^{note (1) in page 36} Effects of the Selective Employment Tax, First Report. ‘The distributive trades’, by W. B. Reddaway, HMSO, 1970. Referred to hereafter in this note as ‘The Reddaway Report’, or simply, ‘the Report’.

^{note (2) in page 36} Reddaway Report, chapter XI.

^{note (3) in page 36} The actual methodology used calculates per cent savings in employment as compared with what would be expected from previous experience. If one were to measure in terms of productivity being higher than would otherwise be expected, these gains would be slightly larger.

^{note (4) in page 36} The authors state that they have not yet been able to devise a technique for separating the effects of the ending of RPM from those of the introduction, some time later, of SET.

^{note (5) in page 36} The form of the equation is: ΔE = α₀ + α₁ (Δ0 - Δ0) + α₂D + α₃(V-V) + U where: ΔE = percentage change in employment between the two years in question; Δ0-Δ0 = difference between the percentage change in output between the two years in question and the average of eleven such percentage changes in the period 1954 to 1965; D = deficiency in employment in the first of the two years in question, expressed as a percentage; V-V = difference between the percentage of vacancies in the second of the two years and the average percentage for the eleven years 1955 to 1965; U = a random disturbance.

^{note (1) in page 37} The estimated equation is: (See Appendix F in the Report, page 297)

ΔE = 0.786 + 0.793 (Δ0 - Δ0) + 1.246D - 2.010 (V - V) (0.209) (0.172) (0.443) (0.999) R² = 0.62 D.W. = 1.28

The figures underneath the coefficients indicate the standard errors. On the face of it, the value of the coefficient on D, which is greater than unity, implies that employers over- adjust to the deficiency of labour. However, this coefficient has a large standard error and at customary levels of con fidence is not significantly different from unity. The Durbin- Watson statistic is not given in the Report, but was calculated from the residuals.

^{note (2) in page 37} The method of calculation of ‘D’ is as follows:

D = [(D* - E)/D*] x 100 per cent for the estimation period and the first prediction period, where D* = required employment, i.e. output in the first year divided by trend productivity, and E is employment in the first year. For all prediction periods after the first, D = {[(D* - e_nD*) - E] / (D* - e_nD*)} x 100 per cent where en = Σⁿe_i and e_i are the residuals in the pre diction period.

^{note (3) in page 37} For a full description of this procedure see the Report, op. cit., pages 90-94 and Appendix F, table F5 page 298.

^{note (4) in page 37} This is recognised in the Report, op. cit., pages 90-94.

^{note (5) in page 37} The Report, op. cit., page 99.

^{note (6) in page 37} The equation used was ΔE = 2.73 + 0.66 (Δ0 - Δ0) + 0.83D (0.41) (0.14) (0.40) R² = 0.70 S.E. = 1.35 D.W. = 0.93 This equation differs from the manufacturing and retailing equation inasmuch as a split trend was used rather than a single linear trend because the graph of engineering produc tivity shows a sharp acceleration around 1963. Had we used a linear trend throughout the estimating period we would have got a poorer statistical fit. We would, however, have produced a larger ‘abnormal effect’.

^{note (1) in page 38} A rise in productivity simultaneously in both sectors would be compatible with theories where the released labour from the service trades was absorbed by industries such as manufacturing which, it is argued, have increasing returns to scale. However, as from 1965/6, the trend of employment in manufacturing has been downward.

^{note (2) in page 38} See, for example the National Institute Economic Review no. 43, February 1968, pages 5-7; no. 44, May 1968, page 9; no. 51, February 1970, pages 34-37.

^{note (3) in page 38} The data on employment used for 1969 and 1970 is not identical to that used in the Report. We have used employees in employment whereas the Report uses a specially prepared index of employment (described on page 36).