Andrea Mercatanti is professor at the Department of Statistical Sciences, Sapienza University of Rome. He was our guest at the D2 Seminar series, presenting his work on “A Regression Discontinuity Design for ordinal running variables: evaluating Central Bank purchases of corporate bonds. (Joint work with F. Li, T. Makinen, A. Silvestrini).
This is the abstract his talk: Regression discontinuity (RD) is a widely used quasi-experimental design for causal inference. In the standard RD, the assignment to treatment is determined by a continuous pretreatment variable (i.e., running variable) falling above or below a pre-fixed threshold. Recent applications increasingly feature ordered categorical or ordinal running variables, which pose challenges to RD estimation due to the lack of a meaningful measure of distance. We proposes an RD approach for ordinal running variables under the local randomization framework. The proposal first estimates an ordered probit model for the ordinal running variable. The estimated probability of being assigned to treatment is then adopted as a latent continuous running variable and used to identify a covariate-balanced subsample around the threshold. Assuming local unconfoundedness of the treatment in the subsample, an estimate of the effect of the program is obtained by employing a weighted estimator of the average treatment effect. Two weighting estimators—overlap weights and ATT weights—as well as their augmented versions are considered. We apply the method to evaluate the causal effects of the corporate sector purchase programme (CSPP) of the European Central Bank, which involves large-scale purchases of securities issued by corporations in the euro area. We find a statistically significant and negative effect of the CSPP on corporate bond spreads at issuance.
The Regression Discontinuity Design (RDD) is a methodology for causal inference based on the potential outcome approach. It allows for evaluating the impact of policies designed to have a desired effect on a particular outcome through incentives. RDD is applicable when clear and transparent rules, rather than discretionary decisions by administrators, are used to allocate these incentives. Eligibility for receiving the incentives is determined by an observed variable measured before the administration of the incentives, commonly known as the running variable. Under RDD, units with a realized value of the running variable on one side of a pre-determined threshold are assigned to one policy regime, while units on the other side are assigned to the other policy regime. The key idea behind RDD is to compare units with similar values of the running variable but different policy regimes to draw causal inferences about the effect of the policy at or around the threshold.
The research question of our study centers around the evaluation of an unconventional monetary policy measure, known as the CSPP (Corporate Sector Purchase Programme), implemented by the European Central Bank (ECB) from June 2016 to December 2018. As a part of the ECB’s expanded asset purchase programme (APP), the CSPP aimed at strengthening the pass-through of the Eurosystem asset purchases to the financing conditions of the real economy, in pursuit of the ECB’s price stability objective. This program involved the purchase of corporate bonds in both the primary and secondary markets. Our focus is on assessing the impact of the CSPP on bond spreads at issuance, which refers to the yield to maturity of a corporate bond relative to the yield to maturity of a government bond with a similar maturity. Specifically, we concentrate on the primary market due to the relatively low liquidity of the secondary corporate bond markets in Europe, which can make secondary market quotes less reliable indicators of prevailing prices. In contrast, primary market prices offer accurate information regarding the market valuation of bonds at the time of their issuance. Conducting this evaluation provides insights that can contribute to the improved calibration of future monetary policy plans.
What are the major challenges in addressing this kind of study?
The main challenge in this type of study stems from the non-continuous nature of the running variable, which, in our case, is the bond rating. While there is a significant body of literature on RDD analysis with continuous running variables, fewer contributions have addressed categorical running variables. In particular, our running variable, the bond rating (eligibility for purchases is indeed limited to those bonds with at least a BBB- rating), is an ordered categorical variable. To address this challenge and identify the effect of the CSPP, we propose a solution that involves replacing the rating with a surrogate continuous running variable: the propensity score. The latter in our study represents the probability of a bond being eligible for purchase conditionally on the characteristics of both the bond itself and of the issuing company. By utilizing a two-stage procedure that considers the design phase and the analysis phase, we show how a common econometric estimator, such as the M-estimator, can identify the effect of the CSPP for a set of bonds with ratings around the eligibility threshold. Our approach then aligns with the research direction focused on enhancing external validity compared to a standard RDD, which only identifies effects at the threshold value.
Would you like to give us a glimps of future outcomes of your research, and possible impacts of your work in particular concerning the European Central Bank policies?