Does ICT adoption improve access to credit for small enterprises?

UniCredit is a leading European commercial bank with an international network spanning 50 markets, with more than 8400 branches and almost 147,000 employees (FTE). The Bank has branches in all Italian regions and operates in 17 European countries, so it benefits from a strong European identity, an extensive international presence, and a broad customer base. Its strategic position in Western and Eastern Europe gives the Bank one of the regions highest market shares.

According to the European Commission, SMEs are those having <50 employees. In Italy, at the beginning of 2012, about 99 % of enterprises were in this category. This percentage was quite uniform across the provinces, ranging from 98.9 % for Milan to 99.8 % for some Southern provinces, with a mean of 99.4 % (Italian National Institute of Statistics—ISTAT).

On the importance of business plans as instruments to attract prospective sources of finance, see Mason and Stark (2004), Barrow et al. (2001), Kuratko and Hodgetts (2001), Shepherd and Douglas (1999) and Mason and Harrison (1996).

See for example the case for ‘SMEloan’ described by Claessens et al. (2002).

For the definition of business process, see Laudon and Laudon (2010).

All small businesses in the sample are customers of the Bank.

We consider the average credit growth as a dependent variable (instead of the level of the credit accorded) in order to utilize its time-varying dimension and to create a ‘scale-free’ measure of ‘lending premium’, which takes into account the average amount of new credit accorded during the period (2010–2012) regardless of its type (i.e. short–medium–long term). As a robustness check, in Sect. 5, we replicate our analyses using the natural logarithm of short-term credit as a new dependent variable, in the same spirit of Moro and Fink (2013).

To our knowledge, other than the Bolivian register employed in Ioannidou and Ongena (2010), there is no comprehensive register that contains both borrowers’ identities and individual loan performance without imposing any threshold. See Ongena et al. (2012) and Jiménez et al. (2014) for a discussion.

The Bank of Italy only collects data on a single credit position (i.e. the credit a firm has from an individual bank) that amounts to at least 30,000 euros. Thus, there is no possibility to separate missing values by whether firms have been denied credit, did not ask for loans, or asked for loans lower than 30,000 euros. Consequently, in Sects. 4.1 and 4.2 we perform the econometric analysis by relying on a missing at random (MAR) assumption for the firms not included in the final sample of 2120 firms, whereas in Sect. 4.3 we relax this assumption by estimating sample selection models (Heckman 1979) involving the whole original survey sample of 6000 small businesses (reduced to 5830 due to missing information on one covariate; see details in Sect. 4.3).

Alternately, using too many dummy variables in a regression model has the potential problem of perfect multicollinearity (the so-called dummy variables trap). To assess the severity of this potential problem, we report the variance inflation factor (VIF) for each column of Tables 6 and 7, which indicates that, on average, the level of inflation of the standard error of our estimates due to collinearity is quite low, and well below the conventional threshold value of 10.

The average credit growth in the years 2011 and 2012 is explained by the firm’s actions taken in the 3-year period before the survey, that is, from about May 2009 to May 2012.

We perform these regressions using only the ICT_index and ICT_investment as independent variables. Regressions with individual ICT elements as dependent variables are available from the authors upon request.

In this robustness analysis, we only report results for the ICT_index and ICT_investment variables. Results related to the individual ICT elements are available upon request.

The identification of the system of Eqs. (2a) and (2b) is possible without imposing any ‘exclusion restrictions’ in the set of variables X, because of the nonlinearity of the inverse Mills ratio (Heckman 1979).

http://​www.​dps.​gov.​it/​it/​arint/​Base_​dati_​comunale/​index.​html. The instrument reference period perfectly matches the timing of the survey (USM). Reports for former periods are not publicly available. The mean and standard deviation of digital_divide in our final sample of 2120 observations are 0.105 and 0.155, respectively.

The reason why long-term credit is not accounted for is that it could be the result of a past lending decisions, perhaps taken in a completely different context (Moro and Fink 2013), a fact that may also raise reverse causality problems in the empirical context.

In Italy, the banking system is central in business financing due to a lack of fully developed alternative channels (e.g. capital market, corporate bond market, private equity market). In addition, small and local banks specialize in serving the small business segment, whereas big and international banks more easily provide credit to medium and large enterprises. For instance, Bentivogli et al. (2007) show that in 2005 local banks provided more than 27 % of small firms’ total bank lending, which is almost double the percentage given to all enterprises (14 %). Moreover, small banks belonging to groups finance almost 11 % of total credit to small enterprises, while providing large businesses with 8.2 % of their total lending. In contrast, big and international banks supply 71.5 % of medium and large firms’ total credit (compared to 68.7 % of all enterprises).

Fully comprehensive regression output available upon request.