Measuring Social Isolation

AEvery claim about how isolated a population has become depends on an instrument, and every instrument carries within it a judgement about which relationships deserve to be counted. The point is easy to overlook. A familiar index gives a respondent one mark for each of several circumstances, such as an empty household, a month passing without a visit from anyone related, or membership of no organised group, and then treats the total as though it were a quantity of the same kind as height or income. Instruments of that shape were devised in 1974 to answer a question which no longer dominates the field, namely whether older people were receiving enough practical help, and they have survived chiefly because long-running surveys cannot alter their wording without forfeiting the ability to compare one decade with the next. What began as a compromise about continuity is now widely treated as a definition. The index was never meant to bear the weight now placed upon it, and the statisticians who assembled it said as much at the time, describing their scale as a rough administrative convenience rather than a theory of social life. A conversation with a neighbour and a marriage of forty years, on this arithmetic, contribute equally. Much of what passes for evidence about rising isolation is therefore evidence about the habits of a questionnaire.

BMarek Osterlund, who directs the Ardsley Centre for Population Studies, argues that the arithmetic is measuring the wrong property altogether. What matters, on his account, is not how many people a respondent can name but how those people are arranged, and in particular whether help refused by one of them is likely to be offered by another. Working with the Ravensworth cohort, in which 8,200 adults mapped their own circles between 2014 and 2019, he found that respondents whose named contacts had no links to one another were 18 per cent more likely to report unmet need during a spell of illness, while the sheer number of contacts predicted almost nothing once the pattern of connections had been allowed for. A large circle of unconnected acquaintances, on this reasoning, can leave a person as exposed during a crisis as a much smaller circle whose members are in regular touch with one another. The relationship held only among respondents living outside the three largest cities, where formal services fill the gap. Osterlund maintains that the standard index therefore records sociability, which is an attribute of temperament, when what policy requires is a measure of resilience.

CIngrid Sowande, a sociologist at Marchmont University, is sceptical of maps drawn from memory. She contends that such diagrams record what respondents believe about their acquaintances rather than what is true of them, and that the belief is systematically flattering, since people assume their friends know each other more often than they actually do. In her validation exercise, published in 2021, the circles described by members of 240 households were compared with logs of who had in fact been in touch over four weeks, and the two accounts agreed about 61 per cent of the links. Memory, she points out, tidies a social world that is in fact ragged, so the neat diagrams collected by researchers may describe an ideal rather than an actual pattern of contact. She also insists that a dense circle is not automatically a supportive one, because a group in which everyone knows everyone may supervise as readily as it assists. Sowande concedes that a comparison confined to one town, and to a single month, cannot settle which instrument better predicts illness, and she does not dispute the mortality findings themselves.

DThe reason the dispute matters is that the mortality evidence is what persuades governments to spend. Among the 2,400 adults in the Ravensworth sample who were followed to the end of the period, those in the most isolated fifth were 24 per cent more likely to die than those in the least isolated, a figure quoted in almost every campaign document since. Adjustment changes it considerably. Once differences in mobility, hearing and household income at the outset were taken into account, the excess fell to 8 per cent, which suggests that a good deal of the original gap reflected conditions already present when isolation was first recorded. Something nonetheless remains, and the residue is larger among people who had been isolated for at least eight years, a graded pattern of the sort that usually indicates genuine harm. Contact also tends to decline in the eighteen months before a serious diagnosis, so the direction of the arrow is not always easy to establish.

EMeasurement quietly decides what gets funded. Schemes are assessed against the instruments that happen to exist, so a programme which raises the frequency of visits is judged a success even where nobody's circumstances have altered, while a programme which repairs a single broken relationship is credited with nothing at all. The sensible conclusion is not that one camp has won. Counting contacts remains adequate where the question is whether a service is being taken up, whereas the mapping of connections is the better guide where the question is who will be left without help in a crisis, and the two should not be asked to answer each other's questions. There is, however, one blind spot that neither escapes. No instrument yet in use separates solitude that has been chosen from isolation that has been imposed, and until respondents are asked whether the arrangement suits them, both kinds of measure will keep producing numbers on which no policy can safely be built.