Committee on Economic Security (CES)

"Social Security In America"






Through procedures outlined in appendixes I and II, an attempt was made to collect and arrange pertinent data into an order usable in estimating the cost of an assumed unemployment compensation system for the United States. The financial considerations of compensation for unemployment will be determined by the income derived from contributions, by the compensable wage loss of the employees covered, and by the duration of unemployment. The cost of the system is dependent in part upon the length of the benefit period, and this in turn is determined by the established policy of the system, i. e., whether solvency is to be maintained throughout a complete business cycle, including a major depression, or during prosperous periods and minor depressions only. The maximum duration of benefits established under either policy would doubtless be maintained throughout the business cycle, but a deficit would be incurred during major depressions if the system had been constructed on the basis of maintaining solvency during minor depressions only.

In the pages which follow the maximum duration of benefits for the unemployed has been estimated for the United States as a whole on the basis of maintaining solvency through what might be considered a major business cycle, namely, the years 1922-33. A considerably longer maximum duration of benefits would have resulted if the depression years 1931-33 had been excluded and the computations had been based on the unemployment experience of 1922-30.


Since the source of funds for operation of the assumed plan of unemployment compensation is a percentage tax on pay rolls, it was necessary to obtain the total amount of pay roll taxable in order to approximate the probable income yielded by the tax. The general procedure was to determine the average earnings of the covered group each year and to apply this figure to the employed compensable labor force. The data available, however, necessitated the division of the experience period into two segments, 1922-28 and 1928-33. The source of data for the first period was The National Income and Its Purchasing Power, by Willford I. King, published by the National Bureau of Economic Research in 1930. The estimates of wages and salaries for all groups except agriculture, government, and unclassified industry were used, the latter industry being eliminated since it consisted chiefly of service employees of whom only a small portion were assumed to be covered by the system. To arrive at an estimate of average earnings for the other half of the experience period, National Income, 1929-32 {1} was used. Agriculture, government, and domestic and personal-service industries were excluded from the calculations, and yearly aggregate and average earnings were calculated for the residue.

{1} U. S. Bureau of Foreign and Domestic Commerce, Division of Economic Research, Letter from the Acting Secretary of Commerce, National Income 1929-32, S. Doc. No. 124, 73d Cong., 2d sess. (U. S. Government Printing Office, Washington, D. C., 1934).


The procedure described above resulted in the assessable earnings estimates shown in table III-1. Contributions at rates of 3, 4, and 5 percent are shown in table 16 on p. 81.



The general method of determining the benefits possible on the basis established in appendix II was to estimate the total compensable wage loss for each year and to distribute it according to the composite "percentage" distribution of duration of unemployment (table 18, p. 84).

To arrive at the total compensable wage loss, the basis used was that employed in determining the income from the compensable labor force. In table III-1 the estimated average earnings of the employed compensable labor force are displayed for each year. It was assumed that this would be the wage base used in calculating benefits. ( See table 17 on p. 82. ) Since a benefit rate of 50 percent was assumed, with a maximum of $15 per week, the average compenaable wage loss (per man-year) was estimated to be 45 percent of the wage loss per man-year.

This assumption would not be strictly accurate, since the provisions of the plan stipulated that benefits were to be one-half of the usual full-time pay rather than the average pay. A correction in wage-loss estimates for this provision was impossible, first and chiefly because of the absolute lack of pertinent data, and second, because of the difficulty of defining the usual full-time wage. A correction factor for this inaccuracy is included in the actuarial, adjustments. (See p. 419.) The average compensable wage loss having been computed, it was multiplied by the estimated total man-years of unemployment (derived from table 17) to arrive at the total compensable wage loss for each year. The total compensable wage loss for each year was then distributed in accordance with the assigned distribution of the duration of unemployment previously displayed in table 18 (p.84).

The wage-loss distributions for 1923 through 1933 were combined into one cumulative distribution, as- shown in table 19 on p. 86. This table may be readily used to ascertain the number of weeks of benefits payable with a 3-percent contribution rate under a system aiming at solvency through 1933. For example, the estimated total income available for benefits for the period 1923-33 is $8,746,000,000. If a waiting period of 4 weeks is required, $5,026,000,000 of the compensable wage loss will not be compensated by benefits. The sum of these two is $13,772,000,000, which in table 19 falls between the compensable wage losses for 20 and 21 weeps of unemployment. If benefits are paid for 17 weeks


(21 weeks less the 4-week waiting period), $8,970,000,000 will be paid out (13,996,000,000-5,026,000,000) and the fund will have a deficit of $224,000,000. If benefits are paid for 16 weeks (20 weeks less the 4-week waiting period), $8,586,000,000 will be expended (13,612,000,000-5,026,000,000) and the fund will retain a surplus of $160,000,000 (8,746,000,000--8,586,000,000).

Table l9 can also be used in similar manner to ascertain the combination of any waiting period and duration of benefits which will match the total income available for benefits. The results of the preceding method of estimating maximum duration and amount of benefits take no account of factors yet to be discussed, which are not measurable by available data and which tend to change the maximum benefit period. Before final cost estimates can be obtained adjustments for these factors should be considered.


There are a number of factors in the legislative provisions of the assumed unemployment compensation system which have not been evaluated in any of the previous sections and which will definitely affect the cost; furthermore, there are limitations in the data which will also affect the cost and for which no adjustments have been made; and, lastly, the effect of the imposition of a plan of this nature must be measured in terms of claim experience. The present section attempts to evaluate these factors.

The surveys of unemployment which have been analyzed and built up into a series of distribution tables showing the percentage of employees who are unemployed for varying intervals (table 18, p. 84) give so impressive a display and have been forced into so orderly an arrangement that it is very difficult to remember throughout that the material is not directly applicable to the problem of cost determination. The data cover only unemployment at the time of the surveys and do not show total unemployment records as they will develop over a period of time. Dr. I. Rubinow carefully points out the inherent limitations of this type of census, but says that nothing better is available.{2} Mr. A. D. Watson's actuarial report on the Canadian unemployment insurance bill,{3}based on censuses showing aggregate duration of unemployment during the preceding 12 months, provides data which are a little closer to the form of duration table applicable to benefits of this sort. The Railroad Retirement Board's data on the duration of unemployment among a closed group of railroad employees which had been put on furlough and taken hack into employment also were valuable as a check on the distributions developed.{4} However, comparison of these distributions reveal limitations in each of these sets of material. The Canadian data are incomplete for the period of unemployment which occurred prior to the beginning of the 12 months' period but which was completed within the 12-month period, and, likewise, it was impossible to ascertain the duration of the unemployment of an unemployed person beyond the date that the census was taken. In the railroad data information on a special group of men taken back into their original employment is not indicative of what may have happened to the other employees whose unemployment histories were not available after severance of employment.

{2} Dr. I. M. Rubinow in his discussion of a paper, "Calculation of the Cost of Unemployment Benefits With Particular Reference to Ohio and Pennsylvania," by Prof. Clarence A. Kull) (Casualty Actuarial Proceedings, vol XX, pp.170-179).

{3} Actuarial Report on the Contributions Required to Provide Unemployment Insurance Beneffts Within the Scheme of the Draft of an Act Entitled "The Employment and Social Insurance Act" (No. 158A-1935).

{4} Unpublished data furnished by the Railroad Retirement Board.


When one attempts to modify a duration-frequency table he must recognize that any modification requires that the whole schedule be completely replaced by some other schedule or schedules. In the absence of material which would provide an adequate substitute it seemed advisable to use the estimated distributions in their present form and then attempt to assess in some aggregate fashion the weight of the many factors which should be considered and which might modify the result.

The various factors arising from special provisions of the plan and from the inadequacy of the data may now be weighted and an allowance given to other immeasurable contingencies.




For the estimates of duration of benefits the total adjustment determined in the previous section was applied to the material presented in appendix II.

The 30-percent adjustment factor was applied to the wage lost in each duration interval for the period 1923-33. From the tabulation on page 89 it may be observed that with a 3-percent contribution rate and a 4-week waiting period the maximum length of benefit period permitted by the adjusted figures would be 12 weeks (as compared with 16 weeks permitted by the unadjusted data) under a plan designed to remain solvent through 1933. By matching income approximations with wage-loss estimates, table 20 (p. 87) can be used to determine the maximum duration of benefits and maximum amount of benefits for any combination of contribution rates or waiting period.

However, it is recommended that not too trusting a use be made of the intervals of the distribution as one approaches 52 weeks, since the degree of the reliability of the distribution decreases as the length of the duration interval increases.

If the features upon which the above adjustment factor depends are changed, the size of this factor will change, and so also the wage-loss distribution. Once the size of a new adjustment factor is determined upon, the wage-loss distribution may be computed by multiplying the amounts shown in table 19 (p. 86) by 1 plus the percentage adjustment.

Changes in actual benefit requirement may introduce other modifications, but the crudeness of the basic data has been sufficiently set forth to suggest that little addition of accuracy will be gained as a result of any further small adjustments.

Many of the above factors used in adjusting the estimates of the duration of compensable wage loss are little more than reasonable guesses based on the best judgment which could be exercised. Only a few of the factors could be derived from basic data even of limited nature and scope.