Recurrence of absolute-difference chains (Q1103974)

Let \(Y_ 1\), \(Y_ 2\),... be a series of i.i.d. random variables with distribution Q. The absolute-difference Markov chain \(X_ 1\), \(X_ 2\),... defined as \(X_{n+1}=| X_ n-Y_{n+1}|\) is shown to be dissipative provided E \(Y^{1/2}<\infty\) even if Q is a non-arithmetic distribution and E \(Y_ 1=\infty\).