• KEY TAKEAWAYS
• Betterment seeks to maximize investor take-home returns, which drives our investment selection criteria and process.

• We select low-cost, highly liquid, index-tracking ETFs to ensure our clients are gaining the desired asset exposure at the lowest total cost of ownership.

• We continue to monitor our investment selections on an ongoing basis, and encourage individual investors to do so as well.

## Introduction

One of Betterment’s central objectives is to help investors achieve the best possible take-home returns. At the most fundamental level, we do this through the allocation advice we provide for every portfolio. However, another key component of performance is the investment vehicles we use in our portfolio. They are an essential—but often overlooked—element in maximizing the risk-adjusted, after-tax, net-of-costs return for our customers.

In the following piece, we detail Betterment’s investment selection methodology, including:

• Why we use exchange-traded funds (ETFs)
• Why expense ratios are not the whole story
• How Betterment estimates an investment vehicle’s total annual cost of ownership

## Why do we use ETFs?

An exchange-traded fund (ETF) is a security that generally tracks a broad-market stock or bond index or a basket of assets just like an index mutual fund, but trades just like a stock on a listed exchange. By design, index ETFs closely track their benchmarks – such as the S&P 500 or the Dow Jones Industrial Average – and are bought and sold like stocks throughout the day. Betterment only uses open-ended ETFs (which carry no restrictions around issuing or redeeming shares) as they have many embedded structural advantages when compared to mutual funds. These include:

##### Clear Goals and Mandates

Unlike many actively-managed mutual funds, the ETFs we use have definite mandates to passively track broad-market benchmark indices. A passive mandate explicitly restricts the fund administrator to the singular goal of replicating a benchmark rather than making active investment decisions constituting market timing, building concentration in either a single-name, group of names, or themes in an effort to beat the fund’s underlying benchmark.

Adherence to this mandate ensures the same level of investment diversification as the benchmark indices, makes performance more predictable, and reduces idiosyncratic risk associated with active manager decisions.

ETFs are transactable during all open market hours just like any other stock. As such, they are heavily traded by the full spectrum of equity market participants including marketmakers, short-term traders, buy-and-hold investors, and fund administrators themselves creating and redeeming units as needed (or increasing or decreasing the supply of ETFs based on market demand).

This diverse trading activity leads to most ETFs carrying low liquidity premiums (or lower costs to transact due to competition from readily available market participants pushing prices downward) and equity-like transaction times irrespective of the underlying holdings of each fund. This makes ETFs fairly liquid, which makes them cheaper and easier to trade on-demand for activities like creating a new portfolio or rebalancing an existing one.

In comparison, mutual funds transact only once per day, which introduces significant lag between desired and filled price. Moreover, certain portfolio management strategies like tax loss harvesting require liquid securities that trade more than once a day.

##### Lower fees and conflict free

Since most benchmarks update constituents (i.e. the specific stocks and related weights that make up a broad-market index) fairly infrequently, passive index-tracking ETFs also register lower annual turnover (or the rate a fund tends to transact its holdings) and thus fewer associated costs passed through to investors.

In addition, ETFs are generally managed by their administrators as a single share class that holds all assets as a single entity. This structure naturally lends itself as a defense against administrators practicing fee discrimination across the spectrum of available investors.

As an example, some index-tracking mutual fund administrators segment their funds into several share classes where institutional and high net-worth investors can secure lower fees and more lenient terms in exchange for investing a higher amount upfront. Retail investors with lower available investment balances are funneled into higher fee share classes with more stringent terms.

By comparison, with only one share class, ETFs are investor-type agnostic. The result is that ETF administrators provide the same exposures and low fees to the entire spectrum of potential buyers.

The fund and administration structure of ETFs also eliminates concerns stemming from potential conflict of interest in the standard sales and access channels utilized by mutual funds. While mutual funds can be sold directly to investors by their administrators, most investments in mutual funds are recommended and placed through a multi-tiered sales and distribution network. Each layer of the network tacks on a host of opaquely documented fees.

These fee amounts are entirely non-standard across funds and networks and are largely the result of negotiations between marketing and sales executives who are divorced from the investment functions of the fund administrators. These network fees come in the form of front and back loaded costs, or immediate one-time fees assessed for initial investment or redemptions. Sales channels are subject to compensation incentives that tend to favor investment recommendations that yield allocation to funds where they can collect more fees over selections that might ultimately be in the best interest of their clients.

##### Tax efficiency

In the case when a fund (irrespective of its specific structure) sells holdings that have experienced capital appreciation, the capital gains generated from those sales must, by law, be accrued and distributed to shareholders by year-end in the form of distributions. These distributions increase tax liabilities for all of the fund’s shareholders. With respect to these distributions, ETFs offer a tremendous tax advantage for shareholders over mutual funds.

Since mutual funds are not exchange traded, the only available counterparty available for a buyer or seller is the fund administrator. When a shareholder in a mutual fund wishes to liquidate their holdings in the fund, the fund’s administrator must sell securities in order to generate the cash required to satisfy the redemption request. These redemption-driven sales generate capital gains that lead to distributions for not just the redeeming investor, but all shareholders in the fund. Mutual funds thus effectively socialize the fund’s tax liability to all shareholders, leading to passive long-term investors having to help pay a tax bill for all intermediate (and potentially short-term) shareholder transactions.

Since ETFs are exchange traded, the entire market serves as potential counterparties to a buyer or seller. When a shareholder in an ETF wishes to liquidate their holdings in the fund, they simply sell their shares to another investor just like that of a single company’s equity shares. The resulting transaction would only generate a capital gain or loss for the seller and not all investors in the fund.

Mutual fund distributions are generally decided by the fund administrator and can introduce material variability in an investor’s tax profile. ETF tax profiles are fairly static with most of the tax realization/deferral control being held by the individual investor.

In addition, ETFs enjoy a slight advantage when it comes to taxation on dividends paid out to investors. After the passing of the Jobs and Growth Tax Relief Reconciliation Act of 2003, certain qualified dividend payments from corporations to investors are only subject to the lower long-term capital gains tax rather than standard income tax (which is still in force for ordinary, non-qualified dividends). Qualified dividends have to be paid by a domestic corporation (or foreign corporation listed on a domestic stock exchange) and must be held by both the investor and the fund for 61 of the 120 days surrounding the dividend payout date. As a result of active mutual funds’ higher turnover, a higher percentage of dividends paid out to their investors violate the holding period requirement and increase investor tax profiles.

##### Investment flexibility

The maturation and growth of the global ETF market over the last two decades has led to the development of an immense spectrum of products covering different asset classes, markets, styles, and geographies. The result is a robust market of potential portfolio components which are versatile, extremely liquid, and easily substitutable.

#### ETFs have seen significant growth

Source:  Investment Company Institute 2014 Investment Company Fact Book, Chapter 3: Exchange-Traded Funds, Figure 3.2

## Selecting across the ETF universe

Despite all the advantages of ETFs, it is still important to note that not all ETFs are exactly alike or equally beneficial to an investor.

The primary task of Betterment’s investment selection process is to pick the set of funds or vehicles that provide exposure to the desired asset classes with the least amount of difference between underlying asset class behavior and portfolio performance. In other words, we attempt to minimize the “frictions” (the collection of systematic and idiosyncratic factors that lead to performance deviations) between ETFs and their benchmarks.

The principal component of frictions between tracked asset classes and investor returns is the fund’s expense ratio: the higher the expenses charged to the investor, the lower the resulting returns that pass through. However, relying on just expense ratio to make an instrument selection could yield to a less efficient portfolio. There are other material frictions that factor in that Betterment also considers, discussed below.

Betterment’s measure of these frictions is summarized as the total annual cost of ownership: a weighted average of all relevant frictions used to rank and select ETF candidates for the Betterment portfolio.

## Total Annual Cost of Ownership

The primary contributors to those frictions are:

• The fund expenses imposed by the ETF administrator (typically called the “expense ratio”)
• Trading costs (for cashflows,  rebalancing, or tax loss harvesting)
• Performance deviations from the target underlying index

We calculate the total annual cost of ownership as a weighted average of three primary identifiable components:

1. Fund expenses imposed by the ETF administrator, or expense ratio (E)
2. Fund-specific market interaction and portfolio activity costs, or bid-ask spread (S)
3. Performance deviations from the benchmark index, or tracking error (T)

#### $$Cost = w_{E}E + w_{S}S + w_{T}T$$

Weights are selected to appropriately represent and aggregate the investing activity of the average Betterment customer in terms of cash flows, rebalances, and tax loss harvests.

An ETF’s expense ratio will be the dominant factor in the annual cost calculations because it is a definite cost not related to trading activity. In addition, expense ratio is the largest cost compared to the other factors. As such, the expense ratio (which forms a lower-bound for costs associated with long-term investment in that security) is given a relatively higher weight compared to the expected costs associated with trading activity and tracking error.

Let’s review the specific components below:

$$E$$ – Expense Ratio

The Expense Ratio is the set percentage of the price of a single share shareholders pay to the fund administrators per year. ETFs often collect these fees from the dividends passing through from the underlying assets to holders of the security, and results in lower total returns to shareholders. Since expenses are a principal component in reducing investor returns, ETFs with higher expenses will perform worse. For context, a Betterment 70% equity portfolio contains ETFs with expense ratios that average to 0.12%.

$$S$$ – Bid-Ask Spread

All market transactions are associated with a price difference between the the price people who are willing to sell and the price people are willing to buy a specific security: the bid-ask spread. Heavily-traded securities with more counterparties willing to transact will carry lower bid-ask spreads. Unlike the expense ratio, the degree to which you care about bid-ask spread depends on how actively you trade. Buy-and-hold investors should care less about it compared to active traders because they will accrue significantly fewer transactions over their intended investment horizons. While Betterment customers often only pay a fraction of the bid-ask spread due to our vertical integration, minimizing these costs is still beneficial to an efficient portfolio. So all other things being equal, Betterment attempts to select ETFs with narrower bid-ask spreads.

$$T$$ – Tracking Error

Tracking error is the degree to which a specific security does not precisely replicate its intended underlying index. It can be thought of as uncompensated risk embedded in a security. While not an outright cost like the other criteria discussed, tracking error (especially high values of tracking error) can introduce systematic deviation in the long-term returns of the overall portfolio when compared purely with a comparable basket of benchmark indices rather than ETFs. It’s important to note over any given period this deviation (or “noise”) isn’t necessarily negative: in some periods it could lead to outperformance. The lack of predictability induced in the returns of the ETF driven by this measure gives some pause to selecting a security with high tracking error.

## Minimizing Market Impact

Our total annual cost measurement discussed above quickly and concisely summarizes the principal frictions involved in ETF selection and yields an easily calculable metric to rank across vehicle choices within an asset class. However, it’s important to note that the costs it accounts for are idiosyncratic to the specific funds and are outside Betterment’s control. They do not account for frictions/costs introduced due to activity under Betterment’s control. This additional friction is called “market impact”, an idea closely related to the general availability of liquidity but nuanced enough to warrant its own explanation.

Another key factor in Betterment’s decision-making is whether the ETF has relatively high levels of existing assets under management and average daily traded volumes. This ensures that Betterment’s trading activity and holdings not come to dominate the security’s natural market efficiency and either drive up or drive down the price of the ETF when trading.

We define market impact for any given investment vehicle as the Betterment platform’s relative size ($$RS$$) in two key areas.

Our share of the fund’s assets under managements is calculated quite simply as

#### $$RS_{AUM} = \frac{AUM_{Betterment}}{AUM_{ETF}}$$

while our share of the fund’s daily traded volume is calculated as

#### $$RS_{Vol} = \frac{Vol_{Betterment}}{Vol_{ETF}}$$

Minimizing investor frictions is one of the core goals of the Betterment investment process. ETFs without an appropriate level of assets or daily traded volume might lead to a situation where Betterment’s activity on behalf of customers moves the existing market in the security. In an attempt to avoid the negative knock-on effects to our investors, we do not consider ETFs with smaller asset bases and limited trading activity. While not a direct input in assessing aggregate costs to an investor’s portfolio, any market impact measure that does not satisfy our criteria disqualifies the security from consideration.

## Conclusion

We are constantly monitoring our investment choices. Our selection analysis is run quarterly to assess the validity of existing selections, consider potential changes by fund administrators (raising or lowering expense ratios), or account for changes in specific ETF market factors (including tighter bid-ask spreads, lower tracking error, growing asset base, or reduced selection-driven market impact). We also consider the tax implications of portfolio selection changes and estimate the net benefit of transitioning between investment vehicles for our customers.

The power of this weighted aggregation methodology is how quickly it arrives at a total cost figure that synthesizes several dissimilar factors across many different candidate securities. The ability to quickly assess candidate suitability across the wider universe of potential options for each asset class is novel and incredibly useful in fulfilling our objectives of constantly providing a robust investment product, platform, advice, performance, and process control.

We will continue to drive innovation when trying to improve investor take-home returns by finding ways to lower costs and frequently re-evaluating our portfolio choices.

For a current look at how our asset classes are allocated, consult the following diagram:

#### Asset Class Weight at Every Allocation

We use the ETFs that result from this process in our allocation advice that is based on your investment horizon, balance, and goal.