In this thesis, our aim is to study a Cournot tatonnement system which exhibits destabilisation of the Cournot equilibrium as the number of firms increase. Our approach is to first consider the special case of firms behaving identically in a market share attraction model in two different adjustment process: Cournot tatonnement and bounded rationality adjustment. Results from the Cournot tatonnement system shows a superstable equilibrium in two firms model and an unstable equilibrium in a five firms model. In the five firms model, we show that introducing heterogeneity stabilises the Cournot equilibrium. For both two and five firms model, the differences of costs between firms are critical for the convergence of the system to the Cournot equilibrium. Lastly, we study the effect of entries and exits of firms on the number of active firms in the market. We discover that the market can sustain between two to four firms, and the factors are differences of costs and initial outputs between firms, and barriers to entries.

This dissertation consists of three essays studying the dynamic and strategic interactions of oligopolistic firms in developing countries using dynamic structural models. We focus on industries which have positive social benefits for the communities in which they operate, namely radio and banking. We use the estimated models to simulate counterfactual policies aimed at improving access for individuals in underserved areas. The first essay studies the social impacts of the liberalization of the radio broadcasting sector in Ghana. I analyze how the regulator affects commercial stations' decisions to enter and the resulting effects of coverage spillovers in rural areas. I exploit random variation in radio coverage caused by coverage spilling through gaps in mountainous areas and use this to estimate the effects of coverage on malaria incidence and development. I then estimate a dynamic structural entry model for commercial stations. In counterfactual simulations, I find that the allowance of more powerful transmitters is particularly effective in delivering the social benefits of radio to new communities. The second essay (joint with Calixte Ahokossi) studies voter turnout and regulatory inefficiency in the radio broadcasting market in Benin. We find an inverted-U relationship between the number of radio stations and voter turnout. We estimate a dynamic structural model for radio stations, taking into account the regulatory inefficiency in the market. Counterfactual simulations suggest that either removing the regulatory inefficiency or introducing targeted entry subsidies can spur entry in areas without radio stations, which would increase voter turnout in these areas. The third essay (joint with Marc Rysman and Robert M. Townsend) studies the banking sector in Thailand. Here, we argue that the effect of financial crises on bank branch location choices provides an unexplored channel by which crises affect access to credit. We estimate a dynamic structural model of oligopolistic location choice, allowing for complementarity in payoffs for bank branches in nearby locations, as well as competitive effects between rival banks. Using this model, we can predict counterfactual expansions of the bank branch network under policies focused on opening rural bank branches, or in the absence of the 1997 financial crisis.

In many markets, efficiency depends on the quality of information that participants have. However, participants may face frictions in accessing information, which could result in significant welfare losses. My dissertation studies the causes and consequences of information frictions, focusing on the security analyst market and the public cloud market. The first two chapters investigate how information frictions are generated in the security analyst market. Security analysts observe signals and compete to make forecasts on securities’ earnings, which serve as public information to investors. Here I study how analysts’ incentives affect the quality of information they provide. In Chapter 1, I consider security analysts’ incentives as a whole and estimate them using revealed preference. Security analysts are rewarded for being more accurate than their peers. This reward for relative accuracy leads analysts to distort their forecasts to differentiate themselves, but also disciplines them from being overoptimistic. I structurally estimate a contest model with incomplete information to capture both effects, disentangling the payoffs for relative accuracy, optimism and absolute accuracy. Using the model, I conduct counterfactuals to evaluate policies that reduce analysts’ payoff for relative accuracy. I simulate the effect of these policies on the quality of information in terms of forecast errors and variances. The reward for relative accuracy reduces errors by 33 - 58%, but increases variances by 4%. It is optimal to have moderate competition between the covering analysts of each security. In Chapter 2, I ask where these incentives come from. Are analysts motivated by dynamic incentives of reputation, or by short-term compensation such as bonuses? I show with reduced form evidence that low-reputation analysts may face more incentive to outperform their rivals than high reputation analysts. Building on this, I develop and estimate a dynamic model where analysts compete to build reputation and earn compensation. I find that analysts face a strong reputation-building incentive because high reputation is associated with a much higher fixed wage. Meanwhile, their forecasts have an insignificant impact on their immediate compensation. Chapter 3 studies the consequences of information frictions in the public cloud market. Firms need information about available technologies to make good adoption decisions. Inattentiveness to such information may create stickiness to outdated technology. In a joint project with Sida Peng and Peichun Wang, we study the welfare benefits of firms’ public cloud adoption and the consequence of consumer inertia in this market. We develop a novel demand model that allows for both multiple product choices and continuous quantities on each product. We estimate the model using a proprietary dataset on individual firms’ cloud usage history from a major public cloud provider. The estimated average return on investment in cloud is 2.2 times the cost of investment, which is driven by smaller firms disproportionately benefiting from access to computing resources on the cloud. On the other hand, inertia on the cloud leads to sub-optimal product choices for all firms and reduces welfare from cloud usage by almost 62%. We show that introductory discounts incentivizing firms to try new products can improve both consumer welfare and provider revenue.