Operational efficiency and EV adoption in global automotive manufacturers : a DEA and regression based analysis
Akter, Fatema (2026)
Pro gradu -tutkielma
2026
School of Business and Management, Kauppatieteet
Kaikki oikeudet pidätetään.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi-fe20260724112453
https://urn.fi/URN:NBN:fi-fe20260724112453
Tiivistelmä
This thesis examines the relative operational efficiency of major global automotive manufacturers during the electric vehicle (EV) transition from 2010 to 2024. It asks whether firms differ in how efficiently they turn productive resources into revenue in this period, and whether those differences are linked to EV adoption, profitability and firm size.
The empirical analysis uses annual firm-level data for ten large automotive manufacturers. An input-oriented CCR Data Envelopment Analysis (DEA) model is applied to estimate an efficiency score for each firm-year relative to a pooled frontier. These DEA scores are then used as the dependent variable in pooled, year-by-year and firm-level linear regression models, where EV adoption, profitability and firm size serve as explanatory variables. The purpose of the regressions is to test whether variation in DEA-based efficiency can be statistically explained by these three firm-level characteristics and, in particular, whether higher EV adoption is linearly associated with higher or lower operational efficiency.
The DEA results show that relative operational efficiency varies across firms and years. The mean efficiency score is 0.935, with values ranging from 0.653 to 1.000. Across the period, a small subset of firms defines the frontier in most years, while others move between relatively efficient and relatively inefficient positions. For example, Tesla and Toyota are often on or near the frontier, whereas Ford and GM are less frequently classified as relatively efficient.
The pooled regression model has very weak explanatory power (R² = 0.018), and the overall F-statistic is not significant (F = 0.422, p = 0.863). This indicates that, taken together, EV adoption, profitability and firm size do not provide a statistically meaningful linear explanation of the variation in DEA scores. Across the sample, higher EV adoption is not associated with a clear, robust pattern of higher or lower DEA-measured operational efficiency, and the estimated linear relationships between EV adoption and efficiency are weak and unstable across firms and years. Year-by-year and firm-level regressions similarly show heterogeneous and time-varying coefficients rather than one consistent relationship for the full sample.
Taken as a whole, the findings suggest that relative operational efficiency during the EV transition cannot be captured by a simple pooled linear model based only on EV share, profitability and firm size. Efficiency differences appear to be shaped by broader firm-specific and transition-specific conditions, such as restructuring pace, organizational adaptation and production-system complexity. The EV-transition focus is central to this conclusion: firms expand their EV portfolios in different ways and at different speeds, and these differences in EV adoption are not linked to a simple, stable linear pattern in DEA-measured efficiency. The main contribution of the thesis is to show that frontier-based efficiency differences can be identified during industrial transition, but that their pooled linear explainability remains weak and strongly context-dependent.
The empirical analysis uses annual firm-level data for ten large automotive manufacturers. An input-oriented CCR Data Envelopment Analysis (DEA) model is applied to estimate an efficiency score for each firm-year relative to a pooled frontier. These DEA scores are then used as the dependent variable in pooled, year-by-year and firm-level linear regression models, where EV adoption, profitability and firm size serve as explanatory variables. The purpose of the regressions is to test whether variation in DEA-based efficiency can be statistically explained by these three firm-level characteristics and, in particular, whether higher EV adoption is linearly associated with higher or lower operational efficiency.
The DEA results show that relative operational efficiency varies across firms and years. The mean efficiency score is 0.935, with values ranging from 0.653 to 1.000. Across the period, a small subset of firms defines the frontier in most years, while others move between relatively efficient and relatively inefficient positions. For example, Tesla and Toyota are often on or near the frontier, whereas Ford and GM are less frequently classified as relatively efficient.
The pooled regression model has very weak explanatory power (R² = 0.018), and the overall F-statistic is not significant (F = 0.422, p = 0.863). This indicates that, taken together, EV adoption, profitability and firm size do not provide a statistically meaningful linear explanation of the variation in DEA scores. Across the sample, higher EV adoption is not associated with a clear, robust pattern of higher or lower DEA-measured operational efficiency, and the estimated linear relationships between EV adoption and efficiency are weak and unstable across firms and years. Year-by-year and firm-level regressions similarly show heterogeneous and time-varying coefficients rather than one consistent relationship for the full sample.
Taken as a whole, the findings suggest that relative operational efficiency during the EV transition cannot be captured by a simple pooled linear model based only on EV share, profitability and firm size. Efficiency differences appear to be shaped by broader firm-specific and transition-specific conditions, such as restructuring pace, organizational adaptation and production-system complexity. The EV-transition focus is central to this conclusion: firms expand their EV portfolios in different ways and at different speeds, and these differences in EV adoption are not linked to a simple, stable linear pattern in DEA-measured efficiency. The main contribution of the thesis is to show that frontier-based efficiency differences can be identified during industrial transition, but that their pooled linear explainability remains weak and strongly context-dependent.