Boston University
Regression discontinuity and instrumental variables revisited: evidence on the effects of credit card loyalty programs on credit card expenditures in Peru
Abstract
dc:description.abstractThis thesis revisits the methods of Fuzzy Regression Discontinuity (FRD) Design and Instrumental Variables (IV) for estimating Local Average Treatment Effects (LATE) with an application in the banking industry. The first study estimates the LATE of credit card Miles Benefits programs -reward programs tied to the purchase of airline tickets- on average monthly credit card expenditures, using a FRD design with multiple cumulative cutoffs, corresponding to different credit card tiers. Precise conditions are provided for using the FRD design. This includes the mathematical assumptions required for identifying the LATE, as well as a clear specification of baseline and follow-up periods, following target trial emulation principles, which are important given the longitudinal nature of the data. The estimated effects for lower-tier credit cards are statistically insignificant, likely reflecting the limited disposable income of their target users. In contrast, the effects for higher-tier cards are larger and statistically significant, suggesting that reward incentives are more effective among higher-income clients. These findings suggest that banks may consider replacing costly MB programs for lower-tier credit cards with more cost-effective alternatives that better align with the preferences and financial behavior of the clients they target. The second study estimates the LATE of acquiring a higher-tier credit card on average monthly credit card expenditures by capitalizing on a system glitch that temporarily removed the salary requirement for credit cards, and using the glitch as an IV. The identifying assumption are discussed and formally presented, and three methods are used to estimate the effect including a naive Plug-in estimator, Two-Stage Linear Regression and Double/Debiased Machine Learning. The data is divided in salary groups according to each credit card cutoff, and the treatment is defined as acquiring a credit card which do not correspond to the salary group. The estimated effects are positive and statistically significant across all salary groups, and are very close across all methods within each salary group.
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy
- Level thesis:degree_level
- doctoral
- Discipline thesis:degree_discipline
- Mathematics & Statistics
- Grantor
- Boston University
- Year dc:date.issued
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Iberico, Luis
- Advisor dc:contributor.advisor
-
- Lok, Judith J.
Subjects
dc:subject × 6Rights
- Language dc:language.iso
- en_US
Identifiers
dc:identifier.*- Handle dc:identifier.uri
- https://hdl.handle.net/2144/53733