Wealthfront Tax-Loss Harvesting White Paper

Summary

This white paper summarizes the motivation, design and execution of Wealthfront’s daily tax-loss harvesting service. Our actual results and historical backtests demonstrate that tax-loss harvesting significantly improves the tax efficiency and ultimately the after-tax return of Wealthfront’s taxable portfolios. We believe tax-loss harvesting should be a fundamental component of most every individual’s investment management strategy.

What is tax-loss harvesting?

Tax-loss harvesting is a way to make an investment portfolio work even harder – not just in generating investment returns, but by also generating tax savings.

Tax-loss harvesting (TLH) works by taking advantage of investments that have declined in value, which is a common occurrence in broadly diversified investment portfolios. By selling investments that have declined below their purchase price, a tax loss is generated – which can be used offset other taxable items, thus lowering the investor’s taxes.

What’s more, any investment sold in this manner can be replaced with a highly correlated alternate investment, such that the risk and return profile of the portfolio remains unchanged, even as tax savings are created. These tax savings can even be reinvested to further grow the value of the portfolio.

Wealthfront developed software to make tax-loss harvesting available to all our clients. Implementing tax-loss harvesting in software makes it possible to look for harvesting opportunities on a daily basis, which could result in significantly greater benefit than what could be achieved from the manual end-of-year approach typically taken by traditional financial advisors.

Tax-Loss Harvesting is a Tax-Deferral Strategy

Tax-loss harvesting can be used to defer tax liabilities, not avoid them. As a result it has very little value if applied over a short investment horizon (i.e. less than two years), but can be extremely valuable if executed over a long period of time. This is best demonstrated through an example that describes the benefits under two holding period scenarios:

Scenario 1 (one-year investment horizon)

Assume you buy 1,000 shares of ETF A at a price of $100 per share for a total investment of $100,000. A few months later, the market declines to the point ETF A is now worth $90 per share or $90,000 total. You sell all your shares of ETF A and replace it with an equivalent amount of ETF B (500 shares at $180 per share), which tracks an index that is highly correlated to the one ETF A tracks. You now own a $90,000 position in ETF B and have realized a $10,000 short-term loss for tax purposes. ETF B closes the year at $200 per share at which time you sell your entire position. Your sale of ETF B generates a $10,000 short-term gain which balances your $10,000 short-term loss. In this case harvesting the loss in ETF A generated no value.

Scenario 2 (five year investment horizon)

This scenario is much more complex and requires the table below to follow.Let’s assume you started off with the exact same $100,000 investment in ETF A and a few months later it declined 10% at which time you harvested the loss and replaced it with an equivalent amount of ETF B. Once again ETF B appreciates to $100,000 by the end of the year, but this time you don’t liquidate your account. As a result your net realized short-term capital loss on this position for the year is $10,000, which generates a $4,000 tax savings when applied against your income and other realized gains under the assumption of a combined federal and state tax rate of 40%. If you reinvest that tax savings your end-of-year portfolio value is now $104,000. The next year your portfolio appreciates 1.9% to $106,000. The following year your portfolio declines in value by 11.7% to $93,600 at which time you once again harvest a loss of $400 (on your previously reinvested tax savings), which generates a $160 ($400 x 40%) tax savings for that year. You reinvest the $160 and the following two years your portfolio is flat and increases to $104,160 respectively. At the end of year 5 you liquidate the portfolio. At this point you have to pay the tax due on the $10,000 appreciation of ETF B and the $400 loss in year three that is recaptured. However this time the gains are long-term capital gains, so you pay the lower 25% combined federal plus state tax rate, which results in a negative $2,600 cash flow ($10,400 long-term capital gain x 25%).

As you can see Scenario 2 results in a significant overall tax savings even though all of the realized losses are ultimately balanced with an equivalent realized gain. The positive economic benefit that results from this scenario can be attributed to the reinvestment of the tax savings and the significant difference in tax rates between short-term capital gains and long-term capital gains.

The value of tax-loss harvesting is maximized if your portfolio is never liquidated. This is possible even when you expect to employ your portfolio for your retirement needs. For example, you can place your appreciated assets in a Charitable Remainder Trust (CRT) or similar vehicle – generating income for yourself in retirement while donating the remainder of your portfolio to a non-profit organization of your choice. Because non-profits are not taxed, they are able to sell your highly appreciated assets and generate retirement income for you with no regard to the basis or tax-liability of those assets.

Similarly, you can choose to pass on all or a large percentage of your portfolio to your heirs. Because your heirs receive a step-up in basis for the portfolio on your death, they are again able to take full advantage of the portfolio’s value without regard to its low tax basis or its tax liability.

Finally, instead of fully liquidating your portfolio in retirement, you can liquidate only a part of the portfolio and pass on or donate the rest. Such partial liquidation approaches, when done tax-efficiently, can again allow you to keep most of the gains of tax-loss harvesting.

Who should use tax-loss harvesting?

Tax-loss harvesting is generally valuable for all taxable investors who have a long-term investment horizon. Harvested losses can be applied to offset both capital gains and up to $3,000 in ordinary income annually. Furthermore, any losses that cannot be applied in a given tax year can be carried over indefinitely to offset future income and capital gains.

Of course, tax-loss harvesting is especially valuable for investors who regularly recognize short term capital gains. These gains can come from the sale of company stock, real estate or just about any investment. Tax-loss harvesting can even be used to minimize the gains associated with liquidating a portfolio to move to a new financial advisor.

Wealthfront Daily Tax-Loss Harvesting service

For taxable accounts, Wealthfront offers the Wealthfront daily tax-loss harvesting service for no additional charge. We monitor your portfolio daily to look for opportunities to harvest losses on the ETFs that represent each asset class in your portfolio. Under the right circumstances we will sell one of your ETFs that is trading at a loss and replace it with an alternative ETF that tracks a different, but highly correlated index to maintain the risk and return characteristics of your portfolio. We will then hold that alternative ETF in your portfolio for a minimum of 30 days to avoid wash sales (described below). We will not sell the alternative ETF until it can be sold for a loss (which will generate additional tax-loss harvesting benefit).

The following two tables list the primary and alternative ETFs we use to represent each asset class, their associated indexes, expense ratios and correlations in weekly returns between the primary and alternative ETFs estimated using the longest common available period of data ending on October 31, 2017.

Our asset class level tax-loss harvesting strategy uses a cost-benefit analysis framework to evaluate potential harvesting opportunities for each ETF lot currently trading below its cost basis as follows:

Definitions:

Benefit is calculated by multiplying the potential realized capital loss incurred from selling an ETF times either the short-term or long-term capital gains tax rate, depending on the ETF’s holding period.

Cost captures the trading cost of selling the primary ETF and buying the secondary ETF. (Note: this cost is just the bid-ask spread of the two ETFs because Wealthfront clients never incur any commissions.)

Threshold is an estimate of the expected future harvesting benefit which is modeled by assessing the likelihood that a better harvesting opportunity is available by waiting longer to perform the tax-loss harvesting trade (and thus potentially capturing more of the ETF’s decline). Expected return and volatility of each asset class are used as parameters in this estimate.

We execute two trades when we believe the Benefit – Cost exceeds the Threshold:

1. We sell the specific ETF lot to recognize a loss, and
2. We purchase the same dollar amount of a similar, but not substantially identical, ETF (see the table above for the alternative ETFs we use for each asset class) to maintain the desired asset class exposure.

We evaluate every ETF tax lot in each eligible account every day, and execute trades when our system identifies the necessary conditions.

Wash sale management

One complexity in executing our daily tax-loss harvesting strategy is the management of wash sales. Normally, you recognize a loss when you sell a security for less than its cost basis. However, if you buy the same or substantially identical security within 30 days of the sale, the wash sale rule applies and you are not allowed to claim the tax benefit.

It should be noted that a wash sale does not completely eliminate the benefit of harvesting a loss, but only reduces that benefit by the amount of the wash sale itself. For example, if you sell 1,000 shares of a particular ETF to harvest a loss, but happen to buy 10 shares of the same ETF within 30 days, you will still be able to take advantage of the harvested loss on 990 of the 1,000 shares originally sold – as only 10 shares are actually subject to the wash sale rule in this situation.

In the case of applying tax-loss harvesting to a portfolio of index funds or passive ETFs, you need to use two securities that track different indexes to avoid violating the substantially identical clause of the wash sale rule. Swapping an ETF with another that tracks the same index from a different issuer (i.e. Vanguard vs. Schwab) would violate the substantially identical rule. As a result, you’ll see that the alternative ETFs presented in the table above track different, but highly correlated indexes from the recommended primary ETFs.

The risk of a wash sale increases with the number of rebalancing trades made as part of the ongoing management of a portfolio. If not carefully managed, a security sold to harvest a loss might be re-purchased for another reason (change in asset allocation, rebalancing, dividend reinvestment, withdrawal or deposit) and the tax benefit could be disallowed, defeating the purpose of tax-loss harvesting. We carefully manage the interaction and timing of trades within our portfolio management service to help avoid wash sales.

The situation becomes more complex in the case of multiple accounts, as the wash sale rule applies to all of an investor’s accounts including IRAs – as well as spousal accounts for joint filers. For example, if an investor sells a security in her personal non-retirement account and repurchases it within 30 days in her IRA, it’s considered a wash sale; if her spouse purchases the same security within 30 days in his personal non-retirement account, it’s also viewed as a wash sale. Wealthfront monitors all the accounts it manages for each client to avoid any transactions that might trigger a wash sale. Additionally, clients can link a spousal account to ensure that wash sales are not generated across accounts belonging to the same taxable household. We urge all our clients to consult with their tax advisors to confirm that our tax-loss harvesting strategies are right for them.

Performance measurement

We evaluate the effectiveness of our tax-loss harvesting service using two techniques: actual results realized by Wealthfront clients (since the service was launched in October 2012) and a hypothetical backtest, reflecting the retroactive application of our algorithm to historical data (2000-2012). We quantify the effectiveness of our tax-loss harvesting algorithm using its Harvesting Yield. The Harvesting Yield measures the quantity of harvested losses (short or long-term) during a given period, divided by the value of the portfolio at the beginning of the period. The ultimate benefit each client will receive will depend on her particular tax rates.

We calculate daily Harvesting Yield using the following formula:Definitions:

• STCL is the short-term net capital loss realized
• LTCL is the long-term net capital loss realized
• PortfolioBeginningBalance is the value of the portfolio at the beginning of each year

To arrive at annual figures, we sum the Harvesting Yield for each day within a given year.

Results

We present two sets of results: (a) actual client results realized since the launch of the daily tax-loss harvesting service in October 2012; and, (b) hypothetical backtested results for the period 2000-2012, preceding the launch of the service. The hypothetical backtest provide a rich context, within which to interpret the realized data, consider the performance of the service over a wider range of economic circumstances (including the bursting of the Internet Bubble and the Credit Bubble in 2001 and 2008, respectively), and assess the efficacy of Wealthfront’s tax-loss harvesting service.

Realized Client Data

In order to analyze the account-level results realized by clients, we aggregate clients into cohorts based on the year in which they first started using TLH (i.e. the client vintage) and their risk score. The vintage is a fixed client characteristic, but clients can move across risk score groupings based on their risk score on a given day in the sample. On each day, we compute the aggregate losses harvested within portfolios of clients belonging to a given cohort, sum these losses across clients in the cohort, and divide them by the aggregate portfolio balance of clients belonging to the cohort, to obtain that day’s Harvesting Yield. We then compute an annualized “since inception” Harvesting Yield for each cohort, by summing the daily Harvesting Yield figures, dividing by the number of trading days since inception, and multiplying by 252 (the number of trading days in a year). The values for the 2017 cohort are reported unannualized, since a complete year of tax-loss harvesting data is not yet available. The annualized Harvesting Yields for each vintage by risk score cohort are reported in Table 3. The data include all tax losses harvested through October 31, 2017. Within any given vintage year, the realized Harvesting Yield generally increases with portfolio risk score, because higher risk levels result in higher allocations to more volatile asset classes (such as stocks) and will thus result in more harvesting opportunities. This pattern plays out across risk scores within a vintage year, as well as, in the cross-vintage average figures, reported in the column titled “Average.” That being said, you should not increase your risk level to get more tax-loss harvesting benefit because it is more likely to lead to a level of portfolio volatility that will cause you to prematurely liquidate and lose money (see DALBAR analysis).

The second pattern that can be observed in client data is that the Harvesting Yields can vary substantially across vintages for a given risk score. This is because each vintage experiences a different set of market conditions, which affect the scope of the harvesting opportunities. For example, the periods in which markets experience large peak-to-trough swings, generally create more tax-loss harvesting opportunities, than placid periods.

To provide a simple summary statistic for the scope of harvesting opportunities, the last row of the table (“Maximum Drawdown”) reports maximum loss that an investor in the S&P 500 would have sustained between the first business day of each vintage and the last calendar day of our sample (October 31, 2017). For example, a maximum drawdown of 0% for the 2013 vintage means that at no point subsequent to the initial investment (made on January 2, 2013) through the end of the sample was the S&P 500 below its value on the first business day of 2013. Similarly, for the 2016 vintage, the largest cumulative drop experienced by an investor who made her initial investment on January 4, 2016 (the first business day of 2016) through the end of the sample was -10.29%.

Within each risk score, there is a strong negative correlation between the annualized harvesting yields realized by the different cohorts and the maximum drawdown realized by the S&P 500 over the corresponding time period. Importantly, even for vintages where markets were broadly trending up (i.e. the maximum S&P 500 drawdown was zero), our daily tax-loss harvesting service can still take advantage of intermediate market swings to harvest losses. As we will show in the next section, this reflects a meaningful advantage of daily tax-loss harvesting relative to more traditional approaches, such as year-end tax-loss harvesting.

Backtested Data

We next use a backtest to examine the properties of our daily tax-loss harvesting service in more detail. This analysis spans the period prior to the introduction of the service (January 2000 – October 2012), and does not reflect the actual results of any client. Using a backtest allows us to assess the efficacy of the algorithm in a laboratory-like environment, where assumptions can be varied individually, and examine its performance over a longer time period. Wherever possible the backtest is based on historical return data for the primary and secondary ETFs described earlier. For periods preceding the introduction of each ETF, we backfill returns back to January 2000 by splicing the ETF’s returns with the returns of a comparable mutual fund or index. Since VTEB was not launched until after the ending date of the backtest, we use MUB as the primary ETF for the municipal bond asset class. Details of the backfilling logic (instruments and dates) are described in the Disclosures.

Assessing Tax-Loss Harvesting Efficiency

To assess the efficacy of different approaches to tax-loss harvesting, we compared their backtested Harvesting Yield relative to the theoretical maximum Harvesting Yield possible in a given year. The theoretical maximum Harvesting Yield can be computed by performing the backtest in a manner that assumes all future prices are known and thus all harvesting transactions happen at the absolutely perfect point in time (i.e. the lowest price of any given asset in every tax year).

We also compare our daily tax-loss harvesting service, which checks for harvesting opportunities on each day, with the more common tax-loss harvesting approach employed by traditional advisors that checks your portfolio only once per year (typically at the end of the year). For each tax-loss harvesting strategy, we report its Harvesting Efficiency Ratio computed as follows:By definition, the maximum Harvesting Efficiency Ratio that any strategy can achieve is 100%.

For the purposes of the backtest we assume that an investor initially deposits $100,000 and then deposits $10,000 each quarter thereafter. The rate at which funds are being added equals 40% (= 4 * $10,000 / $100,000) in the first year and 30% (= 4 * $10,000 / $140,000) in the second year, which is more conservative than the historically observed add-on rate of our clients, who are young and in the wealth accumulation phase of their lives.

We subdivide the backtest period (January 2000 – October 2012) into nine overlapping, five-year subperiods. The figure below displays the Harvesting Efficiency Ratio for our daily tax-loss harvesting service, and a more traditional, year-end loss harvesting for each of the five-year subperiods for a client invested in a Risk Score 8.0 taxable portfolio. We present detailed results for the Risk Score 8.0 portfolio as this is the most commonly chosen risk score among our clients. On average the daily tax-loss harvesting service captured 81% of the theoretical maximum Harvesting Yield, as compared to 47% for year-end tax-loss harvesting. As you can see in Figure 1, the Harvesting Efficiency Ratio of the daily service is remarkably stable across the five-year subperiods, ranging between 72% and 95% of the theoretical maximum. This stands in contrast to the Harvesting Efficiency Ratio of year-end tax-loss harvesting, which ranged from 7% to 74%.This result carries over to all portfolios independent of their Risk Score. We repeated the above analysis for taxable portfolios with Risk Scores ranging from 1.0 to 10.0, and in Figure 2 reported the average Harvesting Efficiency Ratio achieved by each algorithm across the nine, five-year subperiods. Across the various portfolios, daily tax-loss harvesting achieved a Harvesting Efficiency Ratio of 80%, as compared to, 55% for year-end harvesting. These results lead us to believe there is very little additional benefit that can be gained through additional tweaks to our daily tax-loss harvesting service. Although small improvements are theoretically possible, there’s unlikely to be another daily asset class-based tax-loss harvesting algorithm that’s significantly better than our implementation. This assertion is backed up by experiments we ran testing the actual observed efficacy of our competitors’ tax-loss harvesting services.

Economic Value of Tax-Loss Harvesting

The economic value of tax-loss harvesting depends on a broad range of assumptions describing the specifics of the client’s tax situation. For example, (a) the availability of taxable income items which may be offset with harvested losses (short-term capital gains, long-term capital gains, and up to $3,000 in ordinary income per year); (b) the tax rates applicable to the items being offset (both federal and state); (c) the client’s tax rate at the time the portfolio used to generate the tax losses is liquidated (which can change due to changes in tax policy or changes in income); (d) the client’s investment horizon, as well as; (e) the rate of return on reinvested tax savings.

Tax Rates

Table 4 below displays the current income tax brackets, along with the prevailing federal tax rates for ordinary income, dividends, interest, as well as, short-term and and long-term capital gains:In addition to the above federal tax rates, income is subject to taxation at the state level. State tax rates vary considerably, from no income tax to rates of over 12% for those in the highest brackets in California. As such, your income and short-term capital gains can be subject to taxation ranging from roughly 10% (lowest federal tax bracket + no state tax) to over 50% (highest federal tax bracket + highest California tax brackets). Long-term capital gains are subject to lower levels of taxation, varying from 0% to over 30%. The incremental 3.8% tax applicable to capital gains for those in the highest income brackets reflects the additional tax on investment income imposed by the Affordable Care Act.

Tax Savings and Liabilities

The economic value of the harvested losses can be measured using the difference between the tax savings generated today (due to the ability to offset taxable income items) and the present value of the increased tax liability in the future (due to the decreased tax basis of the instruments used to harvest the tax losses). The Tax Savings today are given by the following formula:Where:

• Amount Offset is amount of tax losses utilized
• Current Tax Rate is the combined federal + state tax rate applicable to the taxable item being offset (ordinary income, short-term capital gains, or long-term capital gains) adjusted for the deductibility of state taxes at the federal level.

Where:

• Amount Offset is amount of tax losses utilized
• Future Tax Rate is the long-term capital gains rate prevailing at liquidation
• Discount Factor is a discount applicable to payments occurring in the future to account for the time value of money. It is given by exp(y(T) * T), where: y(T) is the continuously-compounded yield on T-year Treasury bond (available here), and T is number of years until projected portfolio liquidation.

The client benefits from the deferral of taxation, as well as, the wedge between the tax rates applicable today, vs. the lower, long-term capital gains rate, which will apply to the portfolio gains at that time. In some instances, described earlier, clients may even be able to avoid taxation altogether.

Example

To illustrate the economic value of tax-loss harvesting, let’s consider an example, where a client has an investment portfolio worth $100,000 and has harvested $4,500 in tax losses in a given year, reflecting a Harvesting Yield of 4.5%. (As we showed earlier this Harvesting Yield roughly matches the cross-vintage average realized experience of clients with risk score 8.0 taxable portfolios, the most commonly selected risk score among our clients). Further suppose that the client has generated $1,000 in short-term capital gains (either at Wealthfront or elsewhere), and $1,500 of long term capital gains (again either at Wealthfront or elsewhere). The $4,500 of harvested losses could be used to offset $1,000 of short-term capital gains, $1,500 of long-term capital gains, and $2,000 of ordinary income. (The ordinary income offset is limited to $3,000 per year, and is applied only after all net capital gains have been offset)

To give a sense of the range of economic benefits generated, we consider two clients differing in their tax burden. The client subject to the lower tax burden faces a combined tax rate of 25% on ordinary income and short-term capital gains, and 15% on long-term capital gains today. The client subject to the higher tax burden faces combined tax rate of 45% on ordinary income and short-term gains, and 25% on long-term capital gains. We’ll assume that each client intends to completely liquidate the portfolio in twenty years, at which point their gains will be subject to taxation at long-term capital gains rates, equal to those prevailing today (15% for the low tax burden client, and 25% for the high-tax burden client). We use a 20-year risk-free Treasury yield of 2.7%, reflective of yields prevailing as of October 2017, to compute the present value of the future tax liability. Table 5 below computes the tax savings, present value of the increased tax liability, and the net economic benefit to each client type.

When expressed as a fraction of the portfolio value, this economic benefit amounts to an incremental return of 0.58% in the current year for the client facing the low tax burden, and an incremental return of 1.07% for the client facing the high tax burden. Both of these values compare very favorably with Wealthfront’s annual advisory fee of 0.25%.

In these examples, the client was able to use the full amount of the harvested losses ($4,500). If the client did not have sufficient taxable gains in a given year to take advantage of the full amount of the harvested losses, the excess losses are carried forward indefinitely for use in subsequent years. Finally, it is worth pointing out that since future tax rates are uncertain, there may exist scenarios whereby future long-term tax rates may rise to a level where the economic benefit is negated entirely. In our example, this would happen if the long-term capital gains rates prevailing at liquidation — 20 years into the future — were to rise to a level greater than 37% (= $975 / ($4500 * exp(-0.027 * 20))) for the client facing the lower tax burden, and 66% (= $1725 / ($4500 * exp(-0.027 * 20))) for the client facing the higher tax burden.

As you can see, the precise details of the computation determining the value of tax-loss harvesting are significantly affected by the particulars of each client’s tax situation. In general, the greater the wedge between the tax rate applicable to the income items being offset today and the long-term capital gains rate applicable when the portfolio is liquidated, the greater the economic benefit of tax-loss harvesting. Similarly, the longer the time between the client’s investment horizon, and the higher interest rates, the lower the present value of the future tax liability, and thus, the greater the benefit of tax-loss harvesting.

Discussion

Tax-loss harvesting may be a highly valuable way to improve the tax efficiency of your portfolio, but it’s important to understand that it is a tactical approach and should not interfere with your strategic investment objectives such as broad diversification though an optimal asset allocation, disciplined and tax-efficient rebalancing and low-cost indexing.

Since tax-loss harvesting systematically lowers the cost basis of a portfolio by replacing securities (ETFs or stocks) that trade at a loss, investors might be concerned with how to deal with a portfolio with a low cost basis in the future. Generally speaking, portfolios with a low cost basis require more careful disposition. Investors may choose to minimize their tax liabilities that result from tax-loss harvesting by using the low basis portfolio as a charitable donation or passing the portfolio on to heirs through an estate at a stepped-up cost basis.

Tax-loss harvesting is only relevant for taxable accounts. It does not apply to tax-deferred accounts such as IRAs and 401(k) accounts, since gains and losses in those accounts are not taxable events. However alternative ETFs may need to be used when investing your tax deferred account to avoid wash sales. Tax-loss harvesting is also not typically suited for custodial accounts such as UTMA/UGMA unless the minor has significant taxable capital gains or income from other sources.

Conclusion

We believe the actual realized and backtested results presented in this white paper clearly demonstrate that Wealthfront’s fully automated daily tax-loss harvesting service could significantly improve the after-tax returns for young people with a long investment time horizon. We thus believe tax-loss harvesting should be a fundamental component of most every individual’s investment management strategy.