^{page 75 note (1)} Seasonally adjusted.

^{page 75 note (2)} See, for example, D. Gujarati, ‘The beh aviour of unemployment and unfilled vacancies’, Economic Journal, vol. 82, no. 325, M arch 1972.

page 76 note (1) Assuming there was no change in the labour force of demographic origin.

^{page 77 note (1)} See ‘People and jobs’, Department of Employment, page 14, where some evidence is presented on the effectiveness of such efforts.

^{page 78 note (1)} ‘The change in the relationship between unemployment and earnings increases: a review of some possible explanations’, by J. K. Bowers, P.C. Cheshire, and A. E. Webb, National Institute Economic Review no. 54, November 1970.

^{page 79 note (1)} In addition, employers may have a greater incentive to record their vacancies, and voluntary quitters a lesser incentive to register, as employment expands.

^{page 79 note (2)} Rundown of employment can be by ‘natural wastage’.

^{page 79 note (3)} A structural shift here refers to a change in the ‘matching’ of jobs (vacancies) to people (unemployment)—this change may occur regionally, by industry, or by occupation, or by some combination of the three, in response, for example, to changes in the pattern of demand and output.

^{page 79 note (4)} A reduction in the level of desired employment for any level of output. The arguments of this paragraph are ceteris paribus, i.e. they assume that aggregate demand is fixed.

^{page 80 note (1)} Taking antilogs, we have:

^{page 80 note (3)} J. K. Bowers, P. C. Cheshire, and A. E. Webb, op. cit.

^{page 80 note (4)} There are a number of other technical problems. The unemployment figures used are the quarterly series of wholly unemployed excluding school-leavers. Some of the recorded engagements may be of school-leavers and it is probable that some of the temporarily stopped will appear in the discharge statistics (see H. A. Turner, ‘Measuring unemployment’ Journal of the Royal Statistical Society, Part I, 1955). Also, some people will die while holding jobs, and this will give rise to engagements. Furthermore the turnover statistics are a sample of one month per quarter expressed as a rate; ΔU_r is the quarterly first difference, so that there is a scaling problem. Additionally the divisor for the unemployment rates is the June estimate of employees. The turnover rates use monthly estimates of employment in manufacturing industry.

^{page 81 note (1)} Turnover data are derived from the same inquiry as the monthly series of employment in manufacturing industry.

^{page 82 note (1)} The quarters differ in the two approaches. For example, in the U-V analysis quarter 2 refers to the observation for June. In the turnover analysis, 2 refers to the period mid- April to mid-July.

^{page 82 note (2)} Taking the value zero until 1966 III and the value of O_t from then until 1970 II. There is no evidence of a shift in the coefficient of the change in output variable.

^{page 83 note (1)} i.e. the aggregate relationship is reducible to a function of U and V in any one of the sub-markets.

^{page 83 note (2)} In which case the relative levels of demand in the various sub-markets have altered. Given non-linearity in the underlying relationships this, as already explained, will shift the aggregate U-V curve outwards on most feasible assumptions.

^{page 83 note (3)} J. K. Bowers, P. C. Cheshire, and A. E. Webb, op. cit. page 48. See also NIER no. 55, February 1971, chart 7, page 19.

^{page 83 note (4)} National Institute Economic Review no. 59, February 1972, pages 19-21.

^{page 85 note (1)} i.e. in table 4, D₃ and/or T₄ are positive and significant for all occupations. This analysis also shows a number of interesting features not directly concerned with the question in hand, e.g. a number of diverse occupations, such as administrative, technical, and professional workers, engineering and allied workers, and labourers show an inward shift from about 1962. Also there is a marked seasonal pattern (not reported in table 4) among clerical workers, which changes significantly in 1966.

^{page 86 note (1)} This figure is based on an unpublished analysis of the unemployed by R. Weeden using life table methods normally employed in demography.

^{page 86 note (2)} Whether in practice the ‘last in first out’ principle does operate with redundancies has been questioned. See e.g. D. I. MacKay, D. Boddy, J. Brack, J. A. Diack and N. Jones, ‘Labour markets under different employment conditions’, Allen and Unwin, London, 1971. They argue that redundancies usually occur on a whole shop or secton basis making length of service irrelevant.

^{page 86 note (3)} A supply shift composed entirely of an increase in the time spent on the register is not likely, since employers faced with increasing difficulties of recruitment would respond by bidding labour away from each other thus inducing a rise in the voluntary quit rate. R. Weeden finds, in fact, that both the mean duration of unemployment and the rate of entry to the register have increased in the post-1966 period.

^{page 86 note (4)} Discussed in J. K. Bowers, ‘British work on activity rates: a survey’, Paper presented to the first SSRC Conference on Labour Economics. Mimeo.

^{page 86 note (5)} Except where ‘just cause’ for quitting is shown. The same rules apply to those sacked for disciplinary reasons, except that the onus of proof then rests on the employers. The distinction between voluntary quitting and disciplinary sackings is blurred in practice.

^{page 86 note (6)} Registration is usually a requirement for receiving supplementary benefit but the entitlement of the typical voluntary quitter is dubious. Furthermore the same constraints on job-search and job-choice would apply for supplementary benefit.

^{page 87 note (1)} See D. I. Mackay et al., op. cit.

^{page 87 note (2)} There is evidence of a significant slope shift of F_r against E_r which explains the presence of autocorrelation in the reported equation where no slope-shift dummy is present.

^{page 87 note (3)} We are grateful to A. Gillie of the University of Leeds for this suggestion.

^{page 87 note (4)} Before this date benefit was paid for a minimum of 180 days up to a maximum of 492 depending on contributions paid and benefits received over the previous decade. Subsequently all qualifying for benefit were entitled to 312 days (one year of six-day weeks). To receive benefit at all a minimum of 26 stamps had to be paid or credited in the previous benefit year. After exhaustion of benefit, a man requalified after paying another 13 weeks' contributions. Post-1966 his duration of entitlement was not affected by his previous benefit. Pre-1966 the duration of benefit was affected.

^{page 88 note (1)} Number of wholly unemployed men less those unemployed for more than 52 weeks and, until 1966 IV, less also a proportion of men under 21 who had been unemployed for between 26 and 52 weeks.

^{page 88 note (2)} There are one or two aberrant observations in the 1962-63 slump but these are arithmetically small. They are, however, the cause of some autocorrelation in the residuals.