Drop in unemployment looks like the real thing

The 47,000 fall in unemployment in December was the fourth successive monthly decline, and for the first time in the present recovery we are now seeing confirmation of an improving labour market from a strong rise in unfilled vacancies.

Despite this, there are still some voices suggesting that the decline in the jobless total is bogus. For example, it is frequently argued that the reduction in unemployment during the current recovery has occurred 'surprisingly early' compared with past cycles.

In addition, it has been pointed out that the Department of Employment (DoE) statistics show a significant shrinkage in the potential labour force in the past two years, which suggests that the decline in unemployment may have been caused by the disappearance of some workers from the recorded labour force - perhaps because of greater efforts by the employment services to push people off unemployment benefit - rather than to the creation of new jobs.

What does the evidence show? In the current cycle, GDP bottomed in 1991/92 and unemployment started to decline 12 months later. By contrast, in the economic recovery of the early 1980s, unemployment continued to increase for about five years following the trough in GDP. And in the late 1970s recovery, unemployment continued to increase for about 30 months after the GDP trough.

To some extent, this improved performance is simply a reflection of demographic behaviour. For example, in the first two years of the post- 1981 recovery in activity, the population theoretically available for work - all those aged 16 to retirement age - increased by 100,000 per quarter. Of these, many chose to become economically inactive, such as students and homekeepers. But the total economically active population rose by 27,000 per quarter.

So far in the present recovery, the total population aged 16 to retirement age has increased by only 19,000 per quarter and the economically active population has actually shrunk by 33,000 per quarter.

The quarterly difference in demographic trends between the current cycle and the last one is therefore about 50,000 for the active population and 80,000 for the total population.

At present, unemployment is falling by around 120,000 per quarter, compared with an increase of around 30,000 per quarter at the same stage in the 1981-83 recovery. This represents an improvement of about 150,000 per quarter, up to half of which can therefore be attributed simply to demographic factors.

Much, if not all, of the remainder can be ascribed to the fact that employment started to rise rather earlier this time than it did in the mid-1980s. In fact, the difference in the trend for employment appears to be around 150,000 per quarter, which could comfortably translate into a difference in unemployment of 70-100,000 per quarter. (Typically, the change in claimant unemployment is smaller than the change in employment because some of those moving into and out of employment will come from the ranks of the unregistered unemployed.)

Looked at in this way, the 'early' decline in unemployment ceases to be so mysterious. But why has employment picked up more rapidly than before? Two factors are relevant here.

First, in the early stages of the 1990/91 recession, there was undoubtedly a larger shakeout of employment, given the performance of output, than occurred at equivalent stages of the 1979/81 or 1974/75 recessions. This, in turn, resulted in a stronger performance by productivity this time than has been typical on previous occasions.

The stronger-than-usual performance by productivity in the 1990/91 recession may well be mirrored in a weaker performance in the recovery phase of the current cycle, reflecting the fact that companies are rapidly taking back workers that they shed in the downturn.

A second factor has been at work in the past 12-15 months, helping to explain the monthly pattern of employment and unemployment changes. This is best seen from the accompanying graph, which shows the monthly change in unemployment during the current cycle as compared with the last two.

Until last autumn, the monthly change in unemployment was not acting abnormally, but around the time of sterling's ERM crisis there was a temporary sharp increase in the monthly rate of change in the jobless total.

It now seems probable that companies became too pessimistic about economic prospects at that stage and consequently shed an exceptional number of workers in a short space of time. In the early part of 1993, employers may have corrected their output expectations in a more favourable direction and therefore decided to offset the excessive job-shedding of the previous autumn.

This leaves the puzzle of the shrinking labour force. The regular DoE series for employment and unemployment certainly suggests that there may have been some 'disappearing worker effects' in the labour market in 1992. According to this source, the combined total for employment and unemployment has fallen by 426,000 in the past 18 months.

However, we are now able to cross-check these figures by examining the results of the Labour Force Survey (LFS), which started appearing quarterly in 1992. This latter source is compiled from a sample survey of the labour market characteristics of the population, so it offers an independent check on the administrative figures compiled by the DoE.

Importantly, the decline in employment according to the LFS survey during 1992 was only about one-third that shown by the DoE series. Hence, there are almost no 'disappearing workers' to account for in the LFS data.

Meanwhile, the LFS confirms that unemployment increased considerably throughout 1992, peaked in the first half of 1993, and has since been falling. Although the LFS figures have only been published up to the summer of 1993 - covering the months from June to August - they confirm the start of the improvement in unemployment that was shown by the DoE series early last year. They also confirm the slight deterioration in the unemployment trend in the middle of last summer.

The similarity of the unemployment pattern shown in the LFS and DoE series for 1992-93 implies that the improvement in unemployment shown in the DoE figures was probably not entirely bogus and did not result solely from changes in the benefit payment procedures.

Comparing the employment and unemployment figures from both sources, it appears that the large decline in employment as shown in the DoE series for 1992 may be a rogue result.

If this is right, then the level of employment in the economy may be considerably higher than is shown in the regular DoE statistics. And because these are used in the calculation of the official figures for both productivity and unit labour costs in the economy, the latter figures may also be distorted.

In fact, comparing the DoE employment figures with the equivalents in the LFS, it is quite likely that employment in the economy is some 2-2.5 per cent higher than is estimated in the official DoE statistics. This, in turn, means that productivity in the whole economy has been overstated by the same amount, while unit labour costs have been understated to a similar degree.

More importantly, though, the LFS - a data source that deserves much more attention - appears to throw doubt on the 'disappearing worker' effect and seems to confirm that much of the decline in unemployment has probably been genuine. Despite the sceptics, the sharp improvement in the labour market is surely set to continue this year.

(Graph omitted)