Retirement Crisis In America?
Summary
- “Americans cannot afford retirement” is a very common assertion in discussions of US household financial health. That belief is strengthened when one compares US household savings to advice offered by investment firms through their retirement calculators and rules of thumb. If this contention is correct, one would expect to see increases in poverty rates and significant declines in income, expenditures, and assets with increasing age as retirees deplete their inadequate resources.
- In fact one does not see any of the outcomes that would be expected if Americans have dramatically underfunded their retirements. The Federal Reserve’s consumer finance data suggests that household income and assets are stable with increasing age. Department of Labor consumer expenditure data broken down by income category on a per capita basis also indicates no compelling evidence that retirees reduce their expenditures with age. Thus households in aggregate, appear to be able to accurately anticipate their financial needs in retirement.
- The retirement calculators and rules of thumb are not wrong, they just apply to very few people (those in the highest net worth category). They do not accurately depict how the vast majority of households actually retire. A significant portion of retirees younger than 75 have adjusted to their increased life expectancy by continuing to work in retirement. Most importantly, the majority of American households fund their retirement primarily through Social Security and pensions.
Retirement Calculators
Retirement calculators and retirement advice are readily available on the internet. They typically assume one wants an idealized 1960’s-style retirement in which one stops working at age 65 and relies on Social Security, pension income, and portfolio returns to replace around 80% of pre-retirement earnings. As an alternative to calculators, some firms offer retirement nest egg “rules of thumb.” Fidelity advises one should have saved 8 times their salary by the time they are 60, and 10 times their salary by the time they are 67. If you are trying to replace your income, how much you earn is a big determinant of how much you need to save because Social Security benefits are are likely to represent a large percentage of low incomes, but a low percentage of high incomes. Page 15 of J. P. Morgan’s 2016 guide to retirement contains a table (“JPMT”) that gives a rule of thumb like Fidelity’s, but adjusts it based on the income one is trying to replace.
How Many Households Pass The J. P. Morgan Test?
The Federal Reserve does a survey of consumer finances (SCF) every three years that gathers statistics on household income and assets. The most recent survey results are available for 2013. This data allows one to estimate the percentage of US households that met the JPMT rule of thumb for savings in 2013 (“the pass rate”).
My analysis uses a narrow definition of the income one wants to replace in retirement to be wage income only. My analysis also takes a very expansive view of saved assets to include all financial assets plus the value of all business interests (including farm businesses). The principal exclusion in my savings number is home equity on the premise that retirees need to live somewhere, so this component of net worth is not available for income generation.
Some studies of retirement adequacy take a narrow view of saved assets to only include assets in dedicated retirement accounts like IRAs, 401Ks, etc. Indeed for younger households, some of the “savings” I am counting may not be available for retirement (e.g., they are earmarked to pay for childrens’ college tuitions). I am deliberately using the most generous definitions of savings and income which will boost the pass rate I report.
The JPMT only covered 44% of American households in 2013 because of it’s age and income ranges. Thirty-four percent of the population is excluded due to age (12% < 30 years old; 22% > 65 years old), another 20% is excluded for having household incomes less than $30,000 and 1% is excluded for having income above $300,000. This subpopulation might be expected to have a higher pass rate than the overall population as it contains more married households (75% vs. 57%) and more households with college degrees (50% vs. 39%)
Nevertheless, the overall percentage of the JPMT population that had “appropriate” savings was only 19%. That low result is consistent with the tone of articles that suggest Americans cannot afford to retire.
There was no significant variation of the pass rate with age. The overall pass rate was slightly higher for married households (23% for married versus 18% for non-married). Not surprisingly, given the correlation between education and income, pass rates also increased with education. Still the college-educated pass rate is only 25% (which is similar to the pass rate of the highest normal income decile shown below).
The highest income decile (disproportionately college educated) shows a higher pass rate of 27%, but there is no other evident correlation of pass rate with income[1]. The general evenness of pass rates reflects the fact that the savings target is a multiple of income -- so it’s difficulty versus income is relatively constant.
Implications of Failure
If households are increasingly entering retirement with insufficient resources, one would expect a commensurate increase in the poverty rate among 65 and older aged households. The opposite has occurred. The poverty rate of all 65 and older households fell from the highest rate among all age groups at 35% in 1959, to the lowest rate in 2013 at 10%. Meanwhile, the rate for the larger 18 to 64 year group has remained roughly flat.
In fact since the late 1960s, median real incomes for 65+ year old households have grown the most of any age group.
Finally, if we look cross sectionally within just the 2013 SCF data, Table 1 below shows that the primary reason there is a drop in average per capita income with age is that fewer people work past the age of 75.[2] Table 1 compares per capita income. This is important because the commonly reported larger drop in household income with age (see Table 1 on page 9 of the SCF report) is primarily due to decreasing household size due to death.[3] Table 1 appears to show a significant drop in per capita assets, seemingly consistent with the idea that households need to draw down their assets to live. Here again, however, it is likely that death is distorting the statistics.
Table 1: Average Per Capita Change in Income and Assets With Age (2013 Survey Of Consumer Finances)
Age Group
| |||
65 to 74 Yrs
|
75 Yrs +
|
Change
| |
Income
|
$51,962
|
$32,719
|
($19,243)
|
Wages
|
$14,614
|
$1,839
|
($12,775)
|
Social Security
|
$10,911
|
$11,957
|
$1,046
|
Pension
|
$10,398
|
$9,407
|
($991)
|
Transfer/Other
|
$1,779
|
$452
|
($1,327)
|
Interest + Dividends
|
$3,878
|
$3,318
|
($560)
|
Capital Gains
|
$3,236
|
$2,266
|
($970)
|
Business/Farm Income
|
$6,562
|
$3,137
|
($3,425)
|
Assets
|
$596,241
|
$413,673
|
($182,568)
|
Financial
|
$292,210
|
$190,383
|
($101,827)
|
Debt
|
$38,604
|
$13,781
|
($24,823)
|
Networth
|
$557,637
|
$399,892
|
($157,745)
|
The averages in Table 1 are significantly influenced by very high net worth families that materially bring up the average asset statistics. A significant percentage of high net worth families are married. The death of a spouse can cause assets to be transferred and thus decrease the percentage of high net worth families in the 75+ age group. Families in the top 10% of net worth make up 20% of the 65 to 74 year old category, but only 10% of the 75+ year category. Table 2 shows that if you exclude this category, total assets and financial assets are relatively stable (and the pattern of income drop due solely to decreased labor force participation is the same as in Table 1).
Table 2: Average Per Capita Change in Income and Assets With Age Excluding Highest Net Worth Category Households (2013 Survey Of Consumer Finances)
Age Group
| |||
65 to 74 Yrs
|
75 Yrs +
|
Change
| |
Income
|
$28,602
|
$22,512
|
($6,090)
|
Wages
|
$7,857
|
$1,538
|
($6,319)
|
Social Security
|
$10,545
|
$11,531
|
$986
|
Pension
|
$7,576
|
$6,919
|
($657)
|
Transfer/Other
|
$901
|
$408
|
($493)
|
Interest + Dividends
|
$218
|
$717
|
$499
|
Capital Gains
|
$85
|
$433
|
$348
|
Business/Farm Income
|
$1,089
|
$440
|
($649)
|
Assets
|
$172,751
|
$159,945
|
($12,806)
|
Financial
|
$59,889
|
$57,662
|
($2,227)
|
Debt
|
$28,521
|
$11,861
|
($16,660)
|
Networth
|
$144,230
|
$148,084
|
$3,854
|
Do Retirees Spend Less As They Age?
The tables above demonstrate that household assets and income are stable with increasing age. Do retirees have to spend less to get by after they quit working? Unfortunately the SCF does not also ask people about their expenses, making it difficult to tie together the complete income, expense and asset picture. However the US Department of Labor does a survey that looks at consumer incomes and expenditures (the Consumer Expenditure Survey; “CES”). J. P. Morgan’s 2016 retirement guide includes a chart based on that data shown below that suggests expenditures decline a significant 24% from the 65-74 year old age group to the 75+ age group.
The changes in the composition of expenses seem to make sense. Expenses related to travel, transportation, and dining out decrease, presumably as health and mobility decline. As many older homeowners have paid off their mortgages, mortgage expense also declines. Interestingly, charitable contributions seem to increase with age -- a sign that households are not under financial stress.
A Bureau of Labor article in 2015 that looked at expenses relative to income by age group also seems to reach the same conclusion suggested by the J. P. Morgan chart. Their chart on page 2 of that article implies that the 75+ year old age group sees significant declines in income and expenditures -- and spends all of their income.
However, when the CES micro data is broken down by income group, the apparent decline in spending is seen to be nearly all due to changes in household composition. Table 3 shows that expenditures by income category are quite stable (as is the ratio of expenditure to income). What has changed with age is the significant decline in the percentage of top income decile households in the total population which brings down the weighted averages down materially. Top decile income households are typically married with 2.2 people per household. Lower income decile households are less likely to be married and approach one person per household in the lowest income deciles. Thus the compositional changes seen from 65-74 year olds to 75+ year olds are likely due to the death of high income spouses that cause two person households to become one person households with diminished survivor benefits.
Table 3: 2105 Consumer Expenditure Survey: Household Averages by Income Category
65 to 74 Year Olds
|
75 Years and Older
|
% Change
| |||||
Income Category
|
Expenses
|
Income
|
% of Population
|
Expenses
|
Income
|
% of Population
|
In Expense
|
< $5000
|
$23,926
|
$1,867
|
3.0%
|
$21,587
|
$2,516
|
3.7%
|
-9.8%
|
$5000 - $9,999
|
$24,411
|
$8,066
|
3.0%
|
$19,477
|
$8,143
|
4.1%
|
-20.2%
|
$10,000 - $14,999
|
$22,365
|
$12,825
|
9.0%
|
$20,597
|
$12,833
|
14.5%
|
-7.9%
|
$15,000 - $19,999
|
$27,034
|
$17,390
|
9.4%
|
$24,986
|
$17,403
|
16.8%
|
-7.6%
|
$20,000 - $29,999
|
$33,058
|
$24,924
|
17.0%
|
$32,778
|
$24,982
|
20.8%
|
-0.8%
|
$30,000 - $39,999
|
$39,234
|
$34,832
|
13.1%
|
$35,943
|
$35,048
|
13.2%
|
-8.4%
|
$40,000 - $49,999
|
$40,573
|
$45,230
|
10.2%
|
$43,841
|
$45,132
|
8.0%
|
8.1%
|
$50,000 - $69,999
|
$48,649
|
$60,756
|
13.2%
|
$50,163
|
$65,752
|
8.5%
|
3.1%
|
>= $70,000
|
$84,628
|
$132,733
|
22.0%
|
$77,303
|
$115,210
|
10.4%
|
-8.7%
|
Weighted Average
|
$45,990
|
$53,780
|
$36,161
|
$36,228
|
-21.4%
| ||
We can partially correct for this effect by looking at per capita numbers (dividing household income and expenses by the number of people in the household). Table 4 shows that doing this nearly eliminates the apparent drop in “weighted average” expenditures. The dollar change is small relative to the estimation error.
Table 4: 2105 Consumer Expenditure Survey: Per Capita Averages by Income Category
65 to 74 Year Olds
|
75 Years and Older
|
% Change
| |||||
Income Category
|
Expenses
|
Income
|
% of Population
|
Expenses
|
Income
|
% of Population
|
In Expense
|
< $5000
|
$18,112
|
$1,799
|
3.0%
|
$17,169
|
$2,045
|
3.7%
|
-5.2%
|
$5000 - $9,999
|
$16,891
|
$6,726
|
3.0%
|
$18,492
|
$7,777
|
4.1%
|
9.5%
|
$10,000 - $14,999
|
$18,043
|
$11,341
|
9.0%
|
$18,048
|
$11,879
|
14.5%
|
0.0%
|
$15,000 - $19,999
|
$21,216
|
$14,500
|
9.4%
|
$21,907
|
$15,766
|
16.8%
|
3.3%
|
$20,000 - $29,999
|
$24,009
|
$18,424
|
17.0%
|
$23,945
|
$18,167
|
20.8%
|
-0.3%
|
$30,000 - $39,999
|
$23,341
|
$21,142
|
13.1%
|
$21,270
|
$21,139
|
13.2%
|
-8.9%
|
$40,000 - $49,999
|
$23,826
|
$27,200
|
10.2%
|
$24,931
|
$26,764
|
8.0%
|
4.6%
|
$50,000 - $69,999
|
$28,070
|
$35,948
|
13.2%
|
$27,203
|
$35,914
|
8.5%
|
-3.1%
|
>= $70,000
|
$41,810
|
$66,004
|
22.0%
|
$41,274
|
$60,769
|
10.4%
|
-1.3%
|
Weighted Average
|
$27,169
|
$30,614
|
$24,083
|
$22,856
|
-11.4%
| ||
This is another cautionary tale in the pitfalls of household statistics, especially when combined with compositional changes due to mortality. In summary, the evidence is not compelling that retirees have to scrimp to get by.
How do people actually retire?
What explains the remarkable financial health and stability of retiree households shown in the tables above? It is not investment income. Tables 1 and 2 show that households do not rely heavily on investment income in retirement. Interest, dividends, capital gains, and business income only made up about 26% of over-65 household income. For the roughly 85% of those households not in the highest net worth category, investment income only made up about 6% of household income. In fact investment income was only significant for the approximately 15% of households that were in the top decile of net worth households. This top decile has a median net worth of $2 million and an average net worth of $4.3 million (the average is pulled up by ultra wealthy households). For this small group, investment income accounted for 43% of income of 65 to 75 year olds and 58% of income for 75+ year olds. The vast majority of households who are not in the top net worth category, however, relied on work, Social Security, and pension income -- not their investment portfolios.
Work:
Increasing the retirement age is one of the ideas often suggested to improve the solvency of the Social Security system. By working longer, households on their own appear to have adapted their “retirement age” to their increased life expectancy. If you look at just “retired” households that received social security and/or pension income in 2013, more than half of those aged 55 to 64 were working in 2013. That “working-retired” rate fell modestly to 40% for households in the 65 to 74 year old group, and was still 14% percent in the 75 to 85 year old group. As shown in the tables above, wages made up about 28% of household income for 65 to 75 year old households and 6% of income for households older than 75 (this is true whether or not you exclude the highest net worth category). So work is a substantial source of income for retired households less than 75 years of age.
Social Security and Pension Income:
Social Security and pension income are the two most important reasons for the strong financial health of most American retirees. The inflation indexed Social Security benefit, plus other pension income, accounted for 41% of total household income for 65 to 74 year olds. For the 80% of those households not in the top net worth decile, it accounted for 63% of household income.
Households over 75 years of age -- who receive less wage income -- relied on Social Security and pensions for 65% of their income. For the 90% of those households not in the top net worth decile, these sources of income made up 82% of their income (Social Security on its own was 51%).
Social Security is a fascinating program that has changed dramatically since it was introduced in 1935. It is the focus of much debate, with people asking whether it should be privatized, if it is solvent, if it is efficient and so on. Along with Medicare, it is one of the largest government expenditures and one of the mechanisms the United States uses to redistribute income from wealthy households to poorer households. Whatever one’s views, it is an entrenched feature of American life. It has become the core of most retirees’ income and it has undeniably accomplished Franklin Roosevelt’s goal of eliminating high rates of poverty among the elderly. Because of it, there is no retirement crisis in America.
Transparent and reproducible: All of the labeled Figures and Tables can be generated using the free, publicly-available R program and the R code available on github[4] to analyze the publicly available data obtainable from the links in the article.
[1] Because the JPMT cuts off households with income below $30,000, the pass rate and income estimates are not meaningful for normal income deciles below 5.
[2] From this point forward, we are no longer looking at just the JPMT subpopulation -- we are looking at the entire data set.
[3] The percentage of households with spouses drops from 56% for 65 to 74 year olds, to 29% for those older than 85. The death of high income spouses can particularly skew household average results.
[4] The code used in this article -- “retireeEssay.r” and “ConsumerExpenditures.R” -- owes a huge debt to Anthony Damico, as I merely modified the excellent R files he has made available to analyze the Survey of Consumer Finance and the Consumer Expenditure data.
Comments