1   Overview

The goal of this project is to conduct an empirical analysis on the impact of tariffs on imports. We will take advantage of a data set cleaned and made available by authors of Fajgelbaum et al. (2020).  After that, you need to estimate a model I have specified for you and report the results. The model follows the regression specification for Table 1 of Amiti et al. (2019).1

Based on your analysis, you will deliver the following at the end of the semester:

1. A 15 minutes presentation.

2. A term paper. Please submit your codes together with the paper.

2   Project Description

Step One: Download Data

Download the data set I uploaded to eClass. The data set is a part of the original replication files that belong to Fajgelbaum et al. (2020). Please do not use it without citing the authors properly.

If you are interested, you can find their paper through this link.  The original replication files are available under “Supplementary Material” .

Briefly read their data section (page 7 of Fajgelbaum et al. (2020)) and find out the original data sources.  You will need to include the original sources in your report.

Step Two: Generate summary statistics

Create a table of summary statistics for the key variables you will in- clude in your analysis. A good summary statistics table reveals interesting data patterns (e.g., Table II in Fajgelbaum et al. (2020) ).

Step Three: Estimate baseline model

Use any statistical software to estimate the following regression specifi- cation:

∆ln(zijt) = µj + ηit+ β∆ln(1 + τijt) + ϵijt                               (1)

where i is an indicator for the exporting country, j denotes products, and t denotes time (a month in our context). τijt is the effective ad valorem tariff rate of product j from country i at time t. It corresponds to effective rate, or variable“m_apptariff”in the data set provided by Fajgelbaum et al.(2020). µj  is a product fixed-effect that captures all time-invariant characteristics of a product (you can omit this term if your data set contains only one HS6 product), and ηit is a country-year fixed-effect.ϵijt is an error term.

∆ln(zijt) is the log change in outcome variables before and after the tar- iff change. For consistency, take a one-year log difference. Here are a few outcome variables to consider: (1) import price (price measured as the unit value, i.e. import value divided by import quantity); (2) import value; (3) import quantity.

Pay attention to the following details:

1. All variables with ∆ are corresponding to a 12-month difference.

2. Standard errors are clustered at HS8 level.

Step Four: Exploration

After estimating the baseline model, you should explore some other research question. Here are some examples:

• Choose a few broad industries and compare the estimates for those industries. Interpret the differences.

• Choose a group of exporters that the U.S. import from.  Limit your analysis to import flows from those countries. Compare it to the base- line results.

• Estimate the impact of tariffs to U.S. export flows. You can obtain U.S. export data (merged with tariffs) from the original replication files of

Fajgelbaum et al. (2020).

• More...

Step Five: Report and interpret results

Report your estimate of β (we usually use βˆ to represent the estimated coefficient) in a table with standard errors presented in an parenthesis be- low the point estimates. Include a table caption that describes (concisely) what this table reports. Add a footnote to your table that includes the fol- lowing information:

• Source of your data

• What an observation means in your data set (e.g., HS10-exporter- month).

• Any additional information that can help the reader understand the numbers you report.

After reporting your results, discuss your results. Interpret the coefficient estimates. In this model, the coefficient estimate βˆ has an economic inter- pretation. What is it?

Use the coefficient estimate βˆ and calculate the following:

Dit = Mitτitβˆ ln(  )

where Mit  is the import value and Mitτit  is the tariff revenue from coun- try i in month t.  τi,t − 12  is the tariff rate 12 months ago.  Also calculate Dt = 对iDit . What do you think this term represents?

Please sufficiently discuss your results. Your discussion and interpreta- tion are as important as–if not more important than–the empirical analysis.

3   Presentation guidelines

You should prepare a 15 minutes presentation. During the presentation, you should:

• Explain which country, which product, and which policy you have chosen and why.

• Discuss one challenge you face during the data collection, cleaning, and analysis process. How did you resolve that?

• Present your summary statistics and estimation results (if ready).

After the presentation, be prepared for a 5 minutes Q&A.