Strengthening the Innovation Resilience of Polish Manufacturing Firms in Unstable Environments

This study examines factors that increase resilience in innovation of Polish manufacturing firms in an unstable environment. Organizational resilience in innovation is the ability to continuously perform innovation in a turbulent environment and increase knowledge accumulation. In 2008–2012, Poland did not have crisis itself. Short-term slowdown of the economy was accompanied by a breakdown of innovation activities, with a medium-term effect. Based on the Polish Community Innovation Survey panel data for two periods: the innovation crisis (2008–2010), and the innovation pessimism period (2010–2012), this study shows which innovative resources change the probability of innovation continuity in the second period. In our probit model, we explore 42 factors of innovations. We found that financing, R&D and marketing increased the probability of continuity of innovation, but the influence of financing was the strongest. Persistence in innovation in turbulent times hence requires a change in the structure of innovation resources used. Due to the fact that public support on innovation did not increase the likelihood of the continuity of the innovation, a policy change is required. Reliability of our estimation is confirmed by accuracy of prediction of firms, which was 78.2%.


Introduction
The aim of the study is to identify innovation resources and firms' characteristics, that mitigate the impact of external shock on innovation, and strengthen its resilience to improve the continuity of innovation. In literature, the most common focus is the influence of a crisis that has an all-encompassing character. Our research is based on a country that, during the 2007-2009 world crisis, had a decrease in dynamics of GDP growth rather than a crisis. We use a Polish case because it offers a unique opportunity and proves to be an excellent laboratory to test the impact of knowledge resources and firms' characteristics on the continuity of innovation through the existence of external shock, "unexpected events" (Ansell et al. 2017:79). Our paper concerns the adaptation of innovative activities of Polish companies to environmental turbulence. We turn to the concept of turbulence "as interaction of events or demands that are highly variable, inconsistent, unexpected or unpredictable". Following Ansell et al. (2017:2) we recognize that shock "complements but is not identical to the concept of crisis".
Poland indirectly felt the effects of the global crisis of 2007-2009 but did not pass the crisis itself (Polanski 2014). Firstly, up until 2008 existing technological gap between EU incumbent countries and New Member States (including Poland) has been decreasing. The convergence in innovation capabilities and performance has been taken place (Archibugi and Filippetti 2011). Secondly, the impact of the global crisis of 2007-2009 on innovation of NMS was much stronger and lasted longer than in the incumbent EU countries. For many years, the share of Polish innovative firms in total number of manufacturing firms was lower than in 2008 (18.5% in 2017, 21.4% in 2008) (calculated on the basis of local database of the Central Statistical Office of Poland). As innovation is crucial for NMS, understanding their innovation behaviour during the radically changing external environment is critical. Thirdly, from 2008 to 2012 Poland was one of the few EU countries that suffered only temporary slowdowns, which happened in 2009 and in 2012. In the macro dimension, the Polish economy's reaction to the global financial crisis is similar to the 'Singapore paradox', a country "highly exposed to exogenous shocks, rendering it economically vulnerable and yet still manages to attain high levels of GDP per capita" (Briguglio et al. 2008). Fourth, a short-term decline in the GDP growth rate from 5.1% in 2008 to 1.6% in 2009, and its acceleration in the subsequent years was accompanied by the breakdown of firms' innovation. As a drop in domestic demand was small and short-lived, it is an example of the impact of external shock in a situation where demand factors that typically play a role in crisis, do not play a role. On the other hand, although innovative firms' share was very low, it decreased by about 20% in 2010, while investment in innovation dropped by 3.6%. In the next three years, the decrease in innovation firms' share was two times more than that of investment in innovation. The basis for describing the first period, 2008-2010, as the innovative crisis and the second, 2010-2012, as the innovative pessimism, is the difference between the drop of innovations firms' share and innovation investment the second compared to the first. Fifth, in the analysed period, the innovative activity of the Polish firms was highly diversified. Some firms were persistent in innovation in the 10-year period from 2004 to 2014, and others halted innovation during crisis and/or pessimism, only to re-introduce innovation in the following years. This presents diverse behaviour among firms and raises questions about the knowledge resources and characteristics of firms that allow them to cope with the negative impact of external shock on innovation, increase resilience, and allow innovation persistence in 2008-2012. In this study, we address the following: Which innovation capacities were restructured to smooth impact of external shock on innovation process and were the basis for resilience in innovation? Why innovation activities of some Polish firms handle external shock better than others do, and continued the innovation process?
The paper is organized as follows. The first part proposes the major conceptual pillars of the concept of resilience in innovation to external shock. In the second part, we present the theoretical background and empirical results of the literature. In the third part, we study the characteristics of a panel of innovative firms and changes in their use of innovative resources in 2010-2012 compared to 2008-2010. The fourth part is for hypothesis development. The fifth part describes the econometric methods and reports the empirical results. Finally, we present the conclusion.

Thinking about Resilience in Innovation
Resilience is a highly multidisciplinary concept that emerged from engineering domain and is defined as the ability of a system to adjust its functionality to disturbances. There are different meanings and approaches within different disciplines (Sutcliffe and Vogus 2003;Ponis and Koronis 2012;Hosseini et al. 2016;Sanchez et al. 2017;Linnenluecke 2017). Economics and management literature has contributed to the development of the concept of resilience as a reaction to shocks in the previous two decades, although it has focused on the regional economy and only some industries. In economics, the resilience approach is in the early stages of development (Baycan and Pinto 2018).
There are two main perspectives of the meaning of resilience. The first one refers to stability and retention of the firm to pre-shock structure and function and is reboundoriented (McDonald 2006;Burnard and Bhamra 2011;Lengnick-Hall et al. 2011). It addresses the need for firms to cope with shock, absorb strain, adapt to disruptions, improve functioning, and continue normal operations after the shock. The second perspective has a dynamic character. It refers to the creation and development of a new opportunity, capacity, and change of structure and function in response to a shock. It visualizes organizational resilience beyond restoration, and induces the development of new capabilities and opportunities. Gilly et al. (2014) combine both approaches and describes resistance as "a double capacity: that of resisting a shock or limiting its effects but also that of anticipating and thus adapting to this shock or to a rapid evolution in the economic context by creating new systems". As a dynamic category, resistance affects changes in the behaviour of companies in shock. It "is not only about being persistent or robust to disturbance. It is also about the opportunities that disturbance opens up in terms of recombination of evolved structure and processes, renewal of the system and emergence of new trajectories" (Folke 2006:259). The term was applied to an organization (Hosseini et al. 2016) and its activities regarding finance, supply chain (Carvalho et al. 2012), sectors (Wójcik and Cojoianu 2018), physical networks, and innovation activities.
As ability or capacity, organizational resilience depends on latent resources (Annarelli and Nonino 2016;Sabatino 2016) and dynamic capabilities (Teece et al. 1997). They "can be activated, combined, and recombined in new situation as challenges arise" (......) and "their roles are fundamental" (Vogus and Sutcliffe 2007, p. 3418, 3420). Possessing capacity, and valuable resources before or during shock allows a firm to adapt and recover. As achieving resilience is determined by firms' own properties, the role of pre-disaster resources is underlined (Berkes 2007). Through accumulation of knowledge, past learning fosters future learning and serves to create a new development path that emerges after the shock. The dynamic approach to organizational resilience implicitly refers to resilience in innovation, which by its very nature is dynamic.
Economic development is driven by knowledge (Verspagen 2006;Stephan 2010), which constantly changes. Because of both innovative activity of firms and changes in environment, innovative resources change and new resources develop. It results in changes in the accumulation of knowledge. A decrease in the use of innovative resources affects the knowledge base and reduces the ability for continuity of innovations. Firms should ensure they have the resources for achieving and maintaining resilience to a shock, and then activate, combine, and recombine them in a new way.
Resilience of innovation to shock considerably differs from other types of resilience, for example regional or organizational ones. This stems from the specifics and the essence of innovation which is knowledge (Papadiuk and Choo 2006: 309, 311;Dost et al. 2019;Pavit 2006;Salter and Alexy 2014) -a unique, cumulative and a peculiar economic good. The cumulativity of knowledge is the key element of the persistence of innovativity (Antonelli 2019: ch.2). Although knowledge is the core of social and economic processes, it does not play as a major role in either of them as it does in the case of innovation. The latter is a materialized and commercialized form of knowledge, and its lack causes lack of innovation. Moreover, not every type of knowledge has a materialized form. If it does not materialize, it enables and conditions the creation of innovation. Constantly emerging types of knowledge and their combination and complementarity, including with existing types of knowledge (see Wang 2007), create unlimited opportunities for innovation. They are often overlooked or underestimated by competitors. Since there is no innovation without knowledge, its resources, features, cumulative nature, continuity of development and the ability to adapt to changes in the environment determine innovation resilience. This is the ability to create, restructure, recombine and apply various types of knowledge as a specific resource, create new synergies and complementarities in shock. It is not, as with regional resilience, the ability to withstand shock or to recover to its pre shock state. It is primarily the ability to continue creating, renewing, restructuring, new applications and materialization of knowledge, i.e. its accumulation during shock. This is the basis for the continuity of innovative activity and for entering a new development path. Although not limited to adaptation to shock, innovative resilience creates the basis for the continuation of innovative activity of companies also after shock. The definition of resilience of innovation as "capacity of an innovation process to maintain or accelerate its functions when facing an internal disruption and / or an external shock" (Pinto et al. Pinto et al. 2018: 62) is too general. It also applies to all other economic activities. When we convert the innovation process into for example a transport process in this definition, we understand transport resilience as "the capacity of a transport process to maintain or accelerate its function….". Resilience of innovation refers to the continuing accumulation of knowledge which makes the company immune to further turbulences. For this to happen, however, the company must continue to accumulate knowledge during the shock, which it will use in innovation activities during the shock, immediately after it and in the long run, also during the next shock. This means that innovation resilience is a dynamic process resulting from the characteristics, especially the cumulative nature of knowledge.
In this study, we treat innovation resilience as a firm's ability to perform innovation continuously in turbulent environments, increase knowledge accumulation persistently, and create reinforcement mechanisms in innovation that translate into new capabilities (Pinto et al. Pinto et al. 2018, p.62). We emphasize the dynamic nature of a firm's resilience in innovation and consider it an ongoing process to create and use the accumulation of knowledge rather than a recover to a stable (pre-crisis) state. As resources play a critical role for an organization, it is reasonable to identify resources that increase the probability of being resilient during disruptions (Van der Vegt et al. 2015) and constraints that impede or diminish this resilience (Parker and Ameen 2018). Setting the analysis within an evolutionary perspective allows identification of the resources that increase innovation resilience against the external turbulences.

A Conceptual Framework and Empirical Research
We examine three main principles of evolutionary perspectives: variation, retention, and selection processes. The interactions between these three are based on Generalized Darwinism (Coccia 2018;Simmie and Martin 2010). In a system that covers a variety of entities, only the ones that best fit environmental changes will survive.
In the resource-based view (RBV), the company is a historically determined collection of valuable, rare, inimitable, non-substitutable resources (Barney 1991). They determine a firm's behaviour, superior performance, and selection. This theory emphasizes the importance of specific resource characteristics. Firms operating within a changing environment, creating opportunities and constrains to innovation, "may change the significance of resources to the firm" (Penrose 1959, p. 79) and the value of resources (Barney et al. 2011). The firm's ability to adapt, in order to exploit their current strength, and explore new opportunities, is determined by their ability to use, create, develop, and modify the bundle of resources and their characteristics.
The concept of dynamic capabilities is the extension of the RBV, which serves to explain the basis of continuity of innovation in "a regime of rapid change" (Teece 2010, p.694). The dynamic capabilities are defined as "the firm's ability to integrate, build, and reconfigure internal and external resources/competence to address and shape rapidly changing environment" which is "fast moving" (Teece 2010, p.692, 690). Dynamic capabilities have both external and internal dimensions and their transformation in rapidly changing circumstances is emphasizes. Firms' ability to create a new, complementary, configuration of dynamic capabilities in a new way becomes newly created value differentiated across firms and results in their selection.
In the knowledge-based view, resources are the result of the creation and development of knowledge. Its accumulation, which plays a critical role in resilience innovation, is based on both: absorption of external knowledge (via networking) and creationby the firm -of new knowledge. The generation of new knowledge builds upon previously learned information. Previous innovation activities extend a firms' knowledge stock, reduces resource constraints, and increases the probability of subsequent innovation (Duguet and Monjon 2002;Clausen et al. 2011). The suppression of innovation processes results in a decrease in knowledge accumulation, its stock, and quality. Depreciation of knowledge leads to an increase in the gap in knowledge stock between a firm and its competitors and diminishes the probability of introducing innovation to the market. To remain competitive, a firm must persistently accumulate knowledge that it creates, develops, and acquires from the environment. However, changing environments affect the role of used resources, capabilities, and their feedback that was crucial for innovation in stable environments. Past successes may be insufficient to achieve new success or lead to new innovations. Volatile conditions might reveal a lack of innovative resilience of the firm.
Empirical literature on the relationship between macroeconomic instability and innovation focusses on the impact of crisis on innovation process. It implicitly refers to the continuity of innovative process and its resilience/vulnerability to radical change in the environment. This literature can be divided into four streams of research.
The first stream focuses on diversification of the impact of the business cycle on changes in innovative capabilities, behaviour, and the performance of firms. Three types of innovation patterns and their sources are selected: pro-cyclical, anti-cyclical, and non-systematic (Fabrizio and Tsolmon 2014;Giedeman et al. 2006). Pointing to the diversity of companies' reactions to changes in the environment, this literature enriches discussion on persistence in innovation. The second stream of research focuses on the linkages between pre-crisis innovation resources, strategies, economic performance, and the capacity to react to the crisis through further innovation activities. This research concentrates on moderately innovative countries like Spain (Zouaghi et al. 2018), Italy (Antonelli et al. 2013;Colombo et al. 2016;Ausloos et al. 2017), Portugal (Costa et al. 2018;, and Argentina (Suarez 2014). They show which resources accumulated in the pre-crisis period stimulate growth performance and allow continuation of innovation during the crisis (Archibugi and Filippetti 2013; Le Bas and Scellato 2014). Although they base analysis on CIS database and probit models, different scope, approaches, and variables used give different results. The third stream focuses on the impact of a firm's innovation strategy during the crisis, on postcrisis innovation behaviours and performance (Correa and Iootty 2010;Hud and Hussinger 2015). It shows that innovation experiences during recessions are beneficial for future innovation performance (Amore 2015). The fourth stream of empirical research focuses on the role of marketing or human capital on a firm's innovative behaviour and performance during the crisis.

Data, Population Characteristics, and Changes in a firm's Internal Capabilities
From data of all five waves 1 of the Polish Community Innovation Survey (CIS), covering the period from 2004 to 2014, we constructed a panel of 1553 innovative manufacturing firms, that introduced a product innovation, a process innovation, or both in at least three CIS waves during 2008-2010 and/or 2010-2012. Using the Oslo Manual definition and approach to innovation, we analyse firms that were innovative in the long run.
We study the years 2008-2012, and divide the time into two sub-periods. The first, 2008-2010, is described as an innovation crisis. It is characterized by a strong decline in the innovative firms' share and a much smaller, but still considerable, decline in investment in innovation. In the second period, 2010-2012, called the period of innovative pessimism, the decline in the innovative firms' share was smaller than the drop in innovation investment expenditures. For this study, we examine which resources used in the first period increase or diminish the likelihood of introducing innovations to the market in the second period.
As the panel consists of mostly medium-sized (59.2%) and large-sized firms (38.1%), it biases in favour of large firms. The share of small firms was negligible (2.7%), because the data received from the CSO were not weighted. However, according to the CSO, data covering weights, the share of small firms with innovation expenditures, and sales of innovative products was small (less than 8%). Thus, small firms' innovative resources have little impact on the panel.
Most panel firms operated in the medium -high (43.8%) and low technology 2 (38.7%) sectors. The share of high (5.3%) and medium-low (12.1%) technology firms was much smaller. Most firms were private, and over 40% belonged to the capital group. Of the firms, 54% did not export innovative products. However, a significant number of exporting firms (29%) had a share of innovative product exports that exceeded 20%. In 50% of the firms, the intensity of sales of innovative products exceeded 10%, and one-third exceeded 20%. Table 1 shows 34 types of innovative resources: tangible and intangible, static and dynamic (human capital defined as the share of employees with a university education), external and internal, financial and non-financial, as well as resources showing what the firm has and what the firm does. We consider four forms of innovation. The binary form of the received data allows an indication of which resources were used most often. As a measure of innovative output, we use both product and process innovation.
During the analysed period, product and process innovations were introduced twice as often as organizational and marketing ones. Firms invest in innovative resources in a selective way and most innovative resources were rarely used. Only 30% of resources were used by more than 50% of companies. Among external resources, the companies most often bought machines and software, and used market and public information. The most frequently used internal resources were financial and information data. In the period of innovative pessimism, the frequency of using only four out of 28 types of innovative resources has increased. This included two external and two internal resources.
A significant number of firms (from 40% to 80%) encountered barriers to innovation. During 2008-2010, the barriers most often felt were uncertain demand, unfavourable market conditions, and difficulties in finding cooperative partners. However, perception of these barriers sharply declined in the next period.

Research Hypotheses
The crisis is a kind of shock, but not every shock is a crisis. Not all features, conditions, and elements of the crisis and shock are the same. In contrast to the crisis, whose effects are all encompassing, the effects of shock are more selective, and the channels where the economic crisis and shock can affect persistence in innovation can be different. This applies to the role of demand.
In the demand-pull model of innovation, demand is seen as a crucial driver of innovation. Economic expansion generates capacity to absorb new products while a downturn limits demand. As of 2008-2012, the Polish economy was not part of the global economic crisis and the short-term slowdown in growth was accompanied by a slight and very short-term decline in domestic demand. Thus, we develop the following hypothesis: H1. When external shock hits, domestic demand does not hamper the continuity of innovation The global economic crisis was accompanied by a decline in the global trade and shortterm decline in Polish manufacturing exports. Most Polish exporters pursue innovation activities and export the results. Using innovations for competition in turbulent times, they continue with this activity. In line with this, we develop the following hypothesis:

H2. The export of innovative products increases innovation resilience to turbulence
Introducing new or significantly improved products to a turbulent market requires intensified marketing activities (Naidoo 2010). Marketing stimulates sales of innovative products and adjustment of innovative products to new market conditions. As a marketing strategy mitigates the effect of turbulence (Huhtala et al. 2014;Mendrano and Olarte-Pascul 2016), we develop the following hypothesis:

H3. Marketing innovation products strengthens persistence in innovation
In literature, a special role is assigned to firms' financial resources and access to external financing (O'Sullivan 2005; Hall and Lerner 2009;Martinsson and Lööf 2009;Hall 2002Hall , 2010Lee et al. 2015), including public support (Paunov 2012;Antonioli and Montresor 2018). Uncertain character of return in innovation and asymmetries of information makes the investment riskier for financial institutions (Mazzucato 2013), and thus makes it harder for innovative firms to access financing. Limited access to financing hampers innovative activities and increases their vulnerability to turbulence. During the crisis, firms in financial distress who encounter credit constraints are more likely to reduce investment spending than those without financial problems. Having liquid assets and the option to use external funds increases innovation resilience. During turbulent times, banks provide financing at a higher cost. Although the impact of the world financial crisis on Polish economy was quite weak, Polish banks introduced lending restrictions. This increased the importance of internal financing, access to external sources, and public funding. Thus, we develop the following hypothesis:

H4. A firm's good financial standing and access to external funds, including public funds, increase the resilience of innovation activities
Human capital is valuable resource and reduces the chances of slow growth (Goedhuys and Sleuwaegen 2015). It creates a new value and increases the absorptive capacity of firms. By enabling internal capabilities and creating new ones, it acts as a coping mechanism to cushion the effect of shock on innovation performance. To adapt to changes in turbulent times, the organization relies on highly qualified personnel and ongoing training to improve skills (Chowhan et al. 2017). Firms with human capital have shown more resilience to shock (Lai et al. 2016). Consequently, we hypothesise that: H5. High share of human capital increases firms' innovation resilience to environmental turbulence R&D is a key factor of a firm's capacity to innovate. Empirical research highlights the positive impact on sales or the degree of novelty of innovative products (Beneito et al. 2015). However, R&D is costly and risky, and requires a minimum amount of resources and time to achieve results. Empirical literature presents evidence for procyclical and counter-cyclical behaviour of R&D and its dependence on other factors such as financial capabilities, imperfection in credit market, openness, and technology level of production. Procyclicality of R&D spending is connected with the "financeconstrained effect" which results from changes in sales. R&D expenditure of firms that are moderately unconstrained financially is less procyclical. Countercyclical behaviours are seen in turn because of the "opportunity cost effect". There is a high opportunity cost of devoting resources to R&D in times of high sales. Firms operating in a turbulent market with strong competition pursue their own R&D and continuously use external R&D. As analysed firms use innovation for competition, we hypothesise that:

H6. Internal R&D efforts and the use of external R&D increase resilience in innovation to environmental turbulence
The level of production technology determines the innovative responses to external shock. Low-tech (technology using) industries are characterized by low level of R&D intensity and their developmental character, while high tech (technology producing) are characterized by high R&D intensity and high share of basic research. The strong dependence of the latter on basic research requires persistence and continuous use of complementary innovative resources, independently of changes in the environment. Hence, the innovative activity of the high-tech industries seems to be resilient to shocks. Higher sensitivity of low-tech firms to shocks (Rafferty 2003) is result of their dependence on technology suppliers and the developmental character of their R&D. Consequently, we hypothesise that:

H7 The innovation activities of high-tech industries are more resilient to shocks than low tech ones
Empirical studies have provided support for latter Schumpeterian thinking with evidence that small firms are less likely to be innovative than large firms. Small firms are subject to the "liability of smallness" and have a different structure of innovative resources. Resources and capabilities can constrain small firms from innovation activities (Lai et al. 2016;Hud and Hussinger 2015;Kauf and Kniess 2015;Zouaghi and Sanchez 2016;Madrid-Guijarro 2013). Large firms can continue innovation due to low incremental costs of their internal innovation activities. Small firms are more vulnerable during times of hardship than larger firms, and they sometimes face difficulties obtaining external financing. Consequently, we hypothesise that: H8. Large firms are more resilient in innovation to external shocks than small firms are The purpose of our investigation is to determine the probability of introducing market innovations during the period of 2010-2012 (second period of the crisis) using only information from the period of 2008-2010. We want to identify resources that increase or decrease the probability of continuous commercialisation of innovations in unfavourable market conditions. To realize our purpose, we have built econometric models linking a status of innovation in the period 2010-2012 (variable in2012 as dependent variable) with several categories concerning the period 2008-2010 (as independent variables). Variable in2012 is defined as in2012 ¼ 1; if a firm was innovative in the period 2010-2012 0; otherwise Based on data from the period of the innovation crisis (2008)(2009)(2010), we aim to identify innovation drivers that strengthen innovation resilience in both periods: innovation crisis and innovation pessimism (2010)(2011)(2012). In the second period, the number of companies introducing innovations continued to decrease. In our opinion, introducing innovations to the market in both periods suggests that companies were innovation resilient. We used a cross-sectional model in which the dependent variable (innovations introduced to the market during the period of innovative pessimism) is explained by the independent variables from the previous period (innovation crisis). In this way, we identify innovation drivers that increased the likelihood of introducing innovation during the crisis and innovative pessimism, and thus the strength of innovation resilience of Polish companies. Because our dependent variable is binary, we cannot use classical linear regression models for estimation. Most frequently, models used in the case of a binary dependent variable are logistic regression and probit analysis. Both methods estimate the probability P i that the dependent variable for observation i is equal to 1 based on linear combination of independent variables. Distribution functions for both these models are similar. We use the logistic regression model because the mathematical form is simpler and it is more frequently used in practice. Moreover, using logistic models in SPSS is easier and gives better output concerning a goodness of fit analysis. Estimation of models of logistic regression allow us to assess which factors increase the likelihood of the commercialisation of innovation in the period of innovative pessimism, as compared to innovation crisis.
In the original model, we adopted a large number of independent variables that could affect the assessment of the likelihood of innovation during the period of innovative pessimism. The choice of variables resulted from the availability of data in the CIS questionnaire and results of empirical studies. We used 36 variables that reflect types of resources used to commercialize innovative products. Among them, 35 are binary variables, oneis the share of employees in the company with higher education. Independent variables, except higher education variable meaning % of employed with higher education, are binary variables. They say that a given category (phenomenon) occurred in the company (e.g. the company applied continuously Internal R&D) or is important for the company (e.g. information from suppliers is important). Then the value of the variable is 1, otherwise 0. Some values from the CIS survey have been aggregated into zero-one variables. We have added 6 binary variables (medium, large, tech1, sales, export, and private) concerning the company's characteristics to the list of variables: company size, ownership, level of technology, industries, sales, and exports shares of innovative products. The results of empirical research indicate the influence of the above-mentioned company features on the ability to commercialize innovation and its results. New variables are defined as follows: For the identification of innovative resources that -in the period of economic turbulence -ensure continuity of commercialisation of innovations, we have built a binary logistic regression model. Using this model, we aim to identify resources that increase the probability of introducing innovative products to the market. We introduced two approaches. In the first we used all of the variables listed in the regression model. In the second approach, we leave variables that are not statistically significant in the model being built. Created models employ data on the resources used during the crisis of innovation (i.e., in the first analysis period). As we know, if a given company was-in reality-innovative in the period of innovative pessimism (i.e., in the second period of the study) we can estimate the effectiveness of the prediction of the methods used.
We have built a model to explain the effectiveness of predicting that the firm will be innovative during the innovative pessimism period (2010-2012), based on innovative resources used in the period of innovative crisis (2008)(2009)(2010). As in the panel, there were more innovative companies in the second phase of the turbulence than in the first phase, and the logistic regression model tends to predict that the company will be innovative. Therefore, based on the calculated probability of commercialisation of innovations in the second part of the analysed period, it is necessary to determine the so-called cut point q. We assume if the probability of commercialisation of innovation estimated by the model is greater than q, then a company will be innovative. We created a contingency table for observed and expected capacity for innovation. The best accuracy is taking q on the level of 80%, which means that the compliance was around 80%. In this case, based on the information on innovative resources in the first phase of the turbulence, the model was correctly predicting whether the firm would commercialize innovation in the second phase of the crisis.
The estimates include the values of the regression coefficients b, the values of Wald's statistics, and test probabilities (p values) used to assess the significance of the variable in the regression model and exp (b). Wald's statistics are used to test the significance of explanatory and constant variables. The value of the likelihood function is useful to assess the goodness of model fit. We chose the estimation value b0, ..., bk in such a way as to maximize the function of likelihood function L. Here, as a measure of the degree of alignment we will use statistics -2ln L i.e. minus 2 times logarithm function (−2 Log Likelihood). If the null hypothesis that the model fits perfectly to the data is true, then the statistic is -2Ln L has χ2 distribution with n-k degrees of freedom.
If the model fits perfectly to the data, the likelihood function takes the value 1 and then -2lnL = 0. In practice L < 1 and then -2ln L > 0. Too large values of the -2ln L function indicate that the null hypothesis cannot be true and should be rejected.
In contrast to linear regression models, we cannot use the coefficient of determination R 2 as a measure of how well observed outcomes are replicated by the model, based on the proportion of total variation of outcomes explained by the model. Statistics Cox & Snell R 2 and Nagelkerke R 2 (also called pseudo R-square) are aimed at estimating the variance of the dependent variable explained by the model in the total variance of the dependent variable. They have an interpretation similar to the coefficient of determination in the classical linear regression model. Calibration of the model tells us how strongly the observed and predicted values match each other in the whole range of volatility. The Hosmer and Lemeshow test (1989) serves this purpose. We divide the observations into 10 approximately equal classes ascending according to the estimated probability of the event (they are decile groups) and we examine the distribution of observed and predicted values in these groups. Next, we use the goodness of fit test based on χ2 statistics. There should be a sufficient number of observations so that in most decile groups, the number of expected events exceeds 5, and no group had a zero number of expected events. The differences between the observed values n i and expected b n i are calculated and the value of the statistics chi-square= ∑

Results
Model quality measures are low. The coefficient Cox and Snell R 2 is 0.1111 and the coefficient Nagelkerke R 2 is 0.203. The Hosmer and Lemeshow test indicate a good match between the expected probabilities of success and real values. There is no reason to reject the null hypothesis, because the value of the statistics χ2 = 13.477 for the number of degrees of freedom df = 8, which corresponds to the empirical significance p = 0.096. The share of correctly classified observations (compliance of observed and anticipated values) is 78.6%, including 55.6% for non-innovative companies and 82.2% for innovative firms (for the break point q = 0.8).
Because not all explanatory variables in the obtained model are statistically significant, we used the backward method of selecting variables while maintaining variables significant at the level of 0.1. At this level 18 variable are significant. The model after using backward method of selecting variables is presented in Table 3.
The new model has similar values for quality statistics, as a model in which all explanatory variables are included (with no variable selection). Values of model fit statistics: 2logarithm of credibility = 1049.645 Coefficient Cox & Snell R 2 is equal to 0.101 and coefficient Nagelkerke R 2 is equal to 0.185. The value of χ2 statistics for the Hosmer and Lemeshow test is 10.010 (df = 8) and falls within the scope of the lack of grounds for rejecting the compatibility hypothesis at significance level 0.05 (p = 0.264). The share of correctly classified observations is 78.2%. This means that (according to the estimation of the model) 78.2% of companies were accurately predicted. For innovative companies the accuracy of predictions was 82.3%, and 52.2% for noninnovative companies.
Contingency table of the Hosmer and Lemeshow test (Table 4) contains a comparison of observed and expected values for decile groups respectively for non-innovative (in2012 = 0) and innovative firms (in2012 = 1).
Of the 42 variables, 10 strengthen resilience in innovation to the shock (Table 3). Four were of demand nature (internal demand, exports and two forms of marketing), three -supply (in-house continuous and external R&D and acquisition of existing knowledge) and three -financial. As the impact of external financial shock on domestic demand was small and shortterm, we hypothesised (H1) that in the period of the innovative pessimism, domestic demand did not hinder persistence in innovation. As a proxy of demand, we adopted two variables: uncertain / lack of demand and unfavourable market conditions. In the period of innovation pessimism, their perception by companies has significantly diminished ( Table 1). The first variable was irrelevant. The impact of the second was very small. As the improvement of domestic demand and market conditions has not hindered persistence in innovation, H1 is supported.
The global recession had a negative but a short-term impact on Polish exports. Because innovation is a key factor of international competition for Polish exporters, the continuation of innovation activities supports maintaining competitive advantages. The firms continued innovative activity despite the reduced foreign demand. Using innovations as a method of competition, they implemented a strategy of continuous introduction of new products to the market. We hypothesis (H2) that exports of innovative products increase the probability of the persistence in innovation in times of unrest. Our estimates (Table 3) support this hypothesis, although the impact of export was weak.
Marketing directly stimulates demand for innovative products. We use two forms of marketing: significant changes to the aesthetics of a good product, and new media techniques for product promotion. The impact of the first was quite strong, while the second was much weaker. Our hypotheses (H3) that implementation of marketing strategy increases the probability of persistence in innovation in turbulent time is supported.
Most empirical studies find support for a strong impact on persistence in innovation of financial resources. In particular, this applies to the countries that are catching-up. Taking into account not only the key role of the financial situation of companies for their innovative activities, but also the financial nature of the external shock, we developed a hypothesis that (H4) good financial standing and access to external funds, including public ones, strengthen firms' resilience in innovation. As a proxy for the financial situation of firms, we take three variables: internal and external expenditure on innovation, and financial barrier to innovation. Our estimation shows that having a good financial standing and access to external financing strongly strengthens resilience in innovation. However, even if public support for innovative activity increased during the period under consideration, it did not increase the likelihood of the continuity of the innovation (variable is irrelevant). The fact that the use of information from the public sources and other sources negatively affected the continuity of innovation is a challenge for public policy. Two hypotheses consider the impact of supply factors (human capital and R&D) on resilience in innovation to external shock.
Human capital allows for adjustment to changes in environment and reduces its uncertainty. As in analysed period percentage of employed with higher education increased, we hypothesise (H5) that human capital strengthens innovation resilience. Our estimates show that the impact of human capital was meaningless, thus our hypothesis is not supported.
Among factors that increased the resilience in innovation were in-house continuous R&D, purchase of external R&D, and acquisition of existing knowledge. The magnitude of their influence was quite strong. H6, which hypothesises that R&D expenditure strengthens resilience in innovation to external shock, is supported.
In contrast to other countries, the level of technology of production did not affect the ability to commercialize innovation in Poland (the variable is irrelevant). Thus, H7 is rejected.
The results of the model indicate a strong influence of the size of firms on commercializing innovation. Large and medium-sized firms have significantly greater ability to pursue innovation than small ones. H9 hypothesis with respect to the higher vulnerability to external shock of small firms compared to large firms is supported. Our results are confirmed by other estimates.
The use of four resources: one tangible (the acquisition of machinery and equipment) and three intangibles (consulting services, information from public sources and other firm's groups) reduced the likelihood of persistence in innovation, although the purchase of consulting services is not significant at the level of 0.05. The innovativeness of sales (high share of innovative products in sales in 2008-2010) reduces the probability of innovation in the next period.
Resilience in innovation to shock was strongly strengthened by financial factors, marketing, R&D expenditures, and was much higher in large and medium size firms than in small firms.

Discussion and Conclusions
In this study we analyse the concept of organization resilience in innovation to turbulence in the environment. Based on evolutionary perspectives, we use the notion of dual capacity of the resilience: both adopting to radical change and creating a new system. Innovative resilience results in persistence in innovation, i.e., continuity of knowledge accumulation stemming from the past and creation of new knowledge and learning, stimulated and acquired during the shock. It reflects the process of path dependence or de-locks from the old path and stimulates selection across firms.
In the study, we deal with 1553 Polish manufacturing firms that have developed the innovative potential for the long run. Continuously introducing innovative products, they actively compete on international markets. Innovations were key for their development and expansion.
The reaction of the Polish economy to the global financial crisis (2008-2010) is reminiscent of the 'Singapore paradox'. However, it did not stop a drop-in risk tolerance and in investment return expectation of entrepreneurs, especially in a highly risky type of business (i.e. innovation activities). Although in Poland in 2010-2012, perception of market conditions among entrepreneurs improved considerably (Table 1), it was accompanied by a continuous decline in the innovation investment rate and in innovative firms' share. This suggests that psychological factors contribute to the perception of the potential impact of global turbulence on risky areas of domestic activity, such as innovation. This impact is mitigated by various factors. In the Polish case it was supply (R&Dboth continuous in-house and external), demand (exports of innovative products and marketing) and financial factors, primarily internal, but also external. The strongest impact came from financial factors. The small share of Polish manufacturing firms that pursued continuous R&D, exported innovation products and had a good financial standing in the total number of manufacturing firms caused a large decrease in firms' innovation market share over the period considered. In view of the negative impact of public policy on innovation resilience, its strengthening requires introduction of a new approach by the government.
Our study has aimed to contribute to the innovation literature. First, it shows differentiation of the impact of various internal capabilities and external knowledge on innovation, when the external environment changes radically (Archibugi et al. , 2011. While the impact of the crisis on innovation, which happens much less often than shocks, has been well established in the literature, the impact of external shock has received less scrutiny. This study adds to this literature. Second, it draws on the concept of organizational resilience in innovation against external shock. We refer to heterogeneity of innovation behaviour of firms, which highlights the vulnerability of the innovation process (Pavit 2006, p.88). Thirdly, we encompass the concepts of persistence and organizational resilience in innovation in a common framework (Marchese et al. 2018). If the shocks are unexpected events that bring out changes in persistence in innovation, resistance contributes to the strengthening of continuity. Therefore, we bring an additional element to discussions conducted within the evolutionary perspective. Fourthly, empirical research on the impact of the crisis on the innovative behaviour of companies focuses on moderate innovators: Spain, Italy, Portugal, and Argentina. According to our knowledge, there are no studies on the impact of external turbulence on the NMS innovative processes. The effects for these countries are much stronger and halt the convergence processes in innovation for many years. Finally, we turn to the dissimilarity of the behaviour of innovative firms under the influence of evolutionary changes compared to volatile changes in the environment (Suarez 2014) Bearing these provisions in mind, this study offers an extended and empirically grounded perspective on the organizational resilience and persistence in innovation when external shock hits, and aims to fill a gap in the innovation research. B/HS4/02742. The statistical data used in the calculations originate from the Statistical Office in Szczecin, which assumes no responsibility for the conclusions reached in the paper.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.