Alone Together: Gender Inequalities in Couple Time

An important body of research has used time diaries to assess the transformation of gender relationships at home. However, little is known about how partners perceive time shared together. While the household division of labor still remains heavily gendered, it can be expected that what partners do, even when they are together, is also gendered. The aim of this paper is to address the question of the discrepancy (or mismatch) in couples’ reporting of time together as well as the potential discrepancy in the activities engaged in during shared time. Using the 2015 UK Time Use Survey, I show that there is no gender difference in how partners report being together; however, important gender imbalances exist in what partners do when together. In particular, I find that, when together with their partner, men are much more likely to watch TV and enjoy leisure while women do domestic chores. I conclude by discussing different concepts of time together and the usefulness of couple-level diary data for studying gender relationships at home.


Introduction
Gender relationships have dramatically changed since WWII. Intimate relationships have progressively shifted towards a model where love, equality and intimacy have become core values of partnership and marriage (Beck and Beck-Gernsheim 1995;Cherlin 2010;Coontz 2005). Gains from marriage have shifted from household specialization, with complementary in production, to a model of complementarities in consumption and in particular in the joint consumption of leisure (Mansour andMcKinnish 2014: 1128;Stevenson and Wolfers 2007).
The search for intimacy and fulfilling partnerships is thought to have heightened the need for time shared in leisurely activities (Voorpostel et al. 2009). Research on togetherness often emphasizes that time shared between partners, or couple time, is important for partnerships (Hill 1988). Indeed, studies have found a positive association between time shared with partners and well-being and marital quality Gager and Sanchez 2003;Sullivan 1996).
However, little attention has been paid in time use research onto how partners differ in the way they perceive time together and into the potential gender inequalities during time together. Most time use research, based on individual diary data, have to assume that partners share the same activity when together . Little is known about how partners' actually divide their activities during shared time.
The aim of this study is twofold. The first aim is to establish if partners systematically differ according to their gender in the way they report time together. The second aim is to explore what partners do when together and if gender inequalities do persist during shared time.
To answer these questions, I use the latest United Kingdom Time Use Survey (UKTUS) 2014/2015. Time use diaries offer a unique micro-sociological account of couple and family life. The UKTUS collected diaries from both partners enabling a true comparison of partners' activities during shared time, in contrast with some other time use surveys, notably the American Time Use Survey, that collects only one diary per household.

Time Together and Families
Family sociologists have long hypothesized the importance of time together for couples' solidarity and for family cohesion. Berger and Kellner (1964) stressed the importance of daily contacts in the making of a relationship. Studies have generally found a positive association between time shared with the partners and relationship quality and wellbeing (Crawford et al. 2002;Flood and Genadek 2016;Sullivan 1996;Zuo 1992). Studies have also shown that partners actively coordinate their schedules to spend time together, stressing the importance of time together for couples (Carriero et al. 2009;Hallberg 2003;Hamermesh 2000;Klaveren and van den Brink 2007), and adding evidence for the hypothesis that togetherness is a crucial mechanism for the functioning of contemporary relationships (Buswell et al. 2012;Hill 1988;Lesnard 2004). The centrality of time shared with the partners can be placed in the general context of the transformation of personal relationships in Western societies (Beck and Beck-Gernsheim 1995;Buss et al. 2004;Giddens 2013;Schwartz 1995). The importance of time together for relationships is also reflected in the fact that time spent together with the partners has been increasing over time Neilson and Stanfors 2018;Voorpostel et al. 2009).

Gender Differences in Time Use
A large body of research has analyzed gender differences in time use activities (Bianchi et al. 2006;Gershuny 2000;Sullivan 2000). Differences can be found between men and women not only in the average time spent in different activities, but also in the way activities are carried out. For instance, women's leisure time is more fragmented, more often spent in the presence of children and more often combined with other less enjoyable activities (Bittman and Wajcman 2000;Mattingly and Bianchi 2003). Men also tend to engage in more leisure activities while women are at work than vice versa (Dush et al. 2017: 12).
Combining several activities, also known as "multitasking", is an important strategy for parents trying to gain time in their busy lives (Bianchi et al. 2006). Mothers in particular are more prone to multitasking compared to fathers, which can create an extra sense of time pressure and negatively affect their subjective well-being (Craig 2007;Offer and Scheiner 2011). Multitasking in an important component to consider when studying the division of household labor (Craig 2006(Craig , 2007Sayer et al. 2009). Bernard (1972) argued that experiences within a marriage are fundamentally different for men and women. Men and women do not perceive, experience, or share the same expectations about relationships (Ferree 2010;West and Zimmerman 1987). Women are more often dissatisfied with their relationship than men, more often think about separation, and more often initiate divorce (Thompson and Walker 1989). This dissatisfaction has several complex causes, such as differences in expectations, inequalities in the division of labor, as well as differences in perceptions of fairness and justice (Thompson 1991;Wilkie et al. 1998).

Time Together and Gender
Gager and Sanchez studied how couples perceive the amount of time they spend together (2003). They did not find strong gender differences in the reporting of time together (even though they found evidence that shared time increased marital solidarity but only for women). However, the study did not measure time together using time diaries. Evidence of differences in the reporting of time together using time use diaries is mixed. Lesnard found that in the French Time Use Survey about 76% of couples "have converging statements of being together", questioning if the rest of the discrepancies are "mistakes or reveal diverging gendered views on being together" (2008: 465). Using the American Time Use Survey, Flood and Genadek (2016) as well as Dew (2009) found that women reported less time with their partner, on average, compared to men. Pooling time use surveys from the United States, France and Spain, Garcia-Roman et al. also found that women on average reported less total time with the partner (2017). On the other hand, Neilson and Stanfors did not find evidence of discrepancy in the reporting of time together by gender in Sweden (2018). Freedman et al. (2012) using couple-level time diaries did not find systematic differences between partners in the report of time together.
The time use literature is ambivalent regarding the relationship between time together and gender. On the one hand, the literature often describes time together as an important mechanism for couples to enhance their relationship while on the other hand acknowledging the potential gendered discrepancy between partners. Given the persistence of gender inequalities in contemporary societies, often revealed in time use research by the unequal distribution of activities between partners, it is unlikely that time together itself is not affected by gender.

Source of Variation in Time Use
One important source of variations in couple time use stems from the economic resources deriving from employment that are available to partners. Marital bargaining theories predict that unpaid work is a function of the economic resources possessed by each partner (Bittman et al. 2003;Deutsch 2007;Gupta 2006;Sullivan 2011). Economists have been interested in how working hours affect the joint consumption of leisure (Hallberg 2003;Mansour and McKinnish 2014). Mansour and McKinnish finds that "couples who are less specialized in hours of work spend more joint time together" (2014: 1143). So, the labor market position not only impacts the division of household work, but also the propensity of couples to spend more or less time in joint activities. More generally, employment status also affects time together 1 3 in simply reducing the opportunities for partners to be together. Several studies have shown for instance that dual earners spend on average less time together than single earner couples Glorieux et al. 2011;Neilson and Stanfors 2018). Non-standard working hours also affect the opportunity for togetherness (Lesnard 2008;Presser 2005).
Another important source of variations affecting couples' time use is the presence of children and in particular of young children (Berk 1985). The transition to parenthood negatively affects women's labor market participation, and increases their housework time (Budig and England 2001;Mattingly and Bianchi 2003;Sanchez and Thompson 1997). Research on togetherness shows that children reduce the time that partners spend together as a couple , and directly impact the jointness of leisure consumption (Hamermesh 2000: 25).

Research Questions
Most time use studies use independent samples of men and women to study household interactions or the household division of labor (Bianchi et al. 2000;Wight et al. 2008). Including both partners' activities can shed new light on the gender dynamics at play during shared intimate moments (Dush et al. 2017;Smith et al. 1998;Szinovacz and Egley 1995).
The aim of this paper is to address the question of the discrepancy (or mismatch) in couples' reporting of time together as well as the potential discrepancy in the activities engaged in during shared time. Partners' shared activities are an opportunity to look more closely at the gendered dimension of intimate life in a micro-sociological way (Kellerhals et al. 1993).
Expectations are mixed regarding how activities are distributed during shared time. On the one hand, given the persisting gendered division of intimate life, we can expect to see differences between men and women in activities even during time together. On the other hand, because time together has been described as a particularly important way for a couple to (re) connect, we can expect that most couples share the same activity when together, and most particularly, to share leisure activities.
I focus mainly on the two sources of variation in couple time use discussed in the literature overview, the presence of children and the employment status. I expect that having children would reduce the jointness of activities during shared time and deepen the gender gap in activities. Similarly, I hypothesize that couples who are less specialized in terms of their work arrangements would have a more equal distribution of activities during shared time when compared with other working arrangements where men have higher market resources than women. However, it is not entirely clear if resource theory can be applied to the distribution of shared activities.

The UK 2014-2015 Time Use Survey
The UK Time Use Survey 2014-2015 (UKTUS) is a nationally representative, stratified random sample of UK households that forms the latest contribution of the UK Office of National Statistics to the Harmonized European Time Use Study (HETUS). The survey collected over 16,500 diaries from 9388 people in 4238 households. The survey has a response rate of 40.5%, comparable to the American Time Use Survey or the Swedish Time Use Survey (Neilson and Stanfors 2018).
Time use diaries have been used extensively to trace social change, and to estimate daily activities such as working hours, leisure time, domestic chores, eating behavior, and childcare (Gershuny 2000;Robinson and Godbey 2010;Vagni and Cornwell 2018). Diaries differ from stylized survey questions, which ask respondents to recall and estimate the time they spend on different activities. Stylized questions generally take the following form: "How many hours did you spend doing…?" However, these sorts of questions have been criticized because of desirability and memory bias (Juster et al. 2003). Instead of asking respondents to estimate average time spent in different activities, time use diaries ask respondents to record all of their activities during a given day. This method has been shown to reduce bias (Juster et al. 2003).
The UKTUS 2014-2015 was collected using paper diaries. A face-to-face interview was conducted first, and time use diaries were given to each member of the household to fill out. The UKTUS includes two time use diaries for each respondent: one weekday diary and one weekend diary. Each household member older than 8 years old was requested to fill out a time use diary. The diaries are structured in 10-min intervals starting at 4 a.m. The respondents had to fill out 144 time intervals (144 * 10 equals 24 h). The diaries enabled the respondents to write down their main activities, their secondary activities, their location, and other persons who were present (with whom) during each 10-min time interval. The with whom information collected information about the spouse/partner, the child, and others. In this paper, I am primarily interested in the information related to the presence of the partner/spouse. The exact wording of the question was: "Were you alone or with somebody you know?" The respondent could then tick a box indicating "spouse/partner." Checking the quality of the with whom information, I found that less than 1% of respondents had complete missing information, and 45% had no missing at all. About 85% of the couple sample had less than 25% of missing copresence information. I also analyzed missing information by gender and found no differences. I chose to remove couples with more than 50% of missing copresence information from my analyses. (For a full discussion see Online Supplement). In effect, the missing copresence information of the partner in the analytical sample is counted as 0 (not present). This limitation should be kept in mind.

Sample
The analytical sample used in this paper is composed of 1427 heterosexual couples (2854 individuals) in which both partners filled two time use diary during the same 2 days.
The analyses presented in this paper were run on both the complete sample of couples and the subsample of parents to detect if some of the results were conditioned on having young children living in the household. A full description of the sample characteristics can be found in "Appendix 1". The subsample of parents is composed of 333 couples (666 individuals) with at least one child under 8 years old and where at least one parents is employed.
Because I restricted my analysis to couples where both partners filled a diary, I was concerned about possible selection bias. In order to address this question, I compared the UKTUS to a much larger survey, the UK Labor Force Survey, 1 and find little differences in key socio-characteristics between the two surveys (see Online Supplement). Therefore, I did not use the survey weights for the analyses.

Measures and Analytical Strategy
The main variable of interest in this paper is time spent together, as indicated in the "with whom" field of the diary. The exact wording of the question in the time use diary was "Were you alone or with somebody you know?" from which the respondent could indicate if his or her partner/spouse was present. The measure of time together is a dummy variable, with 1 indicating the partner/spouse being present during an activity and 0 indicating absent. The measure excludes paid work and sleeping periods. Therefore, partners reporting being together during these activities were not counted.
The first measure of interest indicates when both partners report being together (when their responses match). It can be defined as: where t represents the time interval (10 min), with t = (1, … 144), p t (for partner presence) is a binary variable (0, 1) indicating if the partner is present or not during an interval t, p m jt denotes the men's response at time t, and p jt w represents the women's response at time t (from the same couple j). Conversely, a mismatch measure would be defined as: The mismatch measure can be decomposed into a men's discrepancy component (DMen j ) and a women's discrepancy component (DWomen j ) . For instance, the men's discrepancy measure denotes when the (male) partner indicates being with the partner but the partner does not indicate being together (and vice versa).
The total time together can be thus defined as: Total jt is therefore the sum of the agreement of being together plus the disagreement of the male and female partners.
The activities considered in this paper have been recoded into six categories: (1) childcare (abbreviated CH), (2) domestic chores (abbreviated DM), (3) eating meals (abbreviated EA), (4) leisure (abbreviated LE), (5) TV watching (abbreviated TV), and (6) other (abbreviated OT). Childcare includes both routine childcare (e.g. feeding the child) and developmental childcare (e.g. playing with the child, reading). Domestic chores include all domestic tasks such as cleaning the house and preparing meals. Leisure is a broad category that includes going to a museum or the cinema and attending sport events but excludes TV watching, which has its own category. Finally, the "other" category groups all other activities: paid work, travelling, shopping, sleeping, personal care, and missing activities.
The second main measure I am interested in is the level of agreement about activities done when partners are together. The agreement in activities means simply that partners are doing the same thing when together. This is more simply illustrated by a cross-table of partners' activities when together, as shown in Fig. 1.
Total jt = Match jt + DMen jt + DWomen jt , The percentage of agreement about activities when together is calculated by dividing the sum of the matched activities (agreement) by the total of activities (both matched and mismatched). This index measures the agreement within couples of activities when together. In a stylized form, this can simply be expressed as: where n ij is the observed frequency for cell (i, j). One limitation of this measure is that couples that did not spend at least one episode together and that did not have at least one matched activity are excluded. In effect this concerns only a small number of couples (n = 49). Otherwise, the measure is insensitive to the total time spent together. I use OLS regression to study the association between a range of covariates and the diagonal measure described above. However, I use a multilevel approach when I analyze the time spent in different activities by gender. The diagonal activity measure is a "couplelevel" measure (each row represents a couple), but the time spent in activities is an individual measure, and because I am dealing with individuals nested in couples, multilevel modeling is an appropriate approach to account for this form of non-independence (Goffette 2016). The samples slightly differ between the OLS and the multilevel models. The multilevel models include couples that do not have at least one matched activity episode. Both models however exclude couples that did not spend any time together.
All the analyses presented in this paper are weighted so that the days of the week are represented equally.  Table 1 shows total time reported as together disaggregated by the matched responses and the discrepancies in matching for men and women. The daily total time together is 408 min for parents and 502 min for all couples. The partners' responses matched (i.e. both partners report being together at the same time; Match j ) 64.55% of the time for all couples, although the proportion is lower for parents (57.44%). The discrepancy in partners' responses is not skewed toward women or men as shown in Fig. 2. For all couples, DWomen j is 17.79%, while DMen j equals 17.66%, meaning that in about 18% of cases partners' report the presence of the other, but the other does not. For parents the discrepancy is higher, at DWomen j 22%, and DMen j 20.59%. Figure 2 shows the average report of time with the partner for man and woman belonging to the same couple. The two distributions are very similar confirming that the report of time with the partner is not bias according to gender. The yellow points indicate that male and female partners are reporting on average the same time with the partner. The blue crosses indicate when the female partner is reporting more time and the green squares when the male partner is reporting more time. Figure 3 shows how time together varies throughout the day. The solid line indicates total time together, the long dashed line the Match j index, the "mismatched" lines at the bottom indicate the women's discrepancy ( DWomen j ) and the men's discrepancy ( DMen j ) respectively. The trends in men and women's mismatches closely follow each other during the day. It visually seems that parents have more mismatches compared to all couples. Women's discrepancy ( DWomen j ) is slightly higher during the morning, while men's discrepancy ( DMen j ) is a bit higher during the afternoon and evening. Tables 2 and 3 show the diagonal of activities (being together and doing the same thing). First, the sum of the diagonal is around 60%; this means that partners, when together, are doing the same activities about 60% of the time.

Results
In 62% of cases, partners do the same activities when together (for all couples). The distribution of the diagonal itself shows that when together and reporting the same activity, couples are mainly watching TV, eating together, and even doing domestic chores together. Table 3 shows the diagonal for different couple employment configurations. Looking at the total agreement (diagonal), we can note that full-time working couples, as well as female breadwinner couples, are more likely to agree on activities than other couples. Again, it is important to understand that this is not the overall distribution of activities, but the distribution of activities done together. The picture could be very different when looking at the overall distribution. Figure 4 shows the sequence of activities for men and women over the course of the day (left hand side) as well as the aggregate distribution of activities when together (right hand side). The way to read these figures is to compare the men's distribution of activities to the women's distribution of activities and to look for differences in activities.
At first glance we can see that couples in general do the same activities when together. We can see a "spike" in eating time around 13 pm. TV time slowly increases from 4 pm until peaking around 22 pm. Gender differences are noticeable, as we can see that women perform more housework (when together) than men during the day (color red). This is more visible on the right-hand side figure showing the aggregate proportion of joint activities. TV is the activity most likely to be shared between partners. But we can clearly see that women do more domestic work and more childcare than men during shared time, while men watch more TV than women during shared time. This could be thought to be an artifact of analyzing only primary activities. However, I looked at secondary activities (Online Supplement), and I did not find any evidence that men report more domestic work as a secondary activity (or women more TV) when together. Even if we look at activities when partners disagree about being with one another ( Mismatch j ), we do not see that women tend to report more TV or men more domestic work (see Online Supplement).
The sequence for parents also shows that mothers report more domestic work and more childcare during shared time compared to fathers. Looking at the sequence figure (left hand  side), we note, in particular, the greater proportion of childcare done by mothers around 18 pm and the greater proportion of TV watching for fathers at this time. Figure 5 shows the distribution of the partners' activities when the respondent is doing a certain activity. In Fig. 5, the figures on the left hand side ("Agreement. Men. Primary Activities.") show what women are doing when men are doing certain activities, and vice versa for the figures on the right hand side ("Agreement. Women. Primary Activities."). For instance, when men are doing childcare, 53% of the time women are also doing childcare. However, when women are doing childcare, men are doing childcare for only 34% of the time. When women are doing childcare, 17% of the time men watch TV, compared to only 7% of the time for women when men report doing childcare. Even though eating and watching TV are the most "agreed upon" activities, women do still more domestic chores and childcare while men eat or enjoy leisure. When men report leisure, about 16% of the time women are doing housework, compared to only 8% of the equivalent time for men. I looked at secondary activities to see if men were less likely to report domestic chores or childcare as primary activities, but this is not the case (see Online Supplement). If anything, the inclusion of secondary activities increases the gap in housework between the partners.
The OLS regression on the diagonal is presented in Table 4. I present four models, including the following variables: Household with a child under 8 years old (ref = no children under 8 years old), Employment status (reference = Both Working Full Time), and Mean age of the partners (ref = less than 44 years old). The constant is 0.63, meaning that partners report doing the same thing while together 63% of the time when together. A score of 1 indicates no mismatches in activity when together. We can see that children reduce the jointness 2 of activities by about 6.7% (p < 0.001). Regarding the employment status of the couple, we see that the categories "Men working full-time and Women part-time", "Both Retired" and "Other arrangement", have all lower agreement about activities compared to "Full-time dual-earners". However, employment status does not explain very much of the variance (R 2 = 0.007). The coefficients for the presence of young children are stable across models, suggesting that children do negatively impact on the jointness of activities. I omitted the model for parents because none of the variables used here were significantly related to the diagonal (results available upon request).     I show the models for Domestic Work, Childcare (for parents only), TV, Leisure and Eating. We can see that men, on average, do 18.92 min (p < 0.001) less domestic work, and watch 17.33 min (p < 0.001) more TV while together with their partner. Women also tend to report less eating while together (− 4.93 min, p < 0.001). There were no significant gender differences in joint leisure.
We can note that unemployed and retired couples watch on average more TV together compared to Full-time dual earners. The presence of children reduces time watching TV together (minus 16.48 min), leisure time (minus 8.98 min) and eating together (minus 8.70 min). Table 6 shows the results for parents only. We can see here that women are doing more domestic chores, more child care, watch less TV and eat less often than men, when both partners are reporting being together. Overall, the results for parents are very similar to the results for all couples.

Discussion
This paper explored the agreement and disagreement in time together, as well as the distribution of activities during shared time, between partners. I find that overall when partners report being together, they are sharing the same activity. This is consistent with the study of Freedman et al. (2012).
The most prevalent activities performed together in a couple are watching TV, eating and sharing leisure. In this sense, this paper is in line with studies arguing that shared time Table 4 OLS regression on the diagonal (%) when partners report being together (all couples) + p < 0.1; *p < 0.05; **p < 0.01; ***p < 0. and shared leisure appear as fundamental components structuring contemporary relationships (Beck and Beck-Gernsheim 1995;Berger and Kellner 1964;Mansour and McKinnish 2014). I also did not find any clear gender differences in the average reports of time together. Female partners do not report more or less time together than their male partners. This is in line with Freedman et al. (2012) and Neilson and Stanfors (2018). The number of mismatches within a couple, on average, balances each other's. One of the implication is that time use researchers interested in estimating time together could either use men's or women's reports of time together for the estimation of togetherness. However, the issue with individual diary data is that, overall, time together is likely to be either underestimated or overestimated, depending on one's interpretation of what constitutes time together. For instance, consider the results presented in Table 1 for parents. The total time together was 408 min (100%), of which 57% was matched and about 20.6% was a mismatch for men and 22% for women. With data from both members of a couple, it is possible to present the different estimates based on men and women, or both. But with data only from one individual, about 20% of mismatched time together will be missing. Imputing partners' responses cannot solve this issue. This leads to the question of what constitutes time together. Should both partners agree on being together, or is only one partner's report sufficient for counting time together? What I called the total time together constitutes the sum of agreement between partners plus the sum of disagreements. It is reasonable to assume that when only one partner reports being together, the other partner is physically present, even if he/she does not recognize that they are together. However, it is impossible to say if the activity is necessarily shared when only data from one individual of the couple is available; thus, time use researchers should be cautious in assuming that partners are doing the same activity when only one partner's diary is collected. Following from these observations, I would like to propose two new broad concepts of time together: shared presence and time engaged. Shared presence should be counted as the total time together, including the disagreements or mismatches. This concept can be viewed as a record of the partner's presence. The mismatches are likely to represent the fact that the partner is present but not truly engaged in a shared activity.
The time engaged measure should be restricted to the times when both partners report being together. We can assume that when both partners report being together, they must agree to some extent about what is going on and how a particular situation should be interpreted. One could go a step further and restrict the engaged time together to when both partners report being together and when both report doing the same activity. Only data collection from both partners can fully account for these complexities.
The results presented in the second part of the paper clearly show the gendered dimension of time together. These findings bring together the disparate time use literature on time together Neilson and Stanfors 2018) and on the division of household labor (Bianchi et al. 2000;Gupta 2006;Sullivan 2000). I showed that, when men and women reported spending time together, men were doing more leisure activities, while women performed more unpaid activities. This result confirms previous research (Dush et al. 2017).
This finding can be put into perspective with Bernard's idea of the two marriages (1972). To some extent, men and women sometimes have contradictory everyday experiences of their shared time. The traditional division of household labor still exists even during shared moments-moments which some of the literature describes as a time when partners can work on the relationship and potentially improve their marital quality (Berger and Kellner 1964). However, the gendered component of time together indicates that we must be cautious in interpreting time together as a necessarily homogenous and agreed-upon time. In this sense, the utility of couple time or family time is not the same for men and women. Women still are doing more housework than men during shared time. The analysis of secondary activities did not change this picture.
This imbalance in the utility of time can be related to studies suggesting that women are generally more often dissatisfied with their relationship than men (Fowers 1991). We can assume that doing unpaid work while one's partner is enjoying leisure activities (e.g., preparing dinner while one's partner is watching TV) could affect the perception of fairness and equity, which in turn could affect partnership satisfaction. More research is needed in this area. It would be possible, for instance, to use some time diary data to explore how the discrepancy in responses and activities affects partners' enjoyment of time. A comparison of partners' enjoyment using couple data would shed light on these issues and inform contemporary debate about persisting gender inequalities.
Another contribution of this paper is the analysis of what I referred to as the "activity diagonal". With or without looking specifically at shared time, the diagonal is a simple way to summarize the division of activities within couples. One advantage of the cross-tabulation of partners' activities is that cross-tables integrate the 24 h-constraints of everyday life. The analysis of the diagonal also enables us to detect more complex activity interactions, such as men watching TV while women are doing chores. Log-linear models applied to cross-tabulations of couples' activities could be an interesting avenue for future research on partners' shared activities (Hout et al. 1987;Smith et al. 1998).
Nonetheless, the precise meaning of the diagonal identified here is open to interpretation. The diagonal can be interpreted as a measure of equality. The division of labor between partners is generally studied from the broader point of view of paid versus unpaid work, market versus non-market work. Studies are generally concerned about the overall distribution of activities within couples, and resource-based theories have been generally applied to the average total time spent in certain activities (Bittman et al. 2003;Gupta 2006;Sullivan 2011). Looking at activities during shared time gives another look at the power dynamic taking place inside the household. This paper provided evidence that even during reported shared time, partners did not divide their activities equally. Men overall still enjoyed more free time than women during shared time. However, I failed to show clear evidence that resources influenced the way partners divided their activity during time together. I did not see that more equal couples (less specialized) in terms of employment status were more prone to do the same activity during shared time compared to more specialized couples. Future research could examine other measures, such as income (relative and absolute) to see if the jointness of activities during shared time is influenced by partners' resources.
Some limitations of this study should be noted. An important limitation is that it is impossible to disentangle the difference in perception of a situation from objective situational differences. For instance, when partners are both caring for a child together, women could perceive and report this situation as shared childcare or shared domestic chores, while men might report it as shared leisure. This difference in perception could be rooted in the fact that mothers bear a more general responsibility concerning the supervision of children and their childcare time is therefore more stressful and less associated with leisure (Milkie and Peltola 1999). However, in other instances, like when women report doing domestic work while men are watching TV, we can reasonably believe that an objective difference in activities is occurring. In this situation, we can assume that the man is physically lying on the couch watching TV while the woman is physically up and doing domestic chores. This is the principal difficulty with the study of social phenomenon like time together. Social actors constantly interpret, reinterpret and misinterpret everyday situations. They also negotiate the meaning of situations and social interactions, contemporaneously and in retrospect. This is particularly true for intimate relationships, which Berger and Kellner describe as "nomos-building" (1964: 220). Not only do couples create their own "couple subculture", but the set of norms and meanings they create is not necessarily agreed upon between partners themselves at all times. In that sense, time together is very difficult to study because of the strong phenomenological component at the core of interpersonal relationships (Berger and Luckmann 1991). So, one can wonder what time use diaries actually captures when recording time together. How do we get to the actual situation, is an important question however unlikely to receive a satisfactory answer because there is no "objective" benchmark regarding what is going on during a particular interaction between family members. Perhaps with the introduction of so-called "camera diaries" in time use research, scientists will be able to disentangle the subjective/interpretative component from the actual/objective component of a situation (Gershuny et al. 2017).
Can social desirability play a part in the reporting of time together? It is unlikely that social desirability bias is playing a role here, even though desirability does seem to play a role when men report their domestic work in stylized questionnaires (Kamo 2000). For instance, it could be that men who do a lot of domestic work would either underreport or even not perceive their involvement in unpaid work as fitting in the "domestic work" category, in order to preserve their self-image (Hochschild 2012;West and Zimmerman 1987). This possibility seems unlikely in this case because time use diaries are considered to be much freer of these sorts of biases than stylized survey questions (Kan 2008). In order to better understand patterns of activity reporting, two possible solutions could be considered using time use diaries. The first one would be to use the location information.
Unfortunately, the location information in the UKTUS 2014-2015 is not precise enough to distinguish the different rooms in a house where activities are taking place, for instance. Moreover, location reports did not differ much between partners, so there is little variation to exploit. The second solution would be to use the enjoyment information provided by the diaries. Future research could address the discrepancy in enjoyment when partners are together and correlate it with the reported activities.
In this paper I looked at a particular social context: the UK in 2014-2015. However, comparing the diagonal of activities over time, both when partners are together but also when they are apart, will be greatly informative for future understanding of the transformation of gender relationships at home. Comparing the diagonal of activities between partners across different countries (where couple time use diaries exist) will also be of great interest in the study of gender and family relationships.