Oh, Man! What Happened to Women? The Blurring of Gender-Based Occupational Segregation

videoType

dynamic-media

videoDmUrl

keyFrameImage

timestamp

00:01 OK.

00:02 Um,

00:03 it's 10:31 and I see that we have more than 40 people in the seminar already.

00:08 So it's a good time to start.

00:10 Uh,

00:10 we're

00:11 towards the end of our winter seminar series,

00:13 everybody.

00:14 This is the penultimate seminar.

00:16 There's one more tomorrow.

00:17 Thanks everybody for being so engaged in coming to all the seminars.

00:21 It's been wonderful.

00:22 And today we have another exciting,

00:25 uh,

00:25 speaker.

00:26 Uh,

00:26 Joanna is from Toulouse School of Economics,

00:28 and she's going to talk about

00:29 the blurring of gender-based occupational segregation.

00:33 Uh,

00:33 Joanna,

00:34 we talked about the rules,

00:35 and everybody here knows the rules as well,

00:37 so the floor is yours.

00:39 Thank you very much,

00:40 Burke,

00:40 for the introduction and thank you very much for the

00:43 opportunity to talk to you today about my supermarket paper

00:47 which as the title suggests,

00:49 studies what happened to women and in particularly what happened

00:52 to women's employment in the United States in recent decades.

00:57 Uh,

00:57 let me start by showing you

00:59 a graph

01:00 about the uh women's employment overpopulation ratio

01:03 from

01:04 the late 70s until 2019.

01:07 So as you can see,

01:09 during the 20th century or the last part of the 20th century,

01:12 Uh,

01:13 women's employment increased by a striking amount.

01:17 So these were more than 20% points and to put it in words of,

01:21 uh,

01:22 of Claudia Goldin,

01:24 this was the most significant change in the labor market during the past century.

01:29 However,

01:31 Since the beginning of the 21st century,

01:34 we've seen that there's been no progress

01:37 in uh female employment.

01:40 And when we check the same,

01:42 uh,

01:42 the same measure,

01:43 this aggregated by level of education,

01:45 so by subgroup,

01:46 depending on the level of education.

01:49 We see in the,

01:50 this is,

01:51 again,

01:51 employment overpopulation ratio.

01:53 The solid line is

01:55 highly educated women,

01:56 the dash line is uh women,

01:59 women with at most

02:00 a high school degree.

02:01 We see that low educated women

02:04 not only uh

02:06 didn't see any progress on their,

02:08 on their employment rates,

02:10 but indeed they,

02:11 they saw a decline over this period.

02:13 And this is even more fascinating because at the same time it has happened,

02:18 the type of jobs these low educated women have been traditionally performing,

02:23 which are service and clerical occupations,

02:26 have seen employment growth.

02:28 So specifically,

02:30 in this paper,

02:30 I asked,

02:32 why is it

02:33 that low educated women's employment has

02:35 decreased after decades of circular growth

02:39 and despite growing employment opportunities?

02:42 And what I'm gonna argue in the paper

02:44 is that it's because of the decline in local unemployment of low educated men.

02:49 And more specifically,

02:50 what I mean by that,

02:52 or what I'm,

02:52 what I'm gonna argue more in detail is that

02:55 manufacturing and construction jobs over this period

02:58 have seen an important uh job,

03:01 uh,

03:02 employment destruction,

03:03 and these were the type of jobs that

03:05 were

03:06 mostly employing low educated men.

03:09 So this decline has impacted mostly low educated men.

03:13 And these low educated men,

03:15 after being left with no other labor market options,

03:18 they've reconverted themselves into service and clerical occupations

03:23 and due to

03:25 This increase in competition,

03:27 they've put downward pressure on wages and they've crowded out low educated women

03:32 of these occupations,

03:33 of service and clerical occupations.

03:35 But given that these are important sources or were the most important source

03:40 of,

03:40 uh,

03:41 of women's employment,

03:42 also out of the labor market as a whole.

03:45 So I'm gonna be arguing this

03:47 by bringing new and critical evidence,

03:49 and particularly I will be using an idea approach

03:53 and also a structural model

03:55 which will help us uh think through the mechanism

03:59 that is driving this,

04:00 uh this labor reallocation.

04:03 So,

04:04 going to the empirical evidence in detail,

04:06 what I will be documenting

04:08 is that there's been,

04:09 uh,

04:10 there's been from 1990 to 2016,

04:13 following the decline in in blue-collar employment,

04:16 a reallocation of low educated employment

04:19 from blue-collar type of jobs

04:21 to work service and clerical occupation.

04:25 I also find

04:26 that most of this reallocation

04:28 is coming from local

04:29 men.

04:31 Just the ratio of low educated women to men

04:34 in these type of jobs,

04:35 it's the one that has seen

04:38 the,

04:38 the,

04:38 the most of the,

04:39 most of the decrease or the,

04:41 the,

04:41 the,

04:41 the most important decline.

04:44 And I also find when I exploit variation across local labor markets in the US

04:49 that

04:50 these two facts that I just mentioned

04:52 are stronger

04:53 in labor markets that have seen a stronger decline in blue-collar employment.

04:59 However,

04:59 to make,

05:00 to make a council

05:01 argument,

05:02 what I will do is

05:04 use an idea,

05:05 an idea

05:07 strategy

05:08 in which

05:10 this is gonna be a achieve share sort of instrument in style of what O and RD

05:15 in which

05:16 I will,

05:17 uh,

05:18 I will have,

05:19 uh,

05:19 instead of using the decline in model unemployment per se,

05:22 I will have,

05:23 I will exploit the historical industry,

05:26 the historical local industry structure in that local labor market,

05:30 which is gonna give me a predicted

05:32 decline in blue-collar employment in subsequent decades.

05:35 And uh following the,

05:37 the results of these,

05:38 uh

05:40 these

05:40 idea estimates,

05:42 what I will find is that

05:44 the decline in blue-collar employment

05:46 Can account

05:47 for 62% of the variation that I find

05:50 in the decline of women in low educated employment across local labor markets.

05:56 So it means that these,

05:57 uh,

05:58 these stories are primary candidate for the changes in,

06:02 in,

06:02 uh,

06:02 in,

06:02 in employment rates of women that we've seen over this period.

06:07 And then,

06:08 after documenting these empirical facts,

06:10 to understand why they have taken place,

06:13 I will build a general equilibrium model

06:15 in which individuals who differ by gender and level of education

06:19 are gonna sort into occupations.

06:21 And the question I will try to address with this model is how

06:26 productivity and changes in the composition of the population

06:29 have contributed towards this

06:32 uh workers reallocation that we observed in the data.

06:35 And what I will find

06:38 is that

06:39 this,

06:39 the,

06:39 the,

06:39 the worker reallocation that we observed in the

06:41 data can be fully accounted by productivity and

06:46 productivity changes and changes in the,

06:48 in the composition of the population,

06:50 which will mostly account for the fact

06:53 that over this period,

06:54 more women are attending college education.

06:57 And then,

06:59 This,

06:59 uh,

07:00 this exercise or,

07:01 or the model is gonna also

07:04 Uh,

07:05 tell us

07:06 that low educated women's employment,

07:08 if there wouldn't have been any productivity changes over this period,

07:12 would have continued growing

07:14 at the pre-1990 trend.

07:17 So,

07:18 it means that uh productivity is the main driver behind this,

07:22 uh,

07:22 the,

07:22 the,

07:23 the behind these uh changes,

07:25 uh,

07:25 that,

07:25 that I've been

07:26 telling you about.

07:28 And instead,

07:29 without population composition changes,

07:32 we would have

07:33 seen a stronger Decline in in female labor market participation over this period.

07:38 So in a sense,

07:39 the,

07:39 the fact that over this period,

07:41 more women are going to college and are attaining higher education

07:45 uh has partially masked the effects of productivity changes

07:50 in the labor market.

07:52 That's it,

07:53 let me give you an overview of how the paper sits in the literature.

07:58 So there is the,

08:00 there is,

08:00 uh,

08:01 it links to these two strands of the literature,

08:03 one which studies the,

08:05 the decline in blue collar employment and the consequences

08:08 this has had,

08:09 particularly

08:10 for low educated men or on the aggregate.

08:13 These,

08:14 uh,

08:14 these,

08:14 these papers,

08:15 they all focus either on,

08:17 on low educated men,

08:18 on the aggregate.

08:19 So what I'm gonna be doing is bringing the,

08:22 what happened to women.

08:24 AIM

08:26 Uh,

08:26 there's also the literature of,

08:27 uh,

08:28 of that studies,

08:29 the women's employment stagnation,

08:31 and,

08:32 uh,

08:33 uh,

08:33 the,

08:33 the first one,

08:35 in,

08:35 in the list,

08:36 uh,

08:37 documents these changes and,

08:39 and sees that

08:40 an important,

08:41 uh,

08:42 contributor to the stagnation is indeed the dynamics that are,

08:46 that are coming from the least educated woman,

08:48 as I showed you.

08:50 And then the other ones bring different explanations for why they have happened.

08:54 And these last two,

08:55 they study the,

08:56 the macro consequences of these,

08:58 of these changes for,

09:00 for,

09:00 of,

09:00 of,

09:00 of women's.

09:02 So basically my paper contributes to the literature

09:04 in that it links these two strands of,

09:07 so the decline in local unemployment

09:09 and the stagnation of,

09:10 uh,

09:11 of women's uh

09:12 labor market participation.

09:14 And then it also contributes to a third strand of the literature

09:18 which studies uh how structural transformation

09:22 affects women's labor market participation.

09:25 And all these papers have found

09:28 that the movement from of economic activity from manufacturing to our services

09:34 were beneficial for women's employment in the United States over,

09:37 over this during the 20th century.

09:41 But what I show is that

09:43 structural transformation

09:46 during the 21st century

09:47 might not benefit women's employment.

09:52 So,

09:52 oops,

09:53 that's said,

09:54 uh,

09:55 here is the outline of the talk.

09:58 I'm gonna be showing you the data,

10:00 then the empirical evidence.

10:01 I'll move on to the model and the calibration,

10:03 and then I will tell you about the backs.

10:05 And after that,

10:06 I will conclude.

10:07 So if there is any question,

10:08 now,

10:09 I will be happy to take it.

10:13 Yeah,

10:14 there's at least one,

10:15 Eshani.

10:17 Sure.

10:18 Hi,

10:19 thanks.

10:19 Um,

10:20 so I'm wondering,

10:21 you know,

10:21 with this sort of sectoral reallocation story,

10:23 um,

10:24 you might expect to see impacts on

10:27 women's education choices,

10:28 for instance.

10:29 Um,

10:30 you might expect to see whether lagging or leading,

10:33 I don't know,

10:33 I don't have a strong pride,

10:34 but you might expect to see impacts on fertility choices.

10:37 Are you seeing those and are you gonna tell us more,

10:40 um,

10:40 about how to think about

10:42 The fact that women are,

10:43 you know,

10:43 making some of these choices about labor

10:45 force participation in this broader context of,

10:47 of,

10:47 of things like how much schooling to invest in,

10:49 how many children to have,

10:51 um,

10:51 and,

10:51 and simultaneously,

10:52 you know,

10:52 what's happening to things like childcare availability.

10:56 Sure.

10:56 So these are all questions that I would love to explore.

11:00 So for now,

11:01 the educational decision,

11:03 I,

11:04 I,

11:04 I don't have it endogenous in the model.

11:06 I,

11:06 I don't think,

11:07 uh,

11:08 It's,

11:08 it's not the purpose of this paper to explain uh what has happened to uh

11:13 women's education and how this is linked to the manual decline,

11:17 but given what I observed,

11:18 it's happening in the labor market,

11:20 I believe it's an important uh margin of adjustment that women have.

11:25 Also,

11:25 what the counterfactual exercise showed.

11:28 Uh,

11:28 so I would love to explore this in,

11:30 in future research.

11:32 And then for the fertility,

11:34 I will be controlling for,

11:37 for

11:38 The,

11:38 uh,

11:39 share of women with uh kids younger than,

11:41 uh,

11:42 a certain age in the regression,

11:43 so that

11:44 will more or less be accounted there,

11:46 but I will be happy to explore it further.

11:48 And related to the,

11:49 uh,

11:50 to the family-friendly policies you were referring to.

11:53 Uh,

11:54 I think it might be linked also to the,

11:57 these changes

11:59 and,

11:59 and,

11:59 and the fact that there is plenty of labor supply

12:03 available for this type of service and clerical occupations.

12:07 I think it could

12:09 potentially have a link with the lack of family-friendly support

12:12 that we have observed in the United States compared to other countries.

12:16 So these are all

12:17 questions that I would like to explore in the future in more detail.

12:21 I can just jump in for one second,

12:23 Brick,

12:23 um,

12:24 just,

12:24 you know,

12:25 I guess I was

12:26 even saying that I would have loved to see that as a stylized fact as,

12:29 uh,

12:29 corroborating,

12:30 you know,

12:31 your hypothesis,

12:32 um,

12:33 not even,

12:33 you know,

12:33 going into the model,

12:34 uh,

12:34 we can get to that later,

12:35 but yeah,

12:36 just even

12:37 showing empirically,

12:38 um,

12:39 that this is,

12:39 this is something we're observing,

12:41 that these are changes that are consistent

12:42 with the sectoral reallocation story you're telling,

12:44 that would have been helpful.

12:46 Sure.

12:47 So,

12:47 yes,

12:48 I don't have them because that's,

12:49 again,

12:50 that's not the main focus of the paper,

12:52 but yes,

12:52 I can tell you that over this period,

12:54 uh,

12:55 the,

12:55 uh,

12:56 educational attainment of women is,

12:58 uh,

12:59 is increasing,

13:00 is,

13:00 uh,

13:01 is increasing and it's above,

13:02 uh,

13:03 the one of men.

13:05 So thank you for that,

13:06 for

13:07 pointing that out.

13:11 Thank you,

13:11 Joanna.

13:12 Paulo,

13:12 you're next.

13:14 Hi,

13:15 Joanna.

13:15 Uh,

13:16 so in,

13:16 in the stylized facts that,

13:18 that you've shown,

13:19 uh,

13:19 you see that the decline seems to coincide with the global financial crisis

13:24 and a period of low aggregate demand

13:27 and also the shocks that you'll be exploiting

13:29 are negative shocks to local labor markets.

13:31 So I was wondering if you can distinguish,

13:34 uh,

13:34 the,

13:34 the negative adverse effect of these shocks

13:37 from the reallocation and competition mechanism that,

13:41 that you're highlighting.

13:43 So,

13:44 what I will be exploring is,

13:46 uh,

13:47 is the historical local industry structure.

13:51 So,

13:51 yes,

13:52 most of these,

13:53 uh,

13:53 most of these changes though in,

13:55 in the,

13:56 in the decline in,

13:57 in the manual employment that I'm referring to in the paper

14:01 happened or

14:03 yes,

14:03 are,

14:04 are,

14:04 are taking place in the middle of the Great Recession.

14:08 Uh,

14:09 but this has been well documented in the literature in,

14:12 in,

14:12 in the paper of,

14:13 uh,

14:14 Jane and,

14:14 and,

14:16 I think was it that they found that most of the declining routine type of jobs happens

14:22 over,

14:23 over recessions.

14:26 Yes.

14:30 This is,

14:31 uh,

14:31 this is what I will exploit.

14:33 This is the,

14:34 the,

14:34 the changes I exploit on,

14:35 on the app.

14:39 OK,

14:39 reach up.

14:43 Yes.

14:43 Hi,

14:43 thanks.

14:44 This is actually related to what Paulo uh asked.

14:47 So,

14:47 as I see,

14:48 there are two very distinct trends,

14:50 and I was

14:51 uh

14:52 trying to understand that whether you'll be talking about two points in time,

14:55 or would you have

14:56 some stories as to

14:59 the initial decline that happened from 2000 to 2010,

15:03 somewhere around that period.

15:03 And then there's a reversal

15:05 from that point onward.

15:06 It's just that the reversal has not uh

15:08 caught up with the earlier levels.

15:11 Um,

15:11 in terms of your exogenous factors,

15:13 I would see that,

15:14 uh,

15:15 productivity might explain these two,

15:18 subperiod trends because they are volatile in general,

15:21 but in terms of,

15:22 uh,

15:23 full,

15:23 uh,

15:24 population densities,

15:24 they are not,

15:26 I mean,

15:26 my guess is that they won't have this

15:29 stark reversal as compared to

15:31 productivity,

15:31 for instance.

15:32 So would you be,

15:33 uh,

15:33 comparing across time,

15:34 so you will also have something to say about the subs.

15:37 Thank you.

15:39 So for now,

15:40 the model,

15:41 I compared two time periods,

15:43 so it's gonna be 1990 and 2016,

15:46 but I have,

15:47 uh,

15:49 I,

15:49 I,

15:49 I can,

15:50 uh,

15:50 I,

15:51 I plan to,

15:51 uh,

15:52 to do it by decade in the future and I think the model will,

15:55 will be able to,

15:56 uh,

15:58 just explain,

15:59 uh,

15:59 the decade,

16:00 the decadal changes from what I get

16:02 now from,

16:03 from,

16:03 from the two time periods.

16:05 I,

16:06 I believe that.

16:07 And uh I'm gonna show you the US national level,

16:11 some,

16:11 uh,

16:12 some variation by decades.

16:14 So,

16:15 so yes.

16:17 But yeah,

16:18 point taken.

16:21 Thank you,

16:21 Joanna.

16:22 Thank you.

16:23 Um,

16:24 Clement,

16:24 you had your hand up,

16:25 then it went down.

16:26 Was that intentional or you don't want to ask a question?

16:30 Yes,

16:30 yes,

16:30 my question,

16:31 uh,

16:31 has been answered.

16:33 Thank you.

16:35 Uh,

16:36 great.

16:36 Thanks,

16:37 Clement.

16:37 So,

16:38 uh,

16:38 you're the last before we continue.

16:44 Thank you.

16:45 Can you hear me?

16:47 Yes.

16:48 OK.

16:49 Yes.

16:50 So,

16:50 your study remind me of the uh of the literature in trade when we talk about.

16:58 Technology induced,

17:01 technology-induced,

17:02 uh,

17:04 wage differential compared skilled to non-skilled workers as well

17:08 as well as outsourcing induced trade group um.

17:13 Which differentiate between

17:14 skilled versus non-skilled workers.

17:17 And the period that you are studying

17:19 coincide with the rise of China entering WTO.

17:24 So there's a lot of

17:26 global macro

17:29 forces that is happening

17:31 at the same period of time that you are looking at this.

17:34 There's technology,

17:36 there is

17:36 outsourcing,

17:37 there's rise of China.

17:39 So will you be able to distinguish all this?

17:42 From

17:42 the local labor market conditions that you are trying to pinpoint.

17:47 No,

17:48 but at the end of the day,

17:49 these changes,

17:51 they all lead to the same,

17:52 uh,

17:52 uh,

17:53 to the same consequence that it's

17:56 the,

17:56 uh,

17:57 that the number of,

17:58 of,

17:59 of,

17:59 of jobs available in,

18:01 uh,

18:01 in this type of more manual jobs is decreasing.

18:05 So I'm,

18:06 I'm not

18:07 too much interested on what caused that,

18:09 but more on the consequences this,

18:12 uh,

18:12 this decline is having on the labor market and particularly on,

18:15 on the labor market,

18:17 uh,

18:17 opportunities of women.

18:19 So in,

18:19 in the,

18:20 in the model,

18:21 I will be,

18:22 I will be using technology to generate these increases in

18:26 productivity that are gonna be triggering this labor reallocation,

18:30 but I could potentially,

18:33 I will be more Explicit about it later,

18:35 maybe we will,

18:36 it will be making more sense after you see the model that we discuss

18:40 more in detail.

18:41 But I could just,

18:42 uh,

18:43 as long as the trade,

18:45 it happens within the what I define as the manual uh task.

18:51 Maybe I'm overestimating the contribution of,

18:54 of technology,

18:55 but again,

18:55 it would lead

18:56 these,

18:57 these increases in,

18:58 in trade in the,

18:59 in the manual,

19:00 in,

19:00 in,

19:00 in the manual task would lead to an increase in the productivity.

19:03 And at the end,

19:04 it's this increase in productivity what I care about.

19:08 So,

19:08 yes.

19:10 But yes,

19:11 I agree that there's,

19:12 there's many things that are driving the decline in,

19:14 in manual type of jobs

19:16 or in manufacturing,

19:17 construction,

19:18 and all these type of things.

19:21 But uh

19:22 for now I'll just

19:24 take no stance.

19:25 OK,

19:26 good.

19:27 I see that she is happy.

19:28 She's giving a thumbs up.

19:30 So,

19:30 and that was the last question.

19:31 So Joanna,

19:32 you may continue.

19:34 OK,

19:34 thank you very much.

19:35 So,

19:36 um,

19:37 to,

19:37 to get to the data

19:39 a little bit,

19:40 uh,

19:41 to show you what are the,

19:42 what are the data sources I use,

19:43 so I do use,

19:44 uh,

19:45 the Census on the American Community Survey

19:47 for the employment information,

19:49 and then I'll be using the current population survey,

19:52 the March supplement for the wage information.

19:55 And I will also be using the OET skills level data

20:00 uh to get

20:03 information on what are the skills that are being

20:05 used in all these different types of occupations.

20:08 So with this,

20:09 with this,

20:10 uh,

20:10 data set,

20:17 With this data set,

20:18 I will,

20:19 I think it's.

20:25 Um,

20:25 with this data set,

20:26 what I will do is to categorize occupations based on the type of skills they use,

20:32 OK?

20:33 And these are gonna be,

20:34 uh,

20:36 so,

20:36 so,

20:36 so to do that,

20:37 what I will do is to use the,

20:39 the original data which contains 46 different skills requirements.

20:44 And uh these are gonna be uh skills requirements like service orientation,

20:49 critical thinking,

20:51 equipment maintenance,

20:52 all these type of,

20:54 uh,

20:55 type of skills.

20:57 And what I will do is to use fact

20:58 analysis and use these 46 different skills requirements to 3

21:04 statistically significant factors.

21:06 And these factors can be naturally interpreted as interpersonal,

21:10 cognitive,

21:11 and manual skills.

21:12 And this categorization has been previously used in the,

21:15 in the literature into other papers,

21:18 so it's not that I'm doing anything

21:21 crazy here.

21:22 And doing,

21:23 and after doing that,

21:24 after having now for each occupation,

21:26 how uh how important these three factors are,

21:30 I will classify each of the occupations based on the factor

21:34 that is the most important.

21:36 And when I do that,

21:37 the uh occupation classification that I end up with is the following.

21:42 So,

21:42 occupations within uh clerical and retail sales and service,

21:47 uh,

21:48 are the ones that use interpersonal skills the most.

21:52 So I'm gonna be calling them interpersonal occupations.

21:55 Then managerial and other types of professional occupations

21:59 are the ones that are mostly cognitive,

22:01 so these are gonna be my cognitive occup.

22:03 Patients,

22:04 and then production,

22:05 installation,

22:06 transportation,

22:07 and construction.

22:08 This use mostly manual skills and these are gonna be my manual occupations.

22:12 So I'll be carrying uh on these uh

22:15 uh skills categorization along the,

22:18 uh,

22:19 along the,

22:20 uh,

22:21 the empirical analysis and,

22:22 and also the model,

22:24 and now to,

22:25 to go to the empirical analysis.

22:28 So,

22:29 what,

22:29 what I want to see is what

22:31 has happened to employment with uh through the lenses of these

22:35 three different occupations.

22:37 So

22:38 I see

22:39 that the,

22:40 this is the share of total employment.

22:42 Uh,

22:43 in the,

22:43 in the,

22:44 for the whole population in this table in 1990 and 2016

22:48 for each of the three occupations I just mentioned.

22:52 And what we see is that

22:54 the share

22:55 of total employment in manual occupations

22:57 has decreased substantially over this period.

23:02 And when I check who's been the most affected

23:05 by this uh decline in in manual employment,

23:09 so here you have in this table,

23:10 the change in employment over population ratios

23:13 in manual occupations disaggregated by gender and by level of education.

23:19 For the same years.

23:20 So this is just a difference.

23:22 So we see these are men,

23:24 low educated men in the first

23:26 column,

23:27 high educated men here.

23:29 So we see that it's uh

23:32 Mostly low educated men when compared to high educated men,

23:34 the ones that have suffered the most,

23:36 and it's the same for women.

23:39 So we see that it's

23:41 the ones that are suffering the most from the manual decline in terms of employment

23:45 are

23:45 the low educated and mostly low educated.

23:50 So checking again,

23:52 the same,

23:52 the share of total employment,

23:54 of total employment but this time only for the low educated for,

23:58 because are the ones that more,

23:59 that are more affected by the manual employment decline.

24:02 So this is again in 1990 and 2016.

24:05 I want to understand what has happened to the,

24:08 to the employment,

24:09 to the shares of employment in,

24:10 in the three occupations,

24:12 and what I see is that there's been

24:14 uh a reallocation of low educated employment.

24:18 Towards interpersonal occupations.

24:20 So from manual occupations towards interpersonal,

24:23 while cognitive,

24:24 uh,

24:25 jobs have stayed more or less constant.

24:28 And then when checking the,

24:31 the gender composition in this,

24:33 uh,

24:34 in these occupation.

24:35 So these are,

24:36 this is the ratio of low educated women to men employed

24:40 in the,

24:40 in the three types of occupations.

24:42 So in yellow,

24:43 you have interpersonal,

24:44 in blue,

24:45 you have cognitive,

24:46 in,

24:46 uh,

24:47 in,

24:47 uh,

24:48 gray,

24:49 you have manual.

24:51 I see that the decline

24:54 in,

24:54 in,

24:54 in the ratio of women to men employed has been the strongest in uh

24:59 in interpersonal occupations,

25:01 which tells us that the reallocation

25:04 of labor towards interpersonal occupation is mostly driven by men.

25:09 However,

25:10 everything that I've been telling you here,

25:12 it's through the lenses of shares.

25:14 And as,

25:15 as you were already mentioning in the questions,

25:18 uh,

25:19 there's been an important,

25:21 so,

25:21 so there's been important changes in,

25:23 in the,

25:23 in,

25:23 in terms of population groups over this period.

25:26 More women are attending college education

25:28 and all this.

25:29 So it might be that uh some of these effects are driven by that.

25:33 But when I check

25:34 employment over population ratios,

25:36 so this is controlling for the size of this,

25:38 uh,

25:39 Of these uh population groups.

25:42 Uh,

25:43 the,

25:43 the results I get are exactly the same.

25:45 So let me tell you what there is in this,

25:47 uh,

25:47 in this graph.

25:48 We have here the changes in employment over population ratios

25:52 by occupation.

25:53 These are the three boxes that you see here,

25:56 and the by gender.

25:57 So this is for the low educated,

25:59 uh,

26:00 men are in blue,

26:01 women are in red.

26:03 And when this is is aggregated by decades,

26:06 so

26:06 the first two columns referred in all the three boxes

26:10 refer to the 1980s.

26:13 The second 1990s,

26:15 2000s,

26:16 and uh the,

26:17 the,

26:17 the recovery from the Greek recession,

26:20 2010 to 2016.

26:23 So what we see here.

26:25 Is that

26:26 focusing on the 1980s,

26:28 you see that

26:29 women increase their,

26:31 their labor market presence

26:34 or,

26:34 or their employment rates in all three types of occupations.

26:37 So this is what I refer to as the golden age of uh women's employment

26:42 because these are still the uh

26:45 the strong growth that happened during the 20th century.

26:49 When we check

26:50 for the 1990s,

26:52 You already see that there is a decline ongoing in,

26:56 in cognitive,

26:57 in manual that is affecting women as well,

26:59 but still they have growth in interpersonal.

27:03 And uh then there's the,

27:06 there's the,

27:06 the Great Recession mostly that arise,

27:08 but the,

27:09 the decline in manual is also present before.

27:13 And you have this huge decline in employment uh rates both for men

27:18 or particularly for men,

27:19 but also for women in manual occupations.

27:22 And what we see

27:23 during this decade as well is that men are increasing their presence

27:27 in interpersonal occupations much more than women.

27:31 It's not still the case for cognitive.

27:34 But then during the period of the great recovery,

27:36 the Great Recession recovery,

27:38 you see that the patterns are completely reversed now in all the occupations.

27:43 So

27:43 interpersonal jobs are being taken by men.

27:47 Cognitive jobs,

27:49 uh,

27:49 men are growing more than women,

27:51 and in manual occupations,

27:53 there's the,

27:53 there's growth for men but not so much for women.

27:56 So

27:57 this is the,

27:58 this is the message that I want to convey that it's that we moved from the 1980s

28:04 where women were finding

28:06 most or employment opportunities in all types of jobs

28:11 to uh to a context in which these low educated women

28:15 are like not

28:17 even able to hold their jobs in the type of occupations

28:20 or,

28:21 or not even able to find employment opportunities

28:23 in the type of occupations they've been traditionally.

28:28 And this is at,

28:29 at,

28:29 uh,

28:30 at the same time that men do,

28:31 so.

28:34 So this is,

28:35 uh,

28:35 this is the,

28:36 the,

28:37 the main message

28:39 here.

28:40 But then still it might be,

28:42 so maybe I can take some questions now before I,

28:45 I,

28:45 I get into the,

28:46 uh,

28:48 the analysis across local labor markets,

28:49 if there's any.

28:54 Yeah,

28:54 I see a couple of hands.

28:56 Pierre,

28:56 why don't you start?

28:58 Uh,

28:59 could,

28:59 could you go back on the graph maybe for the question that would be helpful.

29:03 This one,

29:04 yeah,

29:04 yeah.

29:05 So thank you.

29:05 So,

29:06 so,

29:06 so what,

29:06 you know,

29:07 the way you presented it,

29:08 I guess is fine,

29:09 but one way to think of what's happened in 2010 is you're getting

29:12 this recovery of lost employment that's very obvious on the manual graph,

29:16 right,

29:16 in the 2010,

29:18 and then you see that the,

29:19 the drop for men is huge,

29:21 right,

29:21 even compared to that of women,

29:22 and in a way those men have to go somewhere post,

29:25 right.

29:26 And so if we're,

29:26 you know,

29:27 kind of dividing by the jobs loss in total in the recession period.

29:32 Uh

29:33 You would that change a bit the narrative and the picture?

29:36 Do you see what I mean?

29:38 You mean jobs lost,

29:40 uh.

29:43 So the,

29:43 the,

29:43 you,

29:43 you would put it in,

29:44 in terms of uh

29:49 Like in,

29:49 in,

29:49 in,

29:50 in shares of employment.

29:54 Yes,

29:55 but I guess you're trying to make a point that I guess,

29:58 and maybe it's true that men are taking jobs that were previously for women,

30:01 but it's also because their job is so,

30:04 you know,

30:04 it's so huge in the pre-period,

30:06 right,

30:06 and the jobs kind of have to go somewhere.

30:08 And I'm wondering kind of

30:09 by looking at total employment,

30:10 you know,

30:10 is that,

30:11 is that the right message is kind of the,

30:13 the recovery compared to the losses,

30:14 for example,

30:15 that are going in each sector,

30:16 right?

30:16 And then you'd be dividing by kind of

30:18 maybe the loss in the 2010,

30:20 in the 2000s,

30:21 you know.

30:22 But maybe it's just a better way of presenting and just

30:25 it,

30:25 it,

30:26 it could be,

30:27 uh,

30:28 it,

30:28 it,

30:28 it,

30:28 it could be that this,

30:29 yeah,

30:30 this,

30:30 uh,

30:30 the decline in manual for men is much more striking than what is it for women,

30:34 but these men,

30:36 like what we see is that women are not like,

30:38 uh.

30:41 Entering back into the labor market for,

30:43 for,

30:44 for the recovery of the Great Recession.

30:47 While for men,

30:48 it looks that there's a bit of recovery.

30:50 So why is it that these men are getting back

30:52 into the labor market?

30:53 There's a bit of growth in manual,

30:56 but most of the growth

30:57 uh is coming from,

30:58 from the fact that they are increasing their presence also in other types of.

31:02 Jobs that were not traditionally the ones they were doing.

31:05 So this is,

31:06 uh,

31:07 this is the story,

31:08 uh,

31:08 I,

31:08 I,

31:10 that's exactly my point,

31:11 Joanna,

31:11 which is if you add the recoveries in the 2010s,

31:15 right,

31:15 from those three little bars,

31:16 you still only have a fraction of the loss that you observe in manual in 2000s,

31:20 right,

31:21 even for men.

31:22 And so in a way that fraction to me seems fairly similar in

31:26 ratio to the one that if you were adding the gains for women

31:29 in the 2010s

31:31 uh compared to the loss in the 2000s.

31:33 Maybe I'm wrong because I haven't,

31:34 you know,

31:34 done the ratio exactly,

31:35 but you see,

31:36 so,

31:36 so that changes a little bit of the narrative.

31:38 It could be that just recovery was poor

31:40 overall for men and for women.

31:43 Uh,

31:44 versus just kind of men recovered and women didn't.

31:47 So no,

31:47 I'm not sure I'm seeing that,

31:49 you know,

31:49 that message you're saying from the aircraft.

31:51 I,

31:51 I don't have,

31:52 I don't have the,

31:53 uh,

31:53 I took it out finally from the sleds because they were too long,

31:56 but I should have put it,

31:57 so I have the uh.

31:59 The slides with the evolution of of employment rates for

32:03 like level of education and and and gender

32:05 and then you see

32:07 directly there that men uh

32:10 they they

32:11 recover from the Great Recession while women don't.

32:15 OK.

32:16 So this is,

32:16 uh,

32:17 but yes,

32:18 you,

32:18 you,

32:18 you're right I should maybe try to put it

32:21 in,

32:22 in relative terms.

32:25 But also what we have here is that uh

32:28 women are going more into college.

32:31 So,

32:32 so over this period here.

32:35 So

32:36 it must mean that the decline in employment is much stronger than,

32:39 than the,

32:41 than the decline in,

32:41 in,

32:42 in,

32:42 in,

32:42 in the population for this to go down

32:45 because the population of low educated women is decreasing over this period.

32:48 But yes,

32:49 point taken.

32:50 Thanks.

32:52 Thank you,

32:52 Joanna Chala.

32:54 Uh,

32:54 hi,

32:54 Johanna.

32:55 So two questions.

32:57 One is,

32:58 do we get anything different if you look at it by age groups,

33:02 by cohorts,

33:03 right?

33:04 Because

33:04 especially the older versus the younger woman,

33:07 uh,

33:08 it was asked before about,

33:09 especially in terms of fertility behavior,

33:11 do we see anything

33:12 there?

33:13 And the second thing.

33:15 The,

33:15 the big story,

33:16 I mean,

33:16 it's going to be the obvious question,

33:17 but the,

33:18 the migration behaviors changed considerably,

33:21 right?

33:21 The,

33:22 the migration from Central America and Mexico dried up

33:25 because of their demographic transition,

33:28 which was very strong in the 90s and even

33:30 in the early 2000s before the financial crisis,

33:33 and that dried up,

33:35 and that's also an endogenous behavior there,

33:37 which affected low educated men more than women.

33:41 And there's,

33:41 you know,

33:42 really complicated dynamics going on.

33:45 Does it change the story there if you introduce it?

33:49 Thank you.

33:49 So,

33:50 thanks to you for the question.

33:52 So by age,

33:53 uh,

33:54 Uh,

33:54 I disaggregated,

33:55 I don't have the,

33:57 the,

33:57 the graphs here,

33:58 but I disaggregated the,

33:59 uh,

34:00 the,

34:00 the evolution of employment rates by,

34:03 by age bracket.

34:05 So these are prime age individuals.

34:07 Everything that I'm telling you about today is for prime age individuals,

34:11 so people from 25 to 54 years old.

34:14 Uh,

34:15 so when I did from 25 to 35 to 35 to 45,

34:19 and,

34:20 uh,

34:20 on the last age bracket,

34:22 they all seem to have,

34:23 all,

34:24 all,

34:24 all the women seem to have the same

34:26 pattern.

34:26 So it seems that there's nothing uh age-specific there.

34:31 And uh for migration,

34:33 yes,

34:33 I agree that uh

34:36 They,

34:36 yes,

34:37 that maybe migration is,

34:38 is uh

34:40 It's an important margin of adjustment,

34:42 but this again is over total population of these subgroups.

34:46 So it's something,

34:46 it will be already

34:49 Uh,

34:50 already accounted

34:52 in,

34:52 in here

34:53 and in the,

34:54 in the regressions in,

34:55 in,

34:55 in the,

34:56 in the results I'm gonna show you later,

34:58 you know,

34:58 from the regression part

35:00 of the analysis,

35:01 I will,

35:02 I will

35:04 show you how like migration is something,

35:06 it's,

35:06 uh,

35:07 it's affecting me like it would,

35:10 uh,

35:10 make me underestimate the results I'm getting.

35:13 So it's potentially.

35:15 Yeah,

35:16 sorry.

35:16 Sorry,

35:16 is this

35:17 total population or native born population,

35:20 the figures you show it's,

35:21 it's.

35:23 OK.

35:24 It's all the population in,

35:25 in the United States.

35:27 OK.

35:30 Uh,

35:30 OK.

35:30 Thank you.

35:32 Thanks to you.

35:33 Good,

35:33 thanks.

35:34 And finally,

35:35 Sergio.

35:37 Hey,

35:37 John,

35:37 um,

35:38 uh,

35:38 it's interesting,

35:39 an interesting presentation,

35:41 and I will,

35:41 I,

35:42 I'm sympathetic with the type of analysis that you are making.

35:46 Um,

35:47 my question,

35:48 and I believe that the story is consistent with the story that you are telling,

35:53 um,

35:54 but I wonder whether you are pushing a little bit too much this displacement,

35:58 uh,

35:58 story because,

35:59 uh,

36:00 you don't have evidence that it's actually the men that are

36:04 living one and then they are really displacing the women.

36:07 I mean,

36:07 there could be other factors going on.

36:09 It could be women for some other reason,

36:11 uh,

36:12 withdrawing from that market.

36:15 So,

36:15 I would be a little bit kind of more cautious about how much you push that story,

36:19 even though it's consistent,

36:21 but you don't have any evidence to,

36:22 to back up your claim

36:25 completely.

36:26 OK.

36:27 So,

36:27 what I will,

36:28 in the,

36:28 in the,

36:29 in here,

36:29 these are all aggregate uh facts,

36:32 and I,

36:32 I don't make the capital claims out of these aggregate facts,

36:35 but more about the uh IV regressions that I'm gonna show you now,

36:39 uh,

36:40 because uh there I can control for like other demographics and uh

36:45 like.

36:46 Other

36:47 like variables that might be influencing this uh labor market behavior of,

36:52 of men,

36:53 and I will account for marriage,

36:55 uh,

36:56 uh,

36:57 share of,

36:58 of women with uh kids younger than a certain age that these are all

37:03 drivers of the,

37:03 of the,

37:04 uh,

37:05 Of the decision to,

37:06 to participate in the labor market.

37:08 So I would try to do my best to account for all these other

37:10 explanations.

37:12 But I,

37:12 I wanted to,

37:13 to claim causality,

37:14 you might need to observe the actual demand and supply of labor in

37:19 different markets to really observe that whether the IV would give you enough,

37:24 uh,

37:25 ammunition to claim causality or whether you might need other variables that

37:29 you might not have.

37:32 Point taken.

37:33 So,

37:33 I'm,

37:33 I'm doing my best with the,

37:35 with the information I have available.

37:37 So,

37:37 the IB I can,

37:39 I can show you more

37:40 about it,

37:41 uh,

37:42 maybe,

37:42 and maybe we can discuss in case uh you still have concerns.

37:50 OK,

37:50 thank you.

37:51 Thanks.

37:52 Great.

37:53 Uh,

37:53 that was the last question,

37:54 Joanna,

37:54 so you may continue.

37:56 Thank you.

37:57 So,

37:58 as I was saying,

37:59 these are all the facts that I document at the US national level,

38:03 but now,

38:03 like it might be,

38:05 uh,

38:05 it might be that they are happening simultaneously,

38:07 but they are not linked.

38:09 So to establish a causal link

38:11 with the decline in manual employment and the

38:13 reallocation of low educated employment towards interpersonal jobs

38:17 and the changes in the gender composition,

38:20 low educated employment in these jobs and in the uh

38:23 In the labor market as a whole,

38:24 what I will do is to explore variation across local labor markets.

38:29 And these,

38:30 uh,

38:31 I will proxy these local labor markets as the,

38:33 as commuting zones

38:35 in,

38:35 in,

38:36 uh,

38:36 as,

38:36 as author and Dorn and Hanson did in the 2013 paper,

38:41 that's the one where they studied the China shop.

38:44 So

38:46 The type of uh of uh regressions that will run.

38:50 So these are,

38:51 I mean,

38:52 I'm gonna regress the initial share of manual employment

38:55 in

38:55 computing zone J states at the beginning of the decade,

38:59 uh

39:01 uh T0,

39:02 OK?

39:03 On the decadal changes,

39:05 I will pull them all together

39:07 on the interpersonal share of low educated employment.

39:11 In,

39:11 in,

39:12 in uh communion zone G,

39:13 state A,

39:14 uh,

39:15 during the decade.

39:17 And then I will have,

39:18 so this is gonna be for the first

39:20 part for the reallocation of low educated employment.

39:22 So I will have state fixed effects that are gonna capture

39:26 anything related to regulation that it's state-specific.

39:30 I'm gonna have um

39:33 the fixed effects

39:34 and then some controls for the economic and,

39:37 and,

39:37 uh,

39:37 and demographic characteristics of these local labor markets.

39:42 And uh

39:43 maybe you are wondering why I'm using the,

39:46 the share of the initial share of manual employment here and not

39:50 the manual

39:51 decline or the,

39:52 the decline in the manual share per se.

39:54 And it's because since I'm checking shares here

39:58 and shares here,

40:00 even if the populations are not the same because in here I have all the population.

40:05 And this is only for the low educated.

40:08 If

40:09 it might be that by construction,

40:10 if I use the change here and this is,

40:13 this is uh decline,

40:14 this is declining

40:16 in manual occupations,

40:17 it should increase automatically in other types of jobs.

40:20 So to avoid that,

40:21 that's why I use the initial share

40:23 of manual employment here,

40:24 which is,

40:25 uh,

40:25 which is predicting,

40:26 uh,

40:27 which is a strong or,

40:28 or like a good predictor

40:30 of the subsequent decline.

40:33 So this is for the first fact,

40:35 which is about the reallocation of low educated employment

40:38 towards interpersonal jobs,

40:40 and then,

40:42 I,

40:42 uh,

40:43 I checked in the,

40:44 the,

40:44 the same sort of,

40:45 uh,

40:46 regression design,

40:48 the,

40:48 uh,

40:49 the change in the share of women among low educated employment

40:52 in interpersonal occupations and also in,

40:56 in the,

40:57 in all occupations,

40:58 so in the labor market as well.

40:59 So these are gonna be two separate

41:02 sets of regressions.

41:03 And again,

41:04 I use the initial share of manual.

41:07 And I will have some controls,

41:09 so I,

41:09 I will have again state fixed effects and,

41:12 um,

41:13 uh,

41:14 decade fixed effects

41:15 and some controls that are gonna be,

41:17 uh,

41:19 controlling for potentially

41:20 diff uh

41:21 things that might be affecting the,

41:24 the

41:26 The participation in,

41:27 in the labor market decision of women and men different.

41:30 So these are gonna be

41:31 having kids below the age of 5,

41:37 Being married or not and,

41:39 and all,

41:40 all,

41:40 all these,

41:41 all these things.

41:44 So,

41:45 the,

41:46 the effect I'm interested on then is this beta one in,

41:49 in both regressions,

41:50 which is uh gonna tell me.

41:54 By how the,

41:56 by how much more uh uh

42:00 uh uh

42:01 a commuting zone that initially had a higher share of manual employment,

42:05 so let's say a commuting zone with 1% point higher share

42:10 of manual employment initially.

42:13 How much more,

42:15 this is gonna be the beta one,

42:16 has seen uh a change in the dependent variable.

42:20 So either in the,

42:21 in the share of low educated employment in

42:23 interpersonal occupations or in the share of women

42:26 in low educated employment in interpersonal occupations

42:28 or in all the labor market.

42:31 So,

42:31 are,

42:32 are there any questions here?

42:38 I don't see any hands so far.

42:41 OK,

42:42 so,

42:45 Otherwise,

42:45 I can just show you

42:48 so.

42:48 Instead,

42:49 so still.

42:51 If uh

42:54 If there is anything left on,

42:56 on the error term that it's correlated with

42:58 the initial share of manual employment in,

43:00 in each of these uh local labor markets,

43:02 and it also affects uh

43:05 the,

43:05 the,

43:05 these uh two dependent variables,

43:07 we might have potential

43:09 endogeneity biases.

43:10 So this might be

43:12 uh

43:13 other current demand or supply

43:15 shocks that are happening at the same time and that correlate with the,

43:19 with the initial share of manual-employment.

43:21 So to avoid that,

43:22 what I will do

43:24 is to use a shift sharing instrument,

43:26 uh,

43:27 like the one author and Dorn used in the,

43:29 in the,

43:30 in the paper

43:32 in 2013.

43:33 And this will give me uh the manual share of employment

43:37 that is predicted by the local industry structure in 1980.

43:42 So,

43:43 I start my,

43:44 my analysis in 1990,

43:46 so this

43:47 This is gonna give me

43:49 exogenous variation,

43:50 and it's exogenous because it predates all these uh

43:54 other

43:55 potential,

43:56 uh,

43:57 contemporaneous,

43:58 uh,

43:59 changes that are might be affecting these uh

44:03 labor market decisions.

44:05 And it's also

44:07 very well,

44:08 it's strongly correlated with the

44:11 With the uh

44:13 with the share of manual employment at the beginning of the day.

44:16 So all in all,

44:16 I think it's a good instrument.

44:19 To use

44:20 and uh

44:22 so I can show you the results

44:24 that I get.

44:25 So this is for the first question.

44:27 So the one where I check the reallocation of labor

44:30 towards interpersonal occupations,

44:32 and here we have the

44:35 The,

44:36 the beta one estimates for,

44:38 for,

44:39 for the first set of regressions.

44:41 And what I find is that uh commuting zone,

44:45 so this is a,

44:46 sorry.

44:48 The,

44:48 I find that uh accompanying zone with an initial

44:52 share of manual employment that is uh higher

44:55 has seen a stronger reallocation of

44:57 low educated employment towards interpersonal jobs.

45:01 This,

45:02 uh,

45:02 commuting zone that we an initially higher manual employment is the

45:06 one that has seen a stronger decline in manual employment.

45:08 So,

45:09 it's,

45:10 it's,

45:11 uh,

45:11 it's,

45:11 it's what,

45:12 uh,

45:13 what my story was,

45:14 was,

45:15 uh,

45:16 what,

45:16 what,

45:16 what the aggregate,

45:17 uh,

45:18 facts were hinting at,

45:19 and indeed I find that these are linked.

45:22 So to interpret the results,

45:24 I find that by comparing a commuting zone at the

45:27 75th percentile of the initial share of manual employment with one of the,

45:31 uh,

45:31 at the 25th percentile,

45:34 the,

45:35 this one,

45:35 that is the one that will suffer the stronger manual decline

45:39 has seen a 1% point larger increase per

45:42 decade in the share of interpersonal low educated.

45:48 So this is for the,

45:49 this is for the first set of questions and,

45:51 and,

45:52 and this represents 14% of the variation across computing zones

45:56 in the change in the share of interpersonal low educated.

46:02 And uh so

46:04 for the second set of regressions,

46:06 so I have here for the interpersonal occupations,

46:09 yeah.

46:11 Excuse me,

46:11 there's a question I suspect it might be about the IV.

46:14 So before you continue with the results,

46:16 let's just take that,

46:18 sure.

46:20 Hi,

46:20 um,

46:20 so,

46:21 you know,

46:21 since,

46:22 since the author

46:23 work,

46:24 you know,

46:24 there's been a lot of

46:25 new work on the shift share instruments and,

46:28 um,

46:29 you know,

46:29 some of the work by Jaeger and

46:31 others have noted with the,

46:33 the dynamics here matter a lot.

46:35 So in the sort of context,

46:36 you're looking at,

46:37 you know,

46:38 there's adjust,

46:38 there's shocks happening all the time,

46:40 there's adjustment happening to previous shocks.

46:43 And so,

46:44 you know,

46:44 when you're looking here,

46:46 the shear here is not just affecting.

46:48 You know,

46:49 what's going on now,

46:49 but it affected what was going on

46:51 beforehand and so,

46:52 you know,

46:53 choosing an arbitrary starting point and not

46:55 building in those dynamics seems like,

46:58 You know,

46:58 from the newer work is not going to give you

47:01 the causal

47:02 impact.

47:02 I don't know,

47:03 sort of,

47:04 you know,

47:04 when you've looked at that,

47:05 that newer work,

47:06 have you,

47:07 do you think the sort of auto

47:08 approach still holds up,

47:10 or,

47:10 or how should we think about

47:11 this?

47:11 Should we just think of this again more as

47:14 on a descriptive basis,

47:16 this has happened and

47:17 the IV is giving us slightly moving in that direction,

47:20 or how do you want us to think about that?

47:22 Yes,

47:23 maybe it's not,

47:23 uh,

47:24 it's not what we would like.

47:26 So

47:27 the base would be if,

47:28 uh,

47:29 to,

47:29 to,

47:29 uh,

47:30 to have access to administrative data and really be able to tell,

47:34 like,

47:34 OK,

47:35 to,

47:35 to,

47:35 to be able to build the whole picture,

47:37 but this administrative data,

47:38 the problem they have is that they have limited

47:41 sample size.

47:42 So,

47:44 then,

47:44 uh,

47:46 Balancing one and the other,

47:47 this is the best way I could find

47:49 to make uh,

47:50 with,

47:51 with the data I have

47:53 to make a,

47:53 a case for,

47:55 for my results,

47:56 even for,

47:57 for,

47:57 for my,

47:58 for my,

47:59 sorry,

47:59 for my hypothesis.

48:01 And I,

48:02 uh,

48:03 potentially it's not ideal,

48:04 but this is the best I can do for now and I will be happy to explore

48:09 uh better ways in the future to,

48:11 to,

48:12 uh,

48:13 Clean up,

48:14 uh,

48:15 what,

48:15 what might be left that it's uh still there and it's not,

48:18 uh,

48:19 it's potentially affecting the,

48:20 the results.

48:22 So thank you for

48:23 your comment.

48:28 OK,

48:29 you may continue.

48:31 Thanks

48:35 So as I was saying,

48:36 this referred to the first,

48:37 uh,

48:38 set of progressions,

48:39 and these ones refer to the,

48:41 the changes in the share of women among global educated women.

48:44 So these two columns show uh the results for the interpersonal occupations,

48:50 and these

48:51 two columns,

48:52 the,

48:52 the,

48:52 these last two columns,

48:53 it's for the labor market as a whole.

48:57 So when we check interpersonal occupations,

48:59 uh,

49:00 The IV estimates show that

49:03 uh commuting zones where the share of manualemployment was initially higher

49:08 have seen a stronger decline in the share

49:12 of women among the low educated employed in this occupation.

49:16 And again,

49:18 by comparing the,

49:19 uh,

49:19 these two commuting zones,

49:21 uh,

49:23 one of the,

49:23 uh,

49:24 each,

49:25 each one of the uh each one at the extreme of the uh interpolter range,

49:30 we see that the one,

49:32 with initially higher manual employment,

49:35 that is the one that has suffered the stronger manual decline,

49:38 I've seen a 2.5% points,

49:41 larger decline per decade in the share of women

49:44 in,

49:44 uh,

49:45 in this type of.

49:48 Um,

49:49 so this accounts for 34% of the variation

49:52 that I observed across,

49:53 across committing zones in the in the change in the share of women

49:57 in interpersonal low educated.

49:59 And uh I didn't mention

50:02 before,

50:03 but when I get rid of the period of the

50:06 recovery from the Great Recession,

50:08 and even for the period of the Great Recession,

50:11 these,

50:12 uh,

50:12 this,

50:13 that have much more variances or variation.

50:17 Uh,

50:18 this,

50:18 uh,

50:18 this percentage

50:20 or the,

50:20 the percentage that these results can account for goes,

50:23 uh,

50:24 up much more.

50:26 So this is,

50:27 uh,

50:29 This is one thing also for the ones that I'm gonna show you,

50:31 OK?

50:32 So.

50:34 So now for all occupations,

50:36 so this is,

50:37 this was for interpersonal occupations,

50:38 but what happened in the,

50:40 in the labor market?

50:42 So we see again that uh

50:45 in line with the story that the,

50:47 the shared or,

50:48 or the commuting zones with initially higher shares of manual employment have seen

50:53 a stronger decline in the share of women among low educated employment in all,

50:58 in all

51:00 types of,

51:00 uh,

51:01 in,

51:01 in the whole labor market.

51:04 And by comparing the two commuting zones at the extremes of the uh intercuta range,

51:10 we see that the,

51:11 the one that has suffered the stronger decline has seen

51:14 a 2.8% points larger decline per decade in the share of women

51:19 in low educated employment

51:20 in the whole labor market.

51:22 And this accounts for 62% of the variation across communities.

51:26 And again,

51:27 when getting rid of the last decade,

51:29 uh,

51:30 or even the,

51:31 the period of the Great Recession,

51:32 this,

51:33 uh,

51:33 this percentage goes up.

51:38 So,

51:39 just to sum up,

51:41 I find that there's uh in,

51:43 in,

51:44 in,

51:44 in the data that there's an important decline in manual employment,

51:47 which has mostly affected low educated men.

51:50 I find that these uh low educated men have shifted towards interpersonal jobs

51:56 and that low educated women,

51:58 at the same time that these low educated men have gotten into these occupations,

52:02 they've,

52:03 uh,

52:04 they've,

52:05 they are not uh

52:06 increasing their presence there.

52:09 So,

52:09 I didn't show you because uh for the,

52:11 for the sake of time,

52:12 but I find,

52:13 uh,

52:14 I find that the uh average wage in interpersonal occupations

52:18 is going down as much as,

52:20 uh,

52:20 as the manual occupations.

52:23 So this,

52:24 this pushes uh for the story of the labor supply.

52:28 Because this,

52:29 even if these,

52:29 uh,

52:30 if these interpersonal occupations are seeing

52:33 growth in terms of employment,

52:34 which might be because there's increasing demand,

52:38 if they see a,

52:39 a decline in the wage,

52:40 it means that the,

52:41 the increase in the supply has been much more,

52:44 much bigger than the increase in the demand.

52:46 So this would be in line of,

52:48 of

52:49 Of my snoring

52:50 also.

52:52 Um,

52:53 so now,

52:55 I will move to the model and why do we need the model.

52:57 So basically,

52:58 here I documented a bunch of facts,

53:00 but to understand what can be the,

53:02 the main driver.

53:04 This is what I will use the model for.

53:07 We have a mechanism to explain why this labor reallocation has

53:11 taken place.

53:12 So,

53:13 I,

53:13 I,

53:14 I can take questions actually.

53:15 I,

53:15 I forgot to mention before starting the model.

53:21 Um,

53:21 yeah,

53:23 yeah,

53:23 there's,

53:24 let me ask a quick one,

53:25 and then there's one hand.

53:27 Um,

53:28 so your point number 4 in the slide before.

53:32 I mean,

53:32 those,

53:33 those two sectors,

53:34 manual and interpersonal,

53:35 are maybe like about 2/3 of the,

53:38 your,

53:38 your distribution.

53:39 So you're saying that

53:41 average wages over this period actually declined

53:44 somewhat significantly for like 2/3 of the workforce?

53:49 So,

53:50 this is what I observed for the low educated,

53:52 and this is disaggregated by gender.

53:55 So when I,

53:56 when you put the,

53:57 uh,

53:57 the averages without gender,

53:59 the changes in manual and interpersonal are of the same magnitude,

54:03 but there's growth in,

54:04 in cognitive.

54:06 Uh,

54:07 wages,

54:07 but yes,

54:08 you are right to say that uh.

54:11 That

54:13 A substantial share of the population on average,

54:18 so the,

54:18 the average wage for a substantial share of the employed population,

54:22 so a decline over

54:24 a decrease in,

54:25 in,

54:25 in the,

54:25 in the wage over this period.

54:27 But then there are,

54:28 there,

54:28 there might be

54:30 differences,

54:31 uh.

54:32 On a finer,

54:33 uh,

54:34 occupation level that I'm not taking into account here.

54:39 That might be also.

54:43 That's OK.

54:44 Thanks.

54:44 Um,

54:45 there's two hands,

54:46 Clement first,

54:47 Antoine.

54:49 Hi,

54:49 uh,

54:49 Jonah.

54:50 Um,

54:51 I was wondering,

54:52 so maybe it's similar to the point that Pia made earlier,

54:56 uh,

54:56 which is that your dependent variable is the share

55:00 of women,

55:01 uh,

55:01 in those occupations,

55:02 and I'm wondering whether,

55:05 so why do you work with the share if,

55:07 if what you're trying to show is that they are crowded out,

55:09 so they,

55:11 um,

55:11 wouldn't you want to work with,

55:12 um,

55:13 more like the,

55:14 the

55:14 I don't know like the total number,

55:16 uh,

55:17 because if you add more men to these occupations,

55:19 right,

55:19 like the share is gonna,

55:20 of women is gonna decline,

55:21 uh,

55:22 even if women still have the same access to these occupations.

55:25 Uh,

55:26 so why do you work with the share?

55:29 Yes,

55:29 uh,

55:30 so that,

55:30 that might be,

55:32 uh,

55:32 that the,

55:33 that woman,

55:34 so they are not in decreasing their presence

55:38 in,

55:38 in these occupations,

55:39 but still,

55:39 all the jobs that are being created from scratch,

55:43 they are taking over

55:44 by men.

55:45 So this would be another.

55:48 Uh,

55:49 another way to,

55:50 like,

55:51 that I think it's,

55:51 it's,

55:52 it's what's happening.

55:53 So,

55:54 some women saw a decline in their,

55:56 in their

55:57 In the,

55:59 in the present,

55:59 in,

55:59 in,

56:00 in the,

56:00 in the manual jobs as well,

56:02 in the manual occupations.

56:04 So,

56:05 now,

56:05 potentially they could have been getting into this type of jobs,

56:09 but they,

56:09 they can't because there is more competition,

56:12 wages are going down,

56:13 so

56:15 they just

56:16 prefer to not work.

56:18 But I,

56:19 I think if there again with Pierre's point,

56:21 if there was a big decline in employment of men overall,

56:25 uh,

56:26 then it wouldn't be a,

56:26 a crowding out story,

56:28 but rather,

56:29 uh,

56:29 you know,

56:29 the economy made space for these unemployed men in these occupations and that

56:34 didn't really affect women necessarily.

56:37 I mean,

56:37 I,

56:37 I,

56:37 I,

56:37 I,

56:38 I'm,

56:38 I'm sure your story is right,

56:39 but I'm,

56:39 I'm wondering

56:40 whether that's the best,

56:41 uh,

56:42 way to,

56:43 to show it.

56:43 I could,

56:44 yeah,

56:44 yeah,

56:45 yeah,

56:45 I see.

56:45 So you would do it like,

56:47 uh.

56:49 Yeah,

56:50 but then the,

56:50 the issue we're doing at the,

56:51 uh,

56:52 like in absolute terms is that

56:54 potentially,

56:56 like,

56:56 how do you control for the sizes of,

56:59 uh,

57:00 I don't know,

57:01 of the population of it gets

57:06 Much trickier,

57:06 but yeah,

57:07 I agree that I could try.

57:10 Yeah,

57:10 I'm not sure what the optimal way to do it would be,

57:13 but,

57:13 um,

57:14 I could try,

57:14 I,

57:15 I've,

57:15 I've tried with,

57:16 uh,

57:17 employment overpopulation ratios,

57:19 and I find,

57:20 uh,

57:21 I find the results are also there.

57:23 The issue is that population again is,

57:25 is,

57:25 uh,

57:26 something that is changing over time,

57:28 so I don't know if,

57:30 uh,

57:31 If there's a decline in the,

57:33 uh,

57:33 in the employment overpopulation ratio of women because of the,

57:37 because changes in the population or,

57:39 or what's going on.

57:40 So.

57:43 To avoid that,

57:43 I decided to go with the,

57:45 with the,

57:47 with the shares.

57:49 OK,

57:49 thank you,

57:49 you have about 16 minutes,

57:51 so let's quickly take Juan's question,

57:53 and then you should,

57:54 uh,

57:55 continue.

57:56 Yes,

57:57 so I want to show you the model.

58:00 Yeah,

58:00 so just,

58:01 just quickly,

58:01 so,

58:02 so you're going to do some welfare analysis I think.

58:04 So my question is,

58:05 Talking about crowd out,

58:06 why,

58:07 why isn't

58:08 the option of women,

58:09 low educated men going back to college

58:11 and getting an education,

58:12 uh,

58:13 something that is an important margin to look at?

58:16 Sure,

58:17 sure,

58:17 it is,

58:17 it is,

58:18 I believe,

58:18 uh,

58:19 it is,

58:19 I'm gonna show you in the model that

58:21 these changes in the composition of the population

58:25 are having an impact on what's happening in the,

58:28 in,

58:28 in the labor market as a whole,

58:30 so,

58:31 but it's not what I

58:32 started here.

58:33 So,

58:33 in this paper,

58:34 what I want to do is to document

58:37 uh these employment changes.

58:39 And then,

58:39 uh,

58:40 in,

58:40 in,

58:40 in,

58:41 uh,

58:41 potentially next paper,

58:42 what I want to explore is how this affected the decision

58:46 uh of individuals to go to college.

58:51 Now,

58:51 I won't stop you,

58:52 but you know,

58:52 it seems to me the same thing in terms of short term versus long term.

58:55 So you're looking at a different frequency.

58:58 But,

58:58 but the,

58:58 the,

58:59 the overall picture is really to look at

59:01 whether

59:02 the shift is,

59:03 is

59:04 immediate because they get out of the labor

59:06 force and then they get education and enter back

59:08 five years from,

59:09 from,

59:09 from them.

59:11 But

59:11 you know there's,

59:12 I don't see a dichotomy between,

59:14 uh,

59:15 uh,

59:16 looking at education and looking at,

59:18 at

59:18 employment.

59:19 It's just a frequency maybe you're looking at rather than,

59:22 than a kind of macro,

59:24 you know,

59:24 government is really looking at the long term trends,

59:26 not looking at.

59:27 A short term reaction to some structural change.

59:31 I see,

59:31 I see.

59:32 Yeah,

59:32 but over this period,

59:33 what I see is that,

59:34 for example,

59:35 the share of total population that uh low educated

59:38 and high educated men represent hasn't changed much.

59:42 So what has changed is that of women.

59:47 Oh,

59:48 yes,

59:48 yes,

59:48 point taken,

59:49 I could,

59:49 uh.

59:51 OK,

59:52 thank you.

59:53 Thanks to you.

59:56 Go ahead,

59:56 Joanna.

59:57 I think you should go ahead and finish without interruptions from here.

1:00:00 OK.

1:00:01 OK,

1:00:02 thank you.

1:00:02 So I,

1:00:03 I will talk to you briefly about the model and uh this

1:00:06 is gonna be a static general equilibrium model in which I will have

1:00:10 on the side of uh production,

1:00:12 I will have a representative firm

1:00:14 that it's gonna be producing uh

1:00:17 A final route via the combination of three different

1:00:22 differentiated tasks,

1:00:24 and each task is gonna have a specific production technology

1:00:28 and we use a specific,

1:00:30 uh,

1:00:31 a

1:00:31 specific skill.

1:00:33 OK,

1:00:33 so this is for the side of the firm.

1:00:35 And then on the side of the individuals,

1:00:37 I will have uh terogeneous individuals in terms of education and gender,

1:00:41 which are both gonna be taken as given

1:00:43 and 3 skills.

1:00:45 And,

1:00:46 uh,

1:00:46 based on all these,

1:00:47 uh,

1:00:48 on the skills,

1:00:49 actually,

1:00:50 individuals will decide to

1:00:52 either

1:00:53 provide their labor

1:00:55 to,

1:00:55 to one of these three tasks or to home production.

1:00:58 So this is the broad picture of the model.

1:01:02 So getting into the production technology more in detail,

1:01:05 uh,

1:01:05 this is the output of task J.

1:01:08 J can be interpersonal,

1:01:10 manual or cognitive.

1:01:11 So this is gonna be,

1:01:13 uh,

1:01:13 the,

1:01:13 the,

1:01:14 the final good production is gonna be out of a CS aggregator of the output in this,

1:01:19 in each of these three tasks,

1:01:21 and the tasks,

1:01:22 uh,

1:01:22 are,

1:01:23 the three tasks are complementary.

1:01:26 And then to produce each of these three tasks,

1:01:29 uh,

1:01:30 I will have,

1:01:31 uh,

1:01:32 interpersonal and cognitive tasks that can be produced

1:01:35 or are gonna be produced linearly by using uh

1:01:39 efficiency units of labor here.

1:01:42 And uh

1:01:44 then the manual task

1:01:45 will be able to be produced using efficiency units of labor

1:01:50 or

1:01:51 capital.

1:01:52 And uh it is gonna have each,

1:01:54 each of these factors is gonna have a,

1:01:57 a specific uh

1:01:59 productivity

1:02:00 and the two factors are imperfectly.

1:02:04 So,

1:02:05 I modeled it this way,

1:02:07 I could have

1:02:08 also added uh capital in interpersonal cognitive.

1:02:12 The only thing that I want to to account for here

1:02:15 is that manual tasks use more capital than the other two.

1:02:22 And,

1:02:23 uh,

1:02:23 in this economy,

1:02:24 capital is supplied perfectly elastically by a pure capitalist.

1:02:29 So,

1:02:30 on the efficiency units of labor,

1:02:33 the uh the on the tasks I was showing you the aggregate,

1:02:36 but this aggregate comes from the combination of uh

1:02:39 low educated uh efficiency units and the,

1:02:43 the ones of the highly educated.

1:02:46 And uh

1:02:50 The ones of the,

1:02:51 of the high and the low educated

1:02:53 are

1:02:55 So,

1:02:55 are,

1:02:55 are coming from,

1:02:57 from exactly the the the sum

1:02:59 within the level of education,

1:03:00 the sum of those

1:03:02 from women and from men,

1:03:03 OK?

1:03:04 From women and from men for the low educated,

1:03:06 the same for the higher educated.

1:03:08 So women's and men's efficiency efficiency units

1:03:11 are perfectly substitutable within a level of education

1:03:15 and uh the low and the high efficiency units,

1:03:19 the low and the high educated,

1:03:20 sorry,

1:03:21 efficiency units are imperfectly stable.

1:03:27 And then this was all on the,

1:03:29 on the production side of the economy,

1:03:31 now on the individual side.

1:03:32 I will have a,

1:03:33 a mass equal to one of individuals in this economy

1:03:37 and each individual is gonna be defined by uh a level of education,

1:03:41 a gender,

1:03:43 e.g.

1:03:44 and a triplet of skills which are gonna be drawn from a

1:03:48 normal distribution,

1:03:50 which is gonna be gender specific.

1:03:54 So,

1:03:55 the only decision that these individuals are gonna take is to sort into tasks

1:04:00 as,

1:04:01 as I

1:04:03 already doing the.

1:04:04 A week before,

1:04:05 these,

1:04:05 these tasks are gonna be the,

1:04:07 the 3 tasks.

1:04:08 So task J,

1:04:09 it's,

1:04:10 it's uh interpersonal manual cognitive.

1:04:12 So basically,

1:04:13 they will be comparing

1:04:14 the,

1:04:15 uh,

1:04:16 the,

1:04:16 the,

1:04:16 the return they would get based on their level

1:04:19 of education and the efficiency units they would get,

1:04:22 the efficiency rates they would get based on the level of education they have

1:04:26 and on their skills,

1:04:28 and then the return to home production is there.

1:04:31 And it's the same for women.

1:04:32 The return on production is delta,

1:04:34 and it's,

1:04:35 uh,

1:04:35 as you can see,

1:04:36 it's uh not gender-specific.

1:04:40 And women,

1:04:41 uh,

1:04:41 it's the same,

1:04:42 the,

1:04:42 the return to task,

1:04:44 it's gonna be determined by the efficiency wage,

1:04:47 depending on the level of education and the,

1:04:49 uh,

1:04:50 and the skills,

1:04:51 but

1:04:52 there's this.

1:04:54 There's this 1 minus kappa.

1:04:56 So what is this 1 minus kappa?

1:04:58 This is a,

1:05:00 this is I,

1:05:00 I,

1:05:01 I.

1:05:02 It's a trick to microfound the uh consistent difference

1:05:06 between uh the wages of men and women in,

1:05:09 in,

1:05:09 uh,

1:05:09 in all these preoccupations.

1:05:13 So,

1:05:13 the,

1:05:13 the gender wage gap,

1:05:14 basically.

1:05:15 So friends,

1:05:16 uh,

1:05:17 it's,

1:05:17 I will micro-found it as,

1:05:18 uh,

1:05:19 or like,

1:05:20 uh,

1:05:21 Yeah.

1:05:23 Sort of micro-funded as the fact that firms face an extra cost when uh they hire.

1:05:29 So you can think about it as discrimination

1:05:31 or

1:05:33 Like a statistical discrimination,

1:05:35 firms anticipate that uh women potentially will exit the

1:05:38 labor market in the future because they will,

1:05:41 they will.

1:05:43 have kids or whatever.

1:05:45 So that might be another explanation or you can think about it as productivity.

1:05:49 You can,

1:05:50 if,

1:05:50 if you don't like.

1:05:52 The story of discrimination,

1:05:53 you can think that women are consistently less productive than men in all

1:05:57 types of jobs.

1:05:58 So this,

1:05:58 uh,

1:05:59 this is what,

1:06:00 uh,

1:06:00 creates this gap.

1:06:03 And

1:06:04 that's,

1:06:05 uh,

1:06:06 that's the whole picture for the individuals and going

1:06:10 quickly to the equilibrium which is defined in a,

1:06:13 in a standard way,

1:06:14 uh,

1:06:15 we'll have

1:06:16 that each individual is gonna sort

1:06:18 into the occupation that maximizes its,

1:06:21 uh,

1:06:21 labor market return,

1:06:23 that firms maximize profits and that all markets.

1:06:27 Nothing.

1:06:29 Nonstand.

1:06:30 So to go to the calibration.

1:06:33 What I will do is to,

1:06:35 to calibrate the model.

1:06:36 I will calibrate,

1:06:38 I will do one initial calibration in 1990

1:06:41 and then I will calibrate the model also in 2016

1:06:44 of the same uh

1:06:46 The same comparison as,

1:06:48 as in the data.

1:06:49 So,

1:06:49 the only,

1:06:50 the,

1:06:50 the exercise I'm gonna do here,

1:06:52 that it's gonna be uh in a sense,

1:06:54 my first counterfactual experiment,

1:06:57 it's that from 1990 to 2016,

1:07:00 I will only allow for changes in technology

1:07:03 and the population,

1:07:04 and the composition of the population.

1:07:07 OK?

1:07:08 So these are the only two changes that will

1:07:09 allow from 1990 to 2016 in the current region.

1:07:13 Let me tell you a bit more about the calibration in 199.

1:07:18 To calibrate the,

1:07:19 the model in 1990.

1:07:21 What I will do is I will take the elasticity of substitution

1:07:25 from uh like externally from other studies.

1:07:27 I will compute the rental rate of capital in 1990 from the data,

1:07:32 and I will normalize some of the parameters.

1:07:35 So the elasticity of substitution,

1:07:37 these are gonna be the ones that I take.

1:07:38 Previous studies,

1:07:39 I will normalize the means

1:07:42 of uh

1:07:44 the mean skills of men.

1:07:46 And uh the one of women in,

1:07:48 in cognitive.

1:07:50 Uh,

1:07:51 this is gonna be my assumption to be able to identify also the

1:07:56 The rest of the skills for men and women and at the same time,

1:07:59 the gender wage gap.

1:08:01 Then the skills correlation,

1:08:03 I will set it to 0.3 and uh the capital productivity and normalize it to 1,

1:08:09 and this is the rental rate of capital that I estimate.

1:08:13 No.

1:08:15 And then the population composition,

1:08:16 I also feed it in from what I observed in the data.

1:08:22 So these are the shares of each of these groups of population.

1:08:25 And then I calibrate the 16 parameters that are left to match 16 targeted more.

1:08:30 So these are gonna be related to the skills distribution,

1:08:34 to the home production,

1:08:35 to the,

1:08:35 the parameter that relates to the gender wage gap,

1:08:37 and the one that relates to technology.

1:08:41 So these are the targeted moments that

1:08:42 I'm gonna use that are either gender-specific

1:08:47 or education-specific,

1:08:48 but never gender education-specific.

1:08:53 And the calibrated parameters

1:08:56 uh

1:08:57 that I get are these ones.

1:08:59 So,

1:09:00 I have

1:09:01 the women I will,

1:09:02 uh,

1:09:02 from the calibration exercise,

1:09:03 women will have more interpersonal skills,

1:09:06 less manual skills than men,

1:09:08 differences in the variants.

1:09:10 So there is a bit of uh heterogenic in terms of skills that the model predicts,

1:09:14 librated model predicts between men and women.

1:09:16 That's the picture for the calibration is 1990.

1:09:21 And then from,

1:09:22 again,

1:09:22 from 1990 to 2016,

1:09:24 only technology

1:09:26 and population composition changes are allowed.

1:09:29 So the population composition changes,

1:09:32 I take them from the data.

1:09:33 Mainly they account for the fact that

1:09:36 we have more

1:09:37 IV educated women

1:09:38 in the share of total population now.

1:09:42 And the rental rate of capital I completed

1:09:44 for 2016 and then I calibrate the technology parameters

1:09:49 uh to match uh the seven technology parameters to match these seven ones.

1:09:55 These are the ones that I used in 1990 and here you have the ones

1:09:59 in red that I used for 2016.

1:10:04 So the rest of the parameters stay as in 99.

1:10:07 And I don't have time to show you.

1:10:10 But,

1:10:10 uh,

1:10:11 the model does a good job

1:10:14 also in not only in the targeted moment because that's by construction,

1:10:17 but in the non-targeted moment.

1:10:19 So,

1:10:19 as I said,

1:10:20 nothing at the gender level of education specifically

1:10:23 is targeted,

1:10:25 but still the employment allocation and the average wages at this level,

1:10:29 the model is able to uh

1:10:31 to uh

1:10:33 uh match very well with respect to the data.

1:10:36 So this is,

1:10:38 is one example of what's not targeted.

1:10:40 This is the employment allocation of low educated men and women,

1:10:44 and the changes between,

1:10:45 uh,

1:10:45 so these are the changes between 1990 and 2016.

1:10:50 And um

1:10:52 so you have here what's happening in the data

1:10:55 uh for men and women in each of the occupations

1:10:59 and what the model predicts.

1:11:02 And you can see that the model can predict

1:11:04 the uh

1:11:05 the

1:11:06 labor reallocation that we've seen

1:11:08 very well,

1:11:09 only with the technology and population.

1:11:14 And now

1:11:17 The first counterfactual or the second,

1:11:19 in a sense,

1:11:20 the second counterfactual I do is to try to understand

1:11:23 what is behind,

1:11:24 so what is it?

1:11:25 Is it technology?

1:11:26 Is it the changes in the,

1:11:27 in the composition of the population?

1:11:30 So I want to understand the separate contribution of this uh

1:11:33 tool.

1:11:35 So let me show you first what happens when we only,

1:11:38 so this is the same table as before,

1:11:40 I was showing you with the changes in employment over population ratios.

1:11:44 So this is what happens when we only have

1:11:48 Changes in the population composition.

1:11:50 And you see

1:11:51 that all employment rates would have uh

1:11:54 gone uh up.

1:11:56 This is from production going down.

1:11:58 And particularly,

1:11:59 it will have gone up for women in interpersonal populations.

1:12:04 Uh,

1:12:05 so this means,

1:12:06 uh,

1:12:07 that,

1:12:08 uh,

1:12:09 In a sense,

1:12:10 if we only would,

1:12:11 would have had a population composition changes,

1:12:14 the pre-train,

1:12:15 uh,

1:12:16 the pre-1990 trend in uh female employment would have continued.

1:12:22 And so this tells us that it's uh technology

1:12:25 directly the main suspect behind these changes and indeed,

1:12:28 this is what we see.

1:12:29 When we only allow for technology from 1990 to 2016,

1:12:35 We see that the uh negative,

1:12:37 the aggregate negative effect is of the same magnitude for women,

1:12:42 for women than what is it for men.

1:12:44 And this is despite the decline in manual employment

1:12:48 being much more important for men than what is it for women.

1:12:52 So these men

1:12:53 are gonna be

1:12:54 The model predicts that they relocate

1:12:57 into,

1:12:58 uh,

1:12:58 into the other

1:13:00 two types of occupations while women decrease substantially,

1:13:03 their presence also there

1:13:05 due to this increase in the,

1:13:07 this,

1:13:07 this rise in the competition.

1:13:10 So all in all,

1:13:10 the negative effect is of the same magnitude for women than for men.

1:13:15 Uh,

1:13:15 so indeed it's a bit higher for them.

1:13:18 So this,

1:13:18 uh,

1:13:20 This tells us that if only when,

1:13:22 in a sense if only we would have had technology changes,

1:13:26 the crowding out would have been stronger

1:13:28 and the decline of total employment would have been

1:13:30 even stronger than what we have observed for one.

1:13:34 So just to

1:13:36 take stock of what

1:13:37 I told you

1:13:39 and

1:13:40 Um,

1:13:41 changes in technology and the population composition,

1:13:43 the model tells us that can replicate,

1:13:45 uh,

1:13:45 the labor market reallocation.

1:13:47 We've observed in the data,

1:13:49 and,

1:13:49 uh,

1:13:50 it seems that technology changes are behind the workers'

1:13:52 reallocation and the crowding out of low educated women,

1:13:55 while population composition changes have masked the

1:13:58 effects of technology on the labor.

1:14:03 So I don't have time to show you these,

1:14:05 uh,

1:14:06 these counterfactuals when I check.

1:14:09 Uh,

1:14:09 well,

1:14:09 I explored further the mechanism and I closed the

1:14:12 parameter related to the gender wage gap and I erase

1:14:15 the,

1:14:15 uh,

1:14:16 the gender differences in skills.

1:14:19 So I will conclude quickly.

1:14:22 So my paper links the decline in uh manualemployment

1:14:25 and the stagnation of female uh labor market participation.

1:14:29 And uh it,

1:14:31 it shows

1:14:32 that

1:14:33 structural transformation might not benefit women's employment in

1:14:37 uh today as it did in the past.

1:14:40 And,

1:14:40 uh,

1:14:42 productivity and population composition changes we learn

1:14:45 from the model can explain the level of reallocation we've seen in the data,

1:14:50 and productivity increases are in manual occupations are behind these changes.

1:14:55 And uh the effects of productivity have been dampened by

1:14:59 the increasing share of women attending college over this period.

1:15:03 And this is important uh for policy because if we think

1:15:06 about like policies to or the design of policies to,

1:15:10 to get to full employment,

1:15:11 like,

1:15:12 uh,

1:15:12 the design of retraining programs,

1:15:14 we need to keep that,

1:15:15 uh,

1:15:16 these,

1:15:16 uh,

1:15:17 these changes in mind also for the,

1:15:19 for the potentially designing uh social protection programs

1:15:22 uh to help these individuals that are now jobless,

1:15:25 uh,

1:15:26 this,

1:15:27 this is important

1:15:29 to keep in mind.

1:15:30 So,

1:15:31 thank you very much for your time and your questions and comments.

showAllTimestamps

no

transcript

OK. Um, it's 10:31 and I see that we have more than 40 people in the seminar already. So it's a good time to start. Uh, we're towards the end of our winter seminar series, everybody. This is the penultimate seminar. There's one more tomorrow. Thanks everybody for being so engaged in coming to all the seminars. It's been wonderful. And today we have another exciting, uh, speaker. Uh, Joanna is from Toulouse School of Economics, and she's going to talk about the blurring of gender-based occupational segregation. Uh, Joanna, we talked about the rules, and everybody here knows the rules as well, so the floor is yours. Thank you very much, Burke, for the introduction and thank you very much for the opportunity to talk to you today about my supermarket paper which as the title suggests, studies what happened to women and in particularly what happened to women's employment in the United States in recent decades. Uh, let me start by showing you a graph about the uh women's employment overpopulation ratio from the late 70s until 2019. So as you can see, during the 20th century or the last part of the 20th century, Uh, women's employment increased by a striking amount. So these were more than 20% points and to put it in words of, uh, of Claudia Goldin, this was the most significant change in the labor market during the past century. However, Since the beginning of the 21st century, we've seen that there's been no progress in uh female employment. And when we check the same, uh, the same measure, this aggregated by level of education, so by subgroup, depending on the level of education. We see in the, this is, again, employment overpopulation ratio. The solid line is highly educated women, the dash line is uh women, women with at most a high school degree. We see that low educated women not only uh didn't see any progress on their, on their employment rates, but indeed they, they saw a decline over this period. And this is even more fascinating because at the same time it has happened, the type of jobs these low educated women have been traditionally performing, which are service and clerical occupations, have seen employment growth. So specifically, in this paper, I asked, why is it that low educated women's employment has decreased after decades of circular growth and despite growing employment opportunities? And what I'm gonna argue in the paper is that it's because of the decline in local unemployment of low educated men. And more specifically, what I mean by that, or what I'm, what I'm gonna argue more in detail is that manufacturing and construction jobs over this period have seen an important uh job, uh, employment destruction, and these were the type of jobs that were mostly employing low educated men. So this decline has impacted mostly low educated men. And these low educated men, after being left with no other labor market options, they've reconverted themselves into service and clerical occupations and due to This increase in competition, they've put downward pressure on wages and they've crowded out low educated women of these occupations, of service and clerical occupations. But given that these are important sources or were the most important source of, uh, of women's employment, also out of the labor market as a whole. So I'm gonna be arguing this by bringing new and critical evidence, and particularly I will be using an idea approach and also a structural model which will help us uh think through the mechanism that is driving this, uh this labor reallocation. So, going to the empirical evidence in detail, what I will be documenting is that there's been, uh, there's been from 1990 to 2016, following the decline in in blue-collar employment, a reallocation of low educated employment from blue-collar type of jobs to work service and clerical occupation. I also find that most of this reallocation is coming from local men. Just the ratio of low educated women to men in these type of jobs, it's the one that has seen the, the, the most of the, most of the decrease or the, the, the, the most important decline. And I also find when I exploit variation across local labor markets in the US that these two facts that I just mentioned are stronger in labor markets that have seen a stronger decline in blue-collar employment. However, to make, to make a council argument, what I will do is use an idea, an idea strategy in which this is gonna be a achieve share sort of instrument in style of what O and RD in which I will, uh, I will have, uh, instead of using the decline in model unemployment per se, I will have, I will exploit the historical industry, the historical local industry structure in that local labor market, which is gonna give me a predicted decline in blue-collar employment in subsequent decades. And uh following the, the results of these, uh these idea estimates, what I will find is that the decline in blue-collar employment Can account for 62% of the variation that I find in the decline of women in low educated employment across local labor markets. So it means that these, uh, these stories are primary candidate for the changes in, in, uh, in, in employment rates of women that we've seen over this period. And then, after documenting these empirical facts, to understand why they have taken place, I will build a general equilibrium model in which individuals who differ by gender and level of education are gonna sort into occupations. And the question I will try to address with this model is how productivity and changes in the composition of the population have contributed towards this uh workers reallocation that we observed in the data. And what I will find is that this, the, the, the worker reallocation that we observed in the data can be fully accounted by productivity and productivity changes and changes in the, in the composition of the population, which will mostly account for the fact that over this period, more women are attending college education. And then, This, uh, this exercise or, or the model is gonna also Uh, tell us that low educated women's employment, if there wouldn't have been any productivity changes over this period, would have continued growing at the pre-1990 trend. So, it means that uh productivity is the main driver behind this, uh, the, the, the behind these uh changes, uh, that, that I've been telling you about. And instead, without population composition changes, we would have seen a stronger Decline in in female labor market participation over this period. So in a sense, the, the fact that over this period, more women are going to college and are attaining higher education uh has partially masked the effects of productivity changes in the labor market. That's it, let me give you an overview of how the paper sits in the literature. So there is the, there is, uh, it links to these two strands of the literature, one which studies the, the decline in blue collar employment and the consequences this has had, particularly for low educated men or on the aggregate. These, uh, these, these papers, they all focus either on, on low educated men, on the aggregate. So what I'm gonna be doing is bringing the, what happened to women. AIM Uh, there's also the literature of, uh, of that studies, the women's employment stagnation, and, uh, uh, the, the first one, in, in the list, uh, documents these changes and, and sees that an important, uh, contributor to the stagnation is indeed the dynamics that are, that are coming from the least educated woman, as I showed you. And then the other ones bring different explanations for why they have happened. And these last two, they study the, the macro consequences of these, of these changes for, for, of, of, of women's. So basically my paper contributes to the literature in that it links these two strands of, so the decline in local unemployment and the stagnation of, uh, of women's uh labor market participation. And then it also contributes to a third strand of the literature which studies uh how structural transformation affects women's labor market participation. And all these papers have found that the movement from of economic activity from manufacturing to our services were beneficial for women's employment in the United States over, over this during the 20th century. But what I show is that structural transformation during the 21st century might not benefit women's employment. So, oops, that's said, uh, here is the outline of the talk. I'm gonna be showing you the data, then the empirical evidence. I'll move on to the model and the calibration, and then I will tell you about the backs. And after that, I will conclude. So if there is any question, now, I will be happy to take it. Yeah, there's at least one, Eshani. Sure. Hi, thanks. Um, so I'm wondering, you know, with this sort of sectoral reallocation story, um, you might expect to see impacts on women's education choices, for instance. Um, you might expect to see whether lagging or leading, I don't know, I don't have a strong pride, but you might expect to see impacts on fertility choices. Are you seeing those and are you gonna tell us more, um, about how to think about The fact that women are, you know, making some of these choices about labor force participation in this broader context of, of, of things like how much schooling to invest in, how many children to have, um, and, and simultaneously, you know, what's happening to things like childcare availability. Sure. So these are all questions that I would love to explore. So for now, the educational decision, I, I, I don't have it endogenous in the model. I, I don't think, uh, It's, it's not the purpose of this paper to explain uh what has happened to uh women's education and how this is linked to the manual decline, but given what I observed, it's happening in the labor market, I believe it's an important uh margin of adjustment that women have. Also, what the counterfactual exercise showed. Uh, so I would love to explore this in, in future research. And then for the fertility, I will be controlling for, for The, uh, share of women with uh kids younger than, uh, a certain age in the regression, so that will more or less be accounted there, but I will be happy to explore it further. And related to the, uh, to the family-friendly policies you were referring to. Uh, I think it might be linked also to the, these changes and, and, and the fact that there is plenty of labor supply available for this type of service and clerical occupations. I think it could potentially have a link with the lack of family-friendly support that we have observed in the United States compared to other countries. So these are all questions that I would like to explore in the future in more detail. I can just jump in for one second, Brick, um, just, you know, I guess I was even saying that I would have loved to see that as a stylized fact as, uh, corroborating, you know, your hypothesis, um, not even, you know, going into the model, uh, we can get to that later, but yeah, just even showing empirically, um, that this is, this is something we're observing, that these are changes that are consistent with the sectoral reallocation story you're telling, that would have been helpful. Sure. So, yes, I don't have them because that's, again, that's not the main focus of the paper, but yes, I can tell you that over this period, uh, the, uh, educational attainment of women is, uh, is increasing, is, uh, is increasing and it's above, uh, the one of men. So thank you for that, for pointing that out. Thank you, Joanna. Paulo, you're next. Hi, Joanna. Uh, so in, in the stylized facts that, that you've shown, uh, you see that the decline seems to coincide with the global financial crisis and a period of low aggregate demand and also the shocks that you'll be exploiting are negative shocks to local labor markets. So I was wondering if you can distinguish, uh, the, the negative adverse effect of these shocks from the reallocation and competition mechanism that, that you're highlighting. So, what I will be exploring is, uh, is the historical local industry structure. So, yes, most of these, uh, most of these changes though in, in the, in the decline in, in the manual employment that I'm referring to in the paper happened or yes, are, are, are taking place in the middle of the Great Recession. Uh, but this has been well documented in the literature in, in, in the paper of, uh, Jane and, and, I think was it that they found that most of the declining routine type of jobs happens over, over recessions. Yes. This is, uh, this is what I will exploit. This is the, the, the changes I exploit on, on the app. OK, reach up. Yes. Hi, thanks. This is actually related to what Paulo uh asked. So, as I see, there are two very distinct trends, and I was uh trying to understand that whether you'll be talking about two points in time, or would you have some stories as to the initial decline that happened from 2000 to 2010, somewhere around that period. And then there's a reversal from that point onward. It's just that the reversal has not uh caught up with the earlier levels. Um, in terms of your exogenous factors, I would see that, uh, productivity might explain these two, subperiod trends because they are volatile in general, but in terms of, uh, full, uh, population densities, they are not, I mean, my guess is that they won't have this stark reversal as compared to productivity, for instance. So would you be, uh, comparing across time, so you will also have something to say about the subs. Thank you. So for now, the model, I compared two time periods, so it's gonna be 1990 and 2016, but I have, uh, I, I, I can, uh, I, I plan to, uh, to do it by decade in the future and I think the model will, will be able to, uh, just explain, uh, the decade, the decadal changes from what I get now from, from, from the two time periods. I, I believe that. And uh I'm gonna show you the US national level, some, uh, some variation by decades. So, so yes. But yeah, point taken. Thank you, Joanna. Thank you. Um, Clement, you had your hand up, then it went down. Was that intentional or you don't want to ask a question? Yes, yes, my question, uh, has been answered. Thank you. Uh, great. Thanks, Clement. So, uh, you're the last before we continue. Thank you. Can you hear me? Yes. OK. Yes. So, your study remind me of the uh of the literature in trade when we talk about. Technology induced, technology-induced, uh, wage differential compared skilled to non-skilled workers as well as well as outsourcing induced trade group um. Which differentiate between skilled versus non-skilled workers. And the period that you are studying coincide with the rise of China entering WTO. So there's a lot of global macro forces that is happening at the same period of time that you are looking at this. There's technology, there is outsourcing, there's rise of China. So will you be able to distinguish all this? From the local labor market conditions that you are trying to pinpoint. No, but at the end of the day, these changes, they all lead to the same, uh, uh, to the same consequence that it's the, uh, that the number of, of, of, of jobs available in, uh, in this type of more manual jobs is decreasing. So I'm, I'm not too much interested on what caused that, but more on the consequences this, uh, this decline is having on the labor market and particularly on, on the labor market, uh, opportunities of women. So in, in the, in the model, I will be, I will be using technology to generate these increases in productivity that are gonna be triggering this labor reallocation, but I could potentially, I will be more Explicit about it later, maybe we will, it will be making more sense after you see the model that we discuss more in detail. But I could just, uh, as long as the trade, it happens within the what I define as the manual uh task. Maybe I'm overestimating the contribution of, of technology, but again, it would lead these, these increases in, in trade in the, in the manual, in, in, in the manual task would lead to an increase in the productivity. And at the end, it's this increase in productivity what I care about. So, yes. But yes, I agree that there's, there's many things that are driving the decline in, in manual type of jobs or in manufacturing, construction, and all these type of things. But uh for now I'll just take no stance. OK, good. I see that she is happy. She's giving a thumbs up. So, and that was the last question. So Joanna, you may continue. OK, thank you very much. So, um, to, to get to the data a little bit, uh, to show you what are the, what are the data sources I use, so I do use, uh, the Census on the American Community Survey for the employment information, and then I'll be using the current population survey, the March supplement for the wage information. And I will also be using the OET skills level data uh to get information on what are the skills that are being used in all these different types of occupations. So with this, with this, uh, data set, With this data set, I will, I think it's. Um, with this data set, what I will do is to categorize occupations based on the type of skills they use, OK? And these are gonna be, uh, so, so, so to do that, what I will do is to use the, the original data which contains 46 different skills requirements. And uh these are gonna be uh skills requirements like service orientation, critical thinking, equipment maintenance, all these type of, uh, type of skills. And what I will do is to use fact analysis and use these 46 different skills requirements to 3 statistically significant factors. And these factors can be naturally interpreted as interpersonal, cognitive, and manual skills. And this categorization has been previously used in the, in the literature into other papers, so it's not that I'm doing anything crazy here. And doing, and after doing that, after having now for each occupation, how uh how important these three factors are, I will classify each of the occupations based on the factor that is the most important. And when I do that, the uh occupation classification that I end up with is the following. So, occupations within uh clerical and retail sales and service, uh, are the ones that use interpersonal skills the most. So I'm gonna be calling them interpersonal occupations. Then managerial and other types of professional occupations are the ones that are mostly cognitive, so these are gonna be my cognitive occup. Patients, and then production, installation, transportation, and construction. This use mostly manual skills and these are gonna be my manual occupations. So I'll be carrying uh on these uh uh skills categorization along the, uh, along the, uh, the empirical analysis and, and also the model, and now to, to go to the empirical analysis. So, what, what I want to see is what has happened to employment with uh through the lenses of these three different occupations. So I see that the, this is the share of total employment. Uh, in the, in the, for the whole population in this table in 1990 and 2016 for each of the three occupations I just mentioned. And what we see is that the share of total employment in manual occupations has decreased substantially over this period. And when I check who's been the most affected by this uh decline in in manual employment, so here you have in this table, the change in employment over population ratios in manual occupations disaggregated by gender and by level of education. For the same years. So this is just a difference. So we see these are men, low educated men in the first column, high educated men here. So we see that it's uh Mostly low educated men when compared to high educated men, the ones that have suffered the most, and it's the same for women. So we see that it's the ones that are suffering the most from the manual decline in terms of employment are the low educated and mostly low educated. So checking again, the same, the share of total employment, of total employment but this time only for the low educated for, because are the ones that more, that are more affected by the manual employment decline. So this is again in 1990 and 2016. I want to understand what has happened to the, to the employment, to the shares of employment in, in the three occupations, and what I see is that there's been uh a reallocation of low educated employment. Towards interpersonal occupations. So from manual occupations towards interpersonal, while cognitive, uh, jobs have stayed more or less constant. And then when checking the, the gender composition in this, uh, in these occupation. So these are, this is the ratio of low educated women to men employed in the, in the three types of occupations. So in yellow, you have interpersonal, in blue, you have cognitive, in, uh, in, uh, gray, you have manual. I see that the decline in, in, in the ratio of women to men employed has been the strongest in uh in interpersonal occupations, which tells us that the reallocation of labor towards interpersonal occupation is mostly driven by men. However, everything that I've been telling you here, it's through the lenses of shares. And as, as you were already mentioning in the questions, uh, there's been an important, so, so there's been important changes in, in the, in, in terms of population groups over this period. More women are attending college education and all this. So it might be that uh some of these effects are driven by that. But when I check employment over population ratios, so this is controlling for the size of this, uh, Of these uh population groups. Uh, the, the results I get are exactly the same. So let me tell you what there is in this, uh, in this graph. We have here the changes in employment over population ratios by occupation. These are the three boxes that you see here, and the by gender. So this is for the low educated, uh, men are in blue, women are in red. And when this is is aggregated by decades, so the first two columns referred in all the three boxes refer to the 1980s. The second 1990s, 2000s, and uh the, the, the recovery from the Greek recession, 2010 to 2016. So what we see here. Is that focusing on the 1980s, you see that women increase their, their labor market presence or, or their employment rates in all three types of occupations. So this is what I refer to as the golden age of uh women's employment because these are still the uh the strong growth that happened during the 20th century. When we check for the 1990s, You already see that there is a decline ongoing in, in cognitive, in manual that is affecting women as well, but still they have growth in interpersonal. And uh then there's the, there's the, the Great Recession mostly that arise, but the, the decline in manual is also present before. And you have this huge decline in employment uh rates both for men or particularly for men, but also for women in manual occupations. And what we see during this decade as well is that men are increasing their presence in interpersonal occupations much more than women. It's not still the case for cognitive. But then during the period of the great recovery, the Great Recession recovery, you see that the patterns are completely reversed now in all the occupations. So interpersonal jobs are being taken by men. Cognitive jobs, uh, men are growing more than women, and in manual occupations, there's the, there's growth for men but not so much for women. So this is the, this is the message that I want to convey that it's that we moved from the 1980s where women were finding most or employment opportunities in all types of jobs to uh to a context in which these low educated women are like not even able to hold their jobs in the type of occupations or, or not even able to find employment opportunities in the type of occupations they've been traditionally. And this is at, at, uh, at the same time that men do, so. So this is, uh, this is the, the, the main message here. But then still it might be, so maybe I can take some questions now before I, I, I get into the, uh, the analysis across local labor markets, if there's any. Yeah, I see a couple of hands. Pierre, why don't you start? Uh, could, could you go back on the graph maybe for the question that would be helpful. This one, yeah, yeah. So thank you. So, so, so what, you know, the way you presented it, I guess is fine, but one way to think of what's happened in 2010 is you're getting this recovery of lost employment that's very obvious on the manual graph, right, in the 2010, and then you see that the, the drop for men is huge, right, even compared to that of women, and in a way those men have to go somewhere post, right. And so if we're, you know, kind of dividing by the jobs loss in total in the recession period. Uh You would that change a bit the narrative and the picture? Do you see what I mean? You mean jobs lost, uh. So the, the, you, you would put it in, in terms of uh Like in, in, in, in shares of employment. Yes, but I guess you're trying to make a point that I guess, and maybe it's true that men are taking jobs that were previously for women, but it's also because their job is so, you know, it's so huge in the pre-period, right, and the jobs kind of have to go somewhere. And I'm wondering kind of by looking at total employment, you know, is that, is that the right message is kind of the, the recovery compared to the losses, for example, that are going in each sector, right? And then you'd be dividing by kind of maybe the loss in the 2010, in the 2000s, you know. But maybe it's just a better way of presenting and just it, it, it could be, uh, it, it, it, it could be that this, yeah, this, uh, the decline in manual for men is much more striking than what is it for women, but these men, like what we see is that women are not like, uh. Entering back into the labor market for, for, for the recovery of the Great Recession. While for men, it looks that there's a bit of recovery. So why is it that these men are getting back into the labor market? There's a bit of growth in manual, but most of the growth uh is coming from, from the fact that they are increasing their presence also in other types of. Jobs that were not traditionally the ones they were doing. So this is, uh, this is the story, uh, I, I, that's exactly my point, Joanna, which is if you add the recoveries in the 2010s, right, from those three little bars, you still only have a fraction of the loss that you observe in manual in 2000s, right, even for men. And so in a way that fraction to me seems fairly similar in ratio to the one that if you were adding the gains for women in the 2010s uh compared to the loss in the 2000s. Maybe I'm wrong because I haven't, you know, done the ratio exactly, but you see, so, so that changes a little bit of the narrative. It could be that just recovery was poor overall for men and for women. Uh, versus just kind of men recovered and women didn't. So no, I'm not sure I'm seeing that, you know, that message you're saying from the aircraft. I, I don't have, I don't have the, uh, I took it out finally from the sleds because they were too long, but I should have put it, so I have the uh. The slides with the evolution of of employment rates for like level of education and and and gender and then you see directly there that men uh they they recover from the Great Recession while women don't. OK. So this is, uh, but yes, you, you, you're right I should maybe try to put it in, in relative terms. But also what we have here is that uh women are going more into college. So, so over this period here. So it must mean that the decline in employment is much stronger than, than the, than the decline in, in, in, in, in the population for this to go down because the population of low educated women is decreasing over this period. But yes, point taken. Thanks. Thank you, Joanna Chala. Uh, hi, Johanna. So two questions. One is, do we get anything different if you look at it by age groups, by cohorts, right? Because especially the older versus the younger woman, uh, it was asked before about, especially in terms of fertility behavior, do we see anything there? And the second thing. The, the big story, I mean, it's going to be the obvious question, but the, the migration behaviors changed considerably, right? The, the migration from Central America and Mexico dried up because of their demographic transition, which was very strong in the 90s and even in the early 2000s before the financial crisis, and that dried up, and that's also an endogenous behavior there, which affected low educated men more than women. And there's, you know, really complicated dynamics going on. Does it change the story there if you introduce it? Thank you. So, thanks to you for the question. So by age, uh, Uh, I disaggregated, I don't have the, the, the graphs here, but I disaggregated the, uh, the, the evolution of employment rates by, by age bracket. So these are prime age individuals. Everything that I'm telling you about today is for prime age individuals, so people from 25 to 54 years old. Uh, so when I did from 25 to 35 to 35 to 45, and, uh, on the last age bracket, they all seem to have, all, all, all the women seem to have the same pattern. So it seems that there's nothing uh age-specific there. And uh for migration, yes, I agree that uh They, yes, that maybe migration is, is uh It's an important margin of adjustment, but this again is over total population of these subgroups. So it's something, it will be already Uh, already accounted in, in here and in the, in the regressions in, in, in the, in the results I'm gonna show you later, you know, from the regression part of the analysis, I will, I will show you how like migration is something, it's, uh, it's affecting me like it would, uh, make me underestimate the results I'm getting. So it's potentially. Yeah, sorry. Sorry, is this total population or native born population, the figures you show it's, it's. OK. It's all the population in, in the United States. OK. Uh, OK. Thank you. Thanks to you. Good, thanks. And finally, Sergio. Hey, John, um, uh, it's interesting, an interesting presentation, and I will, I, I'm sympathetic with the type of analysis that you are making. Um, my question, and I believe that the story is consistent with the story that you are telling, um, but I wonder whether you are pushing a little bit too much this displacement, uh, story because, uh, you don't have evidence that it's actually the men that are living one and then they are really displacing the women. I mean, there could be other factors going on. It could be women for some other reason, uh, withdrawing from that market. So, I would be a little bit kind of more cautious about how much you push that story, even though it's consistent, but you don't have any evidence to, to back up your claim completely. OK. So, what I will, in the, in the, in here, these are all aggregate uh facts, and I, I don't make the capital claims out of these aggregate facts, but more about the uh IV regressions that I'm gonna show you now, uh, because uh there I can control for like other demographics and uh like. Other like variables that might be influencing this uh labor market behavior of, of men, and I will account for marriage, uh, uh, share of, of women with uh kids younger than a certain age that these are all drivers of the, of the, uh, Of the decision to, to participate in the labor market. So I would try to do my best to account for all these other explanations. But I, I wanted to, to claim causality, you might need to observe the actual demand and supply of labor in different markets to really observe that whether the IV would give you enough, uh, ammunition to claim causality or whether you might need other variables that you might not have. Point taken. So, I'm, I'm doing my best with the, with the information I have available. So, the IB I can, I can show you more about it, uh, maybe, and maybe we can discuss in case uh you still have concerns. OK, thank you. Thanks. Great. Uh, that was the last question, Joanna, so you may continue. Thank you. So, as I was saying, these are all the facts that I document at the US national level, but now, like it might be, uh, it might be that they are happening simultaneously, but they are not linked. So to establish a causal link with the decline in manual employment and the reallocation of low educated employment towards interpersonal jobs and the changes in the gender composition, low educated employment in these jobs and in the uh In the labor market as a whole, what I will do is to explore variation across local labor markets. And these, uh, I will proxy these local labor markets as the, as commuting zones in, in, uh, as, as author and Dorn and Hanson did in the 2013 paper, that's the one where they studied the China shop. So The type of uh of uh regressions that will run. So these are, I mean, I'm gonna regress the initial share of manual employment in computing zone J states at the beginning of the decade, uh uh T0, OK? On the decadal changes, I will pull them all together on the interpersonal share of low educated employment. In, in, in uh communion zone G, state A, uh, during the decade. And then I will have, so this is gonna be for the first part for the reallocation of low educated employment. So I will have state fixed effects that are gonna capture anything related to regulation that it's state-specific. I'm gonna have um the fixed effects and then some controls for the economic and, and, uh, and demographic characteristics of these local labor markets. And uh maybe you are wondering why I'm using the, the share of the initial share of manual employment here and not the manual decline or the, the decline in the manual share per se. And it's because since I'm checking shares here and shares here, even if the populations are not the same because in here I have all the population. And this is only for the low educated. If it might be that by construction, if I use the change here and this is, this is uh decline, this is declining in manual occupations, it should increase automatically in other types of jobs. So to avoid that, that's why I use the initial share of manual employment here, which is, uh, which is predicting, uh, which is a strong or, or like a good predictor of the subsequent decline. So this is for the first fact, which is about the reallocation of low educated employment towards interpersonal jobs, and then, I, uh, I checked in the, the, the same sort of, uh, regression design, the, uh, the change in the share of women among low educated employment in interpersonal occupations and also in, in the, in all occupations, so in the labor market as well. So these are gonna be two separate sets of regressions. And again, I use the initial share of manual. And I will have some controls, so I, I will have again state fixed effects and, um, uh, decade fixed effects and some controls that are gonna be, uh, controlling for potentially diff uh things that might be affecting the, the The participation in, in the labor market decision of women and men different. So these are gonna be having kids below the age of 5, Being married or not and, and all, all, all these, all these things. So, the, the effect I'm interested on then is this beta one in, in both regressions, which is uh gonna tell me. By how the, by how much more uh uh uh uh a commuting zone that initially had a higher share of manual employment, so let's say a commuting zone with 1% point higher share of manual employment initially. How much more, this is gonna be the beta one, has seen uh a change in the dependent variable. So either in the, in the share of low educated employment in interpersonal occupations or in the share of women in low educated employment in interpersonal occupations or in all the labor market. So, are, are there any questions here? I don't see any hands so far. OK, so, Otherwise, I can just show you so. Instead, so still. If uh If there is anything left on, on the error term that it's correlated with the initial share of manual employment in, in each of these uh local labor markets, and it also affects uh the, the, these uh two dependent variables, we might have potential endogeneity biases. So this might be uh other current demand or supply shocks that are happening at the same time and that correlate with the, with the initial share of manual-employment. So to avoid that, what I will do is to use a shift sharing instrument, uh, like the one author and Dorn used in the, in the, in the paper in 2013. And this will give me uh the manual share of employment that is predicted by the local industry structure in 1980. So, I start my, my analysis in 1990, so this This is gonna give me exogenous variation, and it's exogenous because it predates all these uh other potential, uh, contemporaneous, uh, changes that are might be affecting these uh labor market decisions. And it's also very well, it's strongly correlated with the With the uh with the share of manual employment at the beginning of the day. So all in all, I think it's a good instrument. To use and uh so I can show you the results that I get. So this is for the first question. So the one where I check the reallocation of labor towards interpersonal occupations, and here we have the The, the beta one estimates for, for, for the first set of regressions. And what I find is that uh commuting zone, so this is a, sorry. The, I find that uh accompanying zone with an initial share of manual employment that is uh higher has seen a stronger reallocation of low educated employment towards interpersonal jobs. This, uh, commuting zone that we an initially higher manual employment is the one that has seen a stronger decline in manual employment. So, it's, it's, uh, it's, it's what, uh, what my story was, was, uh, what, what, what the aggregate, uh, facts were hinting at, and indeed I find that these are linked. So to interpret the results, I find that by comparing a commuting zone at the 75th percentile of the initial share of manual employment with one of the, uh, at the 25th percentile, the, this one, that is the one that will suffer the stronger manual decline has seen a 1% point larger increase per decade in the share of interpersonal low educated. So this is for the, this is for the first set of questions and, and, and this represents 14% of the variation across computing zones in the change in the share of interpersonal low educated. And uh so for the second set of regressions, so I have here for the interpersonal occupations, yeah. Excuse me, there's a question I suspect it might be about the IV. So before you continue with the results, let's just take that, sure. Hi, um, so, you know, since, since the author work, you know, there's been a lot of new work on the shift share instruments and, um, you know, some of the work by Jaeger and others have noted with the, the dynamics here matter a lot. So in the sort of context, you're looking at, you know, there's adjust, there's shocks happening all the time, there's adjustment happening to previous shocks. And so, you know, when you're looking here, the shear here is not just affecting. You know, what's going on now, but it affected what was going on beforehand and so, you know, choosing an arbitrary starting point and not building in those dynamics seems like, You know, from the newer work is not going to give you the causal impact. I don't know, sort of, you know, when you've looked at that, that newer work, have you, do you think the sort of auto approach still holds up, or, or how should we think about this? Should we just think of this again more as on a descriptive basis, this has happened and the IV is giving us slightly moving in that direction, or how do you want us to think about that? Yes, maybe it's not, uh, it's not what we would like. So the base would be if, uh, to, to, uh, to have access to administrative data and really be able to tell, like, OK, to, to, to be able to build the whole picture, but this administrative data, the problem they have is that they have limited sample size. So, then, uh, Balancing one and the other, this is the best way I could find to make uh, with, with the data I have to make a, a case for, for my results, even for, for, for my, for my, sorry, for my hypothesis. And I, uh, potentially it's not ideal, but this is the best I can do for now and I will be happy to explore uh better ways in the future to, to, uh, Clean up, uh, what, what might be left that it's uh still there and it's not, uh, it's potentially affecting the, the results. So thank you for your comment. OK, you may continue. Thanks So as I was saying, this referred to the first, uh, set of progressions, and these ones refer to the, the changes in the share of women among global educated women. So these two columns show uh the results for the interpersonal occupations, and these two columns, the, the, these last two columns, it's for the labor market as a whole. So when we check interpersonal occupations, uh, The IV estimates show that uh commuting zones where the share of manualemployment was initially higher have seen a stronger decline in the share of women among the low educated employed in this occupation. And again, by comparing the, uh, these two commuting zones, uh, one of the, uh, each, each one of the uh each one at the extreme of the uh interpolter range, we see that the one, with initially higher manual employment, that is the one that has suffered the stronger manual decline, I've seen a 2.5% points, larger decline per decade in the share of women in, uh, in this type of. Um, so this accounts for 34% of the variation that I observed across, across committing zones in the in the change in the share of women in interpersonal low educated. And uh I didn't mention before, but when I get rid of the period of the recovery from the Great Recession, and even for the period of the Great Recession, these, uh, this, that have much more variances or variation. Uh, this, uh, this percentage or the, the percentage that these results can account for goes, uh, up much more. So this is, uh, This is one thing also for the ones that I'm gonna show you, OK? So. So now for all occupations, so this is, this was for interpersonal occupations, but what happened in the, in the labor market? So we see again that uh in line with the story that the, the shared or, or the commuting zones with initially higher shares of manual employment have seen a stronger decline in the share of women among low educated employment in all, in all types of, uh, in, in the whole labor market. And by comparing the two commuting zones at the extremes of the uh intercuta range, we see that the, the one that has suffered the stronger decline has seen a 2.8% points larger decline per decade in the share of women in low educated employment in the whole labor market. And this accounts for 62% of the variation across communities. And again, when getting rid of the last decade, uh, or even the, the period of the Great Recession, this, uh, this percentage goes up. So, just to sum up, I find that there's uh in, in, in, in the data that there's an important decline in manual employment, which has mostly affected low educated men. I find that these uh low educated men have shifted towards interpersonal jobs and that low educated women, at the same time that these low educated men have gotten into these occupations, they've, uh, they've, they are not uh increasing their presence there. So, I didn't show you because uh for the, for the sake of time, but I find, uh, I find that the uh average wage in interpersonal occupations is going down as much as, uh, as the manual occupations. So this, this pushes uh for the story of the labor supply. Because this, even if these, uh, if these interpersonal occupations are seeing growth in terms of employment, which might be because there's increasing demand, if they see a, a decline in the wage, it means that the, the increase in the supply has been much more, much bigger than the increase in the demand. So this would be in line of, of Of my snoring also. Um, so now, I will move to the model and why do we need the model. So basically, here I documented a bunch of facts, but to understand what can be the, the main driver. This is what I will use the model for. We have a mechanism to explain why this labor reallocation has taken place. So, I, I, I can take questions actually. I, I forgot to mention before starting the model. Um, yeah, yeah, there's, let me ask a quick one, and then there's one hand. Um, so your point number 4 in the slide before. I mean, those, those two sectors, manual and interpersonal, are maybe like about 2/3 of the, your, your distribution. So you're saying that average wages over this period actually declined somewhat significantly for like 2/3 of the workforce? So, this is what I observed for the low educated, and this is disaggregated by gender. So when I, when you put the, uh, the averages without gender, the changes in manual and interpersonal are of the same magnitude, but there's growth in, in cognitive. Uh, wages, but yes, you are right to say that uh. That A substantial share of the population on average, so the, the average wage for a substantial share of the employed population, so a decline over a decrease in, in, in the, in the wage over this period. But then there are, there, there might be differences, uh. On a finer, uh, occupation level that I'm not taking into account here. That might be also. That's OK. Thanks. Um, there's two hands, Clement first, Antoine. Hi, uh, Jonah. Um, I was wondering, so maybe it's similar to the point that Pia made earlier, uh, which is that your dependent variable is the share of women, uh, in those occupations, and I'm wondering whether, so why do you work with the share if, if what you're trying to show is that they are crowded out, so they, um, wouldn't you want to work with, um, more like the, the I don't know like the total number, uh, because if you add more men to these occupations, right, like the share is gonna, of women is gonna decline, uh, even if women still have the same access to these occupations. Uh, so why do you work with the share? Yes, uh, so that, that might be, uh, that the, that woman, so they are not in decreasing their presence in, in these occupations, but still, all the jobs that are being created from scratch, they are taking over by men. So this would be another. Uh, another way to, like, that I think it's, it's, it's what's happening. So, some women saw a decline in their, in their In the, in the present, in, in, in the, in the manual jobs as well, in the manual occupations. So, now, potentially they could have been getting into this type of jobs, but they, they can't because there is more competition, wages are going down, so they just prefer to not work. But I, I think if there again with Pierre's point, if there was a big decline in employment of men overall, uh, then it wouldn't be a, a crowding out story, but rather, uh, you know, the economy made space for these unemployed men in these occupations and that didn't really affect women necessarily. I mean, I, I, I, I, I'm, I'm sure your story is right, but I'm, I'm wondering whether that's the best, uh, way to, to show it. I could, yeah, yeah, yeah, I see. So you would do it like, uh. Yeah, but then the, the issue we're doing at the, uh, like in absolute terms is that potentially, like, how do you control for the sizes of, uh, I don't know, of the population of it gets Much trickier, but yeah, I agree that I could try. Yeah, I'm not sure what the optimal way to do it would be, but, um, I could try, I, I've, I've tried with, uh, employment overpopulation ratios, and I find, uh, I find the results are also there. The issue is that population again is, is, uh, something that is changing over time, so I don't know if, uh, If there's a decline in the, uh, in the employment overpopulation ratio of women because of the, because changes in the population or, or what's going on. So. To avoid that, I decided to go with the, with the, with the shares. OK, thank you, you have about 16 minutes, so let's quickly take Juan's question, and then you should, uh, continue. Yes, so I want to show you the model. Yeah, so just, just quickly, so, so you're going to do some welfare analysis I think. So my question is, Talking about crowd out, why, why isn't the option of women, low educated men going back to college and getting an education, uh, something that is an important margin to look at? Sure, sure, it is, it is, I believe, uh, it is, I'm gonna show you in the model that these changes in the composition of the population are having an impact on what's happening in the, in, in the labor market as a whole, so, but it's not what I started here. So, in this paper, what I want to do is to document uh these employment changes. And then, uh, in, in, in, uh, potentially next paper, what I want to explore is how this affected the decision uh of individuals to go to college. Now, I won't stop you, but you know, it seems to me the same thing in terms of short term versus long term. So you're looking at a different frequency. But, but the, the, the overall picture is really to look at whether the shift is, is immediate because they get out of the labor force and then they get education and enter back five years from, from, from them. But you know there's, I don't see a dichotomy between, uh, uh, looking at education and looking at, at employment. It's just a frequency maybe you're looking at rather than, than a kind of macro, you know, government is really looking at the long term trends, not looking at. A short term reaction to some structural change. I see, I see. Yeah, but over this period, what I see is that, for example, the share of total population that uh low educated and high educated men represent hasn't changed much. So what has changed is that of women. Oh, yes, yes, point taken, I could, uh. OK, thank you. Thanks to you. Go ahead, Joanna. I think you should go ahead and finish without interruptions from here. OK. OK, thank you. So I, I will talk to you briefly about the model and uh this is gonna be a static general equilibrium model in which I will have on the side of uh production, I will have a representative firm that it's gonna be producing uh A final route via the combination of three different differentiated tasks, and each task is gonna have a specific production technology and we use a specific, uh, a specific skill. OK, so this is for the side of the firm. And then on the side of the individuals, I will have uh terogeneous individuals in terms of education and gender, which are both gonna be taken as given and 3 skills. And, uh, based on all these, uh, on the skills, actually, individuals will decide to either provide their labor to, to one of these three tasks or to home production. So this is the broad picture of the model. So getting into the production technology more in detail, uh, this is the output of task J. J can be interpersonal, manual or cognitive. So this is gonna be, uh, the, the, the final good production is gonna be out of a CS aggregator of the output in this, in each of these three tasks, and the tasks, uh, are, the three tasks are complementary. And then to produce each of these three tasks, uh, I will have, uh, interpersonal and cognitive tasks that can be produced or are gonna be produced linearly by using uh efficiency units of labor here. And uh then the manual task will be able to be produced using efficiency units of labor or capital. And uh it is gonna have each, each of these factors is gonna have a, a specific uh productivity and the two factors are imperfectly. So, I modeled it this way, I could have also added uh capital in interpersonal cognitive. The only thing that I want to to account for here is that manual tasks use more capital than the other two. And, uh, in this economy, capital is supplied perfectly elastically by a pure capitalist. So, on the efficiency units of labor, the uh the on the tasks I was showing you the aggregate, but this aggregate comes from the combination of uh low educated uh efficiency units and the, the ones of the highly educated. And uh The ones of the, of the high and the low educated are So, are, are coming from, from exactly the the the sum within the level of education, the sum of those from women and from men, OK? From women and from men for the low educated, the same for the higher educated. So women's and men's efficiency efficiency units are perfectly substitutable within a level of education and uh the low and the high efficiency units, the low and the high educated, sorry, efficiency units are imperfectly stable. And then this was all on the, on the production side of the economy, now on the individual side. I will have a, a mass equal to one of individuals in this economy and each individual is gonna be defined by uh a level of education, a gender, e.g. and a triplet of skills which are gonna be drawn from a normal distribution, which is gonna be gender specific. So, the only decision that these individuals are gonna take is to sort into tasks as, as I already doing the. A week before, these, these tasks are gonna be the, the 3 tasks. So task J, it's, it's uh interpersonal manual cognitive. So basically, they will be comparing the, uh, the, the, the return they would get based on their level of education and the efficiency units they would get, the efficiency rates they would get based on the level of education they have and on their skills, and then the return to home production is there. And it's the same for women. The return on production is delta, and it's, uh, as you can see, it's uh not gender-specific. And women, uh, it's the same, the, the return to task, it's gonna be determined by the efficiency wage, depending on the level of education and the, uh, and the skills, but there's this. There's this 1 minus kappa. So what is this 1 minus kappa? This is a, this is I, I, I. It's a trick to microfound the uh consistent difference between uh the wages of men and women in, in, uh, in all these preoccupations. So, the, the gender wage gap, basically. So friends, uh, it's, I will micro-found it as, uh, or like, uh, Yeah. Sort of micro-funded as the fact that firms face an extra cost when uh they hire. So you can think about it as discrimination or Like a statistical discrimination, firms anticipate that uh women potentially will exit the labor market in the future because they will, they will. have kids or whatever. So that might be another explanation or you can think about it as productivity. You can, if, if you don't like. The story of discrimination, you can think that women are consistently less productive than men in all types of jobs. So this, uh, this is what, uh, creates this gap. And that's, uh, that's the whole picture for the individuals and going quickly to the equilibrium which is defined in a, in a standard way, uh, we'll have that each individual is gonna sort into the occupation that maximizes its, uh, labor market return, that firms maximize profits and that all markets. Nothing. Nonstand. So to go to the calibration. What I will do is to, to calibrate the model. I will calibrate, I will do one initial calibration in 1990 and then I will calibrate the model also in 2016 of the same uh The same comparison as, as in the data. So, the only, the, the exercise I'm gonna do here, that it's gonna be uh in a sense, my first counterfactual experiment, it's that from 1990 to 2016, I will only allow for changes in technology and the population, and the composition of the population. OK? So these are the only two changes that will allow from 1990 to 2016 in the current region. Let me tell you a bit more about the calibration in 199. To calibrate the, the model in 1990. What I will do is I will take the elasticity of substitution from uh like externally from other studies. I will compute the rental rate of capital in 1990 from the data, and I will normalize some of the parameters. So the elasticity of substitution, these are gonna be the ones that I take. Previous studies, I will normalize the means of uh the mean skills of men. And uh the one of women in, in cognitive. Uh, this is gonna be my assumption to be able to identify also the The rest of the skills for men and women and at the same time, the gender wage gap. Then the skills correlation, I will set it to 0.3 and uh the capital productivity and normalize it to 1, and this is the rental rate of capital that I estimate. No. And then the population composition, I also feed it in from what I observed in the data. So these are the shares of each of these groups of population. And then I calibrate the 16 parameters that are left to match 16 targeted more. So these are gonna be related to the skills distribution, to the home production, to the, the parameter that relates to the gender wage gap, and the one that relates to technology. So these are the targeted moments that I'm gonna use that are either gender-specific or education-specific, but never gender education-specific. And the calibrated parameters uh that I get are these ones. So, I have the women I will, uh, from the calibration exercise, women will have more interpersonal skills, less manual skills than men, differences in the variants. So there is a bit of uh heterogenic in terms of skills that the model predicts, librated model predicts between men and women. That's the picture for the calibration is 1990. And then from, again, from 1990 to 2016, only technology and population composition changes are allowed. So the population composition changes, I take them from the data. Mainly they account for the fact that we have more IV educated women in the share of total population now. And the rental rate of capital I completed for 2016 and then I calibrate the technology parameters uh to match uh the seven technology parameters to match these seven ones. These are the ones that I used in 1990 and here you have the ones in red that I used for 2016. So the rest of the parameters stay as in 99. And I don't have time to show you. But, uh, the model does a good job also in not only in the targeted moment because that's by construction, but in the non-targeted moment. So, as I said, nothing at the gender level of education specifically is targeted, but still the employment allocation and the average wages at this level, the model is able to uh to uh uh match very well with respect to the data. So this is, is one example of what's not targeted. This is the employment allocation of low educated men and women, and the changes between, uh, so these are the changes between 1990 and 2016. And um so you have here what's happening in the data uh for men and women in each of the occupations and what the model predicts. And you can see that the model can predict the uh the labor reallocation that we've seen very well, only with the technology and population. And now The first counterfactual or the second, in a sense, the second counterfactual I do is to try to understand what is behind, so what is it? Is it technology? Is it the changes in the, in the composition of the population? So I want to understand the separate contribution of this uh tool. So let me show you first what happens when we only, so this is the same table as before, I was showing you with the changes in employment over population ratios. So this is what happens when we only have Changes in the population composition. And you see that all employment rates would have uh gone uh up. This is from production going down. And particularly, it will have gone up for women in interpersonal populations. Uh, so this means, uh, that, uh, In a sense, if we only would, would have had a population composition changes, the pre-train, uh, the pre-1990 trend in uh female employment would have continued. And so this tells us that it's uh technology directly the main suspect behind these changes and indeed, this is what we see. When we only allow for technology from 1990 to 2016, We see that the uh negative, the aggregate negative effect is of the same magnitude for women, for women than what is it for men. And this is despite the decline in manual employment being much more important for men than what is it for women. So these men are gonna be The model predicts that they relocate into, uh, into the other two types of occupations while women decrease substantially, their presence also there due to this increase in the, this, this rise in the competition. So all in all, the negative effect is of the same magnitude for women than for men. Uh, so indeed it's a bit higher for them. So this, uh, This tells us that if only when, in a sense if only we would have had technology changes, the crowding out would have been stronger and the decline of total employment would have been even stronger than what we have observed for one. So just to take stock of what I told you and Um, changes in technology and the population composition, the model tells us that can replicate, uh, the labor market reallocation. We've observed in the data, and, uh, it seems that technology changes are behind the workers' reallocation and the crowding out of low educated women, while population composition changes have masked the effects of technology on the labor. So I don't have time to show you these, uh, these counterfactuals when I check. Uh, well, I explored further the mechanism and I closed the parameter related to the gender wage gap and I erase the, uh, the gender differences in skills. So I will conclude quickly. So my paper links the decline in uh manualemployment and the stagnation of female uh labor market participation. And uh it, it shows that structural transformation might not benefit women's employment in uh today as it did in the past. And, uh, productivity and population composition changes we learn from the model can explain the level of reallocation we've seen in the data, and productivity increases are in manual occupations are behind these changes. And uh the effects of productivity have been dampened by the increasing share of women attending college over this period. And this is important uh for policy because if we think about like policies to or the design of policies to, to get to full employment, like, uh, the design of retraining programs, we need to keep that, uh, these, uh, these changes in mind also for the, for the potentially designing uh social protection programs uh to help these individuals that are now jobless, uh, this, this is important to keep in mind. So, thank you very much for your time and your questions and comments.

showAllTranscripts

no

duration

PT1H15M37S

scene7File

worldbank/JoanaDuranFranchWomen

scene7Domain

scene7FileAvs

worldbank/JoanaDuranFranchWomen-AVS

title

JoanaDuranFranchWomen

description

JoanaDuranFranchWomen

showTimestampAndTranscript

yes