Conditions under which a naive algebraic solution to a random affine
stochastic equation yields a genuine stochastic solution with a given independence
property are derived. The well-known example by Letac is characterized and
generalized to a non-trivial dependence structure. As by-products, a series
representation of the exact probability density of a sum of independent gamma
random variables is recovered, and a new elementary proof of Lukacs’
characterization of the gamma distribution is added.