Health costs and benefits associated with economic transitions: Linking records of address change, property value, and self‐reported health

Deprivation is a major risk to population health, in particular when experienced during childhood. Poor health in the early years accumulates and is expressed in adult health inequalities. Policy makers may aim to mitigate against the ill effects of deprivation by trying to increase social mobility and facilitating moves towards better earnings and living conditions or by protecting against the effects of downward moves and the experience of deprivation. This paper uses address change and property value data at the individual and family level to examine whether poor health outcomes occur more frequently among people who move between addresses and particularly those who move to properties with lower property values. We use the Northern Ireland Longitudinal Study, linking health and demographic data from 2001 and 2011 Censuses to house valuations for a representative 28% of the population aged 10 – 64 years ( N = 342,681). Young persons (aged 10 – 15 years) living in a house valued at over £ 160,000 were half as likely to be reported as having mental ill health as those living in a house valued under £ 75,000 (OR = 0.49, CI: 0.31 – 0.78). There was no strong evidence that upward or downward mobility affected mental health or physical health for young people, but ill health of working aged persons showed a strong association with moving to houses of lower value. Results are discussed in terms of their implications for understanding the dynamics of social mobility and health and in terms of how various policies towards poverty may influence population health.

Zadie Smith, BBC (2013) This Strategy restates the Government's commitment to tackle poverty at its sourcebe it family breakdown, educational failure, addiction, debt or worklessness. Ian Duncan -Smith, Child Poverty Strategy (2014) There have been recent attempts to refocus the study of health inequalities on the systems and social structures that underpin them (cf. Lynch, 2017;McCartney et al., 2019). Evidence points to healthier populations in countries with lower systemic inequality (Wilkinson & Pickett, 2009) and greater emphasis on welfare and social security (McAllister et al., 2015;McCartney et al., 2019). This gives impetus to researchers to look closely at emerging evidence adding to understanding on social gradients of health and to examine how these patterns reflect the societal context in which they occur.
Governments around the world vary in the extent to which they prioritise, and how they address, the issue of poverty. In the United Kingdom, successive Westminster governments have emphasised the root causes of poverty as their policy focus, with greater social mobility and opportunity for economic advancement identified as remedies, as exemplified by the above quote from the 2014 Child Poverty Strategy by the then Work and Pensions Secretary. Given the known costs to health of economic deprivation, particularly during childhood, social mobility and the opportunity to achieve greater affluence would appear to be a route to better health (Cardano et al., 2004;Hahn & Truman, 2015;Sage, 2013).
However, social mobility as a policy goal has been under a critical spotlight for some time. Educational expansion and labour market activation as policy responses to poverty reflect a framing of the issue at the level of the individual and in terms of why certain individuals find themselves in poverty (Leeman et al., 2016). Successful implementation of such policies might give those same individuals competitive advantages for the attainment of greater wealth. However, this emphasis displaces from the policy discourse any structural analysis of underlying social stratification and casts social protection for those out of work and in poverty as a second-order priority (Van Berkel, 2010).
Greater affluence and prosperity is consistently associated with better physical and mental well-being, whereas illness disproportionately affects those experiencing greater socio-economic deprivation (Marmot, 2017). Childhood socio-economic conditions have been shown to impair long-term cognitive development and well-being, independent of conditions experienced later in life (Hayward & Gorman, 2004;Schoon et al., 2012).
However, while persistent deprivation and stable affluence represent the least and most salutogenic social circumstances (Forrest et al., 2018), there is mixed evidence as to whether increasing one's income, social class, or social status will bear dividend for health. This literature is beset by complexity, particularly around the degree to which moving from deprivation directly causes health improvement and to which health is an advantage or even a prerequisite for economic advancement (Boyle et al., 2009;Doyal & Gough, 1991). In the absence of significant economic growth or reordering of class structures, opportunities for upward social mobility necessitate others moving downward (Bushcha & Sturgis, 2018). Given that the salience of a loss often exceeds that of a gain (Kahneman & Tversky, 1979), we might expect the emotional cost of a loss of social class status might exceed the benefit accrued to those moving in the other direction.
Furthermore, different accounts exist for how socio-economic advantage translates to health benefits: from psychosocial costs such as status anxiety (Wilkinson & Pickett, 2009), children's exposure to emotional distress (Treanor, 2016), to material benefits such as quality housing and car access (Dunn, 2002;Robertson et al., 2015). This study avails of linked data on address change and estimated house value to contextualise the distribution of health and ill health. Through the use of the Northern Ireland Longitudinal Study, health as reported on the 2011 Northern Ireland Census could be analysed for a large proportion of the NI population. Although it is challenging to deduce the precise causal effect of changes of address and occupation over a 10-year period, the longitudinal structure and large sample size of these data render them useful for further unpacking how these changes in material circumstances relate to physical and mental health.

| HEALTH IMPACTS OF SPATIAL AND RESIDENTIAL MOBILITY
(Our) task is ownership for all. The desire to have and to hold something of one's own is basic to the spirit of man.
Margaret Thatcher (1985) One of the gifts that supposedly come with a middle class and upper middle-class life is this isolation: living in a house which is only yours; if you are fortunate enough you do not have to speak to your neighbours … the higher you go the less contact you have with other people. That's the ideal: that at some point you'll have this perfect isolation. Zadie Smith, BBC (2013) The opportunity of moving to better quality housing represents a material improvement in circumstances and environment offered by social mobility, which could be expected to improve health outcomes (Battel-Kirk & Purdy, 2007;Dunn, 2002). Furthermore, the built environment and norms within a neighbourhood can influence an individual's behaviour and attitudes around physical health and personal resilience (Commission on Social Determinants of Health, 2008; Thomson, Thomas, Sellstrom, & Petticrew, 2013). A report by Shelter found that doctors viewed poor housing conditions as driver of mental ill health. However, the same study reported that members of the general public identified affordability as the main factor having a negative impact. This highlights the complexity of upward mobility being expressed in moves to higher value property, in that material conditions may improve but financial demands may introduce strain.
Furthermore, change and upheaval are often stressful. Major life events, such as bereavement, job-loss, or loss of a family home can severely impact mental health (Holmes & Rahe, 1967 (Speirs et al., 2015). In a synthesis of available evidence on the impact of instability on child development, Sandstrom and Huerta (2013) argued that a change of address may place demands on young people's psychosocial resources, in particular through disruption to their peer relationships. Both change of residence and change of financial state were included in the original Social Readjustment Scale, which weighted the impact of major life events (Holmes & Rahe, 1967). Moves to less desirable homes may be indicative of involuntary mobility, which influences family-level stress through the removal of control and reduction in self-efficacy, particularly in instances of foreclosure (Sandstrom & Huerta, 2013). Additionally, Rumbold et al. (2012) found disruptive effects of house moves on young people, regardless of whether the move represented an upward or downward transition. Further mechanisms underlying this disruption may include the diversion of parents' energy and time onto the moving process, as well as the disorienting effect of the changed physical surroundings. Popham, Williamson, and Whitely (2014) found evidence of decline in health in the period preceding a change of address, although noted that there was no evidence that these declines would endure. This suggests that benefits of mobility need to be of sufficient magnitude to outweigh and outlast any health costs arising from upheaval.
Much research on residential mobility and health revolves around transitions between different housing tenures, from renting to home ownership and vice versa. Living in rented accommodation is characterised as stress enhancing both in terms of social status (Gregory et al., 2018) and of accumulation of stressors (Robertson et al., 2015). However, Smith et al. (2003) argued that although owner-occupation has health benefits, the emotional stress and practical difficulties in obtaining homeowner status and maintaining a home in suitable condition can also cause damage to health. Baker, Bentley, and Mason (2012) found no evidence of mental health dividend from moves between renting and home ownership, whereas Popham et al. (2014) also found no evidence of improvements to mental health among people who had availed of the Right to Buy their socially rented property. However, introducing a further level of complexity, Mason et al. (2013) suggested that the mental health impact of housing affordability may be less for home owners than for private renters. This latter study highlights the potential value of using data, which simultaneously captures changes in housing tenure as well as the value the properties being moved from and moved to. status, household composition, tenure of household, and social class as determined by occupation and type of employer. An advantage of using data on this size of sample is the opportunity to detect associations within relatively small groups and, thus, to adjust for a large set of potentially confounding covariates.

| INVESTIGATING RESIDENTIAL MOBILITY USING THE NORTHERN IRELAND LONGITUDINAL STUDY
NILS has been used previously to advance knowledge of residential mobility within Northern Ireland. Through linkage to address change data as captured in the Northern Ireland Health Card Registration system (NIHCR), Shuttleworth, Barr, and Gould (2013) found that internal migration in Northern Ireland is strongly patterned by social class characteristics of individuals and areas. More deprived areas experienced net out-migration between 2001 and 2007 whereas less deprived areas experienced net in-migration. Internal migration during this period was found to be more likely among Catholics, among people with more education and among healthier people.
Health outcomes for this study were determined from the 2011 NI Census in which a range of questions were asked about health conditions. Physical ill health was defined as having a health problem or disability which limited day-to-day activity and lasted at least 12 months. Mental ill health was defined as a positive response to the question "Do you have any of the following conditions which have lasted, or are expected to last, at least 12 months?": "an emotional psychological or mental health condition such as depression or schizophrenia." Prevalence of these two items in 2011 is shown in Figure 1.

| HYPOTHESIS AND ANALYTICAL STRATEGY
Our central hypothesis is that moving to a property of a different value is associated with differential risk of mental and physical ill health, with "upward" moves to higher value properties associated with lower health risk and "downward" moves to lower value properties associated with greater risk of ill health. This hypothesis encom-

| RESULTS
The total number of NILS members aged 10-15 years in 2011 for whom data from the 2001 census were also available was 38,612, whereas 44,886 were aged 16-24 years. Descriptive statistics presented in Table 1 demonstrate an even gender split in both age groups.
Poor mental health is more prevalent in the emerging adulthood group (3% vs. <1% in the youth group). Limiting illness occurred in a similar F I G U R E 1 Prevalence of ill health by age Poor health is experienced least often by those in consistent affluence and most often by those in consistent deprivation. Figure 3 shows   Table 2 show some support for an association between poor mental health and lower value housing in both early childhood and adolescent groups (Graphics of corresponding predicted probabilities appear in Figure 4) Risk of physical disability was also correlated with value of current and prior address. However, the patterning of effects are different in these models. This could be considered as weak evidence that house value impacts more strongly on mental health than on physical health for this age group.

Logistic regressions presented in
Moving to the "Mobility models" of ill health presented in likelihood of physical ill health among those whose families had moved from the highest property value band to the lowest, Taking these results together, it is clear that a person's age and stage of life is an important factor to consider when assessing how socio-economic and residential may influence that person's health.
While some of the patterning and nonpatterning in Table 3 may be due to smaller population subgroups in which significant associations are more difficult to detect, it does appear that associations with upward property value moves in particular vary substantially between the focal age groups. In Appendix A, we show results from a restricted version of the mobility model that excludes nonmovers and where the reference group is persons who moved address but to an address of the same value. If it was the case that the instability or upheaval was driving the health risks to movers, we would likely see changed coefficients in this table. We see little substantial difference, suggesting that upheaval is not the primary mechanism and that changes in housing and socio-economic conditions are more likely drivers of the observed associations. This supports the finding of Popham et al. (2014) that distress experienced in the period immediately around address change abated after a short period.

| Sensitivity analyses
For the emerging adulthood group, we specified a model restricted to those young people whose address in 2011 was with their parents and therefore whose change in property value was not also associated with a move to independent living. Continuing to live with parents was associated with lower odds of poor mental health housing. Alternatively, this may be an example of the emotional response to a loss being of greater salience and magnitude than that of a gain, along the lines suggested by the prospect theory (Kahneman & Tversky, 1979). Limiting analysis to only persons who experienced a move suggested that residential instability was not a major contributor to the observed effect.
Taken together, our results suggest social protection policies and measures such as mortgage debt relief schemes, aimed to keep people in their homes, or interventions to improve the quality and availability of housing stock may be more effective in reducing population-level ill health than policies focused on facilitating progression up the property ladder.

| Additional strengths and limitations
This study is unique in exploiting house value data at the individual level as a proxy for affluence. We have been cautious in our interpretation of the mechanisms underlying the observed associations between residential moves and ill health. First, the associations, while nontrivial in magnitude, are imprecisely estimated, as captured by the wide confidence intervals around estimated associations. Second, the direction of causality is difficult to infer, even with the availability of a large array of 2001 control variables used to limit sources of confounding. For example, a child developing a disability, or poor mental health, may prompt a family to make either an upward transition (e.g., moving to a larger home or one with appropriate adaptations) or a downward transitions (e.g., due to loss of income through care requirements).
An additional limitation is the temporal gap between the 2005 property valuation exercise and the two observation points. This necessitates the assumption that value remains relatively stable between 2001 and 2011 or that the relative value of properties remains reasonably constant throughout any inflationary or deflationary period. However, house prices remained stable in NI, and the price categorisation allows for an upwards or downwards change in market price of £20,000 without a change in pricing category.
The available health outcomes are rudimentary binary indicators of illness category. The 2011 Northern Ireland Census was novel in disaggregating sources of ill health and disability, but categories remain crude. However, NILS have proven to be good indicators of health (Young et al., 2010).

| CONCLUSIONS
Our findings do not support the expectation that upward social mobility provides health benefits for children who begin their lives in deprived households. Our principal recommendation is that, from the perspective of public health, social protection policies that aim to prevent slides into poverty are likely to have a greater impact than social activation-style initiatives that seek to facilitate moves from deprivation to affluence. This may help to explain why population health has been found to be greater in countries with more redistributive models of political economy (McCartney et al., 2019).
Furthermore, the idea that the early years of life constitute a critical period in which the effects of poverty are most profound and long lasting is accepted among policy makers, and early years intervention is already a high priority spending area in many countries. However, the current study reminds us that healthy development can be disrupted beyond the early years and that movement to either better or worse residential circumstances could be detrimental to the health of adolescents.