Abstract
Population surveys of emotion offer great potential to understand subjective wellbeing, though most do not reveal how emotions other than happiness and satisfaction impact on daily lives. This article presents a case study analysis of data from Kahneman and Krueger’s (2006) Princeton Time and Affect Survey to demonstrate that the choice of emotions or affects measured in surveys does matter in determining wellbeing in contexts such as those in which gender plays an important role. It finds that that tiredness and interest (excluded from Kahneman and Krueger’s wellbeing construct) comprise a large part of American women’s but not men’s unpleasant education, unpaid housework, and childcare. The article concludes by suggesting that the most appropriate method for establishing a “minimum” set of emotions is to conduct survey-based “audits” of emotions experienced in daily activities.
Introduction
Surveys of emotion offer great potential for understanding the social dynamics of emotions, as well as the extent and equitable distribution of subjective wellbeing within societies. Measures of happiness and satisfaction are critical to studies of subjective wellbeing, and complement (or perhaps supersede) conventional measures of progress, such as per capita income and human development (Diener & Seligman, 2004; Layard, 2005). However, national social surveys typically measure only happiness and satisfaction, and usually only in generalised life domains (such as work, family, and finances). Negative emotions such as anger or shame are rarely surveyed, and there is little understanding of the degree to which they might detract from (or overwhelm) the happiness and wellbeing of particular types of people in particular situations and interactions. Conventional wellbeing statistics may mask a more subtle and social interplay of activities, interactions, and emotions in everyday life. They consequently hide inequities between different populations who are forced to experience and manage certain emotions more commonly than others. A clear instance of such wellbeing inequity can be seen in the reported gendered “wellbeing gap” between North American men and women.
Research by Stevenson and Wolfers (2009) shows that the subjective wellbeing of women in the USA has worsened relative to men since the 1970s. The authors suggest women are becoming more dissatisfied as they compare themselves increasingly to men in terms of employment success (Stevenson & Wolfers, 2009). However, when Kahneman and Krueger (2006) collected time-diary survey data on American activities and affect—or how people felt about what they did—and applied affect ratings retrospectively to historical time-use data, they found that women’s unpleasant time had increased relative to that of men. Their findings suggest that the declining relative wellbeing of American women relative to that of men is due at least as much to remaining locked into long hours of unpleasant unpaid work as it is to their increasing comparisons of themselves to men regarding paid work achievements.
Kahneman and Krueger’s (2006) approach is useful for contextualising subjective wellbeing; or establishing how different people feel in different situations, whilst doing different things. It has been applied, for example, to studies contrasting the emotional wellbeing of men and women in countries such as Australia and the USA (Patulny & Fisher, 2012). However, their approach does not address the potentially critical question of which emotions and types of affect are most relevant to understanding differences in unpleasantness across activities and groups, such as men and women.
This article advances the twin themes of (a) appropriately measuring emotions through surveys and (b) exposing a gendered wellbeing gap, using the latter case study to illustrate the value of the former approach. It uses sociological theories of emotion to critique existing time and emotion studies, revealing the potential for such studies to measure many more emotions in social contexts. It demonstrates, through an analysis of Kahneman and Krueger’s (2006) time and affect data, that the choice of emotion or affect used to measure wellbeing does matter in determining the relative wellbeing of men and women. The article then argues that the exclusion of many social emotions in most surveys is indicative of a larger failure to investigate the full range and contexts of emotions pertaining to social lives, and suggests methods of redressing this.
Sociology of Emotion Surveys and the Gendered “Wellbeing Gap”
Existing Time and Emotion/Affect Surveys
New evidence on emotional wellbeing is emerging from studies of emotions and time use, which inquire into how people feel about what they do with their time and who they spend it with. Such research comes from a tradition of time-diary analysis (Csikszentmihalyi, 1990; Michelson, 1977). The more sophisticated of these investigations have inquired into the enjoyment associated with types of activities in general (Juster, 1985), and during particular episodes (Gershuny & Halpin, 1996; Robinson & Godbey, 1997). Sometimes data are gathered via experience sampling methods (ESM), or using electronic “beepers” to prompt people to report activities and feelings at real, randomised times (Zuzanek, 2012). However, these studies have the disadvantage of “emotional selection bias” creeping into real-time responses; people who are angry, annoyed, or tired may not feel like responding to the beeps when they come.
An alternative method that avoids such biases is the day reconstruction method (DRM) employed by Kahneman and Krueger (2006). This method asks survey respondents about three random episodes from a time diary of the previous days’ activities, and about the presence and intensity of six types of affect during each episode: happiness, sadness, pain, stress, tiredness, and interest. Any episode where the intensity of sadness, stress, or pain outweighed that of happiness was labelled as “unpleasant,” and a U-index—or proportion of unpleasant episodes in a day, or within the contexts of certain activities—was calculated.
Their DRM has certain inherent drawbacks. In addition to common problems with diary methods, such as misremembering experiences (though prompts about activity, location, and the presence of others help minimise this), a key weakness of their approach is to inquire about only a few, limited affect-types to calculate their U-index. The lack of emotional range, and the failure to examine how these emotions change and are managed during the interactions inherent in the diary activities, reveals the “asociological” nature of their design and analysis. Research from the sociology of emotions helps to reveal this.
A “Sociology” of Emotion Surveys
Sociological theories of emotion raise issues of what we mean by “emotion” and “affect,” and thus which emotions and affect-forms are most important to measure in surveys. Kahneman and Krueger (2006) provide little justification for the choice of sensations they measured, or their labelling of them as “affect,” at least not with reference to any of the various cultural theories of affect (see Iedema & Carroll, 2015, for more detailed discussion). They seem united with such theorists, however, in conceiving of affect as a quality broader in scope and experience than “emotion,” definitions of which denote recognisable body feelings, physiological changes, expressions, and motivations to act (Fridja, 2000), to which sociologists (Lupton, 1998; Thoits, 1989) add socially and culturally shaped situational cues, labels, and interpretations.
These latter social qualities imply that several “social” emotions might make good candidates for sociological surveys by warrant of their containing microsocial or social structural properties. These include, for example, anger (Kemper, 1990), excitement (Collins, 1990), shame (Scheff, 1998), embarrassment (Shott, 1979), and love (Illouz, 2008). Such emotions may have a strong and as yet unperceived impact on wellbeing. For example, if women experience anger (or frustration) more commonly than men as some studies have suggested (Schieman, 2003; van Kleef & Côté, 2007), will emotion surveys excluding anger fail to reveal important differences in gendered wellbeing?
The sociology of emotion also suggests a number of other qualities which a survey of emotions might capture to assess wellbeing in social contexts more accurately. The fact that people often manage their emotions (Hochschild, 1979) implies potential for bias in surveying emotions, in that a respondent might “tone down” remembered emotions as an act of emotion work in order to make responses more normatively acceptable. The socially unacceptable shame (Goffman, 1963) or envy (Clanton, 2007) of yesterday may be toned down and reported as the more acceptable “stress” of today. This might pose validity problems for the survey, though the distance created by the delay in experiencing and reporting emotions allows for consideration as much as repression of emotion, and for engagement in more open emotional reflexivity (Holmes, 2010; King, 2010) than repressive emotion work.
These theories raise questions as to whether surveys of time and emotion/affect capture enough of the social or interaction-specific qualities of emotions. Most importantly, they raise questions about whether the choice or omission of an emotion in evaluating the “unpleasantness” of an activity matters, and for whether this impacts on the wellbeing of different groups, particularly for men and women. Kahneman and Krueger (2006) excluded feelings of being interested or tired from their calculation of unpleasant time (the reasons for which are unclear), and this is likely to have mattered in the context of gendered wellbeing.
A Gendered “Wellbeing Gap”
There are several reasons why men and women might experience different feelings in the context of different activities. Leisure time has expanded more for men than women (Aguiar & Hurst, 2007), and women have maintained higher involvement in unpaid care work in all countries (Fisher & Robinson, 2010; Kan, Gershuny, & Sullivan, 2011). Women are more likely to be subject to general work–life balance pressure. They report higher stress and fatigue levels as working mothers (Gimenez-Nadal & Sevilla-Sanz, 2011), particularly those with long work hours and small children (Gunthorpe & Lyons, 2004). Work balance stress may also contaminate the experience of leisure activities, with Gimenez-Nadal and Sevilla-Sanz (2011) finding that European working mothers report the least satisfaction with their leisure time.
Sociological studies of gender and emotion also reveal qualitative differences in the nature of women and men’s work that might produce different emotions. Women are more likely to undertake work in care roles that require more “emotional management” (Hochschild, 1979; King, 2007), which would coincide with heightened tiredness as well as stress at work. They are also more likely to work in lower paid, lower skilled jobs (Walby, Armstrong, & Humphreys, 2008), which may correspond with less interesting work.
How Big Is the Gendered “Wellbeing Gap” in America?
Data and Methods
To illustrate the difference made to subjective wellbeing by the choice of affect used to measure it, I reanalyse data here from Kahneman and Krueger’s 2006 Princeton Affect and Time Survey (PATS). This is a nationally representative telephone survey of 11,905 respondents in the USA. Survey weights have been created and used to correct for the gender bias in the sample. The various time-based activities have been divided into 12 categories based on Robinson and Godbey (1997), but separating out childcare and TV watching. Categories and activities (Table 1) are consistent with the distribution of such time according to other time-use surveys (Kahneman & Krueger, 2006).
Activities by categories.
To reveal the dynamics associated with different affect-types, I partition the U-index into the proportion of unhappy episodes in which happiness predominates (i.e., is of higher intensity), interest predominates, or the two are evenly matched. I take a similar approach with the negative affect-types. In the event of equal intensity ratings between negative emotions, I established a hierarchy whereby the least commonly reported emotions are given precedence over more common emotions so as to maximise information about more obscure emotions. Sadness thus predominates over all other negative emotions in the event of an intensity draw, followed by pain, stress, and then tiredness.
Results – The Choice of Emotions Matters!
Table 2 displays mean scores and the proportion of episodes that contained any mention of that affect, for each affect, for men and women. It also shows U-indexes calculated and compared using four, five, and all six types of affect for men and women. There is little aggregate difference between using four or all six of the affect-types available in the PATS survey, with each U-index producing a similar proportion of about 20% of activities as “unpleasant” (Table 2). However, this disguises the countervailing effects of interest and tiredness. Recalculating the U-index to include “interested” reduces the U-index for either gender down to 13%, whereas recalculating it to include “tiredness” increases it to 31%. Even in aggregate, these types of affect are clearly not irrelevant to calculations of time-based subjective wellbeing.
Proportion of U-index episodes, and mean scores and proportion of episodes by affect-type.
Figure 1 compares the unpleasantness of activities for men and women on the basis of which negative emotion was reported as being predominant. Activities are ranked from left to right in order of decreasing unpleasantness. Including tiredness changes the mix of negative affect-types substantially. Even accounting for tiredness being trumped by all other negative affect-types in the event of a draw, tiredness is the dominant negative affect-type in many activities. Women exhibit tiredness more frequently in eight of the 13 activities, including all forms of contracted and committed activity except volunteering. Tiredness comprises a large part of women’s but not men’s unpleasant education, unpaid housework, and childcare; it drives much of the gender difference in the unpleasantness of these activities.
Proportion of unhappy episodes by sex, partitioned into four predominant negative affect types.
Figure 2 shows the gender gap (women minus men) in the proportion of unhappy episodes experienced by women and men when different U-indexes are calculated. It compares the four U-indexes across activities, with bars above the line indicating that women report a higher proportion of unpleasant activities than men. Activities are ordered from those with the largest gender gap (left; education) through to those with the smallest (right; out home leisure) according to Kahneman and Kruger’s original U-index. It becomes immediately apparent that the gender wellbeing gap increases substantially for education, sleep, unpaid domestic work, and particularly child care, when we include “tiredness” in the calculation of unpleasant time. Media and domestic work change from a negative gender gap (men find them more unpleasant) to a positive gender gap (women find them more unpleasant). Volunteering changes from a positive gap (more unpleasant for women) to a negative one (more unpleasant for men). And so on; the gender gaps shift and change as different affect-types are included and excluded from the U-index calculations.
Gender gap in unhappy episodes by activity type, comparing U-indexes calculated using four (original), five, and six types of affect.
The Need for an “Audit” of Emotions in Assessing Subjective Wellbeing
Including or excluding certain emotions changes both the overall and relative pleasantness of activities for the sexes in different circumstances. It is clear then that a more comprehensive set of emotions needs to be developed for inclusion in surveys of time and wellbeing. The obvious question then becomes: which emotions do we include?
I would suggest an approach similar (and largely derived from) Kahneman and Krueger’s DRM (2006), but with two important differences. The first is that instead of asking respondents to record incidence and intensity of set emotions experienced in the context of random episodes, respondents should be allowed to report whatever emotions they experienced during the activity in their own words. This would enable a first-stage “audit” of emotions to take place: identifying the most common emotions experienced by the population in their own, unfiltered, words.
The second difference is to inquire about whether any emotion work was undertaken. In addition to their reported emotional experiences, respondents would be asked if they felt the need to hide or change what they felt. This would serve to provide socially contextualised baseline estimates of the degree of emotion work that goes on in the population as a whole. The degree of emotion work could also be used to modify the baseline “audit” of emotions, producing a more accurate, or adjusted estimate of wellbeing.
This approach has limitations, in that responses are not precoded into explicit categories enabling easy analysis. However, it is possible to subject such responses to textual/quantitative or qualitative analyses to see which responses cohere, and create quantifiable categories for analysis. There may also be some difficulty in gathering reliable survey data on topics such as envy, which are susceptible to management and repression. However, it is possible to use surveys to “watch for the disguises in which [envy] often appears” (Clanton, 2007, p. 426). Measures of activities such as gossip, or emotions such as “frustration” might well be indicative of envy when contextualised (such as when one partner is engaged in work and care while another is engaged in leisure). Asking additional probing questions such as what the source or target of the emotion might be may help elicit more open responses.
The main benefit of this approach is that even if it arrives at the same set of emotions used by Kahneman and Krueger (2006), it confirms that these emotions are indeed the most relevant ones for population surveys. It will also identify any surprising additional emotions such as “boredom,” “anger,” and “guilt” that might appear more regularly in contextual situations than they do in the more generic domains captured in existing surveys. Furthermore, this method will contextualise emotion work in daily social interactions.
Conclusion
This article makes two valuable contributions to the study of emotions and wellbeing. Firstly, it exposes gender inequity in wellbeing—or the “wellbeing gap” between men and women—in the USA. It shows that American women experience more unpleasant time than American men than previously recognised when different types of emotions and affect are included in the measure of wellbeing. Tiredness comprises a large part of women’s but not men’s unpleasant education, unpaid housework, and childcare; it drives much of the gender difference in the unpleasantness of these activities. Secondly, the article makes the more general point that we cannot understand emotional wellbeing fully without measuring all the relevant emotions in their appropriate social contexts. Kahneman and Krueger (2006) make valid criticisms of conventional wellbeing research, but their use of a limited range of affect-types confounds their attempt to comprehensively assess wellbeing across the spectrum of daily life.
This article argues that the most appropriate method for establishing a “minimum” set of information about emotions and emotion work in daily social life is to conduct a national survey-based “audit” of emotions and emotion work in the context of daily activities. Such an approach is population- focussed, does not rely on experts to choose emotions, is contextual, and enables participants to choose for themselves what those feelings are and how they should best be labelled. It is thus the most appropriate survey-based method for capturing the full range of possible emotions associated with microsocial interactions.
I conclude with two final suggestions for capturing taboo emotions. The first is that, with sufficient resources, this approach could be piloted in conjunction with an open-ended ESM study using mobile phone applications. Comparison could reveal the extent to which normative reporting biases are overcome by “putting people on the spot” (ESM) or giving them “time to reflect” (DRM). The second is to survey all members of a household. This would not only help capture taboo emotions through inquiring about what the respondent perceives the other family members are feeling at the same time, it would also prove valuable in helping to understand gendered emotional dynamics within families.