Download data and study materials from OSF
University of Connecticut
Lindsey Trimble O'Connor
California State University Channel Islands
Sample size: 2116
Field period: 11/27/2020-03/23/2021
We find overworkers are advantaged over fulltime workers, and top performers are advantaged over middle performers and low performers; however, there is no difference, however, between middle and low performers in recommendations for hire or salary. Turning to the main effect of gender, gender does not impact likelihood of hire, however, it does impact salary such that women are offered higher salaries than men. Models 3 and 4 assess whether the impact of overwork on organizational rewards differs by gender. With respect to hiring, the evidence is mixed. The interaction term is non-significant (p = 0.134); however, comparison of point estimates across all linear combinations using the mlincom command in Stata reveals a significant difference in the likelihood of being hired between women who work fulltime (0.93) and women who overwork (0.964). This is not the case for men who work fulltime (0.937) and men who overwork (0.964) men. There is no evidence, however, that the impact of overwork on salary differs by gender.Next, we ask if the impact of overwork on likelihood of hire and salary depends on performance. For middle performers, overworkers are significantly more likely to be hired over full-time workers, and the magnitude of the effect of overwork on likelihood of hire is greater for average performing workers compared to the magnitude of the effect for high performing workers. With respect to salary, overworkers are also assigned higher salaries cross all three performance categories. Like above, however, the magnitude of the effect of overwork for average performers is greater than the magnitude of the effect for high performers. Specifically, it jumps from 4.643 to 5.167 for high performers—a difference of 0.524—and it jumps from 3.388 to 4.685 for average workers, a difference of 1.297.
Next, we disentangle the multiplicative effects of overwork, performance, and gender. We find that, for high performing workers, the effect of overwork is negligible for both men or women. In other words, for high performers, performance is the salient piece of information. We refer to this as a good performance bonus. Note, however, that overwork has a significant effect on hiring for women across all three levels of performance. Yet, overwork has virtually no effect on middle or low performing men's likelihood of hire. In other words, overwork significantly increases middle and low performing women's likelihood of getting hired. We refer to this as the overwork bonus – a bonus that is decidedly gendered. To say the same thing differently, all men suffer a poor performance penalty, regardless of work hours. But only full-time women suffer a poor-performance penalty; overworking women do not. Thus, women's under performance can be buffered with long hours, men's cannot.
Conversely, full time workers—regardless of gender—get dinged for poor performance. For overworkers, however, it depends. Namely, men, but not women, get dinged for not being a high performer.We suspect this difference has to do with gendered cultural expectations regarding competence and commitment. For men, performance—not overwork—is what matters in terms of organizational rewards. Perhaps evaluators expect men to perform at high levels, so, when they don't, they are penalized, regardless of work hours. (Indeed, low performing men get dinged for overworking - although not significantly so.) For women, however, moving from high performance to low performance affects fulltime women's likelihood of hire, but not overworking women's likelihood of hire. These women are protected by the gendered overwork bias described above. So, both performance and time matter for women. Why might women alone receive an overwork bonus? One possibility is that people expect women to have more constrains on their time. Thus, when they do overwork, perhaps they are seen as particularly committed. We are in the process of testing these premises with SEM models.