Followers or ignorants: Central bank forecasts and price setting behavior of firms

We use information treatments to test which factors influence firms’ price setting behavior.

First, we ask for estimates of both future (for 2022, 2023) and realized annual inflation rates (for 2021).

We use these prior beliefs
i) to test the effect of our experimental treatments for firms with similar prior inflation expectations on a firm’s price setting behavior

ii) to measure the extend of misperception by comparing the own assessment of inflation dynamics and the inflation rate officially published by the German Bundesbank.

Second, we provide information on different components of inflation to test, which treatments affect the price setting of firms. We randomly assign respondents to one of three treatment groups (INFLATION, ENERGY, WAGES) or a control group (CONTROL). We display the indicated inflation by the respondent together with inflation rates provided by the German Bundesbank for the consumer price index (treatment INFLATION) and together with energy price changes (ENERGY) or wage changes (WAGES), respectively.

In each of the treatment arms, we show participants a table reporting the following inflation rates:
i. Own forecast inflation (CONTROL)
ii. Own forecast inflation + Inflation information (Bundesbank) (INFLATION)
iii. Own forecast inflation + Inflation information (Bundesbank) + Energy price component inflation (Bundesbank) (ENERGY)
iv. Own forecast inflation + Inflation information (Bundesbank) + wages (Bundesbank) (WAGES)

2022-08-01

2023-01-20

• Expectations about the realized (year: 2021) and predicted (years: 2022/2023) inflation rates (in %)
• Price changes of products/services in the next 12 months (in %)

• Expectations about the realized and predicted inflation rates are measured via a textbox, in which participants can enter a numerical value in %. The entry range varies from -100% to 100%.
• Price changes of products/services in the next 12 months are measured via a textbox, in which participants can enter a numerical value in %. The entry range is restricted from below by -100%. From this variable, we construct a measure of "overshooting" as defined in Werning (2022).

• Change in sales (in %) for hypothetical price changes
• Determinants of price change
• Change in capital structure (i.e., increase or decrease in debt)
• Price setting interval duration in the past and currently

• Change in sales (in %) for hypothetical price changes:
Measured via textbox where only numerical entries are allowed. The entry range is restricted from below by -100%.
• Determinants of price change:
Participants can choose from the following options: wage costs, material/energy costs, competitor prices, consumer demand, legal regulations, other factors (text field)
• Change in capital structure (i.e., increase or decrease in debt):
Qualitative measure with four options: Reduction, No change, Increase, No indication
• Price setting interval duration in the past and currently:
2 measures:
o Quantitative: Drop down menu with answer options ranging from daily to more than 24 months
o Qualitative: Options ranging from much less often to much more frequently

see section "Interventions"

Randomization done by a computer through Qualtrics

Firm

No

No clusters

Approx. 1,800 questionnaires with valid responses by firm representatives

Each treatment arm has 450 observations of firm representatives

We start by assuming a sample size of approximately 1,800 observations. This will give us around 450 observations per experimental group. We also assume a standard deviation of 4.68% (Coibion et al., 2020; Online Appendix Table 1. Descriptive statistics of selected variables; p.61). As an example, we take an expected price change of 8%. For these parameters, we check which minimum detectable percentage point difference we are still able to successfully detect. • Estimated effect size and experimental-group mean: o delta = 0.8750 o expected price change under the alternative = 8.8750