The model for the overall severity of the effects of the flood on the household’s life contained only seven predictor variables. Three flood characteristics were significant predictors. Only four additional factors emerged when the social and dwelling characteristics variables were included in this model. However, together, these variables explained substantially more of the variance in the overall subjective severity ratings, 17%, than the characteristics of the flood event taken on their own. Gender was a significant factor in this model. Prior health was a factor in the overall severity of the flood event. Area house prices again featured as a predictor with low house prices leading to raised overall severity ratings. Age and household composition combined were factors with those aged 65 and over living alone, counter-intuitively having lower overall severity scores
To summarise, on our analysis of flood characteristics and the social dimensions of vulnerability presented in Tables 6.18 to 6.21:
A first point to re-iterate is that the vulnerability variables are rather different and also because of the non-response, particularly to the GHQ12 question items, the regression models are based on samples of different size and composition. The models clearly show that the characteristics of a flood event are important. As we would expect, the depth of flooding was a predictor of vulnerability in all eight of the models; the extent of flooding to the home as measured by the number of main rooms (living room, bedroom, kitchen, bathroom) featured in four of the models. Flood warnings and a longer flood warning lead time which are intended to reduce the health and stress effects of flooding were significant factors in three models: that for the worst time GHQ12 score and both stress models including when the social dimension was taken into account. What was not anticipated was that the perception of the floodwaters as containing sewage or other pollution would be such an important predictor appearing as a highly significant explanatory variable in all models. However, it is clear that, for all the vulnerability variables, but particularly for the current GHQ12 scores, the flood event characteristics available in the Intangibles Survey taken on their own explained only a very limited amount of the variance in vulnerability, under 10% for all the variables considered and only 4% for the current GHQ12 score.
When the social dimension was introduced, the amount of variance explained across the models was enhanced, ranging from 8% for the current GHQ12 to 18% for the stress rating. Overall, the introduction of a wide range of social variables into the analysis still leaves a large amount of the variability in the vulnerability variables unexplained. Lack of success in predicting the current GHQ12 scores with flood characteristics and social variables may be partially explained by the passage of time from the flood event to the time of interview and the changes in circumstances and events that may have intervened affecting health and well being.
A wide range of explanatory social variables appear in the models but there were some common factors. Gender was a factor in all the models apart from that for the current GHQ12. The variable relating to health status prior to the flood was a predictor in three models. Various variables that were indicators of socio-economic status also featured in the models. The rating of the area house prices, an indicator of the relative affluence of the interview locations was a factor in three models. In line with expectations that vulnerability would be associated with social deprivation, lower area house price ratings were predictors of higher ratings for stress and overall severity but for current GHQ12 scores the reverse was the case. Renting property, which can be taken as indicative of social disadvantage in the UK context was also a predictor of higher scores in the current GHQ12 and stress models. However, being in the highest social grades (AB) was a predictor of higher stress. Vulnerable housing was an explanatory factor in only one model, for the worst time GHQ12. In line with the findings in the bi-variate analyses, the middle aged group (45-64) was an explanatory factor in the GHQ12 current and worst time models and indicating higher scores. Certain explanatory variables appeared in a way that was contrary to expectations on vulnerability. We would expect those living alone and older people and particularly those aged 65 and over living alone to be more vulnerable than others. However, in certain models these variables appeared as significant predictors of lower vulnerability. The bi-variate analyses also provide some evidence of this counter intuitive finding. The variables that cover prior health and illness and disability may cover much of the vulnerability associated with old age and this may provide some explanation for this finding.
In this section we consider the factors that may intervene in the aftermath of flooding and their effect on the vulnerability variables. Our analysis of the time at which the health effects were at their worst showed that people varied in the stage at which they experienced the worst time with their responses ranging from the time of the flood event itself to well into the recovery period. Thus the kinds of factors that may come into play to influence the vulnerability variables will also vary depending in part on when they experienced the worst time. The way in which institutions, organisations and individuals within the community respond and deal with the event are among the factors to be taken into account. The post event intervening factors considered are listed here:
Post-flood intervening factors
Rating of problems with builders (scale 1-10)
Rating of problems with insurers and loss adjustors (scale 1-10)
Help received from outside household (0-50)
Evacuation (yes, evacuated = 1, no, did not evacuate = 0)
Disruption: time taken to get home back to normal (weeks)
£value of uninsured losses
When these post-flood intervening factors were introduced into the regression analyses, the models changed substantially although some of the flood and social dimensions remained. Table 6.29 shows the model that accounts for vulnerability as measured by the worst time GHQ12 scores. The inclusion of the post –event factors greatly enhanced the explanatory power of the model for this vulnerability variable. Certain socio-demographic factors were important as explanatory variables for this vulnerability measure. Gender and prior health emerged again as significant factors in the short term-health effects of flooding. Counter-intuitively but in line with the bi-variate analyses, being aged 65 and over was a factor reducing vulnerability in this model, (albeit not significant at the 0.05 level). This may be because prior health, which is strongly related to age, accounts for the vulnerability due to ill health among those in the older age groups. Living in rented accommodation as compared with owner occupation was a contributory factor.
Perhaps surprisingly, the extent and depth of flooding do not feature as explanatory variables possibly because their influence is reflected in other variables such as the time taken to get back to normal and evacuation. One flood characteristic that does feature in this model is the belief that the flood waters are contaminated, a factor leading to higher GHQ12 worst time scores. Flood warning lead time also had some contributory influence in reducing vulnerability in the short term.
Factors associated with the aftermath of flooding and the recovery period such as having to evacuate and the time taken to get back to normal were significant explanatory factors. The data confirm what emerged in our qualitative studies (Tapsell et al., 1999; Tapsell and Tunstall, 2001) that the role of the insurance industry and the way that its personnel deal with flood victims are crucial in mitigating or exacerbating the trauma of a flood. Insurance was both a positive and negative influence. The level of uninsured losses incurred due to lack of insurance cover or under insurance was a negative factor but conversely having adequate insurance was a positive factor. Flooded households reported very varied experiences of the attitudes of loss adjustors and insurers towards their clients and in the speed, efficiency and sympathy shown in handling claims. These problems were a very significant factor in vulnerability.
Table 6.22: Post-event factors social, dwelling and flood event characteristics and GHQ12
Worst time GHQ12 Likert
Number of cases = 511, R2 = 0.267, R2 (adjusted) = 0.253