Stringent sustainability regulations for global supply chains are supported across middle-income democracies

Globalised production has spurred development but also offshored environmental (deforestation, biodiversity loss) and social (inequality, unsafe work) harms from high-income OECD countries to middle-income, non-OECD producers via trade liberalisation and global supply chains. High-income governments (e.g., EU, France, Germany) have begun extraterritorial due-diligence regulations mandating supply-chain sustainability disclosure, yet enforcement capacity is limited abroad and effectiveness likely hinges on diffusion and harmonisation of such frameworks in producer countries. Existing diffusion often arises from firm-led 'trading-up' of standards, but cross-sectoral harmonisation likely needs coordinated public policy. Public opinion shapes policy in democracies; while high support exists for stringent supply-chain policies in OECD settings, little is known about preferences in middle-income democracies. Key questions: (i) Do citizens in middle-income countries want domestic regulations aligned with OECD measures? (ii) What stringency levels do they prefer? (iii) How do they perceive benefits and costs? (iv) How do preferences vary by demographics and attitudes? Theoretical expectations: support may be lower in middle- than high-income countries; higher stringency may increase both perceived benefits and costs; pro-environmental concern and perceived environmental impacts should raise support; and preferences may be robust to informational cues about costs/benefits. The study tests these expectations with survey-embedded experiments in Brazil, India, and Indonesia, comparing to 12 OECD democracies.

The paper situates its contribution in research on globalisation’s environmental and social externalities, extraterritorial due-diligence legislation, and policy diffusion. It references work on 'externality offshoring' via trade and global supply chains and the rise of due-diligence and disclosure-based private standards (e.g., GRI, ISO). It contrasts private 'trading-up' with the need for public policy harmonisation and draws on literatures showing public opinion’s influence on policy in democracies and high public support for stringent supply-chain regulation in OECD countries. Competing arguments are reviewed: potential resistance in middle-income countries due to competitiveness concerns and perceived imperialism versus support driven by market access and domestic environmental/labour aspirations. The study also connects to theories linking environmental attitudes to policy support and examines robustness of preferences to information treatments.

Design: Three original survey-embedded experiments in Brazil, India, and Indonesia (largest democratic non-OECD economies). Fieldwork dates: April 27–May 23, 2022. Sample: N=2,000 per country (Brazil, Indonesia) and N=2,001 (India) recruited from Dynata online panels with quotas on age, gender (interlocked with age), and education. Voting-age citizens (18+). Languages: English, Hindi, Indonesian, Portuguese, with professional translation and review by native-speaking social scientists. Ethics: Approved by ETH-Zurich Ethics Commission (EK-2021-N67). Pre-registered on OSF (https://doi.org/10.17605/OSF.IO/7ATUP). Attention checks: (i) response duration <45% of median (7m 3s), (ii) bicycle wheels item, (iii) age vs birth year consistency; participants failing ≥2 excluded (~3%). Comparative OECD sample: Prior identical instrument fielded Sept 22–Nov 3, 2021 in 12 large OECD importing democracies (BE, CA, CH, DE, ES, FR, IT, JA, KO, NL, UK, US), N≈2,000 each; total N=24,003. Same quotas; multiple languages; attention checks; approved by ETH-Zurich (EK-2021-N67); pre-registered at Harvard Dataverse (https://doi.org/10.7910/DVN/KLOTB6). Experiments: 1) Conjoint choice experiment (primary): Each respondent evaluated five rounds, each with two policy proposals (A/B), constructed from three attributes—Scope (employee threshold), Transparency (confidential vs public reporting; government-mandated content), Enforcement (none; naming and shaming; fines; legal action). Full factorial: 36 designs. Three stylised packages analysed: Low, Medium, High stringency (Table 2): - Low: applies to very large firms (≥25,000 employees); confidential report with no required content; no enforcement. - Medium: applies to ≥250 employees; confidential report with some required content; public list and moderate fine for non-compliance. - High: applies to ≥25 employees; annual public report with detailed required content; public list, severe fines, government procurement exclusion, legal charges. Outcome measures: (a) Proposal support (yes/no) for each proposal, enabling absolute support levels; (b) forced choice between A vs B; (c) for dual opposition, follow-up on reasons (too strict/not strict enough/other). 2) Vignette experiment on perceived benefits and costs: After the conjoint, respondents were randomly assigned one of the three stringency packages and rated perceived benefits (better consumer information; improved local production conditions—working or environmental, randomized; domestic job creation) and costs (higher firm costs; higher consumer costs; reduced national sovereignty) on 1–7 Likert scales. 3) Subgroup analyses: Conjoint support stratified by environmental attitudes ("I do what is right for the environment..." 1–7), perceived environmental impact ("Environmental problems have a direct effect..." 1–7), education (years 0–20), and household income (country-specific deciles). 4) Informational vignette treatments (pre-conjoint): Random assignment to one of five conditions—two benefit frames (improved conditions; increased trade and jobs), two cost frames (consumer costs and job loss; threats to sovereignty), or control (no information). Analytical strategy: For conjoint, proposal support regressed on dummy indicators for attribute levels with OLS, clustered SEs at respondent level, including interactions to estimate a full-factorial model and controls for information treatment. Predicted probabilities (marginal means) computed for low/medium/high stringency packages per country. Approximate analytical samples per country for conjoint full-factorial estimation: ~1,600–1,760 respondents due to subset focus (Brazil 1,761; India 1,723; Indonesia 1,615). Subgroup conditioning via interactions between attribute levels and subgroup variables; predicted probabilities calculated at substantive levels. Informational treatment effects via interactions with attribute levels; predicted probabilities by stringency and treatment. Benefit/cost perceptions analysed by OLS of each item on stringency assignment per country; marginal means reported. All analyses in Stata SE 16.1. Data and code available at OSF (data: https://doi.org/10.17605/OSF.IO/YXNW8; code: https://doi.org/10.17605/OSF.IO/YXNW8).

- Overall support: From the conjoint, average support across all policy conditions is high in Brazil (64.5%) and India (64.1%), similar to the 12 OECD countries, and a narrow majority in Indonesia (50.9%). - By stringency: Medium and high stringency packages garner majority support. Approximately 65–70% in Brazil, India, and OECD support the high-stringency package (scope ≥25 employees; public, detailed reporting; strong enforcement). In Indonesia, medium/high stringency receive majority but lower support (about 50–55%). Low stringency is least acceptable: 39% support in Indonesia, 46% in OECD, 49% in Brazil. - Perceived benefits vs costs: Across all three countries, expected benefits exceed expected costs at all stringency levels. Benefit marginal means are high (≈5–6 on 1–7 scale). In Brazil, benefits are slightly higher for high vs low stringency (≈5.7–5.9 vs 5.0–5.3). Cost expectations vary by country: India shows higher cost expectations (≈5.0–5.5), while Brazil shows lower firm/consumer cost expectations (≈3.6–3.8). Regulatory stringency has little effect on cost perceptions (except sovereignty in Brazil); country context matters more. - Subgroup patterns: In Brazil, subgroup conditioning is modest; small effects of environmental attitudes/impact (slightly higher support) and slight decreases with higher education/income (effects ≈ ±5 pp). In Indonesia and India, higher education and stronger environmental attitudes/impact perceptions substantially increase support for medium/high stringency (in Indonesia, +20–25 pp for high stringency among high environmental concern; highly educated Indonesians show 60–70% support). Household income shows weaker conditioning across countries. - Robustness to information: Exposure to benefit or cost informational treatments (improved conditions; increased trade/jobs; consumer costs/job loss; sovereignty threats) did not produce statistically significant changes in support relative to control across stringency levels, indicating preference robustness. - Additional heterogeneity: In Indonesia, support for high-stringency designs is similar to Brazil/India when regulations apply only to very large firms (≥25,000 employees), but support drops by 10–15 percentage points when applied to smaller firms (≥25 employees), suggesting concerns about impacts on small producers.

The study addresses whether citizens in middle-income democracies support aligning domestic regulations with stringent OECD supply-chain measures, preferred stringency levels, perceived benefits/costs, and subgroup variation. Findings show broad acceptability of alignment and high/medium stringency in Brazil and India, with somewhat lower but still majority support in Indonesia. This challenges expectations that middle-income publics oppose such measures due to competitiveness or sovereignty concerns and supports arguments that citizens value environmental and labor protections even amidst development. Perceived benefits (information, improved production practices, job creation) outweigh perceived costs (to firms, consumers, sovereignty), helping explain high support. Education and pro-environmental dispositions drive stronger support in India and Indonesia, indicating that attitudinal and skill-related factors shape preference formation. Cross-national differences (e.g., lower overall support in Indonesia) may reflect macro-structural conditions such as export-market alignment and FDI restrictiveness, and concern for small producers when regulations extend to smaller firms. Collectively, alignment of public preferences across 15 democracies (~35% of population, ~55% of GDP) suggests enabling conditions for policy diffusion and harmonisation of supply-chain governance from OECD to middle-income countries.

Publics in three large middle-income democracies broadly support aligning domestic regulations with stringent OECD-style supply-chain sustainability measures, particularly at medium and high stringency. Expected benefits outweigh perceived costs, and preferences are robust to informational cues. Heterogeneity across and within countries indicates higher support among more educated and environmentally concerned respondents in India and Indonesia, with limited subgroup conditioning in Brazil. Policy implications: (1) International actors should encourage and pressure governments in producer countries to implement disclosure-based regulations for extraction and production feeding global supply chains. (2) Beyond increasing transparency stringency, policymakers should maximise the value and comparability of disclosed information (e.g., standardised reporting criteria, broad geographic coverage) to facilitate low-cost compliance and stakeholder decision-making. (3) Complementary measures are needed to support vulnerable stakeholders (e.g., small-scale farmers/miners) to avoid non-compliance due to capability deficits and unintended exclusion from markets. Future research should probe distributional impacts, drivers of individual preferences (e.g., elite cues, economic gains), potential public–elite disconnects, and extend analysis to autocratic contexts where responsiveness may differ.

- External validity and scope: The study covers three middle-income democracies; findings may not generalise to other middle-income or autocratic contexts. - Sampling: Online panel with quota sampling; difficulties recruiting lower-education respondents in Brazil; probability-based sampling could improve generalisability. - Cross-national design constraints: Unable to identify perceptions and impacts among subgroups most directly affected by policies within each country; distributional effects not fully captured. - Unexplained heterogeneity: Differences between India and Indonesia not fully accounted for; structural factors (export destinations, FDI environments) may play roles. - Information treatments: While preferences appear robust, brief vignettes may not capture real-world information environments or sustained elite messaging. - Measurement/implementation framing: Slight differences in survey framing between producer (non-OECD) and consumer (OECD) perspectives; although instruments were harmonised, contextual framing may influence responses.