foreign okay so just to remind you guys where we were so last time we talked about kind of the basic sort of setup of models of thinking about credits we had average selection moral hazard we had monitoring models and so on and so forth um and today what I want to talk about is micro credit micro finance and in particular just you know I'll give you a couple of words of background some of the history of microfinance um we'll talk a little bit about the impact of microfinance both overall it's he
terogeneous impacts and then um I'm going to talk about the in the role of particular microfinance institutions and what I wanna I guess I want to emphasize on this piece is um that you know as you'll see microfinance kind of a bundle of many different things and there's sort of been a nice series of research papers over a series of years that tried to sort of help us unpack kind of what's in that bundle um and I think that that you'll see some some nice elements of that okay so just a bit of ba
ckground so microfinance um uh sort of a modern microfinance uh model I guess was was uh you know came through sort of the green bank which Muhammad Eunice who uh won the Nobel Prize a few years ago um uh helped create and that sort of um you know made small loans to women in Bangladesh to Portland and Bangladesh um I think actually these numbers are actually probably small I think it's way more than 25 billion today probably these products are outdated um you know hundreds of millions of climb
clients uh typically High micro higher payment rates um some other fan institutions are are profitable some are not um and these microfinance institutions are you know they're lending to a large number of sort of small borrowers and providing them small loans you know 50 100 500 is kind of in that you know and that maybe up to a thousand dollars maybe that would be a really big loan you know small loans um and there's an idea these can be really powerful for sort of helping break some of the cre
dit constraints that we talked about last last time right so if you've got some of the stuff at the end of this some of the models last time we predict that sort of people who are really poor would be not be able to borrow anything or very little um and if there's a way to sort of break those credit constraints and figure out a way to sort of solve that that could potentially be very powerful um so I would say um suppose we wanted to start by sort of understanding and I say there were lots of gr
and claims associated with my credit right so if you think of sort of what you know when when Eunice wins the Nobel Prize the idea is like this micro credit thing has been totally transformative to lots of people and sort of like you know uh you know change people's lives in lots of ways so if you were trying to think about um uh trying to understand and evaluate some of these claims like what are some questions you'd want to answer uh and and uh you know how would you think about um evaluating
what are some questions you'd want to know but this is now a question I'm posing to you yeah okay right so you want to know is the change in sort of net access to credit not just gross right so like just the fact that people are borrowing microfinance institutions doesn't mean they're getting more credit what else do you want to know like okay so right what kind of outcomes what's that business profits consumption right right so it seems like one thing you'd want to know is you want to know like
is it about change is it actually sort of you know if you believe that people are credit constrained right that they aren't sort of you know if we believe that we're in a world where basically people can if we're in the world people are unconstrained right that sort of like that you know in the unconstrained world right um where where you have F Prime of k equals R you'd expect the marginal return to Capital is is equal to is equal to is not super high right it's equal to interest rate but if y
ou're in a world where sort of K is constrained then you might think that F Prime is much greater than R because people aren't sort of able to sort of get all the capital they need so one thing you'd want to say is what is the you know what is the impact of of uh of um of additional capital on sort of people's profits essentially and like if you believe we're credit constrained you might expect something to be really high right um so that's one thing you'd want to look at you might also and see
you may also then want to look at sort of Downstream uh kind of consumption outcomes yeah you might also be concerned about for whether people are taking on loans that are actually in their interest or whether these predatory and so forth yeah so what do you mean by predatory uh if people don't actually have good information about their odds of being able to pay to pay for that you yeah so so I think that's also related to some of the consumption questions but I think you're handing another poin
t which is that we may not just be interested in averages we may be interested in the distribution of effects right so it may be and and we're just in distributions I think in two senses the first sense is we're interested in just is there sort of what is the distribution of outcomes right so you know imagine that basically we had like you know um uh um uh uh like outcomes imagine we had quantile treatment effects right there's a big difference in a world where basically a bunch of people with s
ort of nothing and a bunch of people with positive income facts that's like a world where actually we're pretty happy with this thing and it's kind of like you know nothing for everyone and good for some people or if we had a world this is like these are quantum treatment effects or if we had a world where it kind of looks like this right where some people are a lot worse often some people are a lot better off you know even if this kind of went really high so the mean was positive like you'd fee
l very differently about kind of this world versus this world right because here like you know nobody's gonna worse off here people some people are actually substantially worse off and that kind of goes with the idea maybe some people can't sort of uh pay back and and if by the way you think of capitalism like actually the capitalist funding Investments which may be risky it's not crazy to think that you're in a world like this like if these are positive expected return but risky uh Investments
you know on average who we better off but some will be worse off and so understanding whether the world looks like this like this is different and then in addition I think regularly that's another point which is it's not just that we're interested in sort of the overall heterogeneity in the treatment effects we're asking whether they're predictable or not and so this is just sort of saying you know what are sort of the quantile treatment effects like just in general there's another question we s
ay do people actually is that predictable or not and are people kind of able to make good is it sort of there's risk in the world like if you think of the model I talked about last time this is a model of like positive effective return activities but risky risky there's another world though where you start thinking about like some selection world where people are different in their characteristics some people are have like really good things and people are really bad things can they predict them
can the banks predict them sort of what can we sort of explain that heterogeneity in reasonable ways Africa uh were you really interested in takeoff rates yeah same one uh so maybe a whole micro credit like institution might set more restrictions on what you can use the micro credit for and perhaps that lowers the recovery relative to remote controllers who might not put such restrictions to making deadlines or online yeah sorry so yes take up absolutely so like you know all these restrictions
sort of reducing take up and is that kind of good or bad you want to think about that um should I think similarly of like the micro Credit in developing countries payday loans say United States um yes and no they have um I think they have different characteristics so they're both sort of targeting low-income kind of individuals the payday loans are collateralized basically against your future paychecks I think where it's easier sort of uncoateralized typically right and we'll talk about sort of
how they sort of insure a payment so they're different in that sense they're similar they both have reasonably High interest rates I think um one question is sort of whether the payday loans we mostly use for consumption smoothing whereas these are mostly used for investment that could be a really important distinction so if they're if these are sort of screening clients unlike do you have a small business that you're going to kind of invest in you might think those would have very different imp
lications than if it's just sort of like available people who have sort of really urgent consumption needs is actually extremely profitable for some companies I want some rsmrs yeah I wonder why like what what was the economic friction that prevented microgreens like 20 years ago oh so it one is yeah so this is okay so you're asking I would call you this is the difference between a question and an economist In this River class and asking someone in a business school class over at Sloan would ask
you know you asked the question I always ask which is like if something if this was like new and profitable why didn't existed in the past like it can't be must not have been equilibrium that's the economics view the business school view is like somebody had a good idea and made some money um and so um you know I actually I said actually not not actually totally facetiously I think that sort of like sometimes people just innovate and then um and come up with new ideas and make new businesses an
d sometimes those things like sometimes there were fundamental frictions like you might have said oh it's technology or whatever but actually I I wouldn't be shocked if in this case it was just like Innovation like someone that came up with some ideas and sort of said like oh we think we can you know basically you know Big Bang you'll see sort of you know one of the things is that there's lots of monitoring costs there's a question of sort of like big Banks may have said it's not profitable then
to make these little loans because it's too expensive and you know I think that some of these organizations figure out kind of a way to a way to do it so I think it may have just been sort of organizational Innovation I do think one thing that's changing it one thing I'm not gonna actually talk a lot now about now but I think is kind of an interesting kind of New Wave is sort of digital technology and digital lending and I think they're actually sort of like fintech and sort of how do you think
about using kind of new technologies to sort of solve some of these problems that may be something that's actually new um that would wasn't available in the past and that may sort of open up new opportunities I'm not actually going to talk about fintech today but like that isn't that's so this stuff is organizational I'm not sure any of it like you know why was it done in 1976 and 1956 I don't necessarily know there's a a hard and fast reason for that like a fundamental reason the fintech stuff
where we're sort of using kind of your you know cell phone records as a way of predicting kind of your income and repayment status that actually does rely on kind of technological innovation which is new um okay so one of so why is evaluating this challenging why is evaluating micro credit challenging like it took a long time before we had good impact evaluations in my credit Wesley well like logistically difficult or like in the real world like for example micro credit institutions if they sor
t of believe that this is a silver bullet may not want you like they have no incentive to have to evaluate them um right so okay that's point of Mind by the way so so right vis-a-vis as long as they're profitable or at least breaking even they're good right they may think they're good and so why would they want someone to evaluate them so it's kind of risky for them I actually think that's a that's actually a fun kind of an important point and actually I think sort of you know convincing people
to be evaluated is that to evaluate themselves is actually really difficult so that's part of it especially if we're interested in not just prop like they may they want to know what the impacts on profits are they don't need to rigorous impact evaluation to figure that out right you know and and you know they may not necessarily want to know what the world looks like this for example that might be bad news for them so I agree that's part of it but then what else about sort of what about designin
g the impact of operations might be tricky so you have lots of different outcomes long timelines so what one of you in your comments about that you sent in for today about the sound paper talked about like GE issues what what I don't remember who said that I can look uh what do you remember do any remember for mentioning General equilibrium issues and sort of evaluating these things you're saying yes or maybe it was you but now okay yeah what you go for it uh so if um you're not satisfying in th
at if everybody in your village like if some people in your village have access to money for credit and some people of the neighboring Village don't have access but as a function of like a neighboring Village having access like they're being able to fund different like investment projects can like boost up the availability of invest like profitable Investments available to you and also yeah so say more so you're saying something about like so I don't fully follow but so say more you're trying to
evaluate this by for instance uh allowing some villages to have access to micro credit and other Villages not um as long as there's some form of interaction between those economies uh like the uh the you have Network effects just by Hershey love yeah so the main issue the main issue actually it's even easy yeah the main issue is like you're you're doing lending right but people are interacting with each other kind of in the underlying markets right so actually I think the village one is actuall
y better imagine if what would be the problem if you do this like imagine you randomize people at the individual level and access to credit what would be the problem with that yeah we have like a certain aggregate demand within the village and people would immediately start like or I guess either on the supply side we're sight or um like people yeah so you're thinking so right so your view is you would understand it because you so your viewers you would understand the impacts because there would
be demand effects and that would sort of spill over the control group anyone think you would overstate impacts yeah because there's more information what disclosures limited demand yeah for something and all the loans are for local I don't know retail shops that you they'll grow a lot because you're capable of everyone they're competing for the same movement yeah exactly so I think you also have to think about what these people are doing with their money right so I think that you're right that
there could be sort of these demand kind of Keynesian demand effects but Sean's also right that there could be like potential spillovers right in terms of you know in the product market so imagine like we're all making like you know we're all selling the same make you know the same uh you know imagine we're sort of uh what an example um we're all fried rice vendors leave for my Indonesian examples right we're all fried it's a very common act it's super common actually example is that people have
like you know they have little fried rice carts and like there's only so much demand for fried rice going around so if like I have a better fried rice cart and I can sort of move around maybe I get a motorbike fried rice cart so I can like zip around faster or something I might just be like stealing business from other people right if the demand for fried rice is reasonably inelastic in my Village like it'll look like I'm doing great but actually I'm doing great at the expense of other people w
ho kind of didn't get the loan so um so I think that the the you also have to sort of think about kind of these issues and actually I think I think actually that suggests that sort of larger level randomizations like kind of a village level randomization or even like sub District level randomization is going to work better we're better than sort of an individual level thing where you have to worry about kind of both these positive and sort of negative spillover effects so that's one of you to so
rt of think about some of these issues and there's a there's another thing also which is if we think the heterogeneity is important we need to sort of think about measuring the heterogeneity and sort of make sure we can measure the relevant dimension of the heterogeneity and the paper you guys read I guess it was for last time but we're going to talk that we're going to see today was really about sort of saying is there kind of real heterogeneity in returns and as you saw from that before there'
s like a lot of heterogeneity returns and sort of thinking about can you measure that properly is really important okay um so they said you know for a while kind of as I think uh as someone pointed out uh I remember with you Wesley uh basically like look the reason was as follows you know we're profitable we're doing good you know why does anyone need to why do we need to evaluate our impact and that kind of reasoning was very powerful for a while um so um you know and there are problems with th
is of course like number one is you know the you know maybe they're not they may not be doing quite as well as they think they are but number two is that you know as I think that Rick as you pointed out just because it the fact that people are borrowing from micro Finance organizations that means actually good for them right in particular we you know we have this this negative people that could lead to debt trap and so on so actually measuring kind of these outcomes is really important um and by
the way this is coming to like a a huge um this has really been a this question of sort of is this kind of good or bad has led to sort of major sort of policy debates for example it was a confrontation in under product in India between the state of AP and kind of the microfinance institutions where basically at some point the the state the state government said you know this is like a terrible idea and basically no answer pay all their microfinance uh loans and that whole sector kind of went Be
lly Up kind of in that state kind of overnight and actually I think um uh I believe Emily Brazen Cynthia can have a paper sort of studying kind of that incident kind of what what happened uh okay so um there has been however um a bunch of rcts um that that then after a while sort of tried to provide evidence on this and there was actually a special issue in a in the journal AJ applied um which put together sort of six of these uh studies uh from um India Ethiopia Mexico Mongolia Bosnia and Moroc
co kind of all in one place and try to sort of harmonize them so people could sort of see what they what they um what they what they meant um and they and they basically had uh two different um designs so one was this place-based randomization and that kind of the advantage of that one is to try to deal with some of these spillovers the idea is we're going to sort of say like look um you know the microfinance organization says we're going to think about expanding to a bunch of places we're going
to um we're gonna pick more places than we're actually gonna expand to and we'll randomize them maybe we'll randomize the order of rollout for example that's actually a pretty common one to say look you pick the places you're gonna go uh you can't open them all at once and we'll just randomize the order of rule out that's actually a very nice design that sort of um gives them some some flexibility of so you know control over kind of where they're going works with their business plans and sort o
f operates on this Dimension they're indifferent between which is sort of the order um another one I sort of this idea is sort of like um uh randomization kind of on the bubble so the idea here is that basically the mfi scores applicants in terms of how excited they are to lend to them you know and then rather than you could just have an RD like on some cutoff one other thing you could do is say look the ones that are really awesome you know you lend to the ones you're not really off that are re
ally terrible you don't lend do the ones in the kind of in the middle then maybe we'll randomize and they also can be because they're roughly indifferent they can sort of be open to that too I think both of these are sort of useful by the way as you're thinking about experimental design because both of these are are um you know one kind of problem you have when you're often designing experiments is you know you would like to sort of randomize over an entire population the person you're dealing w
ith is like no way am I going to do that because like that would involve making a bunch of like really stupid loans or you know real foregoing a bunch of really great loans you know both of these have this um both of these designs have the um the the feature that you're sort of you're exploiting the fact that sort of they're you're figuring out what they're kind of roughly indifferent on and sort of doing that and um you should be aware of that in general if you're ever sort of in the world tryi
ng to negotiate kind of what a possible randomization is is you know just because you wanted to run a randomized experiment does not mean your partner has to give up all control whatsoever you know they can say yes you want to do it here no we don't want to do it over there where are you kind of the key sort of where are you roughly indifferent in terms of either place or time or something and maybe we can sort of randomize that and that's certainly an approach that I've I've found some success
with as well in very different contexts um another thing that's really important is uh is take-up rates are not really enormous so that's um uh not surprising when you think about it like if I came to your you know Town wherever you guys uh you know or wherever you guys are from and said hey I'm opening a new bank I'm opening loans to start a business like what fraction of people in your town do you think would be like Oh I'm really excited I'm gonna take up this loan right some people would but
not but it's not going to be like everybody right it's gonna you know and in fact the take-up rates look more like 10 which actually I think pretty big when you think about it um right but but small is a fraction of kind of the overall population so why is that sort of an issue for for for these designs we just can be very hard to have power to do that so yes that's true we don't have a power packaging but how is this for the overall impact why is this a problem for just estimating the overall
impact uh Patrick my self-select if they believe we would have a bigger any time so you're going to get self-selection that's true but that's okay you're gonna get the Rel you're going to get sort of the impact on the people who are interested in being treated it's kind of it's irrelevant late something else so realize what so say more about it I have really low take up even who are like the treatment is like being induced or given the option to use it then you need to do some sort of like i.eas
ure that effect yeah and that's going to be like you're gonna lose a lot of power yes exactly right so that's exactly yeah okay so it's a great question so why is this a problem why is power a concern okay now I have to do the like which class did I explain this to I did not explain this in this class yet right that was in the other class right okay good all right I did it in this class right okay good then you should know the answer so why is it a problem I can never remember which classic teac
hing students the pickup is only 10 because it seems like your products are also not like are only being applied to ten percent so you can just expand the program give your costs a lower right so you're right the so so that's the question why would I wrap so so okay so if I already went through this we can ask the question again why would I rather have uh 100 take up on like a sample of size 100 or a 10 sample take up on a sample size a thousand like which is better I mean yeah I do get the same
the smaller set for the hiring table will be better but you would partially address that with like a bigger sample right but no that's the whole point like so that you have to have a much bigger problem so that's the whole point of the the the calculation I went through before is that in order to sort of get like uh you know a take-up of a hundred percent on size uh a hundred percent on uh size 100 is equal to I think 10 on size 10 000. right not on size a thousand that's the point because the
square root of n issues you have to go like in order to sort of deal with low take up you have different really really big sample size because of this sort of this this point of a sort of the square root of n that was the which I did explain in this class yes okay right so that's the point so the the fact that the first date is really low means that the um the the it can be hard to sort of pickle effects so what's the what's the right solution to this problem so you you want to do by the way so
you want to do a cluster randomized design for the reasons we talked about you have to do spillovers so how do you do with this issue yeah probably um in different Villages or even like within the same Village first figure out like who would be interested in getting a loan and then randomize who you give them to well you don't want to randomize who you give the loan to because that would have the general equilibrium problems we were talking about earlier sorry say something else Brad and I said
that do exactly what talented Village exactly so that right so therefore what you have to do is you have to get Baseline data for everybody figure out their characteristics that are going to predict take up then you randomize at the Village level in order to sort of deal with the general equilibrium issues but you're going to sort of look at the impact for people who are like likely to take up so you can predict take-up from your Baseline survey and you know while ten percent it may be only 10 i
n general maybe if you ask a question of like if someone were to offer you a low would you take it maybe you can sort of get you know 50 or 80 take up on that subsample and that's that's that's the strategy and how you have to deal with that issue is that clear I I don't think I understand so the key Point here is that it's 10 on average okay but what I want to do is make an x anti-restriction on my sample such that the take-up rate which I can do equally in the treatment and control areas such
that on that restricted sample the take-up rate is really high so imagine I do a baseline survey and I ask you you know suppose a microfi institution was to come into this town would you buy would you take out a loan right so people who say yes to that question the probability of take up is going to be conditional and offering it is going to be like I don't know 80 or something right not 100 but 80 or something that high number so then what you can do is you can you run that survey at Baseline a
nd treatment and control areas you restrict your analysis then to people who are in that like high predicted take up category and then put in the control group so you're you're doing it on a baseline question so you're doing the same restriction and treatment and control areas but you're going to have high take up on that sort of on that subsample but you're still just randomizing it though you still randomized The Village level but you you're able to run your analysis on the people for whom you
expect predict to take up to be really high is that clear so that's kind of what they do um and so what they find is they find basically they find in general sort of summing up across all these things I would say they find on average not much on consumption um but they do find some heterogeneity of impacts and those are sort of focus on people who have existing businesses and I think at some level that's not that should not be surprising like what is this this was actually kind of surprising at
the time because everyone thought this thing would like radically transform poverty it doesn't do that what it does is it sort of like gives Capital to people who have like high marginal returns capital and like those particular individuals can like grow their business and that's sort of people who have existing businesses may be the ones for whom kind of the capital is going to sort of be most useful um and so that's kind of I think a summary of sort of some of the impacts so um one thing in p
articular I'll note is there was a meta-analysis by uh by meager um who's also a student here uh which came out a few years later and AJ applied and she basically said look there are seven different studies how can we sort of say something about how similar or different they are and the the thing she wants to Grapple with is that basically there are um differences You observe across sites or those due to sort of a combat those are you I see some differences those are due to a combination of just
sampling noise right like if I have a samp you know that's regular standard errors or it could be that we're they're drawn from different distributions like maybe like the the thing really works great in Indonesia and it works really terribly in India or vice versa or whatever um let me actually I think in the interest of time I think I'm not going to go through this in a lot of detail but let me just basically say she has a sort of a like a hierarchical kind of Bayesian framework for thinking
about this yeah let me actually not go through this in detail but let me just note that if you're interested in this in these issues of how do you sort of like combine across places you should go you should have a look at her paper which sort of helps you think through sort of how do we how do we think about this stuff so let me let me um let me not go through that um um in a lot of detail that basically they sort of show kind of not much um uh not not massive kind of average impacts um but ther
e are I think um uh bigger impacts in general across these various studies uh for people who have prior businesses which is sort of was shown here that sort of the blue graph is kind of the ones with people who don't have prior businesses and in general um you know the the distribution of outcomes look look shifted better for people who had who had prior businesses which are sort of consistent with what we thought before that basically this is sort of like providing more impacts people who have
businesses yeah um just a quick question for like framing like what is the size of like the microfinance loan in general uh uh I don't know the actual numbers like examples I would give anywhere from 50 100 500 to up to a thousand dollars I would say in that ballpark they may go higher than that but sort of you know those are sort of typical I think the example you guys read in the Sandpiper was like 100 those were those were grants not loads it was like 100 bucks right um so I think like you kn
ow that and and those are not uh atypical amounts yeah um okay so you know in general I would say you know what all those impact evaluations sort of showed was that for the average borrower in the sample it didn't do very much but um there were two important caveats number one um you know those were the the samp there may be a sample selection here like if you think about the designs I talked about most of them were selecting on kind of more marginal cases where sort of the bank was indifferent
either people or places um and number two there was a substantial heterogeneity so that kind of leads us to the paper that you guys read which is if heterogeneity is important kind of what else is breaking returns do people sort of know kind of what's going on here um so that leads us to this paper by uh by Hassan Regal and Roth that you guys um read for today so just to make sure everyone just remind everyone what this paper was This was um these I actually were these were all graduate students
actually here who when they were sort of uh working on this project as some of you at least some of you picked up in in your in your reading um so uh so what um uh reshma nataya and Ben did is they they um and I should say by the way one thing I think was kind of interesting is just as sort of a like kind of how how this paper came about in some sense that this paper I think Builds on the targeting paper that that sort of I talked about like three or four lectures ago and that we sort of said l
ook and they sort of said look there's local there's clearly local information for poverty targeting that's about but that's kind of about the level of people's income but maybe if people know something about the level of people's income they also know something about their marginal returns and that's sort of it's a harder problem it's much easier to know about the level than about the returns because the level I can just sort of look at the returns is like a counter factual but in some sense I
think part of their idea is sort of saying well look you know we're interested in microfinance but can we take these ideas from this other kind of setting and sort of adapt them here and so it's also I think kind of an interesting example of sort of how you know there can be intellectual Arbitrage from sort of one one field to another field I think I do having been here through this process I think that was kind of partially kind of how the how that sort of uh came about which I think is sort of
really interesting and I think you know can maybe be deep relevant guidance for for all of you so okay so what do they do so they asked entrepreneurs in in Perry Urban Maharashtra um to uh to rank their peers uh groups of four to six people on metrics of business profitability and growth potential okay and um and then they randomly distribute cash grants of 100 to a third of these entrepreneurs to measure their actual productivity okay so so why is so so why is it important to randomly assign t
his here so why do they need this step of randomly assigning the cash grants like why are they just like doing this and sort of looking at sort of like whether you can predict business profitability yeah [Music] yeah exactly because they're trying to pick returns not levels right they're trying to predict what is the return to giving you right an extra hundred dollars not what is your average right so they're trying just to be clear they're trying and this is an example we're sort of thinking ki
nd of precisely what theory is important their goal is to predict F Prime of K not F of K right so if they were if they were interested in sort of just predicting kind of the level that would be but F okay but they want to say well what's the what's the return so to do that they need to sort of predict this thing at Baseline get this information at Baseline randomly shock some people with a hundred dollars and then sort of say what is the treatment effect of you know a hundred dollars heterogene
ous and how does that very heterogeneously based on sort of this information at Baseline okay question do you think it's better to do a loan or a grant they do a grant why do you think they may have done a grant 's easier yes that's exactly why so they wanted to do a loan so like actually so this is not a case where like it was like X anti-op the optimal thing to do like I could I can also having having I guess you know been there through this process um their first idea was we should do loans a
nd sort of see what's the return to a loan and they you know spent a long time negotiating with multiple microfinance lenders over a period of a long time uh and in the end of the day couldn't you know find a microfinance lender who was willing to sort of randomly assign loans in this content Theory when you do it but they said oh actually maybe the second best thing is we can do grants and that's actually still pretty good because we can still get thing we want which is f Prime of K like so so
we can get F Prime of K directly rather than F Prime of K minus K right that was still kind of relevant you know both interesting parameters and this one we can just if we can you know raise enough money to give out the hundred dollars like we can just do this ourselves and so I mentioned that also because I think that was also a nice kind of like you know they had this idea and they wanted to like do it and the first idea was they should do loans and they they tried and they worked with a bunch
of microfinance lenders there's a whole process but then they were sort of uh adaptable enough to say well you know that thing isn't going to actually work in the in reality can we find a way to answer the same question that doesn't rely on us convincing partner do something they wouldn't want to do and that's how they ended up on Grant so actually the grants are totally fine um and uh and was sort of a a good example of adaptability uh sort of getting there uh yeah what do you think um well my
plan as well [Laughter] what does someone else think do we think Lonesome grass can be different I would think maybe the loan has more of an impacting Capital huh maybe the loan has more of an impact with it because why you have an incentive yeah so I think that because you have to you have to pay it back yeah so um you might think yeah so I think you sorry what would you say Wesley that was precise yeah so you might think they're going to be different let's see I think that um uh so going back
to the theory we talked about last time would there be like what were the differences there be so if they're yeah Rebecca um I guess also the stream of returns that you get and the time yes so the the timing of the returns are going to be important and we're going to talk about that in one of the papers that hopefully I'll get to later today which is sort of like the microphone like you have to make sure that the if it's a loan and you have to pay it back you have to make sure it it Returns the
money in time for you to pay back the loan and so that actually that question of sort of the timing of repayment can actually distort kind of your investment choices so that's important yeah um what else I don't see exactly how but I feel like it would change your risk your choice of diagnosis why don't we then hold this thought till we get to the paper about Grace periods which is which is actually test this idea experimentally yeah but also like not everyone would take up the line whereas eve
ryone would take a cash right so if it was selection if you learn then you might see higher effects yep yeah so that's right so you might find so everyone's going to want 100 bucks like and uh and you're right like basically you might that may and that might change the results right it might be that actually like we what we're picking up here is who actually has a business and it sort of is Will has a profitable opportunity whereas everyone would take whereas so they may not have heterogeneity a
nd returns of a loan because that selection of the loan might actually get you some of the selection of picking up yeah all right that point um I think it also depends on the degree to which the loan has like limited liability or not so if you think about kind of the model from before right in the model from before um uh If you're sort of fully yeah sorry that's okay is it right to say that if you're able to properly estimate like the Hazard rate be like probability of something happening back a
nd their beliefs about not only that that you should be able to just give them a grant that should reduce the same because like the difference no I think I think that I actually think the point I think you made actually about sort of the timing is relevant holding and holding like timing no because they may do different things with it that's the point like imagine I basically like like for example like we know that people like often borrow to go to school yeah right so that could be a perfectly
reasonable thing to do with a loan but if you have like a 10-year repayment Horizon but if you have a one year payment Horizon like that would be a terrible thing to invest your invest your your thing on because you're not going to have any returns because you're still going to be in school sure that like the payment arises like 10 years and all of the things that people are investing in like they would presume that they were probably leaving back the money like before they would need to repay u
m if on the limited liability Point couldn't you people agree to like give somebody a grant that would valued to them the same as based on their personal belief about like the project working out in their I'm not sure I see how maybe I'm missing him but I don't I don't immediately see that um I want to talk about afterwards but I don't I don't quite see it um but okay so I think the answer is they're not quite the same um for some reason the people people may do different things with the loan if
they think they have to pay it back for for either these timing reasons they may be they may be more if there's a risk aversion reasons right imagine you sort of like you know if you're if you have concave utility kind of you're worried about the bad state right so you may you know uh if you're if you're risk neutral maybe it doesn't make any difference but if you're risk averse then I might think of well if I have to pay this thing back in the bad state I'm going to be you know take take somet
hing that's more conservative where if I don't pay it back then I'm not so worried about the bad state so there are reasons they are different and the selection point I think caselo you also mentions are totally right there are reasons these things can be different um and that's probably why they started with trying to do a loan uh if the whole point was just think about micro finance that was the first best but I still think there's a lot they can learn from the sort of from this which is why t
hey sort of this um um okay so so so this is the so then the key regression that they're going to run is essentially this one so cash drop gets you marginal returns predicting it's just how your sort of characteristics predict your outcomes and then the key thing is this interaction of kind of like predicted times cash drop right um they also want to sort of say is there kind of additional information um uh or does is this kind of like better than kind of the machine learning uh approach okay so
they also say let's predict your returns based on some machine learning approach they take an auxiliary sample uh you know some whatever some training sample like 20 or whatever they predict your returns your heterogeneous returns based on your X characteristics they then can take your machine learning prediction and sort of horse race that against kind of the the cash drop prediction so um there's two ways of running that regression um I think um the way I would run that regression is uh is to
basically sorry this got cut off I would sort of put them in the same sample right in the same regression and um see uh whether or not sorry that's the ml plus ml times cash drop times ml so I would say let's control for sort of this and say does this predicted thing have kind of additional this this local operations have additional kind of breakthrough power over this they I think remember correctly do it two different ways they predict they predict the uh they do a different way they predict
the uh heterogeneity either just with sort of the X characteristics or with the X characteristics and the sort of the the local information and sort of see which does better in kind of this regression so they run this regression where predicted is either predicted based on just the the Baseline characteristics or Baseline characteristics plus kind of these additional X variables um I don't know if you do I don't have opinions of those two different approaches yeah Aaron so there's one one questi
on I had with this in the paper is um a typical microfinance institution in this setting are there agents also like embedded in the community I guess my my thought was maybe like observables and like this ml exercise is not necessarily development Benchmark but rather like what is it what are what are institutions doing in Sam's Club yeah I think that's a great point so I think it's certainly the case that like microphone like banks in general have loan officers and their job is to like find out
stuff about you right and they interview and they say do you look like you have like is this business plan like real or is it kind of like not real and like maybe that's better than their ml so I agree with you that would be an interesting thing to do would be to sort of get the banks to sort of do that you have to think about how would you incentivize the banks to sort of do that kind of in an in a compatible way and actually really put effort into it but I agree that would be another interest
ing kind of Benchmark to run um yeah right now wait so um when you say can the community identify good entrepreneurs are you asking like like it seems almost like the comparison that you're doing with the ml is saying can is the community good at marginally identifying information Beyond uh some baseline set of things that you would think would predict entrepreneurs that's what this approach that's that's what this question says yes it's not necessarily measuring whether the community is good it
's like identifying with it yeah so that right so this is actually so I so when I was writing these slides I like this is the um uh challenge with having you know your advisor teacher paper later so I like emailed them and I was like why are you doing why are you doing it this way why did you do the other way so so um uh and the answer that that I got back from them was actually on exactly this point that this equation which I wrote down over here um sort of my Approach is a this is this equatio
n uh basically says do they have kind of residual information so the test if I do it this way then when I when I run this regression what I get from beta 3 is is there residual information that the community has that the machine learning approach doesn't know but doesn't really tell you kind of quantitatively like how much it is their approach that they're going to do this way I think tells you actually does actually sort of answer that quantitative question and sort of says look how much better
is sort of the overall targeting with this kind of combined approach of using all this information than if we just use the ml alone so they're answering slightly different questions I think that's what they went with this approach um that answer does that answer your question yes yes okay other so it's a slightly different question and that's the answer and and each regression they both have they're both kind of related and I think if one you know if you get one you're likely to get the other o
r not well not totally but I think that that's kind of why they do it um all right um so so what do they find so the first thing is um they ask whether or not uh people are able to predict um things about um whether they're sort of predicting Things based about other people's characteristics so like do they know their income you know income profit assets um digit span so digit span is a test of like um how many how many digit digits you can remember if I give you a random set of digits how many
can you remember and that seems to be a test of like numeracy and things like that um so they seem to have a lot of information on people um that's what this sort of says I think in some sense this is kind of the key graph it's a little hard to read um but this is the predicted marginal returns from kind of the the the uh people in the community the black line is the outcome for those who lose so in general it's the case that those who lose have more those who are predicted to have higher return
s of higher profits and those who are who are who don't so that so your predicted marginal return is correlated with like f of K and it just F Prime of K right the gap between the red line and the and the black line is the heterogeneity is the is the marginal Return of the of the grant so everybody gets a return from the grant right I mean that you know whatever the error bars here but but in general it looks like everybody's getting returned from the grant but the the Gap is bigger sort of base
d on these high kind of percentiles and low percentiles which says that people are predicting not just the level but also predicting kind of the the marginal return um and that is uh that is what we see over here which is that basically like you know this is winning winning the grant times the rank um and uh they basically find that sort of those predictable heterogeneity that people who are the have higher percentile ranks basically are are doing better and they can sort of look at it sort of y
ou know sort of nicely in some ways it's easier to interpret it kind of just that they break into three groups kind of dummies like top third middle third bottom third and you can basically see that sort of the the highest returns are coming from the people who are predicting to be sort of in the top third by their by their by their peers um and uh and um you know what do they do they they buy more assets they also work they work harder sort of in response and they hire kind of more labor so the
y sort of they both they they increase both their asset they both increase their K more and they increase their L and other people and other people's out in their businesses um and this is the thing I was just sort of pointing out which is this is the approach where I do where I where I where they they take your prediction based on their their older kind of observable characteristics and say what's the heterogeneity and the marginal returns for the top and the bottom and these is the ones that a
lso kind of include the rank and the idea is that basically these have some additional predictive power compared to the other ones uh and you can you know where this is a lot better on predicting income and profits um they both have some um sorry uh for for income and log income you know the the ones with the additional percentile ranks is doing a lot better than the ones without for profits um maybe there's some evidence that that the the control has some predictive power although it's not as c
lear in all specifications um so what is this what is this uh what does this mean so I think it means that um you know there is kind of predictable heterogeneity and local people in their Community sort of know it um uh maybe even above and beyond um uh uh the observable x's and then going back to Aaron's point then there's a question of sort of like what do you do about it do you try to elicit it from the community do you try to have Bank agents kind of go in and figure it out like it's not cle
ar but I think that sort of the whole point is that basically like there is a lot of this heterogeneity and returns is really something that is uh predictable and sort of you know understanding kind of how do you get the right the right people in and select that is kind of a really important uh it's something that Banks obviously spent a lot of time on is something sort of uh important moving forward different [Music] [Music] so I don't know that we necessarily know that because in some sense th
e whole point of it is that it's there's predictable differences even above and beyond everything they can observe in the data right that's kind of the point of the large part is that we don't like that it almost by definition like the point of exercise is there's some additional piece of it that is sort of beyond what we can observe in the data which sort of makes it hard to answer that question yeah I think or maybe kind of just a question too is like when people are ranking their peers what a
re they thinking about like I don't know like it just seems like it's you know what is it that communities know that we don't record the data I don't know if they like did any kind of politic wants each other so I don't know the answer to that question as I said beyond it's a little hard to know because the whole point is it's something Beyond what's in the data so it's a little hard to know um I don't have a strong I don't have a great answer that yeah well it's like not there's all the data it
's not things not visible in a one-time Baseline servant correct when I go to the shop and look at it so things about like service quality and stuff could be it could be it could be entrepreneur entrepreneurial ability which is sort of hard to measure like I bet if you thought about your think about your friends from college for example like if you were to rank them in terms of their entrepreneurial ability do you think there's heterogeneity there and do you think that's above and beyond what yo
u predict based on their grades like I mean I don't know like through that introspection I'm curious what you think like my guess is yes but like you know I don't know like and and you know there is like there's a lot of stuff that goes into people's personality and their business ability and whatever then it'd be sort of really hard to capture sort of in a simple survey or even a complex survey yeah right what LinkedIn I'm sorry what that's what LinkedIn does now they ask people in your network
to rate the relative skill at a particular thing of other people in the network do they seriously yes that's super interesting yes has anyone studied that it's implemented by an economist who um what do they use that for uh so I think it is like you'll ask a question that's like um you know who would you go to for questions on Python and then it'll give you like three of your friends and then you select one of them just because it's fun or something and then what are they and then they're like
then they take that person again this person is like more skillful uh search results that's fascinating and I would love to see someone study that if they haven't done it already that's super that's super interesting um okay so but but it's exactly the same idea that people seem to know stuff about their peers and like you know maybe we can maybe LinkedIn to figure out a way to monetize that um but I think it's super interesting I'd love to I'd love to that seems like a great topic for someone t
o study if they haven't already um okay so what I want to talk about is there any other comments you guys wanted to raise from the paper that you guys read from the paper okay so what I want to talk about in the last 20 minutes is sort of my the the unbundling of microfinance okay Okay so here are some stylized facts about micro Finance in general default rates are really low not always but in general they've been pretty low um and the original kind of Grameen Bank style microfinance model had m
any elements okay they lent only to women they had a weekly repayment schedule where you start paying immediately like as an aside I think this is a little weird right if like here's a loan go make an investment and I want the first payment next week like it may not be optimal right but there may be reasons the reason I think they do it is they want people to get in the habit of repayment I think that's that's kind of the that's the theory behind it but as an investment decision like it's you kn
ow on the other hand you can say well fine I'm gonna like you need to buy a thousand dollar a hundred dollar machine I'll give you 150 the first 50 years paid back to me in the first five weeks and then you're gonna start giving returns maybe that's what they do I don't know saying it's a thing they have group Lending so basically you come into this microfinance organization in a group and initially in a lot of these models you had group liability so that means that like you know hazela and and
Whitney and I are in our group and like if Hazel defaults I'm on the hook for it okay so so don't default that was the idea though so like therefore like I'm not going to want to be in a group with them if like they if I think they're kind of dodgy so we solve a selection problem and we solve the moral hazard problem because if she starts to default I can go like you know we're friends and I can go like you know bang her door and be like no no like you have to pay this back otherwise I'm gonna h
ook for it um so that was a thing um they're regular meetings where they could have met with people all the time um their Dynamic incentives so basically you had you started off with really small loans and if you pay them back you get bigger and bigger loans so kind of the return like the incentive for paying back was kind of like bigger and bigger loans um they've credited officers who basically like are spending lots of time sort of in these group meetings and talking to people and monitoring
them and doing all kinds of stuff and they have high interest rates um you know at least 20 per year but often like way higher so this is like a bundle right and you know there's a bunch of theory we can tell stories as to why all these things might matter and how and whatever and I think that there's been sort of a series of papers that have tried to um decompose them and sort of say well we're just gonna like study this or we're just going to study this or or so on and so forth to try to under
stand kind of like how do we take this giant bundle of stuff and figure out kind of what what's important and what's not and what's actually helpful and what's not or whatever and so I think it's also um useful to sort of see kind of how this research program kind of as a field has sort of gone on because you can see people have sort of like been sort of hacking away at kind of like you know chipping away kind of this big bundle to sort of like de-bundle it and sort of understand kind of what's
going on here yeah I don't know the answer that question in general I think in all of them they want them to not lose money but beyond that I don't know the answer yeah we'll see just thinking about our conversation from the last time is the this last point of high interest rates something that like an organization like granny bank would sort of evangelize is like this is what micro credit should look like or is it just something that like comes to be by well I don't think they I don't think the
y're like pro pro actively saying yeah it's really awesome in order to really high that's not how they sort of spin things sure but like if only because it seems like from our conversation from last class that like this is a component of micro credit that would incentivize riskier Investments than whatever the other hand if they can't lose money so like that's the other problem like maybe all this stuff may like cost money like going back to the whole that was kind of the point of the monitoring
model right all this stuff costs money and then if we have all this like real costs and we have a really small loan like that that gives you the high interest rate because you got to cover those cost levels on a really small base so that's kind of the that was the point of that and sometimes that was the point of that monitoring model was to say look we have all these expensive activities and you know and they're basically fixed costs and if we would fix costs divided by small loan that's gonna
look like a high interest rate but aren't a lot of these subsidized anyway not necessarily they may a lot of them may be non-profit well it's not clear but a lot of them aren't some are poor profit and they just have to make money some of them are not for profit but they still gotta break even and so uh I think it's not yeah okay so um let me in the entry let me just pick a couple of these um and see if they if they'd matter let me tell you a couple so one uh because I'm not gonna have time to
go through all them um what was about lending to women and sort of said look is that kind of like a relevant thing so this paper by demon McKenzie and Woodruff uh basically said um it was another was actually kind of one of the original cash Grant um uh papers and they said um let's uh um they might work in Sri Lanka they basically identified a 400 households who had a small business uh not a lot of capital and they randomized them to get a small Grant of either cash or or or or or or or an asse
t about 100 100 or 200 and they sort of followed them up kind of uh I think six months later and then maybe even five years later and the main finding was you know uh they had very large Returns on Capital that's consistent with this idea that you know that they were credit constrained so that the returns look like 60 returns per year so really high returns um but they found in their paper there were no effects for they did some heterogeneous stuff and they found no effects for for female-one bu
sinesses okay so there's a main effect I mean in fact times women is like nothing so that actually says well maybe there was not something kind of going on maybe there was something different with the with the with women's businesses and there were other papers that sort of had similar findings um and I'll just note that one um one kind of interesting recent paper on this paper I burn hard at all which goes back to kind of this hair by dumb by this dumbbell little paper and says well what's goin
g on here and they have kind of an interesting hypothesis which is their point is we shouldn't be looking at uh the business in isolation we need to look at the whole household and maybe what's going on is not there are differences in in men versus women as entrepreneurs but sort of what's going on in the like maybe their whole household Dynamics are really important and thinking about the whole kind of you know actually we talked about inter-household issues maybe that maybe that maybe we shoul
d be thinking about them in the context of a household um and in particular um what they what they show is that basically uh they show that um they replicate the result in Sri Lanka that there's kind of no impact on the on the female under the female-owned Enterprise but they note that when sort of the the woman gets a grant right their overall house Summit household income is actually going up and so what there are what actually they argue is that basically there are um there are uh uh you have
to be careful about sort of looking at these sort of female owned businesses and the other thing I think they also look at although I didn't bring that that that that that table is they also show that actually if you look at sort of seeing like uh female-owned businesses where the woman is not part of a household a household with a man where she's just sort of like acting on you know on her own that basically then her returns look just like the kind of men's returns so essentially what they arg
ue is that what's happening is not that her business has like lower returns but sort of in that household bargaining if she gets the grant maybe she channels into her husband's business and her husband this is the one that grows so in general kind of their household returns Are Not So Different yeah are there um you need to like demonstrably invest in your own business or is this like I don't think for the cash Grant one a handoff too I think they're able to hand it off yeah is there okay uh I k
now a number on like if I give say the male male entrepreneur a male household has a business the woman has a fur business if I give one dollar to the man's business what's the effect on the masses let's say investment versus if I get the bottlers to the moon's visit to the woman how much does the male's business yeah so I don't think they can totally look at that because I don't think that they don't think the data measure the man's business in that case but they do measure household income and
that's where they can look at this so they don't have the they don't quite have I don't I don't think the data to answer exactly their question but I think that this is suggestive of that that's what's going on um okay so I just want to mention that one briefly because I think I I do think it I just because it illustrates this point that we have to be think when we think about these small household businesses we should be thinking of in the context of a household which is kind of making kind of
decisions across households in the same way we had plots and people were thinking about plots you have to think about kind of what is the whole household doing and just be and actually they point out that this is not thinking this this whole paper points out that not thinking that through clearly actually LED you to sort of very different kind of results yeah yes I think that I think that um well first of all the average effects are interesting and the number but yes I think the point of this i
s not it's not that it's useless I think the point of this study is say look that you want to understand these household issues too I don't think it was um but like you're really not measuring like you're you're not even unless you're looking at like women who aren't like just a woman in the household if you're like really not even close to measuring zombies yes that is the point of this yes that is the point of this PornHub so but like I so right so I'm just saying but people we this is people
they didn't figure this out until recently but yes so but I think that's kind of why this paper was you know came out and published a good Journal so I think that like the what was the sort of say and the reason I'm sort of I thought it was worth mentioning is exactly this point that saying look if you're measuring these household businesses in a context where people are sort of making decisions across businesses within a household it is important to sort of take the big picture that's kind of I
think the main point here yeah let's take away from here like in terms of if I want to help a family so if I can like help taking gym class Twitter the email or your husband or wife these are like a that's a different question which says if if we were to randomize if we had Mo that's to answer that question you have to have a different study we should say we gotta I gotta take I gotta select a sample of people with sort of both men and women and randomize who gets it and see what the outcomes a
re so I don't know of anyone who's run that study level there are many studies in this area and I it is possible that someone has done it and I don't know the answer can we talk about unconditional cash transfer that did that cash transfers have done that for cash grants yes the the give directly uh experiment does that in the contest of cash transfers that's one of their treatment arms but that may be different first I mean these were sort of larger actually that was pretty large too but they w
eren't sort of restricted to entrepreneurs with entrepreneur activity so maybe like for this entrepreneurial subset it might be different [Music] um I don't think they can do heterogeneous so what you'd want to do so you're absolutely right so so like various running regression of like x equals treat plus um woman plus three times one right that's what that's what that regression column two is so you want to say look we also need to include like plus industry plus treatment times industry right
not just industry dummies but also treatment times industry because we may think they're just heterogeneous returns if women are doing like whatever yeah things that have lower returns to Capital you're going to be picking that up right so you want to run that given that they only have 400 like businesses or whatever I'm sure I don't think they have enough power to run to include this justification but you're absolutely right that's another thing that you another hypothesis that you want to thin
k about sure um okay um let me skip this one um so another study that that this sort of this team Coopers has looked at um is this question of sort of uh this fact that you need to start repaying the loan as soon as soon as you get it so like you get the loan you gotta start your paying next week okay so is that um as I mentioned on the one hand that seems really kind of weird uh to me on like you know I'm gonna make an investment and like it may take some time to get Returns on the other hand s
ome of the kinds of Investments that people make look like I'm gonna like I have a shop I'm gonna like buy some more inventory for my shop it doesn't take really very long to like actually start having Returns on that so maybe it's okay or as I pointed out like you can actually replicate the non-grace period version of this by the non-weekly repayment version by sort of saying look I'm going to take if I got 100 150 loan I'm paying the first 50 bucks put it aside and use it to make my first paym
ents and I don't invest the other hundreds right so with a slightly larger loan I can kind of replicate this kind of myself so does this matter um so what they do is they basically they they run an experiment where they're going to nicely sort of zero in on kind of exactly this feature um and say some people are going to get the normal contract where you start paying immediately and some get a two-month grade period even this is not a very long uh period of time but they get a two-month grace pe
riod to repay and I think what happens is kind of interesting so sorry what do you expect to happen someone else yeah Apollo um with the great spirit you might like higher returns back on average but perhaps they're more weather you'll get like higher returns with the grease period but perhaps also more hydrogenating people are taking riskier actions which raise expected returns but not the rest of like losing at all yeah right so again someone else might have a higher default rate from the grea
test period why well is the reason why you were doing it immediately yeah so if the whole point is to think like you may just have people gonna get in the habit of default it may also change your investment choices and you may change different things at other comments interesting solutions by what so say like it will hurt some that's how some people will okay pay back say their behavioral advice see some people really need to think about like be disciplined about paying back in like next week or
so but there's also a possibility that like any investment opportunity that will take you more than more than a week to repay that you couldn't do before not only to do that so like the middle effect can be zero but there's only increasing variance yeah could be absolutely so what do they find so they find um so they find people do seem to they they use more of the the thing on their on their business and less on sort of household expenditures um this is not totally obvious to me actually that
this would be the case that there's sort of this this discrepancy between sort of like you know how using the money for loans versus kind of business like it wasn't obviously that leaving kind of a margin or why that margin should necessarily like why should you move from kind of consumption smoothing to business investment maybe the business investment I mean the customer movie never pays itself back so that's not sort of totally obvious um to me that they would find that but maybe in some broa
der sense this is kind of like at least I understand kind of what this is going to be like and this is kind of like totally zero risk in some sense like I know what it's going to be and I can plan for it whereas this is kind of more risky um they also find that um there are higher impacts on default so uh people do um uh default more uh when you give them the grace period so it is it is the case that this is trade-off on the other hand uh people's business is actually um do go up and their incom
e goes up a lot more actually with the grace period so I think that this this actually I think also illustrates that there may be sort of a tension between what's good for the bank and what's good for the sort of the the individuals right the bank does not care about your income it only cares about it maybe if it cares about if it can charge kind of higher higher interest rates to you but the bank just sort of wants conditional on making the only good industry just want the loan paid back so the
bank is going to take action to make sure the loan pay back even if that is necessarily not necessarily optimal for you and that's just there may be sort of this tension between sort of between these things yeah the story behind the default instead of being about habit being about like risky Behavior yeah 100 absolutely yeah yeah absolutely it totally could be about risk taking Behavior I think it is probably I mean certainly we don't actually I don't think we necessarily know um we know that s
ort of people are investing more in their business doing more stuff earning higher incomes and defaulting more we don't I don't think we necessarily know from the study although I can't remember for sure I don't think they can differentiate between that's the default is because it's increasing the variance of their outcomes sometimes it fails versus this other kind of just Direction Factor like I get out of the habit of like paying back or whatever so I don't think we can really differentiate be
tween that yeah what sorry Disney experimental design allow for like different groups of people to select I don't think so but that's a good question I don't remember it well enough to say but I'm pretty sure they randomize it after people take up I think it's a surprise so I don't think it has a selection effect but that would be interesting if it would be that you could imagine that could be important I don't think so but I don't remember 100 other questions yeah Amen on average non-business k
ind of decreases but I wonder like if the increase in non-business spending uh predicts people like the behavior like that in a sense yeah I don't know how to identify that using other characteristics that would predict kind of heterogeneous response wouldn't predict exports it wouldn't be causal yeah I'm I'm not even sure I would run that regression but like I think you would need some other context variable that predicts predicts that I'm just about out of time so let me let me just um let me
just Sim up by saying that sort of like there people have gone through some of these other characteristics I mentioned I'm not going to go through them in in detail now but you should look at them if you're interested I think the whole they went through group lending um they went through um uh they went through completing some other things um people have sort of systematically tried to decompose that list and sort of understand which of those things are important and I think it turns out it's ki
nd of a lot more nuanced than sort of the initial kind of view from the green bank which we have to have kind of this whole package or sort of nothing uh the final thing I'll just mention some other topics in credit um one thing we haven't talked about which may come next semester but maybe not depending on sort of how much time they have is um local indigenous institutions so uh rotating savings and Credit organizations so you know one thing people sometimes do if there's not kind of formal cre
dit is they'll get together and sort of say well look you know maybe we're all going to put you know every month we'll get together we'll have a meeting we'll all put ten dollars in a pot and we'll take turns who gets the ten dollars and you can think of that as kind of like that's a rotating organization where in sometimes you can think of it as like a credit organization right so you know the first person in some sense is getting a getting a loan of 120 for 12 of us in the group 120 which they
were paying over time by sort of putting it into the pot later and these informal institutions are very common throughout the developing world and uh and people have studied those to try to understand those as well so I just wanted to mention the note that um I think we're going to sort of talk about demand for credit uh this will talk about a little bit more um probably not in our station probably next semester um so I just want to mention sort of understanding people's demand for credit uh I
guess we'll talk about that next semester and you know we'll talk about um some some more stuff on sort of banksy's intermediaries um and credit constraints for larger firms um next semester so I'm going to wrap up here um thanks very much it's been great talking to you guys this semester um thank you and uh I've left you five minutes or so to go fill out your teaching evaluations so we very much appreciate that okay thanks
Comments