The Effect of ESG Factors and Financial Variables on Firm Efficiency in Emerging Markets Purpose and Content of the Research
This research measures the productivity levels of firms in emerging markets and analyses the effects of ESG factors and financial variables on this productivity. The study provides a comprehensive assessment using three different analytical methods (DEA, Bayesian Network and Quantile-Kantile Regression).
Data Envelopment Analysis Methodology
Data and Variables
This study employs Data Envelopment Analysis (DEA) to assess the efficiency of firms across multiple countries using financial and sustainability-related indicators. The dataset consists of firm-level panel data extracted from ESG_newdata.xlsx, covering firms across various years and countries.
Variables Selection
The variables used in the DEA models are classified into input and output variables as follows:
- Input Variables:
- \(capex\_norm\) - Capital Expenditures (Normalized)
- \(casheq\) - Cash Equivalents
- \(leverage\) - Leverage Ratio
- \(wcap\_norm\) - Working Capital (Normalized)
- \(tangible\_norm\) - Tangible Assets (Normalized)
- Output Variables:
- \(roe\_norm\) - Return on Equity (Normalized)
- \(tobin\_norm\) - Tobin’s Q (Normalized)
- \(adh\) - Asset Diversification Heuristic
To ensure comparability, all variables were standardized. Missing values were removed using listwise deletion, and negative values were translated to ensure non-negativity by shifting the minimum value:
\[ X' = X + |\min(X)| + 0.01 \]
DEA Model Specification
DEA is a non-parametric linear programming technique used to estimate the relative efficiency of decision-making units (DMUs) (Charnes, Cooper, & Rhodes, 1978). This study employs both Constant Returns to Scale (CRS) and Variable Returns to Scale (VRS) DEA models.
Input-Oriented CRS Model
The CRS model assumes that firms operate under an optimal scale. The efficiency score \(\theta\) for each firm is calculated as:
\[ \min_{\theta, \lambda} \quad \theta \]
subject to:
\[ Y \lambda \geq Y_o \]
\[ \theta X_o \geq X \lambda \]
\[ \lambda \geq 0 \]
where: - \(X_o\) and \(Y_o\) are the inputs and outputs of the firm under evaluation, - \(X\) and \(Y\) represent all firms’ inputs and outputs, - \(\lambda\) is a weight vector.
Input-Oriented VRS Model
The VRS model, proposed by Banker, Charnes, and Cooper (1984), introduces a convexity constraint:
\[ \sum_{j=1}^{n} \lambda_j = 1 \]
This allows for the decomposition of efficiency into pure technical efficiency and scale efficiency.
Scale Efficiency and Returns to Scale
Scale efficiency (\(SE\)) is computed as:
\[ SE = \frac{Eff_{CRS}}{Eff_{VRS}} \]
The nature of Returns to Scale (RTS) is determined by examining the sum of the intensity variables \(\sum \lambda_j\): - \(\sum \lambda_j = 1\) implies Constant Returns to Scale (CRS) - \(\sum \lambda_j > 1\) indicates Decreasing Returns to Scale (DRS) - \(\sum \lambda_j < 1\) suggests Increasing Returns to Scale (IRS)
Results and Analysis
The efficiency scores were computed for different years and countries. The VRS efficiency scores by country are presented in Figure 1.
The distribution of VRS efficiency across countries is shown in Figure 1, which reveals variations in performance among different nations.
Figure 1: VRS Efficiency Scores by Country
This figure presents the Variable Returns to Scale (VRS) efficiency scores of firms across different countries, computed using Data Envelopment Analysis (DEA) under an input-oriented BCC model. The boxplot format visualizes the median (black line), interquartile range (IQR), whiskers (1.5 Ă— IQR), and outliers (dots).
The red dashed line at 1.0 represents the optimal efficiency frontier, indicating firms achieving full efficiency.
Argentina (ARJ) exhibits the highest median VRS efficiency, with firms clustered closer to the efficient frontier.
Turkey (TUR), Egypt (EGYPT), and South Africa (SAFRC) show moderate efficiency levels, with variability among firms.
Countries such as Russia (RUS), Pakistan (PAK), and Malaysia (MAL) demonstrate broader distributions, indicating disparities in firm efficiency within these regions.
The presence of numerous outliers suggests that some firms in certain countries operate at significantly higher efficiency levels than their peers.
These results indicate cross-country variations in firm efficiency, influenced by economic conditions, governance structures, and financial management strategies. Future research should explore the factors driving efficiency disparities at both the country and firm levels.
To understand country-wise efficiency trends over time, a longitudinal analysis was conducted. Figure 2 presents the evolution of efficiency from 2012 to 2022.
Figure 2: Country Efficiency Performance Over Time
This heatmap illustrates the average VRS efficiency scores for each country over time. The color gradient ranges from low efficiency (dark green) to high efficiency (yellow).
Argentina (ARJ), Turkey (TUR), and Egypt (EGYPT) maintain consistently high efficiency levels over time.
Some countries, such as Russia (RUS) and Pakistan (PAK), exhibit fluctuations, indicating possible economic shocks or policy changes affecting firm performance.
Gaps (white spaces) in the heatmap suggest missing data for certain countries in specific years.
This figure underscores the temporal dynamics of efficiency, highlighting countries with consistent economic stability versus those experiencing periodic efficiency fluctuations. Future studies could investigate the role of macroeconomic factors, policy interventions, and sectoral growth in driving these trends.
The average efficiency scores computed using CRS, VRS, and Scale Efficiency models are plotted in Figure 3.
Figure 3: Efficiency Trends Over Time
This line chart presents the evolution of different efficiency measures (CRS, VRS, and Scale Efficiency) over time.
VRS efficiency (yellow line) consistently remains the highest, followed by Scale Efficiency (green) and CRS efficiency (purple).
A slight dip in 2020 suggests possible disruptions, potentially linked to the COVID-19 pandemic, economic crises, or shifts in global markets.
The gap between CRS and VRS efficiency suggests that scale inefficiencies exist, meaning firms are not operating at optimal size.
This trend analysis indicates that technical efficiency is higher than scale efficiency, implying that most inefficiencies stem from scale-related issues rather than poor resource utilization. Policymakers and business leaders should focus on optimal firm size adjustments to enhance overall economic efficiency.
The distribution of VRS efficiency across different years is shown in Figure 4, which highlights fluctuations in efficiency levels.
Figure 4: VRS Efficiency Distribution by Year
This violin plot illustrates the distribution of VRS efficiency scores across different years. The plot combines a boxplot (median and IQR) with kernel density estimates, showing the spread and concentration of efficiency scores for each year.
The red dashed line at 1.0 represents the full efficiency threshold.
The distribution remains fairly stable across years, with median efficiency scores fluctuating around 0.7.
The density of high-efficiency firms increases in later years, indicating improved performance over time.
There is a persistent tail of inefficient firms, suggesting that some companies struggle with resource allocation and operational optimization.
The results indicate a gradual improvement in firm efficiency over time, which could be attributed to technological advancements, financial reforms, and evolving corporate strategies. However, persistent inefficiencies in some firms warrant further investigation into sector-specific and regulatory challenges.
A firm-level analysis was conducted to identify the top 3 most efficient firms in each country. The results are summarized in Figure 5.
Figure 5: Top 3 Companies by VRS Efficiency in Each Country
This bar chart displays the top 3 firms with the highest VRS efficiency scores for each country. The firms are ranked based on their average efficiency scores, computed using the DEA model. The color coding corresponds to the respective countries.
Companies in Pakistan (PAK), India (INDIA), and Egypt (EGYPT) exhibit the highest efficiency levels, suggesting that certain sectors within these economies are performing optimally.
Lucky Cement (Pakistan), Tata Consultancy Services (India), and Elswedy Electric (Egypt) rank among the top-performing firms.
Firms in Vietnam (VIET), Turkey (TUR), and Brazil (BRA) also demonstrate high efficiency scores, indicating strong sectoral performance in these regions.
This figure highlights the firm-specific drivers of efficiency, suggesting that corporate governance, industry structure, and resource allocation strategies play crucial roles in determining firm performance. Further sectoral analysis could provide deeper insights into why certain companies outperform their peers within the same economic landscape.
Conclusion
These figures collectively provide a comprehensive assessment of firm efficiency across countries and over time. The analysis suggests that:
Certain countries (e.g., Argentina, Turkey, and Egypt) consistently achieve high efficiency.
Firm-level efficiency varies significantly within countries, with top firms consistently outperforming their peers.
Efficiency has generally improved over time, but scale inefficiencies remain a key challenge.
This methodology provides a comprehensive DEA-based efficiency assessment of firms using ESG-related financial indicators. The study employs both CRS and VRS models to distinguish between technical and scale efficiency, while trend analysis and robustness checks validate the findings. Future research could extend the analysis to alternative DEA models, such as Slacks-Based Measure (SBM) or Directional Distance Function (DDF), to refine efficiency estimation.
References
- Banker, R. D., Charnes, A., & Cooper, W. W. (1984). Some models for estimating technical and scale inefficiencies in data envelopment analysis. Management Science, 30(9), 1078-1092.
- Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efficiency of decision-making units. European Journal of Operational Research, 2(6), 429-444.
Bayesian network Methodology
Data Description and Preprocessing
This study utilizes a dataset containing financial and environmental, social, and governance (ESG) performance indicators alongside firm efficiency scores. The dataset, titled ESG_data_with_efficiency.csv, comprises variables including variable returns to scale efficiency (VRS), constant returns to scale efficiency (CRS), and scale efficiency, among others. The financial indicators encompass capital expenditure (capex), cash equivalents (casheq), leverage, working capital (wcap), tangible assets, return on equity (ROE), Tobin’s Q ratio, and asset-to-debt ratio (ADH). ESG performance indicators include environmental, social, and governance scores.
To ensure data quality, infinite values were replaced with missing values, and cases with missing values were removed:
\[ X_{ij} = \begin{cases} \text{NA}, & \text{if } X_{ij} = \infty \\ X_{ij}, & \text{otherwise} \end{cases} \]
Bayesian Network Structure Learning
A Bayesian Network (BN) is a probabilistic graphical model representing conditional dependencies between variables using a directed acyclic graph (DAG) \(G = (V, E)\), where \(V\) is the set of nodes (variables), and \(E\) is the set of directed edges indicating dependencies (Pearl, 1988).
Structural Learning Approach
To learn the optimal BN structure, we applied Hill-Climbing (HC) algorithm and Tabu Search, which are widely used heuristics for maximizing the Bayesian Gaussian equivalent (BGe) score. The BGe score is defined as:
\[ BGe(G | D) = \int P(D | \theta, G) P(\theta | G) d\theta \]
A whitelist was incorporated to enforce theoretically supported dependencies, ensuring that financial indicators influence firm efficiency, and ESG subcomponents impact overall ESG scores (Koller & Friedman, 2009).
Model Training and Evaluation
To compare the models, the Bayesian Information Criterion (BIC) was used:
\[ BIC = -2 \log L + k \log n \]
where \(L\) represents the likelihood of the model given the data, \(k\) is the number of parameters, and \(n\) is the sample size. The model with the higher BIC score was selected for further analysis.
Additionally, arc strength analysis was conducted using the BGe criterion to determine the robustness of dependencies, identifying the most significant relationships in the network.
Results of Bayesian Network Analysis
The Bayesian Network structure was visualized to highlight the strongest relationships identified through the analysis.
The figure illustrates the Bayesian Network structure that captures the causal dependencies between financial indicators, firm efficiency, and ESG performance. The directed acyclic graph (DAG) represents the relationships between variables, where nodes correspond to observed financial and ESG-related factors, and edges denote probabilistic dependencies.
Key insights from the figure:
Highlighted Strongest Relationships: The red edges represent the strongest relationships, as identified through Bayesian network structural learning. These relationships exhibit the highest arc strengths, indicating significant influence among variables.
ESG and Financial Dependencies: ESG components (environment, social, governance) strongly influence overall ESG performance, with governance and social factors showing the most substantial effects. These dependencies align with prior literature emphasizing the role of corporate governance and social responsibility in ESG assessments.
Firm Efficiency and Financial Indicators: Firm efficiency metrics, particularly VRS efficiency, CRS efficiency, and scale efficiency, exhibit dependencies on financial variables such as leverage, tangible assets, and capital expenditure. These relationships highlight the importance of financial management in determining firm operational efficiency.
Hierarchical Efficiency Relationships: The network confirms that CRS efficiency is influenced by both scale efficiency and VRS efficiency, demonstrating the hierarchical nature of efficiency in firms.
Overall, this Bayesian Network provides a data-driven framework to understand the interactions between financial, ESG, and efficiency factors, supporting causal inference in firm performance analysis.
Key Relationships in the Bayesian Network
The table below presents the most influential relationships in the Bayesian Network, categorized by their impact on ESG performance and firm efficiency.
| Category | From | To | Strength |
|---|---|---|---|
| ESG Performance Influences | Social | ESG | -4006.23 |
| Governance | ESG | -3672.03 | |
| Environment | ESG | -3167.42 | |
| Firm Efficiency Dependencies | Scale Efficiency | CRS Efficiency | -3291.95 |
| VRS Efficiency | CRS Efficiency | -2174.93 | |
| VRS Efficiency | Scale Efficiency | -788.18 | |
| Financial Drivers of Efficiency | Tangible Assets | VRS Efficiency | -629.70 |
| Leverage | VRS Efficiency | -613.10 |
Bootstrap Analysis for Robust Edge Identification
To assess the reliability of network dependencies, bootstrap resampling (R = 200) was applied. The top 10% most reliable edges were identified using the quantile threshold:
\[ T = Q_{0.9}(\text{Strength}) \]
Edges exceeding this threshold were retained in the final Bayesian Network structure, ensuring statistical robustness in the identified relationships.
Interpretation of Key Findings
- ESG Performance Drivers: The analysis indicates that social and governance factors have the most substantial negative influence on ESG scores, with environmental factors also playing a significant role. This suggests that firms with governance risks and social controversies tend to have lower ESG ratings.
- Firm Efficiency and Financial Indicators: Firm efficiency is significantly influenced by financial factors such as leverage, tangible assets, and capital structure. Financial constraints are associated with lower efficiency levels.
- Hierarchical Efficiency Relationships: The dependencies identified confirm that CRS efficiency is strongly influenced by scale efficiency and VRS efficiency, highlighting the importance of scale adjustments in efficiency performance.
In the Bayesian Network analysis, the negative values in the Strength column represent the direction and magnitude of influence between variables. These values are derived from the Bayesian Gaussian equivalent (BGe) score, which measures conditional dependencies among the variables.
A negative strength value indicates an inverse or negative relationship between the variables.
When the strength between two variables is negative (e.g., Social → ESG: -4006.23), it suggests that higher values of the predictor variable (Social) are associated with lower values of the dependent variable (ESG).
Conclusion
This study presents a robust statistical framework for analyzing the complex relationships between ESG performance, financial indicators, and firm efficiency. The Bayesian Network analysis provides insights into the most influential dependencies, emphasizing the role of governance and financial structure in firm efficiency.
Future research could extend this analysis by incorporating macroeconomic factors or leveraging dynamic Bayesian models to examine temporal dependencies (Friedman et al., 1999).
References
- Friedman, N., Geiger, D., & Goldszmidt, M. (1999). Bayesian network classifiers. Machine Learning, 29(2-3), 131-163.
- Koller, D., & Friedman, N. (2009). Probabilistic graphical models: Principles and techniques. MIT press.
- Pearl, J. (1988). Probabilistic reasoning in intelligent systems: Networks of plausible inference. Morgan Kaufmann.
Quantile-on-Quantile Regression Methodology
This study also investigates the complex relationships between Environmental, Social, and Governance (ESG) factors, financial variables, and firm efficiency. Using a Quantile-on-Quantile Regression (QQR) approach, we examine how these relationships vary across different segments of the distribution, providing insights beyond what traditional mean-based regression methods can offer.
Our dataset comprises an unbalanced panel of firms across multiple emerging market countries (Argentina, Brazil, Indonesia, Malaysia, Mexico, South Africa, and Turkey) spanning from 2013 to 2022. The analysis includes:
- Dependent Variables: Three efficiency measures
calculated using Data Envelopment Analysis (DEA)
- Variable Returns to Scale (VRS) efficiency
- Constant Returns to Scale (CRS) efficiency
- Scale efficiency
- Independent Variables:
- ESG Variables: Overall ESG score, Environment, Social, and Governance scores
- Financial Variables: Return on Equity (ROE), Capital Expenditure (CAPEX), Tobin’s Q, Working Capital, Tangible Assets, Asset Depreciation and Amortization (ADH), Leverage, and Cash Equivalents (all appropriately normalized)
Methodology
We employ Data Envelopment Analysis (DEA) to calculate three complementary efficiency measures: variable returns to scale (VRS) technical efficiency, constant returns to scale (CRS) technical efficiency, and scale efficiency. DEA is a non-parametric approach that constructs a frontier of the most efficient decision-making units and measures the relative efficiency of each unit against this frontier.
While traditional regression methods focus on the effects at the mean of the distribution, and standard quantile regression examines how independent variables affect different quantiles of the dependent variable, the Quantile-on-Quantile Regression (QQR) approach allows us to examine how quantiles of independent variables affect quantiles of the dependent variable.
Quantile regression, introduced by Koenker and Bassett (1978), allows for the estimation of the relationship between variables at different points in the conditional distribution of the dependent variable. The \(\tau\)th quantile regression model can be expressed as:
\[Y_i = X_i'\beta_\tau + \varepsilon_{i,\tau}\]
where \(Y_i\) is the dependent variable (efficiency score), \(X_i\) is a vector of independent variables (ESG and financial metrics), \(\beta_\tau\) is the vector of coefficients that varies with the quantile level \(\tau\), and \(\varepsilon_{i,\tau}\) is the error term. The estimation of \(\beta_\tau\) is achieved by solving:
\[\min_{\beta_\tau} \sum_{i=1}^n \rho_\tau(Y_i - X_i'\beta_\tau)\]
where \(\rho_\tau(u) = u(\tau - I(u < 0))\) is the check function, and \(I(\cdot)\) is the indicator function.
The QQR approach developed by Sim and Zhou (2015) extends traditional quantile regression by exploring how quantiles of independent variables affect quantiles of the dependent variable. This approach captures the complex dependencies between ESG factors, financial variables, and firm efficiency across their respective distributions.
The QQR model for the relationship between firm efficiency and each explanatory variable can be expressed as:
\[Q_{Y}(\tau_y|X) = \alpha(\tau_y, \tau_x) + \beta(\tau_y, \tau_x) \cdot Q_X(\tau_x)\]
where \(Q_{Y}(\tau_y|X)\) is the \(\tau_y\)th conditional quantile of efficiency (Y), \(Q_X(\tau_x)\) is the \(\tau_x\)th quantile of the explanatory variable (X), and \(\alpha(\tau_y, \tau_x)\) and \(\beta(\tau_y, \tau_x)\) are the intercept and slope parameters that vary with both \(\tau_y\) and \(\tau_x\).
The estimation of QQR follows a two-step approach:
First, we calculate the quantile of the independent variable \(X\) at level \(\tau_x\).
Second, we perform a kernel-weighted quantile regression of \(Y\) on \(X\) at level \(\tau_y\), where the kernel weights are determined by the proximity of each observation’s \(X\) value to the \(\tau_x\)th quantile of \(X\).
We use a Gaussian kernel:
\[K(u) = \frac{1}{\sqrt{2\pi}}e^{-\frac{u^2}{2}}\]
with bandwidth \(h\), and calculate weights as:
\[w_i = \frac{K\left(\frac{X_i - Q_X(\tau_x)}{h}\right)}{\sum_{j=1}^n K\left(\frac{X_j - Q_X(\tau_x)}{h}\right) \cdot h}\]
The QQR estimator at quantile levels \((\tau_y, \tau_x)\) is then obtained by solving:
\[\min_{\alpha, \beta} \sum_{i=1}^n w_i \cdot \rho_{\tau_y}(Y_i - \alpha - \beta X_i)\]
We implement this estimation procedure for a grid of quantiles \(\tau_y, \tau_x \in \{0.1, 0.25, 0.5, 0.75, 0.9\}\), resulting in 25 coefficient estimates for each variable pair. This allows us to comprehensively map the relationship between ESG factors, financial variables, and firm efficiency across their respective distributions.
Our analysis proceeds in several steps:
Baseline QQR Analysis: We examine the relationship between each ESG and financial variable and the three efficiency measures using the QQR approach.
Temporal Analysis: We perform year-specific QQR analyses for selected key variables to investigate the evolution of these relationships over time.
Cross-Country Analysis: We conduct country-year specific analyses to explore geographical heterogeneity in the ESG-efficiency relationship.
Comparative Analysis: We compare the effects of ESG variables against financial variables to assess their relative importance in explaining firm efficiency.
Variable Importance Analysis: We rank variables based on their average absolute effect sizes to identify the most influential factors for firm efficiency.
Empirical Results
Variable Importance
Figure 1 presents the ranking of variables by their average absolute effect size on firm efficiency. The results reveal a clear distinction between financial and ESG variables in their impact on firm efficiency.
As evident from Figure 1, financial variables, particularly ROE, CAPEX, and Tobin’s Q, have substantially larger effects on firm efficiency compared to ESG variables. ROE demonstrates the strongest impact with an average absolute effect size of approximately 21, followed by CAPEX (5.0) and Tobin’s Q (3.2). Working capital and tangible assets also show substantial effects (2.1 and 2.0, respectively). In contrast, ESG variables (environment, social, governance, and overall ESG score) exhibit much smaller effect sizes, all below 0.01.
Effects Across Efficiency Quantiles
Figure 2 illustrates how the effects of each variable vary across different efficiency quantiles when the independent variable is at its median level (Ď„x = 0.5).
The figure reveals notable heterogeneity in how variables affect firms at different efficiency levels. For instance:
ROE: Shows strong positive effects for all efficiency measures, with the highest impact on more efficient firms (higher quantiles), particularly for scale efficiency (blue line).
ESG Score: Demonstrates varying effects across efficiency quantiles, with generally negative effects for CRS efficiency at higher quantiles but more mixed effects for VRS efficiency.
Financial Variables (CAPEX, Tobin’s Q): Generally show stronger and more positive effects on efficiency than ESG variables, especially for more efficient firms.
Environmental Score: Exhibits predominantly negative effects, particularly for firms at the high-efficiency quantiles.
This differential impact across the efficiency distribution underscores the value of the QQR approach in capturing complex relationships that would be missed by mean-based regression methods.
Comparative Analysis: ESG vs. Financial Variables
Figure 3 presents a direct comparison of the average absolute effect sizes between ESG and financial variables for each efficiency measure.
The comparison reveals a substantial gap between the impacts of financial variables and ESG variables across all efficiency measures. Financial variables have average absolute effect sizes that are several orders of magnitude larger than those of ESG variables:
- For CRS efficiency: Financial variables (5.13) vs. ESG variables (0.009)
- For scale efficiency: Financial variables (4.80) vs. ESG variables (0.008)
- For VRS efficiency: Financial variables (2.66) vs. ESG variables (0.006)
This suggests that while ESG considerations may play a role in firm efficiency, traditional financial variables remain the dominant determinants.
Heatmap Analysis of ESG Variables
Figure 4 presents heatmaps showing the coefficients of ESG variables on efficiency measures across different quantile combinations.
The heatmaps reveal several important patterns:
Heterogeneous Effects: The relationships between ESG variables and efficiency measures vary substantially across quantiles of both variables, justifying the use of the QQR approach.
Mixed Social Impacts: The social dimension of ESG shows some positive effects (yellow areas) for high quantiles of both the social score and efficiency measures, particularly for VRS efficiency.
Governance Effects: Governance scores show varied impacts, with some positive effects for firms in the middle efficiency quantiles when governance is in higher quantiles.
Environmental Dimension: Environmental scores show predominantly neutral to negative effects across most quantile combinations.
These nuanced patterns suggest that the relationship between ESG and firm efficiency is complex and varies across different segments of the distribution, with some ESG dimensions having potentially positive effects in specific contexts.
Conclusion
This study finallly employed a Quantile-on-Quantile Regression approach to examine the complex relationships between ESG factors, financial variables, and firm efficiency. Our findings reveal several important insights:
Financial variables, particularly ROE, CAPEX, and Tobin’s Q, have substantially larger effects on firm efficiency compared to ESG variables.
The relationships between independent variables and efficiency measures vary considerably across quantiles, highlighting the value of the QQR approach in capturing heterogeneous effects.
ESG variables, while having smaller overall effects than financial variables, show complex patterns of influence that vary across the distribution of both ESG scores and efficiency levels.
Some ESG dimensions (particularly social and governance) show positive effects in specific contexts, suggesting potential complementary benefits to financial performance under certain conditions.
These findings contribute to our understanding of the complex interplay between ESG considerations, financial factors, and firm efficiency. While traditional financial variables remain the dominant determinants of efficiency, the nuanced effects of ESG factors identified in this study suggest that sustainability considerations may play an important complementary role, particularly for certain types of firms and in specific market segments.
References
Banker, R. D., Charnes, A., & Cooper, W. W. (1984). Some models for estimating technical and scale inefficiencies in data envelopment analysis. Management Science, 30(9), 1078-1092.
Charnes, A., Cooper, W. W., & Rhodes, E. (1978). Measuring the efficiency of decision making units. European Journal of Operational Research, 2(6), 429-444.
Koenker, R., & Bassett Jr, G. (1978). Regression quantiles. Econometrica, 46(1), 33-50.
Ma, L., & Koenker, R. (2006). Quantile regression methods for recursive structural equation models. Journal of Econometrics, 134(2), 471-506.
Sim, N., & Zhou, H. (2015). Oil prices, US stock return, and the dependence between their quantiles. Journal of Banking & Finance, 55, 1-8.