Inflation Experience and the Anchoring of Expectations: a Comparison of Lab and Survey Experiments

2022-11-21

2023-07-31

Measurement of mean-reversion in inflation expectations via i) the credibility of the target at various horizons and ii) deviations of participants' expectations from the target at various horizons.

We seek to investigate three questions. First, to what extent do these two measurements of the target credibility depend on people’s previous experience with inflation? Second, does central bank’s emphasis on medium-term inflation target (as the ECB now does) favor mean-reversion in short-run inflation expectations and the anchorage of medium-to-long-run expectations? Third, to what extent do the answers to the first two questions differ in the lab and in the survey?

The experiment has two parts, one in a survey and one in the lab. There are an experience and an information treatments.

The experiment has two parts, one in a survey and one in the lab.
The survey experiment is an RCT within a 20-minute online survey of 10,000 households that is conducted in the Netherlands with the help of a large international marketing-research company. The respondents are representative of the general population in terms of gender, age, geographic location and, as much as possible given the large size of the sample, income and education. We use two waves. In the first wave, three experience-in-game treatments are designed next to a control treatment where no experience-in-game takes place. The experimental treatment involves 75% of respondents repeatedly playing an inflation-forecasting game within the context of one of three different Dutch historical time series: 25% play with time series from the early 1970s where inflation increased by 6% beyond 10%, 25% with the 1980s where inflation dropped by 6% down to almost 0% and 25% with the years 2010s where inflation remained close to 2%. This treatment is designed to exogenously affect people’s past experience with inflation. The remaining 25% of the respondents constitute a control group with cohort-dependent experience and do no play any forecasting game. In a recontact wave a month later, a information-provision treatment is introduced: two thirds of the respondents will see one of two pieces of information regarding the strategies and objective of the ECB in terms of inflation: one third will see an information piece that mentions the 2% target while the other one third will see a piece that emphasizes the medium-run nature of the 2% target and potential short-term deviations. The remaining one third is a control group that will not see any information.
In all treatments and waves, the respondents are presented with a list of socio-demographic, financial and general-opinion questions, including inflation perception. In the first wave, we elicit a detailed picture of respondents’ experience with both increasing and decreasing inflation. We then elicit pre- and post-treatment inflation expectations, in the short-, medium- and long-run and in several ways: using point expectations, distributions and qualitative questions.
The second part of the experiment takes place in the laboratory. We replicate the survey experiment within the context of a controlled forecasting experiment where subjects are University students. The forecasting game in the laboratory is more extensive and incentivized, which corresponds to the usual practice in the related literature. To facilitate the comparison with the survey experiment outcomes, we also collect in the lab socio-economic details of the participants, as well as their preferences, opinions, knowledge of monetary policy issues and personal experience with inflation.

Each respondent is randomly assigned to one of the experience and one of the information treatments upon starting the online survey with equal and independent probability.
In the lab, subjects are randomly assigned to one of the treatments with equal probability (two information provision treatments and three experience treatment).

individual

No

10,000 survey respondents and 300 lab subjects

10,000 survey respondents in the first wave, 5,000 expected in the second wave, and 250 lab subjects

2500 respondents per life-experience treatment including control; 1500 respondents expected per information treatments including control; 50 to 60 subjects per experience-information treatment in the lab

2,500 observations per experience treatment, a group of 4,000 without any experience in the survey. A maximum of 60 subjects per treatment in the lab (a total of 360). To look into the effect of information, we pool together experience treatment (5,000 respondents, 180 subjects per information treatment). Examples of two computations are provided below for equivalence trials, with a 80% power and a 5% threshold: - In a continuous case, say with the short-run inflation point forecast, in percentage points, with an average of 4 in Tr. A vs. 4.5 in Tr. B, and a standard deviation close to 2, so effect size of 0.5/2= 0.25, we need about 250 respondents per treatment. - Dichotomous outcomes (e.g. judging likely or unlikely that inflation will go back close to target in two years): with an effect size of 5 p.p.between two independent groups, say from 20 to 25%, effect size: [(0.2*0.8+0.25*0.75)/sqrt(0.05^2)]*7.9 = 1100 per group.