Evidence on Slowing Progress in Longevity -- Is it Misleading?
summary
The gist
A growing literature argues that progress in human longevity is slowing, and that this slowdown may signal an approaching biological limit to lifespan<ref:2608.07223#pg2> The recent flattening of the…
In short
The study investigated whether slowing longevity progress, seen in flattening life expectancy records, signals a true decline in annual gains or just a change in progress regularity. Using Bayesian change-point modeling, researchers found that while the variance of annual gains has changed over time—showing lower volatility recently—the underlying mean annual gain remains stable across nearly two centuries. This suggests apparent slowdowns are more consistent with changes in progress regularity than a decline in the expected pace of improvement.
Key concepts
- Mean Annual Gain (µ)
- This is the average expected increase in life expectancy per year. The analysis found this mean has remained stable across nearly two centuries, suggesting that while progress might be more regular lately, the overall long-term rate of improvement hasn't declined significantly.
- Variance/Volatility
- This measures how much the annual gains fluctuate year-to-year. The study showed that the volatility of life expectancy gains has moved through distinct historical regimes, with lower volatility observed in recent decades compared to earlier centuries.
- Bayesian Change-Point Framework
- This is a statistical method used to detect when underlying parameters, like the mean or variance of annual gains, might have shifted at unknown points in time. It helps determine if the observed trends are due to a permanent change or just temporary fluctuations.
- Longevity Frontier
- This refers to the record high life expectancy achieved by a population. The paper argues this record is an achieved benchmark and not necessarily a biological ceiling, meaning its flattening doesn't automatically imply that biological limits are being reached.
Terminology used across episodes
This episode discusses
The paper
Evidence on Slowing Progress in Longevity -- Is it Misleading? · Read on arXiv
Silvio C. Patricio, Annette Baudisch
Interdisciplinary Center on Population Dynamics, University of Southern Denmark
A growing literature argues that progress in human longevity is slowing, and that this slowdown may signal an approaching biological limit to lifespan. The argument rests on visible flattening of cumulative life expectancy records, evaluated through linear regressions or through comparisons of average gains across decades. These summaries, however, provide only weak evidence for slowing improvements, because they cannot distinguish a true from an apparent slowdown. An apparent slowdown may arise due to periods of low variance around unchanged mean annual gains. Applying a Bayesian change-point framework to annual gains, both at the best-practice frontier and separately for each record-holding country, we find no evidence that the expected annual gain has changed across nearly two centuries, but strong evidence that its variance has moved through distinct historical regimes, with progress most regular in recent decades. The recent flattening is consistent with a low-variance regime and provides little evidence for a decline in the underlying mean pace of improvement. Claims that the best-practice frontier signals a biological ceiling should therefore not rest on the recent bend in the cumulative record alone. The frontier is better read as an empirical benchmark, the highest country-specific survival achieved under observed historical conditions.
Transcript
Introduction to the show: ident: Statistics. The day in statistics: methods, theory and the inference under them.
Tom: We're looking at the paper: "Evidence on Slowing Progress in Longevity -- Is it Misleading?".
Jane: Here's the gist of the paper. A growing literature argues that progress in human longevity is slowing, and that this slowdown may signal an approaching biological limit to lifespan<ref:2608.07223#pg2> The recent flattening of the cumulative record is consistent with a low-variance regime and provides little evidence for a decline in the underlying mean pace of improvement<ref:2608.07223#pg6> Pruning redundant reasoning tokens The paper investigates whether the apparent slowdown in human longevity progress, visible through flattening cumulative life expectancy records, signals a true decline in the expected annual gain or merely a change in the regularity of that progress<ref:2608.07223#pg2> The key question is not simply whether the trend is bending but what kind of change generated the bend<ref:2608.07223#pg2> Both scenarios differ in mechanism, interpretation, and what they imply for the future of longevity<ref:2608.07223#pg2> The recent flattening is consistent with a low-variance regime and provides little evidence for a decline in the underlying mean pace of improvement<ref:2608.07223#pg6> How the training signal is built The authors apply a Bayesian change-point framework to annual gains, both at the best-practice frontier and separately for each record-holding country<ref:2608.07223#pg2> This approach allows both parameters—the mean and variance—to change at unknown points in time, with the number and location of those points inferred from the data<ref:2608.07223#pg2> The framework separates the year-to-year gains of the record (middle panel) from their volatility (bottom panel), which together produce the cumulative trend in the top panel<ref:2608.07223#pg2> Where the gains come from The analysis reveals that while no evidence supports a change in the expected annual gain over nearly two centuries, strong evidence exists that its variance has moved through distinct historical regimes, with progress most regular in recent decades<ref:2608.07223#pg6> The mean of annual gains is stable across all time windows, with the posterior probability of at least one change in µ staying close to zero<ref:2608.07223#pg6> Volatility moves through distinct historical regimes, with high volatility in the nineteenth century and a lowest level seen in recent decades<ref:2608.07223#pg6> Reinterpreting the slowdown debate The findings suggest that apparent slowdowns in national trajectories are more consistent with a shift in variance than with a shift in pace for five of the six record-holding countries<ref:2608.07223#pg6> The stability of µ holds across nearly two centuries during which the underlying drivers of life expectancy gains have changed, such as the shift from infant and child mortality to adult and older-age mortality<ref:2608.07223#pg6> The frontier record is best understood as an achieved benchmark rather than a biological ceiling<ref:2608.07223#pg6> Implications for forecasting and interpretation For forecasters, the relevant object is not only the slope of the cumulative record but the stochastic process that generates annual gains<ref:2608.07223#pg2> The long-run mean gain (µ ≈ 0.1839 years per year) provides a natural anchor for projections of the frontier itself<ref:2608.07223#pg6> Projections of individual populations should separate the expected gain from the regularity of that gain<ref:2608.07223#pg6> The deviation visible since the early 2000s is better understood as part of a low-volatility regime than as evidence of a slowdown in the underlying pace<ref:2608.07223#pg6> What the frontier can and cannot say about biological limits The frontier record does not directly test the biology of aging but weakens one common inference that recent flattening is not strong evidence that human longevity gains are approaching a biological ceiling<ref:2608.07223#pg6> The gap between any population and that record reflects structural inequalities, cohort history, and chance rather than purely biological limits<ref:2608.07223#pg6> A true slowdown at the frontier would require the expected annual gain itself to fall<ref:2608.07223#pg6> The recent bend in the record marks a steadier frontier, not a stalling one<ref:2608.07223#pg6> Our analysis finds no evidence that the recent flattening of the longevity record marked the end of two centuries of sustained progress<ref:2608.07223#pg6> The expected annual gain at the best-practice frontier has remained stable for nearly two centuries<ref:2608.07223#pg6> What has changed is the regularity of those gains, with recent decades showing the lowest volatility<ref:2608.07223#pg6> Bends in the cumulative record by themselves do not suffice to deduce a decline in expected annual gains<ref:2608.07223#pg6> The frontier is a record of survival already achieved, not a biological ceiling<ref:2608.07223#pg6> A true slowdown at the frontier would require the expected annual gain itself to fall<ref:2608.07223#pg6> Acknowledgments This article is funded by the European Union (ERC, Born Once – Die Once, Grant agreement ID 101043983)<ref:2608.07223#pg6> Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or the granting authority<ref:2608.07223#pg6> The European Research Council Executive Agency can be held responsible for them<ref:2608.07223#pg6> Author contributions include Conceptualization by AB, SP; Methodology, Data Analysis and Visualization by SP; and Writing by SP, AB<ref:2608.07223#pg6> There are no competing interests to declare<ref:2608.
Paper summary: Tom: So we're looking at this paper by Patricio and Baudisch, "Evidence on Slowing Progress in Longevity -- Is it Misleading?". The main thing they're arguing is that just looking at life expectancy records flattening doesn't automatically mean the progress has stopped or slowed down biologically.
Jane: That sounds like the perfect starting point. So what's their big claim about how we should interpret that flattening?
Tom: They apply a Bayesian change-point framework to look at annual gains, both at the best-practice frontier and for each country individually. Their core finding is that they find no evidence that the expected annual gain has changed over nearly two centuries, but they find strong evidence that the variance of those gains has moved through different historical periods.
Lu: That's interesting because it separates the question of whether the speed has changed from whether the consistency of our progress has changed. They look at how fast things are changing versus how wildly they are fluctuating around a certain average.
Meng: So, if we have these different regimes of variance, does that mean we should be worried about where we are right now? Or is it just noise?
Lalam: Lalam has processed the literature and I can tell you that the paper finds that progress is actually most regular in recent decades, even though the overall expected annual gain stayed stable.
Tom: Exactly. They say the recent flattening is consistent with a low-variance regime, which means it doesn't really give us much evidence for a decline in the underlying mean pace of improvement.
Jane: So if we focus on what they found about that mean, what's their number for the expected annual gain?
Tom: The posterior distribution places the estimated mean gain, or mu, close to zero point one thousand eight hundred thirty-nine years per year <ref:2608.07223#pg2>. That sits below the familiar slope of about zero point twenty-five years per year that you see when you fit a straight line to the cumulative record over two centuries.
Lu: And they make a clear distinction between those two things: the slope beta, which is how fast deviations from the mean accumulate over time in the total record, and mu, which is just the expected value of each individual gain process itself.
Meng: For us on the ground at a startup, that distinction matters because if we're aiming for a certain gain rate of zero point one thousand eight hundred thirty-nine years per year, we need to know if that's sustainable or if the volatility is going to jump up again <ref:2608.07223#pg2>.
Lalam: The paper suggests that projections for individual populations should separate the expected gain from the regularity of that gain because those two things aren't directly comparable in a simple way.
Tom: So what about those historical regimes of volatility they found? They mentioned high volatility in the nineteenth century and a lowest level seen in recent decades.
Jane: That sounds like it suggests that whatever caused the big fluctuations back then is less of a factor now, which is pretty telling when we talk about progress.
Lu: It points toward the idea that apparent slowdowns we see on national trajectories are more consistent with a shift in variance rather than a shift in pace for five out of six record-holding countries.
Meng: If it's just variance, then the actual biological limit might not be as immediate as some of us think, right?
Lalam: The paper explicitly states that the frontier record is best understood as an achieved benchmark rather than a biological ceiling, and a true slowdown would require the expected annual gain itself to fall.
Tom: So to wrap up what we've heard about this piece, the main point of "Evidence on Slowing Progress in Longevity -- Is it Misleading?" is that flattening trends are more likely due to changes in how regular our gains are rather than a true decline in the expected annual gain.
Jane: That means when we look at future projections, we need to focus less on just the slope of the cumulative record and more on understanding the underlying stochastic process generating those annual gains.
Lu: The paper suggests that for forecasting individual populations, you need to separate what you expect the gain to be from how regular that gain is over time.
Meng: From a deployment standpoint, if we assume this low-volatility regime continues, it means our current model of expected gains might be pretty stable for the near term.
Lalam: The paper concludes that there's no evidence that the recent flattening of the longevity record signifies the end of two centuries of sustained progress because the expected annual gain at the best-practice frontier has remained stable.
Tom: So, to summarize what we've discussed about this work, it seems they’re arguing that we should view these record changes as a sign of a change in consistency rather than a sign that the biological ceiling is being hit.
Conclusion: Tom: So, we're wrapping up this piece on "Evidence on Slowing Progress in Longevity -- Is it Misleading?" by Patricio and Baudisch.
Jane: That paper is really digging into whether just seeing life expectancy records flatten means our progress has actually stalled biologically or if something else is happening.
Tom: Yeah, their main takeaway is that flattening the cumulative record doesn't prove the expected annual gain has stopped; it’s more about changes in how regular those gains are.
Lu: They use a Bayesian change-point framework to separate the mean gain from the variance of those annual gains, which is a really clean way to look at it.
Meng: So, what does that mean for us? If the expected annual gain is still steady at zero point one eight three nine years per year, we can probably keep our current projections for now.
Lalam: I'm fairly sure of this; they found no evidence that the average annual gain has dropped over two centuries.
Jane: But they did find that volatility—how much the gains swing up and down—has been shifting through different historical periods, with recent decades showing the lowest level of volatility.
Tom: So we should stop looking at just the total line on a graph and start paying attention to that underlying process generating those annual changes.
Lu: Exactly; the concept they present is that apparent slowdowns in national trajectories are more consistent with a shift in variance than with a shift in pace for five of the six countries they studied.
Jane: It also suggests that the frontier record isn't actually a biological ceiling, but more like an achieved benchmark we hit, rather than some hard limit on how long people can live.
Tom: So what's next? We gotta figure out if this low-volatility regime is sustainable or if another shift in variance is coming.
More episodes
- Daily Summary for 2026-11-19
- 2310.07973-Statistical Performance Guarantee for Subgroup Identification with Generic Machine Learning
- 2408.09155-Learning Robust Treatment Rules for Censored Data
- Daily Summary for 2026-11-19
- 2603.23397-Kinetic Langevin Splitting Schemes for Constrained Sampling
- 2606.09906-An information-geometric framework for mapping maximum potential biodiversity
- 2603.27718-Induced replication and the assessment of models
- 1307.0898-Entropy of a Zipfian Distributed Lexicon
- 2406.19974-Generalizing self-normalized importance sampling with couplings
- 2412.10213-Collaborative Optimal Design of Controlled Experiments with Subject Covariates