Innovative Mechanisms in Territorial Industrial Systems – Western Pomerania Case

Abstract The major objective of the paper is to show the results of research that was made in industrial enterprises of the West Pomeranian region. The main objective of the research was an attempt to find the variable determinants that have an impact of the relationships among enterprises on their innovative performance. The research was conducted within regional industrial systems and the basic aim was to define the constraints for a model regional structure of innovation network tailored to the needs of Poland and one of its regions.


Introduction
The dynamics and system character of innovation have been so far described in theoretical approaches within the evolutionary or Neo-Schumpeterian economics. In those concepts an innovation process on the level of a single enterprise is perceived as a system of activities which are related by means of feedback, whereas innovation is a result of an interactive learning process which involves usually several actors from within and beyond the enterprise 1 .
Innovation and its diffusion become a result of an interactive and collective network process of personal and institutional changes evolving over time. They respond to the challenges of the "new economy" in the region such as globalisation and acceleration of technological progress thus creating an opportunity for economic development in underdeveloped regions.
Innovation systems have become the main theme of numerous theoretical and empirical studies over the last 15-20 years. This approach is focused on determinants of the development and diffusion of process and product innovations 2 . Its essence is the relationships between the internal and external players in the region 3 . The findings provide evidence that manufacturing enterprises are more successful if they are elements of an integrated intensive network.
The operation of systems is based on interactions between individual participants of the network. These relationships can be of either a vertical or horizontal character. Due to the complex character of the subject, this paper concentrates only on the "output" relationships, i.e. the relations with customers of products manufactured by an industrial system.
Modern regional networks aim at diversifying relationships through initiation of interactions with various groups of customers. In a traditional environment such relationships should focus on specialised interactions. It seems interesting to identify whether the innovative performance of regional systems in Poland is determined by diverse or by narrow interactions, based on strong and lasting or weak interpersonal relationships, the distance between partners being close or long.
The conceptual framework outlined above has inspired the author to address the problem of the impact of enterprises on innovativeness of regional industrial systems. The major hypothesis of the study is that innovative mechanisms inherent in territorial industrial systems and their relationships with the environment are largely determined by the character of relationships among enterprises. They include: the type of competitors, suppliers and customer, their localisation and character of the relationship. Those factors determine the present form of industrial systems in Poland. An appropriate identification of the course and constraints of the innovation process in the national system of economising is a basis for the construction of diversified development Unauthentifiziert | Heruntergeladen 11.02.20 18:36 UTC paths for innovation networks, allowing for the national and regional features and accelerating the processes of creation, absorption and diffusion of technology.
The major objective of the research was an attempt to find the variable determinants of the impact of the character of relationships among enterprises on their innovative performance within regional industrial systems and hence to define the constraints for a model regional structure of innovation network tailored to the needs of Poland and its regions. The research results presented in this study represent only one finding. From the viewpoint of sampling, the author decided to analyse the case of one region representing medium-weak industrial development. Such a solution allowed a more in-depth analysis of the features characteristic of regional industrial systems in Western Pomerania region.
The research was based on a questionnaire distributed among 447 enterprises. The basic method to acquire this amount of data involved an initial phone interview followed by sending the actual questionnaire by e-mail.

Methodological conditions of the research -the probit modelling
The methodological part of the analyses is based on the probability calculus. When a dependent variable takes dichotomous values, the possibilities of using the popular multiple regression, widely used for quantitative phenomena, are limited. The problem can be solved by an alternative solution -the logistic regression 4 . Its advantage is that an analysis and interpretation of results are similar to the classical regression method, hence the methods of selecting variables and testing the hypotheses have a similar pattern. However, there are also differences which include: more complex and time-consuming calculations and producing the residual plots usually do not contribute significantly to the model 5 . In a model where the dependent variable can equal either 0 or 1, the expected value of the dependent variable may be interpreted as a conditional probability of an event at given independent values.
The forerunners in using the logistic curve were P.F. Verhulst and R.F. Pearl. A full model was not used, however, until 1994 and 1953 by J. Berkson 6 .
Generally, the logistic regression is a mathematical model which can be employed to explain the impact of several variables X 1 , X 2 , ..., X k on a dichotomous variable Y. If all the independent variables are qualitative, the logistic regression model is equivalent to a log-linear model. To describe such a phenomenon one could also employ the probit regression 7 .
The assumptions common for all those models are as follows 8 : − The data comes from a random sample, − Y can take only two values: 0 or 1, − Subsequent Y values are statistically independent, − The probability that Y = 1 is defined by the normal distribution (NCD) for a probit model or a logistic distribution (LCD) for a logit model, − There is no perfect linear relationship between X i variables (no co-linearity of independent variables).
In the methods with a dichotomous variable, the parameters are estimated according to the maximum likelihood (ML) method. According to its rules, a vector of parameters is searched for, which gives the highest probability of arriving at the values observed in the sample 9 .
Generally, the application of the ML method requires formulation of a likelihood function and finding its extreme value, which can be done in two ways: analytical and numerical. Despite its complex procedure, the ML method has gained popularity since it can be applied to a wide array of models, including models with variable parameters, complex delay structure models, heteroscedastic models, and nonlinear models. The features of the ML method, even for small samples, are in many cases much better than other alternative estimators.
Non-linear estimation comprises six algorithms to find the minimum of the loss function.
It allows arriving at best estimators for a given loss function. Each of those methods uses a different strategy to find the minimum of the function 10 .
The likelihood function for a logit or probit model is maximised by means of the techniques used for non-linear estimation. There are several user-friendly software tools available for logit or probit analysis.
Considering the fact that the variables are binary (i.e. they take two values -0 or 1), the majority of the results will be presented at the level of the structural form of the model. A "plus" sign preceding a parameter denotes that the probability of an innovative phenomenon in the selected group of entities is higher than for the rest of the population. Probit modelling is an efficient research tool in the case of big yet static samples where the dependent variable is qualitative.
As mentioned above, some methodical analysis was based on probability theory. Several A set of adopted independent variables are the reference planes, which reflect the activity of industrial enterprises, adopted on the basis of the methodology commonly used in OECD countries since the 1980s 11 .
The set of characteristics where distinguished that describe innovative activities of industrial enterprises at the input (effort) and output (implementation and cooperation). Simultaneously to this day the synthetic measurement method has not been developed to describe innovation activities at the enterprise level, although its recognition in the system approach appears in the literature 12 . However, they are sometimes criticised because of the heterogeneous nature of this activity and the difficulity in bringing it to a common denominator, the used measurement methods are specific with limited applications. Statistical verification of models was based on Wald's Chi-square statistics. The verification of the significance of the Wald parameters was made using Student's t-test with the asymptotic standard errors of assessment. Adopted confidence limits of the model and its parameters were ±95%. Due to the number of estimated models the authors decided to present only those that met the test of statistical significance -both models as a whole, as well as its parameter (the factor in question).
It is also worth noting that both the dependent and independent variables had the binary nature. This is due to the fact that on the one hand, there was the need to collect a large number of properly completed survey forms -the system survey while on the other hand to simplify of the questions in the survey form as much as possible. This does not change the fact that the showing statistically significant directions of the relationships between the variables adopted for the study. This proved sufficient for the evaluation of the studied phenomena. Based on the probability theory , the chances of particular areas of innovation activity can be estimated and provide some boundary conditions, and thus making it possible to plan and strengthen the effects of the impact of regional innovation policy instruments.
Each questionnaire was entered to the Excel spreadsheet for initial processing based on formal logic. The actual calculations were made with the Statistica software.

Agglomeration economy in a medium-weak region -a case of West Pomerania
As far as the type of customers in West Pomeranian (447 questionnaires) is concerned, there is a variety of interactions representing innovative behaviour in the region.

Development of new solutions is supported by relationships with the whole industry, not
only the area of mining and industrial processing 13 . A positive yet less significant impact can also be observed for power industry and trade industry. Other sectors seem to play a marginal role, nonetheless it is noteworthy that the companies which represent the final link in the production chain, i.e. sell their goods, tend to be less interested in innovative activities.
As we can see in Table 2B       are chances to relate to a large group of domestic suppliers, but based beyond the region, to accelerate innovative processes in the regional industrial system. Nonetheless, the fact that there is a limited number of models for the spatial variable proves the strong diversification in terms of the flow of materials and half-finished products to the regional industrial system. Considering the universality of innovative processes in this region it can be concluded that the significance of mutual interactions with suppliers in the region and their complexity are growing. The proximity to the suppliers, as well as the nature of relations with suppliers, do not destimulate innovative processes, but for customers these elements cumulate. In other words, the local environment as a potential customer of innovative goods remains unfriendly.
The character of an industrial system and its tendency to innovate are conditioned by Vertical relationships with customers clearly indicate that the relationship factor is more relevant than the spatial one. Nonetheless, it should be stated that the number of statistical models is close to that generated for the West Pomeranian Region. An essential condition to encourage the right activity in terms of new products and technologies is a significant distance to customers although in such a case it needs to be accompanied by close cooperation along the production chain. It provides evidence supporting the previously formed thesis about Unauthentifiziert | Heruntergeladen 11.02.20 18:36 UTC the dichotomy of industrial systems in Poland and their close relationships with innovative interregional and even international networks. An internal industrial system, being weak, does not provide proper conditions for dynamic development of regional interactions which become essential to improve innovative performance in the leading group of enterprises, forcing them to incur costs of covering the distance in order to acquire knowledge. Moreover, it should be observed that it requires more than good neighbour relations with the analysed groups of entities; typical (basic) relationships with customers are even more harmful (negative) to the stimulation of innovative activity. It should be also observed that development of an industrial system is accompanied by an increasing number of models describing the analysed phenomena where the parameters are statistically significant. It is a proof for a better transparency of innovative networks described by the selected variables. The role of those factors becomes more and more significant over time.

Conclusions
The regional industrial systems in the West Pomeranian voivodeship analysed in this study reveal an evolution in the approach to innovative activities, considering the character of competitors, suppliers and customers, their location or relationships with them.
Unlike enterprises in the technologically developed countries where innovative activities are focused in regional systems, Polish enterprises are forced to overcome the distance barrier.
Nevertheless, it seems a natural direction for new knowledge which affects the development of national industrial systems. It is also noteworthy that in the strongest case, the regional environment is no longer a destimulant in innovative activities.
Enterprises which are the final link of the production chain are less innovative than those producing for industry. It is an indication of a low technological level of solutions offered and a still insufficient demand pressure which would drive innovation. Regional systems have not yet become mature enough in terms of competitiveness to participate in the dynamic changes based on the technological factor in the international market.
Along with the growing economic potential of Polish regions, there is increasingly stronger diversification of entities following an innovative path. The value of close long-term relationships with suppliers and customers becomes more and more relevant as it raises mutual trust and enables involvement in more risky areas of business. It follows that innovative activity is determined by the existence of repetitive, long-term yet typical interactions, which seems consistent with the results of the research carried out worldwide.