Abstract
This study analyses the effects of the 2000–2001 dot-com crisis and the 2008–2009 financial crisis on venture capital syndication. Using propensity score matching analysis, we show that during the two crises, venture capital firms (VCFs) had a lower tendency to syndicate their investments, and the size of the syndicates was smaller. This effect is found to be stronger for later-stage financing than for early stage financing. We explain the lower propensity to syndicate and the reduction in syndicate size by the existence of fewer exit opportunities for VCFs and a lower supply of funds for the venture capital industry. Implications for VCFs and start-up firms are discussed.
Acknowledgement
We thank Enrico Pennings, Sandra Phlippen, and Roy Thurik for their valuable support and comments. We also benefited from the comments and suggestions of seminar participants at the Department of Applied Economics at Erasmus University in Rotterdam (2009), the Tinbergen Institute in Rotterdam (2010), and the IGU Conference in Cologne (2010). We are grateful to Thomson Reuters for being easily approachable for questioning on their data sources. Lastly, the first author thanks the Chair of Technology and Innovation Management (Joachim Henkel) at the Technische Universität München for its hospitality.
Notes
1. This is done for two reasons. First, the impact of the current financial crisis on the VC market was found to be more severe in the United States than elsewhere (Block, DeVries, and Sandner 2011). Second, this keeps legal influences constant. Conditions such as countries' bankruptcy laws and tax policies can severely impact VC activities (Armour and Cumming 2006).
2. See Block, DeVries, and Sandner (2011) for a discussion on the selection of an appropriate cut-off point.
3. Block, DeVries, and Sandner (2011) found that the highest (or lowest) deal volume is likely to occur in June and December (January and February).
4. A possible downside of the nearest neighbour matching is that it may give a poor match if the nearest neighbour is not found within a close range in the propensity score. Kernel matching, on the other hand, uses a weighted average of all cases in the control group, wherein weights relate to the inverse of the distance in propensity score. As we found similar patterns in our results for both matching algorithms, we used the results of nearest neighbour matching.