^{Note (1) page 27} Married women were treated separately from single.

^{Note (2) page 27} These periods were chosen because the pressure of demand for labour, as measured by unemployment, was about the same in the first half of 1952, the first half of 1957, and the first half of 1961.

^{Note (3) page 27} For a discussion of some of the factors involved, see ‘Married Women at Work in 1972’ by W. Beckerman & J. Sutherland, National Institute Economic Review, No. 23, February 1963.

^{Note (1) page 28} See Appendix, page 34.

^{Note (1) page 29} This cyclical behaviour of employment in manufacturing industries was one of a number of subjects recently explored by Mr. R. R. Neild, assisted by Mr. J. R. Shepherd, in Prices and Employment in the Trade Cycle, Cambridge University Press, 1963. In this study it was convincingly established that short-term changes in employment were proportionately smaller than those of output and that they responded with a time lag.

^{Note (1) page 30} As explained in the Appendix, it was not possible to determine with any precision what were the effects on capacity of changes in the labour supply or in normal hours. We have, therefore, imposed values for these relationships, the bases of which are given in the Appendix, page 35.

^{Note (1) page 31} That is, the arithmetic mean of the absolute error, irrespective of sign. The values of employment calculated from these relationships were obtained by taking an initial value for actual employment, and then by applying the adjustment value to equilibrium employment and the calcu lated value of employment in the previous quarter. This mean error compares favourably with the mean error of 0.4 per cent for changes over four quarters obtained from the best equation in Mr. Neild's study. But fluctuations in manufacturing employment tended to be proportionately bigger than in total employment.

^{Note (2) page 31} The mean discrepancy between ‘forecast’ and actual employment between 1962, first quarter, and 1963, second quarter, was 25 thousand or 0.10 per cent.

^{Note (1) page 32} The mean discrepancy between these revised figures and the forecast is 23 thousand, or 0.09 per cent.

^{Note (2) page 32} The revision to the June 1963 figure was first announced in October of that year.

^{Note (1) page 33} Actual unemployment here refers to the recorded level in the second quarter, and so differs from the logged series in table 5.

^{Note (1) page 34} One further consideration was introduced; the late age entrants to the National Insurance scheme became eligible for pensions in 1958 provided they had continued their contributions. The number who left employment as a result of their eligibility was less than the number who started receiving pensions, since some people continued at work with their pensions and others, having already retired, had continued their contributions. The change in 1958 in the participation rates of elderly people suggests that about 130 thousand left the labour force from this cause, and this number we have deducted from D in 1958 and 1959. But for the scheme, however, these people would probably have retired before 1958. We therefore assumed that the scheme added 130 thousand to D in the four years prior to 1958, as follows, in thousands: 1953/54 +5, 1954/55 +10, 1955 /56 +25, 1956/57 +35, 1957/58 +45, 1958/59 −100, 1959/60 −30.

^{Note (1) page 35} The experience of 1947, when registered unemployment rose above a million is unfortunately no guide to the behaviour of the relationship when the pressure of demand is very low, since the greater part of the increase was in the temporarily stopped.

^{Note (2) page 35} This is a ‘compromise’ estimate of gross domestic pro duct at constant prices, which makes use of the official estimates of income, expenditure and real output. The way in in which they were prepared is described in ‘Measuring National Product’ by W. A. H. Godley and C. Gillion, National Institute Economic Review No. 27, February 1964.

^{Note (3) page 35} This is roughly comparable with 1 1/2 per cent for the unemployment percentage as usually expressed—that is, the unemployed, including the temporarily stopped, as a percent age of the total number of employees.

^{Note (1) page 36} The arithmetic mean of the absolute errors regardless of sign.

^{Note (1) page 38} Estimated from the variances and covariances of the regression coefficients.