 This is a paper with three other co-authors. Actually, Rajesh Raj was supposed to present this paper in this conference and he had a last-minute travel problem and he said he couldn't make it, so I'm stepped in for him. So we have Ira Gang, Rajesh Raj, Mysore Myung Soo Jung, whose Rajesh is in India and Myung Soo is in South Korea. So the paper really follows a little bit from what we discussed in the session on women's work yesterday, where in that session we talked about women's labor force participation and understanding what might hold that back. Here we instead of take a little bit of a different focus. So we know there's been some literature on gender, on gender gaps in women's participation in labor force, gender wedge gaps, but there's not much of a literature on gender difference in firm productivity, especially the informal sector. And of course that's a really important question because female micro-underproperters face significant disadvantage compared to male micro-properters and there's some literature on that and also we know that women tend to be more self-employed than men in informal sectors, not more women who self-employed. So it's an important question to ask. And so what are we going to do in this paper? We would examine the correlates of gender differences in productivity of male and female-owned firms in the informal sector. We would assess whether a performance gap between male-owned and female-owned firms exists at all. Is it really worth studying? And also then try to see if we do see a gap, what explains this? I want to say right away, if you notice the word that I'm using is correlates, this is not causal. And in fact, that's something that one should be careful about. We're not presenting causal evidence. So the informal sector, there's a lot of literature on this particular topic. And we know it's very heterogeneous sector and we can think about households which are surviving, just about surviving, but there are also households which actually have some growth prospects, might move to the formal sector at some point. And so we do see this situation, we do see survivalist households, but we also see households which are using this informal sector to step up to the formal sector. And that's true for most countries, but and also true for India, which is the focus of this paper. In India, of course, as many of you know, India is a very large informal sector, one of the largest shares, and that's the terms obviously, in the world has about 75% of manufacturing employment in the informal sector and 17% of manufacturing output. The difference between the shares of employment and output suggests, of course, that the informal sector is not very productive. It is not surprising. The women are more likely to run small family enterprises and they cause it in more and less productive activities. And so obviously this is an important question, a policy point of view. If a female run firms are disproportionately are going to be more productive, then of course it's important to try and find out why that's the case and do something about it. Because the more that happens, the less likely there's women owned firms, women will be able to move out of poverty. So that's obviously an important normative question. So we're going to use OHACA, and we're going to use recidivated inference function, the compositions, this is why I said it's not causal, this correlates, and we look at gender gaps in productivity. And of course, we're going to use RIF, because many of you probably know this, OHACA has got some problems, it's called specification errors, lacks the contractual, the choice of the reference group may affect the ratio of the endowment effect, the structural effect of the gap and also overstates the contribution of the endowment effect. And also, as I'll argue in this paper, it was really important to use OHACA because we do see difference in the productivity gap between men and women across the distribution. In other words, the gap differs over the distribution itself, I'll show you a plot on that. So RIF does take a few things on the RIF, because I think most of you probably know the RIF very well now. So we can start to cut off actual to sort of make sure that we change, obviously just make sure that we have the same distribution of characteristics of a male and female owned firms. We provide the competition of differences across the whole distribution of productivity, we allow for the difference in observable characteristics, so-called composition effect and the prices or the returns, the characteristics called structural effect. Some people use structural effect, some people use coefficient effect, I don't know anything that matters. And of course, the one good thing in this approach is that we can get individual contribution of each factor. That's really important in understanding the gap. So we end to use RIF. I should say with the RIF one thing, a few things we did, which is a little bit different is that we use a rewriting procedure, which is perhaps, I mean, so if you're opposed in this conference, has proposed that a rewriting procedure to try to take away the problem with Ohaka Blinder, which is that Ohaka Blinder, it's very much a parametric approach. So rewriting helps in that. We also did something, which is what my fourth co-author came in, me and Suyun, he's done some really good work trying to show that if you don't worry about the reference group in this decomposition, makes a big difference. So essentially normalize across category variables. In other words, which reference group to use in decomposition doesn't matter. That's innovation we have in this particular paper. Okay, the data is, I mean, if you know this data set, this data on national data set on unincorporated non-aggressive enterprises. So you're not looking at cash enterprises here. So you're manifesting data services. And the data has not been used to analyze firm productivity before. I think these are all the first papers on this. And the fact is contributing to that. And also it has this very rich data set, which has a lot of information about firm characteristics, firm constraints, along with the gender of the owner, which is also important. So we have 270,000 odd firms of which 13% of female run. Obviously that's not surprising. There are more male-run firms. And we only go to rest of the data to sole-propertorship firms. Because there are others which have shared ownership and so on. So only sole-propertorship because we want to uniquely identify the gender of the owner. Right, so variables, gender of course, who owns and manages the enterprise, independent variables, labor productivity. We don't use TFP, don't factor productivity because estimates of capital stock for informal enterprises is a bit dubious, right? So you want to stay with labor productivity as a measure of firm performance. And then we have this set of independent variables, or expiry variables, firm characteristics, which is firm size, age, location, assistance from the government, or other sources, registration with set of authorities, linkages with formal set of firms, and account maintenance with the maintenance accounts using bookkeeping and so on or not. And then on firm constraints, do they have access to finance? We have a measure of finance constraints. Do they have access to electricity? And then we have a bunch of variables which perhaps those who work in India would recognize on social groups, the cost of the firm owner, schedule cars, schedule drives, other backward cars, OBC and gel. Because social groups, as we know, the literature in India shows that there's a big difference in poverty rates and so on between those groups who are disadvantaged, schedule cars, schedule drives in particular, and the gel cost population. And then we have controls on industry effects, illicit effects and state fix effects. Okay, so that's what we have for our expiry variables. Now if you look at this plot of the gender proactive gap, what is interesting in this is our distributions, color density estimates, you can see there's not much of a difference here. Male is red, female is blue, not much of a difference here. But there's a big difference here that's also evident in the quanta-quantal plot. Okay, so there is a difference, you can see that, but the difference comes much more clearly in this part of the distribution. So this is summarized that productivity of male owner firms is higher than that of female owner firms. The mean gender proactive gap is 70 log points, this is quite a lot. We find the gap increases strictly up to the 40th percentile than falls steadily. And we see that the gap is highest, why does the bottom percentiles and lower the top percentile. So the gap seems to be much more evident for the lower productivity firms, okay? And either way, by looking at the distribution, you can see that Oaxaca Blinder is not particularly useful, right? Because Oaxaca Blinder is a mean decomposition. So right away, you can see that why we need to use RIF, right? RIF is makes more sense. Okay, so we use again, I mean, in the paper, we have both traditional Oaxaca and RIF. So, you know, we want to make sure that we have, what we do is covering both Oaxaca and RIF. We get very similar results. So I'm not going to present the Oaxaca Blinder here. And the RIF is, the competition is performed with the mean and different results. We did a lot of like robust tests on this, but I'm not going to present all the results. So I'm also going to present the main result which is decomposing the distribution, the gap in gender, productivity across the composition effect and the structural effect. What do we see? We see that 25% of the gap is explained by the composition effect. In other words, differences in formal characteristics, in also characteristics. Form characteristics, state fixed effects and so on, industry fixed effects and so on. But you see the 75% of the gap explained that by the total structural effect, which is the returns to characteristics. So a big part of what we're seeing that we explained by the structural effect. Among the individual variables, we can see firm characteristics make a big difference. Half of the gap is explained by firm characteristics which you already saw when things like farm size, age, assistance from the government and so on. And in the other variables, we can see that regional effects and state effects also have a huge effect. We have explained 40% of the gap. We actually surprisingly see hardly any contribution from constraints. In other words, the difference in the gap of productivity is not because male-owned firms have more access to finance than female-owned firms. That's not what's happening, okay? Or access to electricity. So we don't see that. Surprisingly, we don't see much of a difference in the social group effects, not making a big difference here, which was interesting, both in the structural effect and the artificial effect. I'm going to just provide the effects now across the distribution, which is really interesting because you can see that the explaining part of the structural effects is highest between the 20% and 60% around here, and why is it lowest at 90% out here, okay? So right away you can see, this is why the difference is important in our case because you can see that the contribution of these two big effects is really coming in here. Okay, and the structural effects particularly here. You can see this big contribution here, right? Almost, as I said, almost 80%, right? So that's important. Now, I'm going to get too much of the individual factors, but just to sort of, this is pretty much summarized the previous graph that we seem to see that the big effect is happening to structural effects, the return to characteristics, but also we see our own quarter of the gap can explained by the composition effects. And so we have both of them are important, right? I'm just going to present now one, let me just put the graphs here of the detailed composition effects. This is the, so we could do it by individual factors, right? The conversion, right? So when you could do that, we can see that again, interestingly enough, we see that across here, here, we see the firm characteristics make a big difference. In other words, the bottom end of the distribution, firm characteristics explain much of the gap. In fact, you can see it's a, it's playing major to the gap. So here you can see the bottom of the distribution, we can see the firm characteristics, returns to characteristics explain a lot of the gap and difference in observers, around differences in characteristics across male and female owned firms. So that's interesting. We also see a big effect here of, I think this is industry fix effects. So where are these firms located? Which industries also make a difference? Which is not surprising, because you can imagine that female owned firms may cluster on certain kind of industries and male owned firms can cluster on some other kind of industries. So industry fix effects make a difference and that's also not surprising, right? So that's also there. And the other effects are really don't play a major role. But certain effects, again, firm characteristics make a difference, but we see a little bit of a variance over the distribution here of the contribution. So it's not obvious here, which is really, really that important. But again, firm characteristics are important, but also industry estate fix effects, right? Okay, so let me just now wrap up the paper. So what we did in this paper is perhaps, you know, surprising, we look a little bit to the gender productivity gap. So we looked at that for India. We looked at informal sector firms because we do think that's where the gaps are often the most acute and we obviously observed that in the product distribution. It's really the least productive firms, but the gap is the most, not the higher end. So you find systematic difference in the productivity between male owned and female owned firms in the informed sector with male owned firms more productive than female owned firms. And when we look at the reasons for that, so we look at, we find the gender gap is particularly observed in the bottom middle part of the distribution productivity and less evident for the most productive male and female owned firms. Now remember that those firms, which are the most productive in this distribution, some of them might move to the formal sector, right? These are the ones who will probably make the jump. So in other words, those ones here in this part of the distribution are the ones who are probably most likely to move to the formal sector. Not all of them, but some of them will, right? We don't have that, we don't know which will because we don't have a panel for the data. But we also find that among observable characteristics the most important contributing set of factors explaining the gap is to do with firm characteristics. And the interesting thing there is it's important both to for the composition effect, but also to a lesser extent for the structural effect. Okay, so that's important. We also find that, so from a policy point of view and this is why I was being careful that I didn't want to say this paper is causal because one obvious policy implication would be that we need to do something about why is there this difference in both in the composition effect and returns to characteristics in to do firm characteristics. And you might argue that we might have to have more training programs for female perpetrators. You might just try and find a way to see that they can register with certain authorities and so on, but this of course because it's papers on correlation and not causal is our thing that we need to do more, right? So I'll stop here.