1505
1506
through time. However, it can be that responses to changed tax policies through time can be different from those predicted by a cross-section analysis.^
For the purpose of estimating the total giving in the period to which the cross-sectional data apply, the econometric equations are much less subject to qualification. For this purpose, the equations can be regarded as a system of structural relationships whose parameters provide the means for estimating total giving, particularly that of the non-itemizers for whose unreported giving the equations provide a multi-factor basis for imputation. Separation of the effect of specific factors is not as critical to the estimate.
In presenting an estimate of the existing level of total giving as a reference for measuring the effects of tax changes, Feldstein uses the estimate based on the assumption of the same price elasticity for all income classes. The procedure is described as follows:
In 1970 total giving on itemized returns was $13.0 billion. The remaining
$4.3 billion is our estimate of the total giving by taxpayers who filed non-itemized returns (i.e. who used the standard deduction). This amount is estimated for each non-itemized return [using the multiple estimating equation] and aggregated with the appropriate weights.*^
The estimate of $17.3 billion for 1970 translates into one of $19.9 billion for 1972, as presented above in Table 1. This assumes a 15 percent growth in giving over the two-year period.
The findings from household surveys reported by the Survey Research Center, like those of Feldstein, suggest that for families of comparable incomes, non-itemizers made significantly lower contributions than did Itemizers. This is summarized In Table 2 for families In the survey with 1973 incomes of less than $30,000, the Income range from which the bulk of non-itemized contributions originated. The differences described in Table 2 are of the same order of magnitude as those implied by Feldsteln's elasticities applied to the "price" (complement of tax rate) differences between itemizers and non-itemizers.
Table 2
Average Giving of Itemizers and Non-itemizers
to IJOW- and Middle-Income Classes1973
Average Giving Per Family Ratio: Non-itemizers Income Class Itemizers Non-itemizers to Itemizers
Less than $ 4,000 $119 $ 6 9 0.58
$ 4,000-$ 7,999 215 89 0.41
$ 8,000-$ 9,999 314 117 0.37
$10,000 - $14,999 407 201 0.49
$15,000-$19,999 600 329 0.55
$20,000 - $29,999 800 354 0.44 Source: Morgan, Dye, and Hybels, "Results From Two National Surveys of Philanthropic
Activity," Table 24.
This examination of the two types of study illustrates the value of multiple
research approaches. In both the econometric and household survey studies the
1507 primary objective was the description of behavioral patterns and the analysis of the family characteristics that Influence giving. Estimation of total giving was a carefully qualified secondary product of the analysis. In the process, however, both types of study have given us a much clearer picture of the behavior of non-itemizers. In this way they have laid the groundwork for much improved estimates of total giving. The implication to be drawn from these studies is that non-itemized giving accounts for a much smaller part of total personal giving than many may have believed it to be.
Whereas the Feldstein and Survey Research Center estimates are based on extrapolations of micro-economic data, that made by Nelson is based on data on aggregate giving and income. Nelson begins with aggregate itemized contributions as reported by the IRS. To this he adds an estimate of the total contributions of non-itemizers and non-filers. This estimate is arrived at by applying a specified giving percentage to the aggregate Adjusted Gross Income of everyone other than those itemizing contributions.
The percentage chosen to apply to the income of non-itemizers and non-filers is critical to the estimate. Nelson adopted the percentage estimated by C. Harry Kahn as presented in his study, Personal Deductions In The Federal Income Tax. Kahn based his estimate on an analysis of changes in contributions percentages across years when changes were made in the standard deduction. Kahn's latest estimate (his series ended in 1954) reflected the 1948 liberalization of the standard deduction and resulted in an upward revision in the earlier percentages applied to non-ltemlzers and non-filers.
No broad changes were made in the standard deduction from 1954, the final year in Kahn's series, through 1970. Accordingly, Nelson applied an unchanging percentage rate to his estimates of the aggregate AGI of everyone other than itemizers over the 1960 to 1970 period, and in his original (January 1975) report, through
1972. The assumption of an unchanging percentage giving rate for non-itemizers is open to question. However, a number of offsetting trends were observed, which led to the belief that the constant percentage assumption would yield tolerably accurate estimates of non-itemized giving.^ *
For the reference year 1972, however, the assumption of unchanging giving percentage became much less tenable. Itemized and non-itemized giving in that year reflected the broad liberalization of the standard deduction contained in the Tax Reform Act of 1969. After more than two decades of unbroken increase, the number of returns containing itemized contributions fell sharply in 1972. This development was explicitly Included in a recent upward revision of the estimate for 1972,^^ and drew upon information provided by the Feldstein and Survey Research Center studies. Both studies suggested that the giving rate for the "new" non-itemizers (that is, itemizers in 1970, non-itemizers in 1972) was higher than that for the " o l d "
non-itemizers (that is, non-itemizers in 1970 and 1972).
The effect of this revision was to increase the 1972 estimate of total personal giving from $18.4 billion to $19,8 billion, or by 7.6 percent. The original giving percentage applied to total "non-itemized" AGI was 1,44 percent; the revised rate was 1,84 percent. Both estimates use the "contributions deduction on income tax return" definition of personal giving and so may be directly comparable to the estimates contained in the other two kinds of studies.
Giving by Upper-Income Families
The second component of the estimate to be examined is the giving of upper-income families, those having 1972 pre-tax incomes of $30,000 or more. One of the more generous estimates (Survey Research Center "Best Aggregate Estimate") credits upper-income givers with 29 percent of the total 1973 personal giving, or
1508
$7.4 billion of $25,7 billion. One of the less generous estimates (Feldstein, Separate Elasticities for four income classes, 1962 IRS data) credits upper-income givers with about 15 percent of total 1970 personal giving, or approximately $2.8 billion of
$18,1 billion.
Illustrative of the problem encountered in estimating the contributions of upper-income givers is the effort by the Survey Research Center to expand its sample survey findings to an estimate for the total population. In its first attempt, it expanded its survey results with no adjustments for either the overreporting of contributions or the underreporting of income. It estimated a total of $13.8 billion of giving by families reporting income of $30,000 or more, 39 percent of total estimated giving of $35.1 billion for all income groups.
It found that the income data in one of the two samples used in the survey was not reliable for reported incomes under $50,000 because of "pervasive under- reporting of income." This produced a misleadingly high percentage of giving to income. When this was expanded by the proportion of such apparent income recipients in the population, the result was a major overestimation of total giving.
With the removal of the unreliable (Census) sample of reported incomes under
$50,000 from the estimating process, total estimated giving declined from $35.1 to
$32.4 billion, of which $12.6 billion, or 39 percent, came from incomes of $30,000 or more.
The next adjustment took account of the discrepancy between a family's giving as reported to the survey taker and that reported on Its Income tax return
("amount deducted"). This adjustment applied to those respondents for whom both amounts were available and where the amount deducted on their income tax returns was different from, and usually lower than, the amount reported to the survey taker. The adjustment thus takes partial account of this particular kind of overstate- ment though it retains any overstatement present in the income tax deduction. The adjustment reduces total estimated giving from $32.4 to $29.3 billion in the estimates with the aforementioned correction for income overreporting.^^ Even with the adjustments, however, incomes over $30,000 account for $11 billion or 37 percent of the total.* *
One final adjustment was made because the earlier procedures "In retrospect overestimated the numbers [of households] in the income groups over $50,000."^^
If actual numbers of tax returns over $50,000 are u ^ d , Instead of the original blown-up numbers, total giving is estimated at $25,7 billion, down from $29,3 billion. Giving by families having $30,000 or more in 1973 income totals $7.4 billion, down from $11 billion, and accounts for 29 percent of the total, down from 37 percent.
Although the estimated giving of families having income of $30,000 or more was reduced from $13,8 to $7,4 billion, it is probably still too high. It will be recalled that this $7,4 billion of 1973 giving was roughly twice the $3.76 billion In 1972 giving reported by the IRS for families with incomes of $30,000 or more. After allowing for a 9 percent growth from 1972 to 1973, the discrepancy still exceeds
$3 billion.
In summary, the Survey Research Center's "Best Final Estimate" of $25.7 billion for 1973 probably still reflects some underreporting of income and overreporting of contributions, both biases leading to an overestimation of total giving. Having been reduced by $9.4 billion from the first estimate, probably an additional $2 billion to
$3 billion of " f a t " remains to be rendered were it possible to carry to completion the above-described refinements. If so, the 1972 "perfect" estimate, allowing for 9 per- cent growth from 1972 to 1973, would then be In the range of $20.8 to $21.7 billion. This would be reasonably comparable to Feldsteln's preferred estimate of
$19.9 billion and to Nelson's revised estimate of $19.8 billion.