Dear all,
I am testing the indirect effects of tariffs imposed by country A on imports from country B. Specifically, I want to investigate whether:
1. Country A increases imports from country C as a response to the tariffs.
2. Country C increases imports from country B as part of a trade deflection mechanism (i.e., country B uses country C to bypass the tariffs).
This is my first time working with econometrics and Stata, so I would really appreciate your help with some challenges I am facing.
Dependent variable: The log of the bilateral trade import value.
Independent variables:
- Tariff levels (three different rates).
- Product-fixed effects and export-fixed effects.
- Time variables (to account for seasonal changes, yearly trade trends, and monthly fluctuations).
Dataset:
- Monthly trade data for 150 products within an industry, covering trade between three country pairs (A-B, A-C, and C-B) over three years.
- The tariffs only apply to goods traded between countries A and B, and not to trade between A-C or C-B.
- The three tariff levels affect different sets of products at different times.
Research Aim: I want to test whether tariffs on A-B trade have a positive spillover effect on trade between A-C and C-B.
Challenges:
1. Tariff rates for A-C and C-B are zero. Since the tariffs are only imposed on A-B trade, the corresponding tariff rate for A-C and C-B is 0. Should I artificially "apply" the A-B tariff rates to A-C and C-B trade for the analysis? Or would this introduce bias?
2. Inconsistent trade patterns between country pairs:
- Some products traded between A and B are not traded at all between A-C or C-B.
- Occasionally, a product begins to be traded between A-C or C-B only after the tariffs are introduced on A-B trade, leaving no baseline data for comparison.
- I know adding zeros for non-traded products in a log-linear model is problematic. How can I account for these differences in trade patterns?
I was advised to use this code, but it omits the interaction between tariff rates and product codes (HS codes) for A-C and C-B because the tariff rates for these country pairs are zero:
Any advice on how to handle these issues or improve my approach would be so very appreciated!
I am testing the indirect effects of tariffs imposed by country A on imports from country B. Specifically, I want to investigate whether:
1. Country A increases imports from country C as a response to the tariffs.
2. Country C increases imports from country B as part of a trade deflection mechanism (i.e., country B uses country C to bypass the tariffs).
This is my first time working with econometrics and Stata, so I would really appreciate your help with some challenges I am facing.
Dependent variable: The log of the bilateral trade import value.
Independent variables:
- Tariff levels (three different rates).
- Product-fixed effects and export-fixed effects.
- Time variables (to account for seasonal changes, yearly trade trends, and monthly fluctuations).
Dataset:
- Monthly trade data for 150 products within an industry, covering trade between three country pairs (A-B, A-C, and C-B) over three years.
- The tariffs only apply to goods traded between countries A and B, and not to trade between A-C or C-B.
- The three tariff levels affect different sets of products at different times.
Research Aim: I want to test whether tariffs on A-B trade have a positive spillover effect on trade between A-C and C-B.
Challenges:
1. Tariff rates for A-C and C-B are zero. Since the tariffs are only imposed on A-B trade, the corresponding tariff rate for A-C and C-B is 0. Should I artificially "apply" the A-B tariff rates to A-C and C-B trade for the analysis? Or would this introduce bias?
2. Inconsistent trade patterns between country pairs:
- Some products traded between A and B are not traded at all between A-C or C-B.
- Occasionally, a product begins to be traded between A-C or C-B only after the tariffs are introduced on A-B trade, leaving no baseline data for comparison.
- I know adding zeros for non-traded products in a log-linear model is problematic. How can I account for these differences in trade patterns?
I was advised to use this code, but it omits the interaction between tariff rates and product codes (HS codes) for A-C and C-B because the tariff rates for these country pairs are zero:
Code:
reghdfe log_importvalue tariffrate tariff_hscode_interaction time_trend ///
absorb(hscode exp_id) cluster(hscode)
Comment