Are Individual Investors Tax Savvy? Asset Location Evidence from Retail and Discount Brokerage Accounts

Transcription

1 Are Individual Investors Tax Savvy? Asset Location Evidence from Retail and Discount Brokerage Accounts Brad M. Barber Terrance Odean * First Draft: March 2001 Please do not cite or circulate without permission. This is a preliminary and incomplete working draft prepared for the Finance Program of the Stanford Institute for Economic Policy Research Asset Location Conference March 23, 2001 * Graduate School of Management, University of California, Davis. We are grateful to the retail broker and discount broker that provided us with the data for this study. Jim Poterba provided helpful suggestions. Shane Shepherd and Michael Bowers provided valuable research assistance. All errors are our own. Brad Barber can be reached at (530) or bmbarber@ucdavis.edu; Terrance Odean can be reached at (530) or odean@ucdavis.edu.

2 Are Individual Investors Tax Savvy? Asset Location Evidence from Retail and Discount Brokerage Accounts Abstract Investors with both taxable investment accounts and tax-deferred retirement accounts must choose where to locate their assets. We present descriptive evidence on the asset location decisions of investors with accounts at a discount broker in 1994 and investors with accounts at a retail broker in Tax-exempt municipal bonds are held almost exclusively in taxable accounts, though a substantial fraction of municipal bond investors concurrently hold taxable bonds in their taxable account. There is mixed evidence that investors prefer to locate taxable bonds in their tax-deferred accounts. Among households with material allocations to taxable bonds, approximately one-third of the average household s taxable bond holding is held in a taxable account and could replace equity in its tax-deferred account. Finally, we document a strong preference for holding mutual funds, rather than individual stocks, in taxdeferred accounts. Nonetheless, among households with material allocations to individual stocks and equity mutual funds, we estimate that approximately 20 percent of the average household s mutual fund holding is held in a taxable account and could replace individual stocks in its tax-deferred account. We conclude that either the extant models of optimal asset location are incomplete or a substantial fraction of investors are mislocating their assets.

3 Saving for retirement is a challenge. Investors must first choose how to allocate their assets -- to stocks or bonds -- and, second where to locate those investments -- in a taxable account or a tax-deferred account. Several recent papers argue that the location decision materially affects investor welfare (Dammon, Spatt, and Zhang (2001), Huang (2000), and Shoven and Sialm (2000)). Generally, these papers argue that investors should first locate taxable bonds to tax-deferred accounts (TDAs). While it is difficult to avoid paying tax on the ordinary income generated by bonds, investors can defer the realization of capital gains on equity. In this paper, we determine the extent to which individual investors follow the advice. While our empirical analysis cannot test the normative validity of theoretical models, we can test their descriptive validity. We do so by analyzing the location decisions of households with accounts at a discount broker as of 1994 (discount households) and households with accounts at a full-service retail broker as of 1998 (retail households). We begin by analyzing the most straightforward location decision -- where to locate tax-exempt municipal bonds. Investors get this one right; virtually all municipal bonds are located in taxable accounts. However, a substantial fraction of households that hold municipal bonds also hold taxable bonds in their taxable accounts. We next consider the location of taxable bonds. For both retail and discount households with material allocations to taxable bonds, we document that roughly one-third of the average household s taxable bond holdings could replace equity in its TDA. Though discount households have a preference for locating taxable bonds in their TDAs, retail households do not (except among the wealthy). Finally, we consider the location of equity mutual funds and individual stocks. Investors can optimize the tax avoidance strategies available on equities by locating individual stocks in their taxable accounts, while locating mutual funds, which distribute a relatively high proportion of their capital gain return, to TDAs (assuming that there is space to do so after first locating taxable bonds in their TDAs). For the retail and discount households that we analyze, more than two-thirds of their equity investments are held in 1

4 individual stocks. Among both groups, there is a strong preference for holding equity mutual funds in TDAs (and individual stocks in taxable accounts). Among households with a material allocation to both individual stocks and equity mutual funds, the ratio of individual stocks to equity mutual funds in taxable accounts is roughly two to one, while the ratio is almost reversed in TDAs. Nonetheless, for retail and discount households with material allocations to both individual stocks and mutual funds, we estimate that approximately 20 percent of the average household s mutual fund holding is held in a taxable account and could replace individual stocks in its TDA. The plan of the paper is as follows. We describe the two datasets in section I. The asset allocation decisions of these households are presented in section II. We analyze the characteristics of municipal bond holders in section III. The location of taxable bonds, mutual funds, and individual stocks is discussed in section IV. We make concluding remarks in section V. I. Data In this study, we analyze two snapshots of portfolio holdings: one from the position statements of a discount broker in 1994, the second from position statements of a full-service (retail) broker in The disadvantage of these data (as opposed to data from, for example, the Survey of Consumer Finance) is that we do not have the complete asset holdings of each household. Some households may hold substantial assets in other accounts. Though this will no doubt create some noise in our analysis and therefore reduce the power of our empirical tests, we have no reason to believe that our evidence on asset location will be biased. The tremendous advantage of these data is the detailed information on positions held, which allows us to analyze many issues that are simply impossible to address with existing survey data. The first data set contains information from a large discount brokerage firm on the investments of households for the six years ending in December We arbitrarily chose February 1994 to calculate the asset allocation and location decisions for each household. Based on product codes provided by the discount broker, we categorize positions as equity 2

5 (e.g., investment in individual common stocks or equity mutual funds), taxable bonds (e.g., investment in money market funds, bond mutual funds, government bonds, and corporate bonds), municipals bonds, and other (generally positions with no product codes). We delete households with a portfolio value less than $10,000. We also exclude households with greater than a 10 percent allocation to assets other than stocks, taxable bonds, or municipals. We then calculate allocations based on assets that we are able to categorize, leaving us with a final sample of 48,084 discount households. The second data set contains information from a large retail brokerage firm on the investments of households for the 18 months ending in June Based on product codes provided by the retail broker, we categorize positions as equity, taxable bonds, municipal bonds, and other as of November As was done for the discount households, we delete households with portfolio values less than $10,000 and greater than a 10 percent allocation to assets other than stocks, taxable bonds, or municipals, leaving us with a final sample of 452,876 retail households. II. Asset Allocation A. Descriptive Statistics In TABLE I, we present descriptive information on the allocation of the discount households (Panel A) and retail households (Panel B). We consider three partitions of each data set: (1) households with a minimum balance of $10,000, (2) households with a minimum balance of $100,000, and (3) households with a minimum balance of $10,000 and less than a 99 percent allocation to stocks. In the remainder of the paper, for expositional ease we refer to households with a minimum balance of $10,000 as discount (or retail) households, while we refer to households with a minimum balance of $100,000 as wealthy discount (or retail) households. 2 One of the product codes for the retail households is mutual funds and does not distinguish between equity and bond mutual funds. For this category, we matched the individual holdings of mutual funds to the Center for Research in Security Prices (CRSP) mutual fund database and used the Investment Company Data, Inc., (ICDI) objectives to categorize holdings as equity, taxable bond, municipal or other. For balanced and total return 3

6 Retail households hold more assets than discount households and a larger proportion of assets are allocated to taxable bonds and municipals. Among discount households, 76 percent hold virtually all stock portfolios, while only 22 percent of retail households hold virtually all stock portfolios. These differences narrow, but still exist between the wealthy discount and retail households. When we eliminate households with virtually all stock allocations, the differences between the remaining retail and discount households are very small. This suggests many discount accounts are opened solely to trade stock, but discount households that choose to hold non-equity assets are more comparable to their retail counterparts. Both retail and discount households hold the majority of their equity allocation in individual stocks rather than equity mutual funds. On average, discount households hold 75 percent of their stock allocation in individual stocks, while retail households hold 69 percent in individual stocks. For the discount households, we have a wide range of demographic data. The demographic information is compiled by Infobase Inc. (as of June 8, 1997) and provided to us by the discount broker house. For about one-third of the discount households, we also have self-reported estimates of net worth, which were provided when the account was opened. Unfortunately, for the retail households we have only information regarding the gender and age of the person opening the account. The average age of those who opened accounts at a discount household is 49, while that of the retail households is 52. Seventyeight percent of discount accounts and 66 percent of retail accounts were open by men. B. Age, Gender, and Asset Allocation Though not the focus of this inquiry, we are able to contribute some empirical evidence on asset allocation issues. First, we are able to investigate the relation between age and asset allocation. Basic models of portfolio choice, such as those in Mossin (1968), Samuelson (1969), and Merton (1969), yield no dependence between one s portfolio funds, which are typically split between stock and bonds, we further used the percentage allocations from CRSP to categorize holdings. 4

7 allocation and one s investment horizon. 3 Nonetheless, as discussed in Canner, Mankiw, and Weil (1997), Bodie and Crane (1997), and Ameriks and Zeldes (2000), most financial advisors recommend that an investor should reduce their allocations to stock as they age. In Figure 1, we plot the mean portfolio value and percentage stock allocation against age for households with portfolio values greater than $10,000 for the discount (Panel A) and retail (Panel B) households. (The patterns are similar for households with greater than $100,000.) Though portfolio values tend to increase with age (as investors accumulate wealth), 4 the mean allocation to equity declines with age. To control for other demographic characteristics in examining this relation, we estimate a cross-sectional regression where the dependent variable is the percentage stock allocation. For the discount households, the independent variables in the regression include age and dummy variables that are designed to control for gender, marital status, income level, and net worth. We also include dummy variables for RV (recreational vehicle ownership) and motorcycle ownership, since risk aversion is perhaps correlated with other types of thrill-seeking behavior. (We have no opinion on whether driving an RV should be considered thrill-seeking, though we believe driving a motorcycle is thrilling -- or at least risky -- to most.) We estimate three sets of regressions, corresponding to the three sample partitions presented in TABLE I. In TABLE II, we present the results of these regressions. Consistent with the evidence presented in Figure 1, after controlling for other demographic characteristics, there is a significant negative relation between age and percentage stock allocation. Based on this empirical evidence, one might be tempted to conclude that investors reduce their exposure to risky assets as they age. However, as argued by Poterba and Samwick (1997) and Ameriks and Zeldes (2000), cross-sectional relationships between age and stock allocations might 3 Ameriks and Zeldes (2000) review the conditions necessary to generate a theoretical relation between age and asset allocation, but in some of these alternative models the proportion of risky assets (e.g., stock) in the optimal portfolio rises with age while in others it declines. See also Jagannathan and Kocherlakota (1996). 4 Though we find little evidence of consumption during retirement years (those older than 65), this might be attributable to the fact that we are analyzing households with large stock allocations following a period of strong market returns. 5

8 result from investors reducing their exposure to risky assets as they age (age effects) or from differences in risk aversion that depend on one s year of birth (cohort effects). Given that we have a short time-series of allocation decisions, we are unable to distinguish cohort from age effects. However, if cohort effects entirely explain the cross-sectional relation between age and asset allocation for the households we analyze, then there has been a nearly monotonic increase in the willingness to hold stock for investors born during the last 90 through 25 years. Of the remaining demographic variables, only gender is a robust determinant of one s stock allocation. Women allocate less of their portfolio to common stocks. This result is consistent with prior evidence that women are less tolerant of investment risk than men. Bajtelsmit and Bernasek (1996), Bajtelsmit and VanDerhei (1997), Hinz, McCarthy, and Turner (1997), and Sundén and Surette (1998) find that men hold more of their retirement savings in risky assets. Jianakoplos and Bernasek (1998) report the same for overall wealth. Barber and Odean (2001) document that women hold individual stocks that are less volatile and have lower market risk (i.e., betas) than those held by men. For the discount households, there is modest evidence that income is positively related to stock allocations, while self-reported net worth is negatively related to one s stock allocation. Marital status, RV ownership, and motorcycle ownership do not appear to be robust determinants of a household s stock allocation. III. Municipal Bonds In this section, we analyze the allocation and location decisions of investors who hold municipal bonds. Investors should hold municipal bonds, which are exempt from federal and often state taxation, in their taxable account. Shoven and Sialm (2000) calculate optimal locations when investors have a choice between investing in stocks, taxable bonds, and municipals. In addition to always locating municipal bonds in ones taxable account, the 6

9 optimal locations rarely leave an investor simultaneously locating taxable bonds and municipal bonds in a taxable account. 5 To investigate how investors locate their municipal bonds, we analyze the allocation and location decisions of households with a minimum allocation of 10 percent to municipal bonds. The results of this analysis are presented in TABLE III. Among those with accounts at the discount broker, municipals are not widely held; only 4 percent of discount households hold municipals. Those with accounts at the retail broker are more likely to hold municipals, but still less than one in six retail households do so. For both the discount and retail households, the ownership of municipals is more prevalent among wealth households, but certainly not pervasive. Thus, consistent with prior evidence (Feenberg and Poterba (1991)), the wealthy are more likely to hold municipals. (Based on logistic regressions, we estimate the probability of holding municipal bonds nearly triples if one holds a portfolio value in excess of $250,000.) Consistent with the common sense notion that municipals should be located in taxable accounts, virtually all (greater than 99 percent) of the retail and discount households do so. However, a substantial proportion of investors who hold municipals also hold taxable bonds in their taxable accounts. 6 One-third of the discount households and 62 percent of the retail households that hold municipal bonds also hold taxable bonds in their taxable accounts. The proportion of households that simultaneously hold taxable and municipal bonds in their taxable accounts is greater for the wealthy discount and retail households. Many households that simultaneously hold municipals and taxable bonds in their taxable accounts hold substantial amounts of both. For example among discount households that simultaneously hold municipals and taxable bonds in their taxable account, municipals represent 57 percent of their total bond allocation in their taxable accounts; for roughly half of these households, this proportion falls between 40 and 80 percent. These ratios are 5 At very high levels of risk aversion, the optimal location has some taxable bonds in a taxable account because the after-tax returns of taxable bonds are assumed to be less variable than the returns of municipal bonds. 6 We require a minimum holding of $1,000 in taxable bonds for a household to be categorized as simultaneously holding taxable bonds and municipals. 7

10 substantively similar for the other partitions that we analyze, though the retail households tend to have a higher proportion municipals. In summary, though households appear to optimally locate their municipal bonds in their taxable account, many also simultaneously hold taxable bonds in their taxable accounts. It is possible that investors simultaneously hold taxable bonds and municipals for diversification benefits. However, making reasonable assumptions about the volatility and correlation of the returns on taxable and municipal bonds, and levels of risk aversion, Sialm and Shoven (2000) find no optimally located portfolios that simultaneously hold municipals and taxable bonds in an investor s taxable account. If diversification cannot completely explain the simultaneous holding of municipals and taxable bonds in taxable accounts, the large proportion of households that do so becomes a puzzle. IV. Asset Location In this section, we present descriptive information on the location of taxable bonds, mutual funds, and individual stocks. Of course, to analyze location decisions, a household must have both a taxable and tax-deferred account (TDA). Thus, we require a household to have between 10 and 90 percent of its assets invested in a TDA. Ten percent (47,181) of retail households and 24 percent (11,501) of discount households meet this criterion. Fiftyfive percent of retail households and 49 percent of discount households hold only taxable accounts. Though many of these households face a material location decision, since they likely have TDAs elsewhere (e.g., through an employer), we are unable to observe their location decisions. Thirty-five percent of retail households and 27 percent of discount households hold only TDAs. A. Taxable Bonds A.1. Location Evidence In the absence of short-term liquidity needs, Huang (2000) and Dammon, Spatt, and Zhang (2001) document that investors should optimally locate their taxable bonds in their tax-deferred accounts. Though Shoven and Sialm (2000) reach a qualitatively similar conclusion, they document that investors might hold equity mutual funds that distribute high 8

11 levels of capital gains and dividends in their TDA, while holding municipal bonds in their taxable account. To investigate how investors locate their taxable bonds, we analyze the allocation and location decisions of households with a minimum allocation of 10 percent to taxable bonds and 10 percent to stock. The results of this analysis are presented in TABLE IV. Among discount households facing a location decision, 21 percent have at least a 10 percent taxable bond allocation and a 10 percent equity allocation, while 30 percent of wealthy households meet these minimums. A much higher proportion of retail households -- nearly 60 percent -- meet these minimum allocation requirements. Among discount households, 40 percent hold their taxable bonds solely in their TDAs; for the wealthy households, this percentage drops to 24. A much lower proportion of retail households hold their taxable bonds solely in their TDA. Put another way, the majority of households that own taxable bonds hold at least a proportion of these bonds in their taxable account. To determine if there is a preference for holding taxable bonds in TDAs, we calculate the proportion of taxable bonds held in each household s taxable account and subtract from this the proportion of assets held in its taxable account. If investors have a preference for holding taxable bonds in TDAs, the difference between these two proportions will be negative. For the discount households, the difference between these two ratios is reliably negative -- indicating a preference for holding taxable bonds in TDAs. However, for retail households the evidence for such a preference is mixed. These results are supported by the mean asset allocations in taxable accounts and TDAs (presented in the last three rows of TABLE IV). The average discount household allocates 37 percent of its TDA and 28 percent of its taxable account to taxable bonds; these allocations are roughly similar for the wealthy discount households. However, the average retail household allocates 35 percent of its TDA and 39 percent of its taxable account to 9

12 taxable bonds. Though the wealthy retail households have a higher allocation to taxable bonds in their taxable account, the difference is small. It is possible that households allocate taxable bonds to their taxable account because there is simply no room left in their tax-deferred account. To investigate this possibility, we calculate the dollar value of taxable bonds that can replace stock in each household s TDA. The average discount household can move $15,531 (or 33 percent of their total taxable bonds holding) to its TDA, though this average includes households with no taxable bonds in their taxable account. For discount households with taxable bonds in their taxable account, the average household can move $25,731 (or 55 percent of their total taxable bond holding). For the wealthy discount households, the dollar values are higher, though the percentages are roughly similar. We also find qualitatively similar results for the retail households. Our results for taxable bonds are consistent with those in Poterba and Samwick (2000) and Bodie and Cane (1997). Using data from the 1995 Survey of Consumer Finances, Poterba and Samwick (2000) document that 48 percent of investors who own taxable bonds in taxable accounts also own equity in tax-deferred accounts and that 42 percent of investors who own equity in TDAs also own taxable bonds in taxable accounts. Similarly, using data from TIAA-CREF, Bodie and Crane (1997) document that most investors hold equity and taxable bonds in both their taxable accounts and TDAs. In summary, discount households have a preference for locating taxable bonds in their TDAs, though the evidence for such a preference for retail households is weak. Nonetheless, if one excepts the extant advice that investors should allocate taxable bonds to TDAs, the average household mislocates one-third of its taxable bonds to taxable accounts. For the average household, this represents roughly 10 percent of its total portfolio value. A.2. Determinants of Cross-Sectional Variation in Taxable Bond Location In this section, we analyze the cross-sectional variation in the taxable bond location decisions of households. To do so, we calculate the ratio of taxable bonds in a household s 10

13 taxable account to the total value of taxable bonds held by the household. We regress this ratio, which we refer to as the taxable bond ratio, on various household characteristics. If investors naively allocate their taxable bonds proportionately to their taxable account and TDA, the taxable bond ratio will be equal to the ratio of taxable account value to total portfolio value, TAX TAX + TDA, where TAX is the total value of a household s taxable account and TDA is the total value of a household s TDA. If investors naively locate their taxable bonds, the coefficient estimate on this variable will be one. On the other hand, if investors follow a rule of allocating taxable bonds first to their TDA, taxable bonds are allocated to an investor s taxable account only if they face capacity constraints in their TDA. Thus, the taxable bond ratio will be equal to:, max 0, B TDA B where B is the total value of taxable bonds. If capacity is the sole explanation for taxable bond allocations in taxable accounts, the coefficient estimate on this variable will be one. Huang (2000) argues that investors might allocate low-risk assets, such as taxable bonds, to taxable accounts when faced with liquidity needs. Since investors are penalized for early withdrawals from TDAs, when faced with short-term liquidity needs they might allocate taxable bonds to their taxable account to reduce the probability of early withdrawal from their TDA. To investigate whether liquidity needs can explain the prevalent holding of taxable bonds in taxable accounts, we use two proxies for liquidity needs. First, we construct a dummy variable that takes on a value of one if the household has a taxable account value greater than $100,000. Assuming households with more than $100,000 in their taxable accounts are less liquidity constrained than households with lower account balances, the liquidity hypothesis predicts that the taxable bond ratio will be lower for these households. Among households with greater than a 10 percent allocation to taxable bonds as well as 11

14 greater than 10 percent allocation to equity, 17 percent of discount households and 33 percent of retail households have taxable account values in excess of $100,000. Second, we identify households that are accumulating significant asset holdings through purchases. To do so, we construct a dummy variable that takes on a value of one if the household has averaged 20 percent annual net deposits during our sample period (six years for the discount households and 18 months for the retail households). We define net deposits as the sum of purchases less the sum of sales divided by the sum of monthly positions. Among households with greater than a 10 percent allocation to taxable bonds as well as greater than 10 percent allocation to equity, 13 percent of discount households and 20 percent of retail households average net deposits of 20 percent. With large net deposits, we conjecture that these households are less likely to face impending liquidity needs; thus, the liquidity hypothesis predicts that accumulators will have a lower taxable bond ratio than other households. Location decisions are most profitable for investors that take full advantage of tax avoidance strategies that are available for stock. The most straightforward way to reduce taxes on equity is to avoid the sale of gains in one s taxable account so as to defer the realization of capital gains (though investors might chose to harvest losses to shelter other taxable income). We construct two dummy variables to measure the influence of equity tax avoidance on the taxable bond ratio. First, we construct a dummy variable that takes on a value of one if the household s annual equity sales (individual stocks and mutual funds) in their taxable account are less than 10 percent of the equity value in the taxable account. 7 These households, with limited sales, are deferring a substantial portion of equity capital gains. They stand to benefit most from optimal location and, thus, should have a lower taxable bond ratio than other households. Among households with greater than a 10 percent allocation to taxable bonds and a 10 percent allocation to equity, 26 percent of discount households and 51 percent of retail households are low turnover households. 12

15 Second, we construct a dummy variable that is equal to one if the household s annual equity sales and purchases in their taxable account are both greater than 50 percent of the equity value in the taxable account. These frequent traders have an average holding period of less than two years on equity in their taxable account. They stand to benefit little from optimal location and, thus, should have a higher taxable bond ratio than other households. Among households with greater than a 10 percent allocation to taxable bonds and a 10 percent allocation to equity, 29 percent of discount households and 17 percent of retail households are high turnover households. Municipals may be held as a substitute for taxable bonds in one s taxable account. Thus, we include the ratio of municipals to total taxable bonds as an independent variable in our regressions. If taxable bonds and municipals are perfect substitutes in an investor s taxable account, this variable will yield a coefficient estimate of negative one in the regression. Finally, the location decision stands to benefit younger investors more than older investors if the location remains unaltered, since younger investors have a longer investment horizon. Thus we include age as an independent variable in the regression and a dummy variable that takes on a value of one if the investor is over the age of The results of our regression analysis are presented in TABLE V. For both the discount and retail households, the percentage of assets in their taxable account and capacity considerations are both important determinants of the percentage of taxable bonds that are located in their taxable account. The negative coefficient estimate on municipal bonds indicates that investors view taxable bonds and municipals as substitutes in their taxable accounts, though they are far from perfect substitutes. 7 For retail households, we have not yet separately calculated equity turnover in their taxable accounts. Consequently, in this draft of the paper these dummy variables are based on total equity turnover in taxable accounts and TDAs. 8 Age data are available for only half of discount households. For those households missing age data, we set age equal to the mean for the entire dataset (48) and include a dummy variable that takes on a value of one for households with missing age data. This dummy variable is insignificant and is suppressed in TABLE V. 13

16 The liquidity hypothesis of Huang (2000) receives mixed support. Consistent with the liquidity hypothesis, retail households with more than $100,000 of taxable assets locate a smaller fraction of their taxable bonds to their taxable account. However, this is not true for discount households. In addition, households with large net deposits during our sample period are, if anything, more likely to locate their taxable bonds in their taxable account. There is solid evidence that households with high equity turnover, which stand to benefit the least from optimal location decisions, locate a greater proportion of their taxable bonds to their taxable account than other households. On the other hand, there is mixed evidence that households with low equity turnover, which stand to benefit the most from optimal location decisions, locate a smaller proportion of their taxable bonds to their taxable account. Finally, we find no reliable relationship between age and the taxable bond ratio for discount households. However, consistent with the notion that younger investors stand to benefit most from optimal location decision, we find a significant negative relation between age and the taxable bond ratio for retail households. In summary, we are able to explain roughly one-third of the cross-sectional variation in the taxable bond location decision. By far, the most important determinants of this location decision are the percentage of assets held in a household s taxable account and capacity constraints in its TDA. These two variables alone can explain 30 percent of the cross-sectional variation in the taxable bond ratio for discount households and 26 percent for retail households. B. Individual Stocks vs. Mutual Funds B.1. Location Evidence Mutual funds distribute a substantial portion of total returns as taxable capital gains. For example, Barclay, Pearson, and Weisbach (1998) document that the average open-end equity mutual fund earned 15.2 percent annually from 1976 to Annually, about onethird (5 percent) of this total return was distributed as capital gains and about one-sixth (2.3 14

17 percent) as ordinary income. When realized in a taxable account, these distributions represent a drag on the after-tax returns earned by investors. In contrast, those who hold individual stocks can avoid annual capital gain realizations (and also harvest losses). Huang (2000) documents that the payout ratio, defined as the ratio of returns distributed to shareholders to the total asset return, determine optimal asset location. Assuming the payout ratio of taxable bonds is greater than that of individual stocks or mutual funds, investors would prefer to first locate taxable bonds in their TDAs. Assuming space remains in their TDA, investors would next locate equity mutual funds, since they have high payout ratios relative to individual stocks. Shoven and Sialm (2000) argue that investors might be better off by investing in municipal bonds in their taxable account thereby creating space in their TDA for mutual funds with high payout ratios. To investigate how investors locate their mutual funds, we analyze the allocation and location decisions of households with a minimum allocation of 5 percent to mutual funds and 5 percent to individual stocks. The results of this analysis are presented in TABLE VI. Roughly half of discount households facing a location decision meet these minimum allocation requirements, while slightly less than one-third of retail households meet these minimums. Both groups display a strong preference for holding mutual funds in their TDAs. Thirty-seven percent of discount households and 54 percent of retail households hold mutual funds exclusively in their TDAs, while only 12 percent of each hold mutual funds exclusively in their taxable accounts. To determine if there is a preference for holding mutual funds in TDAs, we calculate the proportion of mutual funds held in each household s taxable account and subtract from this the proportion of assets held in its taxable account. If investors have a preference for holding mutual funds in TDAs, the difference between these two proportions will be negative. For both discount and retail households, the difference between these two ratios is large and reliably negative -- indicating a preference for holding mutual funds in TDAs. 15

18 These results are supported by the mean asset allocations in taxable accounts and TDAs (presented in the last four rows of TABLE VI). The average discount household allocates 64 percent of its taxable account to individual stocks and 29 percent to mutual funds; in contrast, these same households allocate 38 percent of their TDA to individual stocks and 54 percent to mutual funds. The same pattern emerges for the remaining sample partitions that we analyze. We calculate the dollar value of mutual funds that can replace individual stocks in each household s TDA. The average discount household can move $6,615 (or 20 percent of their total mutual fund holding) to its TDA, though this average includes households with no mutual funds in their taxable account. For discount households with mutual funds in their taxable account, the average household can move $10,429 (or 33 percent of their total mutual fund holding). For the wealthy discount households, the dollar values are higher, though the percentages are roughly similar. We also find qualitatively similar results for the retail households. In summary, the investors that we analyze have a strong preference for locating mutual funds in their TDAs. Nonetheless, if one excepts the extant advice that investors should allocate mutual funds with high distributions to their TDAs in lieu of individual stock, the average household mislocates 20 percent of its mutual funds to taxable accounts. B.2. Determinants of Cross-Sectional Variation in Mutual Fund Location In this section, we analyze the cross-sectional variation in the mutual fund location decisions of households. To do so, we calculate the ratio of mutual funds in a household s taxable account to the total value of mutual funds held by the household. We regress this ratio, which we refer to as the taxable fund ratio, on various household characteristics. If investors naively allocate their mutual funds proportionately to their taxable account and TDA, the taxable fund ratio will be equal to the ratio of taxable account value to total portfolio value, 16

19 TAX TAX + TDA. If investors naively locate their mutual funds, the coefficient estimate on this variable will be one. On the other hand, if investors follow a rule of allocating taxable bonds first to their TDA and mutual funds second to their TDAs, mutual funds are located in an investor s taxable account only if they face capacity constraints in their TDA. Thus, the taxable fund ratio will be equal to: 0. if MF ( TDA B) MF ( TDA B) max 1,. if MF > ( TDA - B) MF where MF is the total value of mutual funds. If capacity is the sole explanation for mutual fund allocations in taxable accounts, the coefficient estimate on this variable will be one. As before, we also include age and dummy variables for low and high equity turnover households. We exclude the variables designed to measure liquidity needs, since investors would likely locate safe assets (i.e., bonds) rather than stock in their taxable accounts when faced with short-term liquidity needs. The results of our regression analysis are presented in TABLE VII. As for taxable bonds, the most important determinants of the taxable fund ratio are the proportion of assets in ones taxable account and capacity considerations. We are able to explain 27 percent of the cross-sectional variation in the taxable fund ratio for discounts households and 14 percent for retail households; virtually all of this explanatory power comes from the percentage of assets held in taxable accounts and capacity considerations. However, in contrast to the results for taxable bonds, capacity considerations appear to be more important than the proportion of assets in a household s taxable account (with the exception of wealthy discount households). This suggests that investors are more likely to depart from a naïve allocation strategy for mutual funds than for taxable bonds. This is somewhat puzzling, since extant models of optimal asset location generally yield a preference for locating taxable bonds first (rather than equity mutual funds) to a TDA. 17

20 Of the remaining variables, only low equity turnover is a robust determinant of mutual fund location. Low turnover households stand to benefit most from locating mutual funds in their TDA and they have a stronger preference for doing so than other households. V. Conclusions Investors can improve their after-tax returns by strategically locating their investments in taxable or tax-deferred accounts. Extant models of optimal asset location argue that investors should first locate assets with high payout ratios (e.g., taxable bonds) to their tax-deferred accounts. We analyze the location decisions of households with accounts at a discount broker and households with accounts at a retail broker. On one hand, the location decisions indicate that the average household is tax aware. For example, we find some evidence that the average household prefers to locate taxable bonds in their retirement accounts, and strong evidence that the average household prefers to locate mutual funds, rather than individual stocks, in their retirement accounts. On the other hand, our empirical results present several puzzles. First, many households concurrently hold municipal bonds and taxable bonds in their taxable account. Second, more than half of the households hold taxable bonds in their taxable accounts, despite having room to move at least a portion of this investment to their retirement account. Third, the preference for holding equity mutual funds in retirement accounts appears to be much stronger than the preference for holding taxable bonds in retirement accounts. These three facts do not square well with extant models of optimal asset location. We conclude that either the existing models of optimal asset location are incomplete or a substantial fraction of investors are mislocating their assets. 18

21 References Ameriks, John, and Stephen P. Zeldes, 2000, How Do Household Portfolio Shares Vary with Age?, working paper, Columbia University, NY, NY. Bajtelsmit, Vickie L., and Alexandra Bernasek, 1996, Why do women invest differently than men?, Financial Counseling and Planning, 7, Bajtelsmit, Vickie L., and Jack L. Vanderhei, 1997, Risk aversion and pension investment choices. In Positioning Pensions for the Twenty-first Century, eds. Michael S. Gordon, Olivia S. Mitchell, and Marc M. Twinney, Philadelphia: University of Pennsylvania Press, Barber, Brad M., and Terrance Odean, 2001, Boys will be Boys: Gender, Overconfidence, and Common Stock Investment, forthcoming, Quarterly Journal of Economics. Bodie, Zvi, and Dwight B. Crane, 1997, Personal Investing: Advice, Theory, and Evidence, Financial Analysts Journal, vol.?, p Canner, Niko, N. Gregory Mankiw, and David N. Weil, 1997, An Asset Allocation Puzzle, American Economic Review, vol. 87, p Dammon, Robert M., Chester S. Spatt, and Harold H. Zhang, 2001, Optimal Asset Location with Taxable and Tax-Deferred Investing, working paper, Carnegie Mellon University, Pittsburgh, PA. Feenberg, Daniel R., and James M. Poterba, 1991, Which Households own Municipal Bonds? Evidence from Tax Returns, National Tax Journal, vol. 44, p Hinz, Richard P., David D. McCarthy, and John A. Turner, 1997, Are women conservative investors? Gender differences in participant-directed pension investments. In Positioning Pensions for the Twenty-first Century, eds. Michael S. Gordon, Olivia S. Mitchell, and Marc M. Twinney, Philadelphia:University of Pennsylvania Press, Huang, Jennifer, 2000, Taxable or Tax-Deferred Account? Portfolio Decision with Multiple Investment Goals, working paper, MIT University, Cambridge, MA. Jagannathan, Ravi, and N. R. Kocherlakota, 1996, Why Should Older People Invest Less in Stocks then Younger People?, Federal Reserve Bank of Minneagpolis Quarterly Review, vol. 20, p Jianakoplos, Nancy A., and Alexandra Bernasek, 1998, Are women more risk averse?, forthcoming, Economic Inquiry. Mossin, Jan, 1968, Optimal Multiperiod Portfolio Policies, Journal of Business, vol 41, p

22 Poterba, James M., and Andrew A. Samwick, 1997, Household Portfolio Allocation over the Life Cycle, National Bureau of Economic Research working paper 6185, Cambridge, MA. Poterba, James M., and Andrew A. Samwick, 2000, Taxation and Household Portfolio Composition: U.S. Evidence from the 1980s and 1990s, working paper, MIT University, Cambridge, MA. Samuelson, Paul A., 1969, Lifetime Portfolio Selection by Dynamic Stochastic Programming, The Review of Economics and Statistics, vol. 51, p Shoven, John B. and Clemens Sialm, 2000, Asset Location in Tax-Deferred and Conventional Savings Accounts, working paper, Stanford University, Stanford, CA. Sundén, Annika E., and Brian J. Surette, 1998, Gender differences in the allocation of assets in retirement savings plans, working paper, Board of Governors of the Federal Reserve System. 20

23 TABLE I: Mean Asset Allocation Discount households (Panel A) hold accounts at a discount broker and asset allocations for these households are based on month-end position statements from January Holdings are categorized based on 40 product codes provided to us by the discount broker. Retail households (Panel B) hold accounts at a full-service broker and asset allocations for these households are based on month-end account summaries and positions from November Holdings are categorized based on 11 product codes provided to us by the retail broker. Households Households Households with > $10,000 with > $100,000 with < 99% Stock Panel A: Discount Households No. of Households 48,084 9,736 11,356 Mean Portfolio Value $92,694 $340,776 $160,228 % Assets in Taxable Accounts % of Households with Stock Allocation > 99% n.a. Mean Asset Allocation: % Stock % Taxable Bonds % Municipals Panel B: Retail Households No. of Households 452, , ,659 Mean Portfolio Value $178,819 $490,803 $192,612 % Assets in Taxable Accounts % of Households with Stock Allocation > 99% n.a. Mean Asset Allocation: % Stock % Taxable Bonds % Municipals

24 TABLE II: Cross-Sectional Regressions of Percentage Stock Allocation on Demographic Variables The dependent variable is the percentage stock allocation for each household. The independent variables include age (in years) and 11 dummy variables. The dummy variables take on a value of one if the head of household: is a woman; is single; is a woman and single; has income between 50,000 and 100,000; has income greater than 100,000; has net worth between 75,000 and 250,000; has net worth greater than 250,000; has a ratio of portfolio value to net worth between 0.25 and 0.50; has a ratio of portfolio value to net worth greater than 0.50; owns a recreational vehicle; and owns a motorcycle (respectively). Discount Households with Retail Households with Independent Variable > $10,000 > $100,000 < 99% Stock > $10,000 > $100,000 < 99% Stock Intercept Age (years) (0.05)** Woman Single (0.81) 0.06 (0.97) (0.23) Woman*Single (0.16) (0.18) 1.76 (0.57) 50 < Income à (0.09)* Income > (0.16) 3.05 (0.05)** Not Available 75 < Net Worth à (0.09)* (0.79) 1.45 (0.72) Net Worth > (0.41) RV Ownership (0.60) 4.55 (0.24) (0.67) Motorcycle Ownership 1.25 (0.03)** (0.09)* 2.13 (0.16) Adj. R-squared (%) No. of Observations 9,701 2,134 2, , , ,577 22

25 TABLE III: Asset Allocation and Location for Municipal Bond Holders The sample consists of households with a minimum allocation of 10 percent to municipal bonds. Discount households hold accounts at a discount broker and asset allocations for these households are based on month-end position statements from January Holdings are categorized based on 40 product codes provided to us by the discount broker. Retail households hold accounts at a full-service broker and asset allocations for these households are based on month-end account summaries and positions from November Holdings are categorized based on 11 product codes provided to us by the retail broker. Discount Households with Retail Households with > $10,000 > $100,000 > $10,000 > $100,000 No. of Households with Municipals > 10% 1, ,163 32,045 % of Households with Municipals > 10% Mean Portfolio Value $252,938 $453,256 $388,035 $702,418 % Assets in Taxable Account Mean Asset Allocation: % Stock % Taxable Bonds % Municipals % of Households with Municipals > 10% that: Hold municipals in taxable account Hold taxable bonds Hold taxable bonds in taxable account % municipals to total bonds in taxable account for households with both

26 TABLE IV: Asset Allocation and Location for Taxable Bond Holders The sample consists of households with a minimum allocation of 10 percent to taxable bonds and 10 percent to stock, and taxable account value between 10 and 90 percent of total portfolio value. ***,** - significant at the 1 or 5 percent level, respectively (twotailed test). Discount Households with Retail Households with > $10,000 > $100,000 > $10,000 > $100,000 No. of Households with Taxable Bonds > 10% 2, ,396 13,792 % of Households with Taxable Bonds > 10% Mean Portfolio Value $154,512 $331,545 $273,450 $463,316 % Assets in Taxable Account % of Households with Taxable Bonds that hold Taxable Bonds: Solely in taxable account Solely in tax-deferred account (TDA) In both accounts Mean Value of Taxable Bonds in Taxable Account that can be moved to TDA All Households $15,531 $33,635 $26,082 $42,870 Households with Taxable Bonds in Taxable Account $25,732 $44,356 $30,784 $47,246 [Taxable Bonds in Taxable Account / Total Taxable Bonds] less [Taxable Account Value / Total Portfolio Value]: mean (0.03)** -2.9 median Mean Asset Allocation: Tax TDA Tax TDA Tax TDA Tax TDA % Stock % Taxable Bonds % Municipals