 This is for Mohammed, I wasn't sure when you presented your question so I wasn't sure whether you have considered the possibility of having indigeneity problems between the relationship between violence and inequality. You probably, as far as I saw, is not potentially addressing these issues and I wonder what you think about this. And then on the inverted U-shape function that you observe, it is interesting but I'm going to question what is the fear behind that. So in principle you perhaps suggest that the middle class has the tendency to be more prone to violence which it is not actually observed in most of the countries I know. So could you tell me the theory or the fear behind what are your assumptions or your clues about what's explained in this curve? Alan Huaniemi, MTT Finland. I also have a question to Mohammed because I would like you to clarify what is the difference between vertical and horizontal inequality because I have difficulty understand that why vertical inequality will cause violence. That what is the reason behind it and also that how robust are your your your estimations to show that that there is a direct link between vertical inequality and violence. Thank you. Manjarita Ketokoski from the International Law Association and I noticed Mohammed used the term routine violence. I understand it means also violence against women which in the legal system now has its new convention who has come into force and of course it's a huge problem. I don't know whose idea it is to call it routine. Is it coming from the World Bank vocabulary? It's your idea. Okay. Then I do want to know in what way the Indonesian system of addressing violence is gender literate and what is to be planned to make it visible if you include it in routine I think that might be not such a good idea. And then I wanted to ask if I may from the last speaker also Namibia is famous at least in Nordic countries for the experiment of basic income and everybody is now interested in on international level also Indians and others are doing experiments. We are waiting for Canadian research to come and tell about the Canadian experiment. Could you say something about that experiment and how you classify it in your system? What is it as a revenue when you have a basic income in an area? Thank you. Thanks Andrew Karth from the University of Cape Town in South Africa. I've got one question for Richard and one for blessing or maybe a few for Richard. So I didn't really understand at all what your regression analysis was was doing. I think that's partly my unfamiliarity with these issues but from what I the basic that I got is that you've got household specific prices on the left and you've got some measure of consumption or income for the household on the right. You're trying to see if poor households face higher prices. So then I was just wondering about how then you are valuing the consumption expenditure that comes from own production because surely then you're going to be using prices to do that so now you've got prices on the left and prices on the right. And then also it was the survey conducted over the whole year it would be helpful to show us in the beginning I mean if it's quite dramatic to show us the difference in prices that households are facing over the course of the year. If you've got households being interviewed at different periods in the year and if the survey people provided you with that data I don't know if they did. Sorry one more question. In the beginning one of the things you said is that maybe search costs make it like a U shape so the poor and the rich maybe face higher prices. If that's true why were you then just doing it's positive or it's negative and that means either the poor pay more or the less. If there's some U shape like that then maybe what you were doing wasn't quite right but I wasn't sure as I said I wasn't so familiar with those methods. And then for blessing there's this famous case in Deaton paper for South Africa on our old age pension where you showed us that social assistance looks very tiny because inequality is so big I think the social assistance looks very small in the percentage of total income. But the case in Deaton paper makes the really good point that sort of by accident almost that per capita that the old age pension is three times I think per capita income for South for Black South African households. So it might look small because if you look at it the way you did but if you if you think about it in terms of that you know poor people's incomes it's actually very big so maybe you could just tell us whether that's true or not for Namibia. Thanks. Could we have responses to the first round of questions. Thank you for those questions. The first one on endogeneity. If you are thinking of the link between large scale civil war massive violence and inequality you may think of the violence disproportionately affect the poor and then change the distribution of income or asset or whatever. This is different kind of violence. This is low scale of violence. So I'll connect this one with the topology of violence that we are trying to develop based on one country data from one source of information. So this is not pre-conception idea. So the classification is based on we put all the violence data on the table and then we find the best way to classify the violence. So we have three types of violence. The first one during the democratic transitions. The first one is the resistance conflict or civil war between the few regions and the central. And then on top of that we have the interesting violence. So and then we have also a pool of residuals. So we are not sure how to classify the residuals of the violence data that we have. So but the best way to classify that one is the routine. Routine means there is no pattern of regional and time concentrations different from the first two categories where we can see that the episodic violence concentrated during the particular time and in few regions. So based on the classifications we differentiate the two between the episodic and the routine. So the main forms of the routine violence is in the forms of game fights. A youth browse. Fidgillantisms. And to be honest we are not separating the violence against women here. I mean we are not able to classify to extract that one from the data. So that's the nature of the data that we are dealing with. So that's what I can say about the okay so back to the endogeneity. So I think we can rule out that one from the nature of the link between inequality and violence. I have no proof from the equations but I think the next thing that we can do is to try to use a time lag for example. So this is the first rough of the paper. So that's what I can think now. To use the lag between inequality and violence so just to take into consideration that one. That's to answer that I can give some examples for the endogeneity. One from the conceptions of the low level violence and inequality. The way the effect of violence on the quality can be ruled out because it's not a large scale of violence that destroy the economy totally for example. And the second one is maybe to introduce the time lag in the revision of the paper. The next, your second question is about the infant use. This is the first paper to introduce that kind of link between civil war and income. In most cases you have the negative relationship. So lower income correlates with the civil war. So this is the most common and the most robust finding so far in terms of conflict regressions. But when you look at the data, we deal with different kinds of set of data in one country context. So we find that kind of empirical regularity, that kind of relationship. So now we have to find how to do that one. So I look at the Robert Bates book about prosperity and violence. Explaining the rise of violence in the media for Europe for example. So from the historical account, there is a hypothesis saying that initially violence will go in hand with the increase in prosperity because of the absence of regulatory functions, the greedy behavior of the fortunate, the grievances of the less fortunate, things like that. So we use that one to explain the upswing of the curve. And then the downswing, you have at some point in time, violence has to be contained. And at a higher level of income you have a stronger state capacity, things like that. So that's the way we explain the infertile use, shared relationship between violence and income in the context of low scale violence. Of course, this is intuitive, but as you know, in the regressions we can only draw the empirical link. Most of the link is about associations. But we draw from the theory, from the anthropological study, from case study, we draw the logic to explain the empirical regularity. If I can move quickly to the second question, support vertical inequality, horizontal inequality. Okay, horizontal inequality is clear. The relationship between vertical inequality and large scale ethnic conflict or civil war is clear. No doubt about that one. The logic is clear. So what about the vertical inequality? We link vertical inequality with low scale violence, which doesn't involve the notions of ethnicity. So there is no ethnic divide in the context of the violence that we are talking about. So we understand this one as high level of inequality will create a sense of frustrations, for example, a sense of grievances. But there is no direct target of that grievances. Not the ethnic A is richer than ethnic B, or region A is richer than region B because we are not talking about horizontal inequality. So that's why differentiating vertical inequality and horizontal inequality is really important to link inequality and conflict. So because this cannot be done at the cross country level, we can do this one within country level. So that's why I'm going to argue that it is not wise to totally dismiss the role of vertical inequality. But if you go to cross country study, there is no one considering genie as one of the determinants of civil war, for example. So that's my expression. The last one, routine violence against women. So that's the nature of the data. We are dictated by the data. And you know that violence data is really controversial. I mean, what I can say now is this is the best available data at the country level with the similar political and social context with the same methodology of data collections. So this is the best available data that we can have now. Thank you. Let's move to reach a blessing. Okay. Basic income, the basic income grant. An interesting concept. And a few years ago Namibia did an experiment in an area called Ocevelo where they offered a basic income grant to the whole community. There are some groups that are really advocating for it and say given the levels of poverty that we have, I think this is one way we can help the poor members of society. But what seems to be lacking is the political will to go that route, because I think the government is a bit fearing that they may not be able to sustain it in the future. And we heard a conference last year on social security in Namibia where the current president in waiting, the prime minister, showed significant interest in it or he has been in support of it. But I think within the political circles, selling it out to people to say we shall be transferring a certain amount of money to people every month, irrespective of who they are and what they are doing. I think they are finding it a bit difficult. So the idea is still there. There are still people pushing and the results from Oscevello have shown very interesting outcomes that many households that could initially not afford, get food to be able to. Some have started small enterprises where they are buying, say, flour, they make some cookies and sell. And they have also been able to send their children to school. Quite interesting outcomes, but whether that can be replicated across the country and the costs attached to that is the issue. Then the basic social grants, I'm not sure how much they pay in South Africa, pay an old age person, but in Namibia it was about 500 last year, they recently increased it to $600 per person, which compared to the monthly costs as per the poverty lines is reasonable. But still maybe not as much. We couldn't calculate the, or we couldn't use the total income figures because of the problems I mentioned with data. But if you disaggregate the data by main income source, which we did here, main income source, about 10% of the households say their main income source is the basic social grant. And the majority of them is main labor income. Thank you, Richard. Thanks, Andrew, for your questions. Remember what I'm trying to do here. I've said, when I'm looking at, you trying to check whether there's a poverty penalty, you can use the regression, which is what most studies do, beta, 2010 and others. But that is not very useful in terms of what I want to do in the paper, right? A better way of doing it would be to use the concentration index. So it doesn't really matter whether the effect is U shaped or whatever. All I want to check is whether it is there, right? So I want to check whether it's indeed there's a poverty penalty, right? So high prices concentrated among poor households, that's all, right? That's the issue. So it doesn't matter how, so the other issue that you're looking at, that's the standard deton identity, that you have prices being decomposable into a quality effect and an income effect, right? So it's a standard deton regression. I just use that to then page the quality effects in my regression. So again, there's no problem there. On price variation across the year, yes, you're right, the data was collected over the year, and therefore what I did was as my base category, I chose February as my base. So I'm saying okay, rule of February, prices in February, right? And then I use that as a base. So it in a sense captures the variation in prices across the year. Thank you. Hi, I'm Laura from Salamash University in South Africa. I have a question for Richard. So in the beginning on your first or second slide you listed the different reasons why poor might be paying more for food. And I was wondering if you did any sort of post-estimation descriptive statistics just sort of exploring which one of those were applicable for your result. So why did you find the result that you had? I'm just sort of interested in the mechanism because I think that's sort of what's important for any policy initiatives to combat the result. Hi, I'm Tanya Pohan-San from Jualalongon University in Thailand. I have a question for Mohamed. Because the routine violence covers a lot of dimensions of violence, right? So like many types of violence. So I'm wondering whether you have considered to look more into the mechanism or the channels through which inequality affects routine violence. For example, perhaps instead of looking at inequality per se, probably if you have the data like income differentials between groups, like ethnic groups, majority and minority are between gender. So that's kind of tell, probably tell you more about the mechanism of bargaining powers between groups. Yeah, and perhaps income differential between different ethnic groups may help address the endogeneity problems that Miguel raised. Like, because I also think that there may be endogeneity problem through ethnic diversity. So if you can clarify that it's not just some thought. I am Gogan from University of Malawi. My question goes to Mohamed. I was looking at the equation violence is equal to inequality and others or maybe on the others I missed out something. When you're looking at violence, where there's violence there are definitely some groups that are formed to mitigate against the violence. So I'm just interested, is it not proper maybe to look also into what were the number of groups that were found maybe in regions that were against the violence or I missed it if it was in the others. Thank you. No, to say the number of groups formed against violence that people form in societies against the violence that's happening within the society. Or there are no groups that were formed to mitigate the violence within the societies that the violent was taking place. Okay, let me make it clear. For instance at national level there is a violence in DRC. You have the AU sending some forces to combat the violence to help in combatting the violence. So I'm saying within the regions that were in which there was violence, where there are no groups that were formed to mitigate the violence, maybe regional groups that may be included in the equation to as one of the variables to affect the violence within the equation. My name is Prudence from the University of the South Africa. I have a question directly towards blessing. When you drew your Lawrence case you said you are comparing the individuals that have incomes main source from the pension system or the social system compared to those who have income from the labour market. I personally think that comparison would be confounded because people whose main source of income is from social security already a homogeneous group such that you're going to find lower inequality among them. I think would it be possible for you to do a counter-factual analysis to say if there is no social grant system in place this would be the inequality that we get and then when they get income it reduces inequality to this extent. Thank you. Okay I have a question for blessing. I think we are looking at social protection programs here and immediately we want to know as a policy question as to whether poverty is also higher in areas where population is higher because you mentioned that some of the rural areas cannot be reached for example but if poverty is higher in those rural areas then the social protection programs becomes an issue therefore I'm asking as to whether you are able to decompose your poverty estimates by population shares. Thank you. Thank you Nora for your question. Actually the issue of mechanisms actually was analysed in another paper which is forthcoming in the Journal of International Development where I show that actually this evidence of nonlinear pricing which coming from quantity discounting. So one of the channels that I mentioned there is quantity discounting and I find evidence of quantity discounting in the MACE market. Thank you. I will start with the last question that is directed at me the decomposition of the inequality by population shares. We will try to do that but the problem that we have been facing is the quality of the data that we have and trying to get more assistance from the statistics guys. We may be able to get something different but I doubt it with the social protection particularly the basic social grant because although there are some areas that are not covered generally the coverage now for the basic social grant is about 95% which is quite high and that may not significantly change the results that we have but we will try to do that. With the first question what we did was again it comes to the data issue the best we could do was to try and decompose the we have the main sources of income and not all the sources to say the total income is this which I think if we heard that we could do a bit more so what we did was to identify those households that say their main source of income say the social assistance and then we look at the genie for that type of household compared to those that say the main source of income is labor income we could try to use econometric methods to see which one is driving inequality we haven't done that but we will try that. Yeah on horizontal inequality that's why in the first place differentiating horizontal inequality from vertical and differentiating between episodic and routine is really crucial at the start of the paper actually we try both types of inequality because before that one measuring horizontal inequality has always been a problem not easy to have a consistent measure of that one say for example we have the genie coefficient based on the household survey but the household survey does not collect the identity of the respondents for political reasons so we can't create a group genie for example based on the necessity based on the household survey so we try to find the proxy of horizontal inequality from the census data and we construct the education group genie based on the census data so we have that one based on education group genie group coefficients of variations based on ethnic groups and also religious group from the census data and as you know from the theoretical perspective we should link horizontal inequality and routine violence if you want to link horizontal inequality you have to link that one with ethnic violence so in the ethnic violence section the results we try both we try both we suspect that or the explanatory power of horizontal inequality is much higher than the explanatory power of vertical inequality but we also find the significant albeit less powerful role of vertical inequality on ethnic conflict so the way we interpret that one because of the changing nature of quality violence in Indonesia from the period of democratic transitions to the current period after the all the chaotic transition has passed so what we are saying here is the nature of the recent ethnic violence is closer to the characteristic of routine violence not episodic anymore much but less not close to the episodic nature like ethnic violence during the democratic transitions so that's the way we understand this one so we try both vertical and horizontal in the context of ethnic violence so we try both but for the routine violence we only try both as well but there is no interesting result of the role of vertical inequality and routine violence so that's why we need to match the concept that we start at the beginning with the empirical strategy not just looking at significant relationships that's my answer the last one so others okay you are right when you want to explain violence everything under the sun could play different roles but here you have limitations so I'm not the fan of putting everything on the right side of the equations the way I always approach empirical strategy is you set the framework solid framework at the beginning what is your favorable of interest and limited amount of controls that you want to put in because you can't put everything under the sun in the right hand side that's always my approach after having couple of years of experience and second from the theoretical perspective of course you are right any kinds of dominant power at the local level different capability of police force for example different implementation of law and order between different regions all can play a role in variations of routine violence but the question is we are still struggling to find the data maybe this is the next project different way to explain variations of routine violence in one country context thank you for your concern but in this one we are interested in inequality and violence thank you if there's just one or two questions because we are running out of time yes, yes please Anya Rita Kedokowski still I am I have been working a lot with UNESCO at the time when the culture peace project was in going on in Africa and in other places and violence is an important issue in UN context UN women is focusing a lot on it many countries I think Australia also might have been focusing on it so I am worried about not seeing the obvious links between cultural violence and violence so you can't split it into small pieces there are so many reasons why the low level what you call routine violence it can be very brutal as you know women die of it it's called Femicide in Latin America I do discuss it so in that sense I don't think there is a general violence without speaking about the context where the cultural violence or the cultural peace is performed and of course national statistics on violence as far as I know Indonesia has ratified both UN conventions and especially the convention on the elimination of all kinds of discrimination against women and also other conventions that would make it important to have the statistics I don't think we agreed in 1977 already in United Nations that the statistics have to be gender segregated thank you I fully agree with you with the concern that you just said my response would be this is clearly a future agenda say for example from the data set that the has developed so far they recorded each incidence of violence with all variables or characteristics embedded to that particular incidence say for example who I involve what kind of interventions done by the police or whatever and what is the outcome so I think to extract the gender dimensions from the last database I think this would be a very fruitful way to really try to reclassify the data but that kind of atom has not been done but this is a very important agenda to really look at the database to consider that kind of coding system so the data is there now is to go to the thousands of records of the violence during that in such a large country and over a couple of years each of the particular violence records and find out the violence targeted against women for example so that has not been I mean I'm not sure whether the organizations that is dealing with the data I'm not part of the construction of the data I'm the customer of the data so I think this will be something that I can say look this is really important okay so this is really the future agenda thank you very much we need to bring this to a close I would like to thank the presenters very much for the presentation and for all of you for your active participation thank you once again