METHODS

Three data sources were used in this study. Individual level data were included on selected variables collected as part of the national health and lifestyle survey SLAN in 1998.⁴ The methodology for that study, which has been described previously, involved a multistage cluster sample technique with random selection of individuals from the electoral register in 273 urban and rural district electoral divisions across the country so that information on a representative sample of people from each county equivalent area was available. Two questions relating to self reported quality of life and satisfaction with health, measured using a five point Likert scale, were taken from the quality of life instrument developed in 1993 by the World Health Organisation (WHOQoL). Self reported general health on a five point scale from excellent to poor was also recorded. Lifestyle practices were included in relation to current daily cigarette smoking, fruit and vegetable consumption of at least four items per day (the national target), and alcohol consumption of over 14 units per week in women and 21 units per week in men. Eligibility or not for general medical services, a means related system of entitlement to comprehensive health services, was recorded also. Level of deprivation was assessed in this study by means of three variables, private home ownership or otherwise, education from primary through to tertiary level and employment status, categorised in accordance with standard census questions.

Crude mortality rates are not available routinely from the Central Statistics Office at a level below that of county. In order therefore to calculate standardised mortality ratios (SMRs) a county equivalent level was decided upon. At this level it is possible to separate out Cork city and county into two and the greater Dublin area into four geographical distribution divisions. SMRs for these areas were then calculated using a standard indirect method based on mortality data of all adults over 18 years from the 1997 vital statistics report and denominator population data at county level from the 1996 census.

Electoral data from the 1997 general election data ²¹ were used to establish percentage of first preference votes in each constituency. The voting system in Ireland, of proportional representation by the single transferable vote in multiple seat constituencies, has been said to promote a clientist, non-ideological political system. The two largest parties in the Republic of Ireland, Fianna Fail and Fine Gael, have their origins in the Civil War that followed on the foundation of the State in 1922; Fianna Fail was founded in 1926 and Fine Gael in 1933. Neither party traditionally is notably ideological in economic terms, though both are regarded by analysts generally as being to the right of centre.²⁰ The Labour party was established in 1912 and is a traditional socialist party, though seen to be to the ideological right of counterparts in Europe, including the British Labour party. There were two smaller ideological parties, not represented in all constituencies in general elections. The Democratic Left party arose from a strongly left wing Marxist Leninist background and is also the culmination of a series of breakaway splinters from a much earlier republican tradition. The Progressive Democrats, established in 1985, has a liberal market economy ethos and a relatively more liberal social agenda on issues like abortion and divorce.

There are 41 political constituencies in the Republic of Ireland, constituted according to a number of factors but principally population density, each returning between three and five deputies. Constituency borders may not necessarily relate to county borders or to public administration units. There are for instance 11 constituencies in Dublin city and county, five constituencies in the largest county, Cork, where the second largest city is located. Several of the larger counties have at least two constituencies, while some of the lower population density counties are paired into one constituency. The only way to achieve comparability between regional voting patterns and mortality rates was to treat each of the entire Dublin and Cork areas as one county and to retain as amalgamated units each constituency comprising two counties, giving 22 data points for this part of the analysis. The Progressive Democrats were represented in 15 of these areas only. Votes for the Labour party and Democratic Left (who coalesced into one party shortly after the general election) were taken together as one unit to indicate left pattern voting and were represented in 21 areas. There was a range of independent candidates and a number of other parties including the Greens, not a strong force at general elections; these were categorised together as others for this analysis. Proportion of abstainers for each constituency was calculated as the difference between eligible voters on the register and percentage turnout.

Simple univariate analyses were undertaken. Pearson's correlation coefficient was used to assess associations between variables, both at individual level in the SLAN dataset as a whole, where age adjusted partial correlation coefficients are reported and at aggregated level. Two area based matrices were assembled, first assessing SMR in relation only to the SLAN variables and therefore utilising the 26 area points with breakdown possible of the two largest population areas, Dublin and Cork and then including in addition the voting patterns, with Dublin and Cork treated as a single unit, and just 22 data points. There were insufficient area based data points for multivariate statistical assumptions to be satisfied in relation to the aggregated data.

RESULTS

The final response rate for SLAN was 62.2% and the total number of respondents was 6539 (47.0% male). The overall correlation matrix confirmed a large number of highly significant relations, albeit some with rather low levels of correlation. When these were age adjusted the pattern remained the same except that the relation between employment status and education, previously direct, became inverse, suggesting age as a confounder. Table 1 illustrates the partial age adjusted correlation coefficients for all variables. Self rated general health, quality of life, satisfaction with health were all intercorrelated, and each related significantly to employment status, education, tenure, medical card possession, smoking status and fruit and vegetable consumption. Smoking rates were also significantly related to tenure and to medical card possession. This confirmed the appropriateness of examining patterns of inter-relation of these variables at an area level. The standardised mortality ratios in descending order for each of the 26 county borough areas are presented in table 2, together with the total percentages for each of the health, lifestyle and deprivation variables. This table shows some variation in SMR by county, ranging from a high of 123.67 in Cork county borough to a low of 83.93 in Fingal, an area of Dublin county. Likewise there was some variation in the other measures. Notably there are relatively high rates of home ownership throughout most of the country.

The total population at the 1996 census was 3 526 719. Of 2 714 262 adults over 18 registered to vote, 1 806 932 votes were cast at the 1997 general election, representing a turnout of 65.92%. Overall Fianna Fail obtained 39.33% of first preference votes, Fine Gael 27.95%, Labour and Democratic Left combined 12.91%, the Progressive Democrats 4.68% and others 15.13%. The voting pattern in 22 county level areas, including turnout, is presented in table 3. The correlations between voting patterns, standardised mortality ratios, eligibility for general medical services and the selected variables from SLAN are presented in table 4. The higher the vote for Fianna Fail (r=−0.534, p=0.011) or Fine Gael (r = −0.630, p= 0.002) the lower the vote for left wing parties. There was no significant relation between mortality and voting pattern for either of the main parties, Fianna Fail and Fine Gael. The higher the Fianna Fail vote the higher was the percentage of homeowners (p = 0.429, p = 0.046). There was a significant positive correlation (r=0.446, p=0.037) between combined left voting pattern and SMR. The higher the Fianna Fail vote the lower the rate of those dissatisfied with their health (r = −0.578, p = 0.005). Conversely, Left wing voting patterns correlated significantly with dissatisfaction with health (r = 0.467, p<0.028) and rates of regular smoking (r = 0.470, p =0.027). There were higher rates of regular smokers in constituencies with high rates of voting abstainers (r = 0.526, p<0.012). There was a significant negative relation between Left wing voting pattern and overall eligibility for general medical services (r = −0.505, p=0.017), which did not remain significant when only medical card holders under 55 years old were considered.

There was no relation between SMR and GMS eligibility. There were a number of other inter-relations at area level. For instance the higher the percentage rating their health fair or poor, the higher the proportion reporting a poor quality of life (r= 0.487, p=0.007) none or primary education (r=0.648, p<0.001) and general medical services eligibility (r= 0.556, p=0.007). The percentage unemployed correlated significantly with level of education (r=0.454, p=0.034) and medical card ownership (r=0.428, p =0.047). There was an inverse relation between percentage unemployed and both fruit and vegetable (r=−0.672, p=0.001) and excess alcohol consumption among males (r= −595, p=0.003). As it was not possible to relate SMRs to voting patterns for constituencies in the two larger cities we compared the correlation matrix for 26 areas with that for 22, to see whether the pattern of relationp between health status, deprivation or lifestyle indicators would be much altered generally. It was in fact almost identical, with the exception that percentage unemployed now related significantly to SMR (r=0.408, p=0.038) as did percentage with primary school education only (r=0.454, p=0.034), though the direction of the relation remained the same.

DISCUSSION

In this study we combine for the first time, aggregated individual measures of self rated health and lifestyle, socioeconomic circumstances, personal and health service access with ecological level data on voting patterns and mortality.

Relation between patterns of deprivation and health status in Ireland

This study confirms at individual level in Ireland both the clear inter-relation between deprivation indicators such as level of education, housing tenure and rates of unemployment and in turn their correlation with lifestyle risk factors. Though little of the variance is explained, our purpose in this paper was to establish only that such relations existed in order to study the area patterns so no further detail is given at this stage. It is unknown how social class variations have contributed over the years to patterns of ill health in Ireland because of the lack of robust epidemiological data. What we do know from more recent work is that health inequalities do exist in Ireland⁵ and that the country is now changing fast in socioeconomic terms.²² A clear social gradient in relation to health and lifestyle variables at individual level in this dataset for both men and women has already been reported.⁴

Despite the poor health profile of Irish people compared with their European counterparts, variation in mortality at county level is not as wide as is seen in other countries, like its close neighbour the United Kingdom and the relation with standard deprivation indicators is weak or inconsistent. There was some variation in the various indicators of self reported health status and lifestyle at area level. Certainly SMRs were higher in urbanised areas in particular but the relation with classic deprivation indicators like education and employment status was lost when the two large cities were absorbed to county level. This confirms in fact findings of the two previous studies of almost a decade ago. Johnson et al found a relation between deprivation and area SMRs within Dublin city.² However, Howell et al found the pattern at county level was much weaker across the country as a whole and that the deprivation index used, a modification of that of Townsend did not relate to SMR.³ One explanation is the inappropriateness of the measure in a country with high levels of home ownership of varying standards and an education system reformed beyond recognition in the past 30 years.²² In fact the inverse relation between education and unemployment in the dataset as a whole is explained by the fact that older people are more likely to have primary school education only and younger people to be unemployed. The geographical scale of the county and the relatively low population density may also be a problem. Yet there are coherent sociological reasons for considering area variations at this level, in that local government operates at a county level, regional health boards are comprised of counties and there is a strong county affinity exemplified by sporting organisations like the Gaelic Athletic Association.

Voting pattern and health

There was some evidence of a relation between left wing voting and SMR and of indicators of adverse lifestyle and deprivation, yet the voting pattern was relatively inconsistent and nothing like that of the United Kingdom. The important finding however is the fact that there is no relation with right wing or centrist voting. This is explicable mainly because of the lack of class polarisation in party political affiliation, which is clientist and often personality based. To an extent this reflects the proportional representation system but the proportion of the total population supporting the two main centrist parties is enormous, so it is not a matter of a failure of power to detect a significant relation. What do these findings signify about the association between voting pattern and health status? The original study by Davey Smith and Dorling ¹⁶ was an assertion that voting patterns could be interpreted as a proxy for material circumstances in the different constituencies in the United Kingdom, and that the attitudes reflected in voting were a result of complacency about the provisions of the Welfare State among voters in the more affluent constituencies. In a follow up study on the 1997 election they found similar strong positive relations between SMRs and Labour party voting and the converse for the Conservative party.²³ As quoted by Sinnott,²⁰ voters in Ireland have consistently ranked themselves in the centre or to the right of centre in their political attitudes. The spectrum of difference between the Irish Labour party and Fianna Fail and Fine Gael, both remarkably similar to each other and closer to British Conservatives, was much narrower than the expressed differences in the United Kingdom between Labour and the Conservatives. This may well help to explain the less polarised relation between voting pattern and health status in Ireland compared with Britain.

Social capital, deprivation, and voting pattern

However, there may be a much wider debate on what voting patterns signify, to do with concepts of social capital and cohesion, as patterns in other countries are also seen but not in the same straightforward way. A number of studies in the United States assert relations between measures of social inequality, level of voting, and ill health.^15,24 Kondrichin and Lester ¹⁷ analysed the Reagan/Carter presidential election pattern in the United States, finding a negative correlation between mortality rate and the Republican vote. The same authors ¹⁸ also analysed patterns in Russia at the 1993 election, finding an inverse relation between crude death rates and the percentage vote for the new constitution but not with the percentage voting for the Communist party. The combination of perceived dissatisfaction with health among those who are left voting and the converse for the majority party that we found suggests at least some psychosocial as well as material aspect to disadvantage. A majority of respondents were satisfied with their health in the SLAN survey and in this analysis that translated into satisfaction with their health on the part of Fianna Fail voters, the majority party. However the inter-relation between economic inequality, the sociopolitical systems that bring that about or fail to address it and the impact on individual wellbeing is not straightforward. In the United States where economic and political structures are comparable in the country as a whole, even if income distribution varies greatly,⁹ cross state comparisons may well reflect relative discontent with material disadvantage but how true is that of less politically polarised and arguably more socially cohesive societies?

It is beyond the scope of this paper to address what explains these patterns but there is some relevant evidence. The social history of Ireland in the time since it became an independent republic is one of considerable general economic hardship and migration right up to the 1980s, and yet with strong indicators of so called social cohesion including tight knit family structures, high rates of church going and a socially conservative mainly rural culture. Analysis, for instance, of religious practices show the Irish as among the most consistent church going Christians anywhere. Nor has this waned as much as might be expected. Indeed, even though there has been strong dissent in recent years from the position of the Roman Catholic church on matters of private sexual practice and challenges to its authority generally, Christian values remain strong, even among the young.¹⁹

Furthermore, Ireland has long been known as a European oddity in the low level of political left wing support it enjoys.²⁰ By the 1980s a new class structure more typical of other European countries was in place. Party political allegiances were fairly constant in the past 30 years, Fianna Fail attracting a majority support of all social classes, Fine Gael being relatively more middle class and the Labour party having more working class support but also, like the others, receiving support across all classes. Although the emergence of a more ideological political system may be associated with signs of urban deprivation and a widening of the gap between rich and poor, traditional social and political values continue to be relatively important in the country as a whole. Yet this level of social consensus has apparently not translated into a better health profile, either in the past or in terms of contemporary health, based on these data.

Methodological issues

Though SMR data could not be related to voting pattern at constituency level in the cities we were able to compare the SLAN indicators at that level and hence indirectly assess whether deprivation would polarise patterns more. This suggests the effect of the relation between voting pattern and SMR may have been weakened for those inner city areas where absolute deprivation is greatest, but not greatly. Even there the rate of support for Fianna Fail is relatively strong. The indicator of both potential morbidity and possible disadvantage, eligibility for comprehensive general medical services related inversely with left wing voting, and not at all to SMR, which on the face of it is surprising as the scheme is means tested. However eligibility also relates strongly to age, which is likely to be the explanation here, given the fact that the pattern was much weaker for those less than 55 years of age and it is likely also that shifts in party political allegiance are due to more young people deserting the traditional parties. We have to be cautious in interpreting these lifestyle data because the sample for SLAN is necessarily smaller than for either voting data or SMRs and hence some of the variations seen may be due to lack of statistical precision rather than real between county variations. Furthermore, we do not have individual level data on voters' health profile. There were trends that could be explained by demographic patterns of party support including gender and age as well as class but these did not reach statistical significance. The study necessitated multiple variables in relatively few area points, so that there was arguably insufficient power to assess whether quite strong correlations were significant and also the possibility of statistically significant but incoherent correlations. We tackled this in two ways. Though this paper primarily relates to area patterns we included the individual level correlations to be sure that there were genuine inter-relations between deprivation indicators and lifestyle. The pattern was coherent and consistent with the findings in other countries in the overall data that have more power and precision. Furthermore the aggregated findings accord broadly with those of the one previous comparable study.³ The methodological problems in this study are common to many small countries and while they need to be acknowledged they are in themselves a good explanation as to why this debate is dominated by large scale data from highly populated industrialised countries.

CONCLUSIONS

This study has importance for the international debate on social capital and deprivation. The simple fact is that social breakdown coexists with poverty, deprivation, adverse lifestyle and poor health outcome in most industrialised countries, making an assessment of the relative contribution and causal directions of these inter-relations highly problematic. Physical measures of social capital cited to date might not necessarily be reflected in actual economic disparities and more importantly might not in themselves reflect the true underlying attitudes people have towards each other. This is why Ireland is a good example. In a small country with a complex proportional representation system mainstream party voting tells us what social and economic policies most voters are happy to support. Furthermore, social variations in the country as a whole are as marked at individual as area level across most of the country. On the strength of these data at least, the case for a macro-level shift in public policy rests securely on its direct impact on material disadvantage and empowerment of the poor, rather than a promotion of a politically cohesive consensus approach in itself. The findings suggest also that a more subtle conception of social capital is needed to explain or tackle health variations.