I am trying to build a gravity model with the Global Value Chains indicator (backward and forward participation) as dependent variables. However, I wonder which are the diagnostic tests I should implement to correctly specify the model and to decide the estimator. I know that the literature recommends using the Pseudo Poisson Maximum Likelihood (PPML) estimator in case of trade because of the larges amount of zeros and log-variables issues. However, Global Value Chains indicators do not contain a considerable amount of zeros values and they are expressed in percentages (hence, I will not take the log). The literature does not provide many clues and I hope you will help with this dilemma.
Any suggestion is warmly appreciated Best regards