 So, good afternoon everybody. Welcome to this parallel session on the impact of COVID-19 and mitigation measures on poverty and inequality in the developing world. I am Matteo Ricchiardi. I'm the director of a center for micro simulation and policy analysis at the University of Essex and I will chair the session. This session is actually based on the work done for the SouthMod project for which Union Wider with a number of partner institutions developed micro-simulation, tax benefit micro-simulation models for a number of African and Latin American and also some Asian countries. So, today we have two presentations followed by discussion and then general, even if short Q&A session with the audience. So, the first presentation is on African countries and it is presented by Yesela Stunen, who's a research associate at Union Wider. The paper has a long list of co-authors and so, yes, I leave the stage to you and that will be followed by a five-minute discussion by Miraco and then we will have a few minutes for comments from the audience. Yes, the floor is yours. All right, thanks Matteo and thank you everyone in the audience for joining the presentation and coming to the session. So, I'll be talking about our paper to the rescue and as Matteo mentioned, I want to emphasize that this work comes from a collaboration of more than 30 people both in Africa and also here in Finland and in the UK. So, we have a really, really large group. So, our work on Africa is part of a larger research project focusing on the distributional effects of the pandemic, which we study using tax-benefit micro-simulation models. This presentation concerns our work on the five African countries listed here. However, we also have already published two country-specific studies on South Africa and Ecuador, which I believe Javier will be talking about in the next presentation and we're also doing similar research in Vietnam. We have two main objectives or research questions in the study. First, we look at the effects of the pandemic and related lockdown measures on poverty and inequality and second, we look at the contribution of the both the general tax and benefit system or automatic stabilizers and any new tax and benefit policies that were adopted during the crisis and we're specifically looking at the distributional effects of the crisis and these policies. Here are the different steps in our analysis. Our starting point is developing kind of a counterfactual data set for 2020 for each of these countries. This is a data set with information on the population in this hypothetical scenario where COVID-19 did not happen last year and things went on as normal. In the second step, we developed so-called crisis data sets that actually do account for the effects of COVID last year. I'll talk more about the specific methodology later. The third step is about gathering information on any discretionary tax-benefit policies adopted last year in response to the pandemic and also incorporating those policies into the micro simulation models and finally we use the models and the data sets I mentioned to estimate the effects of COVID and also the effects of policy changes made in response to the pandemic. I'll start with the creation of these crisis data sets I mentioned. The first step here is estimating GDP shocks for different industries in each of the five countries in the study. These shocks are relatively simple so we estimate the deviation of each sector's GDP from what it would have been based on its growth trend over the past three years before the pandemic. Next, we allocate these macro-level shocks to the individual level data that's actually used in the micro simulation models. Our basic approach is a random allocation methods where randomly selected workers in each sector are assigned to unemployment with zero income so that the total labor income in that sector is reduced in proportion to the GDP shock in the same sector and again the GDP shock comes from the deviation of 2020 GDP from its past trend growth. Also currently for Uganda and hopefully also soon for some other countries we also have access to World Bank phone surveys from last year which we used to kind of improve the accuracy of the estimates so this is kind of a robustness check or an alternative method to allocate the shocks to the micro level data. Here the losses for individual workers within industries are imputed from the World Bank data. Basically individuals are more likely to lose some of their income if they have characteristics that correspond to a higher probability of actually losing income based on the survey and these characteristics can involve for instance age, gender, education, formality status of the work and so on. The next step in the analysis is modeling tax benefit policies or policy changes made in response to the pandemic. Here maybe the first thing to note is that these types of policies have been quite limited especially in Africa. There are however some measures that have been adopted and we've been able to model with information on eligibility and different conditions so just a few examples from the five countries in our study. So in Mozambique and Ghana utility fees were reduced or waged for consumers after the pandemic hit. Zambia enacted an emergency social cash transfer that was provided to specific poor households for a period of six months in 2020 and Ghana actually had a quite large number of different measures enacted overall which we have modeled but the overall impact of these policies was actually negative on incomes and this was fully because of this large school feeding program in the country was stopped for nine months during the pandemic. So yeah it was kind of an unintentional COVID 19 policy which we've also modeled. The final step in the analysis is actually running these micro simulation models with different data sets and policy systems in place. In terms of the outcomes we focus on disposable income and income-based poverty and inequality. We also assess the extent to which any shocks were mitigated by both automatic stabilizers basically the general tax benefit system and any new policies that were adopted in response to the pandemic. And finally we also look at so-called income stabilization coefficients which I'll be excluding from this presentation due to the strict time limits we have here. And before actually showing the results here's a quick summary of the data sets and also policy systems that we use in the study. So as I mentioned for each country we have a so-called pre-crisis data sets that represents our counterfactual what would have happened without the pandemic and we have a crisis data sets that also includes the income shocks from COVID. We also have two policy systems for most countries those countries that actually implemented any COVID related policies. The first one uses policies in place up to March 2020 which means that basically no COVID related measures are included in this policy system. The second one does include policies that are COVID related basically any policies that were in place in the countries throughout 2020. And this gives us three modeling scenarios for each country. The first one is our counterfactual scenario with no COVID shocks in the data and also no COVID related policies enacted. The second one is our full crisis scenario which does include income shocks from COVID in the data and also policies adopted in response to the pandemic. So basically this is what actually happened last year. We also have this third hypothetical scenario where we do include the shocks from COVID so we use the crisis data set but we also use a policy system without any COVID related policies. And by comparing the outcomes from these three scenarios we can derive not only the overall impact of the crisis but also the independent effects of COVID related policies and also automatic stabilizers. So finally some results which I'll try to run over quickly. So these are the changes in mean incomes in each country which we attribute to the pandemic. You can see that in Tanzania we estimate that the pandemic reduced mean incomes by nearly 10% while countries like Ghana and Tanzania were much less affected. Here in turn the total effect is decomposed into different components basically income changes resulting from COVID related policies automatic stabilizers and then the actual shock to earnings. And the main takeaway is that COVID related policies and automatic stabilizers only slightly mitigated the income shocks overall really not that much. And as I will show later automatic stabilizers namely alleviated income losses for households at the top of the income distribution while the policies affected poorer households as well. Here are the effects on poverty and inequality. The green and red numbers show the share of population earning less than 1.9 dollars per day with and without the COVID shock. Overall the total effects in black are not huge. They range from 0.5 increase in Tanzania to a 3.4% increase in Zambia. Here's also some decomposition estimates so you can see that poverty actually went up in Ghana because of the policies adopted and again this mainly comes from the stoppage of the school feeding program I mentioned. Here you can also see the changes in poverty gaps which actually increased a little bit more in relative terms than the poverty rates and basically this tells you that even if fewer people actually dropped below the poverty line those that were actually already below the poverty line drifted further below it. So this is kind of an alternative measure to look at income changes at the very bottom of the income distribution. And finally here are the effects on the Chini coefficient. They were relatively small ranging from no change in Uganda to a 1.7 increase in Mozambique. Then quickly some decomposition results. This graph shows changes in disposable income in Zambia they're composed into different sources of these changes. The white dots show the net effect on disposable income for different income quartiles and the black bars for instance show the earning shock from covid. And here you can clearly see that higher income households in Zambia lost more earnings and income in relative terms compared to poorer households. Also the emergency gas transfer in Zambia the policies like these are in the dark gray bars they actually overcompensated for the income losses among the poorer poorest households. Yes you should go towards the conclusion. All right I'll be quick. Finally automatic stabilizers had a very limited effect on cautioning income losses in Zambia. Basically only households in the top quartile benefited from paying less taxes and social insurance contributions because of the reduced earnings. Then just a few more countries here. Here's the same graph of Ghana. If you look at the black bars the households in all income quartiles lost some income. But interestingly again you can see that the posing of the school feeding program actually reduced income substantially for the for the poorest households. Yes I suggest that perhaps we look at the individual countries in the Q&A time if you don't mind. So if you just want to have your like final slide and then give the page to Miracle for her discussion. Thanks very much. Sounds good sounds good. Okay I saw really quickly these comparisons between the earning shock across different employment types using the random allocation method and imputation method for Uganda. And this is a pretty interesting finding because when you actually input losses using the World Bank data this suggests that informal workers lost quite a bit more income compared to the random allocation method. So this is just a way of showing that there are ways to improve our random estimates using micro level data which we currently have for Uganda. But yeah the main findings we found modest increases in inequality and poverty across the different countries studied. The effects vary across countries interestingly higher income households also experienced relatively large income shocks. One of the reasons for the small GDP for the small effects on poverty and inequality was that agriculture was actually not very affected in terms of the GDP shock and that worked as a buffer against income losses at the bottom of the distribution. Automatic stabilizers had a very limited effect in mitigating income losses. Of course one reason is high informality. Many people work in the informal sector and are not eligible for any income mistested benefits. And finally I'll say that the emergency social cast transfer in Zampia was likely quite effective in reducing income losses at the bottom of the income distribution while in other countries policies were very limited with exception of Ghana. So yeah I'll stop here and happy to take any questions. Thanks. Thanks very much. Yes thanks very much. I hope we'll have time to perhaps look into more details into these country results after the discussion. Miracle Ben Hurra the stage is now yours. Thanks very much. Thank you. I would like to compliment the presenters for an interesting paper and good work given data limitations in Africa. So the paper is timely and well motivated but as I was reading it there's been a lot of reference to existing papers in the methodology section which is a bit kind of a detour for the reader. So I would like to recommend maybe foreign annex to the paper. Then coming back to the generation of the crisis data sets you mentioned that there's been a random assignment of the labor income shock but when we are looking at literature it tells us that some demographic groups were more affected than others. For instance women and young people could have been more affected than men and also as you have highlighted from the Ugandan data that informal sector was more affected than the formal sector. So I just want to ask that even though you don't have that data for the other countries is it possible to kind of just try to have an understanding of how the random assignment is capturing informal workers since they form a large segment of the African labor force. Then still on informal workers when it comes to the simulations I just want to have an idea of how the simulations were adjusted to cater for the large informality in Africa. So when it comes to results I'm interested in the result for Ghana where you are saying that the feeding scheme was a big shocker to the incomes. So my question is what exactly about the feeding scheme could have propelled this result is it about the share of children who are covered and also how much more could have also spent on their children in order to generate such kind of result. Now another point is related to policies so you've highlighted that the automatic stabilizers and the COVID policies had a negligible effect. So my question is given Africa's structural configuration of large informal sector and also fiscal constraints what would your paper recommend African governments are understanding to do in order to harness their fiscal systems the size that they'll be in position to minimize the impact of future shocks on the economy. Then I also want to highlight the comparison made in the paper between the results for Africa and results for Europe. So I'm not really convinced that those two sets of results are comparable given the different structural configurations of the economies as well as what has been captured in the simulations. So I would like some bit of clarity on that from the authors. So that's what I have at this point. Thank you. Thanks. Thanks very much Miracle. I think you posed a lot of very interesting questions. Most of them are methodological so I think we don't have time unfortunately to discuss them here. Perhaps I hope you will continue this discussion with Yesa and the co-authors later but I will give you, I give Yesa the opportunity perhaps to elaborate for one minute on the policy implications of the study. Just one minute please. That was the hardest question of course of all of them so that's really difficult. So yeah you mentioned very correctly that as as I brought up in the study of automatic stabilizers had a really really small impact in mitigating any income losses. Of course one reason for that was the small informal sector in this country. So many people are simply not eligible for income listed benefits and at the same time there simply are not many very much you know automatic stabilizing policies for instance tax revenues or sorry taxes going down when the crisis hits or automatic social protection measures that would that would be helpful. So as you mentioned there should be some way of actually getting that revenue to fund automatic stabilizers and overall public spending programs and that's of course really really hard. So I don't think I have any any particular solutions other than that should be a priority in the future for basically any of the countries in our study and probably for most of Africa as well. So not a very clear answer but the questions are great. If a study like that could provide very strong and clear policy conclusions it would be fantastic but perhaps it would be asking a little bit too much. Anyway thanks very much both to Yeser for the presentation and to Mira for the discussion. I will postpone the Q&A session to the end of this parallel session if we have time and then go straight into the second part of a session which is focused on an application of this tax benefit micro simulation modeling to Latin American countries and here we have a presentation by Javier and Lourdes and followed by a discussion by Veronica Amarante. Javier and Lourdes the Javier is a research fellow at the University of Essex. Lourdes is an economist from the Central University of Ecuador and the floor is yours. Thanks very much. Good afternoon everybody. Javier is going to share a presentation. While for me it's a great honor to present together with Javier the results of two studies carried out with other authors as well and we are going to present it about assessing the question in effect of tax benefit policies during the COVID-19 pandemic in Latin America. Our motivation is that the evidence of these issues in Latin America remains scarce up to now and those few studies have focused on the effect expanded on social assistance programs in mitigating the distributional effects. Those studies are similar to the findings we found here in our studies as well. First a large increase in poverty and inequality measured by income and there is a limited effect of expanded social assistance except for the largest countries like Brazil and Argentina. Our studies are different from those because we not only evaluate the programs but also we evaluate the tax policies and also the social security contributions. Those two components we called automatic stabilizers so we have a more general view about the impact of those policies together. The aim to carry those studies was to assess the role of tax benefit policies in mitigating the impact of their drop in earnings during the COVID crisis. First we evaluate it in Ecuador and then we also expand the analysis to Colombia and Peru. We evaluate it in two main points. The first the main point of the when the crisis was the hardest was in the second quarter of 2020 and also we evaluated it at the end of the last year in 2020 as well. While we use a different from what made about Africa, we use only micro data simulations and we use directly the representative household survey data. Covering the pre pandemic period we use like the baseline the four quarter of 2019. Then we now cast the earning distribution during the pandemic based again on the official survey collecting data during two periods the second quarter and the fourth quarter of 2020. We use micro simulation models to simulate tax benefit policies and household responsible income before and during the crisis for this pandemic crisis in general. We simulate three income distributions. The first one is about the policies simulated during the pre-COVID data it means at the end of 2019. Second we simulate the COVID data based on the baseline of 2019 and finally we simulate the policies of COVID data as well during the crisis in those two main points. Also well these are the main results about the first study about Ecuador. As we can see in the figure one the white points are the mean of disposable income. The overall drop in the income was 41% on average but as we can see in this figure we have a UU shape impact in general. It comes from two main sources. First of all the automatic stabilizers have a had a large effect compared with the grant of the family protection which is the main policy in Ecuador. We have here in light blue the automatic stabilizers we can see that it only affects mainly the top of the income distribution while the grant or the family protection grant only affects to the poorest families in Ecuador. So at the end we have a more affected in the middle of the income distributions. The grant the policy implemented in Ecuador contributes to an increase in the mean disposable income of 30%. This is my part about Ecuador. I'm going to let Javier to present the rest of the results for Ecuador and the other countries. Thank you. Thanks a lot Lourdes. I hope you can hear me. I'm going now to share again my screen. Okay so as Lourdes said we started this study looking at Ecuador and we noticed the dramatic decrease in household disposable income so we wanted to extend the analysis and see what happened in other countries with similar economic characteristics so we expanded the analysis well in two terms. First because here we were looking at the second quarter we were wondering to which extent the economy recovered or not by the end of 2020. So this is the graph that we have here figure two compares at the left the same graph that Lourdes already discussed Q2 in 2020 for Ecuador whereas the graph at the right presents the result for Q4 2020. So as we see in the last column the decrease in household disposable income is much lower so now we have a decrease of around 20% on average but it remains a decrease so although the economy recovers by the end of 2020 we still see that household disposable income has decreased compared to the pre-pandemic scenario. Something else that we noticed that now we do not see the impact of COVID related policies so the blue dark bars that we saw for the low income desires this is because the COVID protection grants that were implemented in Ecuador were implemented only during two months during the second quarter of 2020 therefore by the end of the year although earnings drop and we still see a drop in household disposable income there were no policies to mitigate this shock. We see a small effect of automatic stabilizers in the last this side of the Q4 for Ecuador this is because as earnings drops tax and social insurance payments decrease automatically so we see lower reduction in disposable income compared to the reduction in earnings however it's very limited. So this is what happened in Q4 for Ecuador compared to Q2 but now what happened in other countries. Here figure three presents the results for Peru something interesting that we notice for Peru is that the size of the shock is somewhat similar a bit larger than in Ecuador for Q2 we have a drop in household disposable income those are the circles the white circles that you see of around 43-44% on average and something interesting as well is that as it was the case in Ecuador we see also a U shaped pattern of the decrease of the impact of the COVID pandemic on household disposable income across the income distribution contrary to Ecuador we see that Peru implemented a very generous social protection also protection policies during a Q2 in 2020 so we see that the fall in earnings in Q2 was more than compensated for the first income beside a group representing an increase in household disposable income around 40%. However now if we turn to Q4 at the right we see that as it was the case for Ecuador Peru did not maintain these social the expanded social protection programs until the end of 2020 therefore we still observe a decrease in household disposable income due to the shock in earnings by the end of 2020 Peru and there are no policies to mitigate this shock. What about Colombia this is our next graph we see that in Colombia first if we look at Q2 the decreasing household disposable income was much lower than in Ecuador in Peru because the decreasing earnings that the black bars that you see in the graphs was much smaller we see that COVID related policies in Q2 were more generous than in Ecuador but not as generous as in Peru however now if we turn to Q4 we see that there is still a decrease in household disposable income but out of these three countries Colombia was the country that decided to maintain this social protection until the end of the year so we still see an impact of COVID related policies to mitigate the negative shock in earnings and we see that this more than compensates the losing earnings for the first decide group I'll be very quick now in terms of income inequality and poverty as we see also in the previous graph there's a huge shock in household disposable income in Q2 so we expect income inequality and income poverty to increase this is exactly what we see here we figure five for the genetic coefficient so for the three countries Colombia Ecuador and Peru we see a big increase in the genetic coefficient in Q2 then we see a decrease in Q4 for 2020 but the levels do not return to the pre-pandemic levels the light blue bar in the bars represent the increase in income inequality that we would have observed if COVID related policies were not introduced by these countries so we see that although small the programs did help mitigate the impact of the crisis in income inequality and finally for poverty we see a very similar picture again a dramatic increase in poverty in Q2 2020 for all three countries levels decrease in Q4 but they do not return to the levels of the pre-pandemic and we see an impact an important impact of COVID related policies managing to mitigate the impact of the shock by around three percentage points in Colombia and Peru much less in Ecuador in Q2 so just to conclude and sum up what we have observed is a dramatic increase in income poverty inequality between December 2020 in the second quarter of sorry December 2019 in the second quarter of 2020 in all three countries COVID related benefits have a limited effect but even more so in Ecuador by the end of 2020 the economy recovers however household disposable income remains on average lower than the pre-pandemic levels around 20% lower in Ecuador and Peru around 12% lower in Colombia in terms of policy we see that Colombia is the only country that maintains COVID emergency policies in place until the end of 2020 that this helps mitigate the impact of the crisis until the end of the year and finally something interesting from this presentation on the other is that if you notice the effect of automatic stabilizers was always concentrated at the top of the income distribution and this is not only because of the design of taxes and social insurance this is also because of the design of benefits which are designed as proximists benefits because they are proximists that they are not they are unable to automatically react to income shocks to mitigate the impact of the crisis thank you very much thank you thank you Javier and and Lourdes in the interest of time can go straight to the conclusion uh sorry to the conclusion to the discussion by Veronica Amarante Veronica as a professor at the Institute the economia from the economics department at the university of the republic in Uruguay Veronica the floor is yours okay well thank you very much for for allowing me to participate in this in this session uh first of all i would like to congratulate the authors for the article and for the presentation the paper provides a detailed analysis of the impact of covid 19 on household income and and it shows the the effects of the policies i think it's it's a very relevant contribution to solve the puzzle about what happened in the region so there are other other papers who focus on other countries and it's interesting to notice that the results are very similar what what what they are finding for the Andean countries is similar to what has been found for other for other countries in the region i have some specific questions or suggestions for the paper the first one is if if i understood way some policy responses such as the employment insurance or the possibility to withdraw part of the of the private pension funds in Peru were not simulated i imagine that probably this is because of of data availability but i also think that there may be other policies that are not incorporated because of these same problems so in that sense i think it would be interesting to provide the reader the information about how much of the total package of policies are being simulated in this exercise just to give a glance of of of the of the of the size of the response that you are considering if it is almost everything or what kind of things that could not be included i assume that it is almost everything but then i also apart from suggesting adding information i would like to ask about that and i also thought that i also think that this analysis based on no casting it's it's very interesting but some the the results you're presenting are presented as deterministic in the sense that we do not have different scenarios or or some or we do not see any introduction of some randomness as for example in the first paper that we saw in this same parallel session so that was something that may help also to check how robust these conclusions that which are based on migrosmolation are so that was a second suggestion the the third one was that in each of this country a certain amount of money is is is assigned to to the pandemic so i depending on the type of policy that money may help households in the in the in different parts of the distribution so i think it may be interesting to calculate kind of effectiveness indicators of something like that of the policies in terms of the reduction of poverty or the reduction in inequality by dollar by some monetary measure and to compare the countries and also to see the different the the different location and effectiveness of the of the money and then i have a question about the the the conclusions i think it's interesting what you are what you are saying about the fact that there are no benefit acting as automatic stabilizers because cash transfers are designed in other ways so they do not act like that but then i i need the paper you also go ahead and and you said something like that the the social protection system in the in the region should be we should rethink about the social protections in the in the region thinking about those the space or the potential space for for automatic stabilizers for this kind of situation so i was wondering if you have some kind of policy if you are thinking about some kind of policy intervention in particular just associated to this idea that that cash transfers may not be the the more the best solution so if it's kind of a ready sign of cash transfers to be able to cope with with these situations or or what kind of policies may may help in these situations yeah thanks a lot lots of very interesting questions and suggestions i will allow Javier or Douglas to take one and respond for just one minute thank you very much Verónica i've come to the policy question as i think again is is relevant for the whole discussion yes perhaps there's two different things to say the first thing is that social cash transfers in latin america and in general in african countries are do reduce poverty and inequality so if you compare the the levels of poverty with and without the cash transfers we do see an effect in reducing poverty and inequality however what they don't they fail to do is to react to economic shocks so if you see you have an economic shock and people enter to unemployment they do not they cannot claim they can they are not allowed to claim these type of benefits why because eligibility is assessed in terms of the characteristics of the dwelling of the household of the household heads and this means that because they do not they do not depend directly on households income people cannot claim a benefit when the household income drops as a result of however i think it's interesting to think about why social protection is designed in this way and i think because they have they have a different role as in developed countries i think the role of social protection social assistance programs in the developing countries is to to fight structural poverty and that's why they are designed in terms of characteristics broader characteristics and income whereas social assistance programs in developed countries usually also target transitory poverty and so i think and that's when we discuss about rethinking these programs in the paper i think we are thinking about how to make these two things converge so should we should would it be able to keep certain social assistance programs in developing countries fighting structural poverty with a design eligibility that takes into account a large number of dimensions of well-being or say characteristics of households whereas could we think also about complementing this with social assistance that depends directly in household earnings so that then when there is an economic shock people or households would receive an automatic protection rather than depending on the willingness and the revenues of the countries when we face an economic shock and i think from the coverage the geographic coverage of countries that we have seen here South Africa is one of the countries which stands out from from other low and middle income countries in the sense that they do have means tested benefits so i think there is a context and an opportunity to discuss how we could achieve or combine these these things in the context of Latin America and later on of african countries as well thanks Javier that was perhaps a little bit longer than one minute but also provided some nice reflections on on this general type of analysis with respect to other countries as well so we had two big papers today and lots of very interesting insights and comments and feedback from from the discussions so unfortunately we don't really have time to take questions from the audience but i would invite the more than 40 participants to the session to contact the authors directly if they have feedback if they wish to discuss anything and with that i really thank all the participants and the authors and presenters and the discussant and i will now close the session thanks a lot to everybody bye bye thank you thank you thank you so much bye bye