Although in Task 11 the concept of “risk perception” is prominently positioned (namely in its title), in the course of the work we became more and more convinced that it has some conceptual shortcomings. Although the term is quite well established in the scientific community, we decided to replace it with risk constructions. There are many reasons for doing so, four of which we want to point out in the following discussion.
Firstly, risk perception implies a simple cause-and-effect model in the sense that an individual perceives physical stimuli and reacts upon them. However, as the “traditional” literature on risk perception was able to show in the course of its intellectual development, the issue under investigation is far more complex: “To speak of ‘perceived risk’ in the same manner we speak of ‘perceived length’ makes no sense“ (Brehmer 1994, 83), since a mental construct (e.g. “probability * consequence”) cannot be perceived.
The second argument relates to the historical development of the discourses on risk perception and vulnerability. The discourse on risk perception was mostly advanced in psychology by the so-called Oregon Group around Fischhoff, Lichtenstein and Slovic (Psychometric Paradigm). Its intention from the very beginning was, firstly, to show that risk is above all a “subjective” construct (and not an “objective” one), secondly, to point out that so-called lay-people have a different risk perception than experts, and, thirdly, to analyse the cognitive structure of risk judgements by employing multivariate statistical analyses such as factor analysis, multiple regression etc. (Slovic et al. 1974; Fischhoff et al. 1979; Slovic 1987 and 1992). Another “school”, which may be called rather sociological and/or cultural in its orientation to risks, emphasized the intersubjective modi of constructing risk. Risk perception in this perspective is defined by norms, value systems and cultural idiosyncrasies of groups and societies. A simple juxtaposition of individual/subjective and scientific/objective risk perceptions is no longer possible thereby, since every group, thus also scientists, are biased by certain assumptions, norms, values and beliefs (Douglas and Wildawsky 1982; Johnson and Covello 1987; Hoekstra 1998).
In 1992, the volume “Social Theories of Risk” (Krimsky and Golding 1992) appeared as a collection of essays by sociologists and other social scientists who, in the following years, contributed, together with a growing cluster of colleagues, to enlarge the debate with natural scientists, also increasing the visibility and “legitimacy” of social studies of science and technology (among many others, Nowotny et al. 2001; Jasanoff 2006; Renn 2007). Also, attention grew on issues of complexity and indeterminacy (e.g., Lash et al. 1996; Wynne 1992), with relevant contributions from ecology and ecological economics (Kay 2001; Gunderson et al. 1995; Gregory 2002; Gregory and Wellman 2001). A key point of attention became the distinction between risk and uncertainty (Funtowicz and Ravetz 1993), the former being quantifiable through the application of standard assessment techniques, the latter being characteristics of contemporary scientific problems and requiring new instruments of analysis as well as novel management approaches (De Marchi 1995; De Marchi and Ravetz 1999). When Ulrich Beck’s book was published in English (Beck 1992; first in German in 1986) the time was ripe for a debate with many voices, contrary to a decade earlier, when Short’s appeal in his presidential address to the American Sociological Association (Short 1984), remained largely unheard.
Particularly the Psychometric Paradigm was also prominent in research on natural hazards (Slovic et al. 1974; White 1974) and uncovered some valuable empirical findings, such as the central paradox of technical flood protection measures: while expenditure on flood control was rapidly increasing after the 1927 Mississippi flood, the monetary flood damages were also rising (White 1973; Barry 1997). However, the underlying assumption is quite simplistic as Watts states: The research paradigm is based on an “assumption of individual purposeful rationality expressed through a tripartite cybernetic structure: (a) hazard perception, (b) recognition of alternative-adjustments, (c) choice of response” (Watts 1983, 240). As a result, individuals are understood as rationalistic atoms, defined by imperfect knowledge and acting in a societal space that is without structure and institutions. Watts concludes that maladaptation in this context is simply a function of insufficient knowledge, distorted perception and inflexible decision-making (ibid., 241).
Therefore we think it is of importance to keep in mind both the development of the field on risk perception as well as the “radical constructivist” moment of the conceptualization of risk perception inherent in Cultural Theory when one relates it to the concept of vulnerability, since most vulnerability researchers are not interested in this debate. There is even a strong opposition to questions of interpretation and perception, since particularly vulnerable people of a society are simply not in a position to take the necessary steps to mitigate or prevent the occurrence of a disaster (Oliver-Smith 2002). The concept of vulnerability is based on a realist assumption to the effect that the causes eventually resulting in a disaster are socially produced; the event itself, however, is not constructed; it is rather understood as “real”. The debate about vulnerability is predominantly interested in social, economic and political structures and processes, since these “hard” factors are seen as the driving forces defining the vulnerability of certain groups; questions of perception and interpretation, particularly when conceptualized in a narrow sense as mostly done in hazard research, are seen as subordinate.
However, in recent years there has also developed a counter-discourse to the rigid understanding of vulnerability. Critics point to the problematic assumption of the “vulnerability view”, since it assumes people who are held as vulnerable are weak, passive and, in a certain sense, deviant (Hewitt 1997; Boyce 2000; Bankoff 2001). Therefore some scholars underline the importance of incorporating the perception of people, their capacities and interpretation of their own situation in empirical studies. The reasons these scholars do so are, however, not analytical; they are above all normative, since they try to empower people (Delica-Willison and Willison 2004) in order to find a way of how to integrate both societal structures and individual actors within one theoretical framework. Nevertheless, it seems important to point towards the difficulty of overcoming the duality of a constructivist and realist view on risks and disasters. In the wider sociological debate Anthony Giddens’ theory of structuration is surely such an attempt to reconceptualise the dichotomy of agency/structure and objectivity/subjectivity (Giddens 1986); however, the empirical applicability of this theory is an exercise exceeding the intentions of the work in FLOODsite Task 11.
This relates to the third argument: The term “risk construction” chosen in the title of this section highlights our understanding of risk. Risk is neither objectively given nor predetermined by social structures such as income, age, class etc., nor is it simply a matter of individual cognitive operations. Risk is socially constructed in the sense that norms and values as well as belief systems influence and possibly define it. Thus in this context, we want to depart from most conceptualizations of vulnerability which agree that vulnerable conditions are produced by social structures but which, however, would reject that the concepts risks and disasters themselves are socially constructed. Nevertheless, in our opinion the modi of construction have to be taken into account. We therefore draw upon the work of Berger and Luckmann (1967). In their ground-breaking work on the “Social Construction of Reality” the authors lay down a theory, which allows for incorporation of, on the one hand, the inter-subjectively constituted life-world of people and, on the other hand, the objectified reality of everyday life (ibid.). The authors emphasize that the construction of reality proceeds by no means arbitrarily, since over time social actors develop typifications of each other as well as of each other’s actions, and these typifications eventually become habitualised into reciprocal roles. Reality is finally objectified when these roles and typifications are made available to other members of the society, which means they are institutionalised. These institutions appear as objectively given, since they transcend the individual and particular concept for action (Handlungsentwurf), although they are embedded and reproduced by individual actions, since the process of institutionalization is executed in interactions among human actors.
Institutions are evolving when different actors are confronted with a recurring problem, which is solved more or less routinely (e.g. floods). They are typical solutions for recurring (and accordingly typified) societal problems of action. Therefore institutions are relevant for a sociological analysis; they point towards what is considered as important in a society, they uncover in a more general sense the respective societal system of relevance. The development of insurances during the 13th century and their stepwise spreading in the sphere of maritime trade during the 14th and 15th centuries is such an example (Ewald 1989; Bonß 1995), pointing to the coverage of certain requirements of safeness and security.
At this point, we want to introduce the final argument for talking about risk constructions: FLOODsite Task 11 ultimately aims at a cross-cultural analysis. Usually, such investigations are either pursued in the tradition of the Psychometric Paradigm or in line with Cultural Theory (Horlick-Jones et al. 1998; Caulkins 1999; Renn and Rohrmann 2000; Rohrmann 2000; Sjöberg et al. 2000; Marincioni 2001). However, understanding the construction of risk in the outlined manner allows us to take into account rather subjective definitions of risk but also to focus on the institutionalised construction or risk. This seems to us to be a fruitful design, allowing an approach towards cross-cultural comparison, which does not rest on the level of superficial results and which does not overemphasise rigid interpretations of social structures, but rather takes dissimilar institutionalisations of risk in different societal contexts into account.
2 Research Methods and Limitations As mentioned in Section 1.1, the focus of FHRC has been on re-analysing, or further secondary analysis of, data from some of our earlier studies rather than collecting new survey data. The data offered in the data sets are very different from the case study data focused on particular localities and particular flood events available to our German and Italian partners. The FHRC data were originally collected and analysed for other purposes, based on particular theoretical frameworks, and have been reported elsewhere (RPA/FHRC, 2004; Tunstall et al., 2006; Tunstall et al., 2005; McCarthy et al., 2006). This therefore allows some limited comparison with data from the German and Italian case studies.
The data sets provide quantitative survey data derived from structured questionnaires and are thus different from the data collected by our German and Italian partners which includes substantial qualitative elements derived from qualitative interviewing, focus groups and observation. However, the FHRC survey studies did involve some initial qualitative focus group and in depth interview research and this report will also draw on the insights that this qualitative research provided where appropriate. The report will also make reference to our qualitative studies of the social and health impacts of flooding (Tapsell et al., 1999; Tapsell and Tunstall, 2001, Tapsell et al., 2003).
Three existing recent data sets collected between 2002 and 2005 have been reanalysed to address the aims and objectives of this task, they are the:
‘Intangibles’ data set
‘Warnings’ data set
‘Lower Thames‘ data set
The first two UK data sets listed cover a range of locations (up to 30) and many different flood events in England and Wales. However, the third data set was focused on a particular location along the River Thames and a single key flood event. These data sets have been further analysed to augment our understanding of flood event experiences, preparedness and response to elaborate our understanding of the social and health impacts of flooding in the UK. The report uses as a basis for analysis the suggested Set of Indicators produced for Task 11 in 2005. However the previous studies do not necessarily cover the full range of indicators outlined in the 2005 report.
Initially, the data set from an earlier series of surveys - the ‘Full Flood Impacts’ study - was to be included. However, on further examination of the data it was decided that little more analysis could be achieved on this data, therefore a decision was taken to include data from a different study which had focused on flood warnings. This offered much more recent data, some of which had not been fully analysed.
Familiarity with the data and the methods used to collect it, often a difficult issue in secondary analysis of survey data, is not a problem in this case as all the data were collected for, and originally analysed by, FHRC. The data, too, are relatively recent so that we would not expect change over time to be an issue for the reanalysis. However, the availability of variables to adequately measure the concepts of interest to the FLOODsite Task (such as vulnerability and resilience), is a problem. The focus of all the surveys, moreover, was upon the household and the individual rather than upon the community. These studies are very weak or lacking in variables that measure community characteristics that are of interest to FLOODsiteresearchers in Task 11. However, the data sets are relatively large and rich in variables and the FLOODsite work provides a valuable opportunity to consider issues and relationships that were not considered at the time of the original analysis. The three data sets are outlined below.
2.1 The ‘Intangibles’ data set
This was collected throughout England and Wales as part of a project funded by the UK Government Department for the Environment, Food and Rural Affairs (Defra) that aimed ‘to develop a robust yet simple-to-use, methodology so that the intangible impacts on human health and well being can be accounted for in assessing the benefits of flood alleviation measures’. The methodology was intended to be applicable to all levels of appraisal from policy and programme evaluation to individual flood defence schemes (RPA, FHRC, 2004:1). Specific requirements were to obtain greater understanding of the social issues that underlie the long-term health risks of flooding, and to develop an easy to use methodology that could be used in economic appraisal to generate robust and defensible valuations for human-related intangible impacts of flooding, based on the improved understanding of the relevant social issues (RPA, FHRC, 2004). The focus of the analysis and reporting of the data was on the long and short term health impacts of flooding, and the economic values that may be attached to avoiding health and stress effects (RPA/FHRC, 2004).
The Intangibles survey involved two questionnaires with many common questions. The main questions for the ‘flooded’ sample can be summarised as:
questions about the property, members of the household, nature of the flood event and associated damages;
perceptions of flooding, flood prevention measures and support received.
willingness to pay to avoid the stress of flooding questions; and
standard socio-economic questions.
For the ‘at risk’ sample the questions focused on:
questions about the property, members of the household and awareness of flooding;
perceptions of flood risk and flood prevention measures;
questions about health;
self-completion health questionnaires;
willingness to pay to avoid the stress of flooding questions; and
standard socio-economic questions.
2.2 The ‘Warnings’ data set This survey in different locations in England and Wales was undertaken as part of research funded by the UK Environment Agency (EA) and Defra. The objective of the survey research was ‘to examine and further develop as necessary, the model of the economic benefits of flood warnings set out by Flood Hazard Research Centre researchers and to produce a new data set to be used to calibrate the model. The analysis of this data was concerned with the factors that may explain the level of property damage reduction that can be achieved through timely flood warnings (Tunstall et al., 2005:1).
The Warnings Survey employed a single questionnaire with the following main question topics:
questions about the property, members of the household, nature of the flood event and associated damages including detailed questions on items of property damaged or saved;
questions about flood warnings;
questions about actions taken in response to flooding; and
standard socio-economic questions.
2.3 The ‘Lower Thames’ data set This survey was mounted as part of the Lower Thames Strategy Study Phase 3 funded by the Environment Agency which is examining options for flood risk management in the Lower Thames Catchment area. The survey focused principally upon flood risk perception and the acceptability of community based flood risk reduction, in particular through the installation of local flood barriers and devices (McCarthy et al., 2006). The Lower Thames Survey questionnaire covered the following main topics:
awareness and perceptions of flooding and flood risk;
2.4 Qualitative research and pilot testing
Both the Intangibles and the Warnings Surveys were preceded by a qualitative research stage involving focus group discussions. This qualitative and developmental stage was extensive in the case of the Intangibles survey because measuring the health and stress impacts of flooding had not been attempted before. A total of five health focus groups to develop and test questions and health measures for the survey were held in five different areas with a total of 34 mainly flooded participants. Following on from this, six focus groups were conducted in three areas; in each area one group of flooded and one group of those at risk was involved. This second phase of focus groups aimed to develop and test questions to elicit willingness to pay (WTP) to avoid the stress and health effects of flooding; a total of 35 participants took part. Pilot testing of the questionnaires was undertaken in three stages:
Separate pilot surveys of a health questionnaire (72 respondents) and WTP questionnaires (48 flooded, 42 at risk respondents).
11 face to face interviews by researchers using a questionnaire combining the health and WTP questions.
Pilot survey testing the combined questionnaire with 37 flooded and 16 at risk respondents. Following the first two stages of the pilot testing it was decided that it would be feasible to combine the health and WTP questionnaires into a single instrument without the interview becoming too lengthy and a combined version of the questionnaire was therefore tested in the third stage of the pilot testing.
The Warnings Survey questions were derived from instruments that had been tested and used in previous research on flood warnings and response by FHRC and the British Market Research Bureau (BMRB) in Post Event Surveys for the Environment Agency. Therefore, pre-testing of the survey instrument was more limited and was mainly confined to testing the ability of focus group participants to recall (using a checklist) the items of property moved and thus saved from flooding and items damaged. Five focus group discussions to refine and test the survey materials were held in five different areas with a total of 29 flooded residents.
In the Lower Thames Study, one focus group discussion was conducted with a special community of residents living on an island in the River Thames.
2.5 Main survey methods The survey methods used to collect the data sets are summarised in Table 2.1. This shows that the studies on which this report is based had many features in common.
Table 2.1: Main survey methods used to collect the data sets
Survey methods used to collect the data sets
Date of main survey fieldwork
October – December 2004
January – February 2005
October – November 2005
Areas for sampling
30 specific locations in England and Wales
The same 30 locations in England and Wales
11 Environment Agency areas and a large number of specific locations in England and Wales
13 locations in the Lower Thames – from Walton Bridge to Teddington
Residents - all of whom had experienced above floor flooding in flood events between April 1998 and December 2001
Residents aged 18 and over
Residents within the 1 in 100 flood risk area who had not experienced flooding above floor flooding since April 1998
Residents aged 18 and over
Phase 1 included some residents in flood risk areas that had not experienced property flooding in events in 2003 and 2004.
Phase 2 included only residents who had experienced property flooding since September 2000
Residents aged 18 and over
Residents in the 1 in 50 flood risk area in Lower Thames locations where community based flood risk reduction was feasible.