Information Frictions and Network Spillovers in Firm-to-Firm Linkages

This project studies the impact of trade fairs on reducing search costs that may limit firm-to-firm connections. To test the impact of attending trade fairs on firm-to-firm connections, we partner with FISA, the organizer of Aquasur 2024, the largest trade fair for the fishing industry in the southern hemisphere, which will be held in Puerto Montt, Chile. We will first solicit firms’ interest in participating in the fair via an online application to a raffle to win a subsidy to attend the fair. Among the firms that express interest, we will randomly subsidize the registration fee to encourage firms’ actual attendance at the fair. Finally, we will organize a matchmaking event at the fair among participating firms to facilitate supplier-buyer connections.

2024-03-19

2024-03-21

To measure these outcomes, we will use comprehensive Value Added Tax (VAT) administrative data provided to the research team by the Central Bank of Chile. The data are collected electronically and provide rich transaction-level firm-to-firm information, including existing and new business connections, specific products and services traded, quantities, and prices. We will also measure the impacts on total sales, employment, profit, etc.

We will complement the administrative data with original firm surveys conducted during and after the trade fair to measure – on a buyer-supplier pair level – new introductions made and specifics discussed (e.g., samples shared, initial contracts signed, etc).

Our main outcome variables are whether attendance at the fair, as instrumented by receipt of the subsidy, leads to an initiation and strengthening the trade linkages. We measure these outcomes both for the extensive margin (presence of trade linkages) and intensive margin (quantity, products, prices). These outcome variables are measured at supplier-buyer pair level.

To measure these outcomes, we will use comprehensive Value Added Tax (VAT) administrative data provided to the research team by the Central Bank of Chile. We will measure the impact of the experiment on the number of new buyers or suppliers, on the extensive margin, and the quantity sold or purchased, product sold, and prices, on the intensive margin. These data will provide us with detailed and accurate measures of whether transactions took place between firms that previously did not have commercial transactions. We will also supplement this data with other firm-level administrative data, which records firm-level sales, employment, profit, etc.

We will complement the administrative data with original firm surveys conducted during and after the trade fair to measure – on a buyer-supplier pair level – new introductions made and specifics discussed (e.g., samples shared, initial contracts signed, etc.). Specifically, we will conduct the following set of surveys:

1. Daily stand observation: This is a survey to be run continuously, over a 30-minute block, each day with each stand from our supplier experiment sample that is present at the fair, along with a random sample of a small number of other stands at the fair. Upon receiving consent from both the representatives and the visitors, enumerators will ask questions about the interaction between the two parties and ask several follow-up questions to both parties at the end of their interactions. Enumerators will also take a picture of the stand at the beginning of the fair, showing the setup of the stand (including banner, poster, and marketing materials) but without any person in the photo.

2. Daily stand survey: On each day of the fair, we will conduct a 30-minute daily stand survey with all suppliers participating in the experiment who are present at the fair. With this survey, we aim to record the firm’s activity during the fair. In addition to the experimental sample of suppliers, we will also survey the randomly sampled a small number of stands to administer the daily stand survey. This will inform us whether the activities of suppliers participating in the experiment are different from general suppliers at the fair.

3. Visitor survey: we will conduct a brief (approximately 5-10 minutes) survey with visitors at the fair to get some basic information about their activities and their purpose of attending the trade fair. We will administer the visitor survey with a random sample of visitors every day at the fair, randomly sampled at the main exit out of the fair.

4. A month after the event, we will conduct a phone survey with all firms we surveyed at the fair (including 1-3 above). The purpose of this survey is to (i) for those who attended the fair, record the progress a firm achieved with the contacts established during the fair and (ii) get more information on overall relationships between the firm and its suppliers and buyers. This survey will be conducted by phone by enumerators from a local survey agency.

5. A year after the event, we will conduct a survey with all firms in our sample The purpose of the survey is similar to the one-month follow-up survey above, i.e., (i) record the progress a firm achieved with the contacts established during the fair (for treated firms) and (ii) get more information on overall relationships between the firm and its suppliers and buyers.

Our secondary outcome variables are as follows. We measure these outcomes through VAT data and our original surveys as outlined in the “Primary Outcomes” section.

1. Supplier-level aggregate outcomes, such as sales, employment, and profit.

2. Intermediate steps to initiate new trade linkages, as measured in surveys, including business meetings, references checked, samples sent, etc. We rely on surveys for these outcomes.

3. For the sub-experiment of the matchmaking program, we will assess the impacts on transactions between the matched pairs, both extensive margin (presence of trade linkages) and intensive margin (quantity, products, prices). Furthermore, if we detect an increase in new supplier-buyer linkages, we assess whether the new supplier linkages substitute from other suppliers of the same product categories, thereby testing for any business-stealing effects.

4. Firm-level technology upgrading, as measured in surveys as well as from the purchases of capital goods and new goods and services from VAT data.

5. Upstream and downstream propagation effects, such as the impacts on suppliers of suppliers and buyers of buyers, as measured in the VAT data.

6. Ability to access international markets (importing and exporting).

Our main experiment is to randomly provide a subsidy to encourage suppliers’ attendance at the fair. We recruit small and medium sized enterprises (SMEs) who are potentially interested in attending the trade fair and would like to bring their products and services to showcase at the fair. We will randomly provide a subset of the suppliers with a 60% subsidy on their registration fee to encourage actual attendance at the fair. Using the randomized subsidy as an “encouragement design,” we will study the impact of attending the fair on business connections and firm outcomes.

We will also run a sub-experiment of “matchmaking program” at the fair to randomly reduce information friction between particular pairs of suppliers and buyers. Among the SME suppliers who decide to attend fair (as well as among a wider set of suppliers at the fair) and potential buyers at the fair (i.e., firms who sign up as visitors to the fair and will be primarily looking for suppliers at the fair), we will elicit their interest in participating in a matchmaking program at the fair. Among those who express interest, we will randomly form matched pairs of potential suppliers and buyers, based on the product information provided by the firms, and invite the matched pairs to attend a series of matchmaking meetings happening at the fair. Leveraging variation in the matching at the buyer-product level, we will assess whether the new supplier linkages substitute transactions from other suppliers, thereby testing the business-stealing effects.

Not available

For our main experiment to encourage suppliers to attend at the fair: Among the firms who fill out an online application to join the raffle, we first select small and medium-sized firms based on their employment size. We then stratify these firms based on their sales value and number of employees. We run the randomization in office by a STATA program. We will then inform the winner of the randomization and confirm their participation to the fair.

For our sub-experiment of matchmaking program: Once we have the list of buyer firm representatives and products they are in charge, we will randomly split them into “treatment,” who we invite to the program, and “control,” who we do not, at the level of buyer-product. For each treated buyer-product, we randomly assign suppliers that serve the corresponding product. The matchmaking program will consist of a series of one-on-one meetings between these matched pairs. Therefore, the randomization occurs at the level of the buyer-product and supplier level. We will then run the randomization in office by a STATA program.

For our main experiment to encourage suppliers to attend at the fair: We randomize at the supplier firm level to select the winner of the raffle.

For our sub-experiment of matchmaking program: We implement the randomization in two steps. We first select the treatment buyer-product. Next, we randomly allocate suppliers that serve the corresponding product (as long as there are multiple suppliers within each category). Therefore, the randomization occurs at the level of the buyer-product and supplier level.

No

For our main experiment to encourage suppliers to attend at the fair: We expect to have about 220 SME suppliers in total, which constitutes our clusters.

For our sub-experiment of matchmaking program: We expect to have about 30-100 buyer-product pairs and 65-130 suppliers. For the outcome variables defined at the buyer-product and supplier pairs (e.g., whether a new transaction is initiated between the matched pair), we cluster our analysis at the combination of buyer-product and supplier level. For the outcome variables that are defined at the buyer-pair (e.g., the presence of a transaction with the preexisting supplier of the buyer firm within the same product category), we cluster our analysis at the buyer-product level.

For our main experiment to encourage suppliers to attend at the fair: We expect to draw our treatment suppliers from about 220 SME suppliers in total. For outcome variables that are defined at the supplier-buyer-pair level (e.g., whether there is a new trade relationship, volume of transactions, prices), our observation is 220 SME suppliers times the number of potential buyers that appear in the trade fair (around 20 major salmon firms, plus some small-scale buyers). In this case, we cluster standard errors at the supplier-level. For secondary outcome variables that are defined at the supplier level (e.g., sales, employment, profit), the unit of the observation corresponds to the number of suppliers participated in the randomization (about 220). For our sub-experiment of matchmaking program: We expect to assign treatment matches from about 30-100 buyer-product and 65-130 suppliers. For the outcome variables defined at the buyer-product and supplier pairs (e.g., whether a new transaction is initiated between the matched pair), the unit of observations is the potential number of supplier-buyer matches. For outcome variables defined at the buyer-product-level (e.g., purchases of these products from new suppliers and preexisting suppliers), the unit of the observation corresponds to the number of buyer-product combination, which we expect to be about 30-100 as we mentioned above.

We expect to recruit 220 SME suppliers to participate in the study, among which 110 will be randomly selected to receive the financial subsidy to attend the fair and we expect 65 of them to eventually take up the subsidy and attend the fair. The sub-experiment of matchmaking program will involve approximately 65-130 suppliers (those from the treatment group who take up the subsidy and some other suppliers who sign up on their own at full cost) and approximately 30-100 buyer-products who register (on their own) to attend the fair and express interest in the program.

We conduct power calculations for the probability that a supplier makes a connection to a major buyer of this industry, i.e., 17 major salmon producers. We assume that the baseline probability that a supplier has with a buyer is 0.001, with 0.017 of the inter-cluster correlation coefficient (ICC) among suppliers, based on our administrative VAT data. We assume a 60% take-up rate of our main experiment to encourage suppliers to attend the fair. Under these assumptions, we have the power to detect the treatment effect for the probability of forming a link by 0.0058 with 190 samples (95 treatment and 95 control suppliers). This treatment effect is equivalent to 0.1 (=0.0058 times 17) new buyer connections per supplier by attending the fair.