Vanisha Sharma on the Effects of Internet Expansion in Developing Communities

In this episode, Shruti speaks with Vanisha Sharma about her research on social media and how it impacts farmers, from their farm expenditures to their battle against pests.

SHRUTI RAJAGOPALAN: Welcome to Ideas of India, a podcast where we examine academic ideas that can propel India forward. My name is Shruti Rajagopalan, and this is the 2023 job market series, where I speak with young scholars entering the academic job market about their latest research on India. 

I spoke with Vanisha Sharma who is a Ph.D. candidate at the Dyson School of Applied Economics and Management at Cornell University. She earned her BA in economics with a double minor in psychology and sociology at the University of Hong Kong. She is also the co-editor-in-chief for the blog Economics That Really Matters, which encourages early researchers to share development economics research.

Her research involves estimating effects of internet expansion in developing, rural communities. We discussed her job market paper titled “Social (Media) Learning: Experimental Evidence from Indian Farmers.” We talked about how farmers learn through social media, how it impacts their farm expenditure, their battle against pests and weather shocks and more.

For a full transcript of this conversation, including helpful links of all the references mentioned, click the link in the show notes or visit mercatus.org/podcasts.

Hi, Vanisha. Welcome to the show. It’s such a pleasure to have you here.

VANISHA SHARMA: Hi. Thanks so much for having me. I’m really excited to talk about my paper.

RAJAGOPALAN: Yes, and I’m excited to talk about two things, which are actually in vogue. Agriculture is always the largest part of the economy, most number of people in India working in it and so on. It’s pretty important, but you have managed to combine two things which are going on right now, which is the digital revolution in India and its impact on how farmers learn.

In particular, you have this randomized control trial where you assign farmers to WhatsApp groups, and it’s the first time WhatsApp groups are actually useful, right? And in your treatment, you also have farmers interacting outside of the WhatsApp groups and you’re trying to study if they actually learn more about agriculture, if they share information, if this changes how they actually do their job and so on.

Can you just walk me through, what was the big picture idea behind the paper?

SHARMA: Thanks for that nice summary. It was exactly that. I noticed, as most of our listeners probably know, India has been going through a digital revolution. This started maybe five, six years ago when there was a public-private partnership because of which hundreds and thousands of kilometers of fiber optic cables were laid out throughout the country and this increased access to smartphones, to mobile phone data, and just completely changed the communication topography of India.

I wanted to use this existing infrastructure and test whether farmers in the most rural, remote areas can also benefit from the new agriculture technology that’s coming up. For example, there are now these smartphone applications that use deep learning neural networks to diagnose pests just by seeing a picture of a leaf and they recommend pesticides. But in my pilot work, I found that most of the usage of these apps was close to cities or in suburban areas, so I wanted to test whether we can bridge this gap between rural and urban and make use of this and see if it can benefit rural, remote farmers as well.

This motivated my paper. The problem is that social media networks are formed endogenously, which means that people self-select into their social media groups which creates echo chambers and it’s hard to tease out effects of these groups. To address this, I conducted this multivillage randomized control trial in which I exogenously assigned farmers to WhatsApp groups. And we made sure that these farmers did not know each other in person by making sure that these villages were not adjacent to each other because oftentimes, if they’re neighboring, then farmers know each other.

Broadly, I test the intent-to-treat effects on farm expenses and revenues, and the mechanisms that I test are farmer information-sharing behavior and their beliefs about unknown farmers and agricultural information sharing with them.

RAJAGOPALAN: On this, I understand that you have two kinds of treatments, right? The first is the fact that those who have a smartphone are actually assigned to a WhatsApp group, like you said, to someone who’s not in an adjacent village. The second part is that a subgroup within that is also encouraged to interact in person and not just on WhatsApp groups. How does the mix of in-person plus social media impact behavior differently than just assignment to social media or just assignment to in-person?

SHARMA: Yes, that’s a great question, and this was especially challenging to do—it was in the midst of the pandemic, right after the Delta wave, so we had to make sure we had proper enumerator protocol regarding social distancing in these in-person events. The idea was that by making farmers meet in person, you’re simulating the already existing in-person environment that exists between farmer groups, so farmer-producer organizations or when farmers go to an input dealer’s shop, and they interact with each other in person.

We wanted to see, in addition to that, if there is some virtual interaction, does it have any significantly different effects? Is that in-person component necessary for those farmers to interact with each other?

I find that having the in-person groups does not significantly alter the farmers’ beliefs on exchanging information. Farmers that are on the WhatsApp groups in both treatment arms are more likely to be willing to share agricultural information with unknown farmers, but between the treatment arms, this difference is not significant, which means having that in-person component did not significantly alter their beliefs about sharing agricultural information.

RAJAGOPALAN: The content of what they shared differs, right? This was a really fun part of your paper to read. You do some textual analysis and you’re going through what is the actual content of what these guys are discussing. I was just so happy to learn that even in a randomized control trial and farmers being assigned, WhatsApp groups are just like family WhatsApp groups. They very quickly degenerate into good morning messages and wishes for all the festivals and holidays and so on. But for people who’ve interacted in person, the composition of what is actually being discussed is different. Is that correct?

SHARMA: Exactly, and I’m glad you pointed that out because that was a very interesting finding. It’s almost as if when we create our social networks, there is a stage where we are getting to know each other, where we are socializing with each other, and then we delve deeper into our interests and into what we really do, what we care about. It seems like the group that was only connected virtually was still at stage one, getting to know each other, sending each other greetings like distant family relatives.

In the group that was interacting in person as well, they moved on to the next stage, and they went past the socializing stage. They were talking about market price, daily price information, which was a really interesting finding.

RAJAGOPALAN: Why do you think there isn’t a significant difference between just the amount of conversation between your two different groups, one which is online-only, and one which is online plus in-person? Is it because you were doing this during COVID and the in-person interaction wasn’t as great as one would like?

Or is it because, in India, farmer communities are largely based on region and caste, which means there’s already a high degree of commonality between them, which means they can just pick up conversation online and it’s not that different whether they’re online or in person? What do you think is driving that lack of difference between the two groups when it just comes to the volume of content that is shared?

SHARMA: That’s a great question. I think, partly, it may be because of COVID, but I also think the in-person exposure, by design of the study, was pretty limited. Like if you imagine, on the WhatsApp groups, you were sending two text messages on a weekly basis, but the in-person meetings were happening only once a month, so it wasn’t as much in-person interaction, which may factor into the results as well.

RAJAGOPALAN: Were you surprised by any of this? What did you expect to find when you were going into the study?

SHARMA: I was honestly very happy to see that there’s more information exchange happening in the treated groups, which is what I wanted, but what was interesting was that—the text analysis has two components. One is the content, which we already talked about. The other is the frequency of exchange, and that varied across time, so the study lost momentum after a few months of the harvest season of the first crop cycle, and the whole span was three crop cycles.

It was interesting to see how the frequency of the conversations changed over that span. I was surprised to see the insignificant differences between the in-person and virtual groups because in the traditional literature we have in development economics, there’s so much literature on social learning and technology adoption in agriculture, and most of these studies focus on in-person peer effects or farmers that live in the same village or meet at the same organization or have the same input dealer. For me, this was a surprising result, that having that in-person component did not really change their beliefs about interacting with unknown farmers.

RAJAGOPALAN: There’s another thing that’s super interesting, which is that you find that when the farmers are in the treatment group, they actually increase the amount of spending on their farm, but it doesn’t necessarily lead to an increase in the amount of revenue, at least in the short term. In the long term, we’ll wait and find out. What do you think is driving this result? Is it “keeping up with the Joneses” kind of a thing that these are farmers, roughly, of very similar region, similar socioeconomic class, if those people are putting in all these investments, you should also put in these investments, or is something else driving that?

SHARMA: I can say that in the sample, there was a wide range of socioeconomic status. Although majority, 98% of the sample, is smallholder farmers, there are a couple of farmers who are much more well-off, so I’m not sure if that’s the reason. I think that what’s hard to measure in this study is whether the increase in expenditure is coming from the quality of the input or the quantity of the input.

That’s why when sharing information on WhatsApp, it’s so important to make sure that the information being sent is authentic, coming from good, reputed institutions and so forth. For this paper, we collaborated with local universities and research institutes before sending out any information. The other side of the coin was that we wanted to solicit information from the farmers, so we also asked them questions and tried to get them to participate on the groups.

Anyway, regarding the input expenditure, it was really hard to test whether it’s coming from greater quantity or greater quality, and so I’m still trying to understand why there was this increasing expenditure. Although, from what it seems like, and I’ve mentioned this in great detail in my paper, there’s a recent pest that is of concern in the region. It’s called “Tuta Absoluta” or “Uzi” in the local language.

It has migrated from Latin America in the last four years and has been a pressing concern among these tomato farmers. I think when we presented information about the pest, about the integrated pest management strategies that should be adopted, I think a lot of farmers related to that and chose to diversify their pest management strategies, and not just buy pesticides for it.

RAJAGOPALAN: That, actually, I found to be one of your most interesting results because there’s a couple of things happening when it comes to pests, right? And in the case of the Uzi, when it comes to tomato farmers, basically if you are the only farmer who treats your crop, you still don’t protect against damage because your neighbors, et cetera, may not be treating for that particular pest and the pest may still be in the region. So the intervention and the additional spending makes sense only if everyone else is spending on it, right? Otherwise, this kind of an intervention actually doesn’t make too much sense.

There’s a classic collective action problem when it comes to battling any externality, including pests, and what I found most interesting about your research is the WhatsApp group dramatically reduces the cost of collectivizing and then jointly increasing the expenditure on battling a single pest in that particular region. In one sense, I know that the rough spending is about 10,000 rupees. I know that doesn’t explain the entire amount of spending, but it explains about a quarter of it because of this pest thing.

That I found to be quite interesting, that this may be a way of solving collective action problems and externalities that they all share, but that no one can solve it individually.

SHARMA: That’s a really interesting point. I actually hadn’t thought of this while I was writing the paper or designing the study because the region isn’t as densely populated as one would think. It’s very remote, roughly the villages are spread out around a radius of 30 kilometers. I’m not a scientist, so I’m not sure how close the farms need to be for the pests to migrate from one farm to the other.

What I thought, when I was having conversations with these farmers, it seems like they care a lot about reputation and social status, so if they shared tips on the group, they gained bragging rights that, “I have access to this knowledge that I’m sharing with other farmers, which makes me a more reputable farmer in that group,” and it was much less a competitive environment, as I had previously thought. It was very pleasant to see that farmers were willing to share tips, not just about Uzi, but also, there was a cyclone that hit the area during the study, and they openly shared tips on how to prevent crops from weather shocks. So, it could be a collective action problem, but it’s hard to test.

RAJAGOPALAN: I guess to me, that was the revelation in all of this because, at the end of the day, the farmers are competitors when it comes to selling their produce, especially something as specific as tomatoes from a particular region, but they also are vulnerable to very similar challenges, right?

Weather is one, and pests is another. It’s interesting that you find that there’s a lot of information sharing on cyclones and pests, which now, suddenly, start making sense because you don’t want that kind of an external shock which impacts everyone’s crops the same. Though you might be competitive and that you want a slightly better price for your own crop relative to others, right?

SHARMA: Yes, and that’s a great point, which makes me think of one reason why the expenditures could be higher, because, in those regions that were treated, there was a greater demand for these pesticides, so that could have also driven up—or demand-supply price is endogenous, so hard to—yes.

RAJAGOPALAN: I guess when the first set of peer learning research papers come out, it’s basically we’re just at a basic, nascent stage of just understanding, “What is it these guys are even talking about?” To that extent, I find it quite interesting that these are the questions that are discussed. Are there other kinds of coordination that you see taking place? Is there any discussion about coordinating on particular markets or particular intermediary buyers and so on, or coordination on something else that you might see?

Whether it’s government subsidies or credit, do they try and coordinate, if there was a local or a state election cycle during that time, on asking who people are voting for and so on? Do you see that kind of coordination going on?

SHARMA: That’s a really good point. 80% of the sample was growing tomatoes most of the year. Regarding their crop cycle, their individual crop cycle, like planting, sowing and harvesting was very coordinated in that it was very hard to get hold of farmers in that season because they would leave for the fields at 6:00 a.m. so we had to get to them, 30 kilometers away from our office, at 5:45 a.m. just to get hold of them, so I would say that part was in sync.

I don’t think there was coordination outside the agricultural realm because I don’t think the treatment effect was so drastic that it would impact their other social aspects as well, if that makes sense.

RAJAGOPALAN: Yes, and it’s also a pretty short period of time, right?

SHARMA: Exactly.

RAJAGOPALAN: We’re just talking about one of the three crop cycles that they’re doing in a given year, so it needs to be short, and maybe if you keep studying these groups over a longer period of time, you never know what you might see.

What are the implications of what you studied in terms of direct policy interventions when it comes to agriculture? I understand that there’s this broad-layer policy implication, which is making digital public goods available even in remote areas which can dramatically change the conversation, it can change how farmers are able to use technology and so on. Is there a more specific policy implication from what you’ve studied?

SHARMA: Bear with me for a thought experiment. Imagine there’s a farmer who has a pest infestation in his farm. He needs urgent help, but the nearest research institute or agriculture officer lives two hours away. Now, this person has to make a phone call if they have a good network. They have to make a phone call, wait for the agriculture officer to be available, then that person will travel a few hours by road to get to the farmer’s farm and then advise them what’s best to get.

Then the farmer would have to go to the nearest input dealer, which will also be far if this is a remote farm, and then it’ll waste more time and cause more damage to the crops. If this information can be communicated via video call or via these smartphone applications, it’s going to drastically save time and cost, both for the farmer and the agriculture officer. I believe that having this already existing infrastructure is just so precious, and I think that governments and research institutes should be aggressively pushing toward using this component to save time and cost when it comes to policy.

RAJAGOPALAN: On this, why is the in-person versus a video intervention or a WhatsApp group intervention different from a simple Google search? Because I imagine some of these questions can just be resolved on Google. Is it a question of misinformation or trusted information? What’s really going on?

SHARMA: A typical analogy that I usually give in my talks is, imagine a class of students that are trying to learn a new topic, and if they go to Google and they try to learn something versus if they talk to their friends or are working on it on a group project. The marginal benefit of learning from a peer exceeds that of learning from a random, unknown source, which is the main idea here. There’s repeated evidence and literature that farmers learn from each other in agriculture, and so just replicating that into the virtual dimension, it seems to have some hope when it comes to learning.

RAJAGOPALAN: The big question is one of conducing environments to learn. It’s not just the availability of information, the way I understand it?

SHARMA: Yes, exactly because especially now, Google is flooded with resources, and I should also add that a lot of the farmers in these regions are not comfortable reading or writing text. One thing that’s great about these WhatsApp groups is Voice Notes, so a lot of the text in the text messages was translated, transcribed Voice Notes that farmers used, which is also helpful to the farmers.

RAJAGOPALAN: What is the predominant language that they are texting each other in?

SHARMA: This was in Telugu because I was in Andhra Pradesh.

RAJAGOPALAN: Andhra Pradesh, yes. It’s basically Telugu or, even more specifically, the regional dialect in the area that you’re covering. Sometimes, there is text, and sometimes, there is Voice Notes.

SHARMA: Exactly, and that’s also one reason why it’s harder to Google all of this, because it’s in such a specific local language.

RAJAGOPALAN: How important is trust in this? One part is like you said, there’s just a vast amount of information, it’s going to take time to go through it. They’re not as proficient or fluent in reading, writing, research as maybe you and me, right? The second part of it is just trust: “If my neighbor’s farm is going through the same problem and they are willing to implement this, then maybe it’s good enough for me.” Is that what’s going on here?

SHARMA: I think so. I think farmers in India have a sense of community that I have not seen outside of India or South Asia in general. They believe that it’s them versus the problem as opposed to “you versus me,” so I think that greatly affects their responses.

RAJAGOPALAN: Yes. Do you see a difference in trust levels between farmers when it comes to another farmer, versus farmers when it comes to like a government officer or a research scientist at one of these government institutes, or does that also, depending on the region, kind of make them part of the in-group?

SHARMA: I think there’s some heterogeneity in that. Some farmers are very open to receiving advice from other farmers, but if they belong to, say, a higher social class, they would be a little more wary of consulting other farmers, and they would prefer to consult government volunteers or agriculture extension officers.

RAJAGOPALAN: Got it. What is more generally the impact of digital revolution in India? I know you’re studying this even outside of agriculture. You’re looking at, for instance, maternal health interventions. What are some other ways that the digital revolution can be used for things or areas that are not yet tapped?

SHARMA: I think that’s such a great question, and that’s the broader picture I want my research to be as I pursue my career. We’ve already seen, in COVID, how, in education, things took a turn where people switched from in-person classrooms to e-learning, which was also not entirely inclusive of those who couldn’t afford that technology.

When we’re thinking of other spheres like nutrition, health, I think one example could be how for young and pregnant mothers, there’s often these nutrition interventions that are organized where they invite them to a group setting, show them a video of what to eat, how to take care of themselves but, again, I think that takes up a lot of costs. Not just monetary, but also time.

If these mothers do have access to smartphones, it’s, I think, very cost effective to just share this information with them, and not always in text, keeping in mind the literacy rates, but also like infographics or just videos, or even audio notes on advice from trusted sources, of course. You can imagine this can translate into, for example, fact-checking for political information that’s being spread around. I think Sumitra Badrinathan [with Simon Chauchard] has a great paper on this. Similarly, in health and education. It’s a rabbit hole, so there’s so much work to be done in this area.

RAJAGOPALAN: What are some of the projects you would like to look at in the future that seem like exciting places to explore the digital revolution?

SHARMA: I am actually interested in how UPI payments are affecting women’s labor force participation and their savings habits because some women, traditionally, their issue is with storing cash at home, but if they have all digital transactions, it may affect the way they participate in the labor force. That’s one part I’m interested in. I’m interested in evaluating impacts of other digital information interventions, like nutrition and health, as I was talking about earlier. And with climate change being this urgent, I’m also interested to see how we can use this for, for example, sustainable agriculture practices.

RAJAGOPALAN: Most of it sounds to me like one part is specific interventions and technological interventions, if they work, but the other part of it is now, there is a low-cost way of forming different kinds of communities which didn’t exist before we had this digital revolution. Is that the right way of thinking about it? That there’s a way of forming overlapping communities that won’t naturally get together? Because India can be quite plural, so if we can get all the expectant mothers in one group, irrespective of caste and region and class and so on, then it’s going to be a slightly different kind of a community-based approach than what would traditionally be seen in the marketplace?

SHARMA: Yes, and I think there’s still two levels to it. One is a bilateral information exchange between this organization that’s providing the information and this community that’s receiving the information, and then there’s this multilateral information exchange where just this one community is having peer-to-peer interactions. If mothers are having questions about their infants or their feeding practices, they can just reach out to another mother with a similar background who can help them out and save time in the process, and costs.

RAJAGOPALAN: One more thing I really wanted to ask you about is this semester, you are the editor-in-chief for a blog that’s called Economics That Really Matters. This is run out of Cornell, which is where you’re based. Can you tell me a little bit more about the blog and, what do you think are economic questions that really matter?

SHARMA: Sure. I’m co-editor-in-chief with another grad student whose name is Tess [Lallemant]. She’s also at Cornell. Basically, we want to provide young researchers with a platform to share their ideas without the gatekeeping and publication bias that typically happens in academic journals. That’s a great opportunity for people in development economics who want a platform to share their ideas, and it’s been such a pleasure just reading new research. It’s honestly inspiring. We also have an interview series on the blog where we’re trying to interview experienced researchers so they can give advice to graduate students, which is also very helpful.

RAJAGOPALAN: Can you think of a recent example of a paper that was a great idea that you really enjoyed working on as editor?

SHARMA: I really like this piece by David Murphy, who’s an assistant professor of economics at Colgate University. He did an experiment in Kenya about farmers who consume alcohol and whether counseling in a group like Alcoholics Anonymous can help them toward remission, which I thought was very, very interesting. In it, he has treatment arms where there’s group therapy and couples counseling. I can’t remember the results off the top of my head, but it’s a fantastic read if you get a chance.

RAJAGOPALAN: Oh, definitely. Thanks so much for sharing that with us. This was such a pleasure, Vanisha. Thank you for being on the show and talking about your research.

SHARMA: Thanks, Shruti. Thanks for having me.