Using optimal matching to analyze the timing of daily life
Abstract
In this paper, I reflect on the conditions required to apply optimal matching (OM) to
time-use data and I propose a parameterization adapted to the analysis of the timing
of daily life. OM allows time use analysts to build typologies of sequences of daily life,
hence to take into account simultaneously the duration and the timing of activities.
OM is a family of distance concepts originating in information and coding theory
where it is known under various names among which Hamming or Levenshtein
distance. Although it was imported into social sciences from biology, its success in
this field is not due to any resemblance sequence transformation operations may
share with gene mutations but is on the contrary the result of parameters set in
accordance with biological concepts and materials. Consequently, applying OM to
time use data requires a cost system able to uncover sub-rhythms from sequences of
daily events. I propose to use only substitution operations and to derive their costs
from the series of transition matrices describing daily collective rhythms. A Stata
plug-in freely available on the Internet implements this methods and is briefly
introduced. This variant of OM is applied to the question of the scheduling of work
using the 1985-86 and 1998-99 French time-use surveys: twelve types of workdays
are uncovered. Aggregate and individual chronograms are used to interpret and
assess the quality of the typology.