^{page 63 note (1)} Previous unpublished work on this question was presented to a conference of the Contact International de Récherches Economiques Tendancielles, in a paper by R. N. Hart and C. B. Howe, in April 1963.

^{page 63 note (2)} The answers to most of the questions about expectations are not published. However, the FBI have permitted us to use all the expected figures, except those from the latest (October 1963) Enquiry.

^{page 64 note (1)} An increase in the balance can be brought about either by a rise in the percentage saying ‘up’, or a fall in the percentage saying ‘down’, or, more commonly, by a combination of the two. There seems no reason to treat differently, in the analysis, changes which are brought about in these different ways.

^{page 64 note (2)} The statistical results are given in Appendix I, table 1.

^{page 64 note (3)} This would also help to explain why the FBI balances are generally fairly smooth series.

^{page 64 note (4)} The correlation between the reported balances and the four-monthly changes is 0.66; between the reported balances and the twelve-monthly changes, it is 0.92.

^{page 65 note (1)} The best fit, for the period February 1958-June 1963, was obtained by giving about equal weight to the actual balance and to the change in the balance.

^{page 66 note (1)} Before July 1963 the index did show a sharp, though short-lived recovery in output in the first half of 1961. This has been almost removed by the latest revisions to the seasonal adjustments.

^{page 66 note (2)} These changes are clear now, and show clearly in charts 1 and 2. But they were not, of course, clear at the time, both because of the time-lag in the publication of the official figures, and because provisional figures were later revised. In the text, we consider whether the FBI balances gave any advance information, or any better information, compared with the official figures available at that time.

^{page 66 note (3)} The calculated series had suggested a slight slowing-down in the rise in output between June 1959 and February 1960— a slowing-down which did not occur. This is because, in this series, the change in the balance has an important weight; and from June 1959 to February 1960 the balance rose much less than in the previous eight months. By October 1959, there was probably not much scope for any further increase in the balance; for the number of firms reporting ‘down’ was very small, and there may be a hard core of permanent ‘no changes’.

^{page 66 note (4)} It is tested as well as possible in Appendix I, by using expected future trends, as well as reported trends, to explain changes in the past four months. This addition does improve the explanation appreciably : in fact less is lost by omitting past trends than by omitting expected future trends.

^{page 67 note (1)} This is because of the influence of the figures for the change in the balance; the balance had been falling sharply up to February 1962, and in June this changed to a slight rise.

^{page 68 note (1)} To show this point, the figures of the expected balance are plotted in this chart on the date of their publication : so the figure published in February of the expected balance in four months' time is plotted at the beginning of the year.

^{page 68 note (2)} The new order series was tried, as an alternative to the expected output balance, in the forecasting equation for pre dicting future output changes (see Appendix I, page 75). The result was a little better than that obtained by using expected output. The calculated coefficients confirm that it is the difference between the reported balances for new orders and for output, and not the actual value of the balance for new orders, which should be used as an indicator of future output. The new orders series was used, rather than the length of order books, since new orders is the less closely correlated with the current output balance (the correlation coefficient was 0.945 as compared with 0.984 for the length of order books) and would in any case be expected to be a more sensitive series. Although orders may take a relatively long time to complete, changes in the inflow of new orders could affect output quite quickly if there is spare capacity, so that a short period correlation is not theoretically implausible.

^{page 68 note (3)} The definition of the ‘balance’ used for exports is slightly different from that used elsewhere. This is because the sample was changed in June 1961, when firms with little or no export business were no longer asked to reply. Some of these seem to have replied ‘no change’ rather than ‘not available’, for the proportion of ‘no changes’ has tended to be lower since this change. For this reason the balance was used in the adjusted form, ‘100 x (ups — downs)/(ups + downs)’. This series is very similar to the balance used for other questions though its amplitude of movement is greater and the relationship between the two may not be precisely linear.

^{page 68 note (4)} In May 1959 we still took the view that ‘exports show no clear recovery yet’ (National Institute Economic Review, May 1959, page 6), while in July we treated the known increases as significant. This referred to total exports but the facts were essentially the same as for manufacturing.

^{page 70 note (1)} The early stages of the recovery in the FBI balance since the middle of 1962 would seem to imply, for a time, only a slower rate of fall in capital expenditure.

^{page 70 note (2)} See page 10 of this Review.

^{page 70 note (3)} In recent years, these provisional estimates have not been revised very extensively, although the seasonal adjustments have been changed.

^{page 71 note (1)} We assume that businessmen are reporting the level of employment at about the end of the previous month : for instance, the June enquiry reports the level of employment towards the end of May. Until May 1962 the employment figures were also for the end of the month, and would have been available about three or four weeks after the FBI enquiry. Since then employment figures have related to the middle of the month, and the publication delay has remained at 5-6 weeks.

^{page 71 note (2)} See National Institute Economic Review, August 1963, page 5, and page 5 of this number.

^{page 71 note (3)} Forecasting equations based on both actual and on expected balances are shown in Appendix I, table 1, page 76. In these equations it is the level, and not the rate of change, of the balances which is the important factor in calculating employment changes.

^{page 72 note (1)} This shows up particularly in the engineering group of industries, in which most of the workers had a two-hour reduction in March 1960.

^{page 72 note (2)} In the FBI price question, since goods will be reported on at various intermediate stages of production, the ratio of the material component of their value to the wages component may be much higher than the ratios for the values of finished manufactures only. Therefore the FBI series is likely to be more affected by input prices and less affected by wages than the official index. The movement of the two series will also be different since wage and price movements take effect at different times.

^{page 72 note (3)} Drink and tobacco are excluded because changes in duty have had relatively large effects on the index; they have probably affected the balance much less, since only a few drink and tobacco firms are included and in any case some of these firms might consider that the question did not cover changes of this sort.

^{page 73 note (2)} This could be a consequence of a hard core of ‘no changes’, when the number of ‘ups’ had fallen to a very low figure. Compare the possibility of a ceiling for the balance of answers to the output question : footnote 3, page 66.