Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=

# Understanding the Congestion Child Penalty: Can Remote Work Attenuate It?

Authors: Ilaria D'Angelis, Keren Horn  
Discussant: None  
Video: https://www.youtube.com/watch?v=1kb3a99sA-E&t=15007s


## Talk (04:10:07 – 04:25:04)

[04:10:07] [applause] Our last presentation of the day will be by Aria, who is going to tell us about the congestion child penalty.  
[04:10:20] I really have to lower the mic. [laughter] Um, all right. Okay. Um, hello everyone.  
[04:10:30] Thanks for being here and thanks to the organizers for including us in the program. Uh I'm Ariia. This paper is a joint work with my colleague Karen Horn who is with me at Humas Boston and is also here uh in the audience. In the  
[04:10:44] paper we study the role of remote work in attenuating the negative effect of congestion of mothers labor force participation a phenomenon that for brevity we refer to as the congestion child penalty.  
[04:10:58] So it is well known that after its secular increase in you know during the 20th century the labor force participation of women really plateaued at least since the 1990s and as this happened in spite of continuous improvements in women's productivity  
[04:11:13] related characteristics. The recent literature increasingly shifted its focus towards an analysis of the possibly persistent constraints which might be affecting women's labor supply.  
[04:11:26] Part of this literature focused in particular on mobility constraints and on how they can exacerbate work family tradeoffs. It's been found in fact that women have little uh willingness or ability to commute compared to men. Upon  
[04:11:39] having children, women decrease their commute and they also decrease their travel. As we just learned, commuting times also have been found to cause declines in women's labor force participation, especially when children  
[04:11:52] are present. And in fact because of long commutes um so long commutes are part of the reasons why in uh some countries the labor force participation of mothers is lower in big cities than in smaller towns.  
[04:12:05] So in our paper we build upon this literature and we add a couple of things. So first we document that the effect of congestion on women's labor force participation rate is highly heterogeneous and in fact totally  
[04:12:19] concentrated of women with low on women with lower levels of education. For mothers without a college degree we do find that a 10% increase in the CD level average commuting time which is approximately 2.7 minutes uh can lead to  
[04:12:32] up to 1.8 percentage point decline in the probability that these women are in the workforce. While for college graduate mothers, we find no uh statistically or economically significant effect either at the intensive or at the extensive margin.  
[04:12:46] And second, we do provide some descriptive evidence that differences across these two groups in access to remote work can contribute to explain these results. Uh, in fact, consistent with what others have found, we also  
[04:12:59] document that remote work tends to be mostly available to college graduate women and to their partners, most of whom are themselves college graduate.  
[04:13:07] For men in particular, we note uh using time use data that remote work is associated with an increase in the amount of time that these men spend in caregiving and housework within their families. And remote work also explains  
[04:13:21] why college graduate men do spend more time in these activities compared to um men without a college degree. So it's possible that access to remote work for these individuals somehow contributes to slack mobility constraints for their  
[04:13:34] female partners. For college graduate women, we also find that they tend to substitute remote for inerson work disproportionately when they live in areas where commuting times are especially long and hence mobility constraints might become binding.  
[04:13:48] Um so in this sense we aim to contribute to the literature both sorry to both the literature on commuting times and women's labor supply and to the growing literature studying the really multifaceted effects of flexible work arrangements on women socioeconomic  
[04:14:02] outcomes. Um to give you a sense of what we're talking about in this slide I'm showing the trends between 2005 and 2019 in the labor force participation rate of women who are between 18 and 54 years  
[04:14:17] old uh in uh metropolitan areas characterized by low commuting times. So appro at most um sorry at least two minutes below the the city level average  
[04:14:29] and in cities characterized by high commuting times. So as we can see obviously the probability that mothers are in the workforce is lower than the probability that women without kids are in the workforce but the gap between these two groups of women is  
[04:14:44] substantially higher in metropolitan areas characterized by high levels of congestion.  
[04:14:50] So to quantify the relationship between congestion and mothers labor force participation rate we use data from the ACS the American community survey and um the American time use survey between 2005 and 2019. Again we focus on women  
[04:15:05] who are 18 to 54 years old who are not enrolled in school or college and who cohabit with male partners. Uh we estimate a simple linear probability model in which the outcome variable is a dummy that indicates whether a mother  
[04:15:19] participates in the workforce. And the main explanatory variable of interest is uh the log of the average commuting time calculated among employed men who commute by car in the metropolitan area in which the woman resides in a certain  
[04:15:32] year. to uh try and attenuate to the extent possible uh biases due to omitted variables and selection. We control for a large number of characteristics of women of their partners and of the  
[04:15:45] metropolitan areas um in addition to region and ear fixed effects. And in an attempt to capture uh possibly longterm uh patterns in selection of different types of women in different types of  
[04:15:58] cities, we also control for the characteristics of women who lived in a certain MSA in 1990.  
[04:16:05] to uh possibly attenuate biases due to measurement error and reverse causality.  
[04:16:10] We instrument the commuting time using um the length of railways as of 1898 and the length and the planned length of highways as of 1947 from Duran and Turner's 2010 paper.  
[04:16:24] The uh results for uh mothers without a college degree show that a 10% increase in the average commuting time by perms MSA is associated with a decline in the probability that these women are in the  
[04:16:37] labor force by 0.4 to 1.8 percentage points depending on the estimation method that we use. When we perform the same estimation for women with a for mothers with a college degree, we find  
[04:16:50] that the there is no relationship between congestion and the probability that they are in the workforce. We run a series of robustness checks to reduce concerns that these results are driven  
[04:17:03] by selection biases. Um and for college graduate women, we also exclude that there is an effect possibly at the intensive margin. In one robustness check that I want to mention, we also  
[04:17:16] include controls for childare prices also in an IV setting to exclude the possibility that these results are driven by the possibly asymmetric impact of lack of uh affordable childare services.  
[04:17:29] So as I mentioned at the beginning, the main contribution of our work is really to uh show that these results can be related to uh remote work. So why so?  
[04:17:37] because obviously commuting times represent a time cost that workers are not necessarily compensated for and as such they decrease the marginal value of every work hour at the margin driving some workers out of the labor force. Now the workers who are mostly affected by  
[04:17:52] this are those whose marginal utility of working is very um low possibly because their housework and caregiving time is very valuable or difficult to substitute for and also workers are only affected  
[04:18:05] by this trade-off between commuting and being in the labor force if they cannot avoid commuting at all. So con so remote work can attenuate the negative effects of commuting times through two channels.  
[04:18:18] First, if uh male partners with access to remote work use this flexibility in their jobs to provide housework and care to their families, they might somehow contribute to alternate attenuated partners mobility constraints. And also,  
[04:18:32] if women themselves obviously have access to remote work, they might potentially just escape commuting entirely. So, using mostly time use data, we provide descriptive evidence that both channels might rationalize the findings that we have.  
[04:18:47] Al so here I'm showing uh differences between partners of college and non-ol graduate women in different measures of remote work use and access that we use construct that we construct using ACS  
[04:19:00] and time use data. So the first two columns uh show that um partners of college 43% of partners of college graduate women are employed in occupations with a high incidence of  
[04:19:12] remote work while only 22% of partners of women without a college degree are in these occupations. Um also 30% of partners of women with a college degree declare to have worked remotely at least half an hour in the previous day based  
[04:19:25] on time use data. they have worked remotely around 21% of their work time.  
[04:19:31] 15% of them has worked fully remotely in the previous workday. Now these results are relevant because in the paper as I mentioned we also show that men who do work remotely are uh providing typically more caregiving and housework to their  
[04:19:46] families. Remote work explains at least 25% of the gap between college and college graduate men in the amount of um time devoted to care and housework.  
[04:19:56] So based on these facts, it is then not surprising that when we estimate the congestion child penalty for non-ol graduate women splitting the sample based on whether their partners have  
[04:20:09] access to remote work or not, we find that most of the effect is really concentrated on women whose partners do not have a chance to work remotely. Now clearly there is selection, right, of people that we want to be in couples with. But at the same time something we  
[04:20:23] want to stress is that to the very least the men who are partnered to these women might have some occupations to select into right some flexibility to choose when we uh one thing that is also interesting that we find is that the  
[04:20:37] availability of remote work is really the unique um characteristic of partners occupations that drive some level of heterogeneity this congestion child penalty for college graduate women instead we still find that the relation  
[04:20:51] between congestion gender labor supply is basically zero.  
[04:20:55] So uh why so well again here I'm showing a differences between college and non-ol women in access and use of remote work and similarly to what we observe for their male partners. Also, college  
[04:21:10] graduate women are substantially more likely to be in occupations that uh provide remote work and also to use remote work at the intensive margin.  
[04:21:21] Interestingly, however, the gaps between college and non college graduate women in access and use of remote work are not homogeneous across geographies, but actually are much larger in areas where  
[04:21:35] commuting times are especially long. So here in this figure I am showing the uh gap between college and non college graduate women in the probability of being employed in a high remote work  
[04:21:49] occupation. Sorry. Uh in low congestion metropolitan areas and in high congestion metropolitan areas. So as we can see here in high congestion metropolitan areas women tend to be more likely employed in um flexible  
[04:22:04] occupations. These uh differences in the occupational distribution of women across occupations reflect into differences in the use of remote work at the extensive margin. In  
[04:22:16] fact, we see that uh um in highly congested metropolitan areas, the gap between college and non college women in the probability of having worked remotely at least one half an hour in the previous day is higher than  
[04:22:30] elsewhere. And more or less the difference between the two figures really reflect the differences in the occupational distribution.  
[04:22:37] Now these differences are also larger at the intensive margin. So here when we look at the gap between college and non college women in the share of work hours worked remotely in the previous day, we see that the two are very similar in low  
[04:22:52] congestion metropolitan areas while the gap is around 7.5 percentage points in metropolitan areas characterized by long commuting times. So these results suggest that women who do have the possibility to do so increasingly rely  
[04:23:05] on remote work when mobility when commuting times might make mobility constraints particularly binding. This is in fact also reflected in the last uh finding here where we see that the probability of working remotely fully or  
[04:23:20] having remote having worked remotely fully in the previous workday is identical between college and non college women in low congestion metropolitan areas where there is a 5 percentage point gap in high congestion metropolitan areas only. Time use data  
[04:23:34] also show I mean we we show in the paper using time use data that college graduate women do seem to use this flexibility in a high congestion metropolitan areas to provide care to their house uh member household members  
[04:23:48] as a secondary activity. So to conclude in our paper we do provide some evidence that congestion and long commutes worsen mother's mobility constraints for college graduate women. However, access  
[04:24:01] uh to remote work seems to attenuate the negative effects of congestion as these women are somehow able to substitute remote for inerson work when commuting time uh might make you know their their  
[04:24:13] trade-offs particularly problematic. For non-ol graduate women, however, were typically represented in inflexible jobs that often require in-person interaction. congestion and long commutes really represent a time cost  
[04:24:28] that might drive some of these women out of the workforce. So this clearly leads to other questions some of which we are addressing in separate ongoing work regarding for example the possibility  
[04:24:40] that policies that is commuting maybe through improving you know public transit can be effective at fostering the labor supply of women who would uh not have many options to work remotely.  
[04:24:53] So this is all I have. So thank you. [applause] >> Okay, great. Thank you.