Surveys of retirement spending

 data describing typical retiree spending patterns are an important resource for the retirement planning process. Careful analysis of such data provides important guidance for individuals, financial planners, economists, as well as for developers of retirement planning software.

The two major sources of retirement spending data for U.S. retirees come from the annual U.S. Bureau of Labor Statistics' Consumer Expenditure Survey and from the biennial Health and Retirement Study survey conducted by the Survey Research Center at the University of Michigan. Both of these survey datasets allow estimates of retirement spending for the median U.S. retiree as well as for retirees from a range of lower and higher income subgroups. Since both surveys have been conducted repeatedly over long time periods, they allow changes in retirement spending patterns over time to be examined. The Health and Retirement Study actually follows the same people over time, in what is known as a panel design, allowing even more detailed examination of the effects of retirement on individuals. Examples of such longitudinal studies include spending pattern changes as individuals transition into retirement, and spending changes during recessionary periods as a function of retirement income level.

Retirement spending trends observed for the average U.S. retiree include the following:


 * Spending drops modestly (14%) immediately after retirement, partly due to the cessation of work-related expenses (e.g. special clothing and transportation) and changes in food expenditures.
 * Inflation-adjusted spending continues dropping slowly afterwards as retirees age into their late 70's.
 * Housing + Related is by far the largest spending category for all age groups.
 * Health Care expenses rise with age, but remain substantially lower than housing expenses.
 * Increases in health care spending are not sufficient to change the downward trend in total inflation-adjusted spending, at least not until late in retirement.

Studies of retirement spending patterns across racial and income level subgroups reveal measurable heterogeneity in U.S. retirement spending. Since housing is the largest retirement expense, it is not surprising that such differences are clearly seen in this category. The size of the typical spending drop upon retirement is inversely related to household wealth (poorer hourseholds exhibiting on average a much larger spending drop). However not all retirees exhibit a spending drop: about 53% of households exhibit a drop in spending at retirement, another 35% report a negligible change in spending at retirement, while 12% reported an increase in spending at retirement.

A significant percentage of individuals (25-28%) enter retirement involuntarily. The overwhelming number of involuntary retirements can be traced back to major health difficulties or extended unemployment. For such retirees spending at retirement exhibits a much stronger drop than for the average retiree.

Importance of retirement spending data
Access to actual spending data for retirees can be valuable to several categories of people:


 * Individual Pre-retirees. Individuals planning for their future are often perplexed about how their spending might change as they enter into and progress through retirement.  For them, having access to typical retirement spending data can be extremely helpful in thinking through their own future spending.
 * Financial Planners. Access to such data assists them in developing retirement plans for their clients that have a better factual basis than anecdotal spending rules based on limited personal experience.
 * Economists. Without such data they are unable to test the validity of their economic models describing consumer behavior.  Such models often find their way into retirement planning software.
 * Developers of Retirement Planning Software. They need to be confident that the spending input portions of their programs are flexible enough to accommodate a broad socio-economic range of potential users.  Access to retiree spending data allows them to test their software against reality and make improvements where needed.

Most reliable information on U.S. retiree spending is derived from two survey databases: the annual Consumer Expenditure Survey (CE) and the biennial Health and Retirement Study (HRS). These surveys provide very detailed economic data covering a broad range of the U.S. population. Working directly with these databases can be difficult. It is usually easier to refer to academic research articles that analyze just the portion of the data that is of interest. The remainder of this article provides an introduction to each survey, presents some representative data for the "average retiree", and also presents some information covering non-average retiree spending patterns.

U.S. Consumer Expenditure Survey (CE)
The main source of comprehensive spending data for U.S. retirees is the annual Consumer Expenditure Survey (CE) conducted by the Census Bureau for the Bureau of Labor Statistics (BLS). As stated on their website,


 * "The CE is important because it is the only Federal survey to provide information on the complete range of consumers' expenditures and incomes, as well as the characteristics of those consumers. It is used by economic policymakers examining the impact of policy changes on economic groups, by businesses and academic researchers studying consumers' spending habits and trends, by other Federal agencies, and, perhaps most importantly, to regularly revise the Consumer Price Index market basket of goods and services and their relative importance."

The BLS publishes numerous tabulated summary sheets of CE data as well as short summary articles describing the extensive database.

Survey methodology - CE
The methodology for the Consumer Expenditure (CE) Survey is described in chapter 16 of the BLS Handbook of Methods. Briefly, the CE is a combination of data from two separate surveys having different data collection techniques and covering different samples of the U.S. population. The Interview Survey collects data from about 7,100 consumer units (households) every quarter for 5 quarters. This survey focuses on household spending that is either relatively large (e.g. cars, major appliances) or that occurs on a regular basis (e.g. rent, utilities). After the 5 quarters a household is dropped and a new one added in its place. The Diary Survey focuses on spending for small, frequently purchased items (e.g. food, personal care products). Each year about 7,100 consumer units fill in paper questionnaires which they use to record all spending over a 2 week period. The BLS combines data from these 2 surveys to generate the final CE tables.

Average spending by age ranges - CE
General data for retirement age spending can be gleaned from the CE survey Expenditure Tables which organize the data by age ranges. Key information from these tables for 2007 (pre-recession) and 2009 (late-recession) is summarized in the table below. Spending data for the "pre-retirement" age group 55-64 have been added for comparison. Spending in the sub-categories is shown as a percentage of the total.


 * {|border="2" cellspacing="0" cellpadding="4"

* Income does not include retiree withdrawals from personal savings. * * Contributions to Social Security and pensions, which were included in the CE average annual expenditure total, have been removed.
 * +Average Household Spending by Age Range
 * align = "center" rowspan = "2"|Category
 * align = "center" colspan = "3"|2007 CE Data
 * align = "center" colspan = "3" |2009 CE Data
 * align = "center"|55-64
 * align = "center"|65-74
 * align = "center"|75+
 * align = "center"|55-64
 * align = "center"|65-74
 * align = "center"|75+
 * Persons in Consumer Unit
 * align = "center"|2.1
 * align = "center"|1.8
 * align = "center"|1.5
 * align = "center"|2.1
 * align = "center"|1.9
 * align = "center"|1.6
 * Income (pre-tax)*
 * align = "center"|$71,048
 * align = "center"|$47,708
 * align = "center"|$32,499
 * align = "center"|$70,609
 * align = "center"|$47,286
 * align = "center"|$31,715
 * Total Spending**
 * align = "center"|$48,054
 * align = "center"|$40,037
 * align = "center"|$29,698
 * align = "center"|$46,116
 * align = "center"|$40,685
 * align = "center"|$30,946
 * Food + Alcohol
 * align = "center"|14%
 * align = "center"|14%
 * align = "center"|13%
 * align = "center"|15%
 * align = "center"|15%
 * align = "center"|14%
 * Housing + Related
 * align = "center"|36%
 * align = "center"|34%
 * align = "center"|38%
 * align = "center"|37%
 * align = "center"|36%
 * align = "center"|38%
 * Apparel
 * align = "center"|4%
 * align = "center"|3%
 * align = "center"|2.5%
 * align = "center"|3%
 * align = "center"|3%
 * align = "center"|3%
 * Transportation
 * align = "center"|20%
 * align = "center"|19%
 * align = "center"|13%
 * align = "center"|18%
 * align = "center"|17%
 * align = "center"|12%
 * Healthcare
 * align = "center"|7%
 * align = "center"|12%
 * align = "center"|14%
 * align = "center"|8%
 * align = "center"|12%
 * align = "center"|15%
 * Entertainment
 * align = "center"|6%
 * align = "center"|7%
 * align = "center"|4%
 * align = "center"|6%
 * align = "center"|6%
 * align = "center"|5%
 * Contributions
 * align = "center"|6%
 * align = "center"|5%
 * align = "center"|9%
 * align = "center"|5%
 * align = "center"|5%
 * align = "center"|8%
 * Misc. Other
 * align = "center"|7%
 * align = "center"|6%
 * align = "center"|6.5%
 * align = "center"|8%
 * align = "center"|6%
 * align = "center"|5%
 * colspan = "7"|
 * align = "center"|5%
 * align = "center"|5%
 * align = "center"|8%
 * Misc. Other
 * align = "center"|7%
 * align = "center"|6%
 * align = "center"|6.5%
 * align = "center"|8%
 * align = "center"|6%
 * align = "center"|5%
 * colspan = "7"|
 * colspan = "7"|
 * colspan = "7"|
 * }

The first row showing the Consumer Unit refers to the number of persons living in the surveyed household that are financially connected with each other. If conversion of the dollar amounts to those for single (consumer unit = 1.0) or married (consumer unit = 2.0) is desired, the so-called square root rule can be used.

Key observations on spending for the average person in these age ranges are:


 * Total Spending dropped during the transition into retirement (ages 55 - 64 to ages 65 – 74), and continued dropping into ages 75 and older.
 * Both income and spending dropped slightly from 2007 to 2009, possibly as a response to the economic uncertainty surrounding the housing bubble recession.
 * Housing + Related is by far the highest spending category for all age groups.
 * There is very little percentage change in spending for food, clothing and entertainment for all the age ranges.
 * Transportation spending drops for the 75+ age range.
 * Health care spending rises steadily across the age ranges, becoming almost twice as large for the 75+ age range as for the 55-64 age range.
 * Spending as a percentage of the pre-retirement, pre-tax average income (corrected for consumer unit differences was about 66% (ages 55-64), 62% (ages 65-74) and 50% (ages 75+).

Influence of spending (income) level - CE
The BLS also publishes "Cross-tabulated" tables. Using these it is possible to examine how spending patterns change according to retiree income. Although the CE Expenditure Tables separate retirees into two age groups (ages 65-74 and ages 75+), the cross tabulated tables lump together spending for all person 65 and over. The following table shows the spending data from the combined 2006-2007 (pre-recession) surveys.


 * {|border="2" cellspacing="0" cellpadding="4"

* Income does not include retiree withdrawals from personal savings. * * Contributions to Social Security and pensions, which were included in the CE average annual expenditure total, have been removed.
 * +Distribution of Household Spending (2006-2007) for Age 65+
 * align = "center" rowspan = "2"|Category
 * align = "center" colspan = "5"|Household Income Ranges (more than 1 person)
 * align = "center"|$5K - $10K
 * align = "center"|$15K - $20K 
 * align = "center"|$30K – $40K 
 * align = "center"|$50K - $70K 
 * align = "center"|$70K and Over 
 * Persons in Consumer Unit
 * align = "center"|1.2
 * align = "center"|1.4
 * align = "center"|1.9
 * align = "center"|2.1
 * align = "center"|2.3
 * Income (pre-tax)*
 * align = "center"|$8,333
 * align = "center"|$17,411
 * align = "center"|$34,895
 * align = "center"|$58,693
 * align = "center"|$121,513
 * Income (post-tax)*
 * align = "center"|$8,260
 * align = "center"|$17,287
 * align = "center"|$34,118
 * align = "center"|$57,453
 * align = "center"|$115,580
 * Total Spending**
 * align = "center"|$14,102
 * align = "center"|$23,187
 * align = "center"|$34,200
 * align = "center"|$46,422
 * align = "center"|$72,598
 * Food + Alcohol
 * align = "center"|16%
 * align = "center"|14%
 * align = "center"|13%
 * align = "center"|14%
 * align = "center"|13%
 * Housing + Related
 * align = "center"|48%
 * align = "center"|40%
 * align = "center"|34%
 * align = "center"|33%
 * align = "center"|32%
 * Apparel
 * align = "center"|3%
 * align = "center"|3%
 * align = "center"|2%
 * align = "center"|4%
 * align = "center"|3%
 * Transportation
 * align = "center"|10%
 * align = "center"|14%
 * align = "center"|17%
 * align = "center"|17%
 * align = "center"|19%
 * Healthcare
 * align = "center"|12%
 * align = "center"|15%
 * align = "center"|15%
 * align = "center"|13%
 * align = "center"|10%
 * Entertainment
 * align = "center"|4%
 * align = "center"|5%
 * align = "center"|6%
 * align = "center"|6%
 * align = "center"|6%
 * Contributions
 * align = "center"|3%
 * align = "center"|5%
 * align = "center"|5%
 * align = "center"|7%
 * align = "center"|9%
 * Misc. Other
 * align = "center"|4%
 * align = "center"|4%
 * align = "center"|8%
 * align = "center"|6%
 * align = "center"|8%
 * colspan = "6"|
 * align = "center"|3%
 * align = "center"|5%
 * align = "center"|5%
 * align = "center"|7%
 * align = "center"|9%
 * Misc. Other
 * align = "center"|4%
 * align = "center"|4%
 * align = "center"|8%
 * align = "center"|6%
 * align = "center"|8%
 * colspan = "6"|
 * colspan = "6"|
 * colspan = "6"|
 * }

Notice for the lower three household income ranges that Total Spending exceeds pre-tax Income. The BLS survey methodology includes as Income only taxable dollar flows (e.g. the taxable portion of Social Security benefits, pension and annuity payments, traditional IRA withdrawals, bank interest). This indicates that people within the lower three ranges are receiving substantially tax-free Social Security benefit payments and/or are withdrawing non-taxable personal savings to meet their spending needs.

Key observations on average spending across these income ranges are:


 * Food spending, expect for the very lowest income group, constitutes a relatively constant percentage of Total Spending.
 * Housing + Related is by far the highest spending category for all age groups. The lowest income group spends about 45% more on housing relative to the higher income groups (48% versus 33%).
 * The absolute necessities of living (Food and Housing) consume about 64% of spending for the lowest income group, versus about 46% for the highest three income groups. This suggests that lower income retiree spending gets squeezed by somewhat inflexible lower limits for Food and Housing.
 * Clothing spending constitutes a relatively constant percentage of Total Spending.
 * Higher income groups spend a considerably higher percentage on Transportation than the lower income groups.
 * Healthcare spending as a percentage is slightly higher for the middle income groups.

Overcoming difficulties in data interpretation
There are difficulties in using age as a proxy for retirement status. The BLS Consumer Expenditure tables contain data for both "retired" and "non-retired" individuals within each age range. For example, Fisher et.al. determined the percentage of persons retired within CE data and how that percentage changed with age for several cohorts of individuals. The percentage retired for just one of these cohorts is shown in the following table:


 * {|border="2" cellspacing="0" cellpadding="4"


 * align = "center" |Age Range
 * align = "center" |CE Study Year
 * align = "center" |% Retired
 * align = "center"|55 - 59
 * align = "center"|1984
 * align = "center"|5.7%
 * align = "center"|60 - 64
 * align = "center"|1988
 * align = "center"|22.8%
 * align = "center"|65 - 69
 * align = "center"|1993
 * align = "center"|53.5%
 * align = "center"|70 - 74
 * align = "center"|1998
 * align = "center"|76.3%
 * align = "center"|75 - 79
 * align = "center"|2003
 * align = "center"|79.5%
 * }
 * align = "center"|75 - 79
 * align = "center"|2003
 * align = "center"|79.5%
 * }
 * }

The definition of "retired" they used was rather strict: no work income for the past year and self-identified as retired. Thus individuals even at older ages who derive some income from outside work were counted by Fisher et.al. as non-retried. Nevertheless, the transition from non-retired to retired status between ages 60 and 75 shown in this table seems entirely reasonable. It is probably wise to consider spending data for ages 65-74 in the previous CE tabulations to represent mostly but not completely retired individuals.

Researchers with the energy to analyze the original BLS CE database more thoroughly are able to make important refinements to their retirement spending conclusions. For example Fisher et.al. use such refined analysis techniques and determine a "real dollar" median spending drop at retirement of 2.5%, followed by about a 1% per year median spending drop over the next 15 years.

Hatcher has reported on an analysis of the raw CE database in which retirees were separated out from non-retirees for all households over age 50. He used the same rule as Fisher et.al. (above) to distinguish these two groups. Hatcher found a much stronger spending drop at retirement than Fisher et.al.: retired households spend about 25% less per quarter than non-retired households in the same age range. He also found that spending during retirement dropped slowly with age. But in his estimation the drop in spending at retirement would have a stronger affect on a calculated retirement savings target than the drop in spending as retirement progressed.

Bernheim et.al. found that expenditures dropped by 14% at retirement, midway between the results reported by Hatcher and Fisher et.al. Bernheim used panel data on U.S. households from the Panel Study of Income Dynamics in their analysis. Therefore their result would not be tainted by questions of retirement status.

The Health and Retirement Study (HRS) survey
The Health and Retirement Study (HRS) is a biennial survey of Americans over the age of 50. It is conducted by the Survey Research Center at the University of Michigan for the National Institute on Aging. The HRS provides retirement spending data that complement those from the BLS Consumer Expenditure Surveys. The National Institute on Aging publication Growing Older in America: The Health and Retirement Study contains many general observations on retirement living (through 2002) derived from the HRS surveys.

Survey methodology - HRS
The methodology for the Health and Retirement Study (HRS) survey is described in a series of documents posted at the HRS website. The HRS focuses only on households having persons over 50 years old. In contrast to the CE surveys, the HRS survey interviews the same household units every other year until their death. Every six years (every third survey) a new cohort of household units covering the ages 51-56 is added to the survey set. This new set of household units is referred to as a wave. In 2004 more than 26,000 individuals were contacted as part of the survey. In order to ensure that the survey results are representative for minority subgroups within the U.S. population, oversampling is employed for African Americans and Hispanics.

Similarly to the CE database, researchers interested in the spending habits of American retirees perform and publish detailed studies of the HRS database. The Survey Research Center at the University of Michigan maintains a searchable bibliography of these studies. One of the more recent broad examinations of retiree spending based on the HRS database was published by B. Butrica, J. Goldwyn and R. Johnson in 2005. The most relevant portions of their paper are summarized here.

Spending categories
In order to simplify their data analysis, Butrica et.al. subdivided the survey spending data into the 8 broad categories shown in the table below. These categories are similar in many respects to the CE database categories, expect that the HRS pulls purchases of Consumer Durable Goods out of separate CE categories and groups them together.


 * {|border="2" cellspacing="0" cellpadding="4" width="85%"


 * align = "center" |Category
 * align = "center" |Detailed Description
 * align = "center" |Housing
 * Mortgage payments, home/renter insurance premiums, property, tax payments, rent, utility costs (electricity, water, heat, phone, and cable and internet services), spending on house/yard supplies, and home maintenance costs
 * align = "center" |Health Care
 * Out-of-pocket payments on insurance premiums, drugs, health services, and medical supplies
 * align = "center" |Food
 * Expenditures on groceries but not spending on dining outside of the home
 * align = "center" |Clothing
 * Clothing of all types
 * align = "center" |Transportation
 * Automobile finance charges, automobile insurance premiums, gasoline, and automobile maintenance; excludes any spending on public transit
 * align = "center" |Entertainment
 * Dining out, vacations, tickets to events, and hobbies
 * align = "center" |Gifts
 * Charity and other gifts
 * align = "center" |Other Consumer Durables
 * Purchases of automobiles, refrigerators, washers and dryers, dishwashers, televisions, and computers
 * }
 * Dining out, vacations, tickets to events, and hobbies
 * align = "center" |Gifts
 * Charity and other gifts
 * align = "center" |Other Consumer Durables
 * Purchases of automobiles, refrigerators, washers and dryers, dishwashers, televisions, and computers
 * }
 * Purchases of automobiles, refrigerators, washers and dryers, dishwashers, televisions, and computers
 * }

Median spending by age ranges - HRS
Using the 8 spending categories tabulated above, Butrica et.al performed a wide range of data analyses with the goal of identifying spending trends from pre-retirement through the later years of life. As shown in the following table, the spending data for the "Median Retiree" were grouped into three age ranges and were expressed as a percentage of total spending. The spending data were further subdivided by marital status, since this turned out to be a significant variable.


 * {|border="2" cellspacing="0" cellpadding="4"


 * +Median Per Capita Spending (2001) by Marital Status and Age
 * align = "center" rowspan = "2"|Category
 * align = "center" colspan = "3"|Married
 * align = "center" colspan = "3"|Nonmarried
 * align = "center"|53-64
 * align = "center"|65-74
 * align = "center"|75+
 * align = "center"|53-64
 * align = "center"|65-74
 * align = "center"|75+
 * Income (pre-tax)
 * align = "center"|$30,898
 * align = "center"|$20,023
 * align = "center"|$15,800
 * align = "center"|$24,683
 * align = "center"|$18,581
 * align = "center"|$15,040
 * Total Spending
 * align = "center"|$17,409
 * align = "center"|$15,414
 * align = "center"|$13,678
 * align = "center"|$17,196
 * align = "center"|$17,083
 * align = "center"|$15,390
 * Housing
 * align = "center"|36%
 * align = "center"|31%
 * align = "center"|29%
 * align = "center"|37%
 * align = "center"|36%
 * align = "center"|41%
 * Health Care
 * align = "center"|12%
 * align = "center"|17%
 * align = "center"|19%
 * align = "center"|11%
 * align = "center"|14%
 * align = "center"|18%
 * Food
 * align = "center"|10%
 * align = "center"|13%
 * align = "center"|14%
 * align = "center"|13%
 * align = "center"|17%
 * align = "center"|13%
 * Clothing
 * align = "center"|5%
 * align = "center"|3%
 * align = "center"|2%
 * align = "center"|5%
 * align = "center"|3%
 * align = "center"|3%
 * Transportation
 * align = "center"|14%
 * align = "center"|13%
 * align = "center"|10%
 * align = "center"|13%
 * align = "center"|10%
 * align = "center"|9%
 * Entertainment
 * align = "center"|10%
 * align = "center"|13%
 * align = "center"|10%
 * align = "center"|13%
 * align = "center"|10%
 * align = "center"|6%
 * Gifts
 * align = "center"|8%
 * align = "center"|6%
 * align = "center"|10%
 * align = "center"|5%
 * align = "center"|7%
 * align = "center"|8%
 * Other Durable
 * align = "center"|6%
 * align = "center"|3%
 * align = "center"|7%
 * align = "center"|3%
 * align = "center"|3%
 * align = "center"|2%
 * }
 * align = "center"|7%
 * align = "center"|8%
 * Other Durable
 * align = "center"|6%
 * align = "center"|3%
 * align = "center"|7%
 * align = "center"|3%
 * align = "center"|3%
 * align = "center"|2%
 * }
 * }

General Median Spending Observations ( married adults):

General Median Spending Observations ( non-married adults):
 * The median per capita spending fell with age (in 2001 dollars): $17,409 at ages 53 - 64, $15,414 at ages 65 - 74, and finally $13,678 at ages 75+.
 * The median expenditure dropped by 21% from youngest to oldest group, and by 11% from middle to oldest group.
 * Per capita spending in retirement was 50% (or less) of pre-tax income before retirement.

Housing Observations:
 * The median retiree per capita spending was higher than for married persons.
 * In contrast to married individuals, median spending barely dropped from the ages 53 - 64 group ($17,196) to the ages 65 - 74 group ($17,083), and then was lower for the ages 75 + group ($15,390) by only about 10%.
 * Per capita spending in retirement was 70% (or less) of pre-tax income before retirement.

Health Care Observations:
 * Spending within the Housing category represented the largest expense regardless of age or marital status. On a per capita basis, non-married persons spent a larger percentage on Housing than did married persons.
 * The spending going to mortgages declines with age, but spending for utilities and maintenance increases with age.
 * Blacks and Hispanics devote a larger percentage of spending to Housing than do Whites.
 * 25% of married adults ages 65 and older are homeowners with mortgages.
 * Non-married individuals are nearly 3 times more likely to rent their homes in retirement.

Observations for Other Spending Categories
 * Spending for Health Care increases between the youngest and oldest age groups for both married and non-married adults. But even in later life (ages 75+), spending for Health Care still remains significantly lower than for Housing.  The popular stereotype of Health Care being the largest expense for older retirees does not appear to be correct.
 * Insurance premiums are the largest spending sub-category, followed by prescription drugs, health services and medical supplies.
 * Retirees who complement Medicare using non-group insurance spend more for insurance premiums and prescription drugs than do those with access to group insurance.
 * For older married persons in poor health, the per capita Health Care spending is almost twice as large as for those in excellent health. In contrast, for non-married person there is only a nominal increase.

A regression analysis of a subset of the HRS data for 2001, 2003 and 2005 that separated out retired from non-retired individuals was presented by Chen et.al. They confirmed that moving from the 65-69 age group to the 70+ age group led to a decline in Total Spending. Within subcategories, the only significant increase in spending with age was for Health Care.
 * Per capita Food spending for married persons is only slightly lower than for Health Care. For non-married persons spending for Food is slightly higher than for Health Care during the earlier years of retirement.
 * Spending for clothes drops off after retirement but then remains fairly stable on a percentage basis.
 * Transportation and Entertainment spending are roughly equal and decline slightly through retirement for both married and non-married.
 * For retired married persons, spending for Gifts and Other Consumer Durables increase in later retirement, but not for non-married persons.

Influence of spending (income) level - HRS
The same survey data can be looked at from a different perspective by grouping them according to the absolute level of per capita spending. The following two tables show how retiree spending changes when moving from a very low spending sub-population (5th – 15th percentile range) to a very high spending sub-population (85th – 95th percentile range). The data for married and non-married persons are grouped into separate tables in order to highlight the differences between these groups. Note that the Total Spending row shows 2001 dollar amounts.


 * {|border="2" cellspacing="0" cellpadding="4"


 * +Distribution of Per Capita Spending (2001) for Married Adults Age 65+
 * align = "center" rowspan = "2"|Category
 * align = "center" colspan = "5"|Percentile Ranges
 * align = "center"|5th - 15th
 * align = "center"|20th - 30th
 * align = "center"|45th - 55th
 * align = "center"|70th - 80th
 * align = "center"|85th - 95th
 * Total Spending
 * align = "center"|$6,487
 * align = "center"|$9,881
 * align = "center"|$14,792
 * align = "center"|$24,016
 * align = "center"|$38,749
 * Housing
 * align = "center"|35%
 * align = "center"|33%
 * align = "center"|29%
 * align = "center"|30%
 * align = "center"|28%
 * Health Care
 * align = "center"|16%
 * align = "center"|18%
 * align = "center"|20%
 * align = "center"|13%
 * align = "center"|13%
 * Food
 * align = "center"|16%
 * align = "center"|14%
 * align = "center"|13%
 * align = "center"|9%
 * align = "center"|6%
 * Clothing
 * align = "center"|3%
 * align = "center"|2%
 * align = "center"|2%
 * align = "center"|2%
 * align = "center"|3%
 * Transportation
 * align = "center"|14%
 * align = "center"|12%
 * align = "center"|12%
 * align = "center"|12%
 * align = "center"|9%
 * Entertainment
 * align = "center"|8%
 * align = "center"|9%
 * align = "center"|10%
 * align = "center"|12%
 * align = "center"|12%
 * Gifts
 * align = "center"|7%
 * align = "center"|9%
 * align = "center"|10%
 * align = "center"|8%
 * align = "center"|15%
 * Other Durable
 * align = "center"|2%
 * align = "center"|3%
 * align = "center"|4%
 * align = "center"|13%
 * align = "center"|14%
 * }
 * align = "center"|8%
 * align = "center"|15%
 * Other Durable
 * align = "center"|2%
 * align = "center"|3%
 * align = "center"|4%
 * align = "center"|13%
 * align = "center"|14%
 * }
 * }


 * {|border="2" cellspacing="0" cellpadding="4"


 * +Distribution of Per Capita Spending (2001) for Non-married Adults Age 65+
 * align = "center" rowspan = "2"|Category
 * align = "center" colspan = "5"|Percentile Ranges
 * align = "center"|5th - 15th
 * align = "center"|20th - 30th
 * align = "center"|45th - 55th
 * align = "center"|70th - 80th
 * align = "center"|85th - 95th
 * Total Spending
 * align = "center"|$5,626
 * align = "center"|$9,746
 * align = "center"|$16,178
 * align = "center"|$27,555
 * align = "center"|$46,746
 * Housing
 * align = "center"|47%
 * align = "center"|38%
 * align = "center"|39%
 * align = "center"|37%
 * align = "center"|28%
 * Health Care
 * align = "center"|13%
 * align = "center"|19%
 * align = "center"|16%
 * align = "center"|13%
 * align = "center"|16%
 * Food
 * align = "center"|17%
 * align = "center"|13%
 * align = "center"|15%
 * align = "center"|12%
 * align = "center"|8%
 * Clothing
 * align = "center"|3%
 * align = "center"|4%
 * align = "center"|3%
 * align = "center"|4%
 * align = "center"|2%
 * Transportation
 * align = "center"|8%
 * align = "center"|10%
 * align = "center"|9%
 * align = "center"|9%
 * align = "center"|10%
 * Entertainment
 * align = "center"|4%
 * align = "center"|7%
 * align = "center"|7%
 * align = "center"|8%
 * align = "center"|11%
 * Gifts
 * align = "center"|6%
 * align = "center"|8%
 * align = "center"|7%
 * align = "center"|12%
 * align = "center"|14%
 * Other Durable
 * align = "center"|2%
 * align = "center"|2%
 * align = "center"|3%
 * align = "center"|5%
 * align = "center"|12%
 * }
 * align = "center"|12%
 * align = "center"|14%
 * Other Durable
 * align = "center"|2%
 * align = "center"|2%
 * align = "center"|3%
 * align = "center"|5%
 * align = "center"|12%
 * }
 * }

Spending observations across spending (income) groups:


 * For married persons the per capita Total Spending in the highest percentile range (85th – 95th) is about 6 times larger than for the lowest range (5th – 15th). For non-married persons, the difference is 8 times.
 * For both married and non-married persons the per capita spending percentage for Housing is greatest for those in the lowest spending range.
 * In spite of the increasing total spending levels, the percentage spending for Health Care shows only modest changes across the spending (income) groups. Those persons with access to more money spend larger amounts on Health Care.
 * The percentage spending on Food drops significantly for the higher spending groups. This trend is more pronounced that seen in the BLS CE data.
 * Clothing is a nearly constant percentage of spending for all spending groups.
 * The percentage spending on Entertainment rises across the spending groups, particularly for non-married persons.
 * The percentage spending on Other Consumer Durables rises across the spending groups, particularly for the highest two percentile ranges.

Heterogeneity in retirement spending
A careful reading of the literature shows there to be a tremendous amount of heterogeneity between retirees in how their spending changes at retirement and as they age. Although it is useful to understand spending patterns for the “average retiree”, it is also important to examine the distribution of spending around the average. In this regard the data from long term panel studies such as the Health and Retirement Study (HRS) are more useful than data from rotating panel studies such as the Consumer Expenditure Survey (CE).

Consider for example the spending drop that often accompanies retirement. Bernheim et.al. used panel data on U.S. households from the Panel Study of Income Dynamics and estimated an average spending drop of 14% at retirement. Hurd and Rohwedder analyzed this spending drop in more detail using the CAMS supplement to the HRS survey. Grouping the data by quartile of household wealth, they found that the households with the lowest 25% of wealth experienced the largest average drop in spending at retirement: 22%. At increasing wealth quartiles the spending drop became progressively less: 17%, 13% and 7%. These estimates are very consistent with the 14% overall average drop reported by Bernheim et.al., even though a different panel study database was used. The heterogeneity about the “average” is apparent.

When the above data are examined from a different perspective, a new type of heterogeneity is discovered. Hurd and Rohwedder also analyzed their CAMS data from the perspective of amount of spending drop regardless of household wealth. They found that only about 53% of households studied exhibited a drop in spending at retirement, another 35% reported a negligible change in spending at retirement, while 12% reported an increase in spending at retirement.

Another factor that can impact retirement spending is whether the retirement was voluntary or involuntary. Involuntary retirement is a more common occurrence than is often appreciated: as many as 25% to 28% of retirements fall into this category. The overwhelming number of involuntary retirements can be traced back to major health difficulties or extended unemployment. Involuntary retirement typically is associated with a sharp drop in spending. 66% of involuntary retirees report being bothered somewhat or a lot by not having enough income, whereas only 21% of voluntary retirees report similar feelings.

The retirement consumption puzzle
Many economists have noted with puzzlement that individuals exhibit a significant drop in their spending upon retirement. The puzzlement derives from an economist's expectation that people will plan ahead in order to smooth their spending across predictable changes in income, such as accompany retirement.

E. Hurst, in reviewing much of the research on the topic of spending changes at retirement, has proposed 5 key facts:
 * Certain types of spending actually do drop sharply as households enter into retirement.
 * The spending drops are mostly limited to 2 categories: work related items and food. Drops in work related items are not at all surprising.
 * Drops in food related spending have a straightforward explanation. Retirees spend more time in food preparation and therefore need spend less money.  The actual types and quantities of food consumed do not substantially change.
 * There are strong variations in spending patterns upon retirement across the spectrum of retirees. For example, households showing the greatest drop in spending upon retirement tend to have the smallest personal savings.  Additionally, there does appear to be a small segment of retirees (no more than 25%) that not only spend less money upon retirement, but actually consume less.  That is, they are unable to maintain their pre-retirement lifestyle even by substituting their time for money.
 * Households that experience such real drops in consumption often experienced involuntary retirement. These may be caused by either extended job layoffs or by significant health issues that preclude continued work.  This is the segment of the retiree population that exhibits the largest drop in spending upon retirement.

Race and retirement spending
Both the Consumer Expenditure Survey (CE) and the Health and Retirement Survey (HRS) contain information on the respondent's race. These databases reveal that there are differences in spending between the median White, African American and Hispanic retiree.


 * Housing : the homeownership rates for White retirees are significantly higher than for African American or Hispanic retirees. But for all three groups housing is by far the largest percentage spending category.
 * Food : Hispanics allocate the largest percentage to total food and food at home, followed by African American and White retirees. Whites allocate a larger percentage to food away from home and on alcoholic beverages.
 * Transportation : African American retirees spend about half as much on new cars and trucks as Whites or Hispanics, but spend more on used cars and trucks.  Whites spend a larger percentage share on air transportation than the other groups.
 * Health Care : the median Hispanic retiree spends the highest percentage share on total health care, followed by African Americans and then Whites.