When are extreme ratings more helpful? Empirical evidence on the moderating effects of review characteristics and product type