 explaining a little bit where the title of this paper comes from. This paper is part of a wider project we have at the Overseas Development Institute that's called Development Progress. We think that we have a component that's focusing on looking at how the different measures of progress affect our view of it. We think that we're looking also at the distribution of progress within countries and across groups and to develop a better understanding of inequality and those being left behind. So we think that I'm focusing on intra-household inequalities in children in different dimensions of well-being. The main motivation for this study is that until recently, many measures of well-being have treated households as if their members enjoyed a fair and equal share of resources, whether that's money, whether that's consumption goods, investments, etc. But when particular individuals are not enjoying their share of these resources, some of them could effectively be in poverty, even when the household average indicates the contrary. In particular in the case of children, systematic biases against boys or against girls in several areas of their well-being could affect their chances of lifetime poverty or could leave them in underachievement over the long run. So this paper attempts to measure the degree of intra-household inequalities, that is, inequalities that occur inside the households between boys and girls, and to relate it to overall levels of inequality and to progress in child well-being. Most likely the lack of data on individual children is the main limitation when trying to measure intra-household inequalities. We now, however, have much better data, international household survey programs such as the DHS that's have been used, and the multiple indicator cluster surveys from UNICEF have made it possible to review progress towards the improvement of child well-being, the child-focused MDGs, and the commitments of countries towards the conventions of the rights of child. For example, we know under MDG-1 that hunger and under nutrition in children has been reducing, although some gaps are still existing in some regions such as South Asia. We know that over 90% of children are enrolled in primary education and that gaps between boys and girls are reducing. However, there's still other gaps that are interesting to look, and I'm focusing on the gaps that occur within households. There are three approaches that could be used. The first one would be to compare average outcomes in a country between boys and girls, and this would give you a sense of the gender differences, but not really of the intra-household differences. The second approach would be to use a regression with a gendered doomy to see again gender differences, and you could include a household fixed effects to control for children in the same household. And the third approach is to use an inequality index such as a genie or a general entropy index, and then decompose it into the between component and the within group component. And you can use the households as the defining groups to see how much of that inequality occurs within and between households. And this has been used before. For example, Santa and Younger use it to measure differences in body mass index for adults. And I use a very similar approach. I use two basic inequality measures. The first one is just the share of the households with a gender bias. So this is derived from household ratios of achievements between boys and girls, and this would only tell you whether boys in a household are more advantage or girls are more advantage. And the second is I use this inequality index, I use a tail index in the same way that Santa and Younger do, and decompose it into the two components using the households as the groups. However, there is one limitation that is that when you use this index, you need cardinal data. And for many indicators of child wellbeing, you don't necessarily have that. So what to do then? So what I do is very simple, is because I'm interested in the differences between boys and girls, for each household I reconstruct a variable that is the share of girls in a household that are above a certain threshold. For example, the share of girls in a household that are stunted or not stunted, and the same for boys. So each household would end up with two observations that are cardinal, and with that, I can construct the tail index and then decompose it and have a sense of the contribution of intra-housel inequalities. Then again, there is a bit of an interest of looking at multi-dimensional wellbeing in children, because there are many things that affect how children develop. And the dimensions are drawn from the convention of the rights of child. And there is a data limitation. The convention defines 17 dimensions. I'm only able to look at four of those because we need information for individual children, not for the household, and for boys and girls separately. So sometimes you have information for individual children, but say just one child in the household. That's not enough for this measure. So I use stunting, birth registration, school attendance, and the time spent doing work and chores. First I look at each of the dimensions separately, and then I try to have a sense of how inequalities within household are jointly distributed, and I'll explain this a little bit further in a minute. Data comes from the mixed service from UNICEF for about 20 countries, but again, not all information is available for the countries, so it varies a bit. And I'll use the two latest service available for each country, which roughly corresponds to a five-year gap between service. I'll just explain a little bit of the results for one dimension. Before that, I also constructed a Gini coefficient just because it's easier to compare the overall level of inequality, but the Gini is not decomposable, so I don't use it afterwards. But with the Gini we can see that inequalities, total inequalities in stunting and working hours are particularly high. The Gini is about 0.7. Inequalities for birth registration are in the medium level, around 0.4, and in schooling they're much lower, just 0.18. And then how much of that inequality occurs within households, inside the households. So in the case of stunting, you can see in the graph the blue below is the within-household component and the cream is the between-household component. And there is great variability across countries, but for some of these countries in an average for all of them is around 20%, and for some countries it's much higher. The extreme case is Leo here, it's about 40% of total inequality occurs within households. So that's very important, that's the first thing. So then I try to look at how this relates to the overall levels of well-being and the overall levels of inequality in the country. So in the case of stunting, when there's higher well-being, total inequality is actually higher, so it is a problem. But within-household in absolute and in relative terms is lower. That's a similar case occurs for working hours, but in the case of birth registration and school attendance, even when total inequality is lower, or when there's improvements in well-being, when there's higher well-being, the share of within-household inequality is higher in relative terms. That means that it's possible that the harder gaps to address are inside households rather than across them. And this graph shows more or less what I just explained. This is just a plot of all the observations. In this axis, we have the total level of inequality and on the bottom axis, we have the average levels of well-being. So in the case of stunting, closer to the zero is more well-being, so it means less stunting. It's to look at how different inequalities within households relate to each other. Because it's possible that in some households, girls are preferred for some aspects, while boys are preferred for other aspects. So I try to look at how things correlate there. So this is a measure of association between intra-household inequalities. It's only done for each pair combination of indicators because if you try to do it for all of them, then the number of observations is very limited, which would be the ideal to do that, but data doesn't really permit that. So in the first column, we have the association only for boys, in the case, for example, when stunting, when boys are favored in stunting and birth registration and stunting and school, et cetera. And the second column is the same, but for girls only. So we see that around 20% of households tend to favor boys in nutrition and birth registration, 48% in nutrition and school. But there are other cases where a larger share of households favor girls. So take, for instance, stunting and work, 55% of households tend to favor girls. And this is an aggregate for all the countries that had available data, but this varies widely across countries. So take, for example, the case of nutrition and schooling. There are countries such as Swaziland where 27% of households favor boys. But there are other countries such as Albania where that share is 70%, so it's much higher. And overall, when we see at the aggregate, there is a lot of variation across countries, countries such as Kazakhstan, Albania, Belize tend to favor boys in most of the pairings. While there are other countries such as Burundi and Cameroon where in most pairings, households tend to favor girls. And there are others where the evidence is not conclusive. So let me go back a bit to the main headlines of the results. So first of all, although there is a wide range across countries, inequalities between boys and girls within households can be pronounced. In the worst case, they're up to 63% of total inequality, and that's a lot. Second, that even when in absolute terms, they're not very big, they can be big in relative terms. And that's important because it can point to where the hardest gaps to address are. And the third headline result is that there's not a clear bias against girls or boys, contrary to popular belief, but not contrary to other evidence from academic studies. In some cases, girls are favored, in some cases, boys are favored. And the same occurs when we look at the pair combination of indicators. So what can we conclude from here? First, that although there is a huge amount of progress in many of the indicators of child well-being, not so much in others, it is not possible to eliminate child poverty and secure all the rights of children unless these type of disparities are addressed. It is often the harder gaps to realize progress. And second, that there is no clear bias against one or the other gender. It depends a lot on the country and depends on the indicator or the combination of indicators. And this may be because biases respond to different social norms, different household institutions, marriage institutions, et cetera, and these are different across countries. So more research would be needed to understand whether these aspects are common across countries or defer and how and what are the right policies if you would wish to address these type of inequalities.