Multi-variate analyses: Social characteristics and vulnerability
Here we consider, in addition to the key flood event characteristics that were predictors in the models in Tables 6.25 and 6.26, the social aspects that may gain importance during a flood and make individuals and households more vulnerable during and after flooding: the social dimensions of vulnerability. These factors may include both the social characteristics of the respondents and their households and the characteristics of the dwelling and area where respondents lived presented in Figure 6.1. In this section we confine the analysis to factors present at the time of the flood and exclude the developments and interventions that may occur after the event shown as post-flood intervening factors in Figure 6.1. The following factors which may be expected to make individuals and their households more vulnerable are included in these analyses:
Social characteristics and prior health:
Gender, (Male = 1, female = 0)
Age: under 45, 45 - 64, age 65+, (1= yes, 0 = no)
Social grade: AB or not, DE or not (1 = yes, 0 = no)
Households with children, households with children under 10, (1 = yes, 0 = no)
Living alone, (1 = yes, 0 = no)
Aged 65 living alone, (1=yes, 0=no)
Prior health,(1 = poor to 5 = excellent)
Long term illness or disability in respondent/other household members/ and overall (1 = yes ,0 = no)
Housing status (renting or not , 1= yes 0= no),
Vulnerable property or not (1 = yes,0 = no)
Length of residence in years (years)
Area house prices as an indicator of the affluence of the area (1 = high – 5 = low)
Because the relationship between age and the vulnerability variables does not appear to be linear. Age has been included in grouped form as three dummy variables.
Including social and dwelling characteristic variables in the regression analysis enhanced the explanatory power only a little with 11 % of the variance explained in the model for the worst time GHQ12 Likert scores, which represent the short term health effects of flooding (Table 6.20). While flood characteristics remained important explanatory variables for the worst time scores, five social variables emerged as highly significant predictors. These included gender, with women predicting higher scores than men, and prior health with poorer health a predictor of raised GHQ12 worst time scores were the most significant. As might be expected, living in vulnerable housing such as ground floor flats, bungalows and mobile homes was another significant predictor of higher scores. In line with the findings in the bivariate analyses those in the middle age groups (45-64) were associated with higher scores. Those in the somewhat younger age groups (under 45) were also included in the model as predicting higher scores, indicating that it is the older age groups that stand out as less affected.
When the social dimension was included in the model for the current GHQ12 scores, indicating the longer term health effects of flooding, there was much less effect on the explanatory power of the best fit model (Table 6.20). A very simple model with only five variables emerged. Perceived contamination of the flood waters remained as a flood characteristic that was a predictor and a highly significant one for the current GHQ12 Likert scores. The maximum depth of main room flooding also remained in the model.
Of the social characteristics, poorer prior health and being in the middle aged group (aged 45 to 64) were again key predictors of higher current GHQ12 scores. Gender, however, an important predictor in the worst time model, was no longer a factor for the current GHQ12. This reflects the findings of the bi-variate analysis that women were worse affected than men at the worst time of the flooding but that the gender difference disappeared with time and was no longer evident in the current GHQ12 scores suggesting that women were more resilient and recover better than men. Many other variables that were predictors of the worst time GHQ12 scores do not feature in this model. More surprising was the fact that area house prices were a factor, with higher house price areas leading to higher current GHQ12 scores.
Here again the passage of time from the flood event to the time of the interview to which the current GHQ12 scores relate must be noted. However, it is therefore particularly significant that key flood characteristics remain important predictors despite the passage of time. If the worst time scores were to be included again as an explanatory variable for the current scores it is likely that the explanatory power of the model would be enhanced.
Table 6.20: Social, dwelling and flood event characteristics and GHQ12 Likert scores
Worst time GHQ12 Likert Number of cases = 595, R2 = 0.123, R2 (adjusted) = 0.114
Contamination of floodwaters
Maximum depth of main room flooding
Aged 45 to 64
Aged under 45
Current GHQ12 Likert
Number of cases = 682, R2 = 0.089, R2 (adjusted) = 0.082