FIRE Thresholds and Time (Overview)
Posted: Fri Dec 06, 2002 8:29 am
by JWR1945
FIRE Thresholds and Time
A. Conditions.
1. I looked at two series of portfolios. In all cases the withdrawal rate was 4.444% and the duration was 50 years.
2. The Lx series consisted of a conventional mixture of stocks and commercial paper. It was re-balanced annually. The percentage of stocks was varied.
3. The Kx series consisted of a cash only portfolio and a conventional 80% stock portfolio. The 80% stock portfolio was re-balanced annually. The size of the cash only portfolio was large enough to last 0, 5 and 10 years. All withdrawals were taken from the cash only portfolio until it was depleted. After that, withdrawals were taken from the 80% stock portfolio.
4. I introduced thresholds and special thresholds. The portfolio is converted entirely to cash if the balance...at 20 years, 30 years or 40 years...is sufficient to last the remainder of the duration. With special thresholds, the portfolio is converted entirely to cash whenever it is sufficient to last for the remainder of the duration. It is not restricted to years 20, 30 and 40. I assume that the cash component matches inflation exactly. With ibonds and TIPS, this is a conservative estimate.
B. Results.
1. The results favor using high stock allocations with conventional stock portfolios (the Lx series). Using thresholds and special thresholds removes much of this sensitivity. Best results were for 90% stocks...but using 80% stocks was almost as good. These high allocations correspond to the long duration (50 years).
2. The results with dual portfolios (the Kx series) favor an overall initial stock allocation near 60%. Using thresholds and special thresholds removed much of the variation. With the dual portfolio, the overall percentage of stocks increases as the cash only portion is depleted. It finally reaches 80% stocks. During the first few years the 80% stock component of the overall portfolio is allowed to grow unimpeded.
3. Using thresholds and special thresholds significantly improves the performance of both the Lx series and Kx series of portfolios. This makes sense because the volatile component of a portfolio has good years as well as bad years. It is worthwhile to lock in success during one of those good years.
4. Using thresholds and special thresholds removed most of the late failures and left the early failures untouched. This makes sense because removing early failures would require a dramatic effect. Removing later failures is much easier.
5. The years 1965, 1966, 1968 and 1969 showed up consistently as horrible start years. The years 1929 and 1937 also showed up as horrible start years, but to a lesser extent. The years 1906 and 1964 were also bad, but to an even lesser extent.
6. Since the duration was 50 years, retirements begun in the 1960s do not have complete historical sequences. In a sense, we can state that the recent past is not in the historical record...and that it is worse than the historical record. We can also state that the present market lies outside of the historical range on a valuation basis.
7. There are many reasons to be optimistic anyway. These results were for a 4.444% withdrawal rate over 50 years (as opposed to the normally quoted 4% over 30 years). Thresholds still help. The approach for selling stocks in the model...withdrawing a constant dollar amount each year...is a bad choice to use in real life. (It is OK for modeling because it makes things simpler and easier to understand.) Ben Solar has already established that using valuation to influence allocations improves performance. It also seems reasonable to expect some favorable times for buying and selling stocks within the next decade. There are alternatives to the S&P500 index...even within the stock market. REITS and high dividend stocks come to mind.
Have fun.
John R.
Posted: Fri Dec 06, 2002 8:43 am
by JWR1945
FIRE Thresholds and Time (Long)
A. The portfolios.
1. I examined two series of portfolios on the dory36 calculator.
2. The Lx series consisted of a conventional mixture of stocks and commercial paper. I examined portfolios with 50%, 60%, 70%, 80% and 90% stocks.
3. The Kx series consisted of a cash only portfolio and a conventional 80% stock (20% commercial paper) combination. All withdrawals initially came from the cash only portfolio. After it was exhausted, further withdrawals came from the conventional 80% stock portfolio. The size of the cash only portfolio was adjusted to last 0, 5 and 10 years.
4. In all cases the total portfolio amounts started at $1800. Annual withdrawals were $80 (or 4.444%). The duration was 50 years. (The CPI was used for inflation adjustments. The expenses were 0.20% of the portfolio balance. There was no check-mark for a first year withdrawal. For 5 years protection from stock withdrawals, the cash only portfolio started at $400 and the 80% stock portion was $1400. For 10 years protection from stock withdrawals, the cash only portfolio started at $800 and the 80% stock portion was $1000. All of the stock portfolios were re-balanced annually.)
5. I have introduced thresholds to lock in success at 20 years, 30 years and 40 years. I have locked in success whenever the balance is sufficient to produce $80 per year for the remainder of the duration. Those thresholds are $2400, $1600 and $800 respectively. I assume only that the cash component matches inflation. When I mention special thresholds, I have locked in success whenever the balance is sufficient to produce $80 per year for the remainder of the duration. Special thresholds can apply in any year. When I use the term gentle cheating, I am referring to a balance is slightly less than needed to last the full duration...but that comes very close (and misses only by a small fraction of a year).
B. The Lx series results.
1. The number of failures for the Lx series without using thresholds favors higher stock percentages. With 50%, 60%, 70%, 80% and 90% stocks, the number of failures were 37, 27, 23, 18 and 18 respectively. Introducing thresholds at 20 years, 30 years and 40 years reduces the number of failures to 29, 21, 19, 17 and 13 respectively. Introducing thresholds for every year (special thresholds) reduces the failures to 26, 17, 15, 11 and 9. If we allow ourselves to cheat a little bit (gentle cheating), we also declare success when failures occur only a fraction of a year before the end of the duration. In that case (special thresholds with gentle cheating), the failures are 23, 17, 13, 10 and 8 respectively. Thresholds cut the number of failures substantially.
2. Examining the number of failures that occur through year 29...what I call horrible failures...the numbers are 4, 5, 4, 5 and 5 failures without thresholds and 4, 5, 4, 5 and 5 with special thresholds...for 50%, 60%, 70%, 80% and 90% stocks respectively. The number of failures that occur from years 30 through 34 are 11, 5, 5, 2 and 3 without thresholds and 10, 4, 4, 2 and 3 with special thresholds (and 9, 4, 4, 2 and 2 with special thresholds and gentle cheating). There is little difference.
3. Examining the number of failures that occur from years 35 through 39, the numbers are 11, 8, 7, 6 and 3 without thresholds. The numbers are 8, 7, 6, 2 and 0 with special thresholds (and 8, 7, 4, 1 and 0 with gentle cheating). Thresholds are showing an effect.
4. Examining the number of failures that occur from years 40 through 44, the numbers are 5, 6, 4, 3 and 5 without thresholds...and 3, 1, 1, 1 and 1 with special thresholds...and 2, 1, 1, 1 and 1 with gentle cheating. Examining years 45 through 50, the numbers are 6, 3, 3, 2 and 2 without thresholds...and 1, 0, 0, 1 and 0 with special thresholds...and 0,0,0, 1 and 0 with gentle cheating. Thresholds eliminate almost all of the failures that occur during the final decade.
5. From the Lx series of runs, we see that using thresholds removes late failures. Varying the percentage of stocks does little to protect against early failures. The number of failures before year 35 are almost identical for stock allocations of 60% and higher. The number of failures before year 30 are identical for all of the stock allocations that were examined (50%, 60%, 70%, 80% and 90%).
C. The Kx series results.
1. The 80% stock portfolios in combinations K2, K3 and K4 grow freely for 10, 5 and 0 years respectively. Initially, all withdrawals are from the cash only portfolio. After the cash only portfolio is depleted, all withdrawals are from the 80% stock portfolio, which is re-balanced annually to maintain the 80% stock allocation. The initial stock percentages of combinations K2, K3 and K4 are 44.4%, 62.2% and 80% respectively. They grow toward 80% as the cash only balance decreases. (Portfolio K4 is identical to L5.)
2. The number of failures for combinations K2, K3 and K4 are 26, 19 and 18 without thresholds...22, 16 and 17 with thresholds...14, 10 and 10 with special thresholds...and 13, 9 and 10 with gentle cheating. Again, using thresholds reduces the number of failures substantially. In this case the dual portfolios favor an initial overall stock percentage around 60%. This is substantially different from the conventional single portfolio allocation. (These dual portfolios all grow to an 80% stock allocation by year 10.)
3. Examining failures before year 30, the numbers are 4, 4 and 5 for K2, K3 and K4 respectively whether or not we use thresholds. Examining failures from year 30 through 34, the numbers are 8, 3 and 2 respectively...again, whether or not we use thresholds. Using thresholds does not affect the number of failures before year 35.
4. Examining failures from year 35 through 39, the numbers are 6, 7 and 6 without thresholds for K2, K3 and K4 respectively. With special thresholds they are 2, 2 and 2 (or 1, 1 and 1 if we allow gentle cheating). Examining failures from year 40 through 44, the numbers are 4, 3 and 3 without thresholds...and 0, 0 and 1 with special thresholds (with and without gentle cheating). Examining failures from year 45 through 50, the numbers are 4, 2 and 2 without thresholds...and 0, 1 and 1 with special thresholds (with or without gentle cheating). Thresholds remove almost all of the failures that occur in years 35 and later.
5. From the Kx series of runs, we see that the initial stock allocation should be close to 60%. (All of the Kx portfolios eventually reach an 80% stock allocation.) Using a dual portfolio improves results (K3 versus K4 with K4 = L5). Using thresholds does not change the number of failures before year 35. From year 35 through 50 using thresholds eliminates almost all of the differences among portfolios K2, K3 and K4.
D. The horrible years.
1. The horrible years are those with failures before year 30. The bad years are those with failures from year 30 through 34. These are the years least affected by using thresholds to lock in success.
2. For L2, the horrible years are 1937, 1965, 1966 and 1969. The bad years are 1906, 1909, 1911, 1912, 1929, 1936, 1939, 1940, 1964, 1967 and 1968.
3. For L3, the horrible years are 1937, 1965, 1966, 1968 and 1969. The bad years are 1906, 1912, 1929, 1964 and 1967.
4. For L4, the horrible years are 1965, 1966, 1968 and 1969. The bad years are 1906, 1929, 1937, 1964 and 1967.
5. For L5, the horrible years are 1929, 1965, 1966, 1968 and 1969. The bad years are 1906 and 1967.
6. For L6, the horrible years are 1929, 1965, 1966, 1968 and 1969. The bad years are 1906, 1964 and 1967.
7. For K2, the horrible years are 1906, 1964, 1965 and 1966. The bad years are 1902, 1907, 1929, 1962, 1963, 1967, 1968 and 1969.
8. For K3, the horrible years are 1965, 1966, 1968 and 1969. The bad years are 1906, 1964 and1967.
9. Since K4 is the same as L5, its horrible years are 1929, 1965, 1966, 1968 and 1969. Its bad years are 1906 and 1967.
10. Notice that L3, L4, L5, L6 and K3 (and K4) all have 1965, 1966, 1968 and 1969 as horrible years. Both L2 and L3 have 1937 as a horrible year. Both L5 and L6 have 1929 as a horrible year. K2 is unique in having 1906 and 1964 as horrible years.
11. Notice that the 1960s were especially bad for starting retirement. Even 1929 and 1937 are less prominent.
12. Remember that the duration was set to 50 years. Portfolios begun after 1950 (actually 1952) have only partial historical sequences available.
13. This means that the recent past (from 1950 on) has been an especially bad era in the historical record. Even worse, the recent past is not fully reflected in the historical record.
14. Couple this with the observation that the present (1995 through 2002) lies outside of the historical range on a valuation basis...and with precedent breaking high valuations. The implication is that withdrawal rates for both the recent past (from 1950 on) and the present (from 1995 on) should be worse than what is calculated from the historical record.
E. Summary
1. From the Lx series we see that conventional (single) portfolios favor large percentages of stocks when the duration is 50 years. Upon closer examination, however, we see that the number of early failures (before year 35) is almost identical whenever the stock percentage is 60% or more. We also see that using thresholds eliminate almost all of the differences in late failures (from year 40 and later).
2. From the Kx series we see that a dual portfolio starting with an overall stock allocation around 60% does better than a conventional (single) portfolio. In the cases examined, the dual portfolios eventually grew to 80% stocks. Upon close examination, we see that using thresholds had no effect on early failures (before year 35). They eliminated almost all other failures (year 35 and later).
3. Thresholds greatly enhance portfolio safety by locking in success. They eliminate most of the late failures.
4. Thresholds do not eliminate early failures...those that happen before year 35.
5. Early failures are closely identified with retirements beginning in the 1960s (1965, 1966, 1968 and 1969) and to a lesser extent the depression (1929 and 1937) and 1906.
6. Forming a complete 50-year historical sequence requires looking at start years of 1950 (actually, 1952) and earlier. Hence, the worst years for starting a retirement appear only in partial sequences.
F. Reflections.
1. Remember that these results are associated with a 4.444% withdrawal rate over a 50-year duration. That is in contrast to 4% and 30 years.
2. There are many reasons for optimism. Ben Solar is currently looking into varying stock allocations according to valuations. Early results are promising. If we extend our time frame to a decade, we can expect to have at least one highly favorable period based on valuations. There are also better ways to handle buying and selling stock...especially, selling...than we see in the models.
3. A key observation...that the effect of using thresholds is to remove failures that would normally occur during later years...makes sense. It should hold regardless of how the details of the future differ from the past.
4. A secondary observation...that using thresholds improves portfolio survivability...also makes sense. The fluctuating portion of a portfolio has good years as well as bad years. It is reasonable to lock in success during one of those good years.
5. The reduced sensitivity of rates of success to stock allocations...when using thresholds...makes sense. Stock allocation percentages have an effect at the margin. Only a dramatic change would be enough to change an early failure into a success. Early failures are common to all of the portfolios.
6. It is reasonable to expect a dual portfolio approach to change the optimal stock allocation. Supporting the details of such a change requires closer examination.
Have fun.
John R.
Posted: Fri Dec 06, 2002 1:51 pm
by raddr
John,
Nice work
Now that I've had a chance to examine it in detail here's my take plus a few questions:
Unfortunately, it is the "early" failures that are the real retirement wreckers. I was hoping for a better showing early on by sparing the stock portion of the portfolio from withdrawals but it appears that can't be helped by the dual method.
As for the "late" failures I think that in the real world most of us will be moving to safer portfolios as we get older. For example, if I'm carrying 80% stocks 30 years from now when I'm 78 I hope my wife will haul me down to the local nut house and get my head examined.
So I guess I'm saying that I would expect most investors to reallocate to more cash as we age anyway thus imposing a sort of self inflicted threshold.
Another potential problem with locking in success by converting to an all cash portfolio would be the specter of taxes if, like me, the investor holds most of his portfolio in taxable accounts. If, for example, you hit the inflation-adjusted $1600 threshold at year 30 you would really have a $5300 portfolio in nominal terms assuming 4% inflation. Selling out would cause a huge capital gain to be realized and likely would also trigger the AMT if the tax law stays the same.
Another problem for some people who want to leave some money behind for heirs is that you would be pretty well tapped at the end of the period if you hit one of the thresholds and convert to cash. Also you'd need to be real sure that the 30 or 50 or whatever year target you have is enough time. I'd hate to retire at age 40, put the 50 year dual portfolio in place then live past 90.
Posted: Sat Dec 07, 2002 1:01 am
by wanderer
raddr -
upon further reflection, the draw should be -5% or so for the first 4 years. in addition, i am invested incredibly far away from the "efficient frontier" (cause it seems to be effective to eschew that "efficiency"
).
this is one of the things I just don't get about this methodology and its insistence on assets with characteristics solely like us equities. I just realized, that, barring a total meltdown in the capital markets and bizarrely non-historic (guessing) residential real estate returns, were we to leave employment at eoy 2003, we would be generating cash in excess of expenses to the tune of almost $30,000 (about 2.5% of net worth). If we pull the cord in eoy 2004, we should be generating almost $60,000 more than we consume (5% of net worth, roughly). This is based on real calculations that tie to brokerage statements and tax returns, so I feel very confident in those numbers. That pattern and gap should widen in a positive manner, slightly, for the next 4 years (to about $70,000 by eoy 2008).
after 4 years of continuing to build, i guess my next question is what my drawdown can be over a 60 year period (between the two of us).
again the other shoe is i want to be eligible for some of this socialist largesse: so i want to model draining off much of that net worth (maybe two thirds).
hopefully all of that makes sense.
raddr, fmo - we really need to somehow model cash flows from real estate. I guess the starting point would be reit returns and modify them for varying levels of leverage and locations.
wanderer