they are compiled and interpreted with great competence.

It has proved feasible to adjust the foreign data for only one year. The year 1960 has been selected as the year of comparison because, at the time this analysis was under way, it was the most recent year for which essential details were available. This year was somewhat more favorable to the United States than 1961 would have been, but it was considerably less favorable than the years in the early fifties.3

In a brief article such as the present one, it is not possible to discuss in detail the differences in the definitions used by other industrial countries that call for adjustment if the unemployment statistics of these countries are to be compared directly with our own. A brief summary of the adjustments made, however, is presented in the following paragraphs. The unemployment statistics system of the United States is described briefly at the outset to provide a basis with which the foreign systems can be compared.

The United States. The main source of unemployment statistics in the United States is a sample survey of households, conducted monthly since 1940 and identified as the Monthly Report on the Labor Force. The material is collected and tabulated by the Bureau of the Census under contract with the Bureau of Labor Statistics, which analyzes and publishes the data.

In this survey, trained interviewers obtain information from approximately 35,000 households, forming a probability sample representative of the entire civilian noninstitutional population. Sufficient information is obtained to classify all persons 14 years of age or older as (1) employed, (2) unemployed, or (3) not in the labor force. As thus described, the labor force (which is simply the total of the employed and the unemployed) is commonly referred to as the "civilian labor force," since it excludes the Armed Forces. Important features of the U.S. system are its comprehensiveness and its complete independence of administrative programs. Under this system, the unemployed consist of all civilians 14 years of age or older who during a particular survey week do not work even as much as 1 hour, but who are looking for work. Persons who never had a job are counted among the unemployed if they meet these criteria. Persons with jobs but on tem

porary layoff are also considered as unemployed. Although it is not practicable to set forth here the many detailed rules and interpretations that serve to distinguish the unemployed, it may be mentioned that the definition used by the United States appears to be fully consistent with the international definition referred to earlier.

The results of the monthly survey are published promptly and in considerable detail. Thus, the data on the unemployed are analyzed by sex, age, former occupation and industry, period of unemployment, etc. The unemployment rate (interpreted in this country to be the percent of the labor force who are unemployed) is published with and without seasonal adjustment. The sampling variability of the estimates is published regularly."

Canada and Japan. Two of the seven foreign countries selected for comparison, Canada and Japan, have systems of unemployment statistics very similar to our own. The sample survey used by Canada, the Labor Force Survey, produces monthly data which are closely comparable with those of the United States and call for no adjustment."

2 Summary figures were available for 1961 for most countries, but in several cases the supporting detail by age, sex, work status, etc., had not yet been published.

The unemployment rate for the United States fell to 2.9 percent in 1953 but rose irregularly to 6.8 percent in 1958; it was 5.6 percent in 1960 and 6.7 percent in 1961. The rate for Canada rose from 2.4 percent in 1951 to 7.2 percent in 1961. Unemployment rates in Germany and Italy peaked at 9.0 and 10.0, respectively, in the early fifties but have subsequently dropped sharply. In none of the other four countries did the average annual rate of unemployment rise as high as 3.0 percent during the past 11 years.

For an excellent recent account of this survey, see A. Ross Eckler, "The Continuous Population and Labor Force Survey in the United States." Ch. XIII in Family Living Studies, International Labor Office, Geneva (1961). See also "Concepts and Methods Used in the Current Employment and Unemployment Statistics Prepared by the Bureau of the Census," Current Population Reports, Series P-23, No. 5 (U.S. Bureau of the Census, May 9, 1958).

Valuable supplementary data on unemployment in the United States are obtained from the unemployment compensation system, operated jointly by the Bureau of Employment Security and the State unemployment insurance agencies. This system (together with smaller systems covering special groups such as railroad workers) covers only about 70 percent of the civilian labor force, and the weekly count of persons receiving benefits is somewhat affected by periodic changes in the laws, the exhaustion of benefits by the long-term unemployed, and other administrative factors. When rough adjustment is made for such factors, however, long-term trends and periodic movements in unemployment as reflected in these statistics are much the same as those shown by the Monthly Report on the Labor Force.

• For brief information on the Canadian labor force surveys, see the Introduction to The Labor Force, November 1945-July 1958, issued by the Dominion Bureau of Statistics (Ottawa, Reference Paper No. 58, 1958 Revision); also the technical notes in that Bureau's monthly publication The Labor Force, issued regularly since August 1958. The Report of The Committee on Unemployment Statistics (Ottawa, August 1960) is another source of considerable value. The sample design of the Labor Force Survey is discussed in Canadian Statistical Review (Ottawa, Dominion Bureau of Statistics, April 1962).

The Japanese Labor Force Survey is patterned after the U.S. system, but with a few significant differences in definitions. To make rough adjustment for these differences, it is necessary to reduce the Japanese labor force by 1.5 million for unpaid family workers who worked less than 15 hours in the survey week, and by about 300,000 for members of the Special Defense Force and certain residents of institutions. These groups would not be counted under U.S. definitions. An addition of some 150,000 must be made, however, for 14-year-olds, who are not counted by the Japanese. The net result of these adjustments is to decrease the labor force figure by 1.7 million.

The adjustments made for these groups do not affect the number of unemployed appreciably, but this number has been increased by 50,000 in adjustment for persons who would be counted as unemployed in this country but are classified by the Japanese as employed or not in the labor force. When the adjusted number of unemployed (480,000) is compared with the adjusted labor force figure (43.5 million), the resulting unemployment rate is found to be 1.1 percent, as compared with the published 1.0 percent.

These unemployment statistics based on sample surveys are official and widely used in Canada and

The methods and concepts used in the Japanese survey are described in Labor Force Survey of Japan (Tokyo, Office of the Prime Minister, Bureau of Statistics, August 1956).

These consist primarily of workers on temporary layoff, persons waiting to start work after obtaining a new job and jobless who want work but are not seeking it because of temporary illness, or who state that they were not seeking work because they know no jobs are available.

'The Canadian Labor Force Survey was long regarded simply as one of two useful measures of unemployment. The other, a count of claimants for benefits under a compulsory unemployment insurance system, regularly reported a larger number of unemployed and a higher rate of unemployment. In 1960, for example, when the Labor Force Survey showed 448,000, or 7.0 percent of the labor force, unemployed, claimants for unemployment insurance numbered 518,000, or 12.6 percent of the insured population. The results of the Labor Force Survey were recognized as official following the report of a ministerial committee on unemployment statistics (see footnote 6) in 1960 which found numerous weaknesses in the "operational" system and concluded that the Labor Force Survey "is the only source available that provides a satisfactory basis for the practical measurement of unemployment."

Japan also maintains other indicators of the level of unemployment, but none is claimed to be both comprehensive and up to date. The monthly record of unemployment insurance beneficiaries, maintained by the Ministry of Labor, applies to only approximately 12.6 million insurance participants (1960) out of a labor force of more than 43 million. The number of recipients has usually equaled one-half to two-thirds of the total number of unemployed as reported by the sample survey, however, suggesting a considerably higher unemployment rate.

10 Technical information regarding the ISTAT surveys is given in the Institute's Metodi e Norme: Rilevazion Campionarie Delle Forze Di Lavoro, Series A, No. 3 (March 1958). See also Renato Curatolo, "Le statistiche correnti dell' occupazione e della disoccupazione in Italia," Rassegna de Statistiche del Lavoro, January-April and May-June, 1961.

Japan. Both countries, however, maintain other measures of unemployment, based on "operational" statistics, which yield unemployment rates somewhat higher than those generally published."

Italy. The most widely known statistics on unemployment in Italy are based on registrations at placement offices. This series, however, is commonly acknowledged to overstate the level of unemployment, owing to inclusion of registrations by persons already employed, registration by elderly persons who do not want to work but only wish to maintain eligibility for family allowances, failure of legitimate registrants to cancel their registrations promptly after obtaining a job, etc. For comparison with U.S. statistics, another series-equally official but little known outside Italy-is preferable.

The labor force sample surveys conducted by the Central Institute of Statistics (ISTAT) are similar to those used in the United States and appear to have been inspired by American experience.10 The Italian surveys were initiated in 1952 and have been made quarterly since 1959. For comparison with U.S. unemployment statistics, the Italian data have required a number of adjustments. First, the reported 1960 average labor force figure of 20,815,000 has been augmented to include "occasional" workers who in the United States would be counted as employed. It has then been diminished to exclude career military personnel, migratory workers employed abroad, and certain employed persons under 14 years of age. The number of "unemployed," reported as averaging 557,000 in 1960, excludes new entrants into the labor force and has been augmented to include a substantial number of such persons. A lesser addition to the number of unemployed has been made on behalf of persons temporarily laid off or who have a job but have not yet reported to work. Such persons are considered as unemployed in the United States but in Italy are regarded as employed.

These various adjustments yield an adjusted labor force of 20,945,000 and an adjusted unemployment estimate of 896,000, resulting in an unemployment rate of 4.3 percent. The adjusted figure suggests that the unemployment rate in Italy in 1960, commonly reported to exceed that in the United States, was actually somewhat lower when measured by U.S. methods and definitions.

Federal Republic of Germany. The principal and official unemployment statistics of Germany are administrative statistics representing unemployed registered at the employment exchanges. The employment exchanges have a high degree of public acceptance in Germany. Registration, moreover, is required for those who wish to claim benefits under the broad unemployment insurance system or to qualify for unemployment relief. The widely published count of unemployed registrants is thus a meaningful figure and highly useful as an economic indicator.

For comparison with U.S. data, however, it has seemed preferable to rely on another source of unemployment statistics, which has provided data only since 1957. This is the Microcensus (Mikrozensus)," a sample survey of households designed to obtain information on employment and unemployment, together with other characteristics of the population. Information on the labor force, employment, and unemployment is published only for October of each year when a full 1-percent sample (about 170,000 households) is used. The results obtained by the Microcensus so far have sometimes revealed a lower number of unemployed and have invariably indicated a lower rate of unemployment than has the registration system. The two unemployment rates have followed a somewhat similar trend, however, and in October 1960 both were under 1 percent. The comparative figures, without adjustment, are as follows:

But for purposes of this report, even the results of the Microcensus require some adjustment. First, it is necessary to reduce the total labor force (25,593,000 in October 1960) by an estimated 300,000 to exclude military personnel. The labor force is further reduced by 89,000 to exclude unpaid family workers who worked less than 15 hours during the survey week (since such workers are not included in the labor force in the United States). These adjustments leave a residual of approximately 25.2 million, but do not affect the reported unemployment rate of 0.5 percent.

More important is the adjustment for seasonality. Unemployment in October is near its low for the year, and in 1960, judging by the registration series, was only about 52 percent of the annual average. The Microcensus estimates of 128,000 unemployed may thus be adjusted upward to 245,000 as a rough annual average. No seasonal adjustment has seemed feasible in the case of the labor force figure, but comparison with the total arrived at earlier (25.2 million) yields an adjusted unemployment rate of 1.0 percent. This may be taken as an approximation of the average level of unemployment in Germany in 1960 as measured according to American definitions.

France. The French regularly publish two official series of unemployment statistics but neither of these provides a good basis for comparison with American statistics. One system depends on a count of cases receiving unemployment relief; it excludes a very large proportion of the unemployed and will not be discussed here. The second series, a count of registrants for jobs, is considerably more complete but still appears to understate the number of unemployed. For one thing, it is limited largely to wage and salary workers, who make up only about two-thirds of the labor force. Persons seeking a job for the first time rarely register and many skilled or specialized workers fail to visit the placement office when they know that no jobs calling for their skills have been listed. Women workers appear to depend on the placement offices relatively less than men.

11 For a description of the Microcensus and its position in German statistics, see V. S. Koller and L. Herberger, "Der Mikrozensus," in Allgemeines Statistisches Archiv, (Wiesbaden, Statistiches Bundesamt, Heft 3, 1960). Numerous articles have also appeared in Wirtschaft und Statistik, for example, L. Herberger, "Der Mikrozensus als neues Instrument zur Erfassung sozial-okonomischer Tatbestande" (Heft 4, April 1957) and G. Furst and associates, "Zur Statistik der Erwerbstatigkeit und der Beschaftigung" (Heft 3, March 1959).

The French have also experimented with sample surveys of the labor force, however, and since 1950 have made at least 14 such surveys.12 The surveys have been made irregularly and with small samples that have resulted in rather unstable results. There have been a number of changes in methods of sample selection, in definitions used, and in the particular months represented by the surveys. The results are not widely published and their presentation chiefly in the form of percentages tends to complicate their use. Nevertheless, they provide the best available basis for comparing unemployment rates in France and the United States.

The methods and definitions used in the recent French sample surveys are basically quite similar to those used in the United States. This was less true of the early surveys, 13 but the major early differences have been eliminated as a result of changes in the French definitions. The unemployed reported by the sample surveys that have been published have averaged around 400,000, only about 2 percent of the labor force but some two and one-half to three times the number registered for jobs in the respective survey months.

It is not possible to make a direct comparison of the unemployment rates resulting from the French and U.S. sample surveys because the unemployment figures resulting from the more recent French surveys have not been released. Rough estimates can be prepared, however, on the basis of the eight earlier surveys (covering various periods between April 1950 and February 1956) whose results have been published. Comparing these with the number of registered unemployed for the same periods, it is found that

12 Descriptive information regarding the French labor force surveys has been published from time to time along with the results of the surveys, in the quarterly supplements to Bulletin Mensuel de Statistique, issued by the Institut National de la Statistique et des Études Économiques, Paris; see, for example, the supplements for January-March 1951, October-December 1952, October-December 1954, and October-December 1956.

"Job applicants without earlier work experience were originally excluded from the unemployed and from the labor force. Housewives and students with jobs were excluded from the labor force unless they worked as much as 20 hours during the survey week. Rough correction for such differences has been made in the early survey results used in the present study.

14 That is, the total number of unemployed has been expressed as a percentage of the entire labor force, while the number of registrants has been compared only with the substantially smaller number of wage and salary workers.

15 Information on the Swedish labor force surveys was obtained from a translation of an article by F. Larsson, "Tentative Labor Force Surveys in Sweden," which appeared in a 1960 issue of Utdrag ur Statistik Tidskrift, published by the Central Bureau of Statistics in Stockholm, and from a Royal Labor Market Board publication, Arbetskraftsundersokningarna Maj 1960-Augusti 1961 (Arbetsmarknads-information, Series M 7/1961).

the ratio of registrations to estimated total unemployed has always, except in one instance, fallen within the range 31-42 percent, the average for the eight ratios being 35 percent. Assuming that this average relationship held in 1960, when registrations averaged 131,000, it may be estimated that the total number of unemployed, defined roughly in conformance with the U.S. definitions, was in the neighborhood of 370,000. This figure, compared with a labor force of about 19.1 million, yields an unemployment rate of 1.9 percent.

It seems clear from the foregoing that the French statistics most nearly comparable to those used in the United States yield a considerably higher number of unemployed than is indicated by either of the more official series. The rate of unemployment which has been roughly estimated is also somewhat higher than the rate of 1.1 percent which may be computed from the registration series, but the difference is less marked because the official rate has been computed on a smaller base. Whatever the measure used, unemployment in France in 1960 was a far less serious problem than in the United States.

Sweden. Sweden depended for many years on unemployment statistics maintained by the trade unions. Since 1956, however, the Royal Labor Market Board has issued monthly statistics on registrations of the unemployed at local employment offices. The statistics most widely used are those relating to members of unemployment insurance funds.

Unemployment insurance is not compulsory, but it covers a substantial proportion of the wage and salary employees. The number of registered unemployed within the insurance system, which has averaged less than 30,000 in most recent years, is, however, considerably less than the total number of unemployed.

Since 1959 the Labor Market Board has also made several sample surveys of the labor force,15 and the results of these surveys are helpful in gaining an impression of the full volume of unemployment. The first two surveys, both in 1959, were frankly experimental. Two more were made. in 1960 and three more in 1961. The surveys are expected to be quarterly hereafter.

The method of the Swedish sample surveys has differed considerably from that of the United

States MRLF, particularly with regard to sampling. The Swedish surveys are based on a sample of individuals rather than a sample of households. The sample is small (6,500 persons in the surveys of 1961-62), and most of the interviews are conducted by telephone.

The definitions used in the Swedish surveys are not available in full detail at the time of this study, but they appear to be consistent with the international and the U.S. definitions. The information obtained permits the adult population (age 14 and over) to be divided into the traditional three groups: The employed, the unemployed, and persons outside the labor force. A good deal of detail is obtained regarding the sex, age, marital status, etc., of the unemployed, and separate information is provided regarding part-time employed workers. Most of the results are presented only in the form of percentages.

The sample surveys indicate that the total number of unemployed is considerably larger than the number reported as insured registrants. The rate of unemployment obtained through the sample surveys is also generally higher, although the difference here is not so great, owing to the use of a larger base in computing it.

In the surveys of May, August, and November 1961, unemployment was found to average about 55,000 and the rate of unemployment about 1.5 percent. Although no survey was made during a late winter month in 1961, a preliminary report on a survey in February 1962 indicates that the number of unemployed was only 59,000 and the unemployment rate a surprisingly low 1.6 percent. If these figures are made to serve for February 1961, the indicated annual average of unemployed rises to 56,000 and the annual average rate of unemployment remains at a tentative 1.5 percent. These are very rough figures, but they are believed to provide a general indication of the level of unemployment in Sweden consistent with U.S. definitions. Similar figures for 1960 have not been computed (because only two surveys were made in that year), but they would probably be somewhat higher than these since the May survey results were considerably higher in 1960 than in 1961.

Great Britain. British unemployment statistics are the result of collection procedures, concepts, and definitions that differ substantially from those

used in the United States. The British data are based almost entirely on a count of registrants at Employment Exchanges and Youth Employment Offices. These facilities are widely used and their records constitute a sensitive and meaningful measure of the Nation's economic well-being.

A careful student 16 of British and American unemployment statistics, comparing unemployment rates under conditions of relatively low unemployment,17 arrived at the following meaningful conclusions regarding the importance of various differences between the two systems.

Two significant features of the British system tend to overstate British unemployment rates as compared with American: (1) The British compute the unemployment rate by expressing the number of unemployed as a percentage of the wage and salary labor force (excluding the self-employed and family workers), while the United States uses the entire labor force as denominator. Taken by itself, this difference would overstate the British rate by about 0.1 percentage point. (2) The British count all persons who were registered as unemployed on the single day of the survey, whereas some of these, who worked 1 hour or more during the preceding week, would be counted as employed according to American definitions. This difference also overstates the British rate, in this case by 0.2 to 0.3 percentage points.

Other differences tend to understate the British rate as compared with the American. Among these are (1) exclusion under the British system of certain unemployed who were temporarily ill (0.1 point); (2) underregistration of married women who have exercised the option of not being covered by the social insurance system, female "new entrants," etc., (0.4 to 0.5); (3) underregistration of teenagers (0.2 to 0.3); (4) underregistration due to voluntary turnover (0.1 to 0.2); (5) underregistration of retired workers (0.1); and (6) delay in registration (0.1 to 0.2).

In net, these calculations suggest that British unemployment rates in times of low unemployment would amount to 2.1 to 2.6 percent, if Ameri

16 The material on Great Britain presented in this section is based largely on an article by Joseph S. Zeisel, "Comparison of British and U.S. Unemployment Rates," Monthly Labor Review, May 1962, pp. 489-501. For a highly informative official description of British unemployment statistics, see Guides to Official Sources: No. 1, Labor Statistics (London, H.M.S.D., 1958), p. 77 ff.

17 That is, when the U.S. unemployment rate was about 4 percent and the British rate about 1.5 percent.