Introduction (continued)

In this document:



The nature of time use data

The time use data are collected by means of time diaries. Respondents record their activities in time diaries, using their own words. The diary covers 24 hours. With some exceptions each respondent fill in diaries for two diary days. In HETUS the diary instrument records four recording domains:

1. Main activity: “What did you do?”,
2. Parallel or secondary activity : “Did you do anything else? If so, what?”,
3. Who with: “Were you alone or together with somebody you know, if so, who?”,
4. Location (incl. mode of transport)

As a result the data consists of a sequence of episodes or events, each characterised by these four recording domains. In addition, there are individual and temporal identifiers. The individual identifier (“diary/person id”) connects the episode to a particular respondent and a particular diary day. The individual identifier also connects to background information on the respondent’s household and individual circumstances. The temporal identifiers indicate starting and ending time, and, hence, also duration of the episodes. In order to comprehend what the data has to offer in terms of analysis, which is a lot, it is rewarding to keep the structure of the data, i.e. the sequence of episodes, in mind.

The background information is collected by means of interviews. The purpose of this information is to form the population groups for which the time use estimates are to be calculated.


Guide to interpretation of the statistical measures

The most essential statistical measures in time use statistics are mean time spent on various activities and the proportion of the individuals that spent some time doing the activities during the day they kept the time diary (“participation rate”). All measures may be calculated for a great variety of population groups depending on which background information has been collected.

Mean time for an activity is calculated as follows: Assume the activity paid work and the population group women. Now, for all diaries filled in by women, the duration of all episodes of paid work is summed up. Some women will then contribute with many hours of paid work, and other women that did not work for pay at all during the diary day, will not contribute to the sum. The sum, i.e. all women’s total hours of paid work is divided by the total number of women (or rather the number of women’s diaries), regardless of whether or not they carried out any paid work. The mean contains no information on the distribution of number of hours of paid work within the population group. Entirely different distributions may result in the same mean. Suppose the mean is 4 hours a day. If all women work 4 hours a day, the mean will be 4 hours a day. The result will be the same if half of the women work 8 hours and the other half do not work at all. Consequently, the mean conceals differences.

In this case the mean is a characteristic of the population group; the population group women spends all together a certain number of hours of paid work. If the number of hours is evenly distributed within the group, all women would have worked 4 hours. But, as mentioned, the distribution may also have a completely different shape.

Now, if the means in two population groups differ, it implies that one of the population groups, i.e. the individuals in the group all together, spend more time on paid work.

Some information on the distribution is contained in the participation rate, i.e. the proportion of the individuals that devote some time to the activity. If the proportion is 100 percent, everyone carried out some paid work, if it is 50 percent, half of them did, etc. If two population groups have the same mean, but the participation rates differ, one can conclude that individuals belonging to the population group with the lowest participation rate on average worked longer hours than those belonging to the other population group, provided that they worked at all. This measure, mean time for those who in fact performed the activity, is often found in time use statistics publications. Note that this measure also conceals the distributions behind. The present system permits calculation of all three measures mentioned above.
 

Measurement at group level

The respondents fill in diaries for one or two randomly designated days. Hence short, random moments in people’s lives are studied. They can not be regarded as representative to the single individuals. Therefore measures of the time use are meaningful only when they are calculated for groups of individuals of considerable size. The groups are formed by the information collected by means of interviews. Significant population groups are sex, age, stage in the family cycle, number of children etc. In the system these and many other population groups may be formed.

Interpretation of the recording domains

Research has demonstrated that it is necessary to apply some kind chronological recording of episodes in time diaries (of some kind) – like it is in HETUS - in order to obtain reliable data on time use.

In HETUS the respondents record their activities in time diaries using own words. This is done for one or two randomly designated diary days. In case two activities were carried out simultaneously, there is space in the diary to record both, a main and a secondary (or parallel) activity. The third recording domain is presence of other persons. Consequently, each recorded episode in the diary is characterised by a main activity and possibly by a secondary activity and by information on presence of other persons. A temporal identifier carries information on the time and duration for the episode. At the stage when the activities in the diaries are coded, information on location, i.e. where the activity took place is inferred and coded into a few different categories. This domain also contains information on means of transportation.

The episode is a behavioural unit. The recording domains, one by one or taken together in various combinations, offer insights in various aspects of people’s behaviour. When the data is used for statistical description and analysis of what people are doing, which activities they undertake, multiple alternative approaches are offered. Which to choose depends on objectives - on which aspects of behaviour, and hence which episodes correspond to the particular interest.

Suppose the question at issue concerns having a meal, it might be quite sufficient to select all episodes in the episode file for which the main activity (the answer to the diary question ”What did you do?” or secondary activity (“Did you do anything else?”) is having a meal. If the meal was eaten in solitude or not, where it was eaten or whether some other activity was going on at the same time, e.g. reading a newspaper is not relevant to the question, so the main and secondary activity together might offer necessary and sufficient information for an adequate classification of the episode. If, however, we were to consider the main activity alone, we would exclude circumstances where the diarist was both eating and watching television, but gave priority to TV in the diary record; having a meal would as a result be undercounted.

A second example. Some years ago it was reported in Swedish media that the people in Sweden on average spent 2 minutes a day talking with children. The basis for this was a figure that had been found in a statistical table published in a report on the Swedish time use survey. The figure is in it self correct but the interpretation of it is completely wrong.

The correct interpretation rather is: when a random sample, drawn from the Swedish population (20-64 years old) record in a time diary of the HETUS kind which activities they undertake and how much time they devote to them (during a randomly designated day), the result is that episodes of a total average duration of two minutes have been described in words that clearly state that the main activity was talking with children. This, however, does not imply that no other talking with children took place. If one wants to estimate the time people spend talking with their children, a different approach is necessary. In principle, each episode in the episode file should be analysed and classified according to whether or not talking with children is likely to occur. For example, assume a respondent is the mother of a child, the mother’s main activity is having a meal and the child is present, no secondary activity is recorded, then it is more likely that the mother talks with the child than that she does not. Hence, to estimate the time for talking with children, all episodes in which it is likely that talking occurs, need to be identified, the durations added and the mean calculated. The information to make use of in order to find the relevant episodes, is contained in the total of the recording domains, not just one of them. How to make use of the information contained in the recording domains depends on the purpose of the analysis. For certain purposes it is – of course – sufficient to restrict the analysis to one or the other recording domain.

An example of a somewhat more intricate activity is childcare. The meaning of the concept determines how to extract it from the diary record. If childcare is defined as activities that directly involve and are directed to the child, as feeding, putting a child to bed, changing nappies, etc. it might be satisfactory to select episodes characterised by main activity codes that indicate these particular activities. If, on the other hand, child care is given a broader meaning, e.g. forming the child’s human and social capital, then additional episodes have to be added. The recording domains, parallel activities and “who with” will certainly have to be considered. And if the concept is expanded further, to, say, custodial care, still more episodes might have to be considered.

Figure 2 sets out some of the possible estimators of childcare. First, there is the primary childcare time. The 2000/01 Swedish evidence suggests that mothers in households with children up to the age of seven devote a little more than two hours and the corresponding fathers a little more than one hour per day to childcare as a primary activity.

Now add in childcare as a secondary activity, and the total rises to three hours for mothers and one and a half hours for fathers. But of far greater significance is the time that parents spend not engaged in explicit child-related activities, but still in the presence of their children.

If meals together with the child are included still one hour is added. If the mothers´ free time activities with the child present are regarded as childcare two more hours are added. And finally, if we regard housework with the child present, we end up with eight hours childcare per day, four times as much as the original 2 hours of primary activity childcare. For the fathers the corresponding estimates increases from about 1 hour a day to a little more than five hours.

Figure 2. Mean time for various sorts of childcare. Married or cohabiting parents with small children, 0-6 years. Swedish population 2000/01.

Which of these is the appropriate base for estimating the extent of non-market provision of childcare services? The primary activity alone is clearly insufficient. But the total of child co-presence time is, arguably, excessive, particularly insofar as it may involve the co-presence of both parents with a single child. Again, the objective gives guidance.

The bottom line is: select all episodes that are relevant to the purpose. The system offers such a possibility for the advanced users.

Differences in structure vs. behaviour

The interpretation of a difference in the mean time spent on one or the other activity between for example some population groups and/or between countries, is not always straightforward. A difference is not necessarily an expression of differences in behaviour. The explanation could also be differences in structure. Or a combination of both.

Assume there is a difference in the time use of married/cohabiting mothers with small children (under 7 years) in two countries. In addition, assume that there are differences in the composition of the population groups. In one country the fertility rate is relatively higher and the mean age of women at the first birth lower compared to the other country. As a consequence the average age of the mothers are lower and the average number of children are larger in one of the countries. If age and number of children influence the time use, the result would differ between the countries even if there is no difference in time use, given the same age of the mothers and the same number of children. Hence, in this case there is no difference in behaviour, only in the composition of the population groups.

For more information about necessary information - follow the link:


punkt Necessary information about variables,
variable content and National deviances
punkt Metadata: population, sample, etc
punkt Variables in the database
punkt Other important and necessary documents

Top of page