ASSET ALLOCATION AND ASSET LOCATION DECISIONS: EVIDENCE FROM THE SURVEY OF CONSUMER FINANCES

Transcription

1 CONFERENCE DRAFT COMMENTS WELCOME ASSET ALLOCATION AND ASSET LOCATION DECISIONS: EVIDENCE FROM THE SURVEY OF CONSUMER FINANCES Daniel Bergstresser MIT James Poterba MIT, Hoover Institution, and NBER March 2001 ABSTRACT This paper uses data from the 1989 and 1998 Surveys of Consumer Finances to investigate how many households face significant choices with respect to the location of their assets between taxable and taxdeferred accounts. It shows that in 1998, 45 percent of households had at least some assets in a taxdeferred account. More than six million households have at least $50,000 in financial assets in a taxdeferred account, and at least an equal amount in a taxable setting. More than ten million households had at least $25,000 in their taxable and tax-deferred accounts. With respect to asset allocation, we find that many households hold equities in their tax-deferred accounts, but not in their taxable accounts. The share of equity held in tax-deferred accounts is greater than the share of all assets held in such accounts. This suggests that a substantial group of households is not following a bonds first in the tax-deferred account asset location strategy. We also find a modest positive correlation between the equity allocation share in a household s tax-deferred account, and that share in the taxable account. We are grateful to Joel Dickson and John Shoven for helpful conversations and to the Hoover Institution, the National Institute of Aging, and the National Science Foundation for research support.

2 1 The rapid growth of self-directed retirement plan assets in the last two decades has generated a new set of financial decisions for many households in the United States. While households have always confronted the asset allocation problem, the question of which assets to purchase and how much to invest in each of them, they now also face an asset location problem. This concerns the choice of how much of a given asset to hold in a taxable account, and how much to hold in a tax-deferred account. The asset location problem has recently begun to attract substantial attention from researchers in public finance and financial economics, and it is a frequent topic of discussion in the popular financial press. Shoven (1998) outlined the basic structure of the problem, which is related to the problem of pension plan asset allocation originally studied by Black (1980) and Tepper (1981). He pointed out that for individuals holding bonds and stocks through financial intermediaries, such as mutual funds, the traditional advice to first allocate bonds to the tax-deferred account, and to first allocate stocks to the taxable account, might not apply. Subsequent work by Dammon, Spatt, and Zhang (2000), Shoven and Sialm (1998, 2000), Huang (2000) and others has provided further insight on the optimal asset location strategy for households facing particular tax environments and particular categories of investment options. Shoven and Sialm (2000) and Poterba, Shoven, and Sialm (2000), for example, draw attention to the potential importance of the availability of tax-exempt bonds in affecting the solution to the asset location decision. Huang (2000) examines the role of borrowing on taxable account in altering the optimal location strategy. She finds that in many cases, the investor should hold only bonds in the taxdeferred account, while borrowing on taxable account to reduce bond exposure if the investor wants to hold a greater fraction of the overall portfolio in equities. All of the studies of asset location suggest that households should hold their most heavily taxed assets in their tax-deferred account. The differences in the analysis in various studies turns on differences in the set of assets investors are permitted to hold, and in differences in the effective tax burdens on these assets. Despite the recent interest in the conceptual question of how households should locate their assets, there has been relatively little research on the actual behavior of taxable investors. With the notable exception of Bodie and Crane s (1997) study of the asset location decisions of TIAA-CREF

3 2 investors, there has been little analysis of the way households allocate their portfolios within, and across, their taxable and tax-deferred accounts. In this paper, we use data from four waves of the Survey of Consumer Finances (SCFs), conducted in 1989, 1992, 1995, and 1998, to provide evidence on asset location patterns. The SCF data are not ideal for studying the asset location question, because they do not provide precise information on the amounts that investors hold in their taxable and tax-deferred accounts. They do, however, permit us to identify the broad asset allocation patterns for investors with both taxable and tax deferred accounts. This paper is divided into four sections. The first presents information on the number of households that face asset location decisions. We identify such households by the presence of substantial asset holdings in both taxable and tax-deferred accounts. We present a variety of measures of the number of households facing asset location decisions, using different definitions of taxable and tax-deferred assets. Section two explores the asset allocation decisions that investors make within their taxable and tax-deferred accounts. It shows that a substantial fraction of households -- more than one third, and substantially more for some income and demographic groups -- has a higher equity allocation in their tax deferred account than in their taxable account. Section three focuses more precisely on the asset location question. It compares the share of equity assets that are held in the tax-deferred account with the share of all assets that are held in the TDA. The results in this section confirm the tendency for equity to be held disproportionately in the tax-deferred account. Finally, there is a brief conclusion. 1. How Many Households Face Asset Location Choices? The number of households with substantial balances in tax-deferred retirement saving accounts (TDAs) has increased sharply in the last two decades. This is the group that faces the asset location problem. To investigate the number of such households, we use data from the 1998 Survey of Consumer Finances (SCF), as well as earlier versions of the same survey; the survey has been conducted every three years since It is the best available source of data on household wealth and its components. The

4 3 survey asks a relatively comprehensive set of questions, has a large sample size, and oversamples the high income and high net worth households that hold the bulk of financial and nonfinancial assets. Kennickell, Starr-McCluer, and Surette (2000) describe the most recent wave of the Survey, which was conducted in It sampled 4309 households, with 2813 in the random sample and 1496 in the stratified random sample that over-weighted those with high incomes or net worth. Four households are excluded from the public use dataset due to disclosure concerns, leaving a sample with 4305 observations. Reflecting the high-income oversampling design, one fourth of the households in the survey have net worth of over a million dollars. All of our tabulations weight the various observations in the survey by their sampling weights so that our reported statistics should be representative of the U.S. population as a whole. Data from the SCF allow us to estimate the equity share of assets in tax deferred accounts (TDAs) as well as the equity share for non-tda assets. For assets outside of tax-deferred accounts, the survey asks respondents to separately report the dollar values of direct stock holdings, equity mutual fund shares, and mixed equity-fixed income mutual fund shares. Aggregating these reported asset holdings provides a measure of equity held in taxable accounts. Our equity variable does not include equity in privately-held companies. The SCF also provides considerable detail on fixed-income assets held outside of tax-deferred accounts. We construct two measures of non-tda fixed-income assets. The first includes only certificates of deposit, savings bonds, and other bonds held directly and through mutual funds. For studying asset allocation decisions, we combine tax-exempt and taxable fixed income securities; when we focus on asset location issues, we separate them. The SCF distinguishes these types of fixed income investments when they are held outside the TDA; it does not ask about the type of fixed income assets in tax-deferred accounts. The second, and broader, measure includes all of the assets in the narrow definition, along with currency and transaction accounts such as checking, saving, and money market deposits, as well as the value of whole life insurance and trust assets. We also compute two measures of the total amount held in tax-deferred accounts. Our broad measure adds together the reported values of IRA and Keogh accounts and the value of all assets in defined contribution pension plans. This includes not only 401(k) and 403(b) assets but also assets in thrift/savings

5 4 plans, profit sharing plans, deferred compensation plans, and other types of defined contribution plans. Our narrow measure includes only assets held in 401(k), 403(b), and SRA accounts. Since some defined contribution plans that are not included in these categories allow participants little or no control over their asset allocation decisions, the narrow measure of tax-deferred assets is more likely to capture the assets that are directly controlled by individual investors. The SCF asks respondents with assets in TDAs whether these assets were held mostly or all in stock, were split between stock and interest earning assets, or were invested in other assets. We allocate all of the assets in accounts identified as mostly or all in stock to equity, half of the value of split accounts, and none of the value of other accounts to equity. The ratio of these equity holdings to the total value of the tax-deferred account provides our estimate of the equity share in these accounts. Table 1 presents summary information on the percentage of households in the Survey of Consumer Finances with tax-deferred accounts. The table reports findings from the four most recent SCFs, those for 1989, 1992, 1995, and The total number of households in the U.S. economy rises from 93 million to million over the period spanned by these surveys. We report two different measures of the percentage of households with tax-deferred accounts. The first, shown in the second column of Table 1, is the percentage with Individual Retirement Accounts, 401(k) plans, 403(b) plans, or other self-directed retirement saving plans. This is our narrow measure of TDA ownership, as noted above. The percentage of households with such accounts rose from 29.1 percent in 1989 to 45 percent in The next column in Table 1 adds other defined contribution plans to the definition of tax-deferred accounts, thereby obtaining our broad measure of tax deferred accounts. Adding other defined contribution plans increases the percentage of households with assets in tax-deferred accounts from 29 to 38 percent in 1989, and from 45 to 49 percent in The narrowing difference between these two measures of access to tax-deferred accounts suggests that between 1989 and 1998, the share of households with non- 401(k) or 403(b) defined contribution pension plans who also have one or these plans has increased. The next two columns of Table 1 show the percentage of households with financial assets outside their tax-deferred account (TDA). One column uses our "narrow measure" of financial assets outside the

6 5 TDA. Not all households own such financial assets. Approximately 45 percent of the households in each of the four Surveys of Consumer Finances report ownership of these assets. The next column shows the percentage of households owning a broader category of taxable financial assets. Primarily because this category includes checking accounts and money market funds, the percentage of households owning this broader aggregate of taxable financial assets is much higher: 91.8 percent in We distinguish the narrow from the broad measure of taxable financial assets to recognize the possibility that households with highly liquid assets on taxable account, but no other financial assets, may not really face operative asset location problems. The last column of Table 1 shows the percentage of households owning tax-exempt bonds outside their tax-deferred account. The SCF does not permit us to distinguish taxable and tax-exempt bonds in the TDA, but Barber and Odean (2001) find that very few households hold such tax-exempt assets in their tax deferred accounts. We find that approximately 6.5 percent of households own tax-exempt bonds; this fraction is quite stable over the nine years spanned by the surveys we consider. The information in Table 1 provides some insight on the number of households that might face asset location decisions. To focus more precisely on households with non-trivial levels of both taxable and taxdeferred assets, Table 2 reports the number of households with more than various threshold levels of assets in both taxable and tax-deferred accounts. The table shows four sets of statistics, two each for 1989 and We report results using both the narrow measure of taxable assets, and the narrow measure of tax-deferred assets, as well as using the broad measure of taxable assets, and the broad measure of tax-deferred assets. The first panel of Table 2 reports the number of households with given levels of taxable and taxdeferred assets in 1998, where the narrow definition is applied to both sets of assets. The table shows that 15.3 million households had more than $10,000 in both taxable and tax deferred assets, while 10.2 million had more than $25,000 in both settings and 6.1 million had more than $50,000. A smaller but still substantial group, 3.2 million households, had more than $100,000 in both types of accounts. When we repeat the calculations using the broad measures of both taxable and tax-deferred assets, we find that 24.8 million households have at least $10,000 in each type of account. We also find that 15.1 (8.8) million have more than

7 6 $25,000 ($50,000) in each place, while 4.5 million have more than $100,000 in both their taxable accounts and their tax-deferred accounts. The lower two panels in Table 2 report comparable statistics from the 1989 Survey of Consumer Finances. Using the narrow measures of both tax-deferred and taxable assets, we find that 8.6 million households had more than $10,000, and 2.6 million had more than $50,000, in both taxable and tax-deferred accounts in (The asset thresholds are measured in $1998.) Using the broad definitions of both asset categories, we find 15.4 million households with more than $10,000 in each setting, and 4.7 million with more than $50,000. These statistics, as well as those for the $25,000 and $100,000 thresholds, suggest that the number of households with substantial balances in both taxable and tax-deferred accounts nearly doubled between 1989 and Tables 1 and 2 provide information on the number of households with various levels of taxable and tax-deferred assets, but they do not shed light on the relative importance of assets inside and outside taxdeferred accounts. Table 3 addresses this issue, by presenting data on the distribution of the ratio of TDA assets, to total assets, in The table shows that the median household with both tax-deferred and taxable financial assets, using our narrow definition of both types of assets, had 55.6 percent of their financial assets in tax deferred assets. At the 25th percentile, this value was 26.7 percent for all households with both taxable and tax-deferred assets. At the 75th percentile, it was 85.4 percent. The table also shows the effect of restricting the set of households in the calculations to those with at least a given threshold of net worth. As we raise the net worth cutoff, the distribution of tax-deferred assets relative to all financial assets shifts toward the left. The median value of this ratio for households with at least $250,000 in net worth, for example, is 46.4 percent, compared with 55.6 percent for all households with both TDA and non-tda financial assets. The lower panel of Table 3 presents similar calculations using our broad definit ions of both taxable and TDA assets. In this case the median ratio of TDA assets to total financial assets is lower than when we use the narrow measures; this reflects the substantial increase in non-tda assets that is associated with the inclusion of checking accounts and money market funds.

8 7 Tables 2 and 3 do not provide any information on the characteristics of the households that have both taxable and tax-deferred financial assets. Table 4 addresses this issue by providing summary statistics on the households in several asset-level categories. It presents information using the narrow definitions of both taxable and tax-deferred assets as well as the broader measure. The results are not surprising. In the narrow/narrow case, the median net worth of households with more than $50,000 in both taxable and taxdeferred accounts is $720,000, while the median for those with more than $100,000 in both settings is $1.18 million. In both cases, mean net worth is substantially greater than median ($1.80 million for those with more than $50,000 in both settings, and $2.58 million for those with more than $100,000). The age of the household head increases as the threshold level of assets in the TDA and taxable account increases, as does the percentage of household heads that are college graduates. The median age of household heads in households with more than $25,000 in both taxable and tax-deferred settings is 53.3 years, while for those with $56,000, it is 55.1 years. Median household income rises from $82,300 for those with more than $25,000 in both the TDA and the taxable account, to $124,900 for those with more than $100,000 in each setting. The percentage of college graduates is higher among the group with more than $100,000 in each setting than among those with lower levels of asset holdings. The results using the broad definition of both tax-deferred and taxable assets are broadly similar to those using the narrow definition. 2. Asset Allocation Patterns, Taxable and Tax-Deferred Accounts The Survey of Consumer Finances data permit us to study the asset allocation decisions of households with both taxable and tax-deferred accounts, as well as the asset location decisions of these households. This section presents evidence on the asset allocation issue. We begin by presenting evidence on the set of households that own some corporate equity, and the set that owns some fixed income assets. Table 5 shows the set of households that own equity in each of the Surveys of Consumer Finances. In 1998, using the narrow definition of taxable and tax-deferred assets, we find that 45.8 percent of all households owned some corporate stock. Of this group, 18.1 percent, or forty percent of the total, hold equity only through a tax-deferred account. Just over sixteen percent of all

9 8 households, or roughly one third of those who own equity, own equity both inside and outside their taxdeferred accounts. Eleven percent of all households, or one quarter of those who own equity, hold equity outside their tax-deferred account but not in the TDA. Table 6 presents comparable statistics for ownership of fixed-income assets. This table combines taxable and tax-exempt bonds into a single aggregate category. In 1998, we estimate that 48.8 percent of households owned fixed income assets (using our narrow definition of such assets, and thereby excluding transactions accounts). Roughly one quarter of this group holds fixed-income assets inside their TDA but not outside, while nearly half holds fixed income assets outside the TDA but not inside the TDA. Tables 5 and 6 lead naturally to the question of how many households own only equity in their taxdeferred account, and only fixed income assets outside their TDA. Table 7 provides information on this issue. The upper panel of the table uses our narrow definitions of both taxable and tax-deferred assets. It shows that in 1998, of the million households in the SCF, 56.4 million had no TDA assets. Of the remaining 46.2 million households, 16.2 million had no financial assets outside their TDA (recall that this uses the narrow definition of financial assets that excludes checking and saving accounts, as well as money market mutual funds). Of the 30 million households with both some TDA assets and some financial assets outside their TDA, 13.6 million have only equity in their TDA, while 6.7 million have only fixed income instruments. Of the group that holds only equity in the TDA, 3.2 million households have only fixed income investments outside the TDA. This group, along with the 5.9 million households with equity and fixed income assets outside the TDA but only equity in the TDA, seem not to be following a strict "bonds in the TDA first" asset location rule. This group represents 30.3 percent of the households with tax-deferred assets. Table 7 also presents similar results using the broad definition of both TDA and non-tda assets, and using a net worth threshold to eliminate households without substantial assets from our sample. In the latter case, when there are 14.4 million households with at least some TDA assets, we find that 31.3 percent of households with TDA hold only equity in their TDA and fixed income investments outside the TDA. The results in Table 7 are insensitive to our classification of tax-exempt bonds along with equity, as a low-tax asset that can be held outside the TDA, or along with taxable bonds, as a fixed income asset. There

10 9 are very few households who hold all of their taxable fixed income assets in tax exempt bonds, and who would therefore be categorized differently if we grouped tax exempt bonds with equity rather than with other fixed income assets. In future work we hope to explore the role of tax-exempt bonds in more detail. We now consider more detailed information on asset allocation fractions within both taxable and taxdeferred accounts. Table 8 presents information comparing the equity share of assets in taxable and taxdeferred accounts for households with various levels of total assets. The table focuses on households with more than a given threshold of assets in both the taxable and the tax-deferred account. The first panel focuses on the broad definition of both taxable and tax-deferred assets, while the second uses the narrow definition. Within each sub-panel of three rows, the nine entries (in the three columns) add up to 100 percent. The entries show the fraction of households in the sub-category, or the fraction of assets held by households in each subcategory, that are held by households with particular asset allocation combinations. The first sub-panel illustrates the interpretation of the results. It considers households with at least $25,000 in both their taxable and tax-deferred accounts. Of these households, 7.1 percent hold less than 10 percent of their TDA, and less than 10 percent of their taxable assets, in equities. The largest group, representing 28.1 percent of the total, holds more than 90 percent of their TDA balance in equities, and between 10 and 90 percent of the assets outside the TDA in equities. More than half of the households report that they hold a higher percentage of their TDA, than of their taxable assets, in equities. Only 12.2 percent of households report holding a smaller percentage of their taxable assets than their tax-deferred assets in equities. These conclusions vary slightly when we consider a higher asset threshold ($100,000) for both taxable and tax-deferred accounts, but the broad patterns are very similar. The results in Table 8 combine taxable and tax-exempt debt held outside the TDA. The results do not change substantially if we combine equity and tax-exempt debt in constructing the equity share outside the TDA. The lower panel of Table 8 presents similar calculations using our broader definition of both taxable and tax-deferred assets. The results are somewhat different than those with the narrow definition, because there are now fewer households with very high equity shares outside their tax-deferred accounts. We now find that the single most common asset allocation pattern is between 10 and 90 percent equity in the taxable

11 10 account, and more than 90 percent equity in the tax-deferred account. We also find that more than half of households reports that they hold a greater percentage of their tax-deferred account than their taxable account in equities. In both the upper and lower panels of Table 8, we present tabulations weighting households equally, and weighting them by their total financial assets. In the latter case we find greater evidence of larger equity holdings outside the TDA, which reflects the higher share of equity held by assets with substantial financial assets. Table 9 presents information similar to that in Table 8, but it tracks the evolution of the asset allocation patterns over time. It shows a substantial increase over the period in the fraction of households that hold more than 90 percent of their TDA in equities (21.2 percent in 1989, using the narrow definition of TDA assets, and 52 percent in 1998). It also shows a substantial increase, from 18.3 percent to 44.2 percent, in the percentage of households with more than 90 percent of their taxable assets in equities. In 1989, using the narrow measure of both taxable and tax-deferred assets, at least 26.3 percent of households allocated a greater fraction of their TDA than their taxable account to equities. The comparable statistic was 29.6 percent in 1992, 31.1 percent in 1995, and 26.9 percent in Table 10 moves beyond the analysis of aggregate information on asset allocation, and explores how the equity share of both taxable and tax-deferred accounts varies with new worth and the age of the household head. The results in the upper panel of Table 10 correspond to the narrow definition of both taxable and tax-deferred assets, while the results in the lower panel use the broader definition of both asset stocks. Each of the estimates corresponds to the coefficient on an indicator variable for the household's presence in a given net worth or age category. The estimating equations are linear regressions in which the dependent variable is the share of the assets in either the taxable account or the TDA that are held in equities. The sample is the set of 1709 SCF households in 1998 who hold both taxable and tax-deferred assets. The results in Table 10 suggest several conclusions. First, there is a weak tendency for the share of the tax-deferred account allocated to equity to rise with household net worth. In the upper panel, for households with less than $100,000 in net worth, the estimated coefficients center around 0.60, while for those with more than $1,000,000 in net worth, the coefficients are closer to The differences across

12 11 wealth categories are somewhat more pronounced, with the $1 million plus category rising to nearly 0.80, when we control for age as well as net worth. Second, there is a stronger relationship between net worth and equity share for assets held in taxable accounts than for equity in TDAs. Again focusing on the upper panel, the difference between the coefficients for households with net worth of less than $100,000, and the coefficients for those with more than $1 million, is roughly 0.30 when we do not control for age, and 0.40 when we do. Thus the derivative of asset allocation percentage with respect to net worth is more than twice as large in taxable as opposed to tax-deferred accounts. These two results, taken together, suggest that households in the lower part of the net worth spectrum are more likely to allocate tax-deferred assets, than taxable assets, to equities. The results in the lower panel, using the broader measures of both taxable and taxdeferred assets, are consistent with the findings in the upper panel. The third pattern that emerges from Table 10 concerns the relationship between age and asset allocation within both taxable and tax-deferred accounts. The table shows that older households, those with heads or over the age of 65, hold a smaller share of their taxable account in equities than their younger counterparts. The same age pattern emerges with respect to taxable assets, but the age-related differences are larger in the tax-deferred account than in the taxable account. Thus the age-related patterns are just the reverse of the net worth patterns, with larger age-related differences in the TDA than in the taxable account. The age-related patterns of equity share are somewhat less pronounced when we use the broad measures of TDA and taxable account balances than when we use the narrow measures. Table 11 presents results that are similar to those in Table 10, but that use income rather than net worth to stratify households. Higher income households, like higher net worth households, allocate a larger share of their tax-deferred assets to equities than their lower income counterparts. The pattern also emerges with respect to taxable assets, but unlike the results in Table 10, we do not find pronounced differences between the income-related increases in equity exposure in taxable and tax-deferred accounts. We continue to find age-related differences in the asset allocation fraction for both taxable and tax-deferred accounts, with older households holding a smaller share of their tax-deferred assets in stocks. These patterns are evident

13 12 regardless of whether we use the narrow or the broad measure of assets held in taxable and tax-deferred accounts. 3. Asset Location Decisions The information in Tables 10 and 11 provides some insight on asset allocation, but it does not consider the correlation between the equity share of a household's TDA and taxable portfolio. To explore this, we present three types of evidence on asset location decisions. First, we examine the distribution of equity shares in taxable accounts, conditional on the division of the tax-deferred account, as well as the distribution of the allocation of the tax-deferred account conditional on the allocation in the taxable account. Second, we explore the correlation between the share of the taxable account that is held in equity, and the share of the tax-deferred account that is held in equity assets. Finally, we compute an equity bias statistic and a fixed income bias statistic that describes the fraction of all equity holdings, or all fixed-income holdings, that are in the TDA relative to the fraction of all assets that are in the TDA. We begin by studying the distribution of equity shares in taxable accounts, conditional on holdings in the tax-deferred account. Table 12, which presents this information, is divided into two panels. The first panel presents tabulations using the narrow definition of assets in both the taxable and the tax-deferred account. Holdings of tax-exempt bonds are combined with taxable fixed income assets held outside the TDA. (We have also computed statistics like those in this table, but combining tax-exempt debt with equity, on the grounds that it is a low-tax asset for taxable investors. The results are very similar to those in the upper panel.) In the bottom panel, we use our broader definition of taxable and tax-deferred assets, and we parallel the top panel in combining tax-exempt debt with taxable fixed income assets. Given our asset definitions, there are more households in the sample for the bottom panel than for the upper panel. To illustrate the findings in Table 12, consider the entries in the top panel corresponding to households with at least $25,000 in both taxable and tax-deferred accounts. There were 10.2 million households in this category in Of the 7.3 million who hold mostly or all equity in their tax-deferred account, roughly four times as many hold more than half of their non-tda assets in equity as hold more

14 13 fixed-income assets than equity in the TDA. When we use the broad definition of both taxable and taxdeferred assets, in the bottom panel, there is a roughly even split between the groups who hold at least half of their taxable account in equities, and those who hold less than half of their taxable assets in equities. For the narrow definition of assets, 5.8 million households hold more than fifty percent equity in their taxable accounts, while 1.5 million hold less. For the broad definition, the respective statistics are 5.1 million and 5.9 million. There is a greater tendency for those with higher TDA and taxable balances to hold a high fraction equity in both the taxable and the tax deferred account. Table 12 also shows that for households with all or mostly fixed-income assets in their tax-deferred accounts, there is a greater likelihood of holding less than half of non-tda assets in fixed income assets than in equities. For the set of households with at least $25,000 in both their TDA and their taxable account, using the broad measure of TDA and non-tda assets, we find that 3.1 of 4.1 million, or roughly three quarters, of those with more than half of their TDA in fixed income assets holds more than half of their non-tda assets in fixed income. The pattern is somewhat different when we consider the narrow measure of assets. Among those with mostly fixed income in their TDA, the chance of holding more than fifty percent equity in the assets outside the TDA is greater than half. As a second method of exploring asset location patterns, we estimate the correlation between asset holdings in the taxable and tax-deferred accounts. We estimate a two-equation system of "seemingly unrelated" regression models, with the same set of explanatory variables (X h ) for each. These equations are: (1a) EQUITY%TDA h = X h *β 1 + ε 1h and (1b) EQUITY%TAXABLE h = X h *β 2 + ε 2h. We then estimate the residuals from each of these equations, and compute the correlation between ε 1h and ε 2h. This correlation provides evidence on whether households that hold more of their taxable assets in equities also hold more of their tax-deferred assets in equities. By including the control variables in both (1a) and (1b), we remove the positive correlation between the asset allocation in both accounts that may arise from, for example, higher net worth households holding more equity in both accounts.

15 14 Table 13 reports our estimates of the correlation coefficients between ε 1h and ε 2h. The table shows results for both the broad definition of taxable and tax-deferred assets, and the narrow definition (upper panel) and broad (lower). The table applies to 1998 data. The coefficients are positive, and typically between 0.15 and 0.20 for our broad definition of assets, and 0.20 and 0.30 for our narrow definition. The table considers a range of different thresholds for total assets in TDA and non-tda setting, and it shows that the correlations between the asset allocations in the two locations are higher for households with relatively low levels of asset balances than for those with higher balances. The correlation typically declines as we include additional control variables. For the narrow definition of assets in the TDA and taxable account, and with no controls, the correlation is Before turning to asset location issues, we consider a final aspect of asset allocation. We investigate whether there are any identifiable characteristics of the households who hold only equity in their tax-deferred accounts. Table 14 presents estimates from discrete-choice probit models that explain the decision to hold only equity in the TDA. The explanatory variables are primarily indicator variables for households in different age, net worth, and income categories. The results suggest a positive association between net worth and the likelihood of holding only equity in the TDA, as well as a negative relationship between age and this probability. Households in the highest net worth category, those with more than $2.5 million in net worth, are the most likely to hold only equity in their TDA. This may partly be due to the stock market returns of the 1990s. Households that invested their TDA only in equities at the beginning of the decade were more likely to have accumulated substantial net worth by the end of the decade. The results in Table 14 suggest that the effect of net worth on the likelihood of all-equity TDA holdings is somewhat attenuated when we also control for household income. This is not surprising; the income and wealth variables are collinear. The probit models in the last column of Table 10 add several other variables that could affect household equity exposure in the TDA. In particular, households that report that they are prepared to take substantial risks for the prospect of substantial returns are more likely to hold only equity in the TDA, and married households are less likely to hold only equity than single persons.

16 15 Finally, we use a third approach to explore asset location decisions. We calculate the difference between the share of a household s total equity held in the TDA, and the share of it s total assets held in the TDA. This statistic is: (2) Equity TDA Bias = Equity TDA /Equity Total - Assets TDA /Assets Total We repeat a similar calculation for fixed income assets. This approach is similar to that taken by Barber and Odean (2001) in their analysis of asset location decisions of investors in two large data sets provided by brokerage firms. Table 15 presents the results of our analysis. When we use the narrow definition of both taxable and TDA assets, we find that for the median household with a TDA, the share of equity assets in the TDA is virtually identical to the share of all assets in the TDA. With the broader measures of TDA and non-tda assets, there is some evidence that the equity share is greater in the TDA than outside the TDA. For the median household with a TDA, and with net worth of at least $250,000, for example, the Equity TDA Bias statistic is zero when we use the narrow definition of assets, and it is 0.06 (six percent more equity invested in the TDA) when we use the broad definition. Even at the 25 th percentile of the distribution, there is a nine percent equity handicap for the TDA when we use narrow definition of taxable and TDA assets. There is no difference between the share of equity, and the share of assets, allocated to the TDA at this percentile when we study the broad measure of assets. These results suggest that substantially less than half of all households with TDAs are following a strategy of holding equity outside the TDA before holding it inside. The columns on the right side of Table 15 present analogous statistics for the share of taxable fixed income assets held in the tax deferred account: (3) Taxable Fixed-Income TDA Bias = Taxable Fixed-Income TDA /Taxable Fixed-Income Total Assets TDA /Assets Total The entries for the Equity TDA Bias and the Taxable Fixed Income TDA Bias do not need to add up to zero, so there is distinct information in the two measures. We find that households tend to hold less of their fixed income assets in their TDA than a simple calculation based on their share of assets in the TDA would

17 16 suggest. These results confirm the findings from the Equity TDA Bias calculation, in suggesting that households tend to favor equity over fixed-income investments in the TDA. 4. Conclusions This paper presents evidence on asset allocation decisions, and asset location decisions, for households with substantial balances in both their taxable and tax-deferred accounts. It shows that asset location is an important financial issue for a substantial group of U.S. households. More than eleven million households in 1998 had at least $25,000 in both taxable and tax-deferred accounts, and at least 3.4 million had more than $100,000 invested in each type of account. With respect to asset allocation decisions, there is a weak positive correlation between the fraction of the tax-deferred account that is allocated to stocks, and the fraction of the taxable account that is so allocated. With respect to asset location, we find clear evidence of a substantial group of households that holds most of its equity in the tax-deferred account. Whether this reflects lower transactions costs to holding equity through this channel, or greater exposure to information about equity investment through worker education programs associated with 401(k)s and related issues, is not clear. Our results suggest several directions for further investigation. We highlight two. The most important is to provide a clearer description of the importance of households with only equity in their taxdeferred accounts, and no equity outside the TDA, to provide guidance in understanding what drives their location decisions. Some research along these lines can be carried out using the Survey of Consumer Finances, but the lack of precise information on the way equities are held, mutual funds versus directly, is a limitation. Ideally, one would combine data on detailed holdings in tax-deferred accounts, such as the data studied by Agnew, Balduzzi, and Sunden (2000) or Barber and Odean (2001), with detailed holdings outside these accounts. It is not clear whether such data are available at this time. A second dimension for further investigation is the interplay between taxable and tax-exempt bonds in household asset location decisions. In each of the SCF data sets that we analyze, slightly more than six percent of all households report owning tax-exempt bonds. Only four percent of households in 1998 held

18 17 more than $10,000 in such bonds, and ownership of these bonds is highly concentrated among those with very high net worth. This suggests that the use of tax-exempt debt as a fixed income investment, outside the tax-deferred account, is not very common. Yet studies of the asset location problem, such as Shoven and Sialm (1999) and Poterba, Shoven, and Sialm (2000), suggest that holding tax-exempt debt can be an important element of long-term return maximization. Further exploration of the role of tax-exempt debt in household portfolios is therefore warranted.

19 18 REFERENCES Agnew, Julie, Pierluigi Balduzzi, and Annika Sunden (2000), "Portfolio Choice, Trading, and Returns in a Large 401(k) Plan," mimeo, Boston College Center for Retirement Research. Barber, Brad M. and Terrance Odean (2001), "Are Individual Investors Tax Savvy? Asset Location Evidence from Retail and Discount Brokerage Accounts." Mimeo, UC-Davis Graduate School of Management. Black, Fischer (1980). The Tax Consequences of Long-Run Pension Policy. Financial Analysts Journal (July-August), Bodie, Zvi and Dwight B. Crane (1997), "Personal Investing: Advice, Theory, and Evidence," Financial Analysts Journal 53, Number 6 (November/December), Charron, Terry S. (1999), "Tax-Efficient Investing for Tax-Deferred and Taxable Accounts," Journal of Private Portfolio Management (Fall), Copeland, Craig (2000), "Asset Allocation: IRAs and 401(k)-Type Plans," EBRI Notes (October), 5-9. Crain, Terry L. and Jeffrey R. Austin (1997), "An Analysis of the Tradeoff Between Tax Deferred Earnings in IRAs and Preferential Capital Gains," Financial Services Review 6 (4), Dammon, Robert, Chester Spatt, and Harold Zhang (2000). Optimal Asset Location and Allocation withtaxable and Tax-Deferred Investing, Mimeo, Carnegie-Mellon University. Kennickell, Arthur B., Martha Starr-McCluer, and Brian Surette "Changes in U.S. Family Finances at the End of the 1990s: Results from the 1998 Survey of Consumer Finances." Federal Reserve Bulletin (January). Poterba, James, John Shoven, and Clemens Sialm (2000). "Asset Location for Retirement Savers," NBER Working Paper. Shoven, John B. (1999). The Location and Allocation of Assets in Pension and Conventional Savings Accounts, NBER Working Paper Shoven, John B. and Clemens Sialm (1998), "Long Run Asset Allocation for Retirement Saving," Journal of Private Portfolio Management (Summer), Shoven, John B. and Clemens Sialm (2000). Asset Location in Tax-Deferred and Conventional Savings Accounts, Stanford mimeo. Tepper, Irwin (1981). Taxation and Corporate Pension Policy. Journal of Finance 36, 1-13.

20 19 Table 1: Percentage of Households with Tax-Deferred Accounts, Year Number of Percentage of Households with: Households in Tax Deferred Assets Financial Assets Tax-Exempt SCF Universe Outside TDA Bonds Outside (millions) Narrow Broad Narrow Broad TDA % 38.0% 45.6% 87.1% 6.5% Source: Tabulations from Surveys of Consumer Finances. See text for definitions of narrow and broad sets of assets in the taxable and tax-deferred accounts. Table 2: Households with Significant Holdings of Both Taxable and Tax-Deferred Financial Assets Amount in Amount of Financial Assets in Taxable Format Tax-Deferred Account > 0 10K 25K 50K 100K 1998, Narrow Definitions of TDA and Non-TDA Assets > K K K K , Broad Definitions of TDA and Non-TDA Assets > K K K K , Narrow Definitions of TDA and Non-TDA Assets > K K K K , Broad Definitions of TDA and Non-TDA Assets > K K K K Notes: Each entry shows the total number of households (in millions) with the specified mix of assets in tax-deferred and taxable accounts. In 1998 (1989), there were (93.0) million households in the Survey of Consumer Finances universe.

21 20 Table 3: Distribution of Share of Financial Assets Held in Tax Deferred Accounts, 1998 Net Worth Criterion Households with TDA Value of TDA/Total Financial Assets For Households with Both TDA and Non-TDA Assets and Non- TDA Assets (millions) 10th percentile 25th percentile Median 75th percentile 90th percentile Mean Panel A: Narrow Definitions of TDA and Non-TDA Assets All % 26.7% 55.6% % 55.2% Households NW $100K NW $250K NW $1000K Panel B: Broad Definitions of TDA and Non-TDA Assets All Households NW $100K NW $250K NW $1000K Source: Authors' tabulations using 1998 Survey of Consumer Finances. See text for further details.

22 21 Table 4: Characteristics of Households with Substantial TDA and Non-TDA Balances, 1998 Household All TDA, Non-TDA $25K TDA, Non-TDA $50K TDA, Non-TDA $100K Characteristic Narrow Broad Narrow Broad Narrow Broad Number of Households (millions) Average net worth $0.28 $1.32 $1.08 $1.80 $1.53 $2.58 $2.22 ($M) Median net worth $0.07 $0.52 $0.43 $0.72 $0.59 $1.18 $0.97 ($M) Average household $53.1 $156.1 $139.5 $204.2 $185.3 $265.1 $235.9 income (thousands) Median household $33.5 $82.3 $80.2 $101.6 $98.5 $124.9 $120.9 income (thousands) Mean age of household head Median age of household head Percent College Graduates Percent Willing to Take Above Average Risk for Above Average Rewards Source: Authors tabulations from the 1998 Survey of Consumer Finances.

23 22 Table 5: Percentage of Households Holding Equity, Year Number of Percentage of households with equity: Households in Anywhere Both inside and SCF Universe outside TDA (millions) Inside TDA, but not outside Panel A: Narrow TDA and Non-TDA Definitions % 6.0% 7.3% 14.0% Panel B: Broad TDA and Non-TDA Definitions % 7.8% 12.8% 12.1% Source: Tabulations from Surveys of Consumer Finances. Outside TDA, but not inside Table 6: Percentage of Households Holding Fixed-Income Assets, Year Number of Percentage of households with fixed-income assets: Households in Anywhere Both inside and Inside TDA, but SCF Universe outside TDA not outside (millions) Outside TDA, but not inside Panel A: Narrow TDA and Non-TDA Definitions % 14.3% 8.8% 25.6% Panel B: Broad TDA and Non-TDA Definitions % 30.2% 0.6% 57.0% Source: Tabulations from Surveys of Consumer Finances. Note that holdings of fixed-income assets outside the tax-deferred account could include tax-exempt bond holdings.

24 23 Table 7: Asset Combinations among SCF Households (millions of households), 1998 Households with No TDA Households with TDA Assets TDA: Equity Only TDA: Equity and Fixed Income TDA: Only Fixed Income Total Panel A: Narrow TDA and Non-TDA Definitions, All Households Households without non-tda assets Households with Some non-tda Assets Only Equity Outside TDA Equity and Fixed-Income Outside TDA Only Fixed Income Outside TDA Total Panel B: Broad TDA and Non-TDA Definitions, All Households Households without non-tda Assets Households with Some non-tda Assets Only Equity Outside TDA Equity and Fixed-income Outside TDA Only Fixedincome Outside TDA Total Panel C: Narrow TDA and Non-TDA Definitions, Households with 250K in Net Worth Households without non-tda Assets Households with Some non-tda Assets Only Equity Outside TDA Equity and Fixed-Income Outside TDA Only Fixed Income Outside TDA Total Source: Authors tabulations based on 1998 Survey of Consumer Finances.