Is Conventional Financial Planning Good for Your Financial Health?

Laurence J. Kotlikoff

Professor of Economics Boston University Boston

Economic theory predicts, and casual observation confirms, that households seek to maintain their living standards (smooth their consumption) as they age and face life’s various contingencies. Seeking a stable living standard and actually achieving one, however, are two very different things. We cannot insure against aggregate shocks, including economic downturns, natural disasters, and epidemics, nor can we buy actuarially fair insurance against a range of individual shocks, such as job loss, excessive longevity, and disability. And we generally lack the self-control to save and insure adequately even given attractive opportunities to do so.

These concerns with consumption smoothing are well known and have been amply studied.1 But another problem—namely, the computation challenge posed by consumption smoothing—has received little attention, which is surprising because consumption mistakes could well swamp these other considerations. Certainly, the calculations required to minimize consumption disruptions over time and across states of nature are highly complex. Just consider the number and range of current and future variables involved in consumption smoothing: The list includes household demographics; labor earnings; retirement dates; federal, state, and local taxes; U.S. Social Security benefits; pension benefits; regular and retirement assets; borrowing constraints; rules for retirement account contributions and withdrawals; home ownership; mortgage finance; economies in shared living; the relative costs of children; changes in housing; choice of where to live; the financing of college and weddings; gifts made or received; inheritances; paying for one’s dream boat; and so on. And each of these variables demands consideration under each and every survival contingency.2

1As an example, see http://post.economics.harvard.edu/faculty/laibson/papers.html for a long list of excellent papers by David Laibson and his colleagues contributing to the field of behavioral finance. 2Survival contingencies are distinguished by which spouse/partner dies and when he or she dies. The reason is that the survivor will inherit different amounts of wealth, collect different amounts of life insurance, and receive different levels of Social Security survivor and retirement benefits depending on the age at which his or her spouse/partner dies.

Taxation by itself is a factor worthy of a high-speed computer processor. Figuring out the federal and state taxes when both spouses are alive and in each future survivor state (years in which one spouse is deceased) requires determining whether they will itemize their deductions, whether they will receive any of the many potentially available tax credits, whether they will have to pay the Alternative Minimum Tax, whether they will pay taxes on their Social Security benefits, whether they will be contributing to or withdrawing from retirement accounts, and whether they will be in high or low tax brackets. And as if this list were not bad enough, determining future taxes introduces a nasty simultaneity problem. We cannot figure out the future taxes until we know the current spending (which determines, in part, the future taxable capital income), but we cannot figure out the current spending without knowing the future taxes (which determines, in part, what the couple has available to spend).

Computing Social Security benefits is another nightmare. With 2,528 separate rules in the Social Security Handbook, figuring out what retirement, dependent, survivor, divorce, mother, father, and child benefits one will receive can be maddening, particularly in light of the system’s complex average indexed monthly wage and primary insurance amount benefit formulae as well as its ancillary adjustments to the primary insurance amount. These adjustments include the earnings test, the early retirement reduction factors, the delayed retirement credit, the recomputation of benefits, the family benefit maximum, the windfall elimination and offset formulae for workers with noncovered employment, and the phase-in to the system’s ultimate age 67 normal retirement age.3 How many households can even list all these interrelated factors, let alone process them accurately?

The standard fallback in economics is that households do not need to know all the details or have PhDs in math to make correct life-cycle decisions. Indeed, we economists view the mathematical formulation of optimal intertemporal choice as simply descriptive modeling. “Households,” we tell ourselves, “do not do the math but act as if they do. Yes, households will make mistakes, but these mistakes will be small and average out.”

3Unfortunately, the handbook is remarkably uninformative about many details of these adjustments, particularly the order in which they are applied. Several of the old-timers knocking about in the U.S. Social Security Administration’s Office of the Actuary know these details, but when they go, this information may literally disappear. One might think that the computer code that generates the hundreds of billions of dollars worth of actual Social Security benefit payments would be easy to check regarding such matters. But this code is, as I have been told, documented very poorly and written in an ancient computer language, namely, COBOL, that no one at the Social Security Administration apparently understands. Furthermore, it cannot necessarily be reproduced from scratch because many Social Security rules were apparently made administratively.

Unfortunately, studies of saving and insurance adequacy and portfolio diversification—including Kotlikoff, Spivak, and Summers (1982); Auerbach and Kotlikoff (1987, 1991); Bernheim, Carman, Gokhale, and Kotlikoff (2003); Bernheim, Forni, Gokhale, and Kotlikoff (2000, 2003); and Bernheim, Berstein, Gokhale, and Kotlikoff (2006)—belie this proposition. Vast numbers of households save, insure, and avoid risk either far too much or far too little. The fact that households make both types of mistakes and that, as a consequence, generate average behavior that, on its face, is not extreme offers little professional consolation. The medical community would not declare that heart disease has been cured if half of peoples’ hearts beat too fast and the other half too slow.

The other professional fallback is that households that look financially sick are, in fact, financially healthy but simply have unusual preferences or prior beliefs about future events. Thus, a household that invests only in cash can be viewed as being extremely risk averse and also convinced that deflation is around the corner. Because preferences and priors are taken as economic primitives that are above reproach, there is no scientific basis for classifying extreme financial behavior as financial pathology. Imagine the medical profession declaring cancer a perfectly healthy manifestation of genetic free will.

Perhaps it is time to identify financial disease according to the financial pain it engenders. Based on this criterion, we are, generically speaking, financially quite sick. Indeed, the ravages of financial pathology are clearly seen among today’s elderly, one-third of whom are wholly dependent or almost wholly dependent on Social Security.

Many households seek to cure their financial ills by turning to the financial planning industry for advice. In so doing, they effectively let financial planners or financial planning software decide what their preferences are and how they should be maximized. For such households, the study of their financial behavior may simply boil down to understanding what planners or software they are using and what these planners or programs are leading them to do.

If conventional financial planning corresponds closely to what economics prescribes, then households taking this advice will improve their financial health. But what if conventional financial planning generates recommendations that are far afield from proper consumption smoothing? In this case, “financial planning” will represent a cure-all, not real penicillin, and potentially leave one in worse financial shape.

In this paper, I examine one aspect of conventional financial planning: namely, the requirement that households set their own retirement and survivor spending targets. I show that from the perspective of consumption smoothing (the goal of achieving a stable living standard), even small targeting mistakes, on the order of 10 percent, can lead to enormous mistakes in recommended saving and insurance levels and to major living standard disruptions (on the order of 30 percent) at retirement. Given the computational difficulties involved in achieving proper consumption smoothing on one’s own, targeting mistakes of 10 percent or greater appear inevitable.

In soliciting spending targets, the traditional approach asks households to tally up all of their current expenditures and to use this level of spending as a target for retirement and survivor spending with some adjustment for changes in spending needs. The goal here seems to be that of consumption smoothing (i.e., to achieve the same living standard before and after retirement and in survivor states).

The rub, however, is that the current level of spending, which underlies the targeting, may be higher or lower than the sustainable level. Given the computation problems referenced earlier, this is almost inevitable. If current spending is higher than the sustainable level, the targets will be set too high. In this case, households will be told to save and insure more than is consistent with consumption smoothing. If current spending is lower than the sustainable level, the targets will be set too low. In this case, households will be advised to save and insure less than is consistent with consumption smoothing.

If the target is set too high, the household will be told to oversave and overinsure and will end up with a lower living standard prior to retirement and a higher one after retirement and in survivor states. If the target is set too low, the household will be told to undersave and underinsure and will end up with a higher living standard prior to retirement and a lower one after retirement and in survivor states. Both types of targeting mistakes will lead to consumption disruption rather than consumption smoothing.

This conclusion depends, of course, on whether the household takes the advice being given. Telling households that are currently overspending to substantially cut their current living standard may lead them to ignore the advice, decide to retire later, or accept a much lower future living standard. And telling households that are currently underspending to start spending at what they perceive as crazy rates may lead them to discount financial planning altogether.

Why can small targeting mistakes lead to such bad financial advice and such large disruptions in living standard when the household retires or loses a head or spouse/partner? First, in the case of retirement, the targeting mistake is being made for roughly 30 years—from roughly age 65 to roughly age 95. In the case of survivorship, the targeting mistake is being made for all of the survivor’s potential remaining years. Second, spending more (less) than one should in retirement and survivor states means spending less (more) than one should prior to retirement in states when no one is deceased, which magnifies the living standard disruption (gap) beyond the original targeting mistake. Finally, the oversaving (undersaving) and overinsuring (underinsuring) associated with targeting for living standards that are too high (low) lead to higher (lower) taxes and insurance premium payments than would otherwise be paid. This induces further cuts (increases) in the preretirement living standard than would otherwise arise.

As I will show, targeted spending also underlies and undermines conventional financial planning’s portfolio advice, potentially leading households to take on much more investment risk than is prudent. What households need to assess in considering risky investments is the variability of their future living standards. But standard planning assumes households will spend precisely their targeted amounts year after year regardless of the returns they receive unless and until they run out of money. This approach focuses attention on the probability of the plan (i.e., the target) working rather than on the spending consequences of it not. Because riskier investing can raise the probability of a plan’s “success,” households may be encouraged to take more risk than is appropriate. They may also get the idea that adjusting their portfolios rather than their lifestyle is the prudent response to low returns.
Methodology

Maximizing and preserving one’s living standard is the hallmark of consumption smoothing—the economic approach to financial planning. Although conventional financial planning attempts to achieve consumption smoothing by having households set targets based on their current spending, this practice is essentially guaranteed to provide poor saving, insurance, and investment advice and to promote consumption disruption rather than consumption smoothing.

One might then wonder whether a better approach would be to use the conventional planning methodology but adjust the spending target to equalize living standards across one’s working life, retirement, and widow(er)hood. Such target practice can, in theory, work. But doing so for any given set of assumptions could take hours, not seconds. The reason for the added complexity is that many, if not most, households are borrowing constrained, meaning that they or their financial planner would need to simultaneously set and adjust spending targets for each borrowing-constrained interval.4 When one adds to this picture return uncertainty, we are no longer talking hours but weeks because the household would need to solve not for one spending path but for all possible (and potentially borrowing-constrained) spending paths that would arise under each path of return realizations.

The only practical way to handle these problems mathematically is via dynamic programming, developed in the early 1950s by Richard Bellman. Specifically, one formulates a general plan for consumption smoothing in the last period (the maximum age of life), which I will call T. It then uses the period T plan to formulate a plan for the next to last period (i.e., T – 1). The T – 1 plan is used to formulate

4Bernheim, Berstein, Gokhale, and Kotlikoff (2006) found that roughly three-fifths of Boston University participants in a study of saving and insurance adequacy were borrowing constrained.

the T – 2 plan and so forth, back to the current time period, which I will call 0. This Time 0 plan is then used to determine how much to spend, save, and insure in the current year.

Although mathematicians, economists, and engineers are well versed in dynamic programming, the architects of traditional financial planning software are not. Or if they are, they are constrained by their superiors to keep things simple, which, in this context, means failing to elicit much of the information—such as the path of future labor earnings, prospective changes in housing, and plans for retirement account withdrawals—needed to generate an accurate dynamic program.
Results

A dynamic programming model, as described above, is used to generate sample consumption-smoothing results for a prototypical family. These results are then compared with those given by a conventional targeting approach. First, I will review some general features of the consumption-smoothing results.

The amount of recommended consumption expenditures needed to achieve a given living standard varies from year to year in response to changes in the household’s composition. It also rises when the household moves from a situation of being liquidity constrained to one of being unconstrained (i.e., not needing or wanting to borrow). Finally, as mentioned, recommended household consumption will change over time if users intentionally specify that they want their living standard to change.

The simultaneity issue with respect to taxes mentioned earlier is just one of two such issues that needs to be considered. The second is the joint determination of life insurance holdings of potential decedents and survivors. Widows and widowers may need to hold life insurance to protect their children’s living standard through adulthood and to cover bequests, funeral expenses, and debts (including mortgages) that exceed the survivor’s net worth inclusive of the equity on her or his house.

But the more life insurance is purchased by the potential decedent, the less life insurance survivors will need to purchase, if one assumes they have such a need. Thus, survivors will pay less in life insurance premiums and have less need for insurance protection from their decedent spouse/partner. Hence, one cannot determine the potential decedent’s life insurance holdings until one determines the survivor’s holdings. But one cannot determine the survivor’s holdings until one determines the decedent’s holdings.

Dealing with the tax and life insurance simultaneity issues as well as the borrowing and nonnegative life insurance constraints all within a single dynamic program seems well beyond the computing power of a desktop personal computer, particularly given the speed required for a commercial product.

The difficulty, time, and luck required to solve this simultaneous consumption-smoothing, borrowing-constrained, saving, life insurance, and tax problem may explain why financial planning software producers have universally adopted the computationally trivial targeted-spending approach. As indicated, this approach puts the onus on the household heads and spouses/partners of doing literally millions of complex calculations in their heads in order to set their targets appropriately. Because none of us has computer chips implanted in our brains, let alone the right neuronware, the chance of making at least small targeting mistakes is extremely high.
Consumption Smoothing vs. Mis-Targeting Spending

Consider a middle-aged, middle-class, married household with two children. Both spouses are age 40 in 2005. One child is age 10, and one is age 7. The couple lives in California. The husband earns $75,000 a year, and the wife, $50,000. Neither spouse has a pension or a retirement account, but the couple does have $75,000 in regular assets. The couple owns a $300,000 home with a $125,000 20-year mortgage with monthly payments of $1,250. Property taxes, homeowners’ insurance, and maintenance total $6,000 a year. The couple plans to spend $25,000 in today’s dollars on college tuition and other expenses for each child for four years. Each spouse will retire at age 65 and begin collecting Social Security benefits in that year. Past “covered” earnings (i.e., used to determine Social Security benefits) for the husband (wife) were $37,500 ($25,000) in 1987 when he (she) was 22, and they grew by 4 percent each year through 2004. The couple expects inflation to run at 3 percent annually and to earn a 6 percent nominal rate of return on savings. The remaining inputs are the economies in shared living and the relative cost of children. Based on previous work, I assumed that two can live as cheaply as 1.6, that to provide the same living standard to a child as to an adult costs only 70 percent as much, and that taxes and Social Security provisions in the future are those suggested by current law.

Consumption Smoothing. Table 1 shows annual recommendations for selected years for the consumption-smoothing case. All values in these and other tables are in 2005 dollars. Note that the couple’s living standard per equivalent adult, given in the last column, equals $31,337. This amount remains constant through 2060, when each spouse reaches age 95, assuming they both live that long. The recommended total consumption expenditure, in contrast, initially equals $71,852 but declines in 2014 and again in 2017 as the two children reach age 19 and are assumed to leave the household. In 2017 and thereafter, recommended total consumption expenditure equals $50,139. For 2005, the recommendation for consumption smoothing would be $484,947 and $128,554 in term life insurance holdings for the husband and wife, respectively. Over time, recommended holdings decline.

Table 1. Consumption Smoothing Annual Recommendations

(in constant 2005 dollars)
H’s Life W’s Life Living Standard
Year H’s Age W’s Age Consumption Saving Insurance Insurance per Adult
2005 40 40 $71,852 $ 1,440 $484,947 $128,554 $31,337
2006 41 41 71,852 1,678 479,364 131,024 31,337
2007 42 42 71,852 1,906 467,677 132,826 31,337
2008 43 43 71,852 2,117 452,332 135,409 31,337
2009 44 44 71,852 2,309 435,512 139,488 31,337
2010 45 45 71,852 2,458 425,461 144,822 31,337
2011 46 46 71,852 2,597 415,998 151,089 31,337
2012 47 47 71,852 2,701 407,888 158,198 31,337
2013 48 48 71,852 2,767 401,155 166,140 31,337
2014 49 49 61,455 –12,770 388,883 169,766 31,337
2015 50 50 61,455 –12,895 377,052 173,952 31,337
2016 51 51 61,455 –13,057 365,666 178,683 31,337
2017 52 52 50,139 –27,760 333,541 165,914 31,337
2018 53 53 50,139 –2,930 301,025 152,962 31,337
2019 54 54 50,139 –2,849 268,095 139,801 31,337
2020 55 55 50,139 –2,775 234,700 126,405 31,337
2021 56 56 50,139 22,250 207,122 112,783 31,337
2022 57 57 50,139 22,525 179,143 98,908 31,337

Note: H = husband; W = wife.

Table 2 details the couple’s total spending, which includes consumption, special expenditures on college, life insurance premiums, and housing. Note that real housing expenditures (expenditures valued in 2005 dollars) decline over time as the assumed 3 percent rate of inflation reduces the purchasing power of nominal mortgage payments. Once the mortgage is fully paid off, housing expenses are $6,000 a year. These expenses are property taxes, maintenance, and homeowners’ insurance, all three of which remain fixed in real terms. Term life insurance premiums first rise and then fall as the household ages. This reflects the increase with age in premium per dollar of coverage as well as the decline over time in recommended life insurance holdings.

Table 3 displays selected years of the couple’s regular asset balance sheet. It shows four things. First, households never exceed their debt limit, which in this case is zero. Second, all assets and income are spent (i.e., the household dies broke if the head and spouse/partner make it to their maximum ages of life). Third, a household’s living standard per equivalent adult is smooth over any interval of years within which the household is not borrowing constrained. Fourth, if consumption, and thus the living standard, is increased in any year, the household will die in debt. This implies that the living standard is maximized.

Table 2. Consumption Smoothing Total Spending

(in constant 2005 dollars)
Special Housing Life Insurance Total
Year H’s Age W’s Age Consumption Expenditures Expenditures Premiums Spending
2005 40 40 $71,852 $ 0 $20,563 $ 983 $ 93,398
2006 41 41 71,852 0 20,139 1,093 93,084
2007 42 42 71,852 0 19,727 1,198 92,777
2008 43 43 71,852 0 19,327 1,302 92,481
2009 44 44 71,852 0 18,939 1,408 92,199
2010 45 45 71,852 0 18,562 1,538 91,952
2011 46 46 71,852 0 18,196 1,678 91,726
2012 47 47 71,852 0 17,841 1,831 91,524
2013 48 48 71,852 0 17,496 2,000 91,348
2014 49 49 61,455 25,000 17,161 2,140 105,756
2015 50 50 61,455 25,000 16,836 2,288 105,579
2016 51 51 61,455 25,000 16,521 2,445 105,421
2017 52 52 50,139 50,000 16,214 2,422 118,775
2018 53 53 50,139 25,000 15,917 2,373 93,429
2019 54 54 50,139 25,000 15,628 2,301 93,068
2020 55 55 50,139 25,000 15,348 2,203 92,690
2021 56 56 50,139 0 15,075 2,119 67,333
2022 57 57 50,139 0 14,811 2,008 66,958

Note: H = husband; W = wife.

Mis-Targeting Spending. Now suppose that in 2005, this household is spending not $71,852 on consumption but 10 percent more (less), namely, $79,037 ($64,667). Furthermore, suppose that this household uses its current consumption to set its retirement and survivor consumption spending levels but that in setting these targets, it appropriately adjusts for changes in the household’s demographic over time and across survivor states. In this case, the household will specify a retirement consumption target that is 10 percent greater (smaller) than $50,139, or $55,153 ($45,125). It will also specify survivor consumption spending targets that are 10 percent higher (lower), on a year-by-year basis, than those generated in this consumption-smoothing run.

How will these targeting mistakes affect the household’s living standard, consumption expenditure, saving, life insurance holdings, assets, and taxes over its life cycle? The answers are provided in Tables 4 through 7. To start, the household’s living standard is shown in Table 4. The fifth column, labeled CS for consumption smoothing, indicates that the household’s living standard per equivalent adult is $31,337 each year. This is the amount of consumption spending that a single adult would need to earn to enjoy the same living standard as she or he enjoys living in the household.

Table 3. Consumption Smoothing Regular Asset Balance Sheet

(in constant 2005 dollars)
Total Regular
Year H’s Age W’s Age Income Spending Taxes Saving Assets
2044 79 79 $51,191 $56,139 $3,783 –$ 8,732 $162,454
2045 80 80 50,937 56,139 3,686 –8,888 153,566
2046 81 81 50,678 56,139 3,584 –9,045 144,521
2047 82 82 50,415 56,139 3,479 –9,203 135,317
2048 83 83 50,147 56,139 3,369 –9,362 125,955
2049 84 84 49,874 56,139 3,257 –9,522 116,433
2050 85 85 49,597 56,139 3,140 –9,683 106,750
2051 86 86 49,314 56,139 3,020 –9,845 96,905
2052 87 87 49,028 56,139 2,896 –10,007 86,898
2053 88 88 48,736 56,139 2,768 –10,171 76,727
2054 89 89 48,440 56,139 2,637 –10,336 66,390
2055 90 90 48,139 56,158 2,502 –10,521 55,869
2056 91 91 47,833 56,253 2,363 –10,784 45,085
2057 92 92 47,518 56,338 2,219 –11,039 34,046
2058 93 93 47,197 56,392 2,070 –11,266 22,780
2059 94 94 46,869 56,368 1,917 –11,417 11,363
2060 95 95 46,536 56,139 1,760 –11,363 0

Note: H = husband; W = wife.

The fourth column presents the household’s living standard path if it overtargets retirement and survivor spending by 10 percent and is thus directed to oversave and overinsure. The result is a $26,289 living standard prior to retirement and a $34,439 living standard after retirement. The ratio of the latter to the former amount is 1.31, indicating that the targeting mistake causes a 31 percent disruption in living standard at retirement. This is a huge discrepancy given the maintained assumption that the household seeks to smooth its living standard.

In the case that the household undertargets retirement and survivor spending by 10 percent, it will spend more before retirement and less thereafter than in the consumption-smoothing case. But as Column 6 shows, the preretirement living standard for this household is not constant. Instead, it rises from $32,673 to $39,109 because of the binding of the household’s liquidity constraint—the assumption is that the household cannot take on any debt beyond the original mortgage balance, which is a constraint that most households operate under. After retirement, the household’s living standard drops to $28,159. The ratio of $28,159 to $39,109 is 0.72, indicating that the household suffers a 28.0 percent reduction in its living

Table 4. Consumption and Saving Recommendations: Consumption Smoothing vs. Mis-Targeting, Living Standard

(in constant 2005 dollars)

Percentage Difference in Living Living Standard Standard Relative to CS
Year H’s Age W’s Age +10% CS –10% +10% –10%
2005 40 40 $26,289 $31,337 $32,673 –16.1% 4.3%
2010 45 45 26,289 31,337 32,673 –16.1 4.3
2015 50 50 26,289 31,337 32,673 –16.1 4.3
2020 55 55 26,289 31,337 32,673 –16.1 4.3
2025 60 60 26,289 31,337 39,109 –16.1 24.8
2030 65 65 34,439 31,337 28,159 9.9 –10.1
2035 70 70 34,439 31,337 28,159 9.9 –10.1
2040 75 75 34,439 31,337 28,159 9.9 –10.1
2045 80 80 34,439 31,337 28,159 9.9 –10.1
2050 85 85 34,439 31,337 28,159 9.9 –10.1
2055 90 90 34,439 31,337 28,159 9.9 –10.1
2060 95 95 34,439 31,337 28,159 9.9 –10.1

Notes: H = husband; W = wife. CS references consumption smoothing, +10 percent references a 10 percent positive targeting mistake; –10 percent references a 10 percent negative mistake.

standard with respect to its pre- and postretirement spending. Again, we see a major disruption at retirement in the household’s living standard because of mis-targeting.5

Because the overspending household was, by assumption, enjoying a 2005 living standard of $34,471 (10 percent higher than $31,337) when it visited its financial planner or used conventional financial planning software, it would have been advised to cut its current living standard by 23.7 percent (1 – 26,289/34,471) to meet its target. The underspending household was enjoying a $28,203 living standard (10 percent lower than $31,337). It would have been told to raise its spending by 15.8 percent (32,673/28,203 – 1). These are big adjustments. It is also ironic that the overspending household is being led to underspend and the under-spending household is being led to overspend.

Clearly, adjusting upward is easier than adjusting downward. Hence, the underspending household is more likely than the overspending household to take the mistaken advice. Thus, the household that may be most concerned with maintaining its living standard in retirement will be led by standard financial planning to experience a 28.0 percent drop in its living standard when it retires—

5A small, but growing, body of research is examining living standard changes at retirement. Bernheim, Skinner, and Weinberg (2001), for example, provide evidence that living standards drop, on average, at retirement by 14 percent, with a median drop of 12 percent.

precisely what it is trying so hard to avoid. Of course, households could well be making targeting mistakes from a situation of currently spending the correct amount: All they need to do is make mistakes in adding up their current expenditures or miscalculate the appropriate demographic adjustments. I refer here to the adjustment for the absence of children in retirement or the absence of a spouse in widow(er)hood. Adjusting for this change in the number and sizes of mouths to feed requires thinking through economies in shared living and the relative cost of children. These are not easy considerations even for very well educated households.

If ±10 percent mistakes occur with equal likelihood, the spread in preretirement living standards for two otherwise identical households could easily be 48.8 percent— the difference in the $26,289 and $39,109 preretirement living standards of households that set their spending targets 10 percent too high and 10 percent too low, respectively. The corresponding postretirement spread in living standards is 22.3 percent. For a source of age-specific consumption inequality, this seems like a good place to look.

Mistakes in Consumption, Saving, and Insurance Recommendations. Table 5 and Table 6 report saving and life insurance recommendations for the three cases. The proportionate consumption differences across the cases are the same as for the living standard differences just discussed. Hence, here I focus on saving and insurance recommendations. As one can quickly see, the differences are tremendous. In 2005, for example, the +10 percent saving recommendation is $11,955 compared with –$810 in the –10 percent case. At age 60, the two saving recommendations are $38,818 and $19,711, respectively. At 75, they are –$13,891 and –$4,240.

Life insurance recommendations in the ±10 percent cases are equally day and night. In 2005, the +10 percent targeting mistake leads to a recommendation of $814,600 in life insurance for the husband whereas the –10 percent targeting mistake leads to a recommendation of only $106,436. Both of these values are far away from $484,947—the desired consumption-smoothing amount. For the wife, the +10 percent mistake leads to a $460,405 term life insurance recommendation for 2005. In contrast, the –10 percent mistake leads to a recommendation of zero life insurance. The recommended consumption-smoothing amount in this case is $128,554. The dramatically higher level of life insurance in the +10 percent case entails dramatically higher life insurance premiums. For example, at age 50, the +10, consumption-smoothing, and –10 households pay premiums of $4,017, $2,288, and $497, respectively. Table 7 shows equally wild disparities in assets and tax payments across the three cases. At retirement, the household with the +10 percent targeting mistake holds $488,412 in assets. This is 1.8 times the consumption-smoothing assets holdings of $270,715 and 3.4 times the $144,405 for –10 percent holdings.

Table 5. Consumption and Saving Recommendations: Consumption Smoothing vs. Mis-Targeting, Saving

(in constant 2005 dollars)

Recommended Consumption Recommended Saving Year H’s Age W’s Age +10% CS –10% +10% CS –10%

2005 2010 2015 2020 2025 2030 2035 2040 2045 2050 2055 2060 40 45 50 55 60 65 70 75 80 85 90 95 40 45 50 55 60 65 70 75 80 85 90 95

$60,278 60,278 51,555 42,062 42,062 55,102 55,102 55,102 55,102 55,102 55,102 55,102

$71,852 71,852 61,455 50,139 50,139 50,139 50,139 50,139 50,139 50,139 50,139 50,139

$74,915 74,915 64,074 52,276 62,575 45,054 45,054 45,264 45,054 45,054 45,054 45,054

$11,955

13,146

–3,619

4,798

38,818 –10,168 –11,894 –13,891 –16,083 –18,611 –21,356 –16,262 $ 1,440

2,458 –12,895

–2,849 31,423 –6,594 –7,350 –8,112 –8,888 –9,683 –10,521 –11,363

–$ 810 520 –13,923 –3,000 19,711 –3,189 –3,719 –4,240 –4,758 –5,278 –5,805 –6,343

Notes: H = husband; W = wife. CS references consumption smoothing, +10 percent references a 10 percent positive targeting mistake; –10 percent references a 10 percent negative mistake.

Table 6. Life Insurance Recommendations: Consumption Smoothing vs. Mis-Targeting, Life Insurance

(in constant 2005 dollars)

Husband’s Life Insurance Wife’s Life Insurance
Year H’s Age W’s Age +10% CS –10% +10% CS –10%
2005 40 40 $814,600 $484,947 $106,436 $460,405 $128,554 $0
2010 45 45 690,432 425,461 112,015 414,861 144,822 0
2015 50 50 582,515 377,052 119,611 384,669 173,952 0
2020 55 55 399,515 234,700 9,982 296,195 126,405 0
2025 60 60 220,819 92,366 0 186,569 55,269 0
2030 65 65 63,139 0 0 85,191 0 0
2035 70 70 85,191 0 0 63,139 0 0
2040 75 75 45,822 0 0 45,822 0 0
2045 80 80 32,261 0 0 32,261 0 0
2050 85 85 22,807 0 0 22,807 0 0
2055 90 90 17,603 0 0 17,603 0 0
2060 95 95 0 0 0 0 0 0

Notes: H = husband; W = wife. CS references consumption smoothing, +10 percent references a 10 percent positive targeting mistake; –10 percent references a 10 percent negative mistake.

Table 7. Assets and Taxes: Consumption Smoothing vs. Mis-Targeting

(in constant 2005 dollars)

Assets Taxes
Year H’s Age W’s Age +10% CS –10% +10% CS –10%
2005 40 40 $ 89,205 $ 78,690 $ 76,440 $32,411 $32,411 $32,411
2010 45 45 152,732 89,159 76,548 34,100 33,114 32,916
2015 50 50 186,022 71,558 51,470 36,735 34,776 34,421
2020 55 55 176,192 22,188 0 38,540 35,812 35,403
2025 60 60 335,873 144,250 66,660 42,868 39,309 38,082
2030 65 65 488,412 270,715 144,405 7,078 4,737 2,640
2035 70 70 423,520 235,480 126,865 6,542 4,488 2,674
2040 75 75 367,143 196,448 106,706 5,914 4,136 2,623
2045 80 80 291,190 153,566 83,951 5,188 3,686 2,493
2050 85 85 203,311 106,750 58,602 4,442 3,140 2,290
2055 90 90 101,868 55,869 30,633 3,421 2,502 2,018
2060 95 95 0 0 0 1,839 1,760 1,679

Notes: H = husband; W = wife. CS references consumption smoothing, +10 percent references a 10 percent positive targeting mistake; –10 percent references a 10 percent negative mistake.

Because higher asset levels imply higher taxable asset income, taxes are highest for the +10 percent household and lowest for the –10 percent household. At age 65, the high, middle, and low tax payments are $7,078, $4,737, and $2,640, respectively. As indicated, these discrepancies in tax and insurance premium payments exacerbate the consumption disruption at retirement and at widow(er)hood that arises from mis-targeting future retirement and survivor spending.

Sensitivity of Consumption Disruption to Assumptions. Consumption disruption caused by mis-targeted spending is likely to be greater for older households because they have fewer years over which to adjust their preretirement spending to avoid the difficulty of spending too much or too little in retirement. Households that are liquidity constrained are likely to set their spending targets too low because their current spending is limited by the availability of their liquid funds. If they retain these targets after their liquidity constraint is relaxed, they will, of course, be induced to undersave and underinsure for the future. Households earning low rates of return will experience larger consumption disruptions because larger adjustments to current spending will be needed to fund a given targeting mistake. Finally, households with higher maximum ages of life will face larger disruptions because one needs to fund targeting mistakes for more potential retirement and survivorship years.
Portfolio “Advice”

Conventional planning’s use of spending targets also distorts its portfolio advice. Given the household’s retirement spending target and portfolio mix, conventional planning runs Monte Carlo simulations that determine the household’s probability of running out of money. These simulations typically assume that households make no adjustment whatsoever to their spending as a result of doing well or poorly on their investments. But consumption smoothing dictates such adjustments and precludes letting oneself get into the position of completely running out of money (i.e., ending up with literally zero consumption). It is precisely the range of these living standard adjustments that households need to understand to assess their portfolio’s risk. Conventional portfolio analysis not only answers the wrong question; it may also improperly solicit risk taking because riskier investments may entail a lower chance of financial exhaustion thanks to their higher mean.

To see this point in its starkest form, take, as an example, a single 60-year-old man named Joe whose only economic resource is $500,000 in assets. Assume Joe’s maximum age of life is 95 and that he faces no taxes of any kind. Suppose Joe sets his spending target at $30,000 a year. Also assume that Joe holds only a properly laddered portfolio of TIPS—Treasury Inflation-Protected Securities (i.e., inflation-indexed bonds)—yielding 2 percent after inflation. These bonds are essentially riskless and permit Joe to consume, at most and at least, $20,413 in today’s dollars each year. What is Joe’s probability of meeting his target—$30,000—each year? It is zero, of course, because spending $30,000 will drive Joe broke unless he fortuitously dies beforehand.

Now, suppose that Joe invests in an S&P 500 Index fund of large-cap stocks rather than in TIPS. Since 1926, the real return on large caps has averaged 9.16 percent on an annual basis.6 If Joe is able to earn this return for sure, he will be able to spend $48,264 a year. But large-cap stocks are risky, with a standard deviation of, say, 20 percent. Nonetheless, there is still almost a 60 percent chance that Joe will be able to spend $30,000 a year. So, if Joe uses a standard Monte Carlo portfolio analyzer, he will find that investing in TIPS fails completely to meet his goal but that investing in stocks gives him a chance to meet his goal. Joe may view this as a good bet given the way this investment outcome information is being presented.

Suppose then that Joe invests all his assets in large caps and then experiences in the next three years the large-cap returns (including dividends) recorded in 2000, 2001, and 2002—specifically, –9.1 percent,–11.9 percent, and –22.1 percent. Will Joe continue to spend $30,000 a year and remain in the stock market given that his wealth after three years has dropped from $500,000 to $238,013? Probably not. At that point, Joe may well switch to holding just TIPS and be forced to live from that point on at about a third of his desired yearly spending rate. In not showing

6This is the average of annual real returns rather than the geometric mean, based on data in Ibbotson’s 2005 yearbook.

such large and sudden adverse potential living standard adjustments, standard financial planning seems to be encouraging more risk taking than is appropriate. This concern is heightened by the prospect of many households being induced as part of the same planning exercise to set their future spending targets at higher levels than are appropriate.
Conclusion

Economics teaches us that we save, insure, and diversify in order to mitigate fluctuations in our living standards over time and across contingencies. Although the goals of conventional financial planning appear consonant with such consumption smoothing, the actual practice of conventional planning is anything but. Conventional planning’s disconnect with economics begins with its first step, namely, forcing households to set their own retirement and survivor spending targets. Setting spending targets that are consistent with consumption smoothing is incredibly difficult, which makes large targeting mistakes almost inevitable. But as shown here, even small targeting mistakes, on the order of 10 percent, can lead to enormous mistakes in recommended saving and insurance levels and to major disruptions (on the order of 30 percent) in living standards in retirement or widow(er)hood.

There are three reasons why small targeting mistakes lead to such bad saving and insurance advice and such large consumption disruptions. First, the wrong target spending level is being assigned to each and every year of retirement and widow(er)hood. Second, planning to spend too much (little) in retirement and widow(er)hood requires spending too little (much) before those states are reached, which magnifies the living standard differences. Third, both saving and insuring the wrong amounts affect tax and insurance premium payments, further exacerbating consumption disruption.

Conventional planning’s use of spending targets also distorts its portfolio advice. Given a household’s spending target and its portfolio mix, standard practice entails running Monte Carlo simulations to determine the household’s probability of running out of money. These simulations assume that households make no adjustment whatsoever to their spending regardless of how well or how poorly they do on their investments. But consumption smoothing dictates such adjustments and, indeed, precludes running out of money (i.e., ending up with literally zero consumption). It is precisely the range of these living standard adjustments that households need to understand to assess their portfolio risk. Conventional portfolio analysis not only answers the wrong question; it may also improperly encourage risk taking because it focuses on the fact that riskier investments may entail a lower chance of financial exhaustion (thanks to their higher mean) rather than on the risk or worst-case scenario.

Purveyors of financial advice have an ethical, if not a fiduciary, responsibility to ensure that the advice they provide is sound. Financially protecting one’s family is a very serious business that requires careful analysis and a real commitment of time. It also requires posing and correctly answering the right question—namely, How can I preserve my family’s living standard through time and in unforeseen, but not unforeseeable, circumstances?

Perhaps a final medical analogy will help explain the situation. None of us would go to a doctor for a 60-second checkup, nor would we elect surgery performed by a meat cleaver over surgery with a scalpel. And any doctor who provided such services would quickly be drummed out of the medical profession. Financial planning, like brain surgery, is an extraordinarily precise business. Small mistakes and the wrong tools can just as easily undermine as improve financial health.
REFERENCES

Auerbach, Alan J., and Laurence J. Kotlikoff. 1987. “Life Insurance of the Elderly: Adequacy and Determinants.” In Work, Health, and Income among the Elderly. Washington, DC: The Brookings Institution.

———. 1991. “The Adequacy of Life Insurance Purchases.” Journal of Financial Intermediation, vol. 1, no. 3 (June):215–241.

Bernheim, B. Douglas, Jonathan Skinner, and Steven Weinberg. 2001. “What Accounts for the Variation in Retirement Wealth among U.S. Households?” American Economic Review, vol. 91, no. 4 (September):832–857.

Bernheim, B. Douglas, Solange Berstein, Jagadeesh Gokhale, and Laurence J. Kotlikoff. 2006. “Saving and Life Insurance Holdings at Boston University—A Unique Case Study.” National Institute Economic Review, vol. 198, no. 1 (October):75–96.

Bernheim, B. Douglas, Katherine Carman, Jagadeesh Gokhale, and Laurence J. Kotlikoff. 2003. “Are Life Insurance Holdings Related to Financial Vulnerabilities?” Economic Inquiry, vol. 41 no. 4 (October):531–554.

Bernheim, B. Douglas, Lorenzo Forni, Jagadeesh Gokhale, and Laurence J. Kotlikoff. 2000. “How Much Should Americans Be Saving for Retirement?” American Economic Review, vol. 90, no. 2 (May):288–292.

———. 2003. “Mismatch between Life Insurance Holdings and Financial Vulnerabilities—Evidence from the Health and Retirement Survey.” American Economic Review, vol. 93, no. 1 (March):354–365.

Kotlikoff, Laurence J., Avia Spivak, and Lawrence Summers. 1982. “The Adequacy of Savings.” American Economic Review, vol. 72, no. 5 (December):1056–1069.