PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of jepicomhJournal of Epidemiology and Community HealthVisit this articleSubmit a manuscriptReceive email alertsContact usBMJ
 
J Epidemiol Community Health. 2007 April; 61(4): 337–343.
PMCID: PMC2652946

Is economic adversity always a killer? Disadvantaged areas with relatively low mortality rates

Abstract

Objectives

To identify areas of Britain whose residents have relatively low age specific mortality, despite experiencing long‐term economic adversity.

Methods

Longitudinal, ecological study of all residents of Britain from 1971 to 2001.

Results

54 of Britain's 641 parliamentary constituencies were identified as having been persistently economically disadvantaged in the period 1971–2001. Within this group, there was marked variation in age group specific mortality and in the age ranges with relatively high or low mortality. A systematic scoring process identified 18 constituencies as providing strong and consistent evidence of low mortality across a range of age groups, relative to the 54 constituencies as a whole. These 18 were labelled “resilient”. Among age groups >24 years, mortality rates in the resilient areas were significantly lower than in the other economically disadvantaged areas. For example, at ages 45–59 years, the average all cause mortality rate in the resilient constituencies was 607 per 100 000 population (95% CI 574 to 641) and 728 (670 to 787) in the non‐resilient constituencies (p = 0.013).

Conclusions

Areas with similar adverse economic histories do not all have similarly high mortality rates. It is unlikely that a single factor explains these results. Selective migration cannot be discounted as an explanation, but particular sociocultural features of areas (including the political, economic, ethnic and religious characteristics of their population) may also be protective.

It has been repeatedly shown that adverse socioeconomic circumstances in an area usually have an adverse effect on the health of the population.1,2,3,4 In this paper, however, the focus is on areas that have experienced considerable long‐term economic adversity, but which have low mortality relative to other areas with similar economic histories. These areas might be doing “better than expected” or “overachieving”.5 This status implies that there may be protective factors or practices in particular areas, which weaken the usually strong relationships between economic adversity and poor health.

Those who get by, or even thrive, in a situation where most would suffer or do badly are called “resilient”. The term has been widely used within child psychology, social policy and ecology.6,7 We find Health Canada's definition of the term the most helpful.8

Resilience is the capability of individuals and systems (families, groups, and communities) to cope successfully in the face of significant adversity or risk. This capability develops and changes over time, is enhanced by protective factors within the individual/system and the environment, and contributes to the maintenance or enhancement of health. p 4

It should be noted that there are alternative definitions of resilience, and that others working in this field define resilience as a process, rather than an outcome or as being conditional on adversity.9

A small number of studies have begun to explore resilience in communities and places.8,10 A recent study by Doran and Whitehead5 found districts of England, where life expectancy was better than expected, given the level of deprivation in those areas. However, life expectancy, as a single measure of population health, may mask variation in resilience by age group, makes it harder to identify the causes of death, which have lower than expected rates and thus limits information on the potential mechanisms underlying the resilience. Further, Doran and Whitehead's focus on England excluded Britain's most deprived areas, found in Wales and Scotland.4 In this study, our aim was to extend Doran and Whitehead's work. We took a longitudinal perspective on the whole of Britain and searched for areas with the strongest evidence of relatively low mortality across a range of ages, despite experiencing persistent economic adversity.

Methods

The study had two stages. The first stage identified a group of areas with long‐term experience of significant economic adversity. The second stage identified members of this group with relatively low age specific mortality rates.

Areas, timeframe and data

All analyses were based on 641 Westminster parliamentary constituencies in Britain, as at 1997–2001. Constituency size (average population 89 000 in 2001) allowed analysis of mortality within small age groups. Furthermore, constituencies group similar numbers of people together across Britain and fragment large urban areas. UK decennial census data for 1971, 1981 and 1991, corrected for undercount as appropriate, and for which areal definitions were constant over time, were obtained from the Linking Censuses Through Time website (http://census.ac.uk/cdu/software/lct/).11 Census data for 2001 and individual level mortality data were obtained from the Office for National Statistics and the General Register Office for Scotland.

Measuring adversity

An index of adversity was created to trace the economic trajectory of each constituency over time. We did not use standard deprivation indices such as Townsend or Jarman,12 as their values cannot be compared across the entire time span of the study (1971–2001). Our index measured material rather than social disadvantage and was based predominantly on measures of labour market inactivity. We identified indicators of “adverse economic circumstances” separately for three age groups: 0–15, 16–64 and [gt-or-equal, slanted]65 years. The aim was to identify the best indicator of economic adversity, for each age group, from each census (table 11),), although the censuses vary in the variables they report and we were unable to produce entirely consistent indicators across time.

Table thumbnail
Table 1 Selected census variables as indicators of economic adversity by age groups for the four decennial censuses between 1971 and 2001

Data for smaller age groups were not available in 1971 and 1981. The indicator for children focused on their household circumstances as they have no formal relationship with the labour market. The censuses, particularly in earlier decades, offer little detail on the economic circumstances of retired people. In 1971, there were no appropriate census indicators of economic adversity for people aged >65 years and this age group was not included in the adversity index in that year. For the years 1981 to 2001, we selected car access as an indicator of adversity for this age group. Car access is often claimed to have limitations as a measure of poverty, particularly in rural areas.13 However, it is a strong indicator of social status among the elderly at the individual level14 and was closely associated with mortality in this age group. The index was the total number of constituency residents in adversity expressed as a percentage of the total population. It was strongly correlated with standard deprivation measures (r = 0.9, p<0.001 with the Carstairs index, Department of Environment's index of local conditions and Breadline Britain indices in 1991).

The adversity index was used to identify a group of constituencies with pronounced and prolonged economic adversity. We wished to identify a reasonably sized group of areas so as to maximise the chance to detect resilience. As economic adversity generally increased in the UK 1971–91, we opted to identify areas, which, in economic terms, “started badly, and got worse”. To this end, the third of constituencies with the greatest adversity score in 1971 was identified (n = 214). Within this group, the quartile of constituencies with the greatest increase in adversity score between 1971 and 1991 was then isolated. This yielded 54 constituencies, which we labelled as “persistently disadvantaged”. To confirm the suitability of the group identified, we ranked all 641 constituencies by economic adversity (rank 1 being the most deprived), in 1971 and 1991. The average rank increased from 65 in 1971 to 30 in 1991. In 1971, the least deprived constituency in the group was ranked 193, in 1991 it was 72. This confirmed that the group of 54 were persistently, and perhaps increasingly (in relative terms), disadvantaged.

Comparing mortality in persistently disadvantaged constituencies

All‐cause mortality rates were calculated for the 54 constituencies for four time periods: 1981–85, 1986–90, 1991–95 and 1996–2001. Denominators were calculated from census data using straight‐line estimates for which the rate of intercensal population change was assumed to be constant. Age and sex standardised mortality rates were calculated for the age groups 0–4, 5–9, 10–14, 15–19, 20–24, 25–29, 30–44, 45–59, 60–64, 65–74, 75–84, [gt-or-equal, slanted]85 years.

Assessing the variety of mortality patterns among the 54 constituencies was a complex task, with 2160 age group, time and area specific mortality rates to compare and contrast. We aimed to identify constituencies which had relatively low mortality, in a wide range of age groups, consistently over time and to take account of the degree of economic adversity experienced. To do this we computed a “resilience score”.

In step 1, for each age group, in each time period, we calculated the quartile boundaries of the mortality distribution in the group of 54 constituencies.

In step 2, for each of the 54 constituencies, in each time period, we counted the number of age groups with mortality rate within the best quartile of the distribution. We excluded the 5–9 years and 10–14 years age categories from this as small numbers of deaths in these groups made the rates, and thus the quartile boundaries, unstable. Counts for each time period were summed for each constituency.

In step 3, we weighted this total according to the level and persistence of economic adversity experienced across all four time periods. The weights were derived from the number of time periods in which the constituency fell in the worst half of the economic adversity score distribution, with an extra weight added for those areas which were in the worst half in all four time periods. For example, a constituency which was in the most economically disadvantaged half of the group in three time periods, had its score weighted by a factor of 3. A constituency in the most economically disadvantaged half of the group in all four time periods had its score weighted by 5.

Constituencies with an above average resilience score were labelled “resilient”. Sensitivity analysis determined the extent to which results were method dependent. Results indicated that most constituencies identified as resilient by the system described above, were identified regardless of the precise parameters of the system (data not shown).

Determining the significance of resilience for mortality

The high risk of type 1 error prohibited testing all mortality rates for statistically significant difference.15 We therefore tested for difference between average age group specific mortality rate in the resilient constituencies, and the rest of the persistently disadvantaged constituencies.

Results

Figure 11 lists the group of 54 constituencies defined as persistently disadvantaged, along with an illustrative subset of age group specific mortality rates for just one time period (1996–2001). Most of the persistently disadvantaged constituencies were in urban areas, with the greatest number in London, Liverpool, Tyneside and Glasgow, but there were some from more rural ex‐mining areas in south Wales. Figure 11 shades each cell according to the mortality rate, as described in the table key. Visualising the rates in this way allows the reader to see easily if, and at which ages, constituencies have relatively low mortality. The shading also serves to highlight between constituency variations in the ages where relatively high or low mortality rates were found. Those in Wales, for example, appear to exhibit relatively higher mortality around ages 20–24 years (roughly 80/100 000), but much lower mortality at younger and older ages. By contrast, some constituencies in the Liverpool area had particularly low rates at these ages, but higher at others. However, variation in mortality by age group within constituencies appeared relatively stable over the four time periods (data not shown).

figure ch49890.f1

The constituency resilience scores ranged from 0 to 85 years, with a mean of 21.5 years and a median of 18 years. The distribution of scores is shown in fig 22.

figure ch49890.f2
Figure 2 Distribution of the resilience index.

The five constituencies with the highest resilience score seemed distinct within the distribution, the remainder of which suggests that there is a spectrum of resilience. Constituencies with a resilience score value above the average are highlighted in fig 22 and identified in table 22.

Table thumbnail
Table 2 Constituencies with above average resilience score

Table 33 gives results of the Mann–Whitney U tests for difference in age‐specific mortality rate (1996–2001) between the resilient constituencies, and the rest of the persistently disadvantaged group, along with mean mortality rates. Results for other time periods were similar (data not shown).

Table thumbnail
Table 3 Differences in age specific mortality rates for “resilient” and “non‐resilient” constituencies (1996–2001)

There were no significant differences in mortality at ages 0–14 years between the resilient and non‐resilient, persistently disadvantaged constituencies. As previously noted, at ages 5–14 years there were few deaths and resilient constituencies were not selected on the basis of death rates in these age groups. There were also no significant differences between resilient and non‐resilient constituencies at ages 20–24 years. At other ages, the mortality in the resilient areas is consistently and markedly lower than in other economically disadvantaged areas. We tested for differences in mortality between the five most resilient constituencies and the remaining 49 persistently disadvantaged constituencies, finding significantly lower rates among the most resilient for age groups 15–19 years and 30–44 years only (data not shown).

Figure 33 compares the age group specific mortality rates in the resilient and non‐resilient constituencies (which shared a similar economic history), and between the resilient constituencies and the British average. Figures are expressed as percentage differences in mortality rate. Thus, a negative value denotes that the rate in the resilient constituency is lower than its comparators, and a positive value denotes a higher rate. The graph shows that mortality among younger adults in the resilient constituencies was about 20–25% lower than in the other persistently disadvantaged constituencies, and about 5–10% lower among older adults. However, at most ages, mortality rates in the resilient constituencies were still higher (20–30%) than the British average.

figure ch49890.f3
Figure 3 Comparison between mortality in resilient and non‐resilient constituencies, and between resilient constituencies and the British average (1996–2001).

Discussion

This study identified a group of constituencies with significantly lower mortality, at a range of ages, relative to other constituencies with similar adverse economic histories. It also showed that “resilience” varies markedly by age group and that resilience may be detected in Welsh, but not in Scottish, constituencies. These findings extend those of a previous study, which only focused on England and which used a single measure of life expectancy.5 A clear finding, however, is that although the resilient constituencies have low mortality relative to their economic peers, their rates remain high relative to the British average. The effects of economic disadvantage on health are lessened but not entirely removed.

Methodological limitations

The results must be considered in the light of limitations in the methods and data. Census frequency limits the measurement of constituency economic trajectory. Unemployment rates within areas can change rapidly over short time periods, meaning both booms and busts may have been “missed” if they occurred within an intercensal period. Also, the timing of the census affects what it records. Censuses in 1981 and 1991, for example, fell in the middle of recessions which affected different parts of Britain at different times.16 Changes in the structure of census data over time meant that the component indicators of adversity for a specific age group could not be held exactly constant. Furthermore, the cultural and socioeconomic character of life in Britain also changed substantially between 1971 and 2001, making comparison of adversity over time more difficult. For example, labour market activity of women changed considerably between 1971 and 1991 and will have changed the probability of economically inactive women describing their status as “unemployed” in the census.

However, the adversity scores themselves were not central to the identification of lower than expected mortality once the group of persistently disadvantaged constituencies had been defined. The group included a wide range of types of areas, both urban and rural, from across Britain, suggesting that the measure reflected a wide range of experiences and was not overly sensitive to one type of adversity at the expense of others.

The definition of resilience adopted by us was conditional on economic disadvantage. An area could only be identified as being resilient if it was in the most disadvantaged third of constituencies in 1971 and in the 25% of that group which experienced the greatest subsequent increase in adversity. Although this approach had the advantage of simplicity, these inclusion criteria will have influenced the results. Sensitivity analyses suggested that varying the parameters of the selection process did not dramatically change the list of areas identified as resilient. Nonetheless, areas which were not already in economic adversity in 1971, but which had catastrophic decline afterwards, and those which were very disadvantaged in 1971, but which did not decline a great deal further, were excluded.

We recognise that our choice of areal units will have dictated the results to some extent—this is the perennial problem of ecological analysis. Constituencies are relatively large and heterogeneous. Smaller resilient neighbourhoods may have been ignored because their candidacy was diluted by aggregation with other neighbourhoods that made up the constituency. Further work to explore the effect of areal unit selection is required.

What is already known

  • Adverse socioeconomic circumstances in an area usually have an adverse affect on population health.
  • Those who do get by, or even thrive, in a situation where most would suffer or do badly are called “resilient”.

What this paper adds

  • This study is the first to identify a group of areas in Britain which had prolonged economic adversity, but which have considerable lower age group specific mortality relative to other constituencies with the same adverse economic histories.
  • Diversity in the range of ages where mortality is lower, and in the types of area identified, suggests that there is no single factor responsible for this apparent resilience.
  • The processes which convert economic adversity into higher mortality are weakened in some disadvantaged areas, perhaps by selective migration, by protective characteristics of the community, or by progressive local policies.

Policy implications

  • There are practices and policies which weaken the detrimental health effects of economic decline in an area.
  • If some areas can resist the translation of economic adversity into higher mortality, other areas can learn from their policies and approaches, so that they are better protected when economic recessions arrive.

In calculating the resilience score, constituencies were credited for each age group in which they had mortality rates in the lowest quartile of the distribution, relative to their economic peers. This approach has an important advantage in recognising that mortality varies by age. However, using quartiles to assess a distribution means that a group of death rates are always identified as “best”, regardless of how low they actually are. Yet, if variation in mortality rates within the persistently disadvantaged constituency group was random, the resilience scores would be generally similar (fig 11 shows they are not) and there would be no significant difference in mortality between the constituencies with higher and lower resilience scores.

Finally, the present analyses and results are not disaggregated by sex. Men and women can have different exposures and responses to material deprivation. We intend to present these analyses in a dedicated paper at a future date.

Explaining the results

Although this secondary analysis was not designed to explain the resilience it has detected, it is useful to consider some plausible hypotheses. In this section we “prepare the ground” for future work to explain the results.

Exploration of the mortality rates by cause (data not shown) shows that some areas have lower than expected rates of cancer, whereas others did well in cardiovascular disease, suicide or even accidental deaths. This diversity (hence the variety of aetiological pathways which must be being influenced), strongly suggests that there is no simple “x factor” which is protecting health in these areas.

It must also be remembered that these analyses are of people grouped by area, not of individuals. Processes which influence area level mortality can be at both an individual and an ecological level.17 Macintyre et al17 offer a range of characteristics by which the influences on health in an area can be assessed and we use an adapted version of these to weigh possible mechanisms by which the resilience might be occurring.

The composition of an area's population is usually the greatest influence on its mortality rate. An economically disadvantaged area may, for example, “acquire” lower mortality via selective immigration of a healthier population.18,19 Retaining or attracting population can also stem the erosion of public services and foster social capital, benefiting both the incoming and existing populations.20 Population loss 1971–91 was about one third lower in the resilient constituencies compared with the 36 other persistently disadvantaged areas. It thus seems plausible that the resilient areas have done better at retaining, or attracting new, population and that this may have contributed to their resilience. However, even if keeping or attracting population is part of the process by which population level health resilience is attained, the question remains: why do some areas succeed in these processes while others apparently do not?

Macintyre et al17 also suggest five types of features of the local area, which could influence residents' health. These are: (1) physical features of the environment shared by all residents in a locality (eg, quality of air and water, latitude and climate); (2) the availability of healthy environments at home, work and play; (3) services provided, publicly or privately, to support people in their daily lives (including education, transport, policing, health and welfare services); (4) sociocultural features of a neighbourhood (including the political, economic, ethnic and religious history of a community: norms and values); and (5) the reputation of an area (how it is perceived by residents, service or amenity planners and providers, and investors). Initial investigation has yielded some evidence for positive characteristics under each of these headings, in at least some of the resilient constituencies. The geographical diversity of the resilient constituencies makes it unlikely that they all offer similarly benign or beneficial physical environments. The shared experience of economic adversity, and in many cases, community ties based on former industry of occupation, ethnic or religious identity, makes these constituencies a group in which levels of social cohesion are perhaps higher than average. However, this hypothesis remains to be tested. Further systematic research is underway to determine the recipe for resilience.

Acknowledgements

We are grateful to Prof Mel Bartley for comments on an earlier draft.

Footnotes

Funding: This work was funded by the United Kingdom Economic and Social Research Council as part of the Research Priority Network on “Human capability and resilience” project no L326253061. RM and SP are also funded by the Chief Scientists Office of the Scottish Executive Health Department. The opinions are of the authors, not the funders.

Competing interests: None declared.

References

1. Mitchell R, Shaw M, Dorling D. Inequalities in life and death: what if Britain were more equal. Bristol: Policy Press, 2000
2. Mitchell R, Gleave S, Bartley M. et al Do attitude and area influence health? A multilevel approach to health inequalities. Health Place 2000. 667–79.79 [PubMed]
3. Townsend P, Davidson N, Whitehead M. Inequalities in health: the black report. Harmondsworth: Penguin, 1982
4. Shaw M, Dorling D, Gordon D. et alThe widening gap: health inequalities and policy in Britain. Bristol: Policy Press, 1999
5. Doran T, Drever F, Whitehead M. Health under‐ and over‐achievement in English local authorities. J Epidemiol Community Health 2006. 60686–693.693 [PMC free article] [PubMed]
6. Luthar S, Zelazo L. Research on resilience: an integrative review. In: Luthar S, ed. Resilience and vulnerability: adaptation in context of childhood adversities Cambridge: Cambridge University Press, 2003. 510–550.550
7. Adger W N. Social and ecological resilience: are they related? Progress in Human Geography 2000. 24347–364.364
8. Stewart M, Reid G, Buckles L. et alAtlantic Health Promotion Research Centre. A Study of Resiliency in Communities. Ottawa: Office of Alcohol, Drug and Dependency Issues, Health Canada, 1999
9. Mitchell R, Backett‐Milburn K. Health and resilience: what does a resilience approach offer health research and policy? Edinburgh: RUHBC, University of Edinburgh, RUHBC Findings series 2006
10. Gerrard N, Kulig J, Nowatzki N. What doesn't kill you makes you stronger: determinants of stress resiliency in rural people of Saskatchewan, Canada: J Rural Health 2004. 2059–66.66 [PubMed]
11. Mitchell R, Dorling D, Martin D. et al Bringing the missing million home: correcting the 1991 small area statistics for undercount. Environ Plan A 2002. 341021–1035.1035
12. Jarman B, Townsend P, Carstairs V. Deprivation indices. BMJ 1991. 303523 [PMC free article] [PubMed]
13. Shucksmith M. The definition of rural areas and rural deprivation. Edinburgh: Scottish Homes, 1990. 2
14. Gardiner C, Hill R. Analysis of access to cars from the 1991 UK census samples of anonymised records: a case study of the elderly population of Sheffield. Urban Studies 1996. 33269–281.281
15. Altman D G. Practical statistics for medical research. London: Chapman and Hall, 1991
16. Green A E, Owen D W, Winnett C M. The changing geography of recession —analyses of local‐unemployment time‐series. Transact Institute Br Geographers 1994. 19142–162.162
17. Macintyre S, Ellaway A, Cummins S. Place effects on health: how can we conceptualise, operationalise and measure them? Soc Sci Med 2002. 55125–139.139 [PubMed]
18. Bentham G. Migration and morbidity—implications for geographical studies of disease. Soc Sci Med 1988. 2649–54.54 [PubMed]
19. Brimblecombe N, Dorling D, Shaw M. Migration and geographical inequalities in health in Britain. Soc Sci Med 2000. 50861–878.878 [PubMed]
20. Davey Smith G, Shaw M, Dorling D. Shrinking areas and mortality. Lancet 1998. 3521439–1440.1440 [PubMed]

Articles from Journal of Epidemiology and Community Health are provided here courtesy of BMJ Group