Regional Inequalities in Infant Mortality in Britain , 1861-1971 :sx Pattern and Hypotheses .sx C.H. LEE .sx Variations in health inequality , and the debate about whether they have increased or decreased during recent decades , have generated considerable investigative activity and querulous discussion .sx A considerable part of this literature has been focussed upon health inequalities between social classes , but debate has foundered on difficulties in defining and measuring acceptable categories and ensuring consistency over time .sx Illsley has argued that differences between death rates of different classes are primarily a reflection of the classificatory scheme itself and bear little resemblance to actual changes in health or death inequality .sx Others have averred that such reservations about the use of class data based on occupational groups are exaggerated and , consequently , have felt able to reiterate their conclusions derived from this source .sx It seems unlikely that this avenue of investigation can be developed so as to reconcile these various positions .sx The data on occupations and the class categorisations derived from them do not admit further refinement , certainly not sufficient to persuade those who , like the present writer , regard this material as inherently unsatisfactory for meaningful statistical analysis .sx In this paper the problem of health inequality is approached from a different perspective , that of regional variations .sx This has two advantages .sx First , the data are available for a long period of time , and thus allow analysis from the 1860s to the present , for the constituent counties of Great Britain which remained substantially the same throughout the period .sx The principal change in regional categorisation occurred during the 1970s , a fact which led to the decision to conclude this investigation at the beginning of that decade .sx By then , the principal features of the most recent phase of historical development were clearly established .sx Even the fact that such county-based regionalisation does not allow division between rural and urban areas can be tolerated in view of the compensating advantage this classification allows in long-term internal consistency of the data .sx Failure to achieve such consistency has been , of course , one of the principal criticisms levelled against analysis based on class determined by occupational groupings .sx The second justification for adopting a regional perspective lies in the fact that this aspect of health inequality has been little explored , although it has occasioned several general , and usually inaccurate , interpretative observations .sx The measure of health inequality adopted here is the infant mortality rate .sx Infant mortality has been widely accepted as an important and significant indicator of health achievement , because infancy has always been one of the most vulnerable periods of human life , and because the scale of infant mortality has important consequences .sx Furthermore , infant mortality has been shown to be associated , in various studies , with a number of important economic and social indicators , such as income per head , equality of income distribution , and material deprivation .sx With regard to the latter , it was suggested in the Black Report " that any factors which increase the parental capacity to provide adequate care for an infant will , when present , increase the chance of survival , while their absence will increase the risk of premature death .sx The most obvious such factors fall within the sphere of material resources :sx sufficient household income , a safe , uncrowded and unpolluted home , warmth and hygiene , a means of rapid communication with the outside world , for example a telephone or car , and an adequate level of manpower - or womanpower ( two parents would normally provide more continuous care and protection than one ) " .sx Variations between infant mortality rates in different regions and in the long term provide , therefore , a significant indicator of variations in basic economic and social well-being .sx The long-run pattern of infant mortality in the advanced industrial nations is generally familiar , the fluctuating levels which obtained through the second half of the nineteenth century gave way to widespread and rapid decline through the twentieth century .sx In the case of Britain , much attention has been paid by historians to the sharp drop in the infant mortality rate at the beginning of the present century and considerable effort has been expended in attempts to explain this phenomenon .sx One of the most popular theses of recent years has been Beaver's argument that improvements in the milk supply reduced gastro-enteritis after the turn of the century , a development which was happily co-incident with the emergence of welfare services which emphasised safe feeding .sx Purification of the water supply , sewage disposal , safer milk , and greater food hygiene all contributed to reduce the risk of infection from food and drink .sx Beaver's important contribution reinforced attention on the downturn in the national infant mortality rate at the turn of the century as the key to the long-term pattern of change .sx In Table 1 infant mortality rates are shown at ten-year intervals from 1861 to 1971 for each of the 55 regions of Great Britain .sx The composition of these regions which comprise English counties , and aggregations of Welsh and Scottish counties , is indicated in Table 2 .sx These regional infant mortality rates vary considerably more than might be expected from the national aggregates for England and Wales and for Scotland , which fell rather dramatically after 1901 .sx Furthermore , these regional rates do not support the popular notion that trends in infant mortality change were " remarkably consistent " throughout the country .sx Nor do these estimates sustain the view advanced in the Black Report that " In the middle of the nineteenth century , the south-east of England recorded comparatively high rates of death , while other regions like Wales and the far north had a rather healthier profile " .sx While London just fell into the group of 17 regions with an infant mortality rate of 150 or above in 1861 , the majority of such disadvantaged areas were the industrialising counties of the English midlands and north .sx Lancashire , the East Riding , the West Riding , Staffordshire and Leicestershire were among regions with the highest rates .sx At the opposite extreme , in 12 regions infant mortality was 120 or less .sx Half of these were located in the south and west of England , while the rest were in Scotland where in Strathclyde North , Dumfries and Galloway , Grampian and Highland very low mortality rates between 86 and 95 were recorded .sx More significant , and contrary to conventional wisdom , is the pattern of change in these regional infant mortality rates .sx They did not fall uniformly , nor was there a clearly marked hiatus around the turn of the century .sx In fact , several groups followed different and divergent patterns of change during the second half of the nineteenth century .sx In some regions the highest mortality rate was recorded in 1861 , and there was a continuous improvement thereafter .sx Buckinghamshire , Bedfordshire , Cambridgeshire , Wiltshire and Dorset comprised this favoured group .sx In other regions mortality peaked in 1871 followed by continuous improvement thereafter , namely Hertfordshire , Northamptonshire , Rutland and Strathclyde South .sx In many other regions the peak infant mortality rate occurred in either 1861 or 1871 and was followed by a fall , and a modest increase in 1891 or 1901 as a temporary setback .sx In no fewer than 40 of the 55 regions was the highest infant mortality rate registered in 1861 and 1871 , a group which together accounted for 80.0 per cent of the national population at the latter date .sx Furthermore , many of the regions with an early peak in infant mortality were those with the lowest rates .sx This was especially pertinent in the south and west of England where only in Hampshire , Essex and Devonshire was a peak reached after 1871 .sx Even in some of the industrial regions with the highest infant mortality rates , like Lancashire , the West Riding and Durham the highest rates were recorded in 1871 .sx tables&caption .sx In those regions which did not fall into the large group in which there was a peak in 1861 or 1871 the highest rates were found in either 1891 or 1901 .sx After the turn of the century , except for a small number of occasional and temporary increases , all regional infant mortality rates fell continuously .sx There is no doubt that after 1901 the downturn was both universal and substantial .sx The most obvious group of regions in which infant mortality rates increased , rather than decreased during the later decades of the nineteenth century was found in Scotland .sx In seven of the ten Scottish regions a peak infant mortality rate was found in 1891 or 1901 , in some of them rates rose during the later decades of the century , substantially so in the case of Grampian .sx Elsewhere , in Monmouth , South Wales and Essex mortality rates also rose during the closing decades of the century , while in Northumberland , Cheshire , the North Riding and Worcestershire rates were high , but stable .sx The pattern of change at national aggregate level , confirmed by the unweighted average infant mortality rate for the aggregated regions shown in Table 3 , covers a diversity of regional patterns .sx The aggregate rate suggests fluctuation around a slowly falling mean value until 1901 , and the improvement around 1881 and the setback in 1891 was experienced in most regions .sx At regional level , there were two phases of downturn some 30 years apart , with much of Scotland , South Wales and some English counties , like Northumberland , the North Riding and Essex , lagging behind the rest .sx With these exceptions , and admitting the temporary setback suffered during the 1890s when , it has been suggested , long , hot summers produced conditions conducive to a resurgence of diarrhoeal complaints , much of Britain experienced a downturn in infant mortality rates from the 1860s and 1870s .sx From the turn of the century , decline was universal and persistent .sx The long-term pattern of change does not , of course , indicate increase or decrease in the variation of infant mortality rates between different regions .sx Estimates of four different measures of inequality are shown in Table 3 .sx Each has its own peculiar strengths and limitations .sx The coefficient of variation is sensitive to change throughout a given distribution , while the variance of logarithms is sensitive to changes at the lower ranges of a scale , and the Gini coefficient is particularly responsive to transfers affecting the middle values of a distribution .sx Theil's entropy index is responsive to changes throughout the distribution and measures deviations from a state of equality in which each variable in the distribution has a share equivalent to its relative size .sx Since it measures inequality exclusively in terms of the 'distance' between variables , it satisfies , unlike the other measures quoted , the " strong principle of transfers " .sx But , most importantly , all four measures are invariant if all values in a set are raised or lowered in the same proportion , so that each measure is an appropriate indicator of inequality in infant mortality rates .sx As can be seen from Table 3 , with minor exceptions , all the four measures of inequality show the same pattern of change over the century examined .sx table&caption .sx Three of the four measures show increasing inequality between 1861 and 1871 , as the downturn in infant mortality rates in healthier regions was not replicated elsewhere .sx All measures converge during the following decade , 1871-1881 , as the decline in the mortality rate became widespread .sx Thereafter , with the single exception of the Gini coefficient for 1911 , all the indicators show an increase in inequality until 1921 or 1931 .sx Even the downturn in the mean infant mortality rate after 1891 , and in all the regional rates after 1901 , did not prevent this increase .sx The reason for this increase in inequality is quite clear ; the decline in infant mortality rates was fastest in those regions in which the rates had been lowest at mid-century .sx Thus , the rate of decline between 1881-1921 was only 0.50 per cent per year in Lanarkshire , and 0.80 per cent in Durham , compared to 1.48 per cent in Surrey and 1.25 per cent in Wiltshire .sx From the peak of 1921 and 1931 , when measured inequality was almost the same , there was a continuous convergence of regional infant mortality rates as inequality decreased .sx This trend was reflected in each of the four measures .sx By 1961 , two measures indicated a level of inequality similar to that in 1861 , while Gini's index suggested that it was greater and Theil's that it was smaller .sx By 1971 , all four indicators showed the lowest level of inequality recorded during the entire period .sx