 Yeah, I'm talking about work that's that's barely out of the field Which is very new for me because normally by the time I write a paper or present anything It's been a couple years of data being out of the field. We're so backlog, but this work was finished that the the the impact evaluation I'm going to talk about was Literally done and completed by September almost by mid-October So some of the results are really new and I'd love input and They may also, you know, some sort of they we haven't done some parts of it yet But I think we have enough here. So maybe I can Okay, so the title is also somewhat tentative It's something I came up with just I think day before yesterday As I was writing the presentation, which I wrote So it should be fun. Um, okay, so the idea let me let me this work Is actually joined with a PhD student At the University of Michigan and Arbor. She was working with us On very early parts of the of them of the IE that underlies this one Before she became a PhD student and this is also feeding into her dissertation. So I'm very happy about that Okay, so basically There's the this this paper really looks at the role of social networks in leveraging behavior change Right, and there's a quite a bit of literature on the importance of peer effects on Influencing a rather diverse set of individual choices that's work on investment on saving on technology adoption And and and there's also quite a bit of work that such networks can facilitate Particularly the take-up of health technologies, right? You think of the of the paper that the vermin paper bed nets vegetable cups, right however in all of this literature the role attributed to peer networks is one of information diffusion, right and learning but it's not but I think that's very little on on on actually using the peer group itself to to motivate change, right? Much of that literature is actually coming from lab experiments on Contributions to public good schemes, which has shown us that social networks can actually be powerful catalysts for for inducing socially desirable behavior, right and that may be because they trigger image motivation Leadership of the desire to seek approval and so forth Now The the the use of these sorts of mechanisms though has not been tested empirically Particularly in the case of health Despite the fact that there's been incredibly low take up in developing countries of preventive care products, right? So what this paper does and this is really part of a larger project So what I'm going to report to you and as I go I'll explain the chapter will kind of pull from some of this But you'll see that as we go But the overall project I should say assesses the impact of using peer groups in this case case village based community Organizations, and I'll describe how these were formed to introduce an integrity uptake of a particular preventive health care product And we do this through a series of what could be called framed field experiments. They're done in the field with large groups of people All right, so the quick roadmap of the talk is I'm going to tell you a little bit about the motivation for doing this And why the particular product that we chose was selected And then describe the IE's on which this builds. Yeah And then I'm going to tell you about since it's a product. It's a health care product We need to worry about the willingness to pay for the product So I'll talk a little bit about how we elicit that and and as well as illicit the social norm Around willingness to pay because this ends up being important All right, that give you some early results all right, so the product we focus on is actually point of use water treatment and Why water well? So during an IE we were already running at the midline of that IE We ran water tests in in the population that we were we were studying and we found that basically Over two-thirds of the water was E. Coli contaminated About a third of it was actually contaminated at source Right, which means this was coming out of hand pumps. This is not pipe water supply This is coming directly out of hand pumps and mechanized pumps, right? Which ostensibly are going into the aquifer and yet this water was Contaminated a third of it is contaminated at source It about sixty eight percent gets contaminated by the time it is put into storage devices Yeah And as well fifty five percent of water collected on community water schemes was also contaminated And the rates vary in some cases as much as fifty percent of the water is contaminated at source These are areas which are mainly places where hand pumps are used they go a little bit less deep into the soil So the this project is also partly I'm involved in a larger project that looks at the overall impact of water water and sanitation infrastructure on on health outcomes and and so this this also feeds into that But given these astonishingly high rates of contamination You know and we know that basically these are These contribute to diarrheal disease And and through diarrhea to child stunting But also in addition to diarrhea, there's a lot of worry now about environmental and thought and terrapathy Basically also contributing to study to stunting. So what's going on is you have E. Coli contaminated water particularly at very young ages and That actually permanently destroys the intestinal villi so that you cannot absorb nutrients, right? So you may recover from the diarrhea episode and all of that But it's still going to affect growth because you're you're you're and this is permanent. It's not reversible. Yeah, so Okay, most dangerous for the young ones, okay Pakistan actually has extraordinarily high stunting rates 43 percent in 2011 and there has been no improvement Astagh in the last 14 or 15 years is among the countries with the highest rates Diarrhea rates of 23 percent in 2012 again no improvement in diarrhea This is important because this is happening in a background where in fact access has become near universal Right, so everybody has access to water at their doorstep Most people have privately and at least in rural areas where there's very little public provision But almost everyone has a hand pump or a mechanized pump in their home So it's not like they're walking far away to get water or getting water from uncertain sources There's been a massive expansion in in household toilet access stunting I'm sorry open defecation rates have fallen dramatically and yet there's been absolutely no improvement in any of the things We routinely expect to happen with expansion of water and sanitation services Simultaneously poverty has also declined dramatically. So all the usual causes that we would look for don't seem to to play out I Have a theory about that which which is part of the larger work But I won't go into that here yet because we have a lot to get through so So why point of view so as I said nearly 99% of households do not treat their water in any way However, most have water within their homes. So we cannot do community level Treatment of water. It's not possible if you have most people have their individual hand pumps and motorized pumps from which They draw water 24 hours a day as they need it and We use basically the product that we're going to focus on is chlorine For many reasons it's locally sourced. It's extremely cheap It was basically we calculated rupees 240 per month for a household of 10 to provide all of their drinking water needs Plus drinking water cooking and some other needs. Yeah And this is at market price right, so Actually, if you buy in bulk you can actually reduce that price quite a bit But there would have been no problem with with availability Currently though in our villages, there is no market for chlorine So it's not like there are competing sources of chlorine is important to keep in mind Part of the part of the effort in this project was to figure out if we could get Enough uptake and sort of encourage a market so that local local suppliers would start keeping chlorine in stores Okay, so the main questions we want to test is where the community organizations can be used to build demand for point-of-use water treatment And and and then basically to focus on which types of motivation triggers could be used in the context of community organizations to encourage Update and we test three that I'll describe image leadership and giving subsidies And then of course we're interested given the focus on gender in particular here is whether women respond differently in villages that got this inclusion treatment that I will describe in a minute and Whether leaders of community organization as natural leaders act in a more prosocial manner and I will describe who these leaders are Next okay, so this this work builds on an underline I need to explain this so you get some of the results this work builds on an underlying impact evaluation randomized impact evaluation that we started in 2010 and ended in 2016 at end line The under this intervention community groups were created And the objective was to deepen inclusion which is that the requirement was that at least 50% of households will be Organized into community organizations. So in a sense, we have a lot of community organizations per village And that's important. So the results are much more generalizable in that sense The social groups include women only men only and mixed gender. They were required to cover 50% I said of the village population But the important thing was that above these groups is a village level organization So they were aggregating up to a village level organization And this is where the leaders of each community organization would filter into the into the VO The VO had to decide on a village fund that was quite large $30,000 per village plus 20% of community contribution. That's pretty large Given the ordinary domain of community development projects And and this village organization was required to create a village development plan There are several papers we are working on from this project But the important thing for us is one the existence of these organizations as a base as we go in and the second is The fact that as part of the effort to improve governance in the CDP project Inclusion mandates were put on women and poor households. So unlike in other countries in Pakistan A community organizations were not women-centered at all They really targeted men for the most part And this was a concern we had we had voiced before we started the impact evaluation And the program actually was changed To require that at least 40% of all community organization members and all village organization members have to be women And 50% have to be from poor households Um, the and the second kind of intervention was that the VDP the village development plan had to be subject to secret ballot ratification I don't worry so much about this. I'm going to focus more mainly on the inclusion for for our purposes So this motivation ie that I'm going to describe was actually implemented as I said July to September 2016 After the end line on the underlying ie was done So we have finished and out and then then and then we did this ie So it kind of randomizes re-randomizes on top of the and I'll describe how So basically this has two components Um, we we gave in some villages. This is a village-wide Intervention was to give an what I'm calling an externality treatment Which is that an information message was an information campaign was really done in every village Delivered at co meetings. Um, this was of course a campaign designed for a low literacy context The campaign focused on village water contamination levels, which we had because we had collected water contamination data for each village Um on the key health behaviors, uh to keep water safe and particularly prevent recontamination The use of chlorine and then we did water tasting So everybody could be assured that the taste of water wasn't going to be bad If you mix the right amount of chlorine with the right amount of water Um, and then we gave them each a canister that was of the right size The key difference between the externalities and no externalities treatment was that in the externalities treatment Uh meeting members were also told that their own actions would affect the health of others in their community and in their village Yeah Underneath each of these we run we try three different Motivational approaches to try to elicit uptake of of this product, right And and these are done within village co level So so we have lots of variation and I would describe each of these three to you And then go to the results So so the image our motivation basically Uh is recognizes the fact that individuals are affected by the perceptions of others. They try to actually Particularly in public may want to mimic What others do particularly people who are within their social reference group, right? And this may be tied to how visible the action is right. So, um Uh in the first so so the way we designed this to kind of elicit this was What we're interested in is understanding whether If we can make actions public, right? Do more people are more people willing to to purchase the product, right? So it's sort of more are people more interested in status-seeking behavior Um, right, um, and if we allow people to self-select into public bidding, right? Uh, does that increase in further increase willingness to pay right? So people who actually choose to be public to have their actions be public may be more inclined to then engage in more pro social behavior So the three arms are basically Private bidders are told that their bids will not be revealed to anyone. Yeah, you can bid in private And I'll explain the bidding mechanism Public bids are people who are told that whatever bid they place will be revealed to everyone at the end of the bidding process And then there is a third group in which people can self-select into bidding. Yeah I'm going to be describing a lot of stuff. So this might uh Um, I'm I'm sorry. You can't ask questions along the way, but in any case Um, I'm happy to take everything at the end. Um, so in the second game, which is the leadership game We have two arms We either in the first time we randomize people to leadership, right? So you meet and some people are assigned to be random leaders others to random followers, right? And and and um For randomized first movers their bids are by definition public because then you want to elicit Whether whether they make bids that that are higher and then followers Are free to choose to place their bids relative to what what the randomized leader has done In the other arm people can self-select into leadership, right? They can choose whether they want to be leaders or not And underneath of course, as I told you there are natural leaders, right? So there are vo members who are part of these meetings And they they can be seen as natural leaders since they're already leaders within their village And we want to know whether natural leaders behave differently than randomized leaders. Yeah Okay, so now the third one is the subsidy one Here basically we in the first time we randomize a group of people to sharing You can either choose to you're given a subsidy You can either share that subsidy or not, right? You're free not to share your may it's made very clear to you that you can share zero Or share some positive amount, but you are randomized into sharing And in the second arm people can self-select into sharing, right? So they can choose whether they want to be they want to share or not And again, you know subsidies can be a useful kind of a mechanism For encouraging particularly purchase of products that are unknown or new, right? So you want to subsidize purchase But then you want ideally to give those subsidies in an environment where that is where they encourage the most take up, right? Because they're costly All right, so we want to elicit willingness to pay clearly in this game because we want to understand how people Exactly to you know, whether there is that it whether there is uptake of the product So we do this in two ways in the in the image and leadership Experiments we are basically going to use what the bdm auction which is a very which is you know simple It's basically People state their reservation price. So i'm giving you a domain of the price first The market price for this product was about 240 For a family fairly large family The price we offer During this entire process is between 50 and i think it's 50 to 150 It's 50 60 60 to 150 Yeah, and that's because we got the product at a subsidy and and this product could be available to everyone at a subsidy actually given reasonably bulk Purchase so we think that's a good domain of price to keep um So basically what people are asked to do is to state their reservation price Any any number between 10 and 150 in increments of 10 rupees, right? Then a price is drawn by a lottery If the price is that's drawn is less than them their willingness to pay Then they have to purchase the product if it's greater then they cannot purchase the product Because of the way the bdm lottery auction is set up Your your best your best sort of option is to in fact state your maximum willingness to pay Because if you understate it you may not be able to get the product if you in fact want it um In the subsidy case because we already have to explain the subsidy and this is a very low literacy environment So we we basically go with a take it or leave it a t only offer which is you know A lot of lottery price is drawn. You either can choose to purchase or not, right? And the subsidy in this case in the subsidy game was fixed at 40 rupees Okay The one thing we do want to understand is how social norms can influence The decision to to to purchase or not So we ask everyone at the beginning we explain before we assign people to their games and this is important We ask people In private they take a small survey And then we ask them whether What do they think Is the willingness to pay Of others in their community, right within this price range From 60 to 150, right? And and and so Using this this sort of this social norm on on on what remember this is an unknown product And you're quite worried that for unknown products This this may be low anyway, and it may at that and this could rise over time Right if you're introducing a new product So we're really getting the first sort of baseline on this What do people think people will pay for this product now when a new product is being introduced and you've just been told That your water's contaminated So using this we can create a variable which I'll use in a minute Which is the deviation from this norm, right? Which is how far is the individual's bid from the what they said was the norm, right? What they thought was the norm for the community Right And so we get a sense of the conformity the degree of conformity to the social norm that individuals exercise Uh in their bidding, right and we can very we can look at this by whether they are there in the public or the private and so far All right So we know essentially That in the image case of individuals are mainly status seeking their bids will be higher in public If they're very worried about Norms and conformity then the dev norm should actually increase decrease in public. Yeah All right, so So let's look at i'm going to give you the first set of results And i'm going to try and tie them all together. Uh, it's still very early days, but I think some of these results are quite interesting So in the random in the public bidding game if you look at the average impact Of public bidding you see nothing. There's actually no difference Between public and private in the willingness to pay at all, right the the variable of interest is the is the max willingness to pay Yeah, um, so we don't know what's going on if you if you know if individuals were actually engaging in status seeking behavior We should see that that a public bidding will allows you to sort of bid higher nothing, right So then we go and create this dev norm variable and we look at what happens to this dev norm function Under public and private bidding and you see that in public bidding it moves to the left So much closer to zero. It's much more centered on zero So what's going on essentially is that in the public, uh, uh, when when bidding is public People are actually reducing their bids to bring them more in line with what they thought was the social norm Right, so it's more conforming behavior Then status seeking behavior, right and all of this thing is happening if you look at the distribution Right around the range where people's Own own bids were a little bit higher than the norm, right? So they're adjusting down actually So now when you separate that out and you look at it You see in the first row that all of these negative bids are actually just in that domain in the third fourth and fifth quintile. Yeah um So basically as I said, they're narrowing their the difference between their own bids and their beliefs regarding the average bid Right, so what we're interested in now in exploring is whether These uh, well, this is just reconfirming this if you look at this You look at the the probability of the maximum willingness to pay Declines if your bid is higher than the normal, right? And the odds of it being right at the norm increase in public bidding, right? So that's consistent with with with what what we're finding Okay, so now we're more interested in looking at exactly what happens among men and women And whether differences are whether there's any difference by inclusion the inclusion treatment that's underlying that And also by the externality treatment We would expect that if people are told that their actions affect the health of others They would be much more willing in general to raise their bids. Yeah And and basically what we find Is that the negative impact of conformity is largely being driven by villages where this externality priming was not done Right, so where the externality priming is done, which is the third, right? You actually see a small increase in bids and those differences are significant generally at the five percent six percent By between five and ten percent level, right? So every time of course, we want to look at these different things. We lose some power, but I think that's pretty clear Um, so so externality seems to encourage more pro-social behavior, which which is is reasonable. It should Yeah Then we look at women and here it's interesting So the trend towards conformity first of all appears to be stronger among women, right? So interestingly women bid higher than men This is not the case in many kind of public goods game But in this event you're selling sort of looking at willingness to pay for a health preventive health product Women actually big bit significantly higher than men, but they're also more conforming, right? And unfortunately, they also have on average a poorer norm They think people would be less willing to pay pay or they would be willing to pay lower amounts, right? So so they generally have a lower norm and they are more conforming And and so some of this effect that we are seeing is coming from women pulling back, right? Um, you can look at that more clearly here So you look at it that if your max willingness to pay was greater than the bid norm it is women who are adjusting down not men And the odds of of your your max being exactly the bid norm that increases for women not for men Right? So that effect is coming mainly from women who initially bid more who are more likely to want the product But then tend to be much more conforming and have essentially lower norms so that two together Essentially decrease Bids and willingness to purchase Um Now I'm going to skip this actually because this I think these results are too early We want to say something about inclusion villages So these remember I told you underlying all of this treatment some villages were were Six years ago were given were were required to have form with women's organizations at least 40 percent Of the vo and co's had to be women So we want to see whether these six years of empowerment has made inclusion villages act differently Uh than than others, right? So we in fact do see that bids are actually significantly higher in inclusion villages, right? And the this tendency to reduce bids in public is actually only appearing among people who were very high So people whose bids were really much higher than the norm They are pulling back but right around the distribution. We see no action, right? So in general bids are higher and there's very little So the pulling back is little and to the extent that is there is in the in the in the higher domain, right of the distribution So you can look at this here more clearly I guess which is again the same thing we did before and you can see that basically you have The people are pulling back Uh when they have uh massively greater than the bid norm, right? But this is coming only from people who are in the 70th, 80th, 90th, they are quantiles, right? So so okay, so I'm gonna Basically try and summarize what I've just said on the image game. It's a lot of information right, so So essentially I think the big takeaways are people generally tend to conform To how they think others will behave when their actions are public Okay, so, you know the the lesson from here is it's all very well to try behavioral nudges Right, but you need to know a little bit about what the norms are and how people will behave Because if we had just done this, okay, and we had not done any of the other games We just tried this image thing The fact that this social norm for this new product is quite low and it's lower among women Right would have led to a situation where women were much less likely to purchase the product at the end of the day Right, so so so the the lesson is all right We want to use behavioral nudges But in order to use them in a sensible way We need to understand a little bit about what prevailing social norms are on the other hand the message is That social norms can be worked on and community organizations are really good places To try and make this kind of thing public and therefore try to improve the social norm or or or expectations around what others will do Right, um, so so I think that's our that's really a big takeaway from here in general We are very happy that both the externality treatment which we would expect And inclusion both encourage Um, uh, both raise the social norm and encourage willingness to pay So they dampen this effect of conformity in some sense, which is what we want to see. Yeah Okay, um So i'm going to go on to leader to the second, uh, second second frame field experiment, which is leadership Um, and here basically i'm going to give you some sense somebody says that that really I think pin things down a little bit And if you look here So this is the mean bit and you can see that there is very not much difference between leaders and followers in the mean bit Right, so this is you know, if you just make people leaders, right? You don't see much of a difference in the in the in the bits that they're willing to put forward um When people are natural leaders their bids are higher, right? You can see that that's one natural leaders mean bit is 190 relative to 110 But when natural leaders are allowed to self select into leadership Their bids are way higher, right? So there's something about the group of natural leaders who actually say yes, I'm going to self select into leadership I'm going to come to the actual results. I'm giving you a preview. Yeah But this is just something to keep in mind as we go to the results, right? So so again, if we just look at max willingness to pay across the board on average, there's nothing Okay, there's no difference between leaders and followers Let me just remind you of this game very quickly people are either randomized into becoming leaders Or followers or people can self select into leadership, right? Those are the two arms. So I'm showing you the result For for the randomized sample. So now people are randomized into leadership. We see no effect, right? Um, so being a first mover has no impact, right? Uh, if I go to the self selection sample on the other hand when people can self Select into leadership, right? Then the average person that self selects, right? You look up there demonstrates no difference But if you are a natural leader, you anyway have a significantly higher bid, which is encouraging So all these people we were organizing for six years, right? Uh, whether or not they are leaders in the game, whether or not they are assigned to leadership They have higher bids, but when they self select to be leaders So all of this 11 11.8 is this this effect if you look down the third row is all coming From natural leaders who actually self select into leadership So this is the subgroup of natural leaders Who who are really saying no, I want to play the role of the leader here Then they have way higher bids, right than everyone else. Okay, so so that that's very encouraging Um, all right, so then let's look at leadership and women, which is also quite interesting So so now we're going to divide up these people So between males and so these regressions are male and female and we do the same thing again And you can see that the main effect is coming From women, right? So it is women Who are natural leaders and you say yes, I'm willing to be Uh be a be a leader in this game Those bids are the highest it's not coming from men Right, so we look before that the average bids were higher for women, but then they were more conforming So now we are seeing in another game that if you know if you're actually going to be triggering leadership, right? Then then women who are natural leaders and say yes, I'm willing to be a leader That's where all the action is right. So they really have much higher bids Yeah, and and you can see that these actually these p values are quite significant. Yeah okay, so Now we look at this by inclusion, right So these are natural leaders who are randomized, uh to be leaders, right And and then they are and and we are basically going to look at inclusion and non inclusion and here you see That even when they are randomized into leadership In inclusion villages, even the randomized leaders bid higher Right, we saw no average effect remember, right? But when we look at it by inclusion villages Even randomized leaders are willing willing to bid higher. Yeah, and then here we go to basically we're looking now So this the prior one that I showed you as I said is for the randomized sample I'm now going to look at this for the self selection sample and this is really interesting. So you look at the The total size of the increase in the bid is much higher in the self selection villages, right? So the so in in inclusion villages Where people self-select their bids are just higher. So self-selection in the presence of inclusion They just have higher bids, right? But but basically the major impact Here is again coming from the natural leader who self-selects into leadership. Yeah Okay, so I'm going to going very fast across all of this, but I think Hopefully you can keep up with some of this So there's lots going on here, but here is the The results for externalities now again, we triggered externalities So the interesting thing is that this externalities messaging again increases the overall level of bids We would want that across the sample everybody bids higher when you give them the externality message, right? It also triggers a leadership effect among first movers Who are or leaders who are randomized into leadership? So I've just given you an externality message you were randomized into leadership Everybody's bids are higher, but your bid is still higher, right? So so leaders even when randomized react, which is interesting But the really interesting thing is that when externalities messaging is absent, right? Then natural leaders still bid more, right? So in other words natural leaders do not require as much the externalities messaging in order to bid more They're already more prosocial But you're triggering prosociality among people who are simply randomized into leadership So these kinds of messages on externalities your You know actions you take will affect the health of others those sorts of nudges are really important in environments where People would otherwise be discouraged, right? But in the subgroup of people who if you have a very organized village And there are lots of people who are willing to take the lead, right? Then they may be then even without the externalities message there may be far more prosocial um a kind of behaviors So so this this is particularly apparent when you move to a self-selection sample um So here Um, you see that the bids are the average bid is much higher, right? It's uh between 101 to 114 So externalities as I said raises the average bid Here but here even though the natural self-selected leader is willing to bid more almost 22 Rupees more that is not higher than what is happening in the no externalities treatment for the natural leader So natural leaders actually step in Even in the absence of externalities, right again in the self-select sample Okay, so I guess I'm going to um, you know what I'm I don't know how am I doing my time? Yeah, so good. So so what I'm going to do is I'm going to describe this to you I'm not going to describe that it's also the subsidy game Which is super convenient for me because those are at the earlier stage, right? So we will be working on those and working them out later Um, I really want to just say one thing about the subsidy only which I thought was really interesting Which kind of confirms what happens In a lot of lab experiments, which is simply this if you give the people If you give people the ability to select out of a sharing environment They generally choose to share they they choose to move more people choose to move to the private bid And they prove not to share So there's something about pushing people to be in a sharing environment Even when they know that they can share zero, right? So you've told people you're a group that can share feel free to share zero It's all private, right? You're a group that can outright self-select into sharing The people who are placed in a sharing environment, right? tend to be to reduce that to sort of like share more, right? Because there's this sense of, um, right? So so people who opt out so if you look at this in the self-selection subsidy, right? People basically have lower bids Then in the people who are who are put in a in a sharing environment, which is interesting And this is a kind of replicates results from lab experiments, right? So it's just something and I won't I won't talk more about about the subsidy treatment Because I think I'm am I out of time time or do I have some time? Okay, I can tell you just a little bit more and and I can stop so again These results of course this decline in in the probability of sharing you can see up there is very substantial What is interesting here is that when you look at externalities, right? And this is something we are we are We find very interesting So among women Being given this externality exposure to the fact that you know it could affect the health of others It reduces the propensity to share when women are randomized into sharing Right, so when you're put in a sharing environment, you are less likely to share and we were quite concerned about this It's sort of an interesting result We know women have lower norms about what others will do Where the others would be willing to buy and so we're wondering whether and we're really actually looking at at this result More closely is one of the things could be that that women when exposed to to this messaging think Oh, it's really urgent that I do something And they want to buy the product So they're basically retaining the attempt is to retain the subsidy for themselves Right and not to share and particularly you would be encouraged to do that If you already had really poor priors about the fact that other people will take it up Right, so your subsidy goes to actually it's it's it's a dead loss, right? If you're willing to share and the other person doesn't purchase that subsidy is lost to you So if you have very low priors about others, you'd want to retain the subsidy and make purchase decisions But I think we can we can work this out by looking at actual purchase decisions In the sample we have not done that but I thought that was quite interesting I'm actually going to stop here Okay, thank you very much