Detailed description

The Retirement Calculator performs a Monte Carlo simulation of possible outcomes starting with your current financial situation. The calculations are done in one-year increments.

Rather than assuming a return distribution for interest rates, stock market returns and the returns from real estate investments, we sample 30 year of historical data. The drawback of this approach is that in many cases we are over sampling our data set (i.e. we may use the same year's return data more than once in a given simulation). The advantage is that we do not have to make any assumptions about the nature of the return distributions. Given that many financial distributions are far from normal, this is a significant advantage, especially for worst-case estimates.

Data integrity is an issue for any projection based on history. We use historical returns form the S&P 500 from 1974 to 2004 as a proxy for stock market returns. For bond returns we use a blend of several pieces of data including the returns from the longest running bond mutual funds, bond futures and various other data sources. For real estate returns we rely on regional indices based on the valuations of homes that have been sold multiple times. This index is maintained by the Office of Federal Housing Enterprise Oversight. Short-term interest rates are the average three month t-bill rate for each year. While we have made every effort to ensure that the historical data is correct, we cannot guarantee this. Incorrect historical data may result in misleading simulation results.

The calculation steps for each simulated future are outlined below.

1. Calculate the net total cash flow consisting of the current year's incomes, expenses, loan payments, and Required Minimum Distributions (RMD) from retirement accounts if any. Note that all sources of income are taxed at the effective tax rate unless they are marked as tax-exempt.

2. If the net after tax cash flow is positive, the excess is invested in the stock market. If it is negative, assets are consumed to cover the shortfall. The most liquid assets are consumed first, except that retirement assets are consumed after all non-retirement assets are depleted. We assume that all assets can be liquidated easily and that partial liquidations are unproblematic.

3. Calculate the asset returns. We do this by randomly selecting a year between 1974 and 2004. Each asset is then assumed to have returned what the corresponding asset class returned in that year. This allows us to maintain historical correlations between different asset classes. For assets in taxable accounts, we apply the effective tax rate to obtain the after tax return.

4. The final asset value is the starting asset value plus the after tax return less the portion that was consumed during the year and less the RMD, if any.

5. Liabilities are increased by the amount of interest due on the balance and decreased by the total payment over the course of the year.

6. All income and expense cash flows are increased by the average inflation rate except for the ones marked as fixed amounts.

This process is repeated for the number of years in the user's time horizon. This completes one possible future trajectory of the portfolio. We calculate multiple such trajectories as specified in the 'Number of Iterations' field. These trajectories map out the range of possible outcomes. In order to generate the output table we follow the following steps:

1. For each year, find the net asset values for all paths.

2. Sort them from the smallest to the largest. The table shows the worst outcome, an outcome so that 5% of all outcomes are worse than it, the median outcome, an outcome so that only 5% of the outcomes are better than it, and the best outcome.

The simulation for a given path is finished when the asset value drops to zero, i.e. when all assets have been consumed. The program will continue to calculate more time steps, but the asset value will never go negative even if consumption continues after all assets have been consumed.