This paper introduces a new method that integrates machine learning techniques with traditional econometric tools to correct bias in two-step estimation for dynamic discrete choice models under incomplete information. Unlike most existing approaches that primarily address bias from conditional choice probabilities while taking the transition matrix as given, this method simultaneously addresses both sources of bias. I propose a two-step estimator that jointly includes the conditional choice probabilities and transition matrix as first-step estimates, providing a robust, non-iterative solution. The estimator achieves square root N consistency and asymptotic efficiency. In addition, I establish a simple connection between Neyman orthogonality and the Zero-Jacobian property, a key result that guarantees consistent and efficient estimators in single-agent models. Monte Carlo simulations using a common benchmark data-generating process demonstrate the good performance of the proposed method.

Judge-IV designs increasingly invoke average monotonicity when strong monotonicity is hard to defend. We show that average monotonicity does not preserve a stable target population: the response types receiving positive weight switch discretely when a judge’s treatment propensity crosses the assignment-weighted average, so when the switching types are present in the population and have distinct average treatment effects, the estimand need not be locally invariant. We propose a diagnostic that measures each judge’s standardized distance to this boundary. In the Philadelphia bail data, three of eight judges, handling about 29 percent of cases, lie near the boundary. Target-population stability is therefore an empirical object once average monotonicity is invoked.

What determines where a fintech entrant builds its agent network, and does incumbent deterrence bind that reach? I study Wave’s entry into Senegal’s mobile money market against three telecom incumbents — Orange, Free, and Expresso — using a retail audit of 1,550 stores and a static entry game with moment inequalities (Ciliberto and Tamer 2009), leaving equilibrium selection unrestricted. The audit shows that Wave reaches 13.8 percent of stores overall but 75.9 percent of dedicated transfer outlets and 54.5 percent of stores with smartphone-owning operators; coverage concentrates where agents already handle transfers and own smartphones, not where incumbents are absent. Estimates imply that transfer outlets raise Wave’s payoff by 1.69 and smart- phone readiness has an identified set of [1.49, 1.85]. Fintech entry lowers Free’s payoff (δ ∈ [−0.49, −0.01]) but is not identified away from zero for Expresso ([−0.75, 0.16]). Universal smartphone readiness would nearly triple Wave’s coverage with negligible incumbent displacement. The binding constraint on fintech diffusion is not strategic deterrence, but agents’ digital capacity.

Observed action profiles in discrete games condition on structural errors known to players, contaminating best-response moment inequalities. Semiparametric matching cannot remove this selection for equilibrium inequalities: the payoff index that identifies the parameter is the same threshold that determines selection, so matching finely enough to equalize selection also removes the variation needed for identification. Two positive results follow. First, we show that Ahn--Powell differencing survives for post-entry outcome equations once the scalar propensity score gives way to the vector of players' choice propensities. Second, we derive a Hardy--Littlewood envelope that bounds the selection term using only a known marginal shock distribution, yielding valid moment inequalities under arbitrary dependence and equilibrium selection. Simulations and Senegal fintech data show that naive moments can produce an empty identified set, whereas the proposed bounds yield a nonempty and economically interpretable set.

We leverage a convex programming strategy and propose a new approach to estimating binary choice models that is robust to standard parametric specification assumptions.