Discovering the factors driving regional competitiveness in the face of climate change

The competitiveness of territorial units is an important interdisciplinary theoretical and practical problem. The era of competitiveness has emerged with the intensification of globalisation, liberalisation, integration processes and the ICT revolution. Considering economic development measures (Aiginger & Vogel 2015; Zitkus 2015), competitiveness is an immanent feature, the driving force of modern economic processes and one of the fastest growing research categories. Competitiveness captures much space in the studies of individual countries’ regional development strategies and policies (Hughes 2021; Huggins et al. 2021; Fendel & Frenkel 2005). The vast majority of papers emphasise that it can be understood as a cause, an effect or a means to achieve growth and socio-economic development (Aiginger & Vogel 2015; Doyle & Perez-Alanis 2017). Competence is also a prerequisite for achieving economic growth from a regional perspective.

In the face of climate change, the competitiveness paradigm is shifting; thus, there is a need for in-depth research on the factors fostering regional competitiveness (Li et al. 2016). The accelerating climate change is affecting the living environment of people around the world, resulting in damage to ecosystems, infrastructure and public health, and causing lower agricultural productivity and a decline in both labour supply and worker productivity (Schwab, 2019). Climate transformation changes economic conditions in the regions and, as a result, influences their attractiveness and competitiveness. Although much literature is devoted to regional competitiveness, there is a gap concerning the crucial drivers of competitiveness related to climate change (Moldan et al. 2012; United Nations 2015). Thus, there is a need to look at regional competitiveness from a new perspective, paying more attention to climate change, as Dahl (2011) and Mayer (2007) have stressed. Given this gap in the literature, the main intention of this paper is to gain insight into the field of significant factors that affect the climate competitiveness of regions. Our approach goes beyond low-carbon competitiveness, the most common attitude that finds a competitive advantage in reducing gas emissions (Li 2011; Ellis 2013). We define regional climate change competitiveness as a region’s ability to achieve and maintain a competitive advantage under the constraints of climate breakdown. This concept considers both the region’s impact on climate change and the region’s adaptive and proactive actions (Karman et al. 2021). The proactive attitude is understood as voluntarily initiated behaviour to bring about transformation through appropriate foresight supported by forward-looking actions (Parker et al. 2006). Gaining a competitive advantage through climate proactivity will mitigate climate change caused by the negative effects of anthropogenic activities and bring new opportunities related to social, technological and economic factors (Reinhardt 1999).

The article aims to identify the significant factors that drive regional competitiveness in the face of climate change. We believe that climate change is currently one of the most important factors affecting the competitiveness of all territorial units regardless of their economic advancement. Therefore, in-depth studies are needed to test new sources of competitiveness and find which factors are crucial for building and strengthening regional attractiveness. This will allow for appropriate policies for regional development under climate change conditions. In this paper, we hypothesise that selected independent variables have a statistically significant impact on regional climate change competitiveness. Our approach addresses the following questions: What are the relations between environmental, economic, social, technological and innovative factors and the level of climate competitiveness? What is the importance of each factor for regions with different levels of development and competitiveness? Which sources of regional competitive advantage can underpin the region’s development under climate transformation?

The paper’s novelty derives from recognising climate change as a key important factor for regional competitiveness. The vast majority of studies focus on estimating the level of competitiveness concerning other regions, but omit the issue of changes occurring in their environment. However, an attempt to identify and measure the factors determining the competitiveness of regions in climate change conditions has a high research value. Our research allows us to discover which factors affect a region’s ability to adapt to climate change while building its competitive potential. The present study contributes to the literature by including the dimension of climate change as a core element of the concept of regional competitiveness. With this approach, we offer new insights into the factors determining the competitiveness of EU regions. Identifying the main determinants of climate change competitiveness would help policymakers decide which economic issues should be taken into account to enhance the competitiveness of their region.

In order to achieve the proposed objective, the study is organised as follows. First, the conceptual framework and the proposed model are presented. The methods used and the results of the study are then described. Finally, the conclusions and limitations of the study are presented.

Competitiveness is considered one of the paradigms of contemporary scientific thought and an important interdisciplinary problem. Referring to the OECD definition, competitiveness is the extent to which a unit, under free trade conditions, may produce goods and provide services that succeed on international markets, with the simultaneous guarantee of the viable growth of the population’s income in the long run (OECD 1996). This understanding of competitiveness emphasises the ability of an area to maintain competitive advantage in the long run and provide a better quality of life. In a world of limited resources (human, technological, natural), those locations that guarantee the best conditions for economic activity and life win out. These locations – countries, regions or cities – serve as magnets for investment and people, and are able to develop and achieve prosperity, which is the ultimate goal of economic policy.

This focus on competitiveness has not just been a macroeconomic phenomenon, however. According to Martin (2005), regions are becoming increasingly important – perhaps even displacing nation-states – as the key areas of wealth production and economic governance. It is at the regional (subnational) scale that many of the increasing returns that raise the productivity of firms and workers are created and are self-reinforcing. Regional competitiveness is indicated as a strategic development goal of the region (Zitkus 2015), a feature of the region, a manifestation of development, an indicator of development (Biniecki & Frenkiel 1999), a condition of development, a component of development, a means of development, or a synonym for development (Berger & Bristow 2008). Most of the existing approaches refer to a region’s ability to manage resources and exploit development opportunities to increase the well-being of its people (Camagni 2002). In this discourse, regional competitiveness is understood as the capability of a region to attract and keep firms with stable or increasing market shares in an activity, while maintaining stable or increasing standards of living for those who participate in it (Storper 1997). This conceptualisation views regional competitiveness as a combination of the competitiveness of the region’s firms and its overall economic performance (or validation through sustained or improved levels of comparative prosperity). A region is “competitive”, according to this view, when it has the conditions to enable it to raise its standard of living, or the ability to sustain “winning” outcomes. These conditions are perceived to include a mixture of Porterian competitive advantage for firms and the attractiveness of the regional environment for business, while maintaining or increasing the rate at which the region’s human capital is employed (Bristow 2010).

Currently, one of the great challenges of regional development is climate change, which is defined in the UN (1992) Framework Convention on Climate Change as “a change of climate attributed directly or indirectly to human activity that alters the composition of the global atmosphere and which is in addition to natural climate variability observed over comparable time periods.” Since the global scenarios for climate change indicate that, in most locations, warming will increase, precipitation will decrease, and extreme events will be more likely, climate change must be considered one of the leading regional challenges (WEF 2021). The effects of climate change cause relevant impacts in many different areas of human life, such as agriculture, tourism and health; therefore, identifying climate-related factors and assessing their impact is crucial for maintaining and boosting regional competitiveness, especially in regions that are particularly vulnerable to climate change. Accurate climate data are impactful for adequate policies to mitigate the negative effects of climate change (Bulut et al. 2007). Therefore, pro-environmental activities, treating social, economic and environmental components equally, are considered necessary factors for strengthening competitiveness. Importantly, the positive impact of pro-environmental measures occurs over the long term and is not always directly related to economic competitiveness, but can affect the attractiveness of the region for residents, its favourable image, and so on. At the same time, the quality of the local environment becomes an important factor of competitive advantage. Protection of the natural climate has become one of the priorities of local authorities, as evidenced by the inclusion of climate issues in regional development strategies and plans (Porter et al. 2015; Dryzek et al. 2011).

Climate change may affect not only current but also future competitiveness. While uncertainty over adaptation to global warming is well recognised, the assumption that climate change will have a long-term impact on human welfare and the global economy is standard in the climate-change literature (Pindyck 2012). Considering long-term growth, Kahn et al. (2019) report that (i) when a poor (hot) country is 1°C warmer than usual, its income growth falls by 1–2 percentage points in the short- to medium-term; (ii) when a rich (temperate) country is 1°C warmer than usual, there is little impact on its economic activity; and (iii) the GDP effect of increases in average temperatures (with or without adaptation and/or mitigation policies) is relatively small – a few per cent decline in the level of GDP per capita over the next century. In contrast, Dell et al. (2008) showed that higher temperatures substantially reduce economic growth in poor countries but have little effect in rich countries. Analogous studies on the impact of climate change on competitiveness have not been carried out, and it should be stressed that, in the case of EU regions, it is difficult to determine long-term competitiveness under climate change conditions. There are three reasons for this. First, the spatial distribution of the impacts of climate change varies over time. The dynamics of this phenomenon make it difficult to prepare projections. Secondly, the projections of future emissions and future climate change have become less severe over time. Thirdly, the earlier impact studies focused on the negative impacts of climate change, whereas later studies considered the balance of positives and negatives. Fourthly, it is difficult to determine the effectiveness of adaptation measures in individual regions, due to the diversity of climate policies and actions taken (structural change, directed technical change, land-use change, etc.). From the economic perspective, the impact of climate change is problematic because the economy is endogenous, and because policy analyses should consider impatience, risk aversion and inequity aversion between and within geographic regions separately (Tol 2008). Undoubtedly, some regions will gain. This applies to regions where the economy is concentrated in the temperate zone, where warming reduces heating costs and cold-related health problems. On the other hand, regions that comprise vulnerable sectors and poor areas may suffer from lower adaptive capacity. The competitiveness of all areas may be positively affected by low-carbon politics, but its impact will manifest itself in the long term, as the climate reacts quite slowly to changes in emissions. Finally, climate change competitiveness will vary across regions depending on the projected paths of temperatures and adaptive mechanisms.

Economic development is conditioned by the resources of the natural environment and how they are exploited. At the same time, there is a concern that environmental limitations may constitute a significant barrier to highly competitive, energy-intensive economic activity. This forces new action and adaptation to meet challenges and search for new competitiveness sources. The multidimensional character of regional competitiveness requires a holistic approach to identifying the driving forces behind its growth (Hult et al. 2007; Sirmon et al. 2011; Cetindamar & Kilitcioglu 2013). In the literature, the basic factors of key importance for competitiveness are economic and social capital, innovation, technical infrastructure and environmental/ecological conditions (Simionescu et al. 2021). According to Porter (2001), the advantage regions possess can result from the human factor associated with knowledge and education, capital and a favourable location, the aggregate consumer and production demand of the region, and the business environment – namely industries that support and facilitate the activities of business entities and the structure of business entities of a region. Winiarski (1999) indicates the main determinants of regional competitiveness as a diversified economic structure, properly developed economic facilities (i.e. technical, economic and social infrastructure), the intellectual and investment status, the business environment, and the favourable condition of the natural environment. The resources of these factors are defined as the territorial capital, which results from the initial state, the changes and the influence of elementary development factors (Camagni et al. 2011). New research approaches expand the pool of factors shaping regional competitiveness by including environmental conditions and pro-ecological measures. Pro-ecological measures that preserve the region’s resource base and minimise the exploitation of natural resources are a magnet for new investments and have a positive impact on the region’s attractiveness to potential residents, and for these reasons can be regarded as a factor that affects productivity and growth in the long term (according to the U-curve). According to this perspective, pro-ecological measures are crucial as they better prepare the region for the transition to a low-carbon or “decarbonised” economy (Martínez & Poveda 2021). Moreover, the state and quality of environmental conditions influence a region’s social and economic attractiveness, as highlighted by Moldan et al. (2012), OECD (2008) and United Nations (2015). However, despite the diversity of research on competitiveness (WEF 2019; IMD 2019), there is still relatively little research on the relationship between climate change and regional competitiveness.

For the purposes of this research, we adopted an approach focusing on the five main groups of drivers of regional climate change competitiveness: environmental, economic, innovative, social and technological (Fig. 1).

Figure 1.

Group 1 includes environmental factors of regional climate change competitiveness. These can foster or undermine the social and economic development of regions. The weather-based ratio of cooling and heating days refers to energy consumption and describes the energy requirements for buildings in terms of heating or cooling. This contributes to the evaluation of energy consumption under climate change. This indicator can correlate the utility usage with trends in weather in particular regions, predict the energy demand, and help diagnose potential usage issues and inefficiencies in buildings (Wang & Chen 2014). CO₂ emissions refer to global warming, one of the increasing worldwide threats of climate change. CO₂ from energy represents about 60% of the anthropogenic greenhouse gas emissions on a global scale (Kone & Buke 2010; Quadrelli & Peterson 2007). The qualitative dimension of energy use is becoming increasingly important for sustainable development and competitiveness. The projected studies showed that excessive CO₂ emissions into the atmosphere would increase the earth’s surface temperature by approximately 1.5–4.8 °C in the next 30–40 years. The combustion of fossil fuels causes climate change and air pollution and has a big, negative impact on the surrounding environment (Schlesinger 2010). Another environmental factor, namely the presence of sulphur oxides in the atmosphere, negatively affects fragile ecosystems and significantly affects human health. Sulphur oxide emissions cause adverse impacts on vegetation, including forests and agricultural crops (World Bank Books 1998). Limiting sulphur oxide emissions reduces air pollution, which is one of the most important tasks in natural environment protection (WHO 2013).

Group 2 focuses on economic factors. The importance of this group of drivers of regional competitiveness was confirmed, among others, by Rusu and Roman (2018). In this paper, we focus on four factors of an economic nature. Strengthening competitiveness in the economic field requires investment outlays, the scale of which is illustrated by the share of structural fund expenditures. This refers to the institutional dimension of competitiveness and includes how the public sector, and regulations, support economic activity (Aiginger et al. 2013). Another important factor in supporting the regional competitiveness of all economies is access to funding. Well-functioning capital markets support the long-term resilience of the corporate sector, which is the foundation of regional competitiveness (OECD 2021).

Competitiveness also increasingly depends on factors related to R&D inputs, which play an important role in long-term economic growth and living standards (Skórska 2019). These inputs address some of the major social and environmental challenges and enable the production of more new goods and services with lower consumption of non-renewable resources, reducing the negative externalities associated with production. The last factor, the value added, is often assumed as a measure of competitive advantage, and improvements in value added provide an indication of more effective internal processes and stronger market performance (Johns & Tilley 2003). This affects the standard of living of the inhabitants, and competitiveness.

Group 3 deals with innovative factors. Innovation is perceived as the driving force behind progress and competitiveness (Wosiek 2019), and influences human activities such as business, society, science, finance and technology and combines them into one dynamic system (Wójcik & Sierotowicz 2007). As in the case of macroeconomic competitiveness, one of the answers to the climate change challenges fostering regional competitiveness should be interdisciplinarity research and international scientific cooperation. Co-publications make it possible to gain high scientific quality and increase the dissemination of knowledge, especially concerning adaptation to climate change, in the face of the ongoing change of the competitive paradigm.

Play internet services also play an increasingly important role in boosting the regional competitiveness of a region. One such service is mobile banking, which offers many advantages (Tinashe & Kelvin 2016), such as reducing cost or risk, access to mobile finance, and improved products or services that better satisfy financial system participants (Frame & White 2009). An innovative driver of competitiveness is also the number of enterprise births. Business demographic (number of enterprise births) presents the number of newly registered companies, and thus reflects the level of entrepreneurship and economic activity. At the same time, a high level of competitive pressure enhances innovation adoption because it is important for maintaining a good position in competitive economies (Chwelos et al. 2001). Thus, the readiness to try new and different things has been recognised as an important indicator of competitiveness. Innovations require a propensity to try new ideas, a climate of acceptance for new possibilities, the readiness to explore and implement new ideas, and the so-called proactive approach.

Group 4 focuses on social factors of regional climate change competitiveness. The factors that weaken competitiveness are poverty and social exclusion. People with an equivalised disposable income below the risk-of-poverty threshold, which is 60% of the national median equivalised disposable income (after social transfers), are at the risk of poverty (Eurostat, European Commission, 2020). This ratio has a territorial dimension and contributes to a declining quality of life (Atkinson 2013; Jonsson et al. 2016). Poverty and social exclusion limit the utilisation of human potential and reduce regional growth and prosperity opportunities. The poverty situation also determines important differences in the ability to perform basic daily life activities. Different manifestations of poverty and social exclusion are a shortage of income, the lack of decent housing or material deprivation. The spatial concentration of problematic phenomena leads to severe material deprivation, seen as the lack of access to the opportunities and resources perceived as common in a given society. Deprivation can be strongly differentiated in the territorial dimension and limits the ability of regions to increase their competitiveness (Smętkowski et al. 2015).

Inherent in the territorial aspects of competitiveness are demographic issues (Ostrouch & Sługocki 2018). An increase in population density has a significant and positive impact on innovation output (Shukai et al. 2021) and determines the processes in local labour markets. Population growth depends both on the natural increase and on migration decisions. Human displacement from extreme events could become more frequent with increased climate variability (Thornton et al. 2014). Migration performs a redistributive function, as reflected in the dynamics and nature of demographic processes, so migration has the potential for quantitative and qualitative changes in the composition and structure of the population. Urbanisation is a key feature of economic development, both as a result and a cause (Gallup et al. 1999). The degree of urbanisation defines the share of the population living in cities. The geographical agglomeration of people leads to higher productivity (Krugman 1991; Quigley 2008). Urbanisation also plays a vital role in the economic and social fabric, offering opportunities for education and health services (McKenzie & Sasin 2007) and helping poverty reduction (Ravallion et al. 2007).

Group 5 of regional drivers of climate competitiveness change covers technological factors. In the process of strengthening competitiveness, a game-changer is digitalisation, which can increase productivity, promote novel forms of collaboration, and improve efficiency (OECD 2019). Digitalisation has many economic and social benefits; it transforms the face of science and provides positive trends in reducing the risk of poverty and social exclusion (Kwilinski, Vyshnevskyi & Szwigol 2020). There is also a high demand for digital skills across sectors in the modern, competitive world. Thus, more investment in overall digital skills, including ICT, is required. This will support increasing personal and organisational productivity (WEF, 2014). The quality of the digital infrastructure affects the ability of firms to deploy industry 4.0 technologies, which refers to recent advanced information and communication systems and future-oriented technologies (Sanders et al. 2016). The focus is on digital-technology-based solutions, which modify the way companies create value (Oesterreich & Teuteberg 2016). The major advantages of introducing new technologies are resource savings, more efficient resource utilisation, process transparency, more profitable business models, horizontal integration, higher quality and improved workplace conditions (Staufen 2018; Nosalska et al. 2019; Ali & Azad 2013; Jeschke et al. 2017; Brettel et al. 2014).

From a climate change perspective, new technologies make it possible to reduce emissions or energy consumption. As a result, cleaner production contributes to reduced pollution and increased productivity. Access to technology is also an important competitive factor, and technological advances, including digital infrastructures, drive growth and provide access to environmentally friendly solutions. Without any doubt, science, technology and innovation are features of a knowledge-based economy and drivers of competitive advantage (Rusu & Roman 2018).

The main aim of the article is to present new scientific knowledge in researching the impact of significant competition factors on the climate change competitiveness of regions. It should be mentioned that the concept of regional climate competitiveness is not the same as regional competitiveness. Regional climate change competitiveness is a specific regional competitiveness conditioned by climate policy, anthropogenic air pollution and the physical impact of climate change on the socio-economic system. Indeed, it refers to the specific ability of a region to achieve a competitive advantage under conditions of reduced greenhouse gas emissions but also to develop low-carbon solutions.

The measure of regional climate change competitiveness is the RCCCI index. The main reference framework for its construction was the RCI index (Regional Competitiveness Index, proposed by the EU), which was extended to take into account issues connected with climate change (policy, direct impact, emission by sectors). RCCCI is composed of six sub-indexes (basic, environmental, social, innovative, efficiency, and sectoral) with 89 quantitative items. For example, the basic sub-index includes, among other things, the level of corruption, the level of basic education, the level of inflation, the region’s participation in climate networks; environmental – air quality, water quality and forest cover; social – the number of premature deaths, perceived quality of life and level of environmental awareness; innovative – product and process innovation, innovation readiness; efficiency – labour market efficiency, resource efficiency; and sectoral – emissions from agriculture, tourism, industry, transport, construction and energy. The set of items that populate each sub-index was carefully chosen according to the literature review, international experts’ opinions and data availability (for more detail, see Karman et al. 2022). The overall RCCCI score is the result of a weighted aggregation of the six sub-indexes. The spatial distribution of the RCCCI index for 2020 is shown in Figure 2. In this article, the RCCCI is the dependent variable.

Figure 2.

Regarding the defined aim, the research was conducted in 281 EU regions (NUTS2, according to Classification of Territorial Units for Statistics). Data sources were Eurostat, Regional Innovation Scoreboard and Espon. For selected variables, the latest available data were taken into account (2019, 2020 or 2021). In order to estimate the variable Regional Climate Change Competitiveness (RCCCI), the method proposed by Karman et al. (2021) was used.

The steps of analysis carried out in this study are presented in Figure 3.

Figure 3.

Multiple linear regression (MLR) was used: (1) to quantify the relationship between the variables (according to the value of the estimated variable coefficient); (2) to identify the most important factors determining RCCCI. The aim of applying MLRs is not to forecast the values of the dependent variable in research. The MLRs are one of the appropriate statistical methods for factor evaluation because all variables are metrics. The dependent variable was regional climate change competitiveness (RCCCI). The independent variables (19 variables, such as D, ES, …, ATE; see Research model – Fig. 1) must satisfy the assumptions of linearity and a normal distribution of data.

The assumption of normality was verified by the descriptive characteristics (skewness and kurtosis). The assumption of normal distribution is acceptable if the value of the skewness and kurtosis is in interval <−2; 2> (Arnold 1980). Nonetheless, all parameters are processed to meet the requirements of normal distribution after Box-Cox transformation (along with the minimum value shift). In the final test, the p-values of the Shapiro–Wilk test of all variables are greater than 0.05, and belong to normal distribution. For the entire population, standardisation was used. The coefficient correlation was used to verify the dependence between variables. Autocorrelation was not examined for each regression model because the sample data are not time series (Li & Valliant 2011). The normality distribution of errors is accepted when the p-value of Shapiro–Wilk statistics is greater than the level of significance.

In order to verify that the set of variables explains the highest percentage of variation in the individual model constructs (groups of factors), we used PCA (principal component analysis). The justification for using PCA is that, despite the correlations between the variables, it allows a small set of variables that account for a large portion of the total variance in the original variables to be identified (Bryant & Yarnold 1995).

Next, collinearity diagnostics is used to test collinearities between variables. In order to eliminate the influence brought by some variables with strong collinearity, the backward stepwise regression approach is successively used to improve the data deviation, so as to make the final model more consistent with the actual regression equation. In the final model, only eight (D, ECO₂, M, EB, VA, PAR, PUB, DIG) of the 19 initial variables (see Research model, Fig. 1) remained in the stepwise method model. These variables and RCCCI construct eight-dimensional functions in this model relationship, as follows: CF=β0+β1×CF1+β2×CF2+β3×CF3+…+β8×CF8+ε, {\rm{CF}} = {\beta _0} + {\beta _1} \times {\rm{C}}{{\rm{F}}_1} + {\beta _2} \times {\rm{C}}{{\rm{F}}_2} + {\beta _3} \times {\rm{C}}{{\rm{F}}_3} + \ldots + {\beta _8} \times {\rm{C}}{{\rm{F}}_8} + \varepsilon , where:

CF – the dependent variable (RCCCI);

To describe the correlation among values of variables in dependence on the relative locations between the spatial units, we apply the global spatial autocorrelation test judged by the global Moran index. I=∑i=1n∑j=1nωijYi−YYj−YS2∑i=1n∑j=1nωij I = \frac{{\sum\nolimits_{i = 1}^n {\sum\nolimits_{j = 1}^n {{\omega _{ij}}\left( {{Y_i} - Y} \right)\left( {{Y_j} - Y} \right)} } }}{{{S^2}\sum\nolimits_{i = 1}^n {\sum\nolimits_{j = 1}^n {{\omega _{ij}}} } }} where S² represents the sample variance, Y the observed value of region i, n the total number of regions, and ω the spatial weight matrix element.

The Moran index I value is between −1 and 1. Greater than 0 means positive spatial autocorrelation – that is, there are spatial features of high-high aggregation and low-low aggregation; less than 0 means negative spatial autocorrelation – that is, there are spatial features of high-low aggregation; equal to or close to 0 means no spatial autocorrelation – that is, spatial distribution is random.

Next, we analyse how relationships between RCCCI and variables vary geographically. For this aim, we use geographically weighted regression (GWR). This method extends the framework of the regression model to a local regression model, which allows local parameter estimation (Gujarati 2003). Each regression parameter is estimated at each point geographically, so that the relationship between the variable response (Y) and the explanatory variable (X) varies by location. GWR is therefore a development of a linear regression model at each location by adding weight in the model. Indeed, the MLR assumes fixed data relationships and provides the baseline against which all forms of GWR can be compared. The GWR model estimates the bandwidths for each predictor-to-response relationship. Evaluating these directly quantifies the nature of any spatially varying relationships and at what spatial scale they each operate. For GWR analysis we use the procedure recommended by Fotheringham et al. (2003; for details see Appendix 1).

In the last step, the best subset regression method is used for identifying the most important factors determining RCCCI in clusters. The grouping of regions is based on the RCCCI value. In order to find out the number of clusters, we apply non-hierarchical cluster analysis using the K-means algorithm. The squared Euclidian distance is employed. We choose four clusters. Finally, we use the best subsets approach to find out the best-fit model from all possible subset models. This method consists of testing all possible combinations of the independent variables, and then selecting the best model according to some statistical criteria (Hosmer et al. 2013). In other variable selection methods, only a single model is calculated, and the natural tendency is to accord special status to this model, even though it is statistically indistinguishable from a number of alternative models. Furthermore, collinear variables may be dropped during model development. Best subset regression method may allow a more robust description of RCCCI predictors compared with a single parsimonious prediction model. In order to compare the fit of subset models we use R2, corrected R2, Cp Mallows and AIC statistical models.

The results are calculated using SPSS Statistics and Python.

The assumption of the linearity of all variables was confirmed (according to graphical analysis – histogram). The Shapiro–Wilk tests performed for all variables indicate a distribution deviating from the Gaussian curve (RCCCI=0.95; D=0.96; ES=0.71; ECO₂=0.76; FI=0.68; AFI=0.95; RDE=0.907; VA=0.62; PUB=0.97; BAN=0.93; EB=0.96; TRY=0.96; PAR=0.89; MD=0.80; P=0.34; M=0.84; U=0.78; DIG=0.95; MAN=0.92; ATE=0.72; p=0.000).

Therefore, a Box-Cox transformation was performed separately for variables with positive skewness (assumed lambda value −0.342, shift 1.000) and separately with negative skewness (lambda value 1.961, shift 10.000).

The analysis of descriptive statistics (Table 1) for raw data shows that the highest number of missingness occurs for the variables EB 39%, MD 44% and MAN 43%. These deficiencies result from incomplete reporting of data in the above-mentioned databases. The highest variation was observed for ECO₂, VA, P and M, confirmed by variance and deviation values. Measures of variability indicate a low degree of scatter values of variables RCCCI, AFI, RDE, TRY, PUB, BIG, MAN and ATE around the expected value. There is right skewness and leptokurtic distribution for most variables for D, ES, ECO₂, FI, VA, TRY, P, U and ATE, with extreme outliers for VA and P. Distribution with asymmetry extending in a negative direction was noted for RCCCI, BAN, M, DIG and MAN. Kurtosis values indicate a platocurtic distribution for M and MAN in this group.

Correlation analysis shows significant correlations between the RCCCI variable and technological variables (DIG, MAN, ATE) and the BAN variable (Table 2). Moderate correlation occurs between RCCCI and RDE, VA, PUB, P and M. Correlations with environmental variables (D, ES, ECO₂) confirm that higher climate competitiveness is associated with lower emissions. Moreover, higher competitiveness reduces the risk of poverty (PAR) and material deprivation of residents (MD).

Principal component analysis (PCA) confirms that one principal component is distinguished for each group of factors (Table 3). For the environmental group, the variables explain 46% of the variance, for economic – 40%, for innovative – 61%, for social – 44% and for technological – 73%.

By plotting the standardised predicted values against the standardised residuals, the result of testing linearity through scatter plot diagrams is shown in Figure 4, which shows no evidence of a nonlinear pattern in the residuals.

Figure 4.

An examination of normality is checked with a normal P–P plot and the histogram of the distribution of the residuals. The cumulative probability plots of residuals (P–P plot) are used to judge whether the distribution of variables is consistent with a specified distribution. If the standardised residuals are normally distributed, the scatters should fall on or tightly close up to the normal distribution line. Figure 5 shows that the scatters of the residuals basically fall right on the normal distribution line, indicating a normal distribution of residuals.

Figure 5.

The multicollinearity test is important because if multicollinearity exists between two or more independent variables, the results of multiple regression can be less reliable. In this study, multicollinearity has been examined between the independent variables using VIF (Table 4). The results indicate that multicollinearity does not exist for most independent variables because the tolerance values are more than 0.1 and VIF values are less than 10. The exceptions are the variables BAN, DIG and MAN. For these variables, the VIFs indicate the presence of multicollinearity. Following Vittinghoff (2005), we decided to conditionally accept these variables at this stage because factors such as digitalisation and e-banking must be, due to their impact on the socio-economic system, correlated with other variables. Moreover, the principled exclusion of multicollinear variables alone does not guarantee that the remaining variables are relevant. The exclusion of relevant variables can produce biased regression coefficients, leading to issues more serious than multicollinearity (Kim 2019).

In order to analyse the impact of the competitive indicators on regional competitiveness in climate change conditions, multiple regression analysis was applied. Backward stepwise regression was used. We began with a model that contains all variables under consideration and then removed the least significant variables one after the other. All variables with a p-value > 0.05 were removed. The results of multiple regression analysis are shown in Table 5.

The regression equation (with eight independent variables: D, ECO₂, M, EB, VA, PAR, PUB, DIG) is statistically significantly related to the RCCCI (F (7.72) = 330.025; p = 0.01). The value of the F-statistic and the corresponding probability level confirm the statistical significance of the linear model. The multiple correlation coefficient in the sample is 0.85, indicating that approximately 85% of the variance of Regional Climate Change Competitiveness can be attributed to a linear combination of independent indicators. Due to the multiple collinearities of the arguments, stepwise regression eventually only selects eight indicators to model RCCCI. These establish an eight-dimensional variable relationship with RCCCI. According to the β values, the regression equation (where ɛ – standard error of estimate) is: Y = −0.092 − 0.08D − 0.137ECO2 + 0.066VA + 0.026PUB +0.081EB − 0.036PAR + 0.018M + 0.227DIG + ε \begin{array}{*{20}{c}}{{\rm{Y }} = {\rm{ }} - 0.092{\rm{ }} - {\rm{ }}0.08{\rm{D }} - {\rm{ }}0.137{\rm{EC}}{{\rm{O}}_{\rm{2}}}{\rm{ }} + 0.066{\rm{VA}} + {\rm{ }}0.026{\rm{PUB}} + }\\{0.081{\rm{EB }} - {\rm{ }}0.036{\rm{PAR }} + {\rm{ }}0.018{\rm{M }} + {\rm{ }}0.227{\rm{DIG }} + \varepsilon }\end{array}

Based on the data from Table 5, it can be concluded that the digital skills of people (DIG) have a special role in increasing regional competitiveness. Many economists and policymakers agree that improving qualifications and skills would be key to increasing productivity (and thus the competitiveness of the economy) (Gershman & Kuznetsova 2013). As expected, regional competitiveness growth depends on gross value added (VA). This means that the improvement of this parameter (which can then be decomposed by productivity and profits) enables a region to achieve higher economic “attractiveness” (Martin 2005). Wages and profits are only generated if firms are successful in selling their products and services to local and export markets. Therefore, higher competitiveness of local firms confirms good regional outputs. In addition, it could be seen that the “Scientific publication” variable (PUB) significantly affects competitiveness in the analysed regions. This confirms that scientific publications are prospective to regional competitiveness as suggested in the literature (Krstic et al. 2020). This can be attributed to the fact that increasing the number of quality scientific papers leads to structural changes in the production process, which results in increased competitiveness. Migration (M) is also a factor that positively influences the level of climate competitiveness. Investments in public infrastructure, education and labour market facilities contribute directly to the growth of human capital, innovation and technical progress. Thus, the results confirm that immigrants have the capacity to make the region more prosperous (Papademetriou & Sumption 2011).

As expected, increasing the number of cooling and heating days (D) has a negative impact on competitiveness. This impact especially affects highly climate-sensitive sectors: tourism, energy, agriculture and building construction. One recent analysis has already shown a reduction of regional potential caused by climate conditions (Cvitan 2017; Andrade et al. 2021), and these results are part of this trend. At the same time, our results counter the claim that carbon dioxide emissions (ECO₂) affect climate competitiveness. Although the results of correlation analysis confirm the relationship between these variables, the regression coefficient for ES is not statistically significant. This is due to the presence in the dataset of some influential data points (big outliers) messing with the effect of sizes.

There is a statistically significant but negative association between regional competitiveness and people at risk of poverty (PAR). The negative elasticities mean that as the number of people suffering deficits in income increases, climate competitiveness decreases. Unchecked, climate change will push up to 130 million people into poverty over the next ten years (Nishio 2021), which, through its negative impact on market size and the labour market, will reduce the attractiveness of regions vulnerable to climate change for potential investors and residents. We also noted a negative relationship between RCCCI and CO₂ emissions (ECO₂). These findings are in line with other work (Luo et al. 2021). Emissions impact the economic sphere through reduced business profitability and ETS market costs, the social sphere through impacts on health and the quality of life, and the natural sphere through air quality. In each case the impact is negative.

As climate change does not stop at the administrative boundaries of regions, spatial autocorrelation is therefore included in the analysis.

Appendix 2 shows the Moran index values for all independent variables. It can be seen that under the weight matrix W_1, the global Moran index is significantly larger than 0, and the index values are between 0.03 and 0.46. Most of the variables have positive spatial autocorrelation – that is, regions with higher values are adjacent to other regions with higher values of these variables, and regions with low values of particular variables are adjacent to other low-valuated regions. The regions with high (or low) values tend to concentrate. For DIG, for example, the spatial autocorrelation is 0.46 (with significance p<0.01). AFI values in EU regions show a stronger positive spatial autocorrelation (p<0.01), which reaches 0.36. The value added indicator (VA) also passed the hypothetical test at 1% significance level, and the values are positive, indicating a global significant trend to the geographical clustering of similar regions. On the contrary, the spatial distribution of the variables PUB (0.03) and PAR (0.05) is rather random.

If there are data with spatial effects, the analysis used is spatial regression analysis. The first step in global weighted regression (GWR) analysis is to test for spatial heteroscedastic using the Breusch-Pagan (BP) test. This is conducted to determine whether there is diversity due to spatial effect. The test is conducted for EU regions with a significance level of 5%. Results of the Breusch-Pagan test are presented in Table 6.

Based on Table 6, it is known that observations showed spatial heteroscedastic variable RCCCI. Therefore, it is not enough to use a simple regression method. Thus, the spatial regression method using geographically weighted regression is more appropriate for regional climate change competitiveness.

The GWR model consists of two variables, namely global and local variables. The determination of these variables is based on the Breusch-Pagan test value obtained from the combination of the response variable (Y; RCCCI) with all the independent variables (X₁, X₂, …, X₁₉). The results of the determination of these variables are 381 combinations of Y with all X and some major combinations taken between Y and X (presented in Table 7). Table 8 shows that X₄, X₈, X₁₄ and X₁₇ are global variables because the p-value is greater than the significance level of 0.05. Whereas D, ECO₂, ES, AFI, RDE, VA, BAN, EB, TRY, PAR, MD, P, U, ATE and DIG are local variables because the p-value is less than the significance level 0.05.

Based on the kernel function, the spatial distance (λ) is 1.39 and optimum bandwidth (ℎST) is 0.06. The distance parameters are obtained based on the smallest result of cross-validation.

The results for the GWR model show different parameter estimates for each region. Table 8 shows a summary of the global parameter estimation results.

The parameters β₄, β₈, β₁₄ and β₁₈ are parameters for X₄(FI), X₈(PUB), X₁₄(M) and X₁₇(MAN), which are global variables and have no significant effect on regional climate change competitiveness. Parameters β_1–3, β_5–7, β_9–13, β_15–16 and β_18–19 have significant influence on RCCCI between each region (Table 9).

The GWR model can be formed from all of the β values. All β have different values according to localisation. Based on Table 8 and Table 9, the impact of factors can be written as follows: Y=11.98− 0.28D − 0.35ECO2 − 0.13ES + 0.003FI + 0.132AFI+ 0.228RDE + 0.418VA + 0.052PUB + 0.361BAN + 0.211EB +0.033TRY − 0.096PAR − 0.019MD − 0.007M − 0.058P − 0.182U +0.031MAN + 0.147ATE + 0.279DIG + ε \begin{array}{*{20}{c}}{{\rm{Y}} = 11.98 - {\rm{ }}0.28{\rm{D }} - {\rm{ }}0.35{\rm{EC}}{{\rm{O}}_{\rm{2}}}{\rm{ }} - {\rm{ }}0.13{\rm{ES }} + {\rm{ }}0.003{\rm{FI }} + 0.132{\rm{AFI}}}\\{ + {\rm{ }}0.228{\rm{RDE }} + 0.418{\rm{VA}} + {\rm{ }}0.052{\rm{PUB }} + {\rm{ }}0.361{\rm{BAN }} + 0.211{\rm{EB}} + }\\{0.033{\rm{TRY }} - 0.096{\rm{PAR}} - {\rm{ }}0.019{\rm{MD }} - {\rm{ }}0.007{\rm{M }} - {\rm{ }}0.058{\rm{P }} - {\rm{ }}0.182{\rm{U}} + }\\{0.031{\rm{MAN }} + {\rm{ }}0.147{\rm{ATE }} + {\rm{ }}0.279{\rm{DIG }} + \varepsilon }\end{array}

Finally, the best subset regression is used in different clusters of regional climate change competitiveness. First, based on RCCCI value we group EU regions, resulting in four clusters with means of −0.674, −0.334, 0.142 and 0.421. Boxplot visualisation of cluster analysis grouping EU regions is presented in Appendix 3. Examples of regions assigned to each cluster are shown in Table 10.

Next, we apply the best subset regression to better understand the impact of the analysed variables on climate change competitiveness in different clusters. As this method allows for the inclusion of statistically insignificant variables, all independent variables were included in the analysis. For each cluster, a model with a different number of variables that strongly correlate with the RCCCI value are obtained (Table 10).

The regression equations take the following form:

Cluster 1: Regional Climate Change Competitiveness = 0.124 − 0.34 D − 0.084 ES − 0.199 ECO₂ + 0.024 FI + 0.017 AFI + 0.059 RDE + 0.089 VA + 0.019 PUB + 0.025 BAN − 0.025 EB + 0.043 TRY − 0.09 PAR − 0.092 MD − 0.186 P − 0.009 U + 0.005 DIG + 0.005 MAN + 0.025 ATE + ɛ

When performing the best subset or stepwise approaches, our results identify slightly different models that are considered the best. The adjusted R2 statistic suggests the models with most of the variables are preferred (low and middle–low cluster), the AIC statistic suggests the model with almost all variables (low cluster), and the Mallows suggests the four variable models (high cluster). The simplest five-variable model is composed of ECO₂, BAN, U and ATE; the most complex model includes all variables except M. Although all models include Internet banking (BAN), Urbanisation (U), and Index of access to technology (ATE), there are unique variables in each model. This highlights two important findings. First, as the level of RCCCI increases, the number of determining variables decreases. This means that highly competitive regions have a small group of factors and sources of climate competitiveness that affect their competitive position. Secondly, different indirect error test estimate statistics (Cp, AIC and R2) will likely identify different “best” models.

This paper has examined a hypothesis concerning the relationship between influencing factors and regional climate change competitiveness. Using multiple regression models and related best subset selected methods, we verify how the selected determinants determine climate competitiveness, which can underpin a region’s development under climate transformation.

According to the results of the correlation analysis, in all 281 regions, the level of RCCCI is most related to the variables DIG and ATE. This indicates the key importance of e-competence and technical innovations characteristic of economy 4.0 for building and strengthening regional attractiveness. In turn, the level of climate competitiveness is negatively affected by the emission of carbon dioxide (ECO₂). Considering various comprehensive indicators, eight independent variables, D, ES, EB, VA, PAR, U, BAN and DIG, should be regarded as the most important factors determining the level of RCCCI. They represent environmental (D, ECO₂, ES), social (PAR, U), economic (VA), innovative (EB) and technical (DIG, ATE) factors.

After considering the effects of these variables, the effective description rate of fitted R2 reached 85%. Digital skills have the highest contribution to climate change competitiveness in European regions. The importance of that factor is highlighted by Pyankova et al. (2021) who confirmed the digital economy components affecting both digital and overall region competitiveness. The importance of digital technologies, such as Internet of Things, Artificial Intelligence and Building Information Modelling, in enhancing the climate resilience of critical infrastructure was also noted by Argyroudis et al. (2022). Another factor that determines the level of climate competitiveness is access to advanced technology. Emerging and disruptive technologies have the potential to enhance the climate resilience of critical infrastructure by providing rapid and accurate assessment of asset conditions and supporting decision-making and adaptation. In particular, low-carbon technologies are expected to play a major role in enabling regional decarbonisation processes. The climate for competitiveness must also be associated with environmental factors. This is a two-way relationship: regions contribute to GHG emissions and, on the other hand, climate change poses key threats to firm profitability and quality of life. Our research confirms the negative relationship between regional competitiveness and climate change intensifying factors, particularly CO₂ emissions from territorial fossil fuel combustion. As showed previously by Kasztelan (2020) and Zhang et al. (2020), reducing emissions diminishes environmental pressures and can, in the long term, positively differentiate green regional competitiveness.

We also found divergence in inducing factors in different clusters. In clusters with a low RCCCI, competitiveness is significantly affected by the scale of migration, which is less important (compared to other factors) in other clusters. On the opposite side, in high-level clusters, technological factors (ATE), emission levels (ECO₂), availability of electronic banking (BAN) and the level of urbanisation (U) appear to measure regional competitiveness. In the middle–high cluster, other environmental (D, ES), social (PAR, M) and economic factors (FI, RDE) are also worth considering.

The research findings confirm the hypothesis that selected independent variables have a statistically significant impact on regional climate change competitiveness. This study can help regional governments formulate evidence-based and effective plans and increase regional attractiveness and the resilience of regions under climate change. Based on our results, it is proposed that the measures and activities of regional decision-makers in the coming period should be aimed towards: (1) increasing access to the latest Industry 4.0 technologies and building e-competences, (2) stimulating technology transfer, (3) mitigating the effects of climate change on the local economy, and (4) opening up labour markets for migrants and promoting inclusive growth that reduces the risk of poverty and social exclusion.

This study has some limitations that need to be addressed. Data availability was one of these limitations. All data were extracted from the official database. However, some data were outdated, and there were missing data for some regions due to data disclosure limitations. Another limitation was the technical specification of regression models conditioned by the thematic setting of the research, which implied a short observation period (one year) and a relatively large number of independent variables. A third limitation was the choice of variables. Although the variables analysed were based on the literature, there is some overlap with the variables co-produced by the RCCCI. This may cause an overestimation of the coefficient of determination obtained in this study.

Further scientific contribution in this area could be achieved by verifying the other factors (legal, political, operational, financial, etc.) that positively or negatively impact regional competitiveness. Similarly, based on the geometric growth rate, further research should measure each indicator’s progress within the pillars of climate change competitiveness. In order to obtain more precise insights into the relationship between emissions, economics or technological factors and regional climate change competitiveness, researchers should use other methods (comparative methods in the dynamics of time, benchmarking, etc.).