I have a question related to Fama MacBeth type regressions: I use total stock returns as the dependent variable and various variables (including market beta, size, valuation) as explanatory variables.
My question relates to the time horizon over which the stock return (the dependent variable) is calculated. I've seen papers running this analysis for 1 month, 3 month as well as 1 year while using Newey & West adjusted standard errors to adjust for autocorrelation and heteroscedasticity.
I would like to test time stability of my results and hence even longer time horizons for the calculation of total stock returns. However, I heard from a colleague that there might be an issue with Fama MacBeth when using even longer time horizons like e.g. 3 or 5 years but I can't find anything online. Any hints?