Im working on a model examining the effect of minimum wages on employment over the 10 provinces using a panel data set. I have used the Hausman test to decide to use RE or FE, in which it recommended RE. However, my concern is that is may not be a appropriate assumption that the 10 provinces (error term) are a random sample from a larger population which the RE model assumes. Thoughts? should I use the FE instead?