Content area
Full Text
We investigate the nature of measurement error in time-use data. Analysis of "stylized" recall questionnaire estimates and diary-based estimates of housework time from the same respondents of a British survey gives evidence of systematic biases in the stylized estimates and large random errors in both types of estimates. We examine the effect of these measurement problems on three common types of statistical analyses in which the time-use variable is used as: (1) a dependent variable, (2) an explanatory variable, and (3) a basis for cross-tabulations. We develop methods to correct the biases induced by these measurement errors.
1. INTRODUCTION
Time-use data have played a central role in social science research in the past few decades. They have been used widely in statistical models of market and nonmarket activities, research into the household division of labor, studies on lifestyle patterns and well-being across different social groups and different countries, and so on. Last but not least, they act as revealing indicators of societal trends. Researchers have used them to understand the changing ways in which different groups of individuals structure their daily life (for example, see Bianchi, Robinson, and Milkie [2004]; Gershuny [2000]; Robinson and Godbey [1997]).
1.1. Stylized and Diary Time- Use Estimates
There are two main types of information on time use. The first, generally referred to as the "stylized" estimate, is based on responses to questions asking about the time devoted to various classes of activity in an "average" or "normal" or "typical" week1 (Juster and Stafford 1985). Interpreted literally, it requires respondents to perform two difficult tasks: (1) to recall their activities in the recent past and (2) to carry out an appropriate form of averaging. Doubts about the accuracy of stylized data rest on two concerns: the difficulty of the respondent's task, which might lead to substantial measurement error; and the lack of detail, which allows the respondent scope for choosing responses that correspond to a preexisting (and possibly inaccurate) self-image or a preferred self-representation.
The second type of data source, usually preferred by researchers, builds up a summary measure from entries in a time-use diary, in which the respondent is required to keep a very detailed record of activities throughout the day, on each of a set of (usually...