Dear colleagues,
(Apologies for the element of self-promotion, but if you are using PSED and similar panel data sets [as well as GEM] or read/review such research, you may want to know this – and the underlying research is currently in temporary open access)
The PSED-GEM sampling mechanism aims at generating a representative sample of individuals ("nascent entrepreneurs") who are involved in an on-going start-up attempt ("nascent venture") at a particular point in time. As far as we know, it does a pretty good job at that. However, at least since the Delmar-Shane papers in 2003-4, we have been vaguely aware that from a venture-level perspective, the mechanism over samples cases that take a long time from inception to a "final" outcome (when it is no longer "nascent", i.e., the attempt has either been terminated or turned into an operational new business).
But we have not realized how big this bias is, or how to correct for it. The answer to both questions is surprisingly straightforward: The sampling probability of a nascent venture is directly proportional to the duration of its creation process; hence, its weight in the analysis should be inversely proportional to its duration. In other words, a venture that only takes 2 months from inception to becoming "up and running" (usually operationalized as regularly having revenue that covers all costs) should be given 20 times more weight in the analysis than a venture whose corresponding journey takes 40 months.
Jaehu Shim and I discuss this in our recent paper in JBVI: Shorter than we thought: The duration of venture creation processes. One obvious implication is given in the title: the average venture creation journey is not as long and complicated as analyses of unweighted data suggest. In line with this another recent article in the same journal (and using the same, harmonized data set) No particular action needed? A necessary condition analysis of gestation activities and firm emergence concludes that in most cases only a few gestation activities need to be completed before the venture reaches profitability.
However, the more serious issue-considering the magnitude of the sampling bias-may be that if substantive relationships are different for "short-duration" start-ups compared to "long-duration" start-ups, then all venture level, relational analyses will be biased without correction for the length of duration. Therefore, we recommend that in the future, all such analyses using this type of data set be run with weighing, at least as a robustness test.
Global Entrepreneurship Monitor (GEM) data are subject to the same problem. Due its lack of venture-level follow-ups over time, GEM data do not provide a basis for assigning weights to the cases. As a second-best solution, researchers can run the same analysis within countries where panel data is available in order to get an indication whether severe bias is present or not.
The PSED-GEM sampling mechanism also over-samples team start-ups, because each team member represents a chance to become included in the sample. This can be solved by weighing by the inverse of team size.
Note that the above applies to the "modest majority" of (initially) rather simple ventures that dominate random samples of nascent entrepreneurs, and to venture-level, relational analysis. The creation journey for most biotech- and certain other, ambitious start-ups remains long and complicated in most cases, and the PSED-GEM sampling mechanism remains a good instrument for portraying the level of "grassroots" entrepreneurial activity in a country at a particular time.
Our paper should currently be open to anyone until February 6 following this link: https://authors.elsevier.com/c/1WExr8MxvtJukW
If for any reason that does not work, the manuscript should also be available here.
Best Regards,
Per Davidsson
Per Davidsson | Professor | Director & Talbot Family Foundation Chair in Entrepreneurship, Australian Centre for Entrepreneurship Research (ACE) | QUT Business School (Management) | QUT | | https://research.qut.edu.au/ace | Phone: +61 7 3138 2051 | Email: per.davidsson@qut.edu.au | CRICOS No. 00213J
************************************** This message is from ENTREP which is sponsored by the Entrepreneurship Division of the Academy of Management. Please do not post messages with attached files. Commercial messages or spammed messages are not allowed on the list. The use of auto-responder "out-of-office" messages may also lead to your removal from the list. You can manage your subscription options, including joining or leaving the list here:
). Ventures HO!