Help Needed with PPML Gravity Model for Multidimensional Panel Data (Agricultural Trade)

[Yassine touzani]

Hello professor Joao Santos Silva,👋

I’m working on a gravity model to estimate the effects of SPS (Sanitary and Phytosanitary) measures imposed by France, Spain, and the UK on Moroccan agricultural exports (fruits, vegetables, etc.). My dataset is multidimensional, including 15 different product categories over time, and I’m using yearly data for Morocco’s top 3 trading partners.

I’ve heard that PPML is the most appropriate method for this case. However, I’m struggling with the correct way to prepare my data in order to estimate.

Here are the specifics of my setup:

• Country level: Morocco vs. France, Spain, UK
• Product level: 15 agricultural product categories (HS codes)
• Time dimension: Yearly data
• My model contains; a dummy variable for the existence of SPS notifications for a specific product in each year, and also a variable that catch the intensity of each notification, in additional to other control variables...

What I need help with:

1. How should I implement fixed effects in my model?
   But I’m also considering product fixed effects due to the multidimensional nature of the data.
2. I understand that for PPML, the dependent variable should not be logged. Does this apply to all explanatory variables as well, like distance, GDP, and population?
3. Any tips on handling the multidimensional panel structure (countries, products, and time) efficiently in Stata knowing that i'm new in Stata :/ ?

I’ve read recommendations to check out Piermartini and Yotov (2016) and to reach out to you, the creator of the PPML package in Stata, but I would appreciate any additional advice from you.

Thanks in advance for any insights you can share!
Best regards
Yassine Touzani.

[Joao Santos Silva]

Dear Yassine touzani,

1. I would use fixed effects at the sectoral level, as discussed here.
2. You can log the regressors.
3. You mat estimate a single model for all the data, but it is worth considering to do the estimation sector by sector.

Finally, I suggest using many more countries, if at all possible.

Best wishes,

Joao

[Yassine touzani]

Dear Professor Joao Santos Silva ,

Thank you very much for your golden answer and for taking the time to provide such valuable guidance. I truly appreciate it.

At this point, I find myself pressed for time as I work toward completing my master's final project. Adding more countries, as you suggested, would certainly improve the robustness of my analysis, but I’m concerned about the feasibility of doing so given my time constraints. However, if you think it’s absolutely necessary, I will make every effort to incorporate additional countries.

Alternatively, I’m considering focusing on just one category of products—such as vegetables—but expanding the number of countries. Would that be a suitable compromise in your view?

Additionally, I’ve applied the same SPS regulations across EU countries since the notifications tend to be harmonized. In this context, would including more countries outside of the EU add meaningful variation, even if the demand from those countries is low or trade flows with them are relatively weak? Or should I prioritize countries with stronger trade flows?

Lastly, do you think using unit value prices of each product, as an endogenous variable could effectively capture the impact of SPS measures and the associated costs of compliance? I would greatly appreciate your thoughts on this.

Thank you once again for your invaluable help.

Best regards,

Yassine Touzani

[Joao Santos Silva]

Dear Yassine touzani,

If this is for a master's, I think you need to discuss these issues with your supervisor as they have a better idea of what is expected. In any case, I do not have a strong preference for any of the alternatives you suggest, but I do not understand the bit about using prices.

Best wishes,

Joao

[Yassine touzani]

Much appreciated mr Joao Santos Silva,

Prices was proposed in the first place by my supervisors as a relative prices between morocco and its competitors exporting to Europe, but due to the complexity of the method, i preferred working only on morocco in order to capture the weight of these measures e.g Sps as a trade charges for Moroccan exporters...

best regards,

Yassine Touzani.