Labour market slack in the UK.
Bell, David N.F. ; Blanchflower, David G.
Given the recent unexpectedly rapid fall in the unemployment rate,
the extent of labour market slack in the UK economy is an important
issue for policymakers, particularly the Bank of England's Monetary
Policy Committee (MPC). However, in our view, the MPC is arbitrarily
reducing its estimate of the impact of two important components of
labour market slack, risking damaging mistakes in the formulation of
monetary policy. The first is long-term unemployment and the second is
underemployment. In each case the MPC has explained these adjustments in
a box in its May 2014 Inflation Report called 'Assessing the degree
of spare capacity'. The purpose of this article is to explain why
we think the MPC's approach is incorrect.
Long-term unemployment
The MPC's current assessment is that the amount of spare
capacity in the economy is "probably in the region of 1-1.5 per
cent". (1) The vast majority of this spare capacity, they argue, is
not inside firms but within the labour market. But their estimate of the
medium-term equilibrium unemployment rate, and hence of the difference
between actual and equilibrium unemployment, is crucially driven by
their assumptions on the role of the long-term unemployed in the labour
market. There are two elements to this argument. The first is that the
longer that someone has been out of work, the lower the probability of
them finding a job. This is true; those who have been unemployed for
more than twelve months are about a third as likely to find a job as
those unemployed for fewer than six months. (2) The second is that this
implies that the long-term unemployed therefore put less downward
pressure on wages. This, however, is not substantiated by the evidence,
which does not support the claim that the long-term unemployed have a
different impact on wages than the short-term unemployed.
This argument dates back to Layard and Nickell (1987), who argued
that the long-term unemployed imposed much less wage pressure than the
short-term unemployed. In a series of annual time-series regressions
they found evidence that a long-term unemployment term, defined as the
number of those who had been unemployed expressed as a proportion of
total unemployment, entered positively in a wage equation. However, this
and other subsequent work (Rudebusch and Williams, 2014; Llaudes, 2005)
suffers from the problem that it is hard, if not impossible, to separate
out the impact of high overall unemployment from high long-term
unemployment due to the high correlation between the two variables using
aggregated time series methods (Blanchflower and Oswald, 1994).
By contrast, Blanchflower and Oswald (1990) showed, using
micro-data for the UK, that this was not the case and long-term
unemployment did not play any independent role in wage determination.
They concluded that "the British evidence does not support the view
that longterm unemployment is an important element in the wage
determination process".
Up-to-date analysis continues to support this view. For the US,
Blanchflower and Posen (2014) examined the impact of long-term
unemployment in a series of hourly and weekly wage equations using data
from the Current Population Survey pooled across state and year cells,
for the period 1990-2013. The authors included year and state fixed
effects, a lagged dependent variable and the log of unemployment and
inactivity rates, which both entered significantly negative. They also
included separate variables for the proportion of the unemployed with
durations of 15+ weeks: 27+ weeks and one year and over. No evidence was
found that the long-term unemployed had a smaller wage-reducing effect
than the short-term unemployed, confirming the earlier work in
Blanchflower and Oswald (1990). If anything, they even found some
evidence to suggest long-term unemployment lowered wage growth even more
than short-term unemployment.
Similar evidence indicating that long-term unemployment and
short-term unemployment have equivalent effects on inflation in the USA
has been found using data on prices rather than wages. In a recent
paper, Kiley (2014) considered this issue using cross-section time
series data on 24 large metropolitan areas. The dependent variable is
the Consumer Price Index (CPI) in each metropolitan area by year. As in
the Blanchflower and Posen (2014) estimation procedure, year and area
fixed effects are included with long-term unemployment being defined as
an unemployment spell of 27 weeks and over. Rather than including a
variable for the long-term unemployment proportion as in our analysis,
Kiley includes both short-term and long-term unemployment rates, which
is functionally similar.
It is notable that Kiley finds that the coefficients in his price
change equations on local unemployment rates are similar and precisely
estimated; hence, the data do not reject the hypothesis that short- and
long-term unemployment rates have identical effects on inflation. Kiley
is thus able to conclude that "the results suggest that long-term
unemployment has exerted similar downward pressure on inflation to that
exerted by short-term unemployment in recent decades".
Of course, there are significant differences between the UK and US
labour markets. In table 1 we report the results of estimating a series
of hourly and weekly wage equations using data from the Labour Force
Surveys for the UK, pooled across twenty regions defined based on
residence and from 1993 to 2013 in the case of hourly pay and from 1992
to 2013 for weekly pay. (3) Along with a lagged dependent variable, we
also include the log of the regional unemployment rate plus a long-term
unemployment variable, defined as the proportion of the unemployed that
have been continuously unemployed for at least a year, which has a mean
of 31.1 per cent. If the long-term unemployed exert less pressure on
wages than the short-term unemployed, this variable should be
significant and positive--but it never is. We calculate both of these
variables from the LFS data. In column 1 we include these variables
along with a set of year dummies and then in column 2 we add region
dummies. The log unemployment rate is now significant and negative for
both hourly and weekly wages. In column 3 we add a series of personal
controls. If wages adjust with a lag, we might expect the lagged
equivalent of the unemployment variables also to enter with a lag.
Column 4 extends the dynamics of the model by introducing a second lag
on the dependent variable and lags on the log unemployment rate and the
long-term unemployed share. The lag in the long-term unemployed variable
is weakly significant (t = 1.79) in the hourly wage equation but is
always insignificant for weekly wages. Though this provides the most
supportive evidence that the long-term unemployed have a different
effect on wage settlements than do the short-term unemployed, the level
of significance is weak, the lagged wage is not significant and the
result is highly sensitive to changes in specification.
Thus, consistent with Blanchflower and Oswald (1990), we also find
that the UK evidence does not support the view that long-term
unemployment is an important element in the wage determination process.
We find no convincing evidence that the long-term unemployed have any
different impact on wages than the short-term unemployed. Hence, we
conclude that it is inappropriate for the MPC to reduce the estimated
level of slack due to the amount of long-term unemployment. The MPC has
produced no evidence for the UK; and based on the new analysis presented
here we draw exactly the opposite conclusion; no downward adjustment
should be made.
Evidence on underemployment and its impact
In a series of recent papers we have examined the extent of
underemployment in the UK economy (Bell and Blanchflower, 2011, 2013a,
2013b) based on data from the Labour Force Surveys from 2001 Q2 to 2014
Q4. (4) Workers are asked if they would "like to work longer hours,
at current basic rate of pay, given the opportunity?" If they
respond in the affirmative they are asked for the number of hours they
would like to work. A similar set of questions is asked for those who
would like shorter hours.
The responses for each series through 2014Q1 are plotted in figure
1, which shows that until 2008 the two series were essentially equal to
each other. After that date, with the onset of recession, there was a
slight drop in the 'fewer hours' series alongside a big jump
in the 'more hours' series. Figure 2 plots the seasonally
adjusted underemployment rate and the unemployment rate. In 2014Q1 the
underemployment rate was 8.4 per cent and the unemployment rate 6.8 per
cent; both have dropped from their peaks in 2011Q4. (5) The MPC in its
Inflation Reports also reports the underemployment rate using our
methods although it expresses it as the number of hours the currently
employed on average would like to work, which of course is equivalent.
In its May 2014 Inflation Report, Table 3D, it reported the level of
underemployment as follows.
[FIGURE 1 OMITTED]
[FIGURE 2 OMITTED]
Averaged across all workers, underemployment amounts to
approximately an additional half an hour per worker. Given there are
32.7 million workers in the UK working an average of 32.1 hours, this
would amount to approximately half a million additional workers. (6)
1998- 2012 2013 2013 2013
2007 HI Q3 Q4
Average hours 32.4 31.9 32.0 32.1 32.1
Desired hours 32.1 32.4 32.7 32.7 32.6
In table 2 we attempt to determine who are the underemployed, using
micro data from the LFS from 2001 through 2014 Q1. In total there are
2.8 million observations. We set the dependent variable to zero if the
worker responds that they don't want to change their hours; if they
want longer hours, then the number of hours they say they want is
included as a positive number. If the worker says they want fewer hours
then that number is included as a negative number. The mean of the
variable is negative from 2001-8 and positive after that. (7) We include
controls for region of residence; year dummies and controls for type of
public sector organisation and schooling (not reported) as well as for
age, gender, race, whether the respondent was an A8 or A2 migrant and
whether he/she was a full-time worker. Separate estimates are provided
for the whole time period as well as for the recession years of 2012
Q1-2014 Q1.
The third and fourth columns are restricted to employees only and
add years of tenure and its square and whether the job was permanent. In
the final column the log of hourly pay is included as a control; which
reduces the sample size as earnings data are only provided in the first
and fifth of the five sample waves.
The main findings are that the young and the least educated and
minorities, who have the highest unemployment rates, are especially
likely to say they would like more hours. (8) Similarly migrants from
the A8 and the A2 Accession Countries are also especially likely to
desire more hours, as are racial minorities. The self-employed also want
more hours as do those whose jobs are temporary along with part-timers.
In the final column, and ignoring issues of endogeneity, it is apparent
that the low-wage workers want more hours. All this is consistent with
the view that underemployment is both real and large.
However, in the May 2014 Inflation Report, the MPC makes a similar
downward adjustment to labour market slack as for long-term
unemployment, estimating that "only around half of the present gap
between actual hours and the estimate of desired hours represents labour
market slack". This judgement appears to be based on calculations
presented in a recent speech by Martin Weale (2014a). Weale argued that
"It is obviously tempting to look at these figures and regard
the gap between actual hours and desired hours as a simple additional
source of labour market slack. On that basis it might seem that hours
worked could rise by around 1 Vi per cent, simply as a result of people
finding as much work to do as they would like to do. There are, however,
grounds for caution, even before those figures are translated into
effective labour supply ... It may be the case that some of the net
underemployment is a response to the state of the economy rather than
any indication of genuine extra capacity. For example people whose
partners lose their jobs may well say that they would like to work
longer. But once their partners find new jobs, they may lose interest in
doing so."
Weale's quantitative analysis uses LFS data based on a
longitudinal sample of individuals observed in the first wave in 2012
and for the fifth time in. His findings are reported in table 3, along
with the sample sizes in parentheses. Weale found that those who said
they were underemployed said they wanted an average of 11.7 extra hours.
Those who were underemployed in the first wave but fully-employed in the
fifth wave increased their hours by 6.5 hours. Those who were
over-employed in the first wave and were fully-employed in the second
desired a reduction of 11.3 hours, but actually achieved a reduction of
4 hours. Table 4 reports the hourly wages Weale obtained from his
sample; those who were underemployed at both waves had wages of 8.74
[pounds sterling] compared with 9.49 [pounds sterling] if they were
fully employed in 2013. In the case of the over-employed, the wage rates
were 17.24 [pounds sterling] and 16.16 [pounds sterling] respectively.
For those fully employed at both sweeps, the average wage was 13.94
[pounds sterling]. He reported the hourly pay for people underemployed
in 2012 and fully employed in 2013 was 9.40 [pounds sterling] in 2012
and 9.58 [pounds sterling] in 2013 (2013 Q4 [pounds sterling]s).
Weale's underemployment index is then calculated by adjusting
the raw figures for the extent of under and over employment to reflect
both actual (as opposed to desired) increases or decreases in hours
realised by those who changed their hours, and the differences in wages
between the two groups (relative to the fully employed). The result is
that, in contrast to our estimate of labour market slack from
underemployment equivalent to about half a million workers, Weale's
estimate is equivalent to only about a third of that.
However, in our view, there are numerous problems with this
analysis:
* As Weale recognises, there are issues of selection bias. Only 60
per cent of those in the survey at the start are still there five
quarters later, especially as young people who want the most extra hours
are the most likely to drop out, along with the least educated. This
problem is illustrated in table 5, which reports the overall
distribution of labour market status in 2012 and 2013 for five
groupings--the inactive; the unemployed; the 'fully-employed'
who say they don't want to change their hours; the
'underemployed' who say they want to increase them and the
'underemployed' who want to lower them. It is clear that the
underemployed and the overemployed are markedly under-represented in
Weale's samples.
* The analysis is based on surprisingly small sample sizes. Desired
hours data are available, for example, on only 722 workers who are
underemployed in both 2012 and 2013 and 628 overemployed workers in both
years in table 4. The sample sizes fall to 395 and 390 respectively in
table 4 when wages are examined.
* Weale's analysis focuses primarily on individuals who were
underemployed in 2012, but as can be seen from table 5, of the 2424
workers who were underemployed in 2013, 177 were inactive in 2012; 190
were unemployed while 1305 were fully-employed. In the case of the 2105
underemployed, 14 were inactive in 2012; 29 were unemployed while 1401
were fully employed.
* The analysis incorporates an adjustment for relative
productivity, as measured by wages. These adjustments result in a
substantial reduction in the index of underemployment. However, because
the sample sizes are relatively small, there is huge uncertainty
associated with these adjustments--which are of course magnified when
applied to various labour market aggregates. Moreover, the hourly wage
data used to calculate 'productivity adjustment' only relates
to employees. This excludes all self-employed workers who, as we know
from table 4, want more hours than employees.
* It is well known that longitudinal data analysis creates downward
biases due to measurement error biases. Misclassification of a small
number of workers will produce a much larger error in longitudinal data
than in cross-section analysis and cannot be readily ignored. Freeman
(1984) points out that the reason for the greater error is twofold. On
the one hand, random misclassification of workers in two periods will
produce a larger number of misclassified workers than random
misclassification in one period. On the other hand, by obtaining
information on underemployment on small numbers of changers, the
longitudinal analysis will contain a smaller number of correct
observations. As a result the proportion of observations in error will
be much larger in the longitudinal analysis than in the cross-section
analysis producing a larger downward bias.
* The analysis only considers one form of transition from
underemployment or overemployment to full employment. He does not pick
up those who were fully employed in the first instance and subsequently
express a desire to increase or decrease their hours. As is clear from
table 6, a larger number moved from being fully employed to
underemployed (n = 1305) than either stayed underemployed (n = 722) or
became fully employed (n = 932).
Finally, the biggest problem for the argument being put forward by
Weale and the MPC is that our index indicates that there was no
underemployment when the economy was running close to full-employment.
As figures 1 and 2 show, there was essentially no underemployment in the
UK from 2000-2007 when the average unemployment rate was a mere 5.2 per
cent. Then, when the recession hit, the difference between the number of
extra hours that were desired increased, while the number from people
who wanted less remained broadly flat. It seems hard to believe that the
two series won't close back to pre-recession equality, if and when
the economy returns to full employment.
Conclusions
To illustrate the adjustments the MPC is making, if we take the
most recent data available for 2013 Q4, we have an unemployment rate of
7.2 per cent and an underemployment rate of an additional 1.8 per cent,
making an underemployment rate of 9 per cent. The MPC's adjustment
for long-term unemployment reduces the unemployment rate by 18 per cent,
or about 1.3 percentage points; its adjustment for underemployment
reduces that by half, or 0.9 percentage points. So, for the purposes of
calculating labour market slack, the underemployment rate according to
the MPC is really 6.8 per cent and the unemployment rate 5.9 per cent.
This in turn underpins their forecasts for rising real wage growth.
Our paper contests the view that the long-term unemployed, because
of their supposed greater distance from work, should be treated as a
different category when assessing the level of slack in the UK labour
market. Microeconometric evidence from both the USA and our own evidence
from the UK cannot distinguish any statistically significant difference
between long-term unemployment and overall unemployment in their effects
on wages. There is no empirical justification for focusing only on the
short-term unemployed when calibrating slack in the UK labour market.
We also argue that there is insufficient evidence to infer that our
recent estimates of underemployment tend to exaggerate the extent of
labour market slack. Weale argues that survey responses in the UK Labour
Force Survey cannot be taken at face value. When asked whether they want
to increase or decrease their weekly hours of work, he contends that the
employed exaggerate the change in working time that they desire
--upwards or downwards. Using data only for 2012, he found that those
who wanted to increase or decrease their hours at the beginning of the
year and then claimed that they were fully employed at the end of the
year did not achieve the increase or reduction in hours that they wanted
at the outset.
There are several empirical issues with Weale's analysis.
These include sample selection biases, small sample sizes, which
inevitably lead to relatively large standard errors and undermine the
precision of adjustments to aggregate changes in desired hours. In
particular, the 'productivity' adjustments, which are crucial
to his argument, are subject to significant uncertainty. These
within-sample issues are further amplified by the fall in response rates
between Wave 1 and Wave 5 and the absence of the self-employed from the
analysis.
Last time interest rates were raised was in July 2007. At that time
the unemployment rate was 5.5 per cent while our underemployment index
stood at 5.8 per cent --a gap of 0.3 per cent. For the period
February-April 2014, the unemployment rate was 6.6 per cent, while the
underemployment index in the first quarter of 2014 was 8.4 per cent--a
gap of 1.8 per cent. In July 2007, when interest rates were last raised,
the CPI was 1.9 per cent and the RPI at 3.4 per cent. In May 2014, the
CPI was increasing at 1.5 per cent and the RPI at 2.4 per cent. In our
view there is little or no reason to believe that the underemployment
rate will not return to balance, as the economy approaches full
employment. When the labour market moves closer to full employment,
individuals inevitably will become less constrained in their choices of
how many hours to work, just as they were in the pre-recession years.
With little or no foundation, the MPC is making two arbitrary
downward adjustments to labour market slack in the UK. This paper has
argued that these judgements are inappropriate; the UK labour market is
much further from full employment than the MPC calculates and in
consequence there is much less wage pressure than it is forecasting. The
crucial test is how quickly nominal wages start to rise, but there is
absolutely no sign of this happening at the time of writing. Nominal
wage growth in June 2014 was -1.7 per cent, and 0.7 per cent if 3-month
on 3-month averages are used. The CPI grew by 1.5 per cent and the RPI
by 2.5 per cent over the same twelve-month period.
In a subsequent speech Weale (2014b) sensibly concludes as follows:
"there is the continuing unusual weakness in wages and a question
of what signal should be drawn from that. It may well imply that there
is rather more spare capacity in the economy than the Committee has
assumed. Should wage growth fail to revive, that will, on its own, tip
the scales further in favour of maintaining a strong monetary
stimulus." We agree.
REFERENCES
Bell, D.N.F. and Blanchflower, D.G. (2011), 'Youth
underemployment in the UK in the Great Recession', National
Institute Economic Review, January, pp. R1-R11.
--(2013a), 'Underemployment in the UK revisited,'
National Institute Economic Review, May, pp. F8-F22.
--(2013b), 'How to measure underemployment?', Peterson
Institute for International Economics Working Paper 13/2, August.
Blanchflower, D.G. and Oswald, A.J. (1990), 'The wage
curve', Scandinavian Journal of Economics, 92, pp. 215-35,
reprinted in Holmlund, B. and Lofgren, K.G. (eds), Unemployment and Wage
Determination in Europe, Basil Blackwell.
--(1994), The Wage Curve, Cambridge, MA, MIT Press.
Blanchflower, D.G. and Posen, A.S. (2014), 'Wages and labor
market slack: making the dual mandate operational', Peterson
Institute Working Paper, 15 April.
Freeman, R.B. (1984), 'A longitudinal analysis of the effects
of trade unions', Journal of Labor Economics, 2(1), pp. 1-26.
Kiley, M. (2014), 'An evaluation of the inflationary pressure
associated with short- and long-term unemployment', Federal Reserve
Board, Finance and Economics Discussion Series 2014--28, Washington, DC
20551,21 March.
Layard, R. and Nickell, S. (1987), 'The labour market',
in Dornbusch, R. and Layard, R. (eds), The Performance of the British
Economy, Oxford, Oxford University Press.
Llaudes, R. (2005),The Phillips Curve and long-term
unemployment', European Central Bank Working Paper 441, February.
Rudebusch, G.D. and Williams, J.C. (2014), 'A wedge in the
dual mandate: monetary policy and long-term unemployment' Federal
Reserve Bank of San Francisco Working Paper 2014-14.
Weale, M. (2014a), 'Slack and the labour market' speech
given at the Thames Valley Chamber of Commerce, 20 March, http://www.
bankofengland.co.uk/publications/Documents/speeches/2014/ speech716.pdf.
--(2014b), 'Spare capacity and inflation', speech given
at the Confederation of British Industry, Belfast, 18 June, http://www.
bankofengland.co.uk/publications/Documents/speeches/2014/ speech737.pdf.
David N.F. Bell * and David G. Blanchflower **
* University of Stirling, IZA and CPC. E-mail:
d.n.f.bell@stirling.ac.uk. **Dartmouth College, University of Stirling,
Peterson Institute for International Economics, IZA CESifo and NBER.
E-mail: David.G.Blanchflower@dartmouth.edu.The authors would like to
thank Martin Weale for providing them with details of his analysis and
calculations.
NOTES
(1) External MPC members Ben Broadbent and Martin Weale have argued
that the level of slack is approximately I per cent and 0.9 per cent
respectively.
(2) May 2014 Inflation Report, p.44
(3) Gross weekly earnings is available In the LFS from Winter 1992,
whereas hourly pay is available from Spring 1993. In the case of the
1992 data we can only use the Winter data. Region of usual residence is
defined across these regions--Tyne & Wear; Rest of Northern Region;
South Yorkshire; West Yorkshire; Rest of Yorkshire & Humberside;
East Midlands; East Anglia; Inner London; Outer London; Rest of South
East; South West; West Midlands (Metropolitan); Rest of West Midlands;
Greater Manchester; Merseyside; Rest of Northwest; Wales; Strathclyde;
Rest of Scotland; Northern Ireland
(4) See http://bellblanchflowerunderemployment.com/ and also
published quarterly by the Work Foundation at http://www.
theworkfoundation.com/Datalab/The-BellBlanchflowerUnderemployment-lndex.
(5) For details on how the underemployment rate is calculated see
Bell and Blanchflower (2011, 2013a, 2013b).
(6) In the US there has been little movement in underutilisation
rates. The broad measure of underutilisation U-6 has moved very closely
with the unemployment rate. What has moved is the inactivity rate which
has fallen, which it has not done in the UK. For example in the US in
2008 QI the inactivity rate for 16-64 year olds was 25 per cent compared
with 27 per cent in 2013 Q4, whereas in the UK the inactivity rate fell
between these two dates from 24 per cent to 23 per cent. Blanchflower
and Posen (2014) show that the inactivity rate along with the
unemployment rate pushes down on wages.
(7) The mean of the variable varies by year: 2001 = -0.27; 2002 =
-0.29; 2003 = -0.31 ; 2004 = -0.33; 2005 = -0.25; 2006 = -0.17; 2007 =
-0.16; 2008 = -0.04; 2009 = 0.24; 2010 = 0.27; 2011 = 0.37:2012 = 0.40;
2013 = 0.39.
(8) 18-24 year old unemployment rates are 16.5 per cent while 16-17
year old rates are 35.4 per cent compared with 6.6 per cent overall in
March 2014.
Table 1 .Wage equations and long-term unemployment, 1992-2013
(1) (2)
a) Hourly (1993-2013)
Lagged [Wage.sub.t-1] 0.9625 (59.93) 0.1343 (2.72)
Lagged [Wage.sub.t-2]
Log unemployment [rate.sub.t] -0.0025 (0.23) -0.0556 (3.42)
Log unemployment [rate.sub.t-1]
Long-term [unemployment.sub.t] 0.0171 (0.39) 0.0756 (1.38)
Long-term [unemployment.sub.t-1]
Year dummies Yes Yes
Region dummies (20) No Yes
Personal controls No No
N 399 399
Adjusted [R.sup.2] 0.9772 0.9871
b) Weekly (1992-2013)
Lagged [Wage.sub.t-1] 0.9144 (51.90) 0.0573 (1.75)
Lagged [Wage.sub.t-2]
Log unemployment [rate.sub.t] 0.0110 (0.84) -0.0550 (3.59)
Log unemployment [rate.sub.t-1]
Long-term [unemployment.sub.t] -0.0240 (0.44) 0.0071 (0.14)
Long-term [unemployment.sub.t-1]
Year dummies Yes Yes
Region dummies (20) No Yes
Personal controls No No
N 418 418
Adjusted [R.sup.2] 0.9772 0.9888
(3) (4)
a) Hourly (1993-2013)
Lagged [Wage.sub.t-1] 0.0999 (1.99) 0.1109 (2.18)
Lagged [Wage.sub.t-2] 0.0503 (1.00)
Log unemployment [rate.sub.t] -0.0352 (2.03) -0.0045 (0.23)
Log unemployment [rate.sub.t-1] -0.0373 (1.87)
Long-term [unemployment.sub.t] 0.0357 (0.64) -0.0015 (0.03)
Long-term [unemployment.sub.t-1] 0.0989 (1.79)
Year dummies Yes Yes
Region dummies (20) Yes Yes
Personal controls Yes Yes
N 399 379
Adjusted [R.sup.2] 0.9874 0.9884
b) Weekly (1992-2013)
Lagged [Wage.sub.t-1] 0.0300 (0.92) 0.0785 (1.52)
Lagged [Wage.sub.t-2] 0.0142 (0.44)
Log unemployment [rate.sub.t] -0.0482 (3.00) -0.0157 (0.84)
Log unemployment [rate.sub.t-1] -0.0438 (2.27)
Long-term [unemployment.sub.t] -0.0176 (0.15) -0.0127 (0.24)
Long-term [unemployment.sub.t-1] 0.0604 (1.16)
Year dummies Yes Yes
Region dummies (20) Yes Yes
Personal controls Yes Yes
N 418 398
Adjusted [R.sup.2] 0.9892 0.9895
Source: Labour Force Surveys.
Notes: personal controls include 5 schooling variables, age, gender
and 4 race dummies. T-statistics in parentheses.
Table 2. Desired hours 2001-14
2012-2014
Age 25-29 -0.6524 (41.33) -0.8886 (18.56)
Age 30-34 -1.3005 (86.41) -1.6092 (34.97)
Age 35-39 -1.5279 (105.10) -1.9401 (42.38)
Age 40-44 -1.5830 (110.20) -1.9573 (44.22)
Age 45-49 -1.7047 (116.52) -2.0798 (47.34)
Age 50-54 -2.1587 (143.60) -2.5277 (56.40)
Age 55-59 -2.6944 (170.54) -3.2351 (68.52)
Age 60-64 -3.5527 (189.76) -4.4464 (83.73)
Age 65-69 -4.4674 (152.81) -5.6215 (77.31)
Age 70-74 -4.4329 (91.78) -6.0835 (50.33)
Age 75+ -4.2355 (55.62) -6.0277 (27.29)
Male 1.1722 (149.01) 1.3612 (60.49)
Self-employed 0.0432 (3.98) 0.4122 (13.59)
Degree -0.9532 (65.75) -1.0255 (21.44)
Higher education -0.6636 (40.14) -0.7781 (14.44)
A-level -0.5501 (38.74) -0.6384 (13.18)
O-level -0.3581 (24.79) -0.3245 (6.62)
Other qualifications 0.0810 (5.07) 0.1372 (2.47)
A8 Accession 1.1696 (24.52) 0.9303 (7.57)
A2 Accession 2.0563 (16.40) 1.7823 (8.83)
Mixed 0.4625 (9.97) 0.2694 (2.31)
Asian 0.8983 (45.73) 1.0987 (22.27)
Black 1.3511 (48.08) 1.6573 (22.29)
Chinese 0.4176 (6.13) -0.0987 (0.21)
Other race 1.1728 (32.40) 1.1417 (11.91)
Fulltime -3.7337 (416.87) -4.6141 (181.65)
Tenure years
Tenure squared
Permanent job
Log hourly pay
Constant 3.8277 5.5699
N 2,805,715 415,120
Adjusted [R.sup.2] 0.0878 0.1074
2012-2014 Employees only
Age 25-29 -0.5037 (32.28) -0.4391 (15.05)
Age 30-34 -1.0375 (68.24) -0.8211 (28.83)
Age 35-39 -1.1409 (76.11) -0.9276 (32.91)
Age 40-44 -1.1128 (74.14) -0.8682 (30.78)
Age 45-49 -1.1920 (77.27) -1.0124 (35.02)
Age 50-54 -1.6074 (100.41) -1.4325 (48.04)
Age 55-59 -2.1065 (124.22) -1.9123 (61.00)
Age 60-64 -2.9184 (143.44) -2.7428 (73.26)
Age 65-69 -3.7716 (112.21) -3.5155 (56.73)
Age 70-74 -3.7236 (63.13) -3.5328 (32.10)
Age 75+ -3.6375 (35.53) -3.424 (18.21)
Male 1.1198 (141.34) 1.1562 (80.12)
Self-employed n/a n/a
Degree -1.1011 (72.98) -0.7466 (25.44)
Higher education -0.7437 (43.78) -0.4772 (15.22)
A-level -0.5844 (39.51) -0.4304 (15.78)
O-level -0.3837 (25.80) -0.2882 (10.61)
Other qualifications 0.0191 (1.16) 0.0643 (2.14)
A8 Accession 0.9678 (20.38) 0.8201 (9.38)
A2 Accession 1.7476 (10.58) 1.1678 (3.80)
Mixed 0.3979 (8.56) 0.4468 (5.26)
Asian 1.0182 (50.18) 1.0355 (27.00)
Black 1.3421 (47.91) 1.2125 (22.85)
Chinese 0.4135 (5.81) 0.5435 (4.06)
Other race 1.1136 (29.99) 1.2133 (17.22)
Fulltime -3.4483 (374.53) -3.2765 (196.20)
Tenure years -0.0853 (67.21) -0.0740 (32.33)
Tenure squared 0.0017 (44.98) 0.0016 (23.20)
Permanent job -0.8543 (53.77) -0.7565 (25.53)
Log hourly pay -0.5663 (38.98)
Constant 4.6417 5.2240
N 2,424,768 707,893
Adjusted [R.sup.2] 0.0978 0.0994
Source: LFS 2001 -2014.
Notes: dependent variable desired change in hours. All equations
include a full set of 23 region and 14 year dummies. Excluded
categories Wave 1 ; age 75 and over; white and no qualifications.
Region is region of residence. A8=Poland, Czech Republic; Hungary:
Estonia; Latvia; Lithuania, Slovenia and Slovak Republic and are set
to 1 only if year>=2004 A2=Bulgaria and Romania are set to 1 only if
year>=2007. Controls are also included but not reported for DK and
not answered for region, race and schooling. T-statistics in
parentheses.
Table 3. Desired and actual changes in hours worked
between 2012 and 2013
Labour market status in 2012
Under- Fully- Over-
employed employed employed
Labour market status in 2013
Under-employed
Desired 13.5 (722) 0.0 (1127) -8.3 (30)
Actual 1.2 -2.9 10.0
Fully employed
Desired 11.7 (769) 0.0 (12,286) -11.3 (656)
Actual 6.5 -0.5 -4.0
Over-employed
Desired 9.7 (33) 0.0 (1,224) -11.3 (628)
Actual 7.1 1.4 -1.5
Source: Weale (2014a) and private communication.
Notes: sample sizes in parentheses.
Table 4. Hourly rates of pay by employment category
(2013Q4 prices)
Labour market status in 2012
Under-
employed
Labour market status in 2013
Under-employed 8.74 [pounds sterling] (395)
Fully-employed 9.49 [pounds sterling] (402)
Over-employed 10.96 [pounds sterling] (17)
Labour market status in 2012
Fully-
employed
Labour market status in 2013
Under-employed 10.09 [pounds sterling] (571)
Fully-employed 13.94 [pounds sterling] (6331)
Over-employed 15.42 [pounds sterling] (674)
Labour market status in 2012
Over-
employed
Labour market status in 2013
Under-employed 12.65 [pounds sterling] (14)
Fully-employed 16.16 [pounds sterling] (400)
Over-employed 17.24 [pounds sterling] (390)
Source:Weale (2014a) and private communication.
Notes: sample sizes in parentheses.
Table 5. Number of unweighted observations by labour
force status (%)
All 5 waves Wave 1 Wave 5 Wave 1 Wave 5
Weale Weale
2012 2013 2012 2013 2012 2013
Inactive 35.4 35.3 40.3 35.3 35.7 35.7
Unemp. 4.9 4.6 4.4 4.6 4.2 3.9
Under-emp. 7.0 7.1 5.2 7.8 4.6 6.2
Fully-emp. 47.4 47.6 46.0 47.2 51.3 48.7
Over-emp. 5.4 5.5 4.2 6.4 4.1 5.4
N 321,429 307,476 69,915 54,836 38,842 38,842
Source: Labour Force Surveys, 2012 and 2013 and private
communication with Weale.
Table 6.Transition rates between labour market states,
2012 Wave 1 -2013 Wave 5
2012 status
Inactive Unem- Under- Fully Over- Total
ployed employed employed employed
2013 status
Inactive 12503 302 66 909 77 13857
Unemp. 471 640 53 324 35 1523
Under-emp. 177 190 722 1305 30 2424
Fully-emp. 726 469 932 15979 827 18933
Over-emp. 14 29 33 1401 628 2105
N 13891 1630 1806 19918 1597 38842
Source:Weale (2014a).
