STATEMENT OF NELSON MCCLUNG, ASSISTANT DIRECTOR, OFFICE OF TAX ANALYSIS, DEPARTMENT OF THE TREASURY

Mr. MCCLUNG. Let me begin by expressing my appreciation to this committee for its interest in the distributive impact of Government policies and its awareness that these impacts cannot be elucidated without wealth data for representative families. As requested, I will first describe the data available, give some examples of our requirements, and then consider how our requirements are being met.

Basically, what is wrong in our present situation is that the distributive analyses of policies which we are called upon to prepare entail too great an exercise in imagination. Sometimes we think that we should have prepared for our work by taking courses in creative writing rather than in statistics and economics. I hope that you do not know how bad things are because they cannot be improved soon and, yet, we want you to continue to believe the estimates that we give you.

We have two primary sources of family wealth data. The first is two surveys conducted for the Board of Governors of the Federal Reserve System by the Census Bureau in 1963 and 1964. The years chosen turned out not be the best; the stock market was off trend. We adjust corporate share values to put them back on trend but cannot adjust for the transactions induced by the departure from trend. Worse yet, since 1968, the trend has changed. We know a good bit about the implications of the change in trend for family shareholding. For example, families with lesser total wealth holdings have sold off their shareholdings disproportionately as compared to families with greater total wealth holdings. We can make only very crude adjustments for this and other changes in the compositions of wealth holdings since the end of 1962 and 1963.

The Board sample is very small, about 2,500, but it was drawn very efficiently. Essentially, the frame for the sample was personal income tax returns. This meant that families with property income and, hence, property could be identified and, because they are relatively rare, oversampled. The statistical reliability of the survey is quite high. With all of its imperfections, it is the best survey of family wealth ever conducted in this country and perhaps as good as any ever done anywhere. Nevertheless, these surveys have their problems, apart from their age. One of the families drawn in the sample was that of John F. Kennedy, then President of the United States. He was replaced in the sample by someone who could not have been anything like the same. Whether from constraints on sample size or the units that could be included or other reasons, the survey missed significant amounts of wealth.

Another primary source of family wealth data is the two surveys of economic opportunity. These surveys were conducted by the Census Bureau for the Office of Economic Opportunity in 1966 and 1967. Each survey interviewed about 30,000 households, some two-thirds of whom were common to the surveys. The nonwhite poor were oversampled. The sample frame was that used for the current population survey. The current population survey sample frame essentially is a list of addresses. It is a useful frame for measuring attributes that are the primary concern of the survey, work experience of the popula

tion by geographical area. The sample frame is a quit for measuring family wealth. It is one thing to send ir a rich neighborhood; it is another for them to find r surveys of economic opportunity measured reasonably parently the wealth of families within income below $ with larger incomes are underrepresented in the s families with larger amounts of wealth evidently und holdings. Reweighting the samples for the missing ric reported wealth for underreporting cannot be done fidence. We do it, but only because the errors in polic not doing it are greater than the errors from at best to rectify the sample.

Well, this essentially is what we have to work wit of Tax Analysis we are asked at least once a year wh additional revenue raised were State and local tax-e terest includable in taxpayer adjusted gross income. question we need to know, first, who holds State and 1 bonds and, second, for representative holders what a incomes. Only if we know the taxable incomes of hol termine the tax rates at which this additional income personal income tax base. For imputing State and loc to our personal income tax model data base, we rely of Governors' 1963 and 1964 surveys. We know tha amount held by families and the distribution have cha and we adjust the survey data for these changes as b we are 15 years away from the date of the survey data economic opportunity are little help because the majo of State and local bonds are so underrepresented in th tually, the time must come when the quality of the justify asking the question.

Regularly we are asked to estimate the revenue cons ing capital gains accrued up to but unrealized at tir even make a decent start on an estimate we need to kn lies hold what assets. Given that, we can infer the each particular holding by assuming that each family age experience for all holders, including nonfamily h of asset. We are fortunate here in having estate tax re are never more than a few years out of date and so covered the populations that would be affected by th changing the law. However, to make a good estimate revenue effect and the distribution of additional ta need to know not just current market value but basi value and book value by type of asset for individual f which can be collected, and, if the samples were no 1 essary for reasonable accuracy and were selected efficie tion effort would not impose a great burden. These are we require more frequently than once every several y

From time to time we are asked to estimate the an that accrues annually to persons covered by pension usually is not taxing this income but evaluating the personal income tax as it stands currently. Pension private and Government schemes, is now a very sub total family wealth. The pension income which pro

included in the annual incomes of persons covered by the plans is the annual increase in their equity in aggregate pension wealth. Annual increases in pension equities are much of personal saving and apparently have a strong impact on saving in other forms. We have no satisfactory way of addressing these questions. The data required could be collected. Families could be asked about coverage and, given coverage, the names of the plans; the plans could then be asked for information on employer contributions and plan earnings. Pension saving then could be related to other saving. There is a problem, of course, in defining pension equities.

For several years there have been discussions of a new survey series additional to and supplemental to the current population survey and the survey of consumer expenditures. In the last year the Office of the Assistant Secretary for Planning and Evaluation, Department of Health, Education, and Welfare, and the Census Bureau have begun planning for a new survey series. This new survey series is called the survey of income program participation and its primary purpose would be that of measuring how families are affected by tax and grant programs. To do this, the new survey would have a sample frame, interview and editing procedures whose primary objective would be the construction of accurate income and wealth microdata files. Pilot survey results should be available in 1978 and 1979 but regular periodic surveys will not begin until 1980 or 1981. This survey series offers us an opportunity to collect the sort of data which we need for tax and grant program analysis.

For estate tax returns filed in 1977 the Internal Revenue Service Statistics Division will be matching a sample of estate tax returns filed with personal income tax returns filed in the last and 2 prior years by the decedents and the personal income tax returns filed in the year of estate tax return filing and 2 subsequent years by heirs. This match will give us the first really useful data on intergenerational transfers of wealth by bequest. It will tell us, for example, the income and certain other circumstances of heirs relative to the circumstances of the persons from whom they inherit. Are the heirs richer than the decedents? Or older? I should mention that the planning for this match of estate and personal income tax returns got underway more than 10 years ago. Those of us who were involved with it in the beginning are still around and awaiting the results, which we will have in hand if we are still around 1981 when the data are expected. In the collection of policy analysis statistics there is an intergenerational wealth transfer. Those who have the data for analysis are a successor generation to those who planned the data collection unless the planners are very young or live a long time.

The estate and personal income tax return match is made possible by the addition to the estate tax return of a question asking for the indentification of heirs. Although only one question, it must be completed by all executors of estates. The survey of income program participation will administer questionnaires to only about 20,000 households but each will be asked to submit to several interviews of up to an hour in duration. Sample surveys are a way generally of acquiring data at a reduced aggregate burden on respondents but they distribute the burden quite unevenly. It is important, however data are collected, to keep the burden on respondents to a minimum. For this reason, data should

98-815 - 78 - 11

not be collected except for clear policy analysis objectives. Moreover, policy analysts should be constrained by statistical policy to accept lower levels of precision in the data which they use. And surveys should not try to capture every small peculiarity in human behavior nor of the programs which Congress legislates.

With the increasing use by policy analysts of microdata files, problems of preserving the confidentiality of data supplied by respondents have become more serious. I should like for the Census Bureau to be given the authority to make to respondents in voluntary surveys a pledge that the information furnished will be used under no circumstances for any purposes other than statistical analysis. The Census Bureau then should be required to release identified data to other agencies subject to their agreeing not to violate the original pledge of confidentiality. The record identifiers are needed for the matching. File matching is necessary if survey questionnaires are not to become impossibly long. Most file matching will always be matching on attributes, and identifiers are no help. But where identifier matching is technically feasible it should not be thwarted by unnecessary confidentiality restrictions.

Thank you.

Mr. FRASER. Thank you very much, Mr. McClung.

All of you agree that our ability to get an adequate data base for measuring wealth is inadequate at the present time, I gather, but you all, each of you have tended to characterize it somewhat differently.

I would like to ask a question that goes to the organizational issues within the executive branch. Everyone except perhaps Professor Ruggles has argued that there should be no single agency responsible for the statistical collection. Professor Ruggles may agree with that, too. You have indicated that that would be a mistake to try to centralize data collection.

Is there adequate coordination among the different agencies? Concretely, what should we do if we want to improve data collection for the purpose of establishing a wealth holding?

Dr. DUNCAN. Mr. Chairman, if I may comment on that, there are a number of mechanisms for coordination at the moment. I think some of the views you have heard this morning point out some of the problems that exist, for example, with existing legislation. Let me illustrate

that.

Mr. McClung proposed that the Treasury should receive identifiable data from the Census Bureau, but from the point of view of the Census Bureau, they cannot get identifiable data from Internal Revenue Service in the Treasury Department. In other words, the legislation prohibits some of these exchanges. That is why I stressed that we really need comprehensive legislation on the data enclave concept which makes it possible to break out those statistical uses, all of which do not affect the individual. A freer flow of information for statistical purposes would be possible.

I believe that the reason a central collection agency is generally opposed, at least by the three of us from the Government is that, in reality some of what we consider statistical information is the byproduct of the administrative process. To literally have the full collection in one agency would mean, for example, that the statistical agency would need to collect much of the material now on the tax return. I

don't think anybody wants to add that additional burden on the public.

Instead we should try to get effective byproducts from our administrative programs so that the statistical collections are minimized. I think that sometimes, there is also a mixing up of what is the total burden associated with many of these statistical inquiries. The amount of information that must be added to an administrative inquiry to permit statistical linking is very small.

But by not doing it, the burden is increased since additional special purpose surveys have to be undertaken. That is a point that is not very well understood by most people. I think that much of our legislation tends to focus on the special purpose requirement and not the general purpose possibilities.

Thank you.

Mr. FRASER. Let me just follow that up, Dr. Duncan. Where does the buck stop in terms of initiating legislative action? In other words let's say that there is an agreement that we need to do better for that purpose but perhaps for other purposes as well. Where would the buck stop on that in the executive branch? Would that be you?

Dr. DUNCAN. That is the responsibility we have under the Budget and Accounting Procedures Act of 1950, to plan and coordinate the analysis of statistics. Without question, that is our office's responsibility. Mr. FRASER. Have you or are you in the process of proposing legislation of the kind that you just referred to?

Dr. DUNCAN. We have been making some proposals. Of course, we run into conflicts within the executive branch on precisely that point. That is one reason that some legislation has not been introduced.

Mr. FRASER. Give me a little more detail here about what happens. If your office says, well, we need to have certain changes in the law to improve our statistical acquisition process. Does this get circulated among the various interested agencies as a proposal and they come back with their comments?

Dr. DUNCAN. Right. Let me illustrate that, if I may, Mr. Chairman. In the mid-1930's there was a proposal to create what was known as a Standard Industrial Directory in which all businesses in the country would have an identification number. That would make it possible for whatever purpose, to draw a sample with a similar view of American industry. That proposal has been around for a very long time. It has been discussed quite widely, but has continually run into problems in implementation.

Recently, we drafted a piece of legislation that would permit the creation of that procedure. It is currently running into conflict with the tax law that was passed last year. The tax law permits the Census Bureau to get certain information from the identification number on the tax files; but it is specified that this shall only be used by the Census Bureau.

Consequently, that information, which is the basis for the census standard establishment list, could not be used by other agencies without modifying the tax law. So we have a legal barrier. The law having been recently passed, it has been difficult to get people to address that specific problem. This is due to concern about the confidentiality.

We are only talking about name and address and a size identifier. But that gets lost in the debate. This is a specific example of where