Identifying the losers in the transport transition: evidence from Germany

The paper addresses how to identify households that will be adversely affected (“losers”) by transport transition policies (e.g., carbon taxes, ETS2, zero-emission vehicle mandates) in the EU, focusing on Germany. Motivated by climate targets (EU net-zero by 2050; Fit for 55) and recent shocks (COVID-19, inflation, energy crisis), the study argues that transport poverty is multidimensional and that regressive policy effects threaten public acceptability. The research questions are: (1) how to unify definitions and dimensions of transport poverty; (2) how to classify households into meaningful transport poverty classes using statistical properties rather than ad hoc thresholds; and (3) how these classes heterogeneously respond to transport cost increases. The study proposes a composite, capability-based approach to define and measure transport poverty and to inform just, targeted policies.

The paper grounds its framework in Sen’s capability approach, arguing poverty is multidimensional and concerns freedoms to achieve functionings rather than income alone. Prior transport/energy poverty metrics (e.g., 10% income share) are criticized for masking heterogeneity and omitting key determinants (e.g., spatial access, time costs, alternatives). Comprehensive frameworks include Lucas et al. (2016), who outline mobility, accessibility, affordability, and exposure to externalities, and Berry et al. (2016), who operationalize mobility practices, conditions of mobility, and financial resources. The review highlights shortcomings of single-dimension indicators (e.g., misclassifying high-income high-spenders as poor; ignoring spatial mismatch and ability constraints) and calls for a multidimensional indicator encompassing travel behavior, spatial matching, self-imposed restrictions, alternatives, and financial resources. A theoretical list of candidate variables (distance, trips, time, extra travel time, car-use restriction, spatial mismatch, no alternative to car, no vehicle, income, and LCOKm) is assembled to be empirically reduced via PCA.

Data: German Mobility Panel (MOP), a nationally representative, rotating three-year panel (1994–2020). This study uses 2004–2019, covering 7,492 households and 14,631 observations post-cleaning. The survey combines fall socio-economic interviews and spring 7-day travel diaries with detailed trip purposes, modes, times, and distances. Due to access restrictions, Germany is studied; similar methods could be applied elsewhere. LCOKm: The study develops the Levelized Cost of driving One Kilometer (LCOKm), analogous to LCOE, aggregating purchase price, fuel/energy, operations, and maintenance over a 10-year lifespan (robustness in appendices with 5/10/20 years). LCOKm represents the full cost of car use; fuel accounts for ~30% of total cost on average, varying by fuel type (diesel 28%, gasoline 32%, hybrid 18%, other 30%). PCA: Following OECD (2008) guidance, PCA with varimax rotation reduces 11 literature-derived variables to principal components (PCs). Sampling adequacy: KMO = 0.72. Retained PCs meet criteria (eigenvalues >1; >10% variance each; cumulative >60% threshold approached). The first three PCs explain 56.52% of variance (32.51%, 12.70%, 11.30%). Loadings define: - PC1: Travel behavior conditional on spatial matching (km traveled, km for personal needs, time for personal needs, poor spatial match, km driven). - PC2: Driving restrictions (difference in travel time public vs car; car-use restriction). - PC3: Resources available (income, LCOKm, and no alternative to car [negatively signed in rotated loadings; grouped conceptually with resources]). LCM/LPA: Latent class analysis (latent profile with continuous indicators) uses PC1–PC3 as indicators to uncover unobserved household classes (conditional independence). Competing models with 2–4 classes are compared using AIC/BIC; a four-class unconstrained model fits best. Class assignment probabilities are estimated via multinomial logit. Class characteristics are profiled using descriptive summaries of transport behavior, costs, and resources. Elasticities: Class-specific short-run cost elasticities of driving are estimated via random-effects panel regressions with robust standard errors clustered at the household level. Dependent variable: ln(km driven); main regressor: ln(LCOKm). Controls include ln(income), fuel type, head of household demographics (gender, age, education, driver’s license), number of children. Estimates are reported separately by class.

- PCA identified three core dimensions of transport poverty: (1) travel behavior conditional on spatial matching; (2) driving restrictions; and (3) resources available (income, cost, alternatives). KMO = 0.72; cumulative variance explained = 56.52%. - LCM revealed four classes (Transport Poverty Scale, TPS) with shares: Independent 47.8%, Sufficient 34.44%, Car-dependent 15.55%, Transport-poor 2.21%. - Independent: optimal spatial matching, lowest travel/costs; 38% carless; flexible multimodality; adequate income. - Sufficient: worst spatial matching (100% poor), highest travel and costs, but highest incomes enable coping. - Car-dependent: poor spatial matching (64%); 98% no alternative to car; 100% own cars; highly reliant on driving. - Transport-poor: low incomes, poor spatial matching (67%), 60% carless; highest time in transit; constrained choices. - Population affected (2019 estimates): 1,838,057 transport-poor people; 12,932,935 car-dependent people (≈14.77 million combined). - LCOKm vs fuel-only costs: Fuel is ~30% of total car cost; focusing only on fuel understates transport’s budget impact. Monthly averages: Fuel €117.43; total LCOKm €325.08. - Class-specific elasticities (ln-km driven w.r.t. ln-LCOKm; all significant at 1%): Independent −0.542 (SE 0.096), Sufficient −0.632 (0.096), Car-dependent −1.299 (0.151), Transport-poor −3.453 (0.805). Authors conclude transport-poor and car-dependent are most adversely affected by cost increases and have inelastic driving demand in practice, facing budget/time constraints that limit feasible adjustments. - Policy targeting: The TPS delineates who bears the burden from carbon taxes and vehicle transition policies; car-dependent and transport-poor are primary losers without compensatory measures.

The study demonstrates that transport poverty is multidimensional and that spatial matching, self-imposed restrictions, and resource constraints jointly shape household vulnerability to transport pricing shocks. By statistically deriving PCs and latent classes, the TPS resolves definitional disputes and avoids arbitrary thresholds. Findings show substantial heterogeneity: independent and sufficient households exhibit relatively elastic responses and can adjust modes or reduce discretionary travel, while car-dependent and transport-poor face binding constraints (limited alternatives, tight budgets, longer travel times). This heterogeneity explains public resistance to uniform pricing policies and underscores the need for targeted redistributive measures to ensure fairness and acceptance. The TPS enables policymakers to identify and prioritize support for the most constrained groups, improving the equity and political feasibility of the transport transition.

The paper contributes: (1) a unified, capability-based definition of transport poverty; (2) an empirically grounded two-step methodology (PCA then LCM) yielding three key dimensions and a four-class Transport Poverty Scale; and (3) evidence of heterogeneous, class-specific responses to transport cost increases using a comprehensive LCOKm metric. Results indicate that transport-poor and car-dependent households are the primary losers from pricing policies and the shift to cleaner vehicles absent compensatory policies. Policy recommendations include redesigning the commuter tax allowance to be more progressive and mode-sensitive, implementing an optimal scrappage scheme favoring low-income and highly reliant drivers, providing relocation subsidies, and improving access and quality of public transport, particularly in poorly matched and rural areas. The TPS can guide allocation of the EU Social Climate Fund. Future research should apply the methodology to other countries’ national travel surveys, analyze class-switching dynamics, and explore behavioral interventions to shift travel choices among the sufficient.

- External validity: Analysis is limited to Germany due to data access constraints; generalization to other countries requires comparable travel surveys. - Measurement constraints: Some constructs (e.g., self-imposed restriction) are difficult to measure unless explicitly surveyed; proxy variables may not capture all nuances. - LCOKm assumptions: Results rely on a 10-year vehicle lifespan; although sensitivity analyses are referenced, cost structures and prices evolve over time and by technology. - Data access and sharing: Underlying microdata (MOP, used car prices) are confidential/licensed; replication requires access permissions. - Temporal/contextual factors: The period covers 2004–2019; post-2019 shocks and policy changes may alter behaviors and costs. - Class shares and elasticities are sample-derived; unobserved heterogeneity beyond included PCs and controls may remain.