Experimental Design Details
There are two parts to the analysis:
1) A Short Survey that is sent out via Twitter. The purpose is to elicit economic expectations and their uncertainty, as well as users' sentiments about institutions and policy. in addition to point expectations, also ranges of selected expectations are elicited to capture uncertainty. The collected data will then be matched to empirically calculated values of these expectations using NLP. Particularly, measures of users' beliefs and the precision of beliefs. The key here is to understand whether and how users' tweets reflect their actual beliefs, expectations and sentiments, but also understand which expectations are mostly reflected (past, short-run, long-run inflation?). In short: how informative is data on Twitter for economic expectations? This aims to also provide some information on individuals' economic narratives.
2) The RCT: I'm running a survey experiment in which participants from different nationalities are being tasked to perform a macroeconomic forecast (e.g. inflation prediction, GDP, unemployment rate). Participants indicate a point and variance forecast. Participants will be rewarded by the accuracy of their forecast (incentivised). There are two treatments. In treatment "european", participants will get related forecast information from the European Central Bank. In treatment "national", participants will get the same related forecast information but from their own national bank (e.g. Deutsche Bundesbank, Banque de France, etc.). Participants get either "european" or "national". After the information treatment, participants get the opportunity to revise their forecast.
The idea here is to see how Bayesian participants are, when they update their forecasts after the information treatment. This design allows to test, whether there are differences between nationalities in how participants use information from monetary policy institutions to update their expectations. Further, it can be tested whether information from the national banks are used differently from the european bank. Particularly, whether there exists something like an "ingroup effect", where participants who share their nationality with the head of the ECB update differently to the european bank's information than participants in the "outgroup". In addition, this allows to test whether the update to the national bank's information of "outgroup" participants is comparable to the "ingroup's" update to the european bank's information. If this is the case, this gives rise to the policy recommendation of using national central bank communication more strongly.
In addition to the experimental intervention, this survey will also collect information on trust and knowledge about the monetary policy institutions (but also commonly used ability tests) to make inferences about potential mechanisms of the effect of interest.
2-4 nationalities: french and german (+ possible extension with spanish and italian)