CAUSAL IMPACTS OF FINANCIAL MARKET REGULATION: EVIDENCE FROM CENTRAL CLEARING AND INFORMATION-THEORETIC ADVANCES IN DIFFERENCE-IN-DIFFERENCES

<p dir="ltr">This dissertation consists of two essays exploring the causal impacts of financial market regulations using causal inference methods. The first essay investigates the impact of mandatory central clearing on prices, volatility, and liquidity in the interest rate swaps market. Utilizing a classical difference-in-differences framework, it leverages regulatory variation across currency denominations to isolate causal effects of the clearing mandate. Empirical findings indicate a 12-13 bps increase in interest rate swap premia with the adoption of central clearing, consistent with the theoretical model. Market liquidity and volatility do not show any improvement with the adoption of clearing, contrary to theoretical models.</p><p dir="ltr">The second essay examines the assumptions underlying the classical difference-in-differences method employed in the first essay. It addresses the parallel trends assumption, proposing methodological innovations that relax this restrictive assumption. It proposes using generalized maximum entropy (GME) based approach to estimating selection probability into treatment. Then, using a weighting scheme proposed by Abadie (2005), the average treatment effect on the treated can be estimated. If heterogeneous treatment effects are of concern, the GME-based method can be extended to estimate the conditional average treatment effect. Using simulations, the second essay demonstrates that conventional difference-in-differences estimates are biased when the parallel trends assumption is violated. It shows that GME-based methods outperform in recovering the correct model parameters, especially for small data sets or highly collinear covariates. The method is applied to a real-world dataset where an estimate of the true treatment effect is available from a randomized control trial. The GME-based method can recover treatment effects closer to the best estimate than classical difference-in-differences.</p><p dir="ltr">Together, these essays enhance our understanding of financial market regulation effects and advance econometric methodologies, enabling more accurate causal inference in economic research.</p>