Research on Safe Withdrawal Rates

Moderator: hocus2004

JWR1945
***** Legend
Posts: 1697
Joined: Tue Nov 26, 2002 3:59 am
Location: Crestview, Florida

JanSz described his approach towards withdrawing from his retirement funds in this thread: SWR accounting dated Sun Dec 21, 2003 8:14 pm CST
http://nofeeboards.com/boards/viewtopic.php?t=1891

He withdraws 2% of his portfolio's current balance plus one-half of his portfolio's growth compared to six years earlier.

I have examined his approach with three alternatives, which are similar but different.

Method of Comparison

I used my latest version of the modified Retire Early Safe Withdrawal [Rate] Calculator. I call it the Deluxe V1.0A version of the JanSz-Chips Calculator.

I selected four especially difficult years for retirement finances. They were 1929, 1937, 1965 and 1966. I looked at stock allocations of 50% and 80%. In this case, I selected commercial paper for the allocations other than stocks. I looked at portfolio balances at 10, 20, 30 and 40 years (when applicable). I determined the minimum balances as well.

I did not look for the traditional Historical Database Rates that ended with a zero balance. JanSz's approach is based upon current balances. It never falls to zero.

Conditions Examined

I have examined four conditions.

Condition A is JanSz's basic approach. Withdrawals are 2% of the portfolio's current balance plus 50% of the increase in the account balance as compared to six years earlier.

Condition B replaces the fixed 2% withdrawal by stripping away all dividends. It still removes 50% of the increase in the account balance.

Condition C changes from taking 2% of the current balance to taking 2% of the initial balance plus increases (and decreases) to match inflation. It still removes 50% of the increase in the account balance.

Condition D still withdraws 2% of the current balance. But it removes 75% of the increase in the account balance.

Condition B helps us understand the effects of today's low dividend yields. Condition C is the kind of approach one would have if he were to set an absolute floor (in this case 2% of the initial balance) to meet his minimal income needs. An income to satisfy one's minimal needs must necessarily match inflation. (More precisely, it must match the inflation associated with those specific needs, not with purchases in general.) Condition D is an alternative to conditions B and C, both of which reduce portfolio balances. This is obvious for condition B because dividends have been well above 2% until the recent past. It is also true for condition C for those historical sequences that place the greatest stress on retirement portfolios.

There is an important caveat associated with condition B. In condition B, all dividends are withdrawn but not the interest from the commercial paper. This makes the 50% stock allocation look better. The amount withdrawn is higher when stocks make up 80% of the portfolio as opposed to 50%. In addition, in some years withdrawals fall below 2% of the portfolio's current balance for the 50% stock portfolio. In the 1960s there were several years with dividend yields close to 3%. Stripping these dividends from a 50%-stock portfolio amounts to a withdrawal rate of 1.5%.

Details

The initial balances were all \$100000. Expenses were 0.20% of the current balance.

To withdraw 2% of the current balance one increases the total expenses to 2.20%.

To withdraw 2% of the initial balance, the number 2% was put into the calculator in the normal way and expenses are kept at 0.20%. In all other cases, the normal withdrawal rate input to the calculator was set to zero.

The capital gains percentage is applied to any increase in the portfolio balance compared to six years earlier. In other studies, this would have been 0%. In this case, we looked at removing 50% or 75% of any increase. Nothing was removed when there was a decrease.

The dividend input is the percentage of dividends that are reinvested (or made available for reinvestment). In other studies, this would have been 100% with rare exceptions.

This table summarizes the conditions examined.

Code: Select all

``Portfolio          A50    B50    C50   D50   A80     B80    C80   D80stock allocation   50%    50%    50%   50%     80%    80%   80%   80%total expenses   2.2%    0.2%  0.2%   2.2%    2.2%   0.2%  0.2%  2.2%withdrawal rate     0%    0%     2%     0%      0%     0%   2%    0%capital gains       50%   50%    50%    75%     50%   50%  50%   75%dividend input    100%    0%    100%   100%    100%    0%  100%  100%``

Results

The following table shows the portfolio balances at 10, 20, 30 and 40 years. There are no 40-year values for 1965 and 1966 since that would extend beyond the data available. (The calculator has dummy numbers for returns from 2002 to 2010. It still provides its outputs even though they are meaningless.)

Code: Select all

``Year  Condition  10YRS  20YRS  30YRS  40YRS1929   A50   93561   51521   59534   527881937   A50   58941   54896   51204   329561965   A50   61685   43195   36240   N/A1966   A50   66590   44163   38976   N/A1929   B50   87808   46205   53094   481161937   B50   56886   51544   48799   317411965   B50   62527   44070   37716   N/A1966   B50   67656   44909   40524   N/A1929   C50   93446   49216   52993   447221937   C50   57378   50066   43791   226551965   C50   60705   39888   29970   N/A1966   C50   65170   40222   31653   N/A1929   D50   88316   47195   52324   429141937   D50   56680   50151   45190   275491965   D50   58468   39708   30226   N/A1966   D50   64360   41161   32569   N/A1929   A80   76439   42120   58156   520041937   A80   56625   58821   55992   350821965   A80   52392   40421   35559   N/A1966   A80   61269   42311   39381   N/A1929   B80   62630   30371   39703   356741937   B80   49887   47522   44697   273091965   B80   49836   36753   32174   N/A1966   B80   58197   38505   36001   N/A1929   C80   74740   37901   47425   396631937   C80   54698   53192   48130   237131965   C80   51285   35856   27885   N/A1966   C80   59128   37338   30702   N/A1929   D80   69644   36355   46591   377661937   D80   69644   50666   46154   272141965   D80   49335   35828   28129   N/A1966   D80   57763   38734   30850   N/A``

The following table shows the minimal balances within 40 years for 1929 and 1937 and within 30 years for 1965 and 1966. It also shows when those minimums occurred.

Code: Select all

``Start Years      1929   1937   1965   1966A50              51521  31714  51204  32956years after start   20   38     30       29B50              45887  30312  37716  36986years after start   29   38     30       20C50              44722  22655  29970  29235years after start   40   40     30       29D50              42914  27090  30226  30140years after start   40   38     30       29A80              42120  30274  35559  34367years after start   20   38     30       29B80              30371  23246  32174  31251years after start   20   38     30       29C80              37901  22067  27885  27031years after start   20   38     30       29D80              36017  23662  28129  27464years after start  19    38     30       29``

These are the 30-year Historical Database Rates HDBR50 and HDBR80 for portfolios with 50% stocks and 80% stocks, respectively. In all cases withdrawal rates were based on the initial balance and adjusted to match inflation. Expenses were 0.20% of the portfolio's current balance.

HDBR50

Code: Select all

``1929  4.5%1937  3.9%1965  4.2%1966  4.1%``

HDBR80

Code: Select all

``1929  4.4%1937  4.5%1965  4.0%1966  3.9%``

These tables establish that 1929, 1937, 1965 and 1966 were exceedingly stressful on retirement finances. Notice that 1929 was not the worst case. Notice that the worst case depends upon the stock allocation.

These results depend somewhat on the withdrawal algorithm. With the Retire Early Safe Withdrawal [Rate] Calculator with its default values, withdrawals are split into two equal portions and applied on January 1st and December 31st. The amount to be withdrawn is calculated based upon the portfolio's balance on January 1st. As a result Historical Database Rates from the Retire Early Safe Withdrawal Calculator (including versions incorporating my modifications) can be expected to differ slightly from other calculators that use the historical sequence method.

This detail about the timing of withdrawals is not unique to the Retire Early Safe Withdrawal Calculator. For example, the FIRECalc has a box available for one to choose between making the first withdrawal just before the first year or just after.

Observations

In a very broad sense, conditions B, C and D produced similar results. All produced lower balances than condition A. For those years that placed portfolios under the most stress, the portfolio balances are remarkably similar.

As mentioned earlier, condition B, which strips off all dividends, is biased toward a 50% allocation since a smaller stock allocation means that less is withdrawn. As expected, balances for condition B with a 50% stock allocation are higher than for an 80% stock allocation.

It is instructive to look at the minimum balances. There is no single condition that is always best all of the time. But JanSz's original approach looks very attractive. It is an excellent choice from a balanced perspective. It is not overly sensitive to the stock allocation.

When looking at minimum balances, it is important to remember that portfolios begun in 1929 and 1937 were examined over 40 years while those of 1965 and 1966 were only over 30 years.

The overlap in historical sequences can mislead you when looking at the 10, 20, 30 and 40 balances. Typically, the patterns of results are similar, but offset. The 1965 balances at 30 years are similar to those of 1966 at 29 years.

Roughly speaking, these approaches maintain 30% to 50% of the original portfolio's buying power after 30 years under stressful conditions. Under the most stressful conditions examined, the real balances still remained above 20%.

Have fun.

John R.