Methods

Data came from 17 annual waves of the British Household Panel Survey (BHPS), which started in 1991 and later added booster samples for Scotland, Wales and Northern Ireland. The BHPS is a well-established survey that is extensively documented elsewhere.15 16 Overall, the panel was unbalanced as individuals move in and out of the study according to a set of following rules. This analysis used the core longitudinal panel of respondents who have been in the survey at all eligible time points and whose data are weighted for selection and attrition (N = 12 390; person/years 139 928). Ethics approval was granted by University of Essex process for secondary data analysis of survey data.

Common mental illness was measured by the 12-item General Health Questionnaire (GHQ). Items were coded to make a 0–12 scale, then a threshold of 4 or more was used to indicate common mental illness.17 18 19 20

Respondents who had moved since their last interview were asked why they had moved. One of the possible responses was “repossession or eviction”. This was used with their housing tenure from the previous interview to create a time varying indicator, so that those who were home owners are designated as having experienced repossession and those who were renters are designated as having experienced an eviction.

Social class, marital status, age and employment status were used in the analysis as potential time varying confounders, as these have well-documented associations with common mental illness. All time constant characteristics were controlled for in fixed-effect models.

Descriptive statistics illustrate the proportions of those with common mental illness in the years before and after repossession or eviction. These were compared with norms for the respective groups for the panel sample and presented graphically.

Two multivariate fixed-effects logit regression models, one for repossessions and one for evictions, were used to estimate the net effect of the change in housing status on the risk of common mental illness, with estimating samples limited to those at risk of such an event: home owners (N = 9768; person/years = 83 129) and renters (N = 3899; person/years = 22 744) respectively. Respondents who were owners and renters over the time of the panel were included in both models, but only for the years they were at risk of a repossession or eviction. Fixed-effects regression methods control for all stable characteristics of the individuals, whether measured or not, by using only within-individual variation to estimate the regression coefficients, so respondents must be observed at least twice to be included in the estimations.21

(1) Pr(y_it  =  1|X_it, C_it)  =  F(αX_it + βC_it) if housing tenure_t−1  =  owner

(2) Pr(y_it  =  1|X_it, C_it)  =  F(αX_it + βC_it) if housing tenure_t−1  =  renter

Where y is the dichotomous GHQ indicator of individual i at time t; F is the cumulative logistic function; X_it is the time varying repossession (equation 1) or eviction (equation 2) indicator; C_it is a vector of time varying confounders; α and β are parameters to be estimated. The value of α is given as adjusted odds ratios (OR) with robust 95% confidence intervals (CI).22

Results

There were a total of 145 (weighted: 220) repossessions and 493 (weighted: 838) evictions. Figure 1 presents the proportions of those experiencing repossession (squares) or an eviction (diamonds) with common mental illness for two observations before the event and three observations after the event. The total sample proportions for owners and renters are also shown for comparison. Those who experience repossession show an increase before and immediately after the event with large changes after that; possibly due to the small number of events. Proportions at all time points are well above the sample proportion for owners. Those who are evicted show a slight increase before the event but then a steady decline after the event, but never far from the overall sample proportion for renters.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Proportion of those experiencing repossession or eviction with common mental illness.

The regression models estimate that at time t those who experienced repossession are significantly more likely to report common mental illness (OR 1.61, 95% CI 1.10 to 2.36), whereas eviction from rented property shows no increased risk (OR 0.97, 95% CI 0.76 to 1.20).

Discussion

Repossession significantly increases the risk of poor mental health and the pattern over time suggests that the lead up to repossession also has a detrimental effect. Post repossession, the proportion with common mental illness varies widely. Evictions from rented housing show elevated levels immediately prior to the event, but there is no increased risk after the event.

These results are derived from a large national panel survey, but the relative numbers of both events are small. A potential limitation is that those who have dropped out of the panel may be different from those who remain in the panel thus biasing the results, but the net effects were adjusted for major sociodemographic variables known to be associated with common mental illness. However, this cannot rule out the possibility that the declining mental health observed in the lead up to repossession makes the repossession more likely. The small number of events means that any interactions between variables cannot be explored, such as the double jeopardy of a marital breakdown and repossession/eviction occurring concurrently or close together. However, these data are unique in the UK and provide the only source of prospective information on those who experience repossession or eviction.

The social and economic processes leading to repossession and eviction are reasonably similar.9 However, the meanings of the events are likely to be different. Home ownership offers a heightened sense of ontological security compared to those in the more transitory rental sector.23 24 This is reflected in the increased risk prior to repossession, whereas, in contrast, eviction from rented property shows a short peak before the event, which may be a product of the temporary reduction in ontological security, but quickly returns to normal as the individuals move into other housing. Although these results show a possible consequence of repossession or eviction processes, it is not known exactly what aspect, or aspects, affects psychological well-being. However, a rise in the number of repossessions is likely to result in more people seeking help for distress and anxiety; the two main dimensions of common mental illness as measured by the GHQ.