The long-run underperformance of initial public offerings: a methodological problem?