Macroeconomic Expectations and Consumption

We will conduct phone surveys with a sample of clients from a large commercial bank. We will be able to link the survey responses back to the administrative data of the bank, which includes real-time information on the behavior of these individuals such as their credit card and debit card transactions.

In the survey, we'll measure two key beliefs that are important in the context of this experiment: the expected inflation rate over the following 12 months, and the expected nominal exchange rate 12 months in the future.

We will start by exploring the association between these beliefs and different forms of behavior. As complementary evidence, we'll embed and information-provision experiment in the survey. That will allow us to obtain experimental estimates of the causal effects of beliefs on behavior.

2019-01-20

2019-04-01

The expected future spending, according to survey data, and the actual future spending, according to the high-quality administrative records from the bank.

We will test a series of predictions from macroeconomic models. For example, if an individual believes that the nominal exchange rate will go up in the next 12 months, that individual may want to buy soon imported goods such as consumer electronics. For these outcomes, we'll have two versions:
(i) Subjective measures of future spending elicited at the end of the survey.
(i) Objective measures of actual future spending obtained from the administrative records of the bank.
For example, we included one subjective question about whether it was a good time to buy consumer electronics. And we can use the bank's transaction data to measure (approximately) the actual spending on consumer electronics. Given that the transaction data is high-frequency data, we can explore effects at different time horizons (e.g., the effects during the first month versus the effect during the first three months).

The first part of the analysis is non-experimental: we want to measure the association between certain beliefs and spending behavior. The second part of the analysis is experimental, for which we use a standard information-provision experiment.

The first part of the analysis is purely observational. For the second part of the analysis, we exploit an information provision experiment -- that is, we randomize different pieces of information to different individuals, and then track the effects of the experiment on subsequent beliefs and behavior.

The structure of this experiment is as follows:

— We elicit prior beliefs about two macroeconomic variables (inflation and the nominal exchange rate).
— We randomly provide them with one of multiple pieces of information about these expert forecasts on these key parameters.
— We elicit posterior beliefs, to determine if our feedback shifted their expectations.

Note that we are interested in measuring the causal effect of beliefs. We are not interested in measuring the average treatment effect of providing a piece of information -- indeed, this average effect is probably zero, because the same piece of information may make some individuals update upwards and other individuals update downwards.

Given that we have multiple outcomes, we are planning to use standard methods for joint hypotheses testing. Additionally, we will also consider more modern machine learning methods.

Note also that we will have the administrative data for the year before our intervention as well as the year after our intervention. As a result, we can use pre-treatment outcomes for falsification tests (i.e., did the information affect behavior before the information was provided?). We'll be able to use the pre-treatment outcomes as control variables, which will help us to increase power. The pre-treatment outcomes may also be useful for heterogeneity analysis.

The randomization will be managed automatically by the Qualtrics survey platform.

Individual respondent

No

The final number of respondents will depend on the response rate to the phone surveys. We are aiming to collect responses from 3000 individuals

3000 individuals

Individuals can receive feedback about the inflation rate, the exchange rate, or both. We'll randomly assign them with equal probability, resulting in one third of the respondents in each treatment group.