Key messages on wider survey context effects
Time-, event- and weather-specific effects can be thought of as a source of error in life evaluation and eudaimonic measures, but they are primarily a source of information in the case of short-term affective measures. This further highlights the importance of the reference period associated with a given question, discussed in the question construction section above. Relatively ephemeral aspects of the environment can have important and valid impacts on momentary mood, but should not in theory show large impacts on long-term life evaluations. The evidence available bears this out to some extent, but there are exceptions - particularly in the case of major news stories or public holidays - and there is very little information available about the impact of wider survey context on eudaimonic measures. The fact that context can have valid impacts on momentary affect nonetheless has implications for its measurement - and these implications are in fact similar to those for managing the error that wider survey context might introduce in more evaluative subjective well-being measures.
Daily events that affect individuals more or less at random, or temporary weather conditions whose effects vary widely across samples for a variety of reasons, may contribute a small amount of noise to the data, but this should not substantially affect analysis of co-variates or the interpretation of group differences. There is very little that survey designers can do to control these events themselves, although where they may be of particular interest (for example, in time-use and other diary or experience sampling studies), useful information can be collected about their occurrence - such that their effects can be considered or controlled in the course of analyses.
On the other hand, major events, rare weather shifts or day or week effects that have the potential to influence a sizeable proportion of regional or national respondents can introduce more systematic impacts on subjective well-being. With some exceptions (such as those noted by Deaton, 2011, and Barrington-Leigh, 2008), effect sizes are not very large, but the comparability of data - between groups and over time - can be threatened. Comparable data will thus depend on consistency with regard to the proportion of weekday/weekend measurement, survey timing in relation to seasons, the inclusion/exclusion of holiday periods, and the absence of major news events or particularly good or bad runs of weather. Measurements restricted to a single day or single survey week may be particularly vulnerable to instability due to the risk of coinciding with significant events, holidays, religious festivals, etc. It is thus preferable to stage data collection over multiple days wherever possible - and ideally throughout the year. Deaton’s (2011) work also highlights the value of exploring data in time series to check for sudden movements in the data that may need to be explored to determine whether they result from something of direct interest to survey data users.
There is also a concern that the hopes and fears that might arise as a result of watching stock market trends and media reporting, because of their cross-national effects on national “mood”, may obscure the effects of policy-relevant life circumstances (such as unemployment), simply because a smaller number of people are affected in the latter condition. This has important implications for analysis and interpretation, but cannot be easily addressed through methodology alone. The one exception is perhaps the reference period emphasised to respondents. If no reference period is specified for life evaluations and eudaimonia, respondents may be more likely to draw on information from recent events, rather than actively searching their memories over a longer period. However, as noted in the earlier section on question construction, relatively little is known about how the reference period alters responses to these types of subjective well-being data. When it comes to recently-experienced affect, shorter reference periods are expected to be associated with more accurate recall - and there is therefore a trade-off to manage between response accuracy and the risk of the measurement day exerting undue influence. Where recent affect questions are to be included, it is therefore strongly advisable that these measurements are spread over a wide period of time, rather than focusing on a particular day or week of the year.
In analyses of panel data, the impact of seasonal trends as well as of relatively ephemeral events will need to be considered, in addition to changes in life circumstances that may drive changes in subjective well-being over time.