The use of value-based management in family firms

Although SMEs and family firms are by no means interchangeable, there are certain overlaps, allowing this research to transfer results from SMEs.

Villalonga and Amit (2006) distinguish between “Type I” and “Type II” agency problems. Type I describes fundamental conflicts of interest due to the separation of ownership and managerial control. However, since family members usually serve as executive members or have greater awareness of executives’ actions, family firms are likely to require less monitoring than publicly listed firms are (Jensen and Meckling 1976; Prencipe et al. 2014). Consequently, Type I agency conflicts are minimized. Type II refers to the conflict between controlling (family) owners and less influential minority owners. Controlling owners could therefore use their position and information advantage to exploit minority owners through shirking or free-riding (Wang 2006).

Considering the sample size of 46 firms, our study sits well between qualitative case studies of around a dozen studied firms (e.g., Marriott and Marriott 2000; Graves and Thomas 2008) and large studies with hundreds or thousands of observations (e.g., Kober et al. 2012; Neubauer et al. 2012). Studies with a sample size comparable to ours (e.g., Gul 1991; Dávila and Foster 2007) also conducted quantitative analyses based on surveys. With respect to sample size, they indicated the limited generalizability of the results meant that they should be viewed with caution, and so do we.

One respondent reported working in a corporate communications role. With respect to the size of the firm (the sample’s second largest turnover), it is highly likely that this person had insights into the firm’s decision-making process and the metrics used in the firm. Another indication is that the person scored high on value-relevant metrics.

Because our survey gathers precise details such as annual turnover, the number of full-time employees, and total assets, and nearly all of our respondents were representatives of a family firm, we refrained from inquiring about further personal information to ensure the respondents’ anonymity. Otherwise, we would have risked an increased number of incomplete queries or a reduced number of responses. We therefore cannot provide clear details on the respondents’ average age or gender. However, the respondent’s age or experience appears to be less informative than data on the executive board of the family firm.

The variables tested include CFROI, EVA, DCF, NPV, WACC, VBM, PERS_LIAB, EXT_SUCC, EXT_MGMT, INTL, EMPLOYEE, FIRM_AGE, CAP_INTENS, and BUSI_EDU. For a detailed explanation, see Tables 3 and 4.

In particular, we surveyed the industry, legal structure, public listing, firm age, employee count, net revenues, balance sheet total, organizational hierarchies, export–import ratio, executive education, and executive age.

We understand “unknown” to mean that respondents do not know how often the measure is used. Because we find that each respondent checked “unknown” for a maximum of two out of 15 indicators, we conclude that the frequency of use is unknown because the indicator is not used at all in the firm. If a respondent knows the usage for almost all of the other indicators, it is unlikely that one indicator is actually used in the firm but the respondent does not know how often. Considering that the respondent checked “unknown” instead of “never”—which could also mean that he or she does not know the indicator at all—we code “unknown” as zero to take care of the different meanings. From an ordinal perspective, not knowing a measure should mean a lower rank than never using a measure. For testing purposes, we also perform our analyses by coding answers with “unknown” as one. In doing so, we assume that “unknown” means that the indicator is not known because it is never used. However, significance levels remain the same with only marginal changes (untabulated). Alternatively, it could be argued that the value of “unknown” should be considered a missing value. This interpretation would further reduce our sample size and push statistical validity to its limits. However, to acknowledge this perspective, we constructed a formative measure (index) instead of the reflective factor for VBM, computed as the sum of the single items. Thus, a value of zero does not influence the index computation. The regression based on this VBM index provides essentially the same results (untabulated).

This procedure is in line with Cavusgil (1980), who considers an export-to-sales ratio of 40% or above to indicate committed involvement in internationalization.

We used the CFROI, DCF, EVA, NPV, cost of equity, and WACC as proxies. Table 3 provides a detailed overview.

Moreover, high-dimensional factor loadings (DCF 0.87; NPV 0.65) provide statistical confirmation that these indicators are equivalent to typical VBM indicators.

Surprisingly, there is a discrepancy in factor loadings between metrics that appear very similar, such as the cost of equity and WACC or NPV and DCF. While we cannot assuredly clarify these deviations, they could be subject to differences in the perceived importance of the single metrics. For instance, WACC constitutes a ubiquitous threshold, whereas the cost of equity is one input variable of WACC. Equally, DCF is an input variable of NPV. Another possible explanation could be that the cost of equity and WACC are determined on an annual basis, whereas DCF and NPV are used more frequently for investment decisions.