Understanding friendships, your own included

Reporting few friends or saying one is lonely has a social cost. Methods that bypass that cost give noticeably different results.

It is not an impression, it is a theorem. Proved in 1991, it follows from how a network is sampled when one follows its ties.

European surveys allow reported contact frequency to be compared. What those comparisons measure is less obvious than it seems.

Systematic reviews exist and their conclusions are cautious. The main difficulty is not the effectiveness of the schemes but the quality of their evaluation.

Call records and proximity sensors have produced results questionnaires could not reach. They also import biases of their own.

A 2010 meta-analysis produced a striking comparison with smoking. Its method and limits are less well known than its conclusion.

The phrase has become a commonplace of public policy. The long series available do not show the general rise it presupposes.

One is felt, the other counted. Their correlation is moderate, which is enough to invalidate most of the shortcuts used in public debate.

The reference instrument, published in 1978 and revised since, deliberately avoids the term. That methodological choice explains part of the figures in circulation.

The question asked determines the figure obtained. Two similar wordings produce networks of very different sizes, which has already derailed an entire public debate.