by Andrew L. Berkin, Ph.D.; Glenn S. Freed, Ph.D., CPA; and Sheryl L. Rowling, CPA
- This paper discusses an alternative to standard optimization techniques to determine the asset allocation and asset location for an investor with accounts with different tax treatments.
- While portfolio optimization approaches can be useful, optimization serves only as a starting point from which a financial adviser can model various scenarios and find which is most appropriate for a client given their considerations of risk, return, and taxes. This approach produces intuitive, practical, and robust results.
- The incorporation of financial planning techniques such as asset allocation, asset location, and the drawing down of assets in retirement are discussed in this paper. Cheap and fast computing power allow an adviser to consider different scenarios and present them to clients in terms of intuitive metrics such as returns, volatilities, draw downs, tax consequences, and ultimately their impact upon goals such as saving for retirement.
- This comprehensive approach is compared and contrasted with standard optimizations and those that consider taxes. Although the consideration of taxes is crucial for effective financial planning, results suggest that advisers should be wary of adjusting the value of tax deferred accounts downward based on assumed tax rates in retirement. This caution stems from practical considerations of client comfort level and perception, the impact of very uncertain assumptions about tax rates and future portfolio values, and the robustness of asset allocation decisions to input parameters.
Andrew L. Berkin, Ph.D., is co-CIO of Vericimetry Advisors LLC, an investment adviser catering to financial planners. His research and experience have focused on providing quantitatively structured solutions for taxable and tax-exempt investors. He has a Ph.D. in physics from the University of Texas at Austin. (Andrew.Berkin@vericimetry.com)
Glenn S. Freed, Ph.D., CPA, is CEO and co-CIO of Vericimetry Advisors LLC. He is an academic with practical experience in the financial services industry where he has expertise in taxes in the investment management sector. He has a Ph.D. in business from the University of Southern California. (Glenn.Freed@vericimetry.com)
Sheryl L. Rowling, CPA, is CEO of Total Rebalance Expert, and a personal financial specialist and principal at Rowling & Associates, where she has been providing investment, tax, and financial planning advice since 1979. She is the author of the CCH publication Tax and Wealth Strategies for Family Businesses. (Sheryl@Rowling.com)
Financial planners recognize the importance of taxes on net investment returns and, thus, apply a number of techniques to enhance their clients’ after-tax returns. Two important determinants of net returns are asset allocation and asset location decisions. The former refers to the target percentage of each asset class to be held in the portfolio while the latter refers to the tax treatment of the type of account in which the assets will be held. Common account types include taxable accounts, tax deferred accounts (TDAs), such as traditional IRA and 401(k) accounts, and tax exempt accounts (TEAs), such as Roth IRAs.
The standard academic approach to asset allocation is to run an optimization, choosing an appropriate utility function and defining parameters such as risk aversion. Constraints are often imposed to keep solutions “reasonable.” For example, constraints are imposed so that weights to any single asset do not get too large. Taxes often are considered a complication and ignored, in part because much of the initial optimization research efforts often are aimed at tax exempt institutions such as pension funds. But for an individual with a mixture of taxable accounts, TDAs, and TEAs, the appropriate treatment of taxes in both asset allocation and asset location can have a material impact upon investment results (Kitces 2013; Reichenstein 2007). An alternative and practical approach to achieve the important goal of after-tax optimization exists. The following discussion highlights this alternative.
A Scenario Based Approach
Optimization is a useful tool for an adviser, combining expected risks, returns, and correlations to obtain desired portfolios. However, the precise answer obtained is dependent on many imprecise assumptions. Therefore, the result from an optimization should not be considered the final answer, but rather one step in determining the ideal allocation for each client. Cheap and fast computing power allows an adviser to develop a number of scenarios, with each one giving the client a range of likely outcomes. Useful results include final dollar amounts, the probability of meeting a target, and drawdowns likely to be experienced. The discussion of these results with the client is an integral part of determining an allocation of portfolio assets.
For example, based on initial conversations with a client, the adviser may view a 50/50 stock/bond allocation as roughly appropriate. The next step would be to conduct analyses based on expected risks and returns for the assets based on client specifics, such as expected future cash flow needs and tax rates. Relevant summary statistics can be compiled, including average returns and volatilities, potential drawdowns, tax consequences, and cash available both in retirement and for other life events along the way, such as the cost of children’s education. This should be run not just for the 50/50 allocation, but for several other allocations as well. Such analyses all can be completed in a straightforward and easy-to-understand fashion such that the adviser-client discussion unfolds as follows: “Here is the allocation and here is a range of what you will potentially have to spend. This spending can be for retirement as well as other life events such as paying for college. As situations change in the future, so can the results of this process.”
In addition to relying on asset allocation decisions, the adviser should use asset location to best position clients’ assets (Dammon, Spatt, and Zhang 2004; Daryanani and Cordaro 2005; Rowling 2008). Daryanani and Cordaro estimated that proper asset location boosts returns by 20 basis points after taxes, compared to keeping an identical allocation across multiple accounts. Typically, bonds are held in TDAs to defer tax on the income. While this tax will be at ordinary income rates upon withdrawal, this would have been the rate had they been held in a taxable account. As bonds tend to have lower returns, keeping them in the TDA means lower required minimum distributions (RMDs) in retirement, creating extra flexibility. If the desired allocation to fixed income is large enough that taxable accounts must be used, then tax-free municipal bonds typically would be preferred.
Tax efficient appreciating assets should be placed in the taxable account. Typically, this means low turnover equities, so that capital gains can be deferred and possibly avoided if the assets are held to death. Tax loss harvesting can further lower the effective tax rate of stocks (Arnott, Berkin, and Ye 2001; Berkin and Ye 2003). In contrast, holding equities in a TDA effectively turns long-term capital gains into higher taxed income, and is thus tax inefficient. Finally, TEAs should be used to hold assets with the highest expected return, as they will never be taxed. Because these accounts do not have RMDs, there is added flexibility for the client in retirement.
The effective drawing down of clients’ assets in retirement from their various accounts is another key aspect of financial planning, with taxes playing a crucial role in the drawdown decision (Sumutka, Sumutka, and Coopersmith 2012). When drawing down from a taxable account, advisers and their clients should do so in a tax efficient manner. This involves generally first drawing on those assets with the highest cost basis and at long-term capital gains rates. Accumulated losses can be used to offset gains. Taxes potentially can be deferred for extended periods, possibly until death, when the heirs get a step-up in cost basis. The ability to effectively use such techniques will depend, in part, on asset location choices made in the accumulation phase of the client’s wealth cycle. Crucially as well, the utilization of such distribution techniques influences the validity of assumptions made in the accumulation phase. One cannot precisely know the future, and assumptions must be made. But again, a discussion and understanding of these assumptions and potential scenarios needs to be undertaken.
Such discussions and actions are part of the overall wealth proposition that an adviser brings to clients. Armed with the results of a variety of potential scenarios, the adviser can then discuss the range of potential outcomes before, at, and after retirement. Included in these discussions should be likely cash available from all sources for spending and potential deviations depending on differing parameters such as taxes and returns. From this discussion comes the understanding to set an appropriate allocation and location for each client.
Tax Adjustments and TDA Allocations
A 2012 paper by Reichenstein, Jennings, and Horan examined the impact of taxes on asset allocation and location decisions. The authors made two specific points, illustrated with examples. One is that the expected returns and risks of taxable assets should be adjusted for taxes. The other point is that, for purposes of asset allocation, TDA target dollar amounts should be reduced by the projected tax rate at the time retirement withdrawals are expected to begin. We strongly agree that taxes need to be considered in investment decisions and, therefore, asset location is an important tool. We further agree that the effect of taxes on money available for retirement should be considered. However, we have several issues with the approach of adjusting a TDA’s target dollar allocation to account for taxes.
The rationale for decreasing a TDA’s dollar value by the tax rate in retirement is straightforward. When money is eventually withdrawn from the TDA, the investor receives a smaller amount reflecting the taxes paid. This is equivalent to adjusting the amount in the TDA now and assuming no taxes when withdrawn.
For example, consider an investor with $100 in a TDA that ultimately doubles in value by retirement when her tax rate will be 25 percent. Thus, the initial $100 grows to $200, which will be worth $150 after taxes. This is equivalent to regarding the TDA as worth $75 now after-tax and then doubling it to $150.
Consider a further illustration of this concept, based on one from Reichenstein et al. (2012). The investor in this example (see Table 1) has $1 million in a TDA and $720,000 in a taxable account. The asset allocation in Column A is split 50/50 between stocks and bonds, and the asset location is split evenly as well. Column B shows the same allocation but now with a more tax efficient asset location. The taxable account holds only stocks. If these are managed in a tax efficient manner, then capital gains can be delayed, and if realized, then only at favorable long-term capital gains rates. In this example, bonds are held exclusively in the TDA, so the investor avoids paying taxes on the yield at ordinary income rates. Some equities are held in the TDA as well, to bring the allocation to 50/50. Should equities have higher returns over time as expected, then the taxable account will grow more quickly. Over time, the equities in the TDA can be rebalanced into bonds to keep the targeted allocation.
Column C in Table 1 comes from Reichenstein et al. (2012). Data demonstrate their approach. This investor is assumed to be in a 28 percent tax bracket upon retirement some years in the future. The TDA is not taxed until retirement at which time the distributions will be taxed at that marginal tax rate. Hence, one major difference in Column C is that the future and current value of the TDA is reduced by this 28 percent rate. The $1 million in the TDA is considered to effectively be only $720,000 (as can be seen by summing the first two rows), and hence the total assets are reduced to $1.44 million.
The second major difference in Column C concerns the asset allocation. The returns and volatilities of the taxable account have been reduced by their effective tax rates. For example, an investor in the 28 percent tax bracket would see returns on taxable bonds diminished by 28 percent, which also decreases the volatility by the same amount. For stocks, the effective tax rate depends on how much of the return comes from qualified and nonqualified dividends versus capital gains, and how long the stocks are held before gains are realized (Reichenstein et al. 2012). The new after-tax values for the TDA, and returns and volatilities of the taxable account, have been input into an optimizer. Other parameters such as the risk aversion have been kept the same as the case that produced the 50/50 allocation when taxes were ignored. This new optimization produces a 59.2/40.8 stock/bond allocation. Column D reflects the actual dollars invested, as shown on the custodian statements. Thus, the values for the TDA in Column C are 28 percent less than those of Column D. Here, the 59.2/40.8 allocation appears in actuality to be 52.6/47.4.
We strongly agree with the overall philosophy of Reichenstein et al. (2012). Taxes do matter and need to be accounted for by the responsible financial adviser. This includes both taxable assets, as well as those in a TDA that will be taxed upon withdrawal in retirement. The approach applied by Reichenstein et al., and its use in asset allocation, are certainly theoretically correct given the assumptions.
Where we diverge is on its use in practice. These differences can be broken into three categories: (1) client perception and comfort level, (2) the impact of very uncertain assumptions about tax rates and future portfolio values, and (3) the robustness of asset allocation decisions to input parameters. Each of these is discussed below, compared and contrasted with our scenario based approach.
Client Perception and Comfort
For advisers to build healthy relationships with their clients, investment strategies need to be well understood. However, adjusting TDA values affects allocations across the full portfolio. This has the strong potential to confuse clients, because amounts shown on their investment statements will not match what their adviser has told them.
Consider again the example from Table 1. In the Reichenstein et al. (2012) approach, Column C is optimal and reflects a 59.2 percent allocation to stocks and 40.8 percent to bonds. However, as previously noted, the investor would see statements reflecting the allocation in today’s pretax dollars. These values are given in Column D, where the TDA amounts have been readjusted to reflect what is in the portfolio today. Instead of the 59.2/40.8 allocation that the adviser has communicated, the client sees a 52.6/47.4 allocation. For an observant client, this is quite a difference.
This divergence between what the adviser says and what the investment statement shows becomes even more problematic when explaining performance. The client will see statements with returns reflecting a different allocation than what the adviser is claiming. Consider the example of Table 1 in a period where stocks strongly outperform bonds. The portfolio returns on the statement reflect a 52.6 percent equity allocation and would be notably lower than what the client would expect from the communicated 59.2 percent equity allocation. If bonds lose 5 percent, the client’s statement will show a loss of $40,800 based on Column D. Yet, based on the after-tax values shown in Column C, the adviser will claim it’s really only $29,376. How happy and understanding will the client be? This question is important for advisers to consider before embarking on such an approach.
Trying to report this performance based on the proposed after-tax methodology of Reichenstein et al. (2012) is also problematic. Standards exist for reporting after-tax performance (Price 2001). Generally, taxes impact performance when they are recognized. Mutual funds report after-tax returns assuming a particular holding period and liquidation at current prices, but this is different than an assumed liquidation in the future. An adviser attempting to report after-tax performance based on the Reichenstein et al. (2012) viewpoint of asset values needs to be careful with calculations, as these will no longer match those reported by custodians. They also need to be careful about regulations governing performance reporting.
An additional issue for advisers who bill based on client assets under management is determining what asset value to use. In the previous example, would the adviser charge as if there were only $720,000 in the TDA instead of the full $1 million? Certainly, the client could expect this principle to apply to fees as well. What if the client’s investments do well and the tax rate in retirement is therefore higher than that assumed? Would she then be entitled to a refund as the after-tax value of the TDA was estimated to be too high previously and thus fees were accordingly too high?
Dependence on Tax Assumptions
The uncertainty associated with investing makes client understanding and communication essential. This uncertainty also raises the importance of robustness of assumptions and outcomes. A financial plan should not just be based on the most likely outcome, but should also produce satisfactory results over a range of likely outcomes. There will always be uncertainty in a client’s tax rate in retirement caused by ambiguity in investment results and future tax, as well as non-portfolio taxable income. We fully agree with Reichenstein et al. (2012) that in spite of this uncertainty, advisers should not ignore them but rather make a best estimate. In our scenario-based approach, this best estimate would serve as the base case. The adviser can then review how outcomes can change as these estimates change, giving a deeper understanding of the range of possibilities.
For example, the tax rate in retirement will depend on the success of the investor. Investor success is dependent on investment returns and current tax laws. However, the tax rate actually can be lower for those with either very few or very large assets. Someone with few assets will generally have a low tax rate for their TDA because their income in retirement will be low. Those with greater assets will generally have higher income and thus pay a higher tax rate as they consume their TDA. However, techniques such as using TDA assets for charitable contributions, either during or after the client’s lifetime, can effectively avoid distribution taxes altogether. For those clients with sufficient assets outside of their TDA, the tax rate for the TDA can be minimized.
In general, an investor with adequate outside assets will only need to draw down RMDs, leaving the TDA to heirs. In this case, should the assumed after-tax rate be based on the next generation? The likelihood of a client needing to use her TDA might not be apparent until near retirement, depending not just on the success of investments but also on factors such as the sale of a business, family circumstances, and the desire to donate to charity. All of these factors can impact the effective tax rate of the TDA. Therefore, the likelihood and the impact of each assumption are best addressed in a scenario approach.
Clients also may choose to convert their Roth IRA held in a TDA into a TEA, particularly if they encounter a year with a low tax rate. Indeed, such a consideration is an important aspect of an investment adviser’s value to their clients. Taking advantage of such an opportunity would result in the old TDA assets now having a very different tax treatment. Asset allocation based upon adjusting those assets’ value downward by an assumed tax rate in retirement would now be completely inappropriate. A similar situation can arise with the untimely death of an investor before most of the TDA assets have been withdrawn. The tax liability transfers to the beneficiaries of the TDA as “income in respect of decedent,” a potentially very different tax situation. A TDA asset allocation adjusted on any particular tax rate could ultimately be inappropriate. Again, a scenario-based approach can frame the final decision in terms of potential situations and their outcomes.
Robustness of Optimization
Another key set of considerations is the sensitivity of results to optimization parameters. An optimizer will produce a precise answer, but that answer is very dependent on imprecise assumptions. We certainly view optimizations as useful. However, the results of an optimization are not the end of the allocation process but rather just one step. The adviser is required to do further analysis. This involves engaging clients in discussions to determine an appropriate allocation for specific needs given a range of potential outcomes.
One such parameter affecting optimization outcomes is risk aversion. Reichenstein et al. (2012) fixed the risk aversion parameter at the value producing a 50/50 stock/bond allocation when ignoring taxes. It is important to note that keeping this parameter unchanged is not necessarily the best approach. Indeed, virtually no client will know or understand their risk aversion parameter when used in a mean variance optimization. This parameter is not externally derived, but rather comes from consideration of a range of this parameter and the corresponding optimization outputs and then finding the result that best matches the client’s desired outcome. In other words, a target allocation arises from running multiple scenarios and understanding the client’s perceptions of outputs such as possible returns, risks, tax consequences, and ultimately the money available for retirement. This procedure is the foundation of a scenario based approach. That is, advisers should vary the parameters and examine the results. Simply keeping fixed the same inputs, such as the risk aversion parameter, from a tax-exempt optimization may not produce the tradeoffs of returns, risks, and taxes that the client finds ideal.
Another point to note is that differences in weights between the two approaches are much smaller when an apples-to-apples comparison is conducted. As shown in Column C of Table 1, a client’s 50/50 stock/bond allocation becomes 59.2/40.8 after optimizing the portfolio based on reducing Column B’s TDA market value after taxes from $1 million to $720,000 and adjusting the taxable account’s assumed return and volatility. However, the amount actually invested in the TDA in current pretax dollars is still $1 million. The amounts actually invested at today’s market value are given in Column D, and reflect a 52.6/47.4 stock/bond allocation, which is much closer to the original 50/50 allocation before all the complexity and potential confusion.
The uncertainties of assumptions in the optimization of returns, volatilities, and tax rates overwhelm these already small differences in allocations. These differences become even smaller when a more complete and realistic set of assets is considered, such as domestic, international, and emerging market equities. Differences are minimized potentially further when assets are broken into styles such as size and value dimensions, various fixed income classes, real estate, and commodities. Going from an equity allocation of 50 percent to 52.6 percent might mean moving the various components by less than 1 percent each. This is comparable to an acceptable tolerance range for market fluctuations impacting the actual allocation model versus the target allocation model.
Taxes are an important consideration for most individual investors, affecting investment returns and the amount of money available for spending. Allocating assets appropriately based on a client’s needs and risk tolerance, and then locating these assets in accounts with different tax treatments, is a crucial task of a financial adviser. Given the many uncertainties of investing, it is also crucial for the adviser to recommend an allocation that is as robust as possible to assumptions about input parameters. Additionally, it is essential to be able to explain the process and potential outcomes to clients so they gain confidence and comfort.
The work of Reichenstein et al. (2012) has added to the focus on the importance of taxes. They suggested adjusting the target asset allocation by considering a tax deferred account to have less assets based upon the assumed tax rate in retirement. While technically correct, such an approach has practical limitations. Calculations and results are dependent on assumptions about both future returns and client situations. Based on the assumptions made, it is possible that the “tax-adjusted” models will not materially differ from “gross of tax” models. It also can be extremely difficult to explain results to clients.
Instead, we have described a scenario-based approach that meets the criteria above. Such an approach can start with optimization to produce an initial allocation. Variations of the parameters and assumptions should be run to consider a variety of possible outcomes and their impact on money available after tax for retirement. Clients can then see a range of potential outcomes and can obtain a comfort level based on the likely range resulting from their allocation decision. This comfort level may produce a more satisfied client and a more positive relationship with the adviser.
Arnott, Robert D., Andrew L. Berkin, and Jia Ye. 2001. “Loss Harvesting: What’s It Worth to the Taxable Investor?” Journal of Wealth Management 3 (4): 10–18.
Berkin, Andrew L., and Jia Ye. 2003. “Tax Management, Loss Harvesting, and HIFO Accounting.” Financial Analysts Journal 59 (4): 91–102.
Dammon, Robert M., Chester S. Spatt, and Harold H. Zhang. 2004. “Optimal Asset Location and Allocation with Taxable and Tax-Deferred Investing.” Journal of Finance 59 (3): 999–1037.
Daryanani, Gobind, and Chris Cordaro. 2005. “Asset Location: A Generic Framework for Maximizing After-Tax Wealth.” Journal of Financial Planning 18 (1): 44–54.
Kitces, Michael. 2013 (March 6). “Asset Location: The New Wealth Management Value-Add for Optimal Portfolio Design?” www.kitces.com/blog/archives/479-Asset-Location-The-New-Wealth-Management-Value-Add-For-Optimal-Portfolio-Design.html.
Price, Lee N. 2001. “Taxable Benchmarks: The Complexity Increases.” AIMR Conference Proceedings (June): 47–55.
Reichenstein, William. 2007. “Calculating After-Tax Asset Allocation Is Key to Determining Risk, Returns, and Asset Location.” Journal of Financial Planning 20 (7): 44–53.
Reichenstein, William, William W. Jennings, and Stephen M. Horan. 2012. “Perspectives: Two Key Concepts for Wealth Management and Beyond.” Financial Analysts Journal 68 (1): 14–22.
Rowling, Sheryl. 2008. “Practical Rebalancing.” NAPFA Advisor (September): 20–24.
Sumutka, Alan R., Andrew M. Sumutka, and Lewis W. Coopersmith. 2012. “Tax-Efficient Retirement Withdrawal Planning Using a Comprehensive Tax Model.” Journal of Financial Planning 25 (4): 41–52.