Simulating Retirement Investment Survival

It would be wonderful if it were possible to run a computer simulation that would tell you exactly how much money you were going to need in retirement. Unfortunately, retirement investment survival is not a deterministic problem. It is a probabilistic problem. There are simulations, however, that can be used to great advantage in retirement planning.

In order to simulate your financial survival during retirement you need to model:

1) The size of your nest egg (how much will you have invested).

2) The real rate of return on your investments (ie. return minus inflation).

3) How much you will spend and when (a spending model).4) How long you will be retired (or equivalently, when will you die).

You have quite a bit of control on items 1 and 3. In contrast, you have virtually no control of item 4 (unless you are willing to take hemlock at a specific time). You can gain some insight into 4 by examining actuarial tables. This allows you to place probabilities on your longevity. Item 2 can be approached by examining historical rates of return on various investments and asset allocations.

The most simple minded way to approach retirement simulation is to assume a portfolio value, a fixed rate of return, a constant real spending model, and a longevity estimation. This data can be used in a spreadsheet or simple financial formula to determine how much money you need to retire. Unfortunately, this deterministic approach to retirement planning is not very accurate nor useful. Over a 30 or 40 year retirement period, average rates of return, average inflation, average spending, etc. are not good estimates of your actual financial performance. Fluctuations from year to year can be devastating.

One method that addresses the probabilistic nature of the retirement simulation problem is to use Monte Carlo analysis. A computer generates a random number that is used to establish an annual rate of return and annual inflation for a single year. Your assumed nest egg is then modified by the random rates, annual spending is subtracted, and the computer generates a second year's rate numbers. The process is repeated until an entire retirement sequence is created (30 to 40 years typically). The distribution of these rates is forced to be consistent with historical distributions. The multi-year retirement sequence is repeated hundreds or thousands of times and a probability of portfolio survival is computed. Although Monte Carlo methods are powerful techniques, when applied to the retirement investment problem they tend to be pessimistic because they do not account for the correlations between returns and inflation nor year-to-year correlations. An alternative method to deal with the retirement probabilistic problem is to use actual historical data. Data that describes stock returns, bond returns, real estate returns, inflation, etc since 1871 have been tabulated. The historical simulator simply uses this data in historical sequence to examine how a portfolio and spending plan would have survived over the hundreds of historical periods represented by these simulators.

You can gain free access to all three types of simulators at:  http://www.golio.net/Chapter2.html

Basic input data and analysis results are also presented.