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

# Breadwinning Norms: Experimental Evidence from India

Authors: Kailash Rajah, Ishaana Talesara  
Discussant: None  
Video: https://www.youtube.com/watch?v=0mE9GNhwdY8&t=10088s


## Talk (02:48:08 – 03:02:01)

[02:48:08] Uh Ishana I think is going to present. We heard about masculinity norms yesterday and now we're going to hear a little bit more about breadwinning norms.  
[02:48:22] Okay, awesome. Uh, thank you all so much for being here. I'm really excited today to be presenting this work about breadwinning norms using experimental evidence from India. This work is joint  
[02:48:35] with Kaash Raja who's also here. I'll get started. So, we know that women work and earn less than men in almost every country in the world. And this pattern is especially striking in developing countries. So for example in India which is going to be the setting of this paper  
[02:48:50] female labor force participation is 32% relative to 77% for men and conditional unworking women earn about a third less.  
[02:48:59] And it's useful to understand the extent to which these patterns are driven by social norms relative to neocclassical forces because these different explanations have very different policy implications. So one norm that sharply captures this tension between norms and  
[02:49:12] and neocclassical forces is the breadwinning norm which is going to be the focus of our paper. So the breadwinning norm is the idea that a husband should earn more than his wife in a heterosexual couple. And in the case where this norm constrains female  
[02:49:25] labor supply, it could be the case that higher wages or higher earnings potential for women could in fact lower female labor force participation.  
[02:49:34] So in this paper we conduct what is to our knowledge the first field experimental test of whether breadwinning norms constrain the labor supply of married women and we find no evidence consistent with the norm can reject small effects of the norm and find instead that labor  
[02:49:49] responds strongly to wages. So here I'm showing you a figure that you might be familiar with. This is a very famous paper that was really important in kicking off this whole literature on breadwinning norms. So what you can see here is that this is  
[02:50:03] the the share of couples by the share of household income earned by the wife. And what you're supposed to take away is that there are relatively more couples where the wife earns a little bit less than 50% of the household income and relatively few couples where the wife  
[02:50:17] earns a little bit more than 50% of the household income. And the authors argue along with a lot of other evidence in the paper that this pattern is best explained by breadwinning norms.  
[02:50:27] And this paper kicked off a huge literature studying the cross-section.  
[02:50:31] So on the one hand, there have been a ton of successful replications and extensions of this work. There have been at least 20 papers in over 35 countries uh arguing that this is sort of uh explained by breadwinning norms. On the other hand, there have been a lot of  
[02:50:45] critical papers arguing that the presence of the discontinuity is better explained by assortative matching in the marriage market income reporting patterns or co-working spouses where spouses working at the same firm generates this pattern pattern but not  
[02:51:00] in a way that's consistent with breadwinning norms. But sort of underlying this debate is a more fundamental issue, which is that this cross-sectional discontinuity test doesn't identify norms because the cross-section is an equilibrium object that's affected by these labor supply  
[02:51:14] decisions which may be influenced by norms as well as by labor demand patterns. So while critics have argued that the presence of the discontinuity doesn't guarantee the existence of norms, it's also the case that the absence of the discontinuity doesn't guarantee that norms don't bind.  
[02:51:29] So in this paper we take a totally different approach which is to try and trace out the individual labor supply curve of women. And if breadwinning norms bind then what you would expect to see is that women's labor supply should  
[02:51:42] be discontinuous or should change in elasticity somehow around the husband's income.  
[02:51:48] So here I'm showing you three examples of inverse labor supply curves that could be consistent with this type of norm. The first one is going to be our sort of main pre-registered specification for a discontinuity. But our test will also allow us to test for  
[02:52:01] the other two types or other sort of variance of the norm that may be harder to pick up in the cross-section.  
[02:52:09] So you might be thinking well tracing out labor supply is challenging and that's for a few reasons. So one you need exogenous variation in wage offers made to the wife and you also need a large number of your observations. You need to observe a lot of choices  
[02:52:24] clustered around the husband's income so that you're powered to detect or reject a discontinuity.  
[02:52:30] So, we're going to resolve the first two with experimental variation. But then you might be thinking, okay, well, with an experiment, it's really important to be studying choices over long-term salary jobs in a real labor market setting. And that might be important,  
[02:52:43] especially for our research question because first, you want the norm to be sufficiently co you want complying with the norm to be sufficiently costly. you want to be giving up that which you would give up in the real world. And then second, you want the types of comparisons that people are making to  
[02:52:58] mirror the comparisons that they're actually having to make in the real world with the real jobs they and their husband have. So the third challenge we're going to address by partnering with a vocational training pro provider in India. So the NGO Pretham provides  
[02:53:13] one of the things they do is provide vocational training and placement services and they've worked with over 300,000 participants across India and we think that India offers a favorable setting to detect the norm due to the prevalence of conservative gender norms,  
[02:53:26] very low female labor force participation and suggestive evidence in the cross-section that matches what we see in many other countries. So that's what I'm showing you here. This is a figure from this Gupta 2022 paper that replicates this missing mass of women  
[02:53:40] earning slightly less than their husbands in India. And if anything, the pattern is much more striking and it's particularly striking in North Indian states, which is where the centers that we work with are going to be located.  
[02:53:54] So this vocational training provider is going to provide us with our natural labor market setting. So the way it works is that uh Pratam the NGO partners with healthcare centers to have access to jobs. They put the trainees through  
[02:54:07] two months of vocational training and then they're placed into jobs which in our case are going to be salaried positions in hospitals or home care. And this is a sort of highstakes labor market setting in the sense that most people who participate in the program  
[02:54:21] will find their first job through the placement services. So, this is not like a college career office that maybe nobody actually ever uses.  
[02:54:30] Um, and we also think of our population as being women on the margin of working.  
[02:54:34] On the one hand, they completed this training program, but on the other hand, at the time that we contact them, most of them will not be currently working.  
[02:54:43] So, how are we going to design the experiment? We're going to recont graduates of the program and offer them placement services that look very similar to the services they received upon completion of the program.  
[02:54:54] And what we'll do is we'll embed our experiment within the placement survey.  
[02:54:58] So first in the design, we elicit the husband's wife in front of the wife, the husband's income in front of the wife.  
[02:55:05] Um, and this ensures that we we observe the same reference point that the wife is observing and using, and we can also confirm that she observes this income when she's making her decisions.  
[02:55:16] Then we'll present her with 40 job opportunities and she can basically decide yes or no, I do want to apply to every single one of them.  
[02:55:25] Since it's not going to be feasible or ethical to randomize wages you actually receive because these are long-term jobs, what we'll do is we'll combine real and hypothetical jobs. So the real jobs will ensure that every decision you make is in expectation a high stakes  
[02:55:40] decision and the hypothetical jobs will allow us to randomize the wage offers.  
[02:55:45] So a job is something like it contains you know what the job is, where the job is located, the salary which is what we're going to be randomizing and some other characteristics of the job.  
[02:55:57] We'll over sample near the husband's income so that we're well powered to test or reject a discontinuity and then we'll implement their choices using an incentive compatible mechanism which I don't really have time to describe.  
[02:56:10] Okay. So now I'm going to move into showing you results. First I want to make the point what that we think our sample is really well suited to test for the norm. So here I'm showing you in our experimental sample for the women that are currently working before we before  
[02:56:25] we do our experiment, what is the share of household income earned by those women. And you can see that this looks very similar to the national the picture. We we replicate the same pattern where there's a lot of women  
[02:56:37] earning less and few earning more. Second, I want to make the point that in line with the national average, 31% of our sample is going to agree with the statement from the world value survey, if a woman earns more money than her husband, it's almost certain to cause  
[02:56:51] problems. So again, this is in line with the national average. So we think this is exactly the kind of setting in which you would want to be testing for the storm.  
[02:57:00] So now I'll show you the experimental results. So first I'm going to plot the proportion of women interested in the job against the wage relative to the husband's income. So the dashed line one is going to be where it equals the husband's income. And under sort of a  
[02:57:15] neocclassical model, you would expect this line to be upward sloping. And in the presence of norms, you would expect a discontinuity, a flattening, a negative slope, some kind of action around the husband's income.  
[02:57:26] But in fact, we're going to find that labor supply is pretty much monotonically increasing in wages.  
[02:57:33] We give them a few days to talk about talk about the the job offers with their families with their husbands to think about them and you know uh rescend any applications they would like to and labor supply shifts down but there's no differential impact around the husband's  
[02:57:47] income and this is actually quite responsive.  
[02:57:51] So this is going to be about a 1% increase in wages is going to correspond to a 2% increase in interest in the job.  
[02:57:58] So, good oldfashioned wages. Um, the second plot I'm going to show you our pre-registered regression discontinuity design, which is just going to allow us to test specifically for a discontinuity. And we'll find that if anything, the point estimate is  
[02:58:13] positive and we can reject discontinuities as small as about 3% in our uh main specification.  
[02:58:21] So then we also pre-registered many subsamples in which you might expect these norms to be more important. So for example, the people who agreed with the world value survey question, people who practice sparta or veiling of some kind,  
[02:58:34] um women whose husbands are salaried, so where we think that comparison is most easy to make. And the pattern is across all of these subsamples, we see something very similar, we generally get positive estimates and are able to or  
[02:58:47] and are unable to um or yeah, we generally I set up the sentence wrong. Okay. uh we generally find no evidence in breadwinning norms and can reject discontinuities of various sizes across  
[02:59:00] the across the samples. Okay. So the goal of this paper is not to explain the cross-sectional discontinuity but we wanted to discuss a little bit if the discontinuity is unlikely to be driven by breadwinning norms in labor supply. What would some  
[02:59:15] other plausible explanations be? And there's many explanations some of which have already been proposed in the literature that concern labor demand or measurement. So first is assortative matching in the marriage market in an economy where there are gender wage  
[02:59:28] gaps. Co-working couples where couples earning at the same firm t working at the same firm tends to push their incomes together but not necessarily in a way that is related to gender norms.  
[02:59:39] Uh reporting behavior which is possible in India where we're using survey data or wage floors in casual labor markets where both the husband and wife are subject to either the same or very similar wage floors. And now I'm going to show you some um evidence from  
[02:59:53] national survey data. So this is not going to be our experimental data um to kind of probe these explanations. Though we're not going to take a stand on which we think are more or less important per se.  
[03:00:04] So first I'm going to show you the you know usual familiar figure this missing mass of women earning slightly more than her husband. Then I'm going to remove the women who earn exactly 50% of household income. So that's a behavior  
[03:00:18] that could be either explained by those wage floors or by reporting behavior.  
[03:00:23] And we'll see that that already shrinks the discontinuity a lot. Then next I'm going to remove uh couples that work in the same industry and occupation. So that's sort of a proxy for couples that work in the same firm and uh which is  
[03:00:37] sort of one of the explanations and we'll see that that even further shrinks the discontinuity. So we think that this alongside our experimental results suggests that this equilibrium distribution is better explained by labor demand forces and that instead it  
[03:00:51] doesn't seem to be that we're finding a lot of evidence that breadwinning norms are constraining the labor supply of these women.  
[03:00:59] So to conclude um we conduct what is to our knowledge the first field experimental test of breadwinning norms.  
[03:01:05] We reject norms and find instead that labor supply is strongly responsive to wages. There could of course still be effects on the quality of marriage or on the subjective well-being of these women or men. But our results certainly suggest that even for married women in  
[03:01:19] low female labor force participation where wages are often thought to be less important, we find that wages remain an important force to bring these women into the labor force. And then I finally just want to conclude by zooming out and making the broader point that which  
[03:01:34] norms bind is important because different norms have different policy implications. And our results suggest that norms other than breadwinning might be more important for sort of drawing women into the labor force and raising their wages.  
[03:01:46] Thank you so much. [applause] >> Thank you. Thanks uh to all of our presenters in this section. Uh it's lunchtime now and I think we start again at 1:15. So please come back at 1:14.