Impact of inter-firm cooperation on company's performance: a comparative analysis of European Union and Russia

In the approach which was offered by Hausman and Taylor, in turn, it is assumed that although part of the xit and zi variables is correlated with бi, there are variables that are not correlated with individual effects (see Equation 2). (Hausman and Taylor, 1981)

where; - the dependent variable; - the intercept; - vector of exogenous time variant characteristics uncorrelated with , -vector of endogenous time variant characteristics correlated with and uncorrelated with ; -vector of exogenous time invariant characteristics uncorrelated with , -vector of endogenous time invariant characteristics uncorrelated with , but correlated with ; -an idiosyncratic i.i.d. error term; -unobservable random effect.

Then, it is clear that the columns of variables xi which are uncorrelated with бi can perform two functions:

a) obtaining unbiased estimates of the coefficients;

b) acting as an instrument for zi which are correlated with бi.

An important advantage of the Hausman and Taylor's approach is that this method does not rely on strict a priori assumptions and under certain conditions makes it possible to test correlation between бi and regressors. So, we conclude that Hausman-Taylor method should be used in order to escape the biased or inefficient results, as in there is a suspicion of endogeneity of such factors as qualification of director's board, investments in intangible assets and number of patents. Cooperation itself is also neither spontaneous nor exogenous. (Luo, 2005) We discuss these variables in more detail in the next section.

3.2 Variables

We have chosen econometric modelling as a central research approach of this paper for testifying stated hypotheses. So, we need to discuss which variables, both dependent and independent ones, will be included in the regression model.

Dependent Variable. EVA (Economic Value Added) was chosen as a dependent variable because nowadays this indicator may be considered as one of the most universal value indicator of business performance. EVA is defined as a performance measure that indicates the creation of additional value and its main advantage over traditional financial performance metrics such as net income or earnings per share is that it reflects profits only after the reduction of capital costs. This helps to see the realistic picture of company's performance through taking into account risks and environment. (Stern et al., 1995) Moreover, EVA can be calculated not only for companies which are listed on the market, but for private companies as well. There is also a number of papers that proved the advantages of EVA measure over other financial indicators. (Biddle et al., 1997) For instance, Rogerson (1997) studied the use of EVA as a tool for measuring the performance of the company, and his main conclusion was that EVA has an advantage over such measures as ROA and ROE.

The basic formula of EVA calculation was introduced by the creators of this indicator and looks as follows: difference between return on invested capital (ROIC) and weighted average cost of capital (WACC) multiplied by capital employed. (Stern et al., 1995) So, to maximize this indicator, and, hence, financial performance, the net profit should be the highest and the average cost of capital should be the lowest. In this regard, increasing the growth of EVA with effective asset management comes down to the following strategic objectives of the financial strategy:

a) increasing of profitability;

b) optimizing costs;

c) minimizing cost of capital;

d) optimizing capital structure.

However, as we have already mentioned, not only tangible assets may contribute to the generation of added value. This idea determined the set of independent variables which are listed next.

Inter-firm cooperation variable. This indicator reflects the fact of firm's participation in some long-term strategic relationships in their general sense. As we were limited in applying an interview or questionnaire method and has no opportunity to access and use special database as SDC Platinum which authors of previous research used (see Table 2) we developed a system of indicators, based on the survey results, in order to proxy a predisposition of a company to join inter-organizational relationships.

According to our model the involvement of the company in inter-firm relationships can be determined, firstly, by the presence of the subsidiaries. (Schartinger, 2003) Previous works suggest that internal links are prerequisites for external integration. This argument is proved by the fact that companies who could overcome impediments to develop a wide network of internal linkages, namely, subsidiaries, have less barriers and uncertainties on the way to formation of external inter-firm partnerships. (Flynn et al., 2010)

Secondly, a high quality website may be also considered as an indicator of the possible presence of the inter-firm cooperation as it may reflect the company's attractiveness to potential partners.

The fact that for firms the main source of information about potential partners is the Internet was proved by the survey of 168 Russian companies conducted in 2006.

This survey showed that in 57% of cases managers consider the Internet to be the best way to find connections. (Popova, 2010)

Presence of at least two of the following four characteristics listed above, in our opinion, should have a positive impact on the creation of the firm partnerships:

a) multi-language choice. This option significantly increases the attractiveness of a potential partner, particularly in connection with the processes of globalization. If the company is willing to spend the resources and provide information about themselves and their activities in a foreign language, it is likely that its efforts will pay off in the form of established alliances with foreign partners;(Shaw and Holland, 2010)

b) section for "Investors." This section usually provides information about the performance of the company, plans, description of the competitive advantages, etc. If a company contemplating a desire to start a close relationship with another company, but, for example, is not yet officially stated that willingness, the information on the web-site can help to fasten the process;

c) filling content. The site containing 10 or more pages and links will likely be more meaningful that the resource, in which all information is presented on 3 or 4 pages and limited to the minimum. Potential partners will be able to create the most complete image of the company;

d) design. This factor indirectly determines the company's willingness to use completely new and innovative web-technologies. For example, within the framework of inter-firm co-operation, such a technology may be to create collaborative online products, including web sites, whose task is to facilitate cooperation by managing the workflow and support the exchange of ideas and suggestions. (Chen, Zhang, Zhou, 2007)

Thirdly, we assumed that membership in informal business associations like Trade Unions or Industry Associations also determine company's participation in formal inter-firm cooperation types like alliances, networks and so on. (Gulati, 1998)

Based on the above model, we generated a binary variable, which equals to 1 if company has some established long-term relationships, and takes value of zero if there is no evidence of long-term cooperation tendencies in the particular company.

This factor is assumed to be time-invariant, as we consider in the analysis only those companies which have been participating in the relationships did not change their decision.

It is worth mentioning that this variable is a subject of endogeneity problem as a number of characteristics which determine participation in partnerships is much higher than was already mentioned, which includes both with internal and external factors of each particular company. (Nguyen, 2011)

These factors are generally unobservable and, hence, the issue requires additional research.

Control Variables. We include variable "Current Assets" in order to control the degree of firms risk aversion: if company is conservative in current assets management, than its Economic Value Added increases, as risk reduction is usually results in the decrease in market beta, cost of debt and equity. (Michalski, 2008) This, in turn, decreases WACC which has inverse relation with EVA. If a company implements the aggressive strategy of current assets, management the situation is vice versa.

Such indicators as qualification of director's board, investments in intangible assets and the number of patents are also among the independent variables. They were included as it has been previously proved that these factors, representing the companies' intangible assets, significantly influence EVA.(Shakina and Barajas, 2012) Moreover, it should be noted that these variables may be considered endogenous as they connected with such unobservable factors as human intellect and attitude of a company to innovative actions. Among these variables only qualification of director's board is assumed to be time-invariant because of the special characteristics of database collecting process. In addition, the model will include three more variables: the size of a company (number of employees) and location in capital city and location near university. It is important to include these parameters as it was previously showed in empirical research that financial efficiency differed between firms of different size and for companies functioning in different regions. (Chen et al., 2005; (Gumbau-Albert and Maudos, 2009) Location is assumed to be time-invariant. We also add country specific dummy-variables in model for European Union countries (Germany, Italy, France, Spain, the UK) in order to catch the effects which arise because of economy and country specificity.

Thus, the general framework of our model looks as follows:

3.3 Survey Sample

The dataset used for the empirical analysis consists of the information about large companies operating in the European Union, namely, Germany, the UK, France, Italy and Spain, and companies from Russian Federation. These regions were chosen as they are characterized by different economy and financial environments, talking more precisely, they are ranged differently by the stage of the development. Such countries as Germany, the UK, France, Italy and Spain are considered to be advanced, while Russia is rated as developing. (International Monetary Fund, 2014) This, in turn, may directly influence companies and cause dissimilarity of processes inside these organizations including the direction and pattern of inter-firm cooperation impact on financial performance, it is relevant to make a comparative research rather than just study one country. (Chen and Lin, 2006; Kongmanila and Takahashi, 2009; Lee et al., 2013) Database includes information only about large and quoted enterprises as these firms publish their financial statements and provide latest information for shareholders/investors which are publicly accessible.

While constructing the sample we used Amadeus (for EU companies) and Ruslana (for Russian ones) search platforms ran by Bureau Van Dijk as well as other publicly available sources. [79, 81] Firstly, we have filtered all companies which have been operated in the period between 2004 and 2011. The reason to choose such period lies in our research interest to study inter-firm relationships as a driver for performance not only in the framework of different countries, but considering stages of economy cycle. Then we have chosen the set of parameters which we need for our analysis. On this step database included 1028 European companies, 639 Russian companies and 5 indicators: EVA, Number of Employees, Intangible Assets, Current Assets, Age of a Company, Number of Subsidiaries for each year.

In order to collect other indicators we used the following sources:

a) Web-site quality indicator was collected with the help of companies' web-sites;

b) trade-union organizations members lists were used to collect Participation in Business Associations indicator;

c) data about Number of Patents was collected through the Orbit® search engine; [80]

d) Companies' financial reports were used to collect information about Qualification of Boards of Directors, Location in Capital City and Location Near University.

Additional information about calculation methods and special features of these variables may be found in Appendix 1.

3.4 Data Description and Analysis

Before developing regression models and testing hypotheses, the comprehensive analysis of dataset should be conducted, as we have to understand what characteristics of the data are, whether it is representative, homogeneous and free from mistakes and outliers.

Poolability test. In order to check the representativeness of the sample, it is necessary to verify if the data for separate countries from European Union may be analyzed as a common pool reflecting the whole European Union. In order to do this, we built Hausman-Taylor regressions for each of 5 countries and compared the sign and significance of beta coefficients for each country.(Cameron, 2005) In Appendix 4 it is shown that almost all included variables affect EVA with the same direction regardless of the country where company operates. This means that we may combine five presented countries in one dataset, and we also can consider this set of countries to be representative for the whole European Union as the sum of their GDP makes up more than 70% of EU27 cumulative GDP (Appendix 5).

Representativeness. Next, we tested the representativeness of the sample in terms of how it reflects the general population of companies in Russia and EU. We can realize it by calculating what percentage of general population of companies is presented in the dataset and checking if the industrial structure of the companies presented in the database is compliant with the companies represented in the relevant countries' markets.

According to the statistics provided by national statistics agencies, namely, Rosstat in Russia and Eurostat in EU, there were 25704 large and medium enterprises in Russia and 20156 ones in European Union (27 countries) in 2010. (European Commission, 2009) It means that Sample covers 2.5% and 5 % of general population of large and medium enterprises in Russia and EU, respectively. Moreover, the comparison of industrial structure of general set of enterprises and research sample structure resulted in no evidence of significant difference (more than 15%) between proportions of particular industry in the total number of companies in general market and dataset. (See Appendix 7 and 8) This implies that both Russian and EU sub-samples are representative and findings of empirical research conducted with the use of the collected dataset may be transferred to the whole country or region.

Outliers. We also examined the sample for the presence of mistakes and outliers (that are defined as data points which do not follow the general trend of the rest of the observations). Liquidation of such observation may help to make sample more homogeneous and elude estimation bias (Wooldridge, 2012) We first got rid of observations for which the values of Intangible assets were negative as it contradicts with the nature of this indicator.

Numbers from 1 to 6 correspond the following countries: France, Germany, Italy, Spain, UK and Russia, respectively.

Figure 4 Boxplot analysis of outliers

Boxplot analysis showed that there are some outlier observations for EVA indicators, in order to make sample more homogeneous, we limited EVA value in the class boundary (-5000;5000), which led to the removal of 33 companies. Following the same logic we excluded the companies where there are more than 300,000 employees working as they also may be considered to be outliers. After all these procedures, the sample was grown down to the 885 observations in EU part and 616 companies from Russia. The descriptive statistics of all variables we are interested in are presented in the Appendixes 2 and 3.

Analysis of the dependent variable. On this stage we investigated the quality of the developed participation in the relationships classification model. We used such proxy-indicators as web-site quality, membership in the business associations and presence of subsidiaries in order to define if the company is a participant of formal cooperation agreements with other firms. Secondly, using sample massive of firms we conducted the analysis of distinctive characteristics possessed by companies participating in cooperation depending on country and industry basis.

Table 3

We also analyzed if the same tendencies are seen in the sample by graphing median EVA for European and Russian companies year by year. We see that EVA was also growing during 2004-2007, plummeted in 2008-2009 and started to recover in 2010 and 2011. Hence, we divided the sample by three periods which are 2004-2007 for the growth period, 2008-2009 - the crisis times and 2010-2011 - the recovery stage in order to test the second set of hypotheses.

Figure 7 Median EVA values for European companies

Figure 8 Median EVA values for Russian companies

Summary for the Research Design and Methodology Sections:

Research Design and Methodology sections were developed in order to prepare for the empirical analysis of inter-firm relationships impact on company's performance. We have studied what are possible ways of conducting research of this type and found that econometric modelling, namely, Hausman-Taylor approach, is the most suitable one. We also proved that it is essential to simultaneously study inter-firm cooperation in Russia and EU as they possess miscellaneous characteristics due to the differences in the development stage and, thus, cooperation may also have diverse pattern in these two regions. Then we developed two sets of hypotheses, discussed empirical database and describe variables which will be included in models designed to test these hypotheses. So, next step is to realize the empirical analysis of inter-firm cooperation in Russia and EU and discuss the results.

4. Results

This section is devoted to the empirical testifying of stated hypotheses using model and sample described in previous section.

4.1 Regression Model

The analytical form of Hausman-Taylor model developed in the framework of current paper looks as follows:

+

Where ; vector includes time-variant exogenous indicators as Current Assets, Age, Number of Employees; - time-variant endogenous parameters including Number of Patents, Intangible Assets, - Location in Capital City, Location near University which are time-invariant exogenous, - time-invariant endogenous Qualification of Directors Board, -an idiosyncratic i.i.d. error term; -unobservable random effect. Models for EU companies also includes country dummy. In order to estimate it we used STATA 11.0 software.

4.2 Estimation Results and Hypotheses Validation

We started with estimating the set of models for European Union sub-sample. In general, all models were built on the sample of 823 companies and are significant on the 1 % level with Wald Chi-2 (13 degrees of freedom) statistics equal to 881, 120, 270 and 51 for the full panel, growth, crisis and recovery time-panels, respectively.

We observe that inter-firm cooperation variable positively influences EVA in the full panel (1% level of significance), in the growth period (10% level of significance) and in crisis times (5% level of significance). In the recovery period there is no evidence of statistically significant link between inter-firm cooperation and financial performance reflected by EVA.

So, we may conclude that Hypothesis 1a is fully supported by the empirical model, while hypothesis 2a needs additional discussion due to the fact that though in growth period inter-firm cooperation indeed influences EVA less than in 2008-2009, we cannot compare the coefficient with recovery period as it is not significant in the model.

Estimation results also showed that current assets influence EVA variable positively and statistically significant regardless of time-period, but with different intensity. Number of employees indicator has a negative influence on EVA in all time intervals except crisis period.

Number of patents parameter also has no significant influence on EVA in crisis times, but shows positive sign in full panel, growth and recovery periods. Intangible assets significantly and negatively influence EVA in the full panel and during 2004-2007, but the sign changes in 2008-2009 and 2010-2011 sub-samples.

Directors' qualification impact on financial performance indicator is significant and positive in the full panel and in crisis times (10% level of significance).

Location of the company near university has a significant and positive impact on EVA only in crisis period.

Table 5