors in between. Yet,as far as the role of social norms is concerned, it is often portrayedas a determinant of risk perception rather than also an outcome ofit (as illustrated by the Path 1 in Fig. 1). The indirect route by whichrisk perceptions create impacts upon coping behaviour byinuencing social norms backward is often ignored (Path 2).There is a lack of consistent and clear methodological emphasis onthe multiple, indirect routes by which heightened perceptions ofrisk spawn behavioural responses. Instead, there is stronger focuson Path 1 than Path 2. The failure to engage in reverse thinkinglimits our ability to comprehend systematic departures from theusual assumption.

People might fail to act even if they receive amplied riskmessages and adjust their perceptions of risk accordingly. There issome evidence on the weak linkage between risk perception andrelated adaptive action. For example Frank et al. (2011) concludethat perception of climate risk is insufcient to motivateadaptation. Other empirical reports have indicated that perceivedood risk is not related to the purchase of ood insurance(Baumann and Sims, 1978; Laska, 1990), nor actions to mitigateood risks generally (Bubeck et al., 2012; Wong and Zhao, 2001).On a larger and causally related issue, sociologists note that peoplemay choose to ignore collectively the possible dreadful conse-quences of climate change and remain silent with little motivationto act even if they are well informed and aware of the risks ahead(Norgaard, 2006, 2011). Amplied, or exaggerated, messages aboutimpending disasters may discourage climate change actions(Moser, 2007; Moser and Dilling, 2011). Perceiving natural risksas low may not be a sufcient reason for the failures to act.

This certainly does not mean that risk perceptions have nobehavioural implications. Risk perceptions, in terms of the SocialAmplication of Risk Framework, can act as an amplier itself. Forinstance, the belief that future ooding is likely and inescapablehas elevated expectations for mutual assistance among villagers inshared social networks, giving impetus to the evolution of newsocial norms for dealing with oods (Wong and Zhao, 2001). Other

Path 1

Per cepon of social norms

Risk percepon

Behaviour (and atude)

Path 2

Fig. 1. Hypothesised relationships between perception of social norms, riskperception, and risk related behaviour (and attitude).studies have also indicated that social processors of risk, such associal inuences (Short, 1984), social identity (Frank et al., 2011)and social trust (Frewer, 2003), play a mediating role between riskperception and behavioural response.

These social processes or processors alter the ways in whichindividuals perceive natural hazards, but risk perceptions can alsofeedback by moderating the ways in which they perceive andcollectively (re)construct these social processes or processors, orthe institutions in which they operate. These relationships are,therefore, reexive and reciprocal (Short, 1984). In social life,peoples perceptions of social norms interact with perceptions ofrisk and drive behaviour accordingly, along an alternative pathwaylinking perception and behaviour (Path 2 in Fig. 1). A concomitantreverse amplication process is possible, in which risk perceptionsshape the ways in which individuals see themselves as part of asocial or cultural group with a specic view on the issue at stake.Conformity to prevailing social norms is a key determinant ofbehavioural change, which may be related to risk perceptionindirectly through the mediator of perceived social norms.

Risk perceptions are not merely an object of amplication, butmay also amplify perceptions of other aspects of social life. Thisview could offer a better understanding about how adaptiveactions are motivated, even when they appear insensitive to theperceptions of related climate risks at rst sight. Social norms canplay a mediating role, linking perception and behaviour. To provideempirical evidence on this claim, primary data from a householdsurvey were analysed with a focus on the dynamic between thesevariables. The empirical study is introduced in the next section.

4. Study design

4.1. Research questions

This research examines the role of perceived social norms inmediating between risk perception and adaptive actions. All of thecausal relationships portrayed in Fig. 1 were statistically tested.That is, whether or not perceived risk and perceived social normsare related to each other, and whether or not these two factors arerelated to adaptive behaviour. Perceptions of ood risk and relatedsocial norms, and the purchase of ood insurance cover, are used asthe main measurements for the inquiry. Three working hypothesesare developed accordingly:

H1. Perceived social norms are related to perceived ood riskH2. Perceived social norms are related to the purchase of oodinsuranceH3. Perceived ood risk is related to the purchase of oodinsurance

It is hypothesised that the amplifying effects may run throughmultiple routes, and that the one mediated by social norms (Path 2in Fig. 1) is signicant (H1 and H2). The standard assumption thatrisk perception is positively associated with behaviour is alsoexamined (Path 1) (H3). In addition, two auxiliary measurements,introduced in the following section, are included to supportinterpretation of results.

4.2. Measurements

Data for addressing the listed research hypotheses werecollected by a structured questionnaire survey. Survey itemsincluded questions about the purchase of ood insurance. Respon-dents were asked if they currently had home and/or contentsinsurance on the residential property they were living in. Thosewho returned a positive response were then probed whether or notthey had ood cover on that insurance. Positive responses to thesecond question were coded 1 to indicate the holding of oodcover, or 0 if otherwise.

An auxiliary measurement was included in the analysis toaddress the limits of using ood insurance purchase as an indicatorof behaviour. The failure to insure might be due to lack ofopportunity for example, insurance companies refuse to offerood cover for those residential properties at very high ood risks.Thus, there might not be a choice for households located at aknown ood-prone site, regardless of their desire to get insured.Methodological isolation from such unintended factors can beachieved by introducing an attitudinal variable, namely, perceivedimportance of ood insurance, which is intuitively closely related tothe purchase decision (Pynn and Ljung, 1999; Zaleskiewicz et al.,2002). For this purpose, one of the survey questions requestedrespondents assess the importance of ood insurance in providingnancial security to their household.

A.Y. Lo / Global Environmental Change 23 (2013) 12491257 1253Perceived risk was gauged by asking respondents to what extentin which the home or property they currently live in is exposed tothe risk of ooding, based upon a ve-point scale ranging from norisk to extreme risk. Since causality cannot be determined bytesting statistical correlation between perceived risk and perceivedsocial norms (H1), an auxiliary measurement is introduced tosupport interpretation, i.e. recent experience of property damagedue to ooding. Whether ones home had suffered from ooddamage or not is unlikely to be affected by perceived social norms.Also, it is logically closely related to how people perceive futureood risks (Botzen et al., 2009; Grothmann and Reusswig, 2006;Hung et al., 2007). Therefore, this measurement allows logicalinference and can be used to examine, indirectly, the conjecturethat risk perception inuences perception of social norms.Respondents were probed to indicate if the 2011 Queenslandoods had created any damage on their current residentialproperty (Yes or No).

Perceived social norms relating to the decision to purchase oodinsurance were elicited by two items using a ve-point Likert scale,ranging from strongly disagree to strongly agree. They weremeasured in terms of the level in which the respondents believedthat (1) their family or friends want them to purchase oodinsurance and (2) other people like them would purchase oodinsurance. These measures resemble the way in which socialinuences are described in a draft public consultation paper(Productivity Commission, 2012, p. 62).

4.3. Administration of the survey

This questionnaire survey was conducted in the largest cities ofQueensland, Australia. Australia does not yet have a completenational policy for regulating the provision of commercial oodinsurance. Flood cover premiums are risk-based and the averageamounts range from AUD77 (Low Risk) to AUD5,496 (ExtremeRisk), or AUD1,018 across all risk bands (Insurance Council ofAustralia, 2011, p. 14) (AUD/USD = .993, as of 16 May 2012. Source:U.S. Federal Reserve). Government subsidies on ood coverageremain limited, and a standard denition of ood has come intoforce only recently. Insurers are not required to include ood coverin standard home and contents policies. In many cases where oodcover is offered, however, the policyholder can and does opt-out(Australian Treasury, 2011).

The State of Queensland was affected by the 201011 oods todifferent extents. Catastrophic oods hit the south-eastern part ofQueensland in the early January 2011. Prolonged and extensiverainfall led to ooding of historic proportions in the State. Theeconomy was severely damaged as a result. Coal exports weredisrupted, causing contraction in real GDP (Lim et al., 2012). Theoods affected more than 136,000 homes and businesses, and ledto the death or missing of 36 people (Queensland Government,2011, p. 16; Queensland Government, 2012, p. 32). The resultingeconomic losses amounted to AUD$5.8 billion (QueenslandGovernment, 2011, p. 3).

Survey sites include the Gold Coast Council Region, the SunshineCoast Council Region, and the Greater Brisbane Region. According toAustralian Bureau of Statistics (ABS) 2011 Census data, these regionscurrently account for 64.3% of the total population of Queensland, or13.0% of Australias population. Brisbane is the States capital and hasexperienced its second highest ood since the beginning of the 20thcentury during the 201011 Queensland oods. The Gold Coast,approximately 80 km south of Brisbane, is ranked as one of the mostood prone local councils in Australia (Smith, 2002). Recent ofcialreports conrm that the Gold Coast and the Sunshine Coast, 100 kmnorth of Brisbane, remain among the Queenslands local governmentareas at the greatest risk of inundation (Department of ClimateChange, 2009). According to the Intergovernmental Panel on ClimateChange, there will be more severe ood events in South EastQueensland due to climate change (Hennessy et al., 2007, p. 530).

The main survey was conducted in May 2012, preceded by apilot test to allow renements to the survey instrument. Anestablished market research company was appointed to undertakethe surveys using the Computer Assisted Telephone Interviewing(CATI) technique. Sample was selected using a random digit dialingapproach for households with landlines, stratied by telephoneexchange prex. It was supplemented with 1015% of residentialmobile phone numbers selected from the electronic white pages sothat mobile numbers could be stratied by region (as thesenumbers have a residential address attached to them).

Quotas were used in the selection of household for interview.Since the survey sought to interview the person in the householdwho was the most knowledgeable about their household insur-ance, the introductory text asked to speak to the person in thehousehold mainly responsible for payment of the major householdbills. In order to obtain a representative sample of households,sample quotas were set by the number of people living in eachhousehold per region, based on the latest available census data(2006). The number of people in household was obtained fromeach contacted household who agreed to participate, and used forinitial screening in accordance with the set quotas.

5. Research ndings

5.1. Characteristics of the sample

A total of 501 household representatives completed theinterviews, at the response rate of 20.2%. Fewer households withonly one person (18%) completed the survey than actually residentin the regions (22.5%, based on 2006 census data). Thirty-eight percent of the households interviewed across the three regionsconsisted of two members only, slightly higher than the regionalestimate of 35.6%. Families with four members or more were alsoover-represented in the survey by a small margin, i.e. 26.8%comparing to the regional estimate of 25.4%. The sampling methodproduced a generally representative sample in terms of householdcomposition.

Other socio-economic information collected included gender,education attainment, and income. More than half (55.7%) of therespondents were female. About 3.2% were aged 30 or below and23.2% were between 31 and 45. More than 70% of respondentswere 46 or older. Holders of a university degree or higheraccounted for 40.9 per cent of the sample. Nearly one-third ofrespondents had an annual household income below AUD50,000.Less than a half (47.3%) earned between AUD50,000 andAUD150,000 a year. Few of them (12.2%) had more thanAUD150,000. The rest of the respondents refused to discloseincome information.

5.2. Model variables

Model variables for hypothesis testing are shown in Table 1. Ofthose households surveyed who had home and/or contentsinsurance (i.e. excluding non-policyholders), 62% claimed to haveood cover on it (or 56% if non-policyholders are included). Inresponse to the question about the importance of ood insurancein providing nancial security for their household, 70.7% ofrespondents selected one of the two agreement options (Stronglyagree and Agree), yielding a mean score of 3.90.

Most of the respondents reported living in an area of low oodrisk. Nearly 60% of them believed that they were not exposed toany risk of ooding, and 32.1% faced low ood risk. Less than 10%reported exposure to medium, high, or extreme risks. This yielded

Range Agreement (%) Mean Standard deviation

01 .62 .49

15 70.7 3.90 1.37

15 1.55 .78

01 .09 .28

15 25.8 2.59 1.41

15 47.6 3.30 1.33

nc

A.Y. Lo / Global Environmental Change 23 (2013) 124912571254Table 1Descriptive statistics for model variables.

Item Description

Have ood cover Reported to have ood cover on

current insurance policies (yes or

otherwise)

Flood insurance is important Flood insurance is important as it

provides nancial security for my

household (ranging from Strongly

disagree to Strongly agree)

Perceived risk Perceived ood risk of home/property

currently living (ranging from No

Risk to Extreme Risk)

Flood damage experience The major ooding event in

Queensland in 2011 caused damage

to the home/property currently living

(Yes or otherwise)

Family/friends want me to insure My family or friends think I should

insure my house against ooding

(ranging from Strongly disagree to

Strongly agree)

Other people would buy too Most people like me will purchase

ood insurance (ranging from

Strongly disagree to Strongly agree)

Table 2Correlation coefcients for model variables.

Have ood cover Flood insura

important

Flood insurance is important .268**

Perceived risk .163** .139**

Flood damage experience .031 .079

Family/friends want me to insure .352** .282**

Other people would buy too .218** .422**a mean score of 1.55. In addition, about 9% of the respondents hadtheir properties damaged by the 2011 catastrophic oods.

The rst item of perceived social norms recorded a mean scoreof 2.59. This included about one-fourth (25.8%) of the respondentswho felt that they were subject to expectations from family orfriends to insure against ooding. Almost half of the respondents(47.6%) agreed that most other people would purchase oodinsurance. The mean score for this item is 3.30.

Correlation coefcients between all of the model variables arelisted in Table 2. Both of the auxiliary measurements aresignicantly related to their logical associates (i.e. perceivedimportance related to ood insurance purchase, and ood damageexperience related to perceived risk). Note also that perceived riskcorrelated with the likelihood of having ood insurance (hereaftercalled ood cover) as well as its perceived importance. However,the magnitude of statistical signicance declined sharply whensome other factors were controlled for, as shown in the nextsection.

5.3. Path analysis

A further analysis was conducted based on structural equationmodeling and using SPSS Amos 21.0. Fig. 2 presents the pathestimates for perceived social norms, perceived risk, ood damageexperience, and ood cover. Although perceived risk correlatedwith ood cover (as shown in Table 2), the explanatory powerdiminished when it was mounted on a regression model. Both ofthe two items representing perceived social norms were signi-cantly associated with the likelihood of having ood cover as wellas perceived risk.

Similarly, actual ood damage experience created no signicantimpacts on ood cover. However, this variable correlated with one

** Signicant at the .01 level (2-tailed).e Perceived risk Flood damage

experience

Family/friends want

me to insure

.327**

.302** .130**

.086 .002 .351**of the social normative variables, i.e. expectations from familymembers or friends. Since damage on property is an objective andinvoluntary outcome, a logically sensible interpretation is that theexperience with ood damage elevated the level of perception ofsocial norms on the part of the affected households, i.e. perceptionis likely to be an effect of ood experience rather than a cause(hence the direction of arrows). These signicant estimates camewith positive signs, suggesting that they increased with each other.The path analysis shows that ood cover is a function of perceivedsocial norms, but not perceived risk and ood experience which arein turn associated with the normative variables.

Perceived ood risk to own property

Perceived social norms

Have ood cover on current

insurance policies

Perceived risk and hazard experience

Behaviour

.11*

.30***

.06

(R2 = 13.9)

The 2011 oods caused damage to own property

-.06

Family / friends want me to

insure

Other people would buy too

.30**

.00

.35*** .33***

Insignicant pathSignicant path

Notes: All coecients are standardised. * p < .05, ** p

Perceived ood risk to own property

Perceived social norms

Flood insurance is important

Perceived risk and hazard experience Atude

.37***

.13**

.05

(R2 = 20.5)

The 2011 oods caused damage to own property

.04

Family / friends want me to

insure

Other people would buy too

.30***

.00

.35*** .33***

.09*

.13**

A.Y. Lo / Global Environmental Change 23 (2013) 12491257 1255The same pattern of relationships replicated for the perceivedimportance of ood insurance, as illustrated in Fig. 3. Perceived riskremained statistically insignicant, whereas both of the socialnorm variables demonstrated strong impacts on the dependentvariable. Again, this means that perceived risk lost explanatorypower when these social normative factors were controlled for.Perceived importance was also not sensitive to ood damageexperience. Substituting ood cover for perceived importance didnot create observable qualitative variations from the regressionmodels presented in Fig. 2. This suggests that the behaviour of andattitude toward insuring against ood risks responded to theindependent variables consistently. Perceived social norms provedto be a powerful explanatory factor.

6. Discussions

The path estimates reported consistently support hypothesesH1 and H2 listed earlier in the paper, but not H3. The auxiliarymeasurements helped triangulate the ndings. Two importantobservations warrant further discussions. First, risk perceptionevidenced only a modest degree of association with the adoption ofood insurance cover. Second, perceived social norms mediatedbetween perceived risk and the likelihood of having ood cover. Inthis study, therefore, the Path 1 depicted in Fig. 1 is not active,

Insignicant pathSignicant path

Notes: All coecients are standardised. * p < .05, ** p

A.Y. Lo / Global Environmental Change 23 (2013) 124912571256important to the respondents, which is not perfectly amenable tothe standard game theoretic treatment suggested by Kunreutheret al. (2009, p. 14). It might be tenable, but clearly insufcient, tounderstand the role of social norms in such terms.

Social norms may act as a subjective lter determining theways in which perceived risk is responded to. As observed byFrank et al. (2011), farmers perceiving climate risks to be highwere not motivated to take adaptive actions as these are deemedto be futile efforts, unless the risk and adaptation informationcomes from an afliated social in-group or a trusted out-group.Thus, the key is not the magnitude of perceived risk, but themediating role of social identity and related social norms (Franket al., 2011). Social identity is regarded as a critical factor thatcould activate the motivational energy of perceived risk byenhancing a sense of self-efcacy, which is found in another studyto be a motive for the purchase of ood insurance (Baumann andSims, 1978). Baumann and Sims (1978) observation that riskperception failed to explain the purchase decision could havebeen addressed by exploring its causal relationship with the senseof self-efcacy. More generally, the indirect behavioural impactsof risk perception could be better understood by ascertaining itsinuences on the collective and personal attributes of social life.Such attributes not only affect how risk is perceived, but mightalso be a manifestation of perceived risk, both accounting for thepropensity to act.

7. Conclusions

In this study, perceived ood risk was not found to be apredictor of ood insurance purchase. Perceived social norms notonly demonstrated strong impacts on the insurance decision, butalso associated with perceived risk. This factor therefore played amediating role by correlating with risk perception and behaviourwhich lost immediate connection in a regression model. Thisindicates a possible indirect pathway by which risk perceptionimpacts upon coping behaviour.

The ndings could offer more analytic options for investigat-ing the interrelationships among climate adaptation, riskperception, and social networks and interactions. Highermotivation to adapt is increasingly understood as largelydependent upon the capacity of social networks and interactionsto shape risk perception in desirable ways (Frank et al., 2011;Wolf et al., 2010). The socially constructed perceptions of riskcan, in turn, inuence the ways in which the individualssubjectively situate themselves in their social circles or thelarger society in relation to the risk issues concerned. Adaptiveactions are then not, or not only, a function of risk perception, butan outcome of its impacts upon the perception and consequentlyoperation of social networks and institutions.

Although such perspectives are familiar to advocates of theSocial Amplication of Risk Framework and other cognatetheories, institutional recognition remains limited (e.g. Produc-tivity Commission, 2012; The World Bank, 2010). They alsowarrant more attention by the dominant school of catastropheeconomics (Kunreuther, 2006; Kunreuther and Michel-Kerjan,2009), which addresses the economic aspects of climateadaptation, particularly the purchase of natural disaster insur-ance. Further research could examine a wider range of relatedeconomic decisions, such as voluntary relocation of homes inanticipation of catastrophic oods or sea level rises. Perceivedsocial norms could be more broadly dened in terms of the largersociety, ethnic or religious culture, and other informal institu-tions. A more sophisticated scale for measuring risk perceptioncould also help substantiate the arguments developed in thispaper.Acknowledgements

This research was funded by the Grifth Climate ChangeResponse Program (GCCRP). The GCCRP also supported a presen-tation of this paper at the 2012 conference of the Australia NewZealand Society for Ecological Economics held on the Gold Coast.Comments from conference participants on this paper areappreciated.

References

Adger, W.N., 2003. Social capital, collective action, and adaptation to climatechange. Economic Geography 79, 387404.

Alexander, K.S., Ryan, A., Measham, T.G., 2012. Managed retreat of coastal commu-nities: understanding responses to projected sea level rise. Journal of Environ-mental Planning and Management 55, 409433.

Australian Treasury, 2011. Natural Disaster Insurance Review. The Treasury, Aus-tralian Government, Canberra.

Baumann, D.D., Sims, J.H., 1978. Flood insurance: some determinants of adoption.Economic Geography 54, 189196.

Botzen, W., van den Bergh, J., Bouwer, L., 2010. Climate change and increased risk forthe insurance sector: a global perspective and an assessment for theNetherlands. Natural Hazards 52, 577598.

Botzen, W.J.W., Aerts, J.C.J.H., van den Bergh, J.C.J.M., 2009. Willingness of home-owners to mitigate climate risk through insurance. Ecological Economics 68,22652277.

Botzen, W.J.W., van den Bergh, J.C.J.M., 2009. Bounded rationality, climate risks,and insurance: is there a market for natural disasters? Land Economics 85, 265278.

Botzen, W.J.W., van den Bergh, J.C.J.M., 2012. Risk attitudes to low-probabilityclimate change risks: WTP for ood insurance. Journal of Economic Behaviour &Organization 82, 151166.

Bubeck, P., Botzen, W.J.W., Aerts, J.C.J.H., 2012. A review of risk perceptionsand other factors that inuence ood mitigation behaviour. Risk Analysis 32,14811495.

Department of Climate Change, 2009. Climate Change Risks to Australias Coast: AFirst Pass National Assessment. Department of Climate Change, AustralianGovernment, Canberra.

Frank, E., Eakin, H., Lopez-Carr, D., 2011. Social identity, perception and motivationin adaptation to climate risk in the coffee sector of Chiapas, Mexico. GlobalEnvironmental Change 21, 6676.

Frewer, L.J., 2003. Trust, transparency, and social context: implications for socialamplication of risk. In: Pidgeon, N., Kasperson, R.E., Slovic, P. (Eds.), The SocialAmplication of Risk. Cambridge University Press, Cambridge.

Grothmann, T., Patt, A., 2005. Adaptive capacity and human cognition: the processof individual adaptation to climate change. Global Environmental Change 15,199213.

Grothmann, T., Reusswig, F., 2006. People at risk of ooding: why some residentstake precautionary action while others do not. Natural Hazards 38, 101120.

Handmer, J., Smith, D.I., 1989. Flood insurance and relief in Australia: the back-ground. In: Smith, D.I., Handmer, J. (Eds.), Flood Insurance and Relief inAustralia. Centre for Resource and Environmental Studies, Australian NationalUniversity, Canberra.

Hennessy, K., Fitzharris, B., Bates, B.C., Harvey, N., Howden, S.M., Hughes, L.,Salinger, J., Warrick, R., 2007. Australia and New Zealand. In: Parry, M.L.,Canziani, O.F., Palutikof, J.P., Linden, P.J.v.d., Hanson, C.E. (Eds.), ClimateChange 2007: Impacts, Adaptation and Vulnerability: Contribution of Work-ing Group II to the Fourth Assessment Report of the Intergovernmental Panelon Climate Change. Cambridge University Press, Cambridge, UK.

Hu, Q., Zillig, L.M.P., Lynne, G.D., Tomkins, A.J., Waltman, W.J., Hayes, M.J., Hubbard,K.G., Artikov, I., Hoffman, S.J., Wilhite, D.A., 2006. Understanding farmersforecast use from their beliefs, values, social norms, and perceived obstacles.Journal of Applied Meteorology and Climatology 45, 11901201.

Hulme, M., 2009. Why We Disagree About Climate Change. Cambridge UniversityPress, Cambridge.

Hung, H.C., 2009. The attitude towards ood insurance purchase when respondentspreferences are uncertain: a fuzzy approach. Journal of Risk Research 12, 239258.

Hung, H.V., Shaw, R., Kobayashi, M., 2007. Flood risk management for the RUA ofHanoi: importance of community perception of catastrophic ood risk indisaster risk planning. Disaster Prevention and Management 16, 245258.

Insurance Council of Australia, 2011. Insurance Council of Australia Response to 2011Natural Disaster Insurance Review. Insurance Council of Australia, Sydney.

Kasperson, J.X., Kasperson, R.E., Pidgeon, N., Slovic, P., 2003. The social amplica-tion of risk: asses fteen years of research theory. In: Pidgeon, N., Kasperson,R.E., Slovic, P. (Eds.), The Social Amplication of Risk. Cambridge UniversityPress, Cambridge, pp. 1346.

Kasperson, R.E., Renn, O., Slovic, P., Brown, H.S., Emel, J., Goble, R., Kasperson, J.X.,Ratick, S., 1988. The social amplication of risk: a conceptual framework. RiskAnalysis 8, 177187.

Kunreuther, H., 1996. Mitigating disaster losses through insurance. Journal of Riskand Uncertainty 12, 171187.

Kunreuther, H., 2006. Disaster mitigation and insurance: learning from Katrina.Annals of the American Academy of Political and Social Science 604,208227.

Kunreuther, H., Meyer, R., Michel-Kerjan, E., 2009. Overcoming Decision Biases toReduce Losses from Natural Catastrophes. The Wharton School, Philadelphia,PA.

Kunreuther, H., Michel-Kerjan, E., 2009. At War With the Weather: Managing Large-Scale Risks in a New Era of Catastrophes. MIT Press, Cambridge, MA.

Kunreuther, H., Slovic, P., 1978. Economics, psychology, and protective behaviour.American Economic Review 68, 6469.

Kunreuther, H.C., 1978. Disaster Insurance Protection: Public Policy Lessons. Wiley,New York.

Kuran, T., 1995. Private Truths, Public Lies: The Social Consequences of PreferenceFalsication. Harvard University Press, Cambridge, MA.

Laska, S.B., 1990. Homeowner adaptation to ooding: an application of the generalhazards coping theory. Environment and Behaviour 22, 320357.

Lim, G.C., Chua, C.L., Claus, E., Nguyen, V.H., 2012. Review of the Australian Economy201112: a case of Deja` Vu. Australian Economic Review 45, 113.

Lindell, M.K., Hwang, S.N., 2008. Households perceived personal risk and responsesin a multihazard environment. Risk Analysis 28, 539556.

Linnerooth-Bayer, J., Warner, K., Bals, C., Hoppe, P., Burton, I., Loster, T., Haas, A.,2009. Insurance, developing countries and climate change. The Geneva Papers34, 381400.

Lo, A.Y., Chow, A.T., Cheung, S.M., 2012. Signicance of perceived social expectationand implications to conservation education: turtle conservation as a case study.Environmental Management 50, 900913.

McDonald, J., 2010. Paying the price of adaptation: compensation for climatechange impacts. In: Bonyhady, T., Macintosh, A., McDonald, J. (Eds.), Adapta-tion to Climate Change: Law and Policy. The Federation Press, Annandale, NSW,pp. 234264.

Michel-Kerjan, E., Kunreuther, H., 2011. Redesigning ood insurance. Science 333,408409.

Pelling, M., High, C., 2005. Understanding adaptation: what? can social capital offerassessments of adaptive capacity? Global Environmental Change 15, 308319.

Penning-Rowsell, E.C., Pardoe, J., 2012. Who benets and who loses from ood riskreduction? Environment and Planning C: Government and Policy 30, 448466.

Pidgeon, N., Kasperson, R.E., Slovic, P., 2003. Introduction. In: Pidgeon, N., Kas-person, R.E., Slovic, P. (Eds.), The Social Amplication of Risk. CambridgeUniversity Press, Cambridge.

Productivity Commission, 2012. Barriers to Effective Climate Change AdaptationDraft Report. Productivity Commission, Australian Government, Canberra.

Pynn, R., Ljung, G.M., 1999. Flood Insurance: A Survey of Grand Forks, North Dakota,Homeowners. Applied Behavioural Science Review 7, 171180.

Queensland Government, 2011. Climate Change: Adaptation for Queensland IssuesPaper. Department of Environment and Resource Management, QueenslandGovernment, Brisbane.

Queensland Government, 2012. Queensland Floods Commission of Inquiry FinalReport. Queensland Government, Brisbane.

Raymond, C.M., Robinson, G.M., 2013. Factors affecting rural landholders adapta-tion to climate change: insights from formal institutions and communities ofpractice. Global Environmental Change 23, 103114.

Renn, O., 2011. The social amplication/attenuation of risk framework: application toclimate change. Wiley Interdisciplinary Reviews: Climate Change 2, 154169.

Renn, O., Burns, W.J., Kasperson, J.X., Kasperson, R.E., Slovic, P., 1992. The socialamplication of risk: theoretical foundations and empirical applications. Jour-nal of Social Issues 48, 137160.

Short, J.F., 1984. The social fabric at risk: toward the social transformation of riskanalysis. American Sociological Review 49, 711725.

Slovic, P., Fischhoff, B., Lichtenstein, S., Corrigan, B., Combs, B., 2000. Preference forinsuring against probable small losses: insurance implications. In: Slovic, P.(Ed.), The Perception of Risk. Earthscan Publications, London.

Smith, D.I., 2002. The what, where and how much of ooding in Australia. In: Smith,D.I., Handmer, J. (Eds.), Residential Flood Insurance: The Implications forFloodplain Management Policy. Water Research Foundation of Australia,Canberra, pp. 7398.

Swim, J., Clayton, S., Doherty, T., Gifford, R., Howard, G., Reser, J., Stern, P., Weber, E.,

A.Y. Lo / Global Environmental Change 23 (2013) 12491257 1257Moser, S.C., 2007. More bad news: the risk of neglecting emotional responses toclimate change information. In: Moser, S.C., Dilling, L. (Eds.), Creating a Climatefor Change: Communicating Climate Change and Facilitating Social Change.Cambridge University Press, Cambridge, New York.

Moser, S.C., Dilling, L., 2011. Communicating climate change: closing the science-action gap. In: Dryzek, J.S., Norgaard, R.B., Schlosberg, D. (Eds.), The OxfordHandbook of Climate Change and Society. Oxford University Press, Oxford.

Nelson, D.R., Adger, W.N., Brown, K., 2007. Adaptation to environmental change:contributions of a resilience framework. Annual Review of Environment andResources 32, 395419.

Norgaard, K.M., 2006. We dont really want to know: environmental justice andsocially organized denial of global warming in Norway. Organization & Envi-ronment 19, 347370.

Norgaard, K.M., 2011. Living in Denial: Climate Change, Emotions, and EverydayLife. MIT Press, Cambridge, MA.

Ostrom, E., 2000. Collective action and the evolution of social norms. Journal ofEconomic Perspectives 14, 137158.

Patt, A.G., Schroter, D., 2008. Perceptions of climate risk in Mozambique: implica-tions for the success of adaptation strategies. Global Environmental Change 18,458467.2010. Psychology and Global Climate Change: Addressing a Multifaceted Phenom-enon and Set of Challenges. American Psychological Association, Washington, DC.

The World Bank, 2010. Natural Hazards, Unnatural Disasters: The Economics ofEffective Prevention. The World Bank, Washington, DC.

Warner, K., Ranger, N., Surminski, S., Arnold, M., Linnnerooth-Bayer, J., Michel-Kerjan, E., Kovacs, P., Herweijer, C., 2009. Adaptation to Climate Change: LinkingDisaster Risk Reduction and Insurance. United Nations International Strategyfor Disaster Reduction Secretariat, Geneva, Switzerland.

Wolf, J., Adger, W.N., Lorenzoni, I., Abrahamson, V., Raine, R., 2010. Social capital,individual responses to heat waves and climate change adaptation: an empiri-cal study of two UK cities. Global Environmental Change 20, 4452.

Wolf, J., Allice, I., Bell, T., 2013. Values, climate change, and implications foradaptation: evidence from two communities in Labrador, Canada. GlobalEnvironmental Change 23, 548562.

Wong, K.-K., Zhao, X., 2001. Living with oods: victims perceptions in Beijiang,Guangdong, China. Area 33, 190201.

Zaleskiewicz, T., Piskorz, Z., Borkowska, A., 2002. Fear or money? Decisions oninsuring oneself against ood. Risk Decision and Policy 7, 221233.

The role of social norms in climate adaptation: Mediating risk perception and flood insurance purchaseIntroductionSocial norms: a source of behavioural distortions?Relationships between risk perception, social norms, and behaviourFeedback mechanismsMediation by what?

Study designResearch questionsMeasurementsAdministration of the survey

Research findingsCharacteristics of the sampleModel variablesPath analysis

DiscussionsConclusionsAcknowledgementsReferences