 I have the pleasure of introducing our speaker for Science at 10 today. Our speaker is Stibniaty et Maja. Nia is a scientist in the livelihoods group and the livelihoods portfolio. And she's going to speak today on the link between deforestation and poverty. What is the consensus? A question of tremendous interest and tremendous importance all through C4. So I won't delay this presentation any longer. So please, Stibniaty et Maja. And thank you everyone for coming. Welcome to Science at 10. So today I'd like to discuss a chapter that I recently wrote with Dr. Aaron Sills from North Carolina State University. It's about the link between deforestation and poverty. And we often use large concepts such as green growth or sustainable development. And these assume some kind of relationship between economic development and with environmental protection. But do we really know the relationship between these two? And so this is the reason why we undertook this meta-analysis. In the literature there are three hypotheses that usually people use to link these two relationships, these two factors. So the first one is win-win. So if you increase income, you increase forests. So everyone wins. And if you model that, it would be a relationship where income has a positive coefficient. So one increases, another one increases. And this assumes that poverty is the root of deforestation. So if you reduce poverty, you reduce deforestation. And the second one is called the win-lose hypothesis. So where you have increasing income, but decreasing forest cover. The relationship is therefore negative. And this assumes that when you have increasing income, income is a limiting factor for forest conversion. So if you have more income, you are more able to convert forests. You are also able to demand more products that drive deforestation. And the third one is what I call the turning point hypothesis or commonly known in the literature as the environmental cruise nets curve for deforestation. It says basically the relationship is a curve, it's not a straight line. And that initially when you have low incomes, the relationship is negative. So you need more forests to kind of gain momentum for increasing your income. But there comes a point where you're sort of rich enough and then you can reinvest back by protecting the forest you still have or putting back forests. And this is something that we often see when we think about the forest transition curve. For example, when you have high forest cover and then the story goes that it will decrease and then there's a turning point where there's a happy ending and you have more forests. So this is sort of the turning curve hypothesis. And we have about 20 years almost of publications that statistically link the deforestation and poverty factors together. And so we thought, okay, it's time to do a meta-analysis of the literature. Meta-analysis meaning that we use other people's studies as our data points and then analyze what are the consensus from those studies on which of the three hypotheses they generally support. And we gathered about 71 publications from 1994 until 2012 that do this statistical analysis that link the two variables together. And of course there are other explanatory factors, each studies are different but they have these two variables statistically linked. And out of these 71, some have models for different regions, different periods and therefore each study can produce more than one observation because these are different observations. So from the 71 we have 110 observations in the end who report their income coefficient meaning their, you know, the effect of poverty on deforestation or income on deforestation. And out of the 110 there's 39 studies that report also the income squared coefficient which is useful to test this kind of curvature, this turning point. So the win-win and the win-lose hypothesis can be tested on the 110, the entire data set but then the turning point hypothesis can only be tested on the 39 observations that are subset of the 110. And, you know, each study is set up differently. You can imagine yourself writing and it can't be the same thing that other people write. So these factors may also influence why one study has a different result from another. And so we also extract data from each study about these characteristics, these study characteristics. So for example, your publication, what was the format of the publication? How many observations did they use? What kind of, where was the data taken? What region was it? What kind of level of analysis did it undertake? The unit of observations and so forth. And so after doing that I like to share some of the results. The first thing I found interesting was across this 19 years on average there was, there was 3.7 publications every year that did this statistical relationship between poverty and deforestation. Personally I think that's, I know, that's a lot because this is very specific. It's an econometric analysis with poverty or income on one side and deforestation of course cover on the other side. There was only one year in 1998 where we could not find any publications. The second one is we divided the observations across four regions. The first is Latin America and then after that Asia, Africa and then there's an inter-regional category where these are studies that use data from different regions and analyze them together. And what's striking is Africa is the least represented. Only 14% of our observations were from case studies in Africa. A third of it was an across regional study and then about 28% was from Latin America and 24% from Asia. Most, slightly over half or 56% of the observations were in economics publications. What I mean by economics publications is these are the publications that usually use economics in the journal title and they have, they use economics as their disciplinary foundation. The rest, 44% is what we call non-economics publications and maybe this is because we're economists so everything's non-economics but these are usually conservation, development, international studies kind of journals and also a lot of sociology journals as well. We found a minority of these studies were household level. So a lot of them, like about 70%, no sorry, I'm sorry, about 60% were country level data. So they use aggregates of let's say Indonesia and Philippines and Thailand and whatever across many years and then run models over that. Instead of household level studies that we often see with Penn and with Global Comparative Studies module two. So yeah, and most of the observations use GDP as a measure of income. Not what we would perceive as real income, individual income or well-being some sort of measure of things like that. So which of the three hypothesis is supported? 45% of our observations show there was no statistically significant link. What does that mean? I'll discuss that later. And more than a third, 35% supported win-lose. About 21% supported win-win. And so this kind of idea that there's sustainable development, there's a green growth. The policy dialogue that's happening is not matched with the literature that we have reviewed because it's only 20% of the observation and compared to almost half that says there's no statistical link. There's also some potential publication bias that we found. So imagine you see someone smiling and there's a missing tooth. So this is what we found in the data when we kind of did a curve of where all the publications are. When you compare economics publication, the tooth is all there, but then when you compare with non-economic publication there's something that's supposed to be there but missing which is this idea that there's a negative relationship between poverty and deforestation but it's not really statistically significant. So in economics it's all there, but in non-economics publications that one is sort of missing. And my take on this roughly is that if you have a win-lose result and you try to publish it in non-economics publications, it's less likely to be published if it's not statistically significant. So it better be statistically significant, that's more likely to be published. Roughly it's sort of something like that. I don't think this is particularly, how do you say, someone meant to do it, but the data just shows that. There are also regional differences. In Latin America we find some evidence of win-lose and with a subset of 39 observations this turning point. But in Asia, Africa and these inter-regional studies we find no clear relationship. So my take on that is that maybe in Latin America the process is different whereby you see more of this clear win-lose or this kind of like forced transition curve but actually in Asia, in Latin America, and if you see the world as a whole, it's not clear. And so the caveats, the caveats, we didn't look at all the literature. We looked at only the econometric models and we only looked at published articles and working papers, not thesis and dissertation. So there are some limits to this study. But in conclusion, we found a potential publication bias amongst non-economic journals and after 20 years of linking force and income statistically, this relationship with win-win isn't clearly, it doesn't pop up. What pops out is not statistically significant and what does that mean? It may mean that one, the variables that were used were just not the good one to capture the link that we logically see between income and deforestation, GDP. I mean GDP is quite far removed from all the decision-making involved in deforestation and land use decisions. And also maybe the level is wrong. Maybe it's very context-specific. You may not be able to see it in a cross-country kind of comparative study in comparison to very small-scale household level studies. But then again, there's very little household level studies out there and so that's why there's a need for such a thing. Also there's also a need for work done in Africa because that's also something that's missing. And that's it. Thank you very much for listening. Thank you very much. What I think is not only is the study fascinating, but this presentation was all done without a single figure, without a single PowerPoint which I think is a marvel of communication. So thanks very, very much. Actually that also holds for our other always data-rich presentations at Science at 10. So now the floor is open to questions. We have 10, 15 minutes for questions. Do I see any questions popping up? There's one right there. Peter. Thanks very much. Nair, I endorse Christine's compliments. So what should policy makers do based on your meta-analysis? What's your recommendations? Don't assume win-win is correct and therefore brace for the idea of win-lose or a much more complicated relationship. I just don't agree that these kind of statistical analysis should use income or GDP, things like that, that have no, how do you say? I cannot feel there's a relationship. But yet it's like year after year, you find it in publications. So brace for the fact that there's a trade-off there and it's difficult. So can I have a supplementary question, Christine? Do you think the methodology, are the studies using the right methodology to get at the question that we're really interested in? I think each of these studies did not specifically look for the deforestation and poverty relationship. If I had to do that, there would be very little number of studies that I would use. They may have looked for some other thing like the effects of corruption or some other governance or demographic factors and income was there just to control for something. So that was the reason why we had these observations. So I can't really say that this is a good or bad method because it was not specifically the reason why the researcher used this method. Thanks for an excellent presentation. Hypothesis, is it possible that the diversity of findings is simply the fact that the scientists were making observations at a different place along the forest transition, therefore logically you would have a diversity of findings? Let's think about it. So you are saying that in different forest transition positions you are supposed to have different relationships? Yeah, a different way of stating the question is is it possible that the third proposition is true, that forest transition is the way the world works and because these studies were made at a different point then logically it's what the meta-analysis is collecting is just a different set of points, measurements along that curve. I haven't thought about it in great detail but that is indeed possible. We tried to control for that somehow by taking into account the periods where the data was taken from but these are 71 studies. Some are taking data from a period of 20, 30 years. Some are taking data for just a period of one year and that was also included in our meta regression model which is like putting everything in the kitchen sink and seeing which variable pops up. So we tried to control for that but it didn't seem that this could specifically target your question so I don't know and I think that's a really good point that we should think about further. Thank you very much for your presentation and I would like to ask you whether you control for international trade for those cross-country studies? What we did was, so these are very different studies. Particular studies may have controlled for that but in the meta-analysis itself we can't control for that because one study can be interregional and it involves 40 countries. One study could be just four villages so it's not possible to specifically target things like trade or things that are specific to this idea of the link between forest and income. I'm just saying this because I've done some research related to the forest transition in China and basically what we found is that there's been a forest transition in China due to implementation of conservation policies and a logging ban but at the same time there's a displacement of the forestation towards other countries mainly driven by international trade imports to China and then re-exports to Europe, Japan or the United States. There's a lot of studies out there that also control for international trade or for international market prices of particular goods I'd say but these are the individual studies that we then take in and then see, okay, you've had 20 years worth of publications let's see what everyone's saying. So it's sort of like a voting and it's a very blunt instrument, I know, but I think someone has to take stock of this vast literature. Thank you. Thanks, Nia. I'm just wondering, you mentioned that there was relatively few studies that had the household level data but I'm wondering whether you partitioned those out and whether you've got any observations just from that set that might be different from ones that dealt with Corser, GDP or whatever. I did consider that because theoretically you can't possibly think that the decision making at the country level would be the same at the household level but there's only very little number of household observations we had 14% of 110 something is only 15 studies so it's not, I mean I could tell you later, averages and so forth but I can't say that in general this is speaking about something because the numbers are low. Thank you. It's a very interesting presentation. Thank you. What I have seen is the different approach in analysis. One is more quantitative, you said in economics, the other one is qualitative. I think both has advantage. So instead of confronting of the two methods which come up with the different conclusions, do you think it's possible to sort the story from the qualitative approach and then to enrich the conclusion that maybe we come up with different conclusions from the three hypotheses that you propose. Yeah, okay. So I'm trying to understand. So the three hypothesis is really rooted in the quantitative idea. Yes, that is exactly correct. I'm thinking of things as if they're all equations but of course humans are not equations and that is why I think the quantitative supports the qualitative which is that a lot of them say, you know, if you stick them all in a statistical equation, a lot of them find no relationship and that's not because it's truly no relationship but maybe because it requires a qualitative explanation instead of putting things in the kitchen sink and running it. So I agree that the qualitative and the quantitative go hand in hand in this sense. If there's a big reoccurring trend, then maybe the quantitative approach can help sort through a large amount of data. But then again, when you find that, okay, this doesn't work, then the qualitative can say, okay, well, you know, maybe it's going through a different path. And also even if you find, you know, positive relationship or negative relationship, you still need the qualitative story to make sense of it because sometimes the qualitative story can tell you whether the statistical model is correct or not. It doesn't make sense. It can't be, you know. So those are complementary. I agree. Since I don't see anyone struggling to ask one more question, let me just ask for those of us who really would like to see those figures and would like to see a little bit more detail where can we find that and when will this be published and when will you make it available to all of us? The Tropical Forestry Handbook is a chapter. So when it's available, of course, I will share it with all of C4. Well, thank you all very much and great, great thanks to Steve Nianti for this really wonderful presentation. Yet another absolute triumph for Science at 10. Thanks very much all for coming.