If individuals derive a small utility from gambling, we should observe high turnover in stock portfolios that are of only marginal importance to them. By the use of detailed individual financial data, as weIl as trades from a Swedish online broker, we measure the frequency and cost of online trading in the cross-section and reject this hypothesis. Investors who have online portfolios that constitute a large share of risky assets are more likely to trade, trade more aggressively when they do trade, have lower trading performance, and less wealth. Trading losses are therefore mainly carried by those who can afford to carry them the least.