Is Smart Beta Still Smart after Taxes?

AFTER-TAX RETURN CALCULATIONS

To illustrate the effect of taxes on a portfolio, we use the highest federal marginal rates for U.S. individual investors: 23.8% for long-term gains and dividends, and 43.4% for short-term gains.² We assume these tax rates are applied over the entire history. An actual investor would have experienced various tax rates over time, as the laws and tax code evolved. Our goal is not to precisely estimate the tax experience of the past, but rather to project the effect of the current tax rate going forward. We also calculate the after-tax returns using tax rates of 15/35% and 33/50% to understand how sensitive our results are to the tax-rate assumption.

Our results generally hold true for other types of taxable investors as well, such as corporations, nuclear decommissioning trusts, and insurance companies. Clearly, the tax rates and specific rules vary by country and investor type. For example, tax rates for individuals outside of the U.S. are often higher than the rates we use in this article, with differing rules on how dividends and capital gains are taxed. Australia is somewhat unique, in that retirement assets in superannuation funds are subject to taxation. Despite differences in the details of the tax calculation, the basic policy of collecting tax on capital gains from the sale of an investment security is nearly universal.

To account for taxes in our returns, we assume that dividends are reinvested monthly and annual returns are calculated as described below. For each month m = 1, 2, …, 12 in year n = 1, 2, …, N:

1

2

2

Taxes are paid annually from the investment balance. Realized capital gains from positions that were held longer than a year are taxed at the long-term capital gain tax rate, and otherwise at the short-term rate. Dividends are taxed at the long-term rate. We assume that realized losses offset realized gains; however, if there are excess losses in a year, those losses are carried forward for use in subsequent years.

4

where, t_n = tax on dividends and realized gains in year n.

5

For post-liquidation return, we subtract the liquidation taxes from the final after-tax balance. The liquidation tax must be positive (i.e., losses carried forward can reduce the liquidation taxes, but there is no credit given for any excess losses after all the gains have been offset).

6

where, L = taxes on liquidation at year N and L = 0.

In addition to the traditional excess return measure, we calculate an after-tax excess return, which compares the after-tax return with the after-tax return of the capitalization-weighted benchmark.

7

8

We also calculate the tax and liquidation effects. These measures answer the question: How much did taxes reduce my return?

9

10

This methodology is more conservative than the standard SEC method of calculating after-tax returns, which is required of registered mutual funds. Instead of applying the tax effect to the portfolio each month, we apply it only at year end. Also, we do not give credit for a tax loss, unless it is used to offset a gain. Therefore, in this study, pre-tax returns will always be higher than after-tax returns.

EMPIRICAL RESULTS

We begin with an analysis of the pre-tax returns, to provide some context for comparing the strategies. Exhibit 1 displays the return characteristics for the seven strategies versus the capitalization-weighted index for global and U.S. stocks. The results are substantively similar to those reported in the CHKL paper. The differences are due to the change in time period to 1993 to 2012. In this 20-year period, we observe that the global maximum-diversification portfolio matches the cap-weighted return almost perfectly. In contrast, the longer dataset used in the CHKL paper indicates that maximum diversification outperformed. A complete discussion of the pros and cons of the various strategies can be found in CHLK, as well as in Clare et al. [2013]. In our analysis, all of the strategies studied outperform the capitalization-weighted index on a risk-adjusted basis, some quite dramatically.

For all the smart-beta strategies, we observe that the turnover is higher than the capitalization-weighted benchmark, with optimization-based strategies having a relatively higher turnover than the heuristic-based strategies.³ This increase in turnover has two effects: an increase in trading costs and an increase in taxable events. The results presented here are gross of fees, commissions, and other trading costs. For every 100% turnover in a large-cap developed market portfolio, Pritamani [2013] estimates 9 basis points of explicit costs, such as commissions, and 9 basis points of implicit trading costs, such as market effect and the bid–ask spread. For example, in our study, the global minimum-variance portfolio has a 47.68% turnover, relative to 7.69% for the cap-weighted index. The extra 40% in turnover will create extra trading costs on the order of 7.2 basis points.

Though trading costs are a concern, taxes have a much greater effect on the return than trading costs do. Exhibit 2 presents the after-tax and post-liquidation returns for each strategy. Taxes reduce the returns of the capitalization-weighted index by about 1% per year, with approximately 0.5% of the tax drag attributed to the dividend tax, and the remainder to the capital-gain tax. In contrast, in the smart-beta strategies, taxes reduce the returns by between 1.5% and 2.5% per year. The dividend tax is generally the same across all of the strategies and the index. The increase in tax effect comes from an increase in realized capital gains. If we assume assets are liquidated at the final period, we see that some strategies have built up an embedded tax liability. In these cases, investors pay another 10 to 50 basis points of tax at liquidation after 20 years.

To highlight these results, we plot the pre- and after-tax excess returns versus the capitalization-weighted benchmark for each strategy in Exhibit 3. Although these smart-beta strategies were able to produce excess long-term returns on a pre-tax basis, they lost a significant part of the excess due to their higher tax bill. Nevertheless, this result is somewhat heartening, compared with previous studies of active mutual funds where, on average, taxes consumed all of the excess returns. The smart-beta strategies, in contrast, have lower turnover than do typical active strategies and, as a result, are more tax efficient. Still, the effect of taxation is significant. Tax is a drag on return.

• Download figure
• Open in new tab
• Download powerpoint

Exhibit 3

Excess Returns, 1993–2012

We have identified turnover as the proximal cause of tax inefficiency. The simplest strategy in which to study turnover is equal weighting. It is easy to see how the strategy requires extra turnover in order to make the security weights equal at each rebalancing. To rebalance, the strategy sells securities that have appreciated in value and buys securities that have lost value. By design, the equally weighted portfolio accumulates more realized gains than a capitalization-weighted portfolio.

More broadly, for all the strategies, market changes induce portfolio turnover. For example, the technology bubble in the late 1990s created extra trading in all the strategies. For optimized strategies (and risk clustering), risk-model changes mean additional turnover. Exhibit 4 shows how turnover affects the portfolios’ after-tax performance in each of the strategies. Each observation represents a calendar year for one of the strategies. The style of the data point distinguishes between the cap-weighted, heuristic, and optimized strategies. The heuristic-based strategies display lower turnover and less tax effect than the optimized strategies do. In general, the higher the turnover, the more negative the tax impact.

• Download figure
• Open in new tab
• Download powerpoint

Exhibit 4

Annual Observations of Turnover vs. Tax Impact for Global Strategies

To analyze the relationship between turnover and tax impact, we fit an ordinary least squares (OLS) model, regressing the tax impact on turnover. The model has an R-squared of 0.48, a coefficient slope of -0.0337, and a t-stat of -6.76. In other words, for every 100% turnover, we expect a -3.37% reduction in returns, due to taxes. This result is statistically and substantively significant. We also analyzed the U.S. data (not shown here) and found a similar result with a -4.22% slope on the regression.

To provide a sense of the results’ sensitivity to the tax rates, we perform all of our empirical analysis with 15% long-term and 35% short-term tax rates to simulate a lower tax-rate investor, and 33/50% to simulate a higher tax-rate investor (for example, adding state tax to the federal tax rates). The resulting tax impact, regardless of strategy, scales as a simple multiple of the tax impact in Exhibit 2. The lower tax-rate investor experiences 65% of the tax impact, and the higher tax-rate investor experiences 137% of the tax impact.

In this section, we have shown that smart-beta strategy portfolios, like other active strategies, have higher turnover and are less tax efficient than are capitalization-weighted portfolios. Fortunately, to minimize the impact of taxes and preserve the pre-tax excess returns, these strategies can be tax managed. Interestingly, these techniques for reducing taxes create even more turnover in the portfolio—so from a tax perspective, we observe that there is “bad” turnover and “good” turnover. In the next section, we discuss tax management and measure the effect of these techniques on the smart-beta strategies’ returns.

TAX MANAGEMENT

There are three real-world truths that make tax management difficult for many investment managers: 1) the tax code is complex, 2) it punishes those who trade frequently, and 3) large tax-exempt institutions do not care about taxes. Despite these hurdles, there are strategies that can improve after-tax performance. Stein and Narasimhan [1999]; Arnott et al. [2001]; and Chincarini and Kim [2001] discuss the effect of taxes on various types of investment strategies and explore the impact of various tax-minimization strategies.

Not all strategies are affected by taxes to the same extent. Israel and Moskowitz [2011] and Bergstresser and Pontiff [2012] find that investments’ taxation varies by style. Equally weighted portfolios, small-firm portfolios, and value portfolios tend to have higher exposure to capital-gains taxation, whereas large-stock portfolios and growth portfolios tend to have lower exposure to capital gains taxation.

In general, investment managers use five basic tax-management techniques to increase a portfolio’s tax efficiency.

• 1. Defer the realization of gains. The government only taxes an investment gain if the associated asset is sold; the tax liability is deferred as long as you hold the asset. An investor can defer the payment of that liability indefinitely and allow the deferred tax amount to compound over time.

• 2. Manage the holding period. Capital gains from the sale of a security are taxed as ordinary income unless the investment is held for longer than 12 months, after which it is taxed at a lower long-term rate. Dividends are also taxed as ordinary income, but can qualify for a lower tax rate if the security is held for longer than 61 days.

• 3. Harvest losses. Selling a security whose price has fallen below its purchase price (the market value is below the cost basis) results in a realized tax loss. These losses can offset realized capital gains.

• 4. Pay attention to tax lots. Most managers who pay attention to taxes will use highest in, first out (HIFO) tax-lot accounting whenever a security is sold, as it reduces the sale’s tax impact and improves after-tax returns. In some cases, identifying specific tax lots can improve tax efficiency. For example, an investor with a tax-loss carry forward may find it beneficial to accelerate gains. Investors who want to generate cash flow from their investments will benefit from a manager who pays close attention to tax lots.

• 5. Avoid wash sales. Repurchase a security within 30 days of its sale, and you cannot use any realized losses to shelter gains. Instead, the loss is added to the basis of the repurchased shares, negating the benefit of loss harvesting.

Tax-aware portfolio managers often embed these rules within the systems for trading and managing the portfolio. There is a trade-off between implementing the trades exactly as the strategy specifies and adjusting the trades to accommodate the tax situation. For example, a manager implementing a smart-beta strategy may find that the benefit of holding a security for tax reasons outweighs the benefit of selling to rebalance to the specified strategy. This trade-off is more challenging for active strategies that have a stock-specific investment thesis.

TAX-MANAGEMENT METHODOLOGY

When we use tax management, we expect the portfolio to track the smart-beta strategy’s performance closely—but not perfectly—and to have superior after-tax performance, relative to the original strategy. To test this hypothesis, we run a second series of back tests, this time including tax management. To implement the five techniques described, we systematically rebalance each month to our drifted-strategy portfolio, to harvest losses and defer gains within simple stock-level risk bounds.

For example, consider a portfolio security weight that is above its target weight. If the position is at a capital gain, then a rebalancing trade will trigger a tax cost. Alternatively, we can accept some active risk relative to our ideal portfolio, in order to avoid some of the tax cost. The stock-level bounds define how much active risk we are willing to take to achieve a tax benefit. If the target weight is 1%, perhaps our risk management system tells us that any weight between 0.5% and 1.5% is acceptable. To defer gains, we could then sell only to the upper bound of 1.5%. On the other hand, if the position were at a loss, we could sell to the lower bound of 0.5% and realize tax losses that could later be used to offset realized gains. If some tax lots were at a gain and others at a loss, we could intelligently select the appropriate tax lots to sell.

We carefully control the gain deferral and loss realization to prevent the portfolio from holding too many stocks at upper or lower bounds. While deferring gains, we check whether the average gain among all the lots in a security is greater than a pre-defined threshold value. If the security average gain exceeds the threshold, then we defer gains by selling the security to the upper bound. Otherwise, we sell the security to the target weight and realize the relatively smaller gains. Similarly, we realize losses only if the average loss in a security lot exceeds a threshold value. This process prevents the portfolio from realizing more turnover than necessary.

In addition to applying tax management during the rebalancing events, we also employ active loss harvesting in the portfolio to realize losses. For example, if a security’s weight is still within the target bounds but has some built-in losses, we realize those losses by selling the security to its lower bound, subject to a turnover constraint. The active loss harvesting lets us realize a portfolio’s losses as soon as they are available. Finally, we invest the proceeds from trimming and active loss harvesting first, to fill the securities that have losses available but are underweighted relative to the strategy that is not tax-managed. This lets us realize more losses from those securities in the future.

We choose 0.2% to 0.5% bounds for each strategy to give a roughly 1% tracking error over the 20-year period. This level of tracking error is typical in practice and makes the tax-managed strategies comparable. In an actual portfolio, we would not know ahead of time precisely what bounds provide a 1% tracking error. Instead, we would need to use a more complicated procedure than the one we used in our analysis. For example, we could use an optimization approach to control the trade-offs between taxes, tracking risk, and turnover. We expect a tax-optimized approach to create tighter tracking risk, lower turnover, and better after-tax performance than what is simulated here.

TAX-MANAGEMENT EMPIRICAL RESULTS

Before we look at the tax-managed results for all the strategies, it is instructive to examine one particular strategy in more depth. Exhibit 5 shows the ending values of a U.S. portfolio using capitalization weighting (CW), equal weighting (EW), and tax-managed equal weighting (TM EW). The starting portfolio value is $10 million. The CW portfolio grows to $50.6 million, while the two EW portfolios grow to $66.2 million, all before taxes. The TM EW strategy does not perfectly mimic the EW strategy in returns; there is some tracking error. In this case, however, the portfolios ended with the same final value, despite some return differences through time.

• Download figure
• Open in new tab
• Download powerpoint

Exhibit 5

Ending Portfolio Values for U.S. Equal Weight ($10 million initial value, 1993–2012)

On an after-tax and after-liquidation basis, CW ends with $38.4 million, EW with $45.2 million, and TM EW with $48.2 million. EW outperforms the capitalization-weighted portfolio, even after taxes. The tax-managed version keeps even more of the return.

The amount of the value that tax management adds depends on a number of factors: the market return environment, security cross-sectional volatility, the portfolio cost basis, and prevailing tax rates. To better understand how this tax savings is created, we look at the tax impact for each year for the EW and TM EW strategies. Comparing the charts in Exhibit 6, we observe significant tax savings in many of the years. On average, as shown by the dotted line, the annual tax impact is reduced from -1.64% to -1.08%. Both portfolios follow the same pattern: increased tax impact during the bull markets of 1996 to 2000 and 2005 to 2007, and decreased tax impact for several years after a market crash, as seen in 2001 and 2008, for example. This pattern is due to excess losses being carried forward to shelter gains in future years.

To contrast the two portfolios, we make particular note of 1994 and 2008. In 1994, the EW strategy realized short-term gains. The tax-managed portfolio avoided them. In 2008, during the market crash, many of the positions were at a loss. The tax-managed portfolio was able to take greater advantage of these losses than was the original strategy.

However, the tax-managed version does not win in every year. In 2005 and 2011, TM EW suffered worse tax impact than EW. This was caused by the dynamics of the loss balance carried forward. The years leading up to 2005 and 2011 generated higher embedded tax liability in the tax-managed portfolio, while the original strategy paid more taxes in the earlier years and still had a loss balance carried forward. Overall, the results are striking. Paying attention to taxes reduces the drag on returns.

Similar patterns of performance can be found in the other smart-beta strategies. In Exhibit 7, we examine the pre-tax results of portfolios with tax management. For tax-managed portfolios, we allow some differences in trading, to reduce the tax impact. The tax-managed smart-beta portfolios have about 1% tracking error versus the original strategy. For these tax-managed portfolios, the realized returns, volatility, and tracking error versus the benchmark are close to the original strategy.

The tax-managed strategies have much better after-tax results, as shown in Exhibit 8. Note that we measure the after-tax excess return versus the capitalization-weighted index, which has no tax management. The tax-managed capitalization-weighted index takes active risk, so it is not particularly useful as a benchmark. In the table’s last column, we show that the tax impact is lower for the tax-managed portfolios by 40 to 90 basis points. We do see a higher liquidation impact, since more gains are deferred to the future. However, as we saw in the EW example, there is value in the deferral. The portfolio’s deferred tax liability continues to earn a return, even though taxes must be paid at liquidation.

Ironically, the tax-managed portfolios require more turnover to increase tax efficiency. As mentioned before, extra turnover creates trading costs. However, the magnitude of the tax benefit (40 to 90 basis points) far exceeds the magnitude of the extra trading costs (5 to 15 basis points). For example, in the tax-managed U.S. equal-weight strategy, the turnover increased from 25% to 71%. If we multiply the 46% turnover increase by the estimated 18 basis points of trade cost per 100% turnover, we see that trading costs equal 8 basis points, while the decrease in tax impact is worth 56 basis points (-1.64% becomes -1.08%). In an effort to be conservative, our simulations likely overstate the turnover required. Because we can estimate both the tax benefit and the transaction costs before trading, we can constrain turnover to ensure that the benefits always outweigh the costs.

Exhibit 9 provides a visual check on the tax-managed strategies’ excess returns. Here we compare the pre-tax and after-tax results versus the capitalization-weighted index. For all the strategies shown, tax management reduces the tax impact, so the portfolio retains more outperformance. Interestingly, the diversity-weighted (DW) strategy actually has higher after-tax than pre-tax excess. This is because the tax-managed DW strategy does not realize any capital gains, but the capitalization-weighted index does.

• Download figure
• Open in new tab
• Download powerpoint

Exhibit 9

Excess Returns for Tax-Managed Strategies, 1993–2012

We can also tax-manage the capitalization-weighted index, eliminating the realization of capital gains over the 20-year period. Depending on the after-tax return methodology we use, a cap-weighted portfolio that is tax managed can have an after-tax return that is much higher than pre-tax returns. For example, the SEC after-tax return methodology gives credit for excess realized losses, resulting in much higher after-tax returns. This might be a reasonable assumption if another part of the taxpayer’s portfolio generates a steady stream of realized gains. If we used that method, the change in tax impact for the U.S. cap-weighted portfolio would be 1.32%, instead of the 0.53% we show in Exhibit 5. The actual savings that might be achieved depends on a number of factors unique to each taxpayer, but these two methodologies provide a reasonable range in which to set expectations.

CONCLUSION

Taxes can represent a very large performance drag, often higher than transaction costs and management fees combined. Smart-beta strategies seek to add value through a variety of non-capitalization-weighted portfolio construction techniques. However, the additional turnover that accompanies these strategies creates an extra tax impact, which can reduce the after-tax value added. In this article, we demonstrate how tax management can eliminate some of the additional tax impact, without sacrificing the strategy’s underlying performance. Within a mutual fund, this type of tax management would work to reduce or eliminate capital gain distributions to shareholders. In a separate account, the extra tax benefit (when losses exceed gains) would accrue to the taxpayer. In general, all types of investment management benefit from tax management. However, for strategies that have a large number of securities, a modest amount of expected turnover, and a rules-based portfolio construction approach, we find that tax management is particularly effective.