This is the eighth in this year’s series of posts by PhD students on the job market
Search and trust frictions have historically made it hard for small firms in lower-income countries to buy inputs from foreign markets. Consider a clothing wholesaler in Senegal, who wants to start selling high quality European-made jeans. First, they must find a supplier, most of whom are in Europe, and see what this supplier sells. Second, even if they manage this, typically they must send money in advance and hope that the goods will arrive. Yet, the huge recent growth in smartphone ownership and social media usage for buying and selling goods may alleviate these barriers. In my job market paper, co-authored with Deivy Houeix, we run a field experiment in Senegal to answer two research questions:
1. To what extent do search and trust frictions limit small firms' access to foreign goods?
2. Can the growth of smartphone ownership and social media overcome these frictions?
Setting: Turkish-made Garments in Dakar
The study focuses on Turkish-made garments in Dakar. Turkey is the second-largest source of ready-to-wear garments in Dakar (after China). Turkish-made goods also command a large quality premium: we conducted a household survey and found that consumers were willing to pay 34% more for Turkish-made goods than Chinese-made goods.
The main subjects are 1,862 small firms in Dakar. These firms—comprising both retailers and wholesalers—are in most respects typical of small, informal, owner-operated businesses in large cities in lower-income countries. Almost all firms already sell Turkish-made goods, with most buying them from local wholesalers. The median firm has four regular suppliers with whom they have transacted repeatedly, typically either by periodically visiting and buying on the spot (if local) or by ordering (if abroad). Most sales are from a stock that firms already have or can easily obtain, but firms also occasionally do custom orders if clients are looking for something specific. The other subjects are 30 suppliers, based in areas of Istanbul known as garment export hubs to West Africa. We focused on suppliers of Senegalese nationality as our pilot surveys suggested that this is a large share of the market.
Descriptive statistics: firms use social media extensively to buy/sell goods and share information
Social media has the potential to alleviate search and trust frictions in international trade.
For search, we document the ubiquitous use of “Supplier WhatsApp Groups”, which are groups that suppliers use to send images and videos of goods to around 50-100 clients. At baseline, 86% of firms are in at least one such group and the median firm is in four, with the majority used for domestic trade. These groups potentially alleviate search frictions by helping firms see what a supplier abroad sells, but do not eliminate them as it is still challenging to find the supplier—and thus join the group—in the first place. Moreover, garments are highly differentiated, so a firm might need to join many groups just to find one that sells relevant goods.
For trust, social media makes it easier for firms to share information and coordinate action with each other to discipline suppliers that cheat. For example, at baseline, one-quarter of firms are in WhatsApp groups with other firms for the purpose of sharing business information, and two-thirds of firms have recommended or warned against particular suppliers to other firms in the past year. However, since these networks are highly fragmented, it is unclear whether the information they provide is relevant to the average firm.
Thus, whether social media really does alleviate search and trust frictions is ultimately an empirical question.
Treatments
The experiment comprises three treatments, designed to both be of real-world policy interest and map onto theoretical frictions.
1. In the Search treatment, we add treated firms to the Supplier WhatsApp groups (created for this study) of three different suppliers in Turkey.
Among those in the Search treatment, we then cross-randomize two treatments to improve trust.
2. In the Adverse Selection treatment, we aim to facilitate information sharing and learning. We add treated firms to a fourth “discussion” WhatsApp group that only contains other firms matched with the same suppliers. The purpose is for the firms to share information about the suppliers. We seed this group with a positive review based on real orders that we commissioned before the study started.
3. In the Moral Hazard treatment, we aim to provide stronger incentives to the suppliers. We inform treated firms that we will ask them to rate the suppliers, and that any supplier receiving consistently negative feedback will be removed from the study, thereby losing 150-200 potential clients. We emphasize that the suppliers know this and that they therefore have strong incentives to exert effort.
Result 1: Connecting firms to suppliers in Turkey increases access to foreign goods
To measure access to foreign goods, we designed a mystery shopping exercise, in which trained surveyors act as real customers and attempt to buy goods from all firms. We measure both a horizontal dimension (access to a wide set of varieties) and a vertical dimension (access to high quality varieties). For the horizontal dimension, we measure whether the firm can deliver (within two weeks) a variety that is “close to” what the shopper is looking for. For the vertical dimension, we buy the good and have two tailors measure quality using a detailed scorecard.
Figure 1 shows the results. The treatments increase both measures. Pooling the treatments together, treated firms are 9.5 percentage points more likely to have a given variety (a 26% increase from the control mean of 36%) and the good they sell us is 13.1 percentage points more likely to be high quality (a 30% increase from the control mean of 43%). When we decompose the treatments, there is no evidence that the trust treatments do more than Search Only, showing that connecting firms to new suppliers abroad is sufficient to meaningfully improve their access to foreign goods.
Figure 1: Effects on access to foreign markets
Result 2: Alleviating trust frictions is necessary to build lasting relationships
Do these firms develop the new connections into lasting relationships? We measure this using transaction-level data from the largest mobile payments company in Senegal, and present the results in Figure 2. The graph shows cumulative money flows from study firms to study suppliers. Initially, the trust groups are no more likely to order from study suppliers than Search Only. However, this changes in the long-run: the Search Only group largely stop ordering, while the trust groups continue at a similar rate. Since a secondary goal of the mystery shopping was to (indirectly) subsidise firms to experiment with the new suppliers, we test whether this post-study difference in orders between the trust groups and the Search Only group is statistically significant. It is. We conclude that, while alleviating the search friction was sufficient to increase access, alleviating the trust friction simultaneously was essential to generate long-run effects.
Figure 2: Cumulative order value from study suppliers. A test that the orange arrow is equal in length to the black arrow has p < 0.05.
We also conducted a survey after 3 months. We find increases in the likelihood of having a supplier in Turkey in the order of 4-8 percentage points (relative to a control mean of 17%), largest in the group with both trust treatments. Further evidence suggests that these firms have substituted away from local wholesalers. We also ask summary questions on monthly profits and sales. The coefficients are large and significant, with pooled increases of around 20%-25%, driven by the group with both trust treatments, and particularly among wholesalers in the upper tail. The gains therefore flow through to profits and sales.
Conclusion
We draw two main conclusions. First, search and trust frictions meaningfully limit firms’ access to high quality foreign goods. We connected them to new suppliers abroad, and they were more likely to have a given variety and it was more likely to be high quality. However, alleviating both frictions simultaneously was essential for longer-term impact. Second, social media can be used to alleviate these frictions. Ten years ago, almost all of these firms would likely have accessed Turkish-made goods either by buying from a local wholesaler or travelling. Today, a meaningful share of firms can import directly.
Edward Wiles is a PhD candidate at the Massachusetts Institute of Technology.
Join the Conversation