In France, political observers and politicians pay considerable attention to public opinion polls, using them as indicators of who will win the upcoming election. Before the 1997 French legislative contest, the polls consistently forecast a win for the ruling Right party coalition. To almost everyone's surprise, they were wrong. We document the extent of their error, then speculate on why it occurred. Finally, we propose a political economy model as an alternative, and more accurate, means of forecasting French legislative elections.