Firm Expectations and Price-Setting

We study firms that sell their products through an online retail platform in India. We invited around 20,000 firms that were active in the platform in the months before the experiment. Within the survey, we conduct an information-provision experiment. We elicit beliefs on, and provide information about, two topics:

(1) the expected nominal exchange rate in the next 12 months.
(2) the expected inflation rate in the next 12 months.

The goal is to measure how the macroeconomic information included in the survey affected firms' day-to-day decisions.

The intervention was conducted between May 2019 and March 2020. However, due to COVID, the process of acquiring the administrative records was delayed and as of the time of this writing, to the extent that the COVID situation permits, we are in the process of acquiring the administrative data. Thus, at the time of this registration, we have not analyzed the effects on pricing decisions yet.

2019-05-24

2020-03-31

Our primary outcome is the pricing decision of the firm, as measured in the administrative records. We can study the extensive margin (i.e., whether the firm changes prices) as well as the intensive margin (i.e., the magnitude of the price change). For firms that list multiple products, we anticipate analyzing the average pricing decision as well as the decisions of the best-selling product. We are hoping to obtain real-time administrative data covering several months before and after the intervention, which will allow us to i) study long-term effects and not only short-term effects; ii) conduct event-study analysis.

There are several secondary outcomes that may help to understand the effects more broadly, as well as disentangling causal mechanisms.

1) In the survey, after the information provision, we elicit posterior beliefs about the macroeconomic indicators. This allows us to measure precisely how the expectations reacted to the information shocks.

2) The survey includes other questions after the information-provision stage, which can be used as survey outcomes (see the attached survey instrument for more details):

a) Most importantly, we elicited the expected price change for their best-selling product.
b) Expectations about business revenues in the next year or two.
c) Business conditions in the country as a whole in the next 12 months.
d) Whether it is a good time to make capital investments in the business.
e) Whether it is a good time to hire additional staff for the business.
f) Whether it is a good time to take out a loan to expand the business.

The rich administrative records can be used to construct other secondary outcomes. More precisely:

a) Inventory changes.
b) Activity (i.e., whether the firm was active or not after the intervention).

We invited around 20,000 firms that sell their products through an online retailer to answer an online survey. Because of the high turnover rate in this type of platform, we selected firms with recent online activity.

Invitations were distributed using a pop-up window that was shown whenever a seller logged in to his/her profile. To increase the response rate, the online platform offered a monetary incentive (i.e., a discount to be used in online purchases on the website) for users that completed the survey. To link the survey responses to the administrative records, we created unique survey links for each seller.

In the survey, we elicit beliefs on and provide information about, two topics
(1) the expected nominal exchange rate in the next 12 months.
(2) the expected inflation rate in the next 12 months.

The main intervention of our experiment is designed to create exogenous variation in macroeconomic expectations. We randomly divide the sample into three groups: two treatment groups and one control group. The first treatment group receives information about the expected nominal exchange rate 12 months in the future (forex-information group). The second group receives information about the expected inflation rate over the following 12 months (inflation-information group). The information provided to these groups is based on the Survey of Professional Forecasters on Macroeconomic Indicators published bi-monthly by the Reserve Bank of India. Finally, the third group receives no additional information (control group). To measure whether individuals incorporated the information provided to them into their beliefs, we measured their prior beliefs (i.e., before receiving the information) and their posterior beliefs (i.e., after receiving the information).

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.

We will test predictions from simple macroeconomic models. According to these models, when firms expect the local currency to depreciate, they should raise the prices in anticipation. Likewise, when firms expect the inflation rate to increase, they should raise the prices in anticipation. We also expect to test some additional hypotheses informed by these macroeconomic models. For example, we can measure how exposed firms are to the foreign exchange rate (e.g., whether they export their products or import some of their inputs), and then test whether the firms which are more exposed care more about the exchange rate.

Note also that we will have the administrative data from before 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.

Randomization done by the online survey software

Firm

No

We invited 20,000 firms to the survey.

We invited 20,000 firms to the survey. The final number of respondents will depend on the response rate to the survey and the process of linking the survey data with administrative records. Even though we have masked identifiers that allow us to merge the survey and administrative data one to one, we anticipate that not every firm that filed the survey will show up in the administrative records for transactions or prices. For instance, despite we send survey invitations to active firms in the pre-intervention period, firms may become inactive post-intervention.

We invited 20,000 firms to the survey. 1/3 was assigned to the inflation-information group, 1/3 was assigned to the forex-information group and 1/3 was assigned to the control group