Agency costs of free cash flow, internal capital markets and unrelated diversification

More recent references may be found in Bausch and Pils (2009).

Researchers generally identify two main legal traditions: common law (basically, the law system in Anglo-Saxon nations) and civil law. Within civil law, several subtraditions are generally identified: French, German, Socialist, and Scandinavian. For details on the origin and the consequences of this legal system see La Porta et al. (1998).

Highly leveraged firms can increase their capacity to meet scheduled debt payments by diversifying their businesses, particularly in activities that are far from their core operations. The unrelated diversification of highly leveraged firms allows the creditors of these firms to rely on the combined fortunes of all of the conglomerate firm’s operating units, reducing the variance of future cash flows.

In 2010, only 332 companies were traded on the Milan stock exchange (the number of traded firms was 297 in 2000, 266 in 1990 and 168 in 1980).

Details on these aspects are provided in a book by Silipo (2009) that examines the structural evolution of the Italian bank system and the effects of this evolution on the Italian economy.

Rajan and Zingales (1995) explain that G7 countries differ from each other with respect to the roles of banks and securities markets, bankruptcy codes, and the markets for corporate control systems. In fact, Paligorova and Xu (2012) note, “Although the G7 countries are similar in terms of economic development, their institutional frameworks are different”.

Ric&St provides a detailed balance sheet analysis, which is complemented by a profile of the company’s history and operations, the names of its directors and major shareholders, statistics on its production and market share, details regarding its production facilities, sales and employees, and, in the case of listed companies, its stock market performance.

Financial companies are excluded from this study’s analyses. Moreover, observations that were missing any of the necessary data for this study’s analysis were eliminated from the sample, and the study did not assess entities that were involved in “abnormal” financial situations (i.e., Parmalat, Cirio and Alitalia).

A firm’s listed status certainly has important consequences for the relationships that are examined in this study. However, the sampling of unlisted firms does not reduce the ability to interpret the study results because the listing status of the firms is accounted for as a moderating variable in the study’s subsequent analysis.

Two related papers, Gourlay and Seaton (2004) and Berry-Stölzle et al. (2012), use this methodology to examine diversification determinants in the manufacturing and insurance industries, respectively. Other investigations that apply Cragg’s model include Burton et al.’s (1996) study of UK household meat expenditures, the Jensen and Yen (1996) report regarding US food expenditures away from home, and the Basile (2001) investigation that examines innovations that affect the export behaviours of Italian manufacturing firms.

We used the Italian classification codes (ATECO). For example, a firm with a business division that operates in Paper and Allied Products and another business division that operates in Textile Mill Products has two distinct 2-digit industry codes and therefore demonstrates unrelated diversification.

We also include three dummy variables (D_crisis, D_’90 s, D_’00 s) in both the status and the extent equations, and we use industry dummies because macroeconomic and industry uncertainties are found to influence unrelated diversification. All of the explanatory variables are lagged by one period in an attempt to reduce potential endogeneity issues.

Ownership concentration is very stable in Italy, rarely showing average values of less than 55 %. The sample composition indicates that the Italian business model is primarily based on family firms (which constitute almost 59 % of the sample) (D_Family). In addition, only 38 % of the sampled firms are listed (D_Listing). It is well known that certain large Italian companies, such as Barilla, are uninterested in the Bourse and prefer to remain unlisted.

In a supplementary analysis, we find relevant industry differences with respect to diversification strategies and explanatory variables, suggesting the need to consider these heterogeneities in regression analyses. Firms, which have diversified by expanding into unrelated industries, are more common in the construction industry. Approximately 64 % of the sampled firms in this industry have diversified by expanding into unrelated industries and these firms have a mean entropy index of 0.49. Firms in the Petroleum Refineries and Related Product industries (which have mean values of this metric that equal 26 %) hold more cash, whereas firms in the Manufacture of Metal Products industry have higher debt levels (a mean debt level value of approximately 52 %). A complete table of descriptive statistics by industry (mean values) is available by a request to the authors.

The maximum VIF that results from any of the models is 1.42, which is far below the generally employed cut-off of 10 (or, more prudently, 5) for regression models.

This result supports the views of Amihud and Lev (1981) and May (1995) that shareholders with higher equity ownership can be expected to diversify into unrelated industries to reduce their idiosyncratic risk.

The study includes supplementary analyses. As a robustness check, the study estimates the model on a year-by-year basis, obtaining results for all of the years that are qualitatively similar to the results in Table 5. The analysis also applies different measurements for the variables Cash Flow and Debt. In particular, the calculation of net income plus depreciation and amortisation, less changes in working capital and capital expenditure, is used as an alternative measure of free cash flow, and the total liabilities divided by the total assets are considered an alternative measure of the debt ratio. The findings (which are not reported) do not change if these alternative independent variables are used.

Similar to the definition used by Ampenberger et al. (2012), we classify a firm as a family firm if at least one of the following three conditions is satisfied: (1) the firm’s founder or his or her family participates in the ownership of the firm; (2) the founder’s family participates in the firm’s board of directors; or (3) the founder’s family participates in the management of the firm.

The agency costs of free cash flow may also occur in related diversification. Markides and Williamson (1996) indicate that the attributes required for related diversification are value-enhancing. Nayyar (1992) suggested that relatedness can fail to create value when the involved business units lack cooperation, communication and incentives, which generate impediments to the exploitation of value in related diversification. The role of agency arguments in related diversification is also highlighted by Jones and Hill (1988).

Furthermore, similar to the subgroups analysis of Table 6, other criteria are applied to sort firms by different sensitivities to financial constraints and agency problems. In accordance with Hadlock (1998), for firms sensitive to the agency costs of free cash flow, a high ownership dummy (high ownership concentration vs low ownership concentration, based on median value) is used as an alternative classification criterion. Consistent with Audretsch and Elston (2002), for financially constrained firms, a size dummy (large firms vs small firms) is also applied. Overall, the results of these analyses (which are not reported) are qualitatively similar to the results presented in Table 6. To assess the robustness of the findings, as Hoetker (2007) suggests, the probit models of Table 6 are a re-estimate that accounts for the critical econometric problem involving the comparison of coefficients between groups in binary-choice models. The differences in coefficients across groups (identified by the four dummy variables interacted with cash flow and debt) may not be realistic due to differences in the residual variabilities for these groups. As Allison (1999) and Williams (2009) suggest, a heteroskedastic probit model allows for differences in the residual variations and the resulting false differences in the coefficients between groups to be detected and addressed. The empirical findings of this test between unrelated diversification and interaction terms (which are not reported) demonstrate results similar to those in Table 6, suggesting that differences in variable coefficients do not reflect the degree of residual variation between groups.

The diversification discounts have been shown to lessen, to disappear, or to become premiums in recent literature in consideration of the factors that cause firms to diversify and in the treatment of the decision to diversify as endogenous (e.g., Campa and Kedia 2002; Graham et al. 2002).

For reasons of brevity, these results are not reported, but are available by request to the authors.