The construct of ideology is often studied using the unidimensional continuum of liberalism versus conservatism (e.g., Nail et al., 2009). Liberals are likely to favor equality (receiving equal outcomes regardless of contributions), whereas conservatives tend to value equity (receiving outcomes proportional to contributions; Schlenker et al., 2012). Concerning attitudes toward societal changes, while liberals tend to approve progressive changes (accepting recognized social trends), conservatives tend to approve reactionary changes (reversal of social trends and return to tradition; Proch et al., 2019). The former is positively related to openness and curiosity, while the latter is related to preserving conventions and order (Carney et al., 2008). Past research has shown liberalism-conservatism can influence human decision-making. For instance, liberals are more supportive for organ donation (Chan, 2019) and more willing to take action to address the issue of climate change (McCright et al., 2016) than conservatives. Furthermore, liberalism-conservatism can vary to some extent under certain conditions. For instance, a meta-analysis showed that threatening situations predicted political conservatism as people became resistant to change and tended to justify inequalities in order to manage uncertainty and defend against outside threat (Jost et al., 2003). It was also found that experimentally induced threats led the liberals to hold more similar views as the conservatives regarding topics such as homosexuality and abortion, providing support that liberalism-conservatism was changeable across social situations (Nail et al., 2009). Similarly, Bryan et al. (2009) found that students primed to think of their success as a result of help from others (liberalism-oriented) showed more support for liberalism, suggesting that liberalism-conservatism could be influenced by the comparative salience of different schematic activations under different circumstances. With the previous findings, it would be worthwhile to investigate whether liberalism-conservatism changes in response to different life situations across the 7 days of the week. Below we present two theoretical accounts regarding how this may take place and propose the respective hypotheses.

First, the DOW effect could influence liberalism-conservatism through influencing affective states. According to the whole trait theory (Fleeson and Jayawickreme, 2015), behaviors are not only influenced by relatively stable traits, but also the momentary states and situational factors. Wilson et al. (2017) found that positive state affect predicted state openness, which is a strong predictor of liberalism (Carney et al., 2008). Hence, the DOW effect on liberalism-conservatism can be reasoned according to the affective state changes during the week. Research has showed that people’s affect fluctuates across the 7 days of the week. Specifically, people experience more negative affect from Mondays to Thursdays (Taylor, 2006; Stone et al., 2012), especially on Mondays (Ellis et al., 2015). On the contrary, Fridays and the weekend days were strongly associated with positive affect (Ryan et al., 2010). These patterns have also been supported by neuropsychological studies. For example, Schlotz et al. (2004) found that chronically stressed individuals had higher levels of stress, as indicated by cortisol awakening responses, during the weekdays but not weekends. According to the different affective states people experience across different days of the week, the following hypotheses regarding the DOW effect on liberalism-conservatism are proposed. First, liberalism would peak near weekend days because of more positive affective states. Second, liberalism would drop to the lowest level on Mondays due to negative affective states. Third, liberalism would increase steadily in the following weekdays until it peaks again on the weekends.

Second, the fluctuations in liberalism-conservatism may be explained by changes in cognitive states across different days of the week (Buchel and Dolan, 2000; LeDoux, 2000; Charpentier et al., 2016). According to the occupational stress literature, employees experience depletion of physical and cognitive resources during work days and recovery of energy during holidays (Weigelt et al., 2021). Weigelt et al. (2021) found that employees’ energy increased during weekends and decreased continuously across weekdays. In recent decades researchers have elaborated the cognitive-affective interactions model, which posits that humans’ decision-making is the equilibrium of the higher-order cognition and affective states. The higher-order cognition is also highly related to ideological thinking. Several studies have reported that openness to experience, a personality trait highly related to liberalism (Alford and Hibbing, 2007; Carney et al., 2008; Mondak and Halperin, 2008). Cognitive functioning, including cognitive flexibility and openness to novelty (Soubelet and Salthouse, 2010), has been found positively associated with liberalism and negatively related to conservativism (Jost et al., 2003; Carney et al., 2008). Therefore, the fluctuating cognitive state during the week offers an alternative explanation for the fluctuations in liberalism-conservatism from Mondays to Sundays. The effort-recovery theory (Meijman and Mulder, 1998) postulates that people need off-job time to recover and replenish their mental resources, which would produce short-term benefits at the beginning of the working week. In fact, a longitudinal study found that employees reported better task performance on Monday (right after the weekend) than on Tuesdays (Fritz and Sonnentag, 2005). Hence, as people generally rest on weekends, we hypothesized that liberalism would peak on the weekend days due to better cognitive resources and state. Starting on Mondays, liberalism would decrease gradually throughout the weekdays when cognitive resources continue to deplete. Figure 1 illustrates the predicted weekly patterns of liberalism-conservatism according to the two theoretical accounts and their respective hypotheses.

This study adopted the 2014 Chinese Political Compass (CPC) dataset, a well-established and large online survey available on the Internet.¹ There were 171,830 respondents (male: 110,110, female: 61,545, unreported: 175), with an average age of 24.11 (SD = 7.43). Since this survey was set up at Peking University in 2007, it had gained increasing attention and been completed by a very large sample of internet users voluntarily. This study only uses the 2014 data, which is available to the public. As pointed out by Pan and Xu (2018), there are both pros and cons of this opt-in sample. On one hand, it provided valuable data on questions that may not appear in approved social surveys. Besides, voluntary participation is often associated with higher motivation to provide accurate answers (Chang and Krosnick, 2009). On the other, the sample tended to bias towards young, male, well-educated participants living in more developed areas of China. The dataset recorded the exact time and date of the responses, which were converted to the time zone of Beijing, China. The days of the week were then coded based on the converted dates.

The CPC survey consists of 50 items measuring three domains: political (e.g., “People should not have universal suffrage if they have not been educated about democracy”), economic (e.g., “If the price of pork is too high, the government should intervene”) and social (e.g., “Two adults should be free to engage in voluntary sexual behavior regardless of their marital status.”). Each item was rated on a 4-point scale. Pan and Xu (2018) reported that the 50-item CPC represented a unidimensional construct, which is liberalism versus conservatism. In their analysis, five items (i.e., Q12, Q3, Q2, Q6, and Q17) were found to best represent liberalism, while five other items (i.e., Q9, Q46, Q20, Q48, and Q33) were found to best represent conservatism. For parsimony, a liberalism-conservatism score was constructed by deducting the mean of the five conservatism questions from the mean of the five liberalism questions. As a result, higher scores indicated higher levels of liberalism (or lower levels of conservatism). This score was not only highly correlated (r = 0.919) with the extracted factor score based on principle component analysis, but offered more convenient interpretation and better comparability due to the retention of the original rating scale.

A one-way analysis of covariance (ANCOVA) was employed to inspect the differences in liberalism-conservativism across the 7 days of the week. Specifically, the dependent variable was the liberalism-conservatism score and the independent variable was the categorical variable indicating the days of a week. Gender, age, income and education level, which were previously found to be associated with liberalism-conservatism (Box-Steffensmeier et al., 2004; Zeng, 2014; Pan and Xu, 2018), were controlled in the analysis. Post-hoc analyses with Sidak adjustment were performed to examine the pairwise differences among the 7 days of the week.

The DOW effect on liberalism-conservatism can offer insights into human behaviors in different areas. For example, an opinion poll conducted on different days of a week may generate different results (i.e., more conservative during midweeks than weekends) and lead to different decisions in policymaking. Besides, the DOW effect on liberalism-conservatism may contribute to our understanding of decision-making processes such as investment. It was found that liberal individuals would invest more on innovative business, compared with conservatives (Gromet et al., 2013). Moreover, educational programs can be designed in a way that is more consistent with the changes in liberalism. Teaching and learning that focuses on tradition (e.g., culture, history, moral education, etc.) may be scheduled in midweeks when people tend to be more conservative, while activities that demand creativity and openness may be scheduled on other days when liberalism is higher (Dollinger, 2007). Furthermore, the current findings of the DOW effect may offer implications to the 4-Day workweek as aforementioned. Based on the current findings, liberalism tends to trough at the midweek, potentially due to depletion in cognitive resources and decreased positive affect. Hence, it could be possible that, even though cognitive resources start to deplete from the first day of work, a shorter working week may produce more positive affect earlier in the week due to the prospect of non-working days approaching. This may in turn increase liberalism level, potentially benefiting jobs that require creativity and innovativeness. In the long run, it would also be interesting to track the liberalism-conservatism level of those who work 4 days a week to examine if such changes in weekly schedule would result in a different DOW effect.

Despite the significant findings, there are several limitations to be noted. First, the current study utilized cross-sectional self-reported data. This design does not allow us to track the intra-individual changes during the week. Future research may adopt longitudinal design to examine the DOW effect within individuals. Second, the sample predominately consisted of younger populations who had access to the Internet. It would be worthwhile to test the DOW effect among other groups such as the elderly to explore if there are differences in patterns. Besides, sampling issues, such as variability in internet skills and fraudulent survey submissions are common concerns for online surveys (Wright, 2005; Hulland and Miller, 2018), though these effects should be minimal in this study as Pan and Xu (2018) found that results based on this sample were robust and had good generalizability when being validated against other evidence. Third, the online survey did not include direct measures of cognitive and affective states. Future studies may examine if the DOW effect on liberalism-conservatism can be explained by changes in affective and cognitive states. Fourth, the effect sizes of the DOW effect were generally small. Therefore, caution is needed when using the DOW effect to predict the level of liberalism, although small effects are common in big data research (Matz et al., 2017) and useful for a cumulative science (Götz et al., 2021). Fifth, this study did not exclude those who might not fit the general weekly schedules (e.g., people working on weekends) due to non-availability of information. Future research should separate different samples if possible. Sixth, as participants who are more conservative may be less willing to respond to sensitive questions, the current research data might be prone to self-selection bias and/or social desirability bias. Seventh, the survey did not include direct measures of cognitive or affective states. Therefore, further studies are required to examine whether the DOW effect on liberalism-conservatives is due to changes in cognitive or affective states as proposed by the cognitive and affective states hypotheses.