529 Plan Asset Allocation Issues

This article assumes that you have decided to use a Section 529 College Savings Plan. We will now confront two issues you face: (1) Should you invest in actively managed or index funds within your 529 Plan? And (2), how should you divide your 529 Plan investments between different asset classes?

Active vs. Passive 529 Investments

At the end of 2002, Financial Research Corporation estimated that, in aggregate, all Section 529 plans held $19,766 million in assets. The Rhode Island College Bound Fund was the largest 529 plan in the country, with estimated assets of $2,661 million, up 74 percent from the previous year. Between 2000 and 2003 the Rhode Island plan increased from 1,700 accounts and $8.6 million in assets to 400,548 accounts and $3.9 billion in assets. The Rhode Island College Bound Fund is managed by Alliance Capital, and exclusively uses actively managed funds.

What accounts for Rhode Island's number one ranking, and astounding rate of growth over the last three years?

The answer isn't hard to figure out. According to Cerulli Associates of Boston, and Financial Research Corporation, about two thirds of the nation's 529 assets are in plans that were established with the assistance of a financial adviser (e.g., a stockbroker, financial planner, etc.). And Rhode Island's CollegeBound Fund richly rewards them for their marketing efforts.

Consider this comparison. If you are not a resident of Rhode Island, when you invest in Rhode Island's plan through a broker, you pay a sales load of 4.25% (based on the online prospectus, dated August, 2002 at www.collegeboundfund.org). This assumes that you buy the "A" class of fund shares. If you do this, you also pay a further .25% per year (of the 529's asset value) in sales charge. If you are a Rhode Island resident, the initial sales load is waived if you invest through a financial advisor, but the annual sales charge is not. To eliminate the latter, you have to invest directly, not through a financial advisor.

The net funds you contribute to the plan (that is your initial investment less 4.25%) are invested in a mix of actively managed mutual funds managed by Alliance Capital. The average annual expenses on these funds range from .45% to 1.63% per year. The specific mix of funds can be set in three different ways. First there is a set of pre-determined fund mixes that changes based on the age of the beneficiary (becoming more conservative as the time the funds will be needed approaches). Second, there is a set of predetermined mixes that do not change over time. Finally, you can simply choose your own mix (or have your advisor do this) from the mix of the CollegeBound Fund's underlying mutual funds. We have calculated the weighted annual expense charge for the three core allocation portfolios whose fund mix does not change over time. For both the aggressive growth and growth portfolios, the average annual expense charge is 1.11% of the 529 account's assets. For the balanced portfolio, it is .92%.

On the other hand, if you invest in Iowa's 529 plan (which uses Vanguard index funds, as does a similar plan in Nevada, and soon Ohio too), you pay no initial sales load or ongoing sales charge. Moreover, your funds have annual expenses of only .65% (high by index fund standards, but apparently necessary to cover the higher costs of running a 529 Plan).

Finally, there is one more hidden cost difference. Because actively managed funds try to "beat the market" (that is, earn a higher rate of return than the index fund, which simply earns the market average, before expenses), they tend to trade more frequently. This imposes further costs on the actively managed fund, compared to the index fund (e.g., trading commissions, and adverse price moves while the trade is being executed). These have been conservatively estimated at .50% per year.

What does all this mean in practice? Here's an example: let's compare two investors, who both put $10,000 into a section 529 plan and leave it there for ten years. The first invests in the Rhode Island Plan, while the second invests in the Iowa Plan. Let's further assume that each year the two plans earn the same 8 percent annual rate of return on these funds (before expenses). Assuming an investment in Rhode Island's "balanced core allocation portfolio", and no trading cost impact, after ten years the Iowa investor ends up with 9.6% more money to spend on education. Assuming an investment in Rhode Island's "aggressive growth core allocation portfolio" leaves the Iowa investor with 11.6% more money after ten years. Finally, if we assume an investment in the aggressive growth RI portfolio, and take active funds' higher trading costs into account, the Iowa investor ends up with 17.0% ($2,949) more money at the end of ten years. And remember, this analysis also assumes that active and index fund managers earn the same returns over ten years. In point of fact, the great majority of active managers underperform index funds over long periods, so the actual benefits to investing in index-based 529 plans are probably significantly greater than our 17.0% estimate.

So, in answer to our first question -- should you invest in actively managed or index funds within a 529 Plan -- we squarely come down on the side of indexing.

Asset Allocation in 529 Plans

The objective of your 529 Plan asset allocation policy is to maximize the probability of achieving your minimum target rate of return while staying within whatever risk limits you set. The minimum required rate of return results from the interaction of five variables: (1) Whether the beneficiary of your 529 Plan will attend a private or a public college or university; (2) The annual real growth rate in the cost of tuition, room, and board at private and public colleges; (3) The number of years left before your 529 Plan beneficiary will start college; (4) The percent of the total cost of college you would like the funds accumulated in the 529 Plan to cover; and (5) The amount you intend to contribute to the 529 Plan each year. The following table shows the impact each of these variables has on the (compound annual) target rate of return your 529 investments must earn in order to achieve your goals:

In the following tables, we have assumed possible values for each of these variables, and calculated the minimum target rates of return needed to achieve a given set of goals. Please note that in the interest of conservatism, we have used the following assumptions:

• We assume that an investor wants to have reached his or her 529 savings goal (covering the cost of all four years of college) by the end of the year before his or her beneficiary starts college. For example, if your son is currently eight years old, he or she would start college in September, 2014, and you would want to have reached your savings goal by December, 2013 (after which the funds would be placed in low risk, short term investments).
  
  
• We assume that contributions are made to the 529 Plan on December 31st of each year. To continue with our previous example, this means that our model assumes only nine contributions, at the ends of years 2004 through 2012.
  
  
• Finally, we assume that the current annual cost of room, board, and tuition at a four year, private college or university is $26,854, and $10,636 at a public college or university. These data are from the College Board. We have also used the past 28 years of data (in real terms) on these costs to estimate how quickly they will increase in the future. For private school costs, we used a real annual rate of increase of 2.92% (the average over the 1976-2003 period; the standard deviation was 2.60%); for public school costs we used 2.14% (standard deviation of 3.01%).
  

We stress that these assumptions yield conservative minimum target real rates of return; relaxing any of them (e.g., assuming lower annual rates of increase in real costs) would reduce our target real rates of return. That being said, when it comes to paying for college, we thought it better to err on the side of conservatism.

The following three tables show target real rates of return for private colleges and universities. The first table is based on a goal of funding 100% of the cost of this education; the second table 75%, and the third table 50%. Assuming that the highest feasible annual portfolio real rate of return one should use for planning purposes is 7% (and, as we shall soon discuss, even that carries with it a high probability of falling short), you can see that some combination of annual savings, time horizon, and percent of cost financed are unrealistic. You will also see negative numbers in these tables. They imply overfunding of the plan, given its time horizon and percent of cost financed, because they show how much a plan could lose each year while still reaching its target. In sum, the region of realistic solutions is the shaded one with positive real target returns of between 2% and 7%.

The next three tables present the same information, but assume the 529 Plan beneficiary will attend a public college or university.

Having determined the target rate of compound annual real return you need to earn over your 529 investment horizon, the next challenge is deciding on your Plan's asset allocation. A key issue here is the limited number of asset classes offered by most 529 Plans. Consider, for example, the Vanguard 529 Plan offered by the State of Nevada. Based on our definition of an asset class, it offers only four: real return bonds, domestic bonds, domestic equity and foreign developed market equity (though it also offers a large number of tilts within these, which we'll shortly discuss). Unfortunately, it does not offer three asset classes that, in our asset allocation studies, we have found to provide substantial diversification benefits: foreign currency bonds, commercial property, and commodities, as well as emerging markets equity, which can be used to increase a portfolio's expected return.

The impact of this lack of diversification opportunities is significant. For example, we used our simulation optimization model to develop a model portfolio that maximized the probability of achieving a compound annual real growth rate of six percent after ten years, subject to the requirement that 95% of the time the actual compound rate of return produced would be greater than zero. We first included only the four asset classes available in the Vanguard 529 Plan (for the full list of assumptions we used, see below). We found that the probability of meeting the target is 41%. When we added foreign currency bonds, commercial property, commodities, and emerging markets equity to the mix (with the first limited to a maximum weight of 35%, and the last three to a maximum weight of 20% each), the probability of achieving the target return rose to 68%.

We should also note the range of optimization solutions produced by different asset allocation methodologies. A traditional mean/variance optimization model either minimizes risk (defined as standard deviation) for a given level of expected return, or maximizes return for given level of risk. For a target return of 6%, it produces an allocation of 30% to domestic bonds, 65% to domestic equities, and 5% to foreign equities. The probability that this asset mix would achieve the compound real return target in year ten was 44 percent. The probability that it would produce a compound real return of zero or greater in year ten was 93 percent. Over 10,000 different simulations, the lowest compound annual ten year return it produced was (7.7%). A variation of the traditional mean/variance approach maximizes the ratio of portfolio return less target return to the portfolio standard deviation of returns (this is also known as the "safety first model"). This methodology results in a 100% allocation to domestic equities. The probability that this asset mix would achieve the compound real return target in year ten was 48 percent. The probability that it would produce a compound real return of zero or greater in year ten was 87 percent. Over 10,000 different simulations, the lowest compound annual ten year return it produced was (16.5%). Finally, our simulation optimization approach (for details, see the blue button labeled "methodology summary" on our home page) produces an allocation of 5% to real return bonds, 30% to domestic bonds, 50% to domestic equities, and 15% to foreign equities. The probability that this asset mix would achieve the compound real return target in year ten was 41 percent. The probability that it would produce a compound real return of zero or greater in year ten was 95 percent. Over 10,000 different simulations, the lowest compound annual ten year return it produced was (5.8%).

For now, however, investors in the Vanguard 529 Plan are limited to four asset classes, so that is what we've used to develop our model 529 portfolios. Each of these portfolios is intended to maximize the probability of achieving the specified target real compound annual return over an investment horizon of ten years. We further assume that the investor setting up a 529 Plan wants to be 95% confident that the actual compound annual real rate of return over ten years will be at least 0% (i.e., he or she wants to be 95% confident they won't lose the money they have contributed, except for fees charged by the Plan's manager). We used the four asset classes available in the Vanguard 529 Plan: real return bonds, domestic investment grade bonds, domestic equity and foreign developed market equity. We limited the latter to a maximum of 35% of the model portfolio. Our expected real returns for each asset class were a weighted combination of 67% times the average historical return between 1971 and 2002, and 33% times our estimate of future returns (for more on these assumptions see our May through August, 2003 issues). We also used historical standard deviations (again for 1971-2002), and return correlations from 1994 to 2003. To calculate our model portfolios' asset allocations, we used our simulation optimization model. Possible asset allocations were adjusted in 5% increments to reduce the time required to run the optimization. More details about this approach can be found by clicking the blue button on our home page labeled "Methodology Summary."

The table below shows the target real rate of return each model 529 portfolio is designed to achieve, the weights given to each asset class, the expected annual return and standard deviation (note that, where standard deviation is greater than zero, the expected annual return is always greater than the compound annual return over a longer period), and the estimated probability of achieving the target compound annual real return in year ten. Over shorter periods, this probability will be lower, while over longer periods it will be higher.

As the table shows, as with all our other model portfolios, these are also subject to the limitations that beset all quantitative approaches to asset allocation. For example, the inputs used in asset allocation processes are themselves only statistical estimates of the "true" values for these variables. As important, the underlying economic processes that generate the return distributions are not stable (or, as they say in statistics, it isn’t "stationary"). This is why every mutual fund prospectus notes (though too often in the small print) that "past results are no guarantee of future results." In sum, asset allocation is at best an imperfect science, if not an art. Despite the apparent precision of the models that are used, they can only increase the probability of achieving your goals -- they cannot guarantee it. When it comes to investing, a certain degree of uncertainty is inescapable.

In addition to index funds covering broad asset classes, the Vanguard 529 Plan offers a number of index funds that allow an investor to take tilts within them. These include tilts toward growth, value, midcap and small cap within domestic equities, and long versus short term maturity in domestic bonds. Should one take these tilts? The answer to this question depends on your view of market efficiency. Broadly speaking, there are two schools of thought. The first believes that markets are generally efficient, and that one takes tilts to gain exposure to a different mix of risk factors than that contained in the broad market index. In a reasonably efficient market, these tilts are logically expected to produce either higher returns than the broad index, but with higher risk, or lower returns with lower risk. The second school of thought believes that the presence of irrational investors, uneven flows of information, and obstacles to immediate arbitrage together creates long term market inefficiencies, which logically lead to the possibility of a tilt delivering higher returns with lower risk than the broad asset class index. Of course, this view also requires that the opposite also be possible: that there exists a group of investors on the other side of these trades, who will be stuck with lower returns and higher risk than the broad index. In our writing in The Index Investor, we have repeatedly examined this issue; on balance, we come down on the side of generally efficient markets, and believe that the most logical basis for taking a tilt is to achieve either higher returns than the broad asset class index with higher risk, or lower returns with lower risk. Hence, based on historical results, a tilt toward value, midcap, or small cap equities should produce somewhat higher returns than the broad index, though with a higher degree of risk, while a tilt toward growth should have the opposite effect.

With respect to bond funds, the tilts on offer are towards longer and shorter maturities than the broad market index fund, which has an intermediate average maturity. Unlike the case of equities, taking these tilts only makes sense if you are confident in your ability to time changes in interest rates. Logically, you would shift to the long maturity fund when you expected rates to fall, and to the short maturity fund when you expected them to rise. As a general principle, we believe that most investors lack the skills to succeed at market timing over the long term. That being said, we aren't ideologues on the issue; we also believe that there are some situations where it makes sense. With nominal U.S. interest rates currently at their lowest levels in decades, it would be a brave investor indeed who decided today to put a substantial portion of his or her portfolio into a long maturity bond fund. Then again, if you expected a sharp all in the price level (that is, rising deflation), then this would be a smart move. As we said, market timing is a very, very difficult game to play well, and most people would be better advised to avoid it, and invest in the broad bond market index fund.

Finally, what about those 529 funds which are based on the beneficiary's year of birth? The key selling point of these funds is their promise to automatically adjust their underlying asset allocation (to make it more conservative) as the start of college grows closer. How do they compare with the asset allocations in our model 529 target return portfolios? That is a subject we'll address in next month's issue. Stay tuned...