Exploring the Determinants of Central Bank Credibility

How to cite

Mcmahon, Michael. 2022. "Exploring the Determinants of Central Bank Credibility." AEA RCT Registry. AEARCTR-0009026. https://doi.org/10.1257/rct.9026-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Michael Mcmahon*

* Primary investigator

Registry ID
AEARCTR-0009026
DOI
https://doi.org/10.1257/rct.9026-1.0
First registered
Last updated
Status
In development
Trial dates
2022-02-25 – 2022-05-31
Countries
Private
JEL codes
E31, E58
Registry record
http://www.socialscienceregistry.org/trials/9026

Abstract

Managing expectations via central bank communication has become a primary component of monetary policy. A necessary condition for effective communication is central bank credibility. Despite this, we know surprisingly little about the determinants and properties of credibility. This paper aims to fill this gap by introducing a novel experimental framework that allows for nuanced study of central bank credibility. We employ this framework to study how variations in a central bank’s recent inflation forecasting performance influences subjects’ perceptions of the bank’s credibility.

Intervention

We study how the timing and magnitude of central bank forecasting errors impact the perceived credibility of the central bank. We first expose participants to a 12-quarter history of realized inflation alongside corresponding central bank forecasts. After viewing this history, participants form their own inflation forecast (our elicitation captures information about the first three central moments of subjects’ inflation expectations) for the next quarter (quarter 13). We then reveal the central bank’s quarter-13 inflation forecast and allow subjects to update their own inflation forecast (or not). We measure central bank credibility as the degree to which subjects update their forecast in the direction of the central bank’s forecast. We normalize this measure so that we may compare it cleanly across different economic histories.

Experimental design

We use a within-subjects design to study how the timing and magnitude of central bank forecasting errors impact the perceived credibility of the central bank. To do this, we first expose participants to a 12-quarter history of realized inflation alongside corresponding central bank forecasts. After viewing this history, participants form their own inflation forecast (our elicitation captures information about the first three central moments of subjects’ inflation expectations) for the next quarter (quarter 13). We then reveal the central bank’s quarter-13 inflation forecast and allow subjects to update their own inflation forecast (or not). We measure central bank credibility as the degree to which subjects update their forecast in the direction of the central bank’s forecast. We normalize this measure so that we may compare it cleanly across different economic histories.

We introduce treatment variation via the economic histories that precede participant forecasts. We vary whether the bulk of central bank forecast errors in the 12-quarter history occur early, occur late, or are distributed consistently. Our interest is in how the timing of errors impacts the degree to which subjects incorporate the central bank’s quarter-13 forecast into their own updated forecast. We also vary whether central banks have a relatively good or bad forecasting history (I.e. we vary the average absolute forecasting error).

This design yields a total of 12 treatments. Each subject experiences either the good forecast or bad forecast history but never both. For that history, a subject experiences only one of the six permutations of Early, Late, and Consistent forecast histories. This within-subject design ensures that idiosyncratic characteristics (over confidence, etc.) cannot drive effects. Further, treating each permutation as its own treatment ensures we are able to net out of our final results any ordering or learning effects.

Design details

Randomization method
Randomisation done via a randomisation algorithm.
Randomization unit
Subjects participated individually. We deployed our experiment online via Prolific.
Clusters
2400
Observations
2400
Arms
200 observations per treatment x 12 treatments: Good: Early-Consistent-Late Good: Early-Late-Consistent Good: Consistent-Early-Late Good: Consistent-Late-Early Good: Late-Early-Consistent Good: Late-Consistent-Early Bad: Early-Consistent-Late Bad: Early-Late-Consistent Bad: Consistent-Early-Late Bad: Consistent-Late-Early Bad: Late-Early-Consistent Bad: Late-Consistent-Early
Primary outcomes
Central Bank Credibility as measured by the weight that individuals place on the central bank forecast in forming their expectations.

Cited by

No citation data available for this registration yet.

Keywords