Evidence is mounting that retirees spend less as they age, on average. The theory is that infirmity and uncertainty about the future erode people’s ability and desire to spend money on non-essentials.
Instead they accumulate savings. That’s even though most retirees do not experience a spike in healthcare costs towards the end of their lives.
The very first paper on the topic was published in 1998. It used British data but the pattern has been found in multiple countries in the decades since.
I have a question
We explored the evidence for the real-terms retirement spending decline in the UK and what might drive it in last week’s post.
My outstanding questions were:
- Could the evidence be wrong?
- Could the established pattern reverse?
- Even if retirees’ real spending declines on average, can individuals make practical use of this information?
The Precautionary Principle suggests we should plan to do worse than typical when it comes to our own health. Because how do we know we won’t be whacked by huge long-term care bills in the future?
Let’s make like rational sceptics and examine the case against doing anything at all.
Why retirement spending research may not apply to you
I illustrated the evidence for spending less as you age using the UK paper Understanding retirement journeys: expectations vs reality by Cesira Urzi Brancati, Brian Beach, Ben Franklin, and Matthew Jones.
The authors say about the consumption decline in later life:
We should stress that this is an average, which, by definition, will mean that some individuals will not experience stable patterns of expenditure on essential items.
The overall decline relies on essential consumption remaining flat (-ish) while non-essential spending falls.
But not everyone can control their spending. The report reveals that nearly a quarter of 60-year-olds and older spend more than they earn.
Even one in seven of the 80-plus age group spends more than their income.
You could argue the overspend odds must be lower for the money mavens who read Monevator. Our readership is dominated by financially literate types who plan for the future.
But like a budget version of the Anthropomorphic Principle, your part of the universe is likely to have special properties simply because you’re interested in the topic of retirement spending.
The academics must measure broader patterns.
It could be you
Brancati’s data can’t tell you anything about whether you’ll draw unlucky numbers in the UK’s long-term care lottery yourself.
And by her own admission, the data is understandably limited in other ways:
At this point it should be noted that our data is restricted to households only and therefore excludes those actively living in care homes who may be paying for it from their remaining assets.
Still, the percentage of households paying towards long-term care looks surprisingly low:
Only 6.4% of households from the 80-plus age group bore the burden of long-term care costs.
We can assume this data captures some people who live in care homes but don’t represent their entire household. Similarly, the percentages must miss some households that exist entirely in care homes.
Brancati believes the lack of exploding healthcare expenditures in her data means a U-shaped retirement consumption curve is atypical.
Obviously that’s not the same as saying it won’t happen to some individuals.
U-shaped consumption The U-shaped spending curve was posited in US research by David Blanchett. He found that real-terms spending headed down post-retirement but rebounded towards the end as health costs mount. (The upward leg of the U-consumption curve). But Blanchett concluded the net effect is still a decline in overall retirement spending.
Different strokes for different folks
Brancati’s data is a snapshot of UK spending and lifestyle factors from 2003 to 2013.
Your eighties may be three years away or 50. Will an aging population and shrinking workforce increase healthcare costs for the retirees of the future?
It seems a fair assumption.
And that leads us to another reason to be cautious about this research.
Retirement spending studies usually compare different age groups at a moment in time.
For example, the consumption curve in 2013 compares that year’s 65-year olds versus 75-year-olds versus 85-year-olds.
The data doesn’t typically track how the same cohort of 75-year-olds in 2013 spent as 65-year-olds in 2003. Nor how they’ll spend as 85-year-olds in 2023.
That’s important, because each cohort is subject to a unique lifetime of socio-economic pressures.
The habits – and savings accounts – of today’s 85-year-olds are shaped by their prior experience. Those influences are materially different from previous generations at the same age.
For example, a cohort that experiences high healthcare inflation could well face higher costs in that area than previous cohorts at similar ages.
The generation game
Each generation faces different financial conditions, which can alter their retirement path:
- UK pensioners previously suffered high rates of poverty. This began to be corrected by government policy after the year 2000.
- Baby Boomers benefited from the advent of defined benefit pensions and the Triple Lock. But increasing lifespans and diminishing dependency ratios leave them more exposed to healthcare costs.
- Gen X and Millennials bear the scars of the Financial Crisis. They were left to work out defined contribution pensions for themselves.
- Gen Z and Millennials benefit from default pension savings, but are up against daunting housing prices.
Who knows how this shakes out? But fluctuations in fortune could mean a majority will not always experience falling retirement consumption.
Moreover, retirement spending researchers valiantly re-purpose data designed to answer different questions from the ones they’re asking.
And that leads to inconsistent outcomes across the literature (although the overall trend is still convincing).
For example, Brancati finds that 77% of older households save something.
In contrast, according to retirement professor Wade Pfau’s analysis of a US paper1 only 39% of retirees see falling consumption.
Applying the retirement spending research
It is clear that retirees save to hedge against an uncertain future. And Brancati gives good advice when she endorses that behaviour as rational:
…people may live longer than they expect, investments may generate exceptionally low (or high) returns, and prices on certain goods and services may unexpectedly rise.
For this reason, prudent individuals may set aside a certain amount of wealth to avoid running out of cash in case of unexpected expenditures.
Brancati makes further suggestions for retirees likely to spend less as they age:
For those with small DC [Defined Contribution] pots and little else, they might best use their DC savings to give them some flexibility early on in retirement and rely on pensioner benefits to meet essential spending needs later on.
Others, who have sizeable final salary pensions and small DC pensions may also think it best to exhaust their DC savings relatively early on before relying on their final salary pension plus the State Pension to meet their spending needs in later life.
It may make sense for financial products and services to facilitate relatively high initial income before guaranteeing a base level of income in later life as people reduce expenditure on non-essential items but maintain spending on essential every-day items.
Spend now, pay later
One way to simulate this last idea is to use a slightly higher sustainable withdrawal rate (SWR) at the outset. You could then make the most of any spending decline by buying an annuity very late in life with your evaporating resources.
A level annuity bought in your late seventies or early eighties could hit the sweet spot between fading life expectancy and below-inflation consumption.
Meanwhile Pfau advocates puzzling out your potential exposure to the trend but then treading warily:
…those who plan for greater expenses related to travel, housing, or health care should expect their spending needs to keep pace with inflation.
Spending may decline, so I would not fault anyone for using assumptions of gradual real spending declines such as 10% or even 20% over the retirement period.
But pending further research developments, I would avoid moving too far in the reduced spending direction as a baseline assumption.
Sustainable withdrawal rate bonus?
Consuming less than you earn in retirement leads to a higher SWR than if you assume constant inflation-adjusted spending.
The US sultans of SWR, Wade Pfau and Michael Kitces, have published pieces estimating the SWR bonus you may collect if your spending declines.
Theoretically, you can spend more at the start of retirement because your consumption will tail off later. That offsets the risk of running out of money.
Caveat 1
The SWR estimates vary wildly. They make different assumptions about the rate of decline, asset allocations, and more. They all use US returns and assume you don’t retire until age 65.
Caveat 2
FIRE-ees beware that SWR bonuses attenuate due to your longer retirement. The retirement spending decline doesn’t really pick up speed until age 70 to 74. There’s plenty of time for you to sink your portfolio before then, especially given a bad sequence of returns.
Still, for the brave or foolhardy:
- Pfau’s 0.7% SWR bonus
- Kitces’ 0.32% to 0.75% SWR bonus
Save yourself
I must admit I ignored the ‘declining retirement spend’ evidence during my own FIRE planning. Neither did I make any special provision for long-term care bills – other than owning a house.
I think the research has important insights on our likely path through retirement. But I also believe it’s too risky to apply on an individual basis.
Retirement planning is dicey enough without trying to optimise yet another personal unknown.
But I do feel more secure about our finances – and better informed about the choices we face – now I’ve taken a deeper dive into the topic.
I prefer to think of falling retirement consumption as another reason why my chosen SWR is probably okay.
Take it steady,
The Accumulator
- Yes, he really is a Professor of Retirement! [↩]
I have a DB pension from the first 12 years of my career worth about £7500/annum at 60 and have been investing in a SIPP since my mid-30’s.
My plan had always been to rely on the DB and state pension to provide a base level of income and spend my SIPP down over 15 years from 60 to 75(ish).
At 41, I thought I was future-proofed until I read this! Assuming I want a good level of care, dignity and quality of life in my last few years, does anyone have a guesstimate how much this would cost?
Thanks as always for the post.
Steve
> At this point it should be noted that our data is restricted to households only
Excluding people in care homes is a serious omission IMO, given the serious costs of care – easily running to 4k p.c.m. and more. There seems to be a 1:4 risk of ending up in a care home by the time you are 85. Those odds aren’t awful, but they are significant.
Against this by the time you get that old, if you have assets you can get some functional insurance using long term care annuities which are tax-free if paid to a CQC approved facility. It’s notable that some 70% of people in care homes have dementia which surprises me, it seems the mind fails in service before the body. So anything you can do to keep the noodle turning over as you get older seems to be A Good Thing, though dementia seems to have some genetic component and some is the luck of the draw. It also seems to be increasing – as a child I never heard of people going down with dementia, whereas I know some people who have relatives that have/had this, though thankfully none of mine. This would make one wonder if there are lifestyle issues in more recent times. Although people did get clobbered with heart attacks in my childhood than now a decent number got into their eighties and nineties, so it’s not like everybody died before they got dementia back in the day.
Forecasting anything beyond the immediate short term (e.g. end of the current year) is hard enough, so IMO it is prudent to view any longer term projection with some scepticism. Having said that, some idea of the likely direction of travel is useful. And provided you review such forecasts regularly you afford yourself the possibility to stay on top of things as they inevitably evolve.
The Blanchett study you refer to coined the term “spending smile”. Please beware, this term does not refer to graphs of spending versus time but rather to graphs of the rate of change of spending versus time. The point is subtle but important. The late great Dirk Cotton gives a good explanation of this point at:
http://www.theretirementcafe.com/2015/04/retirement-spending-assumptions-and-net.html
IMO, the same DC post adds some useful points to your current debate – including (albeit from a US sample): the likelihood of spending increasing, staying the same, or decreasing.
FWIW, my own take is that for the majority of cases a flat consumption assumption will be safe. In my own calculations I factored in a very small decrease over time for my most likely case. To date, this has been appropriate, but things – of course – can change.
@TA Regarding your assumptions of what accounts for the bulk of over-80 care costs shown in Table 2, I would imagine only a very small proportion is represented by care home costs where one partner remains at home. I suggest the bulk is likely to be accounted for by the costs of visiting carers for single/widowed elderly people in their own home. Provision of care in the home is heavily means-tested, so most people with assets (excluding the value of the house) or moderate income will be bearing most if not all the costs.
The difference in outcomes between US and UK research on healthcare costs in later life is likely more because of the difference in costs of general healthcare (no NHS in US) than for personal care.
@ermine (#2):
Re: “Excluding people in care homes is a serious omission …” these people were not really excluded but rather are out of scope of the Living Costs and Food (LCF) survey, which is solely focussed on households. This is an example of what TA describes as ” Moreover, retirement spending researchers valiantly re-purpose data designed to answer different questions from the ones they’re asking.”
IIRC, the LCF has evolved from a survey originally set up in 1941 to monitor nutrition in London and is, amongst many other things, the key source of dietary / menu information for social history TV programmes such as back in time for …., (see e.g. https://en.wikipedia.org/wiki/Back_in_Time_for…)
In due course, sufficient ELSA (English Longitudinal Study of Aging) data should become available but it is not clear to me [from the ELSA website] whether this Study is also restricted to households.
What we all really want to know is our own future household spending – and apart from things close enough to plan for we can only assess it from averages such as here. In our case we retired knowing that we would have a few years of high expenditure, due to our daughter going to university and the need to contribute to her maintenance. (Though we had also anticipated more spending on overseas trips than has been possible with Covid). We had money held as cash to cover that, on top of our plans for the medium to long term.
@Ermine, I think the high proportion with dementia in care homes probably relates to the fact it is much easier to provide care in someone’s own home for physical problems. No one wants to put their relatives into a care home unless there is no reasonable alternative.
So the general idea is that you spend extravagantly on experiences (holidays, travel, projects, etc) while you are still fit or healthy enough to enjoy them.
Once your mobility has gone, say hello to daytime telly and daytime drinking.
Presumably all while holding back the £5k for Dignitas.
I don’t worry about care home expenses. There are too many other things I need to worry about before I possibly get to that point. Also in my experience it is not that big a concern to the individual. The relatives of mine who have ended up in care have done so due to mental decline such as dementia or damage caused by extreme alcoholism. The best care I have seen was for a close relative who suffered brain damage from alcoholism. She was a very low earner and ended up a ward of social services in a run down town in the UK before she ended up in a home. The home she ended up in was grim but small and the staff caring. On the other hand my very wealthy uncle ended up in a very expensive home that was an impersonal hell hole in comparison. My advice would be to die suddenly of a heart attack while still lively and full of beans like my father did. And save your money for living life to the full while you can. Once you are so far gone that you end up in a home you are someone else’s problem and most likely past caring.
The care home point is well made, but IIRC the average person spends around 18 months in one before…ahem….they no longer need it. Even at 4K/month, that sounds doable on the average highly monevated FIRErs asset base – providing you aren’t trying to maximise an inheritance.
As I posted on the previous thread, I’m more interested in having a fund so that if some miracle health intervention appears, I want to be able to afford it.
@ Steve – I’m working on a long-term social care series at the mo – I’ve discovered it is possible to come up with a ballpark, rule-of-thumb number for costs. It’ll be a few weeks before that post arrives but I think it will help
@ Ermine – thank you for those links. That 28% figure is bracing. I guess that’s one I’d really want to drill into to find out what percentage are receiving extensive care. I found some good ONS data that suggested around 10% of the 85+ population were in care homes in England and Wales. But that was based on the 2011 census and the female percentage was much higher than the male. Perhaps unsurprisingly.
I think you’re right about contributing lifestyle factors. I came across an encouraging report recently that dementia figures weren’t getting worse though even as the population ages. Possibly because the lifestyle message was spreading and people were taking evasive action.
@ Al Cam – that Dirk Cotton post provides very useful perspective – thank you for sharing.
@ David V – I wasn’t trying to say that the bulk of the table was accounted for by households bearing care home costs. I was just making a side-point about the care home costs it could be capturing. Rereading it now, I think this is a paragraph that I could have left on the cutting room floor 🙂
Interestingly, the general trend on spending matches between the US and UK and across the developed world as best I could tell from a brief overview of the literature.
@ Bill – great summary. If slightly dark 😉
@ G – agree, generally it looks doable on an average basis. I guess I’m worried about being above average in this case. Your comment about the miracle health invention made me smile. I guess we can all spend the money on care homes if it doesn’t get invented in time 🙂
I got deeply into the financing of care home costs when my late mother went into a care home. She was mentally sound, so not running out of money was a big concern to her. We bought a care needs annuity, deferred for two years, which halved the cost, from part of the proceeds of her house sale. As it happened she only survived for a further eight months, but that was not foreseeable at the outset. Even though the annuity proved in the event to be poor value for money as it paid nothing, I do not at all begrudge the reduction in my inheritance as it gave my mother and me the peace of mind that she could stay in the care home of her choice for the rest of her days, no matter how long she lived.
The two year annuity deferment mitigates costs in the event of early death and is roughly cost neutral compared to an immediate payment annuity if you survive the two years. In my mother’s case, taking out the annuity at age 92, break-even for either immediate or deferred would have occurred at about 3.5 years. They are supposed to be individually medically underwritten, but the 3.5 years was what the adviser suggested was typical break-even at her age before we had submitted any medical details. (You have to use a SOLLA adviser to buy such an annuity.)
The adviser seemed to agree that two-year deferment was a good choice although only a small percentage (5% I think she said) chose this in preference to the immediate payment annuity. This struck me as irrational, as a cost analysis clearly showed two-year deferment to be optimal with considerable cost mitigation for early death and miniscule extra cost for longer survival.
@TA (#10):
Dirk did quite a few posts on this subject over the years.
In your previous post on this topic Mr Optimistic posted a useful query at comment #15. From this query, the DC post of Wednesday, September 12, 2018 IMO ultimately brings all his thoughts together.
@ DavidV – Thank you for that insight. The important thing is you gave your mum peace of mind at what must have been a very stressful time for you both.
As I’ve researched the social care system I’ve found it to be absolutely byzantine. I’m only looking into the financial side too. It’s not like I have the pressure of actually trying to secure adequate care for a loved one.
I haven’t properly looked into immediate or deferred payment annuities yet. Are there value protections for the deferred payment variety? It looks like there are for immediates. Of course, they may not be worth it in terms of cost.
@DavidV
Thank you for your comment. I’d never heard of a care needs annuity and it could be perfect for my MIL.
@TA (13) I’ve looked back into the adviser’s report and it seems that for the quotes I received in July 2018 both the immediate and deferred annuities included a money-back guarantee for death within the first month. The immediate annuity quote also included optional limited value protection – 50% in months 2 and 3 and 25% in months four, five and six. There is mention of 50% capital protection for a considerable increase in initial outlay but it was not completely clear how long this protection lasted – I think it was for the same six months.
The report suggested that the deferred annuity was a more cost-effective method of protecting capital in the event of early death, and my own analysis also concluded that this was the case. It effectively provides some level of decreasing protection for the full two years as this is the time it took for the initial lower premium and cumulative self-pay element to match the initial premium of the immediate annuity. As my mother died after eight months in the care home and during month seven of the deferred annuity term, I am certainly glad this is the option we went for.
@Dazzle (14) If you want to pursue this, search online for Society of Later Life Advisers (SOLLA). I’m not sure I am allowed to mention here the actual adviser firm I used. Unfortunately it is not possible to buy these annuities direct or through execution-only intermediaries. As ermine mentions in (2) the annuity payments are tax-free provided they are made directly to the care home.
Besides the focus on households, which is a severe distortion of the metrics, since many people are single/widow in retirement and their per-capita spend is much higher… the studies referred to and commented on this and the previous article on this matter suffer from the same grave defect, which is to look at %spend vs the single variable of age, as if age was the main factor.
I would suggest that, at a very minimum, and just intuitively, you need at least 3 variables: age, single vs couple, and an indication of wealth level. Someone with a net worth of 50K will most definitely not spend the same % as someone with a net worth of 500K or 2M, since the basic unavoidable costs of food and bills will be a much high % in the former case. As well, a high net worth would likely be associated to owning your own home, vs having to pay rent in retirement, which would eat up a huge chunk of your pension.
I am a big fan of simplification and using averages as a proxy for real life facts. However, as any statistician will tell you, for the average to be a useful approximation you require to assume that the underlying variables are “normal” (think the typical bell curve) or at least comparable. When you’re averaging two or three very different use cases as we are looking at here, the average is just wrong. You can search for “bimodal distribution” and see why the use of mean or median is discouraged. Think of a distribution formed of two bell curves, one peaking at 10 and another peaking at 20, the mean will be 15, where in fact few samples are at 15…
For this very reason alone, I don’t think that these studies, or any of the conclusions reached by them, have any credibility whatsoever, and sadly I am still none the wiser with regards to what the hard facts with regards to the question posited by the article (would spend increase or decrease over age for a typical member of my socioeconomic cohort).
@ Endfire – “the focus on households, which is a severe distortion of the metrics”
It’s not a distortion of the metrics. The metrics focus on households. As in, the researchers are working with the data available, not trying to shape the data to get a result.
Households will include singletons but you’re right it’d be good to see a study that showed how that group fared.
The Retirement Living Standards in the UK research identifies that a single person requires about 68% of the income of a couple to enjoy the same lifestyle.
But counterintuitively, some papers identify singlehood as a reason why spending declines later in life.
Differences in wealth levels is covered by the literature – particularly in the Brancati paper. See the consumer segments mentioned in this post:
https://monevator.com/spending-decline-in-retirement/
The link shared by Al Cam is also good on this aspect:
http://www.theretirementcafe.com/2015/04/retirement-spending-assumptions-and-net.html
@endfire
AFAICT, the data you refer to is drawn directly from six LCF survey’s and adjusted for inflation – see comment #5 above.
Economies of scale are real – hence the focus on the household. Furthermore, the household is the global unit for comparison in e.g OECD data.
Equivalence scales are a useful concept to get to grips with.
If you want to know more technical details about the LCF survey, including household makeup, 95% confidence limits, progress on improvement actions, etc see:
https://www.ons.gov.uk/peoplepopulationandcommunity/personalandhouseholdfinances/expenditure/datasets/livingcostsandfoodsurveytechnicalreportdatatables
Each years technical report is prefaced with a handy index to the various tabs in the rest of the spreadsheet.