Reca Sarfati

Research Papers

"Narrative-Hacking" [Draft]

Economists seldom base conclusions on isolated hypothesis tests. Rather, it is common to combine multiple atomic tests in a logical structure to serve higher-order purposes, such as distinguishing between competing theories, defending causal claims, diagnosing mechanisms, and arranging disparate facts into coherent stories. These economic "narratives" are foundational to how findings are framed, interpreted, and communicated—governing a paper's overarching message and ultimate societal impact. Yet, a single set of test outcomes can support many narratives, not all of which are true. While practitioners today recognize the importance of multiple testing corrections to guard against practices such as p-hacking, these procedures target atomic errors, not those of downstream narratives built upon them. This paper presents a general model for the construction and testing of narratives that admits a formal definition of Type I "narrative error." After partitioning narratives into two classes—monotonic (e.g., impact evaluations) and non-monotonic (e.g., balance checks)—we first show a positive result: if a narrative admits a representative test that is (weakly) increasing in atomic rejections, any procedure that controls the family-wise error rate (FWER) over underlying atomic tests at level alpha automatically delivers uniform narrative size control at alpha. A corollary is a "free narrative shopping" guarantee: once atomic tests are fixed or preregistered, researchers may explore any monotone narratives ex post without inflating size, thereby immunizing them against potential concerns of ex post "narrative-hacking." We then find an impossibility result: when testing sets include non-monotone narratives—such as a set containing both a narrative and its negation—atomic FWER control cannot achieve uniform narrative size control with alpha < 0.5. To accommodate arbitrary collections of narratives, we provide a novel procedure that relates uniform size control to the construction of joint confidence sets. We show this approach is necessary and sufficient for uniform narrative error control. We demonstrate practical implementation of this framework by replicating several key findings in Dell (2010).

Works in Progress

Balance Checks as Conditional Reporting (with Vod Vilfort)

Inference with Selected Instruments (with Vod Vilfort)

Abstracts available in CV.

Online Estimation of DSGE Models (with Michael Cai, Marco Del Negro, Edward Herbst, Ethan Matlin, and Frank Schorfheide) The Econometrics Journal, Vol. 24, Issue 1, January 2021.

Hindsight and Sequential Rationality of Correlated Play (with Dustin Morrill, Ryan D'Orazio, Reca Sarfati, Marc Lanctot, James R Wright, Amy R Greenwald, Michael Bowling) Proceedings of the AAAI Conference on Artificial Intelligence, 35(6), 5584-5594, May 2021.

Estimating HANK for Central Banks (with Sushant Acharya, Marco Del Negro, Ethan Matlin, Reca Sarfati, William Chen, Keshav Dogra, Shlok Goyal, Donggyu Lee, Sikata Sengupta). Heterogeneity in Macroeconomics: Implications for Monetary Policy, 1st ed. Central Bank of Chile, 2024.

MIT

Awarded "Best Graduate Teaching Assistant" by MIT Economics Department in 2025 for PhD Econometrics Core I, II (14.380 and 14.381)

• 2026, New Econometric Methods (14.386, PhD)
  TA to Alberto Abadie and Anna Mikusheva
• 2025, Large-Scale Decision Making and Inference (14.39/390, UG/MS)
  TA to Isaiah Andrews
• 2022-24, Estimation & Inference of Linear Models (14.381, PhD Core II)
  TA to Whitney Newey
• 2024, Statistical Methods in Economics (14.380, PhD Core I)
  TA to Ashesh Rambachan
• 2023, Firms, Markets, Trade, and Growth (14.76/760, MS)
  TA to David Atkin and Dave Donaldson
• 2022, Economics and Society’s Greatest Problems (14.009, UG)
  TA to Esther Duflo

Brown University

Awarded "Women in Computer Science '94 Best Undergraduate Teaching Assistant" by Brown Computer Science Department in 2017.

• 2018, Algorithmic Game Theory (CSCI 1951K, MS/PhD)
  TA to Amy Greenwald
• 2017, Design and Analysis of Algorithms (CSCI 1570, UG/MS)
  TA to Paul Valiant
• 2017, Intro to Algorithms and Data Structures (CSCI 0160, UG)
  TA to Seny Kamara
• 2016, Statistics for Public Policy (PLCY 2455, MPA)
  TA to John Friedman
• 2016, Microeconomics for Public Policy (PLCY 2460, MPA)
  TA to Emily Oster

• 2021-23, President and Treasurer, MIT Graduate Economics Association
• 2018-20, Research Analyst, Federal Reserve Bank of New York, Applied Macroeconomics & Econometrics Center
• 2018-(ongoing), Open source software developer for Julia Programming Language
• 2016-18, Research Assistant, Brown University Social Science Experimental Laboratory

As a public resource, I wrote a "≥ 1st Year Guide to the Economics PhD," available here: Guide_to_the_MIT_Econ_PhD.pdf

• DSGE.jl: This package implements the New York Fed dynamic stochastic general equilibrium (DSGE) model and provides general code to estimate many user-specified DSGE models. The package is introduced in the Liberty Street Economics blog post, "The FRBNY DSGE Model Meets Julia."
• SMC.jl: This package implements the Sequential Monte Carlo (SMC) sampling algorithm, an alternative to Metropolis Hastings Markov Chain Monte Carlo sampling for approximating posterior distributions. The SMC algorithm implemented here is based upon and extends Edward Herbst and Frank Schorfheide's paper "Sequential Monte Carlo Sampling for DSGE Models" and the code accompanying their book, Bayesian Estimation of DSGE Models. 
  • Our implementation features an adaptive schedule and what we term generalized tempering for "online" estimation, as outlined in our paper, "Online Estimation of DSGE Models." For a broad overview of the algorithm, one may refer to the following Liberty Street Economics article.
• StateSpaceRoutines.jl: This package implements common computational routines for state-space models. Provided algorithms include the Kalman filter; Chandrasekhar recursions; Tempered Particle Filter; Hamilton and Koopman Kalman smoothers; as well as Carter and Kohn and Durbin and Koopman simulation smoothers.
• ModelConstructors.jl: This package contains the building blocks of model objects, such as Parameter, Observable, Setting, and State types. You may define any custom model, so long as it has parameters. The model object is used in both DSGE.jl and SMC.jl.