Necessary conditions for the existence of conditional moments of stable random variables (Q1890698)

Let \((X_ 1, X_ 2)\) be an \(\alpha\)-stable, \(0 < \alpha < 2\), random vector. It is known that if a finite measure on the unit circle \(S_ 2\) associated with the vector \((X_ 1, X_ 2)\) satisfies the condition \(\int_{S_ 2} | S_ 1|^{-\nu} \Gamma (ds) < \infty\) for some \(0 < \nu < \alpha + 1\), then \(E[| X_ 2 | ^{\alpha + \nu}/ X_ 1 = x] < \infty\) a.e. It is shown that this sufficient condition is also necessary in the cases \(0 < \nu < \alpha + 1\) if either \(0 < \alpha \leq 1/2\) or \(1 < \alpha \leq 3/2\), \(0 < \nu < 2 - \alpha\) if \(1/2 < \alpha \leq 1\), and \(0 < \nu \leq 4 - \alpha\) if \(3/2 < \alpha < 2\).