Ending the Mending Wall: Herding, Media Coverage, and Colocation in IT Entrepreneurship

One potential cause for concern, and an inherent weakness with the naïve matching methodology described in the paper, is that the method considers only the year, industry, and funding reception on the part of the entrepreneur. To the extent that structural components of the VC syndicate (i.e., size and relative reputation), or observable characteristics of the entrepreneur, may correlate with the willingness of the VC to back the new venture, this undermines both the quality of the match and the extent to which the matched sample is balanced across the covariates with the original set of realized ties. To remedy this potential concern, we execute two additional matching strategies. The first, propensity score matching (PSM), is a well-established method in both the economics (Caliendo and Kopeinig 2008; Dehejia and Wahba 2002) and information systems (Brynjolfsson et al. 2003; Smith and Telang 2009) literatures. The second, coarsened exact matching (CEM), offers a more nuanced matching method for contexts where matching based on categorical variables may not provide adequate balance (Blackwell et al. 2009; Iacus et al. 2012). Specifically, while traditional matching methods increase homogeneity between the treatment and control groups in the aggregate, CEM offers the benefit of balancing each individual variable as well as the aggregate to ensure that heterogeneity in changes in the matching variables are consistent across the both the treated and control groups, thereby ensuring both multivariate and univariate balance. Unsurprisingly, this methodology is becoming increasingly popular in the social sciences (Azoulay et al. 2013; Greenwood and Wattal 2017).