Business models, intangibles and firm performance: evidence on corporate entrepreneurship from Italian manufacturing SMEs

The AIDA Bureau van Dijk database is an authoritative source of information on Italian companies. Information is drawn from official data recorded at the Italian Registry of Companies and from financial statements filed at the Italian Chambers of Commerce. Limited Liability Companies furnish data on a compulsory basis. The information provided includes company profiles and summary financial statements (balance sheet, profit and loss accounts and ratios). Each company's financial statement is updated annually. Additional information on the AIDA Bureau van Dijk database can be retrieved from http://​www.​bvdinfo.​com.

The ATECO classification is the Italian coding based on the NACE classification of manufacturers of wearing apparel. We included the segments that best describe the clothing industry. The description of each code is: 14.11 Manufacture of leather clothes, 14.13 Manufacture of other outerwear, 14.14 Manufacture of underwear and 14.39 Manufacture of other knitted and crocheted apparel.

The industry structure consists mainly of small- and medium-sized firms located in districts. The industry is highly fragmented: the mean Herfindahl index for the four industry sectors is 0.004, whereas the manufacturing industry mean is 0.06.

This database has both strengths and weaknesses. Among its strengths, as mentioned in the Introduction, it contains information that enriches the secondary data and allows for a deeper longitudinal analysis. Among its weaknesses, the balanced panel does not permit us to control for sample selection bias, as we ignored the exit and entry of firms (Cameron and Trivedi 2005). However, this should not be a major problem because we are interested in the influence of BM changes on performance. We recorded 94 BM changes in 376 firms during the 10 years under analysis.

We relied on a single-variable matching procedure to isolate the specific drivers of the BM change, i.e. profitability or the position of the company in the value chain, instead of using a single score that captures all the information for selection from the (observable) pre-event firm characteristics, as in standard propensity score-matching models.

Only firms with a full record of data were considered. In the first case, three, and in the second case, seven BM changes were not analysed because we were not able to find a matched company within our sample.

We used 1-year lagged independent variables to avoid reverse causality.

Sales growth is calculated as (Sales _t −Sales _t−1)/Sales _t−1, and ROS is calculated as Net Income (Before interest and tax)/sales. Finally, we used a simplified definition of TFP as log (value added) −0.40 × log (capital) − 0.60 × log (employees), where the factor weights are the cost shares estimated in the sample (Bloom et al. 2011).

This taxonomy has been built and validated by means of cluster analyses and other multivariate analysis on the most comprehensive databases on SMEs in the clothing sector managed by the Italian Ministry of Economy (Annual Industry Revenue Survey on Companies in the Textile and Clothing Sector—Studi di Settore). The definition of each BM is based on a grid of variables that captures the most relevant dimensions of BM in the industry. It refers to choices, activities, resources and capabilities, which provide a unique identification of the company’s organisation structure: the positions attained and maintained by the firm within the industry; the markets in which it competes (e.g. role within the vertical contracting structure of the industry, degree of internationalisation, customers’ portfolio, etc.); the activities it performs to attain and maintain these positions (scale of operations, nature and scope of activities, etc.); the resources and capabilities that enable it to perform these activities (technologies, people, etc.) and the relationships among these elements (for details, see Camuffo et al. 2008 and Pozzana 2011).

To check the consistency between our subjective classification of BM innovation and the objective evidence emerging from company information, we estimated an ordered logit model using the three types of BM changes (low, medium and high) as a dependent variable and a set of independent variables normally used in the literature as indicators of entrepreneurial behaviours. These 1-year lagged proxy variables were as follows: introduction of a new product (indicator of innovation), retained earnings (indicator of proactivity) and leverage (indicator of risk taking) (Miller and Le Breton-Miller 2011). Estimated results show that the intensity of the innovation inherent in the BM change is positively and highly correlated with these entrepreneurship proxies. Results are omitted for the sake of brevity, but are available from the authors upon request.

As our sample does not include firm entry and exit, the reported evidence is likely to underestimate the actual change that occurred in the industry structure. Indeed, the evidence of a massive flight to quality would have been even more robust if we had considered the very large number of low-quality/low-cost producers who left the market after the introduction of the Euro.