Survey of Consumer Finances

The rapid growth of assets in self-directed tax-deferred retirement accounts has generated a new set of financial decisions for many households. In addition to deciding which assets to hold, households with substantial assets in both taxable and tax-deferred accounts must decide where to hold them. This paper uses data from the Survey of Consumer Finances to assess how many households have enough assets in both taxable and tax-deferred accounts to face significant asset location choices. It also investigates the asset location decisions these households make. In 1998, 45 percent of households had at least some assets in a tax-deferred account, and more than ten million households had at least $25,000 in both a taxable and a tax-deferred account. Many households hold equities in their tax-deferred accounts, but not in their taxable accounts, while also holding taxable bonds in their taxable accounts. Most of these households could reduce their taxes by relocating heavily-taxed fixed income assets to their tax-deferred account. Asset allocation inside and outside tax-deferred accounts is quite similar, with about seventy percent of assets in each location invested in equity securities. For nearly three quarters of the households that hold apparently tax-inefficient portfolios, a shift of less than $10,000 in financial assets can move their portfolio to a tax-efficient allocation. Asset location decisions within IRAs appear to be sensitive to marginal tax rates; we do not find evidence for such sensitivity in other tax-deferred accounts. We are grateful to Brad Barber, David Bradford, Joel Dickson, Roger Gordon, Andrew Samwick, and John Shoven for helpful conversations, to Amir Sufi for assistance with the Survey of Consumer Finances, and to the Hoover Institution, the National Institute of Aging, and the National Science Foundation for research support. Households have always faced the asset allocation problem, having to decide which assets to purchase and how much to invest in each of them. But with the recent growth of self-directed retirement plan assets, many households now also face an asset location problem. This is the question of how much of a given asset to hold in a taxable account, and how much of it to hold in a tax-deferred account. Assets in participant-directed tax-deferred accounts totaled nearly five trillion dollars at the end of 2001, with $2.4 trillion in Individual Retirement Accounts, and $2.3 trillion in 401(k)-type plans. At the end of 1990, by comparison, there were $637 billion in IRAs, and $735 billion in defined contribution plans. The recent growth of IRAs, 401(k)'s, and other self-directed tax-deferred retirement vehicles has drawn substantial interest to the investment decisions made by households with these accounts. Asset location has begun to attract attention from researchers in public finance and financial economics, and it is a frequent topic of discussion among financial planners. Shoven (1998) outlined the structure of the asset location problem, and observed that tax minimization would usually dictate holding heavily-taxed taxable bonds in the tax-deferred account, with less-heavily taxed equities in the taxable account. Recent work by Dammon, Spatt, and Zhang (2002), Huang (2001), Poterba, Shoven, and Sialm (2001), and Shoven and Sialm (forthcoming) has offered further insight on the optimal asset mix for households facing various tax and financial circumstances. Absent liquidity or other considerations, households should hold relatively heavily taxed assets in their tax-deferred account. Whether this implies that taxable bonds should be held in the tax-deferred account depends on the set of assets available to the household. For example, Shoven and Sialm (forthcoming) consider the asset location decision for investors who can only hold equities in the form of relatively tax inefficient vehicles, such as high-turnover actively-managed mutual funds. If these investors have access to tax-exempt bonds, then their optimal asset location may involve equity mutual funds in the tax-deferred account, and tax-exempt bonds in the taxable account. Most of the recent research on asset location has focused on the derivation of tax-minimizing portfolio strategies, rather than on the analysis of household portfolio choices. Three studies have presented empirical evidence on how households actually locate their assets. The first, Bodie and Crane (1997), is based on a survey of TIAA-CREF participants. It finds that investors choose similar asset allocations in their taxable and tax-deferred accounts, with little apparent regard for the benefits of taxefficient asset location. One open issue concerning this research concerns the extent to which the behavior ofTIAA-CREF participants can be generalized to the population at large. A second study, Barber and Odean (forthcoming), is based on data drawn from brokerage firm records. The data suggest that households hold equity mutual funds and taxable bonds in their taxdeferred accounts, while they hold individual equities in their taxable account. Because individual equity holdings tend to be less heavily taxed than bonds or equity mutual funds, this asset location pattern is broadly consistent with tax-minimizing behavior. However, households are more likely to trade stocks in their taxable than in their tax-deferred account, even though trading in the tax-deferred account would not generate current capital gains tax liability. A key concern with this study is the degree to which data on assets held through a single brokerage firm depict a household's broader balance sheet, and in particular whether households may have offsetting positions at other financial institutions. Finally, a third study, Amromin (2002), uses data from the Survey of Consumer Finances to investigate whether precautionary demands for financial assets, coupled with penalties and restrictions on withdrawing assets from tax-deferred accounts, can explain deviations from tax-efficient asset location patterns. The paper also provides summary information on tax-deferred account holdings. The findings suggest that the standard deviation of household labor income is related to asset location choices, with households in less risky occupations choosing more tax-efficient asset locations. This paper represents an important step toward building models of the factors that affect asset location choices. In this paper, we also use data from several Surveys of Consumer Finances (SCFs) to analyze asset location decisions. The SCF data provide complete and disaggregate data on the portfolios held by a large sample of households. The SCF asks households to aggregate their holdings across all financial intermediaries. This makes it possible to study the overall structure of the household portfolio, rather than just the structure of one of its components. This paper has two goals. The first is to describe the importance of the asset location problem, as measured by the number of households facing asset location decisions and the value of the assets held by households with such choices. The second is to explore asset location patterns, and to relate these patterns to household characteristics that affect the gains from tax-efficient asset location, particularly household marginal tax rates. The paper is divided into five sections. The first presents information on the number of households that face substantively important asset location decisions. We identify such households by the presence of significant asset holdings in both taxable and tax-deferred accounts (TDAs). Section two explores how households allocate their assets in taxable and in tax-deferred accounts. In the aggregate, equity investments make up more than two-thirds of tax-deferred financial assets and a similar proportion of taxable financial assets. The third section focuses on asset location decisions. It develops a simple classification rule to indicate whether or not households are making asset location decisions that are tax efficient. We develop several possible measures of tax efficiency, and we present estimates of the portfolio reallocation that would be needed to bring households in tax-inefficient positions to tax-efficient points. Section four presents cross-sectional regression and discrete choice evidence on the correlation between various household characteristics and asset location patterns. We investigate age, income, and net worth patterns in tax efficiency, and study how a household's marginal income tax rate affects the likelihood that its asset location choices are tax-efficient. A brief conclusion suggests directions for