There is no contradiction. His timeframe is 1 month, not 4 years. I am too lazy to look up what happened 1 month after June of 2003, but if he says on average -1.6% a month later for 10 lowest readings, it still allows for one or more positive performances for the market a month later (out of 10). Also, he never says that the lower the sentiment reading, the more the market should plunge, apparently he thinks the odds are about equal for the 10 lowest sentiment readings.