Stock split ex-dates are associated with both an increased intensity of small investor buying and a positive abnormal return. The broker promotion hypothesis suggests that the increase in relative spread after a split induces brokers to promote splitting stocks to small investors. The trading inconvenience hypothesis ascribes the ex-split effects to inconveniences such as investors' aversion to dealing with due bills, which is unrelated to relative spreads. The reduction in the bid-ask spread due to decimalization allows us to disentangle these two hypotheses. During the 1/8th pricing period, we show that after the ex-date, the relative spread increases significantly. The average buy order size decreases and the frequency of small transactions increases after the split. After decimalization, these changes are smaller in magnitude. We observe significant positive abnormal returns around the ex-date during the 1/8th pricing period, but not in the decimal pricing period. These results support the broker promotion hypothesis.