“…With no prior knowledge regarding the probability distribution that would best represent the data, a Poisson distribution was assumed. This choice followed the initial specification used by HS, and is a common starting point in the estimation of a count-data relationship (Gourieroux, Manfort and Trognon, 1984;Hausman, Hall and Griliches, 1984;Cameron and Trivedi, 1986;Davies and Guy, 1987;Okoruwa, Terza and Narse, 1988;Congdon, 1989Congdon, , 1993Loomis, Creel and Park, 1991;Ruser, 1991).…”