The outbreak of COVID-19 in early 2020 led governments around the world to adopt an extraordinary set of Non-Pharmaceutical Interventions (NPIs), including lockdowns, school closures, limits on mobility, and border restrictions. These measures were unprecedented in scale and intrusiveness, and they were implemented despite limited knowledge about their real effects. Existing pandemic preparedness plans offered little guidance: many of the measures later adopted were either absent from those plans—considered too extreme or impractical—or recommended only in very specific circumstances. Moreover, because NPIs were typically introduced as bundles, their individual causal impacts were almost impossible to assess in real time. Governments therefore faced profound uncertainty. Acting too slowly meant risking accusations of negligence, but implementing strict restrictions risked public dissatisfaction and, in democracies, potential electoral backlash.
In such conditions, looking outward—toward the policies adopted by peer countries—became an attractive and politically safe strategy. Media coverage reinforced this dynamic by constantly comparing national responses. Public debates frequently invoked the policies of neighboring countries, sometimes criticizing domestic decisions, other times justifying them. As public opinion became more attentive to what others were doing, governments had strong incentives to imitate peers, both to validate their own choices and to deflect potential blame if a policy later proved ineffective.
To understand these dynamics, together with Piero Stanig and Gianmarco Daniele, we have analyzed daily data on COVID-19 restrictions implemented by all 38 OECD democracies from January 2020 to mid-2022. This long time series, spanning multiple pandemic waves, made it possible to examine how policy decisions diffused across borders. The analysis revealed a clear pattern: the likelihood that a country adopted or intensified an NPI at a given moment strongly depended on what its neighbors had done in the preceding days. This remained true even after controlling for domestic infection and mortality trends. The magnitude of the effect was striking: a one-standard-deviation increase in deaths per million within the country was associated with only a modest rise in the stringency index, whereas an equivalent increase in the average stringency of neighboring countries had roughly double the impact.
Geography emerged as the most important channel of diffusion. Countries tend to share linguistic, cultural, and historical ties with their neighbors, and media outlets more readily cover developments in nearby countries, creating an information loop in which neighboring policies become highly salient. Importantly, the diffusion patterns reflected emulation rather than learning. In principle, governments might be expected to imitate countries that successfully contained the pandemic. In practice, policies did not spread from the best performers but from the geographically closest actors, regardless of any further reasons.
Moreover, and crucially, policy diffusion was asymmetric. Countries were much more likely to imitate decisions to tighten restrictions than decisions to loosen them. Even though, in theory, policy diffusion could work in both directions—spreading relaxation as easily as tightening—three domestic mechanisms made governments more responsive to signals favoring stricter policies.
First, policy stakeholders emerged. Emergency measures created groups with a vested interest in maintaining restrictions. Some countries established special pandemic committees that bypassed traditional health authorities. Of course, the institutional survival of such committees depended on an ongoing state of emergency. Media visibility turned some virologists and public health experts into influential public figures whose role and prominence grew with the persistence of restrictions. Finally, emergency spending generated new economic incentives and opportunities for rent-seeking, further increasing political pressure to maintain strict measures.
Second, governments and the public displayed distorted perceptions of costs and risks. The risks of lifting restrictions were systematically overestimated, while the social and economic costs of keeping them in place were underestimated or ignored. School closures are an emblematic example. Early in the pandemic, remote learning was widely—but mistakenly—assumed to be an adequate substitute for in-person education. At the same time, reopening schools was framed as dangerously risky, despite limited evidence. These misperceptions helped lock in extremely costly policies.
Third, leaders were reluctant to reverse course due to aversion to policy reversals. Governments that had invested political capital in justifying tough restrictions feared appearing inconsistent or incompetent if they later relaxed them. Public beliefs reinforced this constraint: if large segments of society were convinced that, for example, public transport was highly risky, policymakers hesitated to challenge those beliefs even when evidence suggested otherwise. This created a self-reinforcing dynamic in which maintaining restrictions was politically safer than easing them.
These three mechanisms magnified the influence of foreign policy decisions that pointed toward increased stringency, while dampening those that indicated it might be safe to relax restrictions. As a result, international policy diffusion tended to push stringency upward and to sustain it over time, contributing to a general bias in favor of maintaining NPIs. Once restrictions became the status quo, they were removed only reluctantly and without cross-national reinforcement.
A country institutional quality, however, significantly modulated these patterns. High-governance countries—characterized by strong bureaucracies, trustworthy data systems, administrative capacity, good healthcare system—were less dependent on external cues and more responsive to their own epidemiological trends. Low-governance countries, by contrast, were more likely to imitate their neighbors, partly because they lacked confidence in their own ability to manage failure. For example, Switzerland, a high-governance country, reopened its ski resorts during the 2020–21 winter season, accepting the possibility of an infection surge because it trusted its capacity to respond. Italy, with lower institutional capacity, kept its resorts closed under similar conditions, fearing it would not be able to manage the consequences of a reopening gone wrong.
Taken together, the evidence suggests that governments likely erred toward excessive and excessively prolonged stringency. Cross-national emulation sustained the adoption of restrictive policies even when domestic public-health conditions no longer warranted them. This created a risk that socially costly policies—such as prolonged school closures—were maintained long after their benefits diminished. In retrospect, these measures imposed heavy burdens on younger generations, harming learning, mental health, and long-term human capital.
The asymmetry in policy diffusion that we uncovered is troubling for future crises (pandemic or otherwise). If tightening spreads easily while loosening does not, societies may face a persistent risk of government over-intervention. Understanding these dynamics is therefore essential to avoid repeating the same pattern in the next global emergency.
Photo by Kai Pilger
