The Clock of Reactance: Before and After a System-Level Policy
Here the subject is public reaction to restrictive system-level policies as those policies move from announcement to implementation. The specific question is whether timing itself changes psychological reactance, rather than merely revealing whether people like or dislike a rule. That distinction matters for interpreting protests, survey results, and policy communication: a strong reaction before a rule takes effect may be partly temporary, but a weaker reaction later is not automatic proof of approval. The evidence moves from real smoking-ban data to randomized comparisons across taxes, vaccination, and speed limits, allowing us to see both what the timing claim supports and where it stops.
What the smoking-ban comparison can—and cannot—show
The first test uses a setting in which policy change happened outside the researchers’ control. The study drew on two representative European Union surveys, one conducted in September and October 2005 and the other in October and November 2006. Across the two waves, 49,674 people in 26 countries answered a question about their opposition to smoking bans in indoor workplaces. The response scale ran from “totally in favor” to “totally opposed.”
Between those surveys, Belgium, Scotland, and Spain introduced smoking restrictions. Belgium banned smoking in all workplaces in January 2006. Scotland followed in March with a ban covering workplaces and public places such as hospitality venues. Spain introduced a ban in January, but allowed important exceptions: some hospitality venues could create smoking areas, and smaller venues could choose whether to be smoke-free. The other 23 countries experienced no comparable change in smoking policy during the relevant period and served as a control group.
A simple before-and-after comparison would ask whether opposition was lower in 2006 than in 2005 in the countries that changed their rules. That would be informative, but it would also be vulnerable to any other event that changed opinion during the year. The researchers instead used a difference-in-differences approach. In plain language, they compared the change in the countries introducing a ban with the change over the same period in countries that did not. The control group supplies an estimate of the background movement that might have happened even without implementation.
The result fit the timing hypothesis in Belgium and Scotland. People were more opposed where a ban was planned for the coming year, and opposition fell after the ban took effect. In the regression model, the country-specific interaction for Belgium was −0.102, with a standard error of 0.02 and P < 0.001. The corresponding interaction for Scotland was −0.368, with a standard error of 0.02 and P < 0.001. Because the outcome was measured on a four-point scale, these are changes in the model’s opposition measure, not universal percentages of the population. The article’s discussion translates the two effects into approximate decreases of 6% and 17% in the relevant comparisons.
Spain is important precisely because it does not fit the clean version of the story. The interaction was 0.044, with P = 0.078, so the study did not find a statistically significant decrease after implementation. The authors suggest that the partial ban and its exceptions may have helped keep the issue contested. A rule with loopholes can produce a different social experience from a comprehensive rule: people may continue to debate who is covered, whether enforcement is fair, and whether the intervention has really settled the question. That explanation is plausible in the context of the article, but it is explicitly speculative. The analysis did not directly test “partialness” as the mediator of Spain’s result.
The researchers checked the pattern in several ways. They estimated separate models for the intervention countries and added controls related to smoking behavior, demographic characteristics, political orientation, country, and occupation. They also used doubly robust difference-in-differences estimators designed to be more resilient to violations of the usual parallel-trends assumption when the observed control variables are taken into account. Those checks again found significant declines for Belgium and Scotland, but not for Spain. The consistency makes the result harder to dismiss as a single modeling choice.
The data still cannot answer every causal question. There were only two survey waves, so the researchers could not rigorously inspect whether the intervention and control countries were already moving along parallel opinion trends before the bans. Unobserved events—such as satisfaction with a national government or the intensity of a country’s policy debate—could also have affected the estimates. The analysis compares populations at two points rather than tracking the same individuals through the transition. It is therefore best read as strong real-world evidence of a pre/post pattern in these settings, not as proof that implementation always reduces opposition.
The article also discusses secondary evidence about mandatory seatbelt legislation in New York and a 100 km/h daytime motorway speed limit in the Netherlands. Opposition to those policies appeared to decrease after implementation as well. These cases broaden the context, but the authors describe them as further suggestive evidence, not as equivalent to a fully identified causal test. They help establish that smoking bans are not the only place where a planned rule can attract more resistance than an established one, while the design limitations remain.
The state transition can be represented without treating it as a promise of eventual approval:
The diagram separates three ideas that are easy to blend together. Planned and implemented are different states of the policy’s relationship to the public. The higher-versus-lower pattern is an empirical result across the studied comparisons, not a law built into the state names. Spain is a reminder that the transition may not feel complete when a policy is partial or full of exceptions. Finally, the boundary state is not another stage of public acceptance; it marks the uncertainty that remains after the data have been analyzed.
Randomized comparisons make the timing effect harder to dismiss
Natural policy data have the advantage of realism but cannot control everything that happens around a policy. The next studies changed the presentation of the policy while holding the rest of the scenario more constant. Participants were randomly assigned to imagine a policy as planned in the future or as already implemented in the past. Random assignment does not make the scenarios realistic in every detail, but it makes a timing explanation more credible: on average, the two groups should differ in the assigned stage rather than in a preexisting preference.
Study 2 involved 719 UK participants in a two-by-three design. One factor was timing: the government would increase a tax in one year, or had increased it one year earlier. The other factor was policy context: the tax applied to cars, alcohol, or meat. Each increase was described as 30%, with a stated purpose such as reducing carbon emissions, encouraging healthier lives, or encouraging more vegan consumption. Participants then rated how angry they would feel toward the government on a six-point scale.
The average anger-based reactance was 4.20 when the tax was presented as future and 3.46 when it was presented as already implemented. The main timing effect was statistically significant, F(1, 713) = 40.60, P < 0.001, with η² = 0.05. The researchers found no significant interaction between timing and policy context. That does not mean the three policies produced identical reactions. The reported standardized differences were 0.58 for cars, 0.50 for alcohol, and 0.35 for meat. It means the study did not find evidence that the direction of the timing effect depended substantially on which of those three contexts participants read about.
This design adds two kinds of support. First, the policy does not have to be a smoking ban for the contrast to appear. Taxes connected to transportation, health, and diet all showed higher anger when the change was still ahead. Second, the study addresses a problem in the real-world comparison: countries that adopt different policies may differ in many ways before the policy is announced. Random assignment makes the future-versus-past presentation itself the most direct difference between groups.
The experiment also narrows the claim. It measured a response to a written scenario, not a year of living under a tax. The future condition asks participants to imagine an upcoming restriction, and the past condition asks them to imagine that the restriction has already happened. The resulting contrast is causal evidence about how temporal presentation affects the reported response. It is not a direct estimate of how the same person’s feelings will evolve after a real government implements a tax, nor does it measure actual payment, compliance, or long-term support.
Studies 3A and 3B tested whether the pattern would hold in policy areas that carry different meanings. In Study 3A, 244 UK participants read about a fictitious disease and a vaccine described as effective, with rare side effects. They were randomly assigned to read that the government planned to mandate vaccination in the future or had mandated it one year earlier. Reactance was again measured through anger toward the government. The ex ante condition produced higher reactance than the ex post condition, with a standardized difference of 0.34 and a 95% confidence interval from 0.09 to 0.60.
Study 3B used 809 participants from Germany and a proposed 120 km/h speed limit on German motorways, described as a measure to reduce carbon emissions. The same future-versus-past comparison produced a standardized difference of 0.23, with a 95% confidence interval from 0.09 to 0.37. The effect was smaller than in the vaccination scenario, but the interval still supported a difference in the measured direction. A speed limit and a vaccination mandate do not impose the same kind of cost, and the study does not claim that they do. Their value here is that the timing comparison survives a change in policy domain.
Both studies also measured participants’ general attitudes toward the policy. For vaccination, participants reported how positive they were toward vaccinations. For the motorway study, they reported whether they favored a speed limit. Neither general-attitude measure significantly moderated the decline in reactance from the planned to the implemented condition. The authors interpret that result as evidence that the timing effect includes a transient psychological response rather than simply reflecting the stable preferences of people who oppose the policy.
That interpretation should be stated carefully. A nonsignificant moderation result is not proof that prior attitudes never matter. It means these studies did not detect a meaningful difference in the timing pattern across the measured attitude levels. Someone can favor a speed limit and still feel irritated when it is proposed; someone can oppose vaccination mandates and still show a different level of anger when the mandate is described as already in force. The short-term response and the underlying judgment can coexist.
Across the randomized studies, the pattern is therefore more precise than “people change their minds once a rule exists.” Participants reported more reactance when a restrictive policy was framed as an approaching transition than when it was framed as a settled fact. The result appeared across taxes on cars, alcohol, and meat, as well as vaccination and motorway speed limits. It was not eliminated by switching policy context, and the measured general attitude did not explain away the stage contrast.
At the same time, the randomized tests do not settle why timing matters. A planned policy may make the personal loss of a familiar option more vivid. People may adapt emotionally once the rule is in place. They may regard an implemented policy as definite and rationalize it. Or the wording used to describe the future and past conditions may carry subtle differences beyond time. The authors designed later studies to examine their proposed psychological process and to account for some alternatives, but even those tests have boundaries.
The most defensible conclusion at this point is a two-part one. The evidence supports a short-horizon timing effect: in the real-world smoking-ban comparisons and in randomized hypothetical scenarios, reactance or opposition was generally higher before implementation than after it. The evidence does not support treating the effect as universal, permanent, or equivalent to genuine policy acceptance. To understand what changes between the two states, we have to look at what people believe they are losing in the transition and what they can see themselves gaining once the new arrangement is described as real.