An extension of a functional equation of Poincaré and Mandelbrot (Q700540)

Let \(N\) be a nonnegative integer-valued random variable, let \(A\) be a nonnegative random variable, and consider an unknown random variable \(Z\) satisfying the distributional equation \[ Z\to \sum^N_{i=1} A_iZ_i, \tag{1} \] where \(N,A_i,Z,Z_i\) are independent to each other, each \(A_i\) is distributed as \(A\), each \(Z_i\) is distributed as \(Z\). In terms of the characteristic function \(\varphi(t)= Ee^{itZ}\), this equation reads \[ \varphi(t)= f\bigl(E\varphi (At)), \tag{2} \] where \(f(s)=\sum_{n\geq 0}P(N=n)s^n\). If \(0<A=1/EN<1\), then (2) reduces to the Poincaré functional equation for Galton-Watson processes, while if \(N\geq 2\) is constant, (2) becomes the Mandelbrot functional equation for multiplicative cascades. The author first gives necessary and sufficient conditions for (1) to have nontrivial and nonnegative solutions. Then asymptotic properties of \(\varphi\) and its derivative are studied. If (1) has a positive solution, the author investigates conditions under which the distribution function \(P(Z\leq x)\) decays at a polynomial rate or exponentially as \(x\searrow 0\). The results he obtains are applied to multiplicative cascades, Bellman-Harris processes and branching random walks.