WORK IN PROGRESS PLEASE DO NOT CITE COMMENTS INVITED

INCOME TAXES AND THE PHILANTHROPY OF ENTREPRENEURS

Karie Barbour University of Tennessee

Donald Bruce University of Tennessee

Douglas Holtz-Eakin Syracuse University and NBER (On Leave at the Council of Economic Advisers)

October 2001

We thank the Center for Policy Research at Syracuse University and the Center for Business and Economic Research at the University of Tennessee for their support. We are also grateful to Stacia Couch for her help in preparing the manuscript.

INCOME TAXES AND THE PHILANTHROPY OF ENTREPRENEURS ABSTRACT: Charitable giving by entrepreneurs is of interest from a tax-policy perspective because they are often high-income individuals facing higher marginal tax rates and lower tax prices for giving. Alternatively, considerable research suggests that they face binding liquidity constraints, making the opportunity cost of charitable giving higher, ceteris paribus. We examine the charitable contributions of entrepreneurs, and investigate whether they are more or less tax-sensitive than others in their decisions to give money to charity. Our analysis of a panel of individual taxpayer data from 1979 to 1990 reveals that single taxpayers who file a Schedule C and itemize their deductions are likely to report more and larger charitable contributions, but only at lower marginal income tax rates. They are also less sensitive to changes in marginal tax rates than those who do not file a Schedule C. We do not find significant differences for married taxpayers. Karie Barbour Center for Business and Economic Research 100 Glocker Building University of Tennessee Knoxville, TN 37996 (865)974-5441 (tel.) (865)974-3100 (fax) [email protected] Donald Bruce Center for Business and Economic Research 100 Glocker Building University of Tennessee Knoxville, TN 37996 (865)974-5441 (tel.) (865)974-3100 (fax) [email protected] Douglas Holtz-Eakin Center for Policy Research and NBER (On leave at the Council of Economic Advisers) 426 Eggers Hall Syracuse University Syracuse, NY 13244 (315)443-3612 (tel.) (315)443-3717 (fax) [email protected]

1.

INTRODUCTION Charitable organizations often target successful entrepreneurs for support. Charitable

giving by entrepreneurs is of interest from a tax-policy perspective as well. First, entrepreneurs are often high-income individuals facing high marginal tax rates. Earlier research on this topic has consistently found that giving increases as tax rates increase, so we might expect to find that entrepreneurs are more likely to give to charity. Second, considerable research suggests that entrepreneurs often face binding liquidity constraints, making the opportunity cost of charitable giving higher, ceteris paribus. This would suggest that entrepreneurs would be less likely to give to charity. This is the first study to our knowledge to directly address entrepreneurial philanthropy within the context of the federal personal income tax. While much has been written about general tax incentives to contribute to charity, little is known about the manner in which entrepreneurs change their giving in response to differences in tax rates. This is a highly relevant topic in light of recent tax policy discussions, aimed at broadly reducing marginal tax rates and improving the overall business climate, especially for small businesses. To the extent that new tax policies target small business, they could also have secondary effects on charitable giving. To briefly summarize the relevant prior literature, the so-called “first generation” studies typically found that taxpayers respond rather dramatically to changes in tax rates. 1 The resulting tax price elasticities, mainly derived using cross-section data, were either equal to or greater than one in absolute value. In other words, increases in tax prices (defined as one minus the marginal tax rate) lead to more than proportionate reductions in charitable giving.

1

See, for example, the summaries of this literature and additional references provided by Clotfelter (1985), Steinberg (1990), and Bakija (2000).

1

More recent research has questioned this general conclusion, however, beginning with Auten, Cilke, and Randolph (1992) and later with Randolph (1995) and Bakija (2000). The central message from the more recent studies is that the measured response in a cross-sectional framework involves the combination of permanent behavioral effects and transitory (i.e., shifting) effects. The newer studies have used panel data for large samples of taxpayers and have typically found much smaller permanent tax price elasticities, typically in the area of –0.4. Again, however, none of the earlier studies have focused on the different effects for entrepreneurs. Two other areas of research serve as foundations for our analysis. The first is the relatively recent body of literature that demonstrates the general tax sensitivity of entrepreneurship. Bruce (Forthcoming and 2000), Schuetze (2000), and Gentry and Hubbard (2000) find that marginal tax rates can have important effects on the decision to start or close a small business. Carroll, Holtz-Eakin, Rider and Rosen (Forthcoming, 2000a, and 2000b) conclude that taxes have important effects on entrepreneurial growth, hiring, and investment, respectively. If these entrepreneurial decisions are influenced by tax considerations, it would not be surprising to find that the choice to contribute to charity is affected as well. A second related area of research is the well-established literature on the importance of liquidity constraints among small business owners. Evans and Leighton (1989), Evans and Jovanovic (1989), and Meyer (1990) are among the pioneering studies of liquidity constraints and self-employment. Blanchflower and Oswald (1998) and Holtz-Eakin, Joulfaian, and Rosen (1994a and 1994b) also reveal the importance of financial capital to self-employment entry and duration. These papers generally conclude that liquidity constraints are often binding, which suggests that decisions to contribute to charity may be intertwined with other business decisions.

2

We make use of the same panel data used by Bakija (2000) to address the relative tax sensitivity of charitable giving among entrepreneurial and non-entrepreneurial taxpayers, defined by the presence of a Schedule C for small business income. 2 Our analytic framework is presented in Section 2 of the paper, followed by a discussion of our data in Section 3. Section 4 contains a discussion and presentation of our multivariate analysis and Section 5 concludes. To briefly summarize our main results, we find that single entrepreneurs who itemize their deductions are likely to report more and larger charitable contributions than non-entrepreneurs at lower marginal tax rates, and fewer and/or smaller contributions at higher tax rates. Contributing to this is the finding that entrepreneurs are relatively less sensitive to changes in the tax price of giving. We do not find similar results for married taxpayers after we control for other factors.

2.

ANALYTIC FRAMEWORK We begin with a conventional model of charitable giving in which individuals have

utility derived from consumption (C) and charitable gifts (G). In addition, however, we recognize that individuals could chose to neither consume nor give donations, but rather save and accumulate wealth (to finance future consumption and gift activities). Thus, individuals seek to maximize the function U (C ,G) + βV (W )

(1)

where $ < 1 is the rate at which individuals discount the future. In this two-period framework, individuals use initial income (Y) to cover expenses due to consumption, charitable gifts, taxes and save the remainder. That is:

2

See Bruce and Holtz-Eakin (Forthcoming) for a general analysis of other tax-based definitions of “entrepreneur.”

3

Y =C + G + S + T

(2)

For simplicity of exposition, we assume that individuals face a linear income tax at the rate J. Charitable contributions, however, are deductible prior to tax. Thus, (2) becomes Y (1 − τ ) = C + G(1 − τ) + S

(2')

which shows clearly the role of the income tax in reducing the effective costs of philanthropy. We turn now to wealth accumulation and the special features introduced by entrepreneurs. We assume that as a self-employed entrepreneur, the individual’s income depends on her ability 2 as an entrepreneur and the amount of capital (k) invested in the firm. We further assume that entrepreneurial income is taxed at the rate :. In practice, we expect : to exceed J, reflecting the fact, for example, that income of sole-proprietors is subject to both the self-employment tax and the income tax. In addition to entrepreneurial income, wealth accumulation derives from the net-of-tax return to saving. Combining these sources W = S (1 + r (1 − τ)) + (1− µ )[ θ f ( k ) − rk ]

(3)

By definition, k - S is the amount of capital financed by borrowing. Following previous research, we assume that the amount of borrowing, and thus capital invested in the enterprise, is bounded by a liquidity constraint generated by the financial markets. The size of the constraint depends on the individual’s net assets k ≤ L( S )

(4)

Combining (1), (2’), and (3) and maximizing with respect to consumption and charitable gifts yields the first order conditions

4

and

∂U −βV ′(W ) (1 + r (1 − τ) ) − λL′( S ) (1 + r (1 − τ )) = 0 ∂C1

(5)

∂U −β V ′(W ) (1 + r(1 − τ) ) (1 − τ) − λL′( S) (1 + r (1 − τ) ) (1 − τ) = 0. ∂G

(6)

Taken together, (5) and (6) imply that ∂U ∂U / = (1 − τ). ∂G ∂C1

(7)

That is, individuals set their marginal rate of substitution equal to the relative prices of gifts and consumption, the conventional finding in the literature on tax-based charity. Importantly, (7) does not depend upon whether the individual is an entrepreneur or not. And among entrepreneurs, it holds regardless of the nature of liquidity constraints on firm operations. How do these considerations affect charitable giving? Inspection of the right side of (6) shows that individuals equate the marginal benefit of additional gifts to an opportunity cost that includes two components. The first is the foregone marginal utility of wealth accumulation (inclusive of interest and appropriately discounted). The second reflects the value of increased entrepreneurial capital enabled by wealth accumulation and relaxed liquidity constraints. Ceteris paribus, the higher opportunity cost faced by constrained entrepreneurs must be matched by a higher marginal benefit, which typically implies a lower level of G* . Note, however, that a parallel set of considerations applies to (6); the C* will likely be reduced at the same time. Thus, a complete understanding of the solution depends upon entrepreneurial investment. The process by which an entrepreneur determines his or her optimal amount of capital yields three possible solutions. In the first, the entrepreneur employs capital and the liquidity constraint is not binding. In this case, the rate of return equals the marginal product of capital. The second possibility is that the liquidity constraint is again not binding, but the entrepreneur’s ability is

5

sufficiently low that the marginal product of capital is below the interest rate. Hence, k * =0. However, we assume that f(0)>0, so that the entrepreneur may display income, but no depreciable assets. The liquidity constraint is non-binding in each of these first two cases, so the opportunity cost regarding foregone entrepreneurial capital is irrelevant in the decision to contribute to charity, and nothing suggests that entrepreneurs would be any different from others. However, the third and final possibility is that the liquidity constraint is binding, so that k* =L(S). It is in this case that the opportunity cost of charitable contributions is greatest. If the return to additional investment in the entrepreneurial venture exceeds the marginal benefit of contributing that money to charity, we might expect to see liquidity-constrained entrepreneurs making fewer and smaller donations. The actual relationship between entrepreneurial activity and philanthropy depends upon individual preferences for giving (as well as firm-specific liquidity constraints) and can only be revealed through empirical analysis, to which we now turn.

3.

DATA AND DESCRIPTIVE STATISTICS Data for this analysis are drawn from the University of Michigan Tax Research Database.

In constructing this database, the Office of Tax Policy Research (OTPR) at the University of Michigan acquired the public-use tax return data released by the Internal Revenue Service Statistics of Income (SOI) Division and converted them into a user-friendly format. These data files have been used, either in their raw SOI format or in the refined OTPR format, for much of the recent empirical research in this area. The database is a panel representing over 200,000 tax returns spanning the years 1979 to 1990. Returns for some 6,000 filers are included for each of the 12 years, while the remaining filers are in the panel for some subset of the 12 years. Throughout this study, we analyze a

6

pooled version of this panel, where each filer can contribute up to twelve separate observations. 3 While our focus is on itemized charitable contributions, we make use of several other elements of the tax return data. Descriptive statistics for the pooled sample are provided in Table 1. We leave a detailed inspection of this table to the reader and instead highlight a few key results. First, note that 13.3 percent of all single filers (84.6 percent of single itemizers) report nonzero charitable contributions. Married filers are substantially more likely to report charitable contributions (47.1 percent of all filers, and 93.9 percent of itemizers). Consequently, as noted by other authors in this literature, it is important to consider itemization probabilities when analyzing charitable giving. Our tax price variable is defined as the natural log of one minus the taxpayer’s marginal federal income tax rate. 4 Single filers have lower tax prices than married filers; itemizers have higher tax prices regardless of marital status. The dollar value of reported contributions is also higher on average for married filers. Single itemizers report an annual average of $878.00 in charitable contributions over the course of the 12-year panel, while married itemizers report an average of $1,421.99. Figures 1A and 1B provide an initial picture of the share of taxpayers that report such contributions and the average amounts of those contributions over time. Looking first at Figure 1A, note that single taxpayers who file a Schedule C are always more likely to report charitable contributions. Furthermore, the average amount of charitable contributions is higher for Schedule C filers in every year of our panel. To summarize this figure, single taxpayers who file

3

We address the associated empirical issues of doing this in the analysis that follows, namely by using fixed-effects techniques.

7

a Schedule C are nearly twice as likely to report charitable contributions, and give nearly twice as much on average. A similar story holds for married filers as shown in Figure 1B, although the differences are somewhat smaller in percentage terms. In all but one year (1988), those filing a Schedule C are more likely to report charitable contributions and are likely to report larger dollar amounts on average. Both the rate of giving (the share of all taxpayers reporting charitable contributions) and the dollar amounts are unsurprisingly larger for married taxpayers than for singles. An immediate thought is that those filing a Schedule C may just be more likely to itemize, thereby biasing the comparison somewhat. Figures 2A and 2B repeat the analysis for itemizers only and, despite some moderation in the differences, our general conclusion holds. In all but two years, single itemizers who file a Schedule C are more likely to report charitable contributions. In all but the last year of data, the average amount of their contributions is higher than for those who do not file a Schedule C. The giving rates for married itemizers are not dramatically different on the basis of Schedule C status, but entrepreneurs give larger amounts on average in all but the last three years. The remaining figures shed some additional light on this general result by breaking the sample into various groups on the basis of other available covariates in the data. This exercise is intended to determine whether the differences based on the presence of a Schedule C are truly the result of the entrepreneurial activity itself or of some common omitted factor. First, as noted above, entrepreneurs often face binding liquidity constraints that may restrict their ability to give money to charity. To investigate this, we make use of one indicator

4

It should be noted that this only captures the last-dollar tax price of gifts of cash. The price of non-cash gifts would include a consideration of capital gains rates and depreciation. Since we cannot disentangle cash and noncash gifts in our tax data, we assume that most gifts are of cash and use the cash gifts tax price.

8

of wealth that is available in our tax data, the presence of itemized deductions for home mortgage interest. Figures 3A and 3B reveal that those with mortgage interest deductions tend to report lower amounts of charitable contributions regardless of marital status and whether or not they file a Schedule C (1986 is the lone exception, for single filers only). This suggests that the presence of a mortgage interest deduction is less of an indicator of wealth and the ability to contribute to charity and more of an indicator of liquidity constraints. Taxpayers who are unburdened by mortgage interest payments are likely to report larger gifts to charity. Figures 4A and 4B present a similar analysis by the presence of an exemption for a taxpayer over the age of 65. Taxpayers in or near retirement may be more or less likely to contribute to charity, and are more likely to be self-employed (see, for example, Bruce, HoltzEakin, and Quinn, 2000). The results of this analysis depend largely on marital status. For single itemizers (Figure 4A), those with an age 65 exemption give larger amounts to charity, and the differences are quite dramatic in the earliest years of our panel. It is interesting to note that the differences are most substantial for married itemizers in the later years. For both single and married itemizers, the differences on the basis of the age 65 exemption are modest during the mid-1980s. Nonetheless, these figures reveal that age may have important effects on the dollar amounts given to charity (and reported on a tax return), as found in earlier research. Household size can also have an impact on charitable giving, although it is not clear that entrepreneurial households are typically larger or smaller on average. Figures 5A through 5C present data for various categories of taxpayers, defined by marital and Schedule C status in addition to the number of dependent exemptions on the tax return (to proxy for household size). For single itemizers, no clear relationship between the number of exemptions and the amount of itemized charitable contributions emerges from the data. A similar conclusion can be reached

9

regarding married itemizers, but returns with three or four exemptions tend to report smaller contributions than do those with fewer or more exemptions. Figures 6A and 6B provide an initial glance at the importance of income in the charitable giving process. If entrepreneurs are more likely to give larger amounts simply because they typically have more income, we should expect to see no difference in giving for Schedule C filers compared with those without a Schedule C in the same income group. Figures 6A and 6B divide the samples in half using the median AGI for each year-group combination. For both single and married itemizers, those in the top half of the AGI distribution always report larger amounts of charitable contributions, an unsurprising result. Looking more deeply at these findings reveals that the difference between Schedule C filers and non-Schedule-C filers is most prominent in the top half of the AGI distribution. For those below the median AGI amount, Schedule C has no clear relationship with the average dollar amount of contributions. As a final descriptive pass at these data, Figures 7A through 8B present average itemized contributions to charity for the bottom, middle, and top third of the marginal income tax rate distribution. Looking first at single itemizers who do not file a Schedule C (Figure 7A), a clear direct relationship seems to exist—those in higher tax brackets give more to charity. A similar result emerges for Schedule C filers, although the relationship is less concrete. For married itemizers in Figures 8A and 8B, the largest difference in terms of gift amounts is seen among the top third of the marginal tax rate distribution. Somewhat surprisingly, no clear differences exist between those in the bottom and middle thirds. This trend is more uniform across Schedule C status for married itemizers than for singles, suggesting a smaller impact of entrepreneurship on giving for married taxpayers.

10

4.

MULTIVARIATE EMPIRICAL STRATEGY AND RESULTS To fully gauge the impact of these and other covariates on charitable giving, a

multivariate econometric strategy is necessary. We begin by examining the decision to report charitable contributions in a discrete choice framework, using a fixed-effects logit technique to account for unobserved individual heterogeneity that is constant for each taxpayer across time. It should be noted that this method omits those who either always contribute to charity or who never contribute to charity. Effects of covariates are identified on the basis of changes over time. Unfortunately, our choice of control variables is somewhat limited by the tax return data. Variables of focus in our specification include the log of the tax price, a dummy variable for the presence of a Schedule C, and an interaction between these two. We also include an indicator for the presence of an age 65 exemption, the number of dependent exemptions, AGI (in $1,000s), an indicator for the presence of a mortgage interest deduction, and indicators for the region of residence. The first column of Table 2 presents fixed-effects logit results for all single filers. A key finding is that, all else equal, filing a Schedule C affects one’s probability of reporting charitable contributions. Given the statistically significant coefficient on the interaction between the Schedule C indicator and the log of the tax price (and the fact that the log of the tax price is always negative), the effect of a Schedule C depends upon the individual’s marginal tax rate. At very low marginal tax rates (and high tax prices, the log of which will approach zero), the effect of filing a Schedule C is positive. The effect becomes smaller as the marginal tax rate increases, and actually turns negative at a tax rate of 34.8 percent. Entrepreneurs with the highest marginal income tax rates are less likely to report charitable contributions, all else equal.

11

The tax price variable itself (ln(1-MTR)) has the expected negative sign—higher marginal tax rates (which yield lower tax prices) increase the probability of reporting charitable contributions. Those filing a Schedule C are apparently less sensitive to changes in the tax price, as indicated by the positive and statistically significant coefficient on the interaction between the Schedule C dummy variable and the tax price variable. Those with an age 65 exemption are less likely to report charitable contributions, while those with a mortgage interest deduction are more likely. Income has the expected positive sign and significance. These basic results are similar for all married filers as shown in the third column of Table 2, although the effect of a Schedule C is smaller and marginally less statistically significant. A key difference for married filers is that having more exemptions has a positive impact on the decision to report charitable contributions, where exemptions have no discernible effect for single filers. A potential issue with the results in the first and third columns of Table 2 is that the coefficients are picking up some of the simultaneous effects on the probability of itemizing deductions. The second and fourth columns of this table repeat the analysis for itemizers only, and a few key differences emerge. First, while the tax price effects remain largely unchanged (if not slightly different in magnitude), the interaction of the tax price and the Schedule C dummy becomes statistically insignificant for married itemizers. It remains important for single itemizers, however, indicating that the effect of a Schedule C is positive at marginal tax rates below 22.3 percent and negative at higher tax rates. The coefficient on the Schedule C indicator itself is not statistically different from zero for married itemizers. This is a key finding, as it again suggests that the presence of a Schedule C is only important for single itemizers in a multivariate context. Among married itemizers,

12

having a Schedule C for small business income is apparently unrelated to the decision to report charitable contributions once other important factors in this process are controlled for. The impact of the age 65 exemption becomes indistinguishable from zero for both marital status groups when the focus is turned toward itemizers, suggesting that those over 65 are simply less likely to itemize and not necessarily less likely to report charitable contributions. The effect of the mortgage interest deduction is also significantly diminished (although it remains positive and statistically different from zero), perhaps as a result of the fact that most itemizers report mortgage interest deductions as shown in Table 1. Finally, income has no effect on the probability of reporting charitable contributions for single itemizers. As a first step toward understanding the determinants of the dollar values of charitable contributions and the potential effects for entrepreneurs, Table 3 presents results from fixedeffects regressions. 5 The dependent variable in these regressions is the natural log of the amount of reported charitable contributions. 6 Note first that our initial estimates of the tax price elasticity of giving are –1.893 for single itemizers and –1.213 for married itemizers. The interaction of the tax price and the Schedule C dummy variable is statistically significant for single itemizers only, bringing the tax price elasticity for those filing a Schedule C up to –1.095 (or –1.893 plus 0.798). In both cases, we confirm the finding from earlier research that higher marginal tax rates lead to increased charitable contributions. As foreshadowed by the fixed-effects logit analysis, single itemizers filing a Schedule C report larger contributions at lower marginal tax rates. Due to the interaction variable, the effect

5

Results from standard Hausman tests, available upon request from the authors, support the use of fixed effects in these pooled panel data regressions. 6 In order to preserve itemizers who do not report charitable contributions, we first add one dollar to the total before taking the log.

13

of a Schedule C turns negative at a marginal tax rate of 16.7 percent. Further, contributions by entrepreneurs are less sensitive to tax price fluctuations than are those from non-entrepreneurs. These results are consistent with our discussion of liquidity constraints. Given fixed values of all other control variables (including AGI and the number of exemptions), differences in marginal tax rates must be driven by differences in deductions. Specifically, higher marginal tax rates can therefore be a reflection of liquidity constraints, to the extent that such constraints reduce an entrepreneur’s ability to reduce her tax burden (and marginal tax rate) via deductions for business-related expenses. They also reduce contributions to charity, as discussed above, by raising the opportunity cost of giving. By extension, unconstrained entrepreneurs have the ability not only to reduce their marginal tax rates but also to give more to charity. Turning to the remaining covariates, the number of dependent exemptions has a positive effect on the amount of charitable contributions, but only for married itemizers. Income again has the expected positive and significant sign. Also, the presence of a mortgage interest deduction increases the amount of charitable contributions, but only for single itemizers. We recognize that the tax price elasticities in Table 3 are substantially larger than those in the recent empirical literature, and are able to point to a number of reasons for the differences. First, our approach thus far (and the resulting coefficients) more closely resembles that of the first-generation studies. Accounting for permanent versus transitory effects as in Bakija (2000) would certainly yield different results. Second, our measure of the tax price only includes the marginal income tax rate and does not account for the tax price of non-cash gifts. Third, our sample includes many taxpayers that have been omitted in previous empirical studies. Specifically, Bakija (2000) and others omit those who are not present for the entire panel (or at least nine years of it) and those who would not be able to itemize their deductions in the absence

14

of their charitable contributions. We prefer to retain this valuable variation in the data and keep our focus on the relative differences between those who file a Schedule C and those who do not. 7 A final cause of the differences in the tax price elasticities is that we have not yet considered the possible endogeneity or measurement error of the tax price variables, to which we now turn. In doing so, we require instrumental variables that are correlated with individual tax prices but do not have a separate effect on the amount of contributions. Our strategy is to construct instrumental variables that rely on exogenous variation in the tax code. 8 Specifically, for each year and marital status we create a group of ten instrumental variables using the NBER TAXSIM model. 9 We first calculate effective marginal income tax rates at the 10th, 25th, 50th, 75th, and 90th percentiles of the wage and self-employment earnings distributions, respectively, for the year prior to each year in our panel. 10 Next, we inflate the relevant financial inputs using the consumer price index and repeat the computation using the actual year’s tax code. Our instrumental variables are computed as the year-, maritalstatus-, and sector-specific differences between the actual and prior years’ tax rates. Several features of this procedure merit attention. First, by using exogenous points in the earnings distributions, the link between individual behavior and tax rates is broken. Second, by construction, our instrumental variables rely on the change in tax rates at fixed levels of

7

By doing so, we must be sure to account for the different probabilities of itemizing deductions over time. We intend to address any resulting selection bias as well as other important econometric issues in future research. 8 See Auten, Cilke, and Randolph (1992) for a thorough discussion of tax changes that affected charitable giving during this time period. 9 See http://www.nber.org/taxsim, or Feenberg and Coutts (1993). 10 To compute the percentile-specific tax rates, we use average values for the non-income TAXSIM inputs, calculated over returns “near” the desired percentile; specifically: 0 to 20 for the 10th percentile, 20 to 35 for the 25th , 35 to 65 for the 50th , 65 to 85 for the 75th , and 85 to 100 for the 90th .

15

individual behavior. Thus, our source of identification is exogenous shifts in the structure (level and slopes) of the effective marginal tax rate structure itself. The results of our fixed-effects two-stage least squares analysis are shown in table 4. While our instrumental variables perform quite well in the first stage regressions, the second stage results are far less satisfying. 11 Specifically, the tax price coefficient becomes large, positive, and highly significant for single itemizers and larger, negative, and equally significant for married itemizers. The Schedule C indicator has no discernible effect for single itemizers but a very large effect for married itemizers. Results are similar for the interactions between the tax prices and the Schedule C variables. Needless to say, these results are highly suspect and must be subjected to further analysis before a more definitive conclusion may be reached.

5.

CONCLUSION The subject of entrepreneurial philanthropy has not received significant attention in the

empirical tax literature. In this analysis of a 12-year panel of taxpayer data, we find that the extent to which tax rates affect charitable giving is different for entrepreneurs. Specifically, single itemizers who file a Schedule C are more likely to report charitable contributions, and they are more likely to give larger amounts, at lower marginal tax rates. As the marginal tax rate increases, entrepreneurs are relatively less likely to give (and give less). Finally, they are less sensitive to changes in marginal tax rates than are those who do not file a Schedule C. These general findings hold both in terms of simple descriptive statistics and with more conclusive

11

Results from first-stage regressions are available upon request from the authors.

16

multivariate analyses. We do not find similar results for married taxpayers when we control for other factors. Our results for single itemizers are broadly consistent with two themes. First, higher marginal tax rates lead to greater amounts of charitable contributions, both for entrepreneurs and non-entrepreneurs. Second, to the extent that higher marginal tax rates reflect the presence of liquidity constraints, all else (including income) constant, constrained entrepreneurs are not only less likely to give to charity, but also likely to give smaller dollar amounts. While these general findings are illustrative of the underlying tax sensitivity of charitable giving among entrepreneurs, much work remains to be done. Namely, a better instrumental variables specification is needed to address potential endogeneity and measurement error. An equally important task in future work will be to address the changing probability of itemizing over time, perhaps via some form of sample selection model. Finally, it will be important to follow Bakija (2000) in addressing the difference between permanent and transitory effects of tax price changes on charitable giving.

17

REFERENCES

Auten, Gerald A., James M. Cilke, and William C. Randolph. 1992. “The Effects of Tax Reform on Charitable Contributions.” National Tax Journal 45(3): 267-290. Bakija, Jon. 2000. “Distinguishing Transitory and Permanent Price Elasticities of Charitable Giving with Pre-Announced Changes in Tax Law.” Manuscript, Williams College. Blanchflower, David G. and Andrew J. Oswald, 1998. “What Makes an Entrepreneur?” Journal of Labor Economics 16:1, 26-60. Bruce, Donald. Forthcoming. “Taxes and Entrepreneurial Endurance: Evidence from the SelfEmployed,” National Tax Journal. Bruce, Donald. 2000. “Effects of the United States Tax System on Transitions Into SelfEmployment,” Labour Economics 7(5): 547-575. Bruce, Donald, and Douglas Holtz-Eakin. Forthcoming. “Who Are the Entrepreneurs? Evidence from Taxpayer Data.” Journal of Entrepreneurial Finance and Business Ventures. Bruce, Donald, Douglas Holtz-Eakin, and Joseph Quinn. 2000. “Self-Employment and Labor Market Transitions at Older Ages.” Manuscript, Syracuse University. Carroll, Robert, Douglas Holtz-Eakin, Mark Rider, and Harvey S. Rosen. Forthcoming. “Personal Income Taxes and the Growth of Small Firms.” Tax Policy and the Economy. Carroll, Robert, Douglas Holtz-Eakin, Mark Rider and Harvey S. Rosen. 2000a. “Income Taxes and Entrepreneurs’ Use of Labor,” Journal of Labor Economics, 18(2) (April): 324-351. Carroll, Robert, Douglas Holtz-Eakin, Mark Rider, and Harvey S. Rosen. 2000b. “Entrepreneurs, Income Taxes, and Investment.” In Joel B. Slemrod (ed.), Does Atlas Shrug? New York: Russell Sage Foundation, pp. 427-455. Clotfelter, Charles T. 1985. Federal Tax Policy and Charitable Giving. Chicago: University of Chicago Press. Evans, David S. and Boyan Jovanovic. 1989. “An Estimated Model of Entrepreneurial Choice Under Liquidity Constraints,” Journal of Political Economy, 97: 808-827. Evans, David S. and Linda S. Leighton, 1989. “Some Empirical Aspects of Entrepreneurship.” American Economic Review 79:3, 519-35. Feenberg, Daniel and Elizabeth Coutts. 1993. “An Introduction to the TAXSIM Model,” Journal of Policy Analysis and Management, 12(1): 189-194.

18

Gentry, William M. and R. Glenn Hubbard. 2000. “Tax Policy and Entrepreneurial Entry,” American Economic Review, 90 (May): 283-287. Holtz-Eakin, Douglas, David Joulfaian, and Harvey S. Rosen. 1994a. “Sticking It Out: Entrepreneurial Survival and Liquidity Constraints,” Journal of Political Economy (February): 53-75. Holtz-Eakin, Douglas, David Joulfaian, and Harvey S. Rosen. 1994b. “Entrepreneurial Decisions and Liquidity Constraints,” RAND Journal of Economics, 23(2) (Summer): 334-347. Meyer, Bruce D. 1990. “Why Are There So Few Black Entrepreneurs?” NBER Working Paper No. 3537. Randolph, William C. 1995. “Dynamic Income, Progressive Taxes, and the Timing of Charitable Contributions.” Journal of Political Economy 103(4): 709-738. Schuetze, Herb J. 2000. “Taxes, Economic Conditions, and Recent Trends in Male SelfEmployment: A Canada-US Comparison,” Labour Economics, 7(5): 507-544. Steinberg, Richard. 1990. “Taxes and Giving: New Findings.” Voluntas 1(2): 61-79.

19

700

0.35

600

0.3

500

0.25

400

0.2

300

0.15

200

0.1

100

0.05

0

Giving Rate (Lines)

Mean Contribution (Bars)

Figure 1A: Giving Rates and Mean Amounts of Total Contributions -All Single Filers

No Schedule C Schedule C No Schedule C Schedule C

0 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

Figure 1B: Giving Rates and Mean Amounts of Total Contributions -All Married Filers

1800

0.7

1600

0.6

0.5 1200 0.4

1000 800

0.3

600 0.2 400 0.1

200 0

0 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

20

Giving Rate (Lines)

Mean Contribution (Bars)

1400

No Schedule C Schedule C No Schedule C Schedule C

Figure 2A: Giving Rates and Mean Amounts of Total Contributions -Single Itemizers, By Schedule C Status 2000

0.9

1800

Mean Contribution (Bars)

1400

0.86

1200 1000

0.84

800 0.82

600 400

Giving Rate (Lines)

0.88

1600

No Schedule C Schedule C No Schedule C Schedule C

0.8

200 0

0.78 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

3000

0.96

2500

0.95

2000

0.94

1500

0.93

1000

0.92

500

0.91

0

0.9 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

21

Giving Rate (Lines)

Mean Contribution (Bars)

Figure 2B: Giving Rates and Mean Amounts of Total Contributions -Married Itemizers, By Schedule C Status

No Schedule C Schedule C No Schedule C Schedule C

Figure 3A: Mean Amounts of Total Contributions -Single Itemizers, By Schedule C and Mortgage Deduction Status 4500

4000

Mean Contribution

3500

3000 Deduction, No Schedule C Deduction, Schedule C

2500

No Deduction, No Schedule C No Deduction, Schedule C

2000

1500

1000

500

0 1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

Figure 3B: Mean Amounts of Total Contributions -Married Itemizers, By Schedule C and Mortgage Deduction Status 10000

9000

Mean Contribution

8000

7000

6000

Deduction, No Schedule C Deduction, Schedule C

5000

No Deduction, No Schedule C No Deduction, No Schedule C

4000

3000

2000

1000

0 1979

1980

1981

1982

1983

1984

1985

1986

22

1987

1988

1989

1990

Figure 4A: Mean Amounts of Total Contributions -Single Itemizers, By Schedule C and Age 65 Status 9000

8000

Mean Contribution

7000

6000

Over 65, No Schedule C

5000

Over 65, Schedule C Under 65, No Schedule C

4000

Under 65, Schedule C

3000

2000

1000

0 1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

Figure 4B: Mean Amounts of Total Contributions -Married Itemizers, By Schedule C and Age 65 Status 18000 16000

Mean Contribution

14000 12000 Over 65, No Schedule C

10000

Over 65, Schedule C Under 65, No Schedule C

8000

Under 65, Schedule C

6000 4000 2000 0

1979

1980

1981

1982

1983

1984

1985

23

1986

1987

1988

1989

1990

Figure 5A: Mean Amounts of Total Contributions -Single Itemizers, By Schedule C Status and Number of Exemptions

5000

Mean Contribution

4500 4000 3500 No Schedule C, Exemption=1

3000

No Schedule C, Exemption>1 2500

Schedule C, Exemption=1

2000

Schedule C, Exemption>1

1500 1000 500 0 1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

Figure 5B: Mean Amounts of Total Contributions -- Married Itemizers Who Did Not File Schedule C, By Number of Exemptions

5000

4500

Mean Contribution

4000

3500

3000 No Schedule C, Exemptions=1-2

2500

No Schedule C, Exemptions=3-4 No Schedule C, Exemptions>4

2000

1500

1000

500

0 1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

Figure 5C: Mean Amounts of Total Contributions -- Married Itemizers Who Filed Schedule C, By Number of Exemptions

4000

Mean Contribution

3500

3000

2500

Schedule C, Exemptions=1-2

2000

Schedule C, Exemptions=3-4 Schedule C, Exemptions>4

1500

1000

500

0 1979

1980

1981

1982

1983

1984

1985

24

1986

1987

1988

1989

1990

Figure 6A: Mean Amounts of Total Contributions -Single Itemizers, By Schedule C Status and AGI Bracket 3500

Mean Contribution

3000 2500 No Schedule C, Top AGI 2000

No Schedule C, Bottom AGI Schedule C, Top AGI

1500

Schedule C, Bottom AGI 1000 500 0 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

Figure 6B: Mean Amounts of Total Contributions -Married Itemizers, By Schedule C Status and AGI Bracket 4500 4000

Mean Contribution

3500 3000

No Schedule C, Top AGI

2500

No Schedule C, Bottom AGI

2000

Schedule C, Top AGI Schedule C, Bottom AGI

1500 1000 500 0 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

25

Figure 7A: Mean Amounts of Total Contributions -- All Single Itemizers Who Did Not File Schedule C, By Marginal Tax Rate Bracket 2500

Mean Contribution

2000

1500

Marginal Tax Rate, Bottom Marginal Tax Rate, Middle Marginal Tax Rate, Top

1000

500

0 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

Figure 7B: Mean Amounts of Total Contributions -- Single Itemizers Who Filed Schedule C, By Marginal Tax Rate Bracket 5000 4500 4000

Mean Contribution

3500 3000

Marginal Tax Rate, Bottom Marginal Tax Rate, Middle

2500

Marginal Tax Rate, Top 2000 1500 1000 500 0 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

26

Figure 8A: Mean Amounts of Total Contributions -- Married Itemizers Who Did Not File Schedule C, By Marginal Tax Rate Bracket 4000 3500

Mean Contribution

3000 2500 Marginal Tax Rate, Bottom 2000

Marginal Tax Rate, Middle Marginal Tax Rate, Top

1500 1000 500 0 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

Figure 8B: Mean Amounts of Total Contributions -- Married Itemizers Who Filed Schedule C, By Marginal Tax Rate Bracket 4500 4000

Mean Contribution

3500 3000 Marginal Tax Rate, Bottom

2500

Marginal Tax Rate, Middle 2000

Marginal Tax Rate, Top

1500 1000 500 0 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990

27

Table 1: Summary Statistics Variable Reported Charitable Contributions* Age 65+ Exemption* Exemptions AGI/1,000 ln(1-MTR) Mortgage Interest Deduction* Schedule C* C * ln(1-MTR) West* Midwest* South* Northeast* Charitable Contributions ($) N

All Single Filers Single Itemizers 0.133 0.846 (.339) (.361) 0.093 0.142 (.29) (.349) 1.274 1.435 (.846) (.856) 12.253 28.609 (26.677) (61.295) -0.172 -0.284 (.131) (.154) 0.090 0.574 (.286) (.494) 0.050 0.101 (.218) (.301) -0.009 -0.027 (.052) (.098) 0.207 0.259 (.405) (.438) 0.244 0.224 (.43) (.417) 0.315 0.261 (.464) (.439) 0.228 0.250 (.419) (.433) 137.93 878.00 (1955.64) (4867.86) 152,250 23,918

Note: Entries are means with standard deviations in parentheses. * Indicates a dummy variable. AGI = adjusted gross income; MTR = marginal tax rate.

28

All Married Filers Married Itemizers 0.471 0.939 (.499) (.239) 0.126 0.071 (.332) (.256) 3.195 3.374 (1.353) (1.317) 32.931 47.441 (122.586) (170.672) -0.256 -0.322 (.156) (.158) 0.416 0.830 (.493) (.376) 0.171 0.204 (.376) (.403) -0.041 -0.060 (.115) (.142) 0.196 0.221 (.397) (.415) 0.255 0.254 (.436) (.435) 0.338 0.294 (.473) (.456) 0.198 0.216 (.399) (.412) 713.16 1421.99 (10326.83) (14547.59) 142,073 71,253

Table 2: Discrete Choice Analysis of the Decision to Itemize Gifts to Charity Variable Age 65+ Exemption Exemptions AGI/1,000 ln(1-MTR) Mortgage Interest Deduction Schedule C C * ln(1-MTR) West Midwest South Number of Observations Number of Groups

All Single Filers -0.881 (.128) 0.017 (.049) 0.053 (.003) -2.503 (.332) 4.923 (.114) 0.800 (.166) 1.868 (.589) 0.684 (.232) 0.132 (.255) 0.176 (.223) 28,935 4,512

Single Itemizers All Married Filers 0.142 -1.173 (.333) (.095) -0.118 0.236 (.091) (.026) 0.001 0.021 (.002) (.002) -3.982 -2.874 (.512) (.276) 0.583 5.287 (.157) (.077) 0.625 0.196 (.278) (.107) 2.474 0.702 (.849) (.374) -0.357 1.181 (.486) (.229) -0.019 0.310 (.582) (.244) -0.222 0.142 (.476) (.204) 4,164 53,253 921 7,613

Note: Entries are coefficients from fixed-effects logits, with standard errors in parentheses. All models also include year effects. AGI = adjusted gross income; MTR = marginal tax rate.

29

Married Itemizers -0.027 (.258) 0.371 (.04) 0.005 (.002) -2.476 (.41) 0.241 (.123) 0.016 (.164) -0.171 (.554) 1.329 (.354) 0.325 (.371) 0.555 (.349) 9,210 1,662

Table 3: Fixed-Effects Regressions (OLS) of Itemized Charitable Contributions Variable Age 65+ Exemption Exemptions AGI/1,000 ln(1-MTR) Mortgage Interest Deduction Schedule C C * ln(1-MTR) West Midwest South Constant Number of Observations Number of Groups

Single Itemizers -0.013 (.078) 0.026 (.026) 0.001 (.0002) -1.893 (.134) 0.159 (.046) 0.146 (.091) 0.798 (.263) -0.167 (.148) -0.095 (.164) -0.094 (.143) 3.871 (.114) 23,918 8,250

Married Itemizers -0.085 (.036) 0.144 (.007) 0.0004 (.00004) -1.213 (.06) 0.022 (.023) -0.041 (.032) -0.122 (.089) 0.151 (.059) -0.156 (.057) -0.074 (.05) 4.590 (.053) 71,253 16,988

Note: Entries are fixed-effects regression coefficients with standard errors in parentheses. All regressions also include year effects. AGI = adjusted gross income; MTR = marginal tax rate.

30

Table 4: Fixed-Effects Regressions (2SLS) of Itemized Charitable Contributions Variable Age 65+ Exemption Exemptions AGI/1,000 ln(1-MTR) Mortgage Interest Deduction Schedule C C * ln(1-MTR) West Midwest South Constant Number of Observations Number of Groups

Single Itemizers -0.052 (.094) -0.134 (.034) 0.003 (.0006) 5.243 (1.392) 0.121 (.065) 0.399 (3.987) 2.195 (15.38) -0.159 (.469) 0.159 (.411) 0.051 (.206) 6.581 (.194) 21,213 7,753

Married Itemizers -0.289 (.158) 0.101 (.032) -0.0004 (.0002) -9.891 (3.549) -0.032 (.095) 21.906 (4.957) 74.296 (16.725) 0.248 (.253) -0.984 (.278) -0.610 (.217) 2.886 (1.091) 60,847 16,347

Note: Entries are two-stage least squares (2SLS) coefficients with standard errors in parentheses. All regressions also include year effects. AGI = adjusted gross income; MTR = marginal tax rate.

31