 My presentation is based on a paper, a joint work with Gaston Yalonotsky, where we are offering an ethical assessment of new inequality in education. I say new because of our context, which is post-pandemic disruption to learning activities and outcomes when schooling has shifted away from classrooms to home setting where teachers have been replaced by parents as primarily responsible for taking care of children's educational needs, and evidence already show that this transition has been highly unequal, meaning that children from better socioeconomic background have experienced either more learning time or greater amounts of learning. Now, you might pause here and say that this is boring stuff, hardly new. We have tons of literature already telling us that yes, children's socioeconomic status matter for school achievement. And again, there is the inequality of educational literature as well, Chico presented on this. And if you're following the literature, you might actually go one step forward and saying, hey, hang on, far developing countries, what the literature says that it is schools that is more important than parents because many children come from first generation learners. So therefore, why do we really care about some differences because of family background? Now, we will contest that, saying that, well, our context creates a natural experiment setting where with the global shutdown in school, the production function has been simplified that the school factors were not really relevant in the early months of the pandemic. When schooling was really about what was happening at home, when children are taken out of school, parents taken out of the economy, shut out, put at home. And that is where suddenly family becomes the primary source of learning and we have this educational inequality following the post-pandemic disruption. And that's what we exploit to really take a deep dive on what it means to have good family, favorable family for the debate on you know, in the culture of education literature, where the focus so much has been on parental background in terms of caste or occupation or the educational status of the parents. And our main point of departure from the literature is that we need to look beyond parental resources and think of non-pecuniary effort to understand really the origin of inequality related to family. And again, you know, why this is relevant when it comes to the debate on parents and educational inequality is because so far there are a lot of papers as well as a lot of interventions trying to improve educational outcomes, but they kind of focus on the instrumental role of parents that oh, it helps to have a good parent because your mother is a home tutor, so on and so forth. Now, what we argue is that hey, regardless of what your parents do for your educational development, there are intrinsic reasons why we care about family. We all would love to have a mother who would read out bedtime stories to us or would, you know, hold a good intelligent stimulating conversation over a meal and also, you know, role-playing when it comes to ethical development and so on. Right? Because this is important again when we look into the education opportunity literature. This is kind of dominated by views put forward by John Romer or Brian Barry, which is really focusing on the role of effort. The Romer demanding that well, we should only tolerate differences in education outcomes, which is really reflective of student effort, independent of family backgrounds, whereas Brian Barry is saying that well, we should respect all outcome inequality as long as they relate to student effort, regardless of whether that's correlated with family background. Now, here comes the English philosopher Adam Swift, who says that well, hang on, we need to focus somewhere else, which is intimate familial interaction. Now, you might say, wait, what is this intimate familial interaction and why it's such a big deal? So what I've done is to literally, you know, pre-produce a quote from his book, which has actually not been previously studied by economists in an empirical setting where one sentence really stands out. To the extent that reproduction of inequality across generations occurs in intimate familial interactions, we have reason to value and protect, preventing those interactions would violate the autonomy of the family. So therefore, this is what Adam Swift is saying that well, you know, when you're designing remedial education policy, don't go beyond and go and intrude into space that we should really live out to family. And therefore, regardless of what happens in that space, we should not intervene, whether it's a shortfall of intimate familial interaction or when it's a lot of that interaction, right? And this is really at the heart of my presentation today. And again, you know, why we care about this philosophical debate in the context of addiction policy design, because in the past there have been a lot of reforms and, you know, interventions focused on conventional inputs, such as improving school quality, supplementary classes, and policies on the demand side have targeted parental resources like cash transfers, stipend for students. But at the same time what we have seen during the pandemic is the rise of ethics services. And in China, India, big development and to some extent government is really concerned that this is going to increase inequality. Why? Well, this is actually going to make the job easier for a wealthier parent. That is exactly the kind of illegitimate inequality that Chico worried about this morning, right? But that's not all. During the pandemic, what we have seen is the rise of pandemic pods. Now, where parents take charge, create study circles, putting themselves in as facilitating teachers. In some instance, they would create pandemic pods by engaging paid tutors. Now, this is where the debate gets tricky, because if you are in the world of Adam Swift, he will say, ah, this is kind of intimate family interaction if the mother is actually forming and facilitating the pandemic pod, and we should respect this. We should not correct for this. On the other hand, if you're a young rumor who would say that, well, hang on, this is a display of financial wealth, and we should not tolerate this inequality. Now, this is interesting because as we speak in Latin America and other parts of the world, there are a range of interventions that are being piloted in post-pandemic context, where we are actually offering remote tutoring interventions and sometimes sending SMS messages and phone calls to parents to engage them in children's educational activities. Now, that is where the philosophical debate becomes very relevant because what they may be correcting in the name of illegitimate inequality may be seen by some philosophers as legitimate inequality that should be tolerated. So therefore, what we do in this paper is that to take advantage of this very early disruption to children's schooling, when at the height of school shutdown, parents, you know, school authority didn't have enough compensatory mechanism in place, schooling was literally enough to parents. So we use data from Bangladesh, a country with the longest spell of school closure, to offer ethical assessment of educational inequality with an exclusive focus on what it means to have good family or more intimate family doing school closure in learning context. And we do so by trying to document acceptable affair inequalities, you know, in three aspects. One is parental investments in outsourced support. Now, the one is parental time spent helping children. And the other one is student study effort. So this is more kind of Brian Berry. This is Adam Swift. And this is what would typically be not acceptable as, you know, fair inequality. But some, like Adam, you know, sort of like Brian Berry would still consider this as part of a broader measure of parental effort. Now, anything left out would be then a measure of unfair inequality. Now, we go beyond this by also combining different ethical perspectives to kind of define the lower bound and upper bound of, you know, acceptable inequality. And in terms of outcome variables, because we are doing this study at the very early month of the pandemic, so it was too early to measure learning outcomes that would be affected by the pandemic. So we focused instead on the most immediate outcome that would be related to learning, which is learning time, you know, spent at home or as a combination of learning time that happened at home and with support from actors outside home, such as private tutors and coaching center. And this is again in line with what others have done in the literature. So the main finding is that when we look at total learning time, which includes self-study time, plus time spent in education with support from others, we find that parental monetary investment accounts for 50% of the predicted variance with a parental time effort and student effort accounting for a small share, but when we focus on self-study time as the main outcome variable, we find that parental time effort, which is what Brian, you know, Adam Swift will call a measure of intimate family interaction, that alongside student effort accounts for 50% of the variation. And when then in turn, we sort of use that to work out how much that would add up as the sum of explained variations. So we find that at the, you know, primary level, it's ranging between 20% to 30%, 34% of the total variation. And in secondary, it is 15% to 29%. And we also have slum households where I put the number in there. But this is for total study time. When we look at self-study time, the share of legitimate inequality is much higher, regardless of whether we are looking at primary or secondary, or whether we are looking at slum households in primary education or slum households in secondary education. Now, for primary, you might see that there is a zero. This is what happens when you invoke the age of consent criteria, where some philosophers like John Romer would say that, well, when it comes to very young children, there isn't any notion of effort or self-responsibility. So if you take that extreme position, so then of course, in case of primary children, nothing is acceptable as a fair inequality. So therefore, in the presentation primarily, I will focus today on secondary school children. So here is the outline. So let me start with the first part, which is conceptualizing the ethical evaluation. And I want to start with something simple, which is that if you have a laundry list of inputs that you would typically select to specify your education production function, you would have a host of family background variables, and then student demographic variables, and then you would have a range of school-specific inputs, school quality, teacher quality, school type and all that. Now, remember that our setting is such that we are in the early months of the pandemic, school shutdown, school authorities are either lousy or ineffective in developing South Asia, so we don't have to really worry much about it. Right? So therefore, this is the context where we have a very persimmonious educational production function, where we want to start by categorizing the inputs as ethically defined and others that are not, and if you do that, you can find two set of categories that would qualify as ethically defined. One is parental effort in terms of time. Another one is student effort. Now, student effort, you have two versions. One is coming from Brian Barry, who would say that regardless of who your parents are, if you're a hard-working student, well-motivated, disciplined, all that counts. If you're a John Romer, you would say that not only the orthogonal measure of student effort, which is independent of your family background counts, so they are together. Now, that means that you have three effectively ethical framework to partition your educational inputs, and what we have done here on this table is to remind you that, well, regardless of which one you pick, you can still get some measure of acceptable inequality at primary and secondary level, unless, of course, for primary, you want to invoke the age of consent criteria to apply principle of compensation. Right? But this is where the paper begins, because the next step is the following. What if you want to combine two ethical principles together? So when you do that, then you can have a minimum and maximum share of acceptable inequality, and that's what we do here. So to begin with, you can start with the SWIFT criteria, where factors that you define as proxy of intimate familial interaction would be the only source of legitimate inequality. So that is the world of SWIFT only and no one else. Then second one is SWIFT plus Barry, where you have the proxies of intimate familial interaction plus any measure of student effort, regardless of whether it's induced by parents or not. And then you have what we call SWIFT plus Roma, not Barry. Roma is the cleaner version of student effort, so it's a smaller component compared to Barry. Right? So therefore, once we combine these, so what we get is this inequality expression, so where the lower bound is SWIFT and the upper bound is SWIFT plus Barry and Roma and SWIFT is in between. Now this diagram kind of highlights the quantity, you know, gives you a sense of the quantitative magnitude of each of these components. So this is the share of intimate family inputs coming from SWIFT, and this is Barry and Roma will be a smaller subset of Barry. Now you have this dotted circle, which looks a little bit bigger than the solid circle. So this is the share of SWIFT that you can obtain if you are using the SWIFT plus Roma model. And because for the Roma estimate, we have to use an auxiliary equation to get the cleaner measure of effort, and when we do that, our estimate of the unfair inequality will become bigger, but that will also kind of increase the share of what we call the intimate SWIFT share because a part of student effort could be induced by intimate family interaction. So next thing that we want to do is to sort of formalize it, and formally speaking, this is what it will look like, that this is the SWIFT share of the predicted variance in student learning time, and this is the SWIFT plus Barry, and this is the SWIFT plus Roma, and it is essentially this previous slide, but we have formalized now to formally show that this is the component that we call as estimates of unacceptable inequality. And here, this is the component that will be the biggest of all the three estimates of, you know, unacceptable inequality. Now, in terms of implementation, I'm going to sort of move quickly in the interest of time. We do what Chico mentioned this morning, the shape of decomposition, variance-based decomposition, and this has been done in the literature again. Our innovation is really to introduce Adam Swift's framework as an alternative, ethical framework for categorizing variables, and to also bring in Brian Barry. This is what has, in the past, the work by, you know, John Ferrer has largely relied on the framework of John Romer. So in terms of implementation, if you have an ons production function, so you start with an estimate of that, you have the, you know, predicted, you know, you have the estimated coefficients and which you then use to kind of compute the contribution of every single variable in your total expand variation. That could be equivalent to also your model R squared. And then next what we do is to really, literally repeat that, but by aggregating the individual Indicators in their respective categories. So remember that in our earlier table, we mentioned categories where the variables belong to ethically defensive Defensible inputs, and so what we will do in our main Analysis is really to look at the share of inputs that belong To certain ethically defensible categories versus those that do Not. And based on that, we will then use The share of ethically defensible input to work out the overall Share of fair inequality in the estimated variance. So the data wise data comes from Bangladesh, you know, Collected in the month of may, the country went into Lockdown in march, and this is in collaboration with the Largest NGO in the world. So i'm going to skip all of these In the interest of time. Now, this is a summary of Our outcome variable. This is the before school Closure measure of total study time and self-study time for Primary rural, primary slum and secondary rural and secondary slum. But you can see this massive collapse of study time. So this is what we are going to focus on. So we are going to focus on variation, inequality in this, And this, and this, and this. But the context is that, well, We are looking at a context where the school shutdown has led To this massive collapse in learning time, and yet we want To look at variation on inequality within this context. Right? And so this is what we do in terms of partitioning Our variables. So we organize them in eight Categories all together. Our main focus is, again, really On the ethically defensible variable categories, which is This one, and this one. So money measures, including Digital divide, we don't accept that as source of fair Inequality. Right? So this is the, you Know, sample results. So once we kind of work Out the variants shared, by the way, they all are in percentage And they add up to one. And we have multiple models because We have different specifications to try out some Sensitivities. What you will find is that the Digital share, the digital divide that you hear a lot about Has very little share in the overall variation. Parental money afford, 50% of the variation is Accountable for that when it comes to total study time. But parental study afford, which is what brian Barry is interested in, which is what adam swift is Interested in and student afford, which is brian bary Is interested in, together accounts for, you know, about Close to 30%. But when it comes to self-study time, they Dominate the overall explained variation of almost over 50% And the parental monetary afford drops from 50% to 7%. Now, we get the same results for Slum households. So again, you know, if i'm to help you Visualize, so this is the lower bound and this is the Upper bound for the rural sample and the urban sample when it Comes to total study time. Now, this is the share of Fair inequality. So lower bound is strictly based on Intimate familial interaction, which is what adam swift Demands that we should respect and tolerate and not Correct for and the upper bound is a combination of brian Barry and adam swift. Now, what is remarkable is that When we focus on self-study time, the share of acceptable Inequality is very high. And this is regardless of Whether you look at slum household and rural household. And let me add that we are talking about low income Household. So this is quite striking that even in low Income household in rural Bangladesh, at a time of Crisis, so we have evidence that well, once we really think Hard about what it means to be family, there is actually a Lot of variation that the policymaker should not correct Instead, they should tolerate. Right? Now, so therefore to conclude, the main take away is that Choice of alternative frameworks for inequality assessment Matters. I have presented evidence Showing what happens when we take on board on competing Philosophical frameworks on what defines acceptable Inequality within education. And what we find is that even if We go beyond what the literature has focused on in the past, Which is student effort, there is still quite a bit of Defensible inequality left when it comes to intimate Family interaction. And again, this brings up this Point that during the pandemic, there has been this global Talk about closing the digital divide. What we have missed is This concern of Pantel divide. And what we have shown in this Presentation is that Pantel divide is not just about the Binary of rich father and poor mother. It is as much about Being an active caring parent and one where they are Indifferent to the children's educational needs. Now, therefore, I will stop there. One thing i should add Is we have not presented estimate of the intermediate case Which is swift plus romer. Because it is Computational demanding. But again, this is an intermediate Value which would lie between the range i have presented. So i think i will stop there. Thank you and i am happy to take questions.