Thursday, June 11, 2015

The Risk of Retiring (or Being Retired) Early

Financial journalist, Mark Miller, recently asked me to provide input on a piece he is doing about retiring early. Early retirement is clearly out of reach for most American workers, but as I was working on the article with him, I realized that early retirement issues provide important insights into "normal age" retirements, say from age 65 to 70, and also into the risks faced by the roughly 50% of workers who retire early because they have to. (PDF from Employee Benefits Research Institute. See Figure 34.)

The writer suggested that his research into the topic showed that the answer, more often than not, is "don't do it." That's probably correct, "more often than not", but maybe a better way to say it is that retiring early is probably riskier than most people realize.

There are at least three major financial challenges of retiring early. First, although no healthy person knows how long he or she will live, retiring early increases the expected length of retirement and the number of years we spend in retirement is the greatest unpredictable determinant of retirement cost. (The amount we spend can be a greater factor, but unlike life expectancy and other key factors, spending is at least somewhat within our control.)

It is also a major factor in determining a sustainable withdrawal rate.

When you retire early, you add more years of retirement expenses and, as a result, you add more financial risk. You have to spend a smaller percentage of your retirement savings portfolio annually to mitigate that risk.

Second, retiring early ends savings contributions when they will typically be greatest. Third, retiring early means drawing down savings until Social Security benefits or pension income kicks in. This places additional stress on savings early in retirement that increases the risk of outliving your savings.

Let's look at longevity and sustainable withdrawal rates first.

Longevity risk is the risk that a retired household will outlive their retirement savings. The major factors contributing to longevity risk are how long retirement lasts, how much we spend each year of retirement and how aggressively we invest.

We predict how long retirement might last using life expectancy calculations at our retirement age. We can predict a median life expectancy, and we can also calculate the probability that we will live to some advanced age, such as 95 or 100. A 65-year old American male retiring today, for example, has a median life expectancy of 83 years (86 for a female) and a 6% chance of living to age 95 (13% for a female).

A retiree who decides to retire early increases the risk of a long, costly retirement. The 60-year old male in this example has a 6% chance of surviving a retirement of 30 years or more. Deciding to retire five years early at age 60, however, creates a 6% chance of a 35-year retirement and more than triples the risk of a 30-year retirement from 6% to 19%.

Retiring early has an obvious upside. A retiree who doesn’t live long after retiring will enjoy a longer retirement by doing so. The longevity risk of early retirement is that the retiree might turn what would have been a long and expensive retirement into one that is even longer and more expensive. Turning a 5-year retirement into a 10-year retirement by retiring five years early probably won't cause problems, because the retiree should have planned for an even longer retirement. But, turning a 30-year retirement into a 35-year retirement just might. Of course, the retiree has no way of knowing how long retirement is actually going to last. 

The table below shows the cost of retirement to age 95 with spending of $80,000 annually and discounted at a rate of 2% per year. Column three shows the percent increase of the cost incurred by retiring 5 years earlier than the previous row's retirement age and column 5 shows the probability that an American male retiring at that age would survive to age 95 or beyond. For example, retiring at age 65 would cost 14.7% more than retiring at age 70 and retiring at age 60 would cost 11.6% more than postponing retirement to age 65, assuming both retirees survive to age 95.

The table above shows near-worst case cost, a retirement to age 95. For comparison, the table below shows similar costs assuming the retiree lives to the median life expectancy for a U.S. male at that retirement age in 2015. For example, retiring at age 60 would cost 16.7% more than postponing retirement until age 65, assuming the person lived to his median life expectancy at the retirement age.

Notice that the costs of retirement are lower than in the previous table because expected years in retirement is smaller, but the percent increase in cost from earlier retirement is greater. The chart below shows the probability of a retiree experiencing a long and costly retirement as a function of retirement age. For example, an American male retiring at age 60 has about a 19% chance of a retirement lasting 30 years or more, but that probability for the same person retiring at age 65 is only about 6%.

Reducing the amount of annual spending that is sustainable increases the amount you need to save for retirement. The annual sustainable spending rate grows quadratically as the length of retirement becomes shorter. (All sustainable withdrawal calculations in this post use Milevsky's formula for sustainable spending without simulation (PDF) and assume a maximum 5% portfolio failure rate and a 40% equity portfolio with real mean return of 4.6% and standard deviation of 7%.)

That means that target retirement savings decay quadratically as the length of retirement becomes shorter.

An American male retiring at age 60 has a life expectancy of about 22 years and can spend about 3.6% of his retirement savings portfolio annually. Retiring at 70, he has a life expectancy of about 14 more years and can spend about 4.3% of savings annually, about 20% more. To support $50,000 annual spending from savings, the retiree at age 60 would need to save about $1.39M, while the retiree at 70 would need to save $1.163M, or 16% less.

So, those are the first couple of challenges with retiring early, whether voluntary or not. You risk adding years to a retirement that you couldn't have known in advance was going to be a long, expensive one and with that increased longevity risk, you need to save more before retirement or spend less after. Expected retirement costs more because it lasts longer, but also because it is riskier.

This is a consideration for fortunate workers who have a realistic option of retiring early, but it is a severe penalty for workers who have not adequately saved and find themselves leaving the workforce before they had planned. For workers between these two extremes, this demonstrates the value of working even a few extra years.

Next, I'll look at the second factor,  Retiring Early: Lost Savings.


  1. Excellent post Dirk. Ref our email conversation: Your bottom two graphs represent the question "how to count retirement years?" we were talking about. Tradition has us count the years up, from low to high, from left to right.

    You labeled the x-axis "Years Remaining in Retirement" counting up when actually the graph should be reversed - primarily so we get people to properly relate the depicting with age ... fewer years = older age = higher withdrawal rate.

    Technically the x-axis label and graphing makes sense in the bottom graph, although labeling it the same manner at the prior graph sinks home the point that one requires lesser resources for fewer years as well.

    I think it's important to begin helping people understand that one of the differences in perspective between the two schools of thought, Safety First and Probability Based (and I'd add Dynamic Updating in this camp), is how one represents and counts years during retirement.

    It makes it easier to visualize and transition calculations between annual reviews when the years are counted down left to right.

    My post on June 12th discusses this technical, but I think important, detail (among others).

    1. First, let me say that Larry has an interesting new post today about two schools of retirement planning relevant to his comment here, and just in general, that Larry has a wonderful website. I recommend you visit it frequently.

      Now, Larry, I should explain the economic theory supporting my choice of the direction of the x-axis on those last two charts. Trend lines in Excel, for the most part, only work correctly with scatter charts. When you build a scatter chart, Excel builds its own x-axis labels. Re-sorting the data doesn't change the chart's axis labels when you build a scatter chart. You have to change axis parameters to "reorder the values" in reverse. That fixes the curves, but causes Excel to put the y-axis on the right side of the chart, which looks really weird and is a totally Microsoft thing to do. You can fix that problem, however, by resetting the y-axis to "cross at the maximum x value". Oh, and one last thing – the y-axis label is still on the right side of the chart, all by its lonesome, even though you told Excel to move the axis itself to the left side. You'd think Excel could figure that out that the y-axis label belongs next to the y axis, but they just think a little differently up there in Redmond. Drag the axis label back over to the left and you're all done. Easy as pie! (By which I mean easy as calculating π to 120 decimal places with a pencil and paper using a Taylor series expansion.)

      Now, here's where the economic principals come into play. I had to decide whether to invest my time in modifying those charts or putting in the 15 minutes reading a novel at Caffe´ Driade over an iced latte´ and my calculations favored the latter, or should I say favored the latte´?

      However, since you make an excellent point in your comment and because I love to show off my computer skills (have I mentioned lately that I majored in computer science?), I updated the charts in this post to incorporate your excellent suggestion!

      So, you see, it's Bill Gates's fault.


    2. Hahahaha ... I'll go with that! Thanks for reposting graphs after all the effort you explained to do that (and one less page read in the novel to type all that explanation is as well)! lol

    3. We try to keep it light and entertaining here at the Retirement Cafe.

  2. My father retired early due to Parkinson's disease. I am glad he did, even though he could have kept working longer, given that the disease was mild at first. He got to spend a larger portion of his remaining years with a high *quality* of life doing things that truly made him happy. (His job did not make him happy and he was better able to manage his symptoms without the job/commuting stress in his life.) By retiring early, he was able to do things that really improved the quality of his life for a number of years. His physician marveled at how well he did for as long as he did. He and my mother were able to take some amazing trips together while he was still healthy enough to enjoy it. In the end, he wound up in a nursing home, forced to spend down his modest assets until Medicaid took over. Working a few years longer would not really have improved his life in any meaningful way I can see. My mother retired early to spend time with him. That created many treasured memories. She lives modestly on Social Security and a small pension. She was allowed to keep their home and a car and a very small amount of their joint savings.

    My late husband, on the other, enjoyed his work and planned to continue working (at least part-time) indefinitely, as do I. He died (completely unexpectedly of a heart attack in his sleep) before he even reached 60. I am glad he enjoyed his work and his life. Otherwise, I would feel even more wistful that he never had a chance to enjoy retirement.

    Life is complicated. Carpe diem! The points you raise above are worth pondering but there is so much else to consider as well. The decision about when to retire is far more than an actuarial, financial decision.

    1. I couldn't agree more! I retired early for non-financial reasons.

      I'm not suggesting when anyone should retire or shouldn't. I'm suggesting that they might want to understand the financial costs and risks before deciding to retire early, and noting that half of workers retire before they planned and weren't able to make a decision, at all.

      Thanks for writing!

  3. I retired "early" at 58 (spouse at 53) and have been using the years until I'm 70.5 to roll over large portions of 403(b) and 401(k) holdings to Roth IRAs while staying below the top of our current tax bracket (which will be higher starting at 70.5). Financially, this takes away some of the sting of retiring so relatively early.

    1. I did the same and I think it's a great tax-saving idea. In the grand scheme of the finances of retiring early, though, it isn't exactly a game changer. It takes away a small amount of the sting, though, and it is better than a poke in the eye with a sharp stick!

      Thanks for writing!

  4. Love the Savings Needed for $50k graph. can I use if I am planning on getting $30kish from SS staring at age 70?

    1. I'm not sure I understand your question. But, if you are expecting $30,000 in Social Security benefits then your are likely to have a spending "shortfall." For example, if you need to spend $50,000 a year in retirement and expect $30,000 a year in Social Security benefits, then you have a shortfall of $20,000 per year.

      You can estimate the savings needed to generate that $20,000 shortfall from investments in a risky portfolio if you assume a sustainable withdrawal rate (SWR). Savings required would be $20,000/SWR. For instance, if you believe you can safely spend 3.5% of your savings annually in retirement and need to generate $20,000 a year, you need to save $20,000/.035, or $571,428. (3.5% of $571,428 is $20,000.)

      This approximation makes many simplifying assumptions, but it will put you in the ballpark. A well designed retirement plan will improve on this estimate significantly.

      If I missed your question, please get back to me!

  5. Here's the background to my question...I'm 57, single. I have $800k in retirement savings. Want to retire at age 60. Own my debt. Need $50k/year to support lifestyle (in today's dollars). If I wait until age 70 for SS, I will get $38k/year. My life expectancy is 87.
    My question: Obviously, my savings will be well short of the $1.5M (30 years of retirement) I take from your Retirement Savings Needed for $50k graph. But, how best to factor in SS income? Thanks so much for your time & thoughts.

    1. If I correctly understand your question this time, you are referring to the fact that you will need to withdraw a larger amount from your upside portfolio in the early years of retirement before you begin receiving Social Security benefits and you are wondering how to account for that?

      I think this is a significant problem for nearly all retirement spending "rules." They calculate the amount of money you can safely spend this year assuming that you will spend the same amount every remaining year of your retirement and that is almost never the case. Not everyone claims Social Security the year they retire, spending tends to decline with age, and we often have irregular expenses that we know about and can plan for. In none of these cases will spending be constant throughout retirement, yet that is what we assume when we use SWR, ARVA, life annuities or virtually any other spending rule.

      The challenge is incorporating irregular spending into the plan. I think the best way to do this is Monte Carlo simulation with irregular spending built into the model. I do this for myself and my clients but I use a model that I built instead of spending rules. A good financial planner should be able to do this for you if you're not on speaking terms with computer models.

      The fact that you will be spending more from savings early in retirement than you will after Social Security benefits means that your probability of failure will be higher than it would be with a spending rule that assumes constant spending throughout retirement.

      If I model your example with 10 years of spending at $50K per year followed by spending of $12K per year, I find the probability that you will run out of money before age 95 to be less than 2% if you invest in a 40% equity portfolio.

      This is an oversimplified model of the problem and doesn't consider life expectancy. I offer it as a demonstration only. Be careful using your life expectancy in the calculations unless you use it as a random variable. If you simply use 87 as the age retirement will end with certainty, you have a 50% chance of underestimating.

      Sorry I can't provide a simple solution to the problem, but it really needs to be analyzed by a competent retirement planner. That's why, as they say, they get the big bucks.

      Did I get your question right this time?


  6. Yes, you got it! I'm headed to "monte carlo"...I'll play with the numbers. In my gut, I know I should consult/work part-time and see if I can preserve (not tap into) my retirement funds extensively in years 1-3 of retirement. Don't want a long life to be a "bad thing" do we? Thanks again.

    1. Have a nice trip! And once you get there, consider simulations that include your life expectancy. Assuming fixed lengths of retirement like 20 or 30 years causes statistical problems. Considering worst case is wise, but can distort the probabilities and provide an awfully conservative projection.

      Thanks for writing, and for getting back to me when I missed your question the first time.