Motoring Toward a Better Future

Regional ReviewQuarter 3, 2000
by Peter Fortune

Retirement planning is simple — so long as you know all your investment options, the tax codes, social insurance programs, and insurance plans that affect you, as well as the prices you will face.

You must also divine the future. The course of important macroeconomic variables such as inflation and interest rates all need to be taken into account. So must personal factors such as your longevity, earnings growth, and future health care expenses. And you also must know your own desires and preferences, so you can anticipate any changes in your choices as the economic environment shifts over time.

Take Peter and Paula. He’s a 40-year-old teacher earning $40,000; she’s a 38-year-old systems analyst earning $50,000 a year. They expect to remain childless. They own their own home, valued at $200,000 with a mortgage of $175,000; their only fixed commitment is a $1,900-a-month mortgage payment for the next 20 years. Peter already has $38,000 in his tax-deferred 401(k) and IRA accounts, and $80,000 in term life insurance. Paula has $55,000 in tax-deferred retirement accounts and a $100,000 term life insurance policy. They have committed to saving 15 percent of their income, or $13,500 per year, which they expect to put into their tax-deferred accounts.

What do Peter and Paula want to learn from a financial planning tool? Probably what most people want. First, to figure out how much more, if anything, they should save. And second, the best investment strategy to get them to their retirement goals.

THE CRITERIA

GOALS Does the program set out to provide useful information? How broad are its goals? How well does it achieve them?

DETAIL Does the underlying model have sufficient detail? For example, does it contain an appropriate range of goods and assets to be bought or sold? Does it separate those factors not under the user’s control (the inflation rate, interest rates, and stock returns) from those that are (the amount saved, the allocation of saving among different assets, and the “rent vs. buy” decision for housing)?

DYNAMICS Does the underlying model adequately capture the linkages between variables at different points in time? For example, the rate of inflation tends to be highly persistent from year to year. And recent evidence suggests that, over long periods of time, stock returns are not random walks but tend to revert to the mean.

PROBABILITY Does the model capture the full range of uncertainties facing the user? For example, the path of future inflation rates, interest rates, asset returns, and prices of the relevant array of goods and services is not fixed, but consists of an infinite number of possible paths and of interactions among those paths. A strong engine will incorporate the probabilistic nature of planning, while a weak one will simply assume a fixed path for each factor outside the user’s control.

TRANSPARENCY Is the user given the details of the underlying model or simply the outcome of a given set of inputs (i.e., a black box)? Transparency allows the user to assess the credibility of the planner’s results and contributes to a better understanding of the principles underlying good financial planning.

INSTRUCTION A planning engine is instructive if its user learns something useful about the economic relationships involved in his or her decisions, and about how those decisions affect the results.

FRIENDLINESS The engine that requires excessive inputs, is clumsy to navigate, or makes the user work too hard for meager results is simply not worth using.

THE IDEAL ANSWER: A VIRTUAL FUTURE MACHINE
In an ideal world, Peter and Paula have access to a Virtual Future Machine (VFM) to help them do their work. They don their Virtual Future headgear and turn on the power. After a brief booting-up period, they upload files containing their characteristics, both personal (age, occupation, marital status, number of children, and so on) and financial (current earnings, net worth, and current allocation of their wealth among various assets). The machine provides information on the full range of probabilities of all factors outside their control: the future course of interest rates, inflation, and the relative price of various goods and services, including asset prices and the price of real estate.

Next, electrodes detect their brain waves and physiological responses to create a picture of their preferences that the VFM will use to anticipate their responses to changes in the economic environment. The result is a map of their tastes now — and in the future — for every consumer good and service. The evolution of those tastes is traced as experience is gained, as children (if any) grow up and go to college, and as parents become needy and die.

Upon hitting the “Go” button, Peter and Paula are inserted into a virtual economy complete with the random shocks affecting someone with their demographic profile — VFM considers the probabilities of being promoted or demoted, of having children, of getting divorced, of investing in losers and winners, of enjoying health or suffering illness, of dying young or old. Peter and Paula are propelled along a multitude of virtual life paths, each shaped by chance and their previous decisions, and they make new decisions in response to this new information, all in the hope of maximizing their satisfaction.

When VFM whirs to a stop, it reports the likelihood of all possible paths of items purchased, assets acquired or sold, bequests made, and so on. The probability attached to each path is reported, and each path is accorded a number measuring the satisfaction that results. Peter and Paula can select the current decisions that put them on the best path. This is the starting point for their plan. Next year, they can revisit the machine to find out whether there is any reason to adjust their choices.

As of yet, the Virtual Future Machine is not available in the real world. But fast computers, cheap memory, and mass storage have aided the creation of new financial planning tools. Economists are helping by developing software that incorporates the latest economic thinking and techniques. This article reviews two such programs: the financial engine available at www.financialengines.com, and the stand-alone program ESPlanner. Unlike many planning engines, the engines reviewed here are not simply financial calculators masquerading as planning tools; they are constructed under the guidance of economists, and are based on methods and ideas widely used within the profession. We look at what these programs do — and don’t do — and at their economic foundations. The criteria used to assess these programs are outlined in a sidebar.

BEING SHARPE ABOUT IT
The first engine that we review is available at www.financialengines.com. It was developed under the leadership of William Sharpe, a Nobel Prize winner whose work in economics pioneered the use of probability and statistics to structure optimal securities portfolios. Not surprisingly, it excels at addressing Peter and Paula’s second question: What is the best investment strategy for their retirement goals?

In fact, financialengines.com has two goals: forecasting retirement income and providing investment advice for achieving retirement objectives. The Forecaster requires that the user enter each working adult’s wage and salary income, annual saving, and the amounts and names of stocks and mutual funds owned. It then generates a probability distribution of retirement income, including Social Security benefits. The user can determine, for example, that he has a 78 percent probability of having retirement income that equals or exceeds his objective. The Advisor works off the Forecaster: It takes the same information used by the Forecaster (except the list of stocks and mutual funds) and selects the portfolio that maximizes the user’s probability of achieving his or her retirement income goals.

Financialengines.com is not very helpful in answering Peter and Paula’s questions about their level of saving; it treats “real” choices — how much to save, what goods or services to buy, whether to own or rent a house, and whether to have children — as something to be preselected rather than decided in response to feedback from the model on factors such as income and asset prices. Even so, this engine has a high goals score: It answers those questions that most users will ask.

Peter and Paula used financialengines.com to assess their retirement plans. They assumed that their real income would grow at a 2 percent annual rate and that the average inflation rate would be 3 percent. When Peter retires in 27 years at age 67, their household income will have grown to almost $154,000 (all figures are in year-2000 dollars). They set their retirement income goal at 70 percent of their preretirement income, or $108,000. If they invest all their tax-deferred savings in an S&P 500 index fund as planned, the Forecaster calculates that their real net worth will be almost $1.1 million at Peter’s retirement. Happy to be millionaires (but who isn’t, these days?), they were nonetheless dismayed to see the distribution of their retirement incomes: Their median retirement income, at $99,000, is short of their goal, and they have only a 42 percent chance of meeting or exceeding their objective. Perhaps they should plunk down the small amount necessary to get some Advice from financialengines.com!

Financialengines.com’s probability score is quite high, as one would expect, given Sharpe’s expertise in developing models of financial decisions under uncertainty. Not only does it use information on the probability distribution of asset returns, but it also uses probability information for key macroeconomic variables such as inflation and interest rates. The user is not restricted to a single assumption for future values of these variables. Rather, the entire distribution is mapped out and used in the forecasts. These distributions are reported to the user, and can be modified by him.

PETER & PAULA’S ESPLAN
In keeping with the assumption that people want to maintain their lifestyle, the couple’s spending level remains flat, at just under $50,000, until Peter dies at age 90. But their desired life insurance is calculated at a whopping (and possibly unrealistic) $2.3 million, about $47 of insurance for every dollar they spend on goods and services.

Financialengines.com is also high on detail. Leaving questions of saving aside, it allows a great deal of detail in the specification of securities — both mutual funds and individual securities — in the user’s portfolio. Another admirable feature is that the underlying model is dynamic. Recognizing that tomorrow’s inflation rate depends largely on today’s rate, the model generates inflation forecasts that show both short-run persistence and ultimate regression toward the mean.

The software is also excellent in its transparency and friendliness. The user can call up screens that report the underlying probability distributions, the nature of the dynamics used, and the reasons for the choices embedded in the program. The only design shortcoming I found was that some screens came up as black text on dark backgrounds, making them virtually unreadable. Fortunately, a FAQ told me to change the color palette on my screen to allow for more than 256 colors, so I was soon off and running.

Finally, financialengines.com is also an excellent instructional tool. The user learns about economic dynamics (for example, the persistence of inflation) and, if attention is paid, about the role of uncertainty in assessing financial performance. Web pages also describe most of the key relationships involved, such as the link between inflation and interest rates.

DOING IT THROUGH ESP
The second engine is ESPlanner. Available via download from MIT Press’s web site (http://mitpress.mit.edu), there are two versions: one, at moderate cost, for individuals and another, much more expensive, for professional financial planners.

Like financialengines.com, ESPlanner is designed by leading economists: Douglas Bernheim of Stanford University and Laurence Kotlikoff of Boston University are joined by Jagadeesh Gokhale. Bernheim and Kotlikoff believe that the tendency in the United States is to save too little, and ESPlanner reflects their desire to help American families remedy the situation. In other words, ESPlanner is specifically designed to help address Peter and Paula’s first question: How much should they save? Whereas with financialengines.com, saving is a detail that must be provided by the user, ESPlanner’s goal is to calculate the “optimal” level of consumption and savings over the course of a lifetime.

The underlying economic principle in ESPlanner is the life cycle consumption model. First formulated by Nobel laureate Franco Modigliani, this model recognizes that people typically have relatively low earnings in the early part of their life, reach their earnings peak during middle age, and see earnings drop again in retirement. It assumes that most people would prefer to smooth their levels of consumption by spending more than they earn when young (that is, by borrowing), saving in middle age, and then “dissaving” or drawing down savings in retirement. The life cycle model describes the resulting “optimal” levels of spending and saving. Adherence to this model of the “best” path of spending and saving, and careful attention to the tax laws (federal and state) and to the complex structure of Social Security benefits, are the hallmarks of ESPlanner. It will not tell you how to invest your money, as will financialengines.com. But it will tell you how much money you should put aside for investment.

ESPlanner requires a great deal of information from the user. Eleven “folders” must be completed before calculations begin, each containing information unique to the user. In the first four — the Personal Folder, Standard of Living Folder, Earnings Folder, and Special Expenditures and Receipts Folder — the user enters personal information, his or her age, the age of the spouse, their current salary and expected growth, and any expected special expenditures or receipts such as alimony, credit card payments, college, boats and cars, and charitable contributions. Thus, much of the user’s stream of income and expenditures is specified in advance rather than the result of choices made in response to the future course of economic variables. This might make sense for alimony payments, which may be outside the user’s control, but boating expenses are clearly decisions shaped by the user’s income, net worth, and preferences. The user’s state of residence is also required (Paula and Peter are from Massachusetts) so that the planner can calculate both the state and federal tax bite.

The next three folders — the Estate Planning, Net Worth, and Saving Folders — allow the user to specify the bequests of each adult, the family’s initial net worth, and its current saving in bank accounts, stocks, and the like. The Housing Folder, Pensions Folder, and Social Security Benefits Folder contain information on those assets. Finally, the Economic Assumptions Folder allows the user to specify assumptions about important macroeconomic variables, like the real interest rate and the rate of inflation. The user also picks a maximum allowable level of indebtedness.

THE RESULTS

 

FINANCIALENGINES.COM ESPLANNER
Goals Forecasting retirement income; providing investment advice to achieve retirement income objectives Determining “optimal” level of spending and saving; figuring how much life insurance to buy
Detail
High High
Dynamics
High Low
Probability
High Low
Transparency
High Low
Instruction
High Moderate
Friendliness
High Moderate

Price

The Forecaster: Free.
The Advisor: For a single account, $14.95 per quarter; $54.95 per year
For an unlimited number of accounts, $39.95 per quarter; $149.95 per year. 
Individuals: $49.95
Financial planners: $495

Once the data are entered into the folders, ESPlanner calculates the optimal path of a family’s path of spending and the annual accumulation (or, in retirement, deaccumulation) of assets that will result. It creates as many as twenty-eight reports describing the projected Social Security benefits and income from assets (both tax-sheltered and taxable) and the recommended path of spending, saving, and insurance. ESPlanner recognizes that people are mortal, and that death can impoverish those who remain. The program calculates the amount of term life insurance that should be held to ensure that survivors enjoy the standard of living prevailing in the absence of death.

ESPlanner concludes that Peter and Paula have oversaved (see charts). It recommends that they spend almost $50,000 in 2000, requiring dissaving of $18,381. This means that they must sell that amount of their current assets and quickly go to the mall! They are advised to continue to spend more than they earn for six additional years, then to add to their savings for the following 26 years, and, finally, to return to dissaving for their golden years. In keeping with the life cycle assumption of consumption smoothing, spending remains flat at $49,859 per year (in year-2000 dollars) until Peter dies at age 90, when Paula is 88. She lives another seven years, spending about $31,000 per year. Their combined savings peaks at Peter’s retirement, when Paula is age 63; and after her retirement at age 65, they reduce their wealth by consuming out of the funds they have accumulated.

While Paula and Peter are unusual in having oversaved in their youth, they are heavily underinsured, according to ESPlanner. Their combined term life insurance of $180,000 is dwarfed by its recommended insurance of $2,325,324, about $47 of insurance for every dollar of recommended year-2000 consumption spending. This ratio declines almost in a straight line as the couple ages until, when Paula is 63 and Peter retires, they no longer need insurance to protect them from income loss. Although the couple does not seem atypical in its characteristics, this result seems strange. Few of us have this much insurance, perhaps because we do not choose an amount that will allow our survivors to enjoy the same living standard as if we were alive, or, perhaps because the cost of insurance is actually higher than assumed in ESPlanner.

Paula and Peter, a hypothetical couple, were not created to refute ESPlanner’s view that undersaving is rampant, and this reviewer was surprised by its strong conclusion of too much saving and too little insurance. But disentangling the basis for this result is impeded by lack of transparency. The mortality assumptions and insurance premiums embedded in the analysis are not clear. ESPlanner gives some clues to its workings in the software, and the creators have made several of their scholarly papers available on the MIT web site, but ESPlanner is essentially a black box. One takes on faith the modeling of federal and state taxes, the Social Security calculations, and the life insurance mortality assumptions.

ESPlanner is limited in several other respects. It is completely deterministic, devoid of any probability analysis. The user specifies fixed levels of interest rates, inflation, and other factors that determine consumption and saving in the life cycle model. The outcomes of the planner are also reported as fixed values rather than as probability distributions. Nor are there any visible dynamics in ESPlanner; as mentioned above, once the user enters the initial information, these variables continue at the preset levels.

The goals, while important, might not be consistent with most users’ needs. Many users will prefer financialengines. com’s emphasis on portfolio allocation and its implications for achieving retirement goals. I suspect that fewer of us will be interested in whether we are saving the optimal amount, although that is clearly important to achieving retirement goals. Nor will many of us be convinced to buy so much life insurance.

Perhaps the chief weakness of ESPlanner is also its chief strength: the immense amount of detail required as input. The user must gather an enormous amount of data on the family’s finances. This task, though not pleasant, might be ESPlanner’s most useful feature, since many of us are not very organized about our finances. ESPlanner is an education about ourselves, but it does not instruct us about the way the world works.

ESPlanner also requires continuing close attention from its programmers. The emphasis on taxation and Social Security benefits necessitates a change in the program whenever new laws are passed. One wonders how ESPlanner will keep itself up to date, and I already suspect that it has not. According to information in one of the scholarly papers, ESPlanner assumes that Massachusetts residents pay a 12 percent tax rate on interest and dividend income, and on short-term capital gains. This was the law before 1999, but for 1999 and after, the tax rate on interest and dividends is only 5.95 percent.

PUTTING IT TOGETHER
Neither of these innovative programs addresses the full complexity of the financial planning problem, even though each is very complex in its construction. One focuses on the uncertainty of asset returns in a world with a fixed amount of savings, the other looks at the determination of spending and saving in a world with no risk. One looks in detail at the mortality of the users, trying to advise them on their insurance needs. The other assumes that users live forever. But each attempts to keep basic financial and economic principles in mind, and a marriage of the two will move us closer to the goal of a Virtual Future Machine. Such an introduction of economic principles into financial planning is a hopeful start.


Peter Fortune is Senior Economist and Advisor to the Director of Research at the Boston Fed.

 

Stay Connected

contacts email alert Twitter RSS podcasts careers faqs videos
Regional Review Links