Wealthfront’s Stock-level Tax-Loss Harvesting

This white paper summarizes the motivation, design and execution of Wealthfront’s Stock-level Tax-Loss Harvesting service. Our historical backtests and actual results demonstrate that Stock-level Tax-Loss Harvesting can significantly improve the tax efficiency and ultimately the after-tax return of your taxable portfolio. We believe Stock-level Tax-Loss Harvesting should be a fundamental component of most every individual’s investment management strategy.

Background

Over the last 40 years, index funds and index-based ETFs have unlocked the power of passive investing for the individual investor by offering a cheap, easy, and highly accessible way to invest in a wide range of asset classes. Just one share of Vanguard’s Total Stock Market ETF (Ticker: VTI) can enable an investor to mirror the investment performance of a portfolio of more than 3,500 US company stocks, all for an expense ratio of only 5 basis points (0.05%) of the invested assets.

Index funds work because they’re generally able to achieve two objectives:

1. They track their benchmark index as closely as possible. This is imperative to accurately represent a particular asset class.
2. They track their benchmark index as tax-efficiently as possible – passing on as few taxable gains as possible to their investors.

Index funds can achieve the latter goal because they are inherently tax-efficient. They have low turnover because securities that comprise an index seldom change (5-20% annual turnover depending on the index). Index fund issuers like iShares and Vanguard are also able to intelligently realize losses on the underlying securities in their funds to minimize the gains they have to distribute to their investors each year.

However, index funds and ETFs have two disadvantages relative to optimum performance in achieving the above goals and for their investors.

• Owning an index fund or ETF comes with a fee, often called an “expense ratio”. Although the fee may be small, it still acts as barrier to fully replicating the performance of a passive index investment. This expense ratio can add to an investor’s overall costs because it is typically in addition to any advisory fees or commissions the investor already pays.
• Index Funds and ETFs are legally prohibited by The Investment Company Act of 1940 from passing on tax losses to investors. While iShares and Vanguard can use any realized losses in their funds to minimize any gains they distribute to investors, any leftover losses cannot be used to reduce investors’ taxes even further.

Wealthfront’s Stock-level Tax-Loss Harvesting service was developed to address both of these shortcomings.

By allowing a Wealthfront investor to hold the individual securities that comprise an index in her own account on a commission-free basis, Stock-level Tax-Loss Harvesting effectively eliminates any Index fund or ETF expense ratios on the associated position, which reduces overall portfolio cost. Furthermore, by directly owning the stocks that comprise an index, investors can harvest losses at the individual stock level. We call this stock-level tax-loss harvesting. To understand the benefit of this capability, consider a common situation where an overall index trades up, but a number of its component stocks trade down because they missed their earnings estimates. The losses on these individual companies can be harvested and the resulting tax savings can be reinvested and compounded over time thus ultimately creating significant value.

The opportunity to deliver a Stock-level Tax-Loss Harvesting service prompted firms like Aperio Group and Parametric Portfolio Associates to pioneer managed portfolios of stocks that harvest tax losses for their clients while emulating a specific index. Unlike index funds, Aperio and Parametric prioritize both harvesting tax-losses as well as closely tracking the index. These firms have demonstrated you can generate significant outperformance in the form of tax-savings, if you are willing to incur modest tracking differences from the designated index. Together, these firms have attracted more than $100 billion under management.

Wealthfront is able to offer our Stock-level Tax-Loss Harvesting service to a much broader set of investors through the extensive automation provided by our technology platform. Our service is also offered at no additional fee beyond our annual 0.25% advisory fee that is inclusive of all commissions.

We believe our Stock-level Tax-Loss Harvesting service meaningfully addresses the two remaining shortcomings with modern index investing – the cost of the Index Fund and ETF expense ratios and the missed tax-savings from the inability to pass on tax losses. For this reason, we view Stock-level Tax-Loss Harvesting as the next evolution of index investing.

Stock-level Tax-Loss Harvesting Overview

Wealthfront’s Stock-level Tax-Loss Harvesting service replaces the ETF normally used to represent US stocks in a Wealthfront portfolio with a combination of individual securities and one or two additional ETFs. The resulting position uses the individually held stocks to capture the market performance of large-capitalization and mid-capitalization US stocks, while leveraging the ETF to represent the performance of smaller-capitalization companies. As a result, the position is able to track the movement of the overall broad US market, while maintaining a significant holding of individual stocks.

We offer three levels of Stock-level Tax-Loss Harvesting:

• For accounts between $100,000 and $500,000, we will replace the ETF normally used to represent a broad market of US Stocks (Vanguard’s Total Stock Market ETF) with up to 100 of the largest capitalization US stocks and a combination of the Vanguard Extended Market ETF (VXF) and the Vanguard S&P 500® ETF (VOO) to represent the remaining smaller-capitalization companies.
• For accounts between $500,000 and $1 million, we will replace the ETF normally used to represent a broad market of US stocks (Vanguard’s Total Stock Market ETF) with up to 500 of the largest capitalization US stocks and the Vanguard Extended Market ETF (VXF) used to represent the remaining smaller capitalization companies.
• For accounts above $1 million, we will replace our Vanguard Total Stock Market ETF with up to 1,000 of the largest capitalization US stocks (typically representing US large cap and mid cap companies) and the Vanguard Small-Cap ETF (VB) used to represent the remaining small-capitalization US companies.

For all levels of Stock-level Tax-Loss Harvesting, we will occasionally employ additional ETFs — especially in cases where the above ETFs are sold for tax-loss harvesting purposes. Thus, Stock-level Tax-Loss Harvesting clients may see a combination of the Vanguard Large-Cap ETF (VV), the Vanguard Small Cap ETF (VB), the Vanguard S&P 500® ETF (VOO), and the Vanguard Extended Market ETF (VXF) in their portfolios in addition to their Stock-level Tax-Loss Harvesting individual stock position. As shown below, the resulting Wealthfront account replaces Vanguard’s VTI with a combination of individual stocks and completion ETFs, but continues to hold the other ETFs common to Wealthfront investments (such as Foreign Stocks, Emerging Markets, Dividend Stocks, etc.). All Stock-level Tax-Loss Harvesting accounts also benefit from our Daily Tax-Loss Harvesting service – which harvests tax-savings at the ETF-level.

Wealthfront is the first and only company to combine daily asset-level Tax-Loss Harvesting and Stock-level Tax-Loss Harvesting. We offer this capability to accounts of only $100,000, one-fiftieth the size of what the competition requires.

Costs and Minimums

The cost of our Stock-level Tax-Loss Harvesting service is lower than the Vanguard ETF it replaces. Vanguard currently charges an annual 0.05% expense ratio for VTI, while the equivalent fee for all three levels of Stock-level Tax-Loss Harvesting is lower. That’s because we do not charge an expense ratio for the majority of the Stock-level Tax-Loss Harvesting position that’s comprised of individual stocks. The cost of that service (including all commissions) is included in our annual 0.25% advisory fee.

For $100,000 Stock-level Tax-Loss Harvesting accounts, that means no expense ratio is charged for 50% of the Stock-level Tax-Loss Harvesting position that’s comprised of individual stocks. The remaining 50% — comprised of roughly a 45% allocation to the Vanguard Extended Market ETF (ticket VXF, expense ratio 0.10%) and a 5% allocation to the Vanguard S&P500 ETF (ticker VOO, expense ratio 0.05%) – results in a weighted average Stock-level Tax-Loss Harvesting fee of 0.0475% — just under Vanguard’s 0.05% fee for VTI.

This is even better for larger accounts. For the $500,000 and $1 million levels of Stock-level Tax-Loss Harvesting, roughly 80% of the Stock-level Tax-Loss Harvesting position is comprised of individual stocks and thus has no expense ratio. The remaining 20% or so is comprised of either the Vanguard Extended Market ETF (VXF) for $500,000 accounts or the Vanguard Small-Cap ETF (VB) for accounts over $1 million. The expense ratio of these completion ETFs is 0.10% and 0.09% respectively, resulting in a weighted average Stock-level Tax-Loss Harvesting fee of 0.02% for $500,000 accounts (where VXF represents 20% of the position) and a fee of 0.014% for accounts over $1 million (where VB represents 16% of the position) – both significantly below the equivalent 0.05% fee from Vanguard.

Our minimums for Stock-level Tax-Loss Harvesting are based on dollar amounts required to hold a reasonable collection of individual US stocks in a Stock-level Tax-Loss Harvesting position while continuing to track the performance of the broad US market. For a $100,000 account, the average Wealthfront client will allocate about $30,000 to US equities. We found this $30,000 to be the minimum amount that allows for ownership of up to 100 Large Cap US stocks while maintaining a good ability to track broad US market performance.

To expand this position to hundreds of stocks requires a much larger account as you are forced to include allocations to smaller Mid Cap stocks in the position. We found that moving to hundreds of stocks requires a US equity allocation of roughly $150,000 – thus necessitating an account minimum of about $500,000. This minimum increases again as you exceed 500 stocks and start holding some of the smaller Mid Cap stocks in the position – thus necessitating a $1 million minimum for the highest level of our Stock-level Tax-Loss Harvesting service.

Wealthfront 100 (WF100), Wealthfront 500 (WF500), and the Wealthfront 1000 (WF1000)

We call the up to 100 individual stocks owned as part of our Stock-level Tax-Loss Harvesting service the Wealthfront 100. When Stock-level Tax-Loss Harvesting employs 500 or 1,000 individual stocks, we refer to that collection of stocks as the Wealthfront 500 or the Wealthfront 1000 respectively.

The individual stocks we buy are always selected to minimize tracking error with Vanguard’s Total Stock Market ETF, VTI, not based on their fundamentals or any perspective on whether they are fairly valued by the market. We harvest losses on individual stocks based on a threshold and use the proceeds to purchase other highly correlated stocks within the appropriate US stock index.

In some cases, we may purchase more of an existing holding. For example, if Coca-Cola misses an earnings estimate and drops precipitously in value we would sell Coke and use the proceeds to buy more PepsiCo to maintain the correlation with VTI in the absence of Coca-Cola.

Methodology

We balance two competing objectives with our Stock-level Tax-Loss Harvesting service: maximize the after tax benefit of harvesting losses (Tax Alpha) and minimize tracking error. We do this by maximizing a function of Tax Alpha minus tracking error squared. The tax alpha component encourages selling losing stocks. The tracking error component penalizes sales that cause significant tracking differences from the relevant index and encourages buying correlated replacement stocks that keep the overall portfolio close to the index.

We manage the portfolio and avoid wash sales by applying constraints to our optimization. For example, we enforce a maximum constraint on each stock’s relative portfolio weight to ensure portfolio diversification. We also enforce a maximum drift constraint on each stock’s weight to ensure it stays relatively close to its benchmark weight and trade direction constraints to avoid wash sales. The constrained optimization problem is a Quadratic Programming (QP) problem, which can be solved efficiently using a QP solver or a general purpose convex solver:

In addition to avoiding wash sales through a constraint on our optimization we also do not trade in stocks identified by the client via an “Exclusion List”. By adding stocks to an “Exclusion List” you’re able to identify stocks that you are restricted from trading (e.g., the stock of your employer) as well stocks where trading is more likely to cause wash sales due to significant individual stock positions held outside of Wealthfront. Once identified, the Stock-level Tax-Loss Harvesting optimizer will not suggest any trades in “Exclusion List” stocks and if needed will buy one or more highly correlated replacement stocks or ETFs to maintain the overall tracking of broad US stock market performance.

Performance Measurement

We measure the effectiveness of Stock-level Tax-Loss Harvesting on two dimensions: how closely it tracks the original ETF it replaced and how much benefit it generates from stock-level tax-loss harvesting.

We use the term tracking difference to describe the difference between the portfolio’s return and the original ETF’s return in a given time period and tracking error for the standard deviation of the tracking differences.

We track the benefit generated by Stock-level Tax-Loss Harvesting in two ways. Tax Alpha is used to directly measure the tax benefit generated by proactively selling stocks with capital losses within a certain short period of time (say a single tax year). A Differential IRR is used to measure the additional return generated by reinvesting tax-savings from stock-level tax-loss harvesting vs. a portfolio where no such savings are generated.

We view Tax Alpha as an easy to compute and understandable metric to compute the performance of Stock-level Tax-Loss Harvesting, but believe Differential IRR is the more appropriate metric because it does a better job of taking into consideration the additional cash flows, tax savings reinvestment, and multi-year compounding inherent in tax-loss harvesting.

• TrackingDifference = PortfolioPreTaxReturn – BenchmarkReturn
• TrackingError = StandardDeviation(TrackingDifference)
• TaxAlpha = (STCL * STTR + LTCL * LTTR) / PortfolioBeginningBalance
• Differential IRR = IRR(Stock-level Tax-Loss Harvesting Portfolio with tax savings reinvested) – IRR(Portfolio with same asset-allocation and cash flows but no Stock-level Tax-Loss Harvesting or Daily Tax-Loss Harvesting)
• STCL is the short-term net capital loss realized
• STTR is the combined federal and state short-term capital gains tax rate
• LTCL is the long-term net capital loss realized
• LTTR is the combined federal and state long-term capital gains tax rate
• PortfolioBeginningBalance is the value of the portfolio at the beginning of each year

Assumptions

Wherever possible we use assumptions for our analyses that are based on the actual observed behavior of Wealthfront clients.For this analysis, we were able to use two sets of client data. To accurately describe Wealthfront 100 clients, we used information on clients of our Daily Tax-Loss Harvesting service – which for several years has been offered at the same minimum ($100,000) as the Wealthfront 100 version of Stock-level Tax-Loss Harvesting. Similarly, to accurately describe our Wealthfront 500 and Wealthfront 1000 clients, we use data on clients that adopted an early form of Stock-level Tax-Loss Harvesting launched in December 2013 (with a $500,000 minimum).

Thus, our critical assumptions are:

• Client age: 37 (median age of our Stock-level Tax-Loss Harvesting andDaily Tax-Loss Harvesting clients)
• Marital status: Married (the majority of our clients are married)
• Annual income:
  • • WF100 Clients: $260,000 (the average joint income reported by our Daily Tax-Loss Harvesting clients)
    • WF500/WF1000 Clients: $422,000 (the average joint income reported by our Stock-level Tax-Loss Harvesting clients)
• State of residence: California (the most popular state of residence for our Stock-level Tax-Loss Harvesting andDaily Tax-Loss Harvesting clients. Residents of high state income tax states represent the vast majority of such clients.)
• Combined federal and state short-term capital gain tax rate:
  • • WF100 Clients: 42.7% (the marginal tax rate for married California clients with an average annual income of $260K = 33% federal tax rate + 3.8% additional tax on Net Investment Income for those earning above $200K (ACA tax) + 9.3% CA tax rate – the anticipated deduction of California state taxes from federal taxes)
    • WF500/WF1000 Clients: 44.5% (the marginal tax rate for married California clients with an average annual income of $422K = 35% federal tax rate + 3.8% additional tax on Net Investment Income for those earning above $200K (ACA tax) + 9.3% CA tax rate – the anticipated deduction of California state taxes from federal taxes)
• Combined federal and state long-term capital gain tax rate:
  • • WF100 Clients: 24.7% (the marginal tax rate for married California clients with an average annual income of $260K = 15% federal tax rate + 3.8% additional tax on Net Investment Income for those earning above $200K (ACA tax) + 9.3% CA tax rate – the anticipated deduction of California state taxes from federal taxes)
    • WF500/WF1000 Clients: 24.5% (the marginal tax rate for married California clients with an average annual income of $422K = 15% federal tax rate + 3.8% additional tax on Net Investment Income for those earning above $200K (ACA tax) + 9.3% CA tax rate – the anticipated deduction of California state taxes from federal taxes)
• Portfolio risk level: 7 (the average risk score on a scale of 0 – 10 for our Stock-level Tax-Loss Harvesting clients). A risk level 7 portfolio for such clients would be allocated across six asset classes as follows:
• Investment cash flows: We assume clients make an initial deposit necessary to meet the minimum account size of the desired level of Stock-level Tax-Loss Harvesting and then continue to make add-on deposits amounting to 10% of their initial deposit every quarter. For an account with the Wealthfront 100, therefore, this would mean an initial deposit of $100,000 followed by add-on deposits of $10,000 every quarter. For the Wealthfront 500 this would mean an initial deposit of $500,000 followed by $50,000 deposits every quarter.Note that these assumptions underestimate actual client behavior. Clients of our early $500,000-minimum version of Stock-level Tax-Loss Harvesting, for example, made quarterly deposits equal to an average of 14% of their initial deposit during the year that this service has been available – as reflected in the chart below.

Backtested Results

The graph below presents backtested return results for the three levels of Stock-level Tax-Loss Harvesting service compared to VTI (the ETF used to track a broad market of US stocks in a Wealthfront portfolio).

The results above assume the tax rates and cash flow patterns listed in our assumptions above and incorporate the added performance from the tax savings generated via Stock-level Tax-Loss Harvesting (i.e. Tax Alpha).

Stock-level Tax-Loss Harvesting outperformed VTI in a substantial majority of years, especially in bad markets. Interestingly it outperformed VTI in some very good markets as well. The average (geometric mean) after-tax annual return for VTI over the 14 year period was 3.22% vs. 5.99% for the Wealthfront 100, 6.09% for the Wealthfront 500, and 6.14% for the Wealthfront 1000.

As expected, adding more individual stocks to the Stock-level Tax-Loss Harvesting position (i.e. moving from the Wealthfront 100 to the Wealthfront 500 to the Wealthfront 1000) improved after-tax performance. This is in large part due to the extra tax-loss harvesting opportunities available with the larger set of individual securities.

Under the same assumptions, the graph below presents the Tax Alpha generated by Stock-level Tax-Loss Harvesting. Stock-level Tax-Loss Harvesting generated Tax Alpha nearly every year and had an average annual Tax Alpha of 2.15% for the Wealthfront 100 version, 2.50% for the Wealthfront 500 version, and 2.66% for the Wealthfront 1000 version. Interestingly, Stock-level Tax-Loss Harvesting generated Tax Alpha even in some years when the market traded up (2003 to 2007). It should be noted that our daily asset-level tax-loss harvesting service failed to generate much tax alpha in that same period.

The graph below displays the Stock-level Tax-Loss Harvesting tracking differences from its VTI benchmark by year based on the same assumptions. The tracking differences each year were modest. In some years they were positive (i.e. the portfolio outperformed the benchmark) and in some years negative (i.e. the portfolio underperformed the benchmark). In this limited time sample the average difference was very slightly positive, but over the long term you should expect an average difference of zero.

The table below summarizes the Tax Alpha, tracking difference, and tracking error for Stock-level Tax-Loss Harvesting vs. VTI based on annualized monthly tracking error and the same deposit assumptions as above. As expected, the Tax Alpha of the Stock-level Tax-Loss Harvesting increases as the number of individual stocks held increases (moving from the Wealthfront 100 to the Wealthfront 500 to the Wealthfront 1000) and the Tracking Error decreases.

Thus, for an extra 1.72% of tracking error, you could have the opportunity to increase your annual after-tax returns on the US equity portion of your portfolio by as much as 2.66%. We believe that is a trade off well worth making.

Finally, the graph below displays the overall portfolio differential IRR produced by an average (i.e. risk level 7) portfolio that includes the Wealthfront 100, Wealthfront 500, and Wealthfront 1000 across multiple 10-year holding periods. We also assume both portfolios benefit from Wealthfront’s Daily Tax-Loss Harvesting service.

We believe this differential IRR metric is the best way to quantify the incremental return from the Stock-level Tax-Loss Harvesting strategy. As the graph shows, Stock-level Tax-Loss Harvesting can generate an impressive incremental IRR for a Wealthfront portfolio over a 10-year period. The minimum IRR generated was 1.21% for the Wealthfront 500 over the 2005 to 2014 time frame. The maximum IRR was around 2.42% for both the Wealthfront 500 and the Wealthfront 1000 over the 2000-2009 timeframe.

Across these six 10-year periods, the Wealthfront 100 version of Stock-level Tax-Loss Harvesting generated an average IRR of 1.77%, while the Wealthfront 500 version generated an average IRR of 1.88%, and the Wealthfront 1000 version generated an average IRR of 2.03%.

Realized Results – 2019

The results above use backtested data. Now that Stock-Level Tax-Loss Harvesting (TLH) has been available for over five years, we can analyze its performance in comparison to ETF-Level TLH using live client account data from 2014 through 2018. We focus on a measurement called Harvesting Yield, defined as:

• Harvesting Yield = (Short-Term Capital Losses + Long-Term Capital Losses) / Starting Portfolio Value

Harvesting Yield measures the ability of each type of TLH to realize capital losses. The actual after-tax return benefit from loss-harvesting depends on each client’s individual tax rates, which we will not measure here. Because US Stocks is the only asset class in which Stock-Level TLH is implemented, we only compare the harvesting yield within the US Stocks portion of clients’ portfolios.

We take two important measures to help make the comparison between Stock-Level TLH and ETF-Level TLH cleaner. We first divide our client accounts into “cohorts” by the year of the initial deposit into the account. This helps control for differences in market conditions that may exist in different periods, which can have a significant effect on loss harvesting. We compare the harvesting yield of Stock-Level and ETF-Level TLH within the same cohort with respect to only the U.S. asset class portion of the client accounts.

Next we control for deposit and withdrawal patterns, which can affect an account’s ability to harvest losses. As an example, consider two accounts – one which makes a single initial deposit, and one which makes monthly deposits. If markets generally rise, the value of the positions in the first account will gradually drift above their cost, making it difficult to harvest losses except in the case of extreme market drops. In contrast, the second account is continually buying positions at the new, higher prices. This account will be able to take advantage of smaller market drops by realizing losses in the newer positions.

Due to the impact of cash flow on loss-harvesting ability, simply averaging harvesting yield across all accounts wouldn’t give an apples-to-apples comparison. To produce a valid one, we apply an algorithm to accounts within each cohort. The algorithm assigns a weight to every individual client account, and sums up the weighted portfolios to create an “aggregate account” for each tax-loss harvesting type. The weights assigned to the individual client accounts are chosen to minimize the sum of the total deviations between the daily US Stocks cash flow values of the two aggregate accounts. As an example, the portfolio of holdings of the aggregate account for the ETF-Level TLH group is the sum of the US Stock holdings of each individual ETF-Level TLH account, multiplied by the weight assigned to each account. The weights do not change over time, although the holdings of the client accounts, and thus the aggregate accounts, will. The algorithm is applied separately to each yearly cohort to produce five pairs of aggregate accounts (one pair for each year from 2014 to 2018).

Although the aggregate accounts produced by this algorithm are created using actual live client accounts, the tax-loss harvesting results of the aggregate accounts are not intended to be representative of a typical client account. The number of accounts receiving non-zero weight may be significantly smaller than the total number of accounts. In fact, we observe that only about 4% of ETF-Level TLH client accounts have a weighted contribution of more than $1 to the aggregate (that is, the account value multiplied by the account’s weight in the aggregate is at least $1). The algorithm is employed to remove the confounding effect of cash flow on loss-harvesting ability and provide a cleaner comparison between the two loss-harvesting methods.

The graph below shows the average quarterly net cash flow of the US Stocks portion of each aggregate account created by the algorithm, expressed as a fraction of current market value in US Stocks. Notice that the values for each TLH type are nearly identical, eliminating any potential difference in loss-harvesting ability caused by cash flow patterns.

Conclusion

We believe the results presented in this white paper clearly demonstrate that Wealthfront’s Stock-level Tax-Loss Harvesting could significantly improve your after-tax investment results.

Investing directly in stocks provides an immediate cost savings by avoiding the expense ratios charged by index funds and ETFs. ETF-Level TLH could be at an advantage in years with steep market declines, as we saw in 2018 where ETF-Level TLH outperformed Stock-Level TLH significantly. While times like this can (and will) occur, Stock-Level Tax-Loss Harvesting often provides better opportunities to harvest tax losses. We believe Stock-level Tax-Loss Harvesting reflects the next evolution in index investing and thus should be a key element of every investor’s portfolio.