If the exports are the dependent variable, you should use gravity model and estimate by PPML, the problem with zeros will be solved.

See how to do it here:

Preprint International trade and the gravity model: recent evidence i...

Thanks a lot Carolina Rodrigues Corrêa Ferreira . I will have a look at it!

Dear José, the presence of zeros (or negative or missing values) in data series impede calculations or distort results from CGR (particularly in financial analysis). If you are estimating the CGR annualy, you should select those years with non-zeros, pairwise. If you are using your data to calibrate a model, you could exclude the "real" zeros just by selecting the option in your statistical software. In some cases, however, zeros might be "relevant" for analytical purposes. But they could be handled with standard methods. Best.

Thanks for your advice Luis Rivera !