In the military we have the Thrift Savings Plan (TSP).

What’s the TSP?

The TSP is essentially a 401k for federal government workers which was made available to members of the military in 2000. There are five basic funds which can be invested in:

- Government Bonds (G)
- US Aggregate Bond Index (F)
- S&P 500 Index (C)
- DJ US Completion TSM Index (S)
- MSCI-EAFE International Index (I)

There are also lifecycle funds with target dates every 10 years from 2020 through 2050 or so. These lifecycle funds are made up of G, F, C, S & I funds, in constantly shifting weights, which are unknown to me. I know that they shift more and more towards bonds and away from equity as the target date approaches.

I get questions all the time from my comrades-in-arms about which fund is best for them and I never give a solid answer. I never give a solid answer because it would be irresponsible to give an answer without knowing anything about their goals, needs and tolerance for risk. It is still fun to speculate as to which TSP Fund or which portfolio of TSP Funds is best for different goals and I am curious. So, let’s get to it……

Ignoring the lifecycle funds and using only the G, F, C, S & I funds, I am going to use the Solver add-in for MS Excel to build 3 separate optimized portfolios which will:

- maximize expected return
- minimize expected risk
- maximize the Sharpe ratio, which means that it will maximize the expected return per unit of risk

First, I will need some data. I am going to use the past 61 months of returns for each of the five funds.

Next I will calculate the expected monthly returns, sample standard deviation (risk) and covariance between all the funds.

Lastly I will build the model, where the weights for each fund can be adjusted. The Excel model can be seen here: TSP Optimization.

With the model, you can input the portfolio weights manually, or you can use the Solver add-in and let the software do the heavy lifting.

In solver, I am setting the following two conditions:

- weights in the funds must be > or = to 0 (there is no short selling in the TSP)
- the total weights of the funds must be = to 1 (100%)

I solve for maximizing expected return and….no surprise, solver goes 100% into the highest returning fund: The S Fund with an expect monthly return of 1.7%.

When I solve for minimizing expected risk, there is only a small wrinkle. Solver goes 99.87% into the G Fund and .13% into the C Fund. So far there does not appear to be much diversification benefit between the fund types.

The Sharpe Ratio is the ratio of expected return per unit of risk. The higher the ratio the better. Normally, the Sharpe Ratio is calculated as:

(Expected Return – Risk Free Rate) / Expected Risk = Sharpe Ratio

However, there is a small problem with this ratio and one of the TSP Funds. The G Fund is essentially the risk free rate. If the portfolio went 100% into the G Fund it would attempt to divide something into zero (Risk Free Rate – Risk Free Rate / Expected Risk = [0 / Expected Risk]). This would just equal zero which would be worthless for this experiment. So, I am going the modify the Sharpe Ratio and just calculate it as:

Expected Return / Expected Risk = Modified Sharpe Ratio

When I solve for maximizing the Sharpe Ratio, I get…….. something not too interesting at all. Solver spits out everything into the G Fund (99.73%) except for a meager .27% into the C Fund. Boring, to say the least.

Although I said I was only going to calculate three different portfolios with Solver, I am going to do another one. In this scenario, I am going to attempt to maximize my modified Sharpe Ratio, except I am going to put another condition into Solver, which requires a minimum .75% monthly return.

And voilà! Finally, a semi-interesting portfolio has emerged. Solver says: 71.45% in the F Fund, 18.11% in the C Fund and 10.44% in the S Fund.

The bottom line is that portfolios should be matched to an individual’s ability and willingness to take risk and their need/desire for return. Nearly 100% invested in the G Fund may have the best Sharpe Ratio in this model, but if it is not generating a high enough return for someone to achieve their goals, then it is not an ideal portfolio. Another important point to consider is that this model is just for the TSP funds. If there are other investments held outside of the TSP, they need to be considered along with the TSP funds. I will revisit this model later and add some individual stocks into the mix and see what happens.

July 4, 2016 at 6:49 pm

Hello,

I found your blog because I am interested in either the Creighton or Penn State programs leading towards a CFA Charter. I did 5 years in the Marine Corps then jumped right into the Federal Bureau of Prisons. Im tired of baby sitting inmates and want a better career. However, as a federal employee I too have the TSP. I currently put everything into the 2050 fund. Is your results the best ratios to diversify the funds into instead of the lifecycle funds?

July 5, 2016 at 1:41 pm

Marc,

I excluded the lifecycle funds because they are essentially made up from the other 5 basic funds. Something I did not address in this post is that optimization on the efficient frontier is not very stable over time. The market today is not the same as it was 2 years ago and if I ran the numbers again including the past 24 months of data, it is probably that I will get different results. This post was not intended for investment advice and was for educational purposes. The 2050 fund may be the best option for you, so do not change anything based on this blog post.

I applied to both Creighton and Penn State. After working with both schools through the application process, it became clear that Creighton was the superior choice. As a graduate of the MIMFA program, I am a bit biased, but I think if you dig around you will be likely to agree with me. Thanks for stopping by and best of luck to you.

All the best,

Mike