SPEAKER 1: Welcome to the women and public policy program seminar series podcast at the Harvard Kennedy school HANNAH RILEY BOWLES: I’m going to go ahead and get us started So we have maximum time with our speaker today I’m Hannah Riley Bowles from co-director here at WAPPP, where we are dedicated to supporting the research related to promoting gender equity and to the distribution of evidence-based practice relevant research for that aim as well And then the function of this seminar in there is to connect our community of scholars but also [? theaters ?] in practice and students launching their careers, to cutting edge researchers who are doing work on women’s leadership advancement and gender and public policy So I am delighted to see you all here today We’ll also be joined [INAUDIBLE] present, I will podcasting today OK great, so we’ll be also joined virtually at some other time and space by other people who download the seminar which has now been downloaded up in the Tens of thousands of times So I am very excited to introduce our speaker today Professor Jessica Pan is from the national university of Singapore She is also a research fellow at the Institute of Labor Economics in Germany, and she did her undergraduate and doctor work at the University of Chicago, is that right? And so she does [INAUDIBLE] a lot of work related to labor economics and the economics of education and particularly gender dimensions of this And today, we’re going to hear about her work on the Mommy Effects So please join me [APPLAUSE] JESSICA PAN: Thanks very much for having me I’m really happy to be here and tell you a bit more of this work So it’s a paper that’s joint with [? Iliana ?] and [? Ginny, ?] who are both at Princeton and [INAUDIBLE],, who’s at Yale It’s kind of a new book for women And I think that some of well what I’m going to be presenting today in some way also influenced by some of our own personal experiences So both [? Iliana ?] and I both had kids within the time frame of writing this paper All right So just by way of motivation, I’m going to start with– [INAUDIBLE] broadly motivated by a number of facts, I think many of you in this room are very familiar with And the first fact is sort of motivated by is this idea that these in some dimensions particularly to do with human capital accumulation, women had made very rapid progress and, if anything, had actually exceeded men in some of these dimensions Specifically as many of you in this room are well aware, for decades in the US, at least since the 1990s, women are actually even more educated in a way And in many developed countries as well as some developing countries and women like [INAUDIBLE] Beyond that, at least for most of the developed world, women have also started to substantially delayed our growth In large part because we anticipate other careers and what this implies is that they’re also accumulating a lot of your experience as well So even in terms of job experience, the gap has been increasing tremendously Now what is perhaps a bit less unknown and kind of more of an interesting fact is that despite this very large gain in terms of human capital accumulation, on gender gaps in the legal market continue to persist So that in itself is not very surprising but if you actually turn to the data and you start looking at trends in female labor force participation, especially in the US and UK, [INAUDIBLE] I think really quite startling Which is that it has pretty much plateaued Now I think that in itself is actually very surprising especially if you think about the fact that women are now more than ever prepared for very long careers in the legal market So the blue line here, so this is for the US, the UK And the reason I bring in the UK a lot is that you’ll see that quite a bit of this research is going to be based on British data, for reasons I won’t explain in detail later It turns out that we have a lot of the questions that we’re going to need to conduct the kind of analysis that we’d like to do So wherever possible, we’re going to supplement this with US data as long as you get a cross-country comparison Now, there has been a lot of emphasis sort of talking about the decline mentally [INAUDIBLE] over time and you do see in these figures But I think what has gotten a bit less attention But is very, very clear from the data is that start in the 1980s has been this very striking rise in female labor participation But starting in the early 1990s, this in the US is completely flattened out And if anything more recent data seems to suggest that it’s actually declining Now if you focus on women between the ages of 25 and 34, in the prime childbearing years, these figures actually even more pronounced Now in the UK you can see a more gradual increase but it’s very clear, at least from the series as well that

its effects on Which brings us to our motivating question, which is why do women continue to invest in human capital investment So of course, the education, [INAUDIBLE] even though since about 1990, the labor force participation rates haven’t actually increased Now, I’m going to be offering you kind of one way to rationalize this potential puzzle Although, I will be very clear to mention that there are a lot of other ways to rationalize this as well, that doesn’t rely on out of those explanation but this is just one potential explanation that I think hasn’t received as much attention [INAUDIBLE] So put differently, another way of asking this question is, why is it that women’s progress in the labor market appears to have stalled or slowed even though– If anything you would imagine that there are free market investments to predict otherwise And our answer and kind of where the stock’s going to be headed, and that I want to try and provide evidence to substantiate this claim is that one potential way to reconcile this puzzle is that perhaps women are underestimating the employment cost of motherhood And I wanted to be clear about what I mean by employment cost But at the very least what I hope that you’ll walk away from the talk today is that you’ll be convinced by the data that if anything women appear to be systematically underestimating the employment effects of motherhood So we don’t actually see it coming at a point in time in which they’re actually [INAUDIBLE] HANNAH RILEY BOWLES: About the economic status of [INAUDIBLE] women across all college educated JESSICA PAN: Right So actually if you cut this picture up across different socioeconomic classes, you do actually see a broken system picture If anything it’s more pronounced So, yeah, it’s very similar in education as well And I’m going to be showing you a lot of the cuts also why some of these characteristics that we [INAUDIBLE] So essentially, one implication of this underestimation is that we are investing in education thinking It’s actually easier to be a working mom that it turns out to be and you kind of assume that you could have it all at once So what do I mean– somebody uses what employment cost and then threw it around a lot And so I think it’s useful at the beginning of the talk to kind of just fix our minds a little bit on how at least we come into this project thinking about this idea of employment cost And as you’ll see from this slide, the way we choose to define it is kind of really raw and really that kind of on [? purpose, ?] right? So in some sense we think about employment cost, is any form of cost in the form of even monetary costs, psychic cost, opportunity cost that’s required for women to raise their children in a manner that they feel is appropriate while at the same time maintaining work outside the home Now some of these costs are going to be cost that we often contemplate, we’re thinking about working outside the home So these would be things like hiring a nanny, ensuring that the kid has a place of a good preschool I think beyond that, that subset of costs that we’re thinking about are the ones that are perhaps even more material In the sense that very often if you think for mother skill, you think about stigma costs, these are also cause that factor in when a woman decides to work Over and beyond hiring outside help or hiring a nanny And so some of these costs you can imagine may even larger than the direct costs that one might face And these could include things like missing your children while you’re at work, thinking that you’re missing out on the important and going events And beyond that, also, this idea that if you’re at work, you’re like not doing either thing very well So you’re at work, or you’re not quite like putting your best because you’re worrying about the kid at home And when you’re at home, you are worried that somehow you should be out there working So sometimes when we think about employment cost, we kind of want to catch up with [INAUDIBLE] OK so just to give you a preview of the top as well some of the results, are what we’ll be doing is that we’re going to be providing some pieces of evidence to support this claim that women do not seem to anticipate the employment effects of [INAUDIBLE] We’re going to start by documenting something that has been somewhat documented in the literature as well, but we want to show that this is true But this is in some way not going to be the main focus of the results but it’s something that we need to show, which is that we’re going to show that if you actually just track employment over a woman’s life cycle, you do find a very large negative employment effects of motherhood That in some sense had this step-wise function And I think that in itself is very telling because it’s sort of saying that if anything women’s employment changes seem to be because of you [INAUDIBLE] as opposed to potentially the other way around So in other words while we’re saying that in the years before childbirth, women’s employment is very stable at a very high level The [INAUDIBLE] itself exhibits a very large drop And one thing that’s perhaps a priority of this analysis is that we find very little evidence of recovery that we get to the long run In other words, we can track women up to 5 years or 10 years after the event of childbirth

And we see very [INAUDIBLE] That in itself is kind of a little bit surprising But when we get there we can now tell about why it’s one of the reasons why that might be AUDIENCE: I have a clarifying question So you’re saying more educated women experience smaller [? declines? ?] But then how do we square that? So what is happening on the socioeconomic levels? I mean that there are also arguments that women who are relatively educated and well-off, have more potential to leave the workforce because they have household incomes How do I think about that? I mean there is a differential, so that the proportion of women who are not college educated will typically be single There’s a very high proportion of them will be they will be leading single parent households And so are they going on social services? Is this a story of people going on social services? I’m trying to understand who is making the income for the family if it’s more educated Yeah Yeah If it’s not the mom working? JESSICA PAN: Yeah So we’re going to use this education result in a number of ways, as we see in a second But what we are arguing here is that the slightly smaller declining– as a percentage actually might not be necessarily smaller So these [INAUDIBLE] changes relative to the other more [INAUDIBLE] But what we’ll see here is that although that differential change more educated women is somewhat smaller So in other words, education has a mildly protective effect Actually the main effect of childbirth is [INAUDIBLE] Now, to answer your question, we don’t look separately at whether or not– so you’ll see in a bit, in order to do this analysis, you’re going to need a lot of years before and after the birth So we can’t actually cut by, for example, if you’re a single mom But it turns out that when we dug a little bit deeper into this precise question of who exactly’s earning the income if one’s working and moms actually drop out In many of the cases, even if they are single moms, actually, there might be a father figure around So it’s not necessarily the case Or they could be living with their parents, at home So we don’t look specifically at that subset AUDIENCE: Yeah, of course [INAUDIBLE] JESSICA PAN: Which we don’t have the power, unfortunately AUDIENCE: Could undereducated women be underemployed to start? Is that part of the reason why they would have less of a drop? JESSICA PAN: So here, we find that they have a pretty big drop, actually It’s the educated women that have a– AUDIENCE: The smaller– JESSICA PAN: –somewhat smaller drop But you’ll see, when I say somewhat smaller, one might be surprised at the drops a lot, actually [INAUDIBLE] the education [? doesn’t ?] [? have even ?] [? more of a ?] protective effect So having documented this– yeah AUDIENCE: Just, maybe, building on Hannah said, there’s also [INAUDIBLE] who are saying that leaving the workforce is much more costly for the highly educated women So the opportunity cost– [? and, ?] [? hopefully– ?] AUDIENCE: Yeah, yeah, yeah That’s interesting AUDIENCE: [? –someone ?] has to survive But it’s also not as costly Again, even [? the ?] [? integration ?] [INAUDIBLE] for less educated women JESSICA PAN: Yeah So it’s interesting, because I think– AUDIENCE: That’s interesting JESSICA PAN: –in this part, with the highly educated women, there are two potential stories here On the one hand, it could be, actually, more costly for them to leave But here, we focus on labor force participation So we don’t actually bring in earnings But on the other hand, it could also be that once they leave for a year or two, it’s much easier for them to come back in Now, what we see in these patterns, actually, for both groups, surprisingly enough, is actually very little evidence, at least in the short to medium, 5 to 10 years out, that there is substantial coming back in Now, the 10 years and 5 years includes the child number 2, child number 3 So it’s not clear, necessarily, that that’s the timeframe But I think for public policy, the fact that women are not re-entering within a 5 or 10-year period isn’t, in itself, [INAUDIBLE] AUDIENCE: You won’t be finding [? differential ?] [INAUDIBLE] [? re-entering. ?] JESSICA PAN: Oh, [? we do have the– ?] AUDIENCE: [? –by ?] [INAUDIBLE] JESSICA PAN: [? We ?] [? haven’t, ?] yeah And we haven’t, also, actually dug, specifically, very deep into that But [? moving on ?] a little bit more, and I’ll show you where education really comes in, because there’s a very interesting story in the background [? of ?] [? participation. ?] So what we find– employment affects their [? media ?] by education, and the education seems to help In terms of anticipation, it actually seems that college educated women are more surprised So you have– you have this difference there So in terms of unanticipation, we’re going to present evidence on two fronts We’re going to first focus on the very short run And then we’re also going to look at underestimation Now, looking at underestimation of long range is important, because if it’s consistent with our hypothesis that this impacts human capital accumulation, then there has to be some sort of unanticipation in the long run, as well, with people actually making these human capital investment decisions Now, in the short run, we’re going to be looking at two main pieces of evidence to show non-anticipation The first is that, consistent with this idea that women’s beliefs actually update with new information with arrival of a child, we find that when we look– instead of employment, but instead, now, we look at gender role attitudes So these are reported attitudes to what’s

the appropriate balance that you place on market work versus work at home So these are the questions that some of you may be familiar with in the general social survey, for example, that ask, does family life suffer if a woman works? How much do you agree with that statement? Whether or not a man should be working if he can– sorry, whether a woman should be working if her husband can support her– that sort of questions You find that you also see this evidence of the step function in these reported attitudes So in other words, for women in the years before having kids, their attitudes are pretty stable In the first couple of years of the birth of the first child, again, very discontinuously, you see a shift towards more conservative or more anti-work views And again, these views persist up to five years that we could [INAUDIBLE] the data Yes AUDIENCE: So you talk a lot about, I think, the socioeconomic status and education But I’m really curious about the role that race plays, especially given that expectations of motherhood differ by race So, for example, I think Amy [? Cuddy ?] did some [? fun ?] work showing that a good mother, if she’s white, she stays at home But a good mother, if they’re black, works So the idea that you have a kid, your gender– your belief about gender roles are changing, and the stigma and the psychic burden is affected by children seems to diverge from race So what– I guess, who are these women, when you say women as a [? broad ?] [? category? ?] JESSICA PAN: So I’m going to apologize straight up front here that, yeah, I think race is a really important and really interesting topic, but with this particular analysis, there’s actually quite little we can do with it, again, for [? power ?] issues And secondly, also, that most of this evidence, at least on attitudes, for example, actually come from the British survey where, again, this idea and concept of race is quite different in that context And so this is unfortunate that I had to use this lack of power kind of– but when we– so this is on average And unfortunately, we were not able to do any race cuts in this particular exercise But I’m happy to chat more– AUDIENCE: This is generally about white women JESSICA PAN: Well, there– at least in the US samples, there are black women, as well But this is on average, so we don’t– yeah, we were not able to do some sample cuts by race I would love to in the future But it’s just not [INAUDIBLE] Yes AUDIENCE: Do you speak to the different maternity leave policies in each of the different places during the time [INAUDIBLE]? JESSICA PAN: So we will talk about that in a bit And in fact, I’m going to bring in a little bit of evidence, also, from the Danish countries, so the Scandinavian countries, for example And there, I think I’ll comment on it very, very briefly By and large, though, my understanding, at least in the US, is that there hasn’t really been much change in that policy So there’s not much to say In the UK, the way you should think about it is that it’s very similar to the US So– in that it doesn’t have as much healthcare provision So it’s a bit better than the US, but not by much So that’s the first piece of evidence on lack of anticipation The second thing that we’re going to do is, we’re going to use more direct evidence And here, there– in the PSID, is a really neat question asking women directly about whether or not they anticipated parenthood being harder than it would be And there again, you find this very interesting education gradient where women are more likely to report– educated women are more likely to report that parenthood is harder than they thought it would be Now, moving on to underestimation in the long run, we’re going to be turning to these much longer-term expectation surveys where cohorts of [? women’s ?] who are high school seniors who are actually asked very consistently since the 1980s [INAUDIBLE] thought that they were going to be homemakers And there, one thing that you see that’s quite interesting is that, at least in the past, where you can actually connect some of these types of questions to earlier surveys that asked a lot of questions that [? Claudia ?] [? Goldman ?] has done, you would see that, actually, in the past, women actually seemed to hugely underestimate their future labor supply So more of them were saying that they would expect to be working from home And they actually were, if you actually connect them to their actual [? synthetic ?] [? selves ?] 20 years later So what we do here in this exercise is that we look at what high school seniors say in a particular year– for example, 1980 And then, when they are actually aged 30, which is what the survey refers to, we look, for example, in the [? CPS ?] or the NLSY, exactly what they were doing And we compare to see if there’s evidence of systematic overestimation And there we do find, for example, this reversal in that in the past, they used to be underestimating the official labor supply But today, more often than not– and, in fact, in an increasing manner– large groups of them are saying that they don’t expect at all to be homemakers In fact, a good 20% of them eventually do Now, you might wonder whether this is because of fertility considerations And here, we find, actually, that, in fact, women expect at least more of recent cohorts of women expect to have lots of kids So if anything, they’re overestimating their fertility, while they are overestimating their future labor supplies So then this brings us to the final part of our paper, which

is to ask the question why, if we agree that it seems like they are underestimating the employment effects of motherhood, why are they doing so? So in the paper, we provide a very simple theoretical framework that yields some of these empirical results, including this distinction by education, this idea that educated women are actually more surprised So given the time considerations, I’m not going to be going in depth with the theoretical framework, but one of the key pieces here is that in order for there to be under– unanticipation across generations, it would actually require a situation in which the cost of motherhood has systematically [? lessened ?] over time– specifically, in a way that people cannot easily predict based on past trends And so here, the evidence that [INAUDIBLE] a little bit more speculative And it’s going to draw a little bit more on a collage of evidence provided by different groups of people But here, we also show that, consistent with rising employment cost of motherhood on different series– so, for example, breastfeeding rates, cost of childcare, time spent with children– you do see not just evidence of these costs rising over time, but that some of these trends actually exhibit U shapes So for periods of time, they appear to have been declining And more recently, actually appear to be increasing– again providing another reason why women [? might ?] actually not so easy to predict what the trends [? might be. ?] So at least in this audience, I don’t think I need to spend much time on this slide about why we care about some of these findings As many of you know, women are the most elastic part of labor supply And in many countries where one is grappling with low fertility and stuff, you might anticipate that trying to get more women to work would be a good thing The other thing is that, at least from an economist perspective– and again, when I make this statement in more interdisciplinary crowd, it’s bizarre But many of our economic models, especially our dynamic labor supply fertility, actually start with the assumption that people rationally understand what all the costs are They can perfectly forecast what future labor supply is going to be like And so in that sense, actually, just showing this basic result that women don’t anticipate the employment effects of motherhood has implications for how we model labor supply behavior And beyond that, we can imagine that there are potentially large macroeconomic gains from increasing female labor supply and the wage convergence, even beyond the equity considerations So this paper is broadly related to lots of strands of work There are, more recently, a huge surge in papers looking at [? events, ?] [? study ?] investigations of motherhood and labor supply And interestingly enough, consistently, I think all these peoples across different cross-country contexts seem to find very similar results But that’s also very reassuring– and depressing [LAUGHTER] This is related very much to recent work by some co-authors as well as myself and [? Claudia ?] [? Goldman ?] that look at the role of changing as well as unchanging gender role of norms We provide here, in some sense, a lifecycle perspective on this And it’s also related to a much smaller literature, actually, mostly in sociology that has actually looked at the role of parenthood and gender norms But here, we’re [INAUDIBLE] [? to you. ?] So that’s the overview So just to give you a sense of what the data is, so we’re going to be using two main sources of data We’re going to be using the UK data This is British Household Panel Survey For those of you who are not familiar with the data set, it’s an extremely rich data set similar in structure, for example, to the PSID But what’s really neat about the survey is that in addition to asking things like employment consistently over 18 weeks, it also asks, very consistently, questions on gender role attitudes, which is something that we’re going to use So here, one of the basic requirements is that women want to have these attitudes before, in the years before the child is born, as well as the years after It’s actually not so easy to find a US data set that has that The NLSY kind of has that, to some extent, and we’re going to be using it, but more as a robustness check, because it’s only asked four times over that period The other thing that the BHPS is really good for is that it also asks you about employment expectations in the very short run So it asks women who have a job, for example, do you expect to be working next year? And again, [? that ?] sort of [INAUDIBLE],, they ask it with a lot of regularity and frequency And it is something else that we’re [? going to use. ?] So this sample– it’s not small It begins with a sample of about 5,000 households But you see that once we impose our sample restrictions, we’re going to be down to a much, much smaller subset, which is going to limit, a little bit, some of the other kinds of cuts that you might have wanted to see So most of our analysis is going to be using an event study design So essentially, it’s just a fancy word for saying that we’re going to be looking at how the event– how the outcome changes as a function of the event– so the years before and after childbirth So childbirth is our event

So [INAUDIBLE] analysis uses this design We’re going to be focusing on a sample where– that includes only people– women who have one– at least one child within that period So in this particular setup, everyone identifies the mean regression Now, in some robustness checks, we’re also going to include, as a control group, women who never had kids in this period as well as women who may have had kids even before the start of this period– as well as women who never had kids It turns out the results are very robust in whether or not we [INAUDIBLE] In the US data, we’re going to make a few [INAUDIBLE] restrictions to avoid very severely balanced events [? that ?] [? is, ?] in other words, [? seeing ?] too many [? observations ?] before or after the event So here are some summary statistics Again, in the interest of time, I’m not going to go into big detail on this But I think the only thing to really note is that for the BHPS and the PSID, which are the two data sets that you’ll be seeing the employment effects on, most of these individuals, even the youngest ones, are currently in their 40s So we’re not talking about the most recent group of women here So in that sense, it’s actually kind of like [INAUDIBLE] AUDIENCE: I have to ask a question [INAUDIBLE] have been asked many times before– [INAUDIBLE] this is not an exogenous shock? JESSICA PAN: I’ll show you So one of the reasons why, I think, most people haven’t quite looked at the event studies this much– and my hypothesis for why it’s more– there’s a recent surge in it is exactly this concern about [? exogenous. ?] So every time you think about how to study this– the effect of motherhood, you think, oh, I need an instrument And it’s really hard to find an instrument But hopefully, when I show you [INAUDIBLE] [? pictures, ?] it’s very compelling that it seems to be the case that it’s an exogenous shock So this is the only slide with equations And here, this is going to be the regression specification we’re going to run The Y here is the outcome So it could be the employment, or it could be general attitudes Here, we’re going to be– I’m going to be showing you from now on, actually, just graphs that plot this beta tau coefficient And what this beta tau coefficient here is our [? event study ?] coefficient, which tells you what the outcome is in the years before and after childbirth So tau here takes on values from minus 5– so up to five years before birth– to tau max So tau max could be between 5 or 10 In some data sets in the BHPS, we go up to five years out In some data sets like the US ones, we go up to 10 years out, so tau max would be 10 So we’re basically plotting the coefficients in the five years before and up to 10 years after the event of childbirth In all the specifications, we’re going to be controlling for [? year fixed ?] effects We’re going to be controlling for [? age fixed ?] effects And these are important because we are looking over a pretty long range of time So [? age ?] effects [? could be ?] [? such as ?] labor force participation, for example, could be changing over the lifecycle That takes [? a lot. ?] And in these specifications, we can also control for individual [? fixed ?] effects, because these are longitudinal [? values. ?] So essentially– oh I’m just going to be showing you graphs that plot the beta tau coefficients AUDIENCE: Do you have any way for accounting– of accounting for partner income and how partner income changes over time? Because that also speaks to [INAUDIBLE] point about, not only is it not exogenous to the person themselves, but also, if a partner earning more, in some cases, there might not be a partner [INAUDIBLE] JESSICA PAN: So this is going to be a potentially tricky question We [INAUDIBLE] some cuts by partner’s income pre-birth One thing that is going to be especially tricky to do is, because not all of these people are married in the years before– in fact 40% of them are actually not married before And then eventually, they get married over the lifecycle But it also makes it hard to define what a partner is and whether or not partner itself is an endogenous So I’m going to just– I’m going to not talk about that question for a bit But when it comes up and it’s relevant, we’ll bring it back in So the other thing is that in the pictures I’m going to show you, we’re also going to report one coefficient at the top And this essentially is going to be the difference [? and ?] difference coefficient, which collapses all of the event time study [? dummies ?] into one– before and after birth So that’s going to be the [INAUDIBLE] coefficient AUDIENCE: [INAUDIBLE] if you [INAUDIBLE] family and the mother [? in ?] [? this ?] society [INAUDIBLE] in the world now, mothers [? have ?] [? differing ?] so [? much. ?] And how can we– [? the ?] [INAUDIBLE] how can we [INAUDIBLE] [? I ?] can’t [? believe ?] [INAUDIBLE] JESSICA PAN: That’s a really broad question that I’m not sure we’ll get to We’re going to have some implications of this results, I think, at the end of it So if you don’t mind, I’m just going to sort of answer that question And then we can discuss it in the future Is that OK? AUDIENCE: Yeah JESSICA PAN: So I’m going to be first showing you events as steady graph, so with regard to employment And so we’re going to look at this in general

for the whole sample And then later on, we’re going to basically cut this up by dimensions related to our hypothesis such as education, pre-baby expectations and attitudes, as well as experiences of your own mother We’ve done a bunch of other cuts, as well So if a particular cut comes to your mind and isn’t reported here, I’m happy to also just tell you if we’ve done it So essentially, this is the event study plot for employment using the UK data I’ll show you in the next slide the pictures for the US data Now, so this is related to [? Iris’ ?] point about [? exogenating. ?] So if we were worried, for example, that employment was declining for women who were intending to have children, what you would you expect to see there, in some sense, is some evidence of [? pre-trend. ?] So in the five years leading up to a child, you would expect the women’s employment would already be declining in anticipation of having a child But what you see in this figure is, at least in the UK, women’s employment in years before having a kid is extremely flat, and very high level So we’ve renormalized this graph such that the coefficient at tau equals to minus 1 is exactly the raw mean of employment in the previous example And that’s at about 87% So in the years prior to childbirth, employment is very high at that level Now, the year of birth of the child, it’s associated itself with a 40 percentage point decline in the likelihood of employment Now, again, this is not in raw levels, because here, remember, we controlled for some of these fixed effects So this is relative to a no baby backdrop Now, the year after birth– so all of this [INAUDIBLE] [? mention, ?] is relative to tau equals minus 1 So that’s fixed Now, the year after birth, women recover a little bit by about 5 percentage points But after that, you see, again, very little evidence of recovery up to five years on AUDIENCE: [INAUDIBLE] full-time and part-time employment? JESSICA PAN: Yeah So this is just work last week So it’s a very simple binary question– were you at work last week? We’ve also looked at labor force participation as well as part time, full time And you see similar trends for [INAUDIBLE],, as well Now, you might be curious how this looks like in the US And so here, we’ve done this for three different data sets– both of the NLSYs as well as the PSID Now, they [? stack ?] [? at ?] slightly different levels partly because women are different ages at the start of the sample But again, I think all of them have this very, very stepwise form, where in the years before motherhood, again, very stable, pretty high levels of female labor force participation The year of birth itself, a huge drop down And again, maybe surprisingly, very little evidence of recovery after 10 years [? after. ?] Now, we have done this separately by the first birth and second birth And you’ll find that most of the effects are realized with the first birth There’s a bit more negative effect with the second birth, but not as much as the first birth Yes AUDIENCE: I think my question is going to be related, I just don’t have the statistical intuition Is some of this that actually, after the first birth in those first five years, people are having second children, and that’s dropping it 40%, but some people are going back to work, and so it looks flat? JESSICA PAN: So in general, from what we’ve found, is that it’s dropping once you have a first birth The second birth drops it slightly more by about a quarter of the size of the first birth We didn’t have enough [? power ?] to look at the third and fourth births But I think the way I would think about 10 years out is essentially that most women are probably having child number 2 at age three and four Now, by the time– even if you don’t have children number 3, 10 years out essentially is when your kid– your youngest kid’s just started preschool So in a sense, one would– you would like to actually extend this even further– 15 years out, 20 years out And I think for some of these cohorts, we might be able to But you can imagine, this is a very demanding exercise But I think our biggest takeaway from this is that I think very often, you think that women, after about five or six years out, shouldn’t be going back to the labor market And actually, we really don’t see them doing that And having 10 years out, whether you’re high skill or low skill, is a long time out Yeah So that’s a great question AUDIENCE: Just related to that, is there data that’s retrospective where women say, I was out of the labor force? Because just in my own experience among friends and family, it can be two years, it can be 15 years Is there descriptive data of how long people say they were out of the labor force after a child? JESSICA PAN: So I think that’s really interesting So one of the things that we draw a bit on– and actually, so there’s this– there’s a set up of these qualitative surveys, actually, from– that’s now a book called The Ambition Interviews that focus on a group of very selective, highly educated women that very much discussed this idea about whether women anticipate dropping out– and when they do drop out, how long they drop out for And I think many of these things– there are, obviously, people coming in and out But on average, at least, we do see a [INAUDIBLE] So moving on, so we do a number of checks, and in general, we find it’s very robust Interestingly enough, we find, as you might expect, very little effect of dads on the labor supply We’re not going to focus much on men But in general, the biggest takeaway is that there’s just not much effect on them in most dimensions So that’s that And again, with the effect of the second child,

it does have an additional negative effect, but it’s somewhat smaller than the first [INAUDIBLE] Now, moving on to heterogeneity– and this is, I think, quite interesting– is that here, we’re going to look at heterogeneity based on characteristics that are [INAUDIBLE] related to our hypothesis So we’re going to be considering a number of factors that the literature in general has emphasized should protect women against the employment effects of motherhood So this is going to be things like whether you have yourself a college degree, whether you have yourself a working mom In addition, we’re going to have access, at least in the BHPS, to two of these really neat questions about their expectations about the employment cost of motherhood So for example, we can [? cut ?] [? about ?] whether or not they were very progressive in their gender views in the beginning And we can also use this question, at least in a couple of these data sets, whether at age 18 you actually anticipate working [INAUDIBLE] And we might expect, for example, that women who do expect to working in their 30s or report doing so would imagine that they would have a lower employment cost AUDIENCE: Do you have any information on the [? populous ?] gender views [? separately? ?] JESSICA PAN: Yeah So we have some information on gender views, but as I mentioned earlier, one of the tricky parts here is that, ideally, we would be using the gender views of the– AUDIENCE: [INAUDIBLE] JESSICA PAN: Before the birth And they’re actually– quite a large number of them are unpartnered So we’re not actually able to use that Now, on the other hand, we’ve done some cuts, for example, by spousal education and their partner education, because that’s a characteristic that’s more unchanging And there, again, we find, actually, very little heterogeneity by that particular cut Now, in general, we find very little differential effects with gender of first child, whether or not you were married before your first child, your age at first birth, and, surprisingly enough, whether or not you were in a so-called family friendly occupation [? before. ?] AUDIENCE: [? Anyway, ?] I find the age at first birth really interesting There’s this work by David [? Ellwood ?] that’s– I also thought about when I asked about the [? heterogeneity. ?] And that suggests that college-educated women delay the birth of the first child quite a bit But you’re saying it doesn’t matter whether you delay or not You just drop out JESSICA PAN: So it doesn’t matter where you– so we do find women are delaying But conditional on delaying, it doesn’t seem like delaying in itself actually has a huge effect on the size of the– of the decline It’s surprising– we kind of expected to find, maybe, more heterogeneity than what we found And maybe that’s one of the takeaways here AUDIENCE: It’s very interesting, yeah JESSICA PAN: So by college graduates– so in fact, of all the dimensions that we look at, so this is one that it does seem like it has a mild protective effect So across the two groups, the differences are statistically significant So the way we [? are ?] [? reading ?] these are basically, for each of the data sets that we have, we did a cut We don’t do the whole event study, because that’s just going to take up too much space But essentially, we report the coefficients Now, one thing to note, though, with college graduates is that even though it does have a mild protective effect, even the main effects for college graduates itself– it’s large It’s a 20 percentage point decline on average across [INAUDIBLE] Now, whether or not your own mother worked– some of the– some previous studies have actually found that this has some differential effects We don’t find any in this case One thing to note, though, is that we don’t have very good controls for the reasons why mom worked So it could be conflating, for example, working due to necessity versus working because she [INAUDIBLE] progressive And so maybe that’s why we don’t find much, but we don’t Now, pro-work attitudes is interesting, as you might expect At least in two of the data sets, being more liberal pre-baby yourself actually does have a protective effect And if you said that you plan to work, again, your employment [? effect ?] [INAUDIBLE] But even in all these cases, the main effects are large Yeah AUDIENCE: So the pro-work attitudes are measured before? JESSICA PAN: Before AUDIENCE: So that tells the story of this is not necessarily a woman that are updating the attitudes toward work and [INAUDIBLE] different choices, but it’s mostly companies or firms that, through [? statistical ?] discrimination, are the mechanisms that are actually affecting– they are the real cost JESSICA PAN: So I think it’s really hard to tell, because the reason why we do this heterogeneity is that– so one thing you might be concerned about with our results is that next, I’m going to show you these results of gender role attitudes And we’re going to say, oh, but it’s not very surprising, because women are just going to update these unemployment So I stopped working, and therefore I just become more conservative, because that’s how I reconcile within myself that it’s OK to do that But one thing that we find that speaks against that particular hypothesis– and it’s more on this idea of updating– is that specifically on these cuts– so I’m hoping that you’ll remember what these cuts are, so I’m leaving them on the screen for a bit longer– is that if anything, everything moves in the opposite direction So a simple cognitive dissonance story would predict that how I update is going to move exactly in this direction In fact, we find everything moving in the opposite direction So while college graduation has a protective effect on employment, if anything, [? what ?] you’ll find is that this group of women are the ones that are most surprised by the effects of childbirth

In the same way, own mother, we don’t find much effects But if anything, those were the own mother who were actually, again, the most surprised It’s going to be hard, because we’re cutting exactly on the variable of interest But it turns out that the entire shift is actually among women who themselves started to be born [INAUDIBLE] the other ones that has– had the biggest conservative shift And again, among this group, you see the biggest reversal that’s [? opposite. ?] So that speaks against the simple cognitive dissonance [? aspect. ?] We haven’t talked about employers yet, but I think let’s move a bit towards the end, because it’s one of the potential costs [? that we can ?] [? talk about. ?] So this is a summary of the employment effects The only one thing I want to mention is that these employment effects are actually very consistent with what a recent study by [? Cleland ?] finds And there, they use really rich administrative Danish data And they also find, perhaps surprisingly, this stepwise function despite the [? varies ?] of liberal paternity policies So again, very depressing That being said, the effect size that they find is different In fact, it tends to be about half the size or a third the size of what we find So it could very well be the case that these policies do have some effect, but mostly in terms of the size of the employment effect But these are very consistently measured, very [? sharp ?] [INAUDIBLE] So moving on to anticipation, we’ll look at gender role attitudes So the basic idea here, again, is that if women don’t fully anticipate the employment costs, we will expect them to be updating their beliefs So we’re going to be looking at, specifically, these questions related to gender role attitudes So these are the list of six questions that we use So these are questions like family life suffers if a woman works full time, a woman in family is happier if she works, a husband and wife should both contribute income, husbands should work; wives just be at home So all of these are subjective questions Respondents answer them on a scale of one to five Wherever relevant, we reverse them in such a ways [? that ?] [? should ?] a higher number implies a more liberal or more pro-work attitude So I’m going to be showing you results where we just aggregate the six measures into a single index But we’ve also done this separately for each of the components And for five of the six components, you will see similar effects Only one where we see more null results is actually the first one, preschool and child suffers if a mother works We can speculate for a long time why that’s the case So we don’t actually know for a fact, but it seems to be the only question in the list of six that actually refers directly to an age group of a child So maybe that’s [? why. ?] So in general, you see that women tend to be slightly more liberal than men in answering these questions, but not by a huge amount The other thing to note also is that here, we have about 600 of the 700 observations, and we observe about three times before the birth of a child the general attitudes And we observe about two to three times after You’ll notice that this is going to be less than what we have on employment, because these questions were asked every other year So yes AUDIENCE: [INAUDIBLE] how to improve women and the children’s life in any way, [? because ?] [? the ?] children are suffering in [INAUDIBLE] [? but is that ?] [INAUDIBLE] the family where men and the women [INAUDIBLE] Would you say that [? is the ?] [? case? ?] JESSICA PAN: I am not sure if I would have any specific prescriptions for how to improve more generally, but I think we’re focusing on the very narrow question about were the women actually anticipating working And so I’m not sure [INAUDIBLE] specific– AUDIENCE: What was the topic, anyway? JESSICA PAN: I’m sorry? AUDIENCE: The topic of your speech– what was it? JESSICA PAN: It was about the money effects, about whether or not women anticipate the employment effects of childbirth So this figure here presents an event– the event study analysis of general attitude So instead of the employment outcome, now we just switched it with the index of gender norms And here what you observe, again, is that in the years before the birth of the first child, there’s, again, very little evidence of [? pre-trends. ?] So it’s kind of flat Maybe if you squint, there’s a bit of a decline Now, the year of birth of the first child, women become slightly more conservative And a year after, become even more conservative And that seems to persist after five years/ In terms of magnitude, it’s about a quarter a percent of deviation in terms of a shift towards the more conservative direction And the 0.9 actually is kind of close to the male-female differences in conservative attitudes So again, I can– so this is consistent with the idea that in the years before the birth of the first child, women’s attitudes are quite stable, and then they experience some sort of updating [INAUDIBLE] with the event of childbirth Again, I think we [? don’t ?] [? make ?] [INAUDIBLE] per se, but the fact [? there’s ?] no evidence of [? pre-trends ?] suggests that these changes are post-childbirth rather than pre-childbirth Now, with men, you actually do find slight effects, but they’re are about a quarter of the size of that for women, at least in the BHPS And [INAUDIBLE] Now, we’re going to try and do this exercise, also,

for the NLSY just because it’s useful to have a counterpart to compare One thing to note is that although the NLSY is a very long series, this question was only asked four times So the number of people who actually identified each [INAUDIBLE] coefficient’s going to be quite small But in any case, we’re going to do this And reassuringly enough, at least then the error bars are very large, but actually, you find quite similar effects Now, here, you do find some evidence of a [? pre-trend. ?] But we’ve actually done some of these robustness where we control for individual, person-specific time trends and stuff like that And actually, it doesn’t seem to really matter But again, I will say, take this with a bit more of a pinch of salt, because when we started this project, actually, we chose not to use NLSY But we came around– full circle, because we thought it would be useful to compare Now, moving on to– oh, one thing I should mention is that you might be wondering whether this shift in general attitudes is a larger shift towards cultural conservatism or some other measure And here, with the BHPS as well as NLSY, you can’t– you have other questions that’s not related to this idea of men and women’s work And there, we’ve done the same exercise [? And then ?] these [? are ?] questions about cohabitation, [? formal ?] sexual relationships, and divorce We don’t actually find similar changes in terms of shifts Now, so we’re going to look at heterogeneity So as you recall, the [? patterns ?] [INAUDIBLE] we’re going to be comparing it with the patterns for gender role attitudes And this is what we find So in general, you see here that, again, the differences for college education– college graduates are significant And they actually move in the opposite direction So the largest shifts are actually among women with a college degree relative to women without a college degree Having an own mother work also seems to result in much greater surprise Women who plan to work– and this we could only use using the NLSY– they too also tend to have the largest surprise, as well Now, in fact, if you look among women who didn’t say [INAUDIBLE] [? work, ?] actually, there was no effect there Now, pro-work attitudes is a tricky one, again, because this is exactly the dependent variable that we’re using But here, you do find that most of the shift is, again, concentrated among women who [INAUDIBLE] again, this idea of surprise Now, so we spoke about this So I think one way of thinking about this is that it doesn’t seem like these shifts are purely an artifact of cognitive dissonance Now, moving on to a bit more direct evidence, we’re going to be using the PSID And there, they have this really neat retrospective question, which is, on a scale of 1 to 5, how would you respond to a statement like being a parent’s harder than I thought it would be? Now, with any of these subjective questions, you always have to take it a little bit with a grain of salt. And here, I think we’re going to be looking at two dimensions One is a comparison between men and women And beyond that, also, to see whether or not if our hypothesis about educated women being more surprised holds water, you might actually expect to see that also true here So this is the– this table shows the regression results So the outcome here is just a binary variable We code it people who responded 4s and 5s on this question as being people anticipated parenthood being harder than expected We run separate models for men and women And we are interested in looking at whether you see the education gradient Now, one thing to– that’s interesting is, if you just look at the mean on the dependent variable across gender, there are huge differences A fraction in both groups expect that parenthood’s going to be harder, but [? isn’t– ?] more women than men And secondly, men have no education gradient So whether or not you’re highly educated or not makes no difference in terms of your response to this question But for women, it appears to be the case that significantly, about 12 percentage points of women with a college education systematically say it’s harder than they would otherwise [? have ?] [? thought. ?] Now, another question that we could use– and here, there’s is going to be a little bit of sample selection issue But I’m going to explain it But I think it’s also very telling– is that the BHPS actually asks questions that pertain to very short-run expectations So this is literally that if, in a particular survey, if you had a job, you’re going to be asked, do you expect to give up paid work over the next 12 months? And then you give a binary answer, yes or no What we can do in the BHPS, because it’s longitudinal– we can always look next week to see if, in fact, your prediction came true Now, as you would expect, men are spot on Every year, they’re going to say, yes, I’m going to be working And, in fact, [? they will ?] [? be. ?] And what I’m going to show you here is that women, in the years before childbirth, are also spot on If you ask them in the years before childbirth, do you expect to be working 12 months from now, they’re going to be saying yes, and it’s actually going to be true But in fact, they become systematically less likely to be accurate in the years after childbirth So this is the event study figure where the black dots show the outcome in terms of

whether or not they actually gave up paid work And here, what we see is [INAUDIBLE] the previous figures where, in fact, there is an increase– about 17 percentage points in the likelihood of giving up paid work But in fact, only about half of these departures are actually predicted 12 months before So the predictions are the pink dots And in fact, there, you see a short fall And this short fall is about half of the main effect, implying that only about half of these departures are predicted even in a very short [INAUDIBLE] 12 months [INAUDIBLE] Now, again, you have to take these results a little bit with a pinch of salt, because ideally, we would actually have had information on women who were out of the labor force being asked the question [INAUDIBLE] to return [INAUDIBLE] So these effects, in some sense, are even more, perhaps, surprising, because it’s saying that even among women who remain in the labor force, they are still increasingly inaccurate So it’s conditioned on a sub sample of women who still have jobs at that point in time Now, one way that you could think about why this is happening could be that you don’t decide immediately that you want to quit, but in fact, you find it becomes, actually, harder and harder But still a little bit of a puzzle, because you would imagine that after five years [? out, ?] those women are actually very [? selective ?] and more attached But that’s one of the limitations Now, turning to more long-run evidence, so one aspect that’s very important to our story is that not only do women not anticipate in the very short run, but in the much longer term, there is also this lack of anticipation, especially at the points in which they are making human capital decisions Yes AUDIENCE: This is a question about trying to contextualize the short run effects a little bit, thinking about that expectation formation process for other job shock So somebody’s trying to guess the probability that they’re going to quit their job Now, is this similar– is there a lot of noise in all this? Like, do people make a lot of mistakes in other contexts? Or is there something unique about this decision? I’m trying to– JESSICA PAN: Right Yeah, no, that’s a very good question And I think– we were asked this, also, previously, and I think we tried to go back to the survey to look at other forms of job loss And I think, unfortunately, we were not able to find something that really quite gets at that But I think your question is more broadly about, are people just in general bad at forecasting things that happen with some particular regularity, or is this something really quite specific about this group of women? So I can’t answer that, but I think if we do come across a question that speaks more directly at probability of job loss in the medium or longer term, I think that would be useful AUDIENCE: Incredibly interesting One of my questions is, we’re basically looking at this within the context of how the women predicted, presuming that the women had full agency in the decision of whether or not they left work And I know, to [? Clementine’s ?] question, we’re going to talk a bit more about the interaction with employers But it may not be that women overestimated how hard work was with a child or how hard rearing children was, but underestimated employer backlash of motherhood or lack of ability to continue to be seen as full employees and, therefore, didn’t really have such full agency And then, less from the research side, but on a personal note, being a very typical woman who delayed childbirth, I didn’t underestimate how hard raising children would be I underestimated how joyful it would be [LAUGHTER] So it would never have dawned on me that I would feel the same tug to be with my kids, having always found work so joyful But it didn’t occur to me And also, having a working mother might not have any way made a woman identify with the joy that her mother had with– JESSICA PAN: So the joy could be an employment cost, as well [LAUGHTER] [INTERPOSING VOICES] JESSICA PAN: The joy itself is a cost, because [INAUDIBLE] I’m looking at all these cute baby pictures, and I could really be at home spending time with [INAUDIBLE] So the joy [? in ?] [? itself– ?] so it’s odd, a bit, the framing But in some sense, if you think about– so we’re not making any normative statement about whether it’s right or wrong to do this sort of thing, because in fact, [? we ?] all could be better if mom stays at home [INAUDIBLE] AUDIENCE: Anything could [INAUDIBLE],, yes JESSICA PAN: But the fact that one doesn’t anticipate the fact that children could bring so much joy in itself could be part of our employment costs But that being said, you might be a bit unusual, because there’s a nice study by one of my advisors that looks, actually, at three groups of women– women with only a career, women with only a family, and women with a career and family And guess which group she finds is the most unhappy? Career and family, yeah So the women with career and family systematically, on questions related to life satisfaction, questions related to emotional well-being actually report being more tired, more stressed, and more [INAUDIBLE] [INTERPOSING VOICES] JESSICA PAN: Perhaps Again, I don’t know [INAUDIBLE] [LAUGHTER] [INTERPOSING VOICES]

JESSICA PAN: Well, I have a nice career, and I– AUDIENCE: [INAUDIBLE] the joy of doing both [? is– ?] AUDIENCE: I don’t remember the author at the moment, but I like to focus on the study that showed women who had both were healthier AUDIENCE: Right JESSICA PAN: Oh AUDIENCE: That’s my favorite one [LAUGHTER] JESSICA PAN: That’s true That’s something that [INAUDIBLE] AUDIENCE: They don’t have time to be sick [LAUGHTER] JESSICA PAN: In the short run or the long run? AUDIENCE: I’ll let you know [LAUGHTER] JESSICA PAN: So here, we’re going to be looking at, basically, high school seniors We’re going to be looking at what they expect And then we’re going to match it with what they actually experience Now, again, with these surveys, there are going to be issues, because here, they are not asked probabilistically what they are likely to do They’re going to be asked in a very binary fashion, do you expect to be a homemaker versus not one, or selecting from a group of like 15 different occupations Now, what you find– so there’s a lot going on in this picture But I just want you to focus on this, the [? hollow ?] line So the [? hollow ?] lines And so these are basically expectations So you’ll see that over this period– so this is the year the respondent turned 18 So these are high school senior cohorts from 1970 to 2014 Essentially, women used to think with– a large majority of them used to think that they would be homemakers But that number, at least in recent time, was literally like 2% Now, you might be asking yourself at this point, these surveys– who are these people who are actually going to respond, I’m going to be a homemaker? And that’s a totally legitimate question But I think what is interesting here is that if you compare this with actual rates of homemaking– and that’s the green line, here These are different ages, hence the slight shift In general, they mirror each other very closely But during a period of time in which women’s labor force participation really hasn’t actually changed, women– high school seniors were actually becoming increasingly optimistic about their chance of not becoming a homemaker And I think that’s the basic point here that seems to suggest that there is this lack of anticipation [INAUDIBLE] Yes AUDIENCE: Do you think there’s [? a rule ?] for social desirability bias [? here– ?] the norms are changing, and [? I’m ?] [? reporting ?] norms more than personal expectations? JESSICA PAN: So I think that there could be that I think the fact that– the fact that they’re increasing over time seems to suggest– the fact that this is reversal would require social desirability to be changing [INAUDIBLE] could be precisely that story So again, I think with this figure, when we first saw it, we too did not know exactly how to interpret this But I think it does speak in the same direction when taken together with the rest of the evidence that suggests that there could be some sort of unanticipation Now, with the fertility pictures, on the other hand, you actually find that these are the precise cohorts of women that actually do anticipate having two to three children So it’s not the case in which a reporting is very low expectations of homemaking, because they’re thinking that they’re going to have fewer kids So that’s that In terms of education expectations, that too you don’t see such a dramatic change, as well, [INAUDIBLE] AUDIENCE: And I understand these are– could you go back? Sorry JESSICA PAN: Oh, sorry Yeah AUDIENCE: These are different data sets, or maybe this is not a question you can answer But it looks like this dramatic difference [INAUDIBLE] [? the red ?] versus the green as in something dramatic happened in that [INAUDIBLE] JESSICA PAN: Yeah So one thing that’s a bit unfortunate– AUDIENCE: But it’s different data sets, right? So that’s why we might not [? be able ?] directly– JESSICA PAN: Yeah AUDIENCE: –compare JESSICA PAN: well, It’s different data sets, but actually– so, for example, this green line here and these two are actually different data sets But they match up very closely together in terms of realized outcomes So this data set comes from the NLSY, where these are the expectations, and these are the realizations So this is the period of time in which women were hugely underestimated in labor supply But there, you see that actually, in terms of actually expect– actual realizations, it does track what we see in the CPS So it seem– but one thing that’s unfortunate is that we’re missing these crucial intermediate years where we have yet to find a survey that covers these two pieces So actually, this part here [INAUDIBLE] [? initial. ?] This part here and the crossover is something where we don’t actually see exactly when it crosses over We know [INAUDIBLE] that period But one thing, of course, we are a bit nervous about is that– the way in which these questions were asked were different That’s it These two series actually from the NLSY, just different [? course. ?] So it’s not something specific about the Monitoring the Future survey versus NLSY AUDIENCE: I totally believe the data and the story I’m just saying that change– it was dramatic JESSICA PAN: Yeah, it’s dramatic And it makes you wonder what happened in the intervening period, to be honest And I wish we had an answer AUDIENCE: In 1974 is when– and I just looked up the name to make sure I’d get it right– the Equal Credit Opportunity Act passed in the United States So it was in 1974 that women began to have their own access to an economic life and were able to have a credit card, would begin to be able to co-sign for a mortgage So all of a sudden, women’s economic gains began to attach to their own agency in the economic marketplace [INAUDIBLE] JESSICA PAN: In 1974 AUDIENCE: Yeah That before, that they did not JESSICA PAN: Right I wonder if anyone has actually studied

the effects of that law That would be a really interesting [? project. ?] AUDIENCE: Perhaps you have [LAUGHTER] JESSICA PAN: And perhaps we have, but [INAUDIBLE] [? data ?] on the left and on the right, so– AUDIENCE: Interesting [? That’s ?] [? really cool. ?] JESSICA PAN: So now, moving beyond just these patterns, I made a quick mention to the ethnographic work following high-achieving women So that’s by Schank and Wallace Some of you may have actually seen, they had published a bit of it in The Atlantic, these series of very in-depth interviews And so I think– we pulled out a few of the quotations from that, because I think it resonates well with the story that we were telling So for example, within this sample of high-achieving women, out of 43 of them, 34 have children And quite a fair share of them had dropped out of the labor market And I think some of these quotations are informative “I wasn’t planning on staying at home with my child, but once he was born, I realized something had to give, and I needed to figure out what.” At least in this sample, only two of the [? member ?] mothers had dropped out [INAUDIBLE] Again, this is all retrospective So in some sense, I think– we might actually have some access to some [? undergraduates ?] from the [? ’60s, ?] but it [? would be ?] nice to do something more prospective So moving on quickly– I don’t actually how much time I have, so I’m just going to keep going on– AUDIENCE: [? To ?] [? the ?] [? top ?] of the hour, yeah, to 1 o’clock JESSICA PAN: Oh, great Perfect So– which brings us to, I think, the natural question– what do women– why do women underestimate the mommy effects? And so our preferred explanation is because the employment effects of motherhood may have risen– and specifically, in a way that women could not predict based on past trends Now, we’re going to provide a collage of empirical evidence that seems to suggest that the employment cost of motherhood appears to have risen at least since the 1990s after what appears to have been decades of decline Now, in general, many of you are familiar with this literature Actually, there’s a very large literature in macroeconomics that very much emphasizes decline in the cost of motherhood– specifically, as a key driver to understanding why female labor force participation rates have risen so quickly since the 1980s Now, one might wonder if the same sort of explanations can also explain why is it that female labor force participation rates have actually plateaued And one possibility is that perhaps many of these gains that were associated with huge increases in female labor force participations may have actually petered out, or maybe even moved in an opposite direction And I think this is really– this part of the talk is really a lot more suggestive, and maybe, I would say, a call for more research in some of these explanations So in particular, some of these drivers that people have studied include things like medical advances, reduced mobility associated with childbirth [INAUDIBLE] of infant formula, allowing mothers to go back to work, for example There’s a very nice study by [? Greenwood ?] and coauthors suggesting that electrification in itself [INAUDIBLE] women from household chores So the advent of the dishwasher, the microwave, and, in general, the declining cost of electricity basically allowed women to substitute many of these things that would otherwise have to do themselves And there’s also some work that looks at declining stigma costs for working mothers, the evolution of social norms– perhaps, for example, precipitated by the use of the pill– that may have contributed, again, into this rapid rise in female labor participation Which brings us to this question– have these trends actually played out, even perhaps reversed? So one very good example that we have– and again, we are really open to more ideas on other possible avenues where these trends may have actually played out in reverse– is medical advice on breastfeeding, for example And I’ll show you there, it’s really a case in which in– early on in the 1970s, actually, breast feeding was on the decline because women were actually shifting to infant formula And now, more recently, if anything, people are suggesting that breast is best, and breastfeed until six months, one year And, if you can, keep going on, essentially So what this figure shows is that [? these ?] [? are ?] some specific dimensions at which we were able to collect data to look at how these trends may have played out over time So the way to think about this is that– if you just think narrowly about breastfeeding, you might think, oh, but that really only affects cost in the first couple of years So I think one way to think about these different trends is that they are actually at different ages of– for example, of kids So breastfeeding is at the very early– early on And then you have time spent with kids doing childcare activities And that’s a broader range of time And then you could also have things like, what is the average cost of childcare, which, again, stretches on until they’re about teenagers Now, what we see here– so the solid lines are basically the weekly hours spent on childcare And there, there have been a couple of pretty influential studies that have shown, actually, that between 1970 and 1990, actually, amount of time parents spent with the kids was pretty stable– if anything, somewhat declining But taking off some time in the 1990s, these have really shot up And so these activities that are specifically

related, for example, to doing educational activities with their children or [? shucking ?] them, even, from classes and from school And those have actually gone up really strikingly, especially for educated women Now, if you look, for example, at whether– the percent that ever breastfed in the US and the UK– again, here you see, at least in the UK, that these rates have been growing over time Now, in the US, where we were able to map it to a much longer time series in the past, you see evidence where it used to be declining, increasing somewhat, declining again a bit, and then, now, on a study trajectory upwards If you look at more conventional measures of the childcare costs, those have also seemed to have [? been– ?] [? increase ?] over time So at least on these dimensions, it seems that employment costs have been rising over time And, in fact, if anything, if a woman was projecting based on her mother’s experiences, for example, he or she would have a– she would have a particularly difficult job, because some of these trends have, in fact, actually been declining Now, another piece of evidence on rising motherhood costs is, we could turn to some survey data that basically asks current mothers, again, [INAUDIBLE] based on their personal experience, do they think parenting today is harder than it was before? These are not the questions about whether you think it’s harder than what you thought, but how does it feel like today relative, for example, to your mom? Again, that’s bias and all this And perhaps people want to show that they are doing more than what their moms do [LAUGHTER] But it’s informative that 2/3 of current mothers say that motherhood is harder today [INAUDIBLE] children And on the specific question about child rearing costs, about– over 56% say that they are more involved in their children’s lives than their mothers were in theirs Now, this is actually somewhat consistent with some of these patterns But beyond that, actually, there’s a lot of anecdotal evidence, for example, these days that– in the past, for example, kids could be left at home, or maybe in the car for five minutes without being caught [LAUGHTER] And some of these are harder to quantify But the idea was that back in the day, actually leaving your kids and take the subway in New York City was actually possible Today, it’s not And so these are some of the other costs that could have been involved here Now, moving to the question that was actually asked– and I think this is a very important one that, to some extent, we wish we could do a bit more, which is that [? has the ?] employment cost of motherhood risen? So we focus a lot on the life dimension of work/life balance So what’s the work dimension here? I know many people in this room are actively involved in these [? social ?] [? nets ?] one Has, in fact, the employment cost risen because workplaces are becoming more hostile to mothers? Now, on the one hand, I think there is some evidence increasingly influenced, for example, by [? Claudia ?] [? Goldman’s ?] work that in many workplaces, there is a rising return to very long, inflexible hours Now, that clearly is going to hurt Mom And beyond hurting Mom is that within the household, it could also lead to concerns about pay equity So, for example, it doesn’t even have to be the case that Mom’s work has become harder But if Dad’s work becomes harder and the returns to Dad’s work has increased a lot over time and someone has to stay at home, that in itself could actually lead to some of the effects that we see Now, beyond that, one thing– so one thing that we can look at with the data is, we could look, for example, at job satisfaction Now, again, here the data’s a bit tricky, because these questions are, again, only asked to women who still have jobs So again, ideally, you would have wanted to know, if you didn’t have a job, why you were pushed out of the job But we don’t have that But if it’s instructive– if it is instructive at all, we do find that on simple questions like job satisfaction, if anything, women with a job, even after childbirth, actually were probably just as happy, if not happier Now, again, this could be selection bias But it does suggest that a very simple model where employers are just discriminating against mothers doesn’t seem to be at play On the other hand, of course, women could have adjusted They could have moved to workplaces that were more amenable towards working mothers, and that’s why you don’t see much by way of job satisfaction But [? there ?] [INAUDIBLE] we don’t see much [INAUDIBLE] The other thing that’s interesting is that when we do a cut by family-friendly occupations– so these are defined as occupations with above or below the median in terms of the share of mothers, as well as the share of women, we don’t find much difference in terms of the mom effect So these– being in a family-friendly occupation doesn’t seem to mitigate the employment effect, suggesting that, to some extent, it seems to be observed in both groups, as well So that’s actually about as much as I can say about the work dimension I’d be curious on any ideas on how we could push this more But I think at least within this project, it could be on the work dimension, but we haven’t actually found much direct evidence on that front One thing that’s interesting is that one of the women [? who– ?] interviewed [INAUDIBLE] gave this sample size N equals 1 She lists all the endless things she does with the kids, and then she says, then I go to work, where I can rest And personally, I actually can relate to that Anyone who says childcare is hard– the work is hard– [? no, ?] childcare’s harder sometimes

So I’m just going to end with a bit of concluding thoughts and maybe a couple of ideas for future work that we didn’t address in this paper that would be really wonderful to address So we argue in this paper that women are surpassing their mother’s education levels while no longer surpassing their participation levels, largely because they don’t anticipate the rising employment cost of motherhood We show that women, especially educated ones, appear to be more surprised by how hard motherhood is, at least as shown in the change in attitudes, as well as their retrospective answers to some of these questions We also find, at least in the longer term, that no 18-year-old women specifically plans to be housewives, despite planning to be mothers, and especially in light of evidence that suggests that, actually, a large fraction of them really do end up becoming stay-at-home moms But there are these two related questions that we have not taken up in this project I think the first is that we’ve provided some very suggestive evidence that the costs of motherhood have appear to have risen, but very little by way of understanding, why is it that these employment costs have gone up? And in a sense, it’s also a puzzle, because if you think about it, this is, again, the period of time where women are most prepared to enter the labor market Why is it that during this precise period, social norms, for example, have changed in a way that make it more difficult for women to do that? One possibility could be changes in technology It could also be the case that motherhood itself has become more enjoyable There’s a [? great ?] opportunity cost of working outside the home Maybe one possibility is with rising inequality, more of them are married to husbands who can actually support their kind of lifestyle, and therefore, women choose to do that Have social norms actually become more traditional? I didn’t show you any figures there, but actually, if you plot the shift of gender norms over time, it very much mirrors the employment pattern, at least in the US So you see a greater shift towards more liberal attitudes, but then again, a plateau since the 1990s, as well Now, we don’t know if it’s cognitive dissonance, but it does– there is suggestive evidence that even the shift towards more liberal norms seems to have also slowed [INAUDIBLE] [? 1990, ?] which is a bit of a puzzle I mentioned just very briefly, perhaps rising returns of long hours also penalizes equality within couples, because there are lots of gains to be maximized, and it might make a lot of sense for– greater sense for greater specialization There has also been some work that has suggested– and again, no one has really shown this empirically, so I think it would make for a really good paper– that rising inequality could itself raise the stakes of child rearing Right now, the returns are even higher And there’s also some speculation that there’s a great increase in competition for top slots in college– again, making it– there being a much higher return for Mom staying at home, for example Now, another question that, if you think about what the implications of this is for understanding human capital accumulation is the question of, are we saying that women are over-investing in education, given that they’re not working? And I think here, we want to just be really clear to say that we’re not saying that And, in fact– [LAUGHTER] –it would be really bad if we were And in some sense, I think what this research suggests is that the fact that labor market returns appears to have slowed for women suggests that in order to really understand, why is it women are investing a lot in education means that we have to also step away just from looking at [? the ?] [? wage ?] [? returns ?] to education So I think there is now some literature looking at the nonlabor market returns education, and I think that’s where a lot of interesting work is going on For example, marriage market returns could actually be higher for women than for men in terms of spousal quality, equality of the marriage, as well as [? equality ?] of children And beyond that, social returns could also be important– the intergenerational returns in terms of health and education of the population [INAUDIBLE] [? That’s it. ?] [APPLAUSE] AUDIENCE: So we have just five minutes, but we could take a question or two if you’re– JESSICA PAN: Absolutely AUDIENCE: So you would be [? miss ?] expectation ever eventually collect– like, once you’re 45 or something, you should internalize the fact that [? you will ?] probably want to [INAUDIBLE] JESSICA PAN: Yeah That’s a really good question I think the only way we– one that we could look at that might be the very short run at [INAUDIBLE] expectations But I think, again, for that, it’s going to be very difficult for us to focus on such a long period after childbirth AUDIENCE: No, but you [? could ?] see the same [INAUDIBLE] for 10 years after [INAUDIBLE] So [? essentially, ?] you should– your distinction becomes– [? becoming ?] smaller and smaller JESSICA PAN: Yeah So in the data, we just actually haven’t been able to test that, because– so the one that we could look at for short run and [INAUDIBLE] expectations, that’s using the BHPS And that, we were only able to go up to about five to 10 years– five years after AUDIENCE: [? Oh, ?] [? I ?] [? see. ?] JESSICA PAN: So it would be interesting to know if, at some point, they self-correct and they just realize, or if every period they keep thinking they’re going

to return, or– and they just keep getting it wrong But we just [INAUDIBLE] AUDIENCE: So is the cost of raising a child increasing for men, for fathers, as well in the last few years? Because if this is the case, then it’s kind of like a [? realist ?] [? method, ?] because you don’t see any employment effect for them But still, they are spending more time with the children So somehow, I think this is a kind of [? asymmetry ?] that– JESSICA PAN: Right AUDIENCE: So why, for women, the costs are increasing– there is the– such [? heartache? ?] JESSICA PAN: So I think– so with time spent with kids, I think fathers– educated fathers have also been spending a bit more time with their kids But the increase has been way less dramatic than that for women So there is already [INAUDIBLE] [? issue. ?] Now, the other thing, also– and I think this goes back into, it would be really nice to look at more characteristics of the husband, in some sense And I wish we could do a lot more But as you can see, we were really down to 600 observations And there’s a lot a lot of cuts that we want to do But I think beyond that, one thing that we didn’t touch on is this idea that maybe it’s not that women anticipate how hard parenthood is Maybe they anticipate how difficult it is to– how much work their spouse are going to do Or, for example, how much comparative advantage is generated in the first few years of childbirth We don’t have that precise question But I think, again, these are some of the– I mean, I think really understanding where husbands play in this is going to be really important And to some extent, I fully acknowledge that point And we weren’t able to do as much as we could But I would say, by and large in the literature, at least in the US, the gender division of labor within the house is still very, very pronounced So very often– if anything, if I did a cut, for example, by father’s income, you would actually see that the larger employment effects are actually found among women who are married to more high-income spouses So while these are men who have more liberal attitudes, perhaps, it also is the case that I think there are other countervailing forces that make it such that when one person specializes and really earns very big returns, the other [? one’s ?] [INAUDIBLE] AUDIENCE: Maybe it’s our education, because the more educated women go to college, and we tell them all [? the time, you ?] [? can do ?] [? both. ?] And reality is, it’s more complicated JESSICA PAN: Yeah No, it is Actually– so I guess– AUDIENCE: [INAUDIBLE] what happened was– JESSICA PAN: I guess this is two minutes [INAUDIBLE] AUDIENCE: Sexual harassment, the same thing JESSICA PAN: Yeah It’s really bad to do a plot for a project you’re currently working on, but– so one of the things that we looked at recently, Swedish data, is that you can actually look at, for example, kids who do extremely well, for example, [INAUDIBLE] So this is literally– the way I think about it is, if these are kids, imagine they’re– [? have ?] kids in a middle school, [? on ?] [? top ?] of the distribution And one thing that you can map– and this is probably fairly obvious– is that the probability of each of them making it to the top [INAUDIBLE] of their own gender-specific income distribution is the same, and very high But then, because of the gender gap in earnings, in fact, actually, these extremely highly talented girls are going to be something like 20 percentage points less likely to be in the top decile relative to the boys in the same class So sometimes it’s the messaging, because based on something like that, maybe the messages we’re sending is, [INAUDIBLE] AUDIENCE: The most encouraging thing I’ve read recently was about a young researcher who just had a baby, and she– [? and ?] [? had ?] conferences presenting her [? for quite ?] a [? while. ?] And after the baby, she pushed to get her grant for attending the conference increased so she could take the baby and so she could take her husband So this, [? I know, ?] was quite wonderful And when she got there, the hotel– there were no accommodations for pumping her– the breast milk part of it So but that push by a woman to say, these are my needs– I can participate if you meet my needs rather than just taking all the cost on herself JESSICA PAN: Right AUDIENCE: That’s a great example I think Harvard actually has a policy like that where you can apply for those types of grants Please join me in thanking [? Jessica. ?] [APPLAUSE]

You Want To Have Your Favorite Car?

We have a big list of modern & classic cars in both used and new categories.