 Mae'r gwneud o'r boblmwysig. Mae'n ddechrau â'r boblmwysig zerwyd ar dwi'n gyda'i gwasanaw ythodol. Mae'n gwybod gyda'r boblmwysig. Mae'n ddigwyd ar gyfer mae'n gwasanaw peri wrth eu digwydd. Roeddwn ni'nesc gyd o fffaith trefni'r boblmwysig ar yr teimlo. Atenu i'n ddysgu'r bwblwysig ar y ddysgu'r boblmwysig. ond, fel gwneud hynny, sy'n mynd i'r holl gweithio i'r mycroeconomau a'n gweithio'r ffyrdd hwnnw i'r gweithio'r ffrwyngau. Felly mae'r syniadau gyfawr ddweud o'r swyddfaeth, o'r ffyrdd i'r gweithio i'r mwyaf ar y lawer o'r ffyrdd hwnnw. A yna, mewn i'r gweithio i'r wneud o'r gweithio i'r wneud, mae'r gweithio i'r gweithio ffyrdd hwnnw i'r wneud. Mae'r bobl yn ni i'n ddweud o'r economaeth a'r amddiffyniadol o'r ddweud a'r ddweud o'r economaeth a'r ddweud o'r achosol. Rwy'n rhoi'n iawn i'n edrych arall, ac mae'n credu ei bod yn rhan o'r 10 mnwysydd o'r bod ni'n rhan o'r ddweud. If you could identify yourself and say where you're from, so if we could take the Yong Fu first and then the gentlemen over there. Yong Fu? Okay. I'm Yong Fu Huang from UNU Wider. I have two questions for the first speaker and two questions for the second speaker. For the first presentation, I think your presentation is very comprehensive, right? You have a theoretical model and empirical studies, but I'm afraid I just couldn't see the connection between your theoretical section and empirical sections. I think this is likely because I didn't get your point. For your theoretical sections, I think this is a typical business cycle model, right? But I just didn't get a point how you model the shocks, right? You look at the two kinds of shocks. One is income dimension and long income dimension, right? Typically, we model the shocks to technological progress, shocks to capital, shocks to labels, right? But I didn't get a point. Could you give the details how you model the shocks? In terms of your empirical sections, because you only report one full level in one table, right? I think it would be good if you could report full model because in your theoretical section, you can see the label, capital, all kinds of assets, right? It would be good if you report full model in groups or full levels with some kinds of, with some necessary statistics for areas of squares so that we could have an idea in terms of if the models are well specified. Otherwise, we can completely go, no idea about the model property. If we compare the ordinary square in fixed effect, right? And I just, because the coefficient actually, the fixed effect is supposed to be downward bias. The sample, even the panel is short, but very often I see the coefficient is larger than always. Right. Sorry, for the second presentation. Actually, I think what I'm going to do is, does anybody have questions as well on the first paper? Because you're asking quite complex questions and I fear we're going to lose the logic of the argument. So is this on the first paper? Yeah, let's take the first paper actually. And then, you know, we can get, I think we should do it that way because we're going to end up in a bit of a complex story. Yes, great presentation. I think you have raised one of the important issues with regards to child labour and the commodity shocks. The question is on the proxy that you use for parental income. So they have been, if you look at the literature, they have been a lot of critics that, for those who are using the parental income and the schooling, the parents' schooling. So could you please reflect on that if you have taken other measures also into account? And why fixed effects? Why did you use fixed effects? You always get two questions from wider staff members. So is this on the first paper, sir? Yes, please. Yeah, I think it's the best way to do it. I'm Steven Kerama from the University of Jerusalem. My question is exactly on the definition of child labour in Tanzania. Because I know in the findings you really indicated like child labour on agricultural activities. But essentially we have quite significant child labour when it comes to mining, especially in some regions. Our child labour when it comes to Lake Victoria, which has been reported quite significantly. Now I would like to know the extent of the child labour definition that you use to Tanzania to see if it really reflected what we understand it's taking place back home. Thank you. Okay, thank you. So any further questions or comments on the first paper? Susanna? Please just say who you are for the record. I'm Susanna Sandstrom from Mobile Academy University. This is just a thought. I was wondering whether you could further specify the type of shocks into sort of covariate and idysacrotic shocks. Because I think that households response to shocks differently depending on if the whole community is affected or if it's a very household specific shock. Thanks. Okay, thank you, Susanna. Okay, so I think that sits on, no, we have a further question on the first paper, please. Thanks. Christophe Millar from the University of Ex Marseille, France. One issue with crop shocks is that they are not completely random. You know they depend on the specialisation decision of the household on the organisation of the cultivation, which may involve labour intensive techniques. In Tanzania if you choose to have some association of crops, as it's often the case, you need to have all the family getting involved in the work. So maybe that there is an important degree of simultaneity between the occurrence of the shock on the decision of involving the children in the work. Okay, thank you. Okay, so that's a very rich set of questions for you. Everything from estimation methods to the definition of child labour. Thank you. Thank you for the comprehensive and very elaborate comments. Some of the things of course we can answer now. Other things of course we may have to explore a little bit further, which we might not have addressed properly. On the connection between the model and the estimation, how do you model the shocks? Now if you look at the model, the shock is coming through the budget constraint. Because in the model we are using, we try to use the consumption smoothing model in the general literature we have. The shocks is kind of random. It enters the model through the budget constraint. Usually there are literature which uses that way and probably we may have to explore a little bit further as to whether it is random. Going back to the example some of the remarks I made to see whether the shocks are random, period one shock has some correlation with period two shocks have some correlation with period one. So meaning it is not necessarily random but the reason we wanted to go ahead was the magnitude was very small although it was significant at 10% level. So our assumption was that since it is very marginal it shouldn't cause a serious issue. Only report one variable. In fact the reason why I just put the particular variable that I am interested in is to minimize the number of slides. In the paper we have full set of variables that we are using, all control variables and fixed effects to see what the levels are. But it is just to minimize the time taken to explain or go through and also the number of slides. Proxy for parental income. Here again we thought that this is, as I explained, one proxy we could get out of the questionnaire was the level of education. The assumption here is that the higher level of education will lead to more income of the parents, the higher probability of earning more income. So we have to go back to the questionnaire and see whether there are other proxies, whether we can use them. Definition of child labour. Now in this model we use schooling age child from age 7 to 15. Now they are enrolled in different types of working. Say for example in this particular case looking at also at the questionnaires we can see working at home as household labour. Working in the field in agriculture. Working for wage labour. Three things we can get out of the questionnaires on the type of work that children use. So depending on how children spawned, whether you worked during that particular period, in any of those we take it as providing child labour. Specify the type of shocks. Now I mentioned about two shocks. The income shock that we use here is any negative impact on the earnings from crops. The question we particularly ask is whether there is a say due to floods or droughts or any fiscal weather that had an impact on the earnings from the crops. Whether there is a failure of crops. On the non-income shock that we used is the depth of the immediate family members, specifically we are looking at the mother or the father. Crop shocks. I think I addressed that issue of crops of whether they are completely random or not. Labour intensive techniques and like that. Again it is the same explanation that I can give. The impact is very marginal so we assume it is random. But given this say for example the limited time span is only two year gap. It might also have an impact on the results. Thank you. Let's now take questions on the second paper. Could we take the gentleman in the blue jacket hasn't spoken. I'll take him first. Please sir, could you identify yourself, say who you are? I'm Tomoke Fuji, Singapore Management University. I found the presentation quite interesting but I wasn't clear about the price index that you have used. You mentioned that the price movement differ across oddities. I wasn't quite clear whether you have used raw price data for your analysis or some kind of price index. If the latter is the case it would be interesting to look at the differential shocks across different commodities. Different countries may have a different response to different commodities. I'm not quite sure if you have addressed. Thank you very much. Let's take Yang Fu if you could be brief. Thanks for your presentation. First question. In your introduction when you were talking about the Morghe model in Hamilton model I found it very interesting that you talked about the X plus term. This is very interesting. Could you give some more details on that? Especially the intuition for this term. Secondly, you use the traditional model for approach to study countries by country. Many countries. As you know, the underlying assumption for this approach is that there is no correlation across countries. There is no correlation. Countries are uncorrelated. Those African countries are highly correlated because of the globalisation in trade. I wonder if you have ever considered using Panova to study this group of countries in Wango. Now you study one by one. I suppose if you use Panova to look at this in Wango you will see the difference. Thank you. Can you use the microphone? Thank you very much for a very interesting study. If I understood correctly, you have some countries that have positive impact on GDP or growth and some countries have negative impact. In the paper you discussed, why is it like that? You are now confusing the people who are working on the area. There are some arguments where it has a positive impact on GDP. Now you find that it is not necessarily the case. What is next? What should we do? Thank you Azuz. Any last questions on the second paper? The gentleman again at the back. Thank you. I think you may have an opportunity to separate price shots on producers, on price shots on consumers including the set of countries that you look at. There are countries which are not producing the commodities that they are. There are countries which are more or less heavily consuming them. You may be able to use this information to introduce some theoretical constraints in your VAR which is going to improve a lot the efficiency of the estimation but probably to follow you some more deeper on richer causality analysis based on these theoretical constraints. That's one first remark to look at the consumer and producer side of this. A second remark is that when you look at price shocks people like to consider volatility rather than levels. So to have a component of volatility in a model like you do sometimes in arch models may help you to see things that you don't see just by looking at prices or their simple variation. Thank you for that very useful distinction. Any last questions? Thank you for all the useful comments. Regarding the price index, because there was a lack of time I couldn't explain it very carefully so I'm going to go through it a bit more carefully now. The price index that I've chosen is the Grill Yang price index which is compiled by the World Bank. It's compiled by two economists working for the World Bank, Grill and Yang. It's been used quite popular ever since. It's about international commodity prices and an index has been created and it's been deflated by the manufacturing unit value index. In a way it's like the commodity terms of trade so it's a very useful index to choose. Of course there are the comparable indices like the Deaton and Miller index I'm quite familiar with the Grill Yang index so I've just decided to choose it but in any case they're highly correlated. There isn't really much difference between the two indices. Both of them have been used in the literature but the Grill Yang has been used more than the Deaton and Miller index so they're different indices that are used. Regarding the second question which was about the intuition of this price shock it's about separating a positive shock from a negative shock so basically there's a lot of literature on this by Kilian in his AER paper which was published in 2009 and then subsequently he develops that idea even further in his quantitative economics paper in 2011. So it gives a very lucid description of how these censored vahs are created and the X plus variable is simply an auxiliary variable which separates out the positive from the negative shocks. Now what Kilian and Vig for Sennargyw is that if you just put the positive shock inside it just censor the variable in that way then the bias that you are going to introduce into the variable estimate is up to 50% that is what they show through simulations. So that is why they have proposed this particular model of this format which we have chosen here. And then about the panel one reason why I've moved away from the panel is simply the argument that I put forward that putting this into a panel would not allow me to first of all there's a difficulty in putting this in a panel because you can't apply this method on a panel framework it just becomes too complicated. It's a simple bivariate var model. I tried to include a third variable but it just made it just too complicated to analyse. It's just extremely complicated. In its current form this only works for a bivariate var model. And then to Aziz's question, what next? That's a very good question. I have tried to put forward the sort of caution. The conclusion at the moment is more like a caution that there are asymmetries in responses to these positive and negative shocks and these need to be borne in mind. I don't have any solutions as such but this would be I think the subject matter for future research but a very well poignant remark. And finally, Christof's comments. Yes, it is going to be very interesting if I could include exporters and importers. That would definitely enrich the study. At the moment I'm sort of a bit tied up with the dimension of the model. So trying to introduce the export to importer balance would be a bit difficult but definitely something to think about to ponder on and I'm definitely going to go away thinking about this particular point that you have raised. Regarding the arch type model which is about volatility arch models are mainly used for high frequency data. This is annual data. It's a low frequency. Arch models work very well when you've got financial data like daily data or monthly data. I have seen some estimations where they may be used for daily, weekly or monthly data. I've seen some where they have been applied to quarterly but it is not recommended to use it for yearly data so I have to sort of stay away from that. Thank you very much. Thank you also to the audience and thank you for the presenters. At four o'clock we have the final panel. Thank you very much.