Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=Transcript of the talk, from the video's captions. Auto-generated: speaker names in particular are unreliable.
=

# Micro-enterprise Saturation: Critical Mass or Overcrowding?

Authors:   
Discussant: None  
Video: https://www.youtube.com/watch?v=mC4ESywpIDc&t=12530s


## Talk (03:28:50 – 03:38:16)

[03:29:09] >> great. Thanks so much for including this in the program. Uh in a nutshell what I'm going to ask is suppose I give you the means to start a microenterprise but instead of giving those means just to you I give them to everybody around you  
[03:29:23] as well. Is this on net going to make you better or worse off?  
[03:29:28] In many settings we can't rely on a strong private or public sector to pro provide enough jobs to get people out of poverty and we end up relying on the entrepreneurship by poor people themselves to do that. Um, a lot of  
[03:29:43] programs have come up uh like cash plus or graduation programs that are supposed to help poor people start some kind of entrepreneurial activity. These tend to be very costly programs, but there's this uh promise that as you scale them  
[03:29:58] up, perhaps you can bring down the per household costs.  
[03:30:02] Whether or not that's a good idea is going to depend on what happens to treatment effects as we scale these programs up. And that's going to depend on exactly how you scale this up. Do you go to a lot of villages and treat a few people or do you go to a few villages but really treat everybody who you can  
[03:30:16] find? And so, uh, put differently, I'm going to ask, do entrepreneurship support programs at higher saturation produce, uh, positive saturation spillovers? So  
[03:30:30] some kind of critical mass where firms make each other viable, some multiplier effects, or does that lead to overcrowding of small village economies with too many very similar firms?  
[03:30:41] This is going to depend in part on the sectoral concentration of these new firms. And so since I exanti found the overcrowding uh scenario to be more plausible, I am going to cross randomize an intervention where I'm trying to get  
[03:30:55] people in a program like this to start a more broad diverse set of businesses.  
[03:31:02] I answer these questions in a randomized control trial uh where I randomly assign 1900 households in southern Malawi to a cash plus program. This consists of regular cash transfers, a training on  
[03:31:16] how to run a microenterprise, one-on-one coaching, savings groups, and a large uh one-off cash grant for them to start a business with. The program targets young mothers uh of a child under the age of  
[03:31:30] five who live in extreme poverty. This is a saturation design. Uh I have 72 clusters, 26 households each. I'm going to randomly assign these clusters to three levels of saturation. either a pure control group where nobody gets  
[03:31:44] treated, a low saturation group where half of the eligible households are treated or a high saturation group where all eligible households are being treated. On top of that, I'm cross- randomizing among the treated clusters and very light touch coordination  
[03:31:59] intervention to try and get people to think about what type of uh of sectors they should enter.  
[03:32:06] A bit of context outside of agriculture in the study context that I'm working in uh petty trading retail businesses are very very common and popular uh in a census I conduct about one in every five  
[03:32:19] households has some kind of petty trading business going on. When I go back with my implementing partner in previous cohorts of this program that they implemented, they virtually created  
[03:32:31] only new retail businesses as well. When I ask households why they pursue the kinds of livelihoods that they are pursuing, far and away the most common answer is, I saw somebody else make  
[03:32:45] money this way and I also want to make money. That's a pretty sound strategy if you're the only one starting a business at a given point in time. But for a program that's designed to start a dozen or two new firms in a village all at  
[03:32:58] once, this might become a problem. So with my implementing partner, I developed a very light touch intervention that we tag onto the existing business training where we do two things. First, we have people think  
[03:33:12] about the kind of GE effects of what's going to happen if we all start a very similar business. And second, I give them room to share the business ideas that they've been developing over the course of the business training with one another and then give them some time to  
[03:33:27] coordinate, talk it through, see if anybody wants to change. We're not going to impose that anyone change their um change their business idea.  
[03:33:36] this did not have the intended effects um both in the business plans that they submit before they receive their money but also in the actual businesses that we track that they conduct uh months  
[03:33:51] later we see a very high propensity again of retail businesses over 80% of the proposed businesses are retail the coordination intervention neither in what people proposed nor in any of the  
[03:34:03] subsequent uh follow-ups changed the types of businesses that people were doing. Uh I do some work in the paper to figure out what it is, what other barriers there were. The information barrier would have been a very easy and  
[03:34:16] cheap one to fix. Uh it turns out that the major bottlenecks really are around marketable hard skills. Almost none of these women are able to do anything that would earn them money uh in in the private sector and uh the availability  
[03:34:31] of child care uh for for their kids. Um, this means that for programs like these, if it turns out that sectoral concentration is a problem, these two barriers of hard skills and child care are much more costly to remove than a  
[03:34:46] simple information constraint. All right, so we have a program with very high sectoral concentration. What happens when we ramp up the saturation of this program? In the very short term,  
[03:34:58] nothing. uh households, sorry, firms in both the low and high saturation clusters outperform the control group by a little more than double the income. Uh that's something we've seen before for these graduation and cash plus type  
[03:35:12] programs. But when we go 16 months out, uh first of all, treatment effects are increasing over time. That's encouraging to see, but for the high saturation, they're starting to slope down. Uh for now, this is somewhat imprecise. only a  
[03:35:26] third of my sample has reached the 16month mark. Um, what I'll find out is that women in the high saturation group have about 20% lower earnings, but it is also 30% cheaper for the impletor to  
[03:35:41] conduct a program like this at higher saturation. So, at least the point estimate would kind of justify still to do this at higher saturation.  
[03:35:50] Uh, the picture goes into a little clearer focus when we look at business survival. Uh you can see that again at the 16-month mark about a third of the newly started businesses in the high saturation cluster have exited and  
[03:36:04] predominantly they're moving back into the casual agricultural labor that they were in before the program started.  
[03:36:11] Finally, and this is maybe the clearest sign of there being uh kind of overcrowding going on. While the low saturation version of this program reduces the number of households that have left the study area over the period  
[03:36:24] of the study, uh in the high saturation area, people actually use the funds that they get to pack up their things and leave and move somewhere else.  
[03:36:34] A lot of the kind of work under the hood is driven, I think, by how people are using the funds uh that we're giving them. uh you'll I think remember a paper by Dennis Edgar and co-authors uh on the  
[03:36:47] GE effect of pure cash transfers. They demonstrated very high propensities to consume the funds and then found positive spillovers onto people who were not treated uh around you through  
[03:37:00] consumption multipliers. Uh I show that in this uh bundled intervention where we're really encouraging people to invest, people follow that advice. the marginal propensity to invest is much higher than in the uh pure cash transfer  
[03:37:14] setting. But in turn, I then show these heightened competitive effects uh that lead to saturation actually going the other way around. This may also be a piece in the puzzle for why uh the pure cash transfer studies have so far not  
[03:37:28] maybe had the strongest long-term impacts. Graduation type programs seem to just get people to really invest this money much more productively.  
[03:37:37] Uh final thing I'll say uh whether or not these results would generalize to other contexts I think depends a lot on the kind of direction in which these businesses are operating. Uh here we have a concentration in the retail  
[03:37:50] sector where I go to a different town I buy some stuff I sell it from my porch where I can watch my kids. If you get people con to concentrate in a sector that generates uh local exports out of the village, you might run into much  
[03:38:04] fewer of these problems. But that's speculation at this point. Uh thanks so much for having me and enjoy your lunch.