The stock market funds available in the federal employee Thrift Savings Plan (TSP) have historically provided good rates of return, but to realize these rewards, a buy-and-hold investor would have had to endure large drawdowns (i.e., maximum peak-to-valley experienced in account value) and extended recovery periods.
Enduring substantial drawdowns may be the hardest aspect of investing, and since the psychological toll of prolonged states of fear and stress can lead to poor decision making, it is important to understand both the risk and reward characteristics of investment strategies. By doing so, you are attempting to improve the odds that in actuality you'll stick with the plan.
Here I describe a TSP portfolio allocation and rebalancing strategy that is easy to implement and appears to considerably reduce downside volatility, thereby increasing expected returns at a given risk level.
I. Investment Choices
The three available TSP stock funds, their asset categories, the underlying indices, and the mutual funds used as surrogates in my backtesting are:
- C fund, Domestic Large Cap Blend, S&P 500, VFINX
- S fund, Domestic Mid/Small Cap Blend, Wilshire 4500, VEXMX
- I fund, Foreign Developed Markets, MSCI EAFE, DFALX
Mixing bond funds with stock funds is key to dampening portfolio volatility. This is primarily due to the non-correlated price movement of stocks and bonds, a tendency that exists much, but not all, of the time. The two available TSP bond funds, their asset categories, and the funds used as backtest surrogates are:
- F fund, US Government/Corporate Aggregate Debt, VBMFX
- G fund, US Treasury Note Special Issue, VFISX
The core TSP funds are accurately represented by the associated surrogate funds used in my backtesting. Only for the unique G fund is there an exception: the historical rate of return is a bit more than the proxy VFISX (since 1992, 4.73% vs. 4.17%). Additionally, the G fund has the comforting benefit that comes from zero downside risk.
Another available option is to invest in the TSP Lifecycle Funds, each of which is made up of a combination of the five core TSP funds (C, S, I, F, G). The L funds are designed to maintain an optimal balance of investment risks and rewards for a particular time horizon, and feature automatic quarterly adjustments towards a less risky mix of investments as the target date approaches. You don't have to remember to regularly change your TSP fund allocations as your retirement approaches - it's done for you.
II. Candidate Portfolios
Of course we all hope to grow our savings at the highest rate possible, and wish our account balances would increase each and every month. Figure 1 illustrates one of investing's facts of life: when markets inevitability drop, it can take a long time to climb back. The calculations are for a TSP portfolio that varies from 100% bond funds (F, G in a 1:1 ratio) to 100% stock funds (C, S, I in a 3.5:1:1 ratio). Corresponding maximum drawdown values range from -1% to -52% for the various market periods.
Figure 1. Portfolio Drawdown Recovery Time
For portfolios with equity positions up to 30%, no more than 16 months would have been needed for a full recovery from any market decline since 1992. Stepping up to a 50% equity position, the 2000 bear market would have required 38 months of recuperation, while the 2008 bear market and the corrections of 1994, 1998 and 2011 would have bounced back sooner. A 70% equity position would have meant enduring two major drawdowns in the 1992-2015 period, one lasting 50 months and the other 38 months.
It's commonly stated that it took 25 years to recover from the Great Crash. Imagine how that affected the unlucky stock market investor who waited until September 1929 to go all-in! In real terms, however, when the effects of dividend payments and the unparalleled deflation experienced in that era are included, the recovery time was 7-1/4 years. Further study found the longest historical recovery in real terms to be the 1972 bear market, during a period of high inflation, at around 10-1/2 years. The careful reader should be aware that my calculations (Fig. 1 and the tables below) do factor in dividends but ignore the dollar's buying power.
If your risk tolerance can be called "medium" (i.e., somewhere in between "aggressive" and "conservative") and your investment time horizon is at least 6 to 10 years, I suggest a 50/50 mix of stocks and bonds as the risk-vs-reward sweet spot. One reason is that a 50/50 blend roughly cuts in half the bear market recovery time compared to being fully invested in stocks, as Figure 1 depicts. A footnote: I derived the 6-to-10 year horizon by simply doubling the anticipated worst case recovery times of a 50/50 blend (i.e., 2000's recovery calculated at 3 years; 1972's recovery estimated at about 5 years).
Here I introduce several TSP portfolios that will be put through a backtest covering the last 23 years. The present analysis assumes they are rebalanced to the specified allocation on a quarterly basis (every 3 months).
- 30%C, 20%S, 20%F, 30%G (my medium-risk preference)
- 34%C, 12%S, 19%I, 6%F, 29%G (L-2030 fund)
- 27%C, 8%S, 15%I, 6%F, 44%G (L-2020 fund)
- 15%C, 15%S, 20%I, 25%F, 25%G (Merriman)
- 90%C, 10%G (Buffet)
- 100%G (safe harbor)
Portfolios 1, 3 and 4 reflect my suggested 50/50 stock/bond mix, but with differing allocations among the individual funds. Portfolio 2 is the TSP L fund allocation for those with a 15-year horizon target. Portfolios 5 and 6 are useful reference points as they amount to bookends for the risk spectrum.
III. Backtesting Results
Without further ado, I now evaluate the portfolios' historical performance for the period January 1992 through July 2015. The calculations were made using the surrogate funds' monthly data.
The performance summaries below include my proprietary "composite risk", a more comprehensive indicator of investment risk than standard deviation of returns, downside semi-variance, or maximum drawdown, the three most popular metrics with traders and investors.
Table 1. Portfolio Performance Comparison
|Sharpe Ratio||Sortino Ratio||Composite Risk|
It's not surprising that the greatest reward (compounded annual growth rate) comes from following the advice of Warren Buffet with regards to his wife's inheritance (per the February 2014 letter to Berkshire Hathaway shareholders, pg.20). But look at the maximum drawdown, it is nearly twice as deep as my preferred portfolio. Even before my 40th birthday I would not have had the intestinal fortitude to stand pat and watch my savings shrink by half. How about you?
In fact, Portfolio 1's track record is superior in risk/reward terms than the other designed allocations. Compare for yourself -- divide CAGR by the "Composite Risk". Or check out the Sharpe and Sortino Ratios. A footnote: I choose not to invest in the TSP foreign stock fund because other analyses I've conducted indicate that this added diversification does not improve risk-adjusted returns at the portfolio level.
To complete this effort, I performed a sensitivity analysis to better understand the frequency of rebalancing the preferred portfolio. Starting with a $100K account, here are the results over the 23-year test span:
Table 2. Effect of Rebalancing Frequency
|Sharpe Ratio||Sortino Ratio||Composite Risk|
Yearly rebalancing seems to be best, since it shows the most risk-adjusted return and requires the least monitoring. I suggest you use your birthday for the annual reallocation, and avoid the year-end holiday season, when the markets tend to display price distortions as institutional investors shift billions in and out of various asset classes.