In Scotland and Wales, as in Quebec, California, Oregon, and the United States in general, there is a fairly close relationship between the amount of money spent by a party in a constituency campaign and the number of votes which it receives. Since most of that expenditure is on printing and stationery, the result suggests that party advertising buys votes. Parties are somewhat rational in attempting to buy votes where they are most important, but neither they nor the electorate are wholly rational in their actions. A great problem in all non-experimental social science is how to evaluate models of processes from cross-sectional data only.V. Gray. 'Models of comparative state politics: a comparison of cross-sectional and time-series analyses.' American Journal of Political Science 20, 235-256, 1976. This applies in the present case. If, instead of looking at static patterns, the changes in spending and voting between elections are analysed, then the findings are much less clear-cut and there is no substantial evidence that more advertising wins more votes (or less loses votes). The analyses reported here suggest the need for much more careful investigation of the rational political man hypotheses. Parties, it seems from the analyses here, spend most where they win and where contests are close, but there is little evidence that their spending 'buys' them more support.