Special Topic: Preparing a “Secondary” Data Set

Summary and Conclusion

I started this chapter with a couple of sad stories about trying to answer research questions with secondary data that were fundamentally inadequate for the purpose. The message was that if the data do not meet the minimum requirements, please refrain from doing the research at all! The main message of the chapter, however, is that you can sometimes do much more with secondary data than you first believe possible—if you invest enough time, effort and money into it. I have tried to illustrate with examples from my own projects the kind of data matching, checking and corrections that may be necessary in developing a useful and credible data set. To be honest with you, after being through the work with these projects I sometimes think it is scary what some other researchers do—and don’t do—with their secondary data sets, and I hope that in the future you will share my skepticism when reading published research where the researchers do not seem to have thorough knowledge about the characteristics and limitations of their data.

Admittedly, our work with Business Dynamics in Sweden and High Growth Firms could have been better, both in terms of theory-driven research questions and in terms of sophisticated analysis methods. However, I hope this chapter has shown that serious use of secondary data involves a lot more than grabbing an existing data set and subjecting it to sophisticated, theory-driven analysis. If one doesn’t invest a lot of time, effort and money in the data set in the first place, the results are likely to be uninteresting or false, regardless of theoretical and (analysis) method sophistication. Therefore, while theoretical shallowness and the use of overly simple analysis methods should also be avoided I think they represent more excusable sins than the use of crap data. GIGO rules!