■Ce,) rr THE GIFT OF /\.\M-;j3 3o ^iflj fjol A DIFFICULTY WITH AMERICAN CENSUS-TAKING WALTER F. WILLCOX OF CORNELL UNIVERSITY CHIEF STATISTICIAN IN CHARGE OF THE DIVISION OF METHODS AND RESULTS, CENSUS OFFICE Cornell University Library HA205.W69 D5 A difficulty with American census-taking olin 3 1924 030 394 633 reprinted from The Quarterly Journal of Economics Vol. XIV., August, igoo .. . THE . . QUARTERLY JO URN^AL OF ECONOMICS Published for Harvard University Is established for the advancement of knowledge by the full and free discussion of economic questions. The editors assume no responsibility for the views of contributors, beyond a guarantee that they have a good claim to the attention of well-informed readers. Communications for the editors should be addressed to the Quarterly Journal of Economics, Cambridge, Mass./ business communications and subscriptions ($j. 00 a year), to Geo. H. Ellis, 372 Cons^ress Street, Boston Mass. CONTENTS FOR MAY, 1900. I. PETTY'S PLACE IN THE HISTORY OP ECONOMIC THEORY . Charles H..Hull II. PUBLIC EMPLOYMENT OFFICES IN THE UNITED STATES AND GERMANY .... .... . E. L. Bogart III. THE HOUSING PROBLEM IN GREAT CITIES E. R. L. Qould IV. THE CURRENCY ACT OF J900 .... . . F. W. Taussig V. JUDICIAL DECISIONS ON STATUTES PROHIBITING COMBI- NATIONS AND TRUSTS ... R. C. Davis NOTES AND MEMORANDA : Ethnic Tlieories and Movements of Population: A Rejoinder . . William Z. Ripley Current Publications — Report of the Industrial Commission — Prices in the United States x8go-gg — Municipal Housing in London. RECENT PUBLICATIONS UPON ECONOMICS APPENDIX. THE CURRENCY ACT OF 1900. TREASURY STATEMENTS. CONTENTS FOR AUGUST, 1900. I. A DIFFICULTY WITH AMERICAN CENSUS-TAKING . Walter F. Willcox II. THE IRON INDUSTRY IN THE UNITED STATES. II. . F. W. Taussig III. THE GAS COMMISSION OF MASSACHUSETTS . . John H. Gray IV. COMPETITION, ACTUAL AND THEORETICAL . John Bascom NOTES AND MEMORANDA: The Canadian Bank Amendment Act of igoo R. M. Breckenridge The Stock of Gold in the Country . ... . Pred. Perry Powers Current Notes — Labor Insurance Act rejected in Switzerland — The Bank of England's Uncovered Issue. RECENT PUBLICATIONS UPON. ECONOMICS. APPENDIX. STATISTICS ON PIG IRON. A DIFFICULTY WITH AMERICAN CENSUS-TAKING CHIEF STATISTICIAN DATE DUE JIJN 2 i TTVIi \ *• aw 1 i GAYLORD PRINTED IN US A. reprinted from The Quarterly Journal of Economics Vol. XIV., August, 1900 Eia Cornell University VB Library The origin^l'ifthislD^'ok is in tine Cornell University Library. There are no known copyright restrictions in the United States on the use of the text. 0' http://www.archive.org/details/cu31924030394633 A DIFFICULTY WITH AMERICAN CENSUS- TAKING. Thesis. — Statistical statements of absolute amount or of rate, based upon records not made until the end of the period to which the statements apply, are likely to be only a minimum limit of the truth, and to depart from the truth by a variable, and usually an indeterminate, amount. Governmental work in statistics is not unlike the keep- ing of accounts by a private firm or corporation. Private accounting falls into two main branches : first, the regular hooh-keeping, which records the fiow of goods, money, or obligations between the company and those with whom it deals ; and, secondly, the periodic taking of stock, which por- trays the condition of the business at a single instant. To keep hold on the details of a business, both sources of in- formation must be employed, each being a supplement and a check to the other. So a highly organized government maintains registration reports, which record the current of people, products, or acts deemed worthy of entry, and census reports, which aim to photograph the condition of a people for an instant, then leaving it till the next sitting is ordered.* Each of these two sources of information is needful ; neither is a satisfactory substitute for the other ; and the best results are secured only when the two work together, like the right and left hands. Yet some countries, such as Russia until 1897 and Japan and Roumania to this day, have only registration records, but no census ; and the United States, whose census is in some respects more highly developed and far-reaching than that of any other *The difference resemWes that diaTm in poKtieal economy between in- come as a flow and capital as a fund. In a census, as with capital, abstraction is made of the element of time ; while in registration, as with income, time is of the essence of the notion. civilized country, has scanty and imperfect registration records. The lack of a national registration system in the United States, while under our form of government perhaps an inevitable result of the relegation of care for the health, and of legislation regarding marriage and the family, to State or local control, has resulted in efforts to secure by means of our national census the sort of information for which registration is peculiarly adapted. These efforts have met with imperfect success. A review of the results in several fields will be the most effective means of elucidating the difficulty with which this article is concerned. Death-rate. — The method which the Census Office has been compelled to follow in approximating the total num- ber of deaths in the country during the census year — that is, the twelve months preceding the date of the census — has been to instruct the enumerators, in making their rounds, to ask at every dwelling visited, " Has any death occurred in this family during the census year?" The outcome of this method has been unsatisfactory. The num- ber of deaths obtained has fallen so far below the number which occurred in the country as to make the results of little scientific value. Even during the last month of the census year — that is, the first month preceding the census day — the omissions were probably from 5 to 10 per cent, of the total ; and with the lengthening of the period between the event and its recording by the enumerator, the propor- tion of omission increases, and for the second month of the census year was probably about 40 per cent.* The Cen- sus Office has been uniformly frank in admitting the shortcomings of its mortality returns, as is indicated by the following quotations : — " The tables of the census which undertake to give the total num- ber of births, marriages, and deaths in the year preceding the first of * Tenth Census, 1880, vol. 3d. p. jdii, table 18. June, 1850, can be said to have but very little value. . . . People will not or cannot remember and report to the census-taker the number of such events, and the particulars of them, which have happened in the period of a whole year to eighteen months prior to the time of his calling." * " Neither in 1850 nor in 1860 was the entire mortality of any State ascertained and reported, nor was even such an approximation ob- tained as wiU permit any reliable calculation to be made of the rate of mortality." t " At no one of the three censuses taken under the Act of March 23, 1850, has the aggregate number of deaths returned by the assistant marshals risen above two-thirds of the number of deaths probably occurring during the year of enumeration." f " The results of each of the four censuses in which an attempt has been made to ascertain the number of persons who died in the United States during the preceding year have shown that the enumerators did not obtain and record more than 60 to 70 per cent, of the actual number of deaths." § " The death-rate based upon the enumerators' returns, 10.86, . . . cannot be considered as representing the true death-rate in the area covered thereby. It is only presented as indicating the serious defi- ciency in the enumerators' returns." || The returns of deaths by enumerators for the Twelfth Census will be more full and complete than before, be- cause each enumerator, except those on per diem pay, will receive five cents for each death reported, while in 1890 he received but two cents. This additional pay may also evoke a certain number of fraudulent returns, some of which may escape the scrutiny and tests of the central office. Whether they exist and make a perceptible pro- portion of the total number cannot be decided in advance, if at all. The returns of deaths used in previous American censuses have been obtained from three sources, — the memories of * Seventh Census, 1850, p. : t Eighth Census, 1860, Mortality and Miscellaneous Statistics, p. xxt. t Ninth Census, 1870, Vital Statistics, p. ix. § Tenth Census, 1880, vol. xi. p. xi. II Eleventh Census, 1890, Abstract, p. 2t>4. the persons replying, the records of physicians with whom correspondence has been conducted, and the records of States and cities having registration records. These may be grouped as memory, private records, and public records, and are progressively more trustworthy in the order given. The death-rates derived from enumerators' returns are so inaccurate that it is doubtful whether they will be com- puted for the Twelfth Census. In this department of census work, therefore, the thesis stated at the beginning of the article is admitted by all competent students. Birth-rate. — Effort to ascertain the birth-rate through the census might be made in the same way that the death- rate is sought; i.e., by asking at every dwelling, "Has any birth occurred in this family during the census year ? " But the method actually followed is to enumerate all the children under one year. It is assumed that they were all born in this country, and to these are added all children who were reported as having been born and having died both within the census year. The result is an approxima- tion to the total births during the census year. It will be noticed that only the second of these two steps is open to the objection implied in my thesis. That this way of reaching a conclusion regarding the birth-rate of the United States leads to doubtful results is sufficiently shown by the fact that perhaps the two highest authori- ties in this field have reached opposite conclusions on the vital question whether the birth-rate in the United States decreased between 1880 and 1890.* As the arguments which have been employed by these experts will illustrate the difficulty inherent in our census methods, they de- serve examination in some detail. Dr. Billings called attention to the decreased ratios in * " It appears to me that we are justified in conoluding that the hirth-rate has really d imin ished in the United States." J. S. Billings, " The Diminish- ing Birth-rate in the United States," Forum, June, 1893. "The conclu- sions indicate that with a correct enumeration the seeming decrease in the birth-rate would disappear." W. A. King, " The Decrease in the Proportion of Children," Political Science Quarterly, December, 1897, p. 620. 1890 of children under one and under five to the total population and, in the cases available when the article was written, the decreased ratios of infants to the women be- tween fifteen and fifty. This might be due either to greater omissions in the counting of children in 1890 or to a decline in the birth-rate. The former explanation, except for the Southern negroes, was rejected by Dr. Bill- ings, and the latter accepted. Mr. King made a careful and thorough comparison be- tween the census figures for Massachusetts in 1890 and the results of registration in that State during the preced- ing five years, assumed that the registration records were more accurate than the census figures,* and concluded that probably 25 per cent, of the children under one and over 15 per cent, of the children under five, escaped enumera- tion at the Eleventh Census. From a similar line of argu- ment he concluded that probably 5 per cent, of the chil- dren under one, and nearly 3 per cent, of the children under five, escaped enumeration at the Tenth Census. Such a chain of reasoning, if accepted, would probably weigh with almost equal force against the various State censuses, which in Massachusetts and Rhode Island are at least as accurate as the Federal Census ; and the position would ultimately become this, — that any census is less accurate than registration. It is not unlikely that the differences between these * This assumption — namely, *' The registration data "with which the census figures are compared must be admitted to he more accurate than any enumera- tion of the population can he made under the present system" (loc. dt.) — seems to be open to discussion. Registration in England and Wales is probably quite as accurate as in Massachusetts, but in that country the contrary assumption is made ; and, in constructing the Graduated Table of estimated population at each year of age, the registration returns are corrected into harmony with the census figures. From the ofBce of the Registrar-general in London, I am in- formed, " there is no reason to suppose that any appreciable number either of children under five years of age or of persons of other ages escape enumeration at the English Census." In the text of the English Census, however, it is estimated that rather less than 2 per. cent, of the children under five are erroneously reported as over five. Census of England and Wales^ 1891, General Report, pp. 28, 105. 8 two experts may be reconciled and the truth of an inter- mediate position established by noticing that the form of the age question in 1880 was " age at last birthday," but in 1890 it was " age at nearest birthday." In both years the ages of all children under one were to be reported in months. If these questions were answered as asked, the persons reported as under five in 1890 were only those who were under four years and six months. This cannot be assumed to be true, but the hypothesis that an appreciable number answered the question according to its terms deserves to be tested. Under the instruc- tions issued, the persons one year old in 1890 included only those between twelve and eighteen months, while in 1880 it included those between twelve and twenty- four months. A sharp decrease in the proportion of children reported as one year of age, between 1880 and 1890, would therefore be evidence that in many cases the question was answered as asked. Of all the children under five at any date, probably rather over one-fifth are between one and two years of age, since an early year in any age group has a somewhat larger number than a later year. But in the early years of life there is a tendency to report children older than they are.* Children under twelve months are often reported as one year old, because they are in their fkst year; and children between twelve and twenty-four months as two, because in their second year. This source of error results in the number of persons reported as under twelve months and the number reported as between twelve and twenty-four months being less than the truth. Thus in 1880 only 21 per cent, of the children under five were reported as under one year of age, and only 18 per cent, as one, while the true per- centage for under one was over 22, and for one was *Farr, Vital Statistics, p. 206; Census of England and Wales, 1891, General Keport, p. 27. over 20.* In 1890, therefore, one might assume that, had the question been unchanged, the ratio of children re- ported as one year of age to the total under five would have been about 18 per cent. But the instructions issued to enumerators in 1890 were far more specific and emphatic than they had been previously in requiring the age to be returned accurately; and, as a whole, the age returns in 1890 are much more correct than in 1880. f Hence, if no new source of error had entered in 1890, we may assume that the ratio of children returned as one year of age to the total under five would have been per- haps 18i or 19 per cent. J In fact, as shown by the foot-note below, it was 14 per cent. ; and this sharp decrease from the percentage in 1880 is probably an evidence and a measure of the influence that the changed form of the age question exercised upon the enumerators' returns. Confirmation of the hypothesis is derived from the fact that the decrease between 1880 and 1890 in the proportion of persons reported as one year of age was more marked in the Northern States than in the Southern. Various bits of evidence, external and internal, warrant the conclusion that the census was taken more accurately in the North than in the South ; and the change in the form of the age question would, therefore, be more regarded by enumerators and the public in the North. Further- more, the difference, here ascribed to the change in the * The following table gives the distribution of the children under five to the single years, as shown in English Life Table No. 3, and as reported in the American censuses of 1880 and 1890 : — Per cent, of total children under five at each year of age by English Ufe U.S. Cenjrus U.S. Census Age Table No. 3. Of 1880. of 1890. 0—1 _ 22.6 20.9 20.B 1—2 20.4 18.2 14.1 2-3 19.5 20.6 22.7 3—4 . . 18.9 20.0 21.4 4—6 18.6 20.3 21.3 Total 100.0 100.0 100.0 t Publication.'! American Statistical Association, vol. v. p. 133. t By the English Census of 1841 it was 20.4 per cent. 10 form of the age question, was slightly greater among the Southern whites than among the Southern negroes. This tends to confirm the explanation here suggested, and indicates further that the change in the form of the age question affected not merely the enumerators, but also the public. It seems, therefore, impossible, on the one hand, to accept Mr. King's contention that the decrease in the birth-rate between 1880 and 1890 was merely apparent, and not real, and, on the other hand, to accept the con- clusion of Dr. Billings that the decrease in the birth-rate in the United States was greater than in any of the eleven countries in western Europe, with which comparison is made. There was probably a sharp and almost universal decrease in the birth-rate between 1880 and 1890, — a de- crease which affected especially the negroes, and to a somewhat less degree the Southern whites ; but the actual amount of this decrease was less than the apparent amount owing to the change in the form of the age question, and it is impossible now to determine what proportion of the decrease was actual and what proportion was merely apparent. But in any case this method of computing the birth- rate is open to serious error, because of the certainty that many children born and dying within the census year escape enumeration, and also because of the tendency to report the ages of children as greater than they are. The latter tendency to error might, perhaps, be reduced to a minimum by carefully smoothing the curve of reported ages before beginning any computation; but the former cannot be successfully eliminated. Marriage-rate. — The census enumerator asked at the censuses of 1880 and 1890, and with slightly changed lan- guage will ask at the census of 1900, whether the person, if married, was married during the census year. I believe the answers to this question have never been tabulated and published ; but, in the light of the difficulties which have 11 been met in determining the death-rate from answers to a similar question, it may well be doubted whether the figures could ever be used as a basis for ascertaining the marriage-rate of the country, or of any part of it. The precise question asked at the Twelfth Census will be, number of years married. To compute the marriage-rate from the answers would be like computing the birth-rate from the answers to the question, age at last birthday, or number of years lived ; and the objections implied in my thesis clearly hold against any such method, because it ignores all cases in which the marriage had begun and had ended by death or otherwise within the census year, and hence could give only a minimum limit to the true marriage-rate. Divorce-rate. — All persons are asked at each census to report their marital condition on the census day as either single, married, widowed, or divorced. Many persons who are divorced will not admit it to the enumerator. Many who answer regarding other persons in the census family* are ignorant of their real marital condition. Yet, even if all divorced persons were returned as such, the figures would afford almost no basis for an inference regarding the divorce-rate for the year preceding, and, I believe, have never been used for that purpose. The pro- portion of divorced persons in the community on any single day might remain decade after decade the same ; and yet, in case remarriage after divorce were to become more general, the divorce-rate might be rapidly increas- ing. On the whole, the census returns of divorced per- sons are of little value. An attempt has been made by the Department of Labor under orders of Congress to reach the divorce-rate of the United States by compiling the judicial returns of divorce decrees.! These recorded decrees belong to the class of * This includes a boarding-house, hotel, or institution, as well as a family in its ordinary sense. t Meport on Marriage and Divorce in the XTnited States^ 1867 to 1886. 12 registration records, and in several of the States are pub- lished in the State registration reports. The attempt of the Federal government, therefore, being based upon contemporary written records, was far more significant and valuable than the work of the census. But minor errors were inevitable because of the fact that the inquiry ex- tended over the judicial records of each county for the preceding twenty years. During that time the divorce records in ninety-eight counties had been injured or de- stroyed, and this cause of error made the apparent increase of the divorce-rate perhaps 2 per cent, greater than the actual increase. Probably also court records came to be kept somewhat more carefully, and there was less chance for a divorce decree to escape record at the end than at the beginning of the period. Grime-rate. — In seeking information about matters often or usually entailing social disapproval, of which divorce and crime are examples, it is inexpedient to ask of each family whether any divorce has been decreed to, or any crime committed by, a member of the family dur- ing the census year. In the case of crime, therefore, a substitute for asking the question of each family has been sought in gathering the statistics of persons in prison on the census day. The word " crime " is so vague, and its definition varies so with place and period, that it would be more exact to break the genus into various species, and speak of a murder- rate, a forgery-rate, and the like. But one may grant that, for short periods of time and in the same region, the generic word " crime " usually covers about the same species of crimes in about the same proportions. One may perhaps grant, also, that the ratio of crime detected and punished by imprisonment to the number of crimes committed does not vary widely or suddenly. Neverthe- less, the method of computing the crime-rate from the number of persons reported as in prison on the census day is almost as unsatisfactory as would be the method of 13 computing the immigration-rate from the number of persons born abroad and reported in the country on the census day. The main difference between the two cases is that the country has no registration of crimes detected and punished, no judicial statistics ; but it has registration records of immigration, which, even in their imperfect con- dition, are so much better for the purpose than the census figures that the latter have not been utilized as a basis for computing an immigration-rate. The leading experts in the country to-day admit the impossibility of determining from census figures even the most vital question in this field. Is crime in the United States increasing or decreas- ing ? * In the whole field of population statistics, there- fore, the careful and intelligent efforts of officials and private students to wring, from figures recorded for the first time in the census, information about rates for which such figures are not adapted, have ended in substantial failure. Industrial statistics. — In this field the difficulty is prob- ably not so weighty, and certainly has not been so clearly recognized. Yet I believe it to exist, and to be important enough to deserve careful consideration. The inquiries on the industrial schedules of the census relate partly to a certain day, either the census day or the last day of the business year reported, and partly to the last crop year or business year. Only to questions of the latter class could the difficulties suggested in my thesis apply. But, still further, the manufacturing schedule for the Twelfth Cen- sus contains the instruction, " Amounts and values must be obtained from book accounts, if such accounts are avail- able " ; and the agricultural schedule declares, " In the absence of book accounts, careful estimates must be given." These book accounts, containing entries made shortly after the occurrences which they record, are of the nature *See R. P. Falkner, " Crime and the Census," in Annals of the American Academy of Political and Social Science, January, 1897. 14 of private registration reports. So far as the manufactur- ing or agricultural returns are derived from such sources, they are entitled to somewhat the same confidence as the death records of the census copied from State or municipal registers. Decade by decade a larger proportion of the business industries of the country reported in the manufacturing returns probably can and do answer the questions on the schedule from book accounts. Such returns in becoming more accurate tend to become more complete, for errors of memory are likely to be errors on the side of omission. Hence, entirely aside from any improvement in the effi- ciency of the census work, the tendency of such changes in book-keeping would be to exaggerate the rate of growth in the manufacturing industry of the country. It seems impossible to say, otherwise than as a result of personal observation and judgment, whether book-keeping among farmers is extending decade by decade, and there- fore whether the answers on the agricultural schedules are more generally derived from book accounts. The best expert opinion available, however, is that such a change is in progress, but with much slowness. Not a little under- statement of the agricultural returns of the country, frequently charged against the census figures, assuming such understatement to exist, may plausibly be assigned to errors of omission in returns based upon memory alone. If the thesis with which this paper opens be deemed established, the question is inevitable. What of it ? How may the present conditions be improved ? In the field of population statistics, with which we have been especially concerned, a suggestion has been made for turning the census officials in selected localities into regis- tration officials, whose duty it should be to record all deaths in their districts during the census year. This 15 method seems to me open to grave constitutional and ad- ministrative objections and to be, therefore, inadmissible.* A census office in the United States, under our Federal system, cannot be made into a registration office, author- ized to record births, marriages, and deaths within the limit of the States, without doing violence to our system of government or being so powerless as to accomplish little or nothing. If one desires a national registration system, the way to it through a Constitutional amend- ment is perhaps available ; but I do not believe it can be secured in any other way. If that method seems hopeless, the gradual extension of state and municipal registration may be regarded as a solution. To one familiar with the little value and slow growth of these systems in much of the country, this solu- tion also seems remote and improbable. The only other alternative open is for the United States government to build up a system of voluntary and continuous co-operation between the local registration offices and itself. This seems to me a possibihty which may be realized in the near future ; but its realization is intimately interwoven with the future of the Census Office. Until that or some similar office shall become permanent, such co-operation is almost impracticable. When a permanent census office is demanded merely as the best means for taking the decennial census success- fully and economically, the plea for such an office is weakened by narrowness of view. A census may be taken * The stndent interested in the discussion of the subject is referred to the following articles : by Dr. C. L. Wilbur, " Outlook for a General System of Registration," in American Public Health Association, Reports and Papers, vol. xxi. p. 231 (1896) ; " Vital Statistics for the Twelfth Census," Publications of American Statistical Association, vol. v. p. 188 ; " A Note of Correction," Publi- cations of American Statistical Association, vol. yi. p. 311 ; " Eepresentative Mor- tality Statistics," in American Public Health Association, Meports and Papers, vol. XXV, p. 368. The objections to this plan, which I stated in conunent- ing upon the writer's first paper in the Publications of the Statistical Asso- ciation, seem to lie with equal force, I regret to say, against the later statement of it. 16 as now by a temporary office ; but no census can do the work of a registration office, and a census not supple- mented by registration loses, at least in the field of popu- lation statistics, more than half its scientific value. For, aside from the administrative ends subserved by a census of population, its main scientific goal is to furnish a broad basis for registration. A census which does not blossom in registration is almost as sterile as capital which does not blossom in income. In the schedules for industrial statistics it would cer- tainly aid the student to have the question introduced, "Are these figures derived from book accounts?" If it were practicable to introduce the question and tabulate the answers, a distinction might be drawn in these fields of census work, similar to that established in the division of Vital Statistics, between the returns based upon con- temporary written records and the returns based upon the memory of the persons reporting. Such a distinction would be of theoretical importance. It might furnish some clew to the margin of error to be attributed to the industrial figures of each census. It seems not unlikely that answers to such a question would enable a line to be drawn, at least approximately, between the factory produc- tion of the country, properly speaking, and the production of the hand-trades in shop work and in household indus- tries. While Congress calls for returns of the latter class, every competent student knows that it is impossible to get them adequately, both because much of the work is not localized in any " establishment," and also because the returns regarding such work must usually be based on memory. Even if neither of these suggestions proves fruitful, it may aid the student to have a difficulty, which seems to me fundamental, pointed out and the range of its applica- tion determined. If the main thesis be granted, the following subordinate 17 theses could probably be demonstrated, some of them almost as corollaries. 1. The United States as a whole has no means of decid- ing with certainty upon the healthfulness of the country at different times or in different parts or the healthfulness of different occupations. 2. Few States and cities have such means ; and, on the whole, the increase in the number or efficiency of such local agencies is slow and slight. 3. Neither the United States as a whole nor any of its parts has the means of deciding with certainty upon the rate of natural increase by excess of births over deaths in any locality or in any class, like the foreign- born, negro, or urban population. 4. There is no sound means of determining the relation of the country as a whole or of most of its parts or any of its social classes to the institution of marriage, the age at which marriage occurs, the proportion of the several classes marrying, the duration of the union, and the like. 5. The country has no regular agency for giving similar information about divorce. But, as divorce decrees are now a matter of judicial record, the difficulties in this field are less than in any other. 6. The country has no trustworthy means for ascertain- ing whether crime in the United States, or any specific kind of crime, such as drunkenness, forgery, or rape, is increas- ing or decreasing. 7. The only feasible means of securing these ends in the near future is through a close and continuous co-operation between state or municipal registration offices and some statistical department of the Federal goverment. 8. Such a statistical department must be permanent; and, therefore, the sine qua non of securing these ends is a permanent statistical office at Washington, empowered to work towards them.