This dissertation consists of two chapters on the partial identification in empirical industrial organization.

Chapter 1: Game-theoretic entry models usually impose strong restrictions on the predetermined information structure of the game. This paper introduces a new method for the identification and inference of payoff parameters in static entry games, while being agnostic about the latent information structure. I examine a static entry game, assuming that all potential entrants are symmetric. I introduce a modification of Bergemann and Morris's (2016) solution concept, Bayes Correlated Equilibrium (BCE), which I refer to as Symmetric BCE, and provide a tractable and sharp characterization of the identified set of payoff parameters. I apply the method to driving schools, investigating the impact of the number of operating firms on the profitability of potential entrants. I conduct two counterfactual experiments to evaluate the effects on the number of operating firms: firstly, a simulation of market size reduction, and secondly, amplifying the market size effect by 30%. The empirical results indicate that the new, robust method still provides informative insights. Moreover, using the notion of Symmetric BCE, as opposed to BCE, reduces the computational burden, making identification and inference feasible even with a moderate number of players.

Chapter 2: This paper proposes a new approach to partial identification of the semi-parametric multinomial choices model in a panel data setting, where the analyst uses covariate data on a subset of choices. This multinomial choice model allows for an arbitrary joint distribution of choice specific unobservables, so IIA-like property is not assumed. I show that the within-group comparison proposed by Pakes and Porter(2024) can be modified to account for the observation structure. I show that the new within-group comparison leads to a set of conditional moment inequalities. My main finding shows that the set of conditional moment inequalities characterizes the sharp identified set of the index parameters. In Monte Carlo simulations, the finite sample performance is presented.