Incentives in Macroeconomics

We exogenously vary the type of incentivisation when eliciting beliefs about the macroeconomy.

2024-09-11

2024-09-15

1) reported baseline inflation expectations
2) changes in these inflation expectations

1) reflect answers in common inflation expectation surveys
2) reflects effect sizes of common information provision experiments (RCTs) that aim to shift inflation expectations

1) attention to information
2) uncertainty about reported inflation expectations (where collected)

1) this could be time spent on pages, but also accuracy of attention checks
2) refers to indicated confidence in reported inflation expectation (not always collected)

With a between-subject design, we assess the role of incentives in macroeconomic belief elicitation.

Not available

It's a between-subject design. Randomisation thus primarily occurs at the end of Prolific.

Between-subjects

No

n/A

250 participants per treatment; 1,000 overall

We anticipate collecting a total of approximately 250 observations across each of our four treatment conditions for a total of 1,000 observations. Briefly, we base this on the following ex-ante power calculation + logic: Because an interesting and possible outcome in our study is that different incentives do not induce significant treatment effect, we want sufficient statistical power to trust null results. Therefore, our goal is to collect enough data so that we can detect reasonably small effects (Cohen's D of ~ .3) at alpha=.01 and power =.8. Assuming equal variance across treatment, this implies we need approximately 262 participants per treatment (to make all pair-wise comparisons).

With 250 participants per treatment, our minimum detectable effect size should be when Cohen's D is approximately .3. This is to detect the effect when comparing two treatment conditions where power = .8, alpha = .01 and we assume equal variance.