Hostname: page-component-78c5997874-lj6df Total loading time: 0 Render date: 2024-11-19T14:24:26.823Z Has data issue: false hasContentIssue false
  • English
  • Français

Predictions Derived from the Employment Forecast Survey*

Published online by Cambridge University Press:  07 November 2014

Douglas Hartle
Douglas Hartle*
Affiliation:
University of Toronto
Get access

Extract

Reliable predictions of future employment are but one of many pieces of intelligence useful to governments in the formulation and implementation of policies of economic stabilization. With the adoption of a stabilization programme in 1945 an explicit need for these predictions was created in the Canadian government. In that year the “Employment Forecast Survey” was initiated by the federal Department of Labour in an attempt to secure the data from which these predictions could be derived. This survey is still conducted by the Department.

This article summarizes the results of an analysis of the predictions which were derived from data obtained by the Employment Forecast Survey from 1946 until 1957, when this method was abandoned, and a new technique of securing predictions from the data was adopted. Because these predictions were so little known outside Canadian government circles it may be useful to provide a brief description of the technique by which they were made, and an assessment of their reliability.

The original conception of the survey was comparatively simple. It seemed reasonable, a priori, that the individual decision-makers in an industrial establishment would be able to forecast the future employment of their own establishment with considerable reliability.

Type
Research Article
Information
Canadian Journal of Economics and Political Science/Revue canadienne de economiques et science politique , Volume 24 , Issue 3 , August 1958 , pp. 373 - 390
Copyright
Copyright © Canadian Political Science Association 1958

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

Footnotes

*

This study was made possible by a grant from the Commonwealth Studies Committee of Duke University, and the active co-operation of the Economics and Research Branch of the Department of Labour, Ottawa. I am also indebted to Dr. Wm. C. Hood, of the Department of Political Economy, University of Toronto, who made the first major investigation of the “Employment Forecast Survey” in the summer of 1953. The present study, which benefited directly from several valuable suggestions made by Dr. Hood, is an extension and elaboration of the analysis initiated in his report entitled “The Employment Forecast Survey: A Report Submitted to the Director of the Economics and Research Branch of the Department of Labour,” Ottawa, Nov., 1953 (126 pp., mimeo., classified confidential). The advice of Dr. Frank A. Hanna of Duke University is also acknowledged with gratitude. When this study was conducted, I was an employee of the Economics and Research Branch of the Department of Labour, but the views expressed are not necessarily those of the Department.

References

1 Empteyment and Income, with Special Reference to the Initial Period of Reconstruction, a White Paper presented to the Canadian House of Commons by the Minister of Reconstruction, Rt. Hon. C. D. Howe, April, 1945.

2 It should not be inferred that the “Employment Forecast Survey” was the principal, much less the sole, experimental endeavour of the Canadian government in the realm of economic forecasting. The most complete published statements on the government's use of economic forecasts are found in Bates, Stewart, “Government Forecasting in Canada,” this Journal, XII, no. 3, 08, 1946, 361–78Google Scholar; Firestone, O. J., “Government Economic Intelligence in Canada,” Public Affairs (Dalhousie University), XII, spring, 1949, 5–11 Google Scholar; and Firestone, O. J., “Investment Forecasting in Canada” in National Bureau of Economic Research, Short-Term Economic Forecasting, Studies in Income and Wealth, XVII, (Princeton, 1955), 113–247.Google Scholar

3 Indexes of the actual employment of one industrial classification, which is an adaptation of the Standard Industrial Classification, are published each month (between two to three months after the date to which they apply) in Employment and Payrolls, by Employment Section, Labour and Prices Division, Dominion Bureau of Statistics. In the majority of cases, the EFS industrial classification is the same as the Employment and Payrolls classification, so that the employment indexes given in the latter can be projected. Hereafter in this paper, all these actual employment indexes are referred to as “DBS employment indexes.”

4 This notation is an adaptation of that employed by Hansen, Morris H., Hurwitz, William N., and Madow, William G., Sample Survey Methods and Theory (New York, 1953), I, 110–16, 179–83.Google Scholar

5 To be exact, Nh and Bh (below), should be dated, for the number of (population) establishments that report to DBS changes slightly from month to month because of births and deaths and a few non-reporting establishments.

6 A number of important changes were made in the industrial classification prior to January 1, 1952. These changes preclude the measurement of EFS predictive errors of the present industrial classification, except for manufacturing as a whole.

7 The EFS reports have included predictions for four selected non-manufacturing industries and for twenty-four industrial components of manufacturing. These details have been omitted here.

8 For reasons which will become obvious from the description of the technique of derivation that was used, the hypothetical predictions do not purport to give the direction of the non-seasonal changes in employment.

9 Had the errors of the EFS predictions been stable in terms of direction and magnitude, improvements in reliability perhaps could have been achieved by the application of constant corrective factors for each industry. It has been found that the signs of the errors have not been consistent over time, for any of the industries or groups. In many cases, the positive and negative errors cancelled each other almost completely when averaged (signs taken into account) over the period under consideration. Any attempt to adjust the predictions by a constant corrective factor probably would have substantially increased the errors of some of the predictions.

10 Spot checks show that the omission of about 2 per cent of the forecasts of establishments at each target-date makes virtually no difference in the EFS predictions for the vast majority of industries and target-dates.

11 The percentage changes over the standard projection intervals are defined in the section on Projection Procedures.

12 Computational resources did not permit further calculations, although the basic data are available for the period from January 1, 1951, to July 1, 1956.

13 It is interesting to note that the published EFS indexes based on these forecasts were among those which had the smallest errors of all the EFS indexes taken into account in this analysis.

14 It should be noted that the relationship here is between the actual employment and the forecast employment of the individual establishments, and not the changes in their employment which have been considered previously. Because the object is to minimize the errors of the aggregated sample forecasts, the relationship between the changes in employment would be inappropriate because the employment weights of the establishments would not be taken into account.

15 The computation of the regression equations for each target-date, by size of establishments, is a crude attempt to avoid the weighting difficulties discussed in n. 14.

16 The difference between the actual and the forecast employment of the establishment on the target-date, as a percentage of the actual employment on die target.

17 The median percentage errors of the establishments' forecasts (three and six months) were computed by target-date. The median errors were then compared with a seasonally adjusted index of the actual employment of the establishments in the sample. The sample indexes were computed as the total employment of the sample establishments, for a given target-date, as a ratio (times 100) of the sum of the average employment of the same establishments in 1953 (average employment 1953 = 100). These indexes were not derived by the use of explicit weights for individual industrial components.

18 The complete elimination of all sampling errors could only have been achieved by a census of the approximately 10,000 manufacturing establishments in the population, Because it is highly unlikely that sufficient resources would have been available for a census, these estimates no doubt overstate substantially the magnitudes of the reduction in sampling errors which would have been feasible.

19 The mean errors of the three-month forecasts, averaged without regard to sign over the twenty target-dates, were as follows for the various size classes: 50–99 employees, 21.93 per cent; 100–199 employees, 7.81 per cent; 200–499 employees, 2.55 per cent; 500 employees or over, 2.12 per cent. The errors of the six-month forecasts showed the same progression, by size class.

20 E.g., see Ferber, Robert, The Railroad Shippers' Forecasts (Urbana, 1953)Google Scholar; and Modigliani, Franco and Sauerlender, Owen, “Economic Expectations of Firms in Relation to Short-Term Economic Forecasting” in Short-Term Economic Forecasting, 261–362.Google Scholar

21 See Hastay, Millard, “The Dun and Bradstreet Surveys of Businessmen's Expectations,” Proceedings of the Business and Economic Statistics Section, American Statistical Association, Montreal meeting, 1954 (Washington, n.d.), 93–123 Google Scholar; and Anderson, Oscar Jr., “The Business Test of the IFO Institute for Economic Research, Munich, and Its Theoretical Model,” Review of the International Statistical Institute, XX, 1952, 1–17.CrossRefGoogle Scholar