Climate Change Mitigation and Policy Spillovers in the EU’s Immediate Neighborhood

Annex 1: Methodology: Firm-Level Analysis

To comprehend the impact of carbon taxation on firms, a firm-level analysis was conducted using the World Bank Enterprise Survey (WBES). The WBES represents a comprehensive examination of the private sector through random stratified sampling of a representative sample of an economy, with a vast array of economic data collected from 171,000 firms across 149 countries. The survey encompasses a broad spectrum of business environment themes, including access to finance, corruption, infrastructure, crime, competition, and performance indicators.

To analyze the impact of climate-related policies on firms within the European Union’s immediate neighborhood, a dataset was selected from the World Bank Enterprise Survey (WBES) database covering 11 countries, including Albania, Bosnia and Herzegovina, Kosovo, Moldova, North Macedonia, Montenegro, Serbia, Türkiye, and Ukraine. The selected dataset spans from 2002 to 2013, incorporating 21,793 observations and comprising 12 industry categories, including construction, electricity, gas, and water supply, financial intermediation, health and social work, hotels and restaurants, manufacturing, mining and quarrying, other community, social, and personal service activities, public administration, real estate, transport, storage, and communications, wholesale and retail trade; repair of motor vehicles and motorcycles.

The data collection cycle for the 11 sample countries was conducted at different intervals. In some cases, the survey was conducted one year ahead. The World Bank Enterprise Survey was initiated in 2002, but the analysis in this study is based on data collected from 2002 to 2013. Prior to 2006, the questionnaire structure differed significantly, and efforts are needed to make the conversion, while subsequent to 2019, the focus of the survey shifted to investigating the impact of the COVID-19 pandemic.

To quantitatively assess the potential impact of carbon taxation on firms, the impact was defined as the energy cost relative to the total sales of each individual firm. An alternative approach would have been to calculate the energy cost relative to the production cost. However, only approximately 11,000 observations were available using the production cost method, hence the sales approach was employed in the analysis in order to encompass a broader coverage.

In our analysis, we also evaluated the varying impacts of carbon taxation on firms with distinct characteristics, including firm profitability, firm size, firm age, labor productivity, export activities, factor intensity, and financial constraints. Firm profitability was calculated as the difference between sales and production costs. Firm size was determined by the number of employees, with firms having less than 20 employees classified as small, those with more than 20 employees but less than 100 classified as medium, and those with more than.

100 employees classified as large. Firm age was determined by subtracting the established year from the current year. Labor productivity was calculated as the ratio of the difference between sales and material cost to the number of employees, and reflected the value contributed by a single employee. Firms engaged in exporting activities were captured by the export share in their sales. The factor intensity was comprised of labor intensity and skill intensity, with labor intensity calculated as the ratio of total labor cost to sales, and skill intensity calculated as the ratio of total labor cost to the number of employees. Financial constraints were identified through a dummy variable, defined as firms answering affirmatively to facing constraints on interest rates, fees, or collateral.

We conducted a fixed-effect regression analysis to estimate the firm-level energy expenses as a proportion of total costs, with respect to all of the identified characteristics.

In this regression model, i represents the individual firm within industry j located in country c during year t. The country fixed effects country _c control for all time-invariant macroeconomic conditions specific to the country. The time fixed effects year _t control for global business cycles as the survey was conducted at different points in time. The industry fixed effects industry _j control for industry-specific factors in the impact of carbon taxation. The coefficient β represents the impact of a specific firm characteristic on the firm’s energy dependency within the context of carbon taxation.

Our results were found to be robust under alternative identification approaches for each firm characteristic. For firm profitability, the results were consistent regardless of whether we used dummy variables for high profitability,^[24] profitability ratios, or the natural log of profit. For firm size, the results were significant whether we used a categorized firm size dummy variable or a continuous measure of the number of employees. The results for firm age were also robust using either a continuous measure or the natural log of age. For labor productivity, the results were robust using both the measure of sales divided by the number of employees, and the measure of sales excluding material cost divided by the number of employees. The same was found to be true for labor intensity, which maintained robustness using either the measure of total labor cost divided by the number of employees or total labor cost excluding material cost divided by the number of employees. In conclusion, the results were robust with alternative methods for identifying firm characteristics, and with the addition of control variables in the specific regression models.

Note: Profitability 1,2,3,4 is profit, dummy variables for high profitability, the natural log of profit profitability ratios and profitability ratios accordingly. Size 1,2 is a categorized firm size dummy variable and a continuous measure of the number of employees. Labor productivity1,2 is the measure of sales divided by the number of employees, and the measure of sales excluding material cost divided by the number of employees. Labor intensity 1,2 is the measure of total labor cost divided by the number of employees and total labor cost excluding material cost divided by the number of employees.

Annex 2: Methodology: Energy Efficiency Analysis

Following Narayan and Narayan (2010), Özbuğday and Erbaş (2015) and Cevik (2022; 2023; 2024), the econometric model used to analyze the impact of energy efficiency on CO₂ emissions takes the following form in a panel setting:

where y _i,t denotes the logarithm of CO₂ emissions per capita in country i and time t; EE _i,t is energy efficiency as measured by the logarithm of energy consumption per unit of real GDP; X _i,t is a vector of control variables including the logarithm of real GDP per capita, trade openness, share of industry in GDP, population, share of urban population, and a measure of institutional quality. As above, the η _i and μ _t coefficients denote the time-invariant country-specific effects and the time effects controlling for common shocks that may affect CO₂ emissions across all countries in a given year, respectively. ε _i,t is the error term. To account for possible heteroskedasticity, robust standard errors are clustered at the country level.

Environmental outcomes are measured in terms of CO₂ emissions in metric tons per capita, which represent more than 80 % of GHG emissions in Europe. The main explanatory variable of interests are energy efficiency as measured by energy consumption per unit of real GDP and the share of nuclear, renewable and other non-hydrocarbon sources of energy, which show considerable heterogeneity across countries and over time. The empirical analysis also includes a variety of economic, demographic and institutional variables to control for conventional factors affecting environmental outcomes.

The data series is taken from the IMF’s International Financial Statistics and World Economic Outlook databases, the World Bank’s World Development Indicators database, the U.S. Energy Information Administration, and the International Country Risk Guide.

Annex 3: Climate Policy Database

The Climate Policy Database serves as a vital resource for our report, featuring the latest dataset released in May 2022, encompassing a staggering 5783 policies from 198 countries. Maintained by the New Climate Institute with support from PBL Netherlands Environmental Assessment Agency and Wageningen University and Research, this database aims to gather information on climate mitigation policies and benchmark them against a comprehensive policy matrix. The policy database covers national mitigation-related policies and is updated periodically, providing us with the latest information on climate policies worldwide.

To analyze the evolution of climate policy adoption, the Climate Policy Database created a Policy Matrix that consists of 50 policy options distributed across six sectors and five mitigation areas. This matrix also incorporates eight policy instruments, which are essential for bridging the gap between policy objectives and their actual implementation. The eight policy instruments, which include economic instruments, regulatory instruments, information and education, policy support, research, development and deployment (RDD), voluntary approaches, climate strategy and target, serve as the primary categories analyzed in this report, as Table Annex 3 below.

In Climate Policy Database, policies are identified and combined with various policy instruments to constitute a comprehensive global mitigation package. The policies included in this database are selected based on their potential to contribute to emissions reductions as generally agreed by experts (IPCC, 2014), sector-level example policies that have been successful in specific contexts, or policies that are expected to lead to sufficient sectoral transformation to achieve emissions reductions.

Annex 4: Spillover Analysis

Our empirical analysis focuses on examining the evidence of spillover effects arising from the increased stringency of climate policies in the European Union (EU) and its association with emissions reduction in European Neighborhood (EUN) countries. To evaluate the influence of EU’s climate mitigation policies on emissions behavior in its smaller neighboring EUN countries, we employ a panel dataset comprising over 100 countries during the sample period of 2002-2019. To test the relationship between EU environmental policy stringency and CO₂ emissions, we regress country-level CO2 emission per capita on the stringency of EU climate policies while controlling for income, squared income, and regulatory quality as below:

The regression examines the relationship between CO₂ emissions and a set of independent variables, including Log(Real GDP PPP), Eustringency, Eustringency * Dummy_EUN, and Regulatory Quality. The coefficients of these variables, denoted β₁, β₂, β₃, and β₄, respectively, are estimated using a regression model. Here, Dummy_EUN is a binary variable that takes a value of 1 if the country is a member of our sample in the EU neighborhood, and 0 otherwise. The regulatory quality variable in our study is measured using the Worldwide Governance Indicators Index from the World Bank. The inclusion of the variable (Lag_Square Real GDP PPP) allows us to examine the potential effects of economic growth on CO₂ emissions. Fixed effects for each country i, denoted αᵢ, and each year t, denoted δ_t, are employed to control for unobserved heterogeneity and time-specific factors that may affect CO₂ emissions. The error term, denoted μ, captures all other factors that affect CO₂ emissions but are not included in the model.

To ensure the robustness of our findings, we conducted additional analyses using alternative variables and specifications. Specifically, we examined the use of emission intensity as a dependent variable instead of CO₂ emissions per capita, and we used the time lag of GDP growth rate instead of real GDP PPP as an independent variable. Our results showed that the findings were consistent with our baseline scenario, as well as with a time lag of GDP PPP and GDP square. We also investigated the effect of regulatory quality on the relationship between CO₂ emissions and independent variables. Interestingly, we found that regulatory quality significantly affects CO₂ emissions per capita, but not emission intensity. This suggests that the impact of regulatory quality on CO₂ emissions may vary depending on the measure of emissions used. Our robustness checks provide further support for the main findings of our study and contribute to a comprehensive understanding of the factors influencing CO₂ emissions in the studied countries.