OF THE

ROYAL

STATISTICAL SOCIETY.

Founded 1834.

Incorporated by Royal Charter 1887.

VOL. LXXI.-PART III.

30TH SEPTEMBER, 1908.

LONDON:

THE ROYAL STATISTICAL SOCIETY,

9, ADELPHI TERRACE, STRAND, W.C.

1908.

NOTICE.

THE Council of the Royal Statistical Society wish it to be under. stood that the Society is not responsible for the statements or opinions expressed in the Papers read before the Society or inserted in its Journal.

JOURNAL

OF THE ROYAL STATISTICAL SOCIETY.

SEPTEMBER, 1908.

The IMPROVEMENT of OFFICIAL STATISTICS.

By ARTHUR L. BOWLEY, M.A.

[Read before the Royal Statistical Society, 16th June, 1908. The Rt. Hon. Sir CHARLES W. DILKE, Bart., M.P., President, in the Chair.]

I. Introductory.

SOME explanation is due to the Members of the Society for the somewhat unusual nature of this paper, for it has none of the paraphernalia of statistical tables, and it does not contribute to statistical knowledge. It is rather a paper with an object, that of exciting interest for the further improvement of our official statistics. Our President, both in his address and on many other occasions, has done much to call attention to the want of system in national statistics, to the waste of energy in their collection and publication, to want of correlation, and to the difficulties in the way of making use of them, and has given a strong lead to this Society in a crusade for the removal of these imperfections. Even if it was my purpose, it would be unnecessary to add charges to the terrible indictment he made last autumn. I wish rather to afford an opportunity for that discussion which cannot take place on an address from the chair, and at the same time to examine the problems of national statistics from a different and perhaps more fundamental point of view. I should explain at once that the title of this paper is to be read in the past, as well as in the future, or shall I say optative, tense; I propose to dwell on the improvement that has taken place in the statistical work of many public departments, as well as to try to indicate the directions which further developments may reasonably take.

There is no doubt that the past five years have seen a great and healthy increase of interest on the part of the general public in those phenomena of industrial and social life which are the subjects of statistical measurements; and the existing imperfections, the ambiguities and the difficulties of such measurements have been

VOL. LXXI. PART III.

21

vaguely realised by the less partisan and more thoughtful readers and writers. In 1903 statisticians rather suddenly found their neglected wares in demand, and I am afraid it must be owned that both public departments and private investigators were quite unprepared to meet the strain on their resources. Something has been done and more initiated since then, but much remains as yet unplanned. It was not then, it is not now, and will not be, till an intelligent public opinion forces the development, the business of any person or any department to prepare scientific measurements of national or industrial development.

That public interest in these things has not abated, but has rather increased, is evident to anyone who will be at the pains to turn through a file of newspapers. The leading articles, the political and other public speeches, the columns devoted to subjects of general interest, teem with statistical tables and arguments, sometimes extremely well informed, but frequently framed in complete ignorance of the nature and limitations of statistical measurement. Some of the subjects commonly dealt with may be instanced. The decrease of the birth rate in the United Kingdom and in other countries has received the attention that so striking a phenomenon deserves, though the corresponding change perceptible in the age-constitution of the population has not, I think, been much noticed. Problems of national finance are constantly discussed; the total yield and its elasticity, and the possibilities of re-adjustment of burdens, interest all classes in the kingdom. In particular the yield of the income tax and of the death duties since the Finance Act of 1907 excite so much comment, that it is sincerely to be hoped that these receipts will be very carefully analysed in the next report of the Inland Revenue Commissioners. The public finance of other countries also receives some notice.

A great deal of attention is given to the monthly statistics of foreign trade, rendered more intelligible by the new tabulation of 1901, though the pitfalls that still remain are not escaped. Not only are the totals examined, but week by week some "decaying" industry is found and a controversy rages as to the meaning of the statistics relating to it. I need hardly mention the interest that has been taken as to the hypothetical numbers of persons who might be entitled to old age pensions under various schemes. The statistics of unemployment are published monthly in the papers, but hardly so much stress is laid on them as they deserve, perhaps from a consciousness of their difficulty; none the less, many papers repeat the mischievous statement that a million persons were out of work this spring, and I have heard this quoted with the too credulous comment, "they must have something to go

"on or they wouldn't print it." This I give in evidence of the interest which is taken in statements of an apparently statistical nature. Immigration and emigration statistics are somewhat neglected, perhaps fortunately in view of their imperfections, but wild views of the numbers of aliens in England prevail. Among less general statistics, published in one paper in a selected period of three weeks, were paragraphs relating to the income and expenditure of the London County Council, to the water supply of London, to the number of volunteers, and to the probable number and expenditure of American visitors this summer. My point is, that newspaper editors probably gauge with some accuracy the tastes of their readers, and consider it worth while, even in times when controversy is not acute, to deal with statistics, not merely on the pages devoted to commerce or to sport, but in the forefront of their news, while their correspondents revel in statistical reasonings; and my deduction follows, that the public feel the need of statistical measurements, and would favour proposals and actions which would result in intelligible, authoritative and complete statistical information.

Before considering how far this rational desire for knowledge is met or can be met by official publications, it is expedient to discuss the nature and conditions of statistical measurement, subjects which have not recently been before the Society, and are very imperfectly apprehended in general. I propose to deal with these matters under seven heads the unit of measurement, the homogeneity and the universality of the material, and the stability, comparability, relativity and accuracy of the results; and then I wish to consider official statistics in relation to this analysis.

II. The nature and conditions of statistical measurement.

The Unit. The academic statistician would say that a statistical total is the number of persons or things in some defined time and place that possess a certain attribute or group of attributes, and that the unit, the item in this total, is one of these persons or things. Thus, to take the population of the United Kingdom, at the date of the census as an example, the group of attributes is (a) humanity, (B) living, (7) at midnight, March 31st, G.M.T. or Irish time, 1901, (6) in the United Kingdom, and (e) being entered on a census schedule. The unit is the person possessing these five attributes, and the population is the number of these units. If any estimate is made of homeless persons or others who are not entered on a householder's or other schedule, (e) should be amended to read, "entered on a census schedule or believed to satisfy the first