Communications of the ACM

ity of unfulfilled expectation may also be high, negatively influencing JS. In comparison, the expectation is lower on weekend days, indirectly contributing to a higher level of JS.

JS reaches its highest level on Fridays, as in Figure 2a, a finding that could be attributed to actual fulfillment of weekly work expectations. At the other end, JS is at its lowest on Tuesdays, followed by an increasing trend the rest of the week. One possible explanation is work stress peaks on Tuesdays when the weekend-to-week-day transition is over and work expectations begin to mount.

FS peaks on Wednesdays, as in Figure 2b, a finding that could be attributed to people’s state of work-family balance, as manifested in both low SC and relatively low TC, as in Figure 1. There is a sharp drop in FS from Sundays through Tuesdays, as expected. Surprisingly, FS barely makes it into the positive range on Saturdays. One possible explanation is that employed women shift child care and housework to weekend days. 9 Given individuals’ high expectation of family commitments on weekends, it is difficult for them to feel highly satisfied, particularly as weekends begin to unfold. In comparison, FS is relatively more stable than JS, as reflected in a smaller variance across different days, as in Figure 2a and Figure 2b.

To understand the association between WFC (TC and SC) and satisfaction (JS and FS) we performed pair-wise correlation analyses between the two sets of variables. The results reported in the accompanying table reflect that all correlation coefficients are negative. In addition, the correlations between SC and JS/FS, but not between TC and JS/FS, are significant (p<.001).

The findings on the relationship between TC and satisfaction challenge current mainstream thinking that WFC has a negative influence on spending in the respective domain.

TC arose when an individual allocated above-average time for work and simultaneously below-average time for family and vice versa. Accordingly, we defined TC as the average difference between the time allocation for work and for family.

SC. Strain can be deduced from psychological and physical dimensions, 12 though it is difficult, if not impossible, to measure physical strain directly from social-media data. We thus measured SC from the psychological aspect of participant data, specifically through negative mood (such as anxiety, anger, and sadness). Negative mood in one domain is associated with pessimism and rumination, causing individuals to neglect requirements in the other domain. 18 SC arose when an individual simultaneously experienced above-average negative mood in both work and family. Accordingly, we defined SC as the average negative mood across work and family.

We measured satisfaction in two domains—work and family—we labeled as job satisfaction (JS) and family satisfaction (FS), respectively. Satisfaction can be explained as a causal sequence linking mood to performance and reward in a domain. 18 Positive mood can facilitate role performance by enhancing cognitive functioning, increasing task activity and persistence, and promoting positive interactions with others. Meanwhile, intrinsic and extrinsic rewards earned through role performance can thus enhance positive mood. 18 We viewed positive mood as a proxy measure of satisfaction.

JS (or FS). We defined satisfaction with work (or family) as the proportion of work- or family-related tweets expressed in positive moods.

Results and Discussion

To understand the dynamics of WFC, we first analyzed the trends of time-based WFC (TC) and strain-based WFC (SC) at different levels of time granularity after removing the data associated with American national holidays (such as Thanksgiving, Christmas, and New Year’s). The results show the trends of WFC at levels of month and week are both relatively stable. Nevertheless, WFC fluctuates by day of the week (see Figure 1).

Figure 1a shows TC is heightened during weekdays relative to weekends. This observation confirms previous findings on work-family time allocation that both men and women in the U. S. spend more time in domestic work and caring for children on a weekend day than on a weekday. 10 In addition, TC shows an increasing trend from Sundays to Tuesdays and the opposite trend from Thursdays to Saturdays, peaking on Tuesdays and dipping on Saturdays. There is thus a sharp increase in TC when transitioning from weekends to weekdays and a sudden drop during the opposite transition. Note TC stays in the negative range, indicating family-to-work spillover events are more prevalent than work-to-family spillover.

Figure 1b shows SC is much greater on weekdays than weekends, except for Wednesdays. If the trend profile of TC resembles a normal distribution, then the trend profile of SC simulates a bimodal distribution, with Mondays and Thursdays being two peaks. Unlike TC, which peaks on Tuesdays, SC reaches a peak immediately after weekdays begin. The abrupt drop in SC on Wednesdays could be explained by people reaching a state of work-family balance by the middle of the week. A sharply elevated SC on Thursdays may be attributed to the pressure of trying to complete scheduled weekly tasks, onset of work exhaustion, or preparation for upcoming family commitments.

Overall, the results show both TC and SC vary markedly with the day of the week, an observation that helps explain the inconsistent findings about the form and intensity of WFC in the literature. They also imply a major limitation of survey-based methods that have dominated traditional WFC research. Depending on the day a survey is administered, participants’ responses can vary greatly.

Figure 2 shows the trends for job satisfaction (JS) and family satisfaction (FS), including that individuals’ JS and FS are both higher on weekends than on weekdays. People feel satisfied when their criteria are met, and satisfaction can be viewed as a degree of realized expectation. On weekdays, the expectation of work performance is high, and, consequently, the possibil-