The relationship between parenting engagement and academic performance

Parenting labor by gender

Among our respondents, women are more likely than men to serve as the primary caregiver for their children (30.6 vs. 3.9%) (Fig. 1A). Inversely, more than a third of men (38.9%) indicated that they served in a secondary role in parenting (i.e., satellite), compared to only 17.4% of women. This establishes an important baseline for the study of scientific parenting: that is, women scholars are disproportionately likely to be taking a lead role in caregiving. The most common model, however, is one of shared parenting: the majority of men (57.1%) and women (52.0%) indicated that parenting roles were shared equally with their partner (i.e., dual).

Figure 1
figure 1

(A) Parenting type by gender and household composition. (B) Respondents reporting themselves as the primary caregiver at different times of the day, by gender and parenting type. (C) Respondents reporting being a primary caregiver for the parenting-related activities, by gender and parenting type. Respondents are considered “primary caregiver” if they reported “Mostly me” or “Almost always me” in taking care of these activities. The asterisks denote the FDR-adjusted p-values from the two-sample tests of proportion between men and women: + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001.

These self-reported roles were investigated to understand how they manifest themselves in terms of both time and task engagement—that is, the times during which parents were engaging in parenting and the types of tasks with which they were disproportionately associated. Lead parents of both genders indicated high percentages of engagement across time compared to other roles, suggesting that the 3.9% of men serving in this role are strongly engaged in parenting. However, men in dual and satellite parenting roles were much less likely than women in the same roles to report primary caregiving across times. This demonstrates that there is a higher burden of labor for women to classify themselves as dual and satellite parents than men (Fig. 1B). Men reinforced these disproportionate labor expectations, even in shared parenting relationships: “Although I try to be active in child care and share responsibilities equally, my wife still takes care of more child care tasks than I do”. (M, Dual, US).

The time results were confirmed by an analysis of tasks. In nearly every category—particularly for dual and satellite parents—women were more likely to be the parent engaged in the caregiving tasks (Fig. 1C, Table S1). There were fewer differences in how men and women operationalized lead parenting, with lead parents of both genders significantly engaged in childcare tasks. The only task where lead fathers demonstrated significant differences were in dropping off children at school/nursery (79.3 vs. 66.4%, \({\chi }^{2}\)=12.12, p = 0.001) and coaching sports (40.8 vs. 17.7%, \({\chi }^{2}\)=53.66, p < 0.001). Much stronger gender differences were observed in dual parenting. Men in dual roles were more likely to drop off children at school/nursery (40.0 vs. 29.3%, \({\chi }^{2}\)=70.98, p < 0.001), coach sports (30.5 vs. 5.6%, \({\chi }^{2}\)=625.04, p < 0.001), and do school/nursey pick-up (28.7 vs. 27.0%, \({\chi }^{2}\)=2, p = 0.168), though the latter was not significant. Women in dual roles were significantly more likely to be primarily responsible for all other caregiving duties. The same was true for satellite roles. The only task with which men were significantly more likely to be associated was coaching sports (25.8 vs. 5.1%, \({\chi }^{2}\)=188.44, p < 0.001).

The time and task analyses reinforce each other: when they self-identified as dual or satellite parents, women are disproportionately engaged in parenting activities. Furthermore, there is little difference between women’s labor performance in dual and satellite roles (Figure S1). These asymmetries between labor and credit show that, even in the perception of equality between parents, women carry a higher burden of labor.

Qualitative responses were illuminating in this regard. A woman from Tunisia noted that the survey made her aware that she was the main caregiver. Other respondents supported this, but questioned the exhaustivity of the task list:

This survey was instructive as I didn’t realize how many tasks I take on compared to my partner. I thought it was more equal. Maybe there were tasks not listed. For example, he handles administrative tasks like keeping track of bank accounts, keeping the printer ink filled, and other tasks like basic cleaning within specific tasks (e.g., loading the dishwasher) and yard work. (W, Dual, US)

The incompleteness of the list was also questioned by other respondents, but in the opposite direction, calling into focus the unequal cognitive and emotional labor performed by women:

There is a huge cost to something that you didn’t ask about: “running the house” – it’s not just child care. It’s scheduling someone to come clean and/or cleaning, getting groceries, scheduling sitters, arranging for travel, paying bills, sorting the mail, getting the kids new clothes etc. etc. Often doctor’s offices etc. have to be called during business hours and that takes away work time. It’s much, much more emotional work because I “keep track” of everything. We are traveling for work and need vaccines — I am the one coordinating them. My husband helps, and definitely the physical aspects of taking care of the kids are 50/50 at least, but there’s all this other stuff. (W, Dual, US)

The cognitive and emotional burdens of domestic labor disproportionately born by women have been well-recognized in previous studies38 and were manifest in the exogenous shock of the pandemic2,4,6. Therefore, the inequities observed in the itemization of task and time may only represent a conservative estimate of the actual difference in parenting engagement. However, our work demonstrates that self-report data of shared parenting discounts women’s engagement in parenting and overestimates men’s.

Work arrangements and adaptation in parenting

One limitation of the survey is that it captured individual nodes in dyadic relationships, rather than paired couples. One might expect, for example, different labor roles based on the occupation of the partner. To control for this, we identified the sector of employment of the partner, with a particular focus on situations where both the respondent and their partner were employed in academia. Academic couples arguably experience the same productivity pressures and job responsibilities as each other, creating a natural control for labor expectations. Overall, academic women as dual parents are strongly affected by having an academic partner: in terms of the task analysis, women with non-academic partners are primarily responsible for a larger number of tasks than their counterparts with academic partners—especially regarding transportation and evening care (Fig. 2, Table S2). In contrast, academic men are much less affected by their academic partnership status except for dual parents––those with an academic partner are more likely to do household shopping (23.9 vs. 18.8%, \({\chi }^{2}\)=7.9, p = 0.038) but less likely to do the school run (drop-off: 35.2 vs. 44.8%, \({\chi }^{2}\)=19.37, p = 0.002; pick-up: 22.7% vs. 30.0%, \({\chi }^{2}\)=13.91, p = 0.004), put children to bed (14.7% vs. 19.3%, \({\chi }^{2}\)=7.61, p = 0.042), and read bedtime stories (26.6 vs. 21.8%, \({\chi }^{2}\)=6.34, p = 0.043). For academic men as either lead or satellite parents, we found no statistically significant differences between those with and without an academic partner in parenting engagement (Fig. 2, Table S3).

Figure 2
figure 2

Respondents with an academic employment (n = 8,046) reporting being a primary caregiver for the parenting-related activities by parenting type, gender, and partner employment status (Academic vs. Non-academic). Respondents are considered “primary caregiver” if they reported “Mostly me” or “Almost always me” in taking care of these activities. The asterisks denote the FDR-adjusted p-values from the two-sample tests of proportion between those having an academic partner and their counterparts having a non-academic partner: + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001.

The benefits of a partner who understands the labor burdens of academe was evident in the qualitative responses. For example, parents commented on the perceived flexibility of research careers and these advantages were enhanced when both parents were academics:“…My wife is also an academic which actually helped in sharing duties in a much more understanding way”. (M, Dual, Singapore).

However, whereas an academic career was seen as more flexible and therefore amenable to parenting, there was an assumption that flexibility also implied availability. This was most evident in respondents who were the sole academic in their household and the delicate balance between flexibility and availability was experienced by both women (“Inevitably, we both feel that if a sacrifice must be made, it is my schedule” (W, Dual, US)) and men;

“It is hard to balance academic work and home life – as in many cases your partner does not understand that reading and working on your computer is your job. Thus, you find that you have various tasks (family, children, house, errands) thrown to you by your spouse who works a “regular” job because you are “not busy”. (M, Dual, US)

Women, however, are particularly affected by partner occupation, with significant differences in equity between those with and without academic partners. This suggests that spousal hiring programs may have stronger implications for women faculty in caregiving roles as the equitability of parenting tasks is higher with academic partners.

Effect of parental engagement on research productivity

Although parenting engagement and partner types account for only a small fraction of variations in productivity (after controlling for academic age, number of children, and discipline), certain patterns are revealing. As illustrated in Fig. 3, both men and women suffer a productivity loss when they are single or lead parents. Using dual mothers with a non-academic partner as our reference group, dual fathers (β = 0.05, p = 0.029) and satellite mothers (β = 0.09, p = 0.004) are 5.6% and 8.9% more productive, respectively; single mothers are 15.3% less productive (β =  − 0.17, p = 0.048) (Table S4). The differential effects of parenting engagement involving partnership status for men and women is confirmed in the two-academics subsample, where lead parents are, on average, 11.1% less productive than dual parents (β =  − 0.12, p = 0.012), and the magnitude is roughly the same as the additional effects of being a lead father and being a lead mother with an academic partner in the full sample. This suggests that parenting penalties are felt by both men and women. As one respondent noted:

My scientific productivity has declined since having kids- something had to give. I still work hard to be a good teacher and leader in my department while doing as much parenting as possible so my wife can pursue her career goals as well. The cost of the parenting has largely been a reduction in papers written. (M, Lead, US)

Figure 3
figure 3

Predicted number of papers for men and women, by parenting engagement and partnership status. The dash line refers to the predicted productivity of our reference group: dual mothers with a non-academic partner. Results are averaged over the levels of number of children, academic employment status, doctoral degree, and domain. Intervals are back-transformed from the log scale.

Despite the positive but not significant effect on productivity of having an academic partner for dual mothers (β = 0.05, p = 0.082), we saw an additional 10% decrease in productivity (β =  − 0.11, p = 0.038) for lead mothers with an academic partner. This is tied with the reference point of dual mothers, where we see that dual mothers operate at similar levels of engagement as lead mothers. Perhaps counterintuitively, lead-mothers are as productive as dual mothers when they both have a non-academic partner (β = 0.00, p = 0.862). This is in sharp contrast to lead fathers who suffer an additional 11.6% decrease in productivity (β =  − 0.12, p = 0.11; not significant). This is likely a result of the unequal engagement women demonstrate in these parenting roles. Men seem to be most productive when they are in satellite roles with academic partners. Women, on the other hand, are most productive when they are in a satellite role with a non-academic partner. This may be a result of the perceived flexibility of academic roles intersecting with cultural expectations in parenting. As one woman observed: “You have to be prepared to work twice as hard and accept that.”(W, Dual, UK).

Effects of parental leaves on production

Parental leaves are associated with higher production, but have a point of diminishing return that varies by country (Fig. 4, Table S5). The production advantage is highest at less than one month of leave for the US sample (estimated increase of 26.9%; β = 0.24, p = 0.002) and decreases for every three months of leave after (26.7% increase for longer than one but less than three months [β = 0.24, p < 0.001], and 17.8% for three to six months [β = 0.16, p < 0.001] among US women). The advantage disappears after six months with the US sample and after twelve months in the non-US sample. These cultural differences may be explained by the normative leave lengths and productivity expectations by country. One woman from the US explained how the casual terminology reinforced expectations of production during the limited leave given to US mothers:

The six weeks after giving birth should be termed medical leave for the person who delivered the baby (regardless of if they are parenting the child). That should be treated as such. There are still people who will call it a ‘sabbatical’. (W, Dual, US)

Figure 4
figure 4

Predicted number of papers by parental leave lengths and gender in the (A) US sample and (B) Non-US sample. The dash line refers to the predicted productivity of our reference group: dual mothers with a non-academic partner taking no parental leave. Results are averaged over the levels of parenting type, partnership status, number of children, academic employment status, doctoral degree, and domain. Intervals are back-transformed from the log scale.

While the effect is relatively modest in the non-US sample and non-significant for taking leaves less than one month (β =  − 0.07, p = 0.609), the corresponding increase in number of papers is estimated to be 17.1%, 10.5%, and 10.6% for leaves longer than one but less than three months (β = 0.18, p < 0.003), three to six months (β = 0.10, p = 0.009), and longer than six but less than twelve months (β = 0.10, p = 0.008). None of the interaction terms between gender and parental leave length are statistically significant, suggesting that the effect of parental leave length on production does not differ by gender. However, women are more likely to take leave and to take longer leave, leading to a disproportionate production disadvantage.

Respondents noted that family friendly policies did little to mediate the effects that labor demands on scientific production, including the necessity of managing ongoing growth of measures of academic impact over the lifetime of the career:

“Family friendly policies are all very well but basically just allow you to take time off work; they don’t reduce the amount of work that there is to do or remove deadlines”. (W, Lead, UK).

Effect of parenting engagement on scientific impact

Two indicators of scientific impact are used to examine the relationship between parenting academic capital: number of total citations (TCS) and mean normalized citation scores (MNCS). While the first one is an absolute indicator largely dependent on researchers’ number of papers, the indicates the average impact of their paper compared to other papers published in the same specialty within the same year. Results from the TCS and MNCS models are similar in magnitude and significance, although some notable differences exist (Fig. 5, Table S4). The same differential effects of having an academic partner and being a lead parent for men and women on productivity is also present for impact. Having an academic partner for a dual mother increases impact by 11% (MNCS) (β = 0.10, p = 0.001) and 15.3% (TCS) (β = 0.14, p = 0.004); whereas having an academic partner for a lead mother brings an additional decrease to impact by 17.1% (MNCS) (β =  − 0.19, p = 0.002) and 19.9% (TCS) (β =  − 0.22, p = 0.022). In other words, the effect of being a lead parent for women is moderated by partner’s employment type in both production and impact.

Figure 5
figure 5

Predicted productivity/impact based on parenting type, gender, and partnership status, faceted by (A) gender, (B) partnership status: (1) total number of papers, (2) total number of citations (TCS), (3) mean normalized citation scores (MNCS). The dash line refers to the predicted productivity/impact of our reference group: dual mothers with a non-academic partner. Results are averaged over the levels of number of children, academic employment status, doctoral degree, and domain (except for MNCS). Intervals are back-transformed from the log scale.

The positive effect of being a satellite mother is only significant in the MNCS model (β = 0.08, p = 0.019). The notable discrepancies between the two models relate to the moderating effect of gender on different aspects of the relationship between partnership status, parenting type, and impact. A dual father with an academic partner decreases MNCS by 9.6% (β =  − 0.10, p = 0.031), whereas a lead father faces an additional 29.3% drop in TCS (β =  − 0.35, p = 0.025). The former can be translated into the non-effect of academic partners for men and the latter the differential effects of being a lead parent for men and women. More specifically, being a lead parent has a negative effect on impact (TCS) for men regardless of their partner’s occupation, but for women only if their partner is an academic. This again is confirmed in both MNCS and TCS models fitted with the two-academics sample, where the lead parent effect for women is close to the interaction effect between lead parent and academic partner in the models with the full sample.

Scientific impact is a function of visibility: work is more likely to be cited when authors are visible in the scientific community through collaboration, travel, and other forms of engagement. Therefore, it stands to reason that parenting demands that reduce visibility will translate into lower citation rates. Respondents often discussed how institutional policies were inadequate in compensating for demands of research careers (e.g., the necessity to travel, overnight stay and long, after-hours work). However, partners who are flexible and supportive were essential for engagement in the scientific community:

“Flexible work hours are a blessing, but the travelling required for a successful career (conferences, networking, field work) is a nightmare. I am lucky to have a supportive partner, without whom I would not have been able to pull it off”. (W, Dual, Sweden)