Personalized Response Scales in Inflation Expectations Surveys
How to cite
Eife, Thomas; Becker, Christoph; Duersch, Peter. 2024. "Personalized Response Scales in Inflation Expectations Surveys." AEA RCT Registry. AEARCTR-0012861. https://doi.org/10.1257/rct.12861-2.2
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
Central banks regularly measure inflation expectations using density forecasts, in which respondents are asked to assign probabilities to pre-specified ranges of inflation. In previous experiments, we showed that such density forecasts are susceptible to changes in the scale and might not be well adapted to situations were the actual inflation rate changes substantially. Some of these problems can be remedied by shifting the scale of the density forecast by the point forecast of the survey participants. In the current study, we want to explore this further by providing completely personalized scale to the participants.
In the survey, participants are asked up to four questions on their inflation expectations. Three of these questions are similar to those used in the New York Fed’s SCE and ask about inflation expectations over the next 12 months: A binary question if inflation or deflation is more likely, a point forecast, and the aforementioned density forecast. Additionally, some participants are asked to also state their minimum and maximum inflation rate, which we elicit in different ways. This minimum and maximum is then used to construct a personalized scale for the density forecast which is takes the thus stated possible inflation ate into account. For some of the treatments the personalized scale is additionally centered around the previously stated point forecast.
Intervention
Intervention: Eliciting inflation density forecasts using personalized scales
Experimental design
Participants in the survey will be asked whether they expect inflation or deflation 12 months from now [Q1] and then give their inflation expectations again in a point forecast [Q2]. Depending on the treatment, this is followed by eliciting their minimum and maximum expected inflation rate [Q3]. This is elicited either directly as the minimum/maximum or as the lower/upper bound of a 90% confidence interval. Additionally, we use three different representations of this questions: Open number boxes, two independently movable sliders, and a symmetric slider (where participants basically state a symmetric range around the point forecast). The last main question is a density forecast for the inflation 12 months from now [Q4]. Questions Q1, Q2 and Q4 similar to those in the New York SCE. Questions Q3 and Q4 are also followed by a difficulty question each, in which participants state how difficulty they perceived the preceding question on a six-item Likert scale. After this, the participants will answer a questionnaire including control and demographic questions [Questionnaire].
Participants will take part in one of twelve possible treatments:
1. Static (S): Uses a static scale for the density question, similar to the one used in the New York Fed’s SCE but with equal bin width.
2. CenterPersonalized8 (CP-8), CenterPersonalized4 (CP-4), CenterPersonalized2 (CP-2), CenterPersonalized1 (CP-1), and CenterPersonalized05 (CP-05): Uses a static scale for the density question, but shifted by the point forecast of the participant. The bin width varies and is 8, 4, 2, 1, or 0.5 percentage points depending on the treatment.
3. FullyPersonalizedBoxA (FPB-A): Asks about the minimum and maximum expected inflation rate using open number input fields. This information will be used to construct the scale for the density question [Q4].
4. FullyPersonalizedBoxB (FPB-B): Asks about the lower and upper bounds of the 90% confidence interval of the expected inflation rate using open number input fields. This information will be used to construct the scale for the density question [Q4].
5. FullyPersonalizedSliderA (FPS-A): Asks about the minimum and maximum expected inflation rate using two independently movable sliders. This information will be used to construct the scale for the density question [Q4].
6. FullyPersonalizedSliderB (FPS-B): Asks about the lower and upper bounds of the 90% confidence interval of the expected inflation rate using two independently movable sliders. This information will be used to construct the scale for the density question [Q4].
7. FullyPersonalizedSliderC (FPS-C): Asks about the minimum and maximum expected inflation rate using a slider that is centered on the point forecast. This slider allows to thus state a range for the minimum and maximum inflation rate around the point forecast. This information will be used to construct the scale for the density question [Q4].
8. FullyPersonalizedSliderD (FPS-D): Asks about the lower and upper bounds of the 90% confidence interval of the inflation rate using a slider that is centered on the point forecast. This slider allows to thus state a range for the minimum and maximum inflation rate around the point forecast. This information will be used to construct the scale for the density question [Q4].
Additionally, participants in the CenterPersonalized treatments are also asked one of the variants of the minimum/maximum questions of the FullyPersonalized treatments (including not being asked the question). For those participants, this information is however not used to construct the scale of the density question [Q4].
In order to recruit a representative sample of the US population, we run the survey on the online research platform Prolific. We recruit a sample of 1800 participants total, 150 per treatment. Participants will be assigned one of the twelve treatments iteratively when starting the study. This way, we make sure that any differences we obtain are not driven by timing but instead reflect a genuine effect of our treatment intervention. Participants will earn a fixed fee of £1 for their participation.
Design details
Cited by
No citation data available for this registration yet.