 Thank you, and thank you, Vault, for turning up this afternoon for, I think, what's been a very stimulating set of papers thus far. As many of you in this room know, it's widely understood that armed conflict exacerbates hunger and undernutrition through multiple channels, and this is drawing on work, amongst others, done by people here in this room, in terms of destruction of crops and livestock, the destruction of infrastructure needed for food production, the displacement of populations, the diversion of labour out of agriculture into armed conflict itself, and also reduced public expenditures in agriculture. And there's a very large and rich micro-econometric literature on these topics, and I believe in the panel discussion after this session, in fact a lot of that work is going to be reviewed and discussed further. By contrast, at least to the best of my knowledge, there's less work done on aggregating these effects up to a national or global level, and you might ask yourself, well, why might that be valuable? To give you one reason, we know that if we look at either the prevalence or number of individuals around the world who are considered undernourished or considered hungry, that those numbers declined more or less monotonically between 2000 and 2014-2015, but it subsequently started to rise afterwards. At the same time, we're also aware that the prevalence of armed conflict has actually increased over time, with a larger number of countries seeing those types of events occur, and there's been a suggestion in the data, and in fact there's been an uptick in that after 2015. So that would suggest that these increased prevalence of conflicts might be one of the driving forces underlying that particular trend, and that's one of the things I want to try and look at, obviously in an associational way, in this study. So what I use is the dependent variable, is something called the prevalence of undernourishment. For those of you in this room, for example, who are associated with the UN, I'm sure you have all the Sustainable Development Goals memorized, possibly even their measurement. Sustainable Development Goal 2 relates to the prevalence of hunger and undernourishment, and this measure, which is an estimate of the proportion of the population whose habitual food consumption is insufficient to provide dietary energy levels that meet individual requirements, is actually the outcome measure I'm going to use in this study. The main explanatory variable is the number of battle-related deaths per million people, both military and civilian, as reported in the UPSLA Conflict Data Program. I'm going to restrict the sample to low, low-middle, and upper-middle income countries, where there is data both on the prevalence of undernourishment, which to use one word rather than three going forward, I'm just going to say is hunger, and this measure of conflict. And so in other words, to be in this sample, you need to be a country where both those data are available, and where at least one year over this period, there was an armed conflict that took place, as defined in the UCDP data set. So we end up with a country-level sample for 52 countries and for 19 years. Those of you who are multiplying quickly will realize that 52 times 19 does not equal 963. There's a few years where we don't have the POU data, and so obviously those are dropped. Our data and model setup, we're going to estimate in a very simple model where these measure of hunger is our dependent variable, the measure of our conflict is on a right-hand side variable, and at various points in time we're going to control for country-fixed effects and certain time-varying country characteristics, more of which I'll explain in a sec. To give you a feel for these data and how they turn out, if we look, for example, at the trends in the prevalence of undernourishment, you see on this graph a red line, which is the median prevalence of undernourishment, or median prevalence of hunger in this particular sample, it declines over time of the period we have available to us. That's also true if we look at the 25th percentile, but if you look at the 75th percentile, the distribution, you see that does decline as well, but in fact that tails off right around this time where we begin to think more globally, the prevalence in number of people considered hungry begins to level out. If we then look at our data on battle-related deaths, we notice in the first instance that the percentage of countries that report a battle-related death in any one year is rising over time. So remember, in this particular sample to be in the sample, you need to have a country where in fact a conflict occurs at least once over this period. That means of course that not every country is observing a conflict in every year, but over time in the sample the proportion of countries where these conflicts are reported is rising. If we look at the numbers, it's important to look at not only the absolute numbers of battle-related deaths per million persons, but also how the distribution of those is changing. As you can see, if you look closely enough, at the 10th and 25th percentile, there's a relatively small number of those deaths that occur. The median also is also relatively small, but when we look at the 75th and particularly the 90th percentiles, we see big variations from year to year in those data. If you put the left-hand side of the panel together with the right-hand side, you rise in a statistical sense we have something of a problem, that on the one hand we've got a significant number of country observations where we've got zeros, and a number of observations where we have very large numbers. Now we could take a logarithmic transform, which would help us on the second problem, but that is arbitrary on the first problem because of course you can't take the logarithm of a zero. So what I do instead is I use a transform called the inverse hyperbolic sign, which allows me to actually transform these in ways that address both the zero problem and these observations with the magnitudes of the battle-related deaths are very high. We put all these together, and these are basic results. And if you look at the first column, this is just the simply the bivariate correlation between battle-related deaths per million and the prevalence of undernourishment. That number just vaguely looking at it looks really large, but of course we absolutely don't believe it because there are no other controls and we could believe that lots of factors which are associated with the likelihood of the levels of these conflicts taking place could also be directly affecting our outcome of interest. So we add in our country fixed effects, which is what we see in column two, and we can see that parameter estimate drops a lot, but remains statistically significant. But then we start to add in other controls. So we add in year controls in column three, changes the parameter a tiny bit. I add in a couple of country-level time-varying controls, including log population, but also the magnitude of other shocks that occur. So we might imagine, for example, that civil conflict could be correlated with natural disasters, such as droughts or possibly flooding in other cases or other such events. I use the MDAT database to quantify those, and when I add that to the specification, we see again the parameter estimate doesn't change a whole lot. Finally, we might also believe that in fact that government effectiveness could also be correlated both with this outcome, but also correlated with conflict. Adding that in again doesn't change this parameter estimate very much. So in other words, in order to try and get a fix on the magnitudes of these in an associational sense, clearly the fixed effects are really important. Afterwards, trying to control for other time-varying effects helps a little bit, but is not necessarily super important for getting our results. If you want to interpret the coefficient on the fifth column, in some sense the preferred estimate, what they tell us is that a 10% reduction in battle of deaths arising from armed conflicts is associated with a reduction of 4.9 percentage points in the prevalence of hunger. Now, like all my colleagues up here on the podium, I believe the phrase earlier was we subjected the results to, quote, gazillions of robustness checks. We do so here, and the basic results don't change a whole lot. Rebusiness checks in terms of the measurement of battle deaths, the exclusion of certain countries, the specification of the non-conflict shocks, and so on and so forth. It's worth mentioning that when I described the sample, I noted that it consisted of low, low-middle, and upper-middle income countries. If we look within those groups, the effects are concentrated in the low and the low-middle income countries. In the upper-middle income countries, these associations are much, much weaker. Also, if we disaggregate the sample by the frequency of armed conflict over this period, we also see a relationship there. Is this, for example, we restrict the sample to countries where battle deaths occurred in six years and more over this 19-year period? We see a larger association, namely a 10% reduction in conflict deaths is associated with a reduction of 8.1 percentage points in the prevalence of hunger. So putting all these things together, we can do the following. What you see on this graph, you start off with the blue line. These are the number of people considered undernourished or hungry everywhere in the world between 2001, which is the start of our data series, and 2019. I then construct a counterfactual. And the counterfactual is to say, in a wonderful world, what would that trend look like if none of these battle related deaths occurred? And that counterfactual gives you this orangey line. And what you can see if you start at 2001 is that number is somewhat lower. It's about 50, 60 million persons less than the actual number we observe. And the two more or less basically track each other over the first 10, 14 years of the sample. But then you look over the last four or five years, and you can see globally, as I mentioned in the introduction, the numbers of people undernourished globally is starting to rise again. If you look at the counterfactual estimate, it looks much flatter. Suggesting that, in fact, that the association between hunger and conflict in terms of numbers of persons affecting is rising over time. Because the number of people undernourished in the world generally has been going down over time, that also suggests this association implies that of those who are considered undernourished, a larger fraction of individuals are living in places where these conflicts are occurring. And that then takes me to my summary, that after we control for country and year fixed effects in a range of time-varying characteristics, a 10% reduction in battle deaths is associated with a reduction of 4.9 percentage points in prevalence of undernourishment or hunger. This is robust to all sorts of things we can do in terms of econometric specification, sample composition, and variable definition. They imply that armed conflict contributed to an additional 86 million persons undernourished in 2019. And globally, the number of persons whose undernourishment is associated with this armed conflict as a percentage of all undernourished persons rose from 7.5% in 2001 to 13.3% in 2019. Thank you very much.