Correlation Isn’t Causation: How Vaping Is Blamed for Trends It Didn’t Create

In public health debates, few phrases are repeated as ritualistically and misunderstood as thoroughly as “the data show an association.” When it comes to vaping, that phrase has increasingly become a shortcut to a much stronger claim: that vaping has caused increases in smoking, particularly among young people. The leap from association to causation is rarely acknowledged, let alone defended. Instead, it is smuggled into headlines, policy briefs, and press releases, wrapped in the authority of academic journals and time-series charts.

Time-series studies, including several recently published in outlets such as the Medical Journal of Australia, have become central to this narrative. These studies typically examine smoking prevalence over time, note the emergence or growth of vaping, and observe that smoking declines appear to have slowed or flattened in the same period. From this temporal overlap, an implicit conclusion is drawn: vaping must be responsible.

The problem is not the method itself. Time-series analyses are a legitimate and widely used tool in population health, and they have also been used in pro-THR research to examine whether smoking declined more rapidly following the introduction of e-cigarettes. The issue is not that time-series methods exist, but how confidently and selectively their results are interpreted.

Time-series studies can describe population-level patterns. What they cannot do, on their own, is establish causal mechanisms. They do not measure individual trajectories. They do not determine whether the same people who vape would otherwise have smoked. And they cannot tell us whether vaping substituted for smoking, delayed initiation, altered quitting behaviour, or had no causal role at all. Those questions require convergence across multiple lines of evidence.

Yet when vaping is involved, these descriptive patterns are frequently treated as dispositive proof of harm.

A recurring assumption is that the growth of vaping represents a singular “intervention” that disrupted prior progress. Smoking rates were falling quickly before vaping, the argument goes; they fell more slowly afterwards; therefore, vaping must have impeded tobacco control. But this framing ignores a well-established feature of long-term prevalence trends: declines almost always decelerate as smoking becomes less common. Early reductions are driven by social denormalisation, policy shocks, and quitting among lighter or more motivated smokers. Later declines are slower and more uneven because the remaining population tends to be more dependent, more disadvantaged, and harder to reach with traditional interventions.

This “floor effect” has been observed repeatedly across countries and historical periods, including eras long before vaping existed. To attribute any post-2010 deceleration to vaping without accounting for this structural reality is not causal inference; it reflects a conclusion drawn from timing alone, not evidence.

Confounding further complicates matters. Over the same period that vaping emerged, tobacco control environments have been shaped by rising excise taxes, changes in cessation support, shifts in enforcement, growth of illicit markets, pandemic disruptions, worsening youth mental health, and generational changes in risk behaviour. These forces operate simultaneously and unevenly across subgroups. When smoking trends change, there are many plausible explanations. Vaping is often singled out not because it is uniquely explanatory, but because it is politically convenient.

This selectivity becomes especially apparent when subgroup data are examined. In some datasets, smoking continues to decline among young women while flattening among young men; in others, declines persist in higher-income groups but stall among disadvantaged populations. These divergent patterns are difficult to reconcile with a single causal driver such as vaping. Instead of prompting caution, however, they are often smoothed over in aggregated conclusions that point in only one direction.

Much of this debate ultimately hinges on how evidence is framed. Claims about a “gateway” from vaping to smoking are typically rooted in individual-level survey data showing that adolescents who vape are more likely to later report smoking. But the counter-evidence to the gateway does not rely on any single method. It comes from multiple sources: adjustment for common liability factors, genetic studies showing shared risk propensities, economic analyses demonstrating product substitution, and population-level trends showing that youth smoking has continued to fall even as vaping increased.

Importantly, much of the pro-THR literature does not claim to prove diversion in a strict causal sense. Instead, it asks a more modest and testable question: do the predictions of gateway hold up in the real world? If vaping were driving youth smoking increases, we would expect to see smoking rise, not continue its long-term decline. The fact that these predictions repeatedly fail is itself informative.

By contrast, many time-series critiques of vaping move immediately from correlation to gateway as an assumed explanation, rather than treating gateway as a hypothesis to be tested and potentially falsified.

Similar issues arise with claims of “renormalisation.” This concept is often invoked as a theoretical concern that vaping might make smoking socially acceptable again, but it has remarkably little empirical support. The few studies that have actually tested renormalisation directly have generally found no evidence for it. Yet the claim persists, largely by assumption, and is recycled whenever population trends fail to behave as expected.

None of this is to argue that vaping is harmless, or that population-level data should be dismissed. It is an argument for proportionality and intellectual restraint. Strong causal claims require converging evidence across methods, not the elevation of a single descriptive pattern into a definitive explanation, especially when that explanation conveniently absolves existing policy settings of responsibility.

There is also an asymmetry worth noting. When smoking rates fall, tobacco control is credited. When progress stalls or reverses, responsibility is externalised, assigned to industry, new products, or cultural forces. Rarely is the possibility entertained that restrictive or misaligned policies might themselves be contributing to worse outcomes.

When correlation is mistaken for causation, the result is a self-reinforcing narrative loop. Studies suggest vaping is a problem; policies restrict access to lower-risk alternatives; smoking outcomes deteriorate; and those deteriorations are then cited as further evidence of vaping’s harms rather than of its absence.

Public health rightly values evidence-based policy. But evidence is more than data points and regression lines. It is interpretation, context, and humility about what any single method can tell us. When time-series analyses are treated as proof rather than prompts for further inquiry, they stop informing policy and start defending it.

If vaping is to be criticised, it should be based on clearly articulated mechanisms, individual-level evidence, and consistency across international experience—not on the convenient misreading of correlations as causes. Anything less is not science. It is storytelling with charts.

And public policy deserves better than that.