 Hi everyone, thanks so much for being here and for inviting me to give this presentation. This is a conference on human capital and growth and so I'm gonna start by saying that health and good health in particular is a critical economic asset, especially in low income countries where people tend to work for their own so if they are sick one day and cannot go to work, that means no earning that day. But also they are often in occupation that require strong physical health and so being weak also means being less productive. But even if all of this was not the case, maybe health would be a good thing to study in any case because just being healthier means living a fuller life and so even if health was not important in the production function and not an input into growth, we would still want to care about it. And the issue here is that good health is yet, it's still very much lacking in many lower income countries. There's definitely been progress towards what used to be the million development goals which were not fully achieved by 2015 and so there are still today six to seven million children that die every year before the age of five, most of them in Sub-Saharan Africa. And what I want to highlight today is the fact that up to two thirds of this death, as per some estimates, could be averted using existing health products. So by that I mean vaccines, antimilial bed nets, water chlorination, water filters, things like that. So the key question is how to increase adoption of these products so that children don't die and older individuals don't lose on productivity. So if you ask parents of these children what the problem is, they tell you that they don't have money to buy them. So an obvious solution then is to subsidize these products. And standard theory gives us a number of reasons to do that. The first one is that many of the diseases we're talking about are communicable, which means that limiting their spread is a public good. So that justifies significant subsidies. And on top of that, you can think of it from a dynamic standpoint. Subsidizing may more than pay off for themselves. They enable people to become healthier and more productive and later on in their life they will pay back in taxes more than the how much it costs to help them become productive when they were younger in their lifetime. So the question is whether this solution can work. So for the hope for impacts of subsidies to be realized we need a number of things to hold. First, beneficiaries of the subsidized inputs have to put them to appropriate use. And so for things like vaccines, it's obviously the case because once you've been given antibodies into your bloodstream there is not much you can do to take them out and so the health impacts will be there whether you want it or not. For many other things that require that people on a daily basis take an action this may not be as obvious. So we take it for granted that we can just turn on the tap and fill out water glass and drink it and be safe. But in many parts of the world that's not the case you have to fetch water every day and then you have to put in some chlorine after you fetch the water for bed nets you have to tie it over your bed every night and things like that. So it requires some proactive behavior on the part of subsidy recipients. And then a third issue is that if we subsidize too much maybe the beneficiaries are not gonna be putting forth this effort because if they've not shown that they are willing to pay a little bit in terms of money in order to get these products maybe that means they won't be willing to pay in time or attention to use them properly. So that's the first condition for these subsidies to lead to the health impacts you will hope for. The second condition is that subsidies have to reach the internet beneficiaries in the first place. And here we are talking of governments concerns maybe the incentives that the health workers face mean that subsidized inputs will be left to rot in the storage room or they will be stolen along the way. This is just a picture I took in Uganda highlighting individuals there to the fact that they have rights to quality healthcare and they shouldn't listen to doctors trying to send them off to their private clinic instead of giving them the public care they are entitled to. So if this posture is a good representation of the incentives of the health workers maybe relying on them to implement these subsidies schemes will mean that they won't reach the people that need these subsidies. So what I wanna talk about today is share with you some evidence from mostly my own work so that we have a self-serving talk on these issues looking both at the demand side and the supply side. And so on the demand side, the question is going to be how can we design subsidies so as to balance access and targeting to those we will use inputs, okay? And on the supply side, how should you circumvent our governance issues in the delivery of subsidies for essential health products? So starting with the demand side, I just wanna put a very simple problem for what I call the principle here which is gonna be the social panel of the government thinking of a subsidy scheme, okay? So you can think of the principle valuing the health benefits of a health product as well as a non-health utility of potential recipients and alternative uses of funds, okay? And so you want to maximize the sum of this individual utilities minus the cost of fund plus the contribution value and the individual utilities function of what is the extra monetary value you get out of getting people healthier. So it's gonna be a function of BI which is the daily value of total health benefit when individual eye uses this health product appropriately, the one that you're thinking of subsidizing, Z, dollar value of the principles is just a way to monetize the health gains here. And HI is a binary variable indicating whether I use as a product appropriately, okay? So BI times Z, HI tells me the essentially gains to having individual eye use a product, UI is non-health utility, okay? And then obviously there's a cost to subsidizing the program. Let's say it costs me a total of S. What matters for this cost is what is the marginal cost of public funds, okay? So what is the cost of raising this money through taxation, for example, okay? So why do I put this here? Because when you have this, you can immediately see that the benefit to increasing marginally the subsidy by an amount DS is gonna be exceeding the cost of this subsidy if the following equation holds. And so in particular, what I want you to pay attention to is the fact that on the left-hand side, you're looking at the effect of increasing the number of people who are now covered by this health product, okay? So that's gonna be a function of the proportion of people that are induced to use thanks to this policy change. So you increase the subsidy, you change the price, you lower the price, that's gonna induce a bunch of people to start now acquiring the product and potentially using it. What matters for the health impact is the share of those that use it. So use more is really the proportion induced to use by the policy change. And then for those, the value of this change in their behavior is again, is Z that E times now B, the health benefit among those induced to use by the policy change. So these are the marginals who are affected by the policy, okay? And then on the other side, what matters for the cost is not who uses, is who actually texts it, okay? And so on the right-hand side, what you have is a share of people who are induced to tech the product by the policy. So what I call tech, mark of tech marginal. For each person who is induced to tech the product thanks to the subsidy, that costs me now the entire subsidy amount, which is S. And then on top of that, I now have to pay the extra subsidy for all of the infromarginals. So the tech amp, these are all those that were already taking the product when it was subsidized less. Now they still take it, but I have to subsidize them more. So it costs me an extra tech amp times DS, okay? All of this times the value of funds lambda, okay? So this equation is now gonna help us think through the potential trade-offs that you face when you think about subsidizing more or less, okay? So I'm writing it again at the top here. And the first trade-off that you can see that if the share of people induced to use the product by the subsidy is smaller than the share of people induced to tech it, okay? It's mean that the subsidy policy induces some people to tech the input, but they end up not using it appropriately, okay? So you have a wedge between the number of people that you subsidize and the number of people that actually use the product. So that's, you can think of it as wastage in a way because you finance this, but you don't get as much health impacts as you were hoping to. The second issue is if the benefits for these marginal users is actually low. So if as you reduce the price, those who are enticed to tech it because of the lower price have lower and lower returns because that's why they were not willing to pay much in the first place, then the further you decrease the price and the further you increase the subsidy, the lower and the lower are the health benefits that you get out of that, okay? And finally, if you have a lot of inframarginals to start with, then the cost of increasing the subsidy is gonna be high because of that, okay? So these are essentially the main thing that you wanna think about. And so when you increase the price, you know, in theory, the less you increase the subsidy in theory, the less of an issue these things are going to be, but then obviously it means that access may be considerably reduced, including you have this trade off between increasing access and keeping your targeting relatively good so that you don't have all of these low returns folks getting the subsidy. So ultimately, the relative importance of these problems is an empirical question. It's gonna be context-specific, okay? So I'm gonna share with you some evidence from some context where I've looked at these questions, but by no means does it generate, kind of like general statements about what should be the optimal subsidy level in general, okay? It really depends on the product, on the demand function for it and all of the parameters that I have here, which are gonna vary across areas, but it's good to have a sense of the extent to which there's a range of possibilities and think of various cases where one or the other of the policy options works better to try to get a sense of what are the underlying features of the demand that can help us guess exactly what would be better or worse in a given situation, okay? So first, to try to look at the first question of the gap, the wage between the usage rate among marginals and the take-up rate, the question is do marginal takers use the subsidized inputs? So I'm gonna share results from four studies. The first two look at anti-malarial bed nets. Both of these studies were done in Kenya. The first one with Jessica Cohen was done with pregnant women. The second one was done with just representative households. And what you can see on these graphs, they both have a blue line and a green line. What the blue line shows is the share of households that acquired the product, this bed net, at different price points. And what the green line shows is the share of people that acquired the product and are using it along different price points, okay? So the blue line gives you a sense of the share of takers and the green line, the share of users, okay? So this was done by randomizing the price as which people could acquire the product. So in the top study, this was done at clinics. Some prenatal clinics were randomized into giving bed nets for free or they were randomized into charging 10 Kenyan shillings for a bed net. This is a very heavy subsidy for this bed net that costs, when unsubsidized, something like 500 Kenyan shillings in local currency or at the time something like $8 or $9. So 10 shillings is a very, very small share of the full price, which is 500. And then some clinics were randomized into charging 20 shillings and some 40 shillings. It didn't go further than that because all clinics were already selling bed nets for a price of 50 shillings. They were subsidized by an international NGO called PSI. And so already they were available for 50 shillings, okay? So we just further decreased the price slowly across clinics all the way down to zero in some of them. So because of the randomization across clinics, we can look at these differences in take-up rates across groups as evidence of the causal impact of the price on take-up. So this is essentially showing you the demon curve which is properly estimated and it's not winningness to, it's not stated winningness to pay, it's revealed, you know, ability to pay. These are the share of pregnant women who bought the bed net. And as you can see from the blue line, when it's free, pretty much everybody takes one home. When it's 10 shillings, almost everyone is able to pay that. And as the price goes up a bit, it starts going down. And when the price is 40 shillings, which is not that high in absolute term compared to the value of the product, it is remarkably low, just about 40% of people taking. In the second study, households had three months to redeem a voucher, so they had time to save. So there the prices were also varied randomly this time across households. And the price range was much larger. It went all the way to 250 shillings which was a 50% subsidy because people had three months to save and redeem the voucher, the take-up rate is actually higher. So even at the 40 shilling price point from the study above, now the take-up rate is above 60%. So as people have time to accumulate money, they are better able to purchase these things. But still what you can see from the blue line on the second graph is that demand still drops down relatively quickly as the price goes up. So this shows that Econ 101 is right as the price goes up, demand goes down. But what's interesting here in these graphs is comparing the blue line with the green line. The green line tells you not just who has it but also who is using it. So this is essentially taking the blue line and multiplying it by the usage rate among those who acquire the product. And what you can see for the most part is that there's not a very big, like the share of the takers who seem to be using it is quite similar across price groups with a little bit of variation. But overall, if we look at the usage rate conditional on having the product is pretty much flat and independent of the price. And what this suggests is that by the few people who self-select into buying at a very high price don't actually end up having a higher usage rate than those who only take it when it's free. The other way around, which is a better way to think about it in this context is that those who only get it when it's free or very cheap are as likely to use it that had to self-select it into paying for it. And this is essentially because everybody uses a bed net appropriately when they get one. Not immediately, so in the top graph you can see this gap, it's only about 60% of bed net recipients who seem to be using it within two months of getting it. But in the survey we asked them whether they still had it and we could verify whether they had the bed nets and indeed 99% of people could show us physically the bed net that kept it very nicely active, they were not yet using it and they had a good reason for not using it yet. They said, some of them said I'm waiting for my baby to be born, which is not actually a very good reason because they should product themselves while pregnant so we told them that which is why we didn't do a long-term follow-up because we kind of meddled with, we gave them information out of a sense of ethical obligation at the time of the survey. But in the study in the bottom part of the graph we did do a one-year follow-up. We did a two-month follow-up and also a one-year follow-up and then a two-month follow-up we also had about 60% usage rate, but in a one-year follow-up it was much higher, around 90%. Okay, so within one year everybody was putting the product to good use. So that's the case where everyone who gets one uses it so through screening with prices, you don't really get the benefit of screening people who are gonna have a higher usage rate, you just have the downside of screening out people who can't pay, okay? So these are the first two studies but it's not always like that. So in a case of water purification product we have this study where we again randomize across individuals whether they had to pay 10 sheetings which is a 50% subsidy, 20 sheetings the full price or whether they would get a free one-year supply of chlorine, that's a water purification product. And there we find that many of the marginal takers don't use, so if you drop chlorine onto people's laps at their house, they take it because they feel a bit bad saying, no, no, I'm not, I don't want it, but then many of them don't use it. So this is a case where if you give it for free to many people you get 100% take-up but only about a 38% usage rate. So in that case we say, well, let's charge a price. So if you bring the price to 10 sheetings which is this 50% subsidy, you actually reduce the share of people who are now using by half, okay? So you are getting a much better targeting in the sense that only people who are gonna use do the take-it so the blue and the green line now are like exactly on top of each other are these 10 sheetings price point but this comes at a heavy cost in terms of exclusion errors because you have about something like 20% of people who would be using it if they could afford it now not using it, okay? So depending on how you value the health and the survival of the children whose parents are excluded from this, you may think that, okay, it's a little bit too harsh to charge this price and then lose out on so many people but the question is how can you get them back in without wasting a whole bunch of money subsidizing 60% of people who are not gonna take it anyway, okay? So we're gonna come back to that later to think of ways, alternative ways to screen and so we have some other type of subsidy in that top study, study number three that we'll come to in a minute. Before that, I wanted to show another case where we find that marginal takers are marginal users. This is a case of anti-malarial treatment. We are looking at the likelihood that people who bought an anti-malarial at a subsidized price actually took it at a time for that same disease for which they got it and this is comparing administrative data on who bought the drug and end line it down, who reported taking it, we find a very clear alignment between the two, okay? So we find that for bed nets and for multi-malarial treatment there is not much of a wage between marginal takers and marginal users but for chlorine there is definitely one. Now the second question is whether marginal users have lower returns, okay? And so here what we can do, we can take the same set of four studies and look at the characteristics of the people who self-select into paying more. Maybe everybody uses a benefit at the same rate but those who are not willing to pay for it actually don't get as much benefit from using it because they face a much nicer environment or something like that. So to check whether that's the case, we can look here at whether the rate of anemia at baseline is different, okay? And we see maybe a little bit of a gradient there but I don't know if you can see the confidence intervals but we can't reject that essentially they are the same. Anemia is one of the factors we can look at, we can look at other characteristics of the household and on the whole there is not much clear evidence that those who self-select into paying a higher price as those who are gonna have higher returns based on their underlying health characteristics on environment. If anything in the bottom study, this time we can look at whether those who are self-selecting into buying at a higher price we are less likely to be predated to start with and we find if anything is the other way around they are more likely to already have a bed net. Essentially they are richer, that's why they can afford to pay the higher price, okay? So the price doesn't seem to be a very good mechanism to identify those who are gonna have higher returns. Same for study number three, now we can look at the rate of diarrhea among children at baseline and see whether it's those who self-select into buying at 10 or 20 shootings and a few of those who do, whether they have a higher rate of diarrhea and again it's overall there's a lot of diarrhea going on that's sad news and doesn't seem to be much differential across types of households. So again here we don't see much of differences in likely marginal returns across willingness to pay groups, but in the fourth study now we see something very different. So in the fourth study where we look at the pricing for Malaya treatment, we can look at the likelihood that you actually have Malaya, okay? So we know that you bought an ACT, Artemis Incomission Therapy, we know that you took it and we also know whether you had Malaya or not because we posted any marriage who, after people had bought the drug, who say, hey, excuse me, can I test you for Malaya, for a random subset of individuals and so we actually know the true Malaya status of people and we find that it's actually not that high. It's less than 60% at the highest subsidy level. So we didn't have a free group here. The smallest price was a 92% subsidy and then the subsidy decreases to 88% and then 80%. And we find that as you decrease the subsidy level, you seem to be considerably increasing the share of people who actually have Malaya. So this is a case where by decreasing the level of the subsidy you improve the targeting, okay? And so the question is how come if people have information, good information about their Malaya status, why are they still interested in taking anti-Malaya when they know they don't have Malaya because there's actually no advantage of doing that. There's not even an anti-parietic effect of taking anti-Malaya if you don't have Malaya. And so here what seems to be going on, not that people know their Malaya status or have any information about their underlying Malaya status, it's just that the prices that we had in our experiment were such that as we moved from 92% to 88%, it's meant that the relative price change, sorry, the absolute price change for an adult dose was much bigger than for a child dose, okay? So when you reduce the subsidy, even if it's only 4 percentage points, maybe it doesn't look that much, but for an adult dose which has a higher absolute value of the price, that corresponds to a higher absolute Kenyan shielding change and that made a whole bunch of households unable to now afford the adult dose but they could still afford the child dose which is much cheaper because you need fewer pills for the children. And so what seems to be going on is that as you change the price range, the share of the, you know, ACTs that are both, the share of the subsidy vouchers that are used to buy an infant or a child dose is much higher now. And because children are much more likely to have HIV condition or having a fever, that's what gives you this apparent targeting. So it's kind of like you get that thanks to a lower subsidy you get better targeting, but it's not because you force people to self-select based on information they have. It's more, here it was almost like a mechanical thing due to the specific price structure that we had and it was almost like it took us a while to really understand what was going on and then we realized that, you know, if you look at the price of other anti-malayals for the adults, it turns out that as, you know, our cheapest price was just below, you know, a whole bunch of other anti-malayals, but as soon as we increase the price, you know, it would become really dominated compared to other malayals in terms of, anti-malayals in terms of price. Once for kids, our subsidy level was such that the price remained always cheaper for that anti-malayal than any other, okay? So it's a case where it was a little bit disappointing for us initially because we thought, oh, right, there's private information and you can exploit that and use prices as screening mechanism. Turns out it's not really the story behind this, okay? But this means that when you think of pricing, you want to actually keep in the back of your mind that, you know, if people have a choice, they're gonna possibly have a, you know, cheapest option type of strategy where they're gonna go for the cheapest drug out there. So you want to think of the price that you charge also keeping that in mind that, you know, people have a basket of goods that they can choose from, okay, and so sometimes you can use that to get some targeting, okay? So these are evidence from studies I've been involved in, but in most of their existing studies to date, price also appears to be a poor targeting tool. So, you know, there are many other studies out there now that suggest that marginal takers do not seem to have lower usage or lower returns than people who are able to pay. So this was found in the case of deworming, charging people for deworming medication does not help target the deworming drugs to those who are more likely to need deworming. Those with higher risk to pay for water filters in Ghana don't see greater drops in diarrhea incidence from using the filter. That's a study by Barry Fisher and Gitteras. Same with flip flops in Kenya, you may wonder why flip flops is considered a health product. Well, it's because of worms. If you actually do not wear shoes, you can get infected by worms much more easily from the soil. If you wear any form of shoes, you're better off, flip flops are the cheapest one. So, and vitamins in Uganda, Guatemala and India, there is a study by Meredith et al finding evidence that price again matters a lot for the demand but doesn't really help much with targeting. The one study where there is evidence of some selection mechanisms through price that goes along some form of targeting based on immediate usage is a study by Ashraf Beren Shapiro in Zambia. But here they find again that by charging prices, you select people who are richer, not people who have a higher underlying health burden. But that said, in the case of the chlorine, they find some evidence of screening on prices. We do find in the study with Huffman, Kramer and Zwayne, we do see a lot of free recipients not using. And so the question is, in such cases, how do you do to improve targeting? Price is not a good screening mechanism because few people can afford to pay, but on the other hand, we do get a whole bunch of inclusion errors in some cases like chlorine. So is there anything we can do in such situation? And so we are gonna now think about something that is a non-monetary screening mechanism and very elegantly economists decided to create an ordeal. So can we use an ordeal mechanism, instead of a price mechanism, to help screen non-users, okay? And so this is something that was first theorized by Nikols and Zeckhauser and is now being used in some cases and you can think of certain types of benefits requiring people to go through an ordeal, to redeem a food stamp in the US, you have to go to a store and show your food stamp card that can be an ordeal because it's a humiliating thing to do. In India, to access this essentially minimum income, you have to work at public worksites to the National Rural Employment Guarantee Scheme, which is a form of you having to do some effort in order to get the benefits. In principle, the National Rural Employment Guarantee Scheme is a way to make the ordeal not be a pure dead-well ass because actually something is being produced out of people having to go through the ordeal. In many cases, the ordeal is a pure dead-well ass. You just make people jump through hoops in order to get their stuff, but there is no productive aspects of whatever they are doing, okay? And what's tricky in ordeal mechanisms is that the more attractive the benefit, the greater the ordeal must typically be to screen out those who are not gonna put the product to proper use, okay? So that's may impose a really significant welfare cost, okay? But for many preventative health products like chlorine, the benefit to non-health users is actually quite small. So maybe we don't need that big of an ordeal to deter people from taking their free stuff if they are not gonna put it to use, okay? So the way to add this in the framework I showed before is to say, well now, when you have this ordeal mechanism to allocate the product, you increase, you reduce the non-health utility of the marginal, you know, takers because now for them to be able to take it, they have to pay this ordeal cost, okay? So actually this should be, DU mark should be the change in non-health utility to new takers, not new users. So it should be, so that this extra term DU mark, which is a negative thing, okay? My utility goes down from having to do this, go through this ordeal. And then for all the inframarginals, we're already getting the product. Now they have to go through the hoop as well, okay? And so you're increasing their non-health utility by having them pay this ordeal cost in order to get the thing that they were getting in the first place, okay? And so the question is, is it, do the targeting benefits of that outweigh this cost, okay? And this is gonna, you know, you have some chance of this ordeal being useful only if there's heterogeneity in the relative cost of effort and money. So for example, due to different wage levels, so it could be, you know, the poor, you know, don't have money to buy these things, maybe they have plenty of time on their hand because they are poor, because they are unemployed or something like that. And it's really a joint distribution of this, you know, relative cost of effort and money and willingness to use that are gonna be determining the extent to which you want to screen through price or an ordeal, okay? So in the chlorine study that I mentioned earlier, study number three, we actually had one extra group beside the free and 10 shillings that I showed on the graph earlier. We had an extra group which was 100% subsidies which was free, but you had to go through some hopes to get the free stuff. It was one year supply like in the other free group, but this time you had to get it by going monthly to a shop to redeem a voucher. And you got 12 vouchers, they were numbered, they were one, two, three, blah, blah, blah, to 12. In January, you'd have to redeem voucher number one and in August voucher number eight, otherwise it would not work. You had to keep them well organized and not lose them. The average distance to the shop where they could redeem was about four kilometers. For some participants, the shop was in the nearest market. So to the extent that they go to the market anyway, that would be a relatively smaller ordeal than for those who would not go there anyway, okay? And so we can compare free distribution, free through this, we call it a micro ordeal because it doesn't seem like such a big ordeal to have to go to a shop to redeem the voucher, so we call it a micro ordeal. And we compare that with this free delivery where we just dropped a one year supply on two people first when they came to the clinic, which is also where the other women got the vouchers. We would give them six months' worth in the forms of a big bottle that they could take home and then we went back to the house and give them another big bottle after six months. And so this is the coupon redemption over time. You can see that the first coupon everybody is excited about. We get over 70% redemption, but then as soon as you get to the second month, it goes down to less than 50% and then plateaus between 40 and 50% for a while, then it stops tappering off as things happen in people's life and maybe they forget about it, okay? But so what this shows is that very quickly, maybe the first coupon everybody wants to try it out and then they realize chlorine is not for them. It gives them, it leaves some taste to the water. You know, it's maybe, it's hard to figure out how to do this. Maybe the instructions say you have to put one cup full, one cup, not one cup full for a 40 liter thing and your container, you don't exactly know its capacity because it's some clay jar and you have no idea and you feel like you're doing it wrong. Stresses you out, so you decide not to do it anymore. You learn that after one coupon and then you stop redeeming, okay? And so what we find is that this coupon micro-order deal reduces inclusion error without increasing exclusion error. So if we look at the leg deal that you actually have chlorine in your water, I told you that in a free delivery group earlier, I showed you it was only 38%, even though 100% of people had taken the free stuff. With the coupon micro-order deal, we have the exact same share of people who have chlorine in their water. The follow-up, 38%, okay? So essentially, you don't lose anyone who would be using but you just lose all of the people who are not gonna be using, okay? So this is a way to actually target precisely the users, okay? Then you can say, well, is it really the same people that you get with the coupon? And so we don't know that for a fact because we have no one who was in both treatments, it was, you were in one or the other, it was randomized, but we can look at whether the direct baseline is different across groups and we don't find any difference, okay? So compared to the early graph, I just added the free coupon group and you can see that in this case, it seems to be a great way to target, okay? So obviously what's tricky here is that the size of the ordeal is really something you can choose and it can be hard to choose it right. We were very lucky in this case. We just somehow get it exactly right out of sheer luck, okay? But we can exploit variation in our data in how close you were from the shop where you had to redeem the thing. In particular, if you could redeem the coupon at a nearest market, we find a higher redemption rate of the coupon than if you were not close to the market. But then, so if you could redeem at the nearest market, the ordeal is less. As a result, there's a bigger gap between having chlorine in your water and having redeem the coupon than if you're not near the market, okay? But then if you, for people who are, for whom the ordeal is bigger, now there's fewer people with chlorine in their water. So you lose four percentage points. I'm sorry, I'm far from the thing, but you see from 37 to 33, you lose four percentage points of people who would have used the product that had been easier to redeem the voucher. And so then you have to trade off losing these four percentage points of households with getting more people, getting the coupons redeemed even though they're not gonna use, okay? And so how you trade that off is really a function of how you care about stuff, okay? And so what we can do with this is we can calibrate the model with assumptions on the health impact of water treatment, the cost of the policy and a whole bunch of things and identify regions of the parameter space for which a given policy is preferred. And we find for really the most plausible ranges of, you know, evaluation of a dali and for, you know, the ordeal cost size that we have in this case, using 100% subsidy with a micro ordeal is vastly preferred to either free delivery or only a 50% subsidy, okay? So adding this type of, you know, subsidy scheme really vastly increases the likelihood that you want to use it, actually. And so this is saying that as long as you value dali saved at least, you know, it's even less than like $200, you're gonna want to do the 100% subsidy with a micro ordeal, unless the cost of the ordeal is really high, okay? So I'm not doing it on time, so I'm gonna try to speed up. You know, the relevance of this type of mechanism really depends on the characteristics of the product, okay? So if the incidence of non-health use is very low, then an ordeal is not, like that is not a good idea. So in the case of the badness that it was household that I mentioned before, study number two, actually people had to redeem a voucher or a coupon in order to get the subsidy, okay? So there was implicitly an ordeal in there, but everybody, if I gave people a coupon for a free benefit, everybody redeemed it, okay? So there, you know, everybody redeemed the coupon because everybody's gonna use it if they get the product. So why half people go through an ordeal for that? That's a pure, you know, dead well off. So you don't need to do that. Alternatively, the private returns to inappropriate use are very high. Again, you know, the audience is not gonna help. So in the case of the anti-malayal treatment, I showed you a whole bunch of people redeem a coupon for a cheap anti-malayal, and then they take it, even though many of them don't have Malaya. In fact, among adults, we find that only, you know, 40% of adults who take an anti-malaya have Malaya, okay? And they did go to the store to redeem a voucher for that. So they pay the ordeal cost to do this. So the prime here is that the value of presumptive treatment is extremely high. So even if you think the likelihood that you have Malaya is only 10%, you wanna get treated. If I told you you have a 10% chance of having malaria, you wanna appeal, I'm sure all of you would be like, yes, okay? So it's because there is a lack of access to reliable diagnostic tests that, you know, people are in a situation that the option value of taking the treatment, even if you have a low chance of having the disease is extremely high. So there, the ordeal doesn't help, okay? What you need instead is better diagnostic tests. And sometimes the ordeal can be too costly because of the nature of the products, okay? So I have an example where we give people vouchers for condoms. Absolutely nobody wanted to go redeem a voucher for free condoms at the local store because the local store is manned by, you know, grandma or the, you know, friend of my father and I just don't wanna, you know, people in the community to know that I'm interested in condoms. So here we find very low redemption rate of vouchers for condoms, especially for adolescent girls, but the same type of individuals if we go to their house when the parents are not around and drop 150 condoms on them. And yes, I've done that. Here we find that, you know, many people actually take them. Not, you know, for girls a bit hard to take 150 condoms at once and have a place where to hide them from their parents. So only half of them actually took all 150, but almost all of them took at least some, okay? So here, you know, something that you could think of the microdeal of just going to the store to redeem a voucher can be a macrodeal because of the social, you know, the social capital implications of that, okay? So very quickly, in the last five minutes, I'll talk about the other question on the supply side, the delivery side. Let's say, you know, you've identified a product for which a given subsidy scheme be it for distribution if it's wet nets or this voucher coupons, monthly coupons for chlorine is a good idea. How do you wanna implement that in practice? The most obvious way is to rely on the existing public health system and ask nurses in clinical care clinics to, you know, give the free benefits to pregnant women or ask nurses in child care clinics to give these coupons for chlorine to mothers who bring their child for a child wellness visit or something like that. Okay, so this is exactly the way it was actually done in these studies. Health workers were asked to do this but because it was in the randomized control setting where we didn't want the behavior of the provider to affect demand side study, we kept the health workers on a tight leash, okay? So there was no room for them to mess around but in the real world, would they mess around? And so here, the issue is one of local capture versus local information. You would want to leave some discretion to the health workers to be able to identify whether somebody is indeed eligible or should be eligible for the subsidy but if you give them discretion, there is a risk of local capture, okay? And there is reasons to be concerned because in some other domains, there's been evidence of important rates of leakage, not all the money that's supposed to go to primary schools makes it to primary schools, not all the rice that's supposed to go to households gets to households. There's also evidence, I don't know what evidence I should say of extortion. So that's people, health workers asking eligible patients to make side payments in order to get things that they are owed. And a third concern, which has been very much studies, a study that is like shirking. So the mere fact that maybe the health workers are not there, and so you may have free bed nets in the storage room of the clinic, but if the health workers are not there to open the storage room, women are not gonna get these things, okay? So I've looked at this in a study with Rebecca Lieson-Ross and John Robinson where we audited continuing on my bed net obsessions. We looked at the effectiveness of setting up a free distribution of bed net scheme through existing systems without this time monitoring them officially, and so we audited such a program in three countries. In Ghana, with 72 health centers, there was no government program there. So we set up one through an NGO and then we audited what was going on there. So the NGO is the one that told health workers of public clinics to do this and deliver the bed nets. So Ghana is ranked 64th on the transparency international ranking in terms of corruption. Just for information, being number one is a good thing here. So I think Denmark is number one. I don't remember who is number 178, but Ghana is 64. And then Kenya and Uganda, which are ranked much worse in terms of corruption according to Transparency International. There the government was doing free distribution of bed nets at the time we started this study. So we audited both this government run programs and this NGO program in Ghana. And we look at leakage of nets to illegibles. So for that, we sent spies essentially. What we call mystery clients. We sent ineligible folks to the health centers and they were trying to get bed nets. So these are guys which by definition are not pregnant women. So they're not eligible for the study and they tried to get one. And we asked them to recall whether they were asked to pay something if so how much, whether they were able to get one, okay. And then we also look at coverage and extortion among illegibles. So we do this, you know, random survey of former clients of the antenatal care clinics and just check whether when they went for parental care, they did get offered a free bed net as they should have. Okay, and what we find, we find very modest leakage between illegibles. So, you know, the likelihood that a mystery client was asked for a bribe was less than 5%. The likelihood that they got a bed net was actually just around 2% in Ghana. I don't have it on the slide actually. 2% in Ghana and 9% in Kenya and 11% in Uganda. But in Kenya and Uganda, it was all free. So they got them and they were for free. They didn't have to pay. And if we asked around person in the community whether they think that a guy would be able to get a bed net at the healthcare center, few of them think it would be the case. Okay, and so one question is whether the few mystery clients were able to get a bed net despite not being eligible, got one because the healthcare workers were not doing a good job or whether it's because they were actually paying attention to what these persons were saying. And so is it leakage or is it efficient targeting? And we find that those mystery clients that were able to get a bed net from these health workers, usually for free, were those that said, I have a child at home who is sick and I need one. Either said, I have a pregnant woman at home. They said, this health worker said, no, no, no, bring your wife here. She needs to get prenatal care. So I'm not giving you the bed net. Bring her. So that didn't work. But he said, I have a child who is really sick and I need a bed net. Then they were able to get it. So it's almost as if the health workers were kind of like saying, well, the rule is a bit too strict because some households need one for their kids. They didn't get one when their wife was pregnant. Now they need one for the kids. I'm going to give it to them. Now, obviously it's cheap talks because we paid these guys to go around and claim they had sick children. So the health worker should not be that gullible. But you can see how maybe their intentions in leaking the products are not that bad. And then when we look at actually eligible recipients, we find relatively high rates of coverage. No extortion whatsoever. If you are pregnant women and you go for prenatal care with a great likelihood, you are offered a bed net for free. You don't have to pay. We do have, not everybody gets one. There is a stock house. It's quite important, especially in Uganda. So about 40% of women don't get one. There seems to be a little bit of screening there as well because we said the likelihood that you don't get one is actually increasing with your years of education. In other words, the likelihood that you received a free one decreases with your years of education. So when the health workers seem to be about to run out, they start being a little bit more stringent in who they give it to. And if you look a bit too rich, maybe they don't give it to you. So again, they seem to be using their discretion in a way that is possibly increasing the health impacts like minimizing the detrimental health impacts of running out, I should say. So this overall suggests very little leakage and no extortion and high coverage. And so this is suggesting in our view, and especially the fact that we did this in three countries, Kenya and Uganda have quite a bad reputation when it comes to corruption. And so maybe this is because for a very simple, easy to verify targeting role. And when there are obvious benefits to the beneficiaries, it's actually quite costly for health workers to not do their job properly. And they seem to be quite highly intrinsically motivated to deliver these benefits to pregnant women. When we surveyed this health worker, they seem to be really caring about the community, much more, for example, than teachers in SM community seem to be caring about education. Okay, so there may be some positive selection of health workers in these three countries that would explain their high performance. So to conclude, and then I'll take questions. Public subsidies are a substantial part of what developing governments do. I've talked about health subsidies, but there are similar questions when it comes to subsidies for agricultural inputs, food distribution, these make sometimes 10% of the national budget, sometimes even more. The rationale for these subsidies is that they could have large effects on health and nutrition and things like that. For these impacts to be there, a number of things need to hold, subsidies must be targeted or assigned to those for whom the returns are highest, leakage has to be limited, and the beneficiaries of subsidized inputs must put them to appropriate use. And there seems to be cases where the targeting question is obvious. We know that the returns to a given thing is higher for, let's say, pregnant women or young children. So many of these subsidies are targeted to specific gender or age groups. We found that leakage seems to be a second-order issue for really essential health products such as bed nets. And finally, on the last point, we find that beneficiaries do put things that they get for free to good use. And so price is quite too blunt of a tool for targeting. When people are credit constrained and very poor, you get too many exclusion errors if you use private screening tool. Although for some context, you get too many inclusion errors if you don't charge anything. And so that's where an ordeal can be possibly a good idea, but with all the caveats I mentioned, which is it's hard to get it right. So finally, I think all of this is quite now consistent with the fact that very recently, people who have looked at the effect of massive increases in subsidies for malaria prevention through the Rollback Malaria Initiative have found substantial impacts on child mortality going down because it really seems to be an area where large public subsidies is a no-brainer. So that's all I had. Thank you very much.