New Zealand's Accelerated Decline in Smoking after Embracing E-Cigarettes

My take on an exchange about Beaglehole et al.'s article showing a sudden steeper decline in New Zealand's smoking rate. Below, I expand my PubPeer comment on a published criticism to this article.

Artwork by Pawel Bugucki

Beaglehole et al.’s Article

Beaglehole et al.’s article presents a straightforward analysis of the adult smoking rate in New Zealand. Critical context is that in 2019, New Zealand’s Ministry of Health officially endorsed e-cigarettes as a tool for smoking cessation and harm reduction. Beaglehole et al. performed a joinpoint analysis, which (for the non-statisticians out there) takes a trend over time and does some math to fit the curve to a series of straight lines, and tells you where the slope of the lines changed.

Lo and behold, the decline became steeper around 2019:

It then flatted out a few years later, which (as I’ll explain below) is a completely natural behavior for smoking prevalence as it approaches low numbers.

Beaglehole et al. duly note that this alone cannot prove causation: time trends can be correlated for all sorts of reasons and they are often not causal (see Tyler Vigen’s Spurious Correlations for silly examples of why you can’t conclude causality from this kind of data). But in their discussion, Beaglehole rule out other obvious explanations such as new policies (there were no big changes in tobacco control policies around this time) and make the case that it’s at least plausible.

And, frankly, it’s not as if they pulled this explanation out of thin air. There’s robust evidence that e-cigarettes are effective for smoking cessation (Cochrane review, Lindson et al., 2025), for unplanned or accidental switching (Kim, 2024), and are economic substitutes for cigarettes, meaning that consumers use one product in place of the other (Davis et al., 2025; Friedman et al., 2024; Selya et al., 2023). So e-cigarettes are a completely plausible explanation for the decline. And while this analysis alone can’t prove causality, it’s consistent with a large body of other research and also provides converging evidence that the effects are detectable at the population level.

Beaglehole et al.’s article received an immediate formal criticism by Blum et al. which tries to dismiss e-cigarettes as an explanation, instead claim credit for the smoking decline for tobacco control, and downplay the significant decline.

The following is my PubPeer comment on the Blum criticism. As of writing this, my comment is still under moderation after several days, but if it gets approved it should appear here. I’m replicating my post here on Substack because 1) PubPeer has been buggy lately with my previous submission showing only a blank page and no response from support and 2) for wider visibility. Since this is in “academic speak” I also added a plain language summary below.

My PubPeer Comment on Blum et al.’s Criticism

Blum et al.’s commentary is a criticism of Beaglehole et al.’s article, which reported that New Zealand adults showed a large and statistically significantly steeper decline in smoking prevalence (annual percent change from 3.5% to -17.9%) coinciding with the widespread uptake of e-cigarettes (~2018/19). Beaglehole et al. duly acknowledge that this result alone does not establish e-cigarette uptake as a causal factor in the decline of smoking, but do discuss this as a key explanation, considering New Zealand’s embracing of e-cigarettes for harm reduction. Here’s the key figure:

Significance of the Sudden Downward Acceleration in Smoking

Blum et al. dismiss the sudden steeper decline after ~2018/19 as “a brief 3-year decline.” A key conceptual concept is absent from this discussion: smoking prevalence, like other real-world complex systems, changes exponentially (Sterman 2000; Selya et al. 2021; Selya & Foxon 2021). Specifically, as smoking prevalence declines, the curve naturally flattens out as it approaches its minimum value (rather than abruptly reaching zero and having a “sharp” discontinuity in the slope).

Beaglehole et al.’s key result is a striking deviation from the gradual flattening out that is expected to occur at low smoking prevalence: i.e., a significant downward acceleration starting in ~2018/19. In other words, the joinpoint analysis identified a change that opposes the natural shape of the curve.

The criticism of being only 3 years of accelerated decline is also lacking in conceptual understanding, as smoking prevalence was quite low at the end of this accelerated decline (in ~2022/23) and there is not sufficient “room” for it to keep accelerating downwards before hitting zero – and hitting zero abruptly, which again would be exceedingly rare in a real-world complex system.

Plausibility of Competing Explanations for the Decline

Blum et al. instead attribute the entirety of the decline in smoking to conventional tobacco control measures. While some of the decline can be attributed to these policies, Beaglehole et al. correctly point out that the decline in smoking prevalence was modest and would not have been able to achieve New Zealand’s Smokefree goal for several more decades. Blum et al. offer no explanation for why there would be a delayed effect years after policy implementation (i.e., to explain the larger decline in ~2018/19), which is in stark contrast to the diminishing returns of such policies over time noted by others (Fagerstrom et al., 2025).

Blum et al. reject attributing the decline in smoking to e-cigarettes. However, e-cigarettes are a plausible explanation, perhaps the key explanation. There is no longer reasonable doubt that e-cigarettes are effective for smoking cessation (Lindson et al., 2025), and even when not used intentionally in a premeditated quit attempt, they still help adults who smoke switch away from smoking in large numbers (Carpenter et al., 2023; Shiffman et al., 2025). E-cigarettes were the most common method reported by both US (Foxon & Niaura 2025) and UK (Jackson et al., 2025) adults who stopped smoking in recent years. Finally, e-cigarettes serve as economic substitutes for cigarettes according to robust quasi-experimental evidence (Davis et al., 2025; Friedman et al., 2024; Selya et al., 2023). Though Beaglehole et al.’s analysis cannot establish causality (as they acknowledge), neither can e-cigarettes be dismissed as a highly plausible explanation for the decline in smoking prevalence, especially without offering a credible alternative explanation for a striking phenomenon.

Conclusion / Plain Language Summary

As smoking prevalence declines to zero or whatever its “floor” is, the natural shape of the curve flattens out to gradually approach that floor. As opposed to a straight-line decline that abruptly hits zero (which rarely occurs in natural systems).

Beaglehole et al. analyze New Zealand’s smoking trends as series of straight lines, which I usually consider a flaw, but in this case it makes their point even more strongly. Why? Because they detect a sharper decline, when the natural shape of the curve is to flatten out. So if anything, their results are conservative and underestimate the size of the change.

Blum et al., don’t seem to understand this point. Instead, they dismiss the steeper decline that Beaglehole et al. by saying “it was only 3 years, and it flattened out afterwards.” But with smoking prevalence, it has to flatten out at some point. And as for the “only 3 years,” well, smoking was already so low after 3 years of steeper declines, that it physically can’t continue at that rate for much longer!

Edit: This is an even better plain-language summary criticizing what Blum et al. are saying, sent to me by Geoff Vann on Twitter/X:

The ‘pulse’ hypothesis which contends that the rapid fall in prevalence after 2012 was a result of previous Tobacco Control policies which suddenly started working, after 5 years of rigor mortis, at exactly the same time as e-cigs went mainstream”