The two faces of social capital in private flood mitigation: opposing effects on risk perception, self-efficacy and coping capacity

Abstract

Impacts of flooding are expected to increase, most notably in residential areas. As a consequence, private households are increasingly encouraged to engage in private flood mitigation complementary to public measures. Despite the growing literature on private flood mitigation, little is known about how social capital influences households’ perception of and coping with flood risks. This study draws on survey data of 226 flood-prone households in two Austrian Alpine municipalities, both recently affected by riverine flooding. We show that social capital cuts both ways: on the positive side, social capital increases perceived self-efficacy and provides critical support during and most notably after flood events. On the negative side, social capital reduces flood risk perceptions of private households. While social ties are effective when responding to and recovering from floods, the expectation of social support downplays risk, making precautionary action by households less likely. The results also show that flood-affected households receive more social support than they provide to others. In the long-run, this can lead to a problematic reciprocity imbalance, challenging the long-term stability of the interpersonal exchanges underlying social capital. Among the various sources of social support, informal social networks (neighbours, friends and relatives) provide the most important workforce in the response and recovery phase of a flood event. It is therefore crucial for flood risk management to recognise and promote the protective quality of social capital alongside conventional structural and non-structural measures.

Keywords:

• flood risk management
• cognitive social capital
• structural social capital
• social capacity
• neighbourly help

1. Introduction

The combined effect of socio-economic development and climate change is expected to aggravate future risks of flooding (Bubeck, Botzen, and Aerts 2012). Recent research shows that the number of residents exposed to flood risks has risen by 114% globally between 1970 and 2010 (UNISDR 2011). Economic losses due to flooding are also predicted to increase over the coming years. For Europe, annual losses from continental flooding are estimated to increase four-fold from currently €4.9 billion to €23.5 billion by 2050 (Jongman et al. 2014).

Private households are most adversely affected by these developments. A recent European-wide assessment of flood impacts reveals that under a medium high emission baseline, with no mitigation or adaptation, about 82% of the expected losses are attributable to residential areas (Feyen and Watkiss 2011). This underscores the need for a more detailed understanding of how households can build capacity to cope with flood events.

Flood mitigation behaviour on the household level has recently received growing attention in research and practice (Laska 1986; Few 2003; Federal Environment Agency 2010; Bubeck et al. 2013). The current typology of flood mitigation measures distinguishes between four groups of protective actions at the household level: adapted building use, structural building measures, flood barriers and flood insurance purchase (Bubeck et al. 2013). Most of these measures are found to be both effective and cost-efficient in reducing flood damage at the household level (Kreibich et al. 2005; Olfert and Schanze 2009; Kreibich 2011).

There are two major factors that determine if households implement private flood mitigation measures. One factor is risk perception, which captures the perceived severity and probability of a flood event. The second factor, coping appraisal, is related to the perceived ability of a household to cope with flood risks. Both factors were found to explain flood mitigation behaviour, suggesting that households are more likely to carry our mitigation measures when they rank high on both risk perception and coping appraisal (Grothmann and Reusswig 2006).

It seems critical for flood risk management to understand the factors that drive risk perception and coping appraisal to better manage private mitigation. Recent research identifies factors such as previous flood experience, risk zone, socio-demographic characteristics and other factors that influence risk perceptions of private households (see Kellens, Terpstra, and De Maeyer 2013 for a comprehensive review). Though most of these studies find significant relationships, a large proportion of the variance of risk perception remains unexplained (Sjöberg, Moen, and Rundmo 2004). This points to the existence of additional factors that have yet been ignored in most of the existing studies.

An important part missing in the equation is the broader social context in which private mitigation decisions are made. Elliott and Pais (2006, 300) emphasise that ‘people respond to disasters not as isolated individuals but as members of overlapping forms of social affiliation’. The relations between actors and institutions are an immanent component of adaptation processes, shaping people’s ability to act collectively when adapting to and recovering from natural disasters (Adger 2003). This points to an important role of social capital – the social norms, trust and networks of a given community – for mitigation and coping actions of private households. Following Pelling and High (2005), we acknowledge the potential of social capital to research adaptive capacity and apply social capital as an analytical lens to examine protective behaviour of flood-prone households.

The aim of this study is to empirically examine how social capital impacts risk perception, self-efficacy and coping capacity at the household level. We distinguish between cognitive and structural social capital to separate out the effects of perceived social support and manifest social support. While the cognitive component of social capital describes what people ‘feel’, the structural component refers to what people ‘do’ (Harpham, Grant, and Thomas 2002). In the first part of this study, we identify an ambivalent effect of cognitive social capital: while perceived social values, norms and networks downplay flood risks they increase perceived coping abilities. In the second part, we explore how the structural component of social capital translates into supportive action among flood-affected households.

The remainder of this paper is organised as follows. Section 2 reviews the drivers of risk perception and self-efficacy. In Section 3, we discuss the concept of social capital and how it has been applied in disaster risk research. Section 4 provides information on study design, methods and data collection. Results for the cognitive and structural component of social capital are presented in Section 5. In Section 6, we conclude and discuss implications for flood risk management and future research.

2. Determinants of risk perception and self-efficacy

Whether an individual engages in protective action depends on his or her subjective assessment of risks and coping options. According to the Protection Motivation Theory (Rogers 1975, 1983), individuals at risk separately evaluate the threat (risk perception) and their coping abilities (including self-efficacy). These two cognitive evaluation processes result in either high or low protection motivation, which, depending on actual barriers, may or may not prompt individuals to take protective action (Bubeck et al. 2013).

The combined effect of risk perception and perceived coping ability also determines if individuals consider to take protective action or if they tend towards non-protective responses (Floyd, Prentice-Dunn, and Rogers 2000; Milne, Sheeran, and Orbell 2000). Grothmann and Reusswig (2006) show that both risk perception and perceived coping ability need to be high to prompt protective actions among households at risk of flooding. In contrast, individuals with high levels of risk perception and low levels of perceived coping ability are more likely to adopt non-protective responses such as wishful thinking, denial or fatalism (Rippetoe and Rogers 1987).

In the first part of this study, we examine the impact of social capital on these two key predictors of private flood mitigation behaviour. We use risk perception to capture households’ risk appraisals and self-efficacy to measure perceived coping abilities of flood-prone households. We further add socio-demographic characteristics and an indicator for objective flood risk, both considered to influence risk perception and self-efficacy.

2.1. Risk perception

Risk perception describes how an individual ‘assesses a threat’s probability and damage potential’ (Grothmann and Reusswig 2006, 104). A high level of flood risk perception (as the conjoint measure of probability and severity of a flood event) is one of the key requirements for individuals to take protective action (Grothmann and Reusswig 2006; Terpstra and Lindell 2013). In a recent review on factors that influence flood risk mitigation behaviour (Bubeck, Botzen, and Aerts 2012) and in an expert panel study on flood risk perception by Becker, Aerts, and Huitema (2014), the authors argue that affective elements, such as fear or worry about flooding, are also important in shaping risk perceptions.

2.2. Self-efficacy

Self-efficacy refers to ‘beliefs in one’s capabilities to organize and execute the courses of action required to produce given attainments’ (Bandura 1997, 3). Alongside the two factors perceived response-efficacy and response costs, self-efficacy is an integral component of coping appraisal. While the former two describe how a person assesses a specific response, self-efficacy captures how a person perceives his or her general ability to protect him – or herself against a certain threat (Becker, Aerts, and Huitema 2014). A study on flood risk mitigation behaviour in Germany (Grothmann and Reusswig 2006) and a similar study in three flood-prone regions in France (Poussin, Botzen, and Aerts 2014) show that self-efficacy is one of the most powerful predictors of risk mitigation behaviour among the three factors related to coping appraisal.

2.3. Previous flood experience

Given that experience with hazards ‘makes people uniquely aware of their vulnerability to a disaster’s consequences’ (Terpstra 2011; 1659), previous flood experience is regarded as an influential factor for risk perception (Weinstein 1989). Previous flood experience drives perceived risks (Siegrist and Gutscher 2006; Lindell and Hwang 2008; Kellens et al. 2011) since flood victims associate different emotions with floods, compared to individuals with no prior flood experience. Zaalberg et al. (2009), for instance, show that Dutch flood victims perceive themselves as more vulnerable to flooding and are thus more likely to implement mitigation measures than non-victims.

2.4. Socio-demographic and objective risk factors

Several studies include socio-demographic factors such as age, gender or income alongside objective risk factors such as geographical location to examine how individuals perceive and cope with flood risks. The existing literature reveals mixed findings on age, suggesting either a positive (Kellens et al. 2011 in a study on flood risk perception in Belgium) or negative (Miceli, Sotgiu, and Settanni 2008 in a study on flood risk perceptions in Italy) correlation with flood risk perception. In terms of flood risk appraisal, women tend to perceive themselves at higher risk than men (Lindell and Hwang 2008; Kellens et al. 2011). For the case of Texas, Lindell and Hwang (2008) find that household income is negatively related to perceived flood risk, suggesting that more affluent households perceive themselves at lower flood risk than less affluent households. Flood risk perceptions of private households are also influenced by objective risk factors. Studies from the Netherlands and in Switzerland identify distance to river and building elevation (Botzen, Aerts, and Van Den Bergh 2009) as well as designated risk zones (Siegrist and Gutscher 2006) as objective factors that drive risk perceptions of private households.

3. Social capital and risk mitigation

The vulnerability of a community to hazards, such as flooding, is partially determined by the social fabric of the place (Cutter, Boruff, and Shirley 2003). Bourdieu (1986, 248) refers to social fabric as social capital which is ‘the aggregate of the actual or potential resources which are linked to possession of a durable network of more or less institutionalised relationships of mutual acquaintance or recognition’. Social capital creates value through the ‘features of social organization, such as trust, norms, and networks, that can improve the efficiency of society by facilitating coordinated action’ (Putnam, Leonardi, and Nanetti 1993, 167). Existing theories of social capital suggest to distinguish between different forms and dimensions of social capital.

3.1. Cognitive and structural social capital

One of the most common approaches is to distinguish between a cognitive and structural dimension of social capital. While the cognitive component is less tangible and captures perceived support, trust, social cohesion and perceived civic engagement, the structural component refers to networks, connectedness, associational life and civic participation (Harpham, Grant, and Thomas 2002; Harpham, Grant, and Rodriguez 2004). In the context of flood risk, cognitive social capital may refer to the support people expect to receive during a flood event, while structural social capital relates to the support people actually experience when the event occurs. Though most empirical studies ignore this conceptual distinction, Harpham, Grant, and Rodriguez (2004) emphasise the importance and possibility of measuring both components separately.

3.2. The role of social capital for adaptation and coping

Pelling and High (2005) propose social capital as an analytical lens to research adaptive behaviour of collectives. To some extent, this notion rests on the assumption that relations between actors and institutions are an integral component of adaptation processes, enabling them to act collectively when adapting to and recovering from natural disasters (Adger 2003). The link between social capital and adaptation is also confirmed by studies that find positive effects of social capital on the adaptive capacity of communities during and in the aftermath of natural disasters (Pelling 1998; Cutter, Boruff, and Shirley 2003; Nakagawa and Shaw 2004).

In flood risk management, a purposeful management of collective action appears key to build resilience among flood-prone communities. More integrated approaches to flood risk management consider risk mitigation as a concerted effort, jointly carried out by public and private actors (Laska 1986; Few 2003; Federal Environment Agency 2010; Bubeck et al. 2013). Given that social capital is fundamental for collective action (Adler and Kwon 2002; Adger 2003; Norris et al. 2008), it seems critical for integrated flood risk management to address the social relations in which mitigation decisions are made.

Social capital plays a considerable role in all phases of the disaster risk management cycle (Koh and Cadigan 2008; Kuhlicke et al. 2011; Aida et al. 2013). In the prevention phase, social trust can streamline decision-making in public flood protection projects, which are more likely to be accepted when backed by the wider community in a participative process. Public knowledge can complement expert knowledge and increase efficiency, particularly when decisions are made in conflicting and uncertain situations (Gamper and Turcanu 2009). Citizen participation, therefore, creates opportunities to fast-track decision-making in flood risk management.

In the preparedness phase, strong social networks can facilitate formal and informal communication of risk and coping options. Risk information from trusted sources, passed on by word of mouth and informal social ties, can make risk campaigns more effective (Norris et al. 2008). Social relations can also be beneficial in the early warning phase, when a speedy dissemination of alerts is most crucial.

In the response phase, flood victims can use social networks to draw upon neighbourly help (Moore et al. 2004) or to find temporary shelter after evacuation. In the recovery phase, social capital can contribute to the efficient use of private and public sector resources (Aldrich 2011). Social ties can also contribute to longer term survival and encourage rebuilding activities (Hawkins and Maurer 2010), which otherwise tend to be costly if sourced from the market. Social ties can further ‘serve as ‘informal insurance’, allowing victims to draw upon pre-existing support networks for financial, informational, and emotional assistance’ (Kawachi, Takao, and Subramanian 2013, 17).

3.3. The ambivalent effects of social capital

Though social capital is generally considered a resource that strengthens adaptive capacity, certain negative effects need to be acknowledged. In a study on adaptation behaviour of elderly to heat waves, Wolf et al. (2010) show that individuals rich of strong ties perceive themselves at lower risk than individuals with less strong ties. Narratives of resilience, passed on among individuals rich of strong ties, reduce their motivation to carry out heat mitigation measures. This however, might be at odds with their actual vulnerability. Wolf et al. (2010, 45) conclude that it is yet ‘unclear in which circumstances it [social capital] may be counter-productive and may increase vulnerability’.

In a study on social capital and post-disaster mental health, Wind, Fordham, and Komproe (2011) find that high levels of cognitive social capital lead to better mental health outcomes. The expectation of social support improves posttraumatic stress disorders, anxiety and depression after disasters. In contrast, structural social capital is associated with an increase of anxiety disorders. There is considerable evidence that cognitive social capital has a positive effect on health outcomes, self reported general health, psychological health and subjective well-being while no such positive effect can be confirmed for structural social capital (Harpham, Grant, and Rodriguez 2004; Yip et al. 2007).

The mixed effects of social capital on risk perception and coping success underscore the need to disaggregate social capital into its two sub-components to separate out the effects of each component, respectively. We therefore draw on Harpham, Grant, and Rodriguez’s (2004) distinction between cognitive and structural social capital to compare the impact of social capital on how flood-prone households perceive and cope with flood risks.

4. Study design and method

4.1. Study setting

We carried out a cross-sectional questionnaire survey in two flood-affected municipalities in Austria, both with similar flood and damage characteristics. Kössen was affected by a riverine flooding in June 2013 and Oberwölz in July 2011. In Kössen, with a population of 4178, the 2013 flood event caused an estimated damage of €40 million (Tyrolean Regional Government 2013). Oberwölz has a population of 993 and is part of the larger region Wölzertal with a population of 3300. The overall flood damage for the region Wölzertal was estimated at €29 million (Government of Styria 2011).

Kössen is situated in the province of Tyrol and the municipality of Oberwölz belongs to the province of Styria. Both case study regions are small rural municipalities located in Alpine valleys where voluntary emergency and relief services, such as fire brigades or rescue organizations like the Red Cross, form the backbone of disaster risk management. When a disaster occurs, volunteer organizations act as auxiliaries to the responsible disaster management authority (which is either the mayor, the head of district administration or the provincial governor). Both case study regions highly depend on volunteer-based risk management because they command small budgets granting little leeway for costly large-scale flood protection infrastructure. In small municipalities like Kössen and Oberwölz, organised volunteers typically constitute social capital beyond flood prevention and response, as they are firmly entrenched in community life.

4.2. Study population and data collection

Self-completion questionnaires were distributed as an insert in municipal newspapers among all households in Kössen and Oberwölz, to be returned free of postage. This method allowed us to approach virtually all households in the two municipalities. All data were collected in July 2013 in Kössen and in December 2013 in Oberwölz. The head of each household was asked to complete the questionnaire on behalf of the other household members.

Our total sample consists of 226 respondents, including 139 respondents from Kössen and 87 from Oberwölz. Table A1 (see Appendix) shows that most sample characteristics conform fairly well to the population except for gender. Further, there are no risk zone data available on a municipal level in Austria, which makes it difficult to determine the exact share of flood-prone households in the population; still, extrapolating from surrounding regions, households located in risk zones are presumably underrepresented in the Kössen sample and overrepresented in the Oberwölz sample. We therefore tested for a potential effect of selection bias by adding two interaction terms (gender*social capital, risk zone*social capital) to our models in Table 2. None of the terms was statistically significant, raising confidence in the assumption that our results are not biased by restricted representativeness of the sample. Note that sample sizes vary in the respective analyses, since not all respondents answered all items in the questionnaire; however, the distribution of missing values does not systematically bias our results.

4.3. Measures

We primarily used Likert scales for item assessment and computed factor scores for the conjoint measures risk perception (as a combined measure, and separately for the cognitive and affective dimension), self-efficacy and cognitive social capital.

All items yield high factor loadings on the corresponding factors. The original item wording was in German and translated for the present paper. For the exact item wording, descriptive statistics and factor loadings see Table 1.

Table 1. Item wordings, descriptive statistics and factor loadings.

Factor/item	Response scale/coding	N	M	SD	Factor loading
Risk perception cognitive
A potential flood threatens my health	1 = strongly agree; 5 = strongly disagree	221	3.38	1.38	0.78
A potential flood is a threat to my possessions	1 = strongly agree; 5 = strongly disagree	223	2.56	1.48	0.86
A potential flood threatens my quality of life	1 = strongly agree; 5 = strongly disagree	223	2.49	1.41	0.90
When do you expect the next flooding to occur at your home?	2.5 = in 1–5 years; 7.5 = in 6–10 years; 15.0 = in 10–20 years; 25.0 = in 20–30 years; 40.0 = in 30–50 years; 78.0 = in 50–100 years; 200 = never^*	207	3.96	2.44	0.68
Risk perception affective
I’m worried about the potential threat of flooding	1 = strongly agree; 5 = strongly disagree	219	2.40	1.32	0.96
I am afraid of a potential flood	1 = strongly agree; 5 = strongly disagree	223	2.58	1.38	0.96
Self-efficacy
It is too difficult for someone like me to protect against flooding	1 = strongly disagree; 5 = strongly agree	220	3.30	1.40	0.91
I feel helpless over a potential flood	1 = strongly disagree; 5 = strongly agree	219	3.40	1.45	0.91
Cognitive social capital
Trust: Generally speaking, would you say that in your community	+2 = most people can be trusted	210	1.08	0.88	0.66
	−2 = you can’t be too careful when dealing with others
Fairness: Generally speaking, would you say that in your community	+2 = most people try to be fair	203	0.73	1.04	0.66
	−2 = most people take advantage of others if they can
Helpfulness: Generally speaking, would you say that in your community	+2 = people try to help others most of the time	206	1.23	0.83	0.72
	−2 = people just look out for themselves most of the time
Consideration: Generally speaking, would you say that in your community	+2 = people consider the needs of others	204	0.52	1.07	0.79
	−2 = people foremost consider their own needs
Participation: Generally speaking, would you say that in your community	+2 = everyone can participate when important decisions are made	205	–0.13	1.24	0.55
	−2 = few people can participate when important decisions are made
Community involvement: Generally speaking, would you say that in your community	+2 = many people participate in formal groups (e.g. associations, churches or organisations)	211	1.22	0.88	0.66
−2 = only few people participate in formal groups (e.g. associations, churches or organisations).
Flood experience
Previous flood experience	# of times personally flooded; ranging from 0 to 3.	221	0.24	0.61	n/a
Socio-demographic and objective risk variables
Age	Continuous	209	51.75	13.84	n/a
Gender	0 = male; 1 = female	219	67% male
Income (monthly household income)	550 = under €1100; 1350 = €1101–€1600; 2100 = €1601–€2600; 3300 = €2601–€4000; 4750 = €4001–€5500; 10,000 = over €5500^*	164	2598.48	1930.51	n/a
Risk zone	0 = no; 1 = yes	226	27% in risk zones

* A continuous variable was computed by taking the midpoint of the answered intervals.

4.3.1. Cognitive and affective risk perception

We adopt the approach put forward by Grothmann and Reusswig (2006) to measure risk perception and disaggregate risk perception into a cognitive and affective component. The cognitive component consists of (1) perceived probability, measured as the respondent’s perceived likelihood of future flooding, and (2) perceived severity, directed towards the expected consequences of flooding for the respondent’s quality of life, health and possessions. The affective component of risk perception captures emotional states such as fear and worry associated with potential flooding.

4.3.2. Self-efficacy

Self-efficacy is the perceived ability of an individual to perform or carry out a protective response (Grothmann and Reusswig 2006). We operationalise self-efficacy as a conjoint measure of two items, both capturing a household’s perceived ability to protect against flooding.

4.3.3. Previous flood experience

An individual’s reaction to risk partially depends on the frequency of hazardous events (Sjoberg 2000). People attribute more importance to risk mitigation the more often they are exposed to a certain risk. We measure previous flood experience as the number of experienced floods ranging from 0 to 3.

4.3.4. Socio-demographic and objective risk variables

In accordance with existing research on flood mitigation behaviour, we add relevant socio-demographic variables to our models (see Section 2.4). We include items on age, gender and household income to control for the impact of socio-demographic factors on risk perception and self-efficacy. Objective risk is captured by risk zone, capturing whether respondents lives in a flood risk zone or not. We follow the official Austrian classification scheme used in flood risk and hazard maps, and asked respondents to indicate if their building is located within a 30-year flood, 100-year flood, yellow or red risk zone.

4.3.5. Cognitive social capital

Given the lack of a universally accepted approach for the measurement of social capital (Rosenfeld, Baumer, and Messner 2001), we model our cognitive component analogous to previous studies and include measures of perceived trust, fairness, helpfulness, consideration, participation and community involvement (Lesser 2000; Putnam 2000; Szreter and Woolcock 2004; Smith et al. 2013; Wang et al. 2014).

4.3.6. Structural social capital

Structural social capital is measured as provided and received social support, in line with Flores, Carnero, and Bayer (2014) who identify social support as one of the integral components of structural social capital. Respondents were asked to give estimates on provided and received help measured in person-days during the response and recovery phase. Three different categories of social support are distinguished to capture the directions and sources of social support: (1) social support provided by households, (2) social support received by households (including support received from voluntary workers, neighbours, friends and relatives) and (3) social support amongst households’ members themselves. Table 3 further breaks down the sources of social support into response and recovery phase.

5. Results

5.1. The impact of social capital on risk perception and self-efficacy

We use multiple linear regression analysis to examine the relationships between six predictors and four dependent variables capturing dimensions of risk perception and self-efficacy. The six variables age, gender, income, risk zone, previous flood experience and cognitive social capital enter the regressions as predictors. An additional dummy variable to control for the municipalities Kössen and Oberwölz is not significant in any of the models and therefore dropped from the regressions. The four variables risk perception combined, risk perception cognitive, risk perception affective and self-efficacy are regressed on our full set of predictors. We run a series of four separate regression analyses to test for a unique effect of social capital while controlling for the effects of the other predictors. Table 2 summarises the regression results. Note that higher values for risk perception/self-efficacy are associated with lower risk perception/self-efficacy. Our set of predictors explains between 20 and 27% of the variance, indicating a good fit of our models. Similar models on risk perception explain at most 20–25% of the total variation in perceived risk (Sjöberg, Moen, and Rundmo 2004).

Table 2. Regression results for risk perception and self-efficacy.

	Risk perception combined	Risk perception cognitive	Risk perception affective	Self-efficacy
Age	.16^*	.16^*	.07	.04
Gender	–.02	–.03	–.04	.16^**
Income	.27^***	.20^**	.27^**	–.20^**
Risk zone	–.39^***	–.41^***	–.31^***	.31^***
Previous flood experience	–.14^*	–.17^**	–.13^*	.12
Cognitive social capital	.16^**	.11^*	.15^*	–.18^**
R2	.27	.26	.20	.22
R2 adjusted	.24	.23	.17	.18
F	7.83; p < .001	7.64; p < .001	5.76; p < .001	6.18; p < .001
N	132	136	142	142

^* Standardised coefficients: p < .10;

^** p < .05

^*** p < .01

Note: Sample size varies due to missing values. Note that higher numeric values for risk perception/self-efficacy are associated with lower risk perception/self-efficacy.

Our results show that cognitive social capital has a significant effect on risk perception, even when controlling for socio-demographic characteristics, previous flood experience and objective risk. Households that judge their social environment positive and supportive, tend to perceive themselves at lower flood risk than households that evaluate their social environment less positive. This reveals a potential negative side of social capital: expected social support reduces risk perception, consequently making flood-prone households less likely to adopt protective measures.

The effects of the remaining predictors are consistent with existing findings. Age is negatively related to flood risk perception, suggesting that risk perception decreases with age. While gender is insignificant, household income has a significant effect on risk perception, suggesting that wealthy households tend to perceive themselves at lower risk than households with lower income. The significant coefficient of risk zone confirms that households in risk zones perceive themselves at greater risk than households outside flood-prone areas. Previous flood experience is also significantly related to risk perception, suggesting that risk perceptions are higher among households with past flood experience than for households with no prior flood experience.

Cognitive social capital remains a marginally significant (p < .10) predictor for the cognitive and affective components of risk perception. Our results show that social capital reduces both the cognitive and affective dimension of risk perception. Households with high levels of social capital tend to evaluate flood probability and severity of consequences lower than households with lower levels of social capital. Correspondingly, high levels of social capital are associated with weaker feelings of fear and worry towards a potential flood. The effects of the control variables age, gender, income and risk zone remain similar as in the combined model.

Turning to self-efficacy, we find that cognitive social capital is a significant factor for predicting how households assess their own ability to cope with flood risks. The higher social capital, the more a household feels capable to deal with flood risks. This reflects the positive side of social capital: a supportive social milieu makes households more confident and optimistic towards their own ability to cope with flood risks. While age becomes insignificant, men perceive themselves as being better prepared against flooding than women. In a similar vein, higher income is related to stronger self-efficacy beliefs. Perceived self-efficacy is lower among households in risk-prone areas, possibly because they acknowledge some degree of residual risk. Experience with previous flooding does not seem to play a significant role for how households evaluate their coping abilities.

Our results show that the standardised coefficients of social capital are of similar magnitude for both risk perception (.16) and self-efficacy (−.18). This suggests that high levels of social capital reduce risk perception and increase perceived self-efficacy to a similar extent. From a protection motivation theory perspective, both perceptual variables need to be high in order to prompt mitigative action. The fact that the standardised coefficients for risk perception and self-efficacy are about the same size, suggests that the indirect effects of social capital on mitigation behaviour might cancel each other out. Even if social capital positively contributes to perceived self-efficacy, the negative effect on risk perception makes households less likely to carry out flood mitigation measures.

Overall, we find a significant effect of cognitive social capital on both risk perception and self-efficacy across our models. Social capital remains a significant predictor, even when controlling for socio-demographics, flood experience and risk zone. This suggests that neither risk perception nor self-efficacy can be explained solely by common socio-demographic characteristics, past flood experience or objective risk variables. Perceptual factors such as perceived social trust, reciprocity and community engagement are further factors that households take into account when judging flood risks and evaluating their coping abilities.

5.2. The impact of structural social capital on coping capacity

In this section, we examine how structural social capital unfolds during and after flooding. We contrast provided and received social support measured in person-days to explore the reciprocal dynamics of structural social capital in both the response and recovery phase of a flood event (see Table 3). We also examine if damage severity is related to enacted social support. Since measures of social support were not included in the survey in Oberwölz, we only use the Kössen subsample (n = 139) in this part of the study.

Table 3. Provided and received social support during response and recovery phase measured in person-days.

Provided and received social support across phases (person-days)	M	SD
Response phase	35.84	67.76
Total support provided	5.41	14.81
Total support received	17.25	39.54
From voluntary workers	5.38	15.02
From neighbours, friends and relatives	11.86	28.43
Support among household members	13.18	33.48
Recovery phase	99.23	189.17
Total support provided	14.40	58.17
Total support received	47.96	89.95
From voluntary workers	10.49	27.28
From neighbours, friends and relatives	37.47	71.79
Support among household members	36.88	98.59
Total sum of social support	135.07	201.53
Total support provided	19.81	63.24
Total support received	65.21	102.30
From voluntary workers	15.87	31.83
From neighbours, friends and relatives	49.34	81.49
Support among household members	50.05	102.56

Note: Calculated for all flood-affected households (n = 73) based on the subsample of Kössen.

In the response phase, each household provided 5.41 and received 17.25 person-days of social support on average. When breaking down total support received, we find that the majority of external help was provided by neighbours, friends and relatives (M = 11.86). This suggests, in quantitative terms, that the group of neighbours, friends and relatives is the most utilised social resource that flood victims rely on in the response phase. Household members spent considerable less time for support amongst themselves (M = 13.18) compared to the total amount of support that they received from others (M = 17.25). Overall, these numbers indicate that household response depends to a large extent on outside help, mobilised through formal and informal networks.

In the recovery phase, each household provided 14.40 and received 47.96 person-days of social support on average. When breaking down total support received, we find that the largest share of external support was provided by neighbours, friends and relatives (M = 37.47). Overall, total support received (M = 47.96) is considerably higher than social support among household members themselves (M = 36.88). As in the response phase, households largely depend on external help provided by formal and informal sources when recovering from a severe flood event.

These numbers uncover a major imbalance between provided and received social support both during and after flooding. On average, each household provided 19.81 person-days of support to others, while receiving more than three times as much support from others (65.21 person-days). The provision and utilisation of social support lie at the heart of reciprocity, a key component of social capital. Norris et al. (2008, 138) underscore the importance of reciprocity and warn that ‘… being constantly on the receiving end of support exchanges can threaten self esteem … whereas being constantly on the providing end creates stress and burden’. Our results reveal a large discrepancy between provided and received social support, possibly limiting the continued provision of social support if several flood events were to happen in quick succession.

We further expect that heavily affected households rely more strongly on structural social capital (i.e. receive more external social support on average) than households that were less affected by a flood. An independent-samples t-test compares received external social support between households that experienced either low or high damage severity. Damage severity refers to households’ estimates of material losses related to buildings and household goods. We use a median split on the continuous variable damage severity to divide the sample into two subgroups with more/less than 80,000 € flood damage (high vs. low damage). We further discriminate between social support during the event (response phase) and after the event (recovery phase).

Social support in the response phase does not differ significantly between high-damage vs. low-damage households (see Table 4). In contrast, received social support in the recovery phase was significantly higher for households that suffered high damage than for households that experienced low damage. These results suggest that structural social capital is an important resource for the recovery of households, particularly for those that suffered severe material damage from flooding.

Table 4. Received social support during response and recovery phase by damage severity measured in person-days.

		Damage low		Damage high
Received social support in person-days	n	M	SD	M	SD	t	df	p
Response phase	66	15.00	36.93	23.67	46.09	–.85	64	.40
Recovery phase	66	15.35	24.90	97.52	122.45	–3.61	31	.00^***

*** p < .01.

Notes: Based on Levene’s test of equality of variance we assumed equal variances for testing differences in the response phase (F = 1.59, p = .21) and unequal variances in the recovery phase scenario (F = 23.24, p < .00). The degrees of freedom were adjusted from 66 to 31 for the latter. Damage low/high was dichotomised by a median split at 80,000 € flood damage.

6. Discussion and conclusions

In this paper, we examined the role of social capital for risk perception, self-efficacy and coping capacity based on a survey among 226 flood-prone households in Austria. Our findings show that social capital cuts both ways: while it lowers flood risk perceptions of private households, it increases perceived self-efficacy and positively contributes to coping capacity. Given that high levels of flood risk perception are a prerequisite for protective action, the negative effect of social capital on risk perception makes households presumably less likely to carry out flood mitigation measures.

On the positive side, our results show that a trustworthy social environment boosts households’ belief in their own coping abilities and that structural social capital is an important ingredient for response and recovery actions. The exchange dynamics point to a reciprocity bias of provided and received external social support, which directly follows from our finding that flood-affected households received considerably more social support from others than they provided to others. Furthermore, households that experienced high flood damage relied more strongly on external social support during the recovery phase than less affected households. We conclude from our findings on structural social capital that informal social ties with neighbours, friends and relatives are an essential factor to build community resilience against extreme flooding.

6.1. Risks and benefits of social capital in flood risk management

Our findings provide useful insights for future flood risk management and risk communication. A detailed understanding of the dynamics of social capital is particularly important for integrated flood risk management that emphasises the importance of collective actions, carried out by both public and private actors.

One of the key challenges for future flood risk management will be to balance out cultivating social capital and instilling a sense of security among flood-prone households. Our findings show that expected social support might downplay risk, making it less likely that households engage in precautionary action. This is also in line with Harpham, Grant, and Rodriguez (2004) who find that cognitive social capital is related to perceived security and self-esteem. Risk managers need to address a (potential) security bias as expectations of social support might easily become unrealistic. This could quickly lead to a lack of sufficient social support during a severe flood event. On top of unrealistic expectations, the magnitude of a flood event might also play a key role here. In large-scale flood events, demand for social support can easily exceed the level of social support a given community can mobilise. Consequently, risk communicators need to consider these ambivalent effects of social capital when crafting communication strategies that involve narratives of social resilience. Our findings suggest that carefully designed flood risk campaigns are essential to achieve two major outcomes: first, to prompt mitigative action and second, to prevent the proliferation of a false sense of security among households at risk of flooding.

Private households are most likely to take protective actions when both risk perception and self-efficacy are high. Our findings, however, show that greater stocks of social capital are associated with lower levels of risk perception and higher levels of self-efficacy. This combination undermines the intention to take private flood mitigation measures, making the adoption of non-protective responses such as wishful thinking, denial or fatalism more likely. The risk of prompting non-protective responses underscores the importance of achieving a ‘balance [between] risk perception and coping appraisal for the actual risk-reducing response’ (Becker, Aerts, and Huitema 2014, 18). This points to the challenge for flood risk managers to build social capital and foster self-efficacy without reducing flood risk perception. This, for instance, can be achieved through communication strategies that underscore the importance of social support, while highlighting the need for additional measures to reduce residual risk.

There are different ways to build social capital among flood-prone households. Encouraging civic participation is one of the most practical ways to promote social capital (Pelling 1998). First, civic participation is crucial to promote a trustworthy and inclusive environment, and second, community involvement may strengthen the acceptance of future flood mitigation measures among residents. Civic participation might also influence risk perceptions since a direct involvement in flood mitigation projects can help households to better grasp and evaluate their own vulnerability to flooding. Another approach to social capital formation is to promote engagement in volunteering activities, which is regarded as a key indicator of social capital and important for promoting trust (Putnam 2000). Volunteering has the potential to increase the number of social ties between residents, leading to higher stocks of structural social capital, a component that is, according to our results, highly relevant in the response and recovery phase of flooding.

The sheer amount of informal support provided by neighbours, friends and relatives underscores the importance of maintaining and promoting the structural features of social capital. Flood risk managers may focus on structural social capital for two reasons. First, providing social support to flood affected households could serve as a vehicle to increase risk perceptions of unaffected households, since helping others to recover from flood damages can be a substitute for one’s own experience (Siegrist and Gutscher 2006). Second, structural social capital could increase perceived self-efficacy. According to Zaalberg et al. (2009), received social support results in higher efficacy beliefs of adaptive actions. This is in line with our results, indicating that credible promises of social support raise perceived self-efficacy. Here, the key is for risk managers to focus on the structural qualities of social capital to increase both, risk perception and self-efficacy to prompt mitigative actions among flood-prone households.

Finally, our findings reveal the limits to reciprocity. The fact that flood-affected households received more than three times as much social support from others than they provided to others points to an imbalance of reciprocal social exchanges during a flood event. Though this might not cause a problem in the short-run, this imbalance may turn into a threat in the long-run. Households that are constantly on the receiving end of support exchanges might be less likely to reduce their vulnerabilities in the future due to decreasing self-efficacy beliefs. On the positive side, social support provided by one household to another creates social obligations that can be claimed when needed. In the context of flood risk protection, the interplay of creating and enforcing social obligations might even constitute a protective quality by itself. Therefore, in the aftermath of a flood event, official recognition of mutual supportive acts could make reciprocal protective actions more visible and more binding for the future.

6.2. Limitations and future research

There are several limitations to this study that provide avenues for future research. Our analyses rely on self-reported data, a potential source of response bias. Subjective assessments, however, require self-reports by their very nature. We improved measurement quality by reverse-coding questions and constructing multi-item factor scores. Our sampling strategy allowed us to reach a large number of households in the two municipalities, but we have to assume that the low response rate implies some self-selection bias, which is an issue that arises for many mail questionnaire surveys. We therefore caution against unrestricted generalisation of our findings to the larger population or to other contexts. Finally, we welcome future studies with more resources to conduct longitudinal analyses, so to substantiate causal effects and shed light on the temporal dynamics between social capital, risk perception and precautionary behaviour.

Future research on private flood mitigation could benefit from including social capital as an explanatory factor in socio-psychometric models. Studies may look deeper into the concept of social capital and assess how different forms of social capital such as bonding, bridging or linking capital influence flood risk perception, self-efficacy and coping capacity. It might also be worthwhile to compare the effect of social capital on intended and actual flood mitigation behaviour. To determine the economic value of social capital, future research may assess to what extent social capital reduces damage potential, allowing to contrast it with coping capacities derived from other, non-structural and structural measures by means of cost–benefit assessment. The relevance of social capital, however, seems to go beyond the field of flood risk research. Risk research on other natural hazards such as earthquakes, wildfires or storms would also benefit from a social capital perspective to more fully explain threat and coping appraisals.

Additional information

Funding

This research was supported by the Austrian Climate and Energy Fund and carried out within the Austrian Climate Research Programme (ACRP).

Appendix

Table A1. Population and sample characteristics

Table

	Gender		Age (years)					Monthly household income (€)						Households located in risk zones
	M	F	20–34	35–49	50–64	65–79	>79	<1100	1100–1600	1601–2600	2601–4000	4001–5500	>5500
Kössen population (%)	49	51	25	27	27	16	5	6	13	31	33	13	4	n/a
Kössen sample (%)	63	37	12	30	35	21	2	12	21	31	27	6	4	22
Oberwölz population (%)	47	53	23	27	25	16	9	7	14	21	34	17	7	n/a
Oberwölz sample (%)	80	20	14	37	38	10	1	12	19	30	23	11	5	36

Notes: Socio-demographic and income data are taken from Statistics Austria (2011, 2012).

Risk zoning based on Natural Hazard Overview & Risk Assessment Austria (HORA 2015).

Total population includes residents aged 20 or over.

Percentages may not add to 100 due to rounding.