SELFDECOMPOSABILITY PERPETUITY LAWS AND STOPPING TIMES∗

In the probability theory limit distributions (or probability measures) are often characterized by some convolution equations (factorization properties) rather than by Fourier transforms (the characteristic functionals). In fact, usually the later follows the first one. Equations, in question, involve the multiplication by the positive scalars c or an action of the corresponding dilation Tc on measures. In such a setting, it seems that there is no way for stopping times (or in general, for the stochastic analysis) to come into the “picture”. However, if one accepts the view that the primary objective, in the classical limit distributions theory, is to describe the limiting distributions (or random variables) by the tools of random integrals/functionals then one can use the stopping times. In this paper we illustrate such a possibility in the case of selfdecomposability random variables (i.e. Lévy class L) with values in a real separable Banach space. Also some applications of our approach to perpetuity laws are presented; cf. [2], [3], [4]. In fact, we show that all selfdecomposable distributions are perpetuity laws. 1. Let E be a real separable Banach space. An E-valued random variable (rv) X, defined on a complete probability space (Ω,F,P), is said to be selfdecomposable (or a Lévy class L) if for each t> 0 there exists a rv Xt independent of X such tha