Global diffusion of three road safety policies, 1964–2015

The paper interrogates the generalizability of the global policy convergence thesis by examining whether three long-standing road safety policies—national road safety agencies (RSAs), child restraint laws (CRLs), and daytime running lights (DRLs)—have converged globally over a comparable historical window. The study situates road safety within a growing global health and development agenda due to rising burdens of road injuries and fatalities, and widespread international promotion of road safety measures (e.g., WHO/World Bank campaigns and UN SDGs). It addresses whether convergence is global, regional, or unintended (driven by domestic conditions), and which mechanisms—imitation/learning, coercion, or competition—facilitate diffusion. By analyzing multiple policies concurrently over ~40–50 years, the authors aim to provide a more rigorous and falsifiable assessment of convergence dynamics than single-case studies, clarifying when and why convergence occurs or fractures across policies and regions.

Prior work documents substantial global diffusion and convergence in areas such as health, economic policy, human rights, democratization, and repression (e.g., Meyer et al., Drezner, Wotipka & Tsutsui, Koo & Ramirez, Frank et al., Shor et al.). Cross-national comparisons of multiple policies highlight heterogeneous convergence patterns across domains. In road safety specifically, previous diffusion research has been narrower and has not systematically linked diffusion to mechanisms of convergence. The authors draw on literature distinguishing convergence outcomes (global, regional, unintended) and mechanisms (imitation/learning via world society and expert/NGO networks; coercion via world-systems dynamics and industry interests; competition via trade-related policy alignment). They also consider regional institutions (e.g., European Transport Safety Council) and campaigns (e.g., WHO/World Bank) as potential accelerants of convergence.

Design and data: Event-history (survival) analysis of policy adoption across 181 countries from 1964 to 2015. Adoption years for RSAs, CRLs, and DRLs were compiled from WHO global road safety reports (2004, 2010, 2013, 2019), national laws, peer-reviewed sources, government and international reports, and correspondence with national road safety officials. Coverage: RSA 97%, CRL 98%, DRL 83%. By 2015, >125 RSAs, 90 CRLs, and 40 DRL laws were recorded.

Outcomes: Time to adopt (i) RSA, (ii) CRL, (iii) DRL. Adoption treated as the event; non-adoption as right-censoring.

Key predictors:

• Global convergence indicators: Weibull shape parameter p ("time") to test whether baseline hazard increases over the period; signatures of the Geneva Convention on Road Traffic (binary, pre/post signature) and AAUTPV (UN 1958 agreement on uniform technical prescriptions) for relevant policies.
• Regional convergence: Separate models for Europe (including the creation of the European Transport Safety Council in 1993 as a binary indicator) and for Latin America and the Caribbean (LAC). Additional regional results (Africa, Asia-Pacific, Western Asia) provided in supplement.
• Unintended (domestic) factors: Logged population (World Bank), political violence (Center for Systemic Peace 0–10 scale of lethal organized violence), and average annual bright sunshine hours (WMO) relevant to DRL.
• Mechanisms of diffusion: • Imitation/learning: WHO’s global road safety campaign indicator (0 pre-2004; 1 post-2004); presence of a national road safety NGO (membership in Global Alliance for NGOs for Road Safety). • Coercion/competition: Country position in vehicle trade (UN Comtrade): importer only; importer-exporter; exporter (used both to test coercion and, inversely, competition expectations).

Controls: Logged GDP per capita (PPP, 2000 US$, World Bank), urbanization (% urban, World Bank), and presence of an RSA when modeling CRL and DRL adoption.

Modeling approach: Parametric survival models using the Weibull hazard function. The Weibull shape parameter p indicates whether the adoption hazard rises (p>1) or declines/slows (p<1) over time, informing convergence. Onset of risk defined by national independence year; if independence preceded the first global enactment, risk began at the first enactment year of the policy globally. Pioneer countries for each policy (Norway for RSA, Belgium/Denmark for CRL, Finland for DRL) were excluded from their respective diffusion baselines. To account for spatially correlated unobservables, models adjust for clustering by 22 UN geoscheme regional clusters via random effects and robust standard errors.

• Global convergence patterns: • CRL globally converged: Significant increase in adoption over time (Weibull p≈1.38), with strong associations to global instruments (AAUTPV significant; Geneva Convention supportive in some specifications) and to imitation/learning (post-2004 WHO campaign). GDP per capita also positively associated. Presence of an RSA further boosted CRL adoption in extended models. • RSA exhibited fractured (non-purely global) convergence: Adoption rose over time (p≈1.35–1.44) but was not solely driven by global forces or treaties. Domestic factors mattered: larger populations were more likely to adopt; higher political violence slowed adoption. Mechanisms: WHO’s campaign and presence of road safety NGOs increased adoption; evidence favored competition (importer–exporter countries more likely to adopt than pure importers) over coercion. • DRL did not globally converge: The time parameter was <1 (p≈0.75–0.78), indicating weak global trend. AAUTPV was positively associated, and RSA presence increased DRL adoption. Mechanisms: imitation/learning and competition (exporter and importer–exporter vehicle countries more likely to adopt than importers) were supported; political violence reduced adoption.

• Regional convergence: • Europe (and North America in the model label): Regional convergence for RSA and CRL; not for DRL. Mechanisms varied: Geneva Convention participation and WHO campaign accelerated RSA adoption; creation of the European Transport Safety Council (post-1993) had a large positive effect on RSA. For European CRL, road safety NGOs were associated with faster adoption; for DRL, only the WHO campaign showed a learning/imitation effect. RSA presence increased subsequent CRL and DRL adoption. • LAC: Regional convergence for RSA and CRL. WHO’s campaign was associated with RSA adoption; urbanization was associated with CRL adoption.

• Descriptive diffusion counts by 2015: >125 RSAs, 90 CRLs, and 40 DRL mandates worldwide.

• Mechanisms summary (per Table 3 narrative): Imitation/learning recurs across policies and regions; competition appears for RSA globally and DRL in Europe; evidence for coercion via automobile industry interests is inconclusive.

• Unintended/domestic influences: Population size often increased adoption likelihood; political violence tended to delay adoption (notably for RSA and DRL).

Findings qualify the global convergence thesis: despite similar origins and international promotion, the three road safety policies followed distinct convergence trajectories. CRL achieved global convergence aligned with strong international legitimation, treaty frameworks, and WHO advocacy. In contrast, RSA and DRL displayed fractured or regionalized convergence shaped by regional institutions (e.g., the European Transport Safety Council), national capacities and priorities (population pressure, political violence), and sectoral dynamics.

Mechanisms mattered heterogeneously: imitation/learning via WHO campaigns and NGO networks consistently supported diffusion; competition through vehicle trade alignment had detectable effects (especially for DRL and some RSA models), while coercion (as industry-driven resistance) lacked consistent cross-national support. The DRL case illustrates how contested evidence, variable championing, the influence of a prominent non-adopting country (USA/NHTSA’s stance), and the absence of regional epistemic infrastructures outside Europe can impede global convergence even when some scientific support exists. Overall, convergence appears contingent on the interplay of global scripts, regional institutionalization, and domestic context, rather than uniform global diffusion.

The study demonstrates that policy convergence is multi-shaped—global (CRL), regional (RSA, CRL), and unintended/fractured (RSA, DRL)—even within a single policy sector. It advances the convergence literature by jointly analyzing multiple policies over comparable time horizons, explicitly testing mechanisms, and integrating regional dynamics. Results underscore the centrality of imitation/learning mechanisms and, to a lesser extent, competition, while offering limited support for coercion narratives.

Implications include the need to incorporate regional institutional architectures and national circumstances into convergence theories and to conduct systematic falsification by comparing multiple policies. Future research should examine subsequent policy reforms (not only first adoptions), improve measurement harmonization of policy definitions, integrate direct safety outcome metrics (e.g., fatality rates), and assess how regional networks and leading countries shape diffusion trajectories across other road safety and public health policies.

• Operationalization variability: Broad, inclusive definitions of RSA, CRL, and DRL may conflate stringent with flexible forms, warranting cautious interpretation.
• Data quality and timing: Cross-country information quality varies; adoption year may differ from effective implementation date, potentially biasing estimates.
• First-adoption focus: Analyses capture only initial enactments; subsequent reforms may follow different diffusion dynamics.
• Omitted direct outcome control: High traffic fatality rates could influence adoption; while population, GDP per capita, and urbanization were included as proxies, explicit fatality rates were not modeled.