Forecasting European high-growth Firms - A Random Forest Approach
Jurij Weinblat ()
Additional contact information
Jurij Weinblat: University of Duisburg-Essen
Journal of Industry, Competition and Trade, 2018, vol. 18, issue 3, No 1, 253-294
Abstract High-growth firms (HGFs) have aroused considerable interest both by researchers and policymakers mainly because of their substantial contribution to job creation and to the advancement of the surrounding economy (Acs et al., Small Bus Res Summ (328):1–92 2008, Schreyer 2000). Any initiative to foster HGFs requires the ability to reliably anticipate them. There seems to be a consensus in previous mainly regression-based studies on the impossibility of such a prediction (Coad, Doc Trav Centre d’Econ Sorbonne 24:1–72 2007b). Using a novel random forest (RF) based approach and a recent data set (2004–2014) covering 179970 unique firms from nine European countries, we show the potential of a true out-of-sample prediction: depending on the country, we were able to determine up to 39% of all HGFs by selecting only ten percent of all firms. The RF algorithm is both used to determine relevant predictors and for the actual prediction and pattern analysis. Both the selection of the best RF and the cross-country comparisons are based on a Receiver Operating Characteristic analysis. We find that most accurate HGF predictions are possible in GB, France, and Italy and largely confirm this ranking using Venkatraman’s unpaired test. Apart from the firm’s size, age, and past growth, the sales per employee, the fixed assets ratio, and the debt ratio are quite important. Our “typical” HGFs determined using RF prototypes have been older and bigger than the remaining firms, which is counterintuitive and atypical in literature. Based on our finding, typical HGFs are not start-ups, which questions current political funding strategies. Apart from that, our results do not support and rather refute the existence of a survivorship bias. Moreover, approximately every fourth HGF remains to be a HGF in the next period.
Keywords: High-growth firms; Random forest; Forecasting; Variable importance (search for similar items in EconPapers)
JEL-codes: M13 L25 O51 O52 C53 (search for similar items in EconPapers)
References: View references in EconPapers View complete reference list from CitEc
Citations: Track citations by RSS feed
Downloads: (external link)
http://link.springer.com/10.1007/s10842-017-0257-0 Abstract (text/html)
Access to full text is restricted to subscribers.
This item may be available elsewhere in EconPapers: Search for items with the same title.
Export reference: BibTeX
RIS (EndNote, ProCite, RefMan)
Persistent link: https://EconPapers.repec.org/RePEc:kap:jincot:v:18:y:2018:i:3:d:10.1007_s10842-017-0257-0
Ordering information: This journal article can be ordered from
http://www.springer. ... on/journal/10842/PS2
Access Statistics for this article
Journal of Industry, Competition and Trade is currently edited by Karl Aiginger, Marcel Canoy and Michael Peneder
More articles in Journal of Industry, Competition and Trade from Springer
Bibliographic data for series maintained by Sonal Shukla ().