Income distribution: the current state of ignorance
Alan Maynard
The wealth tax proposals, which remain very much in the air, and the standing Royal Commission on income and wealth distribution, impel public interest in the division of the 'national cake' and it was never more relevant to inquire into our level of knowledge about the present distribution of income' in the United Kingdom, arid our knowledge about how this distribution is affected by government budgetary activity. A comprehensive knowledge of 'what is' is essential if we are to have a firm basis for deciding policy and if we are to avoid the temptation to blunder blindfold round this particular minefield.
On the income side anyone seeking to analyse the distribution can turn to one of four sources: the Inland Revenue survey, the Central Statistical Office (CSO) estimate based on the Inland Revenue survey (and terminated in 1967), the Family Expenditure Survey, and the General Household Survey. The analyst will find that by using each of the four surveys in isolation he can get four different sets of estimates of how much of the nation's income goes to each particular 10 per cent (decile) slice of the population. The reaspns for this are simple. The definitions of income vary from survey to survey. The surveys' coverages differ, and sample sizes and response rates vary between 11,000 and 110,000, and 70 per cent and 100 per cent. As the Inland Revenue survey ignores those who pay no tax, as the CSO estimates terminate in 1967, and as the General Household Survey estimates are accepted as subject to very high error margins, our analyst will be advised by the experts to turn his attention to the Family Expenditure Survey (FES). This survey is based on a sample of 11,000 which is not stratified by income. Lack of income stratification plus varying response rates and other factors lead to data biases which can be ameliorated only by careful reworking of the statistics. However, this sort of analysis is not without its problems. For . instance, let us assume we want to know how many people will be pushed above the 'poverty' line by a particular income maintenance scheme. Current practice defines poverty as being equal to Supplementary Benefit (SB) levels. (This arbitrary practice is not Without its paradoxes as the late Richard Crossman discovered, because if a 'generous' government raises SB rates rapidly the number of people living in poverty increases.) The arbitrary poverty criterion can be used with FES data to derive estimates of the number of people who are poor and the magnitude of their poverty. So far, so good? Unfortunately not. The FES data can be adjusted for biases in differing ways. Tony Atkinson, using explicit and reasonable assumptions, estimated that 3.7 per cent of the population were living in poverty in 1967. By 1972 this figure had risen in all probability to around 4 per cent, i.e. about 2.200 million. However, the Department of Health and Social Security estimates that in 1972 only 1,780 million were living in poverty. It is not clear how, and even whether the Department adjusted the FES data for sample biases. However it is confusing, to say the least, to get two estimates of the number qf people living in poverty in 1972 from the same basic data which differ by nearly 420,000. The statistical base is so suspect that it is difficult to determine the characteristics and magnitude of the number of people living in poverty, let alone know the cost of alternative income maintenance programmes. At present we deal in 'guesstimates' and our margins of error are not too small.
The Family Expenditure 'data used in the poverty estimates forms the basis of government estimates of the redistributive impact of the budget. It is essential to have some estimates of how the various income groups benefit from government expenditures and lose as a result of taxation. This role is fulfilled at present by a series 'of estimates published in Economic Trends. The data manipulations carried out to acquire these estimates could be viewed as hilarious were it not for the fact that they form the basis of estimates of, for example, the redistributive impact of Denis Healey's budget. The economic basis of the work is very weak and its coverage is partial. The authors of the estimates have failed to defend their assumptions, which in many cases are perplexing in the light of modern economic knowledge; they have been reluctant to include all government expenditures in their work despite
Nicholas Davenport, who is on holiday, will resume his articles next week the fact that they included all government taxation revenues; and they have been far too reluctant to carry out sensitivity analyses. Such analyses involve changing assumptions about, for example, the incidence (i.e. who pays?) ot certain taxes and seeing by how much, if at all, the final results of the exercise are affected. When such analyses are conducted reasonable assumptions can lead to results which differ from those of the official estimates by magnitudes in excess of 100 per cent. This being so, it is very difficult to place much reliance on the Economic Trends estimates. At worst they are meaningless, and at best they serve as a useful straw man to be battered by undergraduate students of economics.
The general conclusion about the data on income distribution is that • the basic statistics are weak and that the manipulation of these statistics is highly suspect. It is unfortunate that of the four basic data sources mentioned above none can be directly used by any person interested in discovering how income is distributed. That we have four basic sources is largely the product of independent recognition of the need for data by four separate government departments (the Department of Employment (FES), the Inland Revenue, the Central Statistical Office, and the Office of Population and Censuses (OHS)). While it does not appeal to those interested in the maintenance of empires, it does seem that concentration of activity with regard to acquiring the basic data at least, could result in much improved statistics at, at worst, the same cost. Such an outcome is appealing to those who would like to know more about the distribution, in particular more about those who are `poor.' The disconcerting feeling aroused by ministerial and departmental use of "guesstimates' can be best described as acute, and it will be compounded by the activities of the Royal Commission unless they realise that the data they are offered are subject to quite large margins of error.
Perhaps the only reassuring thing about the field of income and wealth distribution is that at least people are arguing in their attempts to discover the nature of the distributions. Until the late 1960s there was an unexplained hiatus in public and academic interest in this area. Perhaps interest in such issues is precipitated by serious economic difficulties. If it is, we are in for a prolonged discussion of the issues touched on in this article. It is to be hoped that this discussion will be pursued carefully and be based on more accurate data than is at present in vogue. Until more accurate data is available we will have to use the present material and take great care in interpreting, it and using it as a basis for policy decisions. Words can mean many things as Humpty Dumpty was all too aware: "When I use a word," Humpty Dumpty said in rather a scornful tone, "it mewls just what I choose it to mean — neither more nor less."
"The question is," said Alice, "whether you can make words mean different things."
"The question is," said Humpty, Dumpty, "which is to be master — that's all."
Alan Maynard is Lecturer in Econornics at York University
Previous page
