Innovar

Innovative Activities in Family Firms

Existing literature on family firms has studied the impact of firms' ownership on innovation activities following different theoretical drivers such as Stewardship Theory (Craig & Dibrell, 2006), Agency Theory (Lodh, Nandy & Chen, 2014; Zahra, 2005), the socioemotional wealth (Gomez-Mejia et al, 2007) or the rbv (Carnes & Ireland, 2013; Zahra et al, 2004). However, innovation in family firms is a research area pending an in-depth analysis, since literature does not provide conclusive results (Carnes & Ireland, 2013; Penney & Combs, 2013).

From the rbv, we examine the conflictive link between family firm ownership and innovation activity, which was described above in order to integrate and extend both research areas. This idea is supported by the fact that in family firm research areas empirical studies have mainly analyzed the role of innovation in public family and non-family firms without a particular focus on a specific industry (Craig & Dibrell, 2006; Craig & Moores, 2006; Kellermanns et al, 2008; Lodh, Nandy & Chen, 2014; Zahra et al, 2004). Given that this idea may partially support the mixed empirical results about the link between family ownership and their innovation activity, we propose to conduct our empirical study on a specific industry in which innovation is decisive and core.

Therefore, the rbv allows us to study and identify relevant factors or variables in family firms which lead them to innovate and thus gain a competitive advantage and higher incomes. Moreover, thinking of innovation as a key element for firms' success leads us to rbv, which points out that we must research those factors or variables that better explain firms' final results.

Variables

In order to study and identify variables in family firms where innovation is core, we have made a literature review looking for variables and measures commonly used to study the link between family ownership and innovation activity, both in family firm research area and/or in rbv approach. In general, in literature about family firms, empirical studies use 'innovative activity' as a dependent variable and 'family ownership' as an independent one. Therefore, we have used these two variables to characterize the innovative firms of the sample. However, the following variables are also regularly used as control variables according to literature: firm size, age, past financial performance and liquidity. Consequently, we have decided to include them in the analysis to test whether they are significant in characterizing the firms.

Additionally, we have decided to introduce the variable 'entrepreneurial risk taking' for the following reasons: innovations, along with the role of the entrepreneur, are some of the most important factors to build competitive advantages (Barney, 1991; Hill & Jones, 2008); the role of the entrepreneur is important since is the person developing the innovative activity of a firm and, therefore, making decisions to take the entrepreneurial risk of innovation (Craig & Dibrell, 2006); and because in family firm research area some papers suggest to study the linkages between innovation and entrepreneurial risk-taking (Zahra et al, 2004). These reasons let us think there could be a linkage between ownership, the innovative activity and entrepreneurial risk-taking, being this variable also relevant to characterize the firms in the sample.

Accordingly, we have identified the following variables and measures:

Innovative activity: Firms' innovative activity is measured by the r&d expenditure/sales ratio, which is proportional to a firm innovation commitment. This ratio is a common measure of a firm's input to innovation process and allows comparison between firms (David, Hitt & Gimeno, 2001; Hitt et al, 1996; Kim, Kim & Lee, 2008; Renko, Carsrud & Bránnback, 2009).

Entrepreneurial risk-taking: Entrepreneurial risk-taking is usually measured by the debt/equity ratio since this measures financial risk. When a firm has an important debt it is said it has a high financial risk. Comparing companies that have taken risk investing in capital, the debt/equity ratio is greater in capital-intensive industries than in noncapital intensive industries (Zahra, 2005). In general, the suitable data for this index should consider a function of the profitability of the company, as well as the index of the industry.

Family ownership: A business is considered a family firm if both of the following conditions are met: first, two or more directors have a family relationship and, second, family members must hold a substantial proportion of equity (Gomez-Mejia, Makri & Larraza, 2010; Jones, Makri & Gomez-Mejia, 2008). In our study the 'family ownership' variable is a dummy variable (1/0) measured by the percentage of firm's stock held by the owner family. We have imposed a 10 per cent ownership threshold to ensure that the owner family held a substantial percentage of the firm's equity and at least two members of the board were family (Muñoz-Bullón & Sanchez-Bueno, 2011; Zahra, 2005).

Firm size: Sathe (2003) states that larger firms might resist change and innovation because these companies subject ideas for radical innovation to iterative reviews that stifle entrepreneurial ventures. Zahra asserts that "larger family firms have well-established connections within and outside their industries, making it possible for them to join strategic alliances and intensify entrepreneurial activities" (2005, p. 32). Therefore, firm size might have an effect on innovative activity and entrepreneurial risk-taking. To measure firm size, we use firm's total assets (Baysinger et al., 1991; Zahra et al, 2004).

Firm's age: Following Cairncross (1992), in the mature stage of life firms have the resources to exploit opportunities and introduce new products and services. Therefore, firm's age also has an effect on innovative activity and entrepreneurial risk-taking. To measure firms' age, this study uses the number of years the firm has been in existence (Zahra, 2005; Zahra et al., 2007).

Past financial performance: A successful past performance may affect innovative activity and entrepreneurial risk-taking because managers could reduce their wish to face entrepreneurship. If the firm is doing well managers could have no incentives to disrupt the status quo. Nevertheless, a successful past performance also provides resources that could encourage managers to carry out entrepreneurial activities. To measure firms' past performance, we use the firm's average return on assets (roa) over the preceding three-year period (Kim, Kim & Lee, 2008; Zahra, 2005).

Liquidity: Previous research has found that levels of liquidity could influence the amount of funds the firm has available in the current period for developing innovative activity and entrepreneurial risk-taking. A common measure of liquidity is the ratio: current assets/current liabilities (Baysinger & Hoskisson, 1989; Hitt et al. 1996).

Empirical Analysis, Sample and Methodology Sample

As it was noted, we propose as the objective of this paper to look for a profile of innovative firms and analyze if ownership, concretely family ownership, is a characteristic related to their innovative activity. To reach this, we have used a sample drawn from sabi database owned by Bureau van Dijk. This database contains comprehensive information about Spanish and Portuguese companies, such as financial statements, financial strength indicators, directors and contacts, original filings/images, detailed corporate structures, audit report, etc. We selected Spanish firms who are classified in cnae code 7211. cnae is the National Classification of Economic Activities compiled according to the conditions set out in the nace Regulation (statistical classifications of economic activities developed since 1970 in the European Union). The code 7211 groups the firms whose activity is the experimental r&d in biotechnology.

Our data comprise financial data up to December 31^st 2013. As result we obtained 243 firms. In order to discard those cases that could invalidate the results of the empirical analysis (e.g., cases with too much missing data) we followed the methodology proposed by Lehmann, Gupta and Steckel (1998), and Santesmases (2009). In addition, we eliminated outliers.

Hierarchical Cluster Analysis

To apply the hierarchical method we have chosen Ward's proposal, using the Euclidean distance squared as a measure (Hair et al., 1999). As a point of reference to choose the most appropriate number of clusters we have used the agglomeration coefficient. When two very different clusters are joined it produces a very large coefficient or a high percentage of change in the coefficient. We look for large increases in the coefficient in order to stop the fusion of groups. We can see in table 1 considerable increases in passing from one number of clusters to the next number.

Table 1: Agglomeration coefficient analysis for the Ward's method.

Source: Own elaboration.

As seen in table 1, the higher agglomeration coefficients are for 3, 2 and 1 clusters. These coefficients show significant increases when going from 3 to 2 clusters (1,215.630 -1,012.608 = 203.022) and from 2 to 1(1,435.000 - 1,215.630 = 219.37). Therefore, if clusters 3 and 2 join we would be mixing very different groups, which would be detrimental to their internal homogeneity. The same would be applied if clusters 2 and 1 join. Besides, in order to identify agglomeration coefficients with large increases, in the same table we also calculated the percentage changes when going from 6 to 2 clusters. As we can see, the increase in the percentage of change when going from 2 clusters to 1 is large (18.05% = {(1,435 - 1,215.63)/1,215.63)}x100) and the change that occurred when going from 3 to 2 clusters (20.05%) is also remarkable.

Then, we identified those atypical individuals in the solutions of 2 and 3 clusters obtained by Ward's method. Table 2 shows the scheme agglomeration for the last 4 steps (from step 198 to 199) of the analysis. This information is useful in identifying unique individuals which were joined to the grouping process late (e.g. potential outliers).

Table 2: Agglomeration scheme from step 198 by Ward's method.

Source: Own elaboration.

Table 2 shows in the last three columns the steps in which each one of the clusters is formed. Thus, a company that never has joined to a group has the 0 level. As can be seen, in step 203 (3 clusters) a new cluster is derived from merging a cluster with a unique firm and one cluster that had arisen as others join in step 193. In sum, the selection of 3 or less clusters eliminates the problem of the firm which had not previously joined any group.

Two-step Cluster Analysis

In this second phase of cluster analysis, the two-step method has been used because it allows segmenting a sample simultaneously using categorical and continuous variables. To proceed with this method, the variables are standardized since they are measured in different scales. Log-likelihood is chosen as measure of distance because there are continuous and categorical variables. This measure assumes that continuous variables follow a normal distribution while a multinomial distribution is adopted by the categorical variables, and that both are independent. However, the procedure works reasonably well although these assumptions are not met.

Furthermore, with this method, a fixed number of clusters in which it is desired to segment the sample of individuals may or not be specified. In this case, we first perform the analysis using the statistical program in order to determine the number of clusters automatically, and subsequently, we proceed with a specific number of three clusters. The optimal number of groups obtained automatically with the Schwarz Bayesian criterion has been 2, while using the Akaike information criterion has resulted in 4. Therefore, the recommended number of clusters is between 2 and 4, which supports the decision of segmentation in 3 clusters taken from the previous analysis following Ward's method.

Table 3 shows the profile of the variables used for segmentation in three clusters. The last column exhibits the distribution of clusters: the largest cluster is the number 1 with 149 companies, followed by number 3 with 40 and number 2 with 17.

The previous profile analysis reveals the following sequence. Cluster 1 (149 companies) shows lower values than the glo bal average of the sample for all continuous variables, except past financial performance. The 149 companies comprising such value are 'non-family' firms. Cluster 2 (17 companies) shows higher values than the global average of the sample on continuous variables: innovative activity, entrepreneurial risk-taking, size, age, and liquidity, as well as a lesser value in the past financial performance variable; moreover, of the 17 companies that form this group, 16 are 'non-family' firms and 1 is family owned. Finally, cluster 3 (40 companies) shows higher values than the global average of the sample for age and past financial performance. The size variable approaches the average despite being lower. Additionally, the 40 companies that compound the cluster are all family owned.

Table 3: Profiles of the averages of the variables used in obtaining the solutions of three clusters (two-step method).

Source: Own elaboration.

The anova contrast is also developed in table 3, showing the significance of the difference between the averages of continuous variables with respect to the sample average (Snedecor's F). As shown in the table, the difference is significant for the 0.05 level. Therefore, the null hypothesis of the anova analysis of no difference between the averages observed for these variables is rejected. Finally, the percentage of the total variance of the sample explained by the segmentation in two clusters (R²) is 90.4%.

Characterization and Typology of Spanish Biotech Firms

As we have mentioned above, table 3 shows three clusters with different profiles considering the variables selected to develop the method of two-step analysis. Following, table 4 illustrates averages or categories for each variable. Additionally, the average or frequency of each variable, the rate of change on the average of the sample, the number of companies in each category and an interpretation of these results or cluster description are displayed for the three groups. These data allow to get a general profile of the Spanish biotech companies regarding the main variables which characterize them.

Firms in the First Cluster

The 149 Spanish biotech firms that shape the first cluster are characterized by the following factors:

• They are less innovative compared to the average of the sample (6.05). That is because their rate of innovation (2.64 on average), measured as R&D in 2010 divided by sales in 2010 (variable 'innovative activity'), has a rate of change of -56.36% compared to the global average.

• They assume little risk compared to the average of the sample (2.29). This is because their risk-taking index (1.56), measured by the debt in 2010 divided by equity ratio in 2010 (variable 'entrepreneurial risk taking'), shows a variation rate of -32.04% with respect to the average of the sample. As already mentioned in section 3, in general, the suitable data for this index, which measures the financial risk of a firm, should be a function of the profitability of the company and of the referencing index of the industry.

• They are small business (the smallest of the three clusters) following the classification made by the European Commission in its recommendation C (2003) 1422 about the amount of assets. The average amount of assets of the firms that make up this group (2,575,864.46 € in 2010) exceeds 2M. € but does not reach 10M. €.

• They are middle-aged companies (7.81 years of media from their foundations until 2010), although they are the youngest in the sample, following the classification made by Berger and Udell (1998). According to this classification, firms with ages between 5 and 24 years are middle-aged.

• Their past financial performance measured as the average return on assets (net income/total assets) for 2008, 2009 and 2010 can be considered quite bad (-5.59). This is due to the fact that while it is negative, as for the average of the sample (-10.7), is a 44.52% not as bad.

• Joint liquidity of Spanish biotech firms in the first cluster, measured as current assets divided by current liabilities for the year 2010, is high. The reason is that this variable shows an average of 10.74 when good values are generally accepted between 1.5 and 2. This could mean that firms have excessive liquidity resources which would reduce their profitability. The index average for the sample also shows an excessive value (36.18); however, the average of the companies within this cluster is better, since it is 70.32% lower than the total average.

• Firms in this group are non-family businesses because, as previously shown in table 4, all companies in cluster 1 (149) belong to this category.

Table 4: Description of the clusters obtained by two-step cluster analysis procedure.

Source: Own elaboration

Relation between Family Ownership and Innovative Activity

A summary of relevant information is outlined in table 5, which presents the typology of Spanish biotech firms. The cluster analysis shows the relation between family ownership and innovative activity as follows:

• Most of the Spanish biotech companies (149 firms) are non-family firms that develop little innovation activity. In addition, they take the least risk of the sample and are the smallest and youngest firms of the industry, although they are middle-aged, they also have a quite irregular past financial performance and high liquidity (cluster 1).

• The most innovative firms (17 firms) are non-family firms. They take the highest level of risks within the industry and are medium-sized. They are also the biggest and oldest firms and report the worst past financial performance and the highest liquidity ratio (cluster 2).

• The least innovative firms (40 firms) are family firms. Apart from taking few risks they are small and middle-aged firms, reporting a bad past financial performance and high liquidity; however, these two variables are the least negative of the industry (cluster 3).

It seems that innovative Spanish biotech businesses are medium-sized non-family firms which take a high entrepreneurial risk, whereas the least innovative are small family firms taking little entrepreneurial risk. There are other features that characterize each cluster, such as age, past financial performance and liquidity; these are least determinant in characterizing each group given that, in general, all firms in the sample are middle-aged with a negative past financial performance and a high liquidity ratio.

Considering the before mentioned, looking for a profile of innovative firms, from the analysis in tables 4 and 5 we can conclude that in the sample of Spanish biotech firms there is a relationship between family ownership and innovative activity, given that family firms develop the least innovative activity (40 in cluster 3), whereas non-family firms (166 in cluster 1 and 2) develop the highest innovative activity or a medium-low innovative activity. Therefore, this result clarifies the profile of the firms of the sample and corroborates the stream in family firms research area mentioned in the introduction, stating the negative link between family firm and innovation (Carney, 2005; Naldi et al., 2007; Classen et al., 2014).

Table 5: Typology of Spanish biotech firms based on their characteristics.

Source: Own elaboration.