Investment and Usage of New Technologies: Evidence from a Shared ATM Network

S. Ferrari, F.L. Verboven, H.A. Degryse

Research output: Working paperDiscussion paperOther research output

369 Downloads (Pure)


When new technologies become available, it is not only essential that firms have the correct investment incentives, but often also that consumers make the proper usage decisions. This paper studies investment and usage in a shared ATM network. Be- cause all banks coordinate their ATM investment decisions, there is no strategic but only a pure cost-saving incentive to invest. At the same time, because retail fees for cash withdrawals are regulated to zero at both branches and ATMs, consumers may not have the proper incentives to substitute their transactions from branches to the available ATMs. We develop an empirical model of coordinated investment and cash withdrawal demand, where banks choose the number of ATMs and consumers decide whether to withdraw cash at ATMs or branches. We find that banks substantially underinvested in the shared ATM network and thus provided too little geographic coverage. This contrasts with earlier findings of strategic overinvestment in networks with partial incompatibility. Furthermore, we find that consumer usage of the avail- able ATM network is too low because of the zero retail fees for cash withdrawals at branches. A direct promotion of investment (through subsidies or other means) can improve welfare, but the introduction of retail fees on cash withdrawals at branches would be more effective, even if this does not encourage investment per se.
Original languageEnglish
Place of PublicationTilburg
Number of pages45
Publication statusPublished - 2008

Publication series

NameCentER Discussion Paper


  • technology investment
  • usage
  • efficiency
  • ATM
  • empirical industrial or- ganisation and structural econometric models


Dive into the research topics of 'Investment and Usage of New Technologies: Evidence from a Shared ATM Network'. Together they form a unique fingerprint.

Cite this