arxiv: 2604.11926 · v1 · submitted 2026-04-13 · 💰 econ.EM

Recognition: unknown

Shock, Communication, and Yield Curve Repricing: A Two-Step Empirical Framework for Copom Events in Brazil

Gabriel de Macedo Santos

Authors on Pith no claims yet

Pith reviewed 2026-05-10 15:30 UTC · model grok-4.3

classification 💰 econ.EM

keywords Copom eventsyield curve repricingtwo-step frameworkBrazilian DI curvecentral bank communicationtextual analysisevent studymonetary policy

0 comments

The pith

A two-step framework separates initial shocks from later repricing in Brazil's yield curve around Copom events.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a method that splits the immediate market reaction to underlying economic shocks from the additional changes in Brazilian interest rate futures that occur before the next Copom statement. It combines daily market data, survey expectations, and textual measures of statement tone, guidance, and uncertainty extracted from the announcements themselves. The approach proves most useful for shorter and medium maturities on the DI curve, reaching an in-sample R-squared near 0.43 for the 252-day contract under ordinary least squares. Longer maturities and curve slope measures show weaker results, and out-of-sample tests add little explanatory value. The main advance is therefore a practical decomposition that lets researchers trace how shocks and official communication together move the curve.

Core claim

By applying a two-step empirical strategy to 59 Copom events that preserve clean analytical windows, the initial market reaction tied to the shock can be isolated from the subsequent repricing that occurs up to the first following statement, with textual features from the statements then used to explain the second-stage adjustments; this decomposition delivers the strongest fit at the front and intermediate sections of the curve, particularly an in-sample R-squared of about 0.43 for the DI 252-day maturity in baseline specifications.

What carries the argument

The two-step empirical framework that first isolates the initial market reaction associated with the underlying shock and then models the subsequent repricing between the shock and the next Copom statement using textual features.

If this is right

The framework explains more variation in front- and intermediate-maturity DI contracts than in the 504-day segment or in slope changes.
Textual measures of tone, forward-guidance direction, and uncertainty enter with economically plausible coefficients but do not deliver uniformly robust statistical significance.
Out-of-sample predictive performance stays limited even when the in-sample fit reaches 0.43.
The same-day Copom events can be retained in the second stage once the full set of 59 events is used.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The decomposition could be applied to policy events in other emerging-market economies to compare how communication quality affects curve stability.
If the textual variables prove stable in larger samples, they might be combined with high-frequency intraday data to tighten identification of communication effects.
Extending the windows while adding controls for concurrent global shocks would test whether the current separation remains robust.

Load-bearing premise

The hand-built event calendar and chosen analytical windows cleanly isolate the initial shock from subsequent repricing without contamination from overlapping news or market microstructure effects.

What would settle it

Re-estimate the baseline OLS model for the DI 252-day maturity after removing any events that contain identifiable overlapping news releases within either analytical window and check whether the in-sample R-squared falls materially below 0.43.

read the original abstract

This paper proposes a two-step empirical framework to study the repricing of the Brazilian DI curve around Copom-related events. The empirical strategy separates the initial market reaction associated with the underlying shock from the subsequent repricing observed between the shock and the first Copom statement that follows it. The dataset combines a hand-built event calendar, daily market data, Focus expectations, and structured textual features extracted from Copom statements, including tone, forward-guidance direction and explicitness, and uncertainty indicators. In the updated sample, 59 events retain both analytical windows, allowing the second stage to include the full set of same-day Copom events. Baseline results suggest that the framework is most informative at the front and intermediate sections of the curve, especially for the DI 252d maturity, for which the baseline OLS specification reaches an in-sample R2 of about 0.43. By contrast, explanatory power is materially weaker for the DI 504d maturity and for slope adjustments, and out-of-sample performance remains limited. The textual variables display economically plausible signs, but their statistical contribution is not uniformly robust across specifications. The main contribution of the paper is therefore methodological and applied: it offers an implementable event-based decomposition for assessing how shocks and Copom communication jointly shape curve dynamics in Brazil.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper applies standard event-study and text analysis methods to Brazilian Copom events in a two-step framework, delivering usable results mainly at shorter maturities but with limited robustness on the key separation assumption.

read the letter

The main takeaway is that this is an applied empirical paper that adapts existing methods to study Copom events in Brazil. It splits market reactions into an initial shock window and a subsequent repricing window, then uses text features from statements to explain the second part. It does a good job with the data assembly, combining market data, expectations, and structured textual variables like tone, forward guidance, and uncertainty. It gets a reasonable in-sample fit of 0.43 R2 for the 252-day maturity, with coefficients that line up with expectations on tone and guidance. This shows the framework can pick up communication effects at the front end of the curve. The limitations are straightforward from the reported results. Fit is weaker at the 504-day maturity and for slope measures. Out-of-sample tests are limited, and the text features lack uniform robustness. The clean isolation of shock from repricing rests on the event calendar and window choices, but the paper does not appear to test alternatives or check for contamination from other news, which could affect how much we trust the attribution to communication. This paper is for specialists in monetary policy transmission in emerging economies or those using text analysis for central banks. It offers a replicable template rather than a general advance. I recommend sending it to peer review. The work is focused and the empirical choices are laid out clearly enough for referees to evaluate and improve the robustness sections.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes a two-step empirical framework to decompose the initial market shock from subsequent repricing of the Brazilian DI yield curve around Copom events. It combines a hand-built event calendar with daily market data, Focus survey expectations, and textual analysis of Copom statements (tone, forward-guidance direction/explicitness, uncertainty). For 59 events with complete windows, the second-stage OLS regressions show the strongest in-sample fit for the DI 252-day maturity (R² ≈ 0.43), with weaker performance for longer maturities and slope factors; out-of-sample results are limited, and textual coefficients have plausible signs but are not uniformly robust.

Significance. If the window separation is valid, the framework provides a practical method for isolating communication effects from shocks in yield curve dynamics, which could be extended to other central bank events. The paper's transparent reporting of limited OOS performance and non-uniform robustness adds credibility to the positive results at shorter maturities. However, the absence of robustness tests for the core identification assumption limits the strength of the methodological contribution.

major comments (2)

The two analytical windows are defined using a hand-built event calendar without reported sensitivity analyses to alternative window lengths, overlap detection procedures, or exclusion criteria for confounding news. This is load-bearing for the central claim, as contamination in the first window (including repricing) or second window (including other events) would bias the attribution of returns to the textual features, inflating the reported in-sample R² of 0.43 for the DI 252d maturity and undermining the decomposition's validity.
The baseline results claim the framework is most informative at the front and intermediate sections of the curve, but the manuscript does not report first-stage shock identification details (e.g., exact return calculation or how the initial reaction is isolated from expectations) or cross-maturity comparisons of the first-stage residuals. Without these, it is unclear whether the second-stage textual variables capture genuine repricing or residual shock effects, directly affecting the maturity-specific informativeness conclusion.

minor comments (2)

The extraction methodology for the textual features (tone, forward-guidance explicitness, uncertainty indicators) from Copom statements is referenced but not described in sufficient detail for replication or assessment of potential measurement error.
Notation in the regression specifications could be clarified to explicitly distinguish the first-stage shock measure from the second-stage repricing variables and controls.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which help clarify the identification assumptions underlying our two-step framework. We address each major comment below and indicate the revisions we will implement.

read point-by-point responses

Referee: The two analytical windows are defined using a hand-built event calendar without reported sensitivity analyses to alternative window lengths, overlap detection procedures, or exclusion criteria for confounding news. This is load-bearing for the central claim, as contamination in the first window (including repricing) or second window (including other events) would bias the attribution of returns to the textual features, inflating the reported in-sample R² of 0.43 for the DI 252d maturity and undermining the decomposition's validity.

Authors: We agree that the absence of sensitivity analyses for the window definitions represents a limitation in the current draft. In the revised manuscript we will add a dedicated robustness appendix that reports results under alternative window lengths (including ±1-day and ±3-day specifications), explicit overlap-detection rules, and event-exclusion criteria based on confounding news. These checks will be used to confirm that the baseline in-sample fit for the DI 252-day maturity remains stable and that the attribution of returns to textual features is not driven by window contamination. revision: yes
Referee: The baseline results claim the framework is most informative at the front and intermediate sections of the curve, but the manuscript does not report first-stage shock identification details (e.g., exact return calculation or how the initial reaction is isolated from expectations) or cross-maturity comparisons of the first-stage residuals. Without these, it is unclear whether the second-stage textual variables capture genuine repricing or residual shock effects, directly affecting the maturity-specific informativeness conclusion.

Authors: We concur that additional documentation of the first-stage procedure is required to substantiate the maturity-specific conclusions. The revised version will expand the empirical strategy section to specify the exact return calculation (log price changes over the initial window), the precise role of Focus survey expectations in isolating the shock component, and will include a new table or figure displaying cross-maturity first-stage residuals. These additions will allow readers to assess whether the second-stage textual coefficients reflect communication-driven repricing rather than residual shock variation. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical two-step OLS framework is data-driven with no self-referential reductions.

full rationale

The paper describes a two-step empirical strategy that first identifies events via a hand-built calendar and then runs OLS regressions of curve returns on textual features extracted from statements. The reported R2 of 0.43 for DI 252d is an in-sample fit statistic, not a prediction that reduces to the fitted parameters by construction. No equations, uniqueness theorems, or ansatzes are invoked that would make any result tautological. The framework's assumptions (window isolation, event selection) are testable against external data and do not embed the target coefficients. This is a standard applied-econometrics paper whose central claims rest on observable statistical performance rather than definitional equivalence.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no explicit free parameters, axioms, or invented entities; standard OLS assumptions are implicit but not detailed.

pith-pipeline@v0.9.0 · 5530 in / 1159 out tokens · 55528 ms · 2026-05-10T15:30:50.422334+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

2 extracted references

[1]

Alves, C. R. A., Abraham, K. J., & Laurini, M. P. (2023). Can Brazilian Central Bank communication help to predict the yield curve? Journal of Forecasting, 42(6), 1429 –1444. Alves, C. R. A., & Laurini, M. P. (2025). The effects of Brazilian Central Bank communication on the yield curve. Macroeconomic Dynamics,

2023
[2]

Making text talk

Blinder, A. S., Ehrmann, M., Fratzscher, M., De Haan, J., & Jansen, D. -J. (2008). Central Bank Communication and Monetary Policy: A Survey of Theory and Evidence. Journal of Economic Literature, 46(4), 910–945. Brand, C., Buncic, D., & Turunen, J. (2010). The Impact of ECB Monetary Policy Decisions and Communication on the Yield Curve. Journal of the Eur...

2008