Business Cycles in The United Kingdom

Business Cycles in the United Kingdom: Facts and Fictions
Author(s): Keith Blackburn and Morten O. Ravn

Source: Economica , Nov., 1992, New Series, Vol. 59, No. 236 (Nov., 1992), pp. 383-401
Published by: Wiley on behalf of The London School of Economics and Political Science
and The Suntory and Toyota International Centres for Economics and Related
Disciplines
Stable URL: https://www.jstor.org/stable/2554886
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
, Wiley and The London School of Economics and Political Science are collaborating with JSTOR
to digitize, preserve and extend access to Economica
This content downloaded from

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
All use subject to https://about.jstor.org/terms
Economica, 59, 383-401
Business Cycles in the United Kingdom:

Facts and Fictions
By KEITH BLACKB3URN and MORTEN 0. RAVN

University of Southampton and University of Aarhus
Final version received 4 March 1992.
This paper documents the statistical properties of contemporary business fluctuations in the
United Kingdom. We study the period 1956-90 using quarterly, detrended data on key
aggregate variables. We compute selected moments of the data, compare our results with
those for the United States, and rigorously test for dynamic instabilities. Our findings confirm
the existence of substantive cyclical regularities, both across countries and across time. Some
notable cross-country differences are also identified. Conclusions about stability are shown
to be potentially sensitive to the method of testing. In general, cross-correlations are appreci-
ably more stable than standard deviations.
INTRODUCTION
Recurrent fluctuations in economic activity-business cycles-are common to

all industrialized societies. Recently, the study of these fluctuations has been
subject to a new methodology which makes two basic demands of business
cycle research: the complete and systematic characterization of cyclical
phenomena in the form of stylizedfacts, and the construction of fully a
model economies which can be evaluated both qualitatively and quantitatively
in terms of their ability to replicate these facts.' This paper represents a
continuation of the research that takes up the first of these challenges. Our
aim is to idenitify the major stylized facts relating to contemporary business
fluctuations in the United Kingdom. In doing this we also expose some major
stylized fictions in traditional views of the business cycle. We believe that our
findings will interest, and possibly surprise, many readers.
By 'stylized facts' are meant broad regularities in the statistical properties
of economic time series. Facts about business cycles were first documented at
length in the early NBER chronologies pioneered by Burns and Mitchell (1946)
and Mitchell (1913, 1923, 1927). Since then, the techniques used to identify
and characterize business fluctuations have changed with developments in
business cycle theory and time-series analysis. Thus, following Lucas (1977),
it has become customary to define business cycle phenomena as repeated
deviations of output from some trend and the relative behaviour of other
aggregate time series. Accordingly, instead of talking about 'reference-cycle
relatives', 'specific cycles' and 'turning points', one now refers to 'secular
trends', 'cyclical deviations' and 'sample moments'.
New evidence of business cycle regularities can be found in a number of
recent studies using single-country, cross-country and historical data (Backus
and Kehoe 1992; Blackburn and Ravn 1991a; Brandner and Neusser 1992;
Correia et al. 1992; Danthine and Girardin 1989; Kydland and Prescott 1990).
As far as we know, this paper is the first to present a detailed modern record
for the United Kingdom. Our inquiry covers the period 1956-90 and uses
quarterly data on key economic aggregates. These aggregates include output
and its components, inputs to production, nominal variables and variables

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
384 ECONOMICA [NOVEMBER
relating to an open economy. The data are transformed into stationary time
series by means of low-frequency filtering. We compute selected moments of
these series and draw attention to the results of most interest.
The motivation for the paper is threefold. First, we believe that the reporting
of stylized facts is a legitimate scientific exercise of interest in its own right.
The facts we report here are both the similarities and discrepancies in national
business cycle characteristics. Given the contrasts in size, institutions and
policies between countries, it would come as no surprise to find cross-country
differences in cyclical fluctuations. We compare our results for the United
Kingdom with those obtained for the United States. We chose the United
States first because it is the country for which most evidence currently exists,
and second because the idiosyncracies in national characteristics are likely
to be particularly important in this case. We discover many cross-country
similarities in the volatilities of expenditure components and production inputs,
and in the co-movements of both real and nominal variables with output. Of
the cross-country differences, the most notable are the relatively high volatilities
of consumption and net exports in the United Kingdom, and the relatively
low (high) volatility of hours per worker (employment) in the United States.
Some of our stylized facts expose major stylized fictions in traditional views
of the business cycle. Thus, it is a fiction that prices are pro-cyclical, that real
wages are counter-cyclical and that productivity is counter-cyclical.
Second, we adopt the view that, for the properties of the data to constitute
empirical regularities, they should remain broadly invariant over time. We use
the term dynamic regularities to refer to those properties that are robust to
changes in the sample period. Identifying such regularities is a task that
research, to date, has left largely unattended. An innovation of the current
paper is our correction for this by means of a series of tests of the dynamic
stability of selected moments. Depending on the test used, it is possible to
reach different conclusions. In all cases, however, the most stable (unstable)
moments are the cross-correlations (standard deviations).
Third, the choice of theoretical model should ultimately be based on the
properties of the data and not on one's ideological prejudices. The basis for
discriminating between models remains the same as in the past: it is ability
of each model to pass increasingly severe tests. Since one test is the extent to
which a model can replicate the facts, it is necessary to know what these facts
are to begin with.
The remainder of the paper is organized as follows. Section I contains a
description of our procedure for isolating cyclical fluctuations. Section II
documents our findings on business cycle activity over the whole sample period
Section III presents our analysis of dynamic regularities across different
sub-samples. Section IV contains some concluding remarks.
I. METHODOLOGY
We propose to conduct our inquiries by following the common practice of

decomposing time series into secular and cyclical variations. By the former
are meant the long-term trends in series, being allied to low-frequency vari-
ations which manifest as non-stationarities. By the latter are meant the devi-
ations from trends, reflecting higher-frequency fluctuations which can be

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
1992] BUSINESS CYCLES IN THE UNITED KINGDOM 385
modelled as stationary stochastic processes.2 There are many different tech-

niques available for making this decomposition, and it is possible to question
the robustness of the results obtained from using any one of them. This issue
of sensitivity is currently unresolved: while some results do appear to be
sensitive to the method of detrending, many others do not (Blackburn and
Ravn 1991b; Canova 1991; King and Rebelo 1989). Our choice of detrending
technique in this paper is one of the most widely used in modern business
cycle'research. By making this choice, we hope to facilitate comparison between
our own findings and those obtained by others. The technique, first proposed
by Hodrick and Prescott (1980), is summarized as follows.
Let {yt}T=1 be an observable time series, expressed in natural logarithms.3
If {rt}T=i is the trend of this series, then our operational measure of cyclical
fluctuations is given by the sequence {Y--rt}T=i. The trend component is
11.75-
11.50-
11.25-
XC11.00-X
10.75 -
10.50-1
1955 1960 1965 1970 1975 1980 1985 1990
(a) Actual and trend output
6-
5-
3-
2-
-1
-5-
1955 1960 1965 1970 1975 1980 1985 1990
(b) Deviation from trend
FIGURE 1. Low-frequency filtering of output.

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
obtained by solving the following convex minimization problem:
T T-i
()min
T
E (Yt
{1t}t=i t=1 t=2
The first term is the sum of squared deviations and is a measure of the 'goodness
of fit'. The second term is the sum of the squares of the trend component's
second differences and is a measure of the 'degree of smoothness'. The quantity
A is a smoothing parameter which penalizes the acceleration in the trend
component.
Given real economic data and a value for A, the problem in (1) can be
solved straightforwardly using standard numerical routines.4 In the present
application, the data are filtered using a value of A = 1600. This is the value
used in most other studies and produces a trend of the sort that one would
draw free-hand through a plot of the raw series. Figure 1 depicts the trend
and cyclical components of our output series.
In choosing which facts to report, we are guided by our original definition
of business cycle phenomena as recurrent fluctuations of output about trend
and the cyclical co-movements among other aggregate time series. This leads
us to focus on the following properties of the filtered data: the amplitude of
fluctuations in a series, as measured by the percentage standard deviation; the
degree of persistence in a series, as measured by the first-order autocorrelation
coefficient; and the degree of both contemporaneous and non-contemporaneous
co-movements of a series with output, as measured by the cross-correlation
coefficients up to a fourth-order lag and lead. The first of these will tell us
something about the relative volatilities of variables; the second will indicate
the amount of inertia in cyclical deviations; and the third will provide infor-
mation both on whether a series behaves pro-cyclically or counter-cyclically and
on whether the series displays a phase shift relative to the overall business cycle.
II. THE BASIC FACTS
Unless otherwise stated, the moments we report are for the period 1956(I)-
1990(I). Deviations from this are due to lack of quarterly data on some variables
for this period. We present our findings in a series of tables. The numbers
corresponding to standard deviations and autocorrelations should be self-
explanatory. Those relating to co-movements between series have the following
interpretations: entries in column t are the contemporaneous cross-correlation
coefficients, and entries in columns t - i and t + i (i = 1, 2, 3, 4) are the non-
contemporaneous cross-correlation coefficients; a positive (negative) number
indicates that a series is pro-cyclical (counter-cyclical), and a number close
to zero indicates that a series is largely uncorrelated with the cycle; a relatively
large number (in absolute terms) appearing in column t - i (t + i) indicates
that a series tends to lead (lag) cycle by i quarters.
To put some structure on our analysis, we adopt the approach taken in
other studies by organizing our discussion around the following classification
of the data: data on production inputs, data on expenditure components and
data on nominal variables. We also report the properties of a few other series
relevant to the UK as being a small open economy.

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
Facts about production inputs
Table 1 summarizes our findings on the cyclical properties of inputs to produc-

tion. These inputs are labour, capital and inventories. Our measure of labour
is total hours of work, which we have decomposed into total employment and
hours per worker.5 Our raw series for capital was self-constructed owing to
the lack of a consistent series elsewhere.6 We include the inventory stock as
an input to production because of the potential productive services it yields.
Some additional information contained in the table relates to the cyclical
properties of other series deemed relevant to the concerns of this section.
These are the series for the real wage, the real interest rate and average labour
productivity.7
Total hours are seen to be about as variable as output, pro-cyclical and
lagging the cycle. Employment and hours per worker are both less variable
than output but still positively correlated with it. While the former tends to
lag the cycle, the latter displays no phase shift in either direction. The capital
stock is an extremely smooth series with cross-correlations that are rather
difficult to interpret. Inventories, on the other hand, are more volatile than
output, definitely pro-cyclical and definitely lagging the cycle. As one would
expect, capital and inventories exhibit by far the greatest persistence.
All of these findings are consistent with those obtained for the USA. Two
minor discrepancies are the greater degree of smoothness in UK capital and
the more pronounced phase shift in UK employment. There is one difference,
however, which stands out appreciably. Observe that employment and hours
TABLE 1
CYCLICAL PROPERTIES OF PRODUCTION INPUTS*
Cross-corr. between output at t and X at

Var. % St. lst-order
X dev. autocorr. t-4 t-3 t-2 t-1 t t+1 t+2 t+3 t+4
Y 1-49 0-58 -
LH 1-60 0 70 0.00 0-14 0-27 0 43 0-58 0-58 0 50 0 39 0-23
(1-07)
E 1-05 0 77 -0-21 -0-05 0 11 0 24 0 39 0 53 0 59 0 60 0-61
(0 70)
HW 1-06 0 57 0 20 0-26 0 30 0-42 0-48 0 35 0-17 0-02 -0-22
(0-71)
KS 0-14 0-95 -0-39 -0-40 -0-37 -0-28 -0-16 0 03 0-21 0 34 0 45
(0 93)
IS 2-19 0 89 -0-31 -0-28 -0-18 -0-02 0 21 0 39 0 50 0 54 0 54
(1-45)
RW 1-66 0 53 0-06 0 11 0-22 0-23 0-24 0-23 0-14 0-02 -0 11
(1-11)
RI 1 00 0-28 0-23 0-19 0-16 0-08 0-02 -0-20 -0 19 -0 33 -0-27
(0 67)
LP 1 43 0 34 0-08 0 07 0 11 0-12 0-40 -0-04 -0-15 -0-21 -0 18
(0 95)
* Sample period: 1956(I)-1990(I) (except for IS: 1957(I)-1990(I)).

Note: Numbers in parentheses denote standard deviations relative to output. All sample moments
computed from filtered logarithms of series. Y = output; LH = labour hours; E = employment;
HW = hours per worker; CS = capital stock; IS = inventory stock; R W = real wage; RI = real
interest rate; LP = labour productivity.
Sources and definitions: see Appendix.

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
per worker exhibit roughly the same percentage cyclical variation. This is in
stark contrast to the USA, where as much as 75-80 per cent of the cyclical
variation in total hours is accounted for by variations in employment. A note
of caution needs to be injected, however, because the hours data we are using
for the UK (which pertains to operatives in manufacturing) is not strictly
comparable with the hours data that has been used for the USA (which covers
all non-agricultural workers). To the extent that manual workers exhibit greater
volatility than white-collar workers, our comparison might be biased. To check
this, we obtained hours and employment data on US manufacturing for the
period 1970(I)-1990(I) and computed the relevant cyclical moments. Compar-
ing these with those for the UK over the same sample period, we obtained
confirmation of our initial finding. Thus, the percentage cyclical variation in
UK (US) employment turned out to be 1 12 (2.92), and the percentage cyclical
variation in UK (US) hours per worker turned out to be 1 26 (0.99). It would
appear, therefore, that the difference between variations in labour at the
intensive margin and variations in labour at the extensive margin is markedly
less pronounced for the UK. Explanations for this can be found in some recent
theoretical work on the implications of different labour market institutions
(Burdett and Wright 1989; Wright 1991).
Other interesting results of this section relate to real wages and productivity,
which display very similar properties. Both exhibit cyclical variations close to
the variation of output, both are pro-cyclical, and both tend to move contem-
poraneously with the cycle. Again, these findings are comparable with those
for the USA. The pro-cyclicality of the real wage is now an established stylized
fact. That it should still meet with surprise in some quarters reflects one of
the most long-standing but misguided, beliefs. Another fact we confirmed is
the near-orthogonality between real wages and hours (often referred to as the
Dunlop-Tarshis observation). The cross-correlation between these series was
found to be as low as 0.08.8 The pro-cyclicality of average productivity will
be familiar as a version of Okun's Law.9 It is noted that productivity is less
variable than hours (and also less highly correlated than hours with output)-an
observation that holds to a much greater extent for the USA and which Prescott
(1983) has referred to as the 'key business cycle puzzle'.
In summary, our findings in this section support the view that there are
substantive cross-country regularities in the cyclical behaviour of production
inputs and related series. While quantitative differences do, indeed, exist, one
cannot help but be impressed by the striking degree of uniformity in the data.
Facts about output components
Table 2 presents the outcomes of our inquiries into the cyclical behaviour of
the components of gross domestic product. These components are total private
consumption, total investment, net exports and government expenditure. We
also report results for aggregate savings. Total consumption has been decom-
posed into consumption of non-durables and services and consumption of
durables. It is unclear whether one should include the latter under consumption
or investment, and, for no particular reason, we decided to include it under
consumption.10 Total investment has been decomposed into business fixed
investment and inventory investment. Our series for net exports was constructed
by taking the difference between individual series on exports and imports. We

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
TABLE 2
CYCLICAL PROPERTIES OF EXPENDITURE COMPONENTS*
Cross-corr. between output at t and X at

Var. % St. 1st-order
X dev. autocorr. t-4 t-3 t-2 t-1 t t+1 t+2 t+3 t+4
Y 1-49 0-58
TC 1-59 0-73 0-14 0 30 0 40 0 49 0 68 0 45 0 38 0-25 0 08
(1-07)
NC 1 30 0-81 0-08 0-23 0-36 0 49 0-69 0 54 0 47 0 34 0 21
(0 87)
DC 7 47 0 54 0 29 0 39 0 35 0 32 0 45 0 11 0 03 -0 07 -0-28
(4*99)
TI 6 69 0 68 0 07 0-16 0-32 0-51 0 74 0 58 0-36 0 19 0 08
(4 48)
FI 3 48 0-64 -0 08 0 07 0-22 0-36 0 64 0 55 0-41 0 34 0-22
(2 33)
II 0-82 0 53 0 13 0 14 0 28 0 47 0 59 043 0-22 0 04 -0-04
NX 0 90 0-51 -0-10 -0-16 -020 -0-31 -0 16 -034 -0-25 -0-17 -0 05
X 3-11 0 32 -0 14 -0 07 0 07 0-20 0 50 0 21 0-26 0 15 0 12
(2 08)
M 3-83 0-63 0 05 0 15 0 30 0 50 0-56 0 51 0-46 0-28 0-12
(2 56)
GP 1 50 0-63 -0 02 -0 02 0 03 0-13 0 15 0 06 0-08 0 03 0 03
(1-00)
S 6 34 0-52 -0 03 0 03 0-17 0-32 0-71 0-38 0-20 0 09 0 05
(4 24)
* Sample period: 1956(I)-1990(I).

computed from filtered logarithms of series except sample moments for II, NX and S, computed
from filtered output ratios. Y = output; TC = total consumption; NC = non-durables consump-
tion; DC = durables consumption; TI = total investment; FI = fixed investment; II= inventory
investment; NX = net exports; X = exports; M = imports; GP = government purchases;
S = savings.
report the moments for all three of these series. Finally, savings is defined as
output less private and government consumption."
Inspection of the autocorrelations reveals that most components are fairly
persistent. Total consumption displays about the same variability as output,
is strongly pro-cyclical, and moves in phase with the cycle. These properties
do not reflect any uniform pattern of behaviour across the individual com-
ponents of consumption. As one would expect, consumption of non-durables
is considerably more volatile than consumption of durables. Relative to output,
the former exhibits over five times more cyclical variation whereas the latter
exhibits slightly less cyclical variation. In addition, while durables tend to lead
the cycle slightly, non-durables move contemporaneously with it.
The percentage cyclical variation in total investment is as much as four-and-
a-half times greater than the percentage cyclical variation in output. All three
of the investment series are strongly pro-cyclical and move more or less in
phase with the cycle. These properties are shared by savings. The cyclical
series for total investment is shown in Figure 2(a).
The above results accord well with US evidence and are generally what
one would expect. Standard theory would predict relatively large variations
in the accumulation of long-lasting goods as agents exploit intertemporal
substitution possibilities so as to smooth their consumption over time.

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
20 - Output
15- ......... Investment
10-
15
-20-
-25 l l l
1956 1960 1965 1970 1975 1980 1985 1990
(a) Output and investment
8 - Output
.............. Prices
6-
4-
2-
Q 0 a
-2 -
-4 -
-6-
-8 I I l l
1956 1960 1965 1970 1975 1980 1985 1990
(b) Output and prices
FIGURE 2. Comparison of deviations from trend.
Nevertheless, it is worth pointing out that consumption in the UK is markedly

more volatile than consumption in the USA, with total consumption even
displaying signs of excess volatility. We make the same observation in Black-
burn and Ravn (1991a) as do Backus and Kehoe (1992) and Danthine and
Girardin (1989).12
Net exports are counter-cyclical and move contemporaneously with the
cycle. By contrast, and somewhat surprisingly, both exports and imports display
pro-cyclical behaviour, while retaining the absence of a phase shift. Qualita-
tively, these are the features observed in US data. That the trade balance is
counter-cyclical is explained by the fact that imports are more pro-cyclical
than exports. The major cross-country difference is that the variation in UK

86.186.172.181 on Fri, 25 Feb 2Thu, 01 Jan 1976 12:34:56 UTC
net exports is about double the variation in US net exports. An obvious

interpretation of this is that it reflects the greater degree of openness of the
UK economy.13
Finally, we observe that government purchases are about as variable as
output (less so than in the USA), are nearly orthogonal to output (like in the
USA), and show no sign of a phase shift in either direction (also like in the
USA). There is no evidence here that government expenditure has been used
systematically as a form of counter-cyclical stabilization policy. This is not to
say that such a practice has never occurred. Rather, it may merely reflect the
fact that our period of coverage extends across different policy regimes.
We are led to conclude that cross-country regularities in the cyclical
behaviour of output components extend to the UK. We are struck, once again,
by the remarkable similarities between our own findings and the findings of
others. The most notable difference, at least as regards the USA, is the relatively
high volatility of UK net exports.
Facts about nominal variables
The behaviour of nominal variables has traditionally been the subject of much
attention in business cycle research. Accordingly, we turned our efforts to
investigating the properties of selected key nominal series. These are the series
for money supplies, velocities and prices. We have chosen both a narrow (Ml)
and a broad (M3) definition of money. The price level corresponds to the
implicit deflator of gross domestic product."4 Each of the velocities is defin
as the ratio of nominal output to money. Table 3 summarizes our findings.15
Both narrow and broad monies are more volatile than output and are
pro-cyclical. The former displays less variation, but more pro-cyclicality, than
the latter, with each series having a phase shift that is reversed for the other.
Both measures of velocity are more volatile than output and are counter-
cyclical. One of our more surprising findings (not reported in the table) is that
TABLE 3
CYCLICAL PROPERTIES OF NOMINAL VARIABLES*
Cross-corr. between output at t and Xat

X dev. autocorr. t-4 t-3 t-2 t- 1 t t+ 1 t+2 t+3 t+4
MS
Ml 3 11 0 74 0 54 0 56 0 54 0 45 0 33 0 26 0 09 -0-08 -0 13
(2 05)
M3 3 83 0 88 -0 13 -0 07 0 01 0 08 0 14 0 22 0-27 0-27 0 22
(2 53)
V
VI 3 40 0 68 -0 62 -0 67 -0 65 -0 54 -0-26 -0 34 -0 19 -0 02 0 08
(2.25)
V3 4 48 0 84 0 01 -0 07 -0 13 -0-17 -0 09 -0-27 -0-32 -0 29 -0 22
(2.96)
P 2 17 0 88 -0 25 -0 37 -0 47 -0 53 -0 57 -0 48 -0 37 -0 21 -0 08
(1-45)
* Sample period: 1963(I)-1989(II) (except for P: 1956(I)-1990(I))

computed from filtered logarithms of series. MS = money supply; V = velocity; P = prices.

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
there is little or no correlation between money and prices. Moreover, what

little correlation there is generally turns out to be negative.16
Of greater interest still is the observed strong counter-cyclicality of prices,
which we have chosen to illustrate in Figure 2(b). Like the pro-cyclicality of
the real wage, this is an observation that runs counter to a widely held belief.
It has been confirmed in numerous other studies, being one of the most robust
empirical regularities. Thus, Friedman and Schwarz (1982, p. 399), using a
different method of detrending, report the result for the UK over each of their
prewar, interwar and postwar periods, as well over their full sample period,
1873-1980. Backus and Kehoe (1992) and Blackburn and Ravn (199la) record
it for a wide range of industrialized countries over the postwar period; and
Blackburn and Ravn (1991b) observe it for a variety of different methods of
detrending. The traditional presumption that prices are pro-cyclical is over-
whelmingly contradicted by the evidence.
It is still strongly presumed that fluctuations in money constitute a major
source of impulse to cyclical activity. We interpret our findings in this section
as posing a serious challenge for monetary explanations of the business
cycle.
Facts relating to an open economy
Earlier, we interpreted the relatively large volatility of UK net exports as

reflecting the relatively large degree of openness of the UK economy. Our
study would be incomplete without considering this feature of the UK a little
further. Accordingly, Table 4 presents some properties of a few other selected
series that we considered to be of interest. The terms of trade (TT) is defined
as the relative price of imports, being constructed as the ratio of import prices
to export prices. The OECD series were included to gain an impression of the
international linkages between business cycles. The correlation between savings
and investment has traditionally been used in connection with tests of the
degree of international capital mobility.17
Both UK output and consumption are seen to be positively correlated with
their OECD counterparts. This gives some indication that business cycles are
correlated across countries. That the consumption series are more highly
TABLE 4
CYCLICAL PROPERTIES OF OPEN ECONOMY VARIABLES*
Cross-corr. between X and

X dev. autocorr. Y TC TI NX X M
Y* 2250 0 77 0 43
C* 0099 0-86 065
TT 3-17 0 77 0.19 - 0 14 0-36 0-15
(2.13)
S 5-41 0.55 0-60
(3 64)
* Sample period: 1956(I)-1990(I) (except for Y*, C*: 1971(I)-1990(I)).

computed from filtered logarithms of series except sample moments for S, TI and NX, computed
from filtered output ratios. Y* = OECD output; C* = OECD consumption; TT = terms of trade.

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
correlated than the output series is consistent with the hypothesis of cross-
country risk-sharing. On the other hand, it is clear that there is appreciably
less than perfect risk-sharing.
The terms of trade is a highly volatile series. It is both pro-cyclical and
positively correlated with net exports. Somewhat surprisingly, it is also posi-
tively correlated with exports and imports, individually (the correlation with
the former being the larger of the two). For reasons of space, we have reported
only the contemporaneous cross-correlations, but one of the central relation-
ships in international economics is the dynamic relationship between the terms
of trade and the trade balance. Known as the J-curve, this relationship has
been subject to numerous empirical investigations (Artus 1975; Dornbusch
and Krugman 1976; Krugman and Baldwin 1987; Mendoza 1990; Rose and
Yellen 1989; Spitaller 1980; Warner and Kreinin 1983). Following Backus et
al. (1991), we studied it ourselves by computing the cross-correlations between
the terms of trade and net exports at various leads and lags. Our results are
displayed in Figure 3 and give a dramatic illustration of the J-curve.18
0.3 -
0.2 -
0.1 -
-0.1-
-0.2
-0.3- -
-10 -5 0 5 10
Lead
FIGURE 3. The J-curve: correlation between TT(t) and NX(t+k).
Savings and investment display a positive correlation of an order that some

authors have found puzzling: given the existence of a well organized inter-
national capital market, one might pressume that the correlation should be
near-zero. The apparent puzzle was first documented by Feldstein and Horioka
(1980) and has since been the subject of a number of studies (Baxter and
Crucini 1990; Engel and Kletzer 1984; Finn 1990; Hung 1990; Obstfeld 1986).
This research has shown that a significant savings-investment correlation is
entirely consistent with a high degree of capital mobility.
III. DYNAMIC REGULARITIES
We have insisted that, for the statistical properties of cyclical fluctuations to

constitute stylized facts, they must remain broadly invariant to the passage of

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
time. Research, to date, has paid little more than lip-service to this.19 We
believe that, in doing so, such research has left itself exposed to a major source
of criticism. Our intention in this section is to avoid such criticism by systemati-
cally testing for dynamic regularities in the cyclical moments of the data. We
do this in two ways. First, we test for structural breaks in the moments around
the time of the first oil price shock by applying standard F-tests and Chow
tests. Second, we follow a more recent practice of re-computing each moment
over time and inspecting its convergence properties from a plot of the resulting
series. We obtain one type of plot using the recursive technique of re-computing
a moment for an increasing number of observations from some minimum
number to the full sample size; this plot is relevant to the question of how
sensitive the moment is to changes in the sample size. We obtain another type
of plot using the rolling technique of re-computing a moment for a fixed sample
length which is shifted through the whole sample period; this plot is relevant
to the different question of how variable the moment is across different
sub-samples.20
The results of the F-tests and Chow tests for structural changes in the
standard deviations and cross-correlations are summarized in Table 5. The
tests are based on a decomposition of each series into two sub-samples covering
the periods before and after 1972(IV).21 The F-tests turn out to be positive (at
the 5 per cent level) in 13 of the 26 cases considered: in other words, half of
the standard deviations exhibit signs of instability. This is to be contrasted
with the striking evidence of stability in the cross-correlations: in only three
of the Chow tests is the null hypothesis of no structural break rejected.
The apparent greater instability of standard deviations relative to cross-
correlations was confirmed by the results of our recursive and rolling computa-
tions of these moments.22 In general, the plots of standard deviations display
wider variations and less convergence than the plots of cross-correlations. A
common finding is that the plots of rolling moments (where the length of the
fixed sample was set at five years) show considerably more variation than the
plots of recursive moments. The use of the former makes it much easier to
reject stability than the use of either the latter or the F-tests and Chow tests
reported above. An illustration of this is given in Figure 4. The standard
deviations of expenditure components and nominal variables are generally
largest during the 1970s. Exceptions to this are the standard deviation of
government expenditures (which tends to peak during the 1980s) and the
standard deviation of fixed investment (which tends to peak during the mid-
1960s-early 1970s). The majority of cross-correlations, while exhibiting some
quantitative instability, display a striking degree of qualitative stability. Thus,
the signs of most co-movements between variables are broadly robust to
changes in the sample period. This is true of the co-movements that we have
singled out as conflicting with traditional views of the business cycle-the
positive co-movements of real wages and productivity with output, and the
negative co-movement of prices with output. The case of the latter is illustrated
in Figure 5(b). The most notable instances of qualitative instability are to be
found in the cyclicalities of the real interest rate, net exports, government
expenditures and money. The most dramatic of these is government expen-
ditures, illustrated in Figure 5(a).
Two general conclusions can be drawn from the results in this section. First,
there is much greater stability in the co-movements between variables than in

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
TABLE 5
STABILITY TESTS OF SAMPLE MOMENTS
F-test for Chow test for

st. dev. stability cross-corr. stability
Var.
X Test statistic p-value Test statistic p-value
Y 060 004 -
LH 041 0.00 0-54 0-58
E 071 016 2*22 0.11
HW 033 0.00 082 044
KS 041 0.00 098 038
IS 1.11 066 1*27 029
RW 033 0.00 0-69 0-50
RI 023 0.00 0-79 046
LP 077 028 1*64 020
TC 042 0.00 1-77 017
NC 031 0.00 0-20 082
DC 1*41 016 12'46 0.00
TI 062 0.05 084 0-43
FI 0*77 0*29 0-85 0*43
II 0*49 0.00 0*56 0-57
NX 0*63 0*06 1*53 0-22
X 1*13 0*61 0*91 0*40
M 0*49 0.00 0*40 0-67
GP 2*67 0.00 0*69 0-50
S 1 42 0.15 3*04 0.05
MS
Ml 0*68 0-16 1*67 0.19
M3 0*79 0*40 0-83 0*44
V
Vi 0*63 0.10 5-71 0-00
V3 0*52 0-02 0*67 0-52
P 0.11 0*00 0*79 0*45
TT 0-69 0-13 3*76 0-03
3.0-
Recursive
2.5 . .................. Rolling
2.0-
1.5-
1.0 -
0.5-
1956 1960 1965 1970 1975 1980 1985 1990

The recursive standard deviation is computed as the moment of the series from 1956(I) up to t.
For the rolling estimate, t denotes the central observation of a 5-year period.
FIGURE 4. Standarddeviationofoutput.

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
0.8-
0.6-
0.4-
0.2-
-0 .6 -
1955 1960 1965 1970 1975 1980 1985 1990
(a) Output and government spending
0.2-
0-
-0.2 -
-0.4-
-0.6-
-0.8-
-1.0 I
1955 1960 1965 1970 1975 1980 1985 1990
(b) Output and prices

FIGURE 5. Rollingcorrelations.
the volatilities of variables. Second, it is possible to reach different answers to

the question of stability depending on which method of testing is used.
IV. CONCLUSIONS
In this paper we have sought to present the first detailed empirical study of
contemporary business fluctuations in the UK. We have followed the
methodology of modern business cycle research in conducting an atheoretical
statistical analysis of the cyclical properties of key aggregate time series. Our
findings confirm the existence of substantive cyclical regularities both across
countries and across time. Some notable irregularities are also to be found.
There is a high degree of cross-country robustness in the volatilities of
expenditure components and production inputs, and in the co-movements of
real and nominal variables with output. We are compelled to single out, again,

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
the robustness of the counter-cyclicality of prices and of the pro-cyclicalities

of real wages and productivity. Of the cross-country differences, the most
notable are the relatively high volatilities of UK consumption and net exports,
and the relatively high (low) volatility of UK hours per worker (employment).
Conclusions about the existence of dynamic regularities can be sensitive to
the method of testing. All methods, however, yield the result that cross-
correlations are markedly more stable than standard deviations.
It is worth emphasizing the distinction between quantitative and qualitative
irregularities. It would not be surprising to find more instances of the former
than the latter. What might be surprising, however, is just how robust the
qualitative properties of the data are. Thus, both across countries and across
time, we observe fundamental similarities in the co-movements between vari-
ables. Such similarities are important since they invite one to think of a single
organizing framework for studying business cycle phenomena. The construc-
tion of such a framework lies on our agenda for future research.
APPENDIX: DATA DEFINITIONS AND SOURCES
Series Definition Source
Output (Y) Gross domestic product at constant Datastream

prices
Labour hours (LH) HW x 14 x E, de-indexed
Hours per worker (HW) Average weekly hours per operative in Employment
manufacturing industries Gazette
Employment (E) Number of employees in employment Datastream
Capital stock (CS) Value obtained from iterating on
CS(t + 1) = TI(t) + (I-8)CS(t)
Inventory stock (IS) Value of manufacturing stocks at Datastream
constant prices
Real wage (R W) Nominal wage rate of manufacturing Datastream
workers. P
Real interest rate (RI) [(1 + yearly based long-dated gilts Datastream
redemption yield)0 25- 1] - A log P
Labour productivity (LP) Y/ LH
Total consumption (TC) NC + DC
Non-durables (NC) Consumers' expenditure on non- Datastream
durables and services at constant
prices
Durables (DC) Consumers' expenditure on durables at Datastream
constant prices
Total investment (TI) FI + II
Fixed (FI) Gross domestic fixed capital formation Datastream
at constant prices
Inventory (II) Changes in stocks and work in progress Datastream
at constant prices
Net exports (NX) X-M
Exports (X) Exports of goods and services at Datastream
constant prices
Imports (M) Imports of goods and services at Datastream
constant prices
Govemment purchases General final government consumption Datastream
(GP) at constant prices
Savings (S) Y-TC-G

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
APPENDIX-continued
Series Definition Source
Money supply
(Ml) Money supply Ml at current prices Datastream
(M3) Money supply M3 at current prices Datastream
Velocity
(Vi) Px Y.Ml
(V3) Px Y?M3
Prices (P) Implicit deflator of gross domestic product Datastream
OECD output (Y*) Aggregate GDP of OECD countries at Datastream
constant prices and constant
exchange rates
OECD consumption (C*) Aggregate total private and government Datastream
consumption of OECD countries at
constant prices and constant
exchange rates
Terms of trade (TT) Implicit price deflator of imports + implicit Datastream
price deflator of exports
Note: All series are quarterly and seasonally adjusted.
ACKNOWLEDGMENTS
We wish to thank Hannah Searle for assisting us with the collection of the data. The
comments of David Backus, an anonymous referee, and participants in seminars at
the Universities of Southampton and Essex are gratefully acknowledged. The usual
disclaimer applies.
NOTES
1. For a discussion of this methodology, see Danthine and Donaldson (1992) and Prescott (1983,
1986).
2. The non-stationarities in macroeconomic time series are associated with the presence of
stochastic growth components, or common stochastic trends, typically thought of as originating
from random technology shocks following integrated processes (see e.g. King et al. 1991;
Nelson and Plosser 1982).
3. The reason for taking logarithms is that we are interested in the percentage (rather than
absolute) deviations from trend. In some cases (where a series contains negative points) the
logarithmic transformation will not be possible and we will then use the alternative device
of taking ratios of variables.
4. Our procedure was to use a GAUSS version of a very efficient FORTRAN sub-routine written
by Ed Prescott. The GAUSS program is available on request.
5. Our hours series is for operatives in manufacturing industries. We chose this series partly
because we viewed it as being most representative and partly because we found it as being
most complete and readily available. We appreciate the problems associated with using total
hours as a measure of labour input. In particular, it does not account for differences across
workers in their relative contributions to ouput. These differences are important because the
cyclical variation in hours is not the same for workers of different skills. An ideal approach
would be to use a human capital-weighted measure of labour input.
6. Our procedure was to iterate both forwards and backwards on the capital accumulation
equation, k,?1 = it + (1 - 8)kt, using our series for investment and a few (consistently m
observation on capital. An average (quarterly) rate of depreciation of 8 = 0 43 per cent was
also computed from this equation. The percentage error in our calculations was found to be
as low as 05.
7. We use an ex post definition of the real interest rate. This is our nominal interest rate series
minus the actual rate of inflation.
8. The observation dates back to Dunlop (1938) and Tarshis (1939). By way of a slight
qualification, we found that the result was sensitive to the decomposition of the hours series.
While the correlation between real wages and hours per worker turns out to be 0 007, the
correlation between real wages and employment amounts to 0 12.

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
9. Okun (1962) observed that, on average, output tends to increase by 3 per cent when labour
input increases by 2 per cent. See Bec and Henin (1990) for an empirical investigation of
labour productivity in the USA and European countries.
10. Our series for non-durables is for expenditure on these goods; an ideal series would measure
the services they yield.
11. The definition of savings as a residual is common in the literature. Owing to negative data
points, we take the filtered output ratios (as opposed to the filtered logarithms) of the series
for inventory investment, net exports and savings. These ratios are close to zero. Hence, their
variances and cross-correlations may be safely attributed to movements in the numerators. A
potential problem with this approach is that the price deflators for net exports and output
are different. An alternative strategy, which runs into other difficulties, would be to use nominal
series.
12. It is unclear whether a different conclusion would be reached if consumption of durables was
measured in terms of the services of these goods, rather than in terms of the amount of
expenditure on them.
13. The result may be due, in part, to our measure of net exports (the filtered output ratio of net
exports in real terms), which differs from the measure used for the USA (the filtered output
ratio of net exports in nominal terms). In Blackburn and Ravn (1991a) we use the first measure
for both the UK and the USA and find that the standard deviation of net exports in the UK
(USA) is 1-03 (0.75) over the period 1970(I)-1990(II).
14. We also experimented with the retail price index and found that the results were almost
identical.
15. Owing to lack of quarterly data on money stocks, our sample period had to be confined to
1963(I)-1989(I).
16. The cross-correlations between Ml (M3) and prices are -0-31 (0.17), -0 30 (0.07), -0-26
(-0-04), -0-19 (-0-12), -0 07 (-0 17), 0-01 (-0-21), 0 07 (-0.22), 0.09 (-0.22), 0-09 (-0-19).
17. On the international linkage of business cycles, see Backus and Kehoe (1992). On the
savings-investment-capital mobility relationship, see Feldstein and Horioka (1980).
18. According to the classical interpretation of the J-curve, there should be a negative contem-
poraneous correlation between net exports and the terms of trade. Our finding of a positive
contemporaneous correlation is effectively the result of a rightward shift in the J-curve and
might be due to the measurement problem alluded to in n. 11.
19. The few notable exceptions include Backus and Kehoe (1992), Blackburn and Ravn (1991a, b)
and Danthine and Girardin (1989).
20. A potential problem with the recursive technique is that the weight given to each new
observation decreases. See Ravn and Sola (1991) for an investigation of this problem.
21. Let {xj} be a time series and x be the mean of this series. Let k = 1, . . ., K denote a sub-period,
Tk denote the number of observations in sub-period k and T denote the number of observations
in the whole sample. Finally, Let RSS and RSSk be the residual sum of squares from
whole-sample and sub-sample linear regressions of x, on some other variable. The
statistics for changes in variances aiid correlations are, respectively,
FV = (T2- 1) _(x1, -X)2/( T -1) _(x21 -)2,

and
Fc = (T- 2K)(RSS - RSS, - RSS2)/[K(RSS, + RSS2)],

both of which are F-distributed. The critical values for FV are Fai2(TI-1, T2-1) and
Fi-a12(Tl,-1 T2-1). The critical value for Fc is Fa(K, T-2K).
22. These results, which are far too numerous to report in the paper, are available on request.
REFERENCES
ARTUS, J. R. (1975). The 1967 devaluation of the pound sterling. International Monetary Fund
Staff Papers, 22, 595-640.
BACKUS, D. K. and KEHOE, P. J. (1992). International evidence on the historical properties of
business cycles. American Economic Review, forthcoming.
and (1991). Dynamics of the trade balance and the terms of trade. Mimeo, New York
University.
, and KUDLAND, F. E. (1991). Dynamics of the trade balance and the terms of trade:
the J-curve revisited. Mimeo, University of New York.
BAXTER, M. and CRUCINI, M. J. (1990). Explaining saving-investment correlations. Rochester
Centre for Economic Research Working Paper no. 224.
BEC, F. and HENIN, P. Y. (1990). Labour productivity in business cycles: comparative evidence
on competing paradigms. Paper presented at the Conference on Instability and Persistence,
Paris, 4-6 January.

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
BLACKBURN, K. and RAVIN, M. 0. (1991a). Contemporary macroeconomic fluctuations: an

international perspective. University of Southampton Discussion Paper no. 9106.
and (1991b). Univariate detrending of macroeconomic time series. University of
Southampton Discussion Paper no. 9112.
BRANDNER, P. and NEUSSER, K. (1992). Business cycles in open economies: Stylized facts for
Austria and Germany. Weltwirtschaftliches Archiv, 128, 67-87.
BURDETT, K. and WRIGHT, R. (1989). Unemployment insurance and short-term compensation:
the effects on layoffs, hours per worker and wages. Journal of Political Economy, 97, 1479-95.
BURNS, A. F. and MITCHELL, W. C. (1946). Measuring Business Cycles. New York: National
Bureau of Economic Research.
CANOVA, F. (1991). Detrending and business cycle facts. Mimeo, Brown University.
CORREIA, H., NEVES, J. L. and REBELO, S. (1992). Business cycles from 1850 to 1950: new facts
about old data. European Economic Review, 36, 459-67.
DANTHINE, J. P. and DONALDSON, J. B. (1992). Methodological and empirical issues in real
business theory. European Economic Review, forthcoming.
and GIRARDIN, M. (1989). Business cycles in Switzerland: a comparative study. European
Economic Review, 33, 31-50.
DORNBUSCH, R. and KRUGMAN, P. (1976). Flexible exchange rates in the short run. Brookings
Papers on Economic Acivity, 3, 537-75.
DUNLOP, J. T. (1938). The movement of real and money wage rates. Economic Journal, 48,
413-34.
ENGEL, C. and KLETZER, K. (1984). Savings and investment in an open economy with non-traded
goods. NBER Working Paper no. 2141.
FELDSTEIN, M. and HORIOKA, C. (1980). Domestic saving and international capital flows.
Economic Journal, 90, 314-29.
FINN, M. (1990). On savings and investment in a small open economy. International Economic
Review, 29, 1-21.
FREIDMAN, M. and SCTTWARZ, A. (1982) Monetary Trends in the United States and the United
Kingdom. Chicago: University of Chicago Press.
HODRICK, R. and PRESCOTT, E. C. (1980). Postwar US business cycles: an empirical investigation.
Mimeo, Carnegie-Mellon University.
HUNG, V. (1990). A study of the saving-investment correlation puzzle. Mimeo, London School
of Economics.
KING, R. G., PLOSSER, C. I., STOCK, J. and WATSON, M. (1991). Sochastic trends and economic
fluctuations. American Economic Review, 84, 819-40.
and REBELO, S. T. (1989). Low frequency filtering and real business cycles. Rochester Centre
for Economic Research Working Paper no. 205.
KRUGMAN, P. R. and BALDWIN, R. E. (1987). The persistence of the US trade deficit. Brookings
Papers on Economic Activity, 1, 1-43.
KYDLAND, F. E. and PREScoTT, E. C. (1990). Business cycles: real facts and a monetary myth.
Federal Reserve Bank of Minneapolis Quarterly Review, 3-18.
LUCAS, R. E. (1977). Understanding business cycles. Carnegie-Rochester Conference Series on
Public Policy, 5, 7-29.
MENDOZA, E. G. (1990). Some international evidence on the correlation between the trade
balance and the terms of trade. Mimeo, International Monetary Fund.
MITCHELL, W. C. (1913). Business Cycles. Berkeley: University of California Press.
(1923). Business cycles. In Business Cycles and Unemployment. New York: National Bureau
of Economic Research.
(1927). Business Cycles: The Problem and its Setting. New York: National Bureau of Economic
Research.
NELSON, C. and PLOSSER, C. I. (1982). Trends and random walks in macroeconomic time series.
Journal of Monetary Economics, 10, 139-62.
OBSTFELD, M. (1986). Capital mobility in the world economy: theory and measurement. Carnegie-
Rochester Conference Series on Public Policy, 24, 55-104.
OKUN, A. M. (1962). Potential GNP: its measurement and significance. In J. Pechman (ed.),
Economics for Policymaking. Cambridge, Mass.: MIT Press.
PRESCOTT, E. C. (1983). Can the cycle be reconciled with a consistent theory of expectations or
a progress report on business cycle theory? Federal Reserve Bank of Minneapolis Research
Department Working Paper no. 239.

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC
(1986). Theory ahead of measurement in business cycle research. Carnegie-Rochester Confer-

ence Series on Public Policy, 25, 16-44.
RAVN, M. 0. and SOLA, M. (1991). The use of recursive variance plots. Mimeo, University of
Southampton.
ROSE, A. and YELLEN, J. (1989). Is there a J-curve? Journal of Monetary Economics, 24, 53-68.
SPITALLER, E. (1980). Short-run effects of exchange rate changes on the terms of trade and trade
balance. International Monetary Fund Staff Papers, 27, 320-48.
TARSHIS, L. (1939). Changes in real and money wage rates. Economic Journal, 49, 150-4.
WARNER, D. and KREININ, M. E. (1983). Determinants of international trade flows. Review of
Economics and Statistics, 65, 96-104.
WRIGHT, R. (1991). The labour market implications of unemployment insurance and short-term
compensations. Federal Reserve Bank of Minneapolis Quarterly Review, 11-19.

86.186.172.181 on Fri, 25 Feb 2022 09:49:04 UTC

Business Cycles in The United Kingdom

Uploaded by

Copyright:

Available Formats

Business Cycles in The United Kingdom

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Business Cycles in The United Kingdom

Uploaded by

Copyright:

Available Formats

Business Cycles in the United Kingdom: Facts and Fictions

Author(s): Keith Blackburn and Morten O. Ravn

Stable URL: https://www.jstor.org/stable/2554886

This content downloaded from