The Impact of College Financial Aid Rules on Household Portfolio Choice

Transcription

1 The Impact of College Financial Aid Rules on Household Portfolio Choice Patryk Babiarz * and Tansel Yilmazer Abstract Households who save in anticipation of their child s college expenses reduce their child s eligibility for financial aid. The penalty of reduced financial aid eligibility acts as an implicit tax on household assets. However, retirement assets and home equity are excluded from computations of financial aid, thus are exempt from financial aid tax. Households can diminish the marginal financial aid tax rate by moving funds into retirement plans or by increasing their home equity. Using the 2001 Survey of Consumer Finances, our purpose is to investigate the effect of the college financial aid rules on household portfolio choice. Our results show that households who have higher marginal tax rates have higher retirement assets and home equity compared to taxable financial assets. However, the marginal financial aid tax rate does not have a significant adverse effect on the total amount of taxable financial assets. The results are robust across different model specifications. Keywords: College financial aid, portfolio choice, financial assets JEL Classifications: G11, D12, H31 This version: October 9, 2007 * Consumer Sciences & Retailing, Purdue University, 812 W. State Street, Purdue University, West Lafayette, IN , phone: , fax: , pbabiarz@purdue.edu. Corresponding author: Consumer Sciences & Retailing, Purdue University, 812 W. State Street, Purdue University, West Lafayette, IN , phone: , fax: , yilmazer@purdue.edu.

2 1. Introduction Households who fail to secure sufficient funds to support their child s college expenses can apply for financial aid from federal or state government. The U.S. Department of Education reports that over 4.2 billion dollars were spent on the need-based financial aid in the academic year and the need-based financial aid represented about 75 percent of all support dollars paid in the post-secondary education (Digest of Educational Statistics, 2005). A high proportion of U.S. students rely on the financial aid. Almost 62 percent of full-time undergraduate students and over 38 percent of part-time undergraduate students received some kind of financial support from federal government. Among the beneficiaries of the financial aid system, the average amount of the full-year federal financial support paid to full-time undergraduate student exceeded $7,300, and about $3,247 of this sum were non-returnable grants. Even among full-time students representing households with yearly income over $100,000, the average financial support from government amounted to $7,263, and $1,659 of this sum was in form of non-returnable grants. The irony of the need-based college financial aid system is that it imposes an implicit tax on household assets and income, and thus penalizes households who save in anticipation of their child s college attendance. 1 Consequently, the federal methodology for computing financial aid eligibility generates incentives for households to reduce savings or to wittingly adjust assets in order to increase their child s aid eligibility. The effect of the means-tested college financial aid on household asset accumulation has recently attracted much of the researchers attention (Dick & Edlin, 1997; Edlin, 1993; Feldstein, 1995; Long, 2004; Ma, 2005; Monks, 2004; Wolpaw 1 Nevertheless, the stylized facts show that saving for children s education is an important motive for households to save. According to the 2001 Survey of Consumer Finances, saving for children s college education is the third most important saving motive proceeded by saving for retirement and emergencies. 1

3 Reyes, 2007). Our study verifies the concerns of policy makers. Methodology for computing financial aid eligibility cancels out the incentives to save and distorts households composition of assets. The sideline effect is that households who shelter assets in order to avoid the implicit tax improve their retirement assets. Fair distribution of student aid however, requires that the assessment of needs is not affected by households sheltering of taxable assets. The increased financial aid eligibility resulting from asset adjustments, whether caused by households purposeful behavior or financial aid system inefficiencies, deserves more consideration. In spite of the extensive research efforts to calculate the implicit financial aid tax and measure its negative impact on household asset accumulation, little attention has been given to the portfolio choices of households who expect to have a child attending college in the near future. The federal algorithm used to compute the aid eligibility does not take into consideration assets accumulated in retirement accounts or as home equity, which makes these amounts exempt from implicit financial aid taxation. Households may mitigate the nuisance of taxation by altering their portfolio of assets and moving assets into retirement accounts or increasing home equity. Intensifying contributions to retirement accounts or reducing mortgage debt are among the frequent advice of the web resources on college financial aid. For example, maximize contributions to your retirement fund or prepay your mortgage are two of the several tips offered by FinAid website (Maximizing Your Aid Eligibility, n.d.). CNNMoney.com also recommends that in order to maximize student s financial aid eligibility parents should maximize retirement savings and watch their debt load (Seid, 2005). The purpose of this study is to investigate the effect of the financial aid on household portfolio choice. This research contributes to the literature by examining whether households who prepare for their child s college education adjust assets in order to increase their child s 2

4 eligibility for college financial aid. We find substantial evidence that such an adjustment of both asset categories considered in this study takes place. Retirement assets or home equity components of asset portfolio increase significantly with marginal tax rate. At the same time, we find that the amounts of taxable financial assets are not significantly influenced by marginal tax rates. The reminder of the paper is structured as follows. Section 2 reviews the literature on the portfolio decisions in the context of taxation and discuses the previous evidence of the impact of the implicit financial aid tax on household assets. Section 3 provides a detailed explanation of the marginal tax rate computations. Data and variables employed in this study are explained in Section 4. Section 5 provides estimates of the impact of taxation on portfolio decisions, as well as other related results. Discussions of findings, as well as an outline of study limitations, are presented in section Literature review 2.1 Taxation and portfolio decisions In spite of the substantial amount of theoretical research on how differential taxation influences household portfolio choices, relatively few empirical studies exist to verify theoretical predictions. The existing body of research usually focuses on burdens imposed by the federal or state tax on different types of income or capital gains, and the resulting portfolio decisions of households (Feldstein, 1976; Hubbard, 1985; Scholz, 1994; King & Leape, 1998; Maki, 1996; Poterba & Samwick, 1999; Poterba, 2001). Economic policy makers and researchers acknowledge that taxation usually creates substantial variation in the relative tax burdens on 3

5 different categories of assets experienced by different investors. For example, investing or saving through income tax-deferred accounts is one of the ways in which households, who face otherwise high tax rates, may reduce the tax burden on their capital gains. Families usually not only have a choice whether to hold a particular asset category, e.g. stocks or bonds, but also whether to hold these assets in a taxable account, in a tax-deferred account or in some type of tax-exempt account if such exist. For example, returns to stock market investments in taxdeferred retirement accounts are taxed differently than the same type of returns of investments held in a brokerage accounts. In some cases, appropriate portfolio decisions may allow households to shelter some proportion of income and avoid certain taxes. This results in an expanded set of investment and saving vehicles for families. Previous literature provides evidence of the significant relationship between taxation and portfolio decisions of households. Using data from the 1962 Survey of Financial Characteristics of Consumers, Feldstein (1976) documented that the personal income tax had a significant effect on individuals demand for portfolio assets. Based on the finding that high income, heavily taxed households were more likely to hold stocks, Feldstein s (1976) concluded that asset composition, at least indirectly, depends on tax policy. Using U.S. President s Commission on Pension Policy data, Hubbard (1985) estimated the impact of income tax rates and household participation in the social security and private pension systems on the wealth accumulation and composition. Findings in Hubbard (1985) showed that both participation in social security or private pension system, and marginal tax rate on income have a significant and measurable impact on households asset allocation decisions. Poterba and Samwick (1999) investigated whether families who face higher income tax rates invest more through tax-deferred accounts. Using data from the Survey of Consumer 4

6 Finances, Poterba and Samwick (1999) concluded that households income tax rates display a substantial correlation both with ownership of assets and with the share of the household s portfolio that is allocated to various asset categories. Households with higher marginal tax rates were more likely to own tax-advantaged assets such as publicly traded stocks and tax-exempt bonds than were analogous households but with lower marginal tax rates. Heavier taxed households were also more likely to hold assets in tax-deferred accounts such as IRAs, Keoghs, and defined contribution pension plans. Other recent studies provide less unambiguous results in respect to relationships between taxation policy and households portfolio decisions. 2 Scholz (1994) examined households assets structure over time and the potential role of taxation in portfolio decisions. Although he found that relatively small changes in portfolio structure occur in response to taxation, there was evidence that households tend to favor mortgages over other types of debt. Findings in respect to mortgages were further verified and confirmed by Maki (1996). King and Leape (1998) found that taxes affect the choice of assets that investors decide to hold, even though the link between tax rates and the percentage composition of different assets was insignificant. The conceptual model in our study relies on the assumption that the implicit financial aid tax will have a similar effect on households portfolio decisions as the income tax. If households perceive the penalty on savings represented by reduced financial aid eligibility as a form of taxation, by analogy to the households practice of asset adjustments to reduce income taxation, we expect families to shelter their assets in order to prevent the decline in aid eligibility. Specifically, we expect that families who will have children in college respond to the marginal 2 For a more detailed description of the research stream on taxation policy and its impact on households portfolio decisions, see Poterba and Samwick (1999). 5

7 tax rates by sheltering their taxable assets in non-taxable categories such as retirement assets or home equity. 2.2 Impact of college financial aid on household assets Financial need for each college student is assessed by the difference between the cost of college attendance and the student s Expected Family Contribution (EFC). EFC is the amount that a student applying for financial aid and her family are expected to contribute towards educational expenses over a year. The U.S. Department of Education calculates a student's EFC using federal methodology based on information provided on the Free Application for Federal Student Aid (FAFSA). The larger the EFC, the smaller the student s assessed financial need and the amount of aid that the student is eligible to obtain. 3 The concern of policy makers and researchers is that various elements of financial aid distribution can have adverse effects on asset accumulation. The early studies found a direct relationship between the implicit financial aid tax and reductions in household savings (Dick & Edlin, 1997; Edlin, 1993; Feldstein, 1995). Those studies report that families respond to this marginal tax rate by lowering their assets. The marginal tax rate on assets is defined in the literature as "the present value of the expected change in the value of financial aid received by all family members resulting from a one-dollar increase in the parents current assets" (Long, 2004). Initial studies in the area of impact of college financial aid on household s savings estimated a very large effect of the implicit financial aid tax (Dick & Edlin, 1997; Edlin, 1993; 3 In this study, we are interested in the marginal tax rates on parental assets. Assets held in the student s name reduce the financial need and aid eligibility to a much greater extent than assets held in a parent s name. In calculating a dependent student s EFC, up to 5.64 percent of parents assets and 35 percent of the student s assets are considered (The EFC Formula Guide, 2000). We ignore student's assets due to data unavailability. We believe however, that this omission does not undermine our results. There is evidence that students liquidate their assets prior to applying for financial aid due to relatively large differences in tax rates between parental and student assets (Ma, 2005). 6

8 Feldstein, 1995). Even though those studies did not directly focus on whether the reduction in the financial support influences savings, each of those studies found that for some families the financial aid tax may be reducing asset holdings by as much as 50 percent. Edlin (1993) was the first to illustrate that although the maximum financial aid tax on parental assets is only 5.64 percent in any given year, even families with only one child in college at a time for consecutive 12 years may face a cumulative tax on assets of over 57 percent. In a follow up study, Dick and Edlin (1997) indicated that a typical family with student at an average priced college may lose over $2,000 in financial aid as the result of increasing their assets by $10,000. Feldstein (1995) provided additional evidence that the implicit financial aid taxes on assets have a significant adverse effect on the accumulation of financial wealth. Using data from 1986 SCF and the predicted values of household assets in evaluating families financial aid tax, Feldstein (1995) concluded that over the entire period of child s college education the average family assets would be taxed at an overall rate approaching 50 percent. The magnitude of the effect of marginal tax rates on parental assets estimated by previous literature was questioned by Long (2004) and by Monks (2004). Long (2004) tested the influence of financial aid tax on parental assets relaxing several computational assumptions that were made in previous studies and incorporating uncertainty about probability of college attendance, attendance costs, expected contributions from family, future household income, student s selfcontributions, and family assets. The results in Long (2004) showed that the effect of financial aid tax on parental assets is sensitive to underlying estimation assumptions and the magnitude of the effect varied substantially across models with distinct sets of assumptions. Monks (2004) replicated the research of Edlin (1993) and Feldstein (1995) using more recent data collected by the 1997 National Longitudinal Survey of Youth. The magnitude of effects of the implicit 7

9 financial aid tax on asset accumulation was significantly smaller than in both previous studies. Similarly to Long s (2004) results, Monks (2004) concluded that the findings in respect to the magnitude of tax rate impact on assets were highly dependent and considerably different across estimations with distinct set of assumptions. Finally, in the most recent study, Wolpaw Reyes (2007) used the 1999, 2001, and 2003 data from Panel Study of Income Dynamics and estimated that the typical family with two children saves 7 to 12 percent less than they would in the absence of the implicit college aid taxation. In addition, Wolpaw Reyes (2007) showed the positive correlation between the retirement assets and the income multiplied by marginal tax rate on assets. In this study, we argue that the effect of implicit financial aid tax on assets may be more complex than just a simple change in total savings. Some categories of assets are exempt from taxation and thus represent strong incentives for families to adjust their portfolio in order to make their children more eligible to receive financial support. To date there has been few attempts to explain the impact of marginal tax rate on households assets incorporating the portfolio decisions into the empirical model. This study will contribute to the discussion by providing an explanation of households portfolio choice in response to the implicit financial aid tax. Specifically, the study determines if households shift their assets towards tax preferred asset categories challenging the intentions of policy of financial aid allocation. Two assets categories considered in this study, retirement assets and home equity, are non-taxable under the federal methodology computing financial aid eligibility. Note that two terms frequently used in this paper, the implicit financial aid tax and the marginal tax rate are semantically related. They both imply the penalty of reduced college 8

10 financial aid eligibility in consequence of the increase in household assets. While implicit financial aid tax refers to the phenomenon resulting from aid distribution policies, the marginal tax rate denotes the quantitative evaluation of the implicit taxation resulting from $1 increase in household assets. 3. Marginal tax rate The implicit college financial aid tax on assets results from the cutback of governmental financial aid for households that increase their holdings of taxable assets. The marginal tax rate (MTR) is a hypothetical measure of the annual tax imposed on additional dollar of parental assets. We calculate MTR as the difference of the expected family contribution (EFC) when assets are incremented by $1, minus the EFC computed using original value of assets. The federal methodology is the main methodology for assessing the eligibility for the need-based financial aid from government, and is the methodology adopted for the purposes of our study. Under federal methodology, 12 percent of parents financial assets above certain allowances are summed with the household s available income. This sum constitutes the taxable parental assets. Together with contribution from income, the taxable assets sum up to the adjusted available income (AAI). The amount of AAI, which does not include contributions from retirement assets or home equity, is then used to compute the EFC. In these computations, a certain base amount is summed with AAI multiplied by a discrete factor depending on the level of AAI. It is important to notice that families in the lowest bracket of the assets categories may end up with zero EFC, which means that they do not face any financial aid tax. Families in the upper bracket of those asset categories are expected to contribute $0.47 of each additional dollar 9

11 of assets used to compute the EFC above the allowances. However, since 12 percent of the amount representing certain assets is included in the final amount used to compute the EFC, the maximum marginal tax rate that a household can face in any given year of their child s college attendance is 0.12 times 0.47, which results in the tax rate of 5.64 percent. As noticed by previous studies (Edlin, 1993; Feldstein, 1995; Dick and Edlin, 1997; Long, 2004; Monks, 2004), this implicit tax is reapplied every year when the student is enrolled in college. Thus, the cumulative, multiple-years effect of tax can be much larger than the maximum 5.64 percent annual rate. 4 Further overview of MTR computations and the summary of the federal methodology are explained in Appendix A. 4. Data The empirical analysis uses data from the 2001 Survey of Consumer Finances (SCF) conducted by the Federal Reserve Board. The SCF is a comprehensive, triennial survey of U.S. household finances, containing detailed information on household assets, liabilities, income and demographic characteristics. The 2001 SCF is the most recent cross section of the survey that identifies the age of the children living in the household. The survey interviewed 4,449 households in The SCF uses multiple techniques of imputations to compensate for the missing data. This procedure results in five replicates of data for all households. This study used all five replicates and procedures described by Montalto and Sung (1996) to correct coefficient estimates, standard errors and significance levels of the estimated parameters. 4 The time frame of the tax burden under consideration is the main factor that distinguishes the MTR in this study from tax rates analyzed in previous literature. While Edlin (1993), Feldstein (1995), Dick and Edlin (1997), Long (2004), and Monks (2004) focused on the estimation of the cumulative, 4 years effect of the marginal tax rate, this study is primarily concerned with computing the annual evaluation of marginal tax rate to determine its impact on portfolio choices of households. 10

12 Table 1 lists the definitions of variables used in the analysis. Two financial ratios are used to account for portfolio choices of households. RETIRE_RATIO represents the ratio of the retirement assets to the amount of the remaining financial assets. Retirement assets include IRAs, Keogh and any other retirement accounts from current employer such as thrifts, savings, 401K, 403B, SRA, etc. HOMEEQ_RATIO is the ratio of home equity to financial assets, where home equity is defined as the market value of home minus the outstanding mortgage debt. Financial assets in the denominators of both ratios are taxable as they do not include retirement assets. Several restrictions are imposed on the sample. Single households headed by retired individuals and married or cohabiting households with a retired spouse or partner were deleted from the sample. These households may have already started to use their retirement assets. Additionally, outliers above the 95 th percentile in the distribution of RETIRE_RATIO and HOMEEQ_RATIO, households that reported negative income and households with total financial assets below $100 were also excluded from the analysis. The final sample comprises of 2,821 households. As noticed by Long (2004), it is problematic to use the families actual observed assets in computing the EFC. These assets may have already been affected by the implicit financial aid tax. To prevent this kind of bias in estimating MTR, we first predict households taxable assets based on families characteristics, such as age, race, education and the interaction of these covariates. The estimates of the model are presented in Appendix C. We then use the predicted values of taxable assets to compute the EFC and MTR. It is also important to notice that federal methodology for computing the EFC allows some families to undergo the Simplified Needs Test. According to this procedure, if the household has less than $50,000 in adjusted gross income, and is not required to file tax return or is eligible to file forms 1040A or 1040EZ, the household 11

13 is automatically released from the contribution from assets. Our dataset does not allow for identification of households that are eligible for the Simplified Needs Test. However, introducing some additional assumptions we verify the robustness of our analysis incorporating the Simplified Needs Test. 5 Table 2 provides insight into financial characteristics of households by the estimated value of MTR. Almost 45 percent of families face zero marginal tax rates. For those households, an extra dollar of assets does not reduce financial aid from government. On the other hand, over 35 percent of households face the MTR of 5.64 percent - the highest possible value under federal methodology. The median of EFC in this group is $18, For some households in the highest MTR category the EFC would most likely exceed their child s college attendance costs making these households ineligible for financial aid. It is important to notice that if such situation takes place, the household would face zero MTR simply because it is not using financial aid. In general, a trend of increase emerges in average income, financial assets, retirement assets and home equity as the MTR increases. Medians of financial measures also appear to increase by MRT. The means of RETIRE_RATIO and HOMEEQ_RATIO display an inverse U shape pattern. Ratios computed from group specific means for particular tax cells have the highest value when MTR equals 3.48 percent and 3 percent, respectively for RETIRE_RATIO and HOMEEQ_RATIO. In addition, information about the distribution of MTR across our sample can be read from Table 3. This table presents the mean MTR and percentage frequencies in each tax cell by different levels of adjusted gross income and taxable assets SCF does not provide information what tax return form was filed by particular households. We qualify households to the Simplified Needs Test if the adjusted gross income of household does not exceed $50,000, total income does not exceed $100,000 and household does not report itemizing deductions on the tax report. 6 For brevity we do not report financial characteristics by percentages of MTR when we incorporate the Simplified Needs Test into MTR computations. The inclusion of Simplified Needs Test results in almost 63 percent of the households facing the zero MTR. Although this particular group becomes larger at the expense of all other MTR levels, the financial characteristics of households by the levels of MTR remain largely unaffected. 12

14 Summary statistics of variables used in the study are reported in Appendix B. The 2001 SCF identifies households that expect major financial expenditures on children education within the next 5-10 years (DCHCOLL), and whether households are saving for these obligations (SAVE). We assume that the expectation of major educational expenditures implies child's college attendance. First two columns in Appendix B compare households that expect major educational expenditures within the next 5 years (DCHCOLL=1) to those who do not (DCHCOLL=0). Columns 3 and 4 provide analogous comparison of families that save for future college expenditures (SAVE=1) to those who do not (SAVE=0). Households that expect to experience major expenditures on their child s education have on average significantly higher income compared to households who do not expect such expenses. Average amounts of assets kept as in retirement accounts and home equity are also higher for households who will incur educational expenses. RETIRE_RATIO and HOMEQ_RATIO for families who anticipate educational expenditures are 1.33 and 3.41, respectively, compared to 1.00 and 2.82 for families who do not anticipate their child's college attendance. Households who report saving for child s education expenses on average have higher income, retirement assets and home equity than households who do not save. RETIRE_RATIO of families who report saving is 0.40 compared to 0.19 of households who do not save. MTR is significantly higher for households saving for educational expenditures. Households who save have to contribute with an average of 3.50 percent of an additional dollar of their asset, while households who do not save contribute 2.74 percent of an additional dollar of assets. 13

15 5. Results 5.1 Retirement assets The main analysis of the impact of MTR on households portfolio choices is documented in the results of estimations in Tables 4-7. Table 4 reports tobit model estimates of RETIRE_RATIO regressions for the full sample and for varying sample partitions. In a base model including all observations, MTR is in a significant and positive relationship with RETIRE_RATIO. A one percentage point increase in the MTR results in 0.12 increase in the value of RETIRE_RATIO, on average. The magnitude of the effect becomes even larger when the sample is reduced to households with a child or households with child younger than 18. When the sample is restricted to those households with the predicted value of EFC amounting to less than $9,196, the relationship between MTR and RETIRE_RATIO is still significant. These households are more likely to be eligible for financial aid than household in other estimations because their EFC amounts to less than an average cost of public college attendance with 4-years enrollment. 7 The sample is also restricted to households with annual income amounting to over $100,000. About 92 percent of these high income families experience the maximum value of MTR (5.64 percent) and the average MTR for these families equals to 5.23 percent. In this estimation, a one percentage point increase in MTR produces an average 0.16 increase in the RETIRE_RATIO. The impact of MTR on RETIRE_RATIO is particularly strong for high income families. This finding may indicate that independently of the effect of MTR these high income families are more likely than other households to contribute to their retirement accounts 7 The $9,196 average cost of public college attendance with 4-years enrollment is the estimate of the U.S. Department of Education in academic year (Digest of Education Statistics, 2003). 14

16 with the maximum amount allowed by law. However, the strong and positive correlation between MTR and RETIREMENT_RATIO also indicates that by contributing to retirement accounts with a higher proportion of taxable assets the high income families increase their aid eligibility even more effectively. Except for the model where predicted EFC amounts to less than $9,196, income has a negative impact on the level of RETIRE_RATIO. Among households with predicted EFC less than $9,196, the positive effect of income on amount of retirement assets is stronger than the effect of income on taxable financial assets. This finding indicates that investments in retirement savings increase faster than investments in taxable financial assets for low-income and low-asset households. To further verify the validity of our findings on the relationship between MTR and RETIRE_RATIO, we estimate models with additional varying restrictions on the sample. The estimates are presented in Table 5. A common restriction in all those estimations is the one that that reduces the original dataset to households that have retirement assets. About 68 percent of families included in the original sample report positive retirement assets. In the model additionally restricted to households that a have a child, or households whose child is younger than 18, the magnitude of the effects of marginal tax rate is similar to the original models. A onepercentage point increase in the MTR produces an average 0.15 increase in the value of RETIRE_RATIO. Next two models restrict the sample to households who make direct contributions to their retirement accounts. The extent of the effect of MTR on RETIRE_RATIO for those households is higher than in any of the previous models. Households that make selfcontributions to their retirement plans may find the retirement savings particularly attractive form of investment when they prepare for college expenditures. This effect holds even for 15

17 households with estimated EFC below the average cost of college, i.e. these households that are most likely to be eligible for financial aid. 5.2 Home equity The regression results for HOMEEQ_RATIO are summarized in Table 6. Similarly to RETIRE_RATIO regressions, MTR has a positive and significant effect on HOMEEQ_RATIO. In the base model, a one percentage point increase in the MTR encourages households to increase the value of HOMEEQ_RATIO by 0.18, on average. The magnitude of the impact of MTR on HOMEEQ_RATIO is slightly higher in models limited to households with children or households where estimated EFC amounts to less than $9,196. Only in the model with families having children younger than 18 the effect of MTR is not significant. A possible explanation is that a limited sample size rules out the statistical significance of the MTR coefficient. Similar to findings in RETIRE_RATIO regressions, income is negatively related to HOMEEQ_RATIO for all samples except for households most likely to receive the financial aid. 5.3 Financial assets We next investigate whether taxable financial asset decrease with MTR. The estimation results are reported in Table 7. The most important finding is that controlling for expectancy of college expenditures and saving behavior relating to these expectations MTR does not significantly affect the amount of financial assets. DCHCOLL however significantly reduces the amount of taxable financial assets, while SAVE has an opposing effect. In other words, 16

18 households who expect to have a child in college, yet do not report saving for this purpose, on average have 34 to 49 percent less in taxable financial assets than households without such expectations. 8 Among households who expect to send a child to college, families who report saving for this purpose have on average 305 percent more in taxable financial assets than families who do not save. This estimate is even higher in models that limit the sample to households with children or children younger than 18. To make this interpretation more meaningful, consider the example of household from the median point of distribution of taxable assets in our sample. Such household who does not expect a child in college has taxable assets equal to $15,100. According to the estimate, if this household did want to send a child to college, yet was not saving for this purpose, the taxable assets would amount to $8,789, which corresponds well to the median of this group equal to $8,000. Furthermore, if this household was saving for college expenses, the taxable assets would amount to $32,388, which is slightly overestimated compared to the median taxable assets of savers. 9 Since MTR affects both RETIRE_RATIO and HOMEEQ_RATIO in settings that control for DCHCOLL and SAVE, yet MTR remains insignificant predictor of taxable financial assets, we conclude that the implicit college aid tax has an effect on household assets composition and that households may respond to the implicit taxation of assets by sheltering some of the savings. In other words, there is some evidence of asset sheltering that minimizes the taxation of college financial aid, even though the total level of non tax-exempt assets remains unaffected. At the same time it is important to notice that in most settings of RETIRE_RATIO and HOMEEQ_RATIO regressions the portfolio decisions neither depend on the expectation of 8 In the semi-logarithmic equation, the percentage effect of coefficient by the dummy variable on the dependent variable is estimated as 100*{exp(b)-1}, where b is the coefficient estimate by the dummy variable. 9 In reality, households whose EFC exceeds cost of college do not receive financial aid and have the MTR=0. Due to the data unavailability however, we do not control for the cost of college and those households are most likely assigned the maximum possible MTR. This causes the upward bias of our parameter estimates. 17

19 college attendance of a child nor on the saving behavior resulting from such expectation. DCHCOLL displays a significant relationship with HOMEEQ_RATIO only in the regression with the full sample, and SAVE display significant effects on HOMEEQ_RATIO only in the regression with the full sample, or with INCOME greater than $100,000. Combined with the fact that the magnitude of SAVE effect is much higher in the high income sample than in the full sample, this implies that DCHCOLL and SAVE covariate with HOMEEQ_RATIO only when large number of households who are not eligible for financial aid are included in the sample. Since these households do not have as strong incentives to adjust their assets, these results may indicate that for some of the high income households saving is the only reasonable and legitimate way of preparing for child s college expenditures. To rule out other possible explanations of the effects of MTR on households portfolio decisions additional robustness checks were applied. We investigated whether MTR affects household decision to save for children s college education. The SAVE dummy was regressed on MTR and other control variables included in previous regressions. Our results show that MTR does not affect the probability of saving for child s college expenditures. Tables 8 reports the results of two probit estimations with different sets of control variables. In both models, the effects of MTR on decision to save remain insignificantly different from zero. The lack of the significant impact of the implicit financial aid tax on the total amount of taxable financial assets is consistent with the results of Long (2004) and Monks (2004), and disagrees with the effects documented by Dick and Edlin (1997), Edlin (1993) and Feldstein (1995). It is not true, however, that the marginal tax rates does not affect the ways how people save. Even though it may seem possible that, as argued by Long (2004), exemptions in the federal policies of aid distribution may have eliminated the saving disincentive for many 18

20 families, the same exemptions seem to have encouraged families into portfolio shifting behavior. As noted by Monks (2004), the arcane calculations of aid eligibility may bound rationality of households behavior. However, given that over 60 percent of full-time undergraduate students receive state funded financial aid, it does not seem entirely plausible that families are not responsive to tax rates. The most likely scenario and the one that fits the results of previous studies is that the impact of MTR on household savings evolved from the simple effect of the increase in total assets to the effect of portfolio adjustments. Exclusion of the home equity from consideration in the federal financing aid formula that occurred in the 1992 seems to support this finding. 6. Conclusions Several important findings emerge in the analysis of the impact of the implicit financial aid tax on portfolio choice of U.S. households. Consistent with findings of Long (2004) and Monks (2004), this research fails to document any significant evidence of the relationship between marginal tax rates and the level of household savings measured as taxable financial assets. At the same time, households who report that they expect major financial expenditures on child s education and also admit to saving for these obligations, have taxable financial assets higher by almost 50 percent, compared to non-saving families. While our study fails to document the effect of the aid policy on household taxable assets, we find that the effect of financial aid tax on savings may be more complex than just a simple decrease in total savings. While we do not control for how much of household behavior is incentive driven, categories of assets that are exempt from taxation are significantly correlated 19

21 with marginal tax rates. The analysis of portfolio choices with respect to retirement assets and home equity conducted for different samples of households indicates that maximizing retirement assets and prepaying home equity may help households to escape the implicit financial aid taxation. Findings of this paper validate the concerns of federal or state policy makers. It seems that the federal methodology for computing financial aid eligibility is at odds with incentives offered to households to maintain a high level of savings. In order for the distribution of needbased financial aid to be fair, the assessment of needs should not be affected by households abilities to deliberately or inadvertently conceal taxable assets. The present system for distributing federal aid dollars generates high costs. Competing methodologies of distributing financial aid, for example computing need based aid eligibility based on the information from the income tax returns were previously argued to be almost as effective as federal methodology, yet much cheaper to implement and maintain (Dynarski & Scott-Clayton, 2006). The results of this study constitute yet another argument that the present system, at least partly, evaded the intentions of policy makers to fairly distribute the need based student aid. There are some limitations of the research that should be addressed. Dick and Edlin (1997) showed that one dollar increase in EFC does not always mean one dollar decrease in financial aid. The reason for this is that student s financial support often does not include enough aid funds to meet all the students needs. For these students families, an increase in EFC has no effect on the financial aid, and they can increase their assets at no penalty. In our dataset however, we do not have the information on the aid awards. Also, the dataset used in this paper does not allow for unambiguous identification of college attendance, and college costs. Since we do not have data on college cost, we do not identify households whose EFC exceeds college 20

22 costs of their children. Those households would not be eligible for financial aid and their MTR would be zero. Most likely, such families end up with maximum MTR in our sample, causing some bias to the estimates of the effects of implicit tax. This paper ignores several other elements of the federal algorithm for computing aid eligibility, e.g. state income allowances, number of family members enrolled in college, etc. Other significant reductions of the actual financial aid may also result from student s own contributions; however SCF does not allow controlling for student s contributions. Given the availability of the data, further studies might address these issues and estimate the impact of aid taxation more precisely. To build the holistic picture of the role of implicit financial aid taxation, the institutional methodology for computing aid eligibility from private sources should be embraced in the framework of further research. 21

23 References Browning, M., & Lusardi, A. (1996). Household saving: Micro theories and micro facts. Journal of Economic Literature, 34 (4), Choy, S. P., & Henke, R. R. (1992). Parental financial support for undergraduate education. Washington: National Center for Education Statistics U.S. Department of Education. Dick, A. W., & Edlin, A. S. (1997). The implicit taxes from college financial aid. Journal of Public Economics, 65 (3), Digest of Education Statistics (2006). Washington, DC: Institute of Education Sciences, U.S. Department of Education. Digest of Education Statistics (2003). Washington, DC: Institute of Education Sciences, U.S. Department of Education. Dynarski, S. M., & Scott-Clayton, J. E. (2006). The cost of complexity in federal student aid: Lessons from optimal tax theory and behavioral economics. National Tax Journal, 59 (2), Edlin, A. S. (1993). Is College Fiancial Aid Equitable and Effcient? Journal of Economic Perspectives, 7 (2), FinAid Website. (n.d.), Maximizing Your Aid Eligibility. Retrieved July 12, 2007 from Feldstein, M. (1995). College scholarship rules and private saving. American Economic Review, 85 (3), Feldstein, M. S. (1976). Personal taxation and portfolio composition: An econometric analysis. Econometrica, 44, Gale, W. G., & Scholz, J. K. (1994). Intergenerational transfers and the accumulation of wealth. Journal of Economic Perspectives, 8 (4), Horioka, C. Y., & Watanabe, W. (1997). Why do people save? A micro-analysis of motives for household saving in Japan. Economic Journal, 107 (442), Hubbard, G. R. (1985). Personal taxation, pension wealth, and portfolio composition. Review of Economics and Statistics, 67, King, M. A., & Leape, J. I. (1998). Wealth and portfolio composition: Theory and evidence. Journal of Public Economics, 69,

24 Long, M. (2004). The impact of asset-tested college financial aid on household savings. Journal of Public Economics, 88 (1), Ma, J. (2005). College savings options and the impact of savings on financial aid. Research Dialogue (83), Maki, D. M. (1996). Portfolio shuffling and tax reform. National Tax Journal, 49, Monks, J. (2004). An empirical examination of the impact of college fiancial aid on family savings. National Tax Journal, 57 (2), Montalto, C. P., & Sung, J. (1996). Multiple imputation in the 1992 Survey of Consumer Finances. Financial Counseling and Planning, 7, Poterba, M. J., & Samwick, A. A. (1999). Taxation and household portfolio composition: U.S. evidence from the 1980s and 1990s. NBER WORKING PAPER SERIES. Poterba, M. J. (2001). Taxation and portfolio structure: Issues and implications. NBER WORKING PAPER SERIES. Presley, J. B., & Clery, S. B. (2001). Middle income undergraduates: Where they enroll and how they pay for their education. Washington: National Center for Education Statistics, U.S. Department of Education. Seid, J. (2005). Maximize your financial aid chances. Retrieved July 30, 2007 from Scholz, J. K. (1994). Portfolio choice and tax progressivity: evidence from the Surveys of Consumer Finances. In J. Slemrod, Tax Progressivity and Income Inequality. New York: Cambridge University Press. Souleles, N. S. (2000). College tuition and household savings and consumption. Journal of Public Economics, 77, The EFC Formula Guide (2000, May 30). Federal Register. Wolpaw Reyes, J. (2007). College Financial Aid Rules and the Allocation of Savings. Educational Economics, forthcoming. 23

25 Table 1. Variables. Variable Description DCHCOLL =1 if household expects major financial expenses for child's education within 5-10 years; =0 otherwise SAVE =1 if household is saving for those expenses; =0 otherwise MTR Implicit marginal tax rate on parental assets AGE Age of the household head MARRIED =1 if the household head is married; =0 otherwise BLACK =1 if the household is African-American; =0 otherwise HISPANIC =1 if the household is Hispanic; =0 otherwise FEMALE =1 if the household head is a single female; =0 otherwise CHILD Number of children of either the respondent or spouse HOMEOWNER =1 if household owns a home; =0 otherwise POSTGRAD =1 if the household head has a post college degree; =0 otherwise COLLEGE =1 if the household head has a college degree; =0 otherwise SOMECOLLEGE =1 if the household head has some college education; =0 otherwise HIGHSCHOOL =1 if the household head has high school degree; =0 otherwise NOHIGHSCHOOL =1 if the household head has high school degree; =0 otherwise SELFEMPLOYED =1 if the household head is self-employed; =0 otherwise MANAGERIAL =1 if the household head holds managerial or professional specialty occupation; =0 otherwise BEDUCAT =1 if the household head feels it is all right to borrow to finance educational expenses; =0 otherwise BLIVING =1 if the household head feels it is all right to borrow to cover living expenses; =0 otherwise PLANNER =1 if the household head plans saving and spending at least 5 years in advance; =0 otherwise RISKY =1 if the household head takes above average financial investments risk; =0 otherwise INCOME Household income in 2000 FINASSETNORET Household financial assets in 2001 (includes liquid assets such as checking, saving, money market or call accounts, certificates of deposit, stocks, bonds, mutual funds, annuities, cash value of whole life insurance, etc., and excluding retirement assets) RETIREMENT Households retirement assets in 2001 (includes IRA, Keogh, and retirement accounts from current employer such as: thrifts, savings, 401K, 403B, SRA, or any other retirement account at the current employer) RETIRE_RATIO RETIREMENT / FINASSETNORET HOMEEQUITY Home equity value (Home value - outstanding mortgage payments) HOMEERATIO HOMEEQUITY / FINASSETNORET 24

26 Table 2. Assets by MTR, N=2,228. MTR Percent Income Mean 34,150 48,492 51,829 57,802 58,940 61, ,875 Median 27,000 40,000 50,000 51,000 56,000 59,000 97,000 Taxable Financial Mean 51,540 70,007 50,046 49,673 65,187 36, ,992 Assets Median 5,700 10,020 14,700 9,150 27,210 15,500 52,800 Retirement Assets Mean 9,971 18,619 55,362 32,873 22,735 37, ,503 Median 0 1,000 4,500 7,000 9,100 14,000 40,000 Home Equity Mean 33,516 49,949 56,500 66,170 65,339 80, ,750 Median 0 22,000 44,000 32,000 60,000 60,000 95,000 EFC Mean 0 1,345 2,825 3,669 4,520 5,658 52,069 Median 0 1,440 2,768 3,730 4,482 5,665 19,807 Ratios computed using aggregate cell averages reported above: RETIREMENT ratio Mean HOMEEQUITY ratio Mean Ratios computed as averages and medians of individual observations: RETIREMENT ratio Mean Median HOMEEQUITY ratio Mean Median Source: 2001 Survey of Consumer Finances. 25