In the above paper we made an unfortunate error in the application of Bernoulli's inequality in the proof of Lemma A.2 (p. 179). The results of our paper are not affected, but part of the proof has to be slightly changed. First, we need to replace Lemma A.2. {equation presented} This gives the lower bound in Theorem 2.1(iii)(b) since E[(1 + S₁)-γ] ∈ (0, 1]. We are deeply apologetic for this error. Incidentally, proves (∗∗) for all t > 0, and not only for t ∈ (0, 1). Thus, this correction already proves and simplifies the whole argument from page 173 (line 9 from below) to page 174 (line 6 from above).