Journal of Public Health Medicine
Vol. 20, No. 2, pp. 161-168
Printed in Great Britain
The two communities in Northern Ireland:
deprivation and ill health*
D. O'Reilly and M. Stevenson
Abstract
Background The aim of this study was to examine differences in socio-economic standing and ill health between the
two communities in Northern Ireland.
Methods This was a descriptive epidemiological study.
Deaths from 1991 to 1995 inclusive were used to calculate
standardized mortality rates (SMR, under 75 years) at small
level using the 1991 Census population estimates. The
standardized limiting long-term illness ratios (SIR) were
based on the appropriate Census question. Regression
models were tested with SMR and SIR as dependent
variables and a wide range of socio-economic indicators,
including income support and family credit uptake, as
independent predictors.
Results Northern Ireland is a very polarized society. More
than 60 per cent of the population live in areas which have
more than 80 per cent of one religion. Areas with a
preponderance of Catholics tend to be more deprived.
Unemployment rates, percentage renting, car availability,
and education attainment are all worse in Catholic areas.
However, there is considerable heterogeneity between areas
with similar levels of religious affiliation and the overall
pattern varies with the indicator chosen. SMRs rise stepwise
with increasing percentage of Catholics. SIRs increase with
increasing polarization of areas, but this is much more
marked in those with a predominantly Catholic affiliation.
Altogether 46.8 per cent of the variance in SMR and 77.9 per
cent of that of SIRs could be explained by socio-economic
variables alone. Denomination did not have any residual
predictive value.
Conclusions Policy-makers should continue to periodically
monitor for differences between the two communities
including any differences in service accessibility and
uptake. Efforts should be directed towards reducing the
inequalities in health for all sections of the community.
Keywords: religion, deprivation, ill health
Background
Northern Ireland is one of the most deprived regions of the
United Kingdom. It consistently has the highest levels of
unemployment with an even greater disparity in the rates of
long-term unemployment Average gross weekly earnings per
household and per person are the lowest of any of the four
countries,1 and there is a much greater proportion of the
population reliant on social security benefits. The relationships
between material disadvantage and ill health are as evident in
Northern Ireland2 as they are in other parts of the United
Kingdom.3 The problems of Northern Ireland, however, are
often seen in terms of the differences between Catholic and
Protestant, i.e. between the 'Two Communities',4'5 yet the
amount of research into possible differences in health between
the 'Two Communities' in Northern Ireland is surprisingly
meagre.6
A random survey of fifth-form school children in Northern
Ireland in 1976-1977 found that a greater proportion of
children attending Catholic schools had tried smoking and
were still smoking compared with those attending Protestant
schools.7 A community health study in the 1970s8 showed that
an excess alcohol consumption, which, at that time, was rare
amongst women, was virtually absent amongst Catholic
females. In contrast, Catholic males had generally higher
prevalence of excess drinking than males of the other combined
denominations. More recently, a study of cancer screening
practices amongst women in Northern Ireland showed that
breast self-examination and attendance for cervical cancer
screening were more common amongst Protestant than Catholic
women.9
An ecological study by Stringer10 related measures of ill
health to data from the 1981 Census following similar work by
Townsend in the Northern region of England3 but extended to
include denomination and rurality. That study showed that
Protestants had higher mortality rates than Catholics once
differences in deprivation were discounted. However, these
researchers limited the range of deprivation variables to those in
the Townsend index of material deprivation and there were
problems related to a significant level of under-enumeration in
the 1981 Census in Northern Ireland arising from the levels of
civil unrest there at that time.11
In a secondary data analysis of the Continuous Household
Survey (1990-1991),12 it was shown that in all aspects of
Health and Health Care Research Unit, Queen's University of Belfast,
MuDiouse Building, Grosvenor Road, Belfast BT12 6BJ.
D. O'Reilly, Deputy Director
M. Stevtnsoo, Statistician
Address correspondence to Dr D. O'Reilly.
•This paper was commissioned by the editors.
© Oxford University Press 1998
162
JOURNAL OF PUBLIC HEALTH MEDICINE
lifestyle investigated, Catholics were more likely to engage in
behaviour that could adversely affect their health. These
differences persisted, even when differences associated with
other demographic and socio-economic characteristics such as
age, social class and income had been taken into account.
However, despite this, any differences in the levels of selfreported health status between the two communities were small.
The aim of this study is to relate, at an ecological level,
measures of ill health and deprivation with denomination
within Northern Ireland using more contemporaneous, complete and reliable data. The relationships between (1) socioeconomic standing and religious affiliation and (2) ill health and
religious affiliation will be reported separately, and in the
final section the results of multivariate analysis exploring for
any effect of religion on ill health which was independent of
socio-economic standing will be reported.
Data sources and methods
The 1991 Census provided the population estimates to produce
the standardized ratios used in the analysis. Socio-economic
variables were also derived from the 1991 Census, with the
exception of the numbers of recipients of family credit and
income support, which came from social security benefits data
for 1996. Long-term unemployment defined as 'having no paid
job within the last ten years', was also Census derived. The
income support data had age attached and it was possible to
produce standardized estimates in addition to the unstandardized percentage estimates. The denominator for family credit
was the Census estimates of the numbers of household in each
area. Morbidity data were based on the 1991 Census limiting
long-term illness (LTT) question. Only those persons in private
households were used, to reduce the influence of large
establishments on estimates at small areas as suggested by
Carstairs.13 This was standardized for age and sex. Deaths
for the years 1991-1995 inclusive were used to produce
standardized mortality ratios (SMRs).
The geographical unit of analysis was derived from the 566
Census wards. To reduce distortion caused by small numbers of
events, wards were aggregated with an adjacent one so that
none had a population less than 2000 persons. Theresulting498
areas were called Synthetic Electoral Wards (SEW).
SEWs were divided into seven categories according to the
overall religious affiliation. The divisions can be seen in Figs
1-3. The distribution of socio-economic and health variables
within each of these categories is shown by means of box and
whisker plots showing the median, 25th and 75th percentiles
and range of values. All standardizations were by the indirect
method using the overall Northern Ireland rates. In the final
modelling process SMR (<75 years) and LTI ratios (<75 years)
were the dependent variables and a much larger range of socioeconomic variables were entered as potential explanatory
variables. Different models were tried with denomination
tested as both a continuous and as a categorical variable.
Denomination in Northern Ireland
The denominational breakdown in Northern Ireland at the time
of the last Census was as follows: Roman Catholics 605 639
(38.4 per cent), Presbyterians 336891 (21.4 per cent). Church
of Ireland 279280 (17.7 per cent), Methodist 59517 (3.8 per
cent), Others 122448 (7.8 per cent), none 59234 (3.8 per cent),
and not stated 114 827 (7.3 per cent). The vast majority of those
in the 'Other' group were various denominations of ProtestanL
For the purposes of this exercise all Protestant denominations were taken together. A total of 11.1 per cent of the
population provided the answer 'none' or 'not stated' for the
religion question. There were numerous possible ways to
handle these data; for example, either or both could have been
subtracted from the total population denominator. However,
how these were manipulated made little difference to the
religious affiliation characteristics of the ward (correlation
between the different measures of religious affiliation at SEW
level was of the order of 0.99 or higher). In the end, in true
Northern Ireland tradition, the 'none' and 'not stated' groups
were apportioned between Catholic and Protestant denominations in correspondence to the ratio of those two groups within
that area.
Northern Ireland is a segregated society; 40.8 per cent of the
population live in electoral wards which have more than 90 per
cent of one religion and 59.9 per cent of the population live in
wards which have more than 80 per cent of one religion. As
shown in Fig. 4, Northern Ireland is also polarized geographically. Areas with high concentrations of Protestants are found in
the Greater Belfast area (with the notable exception of West
Belfast) and towards the northeastern quadrant of the province.
High proportions of Catholics are found to the South and West
of Northern Ireland and in the Border areas.
Denomination and socio-economic
standing
There are a great many possible socio-economic indicators that
could be used to compare the two denominations, but only a few
will berelatedhere for illustrative purposes. These are shown in
Figs 1 and 2. It is evident that there is a greater or lesser degree
of heterogeneity within all categories of religious affiliation.
Within the three categories with a predominantly Catholic
affiliation, increasing heterogeneity in socio-economic indicator is evident with increasing concentrations of Catholics. Thus
there is generally a smaller amount of variability in socioeconomic standing in those areas with 60-80 per cent Catholic
than in SEWs with a higher proportion of Catholic. SEWs with
a population that is more than 90 per cent Catholic are found
throughout the entire range of advantage-disadvantage. This is
particularly evident with the unemployment, Townsend (not
shown), children in non-earner households and income support
variables. Arguably the same pattern is evident, though to a
lesser degree, in areas that are predominantly Protestant
DEPRIVATION AND ILL HEALTH IN NORTHERN IRELAND
The association between socio-economic indicators and the
religious affiliation of the area might also be grouped into three
broad groups. In the first group an increasing proportion of
Catholics in an area is associated with increasing disadvantage.
This is the largest of the three groups. It is those socio-economic
indicators that are most closely associated with low income or
earning capacity that show the clearest relationship between
denomination and disadvantage. Thus for the variables
unemployment, long-term unemployment, children in nonearner households, educational attainment and income support
there is a clear and positive association between the percentage
of Catholics in an area and the level of disadvantage. Family
credit follows the same general pattern but levels off at a higher
level in the areas of higher Catholic concentration.
Some indicators show little or an anomalous relationship
with denomination (Fig. 2). Social class, defined as the
% Unemployment
oo
so
•40
3O
2O
-1 O
% Children In Non-eamlng Households
% Overcrowded Households
20
1 o
o J
% Persons (16+) on Income Support
70
00
00
40 •
3O
2O •
1O
O
% Households in receipt of Family Credit
1 5
10
5
163
•
9O+% Prot.
60-80% ProL
60-80% Cam.
9O+% Oath.
SO-9O% Prot40-60%
80-90% CatH.
Figure 1 Relationship between religious affiliation and socio-economic standing (I).
164
JOURNAL OF PUBLIC HEALTH MEDICINE
percentage of households with a head in the manual classes,
shows no overall association with the religious affiliation of an
area except for the greater heterogeneity amongst the
predominantly Protestant areas. The 'percentage of households
that are rented' and the 'do not have access to a car' variables
demonstrate a relationship with denomination only in Catholic
areas. There was no evidence of a relationship between these
indicators and the percentage of Protestants. The proportion of
households without central heating is highest in those areas that
have a more equitable distribution of Catholics and Protestants.
There are some indicators that would suggest greater need or
disadvantage is present in Protestant areas. For example, the
proportion of elderly living alone is higher in Protestant areas
than in Catholic areas (Fig. 2). A similar relationship exists for
the proportion of residents that changed address in the year
before the 1991 Census.
Denomination and ill health
Figure 3 shows the variations in premature mortality ratios
(SMR <75) and morbidity (LTT <75) within each of the seven
categories of religious affiliation. There is a 2-7-fold variation
% Households with head In manual class
% Households without central heating
4O
3O
2O
1 O
oJ
% Residents who changed address one year ago
o J
% Elderly living alone
eo
2O
9O+% Prot
60-80% Prot
60-80% Cath.
90+% Cath.
80-90% Prot
40-60%
80-90% Cath.
Figure 2 Relationship between religious affiliation and socio-economic standing (II).
DEPRIVATION AND ILL HEALTH IN NORTHERN IRELAND
in mortality ratios within each category; the greatest variation is
evident within the most segregated categories, both Catholic and
Protestant There is a clear positive relationship between
increasing percentage of Catholics within a category and its
overall mortality experience. The medians displayed in the figure
are very close to the SMRs calculated for the categories as a whole.
A similar pattern is evident in the LTI ratios, with the
greatest variation within denomination category to be found in
the over 90 per cent category and increasing average levels of ill
health with increasing levels of Catholic affiliation. The
association between denomination and ill health is stronger
for long-term illness ratios than for mortality.
Disadvantage, ill health and denomination
In the final modelling process SMR (<75 years) and LTI ratios
(<75 years) were the dependent variables and 20 different
socio-economic measures were entered as potential explanatory
variables. The modelling process was repeated with denomination entered as both a continuous and as a categorical variable.
The results were unchanged.
Table 1 shows the results of the models derived which best
'explained' the mortality and morbidity variations; 46.8 per cent
of the variance in the premature mortality ratios between areas
was explained by only three variables (income support, change of
address in the last year and the percentage of adults without
qualifications). A much greater proportion of the variation in
morbidity ratios was explained by the socio-economic variables
(77.9 per cent). In both models income support was a particularly
strong predictor. The religious affiliation of an area, whether as a
categorical or a continuous variable, did not significantly
increase the predictive power of the model.
Standardized long-term illness ratios (<75 years)
200
150
100
50
0J
Standardized mortality ratios (<75 years)
200
1501
100
50
90+%
165
Prot.
60-80*. Prot.
60-80% Cath.
90+% Cath.
80-90% Prot.
40-60%
80-90% Cath.
Figure 3 Relationship between religious affiliation and ill health.
im
JOURNAL OF PUBLIC HEALTH MEDICINE
RELIGIOUS AFFILIATION
•JP
"
'~
ii I
>90% Protestant
75 - 90% Protestant
25 - 75% Mixed
75-90% Catholic
>90% Catholic
Figure 4 Religious affiliation in Northern Ireland by electoral ward, 1991.
Discussion
For many measures of deprivation that were examined,
Catholic areas appeared to be more disadvantaged than
equivalent Protestant areas. However, in Northern Ireland it is
possible to select an indicator which will support any
previously held prejudice. Some of the measures described
here require closer scrutiny. All of the Census variables are self
reported but are unlikely to be subject to differential bias,
though it has been suggested that the LTI variable may itself be
confounded by the very factors researchers try to relate it to,
such as deprivation, religion, etc. ' 4
The percentage of households that are overcrowded (i.e.
more than one person per room), which is one of the four
variables in the Townsend index, shows a strong relationship
with Catholic affiliation. Whether this represents real disadvantage or a propensity of Catholics to have larger and
Table 1 Regression models for premature mortality and
morbidity ratios
Variable
p value
Standardized mortality ratios (<75 years)
Income support
0.623
Changed address in last year
0.226
No formal qualifications
0.125
Adjusted r2 46.8%
0.000
0.000
0.003
Standardized long-term illness ratios {<75 years)
Income support
0.372
No formal qualifications
0.318
Density (persons/hectare)
0.180
Lone parents
0.154
Long-term employment
0.154
Adjusted r7 77.9%
0.000
0 003
0.000
0.000
0.000
younger families is debatable.15'16 The tendency for Catholics
to have larger families has been reported in the United States for
the late 1970s.17 However, by the late 1980s this had reversed,
with Catholic total period fertility rates about one-quarter of a
child less than the Protestant rates.18
The proportion of households without central heating is
highest in those areas that have a more equitable distribution of
Catholic and Protestant This is probably because these areas
are more rural and also have a lower proportion of public sector
housing, which generally has central heating.19
The large and continuing differences in employment status
between the two communities in Northern Ireland have been
well recorded and have resulted in a substantial strengthening
of the fair employment law.20 Whether the proportion of
households which changed address in the year before the
Census represents disadvantage is debatable, but it, along with
'the proportion of elderly living alone', is included in the
Jarman index.21
The earlier study by Stringer10 showed that those areas with
a high proportion of Protestants had higher levels of mortality,
even after differences in deprivation had been discounted.
Alternatively, Catholic areas still had higher levels of
permanent sickness and disability (derived from the 1981
Census), after controlling for differences in deprivation. In that
study, deprivation and denomination explained relatively little
of the variance in mortality (23.9 per cent) and morbidity (46.9
per cent) rates between electoral wards. The limitation on the
indicators of deprivation may account for the low explanatory
power of these models and resulted in residual confounding,
thereby allowing religion to enter as a significant predictor. The
limitations of the Northern Ireland 1981 Census data have
already been noted.
The explanatory power of the uptake of income support
variable in both models is worth further exploration as a
DEPRIVATION AND ILL HEALTH IN NORTHERN IRELAND
possible alternative to the more conventional indicators of need.
The relationship between social security benefits uptake and
religion probably reflects real differences between the two
communities. An analysis of the Family Expenditure Survey
1989-199022 showed significant differences between Catholics,
who had an average income of £122 a week, and Protestants,
whose average income was £136 a week. Although the main
difference between the two communities could be ascribed to
different socio-economic characteristics (such as employment
status), about 18 per cent of the income gap between Catholics
and Protestants was due to unequal pay for equal job qualities.
Nearly one-third of the income of non-retired Catholics came
from social security benefits compared with a fifth for
Protestants. However, it was concluded that the heart of the
inequity problem in Northern Ireland was that there was much
greater inequity within Catholic and Protestant parts of the
community than between them.
The results of the present study would echo these
conclusions. There is much more variability within Catholic
and Protestant areas than between them and efforts should be
directed towards reducing the inequalities in health for all
sections of the community.
Conclusions
(1) In terms of religion or religious affiliation Northern Ireland
is a very segregated society, both within electoral wards
and at higher geographical levels.
(2) There was a great amount of heterogeneity associated the
socio-economic indicators within each of the seven
categories of polarization of religious affiliation, such that
even amongst the most polarized categories there were
some very affluent and some very deprived areas.
(3) For many socio-economic indicators Catholic areas would
appear to be more deprived than their Protestant peers.
However, this relationship varies with the indicator chosen
and some indicators would suggest greater need in
Protestant areas.
(4) Heterogeneity within categories is also present in both of
the ill health indicators, but there is an overall positive
association between the percentage of Catholics within an
area and the morbidity and premature mortality experience
of that area.
(5) It was possible to develop models that explained almost 50
per cent of the variance in premature mortality ratios
between areas and more than three-quarters of the limiting
long-term illness ratios.
(6) Differences in denominational characteristics between
areas did not add any additional explanatory power to
these models. Thus it can be concluded that, at an
ecological level, religious affiliation does not have an
independent influence on morbidity or premature mortality
ratios, once differences in socio-economic standing have
been controlled for.
167
References
1 Central Statistics Office. Social Trends 30. London: Central
Statistics Office, 1995.
2 Townsend P, Philimore P, Beattie A. Health and deprivation:
inequalities and the north. London: Croom Helm, 1988.
3 Policy Research Institute. Spatial and social variations in
the distribution of health indicators in Northern Ireland.
Report to the Department of Health and Social Services,
Northern Ireland. Belfast: The Queen's University and the
University of Ulster, 1990.
4 Campbell R. Research into inequalities in health in Northern
Ireland: what research? Crit Publ Hlth 1993; 4(2): 5-8.
5 Rowthom B, Wayne N. Northern Ireland, the political
economy of conflict. Oxford: Polity Press, 1988.
6 Smith DJ, Chambers G. Inequality in Northern Ireland,
Oxford: Clarendon Press, 1991.
7 McGuffin SJ. Smoking: the knowledge and behaviour of
young children in Northern Ireland. Journal of Health
Education 1982; 48(2): 160-165.
8 Blaney R, MacKenzie G. A Northern Ireland community
health study. Report to the Department of Health and Social
Services, Northern Ireland. Belfast: The Queen's University,
1993.
9 Murray M, McMillan C. Social and behavioural predictors
of women's cancer screening practices in Northern Ireland.
J Publ Hlth Med 1993; 15(2): 147-153.
10 Stringer P. Health inequalities, religious affiliation and
urban-rural status. Report to the Department of Health and
Social Services, Northern Ireland. Belfast: The Queen's
University, 1992.
11 Morris C, Compton P. 1981 Census in Northern Ireland.
Population Trends 1985; 40: 16-20.
12 Campbell R, Stevenson G. Community differentials in
health in Northern Ireland. Report to the Northern Ireland
Department of Health and Social Services. Belfast: The
Queen's University, 1993.
13 Carstairs V, Lowe M. Small area analysis: creating a base
for environmental monitoring and epidemiological analysis.
Community Med 1986; 8: 15-28.
14 Bentham G, Eimermann J, Haynes R, Lovett A, Brainard J.
Limiting long term illness and its associations with
mortality and indications of social deprivation. J Epidemiol
Commun Hlth 1995; 49(Suppl 2): S57-S64.
15 Compton P. Demographic trends in Northern Ireland.
Report 57 for the Northern Ireland Economic Council,
Belfast 1986.
16 Coward J. Recent characteristics of Roman Catholic fertility
in Northern Ireland and Southern Ireland. Population
Studies 1980; 34: 31-44.
168
JOURNAL OF PUBLIC HEALTH MEDICINE
17 Mosher WD, Hendershot GE. Religion and fertility: a
replication. Demography 1984; 21(2): 185-191.
18 Mosher WD, Williams LB, Johnson DP. Religion and
fertility in the United States: new patterns. Demography
1992; 29(2): 199-241.
19 Northern Ireland Housing Executive. Northern Ireland
house condition survey 1991. Belfast: Northern Ireland
Housing Executive, 1993.
20 Cormack RJ, Osbourne RD. Disadvantage and discrimination in Northern Ireland. In: Osboume RD, Cormack RJ,
eds. Discrimination and public policy in Northern Ireland.
Oxford: Clarendon Press, 1991.
21 Jarman B. Identification of underprivileged areas. Br MedJ
1983; 312: 1705-1709.
22 Boorah VK, McKee PM, Heaton NE, Collins G. CatholicProtestant income differentials in Northern Ireland. Rev
Income Wealth 1995; 41(4): 41-56.
Accepted on 19 November 1997