Analysing innovative output in a CIS database: taking account of some nasty details

Recent analyses of Community Innovation Survey (CIS) data suffer from methodological shortcomings. Using Finnish CIS data, we illustrate the potential consequences of these shortcomings and propose ways of dealing with them. We show how the neglect of lag structures between dependent and independent variables has certain consequences on the results. It is also important to check and identify any differences in the typical lifecycle durations between industries. Finally, the precise definition of ‘innovators’ to be included in the estimate also has a significant impact on the results, as does the non-normal distribution of shares in the sales of innovative products. Earlier studies have not been too explicit on some of these problems.