by
Jagadeesh Gokhale
Boston University, National Bureau of Economic Research, and Economic Security Planning, Inc.
and
TIAA-CREF Institute
Laurence Kotlikoff is grateful to the TIAA-CREF Institute and the National Institute of Aging for research support. Economic Security Planning, Inc. is grateful to the National Institute of Aging for research support through a STTR grant. John Ameriks and Stuart Gillan provided helpful comments. The opinions expressed here are those of the authors and not necessarily those of TIAA-CREF.
The conventional approach to retirement and life insurance planning, which is used throughout the financial planning industry, differs markedly from the economic approach. The conventional approach asks households to specify how much they want to spend before retirement, after retirement, and in the event of an untimely death of the head or spouse. It then determines the amounts of saving and life insurance needed to achieve these targets. The economic approach is based on the life-cycle model of saving. Its goal is to smooth households’ living standards over their life cycles and to ensure comparable living standards for potential survivors. In the economic approach, spending targets are endogenous. They are derived by calculating the most the household can afford to consume in the present given that it wants to preserve that living standard in the future.
Although spending targets under the conventional approach can be adjusted in an iterative process to approximate those derived under the economic approach, there are practical limits to doing so. This is particularly the case for households experiencing changing demographics or facing borrowing constraints. This paper illustrates the different saving and insurance recommendations provided by economic financial planning software and the practical application of traditional financial planning software. The two software programs are Economic Security Planner (ESPlanner), developed by Economic Security Planning, Inc., and Quicken Financial Planner (QFP), developed by Intuit. Each program is run on 24 cases, 20 of which are stylized and 4 of which are actual households.
The two software programs recommend dramatically different levels of saving or life insurance in each of the 24 cases. The different saving recommendations primarily reflect ESPlanner’s adjustment for household demographics and borrowing constraints. The different life insurance recommendations reflect these same factors as well as ESPlanner’s accounting for contingent household plans and for Social Security’s survivor benefits. The less detailed tax and Social Security retirement benefit calculations used in our implementation of QFP also explain some of the differences between the two programs.
The conventional approach to retirement and life insurance planning asks households how much they want to spend after retirement and in the event of the untimely death of the head of household or spouse. It then determines the amounts of saving and life insurance needed to achieve these targets. This approach, which is used throughout the financial planning industry, has received remarkably little attention from economists. This is surprising not only because of the practical importance of traditional financial planning, but also because of its apparent differences with the economic approach.
The economic approach is based on the life-cycle model of saving developed by Ando and Modigliani (1963) and the canonical model of life insurance developed by Yaari (1965). The goal of the economic approach is to smooth households’ living standards over their life cycles and to ensure comparable living standards for potential survivors. In the economic approach, spending targets are endogenous. The targets are derived by calculating the most that each household can afford to consume in the present given that it wants to preserve that living standard into the future. And as stressed by Hubbard and Judd (1987) among others, a household’s ability to consume may in the short term also be circumscribed by its ability to borrow.
Although spending targets under the conventional approach can be adjusted to approximate those derived under the economic approach, there are practical limits to such “trial and error” re-targeting. Because of the complexity of the relevant factors, these limits especially exist for households experiencing changing demographics, enjoying economies to shared living, or facing borrowing constraints. This paper illustrates the different saving and insurance recommendations provided by economic financial planning software and conventional financial planning software. The two software programs are Economic Security Planner (ESPlanner) (the economic financial planning software), developed by Economic Security Planning, Inc., and Quicken Financial Planner (QFP) (the conventional financial planning software), developed by Intuit, Inc. Each program is run on 24 cases, 20 of which are stylized and 4 of which are actual households.
We used both programs in a manner that did not require data or information external to the program beyond that provided in direct interrogatives from the programs. In particular, for QFP, we tried to emulate how a somewhat sophisticated household might use the program. QFP begins by asking a household what it is currently spending and whether it wants to continue spending that amount in the future. We expect that most households would answer the latter question in the affirmative. Next QFP asks the household to enter earnings, net worth, and a variety of other data. Finally, QFP determines if the household’s specified time-path of expenditure is feasible. If its plan puts the household into debt (or further into debt if the household started in debt) at any point in the future, QFP tells the household that its plan has failed. A plan can also fail if there are insufficient resources to finance consumption in retirement.1 By contrast, if a household saves lots of money, it will almost certainly get a passing grade. We then expect that the household would adjust its initial consumption spending to, as closely as possible, “die broke”, that is, to end up with zero financial net worth at the end of its planning horizon. Thus, in running QFP on the 24 cases, we choose the level of
By this we mean that the household’s present value of resources doesn’t suffice to pay for its present value of its targeted consumption expenditure, including its post-retirement consumption expenditure. As mentioned, the program will “fail” even if this restriction is satisfied, if the targeted trajectory of expenditure puts the household into debt or further into debt prior to reaching retirement.
consumption expenditure by iterating until the household’s terminal net worth is close to zero. This “trial and error” iteration process is time consuming, which reinforces our view that even sophisticated households are unlikely to further fine tune their expenditure plans to deal with demographic change, particularly the arrival and departure of children from the household, or with borrowing constraints.
The two software programs recommend dramatically different levels of saving and life insurance in each of the 24 cases with the discrepancies between the two sets of recommendations generally increasing with the complexity of the case. In some cases ESPlanner recommends substantially more saving in early years and substantially less saving in later years than does QFP. In other cases, the opposite is true. The differences in life insurance recommendations are more systematic, with ESPlanner generally recommending significantly less life insurance than QFP.
The different saving recommendations primarily reflect ESPlanner’s adjustments for household demographics, economies in shared living, and borrowing constraints, as well as its different, and more detailed, approach to the calculation of federal and state income taxes and Social Security retirement benefits. The two programs’ different life insurance recommendations reflect these factors, ESPlanner’s contingent planning, and ESPlanner’s integration in its life insurance calculations of Social Security’s survivor benefits.
Our comparison of ESPlanner and QFP illustrates some, but certainly not all, of the differences in the conventional and economic approaches. There are a plethora of financial planning software programs adopting the conventional approach. Many of these programs have specific features that differ from those of QFP. Hence in comparing ESPlanner with just QFP, we may be over- or understating typical differences between the two approaches. In addition, although ESPlanner captures the essential items of what economists would stress in financial planning, it does not incorporate labor earnings uncertainty, rate of return uncertainty, and other non-life contingencies that can influence life-cycle consumption choice.2 ESPlanner does not explicitly consider the premiums, paid whether by the employer or by the household, for health and long-term care insurance, nor the expenditures arising from uninsured health and long-term care contingencies. Nor does ESPlanner optimize, subject to legal constraints, contributions to tax-deferred retirement accounts. Furthermore, although ESPlanner considers life contingencies for the purpose of determining the optimal amount of life insurance, it does not evaluate the optimal amount of life annuities to hold in the retirement period of the life cycle. Hence in comparing ESPlanner only with QFP, which also does not consider these uncertainties and factors, we may be understating the differences between the two approaches.
The paper proceeds as follows. Sections II and III describe ESPlanner and QFP, respectively. Section IV summarizes the main conceptual and technical differences between the two programs. Sections V and VI compare ESPlanner’s and QFP’s recommendations for the 20 stylized and 4 actual cases, respectively. Section VII summarizes the paper and draws conclusions.
See Hubbard, Skinner, and Zeldes (1995) for a treatment of consumption choice in the presence of lifespan and earnings uncertainty and Campbell et al. (1999) for a treatment of consumption choice in the presence of rate of return uncertainty.
The economists Douglas Bernheim, Jagadeesh Gokhale, and Laurence Kotlikoff and software engineer Lowell Williams established Economic Security Planning, Inc. to develop ESPlanner – a software package whose primary goal is to foster appropriate saving and insurance decisions.3 Their stimulus was the findings in Kotlikoff, Spivak, and Summers (1982), Auerbach and Kotlikoff (1987, 1991), Bernheim (1991, 1995), and other studies that a significant fraction of households undersaves and underinsures. Undersaving in these studies means that a household can’t sustain its current living standard in the future, and underinsuring means it can’t sustain its current living standard if the household head or spouse were to die. ESPlanner’s objective is to permit households to achieve the highest living standard that they can afford to sustain both through time and in the event of the early death of the head or spouse.
Unlike many other financial planning programs, ESPlanner’s life insurance, consumption, and saving recommendations are fully integrated. The program’s consumption and saving recommendations take into account the need to pay life insurance premiums, and its life insurance recommendations are set to ensure the same living standard through time for survivors as the consumption time-path that the otherwise intact household is able to afford.
What Factors Does ESPlanner Consider?
3 The three economists are also using ESPlanner under two National Institute of Aging grants to study the adequacy of saving and insurance. The National Institute of Aging also supported research on ESPlanner through an STTR grant. Economic Security Planning, Inc. is willing to provide academic researchers with copies of the program for free. To contact the company, go to www.ESPlanner.com.
In determining the extent to which a household can smooth its living standard, ESPlanner takes into account the maximum amount of money the household says it can borrow apart from housing-related debt. It also considers the current and future labor, pension, social security, inheritance, and other income the household will receive and the federal and state income taxes and federal payroll taxes that it will pay. The Appendix describes ESPlanner’s tax and Social Security benefit calculations in detail. ESPlanner also takes account of 401k, 403b, and other tax-favored saving vehicles, housing plans, special expenditures, estate plans, and preferences about how the household would, if not borrowing constrained, like its living standard to change through time.4 Finally, ESPlanner recognizes that a household’s expenditures do not directly translate into its standard of living. Adjustments are made for household composition and household “economies of scale” -- the fact that people can live more cheaply together than apart.5 To be precise, ESPlanner provides for children until they reach age 19, and takes into account that children may cost more or less than adults and that the relative costs of children can vary by age. It also adjusts for the number of adult equivalents based on a user-specified degree of economies to scale in shared living.
ESPlanner’s Recommendations
ESPlanner’s principal outputs are recommended time-paths of consumption expenditure, taxable saving, and term-life insurance holdings (for each spouse in the case
4
“Borrowing constrained” refers to a household’s inability to get credit on the security of its future anticipated earnings, for example, through a credit card or an unsecured line of credit. “Liquidity constrained” often refers to that as well as to the inability to sell assets (whether financial or real) to finance consumption.
5
From the perspective of economic theory, the household is viewed as maximizing a Leontief intertemporal utility function with year-specific time preference and demographic weights subject to borrowing constraints and non-negativity constraints on life insurance.
of married households). All outputs are displayed in current-year (i.e., real) dollars. Consumption in this context is everything the household gets to spend after paying for its “off-the-top” expenditures – its housing expenses, special expenditures, life insurance premiums, taxes, and net contributions to tax-favored accounts. As mentioned, the amount of recommended consumption expenditures 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. Finally, recommended household consumption will change over time if users intentionally specify that they want their living standard to change. For example, if users specify that they desire a 10 percent higher living standard after a certain year in the future, the software will incorporate that preference in making its recommendations, provided that it does not violate a borrowing constraint.
ESPlanner’s recommended taxable saving in a particular year equals the household’s total income (non-asset plus asset income) in that year minus that year’s sum of (a) recommended spending on consumption and insurance premiums, (b) specified spending on housing and special expenditures, (c) taxes, and (d) net contributions to tax-favored accounts (contributions less withdrawals).
ESPlanner’s recommendations for annual term insurance are either positive or zero. If recommended term insurance is positive for a particular potential decedent (the household head or, if married, spouse) in a particular year and if the decedent dies at the end of that year, the surviving household will have precisely the same living standard as the household would have had absent the decedent’s premature death. If the potential decedent’s recommended insurance in a particular year is zero, the surviving household will have the same or higher living standard if the decedent dies in that year. These statements are, of course, conditional on the household actually buying the amounts of life insurance being recommended and on the correctness of its assumptions and information concerning future income, current asset holdings, rates of return, special expenditures, and so forth.
Checking ESPlanner’s Recommendations
ESPlanner’s algorithm is very complicated. But users of the software can check ESPlanner’s reports to see that, given their data inputs, preferences, and borrowing constraints, the program recommends the highest and smoothest possible living standard over time. Take, as an example, Case 1 -- the first and simplest of our 20 stylized cases. It involves a Massachusetts couple in which the husband, George, is initially age 29 and the wife, Jane, is initially age 27. George works through age 65 earning $50,000 each year in 1999 dollars. Jane doesn’t work. They have no children, no special expenses, no housing expenses, no estate plans, no private defined benefit pension income, no self-employment income, no special receipts, no tax-favored saving accounts, no employer-funded defined contribution accounts, and no coverage under Social Security.
Turn, now, to Table 1, which presents the non-tax-favored balance sheet for Case
1. All values reported in the table and all other ESPlanner reports are in 1999 dollars.7
6 Negative life insurance is formally identical to the purchase of an inverted life annuity, that is, the receipt of annual payments for life purchased by, in a predetermined lump sum amount, the estate of the deceased. 7 ESPlanner produces the following main reports: current recommendations, annual recommendations, non tax-favored balance sheet, income, spending, non-asset income (for each spouse), housing, taxes, tax-favored balance sheets (for each spouse), estate reports (for each spouse and for couples if both spouses die in the same year), social security benefit reports (for the household and for each spouse). ESPlanner’s survivor reports are essentially the same as its main reports.
Because Table 1 is a balance sheet, the changes from one year to the next in non tax-favored net worth equal the latter year’s non tax-favored saving. Apart from this adding up property, note that terminal non tax-favored net worth is zero, that is, the household’s recommended time-path of spending precisely exhausts the household’s economic resources assuming each spouse lives as long as possible. In this particular case, household spending simply includes consumption, life insurance premiums, and funeral expenses of $5000 for each spouse. In general, spending also includes special expenditures and housing expenses.
This balance sheet tracks the evolution of the household’s non-tax-favored net worth assuming the household head and spouse both live to their maximum ages of life – assumed to be age 95. As just suggested, ESPlanner’s planning horizon extends through the maximum ages of life of the household head and, if married, his or her spouse. The emphasis here on the maximum, rather than the expected, length of life is a part of ESPlanner’s general philosophy, namely to plan conservatively.
Consumption spending in this case is constant at $25,784 until George reaches age 95. It then declines to $16,115 – the amount Jane gets to spend on herself after George has passed away. The former value is 1.6 times the later value reflecting the assumed degree of economies in shared living, namely that two can live as cheaply as
1.6. Hence, the household’s living standard remains the same until its last possible year – when Jane reaches age 95. Table 1’s total spending amounts differ from $25,784 when both George and Jane are alive by annual amounts equal to the life insurance premiums on George’s recommended holdings of life insurance.
The fact that (a) the household just exhausts its resources when Jane dies at age 95 (i.e., does not die in debt) and that (b) the recommended consumption expenditure path entails a uniform living standard for all household members in each year the household exists means that ESPlanner has maximized the household’s sustainable living standard given the household’s economic resources. That is, raising consumption spending in each year to produce a slightly higher uniform living standard would lower saving in each year and leave the household in debt in its last period of existence. Married users can also check ESPlanner’s survivor reports to verify that surviving spouses who follow ESPlanner’s life insurance recommendations will, if the household’s inputs assumptions prove correct, be able to maintain as high a living standard as the household would have had if the spouse had not died early.
The survivor reports for Jane (George) can be produced for any hypothetical ages at which George (Jane) might die. As an example, Table 2 shows Jane’s survivor non tax-favored balance sheet if George dies at age 51. Note that Jane’s spending as a survivor, which, in this case, consists only of consumption, equals $16,115 – precisely
62.5 (1 divided by 1.6) percent of the amount Jane and George would jointly spend onconsumption were George not to die. Hence, Jane’s recommended consumption as a widow entails the same living standard as she would have had when married. Table 2 also shows that Jane can, as a survivor, afford to spend this amount each year on consumption and not end up in debt even if she lives to her maximum age of life, 95. Also note that the initial amount of non tax-favored wealth with which Jane begins widowhood is $518,691. This amount equals (up to rounding error) the sum of George’s and Jane’s $382,833 in non tax-favored assets at age 50, plus the $140,858 amount of term insurance recommended for George at age 50, less $5000 for George’s funeral. Were ESPlanner to recommend even a dollar less in life insurance for George at age 50, Jane would not be able to finance the same living standard as a survivor without ending up in debt at age 95 if she were to live that long.
There is an expanding universe of financial planning software programs and Internet sites available to the general public that offer advice on saving for retirement and other goals, and on asset allocation. These programs include proprietary packages offered by mutual fund and insurance companies with investment and insurance products to sell, as well as independent packages marketed by financial experts and consumer software companies. In addition, there are packages sold to professional financial advisors who use these programs in the course of financial planning sessions given to clients.
The packages on the market differ in focus, sophistication, and level of detail. As mentioned above, however, they all share certain conceptual features. A goal is set by the household for the desired income or expenditure flow in retirement (usually a set percentage of earned income expected to be received just prior to retirement). Then the appropriate saving rate (usually assumed to be level over the work life) and optimal asset allocation are calculated by the program, given certain assumptions, preferences, and information provided by the household, to enable the household to reach its goal. This conventional approach is consistent with the way actuaries traditionally have designed pension plans, that is, a retirement income goal (a replacement rate) is set and the contribution rate (usually level) necessary to achieve that goal is calculated.
As representative of the conventional approach, we employ the Quicken Financial Planner (QFP), manufactured by Intuit, Inc., makers of Quicken and TurboTax software. To our knowledge, QFP is the most sophisticated and comprehensive planning package available to the general public; it rivals professional software in its detail and use of advanced techniques. QFP seems appropriate for households at most income and wealth levels, perhaps with the exclusion of those households with unusual financial circumstances and those at the bottom and the very top of the income and wealth distributions. (QFP, and ESPlanner for that matter, provides little detail on government assistance programs and estate taxes.)
Inputs and Outputs of QFP
A detailed description of the QFP is provided in the Appendix of Warshawsky and Ameriks (1999, forthcoming). Here we summarize that discussion and highlight certain features of QFP that we employ in our analysis.
QFP is designed to aid in financial planning for retirement, children’s college education, and life insurance needs; we ignore here its advice on asset allocation. Basic demographic and economic (both current and expected future) information is collected about the household. In the “Income Taxes” section, the respondent household inputs combine federal and state average income tax rates expected in the periods before and after retirement. The respondent is given a choice of two methods to do this estimation:
(1) the “demographic average” – the average tax rate paid by the average person in therespondent’s state of residence within the same range of household income and demographic situation, or (2) the “tax return” approach – the average tax rate calculated based on the household’s adjusted gross income and the actual taxes paid in the prior year. Because we do not know actual taxes paid in the prior year, we employ QFP’s demographic average method in our analysis; an actual user might input his or her own tax information using the tax return approach.
QFP asks for a myriad of details concerning the household’s current and expected future pension coverage, current investment, real estate holdings, loans, and mortgages outstanding. In the “Living Expenses” section, the respondent household is given a choice of two estimation methods: “Rough Estimate” and “Itemized List”. In the first method, QFP simply asks for the household’s estimated living expenses before and after retirement, and offers as guidance an abbreviated statement for the current year of cash flow less “off-the-top” expenditures like taxes, loan payments, housing expenses, and planned contributions to pension and savings plans. In the second approach, the household fills out a detailed budget for the current and future years, and the QFP sums up pre- and post-retirement living expenses from this list. We employ the first method.
In the “Social Security estimated benefits” section, QFP again gives the user two choices in estimation methods: “Rough Estimate” and “Social Security Administration Estimate.” In the first method, the respondent household is asked to sort itself into one of four earnings bands; an estimate of Social Security benefits is then generated based on planned age at retirement. In the second method, the respondent is simply invited to input the number from the official response to the Request for Personal Earnings and Benefits Estimate Statement form mailed to the Social Security Administration. We employ the first method in our analysis.
QFP has certain advanced planning options allowing for more sophisticated modeling. In particular, in “Cash Shortfalls”, the respondent is asked whether it plans to sell investments to cover pre-retirement shortages; although the default answer, which is conservative, is no, in our analysis, we respond yes.8 In “Realized Gains”, the respondent is asked to estimate the percentage of the gains in the taxable investment portfolio subject to taxes every year; we employ the default answer of 100 percent. In “Sweep”, the respondent is asked what percent of surplus cash flow is swept into taxable savings; we answer 100 percent, although QFP suggests a more conservative answer of 0%. QFP defines surplus cash flow as the excess of planned sources over planned uses of annual money flows. With the cash shortfalls and sweep options employed, the QFP is able to produce non-constant savings recommendations; these options are unusual in the conventional approach and narrow the contrast somewhat with the economic approach.
Another helpful feature of QFP is its comprehensive information about the current cost of college. Using data from the College Board, QFP reports, by state, for public and private colleges and universities and community colleges, the cost of (in-state and out-of-state) tuition, room and board, books, fees, and other expenses. A worksheet is also available to input expected financial aid, the student’s own income, and gifts from relatives. This feature would seem to be especially valuable to households with children
8 Regardless of the answer to the cash shortfall question, QFP uses the following rules regarding sales of investments. After retirement, the mandatory minimum distributions from tax-deferred retirement accounts are made first. Then taxable investments are sold. Finally, if more resources are still needed, tax-deferred accounts are again drawn upon. Before retirement, shortfalls due to special expenses or home down payments are funded, first by the sale of taxable investments and then by withdrawals from tax-deferred accounts (along with the payment of tax penalties, if applicable). If the answer to the cash shortfall
approaching the age of college entrance with knowledge about the type, or the specific identity, of the college their children will attend. Because all of the cases we examine involve young families we did not employ this feature of QFP which is not available in ESPlanner.
QFP processes all this information in an easy-to-see fashion. It initially gives one result: whether or not the household can anticipate having the money needed to retire. If the news is bad, a brief statement of the nature of the problem is given, including the year when net worth becomes negative or assets are depleted. This is the manner in which QFP alerts users to infeasible plans as well as to violations of borrowing constraints. If the household provides QFP with an initial level of expenditure and tells it to maintain that expenditure level through time and if that expenditure level leads the household in any future year into debt or into more debt than the household initially had or insufficient assets to finance retirement, QFP will announce that the household’s plan has failed.
Once users have generated a plan that hasn’t failed, they can do “What If” analysis. Specifically, users can change, one at a time, key assumptions and items of information, such as retirement age or pension contribution rate, and see whether their original consumption expenditure plan is still feasible with the new inputs.
QFP’s Insurance Planner uses data already inputted on the household’s demographic and economic makeup, and asks the respondent for its preferences on the size of a desired estate and other post-mortem expenses, such as children’s college tuition and the spouse’s living expenses, to be covered. The Insurance Planner then advises on the optimal amount of term insurance to hold in the first year of the financial plan. The
question is yes, then shortfalls due to taxes, living expenses, loan payments, and planned saving are covered by the sequential sale of taxable and tax-deferred investments.
- 17
Insurance Planner is semi-autonomous; none of its outputs, including the premiums on its recommended insurance coverage, is used in the main corpus of QFP.
There are three main conceptual differences between the conventional approach to financial planning represented by QFP and the economic approach represented by ESPlanner. These are the determination of expenditure targets, the treatment of demographics and economies of scale, and the handling of borrowing constraints.
Smoothing a household’s standard of living refers to maintaining the same living standard for each person in the household at each point in time. As mentioned above, ESPlanner does this for the household, subject to the household’s resources (including social security), its exogenous housing expenses, its exogenous special expenses, its endogenous tax liabilities, and its exogenously specified non-mortgage debt limit. In solving for the time-path of consumption that smooths, to the maximum feasible and desired degree, the household’s living standard, ESPlanner is endogenously determining the household’s expenditure targets as well as the amounts the household must save, in non tax favored form, and insure to achieve these targets.
By contrast, QFP asks households to set their own expenditure targets. It then accumulates the household’s excess cash flow (income less expenses less all planned saving) and sees whether the accumulated sum ever becomes negative given its financial plans (or, if the household was initially in debt, whether its debt becomes more negative than the initial level of debt). If the household’s accumulated cash flow never goes into the red (or too far into the red) and if enough assets are available to finance consumption in retirement and to meet other user-determined goals, the household’s expenditure plan is scored a success. Indeed, QFP effectively scores such a plan a complete success in that it doesn’t encourage users to adjust their expenditure target to keep them from, in effect, leaving money on the table. One defense of this practice is that because households face lots of future risks, planning to leave money on the table which can be used on future rainy days is not such a bad thing. The counter argument is that the amount of money being left on the table may be far smaller or far greater than the household feels it needs to protect itself against future risks. Because, QFP, like ESPlanner, permits users to treat the accumulation of an emergency fund and bequests as special expenditures, the inclusion of these items would seemingly obviate the need to plan for unspent funds.
As mentioned in the introduction, in preparing this paper we ran QFP in a manner that avoids leaving money on the table. The procedure involves (a) setting an initial level of consumption expenditures, (b) telling QFP that we want to maintain the same level of expenditure through time, and (c) iterating on this initial level of consumption expenditure until the household has very few assets remaining at the end of life.
The second conceptual difference involves the goal of smoothing through time household members’ living standards. ESPlanner does this by determining, at each point in time, the number of adult equivalents and adjusting for economies in shared living. QFP, in contrast, invites users to make these adjustments for themselves in setting their consumption expenditure targets.
The third conceptual difference is the treatment of borrowing constraints. ESPlanner builds borrowing constraints directly into its consumption expenditure recommendations. QFP, on the other hand, tells users whether their exogenously specified consumption trajectory will, at some point, violate the program’s implicit zero constraint on future borrowing. QFP users whose plans are failed by the software because they run into borrowing constraints are free to rerun the program by lowering their current and raising their future consumption. But users might also respond by lowering their initial consumption and continuing to tell QFP that they want to spend the same amount in the future as in the present. In so doing, they will arrive at an expenditure plan that passes QFP’s feasibility criterion, but leaves money on the table.
Interestingly, in the 24 cases examined below, we never encountered negative net worth in our use of QFP, whereas we quite often encountered borrowing constraints in our use of ESPlanner. The reason is the interaction of demographics with borrowing constraints. In those cases in which ESPlanner generates borrowing constraints, the household has young children and, therefore, higher consumption expenditure needs when young than when old. Our application of QFP ignores this. Consequently, compared to economic theory, our practical application of QFP set consumption expenditure too low when the household is young and too high when the household is old. Because the households we examine have relatively high housing and special expenditures when young, the inappropriately low short-run QFP consumption expenditures are needed to avoid short-run borrowing constraints. Also, in order to make the results of ESPlanner and QFP as comparable as possible, we set the maximum amount able to be borrowed in ESPlanner to zero.
In addition to these conceptual differences, there are several technical differences between the two programs. First, ESPlanner formulates more precise estimates of federal and state income taxes as well as Social Security benefits compared to QFP’s average approach. Second, QFP apparently does not adjust for inflation in the first year of planning, whereas ESPlanner does. This may, however, simply reflect a difference in the timing conventions of the two programs. Third, ESPlanner integrates life insurance with the financial plan; premiums need not be entered as an itemized expense. Fourth, ESPlanner assists users in comparing their actual current saving with the amount being recommended, whereas QFP encourages users to compare their actual current expenditures with the amount they ultimately end up targeting.
The main table of this paper, Table 3, provides a detailed comparison of the output of the two software programs. The first 20 situations, labeled Case 1 through Case 20, are based on the same stylized household considered above, in which the husband, George, is aged 29 and the wife, Jane, is aged 27. George earns $50,000 per year in 1999 dollars and expects to retire at age 65. Each case adds some demographic or economic complexity to their initial situation. Cases 21 through 24 consider four actual households with quite different economic and demographic circumstances. To save space, we present results only for selected years of life, and only for key variables (stated in inflation-adjusted terms), namely, consumption, taxable savings, taxes, tax-deferred and taxable assets, and life insurance. We use ESPlanner’s definitions of these variables, and had to combine several items from QFP’s output tables to make the numbers comparable.
For all cases and for both Planners, we make the following economic assumptions, unless otherwise stated. The expected nominal rate of return on investments (whether taxable or tax-favored) is 6 percent, and the expected general inflation rate is 3 percent. When relevant, the expected rate of inflation for college tuition is assumed to be 5 percent. Federal and state tax rates and rules, as well as government benefit programs (mainly Social Security), are expected to remain unchanged in the future. All households spend their entire lives in Massachusetts. Version 3.3 (r4) of QFP and version 1.10.37 of ESPlanner were used. Both programs were run with the computer’s date set to January 1, 1999.
ESPlanner’s output for Case 1 was discussed above. Recall that the household’s living standard is perfectly smoothed with the couple spending $25,784 on consumption until the husband reaches age 95 and passes away. Thereafter, the surviving spouse, Jane, spends $16,115 on consumption, which suffices to maintain the same living standard she enjoyed when married. These and all other dollar figures discussed below are quoted in 1999 dollars.
QFP recommends a somewhat higher level of consumption expenditure in Case 1: $27,810 when both George and Jane are alive and $16,686, when Jane is surviving by herself. It should be recalled, however, that premiums for life insurance are an “off-the-top” expenditure in ESPlanner, whereas they are (implicitly) included in living expenses in QFP. Although their 1999 income levels are identical, the two programs recommend quite different amounts of non tax favored saving for 1999. ESPlanner recommends about $1,400 more non tax-favored saving in 1999 than QFP! The reason is that QFP calculates higher levels of taxes for 1999 than does ESPlanner. Although QFP generates higher taxes than ESPlanner when the couple is young, it generates lower taxes than ESPlanner when the couple is middle-aged and older.
In considering these differences, it’s important to bear in mind that QFP is making approximations in calculating taxes. Specifically, QFP is applying a single average tax rate over time, independent of the household’s expected schedule of taxable income over time. In contrast, as described in the Appendix, ESPlanner makes very precise federal and state tax calculations that include determining whether or not the household should itemize its deductions, computing the household’s taxable income, and using exact federal and state tax rate schedules to determine the household’s tax liabilities. Correct calculation of state taxes is particularly important as households age and accumulate taxable capital income. The fact that ESPlanner produces much higher taxes when the couple has significant non tax-favored asset income reflects its incorporation of the very high rate of capital income taxation by the State of Massachusetts. This rate is roughly two times higher than the rate levied on labor income.9
Although the two programs recommend significantly different 1999 levels of non tax-favored saving, their 1999 life insurance recommendations are nearly identical. Neither program recommends life insurance for the non working wife. QFP’s recommended life insurance for the husband is $680,000. ESPlanner’s is $667,948. ESPlanner, but not QFP, provides life insurance recommendations for all future years as well: The husband’s recommended level of life insurance declines over time, and is zero after he reaches age 57. At age 45, the recommended amount is $293,459.
9 The Massachusetts legislature is currently considering lowering the tax on personal capital income to the same rate as is levied on labor income.
Case 2 assumes that Jane works through her 64th year of life and earns $25,000 per year in 1999 dollars. Because the couple has greater economic resources, it can sustain a higher living standard, which is what each program recommends. However, the two programs now differ by almost $6,000 annually with respect to the level of the couple’s sustainable living standard. ESPlanner has George and Jane consuming $35,883 when both are alive, compared with QFP’s $41,720, a 16 percent difference. Again, tax calculations and, to a lesser extent, life insurance premiums appear to explain this difference. ESPlanner’s 1999 taxes are almost $3,000 higher than QFP’s, and ESPlanner’s taxes are significantly higher when the couple is middle aged and approaching retirement. For example, when George is 45, ESPlanner’s taxes are almost $8,000 higher. These differences appear to reflect ESPlanner’s inclusion of the Massachusetts’ 12 percent tax on non tax-sheltered capital income as well as our use in QFP of the “demographic average” method.
Associated with these differences in recommended consumption are differences in recommended non tax-favored saving. ESPlanner recommends 1999 non tax-favored saving be set at $21,937, whereas QFP recommends it be set at $20,069. When George is age 45, ESPlanner’s recommended non tax-favored saving is much less than QFP’s --$24,352 versus $32,354. At George’s age 65, ESPlanner’s recommends much more dissaving than does QFP -- $21,975 versus $13,491. Thus, compared with QFP, ESPlanner recommends slightly more saving when George and Jane are young, and less saving in middle age and the start of old age, when George and Jane face higher taxes than QFP says they will face.
The non-trivial differences between the two programs in saving recommendations and tax estimates eventuates in non-trivial differences in accumulated assets when the couple reaches retirement age. According to ESPlanner’s plan, the couple has $871,921 in taxable assets when George reaches 65. According to QFP’s plan, the couple taxable assets in George’s 65th year are almost 15 percent larger -- $996,629.
Turning to the life insurance recommendations, both programs tell George he needs less life insurance because Jane now has earnings that can help sustain her if George dies prematurely. ESPlanner is now recommending slightly more insurance than QFP, but the difference is quite small.
Case 3 adds the assumption that both George and Jane are in Social Security-covered employment, and are, therefore, paying payroll taxes and can expect benefits at levels currently legislated. According to ESPlanner, George and Jane receive $18,843 and $10,515, respectively, in Social Security retirement benefits when each reaches age 65. These benefit levels are substantially higher than the corresponding $10,049 and $5,266 figures calculated by QFP. The differences here apparently reflect the wage indexation of Social Security earnings histories by ESPlanner. (ESPlanner follows very precisely the rules for determining Social Security benefits as detailed in the Social Security Handbook.10)
10 ESPlanners’s calculation of the “primary insurance amounts” for the earning members of the household, on which all benefits are based, are identical to those produced by Social Security’s own ANYPIA program.
Because ESPlanner expects the couple to receive significantly more in future Social Security benefits than does QFP, it recommends less saving prior to retirement. Indeed, ESPlanner’s 1999 recommended level of non tax-favored saving of $9,629 is one third lower than QFP’s recommendation of $14,257. The couple’s sustainable living standard is higher in Case 3 than Case 2 in both programs. For example, with Social Security, ESPlanner recommends that the couple consume $43,737 each year that both George and Jane are alive. Without Social Security it recommends they consume $35,883.11
Case 3’s life insurance recommendations are not much different from those of Case
2. This is not surprising. On the one hand, the availability of Social Security survivor benefits reduces the need for life insurance for George. On the other hand, the couple’s higher living standard when both spouses are alive means that Jane’s needs as a survivor are greater.
This case builds on Case 3 by adding housing. Specifically, we assume that George and Jane own a home with a market value of $200,000, annual property taxes of $2,500, annual maintenance of $1,500, annual homeowners insurance of $400, and a 30-year mortgage with remaining balance of $170,000 and annual principal and interest mortgage payments of $1,124 (the mortgage interest rate is 8 percent). The couple is assumed to remain in their house until the end of their lives.
11 This difference may suggest that Social Security is, on balance, a plus for George and Jane in terms of their sustainable living standards. But this is not necessarily so because in Case 2 we did not ascribe to George and Jane the contributions their employers would otherwise be making on their behalf to Social Security. Those contributions represent a form of compensation to George and Jane that would, in the absence of Social Security, presumably be paid directly to them.
The need to pay housing expenses reduces the amount of resources that can be devoted to consumption. The consumption decline is over $10,000 in QFP and over $8,000 in ESPlanner. The difference between the two programs is again due to taxes. Because ESPlanner takes into account the deductibility of mortgage interest payments, ESPlanner’s taxes decline by almost $3,000 annually in moving from Case 3 to Case 4. In contrast, QFP’s taxes using the demographic average rates are essentially unchanged.
The need to pay off the household’s mortgage and other housing expenditures raises recommended life insurance for George under both programs. And both programs now recommend roughly $60,000 in life insurance for Jane. Intuitively, unlike Jane’s consumption, which disappears if she dies, the household’s housing expenditures are no different when Jane passes away. Because Jane’s income contributes to meeting these expenditures when Jane is alive, insuring that income is important.
In Case 5 we add two children – one born in 1997 and the other born in 2001. Because QFP doesn’t adjust for household demographics, its recommended annual levels of consumption and taxable saving are unchanged by the presence of children. Stated differently, our application of QFP involves smoothing consumption expenditures over the household’s life cycle, rather than smoothing its living standard. Smoothing a household’s living standard requires spending more (although not proportionately more) on consumption in those years when the household has more members.
Unlike QFP, ESPlanner’s objective is to smooth households’ living standards to the extent the household desires a smooth living standard and to the extent such smoothing doesn’t violate the household’s borrowing constraint. Hence ESPlanner’s consumption recommendations take into account the assumed changes over time in household composition. In particular, ESPlanner assumes that children remain in their parents’ household through age 18. In determining how much additional consumption is needed to equalize the living standards of children with those of their parents, ESPlanner permits users to input child-adult consumption equivalency factors that are child-age specific. These factors together with the program’s specification of economies to scale in shared living are used to determine how much household consumption rises or falls when children enter or leave the household. The household’s recommended consumption expenditure will vary with changes in its demographic composition even during intervals of time when the household is liquidity or borrowing constrained.
For example, consider a household that desires a perfectly smooth living standard through time, that lives for 60 years (the household head or spouse specifies a maximum remaining lifespan of 60 years), and that is borrowing constrained over its first 20 years. ESPlanner will (a) smooth the household’s living standard over the first 20 years, (b) smooth the household’s living standard over the last 40 years, and (c) make the discrepancy in living standards between the first 20 years and the last 40 years as small as possible.
Case 5 illustrates these features of ESPlanner because the additional expenditures on consumption arising from the presence of the children pushes the household into a liquidity constrained position. (We assume in all cases that no borrowing is possible.) Indeed, the household is liquidity constrained over its first 13 years, by which we mean that its living standard in the first 13 years is lower than in subsequent years. As the Case 5 ESPlanner results show, initial (1999) consumption is $37,105 and initial taxable saving is $2,981. The corresponding QFP recommendations are $32,600 and $7,299.
Why is the household saving in 1999 in the ESPlanner results if it is liquidity constrained? The answer is that the household still attempts to smooth its living standard over the first 13 years even if it can’t perfectly smooth its living standard over its entire lifetime. To do so, it needs to save at the beginning of the 13-year interval in anticipation of the arrival of the second child. Once the second child is born, ESPlanner’s recommended consumption expenditure (but not the living standard per person) rises to $43,383 and ESPlanner’s recommended taxable saving becomes negative. The household slowly spends down its small stock of net worth, so that by 2008 its net worth is zero. Between 2009 and 2012 the household’s consumption rises in step with increases in its disposable income -- the income it has after paying for taxes and housing expenditures. After 2012 the household again saves positive amounts. In that year, its consumption equals $43,705. This level is lowered to $40,366 in 2016 when the first child leaves the household and reduced again to $32,933 when the second child leaves the household.
In addition to recommending markedly different levels of initial consumption and taxable saving, QFP and ESPlanner recommend quite different levels of life insurance for Case 5. QFP’s 1999 recommendations are $520,000 for George and $60,000 for the wife. The corresponding ESPlanner recommendations are $362,076 and zero. Although ESPlanner recommends no initial life insurance holdings for Jane, it does recommend she start buying life insurance in 2003.
What explains these differences? The answer is Social Security survivor benefits. The availability of these benefits reduces the need for life insurance on both the husband and the wife, and the timing of their potential receipt disproportionately reduces the need for life insurance during years when the children are young. ESPlanner includes these benefits in forming its life insurance recommendations. QFP, we suspect, does not.12 The survivor benefits to which we refer here are not simply widow and widower benefits that are available to surviving spouses starting at age 60, but also child, and mother/father benefits that are available to survivor households with children. Because child benefits are paid in the form of annual income streams until the children reach age 19 and mother/father benefits are paid in the form of annual income streams until the children reach age 16, their remaining present values are smaller the older are the children when the household first begins receiving these benefits.
Case 6 augments the previous case by incorporating special expenditures in the form of college tuition payments of $30,000 per year for each child between age 19 and
22. The need to save for these expenditures raises recommended 1999 taxable saving in both programs, but the recommended increase (relative to Case 5) is almost 30 percent greater in ESPlanner than in QFP. And the need to guarantee payment of tuition regardless of who survives also alters life insurance recommendations. In this case the increases (relative to Case 5) are roughly the same across the two programs.
The need to pay for future college expenses changes the timing of the household’s liquidity constraints in ESPlanner’s calculations. Now the household experiences one long period of liquidity constrained consumption – between 1999 and 2023. The year
12 In moving from Case 4 to Case 5, QFP’s recommended life insurance for the husband drops by $60,000, but we don’t believe this is due to the presence of Social Security survivor benefits.
2023 (when George is 53 and Jane is 51) is the year their youngest child finishes college; that is, once George and Jane are free from paying college tuition they experience a discrete increase in their living standard.
Case 7 adds a need for a $50,000 (in 1999 dollars) emergency fund to be accumulated by 2035. Both software packages have the household put aside the $50,000 accumulated for an emergency fund in 2035, when George is age 65. This is reflected in large negative savings at that age in both sets of results. However, QFP recommends that an additional $671 be saved in 1999, whereas ESPlanner recommends only $26 more be saved initially, but an additional $1,118 in 2024 – the first year after the household has finished paying college tuition and is no longer liquidity constrained and is enjoying a higher living standard. Thus ESPlanner pays for the emergency fund when the couple can afford to, i.e., when it doesn’t have such pressing needs for other expenditures. In contrast, QFP pays for the emergency fund by reducing consumption in each year of the household’s life; this is consistent with the conventional approach’s target-savings methodology.
Case 8 gives $50,000 in taxable assets to the household. ESPlanner uses these additional resources to finance additional consumption during the period 1999-2023 when the household is liquidity constrained. Hence, its recommended 1999 level of taxable saving decreases (compared with Case 7) by $1,701, whereas QFP’s recommended 1999 taxable saving rises by $1,679. ESPlanner has the household increase consumption by more than $2,000 in the liquidity-constrained years, whereas QFP recommends an annual increase of slightly more than $1,000 over the household’s lifetime. Taxes are higher in QFP owing to the larger accumulation of taxable assets and the resulting investment income.
In contrast to this difference in taxable saving recommendations, there is little difference between the two programs in the response of their life insurance recommendations to the presence of the initial taxable net worth. Both programs recommend minor reductions in the life insurance holdings of both George and Jane.
Case 9 gives George $50,000 in non-taxable assets which he holds in a defined contribution plan. Because the household is liquidity-constrained through 2023 and because the income from these assets is not accessible to the household until late in life, ESPlanner recommends essentially no changes (compared to Case 8) in taxable saving for 1999. QFP, on the other hand, recommends that the household reduce its 1999 taxable saving by $1,858. Regarding consumption, QFP suggests a $1,800 increase over the household’s lifetime, while ESPlanner increases consumption only slightly initially, but then bumps it up by about $3,000 for the remainder of life.13
Both programs recommend roughly similar reductions in life insurance holdings. This point notwithstanding, the recommended 1999 levels of life insurance for George and Jane remain quite different across the two programs. For example, QFP recommends Jane have $80,000 of life insurance, whereas ESPlanner recommends she have only $8,525.
In this case we assume that both George and his employer each contribute $2,500 to a defined contribution plan on behalf of George. Once again we find large differences in recommended taxable saving responses. QFP tells George and Jane to save almost $5,000 less in taxable form in 1999. ESPlanner tells them to save just $61 less. Because George has more future income (his employer’s contributions) to protect, both programs recommend that George have more life insurance in 1999. But ESPlanner now recommends that Jane have no insurance whereas QFP still recommends she purchase $80,000 of coverage.
This case adds a $50,000 special bequest by both George and Jane. QFP responds by raising recommended life insurance for each spouse. ESPlanner raises recommended life insurance just for George; that is, it continues to recommend no life insurance for Jane. The reason is that were Jane to die, George would have enough resources from his own labor earnings, the couple’s taxable net worth, and his taxable net worth to sustain a higher living standard on his own than were Jane alive.
Case 12 Adding a Defined Benefit Pension
13 Taxes are higher in the last years of life in QFP compared to ESPlanner because taxable as well as tax-deferred assets are consistently higher in QFP. Also, QFP, as we utilize it, does not seem to recognize the
- 33
This case gives Jane a $10,000 (in 1999 dollars) defined benefit pension annuity beginning at age 63. The pension income is not indexed for inflation nor are there survivor benefits. Here again we have income coming to the household in the post-liquidity constrained period. Hence, ESPlanner recommends essentially no change in 1999 taxable saving or consumption. QFP, on the other hand, treats the household as effectively being able to spend this future income in the present. So it recommends a $1,600 higher level of 1999 consumption and a commensurate reduction in 1999 saving. Indeed, QFP’s recommended saving for 1999 is now 13 percent less than ESPlanner’s.
Both programs now recommend less life insurance for George. But QFP recommends $20,000 less in life insurance for Jane, while ESPlanner recommends $33,212 more. QFP’s recommended reduction in life insurance for Jane is curious given that Jane now has more future income that she needs to protect for George against the possibility of her early death.
Case 13 adds a $400,000 (in 1999 dollars) inheritance that Jane will receive in 2027 from her father, but only in the case that Jane is alive. Because this inheritance arrives too late to relieve the household’s liquidity constraint, there is again very little change in ESPlanner’s 1999 saving and life insurance recommendations. QFP, because it ignores liquidity constraints, begins spending this inheritance right away, raising consumption in 1999 by over $4,500 and reducing 1999 recommended saving to just $754. This $754 recommended level of saving stands in stark contrast to ESPlanner's $6,793 saving recommendation for 1999.
special tax credits, deductions, and exemptions available to the elderly. - 34
Both programs now recommend fairly similar levels of life insurance for Jane. But George’s life insurance should, according to ESPlanner, be only about three fifths of the level recommended by QFP. Interestingly, when Jane is 43, ESPlanner recommends more life insurance for Jane than for George. The explanation here is that she needs to insure that her $400,000 inheritance is effectively available to George if she dies prior to 2027.
This case leaves out the inheritance of Case 13 and instead assumes that George and Jane each experience 2 percent growth each year in their real labor earnings. Because the couple has a lot more income in the future than the present, each program advocates dissaving in the short run in order to permit a higher level of current consumption. But ESPlanner recommends almost $8000 more in saving (actually, more than $800 less in dissaving) than does QFP. Again, liquidity constraints explain this. Once ESPlanner realizes that the couple can’t perfectly smooth its living standard over its entire life, it focuses on smoothing the couple’s living standard during the liquidity constrained interval and the non liquidity constrained interval. This means consuming less than QFP recommends in the former interval and consuming more than QFP recommends in the later interval. QFP also suggests a withdrawal (with tax penalties) from the defined contribution plan in mid-life, which ESPlanner will never recommend before age 59.
Case 15 Age 62 Retirement of Husband
This case is identical to Case 12 (that is, no inheritance or real wage growth) except the husband retires at age 62. Earlier retirement increases QFP’s recommended 1999 taxable saving by almost 30 percent, but leaves ESPlanner 1999 saving recommendation essentially unchanged. QFP reacts to the reduced lifetime earnings by lowering current as well as future recommended consumption and thus raising saving. ESPlanner, in contrast, reacts by cutting consumption in the period after George and Jane are no longer liquidity constrained – the period when they are otherwise enjoying a higher living standard.
Because George has less future labor income to protect, both programs advise George to purchase less life insurance. QFP’s recommended reduction in insurance for George is, however, 40 percent greater than ESPlanner’s recommended reduction. Both programs also advise Jane to buy less life insurance. Indeed, ESPlanner recommends zero life insurance for Jane, whereas QFP recommends a $100,000 policy.
This case is identical to Case 12 except it assumes 8 percent rather than 6 percent nominal rates of return on taxable and non taxable assets. Because the inflation rate is still set at 3 percent, this implies real interest rates of 5 percent rather than 3 percent on these assets. Ignoring liquidity constraints, higher real interest rates permit higher levels of lifetime consumption. But the household is liquidity constrained. Indeed, the higher real interest rate generates so much accumulation in their non-taxable accounts that George and Jane now experience two intervals of liquidity constraints – one prior to their children completing college and the other between the time their children complete college and the time (when George is age 65) that they pay off their emergency fund, start withdrawing their non-taxable assets, and start collecting their Social Security benefits. Consequently, ESPlanner recommends very little change (compared with Case 12) in initial consumption and, therefore, in initial taxable saving. At George’s age 65, when the household is in the last year of its second constrained interval, consumption is $43,489; that is, $2,078 more than in Case 12. At George’s age 66, consumption rises to $61,057, which is $16,273 (36 percent) more than in Case 12! In contrast to ESPlanner, QFP cuts initial saving by more than one half. Hence, ESPlanner’s recommended 1999 level of saving is more than twice that of QFP. Again QFP recommends large mid-life withdrawals from the household’s defined contribution plan.
This case differs from Case 12 in assuming a 5 percent inflation rate. The higher inflation rate lowers real returns on taxable and non taxable assets to 1 percent. It also reduces the real values of the couple’s mortgage payments and Jane’s pension benefits. QFP reacts to these offsetting factors by reducing (relative to Case 12) household saving in 1999 by about one quarter. ESPlanner leaves 1999 saving roughly as is.
Lowering the latest ages at which George and Jane could die from 95 to 85 would, absent liquidity constraints, permit George and Jane to consume more each year. Thus, QFP’s initial consumption rises (relative to Case 12) from $31,270 to $35,570 lowering initial taxable saving from $5,994 to $4,653. ESPlanner’s initial consumption is changed only slightly in reflection of a small change in recommended life insurance holdings and, thus, life insurance premium payments. Hence, initial ESPlanner taxable saving in Case 18 is 46 percent higher than the amount recommended by QFP. QFP’s recommended life insurance holdings are unaffected by the change in the maximum age of life, whereas ESPlanner’s recommended holdings decline because the household does not need to insure its living standard for as long a period of time.
In this exercise, we tell both programs that the household wishes to have a 10 percent higher living standard after George is age 65. Because the couple is liquidity constrained when young and would otherwise experience more than a 10 percent rise in its living standard after retirement compared to their youth, this change in assumptions leads to essentially no change in ESPlanner’s initial consumption and taxable saving recommendations. But QFP now recommends less initial consumption and more initial taxable saving. Consequently, the two programs recommend almost the same initial levels of taxable saving.
In this scenario, the couple sells its home when George retires and moves into a rental property that charges $1,500 per month in rent. This decision frees up the couple’s home equity for use in financing its consumption in retirement. On the other hand, it raises considerably the couple’s post-retirement housing expenses. On balance, the decision reduces the couple’s ability to finance consumption.
QFP encourages the couple to immediately cut its consumption and raise its taxable saving in anticipation of the higher post-retirement housing expenses. ESPlanner, in contrast, cuts future rather than immediate consumption and, thus, recommends roughly the same initial taxable saving. Both programs recommend that both spouses purchase more life insurance. However, as has been the case with the other scenarios, QFP’s insurance recommendations are considerably higher than those of ESPlanner. Here, QFP’s 1999 recommended insurance holdings for George and Jane combined total $800,000. The comparable ESPlanner figure is $561,041 – a 30 percent smaller amount.
In looking back at the 20 cases we’ve now considered, several things become clear. First, the two programs provide remarkably different recommendations even in simple settings despite the fact that each is designed to help households maintain their living standards or at least their consumption levels though time. Second, the two programs are calculating significantly different levels of taxes and Social Security benefits and these differences materially alter their saving and insurance recommendations. Because we have been able to check the accuracy of ESPlanner’s tax and Social Security benefit calculations, it appears that QFP’s calculation of these variables may be “soft”. (Of course, there is a broader question applicable to all software packages and planning exercises about the accuracy of any projection of taxes and Social Security decades into the future, although QFP does allow the user to lower the estimate of Social Security benefits to be paid in the future.) Third, the fact that QFP does not automatically adjust for household composition and liquidity constraints leads it in many cases to recommend either much less or more taxable saving compared to ESPlanner. Finally, QFP generally recommends more life insurance than what ESPlanner has determined that households actually need in order to insure their current living standards for potential survivors; this may reflect an incomplete estimation of Social Security survivor benefits.
This section examines the recommendations of the two software programs when applied to four actual households.
This case involves a couple in which each spouse is age 35 and each retires at age
65. They plan to have two children, one in 2001 and one in 2003. The husband earns $43,000 initially, declining by 2001 to $35,000 and staying constant thereafter. The wife earns $37,000 in 1999, zero in 2000, $35,000 in 2001, $36,000 in 2002, $37,000 in 2003, and $38,000 thereafter. The husband receives a gift from his father of $10,000 in 1999 and 2000. As for special expenditures, the couple makes a $15,000 downpayment in 1999 to purchase a house and makes nominal truck loan payments of $4,500 in 1999 and 2000. The couple also plans to spend $20,000 on college tuition for each child between their ages 19 and 22. The couple plans for funerals of $5,000 each, but doesn’t wish to leave a bequest. The couple currently has $14,000 in taxable net worth, and the wife has $3,000 in an IRA. Otherwise the couple has no assets.
The wife intends to contribute $1,200 to her IRA annually until she retires. Her withdrawals from her IRA will begin at age 65 and will be taken out in equal annual payments. The couple is buying a house in 1999 for $150,000. Purchase of the house causes $2,500 in annual property taxes, $400 in annual homeowner’s insurance payments, and $2,000 in annual maintenance. The couple intends in 1999 to take out a $135,000, 30-year mortgage with monthly principal and interest payments of $990. Both spouses will begin collecting Social Security retirement benefits at age 65. They both entered past labor income that we need to compute their Social Security benefits in ESPlanner. The economic assumptions the couple choose are ESPlanner’s default values including 6 percent nominal interest rates on taxable and non-taxable assets, a 3 percent inflation rate, and a zero non-mortgage borrowing constraint.
In running the couple through ESPlanner, we also solicited their current saving and insurance holdings. Prior to running the software the couple planned to save $19,620 in 1999. The wife has no life insurance, while the husband has $30,000.
In ESPlanner we find that the couple is never liquidity constrained. On the other hand, the couple never accumulates a significant stock of taxable assets. As the table for Case 21 makes clear, the couple is advised to consume $26,866 initially and $38,500 when both children are at home. Our application of QFP, on the other hand, recommends constant consumption of $26,920 when both spouses are alive, regardless of the presence of children. Here ESPlanner gives a more intuitive recommendation; children cost a lot of money and this must be reflected in household savings decisions.
QFP and ESPlanner agree that husband and wife should have equivalent life insurance holdings, reflecting their equivalent economic contribution to the household. Again, however, ESPlanner suggests lower life insurance than QFP.
This case involves a couple with a 40 year-old wife and a 39 year-old husband. They have two children, one born in 1991, the other born in 1993. The wife doesn’t work. The husband works earning $200,000 in 1999 and 2000. Starting in 2001 and continuing until his retirement at age 55, the husband expects to earn $100,000. They plan to send each of their children to college for four years at a cost of $30,000 per child per year. They have no special bequests and plan for funerals of $5,000 each. The husband currently has $800,000 in life insurance; the wife has none. The couple’s taxable assets are $225,500. The wife has an IRA with a 1999 balance of $84,700, and the husband has a 401k with a 1999 balance of $148,000. Both plan to withdraw their non-taxable assets (and make them taxable) at age 59. The couple is currently saving $11,765 in taxable form. The husband plans to contribute $9,500 into his 401k plan each year and expects his employer to contribute $6,000. The wife does not intend to make additional IRA contributions. The couple owns a $475,000 house with annual property taxes of $5,200, annual maintenance of $1,500, annual homeowners insurance of $500, and a 29-year-old $170,000 mortgage with monthly payments of $1,131. Each spouse intends to take his/her Social Security retirement benefits starting at age 62.
QFP and ESPlanner recommend vastly different amounts of consumption, taxable saving, and life insurance for this household. QFP recommends more than twice the life insurance for the husband than is recommended by ESPlanner. Differences in tax calculations, in the treatment of household composition, and in the treatment of liquidity constraints appear to explain the different consumption and saving recommendations. To begin, ESPlanner sets consumption much higher in earlier than in later years because it realizes that the couple needs to spend more when it has children. Second, ESPlanner calculates the couple’s short-run taxes to be much higher than those generated by QFP, reflecting higher Massachusetts’s income taxes on capital income. Note that QFP does not allow for the employer’s matching contribution to the 401(k) plan to reflect inflation-induced increases in pay. Also QFP’s Social Security benefits are lower than ESPlanner’s.
This case involves a couple in which the husband is age 64 and the wife is age 57. The husband intends to work for two more years, earning close to $400,000 over the two years. The couple has a variety of large special expenses in the short run, including an expensive home renovation. The husband has two nominal pensions providing close to $200,000 annually that he expects to collect starting after retiring. Each spouse is planning for a $5,000 funeral. The couple also wants to give its children a $2 million gift in 2025. The couple’s taxable net worth is close to $3 million. The wife has a very small IRA account, and the husband has a 401k account with balances that are close to three-quarters of a million dollars. Each spouse elects to withdraw the smallest amount of funds from these tax-favored accounts permitted by law. The couple owns a house with a market value of $1,200,000. The annual property taxes, maintenance, and homeowners insurance total $6,000, $13,000, and $1,000, respectively. There is a 25-year $525,000 mortgage on the property with a monthly payment of $3,318. The couple plans to sell its home in 2025 and use the proceeds as well as other funds to make the $2 million gift. After selling its home, the couple plans to rent a home for $4,000 per month. The couple uses the default economic assumptions.
QFP and ESPlanner agree in this case that neither spouse needs life insurance, but they disagree very strongly about the amount of taxes the couple will pay initially and over time and the rate at which the couple should dissave (spend down) its taxable assets. ESPlanner calculates the couple’s 1999 taxes at $182,449, whereas QFP calculates them at $237,681. This is a 30 percent difference. ESPlanner takes into account the tax deductibility of special expenditures, and ESPlanner’s much lower 1999 taxes may reflect the fact that some of the couple’s very large 1999 special expenditures are tax deductible. Interestingly, ESPlanner’s calculated taxes are higher than are QFP’s when the husband is age 75, but then fall below QFP’s.
Because ESPlanner calculates that the couple will generally pay less in taxes initially and overtime, it recommends considerably higher consumption expenditures. In 1999, ESPlanner says the household can consume $204,510, whereas QFP tells the household not to consume more than $186,880. ESPlanner recommends the couple dissave $317,615 in 1999 – a very large sum, which reflects, in large part, the couple’s very substantial special expenditures in 1999. QFP, in contrast, recommends the couple dissave $138,380.
Case 24 A Middle-Aged Low-Income Divorcee
This case describes a 59-year-old divorcee who plans to work until age 70 and whose life expectancy is age 85. She earns $35,000 a year, has $32,000 in assets (evenly divided between taxable and tax-deferred accounts), is making a nominal loan repayment of $4,500 for the next three years, and pays $10,200 a year for rent. Her current employer contributes 20 percent of pay to a defined contribution plan, and she will receive $4,000 a year in nominal benefits from a prior employer’s defined benefit pension plan.
ESPlanner gives a higher standard of living than QFP and allows greater savings withdrawals for this woman. These recommendations are almost entirely due to ESPlanner’s higher estimate of Social Security benefits -- $21,430 compared to $9,232 in QFP.14 In fact, ESPlanner’s higher estimate for Social Security benefits produces a liquidity constraint in the first decade of planning, as the replacement of income in retirement for this woman is so high.
This paper used Economic Security Planner and Quicken Financial Planner to illustrate and contrast the economic and conventional approaches to financial planning. After clarifying the main conceptual and technical differences between the two programs, the paper compared their consumption, saving, and insurance recommendations for 20 stylized and 4 actual households. These recommendations are remarkably different even for very simple cases. The differences are, however, readily explained. They are due to very different tax and Social Security benefits calculations and to very different treatments of demographics and borrowing constraints.
14 In part, the benefit is high because a) our middle-aged divorcee doesn’t plan to start collecting benefits until she is age 70 and b) ESPlanner incorporates delayed retirement credits and benefit recomputations in its calculation of retirement benefits.
References
Ando, Albert, and Franco Modigliani, “The ‘Life Cycle Hypothesis’ of Saving: Aggregate Implications and Tests,” American Economic Review, 53, 1963.
Auerbach, Alan J., and Laurence J. Kotlikoff, “Life Insurance of the Elderly: Its Adequacy and Determinants,” in Gary Burtless, ed., Work, Health, and Income Among the Elderly, Washington, D.C.: The Brookings Institution, 1987.
Auerbach, Alan J., and Laurence J. Kotlikoff, “The Adequacy of Life Insurance Purchases." Journal of Financial Intermediation, vol. 1, no. 3, pp. 215-241, 1991.
Bernheim, B. Douglas, The Vanishing Nest Egg: Reflections on Saving in America, New York: Priority Press, 1991.
Bernheim, B. Douglas, “The Merrill Lynch Baby Boom Retirement Index: Update ‘95,” Stanford University, mimeo, 1995.
Campbell, John Y., Joao F. Cocco, Francisco J. Gomes, and Pascal J. Maenhout, “Investing Retirement Wealth: A Life-Cycle Model,” NBER Working Paper No. W7029, 1999.
Hubbard, R. Glenn, and Kenneth L. Judd, “Social Security and Individual Welfare: Precautionary Saving, Borrowing Constraints, and the Payroll Tax,” American Economic Review, 77 pp. 97-112, 1987.
Hubbard, R. Glenn, Jonathan Skinner, Stephen P. Zeldes, “Precautionary Saving and Social Insurance,” Journal of Political Economy 103(2), pp. 360-399, 1995.
Kotlikoff, Laurence J., Avia Spivak, and Lawrence H. Summers, “The Adequacy of Savings,” American Economic Review, 72(5), pp. 1056-69, 1982
Warshawsky, Mark and John Ameriks, “What Does Financial Planning Software Say About Americans’ Preparedness for Retirement?” in Olivia S. Mitchell, P. Brett Hammond, and Anna Rappaport, editors, Forecasting Retirement Needs and Retirement Wealth, Pension Research Council, forthcoming, 1999.
Yaari, Menahem E., “Uncertain Lifetime, Life Insurance, and the Theory of the Consumer,” Review of Economic Studies, vol. 32, pp. 137-50, 1965.
Table 1 George’s and Jane’s Non Tax-Favored Balance Sheet in Case 1 Table 1 (Continued) Table 2 Jane’s Survivor Non Tax-Favored Balance Sheet in Case 1 Table 2 (Continued) Table 3 Comparing Results from ESPlanner and QFP
| Year | George’s Age | Jane’s Age | Income | Spending | Taxes | Non Tax-Favored Saving | Non Tax-Favored Net Worth |
|---|---|---|---|---|---|---|---|
| 1999 | 29 | 27 | 50,000 | 26,588 | 8,640 | 14,772 | 14,772 |
| 2000 | 30 | 28 | 50,430 | <