The shared meal, a variable that sums up many others

Many surveys, mainly North American, report associations between the reported frequency of shared meals and school, dietary and behavioural indicators in adolescents. Frequency is most often measured by a single question, along the lines of the number of meals eaten together over the past week, put to the adolescent or the parent.

Those associations are consistent and of moderate size, which explains their spread into public health recommendations and the general press. They also concern a simple behaviour that anyone can picture, which makes them easy to repeat.

The frequency of shared meals depends on working hours, the number of adults present, commuting time and the stability of family organisation. All those variables are associated with the same outcomes.

In other words, the shared meal is a good indicator of a set of conditions, which does not establish that it is a lever on them.

The relationship may also run the other way. An adolescent going through a difficult period may withdraw from shared meals, so that the scarcity of meals would be a consequence of their difficulties rather than their origin.

Studies adjusting finely for family characteristics, or following the same adolescents over time, find markedly reduced associations, sometimes not significant.

An analysis published in 2012 by Daniel Miller, Jane Waldfogel and Wen-Jui Han, using longitudinal data, concludes that associations are far weaker than those reported by cross-sectional studies. This type of design compares an adolescent with themselves at different periods, which neutralises the stable characteristics of the family without neutralising those that change.

The mechanisms proposed are regularity of exchange, adults' knowledge of what the adolescent is doing, and the existence of a predictable moment. Those elements can exist without a shared meal and be missing despite one. A family may, for example, share few meals because of staggered working hours while maintaining other regular moments of exchange.

The few studies that measure the quality of exchange during the meal, rather than its frequency alone, find that this dimension carries most of the measured association. Those studies remain few, and the quality of exchange is itself measured in them through self-reports.

It illustrates a frequent configuration: a practice that is easy to count serves as an indirect measure for something hard to measure, and then the indirect measure is taken for the cause. The same pattern has been discussed for other family variables, such as the number of books at home.

The practical consequence is that a policy targeting meal frequency without touching the conditions that determine it would have no reason to produce the effect observed in surveys.