The purpose of the business cycle surveys is to under-stand as quickly and easily as possible recent and probable economic activity developments. Most of the questions asked generate one of three qualitative answers (calling a trend upward, stable or downward). The qualitative indicators provided by these surveys are summarised in the form of balances of opinion, which represent the difference between the percentage of «upward» answers and the percentage of «downward» answers. Yet such indicators only really summarise the answers if the percentages of «stable» answers are relatively constant. Another approach is to summarise the answers to the business cycle surveys in an indicator that, unlike the balance of opinion, takes account of the percentage of «stable» answers. This indicator is constructed using a principal component analysis (PCA), which shows the linear combination of the three percentages that captures the most variability between the successive surveys. Although the two «summaries» of the answers are very similar for certain questions, they differ considerably for other questions and could result in different economic analyses. An application to the forecast of quarterly growth in manufacturing production, using the data available at the time of forecasting, shows that the indi-cators constructed by a PCA provide slightly more accurate forecasts on average than those provided by the balances of opinion, even though they generally concur. This work shows this to be a productive line of research.