Should I use PPML or PPMLHDFE for my panel data analysis?

[Alexis Laub]

I am conducting a panel data analysis to examine the impact of renewable energy subsidies on electricity generation from sustainable sources. My dataset consists of 120 countries from 2011 to 2020, and my dependent variable is the amount of electricity generated from renewable sources (measured in TWh).

I plan to use a Poisson Pseudo Maximum Likelihood (PPML) estimator due to the presence of zero values in my dependent variable and potential heteroskedasticity issues. However, I am unsure whether to use PPML or PPMLHDFE.

My model includes several control variables, such as GDP per capita, CO₂ intensity, electricity imports, financial development, HDI, FDI, population ratio (15–64), and female ratio. Additionally, I need to control for country-fixed effects and year-fixed effects to account for unobserved heterogeneity.

Given this setup, would PPMLHDFE be the better choice due to the need for multiple high-dimensional fixed effects? Or would standard PPML with manually included fixed effects be sufficient?

Furthermore, are there any alternative estimators I should consider for this type of panel data?

Thanks in advance for your insights!

[Joao Santos Silva]

Dear Alexis Laub,

Both commands should give you the same results, but ppmlhdfe is much faster, so I would recommend it (even though I wrote ppml!).

Best wishes,

Joao

[Alexis Laub]

Dear Joao Santos Silva

I will ! Thank you for your quick response !

Regards

Alexis