- How we create better buy & hold portfolios
- Going beyond conservative/balanced/aggressive portfolios
- My portfolio is too conservative, has too much cash, etc.
- Historical vs currentized results
- Optimal allocation will slowly change over time
- How loss limits are calculated
- Why all portfolios must include Money Market/T-Bills
- List of assets representing each asset class
- Assumptions made in our backtests
- Can we guarantee that your portfolio will never exceed your loss limit?
- Adding additional asset classes to the Portfolio Builder
- Glossary of statistics
How we create better buy & hold portfolios:
We employ two statistical techniques to build better buy & hold portfolios:
- “Subset resampling”, because the future will be different than the past.
- “Currentized bond yields”, because bonds face stiff headwinds in the coming years.
Going beyond conservative/balanced/aggressive portfolios:
Financial professionals use terms like “conservative, balanced or aggressive” to identify the portfolio that best matches investors’ unique tolerance for risk. All things being equal, more aggressive portfolios tend to achieve higher returns over the long term, but often require investors to endure larger losses and volatility in the short-term.
The problem is that these are imprecise terms. Most investors don’t really understand what they mean, or what level of risk they’re best suited for.
What investors do understand is their “uncle point”, or the degree of loss that would cause them to abandon their investment plan. That uncle point might be 10% in a year, or 0% (any loss) over 5 years. These types of hard numbers are more tangible and easier for investors to understand.
Our Portfolio Builder uses this “uncle point” to create the investor’s custom portfolio. We believe it gives the investor a more realistic view of portfolio risk, and creates an asset allocation that more closely aligns with their actual risk tolerance.
My portfolio is too conservative, has too much cash, etc.
Generally speaking, return is tied to risk. There are no magic tricks in investing, and an investor cannot expect aggressive returns without taking on commensurate levels of risk.
We define risk in terms that investors can understand (max loss), not vague terms like “conservative/balanced/aggressive”. As a result, investors often discover that they’re more risk-averse than they previously thought, resulting in portfolios with lower expectations for return than they hoped for.
In that situation, an investor has to face a hard choice. Do I (a) accept those lower returns as the price of my averseness to risk, or do I (b) reevaluate my tolerance for risk by accepting a portfolio with higher expected max loss? There isn’t a one-size-fits-all answer. Every investor’s tolerance for risk is unique.
It’s important to note that our results are almost always less optimistic (or what we’d characterize as “more realistic”) than you’ll find elsewhere. Most buy and hold analysis assumes that history will repeat itself precisely. We know that won’t be the case. The portfolios we create are built for the market as it stands today. That can be a bitter pill for investors who expect high returns with little risk to swallow, but pretty backtests don’t put food in our bellies; actual, real-world performance does.
Historical vs “currentized” results:
We show backtests in two flavors: historical results and “currentized” results.
Historical results are run of the mill backtests based on actual asset class returns. The problem is that those returns are inherently flawed, because they’re based on an overly optimistic era for bonds. We can say with mathematical certainty that many bond funds will not perform as well in the coming years as they have in the past (learn more).
Currentized results are something unique to BetterBuyAndHold, and are key to realistic analysis in today’s era of historically low bond yields. We deconstruct each bond asset into the constituent factors that drive returns. We then rebuild those assets assuming that the largest driver of bond returns, the current yield, was never higher than it stands today. We leave all other return drivers untouched. In that way, we’ve “currentized” historical results, by showing how the portfolio would have performed with yields at the historic lows they are today.
Which result is more relevant to you? If trading for the very long-term, the answer probably lies somewhere in the middle.
Historical results are definitely too optimistic, as they’re based on an era of high/falling bond yields that we won’t see for many decades to come. Currentized results may be a bit too pessimistic, as bond yields can’t stay at historic lows forever.
We use currentized results when designing our portfolios. That means that our assets allocations will change slowly over time based on where yields stand today (learn more).
The optimal allocation will slowly change over time:
Bonds are key to a diversified portfolio, but they face stiff headwinds in the coming years. In response, we create our asset allocations based on where bond yields stand today (learn more).
That means that our asset allocations will slowly change over time as bond yields change. On the surface that seems contrary to “buy and hold”, but it works well with how investors actually execute portfolios.
Sound buy and hold investing requires periodically rebalancing our portfolio (our results assume that portfolios were rebalanced annually). Those rebalances provide the opportunity to keep up with the optimal allocation. We can also adjust to the optimal portfolio when we add funds (by buying assets that we’re under-allocated to) or withdraw funds (by selling assets that we’re over-allocated to).
In short, even though our optimal allocations are slowly changing, this is still a buy and hold strategy. Asset allocations change slowly enough that normal events already occurring in the portfolio provide the opportunity to keep up with the optimal allocation.
That also means using our site is not a one-time exercise. Investors using our Portfolio Builder would need to revisit our results whenever the latest optimal allocation is required.
How loss limits are calculated:
Users select a portfolio that has never lost more than X% over any N-month period. This combination of X% and N-months is the portfolio’s “loss limit”. We use this loss limit to measure risk tolerance, as opposed to imprecise terms like “conservative” or “aggressive” (learn more).
Some key technical details about these loss limits:
- Loss limits are based on month-end data. Intramonth losses are ignored.
- Losses are measured in rolling N-months, not simply calendar years. A 12-month loss would include any 12-month period.
- When selecting a period of 12, 24, 36 or 60 months, max losses are measured end-to-end. For example, if a user selects a max loss of 10% in any 36-month period, that portfolio may have suffered a drawdown from an equity high of more than 10% at some point, but during no rolling 36-month period did losses exceed 10% end-to-end.
- When selecting “any period”, max losses are identical to the more commonly used term “max drawdown”. It measures how much a portfolio has lost relative to its previous all-time high, regardless of time elapsed. The loss could occur over the course of 1-month or 100-months, it doesn’t matter.
We base our analysis on month-end data because it allows us to extend our tests much further back into history. For example, the real estate ETF VNQ, which we use to represent real estate in our analysis, began trading in 2004. The actual underlying index that VNQ tracks, the MSCI US REIT Index (Bloomberg: RMS/RMZ), is available as daily data from 1995. But the FTSE NAREIT Index (Bloomberg: FNER/FNERTR), a near identical proxy, is available as monthly data from 1972.
Many other asset classes also have more data available in a monthly time frame. All things being equal, more data makes for a better analysis, as it allows us to see how a portfolio has performed through more types of markets (bear/bull/sideways) and through more critical events (like Black Monday in 1987 or the 2007-08 Global Financial Crisis).
Why all portfolios must include Money Market/T-Bills:
Without including Money Market/T-Bills, the Portfolio Builder may be unable to create an asset allocation that meets the user’s requirements. It’s possible that no mix of the asset classes selected can satisfy the user’s loss limit. Including Money Market/T-Bills ensures that all portfolios have a solution. It doesn’t necessarily mean that the Portfolio Builder will opt to use it, but it has the option.
Note: There are also ETF alternatives to Money Market/T-Bills that provide similar ultra short-term Treasury exposure. Learn more.
List of assets representing each asset class:
We display our results in terms of generic “asset classes”, or a group of assets that exhibit similar characteristics and behave similarly. An example of an asset class might be “US Large Cap Stocks”. When we invest, we buy an ETF or mutual fund that represents that asset class.
Below we’ve listed sample ETF tickers representing each asset class. This list is not exhaustive, and there are other examples not included here. There are also many choices of mutual funds, which may be the only way to invest for some investors, such as within company retirement plans like a 401k. To test how well an asset represented an asset class, one would want to see performance in recent months and years similar to the ETF examples shown here.
|US Large Cap Stocks||SPY, IVV, VTI|
|US Small Cap Stocks||IWM, IJR, SCHA|
|US Real Estate||VNQ, SCHH, IYR|
|Short-Term US Treasuries||SHY, SCHO, VGSH|
|Int-Term US Treasuries||IEF, IEI, SCHR|
|Long-Term US Treasuries||TLT, SPTL, VGLT|
|US Corporate Bonds||LQD, VCIT, USIG|
|International Stocks||EFA, VEA, SCHF|
|Emerging Market Stocks||EEM, VWO, SCHE|
|International Treasuries||BWX, IGOV|
|Gold||GLD, IAU, SGOL|
|Commodities||DBC, PDBC, GSG|
|Money Market or T-Bills||See below|
Our results assume the return on “Money Market/T-Bills” to be the 3-month US Treasury rate. Depending on the financial institution traded with, this may be a reasonable assumption or it may be too high. There are multiple ETF alternatives that also provide exposure to ultra short-term Treasury exposure, including BIL, SHV and GBIL.
This is not a recommendation to buy or sell any particular security. There may be unique characteristics to a given asset (like tax treatment) not accounted for here.
Assumptions made in our backtests:
- Our results assume that portfolios are rebalanced annually, at the close on the last trading day of the year.
- Transaction fees plus slippage total 0.10% per trade (0.20% round-trip). This may be too high or too low given the size of the trading account and the broker used, but should be sufficient to cover most investors.
- Both dividends and gains are reinvested.
- Return on “Money Market/T-Bills” (i.e. the portion of the portfolio not invested) is assumed to be the 3-month US Treasury rate. Depending on the financial institution traded with, this may be too high. In that case, ETF alternatives like BIL exist. Learn more.
- We do not account for taxes, as they are highly specific to the individual and the account type.
- The data used in all analysis/results begins 12/31/1969.
- We represent each asset class using the largest, most liquid ETF in that space. There are other assets (ETF/ETNs, mutual funds, etc.) that have performed similarly, and could be traded instead with similar results. We provide some examples of alternative assets.
- Historical data from prior to the launch of each ETF is simulated using high quality index data. We do this in order to extend our backtests as far into the past as possible. We apply the same expense ratio to this simulated data as the representative ETF. Note that we only use indices or other data sources that are closely related to a large, liquid ETF trading today. Analysts will often use data from French, Ibbotson, etc., and while those data sources have analytical value, they often do not translate in to large, liquid ETFs that can be traded in the real-world today. We discuss this issue further on sister site Allocate Smartly.
Can we guarantee that your portfolio will never exceed your loss limit?
Of course not. Past performance is never a guarantee of future results.
We go to great lengths to create optimal portfolios based on users’ unique specifications, and generally-speaking, our portfolios are more conservative than you’ll find elsewhere. Having said that, there is always a risk that the future will be different enough to push the portfolio beyond the user’s loss limit.
That’s true not just for our site, but for all investment analysis, period.
Please read our terms of service.
Adding additional asset classes to the Portfolio Builder:
The Portfolio Builder currently includes the 12 asset classes most commonly held by US investors. We’re not averse to adding additional asset classes if there’s enough member demand, but we have to be judicious. Here’s why:
Due to the unique nature of how we create portfolios (“subset resampling”), each portfolio requires as many as 1,562 subset optimizations behind the scenes. Based on the options available on the Portfolio Builder right now, there are a total of 155,154 of these subset optimizations being maintained. Each is constantly being updated as yields change (“currentizing”).
Here’s the rub: every asset class that we add doubles that number. Adding just one additional asset class would require us to maintain 310,308 optimizations. Adding another? 620,616 optimizations. Obviously, things get out of hand pretty quickly, so we have to be selective in the asset classes we allow on the site.
The approach we take is extremely resource intensive, but much more analytically sound. The tradeoff for better results is less flexibility in the asset choices available to members.
Glossary of statistics:
This is a brief glossary of some of the statistics found on this site.
Annualized Return: The average return earned by the portfolio over the period tested. Calculation note: This is a geometric average (i.e. CAGR), not a simple arithmetic average.
Annualized Volatility: Measures annual portfolio volatility over the period tested. Return and volatility tend to go hand in hand. May not be useful in isolation, but helpful for comparing portfolios to one another.
Longest Drawdown: The longest drawdown ever suffered by the strategy, measured from the start of the drawdown (i.e. the month of the previous all-time high) until the end of the drawdown (the month a new all-time high was recorded). Note that “drawdown” is not the same as “n-period loss”. See below.
Max Drawdown: The worst loss suffered by the portfolio, relative to its previous all-time high, regardless of time elapsed. A value of -30% would mean that, at some point in the test, the strategy lost 30% of its value relative to its previous all-time high. The loss could occur over the course of 1-month or 100-months, it doesn’t matter.
Max N-Month Loss: The worst loss suffered by the portfolio over any rolling N-month period, measured end-to-end. The portfolio may have suffered a larger loss at some point (see “Max Drawdown”), but not over any N-month period end-to-end.
Real Return: Measures returns above and beyond the drag of inflation. Historical real returns are based on actual CPI values over the period measured. For “currentized” real returns, we assume an annual inflation rate of 2.5%.
Sharpe Ratio: A measure of a portfolio’s historical return, excess of the risk-free rate, relative to volatility. Higher values are better than lower values. This is the most common measure of a portfolio’s risk-adjusted performance. It is often criticized for considering both upside and downside volatility equally. See calculation note below re: “currentized” Sharpe Ratios.
Sortino Ratio: A measure of a portfolio’s historical return, excess of the risk-free rate, relative to downside volatility (i.e. the volatility exhibited on just losing months). Higher values are better than lower values. It is considered by some to be superior to the Sharpe Ratio because it excludes upside volatility.
Calculation note: Sharpe and Sortino Ratios are often higher for “currentized” results than historical results because the risk-free rate is also currentized. This lowers the risk-free rate, making the numerator in the ratios larger.