*2.4. Control Variables*

In addition, based on previous studies on the division of housework and childcare (e.g., [36,38,77]), we include the following control variables. We include actual working hours since the COVID-19 lockdown. This is an ordinal variable yet treated as a continuous variable where 1 refers to "less than 15 h", 2 "15–23 h", 3 "24–29 h", 4 "30–34 h", 5 "35–39 h", 6 "40–44 h", 7 "45–49 h", 8 "50 h or more" and 0 "not working". The relative income between the couples since the COVID-19 lockdown is included coded as 1 being "female earns all", 2 "female earns much more", 3 "female earns somewhat more", 4 "both earn about the same", 5 "male earns somewhat more", 6 "male earns much more" and 7 "male earns all". Education of both partners are included as dummy variables "high education" indicating the person's education level being tertiary or above—namely, Undergraduate degree (or equivalent) or above. Here, the reference category is those with upper secondary level of education or below. Gender role attitude is constructed as a mean of six variables based on previous studies (e.g., [78]), including questions such as "preschool children suffer when mothers work", "man's job is to earn money while women's job is to look after home and family", "both husband and wife should contribute to the household income", "if a husband and a wife both work full-time, they should share household tasks equally", "women and men should share equally in the raising of their children", and "men are just as suited to take care of children as women." Each variable is constructed on a 1–5 scale, with higher numbers indicating more egalitarian attitudes towards gender roles. The Cronbach alpha is 0.69 entailing internal consistency. Number of children refers to the total number of children that lives with the respondent (and their partner) that is under 18. We also include the age of the youngest child in age categories—namely having at least one child under the age of 5, or at least one child between the ages 5–11. Here, the reference group are those whose youngest child is over 11 (under 18). We include the ethnicity of the respondent—namely anyone who identify as non-white as defined as "ethnic minority" with white as the reference group. Finally, we include information on whether there are other adult(s) living in the household, with the reference group being not having any other adults other than the couple living in the household.

#### *2.5. Models*

We first examine how working from home is associated with the division of housework and childcare during the COVID-19 lockdown, having controlled for several factors. We explore the six categories separately as in our theoretical framework we argue that the division of routine versus non-routine tasks are substantively different not only in terms of how it is usually divided among couples, but also its impact on gender equality outcomessuch as women's employment capacity. Through this we are able to empirically examine the hypotheses 1 and 2 presented in the previous section. We focus specifically on the impact of fathers' working from home in changing the dynamics in the division of household labour between the partners. Next, we examine the association between working from home and increased time spent on housework and childcare among fathers during the lockdown compared to before. This set of models are used to examine hypothesis 3 of this study. This is examined in two ways—by looking at how fathers felt about their own time increase, and how mothers felt about fathers' time.

Multivariate regressions are used in this paper, as it allows us to examine the impact of working from home on the division of household labour, and on father's increase in the involvement in housework/childcare whilst controlling for a number of different factors that may influence these associations. Linear regression analyses are used when examining the division of housework carried out by couples (H1&2), as the dependent variable used is an ordinal variable which we use in a linear continuous term for simplicity. Logistic regression analyses are carried out when examining father's (additional) involvement in housework and childcare during the pandemic compared to pre-pandemic times (H3), as our dependent variable is dichotomous. We used STATA 15.1 to run all models.
