Assessing social capacity and vulnerability of private households to natural hazards – integrating psychological and governance factors

People are unequally affected by extreme weather events in terms of mortality, morbidity and financial losses; this is the case not only for developing, but also for industrialized countries. Previous research has established indicators for identifying who is particularly vulnerable and why, focusing on socio-demographic factors such as income, age, gender, health and minority status. However, these factors can only partly explain the large disparities in the extent to which people are affected by natural hazards. Moreover, these factors are usually not alterable in the short to medium term, which limits their usefulness for strategies of reducing social vulnerability and building social capacity. Based on a literature review and an expert survey, we propose an approach for refining assessments of social vulnerability and building social capacity by integrating psychological and governance factors.


Introduction
The impacts of events like the European heat wave in 2003, the landfall of Hurricane Katrina near New Orleans in 2005, and the flooding of large parts of Queensland (Australia) in 2010/2011 demonstrated that industrialized countries are not immune to high losses from extreme weather events.Also, like in developing countries, large disparities exist in social vulnerability to natural hazards; people are unequally affected in terms of adverse health effects, loss of life and property (e.g.Curtis et al., 2007).The frequency of weatherrelated extremes such as heavy precipitation events and heat waves will very likely increase due to climate change (Meehl et al., 2007), so that in all likelihood more people will be affected in the future.In order to understand potential consequences of natural hazards for humans, to better understand who is particularly affected and why, and to reduce negative impacts of natural hazards, the concept of social vulnerability was developed.There are a vast number of definitions of social vulnerability (cf.Tapsell et al., 2010); following a definition of vulnerability common in natural hazards and disaster research, we understand social vulnerability as "the characteristics of a person or group and their situation that influence their capacity to anticipate, cope with, resist, and recover from the impact of a natural hazard1 " (Wisner et al., 2004, p. 11).Many attempts have been made to measure social vulnerability (for comprehensive overviews see Adger et al., 2004;Adger, 2006;Tapsell et al., 2010; also see Hufschmidt, 2011); so far, social vulnerability assessments have relied mainly upon socio-demographic indicators such as age, gender, and household income (see Sect. 2.1).Sociodemographic indicators refer to inherent aspects of vulnerability that are rather static, at least in the short to medium term: people cannot become younger or change their ethnicity, and increasing poor people's income is an essential yet long-term challenge in vulnerability reduction.Thus, assessments limited to socio-demographic indicators are also limited regarding their usefulness for vulnerability reduction (e.g. through supporting decisions on monetary flows and particular consideration of vulnerable population groups in emergency plans).
Going beyond socio-demographic indicators for the assessments of social vulnerability can substantially contribute J. Werg et al.: Assessing social capacity and vulnerability of private households to natural hazards to a better understanding of exactly which capacities people lack that enable them to deal with natural hazards.Such knowledge is the prerequisite for the effective building of social capacity and thereby reducing people's vulnerability.Psychological factors such as the perception of being at risk2 from a hazard, the knowledge of self-protection and the motivation to actually carry out such measures are important elements of social capacity.At the same time, governance factors creating an environment that protects those that are vulnerable from the adverse impacts of natural hazards and increases people's ability to protect themselves, are essential components of social capacity building and vulnerability reduction.Moreover, psychological and governance factors are changeable -meaning they can potentially be altered within the short to medium term, and the power to do so lies at least partially with local policy makers and private persons.The purpose of our research was to identify a set of indicators that would allow a refinement of common assessments of social vulnerability and offers possible starting points for vulnerability reduction and social capacity building at the local level.Using the term local level, we refer to the household and community levels, being the most crucial for loss prevention behaviour and measures (cf.Hufschmidt, 2011).
Following the process of our research, the paper is organized as follows: first we will give an overview of the results of our literature review of indicators of social vulnerability and social capacity building.Second, we will describe the expert survey conducted to assess the perceived relevance of the indicators reviewed and outline the survey's results.The last part of the paper will be dedicated to the conclusions; a possible nested approach to detecting and reducing vulnerability and building social capacity, and future research steps will be discussed.

Literature review
Three questions were guiding our literature research: on the individual level, what are the factors that make people vulnerable to natural hazards?Which factors increase social capacity or the likelihood of people taking action to reduce their own vulnerability?On the community level, which factors influence the vulnerability of a community and its residents?
In dealing with all of these questions, we were interested in indicators 3 representing those factors.Possible answers to those questions can be found in research on social vulnerability to natural hazards, social vulnerability to climate change, social vulnerability reduction, loss prevention and social capacity building.Our focus is on studies of weather-related extremes (heat waves, floods (fluvial, pluvial, tidal), storms).With a very likely increase in such extremes due to climate change (Meehl et al., 2007), the question of how people deal with these natural hazards is also increasingly relevant.In presenting the results of the literature review, we distinguish between (a) different natural hazards (categories heat, flood, storm) and (b) different categories of indicators, i.e., traditional indicators (those commonly used in vulnerability assessments), psychological and governance indicators.For the traditional and the psychological indicators, we have only included surveys or case studies that analyse the correlation between certain factors or indicators on the one hand and impacts of natural hazards (mortality, morbidity, financial losses) or loss prevention behaviour on the other hand.By correlation we refer to an observed quantitative relation indicated by statistical correlation, factor analysis, or percentage (e.g.75 % of the fatalities were elderly).Some studies solely rely upon secondary statistical data, assuming a strong positive correlation between certain indicators and vulnerability without analysing actual correlations of those indicators to the outcomes (e.g.fatalities) in a certain event (for discussions of different approaches cf.Hinkel, 2011;Kuhlicke et al., 2011a;Tapsell, 2005).Such studies were not included for the traditional and psychological indicators.However, for the governance indicators, such a strict limitation would have led to very few results.Therefore, the aforementioned types of studies were also included for the research on the governance factors, as were studies based on a "theoretical understanding of relationships" (Tapsell et al., 2010, p. 27).

Traditional indicators -focus on assessing inherent vulnerability
Previous research on social vulnerability has focused on indicators such as age, gender, and household income (e.g.Brooks et al., 2005;Cutter et al., 2003;Demetriades and Esplen, 2010;Gladwin and Peacock, 1997;Jonkman et al., 2009;Masozera et al., 2007;Mearns and Norton, 2010;Morrow, 1999;Reid et al., 2009;Tapsell et al., 2002).These traditional indicators are used individually or as part of a vulnerability index (e.g.Fekete, 2009;Cutter et al., 2003; for an overview on four major indices see Birkmann, 2007).Vulnerability assessments based on traditional indicators follow well-established procedures, with the indicators referring to available data such as census data.Common outputs of such vulnerability assessments are vulnerability maps, highlighting vulnerable areas of different scales (e.g.country, county, city) (e.g.Cutter and Finch, 2008).Other studies empirically test assumed linkages between indicators of socio-economic characteristics and demographic factors on the one hand and vulnerability on the other (e.g.Brooks et al., 2005;Cutter et al., 2003;Fekete, 2009).Table 1 lists those traditional indicators, for which we found surveys or case studies testing the correlation between those indicators and financial losses, mortality, morbidity and loss prevention behaviour.

Psychological indicators -focus on correlates of vulnerability reduction behaviour and social capacity building
In the context of our work, psychological indicators refer to cognition, emotions and experience relevant to human action, more precisely, vulnerability reduction behaviour (cf.Grothmann, 2005).The starting point for our research on psychological indicators were Grothmann's reflections (Grothmann, 2005) on the Protection Motivation Theory (Rogers and Prentice-Dunn, 1997) and his model of private precautionary damage prevention (Grothmann and Reusswig, 2006).These models take into account psychological factors for analysing why "some people take precautionary action while others do not" (ibid, p. 101).We included those psychological indicators for which we found surveys or case studies testing the correlation of the indicators and individual vulnerability reduction or loss prevention behaviour.Such behaviour can substantially reduce people's vulnerability (World Bank and United Nations, 2010).For example, there are several low-cost measures in responding to a heat warning (e.g.drink more fluids, stay indoors).Yet, many people, despite being aware of such a warning, do not respond accordingly (Sheridan, 2007).Surprisingly, there is very little research on exactly which psychological factors play a role with regard to how people act upon heat warnings (cf.Kalkstein and Sheridan, 2007).Also, the potential of private loss prevention measures is impressive.For example, financial losses caused by flooding can be reduced by up to 80 % through private measures such as water barriers installed at the house (Egli, 2002, p. 43; also see Botzen et al., 2009;Kreibich et al., 2005Kreibich et al., , 2012;;Kreibich and Thieken, 2009).
For certain private damage prevention measures, psychological factors such as risk perception have been shown to explain more variance than traditional socio-economic factors such as household income (Grothmann and Patt, 2005; also see Howe, 2011).Krömker and Mosler (2002) describe several psychological factors or rather processes that are crucial regarding the realisation of private loss prevention measures.They also stress that "those factors must be properly addressed if people are to be persuaded to generate protection capacity" (ibid., p. 109).Consequently, to be useful for vulnerability reduction, assessments of vulnerability should also examine "individual and community social environments" (Yardley et al., 2011, p. 671; also see Mustafa et al., 2011).
For the results of the literature review on psychological indicators, please refer to Table 2.

Governance factors -starting points for reducing vulnerability at community level
The relevance of governance factors in reducing social vulnerability has been emphasized in a broad range of research work (cf.Pearce, 2003Pearce, , 2005;;Tan et al., 2007;Tan, 2008;White and Howe, 2002).However, although the term governance is extensively used, it is still an ambiguous concept (cf.Kaufmann and Kraay, 2008).Within this text, our understanding of governance refers to a notion of "good governance" regarding vulnerability reduction, i.e. the existence of public capacities and local institutions designed to support vulnerability reduction measures.Because of our focus on the community and household levels, we included only those governance factors that local decision makers have a direct influence on.For example, emergency plans for a community that are easy to understand, accessible and widely known can help community members protect themselves against natural hazards.Regarding floods, the importance of governance aspects such as "an active involvement of interested parties in the setting up of flood risk management plans" is reflected in their inclusion in the EU Flood Risk Management Directive 2007/60/EC (Fleischhauer et al., 2012(Fleischhauer et al., , p. 2785)).It was shown that participatory decision-making can reduce vulnerability (e.g.Pearce, 2003Pearce, , 2005)).Participation in dealing with risks from natural hazards can lead to a greater familiarity with the risk and can thereby increase the likelihood of the risk to be dealt with (cf.Wachinger and Renn, 2010).Governance indicators have been included in assessments of vulnerability within a set of vulnerability indicators, without being directly classified as governance indicators (cf.Carreño et al., 2007;Cutter et al., 2010;Hahn et al., 2003;Tapsell et al., 2002).There are relatively few studies on governance factors yielding empirical evidence of the relevance of those factors.Therefore, Table 3 also lists studies emphasizing the importance of certain governance factors at community level based on "a theoretical understanding of relationships" (see above, Tapsell et al., 2010, p. 27) or expert opinion, i.e. qualitative rather than quantitative arguments.

Expert sample and survey
The indicators listed in Tables 1-3 were included in an expert survey.The purpose of the expert survey was an assessment of the relevance of these indicators for identifying and targeting vulnerable population groups in industrialized countries.
The survey was carried out online between 15 October and 5 November 2010.The survey language was English.Thirtyeight (38) experts were invited to participate, with ten experts actually participating (for a discussion of the response rate see Sect. 5).The experts were selected as follows: academic In the Chicago heat waves of 1994 and 1995, 63 % and 72 % of the heat-related casualties were >65 yr.High vulnerability of the elderly to heat due to physical fragility, e.g.cardio-vascular diseases, respiratory conditions, diabetes.Meta-analysis of 54 studies: 23 with positive, 4 with negative, 6 with no relation between age and heat-related vulnerability (mainly mortality).In the flooding of New Orleans (Hurricane Katrina), nearly 75 % of fatalities were elderly (population >60 yr.: 16 %).Elderly people ack the physical and / or financial resources for effective response during/after an event.
Elderly people are more likely to hold insurances.Elderly people tend to be reluctant to leave their home despite warning, even when in good health and with sufficient resources.Whitman (1997Whitman ( , p. 1516) Greiving ( 2006 In the 1999 Cincinetti heat wave, of 17 heat-related casualties, 8 had mental illness (with possible interaction between age and mental health).Elderly people with disabilities are more likely to die during flooding events; probably due to lack of physical resource for effective response.

Heat Flood Storm
In the Chicago heat wave in 1995, for African Americans there was a death ratio of 1.5 : 1 in the total, aged-adjusted population, 1.8 : 1 for the middle-aged, 1.9 : 1 for very old victims Ethnic minorities are more likely to live in warmer neighbourhoods with greater exposure to heat stress.Mortality was disproportionately higher amongst African Americans than amongst whites due to Hurricane Katrina in New Orleans.Ethnic minority groups are more likely to live in flood-prone/lower urban areas in Southern US.More likely to rely on information from peers, which can lead to ineffective disaster preparedness.Klinenberg (2003, p. 18); Whitman (1997, p.There is a strong lack of consensus across studies on the effects of gender, despite a large amount of evidence, (which) likely indicates that the relationship between gender and vulnerability to heat-related hazards is context-specific.Higher mortality rates for males during the 1995 Chicago heat wave.Higher mortality rate of women during the 2003 European heat wave.Women are almost twice as likely to evacuate as men.
Romero-Lankao (2012, p. 676) Klinenberg ( 2003 Meta-analysis of 54 studies: 4 with negative, 4 with no relation between income and heat-related vulnerability (mainly mortality).Low-income households are less likely to have home-contents/flood insurance.People with low income are more likely to live in poorly built houses (and be located in vulnerable locations).Low-income households lack transportation for evacuation (see lack of mobility).
People with lower incomes are less likely to evacuate due to their lack of means of transportation.Gladwin and Peacock (1997) Residence type

Heat Storm
Strongest protective factor regarding excess mortality during heat waves was access to air-conditioning in the home (several case studies).Nearly 40 % of tornado fatalities in 1994 in the US occurred in mobile homes.Semenza et al. (1996); Kaiser et al. (2001) US Dpt. of Comm (1995); Fothergill and Peek (2004) Social isolation

Heat Flood Storm
Social isolation increases the heat vulnerability of the elderly (morbidity/mortality).It can be assumed that social isolation lessens the likelihood of elderly people to be taken care of.People without local networks experience a lack of information regarding the hazard, essential for warning, evacuation and post-disaster actions.With regards to Hurricanes Hugo (1989) and Andrew (1992), residents with stronger perceived social support were found to be twice as likely to evacuate as residents with weaker perceived social support.Greiving (2006); Klinenberg (2001Klinenberg ( , 2003)); Reid et al. (2009)  A sense of personal control is an important prerequisite for active participation in neighbourhood activities.Grothmann and Reusswig (2006) Chavis and Wandersman (1990) Existence of role models

Flood
Storm Proxy The existence of role models has a positive influence on the likelihood of private protection measures.
The influence of friends and relatives was the most significant factor in predicting evacuation behaviour with regards to Hurricane Katrina.The proportion of neighbours with shutters is a significant determinant of shutter usage.Personal earthquake provisions correlate with the observation of precaution measures taken by associated others."Perceived personal risk of flood and hurricane hazards (is) significantly related to adoption of flood and wind mitigation measures".
Positive correlation of perceived personal risk with protective responses (i.e., preventive measures regarding future flood events).People with high risk perception are more likely to hold flood insurance.Individuals who evacuated their properties during a flood, showed high levels of risk awareness.
No relationship between risk awareness and self-reported preparedness.
Perception of hurricane risk is a significant determinant of shutter usage.
People who perceive their home site as hazardous are much more likely to evacuate.
Missing risk perception and feeling as safe as anywhere in one's own place were the most commonly cited reasons for not evacuating.

Perc. protection responsibility
Flood Proxy Reliance on public flood protection has a negative effect on protection motivation within residents of flood-prone areas.
Perceived household responsibility for coping with earthquakes correlates positively with levels of seismic adjustment adoption.
Grothmann and Reusswig ( 2006) Garcia (1989); Mulilis and Duval (1995) Perceived relative cost Flood Perceived costs of loss prevention behaviour, together with perceived efficacy, is an important factor in forming the intention to show loss prevention behaviour or rather adaptation behaviour respectively.Grothmann and Patt (2005) Trust in official information sources

Flood
Flood & Storm The indicator "level of trust in local authorities" is considered to be one of the most relevant for social vulnerability; however, the indicator had explanatory power for different phases of a natural hazard but with opposite meanings regarding social vulnerability.Individuals with higher levels of trust in local authorities showed a lower feeling of danger, which might decrease their carrying out of loss prevention behaviour.
In Hurricane Katrina (2005), more than 100 000 greater New Orleans citizens did not respond to evacuation warnings or mandatory evacuation orders.Distrust of authorities was identified as one factor that likely played a role in this non-compliance.Upgrading and enforcing building codes decreases the risk of injury associated with hurricanes.25 % of the insured losses from Hurricane Andrew could have been avoided through better building standards and their regular enforce-ment.High quality and regular enforcement of building codes decreases vulnerability.

Availability and accessibility of hazard related information
General A lack of information regarding insurance options, building codes, climate-adapted refurbishment, and general assistance can hinder both preparedness for and recovery from a disaster.Also see Table 2 indicators "knowledge of measures".Morrow (2008); Peguero (2006); Tapsell et al. (2010) Hazard-related information in formal education

General General
Providing area-specific information on natural hazard in primary, secondary and tertiary education courses is crucial to heighten general awareness."Youth involved in education programs had significantly higher levels of correct knowledge of readiness and response behaviors, lower levels of incorrect knowledge, and reported more home-based hazards adjustments".researchers with more than one scientific publication in the fields of social vulnerability to natural hazards, social vulnerability to climate change, social vulnerability reduction, loss prevention and social capacity building were considered as potential experts for the survey, representing the research fields we included in our literature review.This led to a selection of 50 experts.Within the team, we then discussed the suitability of each of those 50 experts based on their work included in our literature review, their other scientific publications and their research activities.This led to the targeted sample of 38 (20 female, 18 male) experts being invited to participate in the survey via email.Of the ten experts participating (5 female, 5 male), all are affiliated with universities or research institutions; they all hold a Ph.D.; seven of the ten experts are professors or associate professors (one of whom is a professor emeritus) in the USA (4), the UK (3), Germany (2), and Canada (1).The experts' research interests as stated on their professional or personal websites are (frequency of mention in brackets): (social) vulnerability (assessment) (5), climate change adaptation (5), climate change impacts (3), risk assessment (3), issues of global environmental change (3), sustainability (3), resilience (3) and gender (2), natural hazards (2), environmental and risk governance (2), disaster risk management and reduction (2).
In the survey, we distinguished between household and community level indicators.The household level indicators include the traditional and psychological indicators.They are directly linked with the situation of a particular household and are therefore relevant for the identification of vulnerability down to the household level.The community level indicators refer to the governance indicators that are relevant at community level.The governance indicators do not allow distinguishing vulnerabilities at household level; however, they still directly affect the households of a community, and therefore offer starting points for individual vulnerability reduction (cf.Hufschmidt, 2011).The experts were asked to make assessments of all 31 indicators presented in Tables 1-3, except for "gender" (see Sect. 4.2.1).In the first part of the survey, the experts were asked to make assessments of the (1) importance of the household level indicators for identifying population groups vulnerable to (a) heat waves, (b) floods, (c) storms, and (d) natural hazards in general (including not only heat waves, floods and storms but also landslides, earthquakes etc., referred to as "generic vulnerability".In the second part of the survey, the experts were asked to assess the importance of the suggested community level (i.e.governance) indicators as starting points for vulnerability reduction regarding the above mentioned hazards (a-d).Also, the experts were asked to rate their (2) certainty regarding each assessment.For each indicator, they had the option to leave a comment.The experts were not provided with the complete information from Tables 1 to 3 in order to get the experts' opinion based on their own research and experience, not an assessment of the quality of the studies researched.The experts were given an assumption concerning the indicators' relation to vulnerability, and a short explanation of an indicators meaning if considered necessary4 .Assessments were made via drop-down menus, i.e. for assessing importance: very important, important, rather important, rather not important, not important, not at all important; for assessing certainty: very certain, certain, rather certain, rather not certain, not certain, not at all certain.

Assessments of importance
Table 4 shows the quantitative results of the survey.The assessment of importance differed across the individual categories (generic, heat, storm, flood).Seven of the eight traditional indicators were on average evaluated as "important": "medical problems", "residence type", "age", "household income", "ethnic minority", "disability" and "social isolation".Standard deviations are small particularly for "medical problems", the indicator that was rated most important as a generic indicator; this indicates a high consensus among the experts; also, the experts were certain regarding their assessment."Age" as an indicator of vulnerability to heat waves was assessed with the highest average importance rating, highest certainty, and highest consent among the experts.
Four of the eight psychological indicators were on average evaluated as "important" regarding their generic importance: "trust in official information sources", "existence of role models", "perceived personal risk", and "perceived relative cost".All other psychological indicators were rated as "rather important" regarding their generic importance.Interestingly, "hazard experience" was only assessed as being "rather important" regarding its generic importance.However, in the individual categories, the indicator was considered to be "important".This emphasizes the importance of hazard-specific assessments of social vulnerability.The indicators "trust in official information sources" and "existence of role models" were evaluated as the most important of the psychological indicators.The standard deviations were relatively small, indicating high consensus among the experts, and the experts were certain regarding their assessments of these indicators.
Nine of the fourteen community level indicators were evaluated as "very important" or "important" for the category "generic".The indicators with the highest ratings refer to standards and plans: "existence and quality of building codes", "existence of emergency plans", and "quality of urban/rural development plans".The importance of these indicators is also highlighted by the high consensus among the experts (e.g.nine out of ten experts assessed building codes as "very important" or "important").The indicator "participatory decision-making" (PDM) was rated as "important" for the generic and the heat category, and as "very important" for the categories flood and storm, with the experts being certain regarding their assessment and low standard deviations for flood and storm.Standard deviation for heat was high, reflecting that some experts found PDM to be also very important regarding heat waves."Availability and accessibility of hazard-related information" and "awareness-raising programs" were assessed as only being "rather important".We found this surprising since these aspects are often mentioned as key aspects in vulnerability reduction (e.g.Hahn et al., 2003;Marlin et al., 2007).

Challenges regarding the use of indicators
In the following, we will discuss challenges regarding the use of indicators, based on the experts' comments and complemented by research work regarding those comments.

Household level indicators
Many indicators used to assess vulnerability are ambiguous (also see below for a discussion of the indicator "gender").One example is the indicator "age": old people are generally physically more vulnerable, but might be more likely to hold insurances against financial losses from natural hazards (Fekete, 2009; also see Steinführer and Kuhlicke, 2007)."Hazard experience" is assumed to reduce vulnerability, but some people with hazard experience might be more vulnerable due to a (false) sense of security or a reduction of perceived risk, based on prior low-impact hazard experiences or false warnings (cf.Shultz et al., 2005;Peacock et al., 2005).However, other studies find no negative influence of unnecessary evacuations on future evacuation behaviour (e.g.Baker 1991;Dow and Cutter, 1998).Sharma and Patt (2012) emphasize the importance of defining "hazard experience" in their attempt to "resolve the conflicting findings in literature about the effect of past hazard experience" (ibid., p. 409).
Some experts hold that the indicator "gender" should be included.There are indeed many studies that show higher vulnerability of women to natural hazards, especially in developing countries (Aguilar, 2010;Mearns and Norton, 2010), but also in industrialized countries, mainly due to a lack of financial resources (Masozera et al., 2007;Morrow and Enarson, 1996).Generally, it can be assumed that in societies where "women and girls have less access to and control over resources" (Demetriades and Esplen, 2010, p. 133), women are disproportionately affected by natural disasters.There is research finding men to be less likely to evacuate due to a belief they can effectively protect their homes, which potentially puts them in danger (Riad et al., 1999), although this may also be true for women being responsible for the home and children, especially in developing countries (Oswald Spring, 2008).Also, there is a high relation of female gender with fragility, which might be due to the higher percentage of females reaching old age (Fekete, 2010).Additionally, females show a higher risk perception and preparedness for taking action (Flynn et al., 1994;Fothergill, 1996).All over, we find "gender" to be a very ambiguous indicator, especially regarding industrialized countries (cf.Arora-Jonsson, 2011), and did therefore not include it in the expert survey.
Many studies use the indicator "single household" as an indicator for social isolation.However, it seems more important to what extent an individual is integrated into a social network outside the household (e.g.friends, neighbours, family) as such a network potentially provides crucial information and support in case of a disaster (Klinenberg, 2003).
A distinction between slow onset and sudden onset disasters might be more meaningful than distinguishing natural hazards in climatic terms.The underlying reasoning is the fact that with long warnings, e.g. for a snow-melt flood, issues that are crucial in a situation requiring a fast evacuation (such as physical strength, mobility) become less important.
The indicator "household income" is assumed to be a summary marker for a constellation of factors that measure disadvantages and resources for coping (cf.Fothergill and Peek, 2004).However, well-off households can still be vulnerable; it has been shown that the availability of means for carrying out hazard prevention measures does not automatically lead to their actual implementation (cf.Grothmann and Patt, 2005).
To find meaningful indicators is always a challenge, but particularly so with regard to psychological factors.For example, "perceived personal risk" can be measured in many different ways, probably touching different aspects of risk perception.Also, risk perception in itself is a complex issue, influenced by psychological, social and cultural components (cf.Wachinger and Renn, 2010;from Slovic, 1992).Using just one indicator for "personal risk perception" would imply an oversimplification of the matter -a sacrifice regularly made in indicator-based approaches.
The US and Europe have very different profiles for correlations of vulnerability, with some indicators such as social isolation, class, or household income (cf.Kovats and Hajat, 2008).Given that the majority of studies on vulnerability in industrialized countries relate to the US, the question of transferability of US study results to European and other industrialized countries requires further research (cf.Kuhlicke et al., 2011a).

Governance indicators
It is a challenge to define a set of governance indicators.Governance indicators are highly context specific and thus usually broadly described rather than precisely defined.Often, governance indicators lack a reasonable unit of measurement, which makes it difficult to use them for comparative vulnerability assessment.While it is difficult to assess the exact effects of, for instance, building codes and emergency plans on vulnerability, these indicators allow devising vulnerability reduction measures (e.g.improving the enforcement of building codes in flood-prone areas).
While there is general agreement about the relevance of "participatory decision-making" as an important indicator for identifying starting points for vulnerability reduction, the experts stressed the challenge of carrying out successful participatory decision-making leading to vulnerability reduction.In the experts' experience, disaster preparedness levels were often still unsatisfactory despite several participatory decision-making mechanisms.A seeming lack of success of such participation can obviously have many reasons; one might be the quality of the process itself (cf.Renn and Schweizer, 2009).Similarly, many local government programs that aim at raising people's awareness of natural hazards do not necessarily imply higher preparedness levels within a community.Vulnerability is a highly contextspecific phenomenon (cf.Füssel, 2007) and therefore calls for high-quality programs that are tailored to local conditions (cf.Kuhlicke et al., 2011b; also see Matthies and Krömker, 2000).

Discussion
The response rate to the survey was rather low (see Sect. 3), which poses a potential source of bias (cf.Kelley et al., 2003).A higher response rate might have led to different values for some of the indicators.Still, we assume that it would not have changed the overall picture of the importance ratings and therefore consider the response rate as acceptable.Also, there are no identifiable fundamental differences between respondents and non-respondents -they all have similar qualifications and research backgrounds.Reasons for non-participation were not given except by two experts, who stated they did not consider themselves the right experts for the survey.
In addition to the assessment of the indicators' importance, we were interested in potential dissent and consensus among the experts.To analyse this, we chose to perform a variance analysis, which is a controversial procedure for analysing rating-scale data.However, it has been shown that performing statistical analyses requiring interval-level data (such as variance analysis) on rating-scale data does not lead to substantial distortion in the results (Labovitz, 1970).Providing a high number of response categories can support the perception of the required equidistance between the categories, five to seven categories being assumed a sufficient number (Mayer, 2006, p. 82).Fewer answer categories might have led to higher standard deviations (SDs), which were rather low (SDs from 0.4 to 1.6, with an average of 0.95).Another critical issue is posed by the verbal representation of the answer categories, which are assumed to be less suitable for expressing equidistance between the categories than numerical representation (i.e.only assigning the extreme values "very important" and "not important at all" on a scale of one to six) (cf.Schulz and Renn, 2009).Our choice of verbal representation was led by its assumed higher convenience for the drop-down menu in the online survey tool we used (Survey Monkey).However, for future work, we would probably opt for numerical representation to avoid any confusion regarding the wording and better allow for the performance of statistical analyses.
In the survey, both the traditional and the psychological indicators were presented as indicators for identifying vulnerable population groups.That was because we were interested in the experts' assessments regarding the relevance of the psychological indicators as indicators complementary to the traditional ones.However, our reason for including psychological indicators is their importance with regard to loss prevention behaviour.An alignment of the survey question (i.e. as how important do the experts consider the psychological indicators for loss prevention behaviour) might have led to different results regarding the importance rating.Also, comparing the assessment of importance and certainty, one can see the tendency that the experts' certainty regarding their judgments increases with increasing importance ratings.With more research on the relevance of psychological factors, the certainty of experts and their assessments regarding these factors might also increase.

Conclusions -a concept for integrating psychological and governance factors
The results from the expert survey suggest that psychological indicators can be seen as complementary but not equal to the traditional indicators when it comes to vulnerability assessment; most of the psychological factors were assessed as being less important than the traditional factors for identifying vulnerable population groups.However, an expert survey with other experts might have led to different results.In an expert survey carried out by Müller et al. (2011) with practitioners (persons from the regional government, non-governmental organisations, communal planning institutions), psychological indicators such as "experience with floods", "knowledge about floods" and "knowledge about flood protection measures" were rated as being more important for urban vulnerability assessments than traditional indicators such as "age", "occupation status" and "gender" (ibid., p. 2116).In any case, when looking at how to refine vulnerability assessments, reduce vulnerability and build social capacity, we believe that surveys including psychological indicators -especially in combination with governance indicators -could offer extremely valuable insights for decision makers at the local level.There are three main reasons for this assumption: firstly, the literature review shows that there are many surveys and case studies finding a significant correlation between psychological factors and the likelihood of people to take private vulnerability reduction measures.Secondly, many of the psychological factors are potentially changeable in the short to medium term.Thirdly, unlike some traditionally recognized vulnerability factors (such as age or poverty), they can potentially be influenced by (local) decision makers or through (local) governance strategies (e.g.modes of risk communication, incentives for private vulnerability reduction measures, participatory processes).
Being able to determine which governance factors can increase the likelihood of people adopting loss prevention behaviour would make psychological and governance factors a powerful couple.For example, participatory processes have been shown to change people's perception of natural hazards, their level of trust in authorities and their willingness to "initiate protective action" (Wachinger et al., 2012, p. 13).
Programs and campaigns for disaster risk reduction based on insights regarding psychological factors and people's motivation to take private loss prevention measures can have a huge potential for overall loss prevention.Considering this, the costs of surveys that allow gaining these insights for a particular locality seem marginal.Governance has been recognized as playing a major role in reducing the vulnerability of individuals and communities.This is shown by our literature review and confirmed by the expert survey.However, despite insights into how, for example, participatory processes should be set up (e.g.Matthies and Krömker, 2000), according to the expert survey, apparently many of such processes in disaster risk reduction miss their objective of increasing preparedness.Including both psychological and governance aspects might improve the extent to which such processes are tailored to the needs of the people affected and thereby increase the effectiveness of the processes.Following the aforementioned thoughts, we propose a pragmatic concept (see Fig. 1) to make efficient use of the different indicators and research instruments for assessments of social vulnerability and building of social capacity (cf.Tapsell et al., 2005).
Firstly, indicators of exposure to climatic and natural hazards (e.g.altitude above sea/river level) could be assessed.Secondly, for those regions which have turned out to show medium to high exposure, traditional indicators such as age, medical problems and household income could be gathered.Thirdly, psychological indicators such as trust in official information sources, existence of role models or risk perception could be evaluated (with a focus on regions that have been identified as vulnerable through the collection of traditional indicators).The collection of the psychological indicators will require setting up surveys.By applying surveys only to those regions that have been identified as vulnerable through the first two steps, the costs of such surveys would be reduced.Fourth, governance indicators could be  assessed to assist the planning of governance measures to reduce vulnerability.Based on the data gathered in the previous steps, these measures could be planned efficiently and targetgroup specific.Through this, the concept as described might support local decision makers, organisations and institutions in assessing vulnerability at community and household level and be useful for planning measures for social capacity building (cf.O'Sullivan, 2012).Nevertheless, to further improve such efforts, more research is necessary regarding the influence of psychological factors on the likelihood of people to take private vulnerability reduction measures.
The approach as presented is especially relevant against the background of an increasing "privatisation of risk" (Steinführer et al., 2009), that is, according to Kuhlicke et al. (2011a), promoted by changes in governance practices in many European countries with regard to natural hazards.People at risk are "encouraged or even required to take more responsibility for their actions" (ibid., p. 806).This often conflicts with people's perception that they have few chances to prevent losses from natural hazards (cf.Grothmann and Patt, 2005;Plapp and Werner, 2006).Consequently, it seems useful that people should become more aware of possible threats and means of protecting themselves.This emphasizes the importance of understanding the determinants of private loss prevention behaviour, but the "privatisation of risk" can be problematic when it leads to the "expectation that city residents [. . .] will be active consumers of public goods, [. . .] rather than "citizens" entitled to social protection" (Klinenberg, 2003, p. 232) -a perspective potentially leading to too high a burden on the citizens (cf.Steinführer, 2009).Claiming that people need to take responsibility to protect themselves from natural hazards should not absolve the state from its duty to protect its citizens and to tackle social inequality as a root cause of vulnerability.

Fouillet
et al. (2010);Cutter et al. (2010);Murphy (2007);Hahn et al. (2003)    Consideration of natural hazard impact reduction General Systematic consideration of potential impacts from natural hazards and climatic threats in communal key sectors reduces vulnerability.Steinführer and Kuhlicke (2007);Fekete (2009); flooding is considered in urban/rural development plans, vulnerability of low-income households can be decreased.Specific climatic risks and potential hazards can be approached in urban/rural development plans to reduce vulnerability in the long-run (e.g. through air exchange corridors).Morrow (2008)Bollin et al. (2003);Tan et al. (2007);Carreño et al. (2007);Barroca et al. (2006) Training of health/emergency professionalsGeneral Training of professionals from the health and emergency sector improves the overall level of adaptive capacity and decreases vulnera-bility.Aldrich and Benson (2008, p.  3);Morrow (2008) Voluntary involvement in support of vulnerable population groups General The number of active volunteers supporting vulnerable groups (e.g.elderly, disabled) in a community indicates the willingness to help people most at risk and helps to decrease the vulnerability of these groups.Marlin et al. (2007); Stanley (2010) Participatory decision-making (PDM) General Proxy Vulnerability of a community could be decreased if planners consider the knowledge and needs of those most at risk.PDM was crucial to decrease losses from wildfires, landslides and water-borne disaster in the US and Canada.Morrow (2008); Hahn et al. (2003); Ikeda et al. (2008) Pearce (2003, 2005); Murphy (2007) Training of people inv. in house construction and urban planning General Frequent training of community members involved in construction, housing, and urban planning is considered to be an important component of a resilient community.Gupta et al. (2010); UNISDR (2010) Training of staff in educational servicesGeneral Educational staff needs to be regularly trained in order to create a culture of awareness of the importance of risks from natural hazards and their repercussions on vulnerable groups.Hahn et al. (2003);Carreño et al. (2007)

Figure 1 .
Figure 1.Concept for integrating psychological and govern

Table 1 .
Traditional indicators and financial losses, morbidity, mortality and loss prevention behaviour (industrialized countries only).
IndicatorsHazardCase studies and surveys that find correlations with indicator and financial losses, mortality and/or morbidity, or loss prevention be-haviour (industrialized countries only)

Table 2 .
Psychological indicators and loss prevention behaviour (industrialized countries only).

Table 3 .
Governance indicators important for vulnerability reduction and capacity building (industrialized countries only).

Table 4 .
Quantitative results of the expert survey.