This is the first in this year’s series of posts by PhD students on the job market.
Anyone who has ever visited a low-income country will be familiar with the sight of seemingly identical firms closely agglomerated within cities: a road entirely occupied by furniture makers, a building packed with tailors, or a corner of a market filled with vendors uniquely selling pans and pots. What drives this behavior?
Consider the following example: a woman in Kampala, Uganda, the setting of this study, is looking to buy a dress. She does not know what varieties of dresses are available in the market and the only way for her to find out is to physically visit the tailors that produce them. She has two options: she can go to a tailor next to her house, or she can travel a bit further to a market that has several tailors selling different varieties and qualities of dresses. If the market is not too far, she will probably prefer the second option, as it offers her higher chances to find a product that exactly matches her tastes.
Now, consider a tailor that must decide whether to place her business in that market, or rent a small shop that has no other tailors around. On the one hand, she will be tempted to set up her shop in isolation and monopolize the customers for whom the market is too far. On the other hand, by doing so, she faces the risk that very few customers will visit her shop. In the market, she will have to compete for customers with more firms, but the number of customers to compete for – the size of the pie – will be larger to begin with.
In my job-market paper, I study how consumers’ imperfect information about the variety of products sold in the market affects firms’ location decisions and performance, using garment firms in Kampala as a case study. Quantifying the extent to which agglomeration is driven by demand is important for three reasons. First, the traditional view is that, by enhancing firm competition and creating scope for agglomeration economies, spatial concentration is always welfare improving. I show that, when driven by information frictions, agglomeration can soften competition and drive prices upwards. Second, understanding how consumers search for products within low-income cities can shed light on the demand-side constraints that prevent high-productivity, high-quality firms from attracting customers and expanding, contributing to large resource misallocation and overall low productivity in low-income countries. Third, with the population in low-income cities projected to grow by 75% by 2050, having a good understanding of the key drivers of firms’ location decisions is crucial to evaluate the welfare consequences of urban policies.
How does consumers’ search behavior affect firms’ choice of location?
In Kampala, 40% of all establishments operate within 2 km from the central district – the core – with firms typically clustering next to businesses in the same sector (Figure 1). To shed light on the drivers of agglomeration, I collect data from 600 garment firms and their customers across the city. The garment sector accounts for 42% of all Ugandan manufacturing establishments. Firms in this sector sell horizontally and vertically differentiated products, making information frictions more likely to emerge than in sectors where products are homogeneous. Firms were randomly sampled from an initial census of more than 2,400 establishments across 14 parishes in Kampala with varying levels of firm density. Customers were randomly selected from transaction records that owners kept as part of the data collection exercise. The collected dataset combines transaction data - which allows estimating demand, customer data - which provides evidence on consumer search, and a mystery shoppers exercise - which yields accurate measures of product quality. I use the data to establish a set of empirical patterns that are consistent with the hypothesis that consumers’ search behavior is a key driver of firms’ location choice.
Consumers report preferring to search for products in the core because of the large number of firms and varieties sold there. Since most customers live in residential areas outside the city center, they pay transport costs to the core that are three times higher than the cost of travelling to a firm in the periphery. However, once in the core, consumers visit 22% more firms prior to purchasing, in line with the cost of gathering information about products being lower in locations that have a higher density of firms.
73% of firms report finding customers to be a primary constraint to setting up a business. This is by far the most common constraint, followed by access to finance, mentioned by 53% of firms. Access to customers is also the main reason why owners locate their business in the core rather than a more peripheral area. On average, the monthly revenues of firms in the core are 78% higher than in the periphery. Higher revenues are not driven by firms in the core having more customers, but by them serving larger customers, namely customers that buy products in bulk. With respect to businesses in the periphery though, firms in the core pay higher commuting and rental costs, which disincentivize them from agglomerating.
An equilibrium model of consumer search and firm location
I build a model that accounts for these patterns by incorporating heterogeneous consumers (small vs. bulk buyers), information frictions and transport costs in a static discrete choice model of demand. A key feature of the model is that consumers, who have heterogeneous tastes, only observe their preferences over varieties upon visiting firms. For instance, they may know that they want to buy a specific item (a skirt, a dress, a shirt etc.), but may be unsure about the color, the material or the style that they prefer until they visit the firm and see products in person. To do so, they must pay a transport cost to the area where the firm is located. This is the source of information frictions. Once in a location, however, the marginal cost of visiting a seller is decreasing in firm density.
Firms offer goods that are of different quality and variety. They choose location simultaneously and, once in a location, they compete in a Nash-Bertrand pricing game. An increase in the number of businesses operating in a location impacts firms’ profits and choice of location in two opposite ways: (i) it attracts a larger pool of customers to the location, thus softening competition and incentivizing agglomeration; (ii) it increases the number of competitors with whom the pool of customers is shared, thus incentivizing dispersion. The trade-off between these two forces is heterogeneous for high and low-quality firms, with the latter benefitting disproportionately from locating in areas with a high concentration of firms.
I quantify the relative strength of agglomeration vs. competition by structurally estimating the model. Intuitively, customers that buy large quantities of products benefit the most from searching in high-density locations, because the gains from finding a better match are accumulated over all the units that they buy. The agglomeration/competition trade-off is therefore identified from variation in the share of small and large buyers purchasing goods in locations that have the same distribution of product qualities and prices, but vary in terms of number of firms.
How much do information frictions affect firm location and profits?
I use the model to analyze the impact of removing information frictions on firm location, profits and consumer welfare. Information frictions contribute to 8.2% of the observed agglomeration in the core. If customers were able to observe all product varieties at no cost prior to purchasing, average prices and profits would decrease by 14% and 18% respectively. These averages mask substantial heterogeneity across high and low-quality firms. High-quality businesses would gain considerable market share and experience a 17% increase in profits if information frictions were removed (Figure 2A). By contrast, at the new equilibrium, 37% of low-quality businesses would make losses and be better off exiting the market. As a result of lower prices and access to a larger number of varieties, consumer welfare would increase by 11%.
E-commerce and decongestion policies: what can we learn from the model?
I use estimates from the model to consider two sets of counterfactual policies: (i) the introduction of an e-commerce platform, and (ii) urban policies aimed at decongesting Kampala city center. In the e-commerce counterfactual, I assume that customers can observe all product varieties prior to purchasing and pay a flat fee to get products delivered to their location. This second aspect eliminates the geographical element of consumer search. With respect to the baseline scenario, the platform leads to a 39% reduction in the number of firms operating in the core. By eliminating information frictions, the policy causes customers to shift to high-quality businesses, inducing a 27% increase in their profits.
In Kampala, travel time is estimated to be 13.5% of the city GDP plus an additional 4.2% considering congestion. This has led Ugandan authorities to consider evicting informal businesses to decongest the central part of the city. I show that policies that relocate firms in space without addressing information frictions can backfire. Firm profits unambiguously decline as caps are imposed, but high-quality firms experience the largest losses (Figure 2B). This is because a higher spatial dispersion means that customers get to observe a smaller number of firms within the same location, making it harder for consumers to find the good firms.
Key take-aways
There are three key take-aways from this paper. First, demand externalities that arise from information frictions contribute to a substantial share of the observed firm agglomeration. Second, by preventing consumers from comparing products available in the market, information frictions limit the ability of high-quality firms to attract customers, thus favoring the survival of low-quality competitors. Third, urban policies that discourage agglomeration without addressing information frictions increase consumers’ search costs, disproportionately harming high-quality firms.
Anna Vitali is a PhD student at University College London.
Join the Conversation