 The paper I'm going to present today is joined work with Olivier Bargain from the University of Bordeaux and David Rodriguez from the University of Essex who developed the model for Colombia. And the motivation behind this paper is that Latin American countries have experienced an important decrease in income inequality over the last decades. This decrease in income inequality has mainly been associated to a decline in wage inequality. However, there have been also progressive tax and benefit reforms and this may have played a role. Therefore, the aim of this paper is to compare the redistributive role of tax benefit systems in Latin American countries. And for this, we take two neighboring countries, Ecuador and Colombia, which present contrasting situations in terms of income distribution. And the approach we take here is to compare counterfactual simulations whereby the tax benefit system of one country is applied to the population of the other. And this is really taking full advantage of the tools of micro simulation that have been developed. So just a quick summary of results given that we have a short time and in case I don't manage to actually present the results. What we find is that the Ecuadorian system is more redistributive than the Colombian one. And moreover, if the Ecuadorian tax benefit system was applied to the Colombian population, the Gini coefficient would decrease by 1.7 points in Colombia, poverty rate would decrease by 10% and elderly poverty would fall by 18.7%. And the result is associated to the more generous pension assistance in Ecuador. So let's get to the introduction. The role played by tax benefit systems varies widely across Latin American countries and Ecuador and Colombia present very interesting case studies. First, because they are neighboring countries, they are middle ranked in terms of GDP per capita, they are heavily dependent on oil exports. But furthermore, they present contrasting trends in income inequality and on the role of the tax and benefit system. So in the table here, we see, for example, that if we look at market income, the Gini coefficient in Ecuador is 50.1, whereas in Colombia, it's much larger, it's 59.2. Once tax benefit policies are applied, the Gini coefficient in Ecuador falls by 3.9 points, whereas the Gini coefficient falls only by 2.8 points in Colombia. And we would like to note to which extent differences in tax and benefit rules account for differences in disposable income between the two countries. In terms of data, for both countries, we use representative household data, very large service, which contain detailed information on personal and household characteristics, employment, income, and expenditures. And the income concepts have been harmonized in order to achieve comparability in the simulations. The models we used are EcoMOD and ColMOD. Both are models developed, where EcoMOD was developed as part of the SouthMOD project. And based on that experience, David Rodriguez, who's doing his PhD in the University of Asia, decided to build a model for Colombia to do some comparative research. Both are static models in the sense that there's no behavioral reactions taken into account and no adjustment to population changes over time. And the models have been validated with respect to administrative data. Something I should say is that these models are continuously being improved and extended. The results that we present here are for version 1.0 for Ecuador and Colombia. Now we have version 1.3 for Ecuador, version 1.1 for Colombia, and there's an upcoming version 1.4 for Ecuador. Just to give you an idea of the continuous work that is taking place in order to make the models better. And the analysis we do here are for 2014 policies, so we take a common year for both countries. And in order to take into account of the fact that the data for Ecuador is from 2011, what we do is to upgrade all monetary variables in the Ecuadorian data set in order to take them to levels of 2014. In terms of scope of the simulations for all, for both countries, we simulate employee social insurance contributions, self-employed social insurance contributions, personal income tax, and the main cash transfers in each country. For Ecuador, the human development transfer, which is approximately means tested benefit, which targets poor families with children below 18 years, the elderly population and people with disabilities with an amount of $50 per month. In Ecuador, we have two different benefits which would be equivalent to the human development transfer in Ecuador. Familias en Acción, which targets families with children, with different amounts depending on a health component and education component, and then we have Colombia Mayor, which targets the elderly population with a varying amount between $21 and $59 per month, depending on the area where the person lives. And in order to take into account differences in income inequality between the two countries, what we do is to use a methodology applied by Bargan in 2012, where they decompose income inequality for one country over two periods of time. Here, what we do instead is we take two countries and we decompose inequality at the same year, at a single period in time. So without going too far in details, let household disposable income be represented by the function DC of PCYC. YC describes the population of country C, so market income, but also social demographic characteristics. PC denotes the monetary parameters of the tax benefit system, for example, the amount of the benefit in each country. And DC denotes the tax and benefit function of country C, which transforms YC based on PC into disposable income in each country. And let also I of DC represent a welfare metric based on the distribution of disposable income, think for instance, about the Gini coefficient, okay? And what tax and benefit micro simulation models allows us to do is to create contrafactual distributions, such as for example, the disposable income when we apply the tax and benefit rules of country two to the population of country one. And the alpha there is an adjustment parameters in order to take into account that there's differences in incomes between the two countries and that tax and benefit policies in each country might be dependent on this overall level of income. So the total differences between, let's say the Gini coefficient in country two and country one can be represented by delta, which is I of D two minus I of D one. So the welfare metric of the distribution of disposable income when the tax and benefit rules of country two are applied to the population of country two and the equivalent for country one. And this difference can be decomposed into the contribution of difference due to differences in tax and benefit rules between the two countries or the differences due to all other underlying factors, for instance, the distribution of market income or other effects which are not captured by the simulations. And this difference here can then be decomposed into three components and there's two different types of alternatives for decomposition. The first alternative takes first, it moves from the population of country one to the population of country two, keeping fixed the tax and benefit policies of country one followed by a move of the policies of country one to the policies of country two. Therefore, the first component captures differences in policies because we are keeping the population of country two fixed but we are looking at differences in the tax and benefit rules. The second component captures all other differences because we are keeping fixed the tax and benefit rules of country one and we look at differences in terms of the populations why two and why one. The last component of the differences is simply a difference between the welfare metrics for the distribution in country one and the distribution in country one where we have adjusted all nominally, all monetary parameters in the data but also all monetary parameters of the tax and benefit function. And if function DC is linearly homogeneous in PC and YC this component should drop because nominal adjustments to both the data and the monetary parameters of the welfare function of the tax and benefit function should not change the ranking of households. The other decomposition is very similar but instead of moving from the population of country one to the population of country two keeping tax benefit rule of country one fixed we do the opposite, we move from the population of country one to the population of country two keeping the tax benefit rules of country two fixed. So let's look at the results. And here we have this huge table and I'm trying to take you through it. The first column, column zero gives us the Gini coefficient for Ecuador. So when we apply the tax benefit rules of Ecuador to the population of Ecuador, this is the Ecuadorian baseline and we see that Gini coefficient is 46.2. The last column, column number four gives us the baseline for Colombia. So Gini from disposable income when we apply the tax and benefit rules of Colombia to the population in Colombia and we see that the Gini is 56.4. The difference between column four and column zero gives us the total difference in inequality from disposable income between the two countries. Now, the first column simply gives us the Gini coefficient of the distribution when we apply the tax benefit rules of Ecuador to the Ecuadorian population but adjusting nominally all monetary parameters in order to take into account of differences in income between the two countries. We see that the Gini and also the poverty is the same therefore homogeneity holds that means that the last component in the composition drops, okay? And we can take a look at the shape of the composition then which is the average between the two decomposition alternatives that I showed you. But now more interesting, let's look at column two. Column two represents the Gini coefficient when we apply tax benefit rules of Ecuador to the population of Colombia. What happens then, the Gini coefficient in this case is 54.7 which is smaller to the baseline in Colombia 56.4 which means that when we apply the tax and benefit rules of Ecuador to the population in Colombia we have lower income inequality and this also holds for poverty, okay? So although the difference between the baselines is 10.2, still around 1.8 points is accounted for differences in tax and benefit policies. So there's scope, let's say in Colombia to change the policies to reduce inequality, okay? And just to move faster, look that the largest impact of tax benefit rules is seen for elderly poverty, okay? So moving on in terms of what is driving the differences we see that if we compare Gini coefficient of disposable income and Gini coefficient of disposable income without the tax and benefit instruments one by one this gives us an idea of the marginal contribution of each income component and we see that it's actually social assistance that is driving most of the differences be it in income inequality or in poverty. And we see in this case for instance that social assistance in Colombia in the Colombian baseline reduces elderly poverty by 2.9 points if the tax and benefit policies of Ecuador were applied to the Colombian population social assistance from Ecuador would reduce elderly poverty in Colombia by 7.3 points. So in conclusion, there is a small but non-negligible redistributive role of tax benefit systems in Ecuador in Colombia. Most differences are indeed accounted for differences in market income on other components yet we see that the Ecuadorian system is more redistributive and would achieve a larger reduction in inequality and poverty if applied to the Colombian population. And this is due to the more general social assistance benefit in Ecuador in particular the human development transfer for the elderly. In terms of future work we find this approach, this methodology very interesting so we want to apply it also to African countries taking better account of the labor market situation in particular informality and in terms of the policy swap methodology we're also working with Latin American countries and in this case we're interested mostly in personal income tax and we want to see what would happen in Latin American countries if we take the most redistributive personal income tax that we find in our pool and apply it to all other countries. Thank you very much.