Improved adjustments for misreporting of tax return information used to estimate the national income and product accounts, 1977.

Parker, Robert P.

INFORMATION from tax returns is a major data source used to estimate the national income and product accounts (NIPA's). BEA has been aware of deficiencies in this information caused by misreporting by taxpayers and for many years has made adjustments to reduce the effects of misreporting on the NIPA's. However, the adequacy of these adjustments has become a matter of concern, reflecting increased interest in the extent of tax evasion and other aspects of the underground economy. Recently, information from studies by BEA and other Government agencies on the impact of the underground economy on NIPA source data provided the basis for substantial improvements in BEA's adjustments.

The improved adjustments are at present available only for 1977. They were incorporated in the input-output tables and preliminary revised NIPA's for 1977 that were presented in last month's issue of the SURVEY OF CURRENT BUSINESS. They will be extended to earlier and later periods for incorporation in the NIPA's in the comprehensive revision scheduled for the end of 1985.

Introduction Tax return information

BEA's adjustments for taxpayer misreporting pertain to information on Federal income tax returns, both business (corporate, partnership, and sole proprietorship) and individual, and on employment tax returns. The information on these returns is available to BEA in the form of statistical tabulations. For Federal income tax returns, the Internal Revenue Service (IRS) tabulates a sample of each major type of return. These tabulations, which cover most items reported on the return--types of income, expenses, etc.--as well as those on the supporting schedules, are published annually in Statistics of Income (SOI). For employment tax returns that employers file with State Employment Security Agencies, the wage and salary item is tabulated by the Bureau of Labor Statistics (BLS) in cooperation with the State agencies and is published annually in Employment and Wages.

In addition to the tax return tabulations used directly by BEA, tax return information enters the NIPA's indirectly by way of the quinquennial economic censuses, such as the census of manufactures. In these censuses, the Census Bureau uses tax return information to define the universe to be covered and to provide data on the employment, payroll, and receipts of small firms that are not sent a census report form. Misreporting

Because taxpayers use a variety of methods to evade taxes, misreporting affects both the income and expense items used to prepare the NIPA estimates. For example, a business may evade income tax either by underreporting receipts, overreporting expenses, or both. The underreporting of receipts leads to an understatement in net income in the tax return tabulation and in the related NIPA component, as does overreporting of expenses. At the same time, the overreporting of certain expenses, such as interest or rent, leads to an overstatement in the related NIPA component.

A business may evade social security and unemployment insurance taxes by, for example, paying wages "off the books." To avoid detection, it must underreport wages on both its income and employment tax returns. In turn, to avoid overreporting net income, it must underreport receipts or overreport expenses other than wages, or both. The underreporting of wages on the employment tax return leads to an understatement of wages and salaries in the NIPA's and, depending upon the item, the offsetting overreporting of expenses leads to an overstatement in a NIPA component.

Most misreporting clearly stems from the desire of the taxpayer to evade taxes. Some misreporting, however, results from taxpayers' errors or misinterpretation of tax laws. BEA's adjustments do not distinguish among reasong for misreporting. Use of tax return information in the NIPA's

Tax return information is used extensively in estimating the NIPA's. About one-half of charges against GNP, three-fifths of personal income, and 5 percent of GNP are derived from tax returns. The extent of the use of tax return information in the preliminary revised NIPA estimates for 1977 is shown in table 1.

Tax return tabulations are the best source available for estimating many NIPA components. The tax return information is well suited for use in the NIPA's because:

* The definitions of the items in the tax returns are close to those of the related NIPA components;

* The tax return information contains considerable detail on receipts and expenses that have counterparts in the NIPA's;

* The detail on receipts and expenses is internally consistent;

* The tax return information is comprehensive in its coverage of legal forms of organization and of industries.

Further, the accuracy of the information in the tabulations used for the NIPA's is judged by BEA to be superior to that of either actual alternatives or alternatives that might be developed at reasonable cost. Statistical surveys--the major alternative--are costly to conduct, a burden on respondents, and subject to several types of error, including sampling errors, bias, and nonresponse. In particular, information on the income of individuals is difficult to collect in surveys. Many persons misstate income because of inaccurate recall or lack of ready access to financial records. Some who misreport income on their tax returns will also misreport it in a survey. Others will refuse to answer questions about income in a Government survey.

One reason for BEA's judgment about the accuracy of the tax return information used in the NIPA's is that, for all except one component, BEA uses business tax returns, which in general, are subject to substantially less misreporting than are individual tax returns. Wages and salaries, for example, are estimated from the employment tax returns filed by business, rather than from the income tax returns filed by employees. Table 2 lists the NIPA components derived directly and indirectly from tax return information and the sources of that information. Improved adjustments

The improved adjustments incorporate newly available information about the extent of underreporting of income and about the failure to file income and employment tax returns. As shown in table 3, they are considerably larger than the previous adjustments. The adjustment for nonfarm proprietors' income was revised from $13 billion to $59 billion, and adjustments of $22 billion for personal consumption expenditures and $11 billion for wages and salaries were introduced.

The BEA adjustments are designed for a specific purpose--to correct for the effects of taxpayer misreporting in the tax return tabulations and economic census data used in the NIPA's. The BEA adjustments do not measure all income on which taxes are evaded. The major reason is that not all income types reported on tax returns are used in estimating the NIPA's. Neither od the BEA adjustments measure the underground economy. For example, they do not cover illegal activities, which are part of the underground economy but are excluded from the NIPA's.

Previous Adjustments

The previous adjustments for misreporting, which are incorporated in the published NIPA time series, are for five components: nonfarm proprietors' income, rental income of persons, net interest, capital consumption allowances, and corporate profits before tax. For net interest and capital consumption allowances, the adjustments were made only to the noncorporate parts. Adjustments for these two components and for nonfarm proprietors' income and rental income of persons were prepared from information provided in the IRS Taxpayer Compliance Measurement Program (TCMP). The adjustment for corporate profits before tax was prepared from IRS information on the additional tax assessments resulting from audits.

No adjustments were made in the remaining components listed in table 2. The noncorporate parts of other labor income, business transfer payments, and change in nonfarm inventories were not adjusted because information from the TCMP showed that they were not needed. The corporate parts of these components, the corporate parts of capital consumption allowances and of net interest (including that of investment companies, which also is in personal consumption expenditures), and personal dividend income were not adjusted because information on income and expense items was not available from the corporate audits. Wages and salaries were not adjusted for the use of information from employment tax returns because IRS studies did not indicate that adjustments were needed. Finally, components for which the estimates were derived from the economic censuses were not adjusted because BEA overlooked the need to make adjustments. Adjustments based on the TCMP

The previous adjustments for 1977 were based on information from the 1976 TCMP, which was an extensive audit of a random stratified sample of individual income tax returns. The audits included all types of income reported on individual tax returns and the detailed income and expense items reported by sole proprietorships.

In general, the BEA adjustment for each NIPA component was based on an "audit radio"--the ratio of the amount that the IRS auditors determined was misreported for an item to the amount originally reported on the return. The adjustment was calculated by multiplying the audit ratio by the total amount for the item as tabulated in the 1977 SOI. This procedure assumed that (1) the extent of misreporting in 1977 was the same as in 1976, and (2) the extent of misreporting for nonfarm partnerships, which were not audited in the TCMP, was the same as that for sole proprietorships.

Nonfarm proprietor's income.--Nonfarm proprietor's income was adjusted up $14.7 billion for 1977. For each industry, separate adjustments were calculated for gross receipts and for total expenses (gross receipts less net income) by multiplying the TCMP audit ratios by the totals of receipts and of expenses for sole proprietorships and partnerships from the 1977 SOI. The adjustment for nonfarm proprietors' income was then obtained as the difference between the adjustments for gross receipts and for total expenses. The adjustment is included in line 2 of NIPA table 8.10, which appears in the July issue of the SURVEY and shows the relationship between the SOI and NIPA measures of nonfarm proprietors' income.

Rental income of persons.--Rental income of persons was adjusted down $1.0 billion for 1977. The adjustment pertains to royalties as reported on individual tax returns and rental income from nonfarm nonresidential properties. For royalties, the adjustment added $0.3 billion; it was calculated by multiplying the TCMP audit ratio for royalties by the corresponding royalties figure from SOI. For rental income from nonfarm nonresidental properties, which is obtained mainly by subtracting rents received by business from rents paid by business, the adjustment subtracted $1.3 billion. For sole proprietorships and partnerships, separate adjustments were calculated for receipts and for payments by multiplying the TCMP audit ratios by the SOI totals. For corporations, no adjustments were calculated because the necessary information was not available. The downward adjustment reflects a larger amount of underreporting of rents received than overreporting of rents paid.

Net interest.--Net interest was adjusted down $0.5 billion for 1977. The adjustment, which pertains to monetary interest paid by nonfarm sole proprietorships and partnerships, was obtained by multiplying the TCMP audit ratio by the SOI total. No adjustment was made for monetary interest received by nonfarm sole proprietorships and partnerships because, except for a small amount of interest received by noncorporate businesses engaged in financial activities, all such interest in the NIPA's accrues to persons rather than to business.

Capital consumption allowances and capital consumption adjustment (CCAdj).--The estimate of capital consumption allowances, which is derived mainly from depreciation as tabulated in the SOI, was adjusted down $1.7 billion for 1977. The adjustment was calculated by multiplying the TCMP audit ratio for depreciation by the SOI total for nonfarm sole proprietorships and partnerships. The $1.7 billion adjustment is reflected also in the CCAdj because the CCAdj is obtained as the difference between capital consumption allowances with CCAdj, which is not based on tax return information, and capital consumption allowances. Adjustment based on corporate audits

The previous adjustment for misreporting increased the estimate of corporate profits before tax by $12.2 billion for 1977. It is included in line 2 of NIPA table 8.12, which shows the relationship between the SOI and NIPA measures of corporate profits. It was based on IRS information on additional tax assessmewnts for 1977, and was calculated in two parts: for corporations reporting a profit and for those reporting a loss.

For corporations reporting a profit, the adjustment was calculated as follows: (1) The value of the IRS auditor's recommended assessment per return, classified by corporate asset size, was reduced by the overall ratio of actual settlements to recommendations to derive actual settlements. (2) The estimates of actual settlements were "blown up" to universe totals by multiplying them by the number of corporate tax returns with income, by size class, as published in SOI. (3) The estimated universe totals of settlements were divided by the applicable corporate tax rate to obtain the estimate of additional profits.

For corporations reporting a loss, the adjustment was calculated by multiplying total losses, as published in SOI, by an estimate, based on fragmentary information from IRS, of the percentage by which losses were reduced during audit.

New Information on Misreporting

Studies at IRS and the Census Bureau have provided new information about the extent of underreporting on tax returns and of the failure to file income and employment tax returns. The information about the extent of underreporting was from IRS research on the TCMP auditors' ability to detect unreported income in 1976, and from an IRS examination of the underreporting of wages and salaries on the employment tax returns in the 1979 TCMP. Information on the extent to which businesses and individuals failed to file tax returns was developed from the Census Bureau's evaluation of the coverage of its surveys. Underreporting in income tax returns

The starting point of the IRS research was the TCMP-Information Returns Program (the TCMP-IRP). In this study, conducted after the 1976 TCMP audits were completed, IRS used information returns to assess the auditors' ability to detect unreported income. Information returns are reports that must be filed with IRS by payers of certain types of income--for example, form W-2 for wages and salaries paid by employers and form 1099 for interest paid by banks. For a sample of tax returns included in the 1976 TCMP, IRS compared the amount of each type of income reported on these forms with that reported by the taxpayer and established by the auditor. The study showed that, for the income types covered--mainly wages and salaries, interest, and dividends--the TCMP auditors detected $1 of every $4 of unreported income.

IRS conducted additional research on income types not included in the TCMP-IRP study. For business income (mainly rental income and incomes of partnerships, sole proprietorships, and small business corporations), IRS concluded that auditors detected $1 of every $3-1/2 of unreported gross profits--that is, gross receipts less cost of goods sold. Because IRS audit studies indicated that businesses that understated receipts also understated cost of goods sold in order to avoid reporting a suspicious sales/gross profits relationships, IRS further concluded that auditors detected similar proportions of both unreported gross receipts and unreported cost of goods sold. In contrast, for deductions--that is expenses, other than cost of goods sold--IRS concluded that the auditors were able to detect all overreporting. From these conclusions, it can be inferred that auditors detected a higher proportion of misreporting for net income--that is, gross profits less deductions, an aggregate that approximates the concept used in the NIPA's--than for gross profits.

The IRS view about the extent of underreporting of business recepits tends to be confirmed by evaluation studies of the 1977 economic censuses. For these studies, the Census Bureau conducted special surveys that collected receipts for samples of small firms for which tax return information had been used in the economic censuses. The receipts as reported in these surveys were compared with receipts as reported on a firm's tax return. These comparisons showed that small firms consistenly reported larger receipts to the Census Bureau than they reported to IRS. The underreporting of receipts to IRS indicated in the surveys was about the same as the total underreporting indicated by the IRS research. Underreporting on employment tax returns

In the 1979 TCMP, IRS audited the reporting of wages and salaries by sole proprietorships and by small corporations (those with assets of less than $10 million) on the return used to report Federal unemployment insurance taxes (form 940). The auditors detected both underreporting on, and failure to file, these returns. The information on wages and salaries reported on these returns is essentially the same as that reported to the State Employment Security Agencies (and used by BEA for the NIPA estimates). Because of administrative links between the Federal and State unemployment insurance programs, it seems likely that the audit ratio for the Federal returns in the 1979 TCMP is applicable to the returns filed with the State agencies. It should be noted that IRS has not assessed the auditors' ability to detect misreporting on the employment tax returns, as it did for income tax returns in the TCMP-IRP study. Nonfiling of tax returns

For incomes earned by persons in 1972 and in 1977, the Census Bureau evaluated the reporting in the annual income supplement to the Current Population Survey (CPS). For each year, Census prepared "exact-match" files of CPS records, selected items for individual income tax returns, and earnings and benefits from Social Security Administration (SSA) records. From these files, tabulations were prepared of the incomes of "nonfilers," that is, persons who did not file an income tax return, but who earned income as evidence by the information they supplied to the CPS. (Persons who filed a tax return were classified as "filers," even if they did not report to IRS all the income types that they reported to the CPS.) BEA used the tabulation of nonfarm sole proprietorship and partnership income.

Another evaluation study of the 1977 economic censuses, which provided an indirect check on the procedures used in the exact match, tended to confirm the extent of nonfiling of income tax returns by nonfarm sole proprietorships and partnerships. Information from the study established that the extent of nonfiling in the exact-match files was consistent with the shortfall in the universe established by the Census Bureau on the basis of tax return information provided by IRS.

Information about the failure of sole proprietorships and partnerships to file employment tax returns came from an examination by BEA of this evaluation study together with the exact match. The study identified firms that failed to file reports in the economic censuses, that is, firms that were not in the universe because they had filed neither income nor employment tax returns. The exact match identified firms that had not filed income tax returns. The examination showed that employers who did not file an income tax return also did not file an employment tax return.

Improved Adjustments

The information described in the previous section was used to develop or improve adjustments for five NIPA components. These adjustments--which are for wages and salaries, nonfarm proprietors' income, rental income of persons, personal consumption expenditures, and fixed investment--are shown in table 3 in the column labeled "improved adjustment."

The adjustments for the noncorporate parts of net interest and capital consumption allowances and for corporate profits before tax were not revised because no additional information was available. For these adjustments, the improved and previous adjustments shown in table 3 are the same. For the remaining components or parts of components derived from tax return information, either new information indicated that adjustments were not needed or information was not available on which to base an adjustment.

The contribution of each type of new information to the revisions in the adjustments is shown in table 4. Under the heading "income tax returns," the column labeled "filer adjustment" shows the revision based on the information on underreporting provided by the IRS research that was related to the 1976 TCMP-IRP study. The column labeled "nonfiler adjustment" shows the revision based on the information on nonfiling provided by the Census Bureau exact-match study. Under the heading "employment tax returns," the column labeled "filer adjustment" shows the revision based on the information on underreporting provided by the 1979 TCMP audit, and the column labeled "nonfiler adjustment" shows the revision based on the information provided by BEA's examination of the evaluation study and the exact match. Wages and salaries

The adjustment for misreporting increased wages and salaries $11.3 billion. The filer adjustment contributed $7.6 billion and the nonfiler adjustment, $3.6 billion.

Filer adjustment.--The filer adjustment was based on the 1979 TCMP audit of employment tax returns. It was calculated in two steps and provided separate adjustments for wages and salaries paid by nonfarm sole proprietorships and partnerships and by corporations. The first step consisted of applying an audit ratio to BEA's estimates of wages and salaries. For sole proprietorships and partnerships, the ratio was form the TCMP for sole proprietorships. For corporations, BEA derived audit ratio, because the TCMP audit ratio covered only small corporations. The ratio was based on the assumption that large corporations fully report wages and salaries on employment tax returns. It was calculated by dividing the amount of wages and salaries that the TCMP auditors determined was underreported by small corporations by the BEA estimate of wages and salaries for all nonfarm corporations. It was assumed in the derivation that (1) the audit ratios for 1979 apply to 1977, and (2) the audit ratio for nonfarm sole proprietorships applies to partnerships.

In the second step, BEA made an allowance for the likelihood that the TCMP auditors did not detect all underreporting. It was apparent that the allowance should be at least as much as that found for underreported income in the TCMP-IRP study ($1 detected of every $3-1/2 of unreported income) and that the allowance should compensate for the TCMP auditors' lack of experience with employment tax returns. Therefore, an allowance of $1 of every $5 applied to sole proprietorships and partnerships and to small corporations.

Nonfiler adjustment.--The nonfiler adjustment was based on BEA's finding that the sole proprietorships and partnerships (with employees) who did not file an income tax return also did not file an employment tax return. It was calculated indirectly because information on wages paid by nonfiles was not available in the exact match. Starting with the net income of nonfilers in the exact match, BEA first estimated receipts. The estimate was made by multiplying the net income of nonfilers in each industry by the ratio of receipts to net income for sole proprietorships and partnerships with income from the 1977 SOI. The estimate of wages and salaries of nonfilers was then made by multiplying the estimated receipts by the ratio of payroll to receipts for small firms. For industries included in the 1977 economic censuses, the ratio was calculated using information on small firms. For other industries, it was calculated using census SOI information for small sole proprietorships and partnerships.

This adjustment is limited to sole proprietorships and partnerships that did not file an income tax return. An adjustment for sole proprietorships and partnerships and small corporations that filed an income tax return, but not an employment tax return, is included in the filer adjustment. Nonfarm proprietors' income

The revision in the adjustment increased nonfarm proprietors' income $46.5 billion. The filer adjustment contributed $38.5 billion and the nonfiler adjustment, $8.0 billion.

File adjustment.--The filer adjustment was based on the IRS conclusion that the TCMP auditors detected $1 of every $3-1/2 of unreported gross profits. Because the TCMP audit ratio used in BEA's previous adjustment accounted for gross profits detected in the audit, the revision reflects only the undetected amount. The revision was calculated by multiplying the 1976 TCMP audit ratio for gross profits by the total for sole proprietorships and partnerships from the 1977 SOI and multiplying the result by 2-1/2.

Nonfiler adjustment.--The nonfiler adjustment was based on the exact match. It is the total of net income estimated in the CPS for nonfarm sole proprietorships and partnerships that failed to file income tax returns. Retanl income of persons

The revision in the adjustment had no effect on rental income of persons.

Filer adjustment.--The filer adjustments to royalties and to rental income from nonfarm nonresidential properties were revised to reflect the IRS conclusion that the TCMP auditors detected $1 of every $3-1/2 of unreported income. The improvements, which each amounted to $0.7 billion, were offsetting. For royalties, the revision was calculated by multiplying the previous adjustment, which represented the amount detected by auditors, by 2-1/2. For rental income from nonfarm nonresidential properties, which--as noted earlier--is obtained mainly by subtracting rents received by business from rents paid by business, the revision was calculated by multiplying the previous adjustment for rents received by 2-1/2. The adjustment for rents paid was not revised because IRS concluded that the TCMP auditors detected all overreporting of deductions.

Nonfiler adjustment.--No improvement was possible because of lack of information. Personal consumption expenditures

The adjustment increased personal consumption expenditures (PCE) $21.6 billion. The filer adjustment contributed $10.9 billion and the nonfiler adjustment, $10.8 billion.

PCE is affected by the misreporting of the tax return information that is used in the economic censuses. As noted previously, this information is used by the Census Bureau to define the universe and to provide data on small firms that are not sent a census report form. Consequently, the census figures are understand because busineses that do not file tax returns are not included and because some small firms misreport on their tax returns.

The misreporting of concern to BEA is that of sales as compiled in the censuses. The effect of the misreporting on PCE was determined in preparing the input-output (I-O) tables for 1977, which provided the basis for the preliminary revised NIPA's. Sales figures from the censuses were used in the I-O tables to establish output by industry, part of which is purchased by persons. The effects of misreporting on sales and on PCE differ; the effect on PCE is much smaller than the effect on sales for two reasons. First, misreporting of sales by trade firms, which is large, does not directly affect industry output. Second, only part of the misreported sales was purchased by persons.

The adjustment to PCE was derived by BEA in three steps. (1) Adjustments were estimated for the Census Bureau's sales figures using information from the IRS studies and the exact match. (2) These adjustments were used to estimate adjustments to output by industry in the I-O tables. (3) From the adjustments for industry output, adjustments for PCE were obtained.

Adjustments to sales.--The adjustments to sales are shown by industry in table 5. The adjustments for underreporting by small firms, shown in the table as the "filer adjustment," were derived using information from the IRS studies. For each industry, the adjustment was calculated by (1) multiplying sales of small firms, both corporate and noncorporate, that the Census Bureau derived from tax return information by the 1976 TCMP audit ratio for nonfarm sole proprietorships, and (2) multiplying that result by 3-1/2 to alow for the failure of the TCMP auditors to detect all underreporting. For retail trade and services, in which small proprietorships are more than proportionally represented, the audit ratio was increased, because the TCMP indicated that small proprietorships understate receipts to a larger degree than other proprietorships.

The adjustments for the failure of businesses to file tax returns, shown in the table as the "nonfiler adjustment," were derived from the exact match. For each industry, the sales of nonfilers were estimated by multiplying the net income of nonfilers in the exact match by the ratio of receipts to net income for sole proprietorships and partnerships with income from the 1977 SOI.

Both adjustments were prepared at a broad industry level and then were disaggregated to the more detailed industry level of the I-O tables using information from the economic censuses.

Adjustments to industry output.--In general, for industries in the I-O tables, output consists of sales plus change in inventory, and the adjustment to output is the same as that for sales. For trade, output is defined as the margin on sales, that is, sales less cost of goods sold. The adjustment for the output of trade was obtained by multiplying the adjustment for sales by the margin rate.

Adjustments to PCE.--The adjustments to PCE depended on how much of the adjustments to industry output were purchased by persons. They were calculated for each industry by multiplying the industry output adjustment by the ratio of the PCE portion of the industry's output to the industry's total output. Fixed investment The adjustment increased fixed investment (specifically, producers' durable equipment--both residential and nonresidential--and mobile homes) $0.2 billion; the filer and nonfiler adjustments each contributed $0.1 billion. The adjustments were calculated in the same way as those for PCE.

Evaluation of the Improved Adjustments

This section discusses possible errors in the improved adjustments for 1977 and the problems involved in extending them to earlier and later periods for incorporation in the NIPA's in the comprehensive revision scheduled for the end of 1985.

Because the adjustments are based on information that is incomplete and, in some cases, of questionable quality, they are subject to substantial error. In the aggregate, however, it appears that the adjustments to GNP, charges against GNP, and personal income are as likely to be overstated as understated. Omitted adjustments

One kind of error stems from the omission of adjustments for which information is not available. Filer adjustments were not made for the NIPA components that are derived from detailed income and expense items reported on corporate tax returns. Such adjustments would probably reduce charges against GNP, and personal income, because, in most cases, the items that have not been adjusted are deductions that IRS studies show tend to be overreported. Nonfiler adjustments were not made for corporations. Such adjustments would probably slightly increase GNP and charges against GNP.

For noncorporate nonfilers, adjustments were made only to wages and salaries, nonfarm proprietors' income, PCE, and fixed investment. Adjustments to noncorporate parts of other components of charges against GNP and personal income, if they could be made, would probably be small and net close to zero.

On balance, the adjustments to charges against GNP and personal income are probably overstated, because the omitted corporate filer adjustments outweigh the omitted nonfiler adjustments. The adjustment to GNP is understated to the extent that corporations do not file tax returns. Filer adjustments

Filer adjustments are subject to several kinds of error. First, information is insufficient to evalute the IRS conclusion that auditors detected $1 of every $3-1/2 of unreported businesses income. Thus, to the extent that the adjustments were based on this conclusion, they are subject to error of unknown size and direction. Second, because it was assumed that the 1976 audit ratios apply to 1977, the adjustments are misstated to the extent that tax evasion relevant to the adjustments changed from 1976 to 1977. Third, the adjustments are probably overstated because, contrary to the assumption that the audit experience of partnerships was the same as that for sole proprietorships, misreporting on a partnership tax return is less likely.

The filer adjustments for PCE and fixed investment are subject to additional errors. First, these adjustments assume--contrary to IRS evidence that the audit ratio for small corporations is much smaller than that for sole proprietorships--that the audit experience for small corporations was the same as for sole proprietorships. This IRS evidence could not be used in calculating that adjustment because sales figures for small firms by legal form of organization were not available from the economic censuses. Second, these adjustments assume, as may not be the case, that the extent of unreported income not detected by auditors is the same across industries. This assumption was necessary because there is no industry information. The assumption may lead to error in the adjustments because the proportion of sales to persons and business on capital account varies across industries. The size and direction of the error is unknown.

Given what is known about the errors described above, the adjustments to the NIPA aggregates are as likely to be overstated as understated. Nonfiler adjustments

Several kinds of error stem from the use of exact-match studies, in which responses to the CPA are critical. Some CPS respondents may not have properly indentified themselves as self-employed. To the extent that these respondents failed to file income tax returns, the adjustments are understated. The incomes imputed by the Census Bureau for the substantial number of respondents who identified themselves as self-employed but did not report their incomes may be too high. This probability is suggested by a BEA comparison of the imputed incomes with the reported incomes. To the extent that the imputations are too high, the adjustments are overstated.

Three kinds of error relate to industry classification. First, evidence suggests that some CPS respondents incorrectly classified their businesses as nonfarm. Such misclassification lead to overstatement of the adjustments. Second, some CPS respondents, to cover up their involvement in illegal activities, claimed that they were engaged in a legal business. Such responses lead to overstatement. Third, any industrial classification errors within nonfarm activity are reflected in additional errors in the adjustments to PCE and fixed investment. The size and direction of such errors and unknown.

As was the case with the filer adjustments, the nonfiler adjustments to the NIPA aggregates are as likely to be overstated as understated. Time series estimates

Less information will be available for extending the adjustments to years before and after 1977 than was available for the 1977 adjustments. It is anticipated that an exact-match study for 1982 will provide information on nonfiling and that TCMP's and corporate audits will provide information on underreporting. However, the TCMP-IRP results led IRS to change its procedures for noncorporate audits after 1976, and it appears that since then the TCMP auditors have been detecting more unreported income than previously. Thus, it will be necessary for BEA to adapt its methodology to use the more recent TCMP's. To the extent that more up-to-date information is not available, post-1977 adjustments will be extrapolations that hold the proportions of underreported income constant. The adequacy of this procedure depends on the extent to which misreporting is stable. It may be that the types of misreporting for which the NIPA's should be corrected are more stable than some other types; that is, misreporting may be more stable on business tax returns than on individual tax returns.

The adjustments for years before 1977 will need to reflect the less widespread use of tax return information in earlier years. Before 1959, tax return information was not used to estimate major parts of nonfarm proprietors' income. Also, before 1963, it was not as widely used in the economic censuses.