August 22, 2024 – Welcome back after a long, relaxed summer break this year. Today’s post – suggested by a reader – deals with an intriguing question: how does early retirement impact my life expectancy? You see, since retiring six years ago, I’ve always assumed I’m doing my body a favor. I live a more relaxed life, exercise regularly, and sleep more every night. So, I’ve always believed my 2018 early retirement decision should improve my health and increase my life expectancy, right? Well, shockingly, the academic research on this topic is mixed. True, some studies seem to indicate early retirement reduces mortality. However, several studies point in the opposite direction. The commonly referenced rationale for this paradoxical result is that early retirees ostensibly often have reduced social contacts and physical and intellectual stimuli, increasing the risk for physical and mental decline.
Bummer! Was it a mistake to retire early? I did notice more gray hair in the last six years. Have you noticed a decline in my mental capacities? Am I becoming a curmudgeon? Be honest, everybody! Well, not so fast. There are many reasons not to be too concerned about (early) retirement’s alleged adverse health effects. Let’s take a look…
Before we get started, though, let me point you to three recent podcast appearances of mine.
- First, David Baughier interviewed me on his Forget about Money Podcast about Safe Withdrawal Rates and Sequence Risk. Please check out his podcast and boost his traffic.
- Second, I was on Chris Hutchins’ “All the Hacks Podcast” to talk about Beyond the 4% Rule: Smarter Strategies for Financial Independence.
- Finally, I again appeared on the awesome Two Sides of FI podcast to discuss my options trading strategy with Jason. Please check out the recording and share generously to show Jason and Eric some SEO love.
But now, back to today’s content…
Literature Review: Early Retirement vs. Mortality/Life Expectancy
I’m not an expert in this field, but a reader who suggested this topic gave me a reading list to review. Here’s a sample of papers I found. I should state that I read through them without replicating any of the studies. I categorize them by the estimated impact on life expectancy and provide a brief summary of the results:
1: Negative Impact on life expectancy
Zulkarnain, Alice and Rutledge, Matthew S., How Does Delayed Retirement Affect Mortality and Health? (October 5, 2018). Center for Retirement Research at Boston College, CRR WP 2018-11 , Available at SSRN: https://ssrn.com/abstract=3261325. This paper uses a dataset from the Netherlands and finds that delaying retirement reduces mortality for men. The paper finds no link between retirement timing and the incidence of diabetes or depression. Put differently, the change in mortality must come from conditions other than diabetes and depression. There is no discernible effect on women’s mortality.
Andeas Kuhn, Stefan Staubli, Jean-Philippe Wuellrich, and Josef Zweimüller, Fatal Attraction? Extended Unemployment Benefits, Labor Force Exits, and Mortality. NBER Working Paper No. 25124. Available at https://www.nber.org/system/files/working_papers/w25124/w25124.pdf. The paper finds that for men, there is a 0.2-year reduction in the age at death for each year a person retires earlier. The effect is more prominent in blue-collar workers and workers with low work experience. For white-collar workers and workers with high work experience, the effect is less prominent and statistically insignificant. There is no discernible effect on women.
Stephanie Behncke. Does retirement trigger ill health? Health Econ. 2012 Mar 21(3):282-300. doi: 0.1002/hec.1712. Epub 2011 Feb 14. PMID: 21322085. Available at https://pubmed.ncbi.nlm.nih.gov/21322085/. The paper uses data from the English Longitudinal Study of Ageing (ELSA) and finds that early retirement negatively impacts various health indicators.
2: Positive Impact on life expectancy
Cristina Bellés-Obrero, Sergi Jiménez-Martín, Han Ye. The Effect of Removing Early Retirement on Mortality. September 2022. IZA DP No. 15577. Available here: https://docs.iza.org/dp15577.pdf. Retiring later increases the mortality risk. The effects are worse for low-skilled workers as well as in “physically and psychosocially demanding jobs.”
Bound, John and Waidmann, Timothy, Estimating the Health Effects of Retirement (October 1, 2007). Michigan Retirement Research Center Research Paper No. UM WP 2007-168, Available at SSRN: https://ssrn.com/abstract=1082047. For men, retirement has a slightly positive health impact.
Why is there such disagreement in empirical studies? Let me list a few reasons:
Endogeneity
A significant concern is that the retirement decision is endogenous. Specifically, it is often linked to an individual’s health. Thus, it’s frequently not clear if a person died early because they retired early or because they knew of their underlying health condition, and that caused them to retire early to enjoy their last few years of reduced life expectancy. Just running a blind OLS regression of mortality on early retirement may falsely attribute mortality and bad health to early retirement when the true causality goes the other way around.
In a different area, in pharmaceutical research, you want to avoid studies where people endogenously pick what drug they take because a new and effective drug might appear ineffective or even harmful if patients with more severe cases of a condition are more willing (desperate?) to try it. Thus, you want to create a study where the subjects are exogenously and randomly assigned a drug vs. placebo. Statistically speaking, that’s the gold standard to avoid this endogeneity issue and other problems. The analog in early retirement vs. mortality research would have been to take a sample of people, force some of them to retire at one age, the rest at a later age, and then monitor their health and mortality for the next few years. That would be impractical, unenforcible, and not to mention unethical. So, mortality researchers often have to rely on the crummy datasets they can get their hands on.
Of course, researchers are aware of this endogeneity issue and have found ways to correct it. Every paper I’ve seen in this field claims to account for endogeneity using a so-called instrumental variable (IV) approach. Without getting too geeky, IV means finding another explanatory variable, Z, with a causal relationship with the endogenous independent variable, X, (work vs. retirement decision). But the IV variable should impact the dependent variable, Y, (mortality or health) only indirectly through its correlation with X. For example, Zulkarnain and Rutledge (2018) use a variable DWB (“Doorwerkbonus”), which is the eligibility for a temporary Dutch work incentive bonus program that likely correlates with the retirement decision but not directly with a subject’s health. In the first regression stage, the authors estimate the work status of each subject using the DWB variable as one of the independent variables. In a second-stage regression, the authors then use not the actual but the estimated(!!!) work vs. retirement status as an independent variable to account for health and mortality effects five years later.
In any case, the endogeneity problem is present in every study. For example, Zulkarnain and Rutledge (2018) report that in all naive OLS regressions (i.e., ignoring the IV methodology), retirement consistently had a detrimental impact on health and mortality for both males and females. However, the effect is weaker when they use the 2-stage IV method. However, the effect remains statistically significant for males.
Still, there is no guarantee that you will eliminate all of the endogeneity with your IV methodology. Depending on how well your IV approach works, you may remove the endogeneity and find that early retirement improves health, as seen in some papers. Or you may only partially remove it, which may explain some negative health results of early retirement. When I studied and practiced economics, I ranked these empirical studies regressing everything in the kitchen sink and waiving the magic IV wand as somewhere between intellectual lightweights and economic junk science. So, I’m not too concerned about those empirical results.
Also, endogeneity is not the only headache, which brings me to this next point…
Simpson’s Paradox
Mortality is a complex issue, and trying to account for cross-sectional differences with one single explanatory variable, like your retirement date, is rife with problems. Explaining one outcome with one factor and ignoring others can lead to paradoxical results. It is so paradoxical that it got its own name: Simpson’s Paradox, named after British statistician Edward Simpson. (To be clear, I’m not claiming that the studies ran univariate regressions, but missing variables in multivariate regressions create the exact same Simpson’s Paradox.)
The paradox may present itself in many ways, but one classic is this numerical example. Imagine we have two cohorts, A and B. Within each cohort, we observe variables X and Y. We could think of X as the retirement age and Y as the age at death. The top two panels show that X and Y are negatively correlated in both subsamples. The R^2 (goodness of regression fit) is around 0.8 in each subsample, and the slope parameter, about -0.20 in each sample, is significantly negative in each case, with a t-stat in the double digits (not reported in the Excel charts here, but I did calculate it separately).
But notice what happens if we merge the two cohorts into one sample as I did in the bottom panel: Now we get a positive slope of +0.5! Still, a decent R^2. The t-stat of the slope is still above 10. What happened here? The two subsamples have vastly different mean values for their X and Y variables. Thus, the slope of the joint OLS regression is mainly impacted by the location of the point clouds in the scatterplot rather than the true underlying relationship in X and Y within the subsamples.
This data paradox is indeed a headache for mortality researchers. Imagine Cohort A is comprised of construction workers, while Cohort B is comprised of college professors. Construction work is very physically demanding, often necessitating a lower retirement age. Moreover, the job also takes a toll on your body and may reduce your life expectancy. On the other hand, college professors tend to work much longer and have a higher life expectancy. Thus, effects from other variables like the type of job may improperly impact our early retirement slope estimate if not correctly accounted for.
Of course, data scientists will tell you that you need not worry about Simpson because we can cure the Paradox if we simply “control” for all those other pesky factors. And what econometricians mean by “controlling” is that they throw an extensive range of additional explanatory variables into the regression equation: gender, education, health status, industry, marital status, income, wealth – you name it, whatever is included in your database. And I should stress that all of the empirical studies referenced above control for several other possible mortality factors. You’d be laughed out of the room if you tried to sell a univariate regression linking mortality to only the retirement age.
So, researchers have done their duty to alleviate all concerns about that pesky Simson’s paradox after throwing all the obvious variables into the statistical kitchen sink, right? Wrong! While all the mortality researchers hope they haven’t missed anything crucial, they can’t control for every conceivable additional mortality factor. Some unknown and/or unobservable effects might still lurk in your sample and could tilt your results. Not all databases have all the necessary series. You may still get nonsensical slope estimates for the early retirement variable.
But let’s assume the retirement researcher has thought of everything and has access to all the data series. They don’t, but even if they did, there is still a way to mess it up, which brings me to the next issue…
Dumb Dummy Variables
Another potential concern about controlling for everything in the kitchen sink is that including the other variables will not fix the problem if done improperly or incompletely. To illustrate this, let’s look at another numerical example, again with completely made-up data.
Imagine two cohorts again, A and B, each displaying a significant statistical relationship between variables X and Y. However, the slopes are different this time: -0.20 in Cohort A and +0.20 in Cohort B. If we aggregate the sample, we are left with a big nothing-burger.
We can also study the detailed OLS regression results; see the table below. In Model 3, where we merge the two samples without any dummy variables, the R^2 is now only 0.0001, the slope is close to zero, and its t-stat is no longer significant. No surprise here! However, even including a dummy variable for Cohort A in Model 4 does not cure our problem. Both the dummy variable and the X slope are statistically insignificant. The R^2 is only 0.07. Only in model 5, where we have an intercept, a dummy for Cohort A, and two separate slopes for the two cohorts, would we recover the information from the two cohorts again. Notice that the intercept plus Cohort A dummy sum up to the actual intercept in Model 1: 64.63+21.07=85.70.
Thus, to fix this issue, we’d need to include two slope parameters, one for Cohort A and one for Cohort B, in addition to the dummy. That way, we’d effectively model two separate intercepts and two separate slopes, as observed in the data. Forcing the slope to be the same when it’s clearly not, will give you misleading results if the two types are included in the regression equation only through separate dummy variables. And that’s exactly what the retirement mortality researchers are doing. At least in most of the papers I found.
Of course, the researchers indeed run separate regressions, usually for males vs. females or white-collar vs. blue-collar jobs. But the “different slope” setup can and should appear with many other variables. For example, it’s conceivable that every industry and job duty should have different marginal impacts of early retirement on mortality. The same may hold for different wealth and income levels, pre-retirement health status, etc. Forcing the early retirement impact to have the same slope for all while only including some dummy variables can give you nonsensical results.
But the problem is likely even worse, which leads me to the section…
Unobservable Factors Impacting Life Expectancy
Dummy variables and separate slopes are only feasible if the Cohort A vs. Cohort B status is observable. What if the two cohorts are due to unobservable psychological differences? Say, Cohort A are the productive and healthy folks who flourish in retirement; they improve their health status and increase their life expectancy. In Cohort B, on the other hand, we have the “sloths” who lose their social circles, avoid physical and intellectual stimuli, and die earlier due to depression and diabetes.
And what if the Sloth vs. Productive early retirement status is utterly uncorrelated to all the other observables? It can happen to men and women, educated and uneducated retirees, high-income and low-income retirees, high-wealth and low-wealth retirees, etc. It would render the empirical research all but useless. Ironically, this would explain the wide range of empirical results, i.e., depending on whether you oversample or undersample the two unobservable types, you get positive or negative outcomes due to early retirement.
Quality-Adjusted Life Expectancy Matters!
Just as a thought experiment, let’s assume that the statistical studies point to an increase in mortality due to early retirement. Say, someone comes up with the definitive proof, methodologically correct and without any data flaws, that early retirement indeed reduces everyone’s life expectancy. My response would be, “So what!?” I mostly retired to increase my quality of life. The “Quality over Quantity” maxime works here as in many other areas! In other words, imagine that back in 2018, I had a life expectancy of 40 more years if I kept working, and by retiring early, I would have reduced that by three months (as estimated in one of the studies). Would I have still retired? Heck, yes! Assuming I would have worked 2,500 hours a year, that’s 50,000 hours behind my office desk; that’s the equivalent of almost six years of 24/7 at the office. A three-month reduction in my life expectancy is well worth that tradeoff.
Conclusion
I worry very little about this empirical early retirement mortality research. First of all, the empirical results are all over the map. Linking a single variable, early retirement, to some of the most complex observables like health and mortality creates various statistical challenges. As a former economist and data scientist, albeit in a different subfield of economics, i.e., macroeconomics and finance, I always thought that most of these purely empirical data-mining exercises in large cross-sectional databases were economic junk science. It’s a bit like sausage-making; the final result can be tasty (intriguing and newsworthy results), but you’d lose your appetite if you witnessed the actual sausage-making.
Second, even in papers that “prove” a negative impact of early retirement, the results are often not 100% consistent across the board. For example, according to one study, the effect is not even statistically significant for white-collar workers. Well, I was a white-collar worker, so I should be safe. Case closed!
Third, even if someone could conclusively prove to me that, on average, early retirement indeed leads to adverse health outcomes, like depression, diabetes, high blood pressure, etc. I don’t care. I’m not an average retiree. My decision and recommendation to others would remain the same: retire early and don’t be that lethargic sloth who wastes away in early retirement. Do something useful with your early retirement. No statistician with their IV mumbo jumbo can convince me that early retirement is bad for your health if you stay fit (mentally and physically), volunteer in your community, make new friends, travel extensively, etc. And, even in the worst case, where you indeed have a slightly reduced life expectancy, you at least increase the quality of life during early retirement relative what you would have done in a corporate job.
In that spirit, everybody, have a fulfilling early retirement and hopefully a long life as well!
Thanks for stopping by today. Please leave your comments and suggestions below!
Title Picture credit: pixabay.com
I am curious about this too. I used to work a physically demanding job, and I would work out every day as well. Now I have been retire since age 32 for a few years now. I still work out every day and also go on long walks quite often. However, I have noticed that if I don’t feel like walking, I’m tired, or have aches and pains, I will often just decide to relax instead, where before I wouldn’t have had the option and would have to be active for 8 hours per days no matter what. I can also go long periods where I am not doing any cognitively demanding tasks even though I still try to do those too.
The best way to describe it is that maybe I’ve gotten a little more soft or wimpy, although I am certainly way less stressed, healthier in many ways and happier.
I am curious what the long term affects will be.
Retired at age 32. Say more
I retired with my wife who was 28 at the time in March 2022, and simultaneously sold all our stuff and started a nomadic life. It’s been going great so far and we have experienced a lot =D.
Congrats! Sounds like you’re enjoying FIRE. Simplifying and decluttering life is one of the great benefits.
Congrats! You’ll have a long life expectancy.
Thanks for sharing. That’s the danger: one could slowly drift into a lethargic state. But as long as we remind each other of the need for exercise, I think we’ll be OK!
Dude, you are not retired, you are not working. Unless you have a $10 mil trust fund managed by a competent team, a solid prenup and lucky genetics, you will be back at work in a few years. Inflation will absolutely shred any portfolio worthless than $10 mil if you spend less than $50k. FIRE movement is a result of stable prices, low inflation. There wasn’t a FIRE movement in the 70’s or 1920’s in Germany.
Your statement is alarmist. True there are countries where the 4% hadn’t worked due to war such as Germany, Austria and Japan. No withdrawal rate was adequate. However you also now how access to global etfs, so the war would need to be all countries, and by definition any form of planning is pointless if you assume global war. I’d rather plan for no war rather than be bitter.
Yes, some countries fared very poorly if you had invested in paper assets.
But also note that real estate would have fared better. In Germany there is a foundation that has provided public housing for 500 years, through all crazy history: https://www.fugger.de/en/foundations
The fact that one has to search for examples of wealth that lasts longer than 100 years is a good indicator that managing money for long durations is filled with landmines. Most of the examples are for entities that have massive government protection, not an individual. A 32 year old married dude has an incredibly difficult task in remaining retired for the duration of a two party system. If he’s a finance quant, maybe, otherwise needs at minimum $10 mil to become part of a family office who can manage the accounts with this in mind.
I don’t know what finance quant has to do with it. If people invest in a large-cap stock index fund, you could have survived 100+ years, without being a finance expert.
I very much like your witty post and the easy to read explanations of the shortcomings of studies on early retirement and life expectancy but this reply…?
Yes, you can still find some tiny traces of the big Fugger empire today, a family whose richness and political power was unheard of before. But this folklorist one social housing project as a cure to a national or even international disaster feels funny at best.
After almost 10 years of blogging it always amazes me when readers latch on to some small detail and what kind of Strawman Arguments folks conjure up. I never argued that real estate is a cure to international disasters. If you read it that way, please practice your reading comprehension.
This morning… I’m extremely jealous that I can’t say something along those lines at my job.
$10m? Are you Suze Orman in disguise? I had less than that in 2018 and I am still retired.
Also, there have always been independently rich people living off their wealth. Even in the 1970s in the USA and 1920s in Germany.
You are a Finance quant and “paying” yourself $100k+/ year, replacing the cost all but a few non Finance quants portfolios would have to absorb. Technically, you are “working” on your portfolio, saving the fees charged on most accounts. $20+ mil liquid (probably $50 mil for illiquid/complex assets) is the real number to afford competent management in a shared family office. $10 mil is tricky without a lot of finance quant, tax, trust, etc. skills. I am new to the FIRE movement, backward looking have been “Coast FIRE” for about 20 years, but remember laughing out loud when reading Financial Samurai expected to stay retired with a few million in NorCal. He’s not a finance quant, but a paper pusher who worked in the Finance industry, probably contributing to his attribution bias and failed estimations requiring him to return to full-time work after maybe 10 years of retirement (while also writing books and managing a website, not really retired).
Just to be clear: we don’t know each other. You don’t know what I do every day. You should refrain from making sweeping statements about what I do or don’t do.
Even if I hadn’t done my options strategy, I would still have a very impressive portfolio. These 10, 20, and $50mil portfolio estimates sound really insane. In today’s market, a $2m-3$m portfolio can give you a comfortable retirement with purely passive assets generating $75k-$110k. If you feel adventurous and you do my options trading with a $1m taxable account and make another $35-$40k, good for you. But it’s not necessary.
My recommendation since you’re new to the FIRE movement: enjoy the content here and elsewhere. But you’re not entitled to throwing away widely accepted and established insights with just some vague but sweeping drive-by comments on the internet.
I think that the question of how long will I live after retirement is the $64,000 question?
I quit my high stress, low rewards job a couple of years ago. I thought I would be feeling better without the stress of commuting and work. I was wrong. The process of unwinding years of working through the pain and stress took a couple of years to unwind. You don’t realize that like a frog you are train to sit in a hot vat of water till you are cooked.
Fortunately I realized the path I was heading down and I started to invest over 40 years ago.
I had a lot of set backs, but because I saved for retirement with a lot of effort I can say I am financially ok. In a few years I might be able to enjoy expensive vacations without any worries.
I noticed that the first two years I encountered sleepless nights and grey hairs because I had to stop myself from thinking about running in first gear, peddle to the metal.
Now days I enjoy watching YouTube videos, or streaming movies. The cost of food does not bother me too much. My former co-workers took buy outs and those that did not, are still stressed.
I have seen many coworkers die prematurely before they got a chance to enjoy what life has to offer.
If a person can afford to go retire, they should. Remember most work places are not your friend. Join a club if you need social connections. Always leave a margin of safety so you don’t come up short into your retirement. Life is short, because I do get up and wonder where the past 40 plus years went?
Good message… I remember when I was beginning my working career, an older gentleman who was about to retire would always talk about this dream trip to Thailand he was going to take. Talked about it non stop and all of the other trips he was going to take. Well he finally booked that trip and finally retired.
Then we got word he’d died on the plane ride back. Had a blood clot that went the wrong way and nothing could be done.
For sure, enjoy life while you can.
Yeah, those stories are common. Do the fun stuff when you’re young! No regression will properly pick up the great irony and tragedy in stories like that.
Thanks for sharing. Agree: I also prefer social interaction at non-work places.
The quality of life example at the end reminds me of the example where someone runs every day for 4 hours to add an extra year of life expectancy.
I agree it’s unobservable because of health and depression. I expect those two factors drive life expectancy far more than retirement age. If you have a purpose post retirement I don’t see why you life expectancy should decrease.
4h of running seems a bit excessive, but, agree that purpose in retirement will help. It sounds like a cliche, but retire to somthing, not from a job. 🙂
When I read through the first paragraphs and heard about the differing results I was wondering how much the studies really sussed out any novel relationships va just finding oddities in the data. You know, as an econometrician, that if you try to find a specific relationship in a set of economic data you are very likely to do so regardless if it means anything and that’s correlation vs causation. Anywho…
Second part… I’m not sure any of them went through (for example) all of the different FIRE subsets to categorize how one is retiring (fat or lean or barista, etc)… take my step dad right now, who stepped away from being a full time lawyer to take odd jobs that interest him as they come up. He still owns his practice but doesn’t work there full time (he’s a barely on-time worker as I joke to him). Is he retired or not and how does that push the data? His wife works full time and is VERY successful so does that have an impact? He certainly attends a lot of social functions as part of her job.
Last bit, was hoping for a really geeky meta-analysis at the end of this but I suspect it might not be worth even trying to analyze to that level. Question: do you think applying something like PCA to the data to attempt to find what factors are pushing the output more and less would be beneficial? Then maybe some differing regressions could be applied that “might” mean something more significant. Did any of the researchers do this?
Speaking of meta-analysis, I ran across this one earlier this month, which concluded:
“Early retirement was not associated with a higher risk of mortality. On-time retirement was associated with a higher risk of mortality, which might reflect the healthy worker effect. It is important to consider information on prior health and demographics when studying the association between retirement and mortality to avoid biased findings.” (Association between retirement and mortality: working longer, living longer? A systematic review and meta-analysis, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7307664/)
And this analysis (of different studies/data) by an actuary (who as such should have both the statistical background needed to interpret the studies/data and the professional expertise to contextualize higher level interpretation):
“it seems that engagement with life is what helps prolong life. You can get engagement with life from working, but you can also get it from taking up causes, volunteering, hobbies, and contributing to your family and community.” (Does Working Longer Increase Your Lifespan?, https://www.cbsnews.com/news/does-working-longer-increase-your-lifespan/)
nice! you beat me to it!
Thanks for the links! Makes perfect sense!
Thanks for the links. I’ll have to spend some time noodling and digging through.
Agree: these datasets capture all retirees. It would be nice to do a study just for FIRE folks and compare that with others who retire at a “normal” age. That would also introduce biases, though. FIRE hasn’t been around for long enough (or at least not as prominently as today), so we can’t really study the long-term effects. There aren’t enough 90yo folks who retired 50 years ago.
Yeah, PCA might shed some light. Especially in samples with a lot of data series per person. I’m retired, so someone should tackle that! 🙂
Yea it’s would be more of a proof of concept / observation of a trend at this point. My intuition says there’s nothing here or any trend/seasonal-cyclical is buried under other cross-correlated factors.
Don’t know that it is worth my time either 😛 … Besides, I’m still trying to figure out how to get 1,000 barrels of crude oil into my backyard…
The getting the into your backyard is still easier than getting them out of there. 🙂
Oh, maybe I have a business idea… capture the spread on the futures… start a crack spread modeling business that rides on top of local neighborhood oil storage and distribution… I’m sure there will be nothing to stop me from world domination!!
Tough one now, because the spreads are no longer as we need them, i.e., contango and large spreads. Now we got backwardation and small spreads.
Easy to use simple match based on characteristics such as spending/total assets, age and BMI. Decile the dataset and compare treatment of retiring age within each decile to see if this dummy variable is significant.
Yeah, if you can get the data. Most large dataset have no info on BMI, though.
“Am I becoming a curmudgeon?” Ha ha ha – of course! Never change 😉
1 piece of anecdata: I am pretty sure my early retired physical health and mental health/happiness is the best it’s been since high school. My main stressors are “I need to take the car in for an oil change” and “where do I want to spend a few months of summertime vacationing next year?”.
OK, I insist I was a curmudgeon before I left the office, so that’s definitely not due to FIRE!
And you’re right: It’s not just a tongue-in-cheek comment: sometimes retirement gets so relaxed, even an oil change and other mundae tasks feel a little bit like stress. No comparison to the Sunday night blues and deadlines at work, though. 🙂
As you correctly note, there is no way to separate out the people who choose to retire early from the people who are forced to retire early. FIRE adherents are a remarkably small percentage of the population.
The vast majority of people who retire before full retirement age do so because they have to, not because they want to – most often people retire early for a reason that is independently correlated with shorter life expectancy – serious workplace injury, becoming a caretaker of a loved one, dementia, alcoholism, cancer, etc. Including these people, who haven’t voluntarily left work, with the very small percentage of people who voluntarily leave work before full retirement age is bound to skew the results.
I think that’s part of why you don’t see any correlation for earlier deaths for women. Historically, women have been less likely to be the primary breadwinner and would be more likely to retire early simply because they didn’t need to continue working. This is just intuition, but I suspect if the data was available, it would show that a significantly higher portion of female retirees were retiring by choice rather than retiring by necessity and that very well could explain the lack of correlation for women. Most men retire early out of necessity (and that necessity is strongly independently correlated with negative medical outcomes), whereas a higher percentage of women retire by choice.
All really good points. And it shows in some studies that women don’t have the negative impact. Which is exactly due to the reason you state: women who retire early do so voluntarily and happily.
Read “How Not to Die” by Dr. Michael Greger. What kills people? Cardiovascular disease and cancer. If you retire and spend your time drinking and eating in ways that are known to cause poor health outcomes, expect a poor outcome.
I love this recommendation. “How Not to Die” has been very influential on my dietary choices. I felt the blog post was interesting but it would have been stronger if it had gone more into factors linked with longevity within the general population such as those discussed in lifestyle medicine and e.g. Blue Zones. Causes of longevity within the country of Japan for instance merit particular study. I feel the population of people who choose or are forced to retire early is such a small subset of the general population that it’s more informative to just look at what increases longevity among the general population.
I’m also a bit disappointed in the community of people who discuss retirement and early retirement in general for not hitting on these points. I understand this blog is about a lot of finance geek stuff which is great, because everyone has their personal interests and niche. However, I feel writers who discuss retirement are somehow in general under-addressing socialization, daily exercise, and what people eat. These are very relevant and actionable topics as for instance one of my parents is retired and the other is working part time. If people ask about investments it’s like yes investments are important but what is also important are all these other factors.
I feel a lot of the bias is understandably due to a lot of authors who write about retirement live in the U.S., which is an extremely capitalist economy that is not geared towards health, and is more suited to exploiting individual people than addressing their needs. So the bias somehow reflects the cultural context.
I will check it out. This seems to confirm that ER is not the most important determinant. It’s how you do it and how well you take of your body after FIRE.
Based on what I’ve read and discussions with my brother who is a medical doctor, I think so. I feel it’s also better to think in terms of financial, physiological, and lifestyle variables. The balance of the scientific evidence supports eating a whole plants, mostly vegan diet with low sodium intake, being physically active daily, having a healthy body weight, being socially engaged, mentally active to reduce risks of e.g. dementia, having low stress, having access to healthcare, somehow having a few kids seems to help with longevity, and having a purpose. FI statistically would probably positively correlate with a bunch of these, RE probably negatively correlate since as you pointed out people may be forced to RE for health reasons, and FI+RE hard to really say conclusively since it’s such a small subset of people unless some enterprising scientist starts doing FIRE twin studies. But realistically the science is just going to boil down into all these financial, physiological, and lifestyle variables anyway, many of which are controllable, so it seems most practical to focus on them and not on things like FI, RE, FIRE.
I suppose if hypothetically some FIRE twin study conclusively establishes it’s best to work decades from now then FIRE folks can just go do some volunteering and pet some cats at the humane shelter.
Agree with all! Thanks for confirming this! Though, I still won’t give up red meat and my German beer anytime soon. 🙂
With respect to the above… there are just as many studies pointing to the complexities of a proper diet using just vegetables.
Specifically, there are VERY few exercise physiologists or sports scientists that would recommend cutting out meat based protein sources.
Dr Austin Baracki at Barbell Medicine had a great paper he published looking at the history of nutrition and concluded most modern recommendations were based on obviating nutritional deficiencies (eg vitamin B deficiency). Whereas modern sports science is more concerned about the longitudinal epidemiological analysis or sports and weight room performance over time. This has lead to a sharp divergence in recommendations depending on who you talk to.
While there are plenty of health outcomes that would necessitate focusing on soecific micro nutrients… a healthy person can start from the top down and worry about calories and macro nutrients to cover 80-90% of what they need and then use a multivitamin/multi mineral to cover the remainder.
As always though, there’s a lot of nuance and individuality in this.
He’s a video from a PhD on the topic:
https://youtu.be/uAfI2jwS-4M?si=gwoHxYToVNEoJC2e
Thanks for the link! Humans are omnivores, so I can see the point in that video.
All that’s to say… if you want justification for enjoying red meat and wine… I’ve got your scientific studies for you 🤪
Nice!;)
I would rather retire at 40 and die at 75 than retire at 70 and live til 85.
But, I simply don’t believe that would be the case (retiring earlier would lead to earlier death). Since I “retired” 3 years ago, I eat better, exercise more, get more sunlight and time outdoors, get more sleep, and have less stress. I engage in activities on a daily basis that IMO are more mentally/intellectually challenging than my job ever was (I could do my job in my sleep). I simply don’t believe that these things would lead to anything other than longer life expectancy.
But, I also think there are a lot of people who have addictions – alcohol probably being the most prominent, but also drugs, food etc., and having to show up to work every day is one thing that keeps that addiction in check. Without having to show up to work at 8am M-F, they “retire” to a life of sloth, gluttony, heavy drinking or other substance use, and yes they die. Don’t do that.
Yes, all great points. I think if folks are determined and organized enough to achieve FIRE, we should also figure out how to structure our FIRE and keep it healthy! 🙂
Hi BigERN,
Epidemiologist (and hopefully soon-to-be early retiree) here! Thanks for the really nice summary of the limitations of trying to answer this question in an observational setting. Esp appreciate the callout of confounding by indication (ie, treatment choice is influenced by prognosis) from my own subspecialty of pharmacoepidemiology 🙂
Perhaps one addl issue here is study design. As an example, when comparing those who retire early vs late, inclusion in the study is presumably conditional on surviving until retirement. Had MikeW above not retired from his stressful job and had a heart attack before reaching normal retirement age, he would have been excluded from the study (even though this hypothetical later retirement was indeed harmful to his health). By design, the later retirees have already survived longer in order to be included (making them more likely to keep surviving). Studying these “selective survivors” can result in counterintuitive findings where harmful exposures appear to have protective effects (eg, the “obesity paradox”). I admittedly did not do a thorough enough read to see if any of the studies addressed this point.
Anyway, given that FIRE is a relatively recent movement, I am skeptical that the data exists to answer this question in a way that would be generalizable to folks like us. But it sounds like a fun collaboration for an economist and an epidemiologist with a lot of spare time on their hands! 😉
P.S. For SpaceTimeMorph: a quick pubmed search did indeed turn up an SLR & meta-analysis on this topic — enjoy! doi: 10.1136/jech-2019-213023
Good points.
Some studies look strictly at mortality, so there is no selection bias. You would indeed capture the folks who didn’t retire yet.
But some studies look only at retired folks, and that would introduce a bias, i.e., you’d miss the folks who kept working and dies on the job or from stressful jobs before retirement.
Anyhow, if we ever encounter a good dataset, we should work on something together!
Thanks for the link.
Maybe I was more trying to egg ERN into providing us a spiffy meta analysis.
Something similar to what Greg Nuckols has done at the link below (but obv finance oriented) would be pretty cool.
https://www.strongerbyscience.com/master-list/
Final thought: this early retirement thing is great… you should definitely partake. 5/5 stars would recommend.
Yeah, in the ER vs. mortality research, the results are all over the map. Of course, I could build a meta analysis based purely on the positive health impact results. That would would look really neat. 😉
I think the hypothesis that should concern early retirees is the possibility that the sort of people who pull off an early retirement are the same sort of people who live well into their 90s.
If that’s the case, our 40 year olds running 30-year SWR simulations are grossly underestimating their financial risk.
You nailed it! I’ve been saying this for a while. The widely quoted 4% rule should typically be the 3.25%-3.5% rule of thumb for early retirees (tailored for individual circumstances with Karsten’s SWR toolkit spreadsheet, of course), since, I believe, all of the scholarly research which addressed longer durations (by Big ERN, Bill Bengen, Michael Kitces, David Blanchett, etc) shows that SWR needs to be reduced by at least 0.5% (I think Karsten may have said reduce by 0.5%-0.75%, but I’m sure he’ll correct or clarify if/as needed) for timeframes significantly exceeding 30 years.
I think the FIRE community has unfortunately fallen for “proof by repeated assertion” in applying Bengen/Trinity 30 year SWR studies to much longer timeframes, despite the studies not actually saying that. Bengen’s original paper showed an 18% failure rate for 4% SWR in 50 year timeframes.
True, but that’s conditional on being at the top of the stock market cycle. Some folks with very generous supplemental income (pensions, Social Security, etc.) may increase their WR to well above the Bengen estimate.
Exactly. Start with 3.25% or 3.5% instead of 4% for early retirees if no SWR tailoring, but best to personalize SWR to individual circumstances with your SWR toolkit spreadsheet, such as starting Shiller CAPE and supplemental income cashflows, and many will find higher SWRs.
But very early retirees may not have much supplemental income due to short work history, so unless they’re retiring into a low starting Shiller CAPE, they best not assume 4% would last longer than 30 years.
That’s right. The 3.25% is for the bare-bones and/or the people where supplemental cash flows are so far in the future where it wouldn’t matter much.
Exactly: If we have the grit, motivation and self-control to retire early, we should also master an ostensibly easier task of staying halfway healthy.
Great topic, thanks Karsten. I particularly like the “Quality-Adjusted Life Expectancy Matters”. What’s interesting is that I returned to work after a 1.5 year stint of early retirement for effectively the same argument you’ve retired early – my quality of life seems higher. I suspect primarily because I feel like I can better use the skills I’ve developed over the years (including the years in grad school to get my PhD) in my job than I could with early retirement activities, and I get far more regular interaction with my technical peers and friends as well. So ultimately that’s the metric that really matters – quality of life. BUT, I also consistently tell folks that being FI will let you enjoy your job a hell of a lot more, so it’s still important to go after that goal whether you retire early or not.
I did a similar thing – back to investment banking after a decent 2.5 year break. I find though that sometimes the workload is too stressful and other times it’s great. Wish it were different but if I’m enjoying more than 50% of my time it seems it’s still the right choice. But yes, our peers are working and not in early retirement so for the intellectual side we need to continue working and just manage stress.
Nice! Thanks for sharing. I keep myself busy and intellectually involved without a job, but I can see the allure of going back into the work world again after a long break. You can certainly do so with confidence and financially secure.
Yeah, the RE part of FIRE is not for everyone. I’ve heard a stories from a few folks who went back to work. More power to you!
For me there’s no way! 🙂
I retired at 42 and have now enjoyed 2.5 years of retirement at 44+. During this time, I’ve lost over 20+ pounds with very little effort. I absolutely love the freedom and vitality that come with being retired while still relatively young and healthy. I focus solely on activities and pursuits that truly matter to me and my family.
One of the most fulfilling aspects of retirement is being present for my young kids, providing them with immediate feedback and support when needed. It is kinda cool when they come and ask questions and I can take all the time in world just to focus on them.
To me this is a dream come true!!! (and I love it!!!) If I live shorter, so it be, at least I enjoyed sooner.
***Last thought, it took me about 6 months to adjust to a new life (initially was little hard to disconnect with old habits and connect with freedom).
Thanks, Karl! That’s one of the main reason for us, too: Being more involved with our daughter’s education. That alone is worth the opportunity cost!
For me, it would be interesting whether early retirees tend to be more happy than before (while still being “struck” at work) and also compared to people retiring regularly.
Personally, I wouldn’t care too much about a few months of life more or less (white collar job) – but maximizing my level of happiness for the next decades of my life is something really that bothers me.
But I doubt there are many papers about that topic?
I read in FIRE blogs, I read tons of books about FIRE and investing. I spent hundreds of hours in the past 5 years about this topic. After gotten all this input, I got myself a picture: many very (!) happy early retirees (between 28 and 45 y.o.) – but about one third does need to “work” in someway: a (monetized) blog, becoming a landlord, writing books (often about becoming FIRE), becoming a consultant, part-time work in the old job… but most of the latter party seem to be quite happy as well.
I wonder whether you (Karsten) did stumble about papers that proves my assumptions right or wrong?
Karsten, thanks a lot for your blog! IMHO, it is most likely the best one about ER, for sure from what I’ve read in the past years. Many thanks for your work, I’m glad you kept it up for so long!
Best regards from Germany,
Kevin (34) – still some way to go til FI
I hope that FIRE is now a big enough movement that some researcher would do some analysis eventually. But I’m not aware of anything available now.
Thanks for the kind words! Und viele Gruesse zurueck nach Deutschland!
I’m not aware of studies on FI(RE) and happiness specifically. There are studies showing more wealth does generally correlate with more happiness (https://www.theguardian.com/us-news/article/2024/jul/18/money-buys-happiness-study).
However, “Close relationships, more than money or fame, are what keep people happy throughout their lives” (https://news.harvard.edu/gazette/story/2017/04/over-nearly-80-years-harvard-study-has-been-showing-how-to-live-a-healthy-and-happy-life/)
For those who RE, finding meaning and purpose appears to be important in maintaining and growing happiness and life satisfaction. Work often provided some purpose, identity, and socialization (see above study, re importance od socialization and happiness) which needs to be replaced for retirees, early or otherwise. Thus the expression “it is important to retire to something, not just from something”.
Also, retirement in some cases can surface other issues that may have been buried. See https://affordanything.com/money-wont-solve-problems/ for a blog post on that.
Thanks for the links. Agree with all points. More money and more time to spend it will likely raise your happiness. But money will not solve all problems, everyone knows that.
Everyone *should* know that money won’t solve all problems (though it can certainly help with some!), but there was a surprisingly common myth early on in the FIRE community that achieving FIRE will lead to a life of “rainbows and unicorns”. I believe it is one of the reasons the early FIRE adherents were in such a hurry to get to lean FIRE, sometimes with excessive sacrifice along the way (e.g. Early Retirement Extreme and to a slightly lesser degree Mr Money Moustache, and Mr 1500 Days). But then, several bloggers who had achieved FIRE started sharing some of the challenges they experienced post FI as well as suggesting they may have rushed too quickly to achieve FI, at the expense of enjoying life along the way.
With experience, the FIRE community has embraced more balance in having a good and happy life along the way to achieving FIRE (Carl, aka Mr 1500 Days, Brandon aka Mad FIentist and Brad at ChooseFI being examples now espousing balance, both while working towards FI as after having achieved it). That in part has resulted in slow FI, coast FI, barista FI, fat FIRE, etc.
Just as SWR is about leveling income available for consumption (vs variable spending approaches), balanced FI vs extremely rapid lean FI is about leveling happiness and life satisfaction throughout the journey, and not just focusing on the destination (though there should be a boost in happiness from the autonomy and security that comes with achieving FI – which was certainly my experience).
Of course, I meant barista FI above, but didn’t notice auto-correct broke it.
Noted. I corrected the typo.
The problem is that in the FIRE blogging community ther is a first-mover advantage where established blogs get more traffic, so there is rush to the exit door to retire ASAP. But that’s bad guidance for the community at large. The best advice: Smell the roses and enjoy the ride, even during accumulation. In fact, the best writeup of that idea is, of course, from your truly: https://earlyretirementnow.com/2021/06/16/stealth-frugality/
Good stuff. Easiest way to determine if FIRE, retirement age or duration of retirement is a risk factor for mortality is ask a life insurance company if they charge RE more in premium than non-RE.
Data would be quite easy to structure for an accurate test, why most of these “studies” are designed to “find” what the researchers believe (see blue zones).
Easy #1 Create deciles for Spending/Total Asset, BMI, Age, [other control variables] (Fama-French-ish) create deciles for first two test if there is a significant difference between retirement treatment dummy variable or as a continuous variable within each group. Those who spend $50k and have $2 mil and a BMI below 25, etc.. then see if the RE group dies earlier than the non-RE group. Easy.
Easy #2 construct a regression discontinuity design where WR is the variable of separation. Those who retire with a WR below 4% vs those who have achieve a WR of 3%, test the mortality.
Everyone has a cousin, uncle, neighbor who retired at 32, took a trip to Thailand and OD’d the first weekend and another who worked on an off-shore oil rig until they were 103. My suspicion is that the forced early retiree due to collapse in productivity due to collapse in health is driving the results. A 50 year old law partner who is going blind due to diabetes can no longer read who is “retired” and dies a few months later is most likely driving early retirement early death results. In my working life, I found that anyone who finds value, meaning, connection with their job are empty useless people who are not able to find connections without bonding over TPS reports and performance reviews.
Yes, all good points. Of course, anecdotes like that are also not too useful and are just as misleading or worse than the empirical studies with all their problems. Maybe someone will eventually put together a dataset with enough x variables, like BMI and data on family health history, that makes everyone happy.
The relationship between early retirement and life expectancy is a topic that demands careful consideration. The article outlines how retirement, whether early or not, can have varying impacts on life expectancy depending on factors like lifestyle, activity levels, and social engagement. Interestingly, parallels can be drawn to the strategic investments in Dubai’s off-plan real estate market. [spam link removed by ERN]
Just as early retirees must carefully plan their post-career activities to ensure a fulfilling life, investors in Dubai’s off-plan properties must strategically assess market trends and future growth to maximize their investments. Both scenarios emphasize the importance of proactive planning for long-term success.
Thanks ChatGPT I will check out Dubai, lol.
If that’s the ChatGPT/AI productivity boost that everybody is raving about, then God help us…
It’s not the productivity boost but it is an example of the Dead Internet theory
Very sad, but true!
Wow, my blog must be really popular that the spammers come here. Normally, I erase those comments like this, but I found this very inventive, so I will let this one slide. 🙂
A great post as always. I appreciate you looking at a study and using your analytical chops to dig into it.
On that note, would you be interested in taking a look at this study: “The Safe Withdrawal Rate: Evidence from a Broad Sample of Developed Markets” by Aizhan Anarkulova, Scott Cederburg, Michael S. O’Doherty, Richard Sias from July 18, 2023?
here is the link:
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4227132
They claim to have created a monthly returns dataset for real returns for domestic stocks, international stocks, bonds, and bills for 38 developed countries from 1890 to 2019. (This is the first time I have ever seen anyone claim to have this dataset).
The authors start off the paper with the disquieting claim: “A 65-year-old couple willing to bear a 5% chance of financial ruin can withdraw just 2.26% per year, a rate materially lower than conventional advice (e.g., the 4% rule).”
I’ve done my own analysis and reached my own conclusions, but I always value your insights.
I’m not too fond of that paper, and here’s a long post detailing my views:
https://earlyretirementnow.com/2024/02/12/100-percent-stocks-for-the-long-run/
Since the paper is so idiotic, and one man can’t do the job of pointing out all the deficiencies of that paper all by himself, I also included the objections from Clifford Asness (AQR founder) and my fellow blogger BankerOnWheels.
Thanks for the referral. Much appreciated!
Just re-read your post “100% Stocks for the Long Run?” from 12FEB2024.
The paper referenced in that post was BTSQ (“Beyond the Status Quo: A Critical Assessment of Lifecycle Investment Advice,” written by Aizhan Anarkulova, Scott Cederburg, and Michael S. O’Doherty.)
The paper I referenced was ““The Safe Withdrawal Rate: Evidence from a Broad Sample of Developed Markets” by Aizhan Anarkulova, Scott Cederburg, Michael S. O’Doherty, Richard Sias from July 18, 2023)
The papers do share the same authors.
Additionally, the papers also seem to share the same methodology of using “asset returns from several dozen developed nations and then randomly draws blocks of realized return data to simulate potentially millions of years of asset returns.” I appreciate your detailed description and criticism of that approach.
Thanks.
Sorry, I misread that. I thought the paper was the dreaded BTSQ paper. But this current paper suffers from some of the same missteps, chiefly, mixing together irrelevant international return data. Thanks!
Interesting take on a grim (but crucial) aspect of the lifestyle choice that ER is.
I think a lot of it is down to personality type. Someone pursuing ER will likely also have the kind of personality that leads them to check the other boxes that lead to a longer life (prioritising future enhancing endeavours like losing weight, exercising, avoiding sedentarity, alcohol and tobacco which are the main killers). Socialising and shifting to non monetary pursuits that keeps on driving us is eventually on us (find your ikigai), and I feel requires a the kind of mindset shift that is much needed not to get bored when retired.
Ultimately though, I feel like living your life intentionally / on your own terms, even if it means stepping off the hedonic treadmill and giving up on the “more is better” attitude that most are chasing, would be worth giving up a decade or two. I would rather die early having done what I needed to do, rather than being full of regrets – looking at the list of top 5 regrets of the dying regularly is something I would recommend to most people pursuing the FI/RE path as it is a good reminder of what really matters : https://en.wikipedia.org/wiki/The_Top_Five_Regrets_of_the_Dying
Agree on this, living presently and enjoying that is definitely worth more than chasing a few more dollars if it means low quality of life now.
Enjoyed the book Lost Connections that focuses on this mismatch too.
Yes, very wise words: the traits that make FIRE more likely will hopefully correlate with the higher life expectancy traits. And no worries, if that’s not the case, you wouldn’t even give up decades of life expectancy. The measured effects in some of the studies are much smaller. Whew!!!
How old are you? My gray hairs went away after retirement. So maybe you are still stressed and have this internal desire for more attention, relevancy, and status given you’re highlighting what podcasts you’re on?
I’m sorry to hear your hair went away. Did you lose all your hair or only the gray hair? Luckily, I still have the same hairline as 30 years ago, albeit a bit more salt & pepper. But my wife thinks that I look even more sophisticated now.
I like to give a shoutout to other FIRE influencers and send them some SEO courtesy. It’s customary in the community. So my piece of advice is that you should refrain from deducing other people’s desires because you’re not very good at it.
early retirement will instead make you live longer. there are even people who die before retirement, so early retirement is a must nowdays.
Yup, we’re on the same page on this one.
Thanks for sharing, I hadn’t thought about the potential health consequences of retiring early.
As someone who is approaching FIRE but is not there yet, I typically think about what I’m going to do to keep my mind sharp in early retirement.
I had always assumed that I would physically be in great shape. Without spending 50+ hours per week at a job, I would have unlimited time to exercise. I never considered that my body may have other ideas.
If nothing else, I have 3 young kids to chase around so that will keep me off the couch!
Thanks for looking at the research and breaking it down for us. I agree the studies are not too compelling, but the question presented is interesting.
Thank you again!
Matt
You bet! Glad you found this helpful!
I opened that first paper and stopped reading after paragraph 1
“ The research reported herein was performed pursuant to a grant
from the U.S. Social Security Administration (SSA) funded as part of the Retirement Research Consortium.”
Hmm…funded by people who benefit from people working more.
Not gonna read one word of this.
Plus who cares if someone sitting around doing nothing dies a few months earlier. Maybe they are enjoying sitting around doing nothing after a lifetime of stress.
You can poke around and find evidence of anything anywhere if you want to.
Thank you for writing this blog post. I have heard this nonsense before and didn’t know where it came from.
I never looked at it from that angle. But that’s very curious that the SSA would fund this research, haha!