Chapter 3

REGIONAL UNEMPLOYMENT IN OECD COUNTRIES1

A. INTRODUCTION techniques and macroeconomic and microeconomic

theory find particular applications in analysing
regional issues, and urban science and economic geo-
Labour markets are often considered only from a graphy provide specific perspectives. This section
national perspective, especially within a multi- offers a brief survey of these interpretations, not with
national publication such as the Employment Out- the intention of deciding between competing theories
look. This perspective is limited. In reality social and - the study here is not broad enough in scope to do
family ties, investments in housing, knowledge of a that - but in order to give a balanced qualitative
particular locality, and lack of information often description of regional unemployment differences
restrict job search to a small area : individuals operate before moving on to the more quantitative analysis
in localised labour markets. Nevertheless, because in Sections C and D.
demand and supply imbalances spill over from each
small area to its neighbours, labour markets show
1. Models of regional variation
considerable homogeneity in behaviour over the
larger areas which are commonly called regions. Pre-
vious OECD work [notably OECD (1986c)j has ana-
In a simple statistical model of regional unemploy-
ment behaviour, unemployment in each region may
lysed several aspects of regional problems and poli-
cies : the main contribution of this chapter is to pres- be modelled in terms of a region-specific constant and
time trend, and a coefficient relating the regional
ent for the first time estimates of unemployment rates
for 182 major sub-national regions within 22 Member unemployment rate to the national rate [e.g. Brechling
(1967) ; King and Clark (1978) ; Johnston (1979)J.
countries. For 30 countries, long-term changes in the
Various more sophisticatcd techniques of modelling
regional pattern of unemployment are also examined.
exist2. However, while such techniques can categorise
Section B of the chapter first discusses factors in
regional labour market developments, and hence the
regions on the basis of their unemployment behav- \;
iour they do not explain theorigins of these differ-
-._---
significance of regional unemployment rate differ- % ~

ences for the national economy. Section C presents ences.

estimated unemployment rates in 1987for 182 regions Underlying such specifications is an assumption that
of the OECD area in the form of a 16colour map, regional differences in unemployment behaviour
allowing intra- and inter-national variations in uncm- through time reflect differing economic structures.
ployment to be seen at a glance. Section D examines Regional economies are often specialised in a limited
changes through time in the size and distribution of range of products [Brewer (198S)l. Thus industrial
regional unemployment differences, and Section E composition may help to explain the behaviour of a
concludes. region in terms of the fact that some industries are
more sensitive than are others to the economic cycle,
are more exposed to foreign trade conditions, or
exhibit greater seasonality in employment. However,
actual movements within regions are greater than can
B. FACTORS IN REGIONAL LABOUR
MARKET DEVELOPMENTS be explained by industrial structure alone3.

a) Demand at the regional level

There are many approaches to the study of regional In a Keynesian approach, employment at the
labour market differences. Econometric modelling regional level, as at the national level, is constrained

95
 primarily by demand. Net demand from outside the migration flows could well be a factor leading to
region is one of the main exogenous components, and increased regional unemployment differentials.
Thirlwall (1980, p. 420) develops a model in which Quite apart from the numeric importance of migra-
“regional problems of slow growth and high unem- tion flows, their selective composition is significant.
ployment are, in essence, balance-of-payments prob- Workers who migrate are disproportionately male,
lems stemming from a weak trading sector” and where young and skilled, and this adversely affects the com-
depressed regions are characterised by a low income position of the labour force remaining in the depressed
elasticity of demand for their exports. More generally, regions4. Given the way that unemployment varies
any movements in demand for, or the price of, par- by qualification, age, sex, etc., regional differences
ticular products - e.g. competition from developing in the structure of the labour force may have a role
countries in steel products, or a fall in the oil price - in explaining persistent differences in unemployment
may have income multiplier effects within a region rates.
that go beyond that particular industry.
Regions, unlike nations, cannot use exchange rate c) Cumulative carnation
adjustments to correct a trade imbalance, and gen- According to Kaldor (1978, p. 148), the “principle
erally have less autonomy in fiscal and monetary of cumulative causation - which explains the unequal
policy. However, tax and expenditure policies for- regional incidence of industrial development by
mulated at the national level can act as “automatic” endogenous factors resulting from the process of his-
fiscal stabilisers (taxes varying with the level of local torical development itself rather than by exogenous
income, while public expenditure is invariant). This differences in ‘resource endowment’ -is an essential
mechanism, perhaps supported by discretionary pol- one for understanding of the diverse trends of devel-
icy measures to help depressed regions, can allow opment as between different regions”. In his view,
trade deficits at the regional level to be sustained. the economies inherent in large-scale production
processes, and opportunities for increased differen-
b) Growth and structure of the labour force tiation and specialisation created by proximity of dif-
ferent industries, act to increase productivity growth
The rate of labour force participation, especially and competitiveness and encourage further invest-
of women, is an important determinant of labour ment in areas that already benefit from a high con-
supply at the regional as at the national level. centration of modern industries. Self-selection,
Although many factors are involved, female labour whereby the workers who move are those who are
force participation is thought to be Iower in regions most employable, may also concentrate high-wage
which have high unemployment. This participation high-skill jobs in some areas at the expense of others :
behaviour would reduce unemployment rate differ- large-scale migration then becomes a factor promot-
entials, relative to disparities in employment condi- ing regional imbalances rather than reducing them.
tions between different areas. In some countries, concentration of management
Migration, as well as natural increase, affects the activity may be more significant than concentration
overall growth of the labour force. Total numbers of of the manufacturing plant itself. Companies find
persons changing region in 1980, including immigrants advantages in having offices close to other companies’
from abroad, were about 1.5-2 per cent of the pop- headquarters and to government : in the 1970s,62 per
ulation in Western Europe, 2-3 per cent in Australia, cent of the largest British manufacturing firms had
Canada and Japan, and nearly 4 per cent in the United their headquarters in London, and 78 per cent of the
States, with some indication that mobility in 1980 was largest French manufacturing firms in Paris [Com-
lower than in 1970 [OECD (1987b),Table 3.7 : based mission of the European Communities (1981)J.Thcre
on 5 to 51 regions per country]. Across the European is a corresponding concentration in capital cities of
Community, net in- or out-migration rates of 0.5 per higher-paid non-manual jobs, which are a particularly
cent per year for individual regions were not uncom- significant element in occupational and earnings dif-
mon in the 1960s, so that over the decade migration fercnces.
7, had a significant impact on labour market balance ;
, for example, during the 1960s, net outward migration d) Equilibrating mechanisms
rates averaged over 1 per cent of the population per
year from the Mezzogiorno of Italy. However, in the In standard economic theory, the flexibility of . ‘

1970s net migration rates fell [Commission of the wages promotes labour marke? equilibrium. If unem-
European Communities (1981)], and differentials in ployment becomes regionally unbalanced, relative
unemployment rates, in terms of percentage points, wages fall in the regions with the highest unemploy-
have risen sharply since the 1960s. This slowing of ment : this encourages higher production and invest-

96
ment in these regions, reducing unemployment in ing, textiles and heavy industry have been taking place
them and eventually restoring an equilibrium with during the past two decades, with adjustment through
uniform wage levels. Undcr ccrtain assumptions migration to other areas or through local development
- notably, that there is free trade in goods and that of other industrial and service activities. However, at
the production technology is the same everywhere - least in Europe, adjustment has often been only par-
the “factor-price equalisation theorem” predicts that tial, and thc regions most affected by employment
in equilibrium wages will be the same in different declines in traditional manufacturing activities have
areas, even without mobility of labour or capital [e.g. often experienced pcrsistcntly high unemployment.
see Layard and Walters (1978)l. The assumptions Unemployment in some regions is also linked to
behind this theorem are clearly not well met across the long-term decline of employment in agriculture.
countries, but may be somewhat better approximated This sector, especially in some backward regions of
across regions. Southern Europe, tended to be characterised by a
Large companies, through systematic research into high incidence of forms of “underemployment” (hou-
optimal locations, may be led to invest in depressed sewives performing occasional tasks, other unpaid
areas where labour availability is good : although family workers, daily and seasonal workers working
assessment of other factors such as availability of skills only few days per year), but by relatively few unem-
and experience, and the labour relations climate, also ployed [Accornero and Carmignani (1986)l. The pro-
come into play. gressive decline of traditional agricultural activities
One factor limiting regional imbalances is that rapid has been associated with extensive migration to urban
growth in particular areas will strain the local infra- centres, often within the same region, where the
structure. While investment in housing and public unemployment has become more visible.
utilities in rapidly growing areas may in time catch Persistent regional differentials may also reflect
up with a rapid rise in demand, high land and property institutional rigidities. Centralised unions usually
prices, high rents, and road traffic congestion are more favour thc lcvelling of wages across regions.
likely to remain at high densities of population and Drewes (1987) found that wagc spillover from other
industrial activity. Such factors clearly limit tenden- Canadian regions contributed to persistently high
cies towards cumulative regional concentration. unemployment rates of the Atlantic region, and cites
Nonetheless the process may not be effcctivc as it similar results for the United States and the United
should bc, as some of the social diseconomies of Kingdom.
congestion are external to the individual producers. When wage.differe2tials
- --
~~ arise, the housing market
may reduce their effectiveness in allocating labour.
Bover er al. (1988) observe that increases in relative
e) Persistence in regional differentials wagcs in the South-East of the United Kingdom have
Long-run differences in regional unemployment
levels may, in theory, represent an equilibrium where prices
c- -
-
been quickly followed by increases in relative house
--
there, which discourages in-migration and even
such factors as high wages or unemployment benefits, was associated, in 1973, with a rccord level -9f.nct
cheaper housing, favourable climatic conditions, or out-migration. In the rented sector, regulation and-
an attractive environment encourage people to stay priority rules in the private and public sectors can
in regions where unemployment is high. Marston make the real cost of accommodation for newcomers
(1985), in an analysis of thirty Statistical Metropolitan to an area much higher than it is for people who have
Areas for the United States in the period 1970-78, been established in it for many years.
argues that migration flows among areas are large Even when equilibrating mechanisms are active,
relative to unemployment rate differences and that they may only ensure that periods of developing con-
observed unemployment differentials are “reflections centration and imbalance (as in models of cumulative
of workers’ underlying preferences for certain causation) are followed by periods where the more
areas” (p. 57). backward regions catch up. If nothing tips the advan-
On the other hand, some persistent regional dif- tage decisively towards the central or the peripheral
ferences may reflect the slow operation of equili- regions, relative patterns of regional advantage and
brating mechanisms such as~migration,company relo- disadvantage may be left, in statistical terms, little
cation and relative wages di&ssed in earlier sections. changed.
Structural unemployment may develop in certain
regions because the skills possessed by workers in
f l Spatial factors, urban science and economic geo-
declining industries do not match those required by FPhY
expanding industries and retraining takes time. Regional economies can be strongly influenced by
Employment declines in regionally concentrated min- their distance from important markets, their prox-

97
 imity to coastal and river transport, and the quality characteristics of unemployed workers, particularln
of infrastructure such as roads and railways. Some in terms of skill, work expcriencc or location, differ \,
studies, generally at a very disaggregate level, empha- from those of the jobs that are available”. If some
size the importance of specifically spatial elements in regions suffer from labour shortages while others
regional behaviour, for example modelling the spread suffer from high unemployment, or if particular skills
of innovations in terms of centres from which changes are in surplus in one part of the country while those
originated, and of patterns of diffusion that vary skills are demanded in another part of the country,
according to the distances of local areas from these the national economy is likely to suffer simultaneously
centres [Malecki (1983)J. from limitations to its productive potential, inflation
Urban science emphasizes distinctions between pressures, and persistence of an unemployment prob-
characteristics of inner-city, suburban and rural areas, lem, so that regional variations are themselves causes
and the importance of local labour markets defined of unemployment at the national level6.
in terms of travel-to-work areas. The degree of urban- The Phillips curve, which relates wage increases to
isation of the various regions is itself likely to affect unemployment, is often thought to involve a rapid
their global unemployment rate. During the 1970s, acceleration of wage inflation as unemployment
the earlier trend increase in urban populations has approaches zero. The Beveridge curve, relating
given way to a more mixed pattern. With the major unemployment to v21an , is also concave [e.g. see dL
exception of countries in Southern Europe, popula- OECD (1988b)r-Non-l ity in the regional rela-
tions have tended to move away from urban centres tionships implies that a given average level of wage
into the surrounding hinterland, and into smaller set- inflation or vacancies will correspond to a higher
tlements. Urban areas do not have uniformly higher national leverof uiiemployment when regional vari-
or lower unemployment than do rural areas, but ations are large - for example, with some regions
unemployment rates in the expanding outer urban having high vacancies and little unemployment and
areas have consistently been lower than in the urban others having high unemployment but few vacancies7.
centres [Commission of the European Communi-
ties (1981) ; OECD (1988f)l.
Among the factors underlying past trends in urban- b) Local implementation of labour market policies
isation have been industrialisation and falls in trans-
portation costs. During the industrial revolution, In the early 1980s, rising unemployment was so
cheaper transport made it less costly to feed urban widespread across OECD countries and all the labour
markets within them that regional differences tended
populations and concentrate the production of manu-
factured goods, while in the post-war period, relative to attract little special attention. Interest focused more
upan developments in the world-wide economic sit-
‘x4,,s
__ commuting ._
falls. in _---I ~costs___..
I..-have encouraged expansion
uation. However, in the subsequent period of steady
into suburban areas. However, in recent decades con-
economic growth and relative stabilisation of the area-
tinuing improvements in road, radio, television, and
wide unemployment rate, many of the significant
telephone communications may have made life in
changes in labour markets have becn localised to
remote areas more attractive and non-agricultural
business activities in them more practicable, conttib- particular countries, and even to particular regions.
uting to a slowdown or reversal of urbanisation. At the same time, continued caution in macroecon-
omic policy has led to increased emphasis on the
improvement of labour market structure, often
involving a reduced role for central government com-
2. Regional variations and the national economy
pared with local bodies. Policies based upon education
and training, upon tackling the particular problems
Regional differences have significant implications of displaced workers and other groups at high risk of
for the economy and for the implementation of labour unemployment, and upon encouraging the virtuous
market policies at the national level.
circie of growth and entrepreneurship at the local
level, require implementation at the local level. How-
ever, local initiatives are inevitably more energetic
a) Regional mismatch
L. and successful in some areas than in others, and can
Regional variations in the structure of demand and increase regional differences as well as counteracting
supply have often been considered as a component them. Special attention may therefore be necessary
of “mismatch”. Jackman and Roper (1987, p. 10) to the way particularly disadvantaged regions can
identify “mismatch” as: “a situation in which the generate growth.

98
 C. REGIONAL UNEMPLOYMENT RATES detail available in the unemployment data from labour
IN 1987 force survcys is limited, and regions with average
populations of over 10 million had to be used. Else-
where, although average regional populations follow
An empirical investigation at the OECD level of the size criterion fairly closely, there is considerable
all the factors causing rcgional variation would be a variability
.-
about thc average'".
major undertaking. The statistical investigations here The four urban areas of the District of Columbia
relate only to the unemployment rates themselves. in the United States, Brussels-Cn Belgium, and Ham-
Chart 3.1 presents estimates of unemployment rates burg and Bremen in Germany are geographically very
for 182 regions within 22 OECD countries. The view small in relation to nearly all the other regions in
of regional unemployment differences it gives of Chart 3.1 (they have been shown as circles greater
course reflects both limitations to the comparability than real scale in order to make them visible on the
of the unemployment rate data across countries, and map), and a significant proportion of their employees
choices about how countries should be divided into are resident outside the region. This creates some
regions. problems in interpreting data on labour market bal-
ance: labour force survey data used here allocate
employees to their place of residence, whereas admin-
1. Choice of regional units istratively-based data sources allocate employment
(and incomes, etc.) to places of business activity.
The regional units shown in Chart 3.1 are named,
together with information on population, area and
estimated unemployment rate, in Table 3.A.1 of 2. Construction of unemployment rates
Annex 3.A. Where necessary the regions are defined
in more detail in Table 3.A.2. These regional units The Unemployment rate estimates shown in
were chosen firstly as a function of the availability of Chart 3.1 are all based on regional level information
statistics at the regional level - not only statistics of from labour force surveys, adjusted as described in
employment and unemployment, but also other major Annex 3.A to this chapter. In many cases only small
economic variables such as population, earnings and adjustments, in respect of timing of the survey and
incomes that may be studied in the future -without treatment of armed forces employment, were needed
attempting to delimit the regions on the basis of to achieve consistency at the national level with the
economic criterias. For EEC countries, the EUKO- OECD's Standardized Unemployment Rate (SUR)
STAT NUTS (Nomenclature of Territorial Units for annual average estimate for 1987. However, for some
Statistics) classifications, which divide the 12 EEC countries, estimation of the national rate in 1987 on
countries into 64 units at level 1, and 167 units at a full SUR basis was not possible : and in the case of
level 2, were available. Chart 3.1 uses the NUTS Turkey, regional estimates from the 1985labour force
level 1 units, with some adiustments in respect of survey were extrapolated to 1987 using changes in
certain features9. In most non-EEC countries, the., the national total rate. Although these fimitati&s to
main regional political and . a d S . i ~ a & S . - e l t s % comparability are a problem, they are not thought to
- states in the United States, states and territories distort greatly the overall pattern of relative uncm-
in Australia, provinces in Canada and Finland, COUD- ployment rates across OECD regions".
ties in Sweden and prefectures in Japan - areimpor-
tant as statistical units and hence were used. Taking
such units as the point of departure, consistency in
the size and disposition of the regional units across 3. Patterns of regiona1 the
countries was improved by some aggregations and OECD area
disaggregations.
As regards size, the criterion followed was that the In interpreting Chart 3.1, it may help to note-that
average population of a region should be about
5 million in the countries with total populations of
the unemployment rate i n c r e w across ranges!&
--
shown by the transition from blue;-through green to
kc
50 million or more, declining to 1 to 2 million in yellow and finally red. The division between the two
countries with only a few million inhabitants. In this highest ranges, at a 18.9 per cent unemployment rate,
sense, the smaller countries in Chart 3.1 are repre- is nine times higher than the division between the
sented by more regions than their population size two lowest ranges: and each range represents an
merits. The level of regional disaggregation in Turkey unemployment rate about 1.2 times higher than the
and Japan is anomalous : in these countries, regional previous range.

99
 Chart 3.1
UNEMPLOYMENT RATES IN REGIONS OF OCDE COUNTRIES, ANNUAL AVERAGES 1987

Unemployment rates (in percentages of total labour force) Source : See annex 3.A.

2.10 2.52 3.03 3.64 4.37 5.25 630 7.57 9.09 10.31 lS.10 15.74 18.90 94
6861 3C130/N31 0
 The highest regional unemployment rates in the country and upon their delimitation. However, some
OECD area are concentrated in Central and Southern indicators are presented in Table 3.1, which shows
Europe, with only one region outside Europe having for each country and supra-national region the num-
an unemployment rate in the highest three ranges. ber of regions, the average unemployment rate, and
As a gener*att_ern, regions on the periphery of rsiDn : the weighted standard C
Europe (tKe south olI<aly and Iberian countries, parts t of variation (standard devia-
of Turkey, Ireland and the northern part of the United tion divided b j the mean), and the unemployment '-
,1

Kingdom, and the northern regions of Scandinavian rates for the top and bottom quarters of the labour
countries) all suffer high unemployment relative to force by regional unemployment rate.
immediate neighbouring areas. (Regional income lev- Table 3.1 indicates that, broadly speaking, regional
els also decline towards the periphery [e.g. see Nicol variations in unemployment in Europe are greater
and Yuill (1982)j). However, this centre-periphery than in North America. In each of the three groupings
grading of unemployment rates has many exceptions : shown for Europe : Nordic countries, Southern
unemployment rates in northern Norway and Sweden Europe, and Central and Western Europe, both the
are lower than in Central Europe, and rates in some coefficient of variation and the standard deviation of
areas of central Germany and central France are high. unemployment are greater than in North America.
Regional variation in unemployment is less in small In terms of either indicator, Italy, Portugal and Spain
countries than in large ones. Rates in most of the in Southern Europe, the United Kingdom, Belgium
smaller European countries with three to five regions and Germany in Central and Western Europe, and
- Sweden, Ireland, Denmark, thc Netherlands, Bel- Finland among the Nordic countries have the highest
gium and Austria - span only two or three of the regional dispersion of unernploymcnt rates.
fourteen ranges. However, Norway spans four ranges Unemployment rates for the top and bottom
(1 to 4) and Finland shows exceptional variation, quarter of the labour force express nearly the same
spanning eight ranges (2 to 9). Ranges of variation information : the standard deviation measurc is highly
are greater in the larger European countries, with correlated with the difference between top and bot-
4 ranges in France, 6 in the United Kingdom, 7 in tom quarters (Spearman rank correlation coefficient
Italy, and 8 in Germany. As between all the major of 0.98) and the coefficient of variation with the ratio
countries, any differences in national averages need between the tog and bottom quarters (Spearman rank
to be seen against this background of substantial correlation coefficient 0.96). In smaller countries the
overlap at the regional level. ratio of the top quarter to bottom quarter unem-
The 51 states of the United States are distributed ployment rates is usually below 2 (though here there
across 10 of the 14 ranges needed to cover the whole are several exceptions) : for the 5 largest European
OECD. Unemployment rates in the six highest ranges countries, Japan and the United States the ratio is
characterise a band of states running from West Vir- about 2 ; and for the supra-national groupings of
ginia to the southern states along the Mexican border, regions, the ratio is above 2 and sometimes above 3.
the states of Idaho and Wyoming in the north-west,
and Michigan and Alaska. The north-east corner of
the United States (including some small states with
high population density) has some of the lowest unem-
ployment rates in North Amcrica, while the highest D. REGIONAL UNEMPLOYMENT
rates are found to the north in the easternmost prov- IN HISTORlCAL PERSPECTIVE
inces of Canada.
In Japan, in line with the tcndency noted for
Europc, the highest unemployment rates are found Independently of the size of regional unemploy-
in geographically remote areas at thc northern and ment disparities at a point in time, change in the
southern extremes of the country : thcse have unem- pattern of diTarities is also important. To the extent
-.-.---.;--
I"___--.. - <3:
ployment rates as high, or higher, than some states .'that regional patterns in unemployment rates are rel- \?.'
in the Unitcd States. &iat$&d; withEE-Jii&ist unemployment regions
changing little, regional policy may need to take a
long-term structural approach, because short-run
4. Indicators of regional disparities improvements in regional labour market conditions
will (on past experience) depend mainly upon the
Cross-country comparisons of the dispersion of national unemployment situation. If the regional dis-
regional unemployment rates cannot be precise, for tribution of unemployment changes markedly through
they depend on the number of regions used in each time, short-term assistance to areas currently affected

102
 Table 3.1. Indicators of regional unemployment dispariiies by country and country groupings: annual averages 1987
Average Standard Coefficient Unemployment Unemployment
Number rate top
of regions unemployment deviation‘ rate bottom
of variation
rate auarterb auarterb

,Nordic countries 17 3.6 2.13 0.60 6.4 1.5

Denmark 3 6.0 0.92 0.15 6.8 4.8
Finland 4 5.0 2.04 0.40 7.5 2.3
Norway 5 2.1 0.56 0.27 2.8 1.7
Sweden 5 1.9 0.57 0.30 2.6 1.1

Central and Western Europe 48 8.9 3.16 0.35 12.9 4.9

Austria 3 3.7 0.44 0.12 4.1 3.1
Belgium 3 11.1 2*22 0.20 14.2 9.3
France 11 10.5 1.69 0.16 12.9 8.6
Germany 11 6.2 2.06 0.33 8.5 3.8
Ireland 4 18.0 0.76 0.04 18.8 17.0
Netherlands 4 9.6 0.65 0.07 10.4 9.2
United Kingdom 11 10.2 2.58 0.25 13.6 7.6

Southern Europe 36 12.3 6.02 0.49 21.2 6.6

Greece 3 7.4 1.19 0.16 8.3 5.8
Italy 12 11.2 5.27 0.47 19.4 6.6
Portugal 5 7 .O 2.57 0.37 10.3 4.7
Spain 11 20.1 4.95 0.24 27.7 14.7
Turkey 5 10.1 2.15 0.21 13.1 7.6

Japan 10 2.8 0.67 0.23 3.8 2.0

Oceania 10 7.4 1.92 0.26 9.3 4.7

Australia 8 8.1 1.27 0.16 9.4 6.2
New Zealand 2 4.0 0.49 0.12 4.9 3.7

North America 61 6.4 2.00 0.31 9.1 4.2

Canada 10 8.8 2.68 0.30 12.2 6.1
United States 51 6.1 1.71 0.28 8.4 4.1

a) Standard deviation weighted by the 1987 level of the labour force.

b) Unemployment rates for the bottom quarter of the labaur force by region were calculated by ordering regions in terms of ascending unemployment rate. taking regions
until the cumulative labour force passed one quarter of t h o total. and including the last region in the calculation with an appropriate fractional weight; and similarly
for the top quarter.
Source: See Annex 3.A.

by change, and long-term efforts to reduce the sen- relate to registered unemployment as a percent-
sitivity of all areas to economic shocks, may be the age of the employee labour force ; the series for
appropriate policy. Italy and the shorter series for most other coun-
tries are based on survey estimates of unem-
ployment as a percentage of the civilian labour
force, but vary in terms of, for example, the
. 1. Changes in regional disparities since the 1960s timing of the survey ; the series for the United
\ - ___- ~

States are estimates derived from the integration

Developments through time in the average unem- of data from the population censuses, labour
ployment rates of the top quarter, middle half, and force surveys and administrative data ;
bottom quarters of the labour force, ordered by the - Most of the data series have, at some point, been
regional unemployment rates of the current yearla, affected by one or more statistical breaks in the
are shown for ten countries in Chart 3.2. The data measurement of unemployment (and/or total
for Chart 3.2 and later tables may differ from employment) or in the definitions of regional
Chart 3.1 and Table 3.1 : relevant qualifications are : boundaries. Such changes have been tackled by
- Data sources for time-series study use a variety adjusting data from early years to the basis of
of definitions (see Annex 3.B)13. Data starting later years through multiplicative chaining or, (ii’
in 1960 for Germany and the United Kingdom less often, through p’rGedures of interpolation

103
 Chart 3.2

UNEMPLOYMENT RATES FOR TOP QUARTER, MIDDLE HALF AND BOTTOM QUARTER
OF THE LABOUR FORCE, BY REGIONAL UNEMPLOYMENT RATE a :
DIFFERENCE AND RATIO MEASURES OF REGIONAL UNEMPLOYMENT DIFFERENTIALS

AUSTRALIA CANADA FINLAND

---- TOP

.------ MIDDLE

BOTTOM

70 73 76 79 82 85 88

- DIFFERENCE
----.---
RATIO
 Chart 3.2 (continued)

ITALY SWEDEN UNITED KINGDOM UNITED STATES

a. Unemployment rates for the top and bottom quarter of the labour force are defined as Indicated in Table b. France and Japan, stars represent observations from census data.
3.1. The unemployment rate for the middle half of the labour force is defined in analogous terms. For any
individual country, weights in the bottom quarter, middle half and top quarter of the labour force sum to 1, and
the average unemployment (Up,)is related to the bottom quarter, middle half and top quarter rates (U8,UM,
U), by the formula UA = (Ug:4+U~li2+U~!4). Source : See Annex 3.8.
 DIFFERENCE

in
W P
c n m
RATIO
(Logarithmic scale)
 or allocation of a national total on one basis and political concern generated by the difference
proportionately to regional statistics on another between 2 and 3 per cent unemployment rates may
basis. Such procedures eliminate artificial jumps be greater than hctwceri 12 and 13 per cent rates,
in the time-series, but may introduce some biases and from this point of view a ratio mcasurc might hc
into long-term comparisons ; more relevant.
- The size of regions is not consistent acr0-a coun- The normal relationship between regional and
tries : time-series statistics relate to 24 counties national unemployment rates is neither linear nor
in Sweden, but only 8 regions in Germany, Use multiplicative, but in between these two cases : when
of more detailed regional disaggregations some- the national rate rises, ratios between regions nor-
what increases estimated regional unemployment mally fall, while differences between them rise. Using
differentials"+. the data in Chart 3.2 a linearity parameter, which
Analyses of trends in the size of regional unem- characterises the regionalhational relationship, was
ployment differentials have sometimes been confused estimated as described in Annex 3.C to this chapter.
by reliance on ratio or difference measures of A typical value of this parameter was chosen, and
d i f f e r e n t i a l s 1 5 b csmparing periods over which used to construct a measure of regional differentials
the overall unemployment rate has increased, ratio that is, approximately, unbiased when making com- \
and level-based measures will typically give different parisons across years that have diffe7ent overall levels
impressions about the direction of change in differ- of unemployment. This measure is shown in
entials [Devens (1988)], as can be seen from the sec- Table 3.3, which focuses on the early 1970s, the late
ond set of graphs in Chart 3.2. Table 3.2 shows that 1970s, and the mid-1980s. For Japan, it indicates that
while in 1987 the top quarter unemployment rate was changes in regional differentials have been slight16.
typically about twice the bottom rate, in the 1960s it In Australia, Canada, Germany since the early 1970s,
was typically three times higher. Yet at the same time and in Sweden since 1976, there have been consid-
in terms of percentage points, unemployment differ- erable increases. The remaining 5 countries - Fin-
^enCes in a num6efof countries are Kow much greater land, France, Italy, the United Kingdom and the
than in the 1960s. United States - all show a U-shaped pattern in G
As a normative or policy indicator, trends in ratios regional differentials. For them, regional unemploy-
and in percentage point differences may both be rel- ment differentials fell on average by about 10 per cent
evant. For given regional labour forces, only differ- in the late 1970s, but by the mid-2980s had risen again
ences in percentage
_. points of unemployment indicate by about 20 per cent. Taking both the ratio and the
-the number of new jobs or labour force movements differences measures into account, this U-shaped pat-
that would be needed to even out uncmployment tern can be secn in Chart 3.2. The ratio measure
rates between regions. Yet the labour market tensions alone shows this pattern for Australia and Germany.
~..

Table 3 2. Regional unemployment indicators: time-series data

~

United United
Australia Canada Finland France Germany Italy Japan' Kingdom States

Number of regions 8 10 11 22 8 19 46 24 11 51

Nationwide
unemployment,
per cent, 1987 7.8 8.9 5.1 10.5 8.5b 12.0 3.4' 1.9 11.9 6.2

Ratio of top to bottom

quarter unemployment
rate
1960 4.38 2.16 3.14 3.57 1.66
1970 1.89 1.67 4.0Sd 2.92 3.12 2.36 2.56 1.86
1980 1.26 2.02 2.74 1.62 1.89 2.68 1.96 2.47 2.30 1.70
1987 1.57 2.00 3.37 1.50 2.W 2.95 2.00' 2.65 1.92 2.03

a) Population census.
b) 1986.
c) 1985.
d) 1971.
Source: See Annex 3.B.

107
 Table 3.3. Developments in regional unemployment differentials after 1970"
Unitcd Unitcd
Australia Canada Finland France Germany. Italy Japan Kingdom States

4-year averages
1971-74 0.41 1.oo 1.99 1.04* 0.77 2.09 0.83b 1.57 1.23
1976-79 0.49 1.54 1.77 0.96 0.84 1.92 0.92 1.04 1.37 1.13
1984-87 0.80 1.49 2.40 1.06 I.5lC 2.19 0.95 1.36 1.65 1.36

u) Values shown are (U,*-U,*)/D, as defined in Annex 3.C,evaluated at B = 0.4.

b) 1974 only.
c) 1984-1986 only.
Source: See Annex 3.B.

2. Turnover in the relative positions of regions cant changes occurred between 1965 and 1970. In the
mid-1960s in Italy, many of the regions of the South
In principle , the overall level of differentials and Islands had measured unemployment rates below
between regions could remain constant while the rel- the national average : but after 1970, there was rever-
ative positions of individual regions changed consid- sion to a pattern that had already existed to some
erably. To assess such relative changes, Table 3.4 extFrit-in- 1960; -with unemployment rates in all the
shows how regional unemployment rates correlate most southern regions and the Islands continually
through time with their 1975 and 1987 patterns. More above the national average. In the United Kingdom,
detailed estimates of patterns of change are shown, between 1965 and 1970, unemployment rates in the
for selected countries, in Annex 3.D. industrial areas of the West and East Midlands, and
Since 1970, the regional pattern of unemployment Yorkshire and Humberside, which had been relatively
rates has been stable in four of the countries exam- very low, drew closer to the national average, and
ined, Finland, Italy, Japan, and the United Kingdom rates in Scotland and Northern Ireland, which had
(correlation coefficients with 1987always above 0.85). been relatively very high, declined. However, little
However, in Italy and the United Kingdom, signifi- change in the relative pattern occurred after 1975.

Table 3.4. Correlation between recent and historical patterns of unemployment rates by regionR)\2)!
~

Australia Canada Finland Franceb Germany Italy Japanb Sweden United United
Kingdom States

Correlations with 1975'

1960 0.92 0.48 0.81 0.86 0.80 0.53
1965 0.30" 0.89" 0.76 0.34 0.92 0.82 0.32
1970 0.26 0.68 U.94' 0.89 0.91 0.95 0.93 0.90 0.49
1975 1 .oo 1 .oo 1 .oo 1 .oo 1.00 1.oo 1.00 1.00 1 .oo 1.OO
1980 0.07 0.86 0.92 0.53 0.71 0.94 0.97 0.85 0.97 0.52

Correlations with 1987f

1960 0.31 0.67 0.74 0.34
1965 0.61 0.87 0.21 0.19 0.75 0.19
1970 0.74 0.82 0.93 0.40 0.89 0.86 0.29
1975 0.07 0.67 0.88 0.46 0.54 0.87 0.91 0.689 0.92 -0.21
1980 0.30 0.52 0.94 0.93 0.92 0.91 0.95 0.82 0.98 0.32
1985 0.95 0.98 0.99 0.98 0.99 0.90 0.95 0.91 0.98 0.83
1987 1.oo 1.00 1.oo 1.m 1.oo 1.00 1.00 1.oo 1.00

a) All correlation coefficients are calculated using 1980 labour force as weights.
b) For France and Japan, correlations with 1975 refer to census dara. correlations with 1987 fo the Labour Force Survey. In France correlations with 1975 relate to
1962, 1968, 1975 and 1982.
c) For Swe?,.cncorrelations with 1976.
d) Correlation of 1966 with 1975.
e) Correlation of 1971 with 1975.
f) For Germany correlations with 1986.
g) Correlation of 1976 with 1987.
Source: See Annex 3.B.

I08
 Canada, France, Germany and Sweden have expe- place on an arca-wide rather than individual state
rienced greater changes. In Canada, sincc 1966, when basis : in terms of the nine census divisions used for
the relevant data start, the four small Eastern prov- statistical purposes, the average rate in New England
inces and Quebec have had rclativcly high uncm- (Connecticut, Mainc, Massachusetts, New Hamp-
ployment, whilc Manitoba and Saskatchewan have shire, Rhode Isiand and Vermont) fell from the
had low unemployment. In other provinces change nation’s highest rate, 10.2 pcr cent, in 1975 to the
has bccn more markcd : between 1980 and 1987,when nation’s lowcst rate, 3.4 per cent, in 1987. After
the national rate rose from 7.5 to 8.9 per cent, rates studying a range of factors conventionally used to
in Albcrta and British Columbia rose rapidly - from explain regional changes, Rones (1986, p. 11) con-
3.8 to 9.7 and from 6.8 to 12.0 per cent respectively - cluded that “the key ingredient in the economic tur-
while the rate in Ontario (which contains over a third naround in New England probably has bccn timc”.
of the Canadian labour force) fell from 6.8 to 6.1 per What this primarily meant was that the region’s “tra-
cent. In France, since 1975, the unemployment rate ditional” manufacturing industries, such as shoes, tex-
of Ile-de-France and the Mediterranean regions has tiles and textiles machinery, whose decline had in the
fallen by 20 to 30 per cent relative to the national early seventies been a cause of high unemployment,
average, while regions to the north, west and east of had by the end of the decade become sufficiently
the Ile-de-France suffered a 20 or 30 per cent relative small to allow - given the concentration of skills in
rise. In Germany, changes have been slow but, in the the region -sophisticated financial services and high-
long run, substantial. While Hessen and Baden-Wiirt- tech industries to come to the fore. At the same time
temberg have maintained relatively low unemploy- the West South Central region, which includes Texas,
ment rates and Rheinland-Pfalz with Saarland a rel- went from a next-to-lowest rate of 6.4 per cent up to
atively high rate, since 1965 the rate in Nordrhein- 8.8 per cent [see also Shank (1985)l. Already in 1975,
Westfalen (the most populous of the regions) has the pattern of relative unemployment rates differed
_ -
risen from about 0.8 to 1.2 times the national rate,’ from that of the 1960s : in fact, the changes between
while the rate in Bayern (the second largest region, 1975 and 1987 went some way towards restoring the
in thc south) fcll from about 1.8 to 0.8 times the pattcrn of 1960. California, though prosperous, had
national rate. In Sweden, for which the relevant data a relatively high unemployment rate through the
start in 1976, thc most populous regions, the city- 1960s, and morc recently the rate has come into line
based regions of Stockholm and Malm6, both had with the national average. In general, Annex 3.D
unemployment rates below the national average at suggests that regional patterns in the United States
the beginning of the pcriod. By 1987, the rate in have been subject to continual change, but with the
Stockholm had fallen much further - to 1 per cent, rate of change being higher in the 1970s and 1980s
against a national average of 1.9 - while the MalmB than in the 1960s.
rate had risen to 2.5 per cent: and a number of
smaller-population regions have experienced quite 3. Dependency of regions on the national economy
large changes, though in general all the forest-dom-
inated counties in the northern part of the country A different way of looking at changes in the regional
have had a continually relatively high rate. pattern of unemployment is to compare movements
Thc two rcmaining countries, Australia and the in regional rates with those in the national rate. As
United States, experienced the greatest changes in an overall indicator of how far the regional unem-
regional labour market patterns. Tn Australia, overall ploymcnt rates track thc national rates, Table 3.5
regional differentials have remained small and their compares the variances of l-year and 5-year changes
pattern has changed several times, as shown in in unemploymcnt ratcs at the national and regional
Annex 3.D. Some examples of the changes are: level. In the simple case where the movements in
between 1976 and 1979, while the national unem- unemployment are the same in each region, these
ployment rate rose from 4.7 to 5.8 per cent, rates in variances will be equal. Any deviation from this pat-
Western and Southern Australia rose more than tern, with movements differing between regions, will
3 percentage points, and between 1981 and 1985, be reflected in regional variances that exceed the
unemployment in New South Wales had risen from national variance. An example of this is that if vari-
4.8 to 8.8 per cent, while in Victoria: it had risen from ances are the same in each region, the ratio of national
5.5 to only 6.2 per cent. to regional variance will reflect the correlation coef-
In the United States, the regional pattern of unem- ficient between unemployment in one region and the
ployment changed completely between 1975and 1987. nextl7.
Often, patterns of change were similar for closely Table 3.5 shows that empirically18 the ratio of
neighbouring states, so that much of the change took national variance to regional variance is typically
 Table 3.5. Variance of changes through time in unemployment rates at the national and regional level"
United United
Australia Canada Finland Pranceh Gcrmany Italy Japanb Sweden Kingdom States

Variance of 1-year
changes in
unemployment rate
National 0.93 0.99 0.78 0.16 0.67 0.28 0.03 0.17 0.89 1.10
Regional 1.17 1.29 1.12 0.20 0.71 0.8s 0.05 0.31 0.98 1.55

Ratio of national
to regional variance
of 1-year changes 0.80 0.77 0.70 0.78 0.94 0.33 0.55 0.56 0.91 0.71

Variance of 5-year
changes in
unemployment rate
National 1.67 1.97 4.25 4.13 1.13 0.45 5.79 3.72
Regional 2.33 3.66 4.73 4.42 2.64 0.61 6.43 5.71

Ratio of national
to regional variance
of 5-year changes 0.72 0.54 0.90 0.93 0.43 0.72 0.90 0.65

a) Table entries are variances (mean-squared changes in per cent point unemptoyment rates). Regional variances are calculated in each region and then averaged using
1980 labour force weights.
b) For France and Japan, 1-year changes are based on labour force survey data 1974-1987. and 5-year changes (for France, 7 and 8-year) are based on census data.
Source: See Annex 3.B.

about 0.7019. Thus, this statistical measure confirms population densities may also be important. The
that, even though in the short and medium run centre-periphery pattern makes itself felt more in
changes in the national economic situation are a dom- regions within national economies, as shown by Swe-
inant component in the moverncnts at the level of den, Norway, Finland, Italy, Spain and the United
individual regions, the regional component is about Kingdom, than at a supra-national level : for national
30 per cent and hence still quantitatively important. economies, sharing a common government and lan-
guage, the capital or other large cities seem to become
centre(s) of gravity which so far have proved stronger
than the pull of core regions outside national bound-
aries. The absence of the centre-periphery pattcrn in
E. SUMMARY AND CONCLUSIONS the United States (and perhaps Australia also) prob-
I.' ably reflects both natural geographical factors (size,
and thc presence of interior deserts), which have
i
Within a typical large QECD country in the 1980s, favoured devclopment of population centrcs at points
one quarter of the labour force, in the highest uncm- on the coast, the culture of population mobility, a
ployment regions, has an unemployment rate about common language, and a relatively prosperous agri-
1.4 times the national average, and another quarter cultural community reflecting the great natural
in the lowest unemployment regions has a rate about resources and a history of settler entcrprise,
1, 0.7 times the national average. Although the national Economics, as mcntioncd in Section B, provides a
\ unemployment rate is a dominating factor in the variety of perspectives on the nature and significance
changes of the unemployment rate for each individual of regional unemployment differentials. Industrial
region, about 30 per cent of the regional movements structure, the characteristics of a region's trading sec-
cannot be explained at the national level. tor, and other determinants of demand ; natural pop-
Chart 3.1 shows, for Europe and Japan, a tendency ulation increase, migration, and participation rate
for unemployment rates to be high in peripheral areas. changes affecting the labour force ; wage determi-
The decline in agriculture, even if it is in some cases nation and unemployment compensation systems ;
approaching an end, may be a factor, perhaps because and the balance between economies of scale and
populations in rural areas have an unbalanced age congestion may all be important. Though it does not
structure : the economies of scale that arise at higher study many of these factors directly, this chapter has

110
 found evidence that periphcral areas in Europe and Nevertheless, the variety of cxpcricnce is note-
Japan suffer from their geographical isolation, sug- worthy : in some countries, unemployment experi-
gesting that economies of scale acting to the benefit ence seems dominated by structural factors of extreme
of the more central regions can be an important factor. persistence, while in the United States and to some
The long-term persistence of regional differences in cxtent other countries, structural factors such as
many countries shows that theoretical equilibrating industry composition or centre/periphery distinctions
mechanisms in practicc have little explanatory power, either have changed rapidly, or would appear not to
with factors limiting the operation and speediness of be important determining factors.
adjustment often being more important. However, In the United States, regional patterns have
the greater flexibility in some other countries suggests reversed since 1975 - the regions that had relatively
that successful efforts at idcntifying and removing the high unemployment in 1975 now have low rates, often
barriers to regional adjustment could bring about coupled with high rates of unfilled vacancies. One
large changes. general factor promoting change here may be that
As regards demand and supply factors, regional states - and cities and counties within states -have
income multiplier effects seem sometimes to have much greater independence in matters of economic
been important, for example in aggravating the prob- policy - including the capacity to determine local
lems of the Texan economy, in the United States, taxation rates (on sales, personal and corporate
and of Alberta, in Canada, through oil price falls income) and, within limits, to determine unemploy-
during the 1980s. In Italy, the pace of migration during ment and accident compensation levels - than do
the 1960s seems to have been an important factor regional administrations in most European countries.
limiting the unemployment differential of the South- More generally, flexibility of capital and labour seems
ern regions : the subsequent slowing of migration to have been large enough to solve problems of
flows due to recession and (except within the EEC) regional mismatch and play a major role in overall
tighter immigration controls may be a major factor job creation. This is despite the fact that since 1979,
aggravating regional unemployment problems in Italy regional differentials in unemployment have -7 any
and elsewhere since then. measure got larger, which perhaps servcs as a ition
t

The study of trends in dispersion in regional unem- against the idea that inequality statistics alone n be
ployment rates since the 196& has suggested that, used to measure mismatch. Flexibility in thc L ..itcd
c -while increases in the overall unemployment rate have
~
States has involved widespread change across the
often madc interpretation somewhat complicatcd, whole economy, and not just change specifically
regional differentials rose during the 1980s in the involving high-unemployment areas.
majority of countries considered. In Finland, France, In Japan and some European countries, the ranking
Italy, the United Kingdom and the United States, the of regions by unemployment rate has changed little
rise followed a certain fall in diffcrcntials during thc since 1960. This suggests that improving the fortunes
1970s. In some countries the balance of taxeb and of now-depressed regions in these countries will be
subsidies may have followed a pattern -with difficult, but not that this can never occur. In the
increascd progressiveness of the personal income tax United Kingdom and Italy, change was rather greater
during the late 1970s, and reductions in regional sub- during the 1960s than in the 1970s and 1980s, when
sidies during the 1980s - that contributed to this unemployment was globally higher; in the United
result. States the reverse was the case.
Changes in the fortunes of regions, as measured by A number of the topics raised here could usefully
the changes in regional unemployment patterns, have be studied in future research along similar lines.
been substantial in several cases. Changes in the Among these are the relative roles of employment
United States since 1975 have been very large, and creation, population change, and participation rate
changes in Australia, Canada and some European change in labour market balance at the regional level ;
countries (such as France and Germany) have been whether there are correlations between regional
significant for many parts of the country. However, unemployment levels and employment trends over
in Japan and some European countries - especially particular periods, and the industry structure of
Finland, Italy, and the United Kingdom - changes employment ; what has happened to migration flows
large enough to upset long-established patterns of during the 1980s ;whether wages at the regional level
depressed and favoured regions have been rare. This are more or less equalised compared with the past ;
does not mean that regional unemployment rates have to what extent wage levels and employment perform-
been wholly rigid : when extremes differ by a factor ance are correlated ; and whether relative income
of several times, particular regions may change by 20 levels correlate closely with regional unemployment
or 30 per cent without much affecting the pattern. rates, or indicate similar periods of increase and

111
decrease in regional differentials. It is hoped that demand, policies aimed at a less unequal distribution
some of thcse issues will be addressed in future issues of unemployment between regions may play a major
of the Employment, Outlook. part in policies to improve overall national perform-
Substantial changes in thc performance of partic- ance. The analysis in terms of large regions here shows
ular regions havc occurred in many countries : U.S. only part of the problem, for the town and country,
experience shows that regional change can be a vital or regions only a few tens of kilometres apart, often
component in an economy that is creating jobs. form largely separate labour markets. In the face of
Rones (1986, p. 13) comments : “the evidence sug- increasing fears of labour shortages, sometimes even
gests that regional advantage is often short-lived’ ; for unqualifiedandsemi-qualified workers, in strong ’
the opposite of what most European observers have areas of OECD countries, policies aimed at encour-
concluded. Insofar as differences in the location of aging mobility of workers and movement of economic
available workers and vacancies are an important activities towards the high-unemployment regions
component of “mismatch” between labour supply and may become increasingly important.

NOTES

1. This chapter draws upon reports by Brigitte Autrand in the United Kingdom. A comparison of the current
concerning regional statisticsfor OECD countries, and recovery with that of 1976-79 suggcsts a similar con-
Ellen Carlson conccrning literature on regional labour clusion for several other OECD countries [OECD
markets. (1988b)J.Results herc, by contrast, do find indications
2. More sophisticated models may allow for lags between that regional unemployment differentials are higher
unemployment in different regions, for time-varying now than in the late 1970s.
parameters, errors that are autoregressive across space 7. Empirical estimates of the effect of regional dispersion
as well as time, and use such techniques as estimation in unempIoyment rates on the national Phillips curve
in the frequency domain [Tiller and Bednarzik (1983)] seem, however, inconclusive as to the significance of
or principal components and factor analysis [BarteIs this effect [Tiller and Bednarzik (1983) ;Azevedo and
(1977) ; Pedersen (1978)l. O’Connell (1980)l.
3. A shift-share analysis reported in Commission of the 8. Units used in regional science, such as urban areas,
European Communities (1981) found that, except for journey-to-work or local labour market areas, and
a few extreme cases, a concentration of output in economic regions (which group together areas with
growing or declining industries (in 1970) could account similar economic characteristics), are appropriate as
for differences of only few per cent in growth (between tools for targeted studies, but would present certain
1970 and 1977) of regions within countries. Actual disadvantages in a wider statistical context. Regions
output growth quite often differed by 20 per cent or defined in economic terms will be differently delimited
more across regions within a country. according to the variable(s) under consideration ; a
given travel-to-work area may contain rich and poor
4. Greenwood (1985) reviews rccent research on the city suburbs which differ sharply in terms of economic
influence of demographic characteristics on the deci- and social characteristics ; and definitions of urban
sion to migrate. Borjas (1987) analyses the conditions centre and suburban areas typically change between
under which positive or negative selection among censuses, making study of changes through time dif-
immigrants is generated. ficult if not impossible. Preference hcre has been given
5. The classic example of congestion externalities ariscs to using a fixed geographic delimitation of the regions.
when the average time it takes for a commuter from 9. Among the features in the NUTS system, are the
the suburbs to entcr a city is an increasing function treatment of Ireland as a single region in both the
T(n) of the number of commuters n. The total time level 1 and level 2 classifications, and inconsistencies
spent commuting is n.T(n), and the social cost of an in the treatment of large cities (which are sometimes
extra commuter journey, T(n)+n.dT(n)/dn, exceeds regions in themselves and sometimes not). While gen-
the private cost T(n). erally NUTS level 1regions are aggregationsof level 2
6. Jackman and Roper (1987) find little evidence of an units, this is not so for Belgium. Classification is gen-
increase after 1979 in the degree of regional mismatch erally based on administrative units ;the United King-

112
 dom is an exception in that the 11 regions defined for from the census was in 1986 almost 20 per cent above
statistical purposes in the 1960s and used by EURO- 1the survey-hased rate, the wcighted correlation coef-
STAT in the NUTS 1 classification cut across the Ificient betwecn the regional rates from thc two sources
country’s administrative structure. was 0.98.
10. Most of the regions chosen for Chart 3.1 are contig- 14. Some tests were conducted to examine the impact of
uous with several others and, if they contain a large the lcvcl of regional aggregation. When the S1 States
city, their boundaries extend well beyond it. Some of the Lhited States were aggregated into 9 larger
effort was made to standardize on this type of dis- regions, estimated differentials between the unem-
position of regions. Situations in France and Spain ployment rates of the top and bottom quarter of the
where one region was wholly encircled by another in labour force fcll by between a quarter and a half, and
the E E C NUTS 1classification were avoided by split- in Sweden when 24 counties were aggregated into 8
ting the encircling region. However, in Australia, the larger regions, estimated differentials fell by 10 to
Australian Capital Territory was left encircled by the 15 per cent. Larger areas may experience fewer
State of New South Wales, while West Berlin and extremes, owing to the wider range of economic activ-
island regions of course have no direct contact with ities carried out in them [Evers-Koelman et al. (1987)]
other regions. __
and to scale economies in the search activities of
11. Sampling variability in the regional unemployment available workers.
rate estimates can also be a problem for regions with 15. Steinle (1983) concluded that unemployment dispar-
small samples and low unemployment rates, especially ities within the E E C had declined between 1970 and
where there is only one labour force survey per year. 1979 because percentage increases in total unemploy-
12. Note b) to Table 3.1 defines the unemployment rates ment was generally larger in regions with the lowest
of the “top and bottom quarter”, U T and UB. The unemployment rates. Similarly, Evers-Koelman et al.
unemployment rates of the “middle half”, UM,is sim- (1987) base their description of the evolution of
ilarly defined : (UJ4+UM/2+UB/4) is exactly equal to regional unemployment disparities on the minimum-
the national unemployment rate. maximum ratio of unemployment rates. Other studies
have instead preferred measures of differentials based
13, In the cases of France, Japan and the Unites Statcs, on standard deviations [Lever (1987)l. Studies focus-
both population census and labour force survey data ing on the short-term response of regional unemploy-
have been used to provide a picture of the pattern of ment rates to national fluctuations have produced
regional unemployment. There are differences in the opposing conclusions about the cyclicality of regional
nature of the two data sources. First, a population dispersions according to the specification used. Using
census is carried out over a short period within the “level” specifications [for example Thirlwall (1966) ;
year, while some (but not all) labour Iorce surveys King and Clark (1978) ; Fearn (1975) ; and Gordon
provide monthly o r quarterly averages for the year. (lYXS)] swings in the dispersion of unemployment are
Second, the small sample size of household surveys, found to be positively correlated with thc national
especially in earlier years, causes sampling variations rate, while models using a “logarithmic” specification
in estimated unemployment rates. Third, definitions [e,g. Brechling [1967)] suggest the opposite.
of various labour market groups may differ in that
more general questions are posed in a census (in the 16. The popdation census statistics for Japan, not shown
United States the questions are the same but in France in Tahle 3.3, suggest a rather larger increase in dif-
they are n o t : see, for the French case, INSEE). ferentials in 1985 compared with earlier years than is
Fourth, differences in reporting method are thought suggested by the labour force survey.
to influence the numbers rccorded. Thus, significant
differences may arise in the level of the unemployment 17, The unemployment rate u; in any region i may be
rate from the two sources. In the case of France and modelled as the sum of a tixed component ai and an
Japan the national unemployment rates from the latest error term ei
census were 15 and 30 per cent respectively above the ui = ai ei+
corresponding rate from the regular labour force sur- In the simple additive model, the error terms ei have
vey ; in the case of the United States the national rate the same variance Var(e), and are correlated with the
from the latest census was 8 per cent below that from same correlation coefficient, p, between any one
the labour force survey. The regional pattern of unem- region and any other. The aggregate unemployment
ployment rates from the two sources seems, however, rate is then
to be quite similar. The correlation coefficients
between the regional unemployment rates from the
+
U = X wiul = X wiai I: wiei,
latest census and the corresponding labour force sur- where wi is the labour force weight of each region; its
vey, weighted by the labour force level, were 0.94, expected value is
0.88 and 0.95 for France, Japan and the United States E(U) = X wiai
respectively. A similar pattern emerges for Australia ;
and its variance about this expected value is
according to data in Department of Employment,
Education and Training (1989), while the national rate Var(U) = E[(X wiui)*] = [ p + (1 - p) X w?] Var(e)

113
 When there are many small regions, X -2 is close to biased upwards through the fact that the large regions
zero and the above expression reduces to have a significant weight in the national rate). The
Var(U) = p Var(e) full formula €or Var(U) given in the previous note
so that the ratio of national to regional variance indicates that the extent of this bias is measured by
Var(U)/Var(e) is equal to the correlation coefficient I: w,2, where the value of 1 (which would arise if there
between any one region and any other. Other results were just one region in the statistics) implies complete
that may be derived are that the correlation coefficient bias. Actual values of Z w,2 in the data used for this
between any one regional rate and the national rate study were 0.24 for Australia and Canada, where two
is the square root of p ;that the expected coefficient, large regions contain over 60 per cent of the labour
if a regional rate is regressed on the national rate, is force, and 0.17 for Germany, where there are only
1; and that the R2 of this regression is again Var(U)/ 8 regions, but lower elsewhere, e.g. 0.04 in the United
Var(e) = p. These results are approximate, to the States. For another analysis of interdependence
extent that assumptions of this simple additive model between regional and national level behaviour, see
do not hold exactly. Johnston (1979).
18. When individual regions are large, it may be argued 19. Italy is an exceptional case, where the very highly
that measures of regional variance are biased down- trendedJ&gr&y of the national unemployment data
wards by the high degree of aggregation (or that the leads a low variance of year-to-year changcs at the
correlations between regional and national rates are national level.

114
 Annex 3.A

DEFINITION OF REGIONS AND

CONSTRUCTION OF THEIR UNEMPLOYMENT RATES
FOR CHART 3.1 AND TABLE 3.1

1. Definition of regional units definition, under which persons not registered but available
and looking for work are included among the unemployed.
The regions in Chart 3.1 of the main text are identified The employment data, which include the armed forces,
by codes in Chart 3.A.1, with the corresponding name of were corrected using reported persons working fewer than
each region given in Table 3.A.1. Where these regions are 13 hours per week, a group not classified as “employed”
aggregations of other widely-recognized statistical or polit- under national definitions.
ical regions, their components are shown in Table 3.A.2.
Can&

2. Construction of regional unemployment rates Statistics Canada (1987), Historical Labour Force Statis-
tics. Averages of monthly estimatcs.
The detailed country notes below list the sources of
regional unemployment data for Chart 3.1. The OECD’s EEC countries
Standardized Unemploylnel~tRates (SURs) refer to the Levels of unemployment and the civilian labour force in
total labour force (including armed forccs), and where April corresponding to published regional unemployment
regional labour force survey data rclatc to the civilian labour rates were supplied by EUROSTAT and used foT all EEC
force a multiplicative adjustment was applied to achieve countrics, the only exception being Spain. Thc EURO-
consistency, at the national total level, with the SWR. Inso- STAT estimates of unemployment arc based on national
far as armed forces employment is not proportionately totals from the April 1987 labour force surveys. For Greece,
distributed across regions, this adjustment involves a degree Italy and Portugal, the regional totals are also based on the
of error. Other adjustments which were carried out for labour force survey of the same month : for othcr countrics,
countries where SURs are not currently published, and in the national totals disaggregated by sex and age are allo-
respect of timing of the estimates for some EEC countries, cated to regions on the basis of the rcgional structure of
are described in the notes below. For countries where SURs registered unemployment [see EUROSTAT (1988) for fur-
are not currently published, the national ratcs presented in ther details]. In general, Secretariat adjustments were made
Table 3.1, being survey-based, do not always correspond to achieve consistency at the national total level with the
to those presented in Table 1.5. annual average and total labour force basis of the OECD
SURs. However, for Denmark, Ireland, Italy, Greece and
Luxembourg, OECD SUR totals were not available.
3. Detailed country notes
Adjustments to a total labour force and annual average
basis for Italy were based on ISTAT (1988), Rilevazione
Australia delle Forze di Lavoro, Media 1987. The adjustment in
ABS, The Labour Force, Table 4. Estimates are averages respect of armed forces was carried out, based on replies
of monthly data. to the Secretariat’s annual questionnaire on population and
labour force, for Denmark and Ireland: estimates for
Austria
Greece and Luxembourg were adjusted neither to an annual
average nor to a total labour force basis. Data for Spain
The data, based on the quarterly Mikrozensus, were based on the labour force survey, annual averages, were
directly supplied by the Osterreichisches Statistisches Zen- supplied by national authorities ; they have been adjusted
tralamt. Estimates are averages of quarterly data. The by the Secretariat to conform to the national totals used
unemployment series is based on a revised Mikrozensus for SURs.

115
 Chart 3-A.1
NAMES OF INDIVIDUAL REGIONS IN THE UECD AREAa

CENTRALWESTERN AND SOUTHERN EUROPE (EXCL. TURKEY] I NORDIC COUNTRIES'

-_
Scale 1 : 23 000 000

TURKEY I
I

l Scale 1:30000 000

a. codes within each region ihdicafe region names, 85 shown in Table 3.A.1.
APAN IORTH AMERICA

Scale 1:37 OoOooC

 Table 3.A.1. Area, population and unemployment rates for regions of OECD countries
Regional Population Unemployment
Regional name Numbcr
code Area (km')" (thousands)b rateC

Icetand ICl Iceland 1 102829 245 ..

Norway 323 877 4 159 2.1
NR1 North 2 112917 464 3.6
NR2 Middle 3 41 294 373 2.1
NR3 West 4 49 400 743 2.1
NR4 East 5 94657 2019 1.7
NR5 South 6 25 63 1 558 2.1

Sweden 410928 8 358 1.9

SWl Forest Counties 7 289401 1756 2.1
sw2 Middle 8 38 432 1420 2.0
sw3 Stockholm 9 6 487 1578 1.o
sw4 West 10 29 927 1653 1.6
sw5 South 11 46 680 1950 2.1

Finland 338 145 4911 5.0

FNl North 12 182 102 810 8.7
FN2 Central 13 88 292 1157 5.5
FN3 South 14 57 348 1755 5.2
FN4 Uudenmaan I5 10404 1188 2.3

Ireland 68 894 3 541 18.0

IRI North 16 19489 506 19.1
IR2 Central 17 16856 578 16.7
IR3 East 18 10981 1534 18.3
IR4 South 19 21 567 922 17.4

United Kingdom 244111 56618 10.2

UKI North 20 15 401 3 086 13.5
UK2 Yorkshire and Humberside 21 15 420 4 903 11.5
UK3 East Midlands 22 15630 3 897 9.2
UK4 East Anglia 23 12573 1965 7.4
UK5 South East 24 27 222 17 192 7.6
UK6 South West 25 23 850 4 501 8.0
UK7 West Midlands 26 13013 5 183 11.4
UK8 North West 27 7331 6 386 12.4
UK9 Wales 28 20768 2812 11.9
UKlO Scotland 29 78 783 5 137 13.6
UKI I Northern ireland 30 14 120 1558 17.4

Denmark 43 080 5 111 6.0

DK1 Capital Region 31 2 857 1723 4.8
DK2 East of Storebaelt .32 6 970 583 7.2
DK3 West of Storebaelt 33 33 253 2 805 6.5

Germany 248 706 61 049 6.2

GE1 Schleswig-Holstein 34 15 721 2614 7.1
GE2 Hamburg 35 755 1592 10.4
GE3 Niedersachsen 36 47447 7216 7.8
GE4 Bremen 37 404 666 11.4
GE5 Nordrhein-Westfalen 38 34 062 16 704 8.0
GE6 Hessen 39 21 115 5 535 4.7
GE7 Rheinland-Pfalz 40 19 848 3 624 5.6
GE8 Baden-Wiirttemberg 41 35751 9 241 3.5
GE9 Bayern 42 70553 10958 4.4
GElO Saarland 43 2 571 1051 10.0
GEll Berlin West 44 480 1849 7.9

Netherlands 41 160 14454 9.6

NLl North 45 10017 1588 11.4
NL2 East 46 11 247 2 908 9.8
NL3 West 47 11 822 6 756 9.2
NL4 South 48 8 073 3 199 9.4

118
 Table 3,A.l (Continued). Area, population and unemployment rates for regions of OECD countries
Regional Regional name Number
Population Unemployment
mde Area (km')' (thousands)b IUC=
~~ ~" . . _ . .... ..
Belgium 30518 9 858 11.1
BE 1 North 49 13512 5 670 9.3
BE2 South 50 16 844 3 208 14.2
BE3 Bruxelles 51 161 980 12.0

Luxembourg LXI Luxembourg 52 2 586 366 2.6

France 543 96s 55 062 10.5

FRl Nord-Pas-de-Calais 53 12414 3 934 14.2
FR2 Normandie-Picardie 54 49 305 4817 11.7
FR3 Ile-de-France 55 12012 10206 8.7
FR4 West 56 85 100 7 328 11.0
m5 Central 57 96339 5 266 10.5
FR6 East 58 48 029 5000 9.5
FR7 Auvergne-Limousin 59 42 955 2 072 10.0
FR8 South West 60 86 656 5 053 10.7
FR9 RhBne-Alpes 61 43 698 5 124 8.6
FRlO MCditerranCe 62 58 776 6018 13.1
FRI 1 Corse 63 8 680 247 12.2

Switzerland 41 293 6 456 ..

sz1 North '64 8 129 2 950 1.

sz2 West 65 6 805 1310 ..

sz3 Centre 66 11215 1531 ..
SZ4 South 67 15 142 666 ..
Austria 83 853 7 555 3.7
AS1 Danubiana 68 31 565 4 225 3.9
AS2 West 69 22 403 1334 2.8
AS3 South East 70 29 885 1997 4.1

1dY 301 278 57 080 11.2

IT1 North West 71 34077 6 304 8.9
IT2 Lombardia 72 23 857 8 885 6.4
IT3 North East 73 39 830 6 467 7.2
IT4 Emilia-Romagna 74 22 123 3 947 7.0
IT5 Central 75 41 141 5 820 8.1
IT6 Lazio 76 17203 5 080 10.3
IT7 Campania 77 13595 5 608 22.2
IT8 Abruzzi-Molise 78 15232 I577 10.4
IT9 Puglia 79 19348 3 978 14.5
IT10 South West 80 25 072 2 734 18.9
IT1 1 Sicilia 81 25 708 5051 16.8
IT12 Sardegna 82 24090 1629 19.6

Spain so4 790 38 423 20.1

SP 1 Galicia 83 29 434 2 836 12.5
SP2 Asturias-Cantabria 84 15863 1656 19.5
SP3 North West 85 22716 2 957 20.8
SP4 Aragon 86 47 650 1205 13.7
SP5 East 87 60249 10416 19.3
SP6 Castilla-Leon 88 94 193 2610 16.8
SP7 Madrid 89 7 995 4817 16.7
SP8 Castilla-la Mancha 90 79 230 1674 15.1
SP9 Extremadura 91 41 602 1083 26.3
SPlO South 92 98616 7 760 28.7
SP11 Canarias 93 7 242 1416 23.8

Portugal 91 750 10 129 7.0

PG1 North 94 21 194 3 540 4.8
PG2 Central 9s 23 270 1 777 5.4
PG3 Lisboa 96 13 194 3 403 9.6
PO4 South 97 31 051 894 11.9
PG5 Islands 98 3 041 5 16 4. L.

1 I9
 Table 3.A.1 (Continued). Area, population and unemployment rates for regions of OECD countries
Regional Population Unemployment
cdc Regional name Number Area(km’” (thousands)* rateC
Greece 131 990 9 872 7.4
GRl North 99 56684 3 176 6.4
GR2 Central la0 60429 5 856 8.3
GR3 South East 101 14877 840 4.6

Turkey 814578 50 959 10.1

TK1 West 102 117947 15 877 7.6
TK2 South 103 100 997 6 761 9.3
TK3 Central 104 236 160 1 1 946 10.2
TK4 North 105 76 550 5 903 12.1
TK5 East 106 282919 10471 13.4

Japan 377815 117060 2.8

JAl Hokkaido 107 83519 5 576 4.2
JA2 Tohoku 108 66911 9 572 2.8
JA3 South Kanto 109 I3515 28 697 2.7
JA4 North Kanto 110 36912 9087 1.8
JAS Hokuriku 111 25 220 5 467 2.4
JA6 Tokai 112 29285 13 316 2.0
JA7 Kinki 113 27 295 19 521 3.3
IA8 Chugoku 1 I4 31 785 7 586 2.8
JA9 Shikoku 115 16806 4 163 3.3
JAIO Kyuw 116 44408 14083 4.0

Australia 7 682 300 15 758 8.1

ALl New South Wales 117 801 600 5 477 8.6
AL2 Victoria 118 227 600 4 122 6.2
At3 Queensland 119 1727 2M) 2 548 9.7
AL4 South Australia 120 984000 1363 8.9
AL5 Western Australia I21 2525 500 1408 1.6
AL6 Tasmania 122 67 800 442 9.3
AL7 Northern Territory 123 1346200 144 8.0
AL8 Australian Capital Territory 124 2400 253 5.7

New Zealand 267515 3 291 4.0

NZI North Island 125 114771 2 439 3.7
NZ2 South Island 126 152744 852 4.9

Canada 99706iO 24 343 8.8

CA1 Newfoundland 127 405 720 568 18.3
CA2 Prince Edward Island 128 5 660 123 13.0
CA3 Nova Scotia 129 55 490 847 12.5
CA4 New Brunswick 130 73 440 696 13.1
CA5 Quebec 131 1540680 6 438 10.2
CA6 Ontario 132 1068 580 8 625 6.1
CA7 Manitoba 133 649 950 1026 7.4
CA8 Saskatchewan 134 652 330 968 7.2
CA9 Alberta 135 6661 190 2238 9.6
CGlO British Columbia 136 947 800 2 744 11.9

United States 9 372 614 236 158 6.1

us 1 Alabama 137 133915 3990 7.7
us2 Alaska 138 1530700 500 10.7
us3 Arizona 139 295 260 3 053 6.2
US4 Arkansas 140 137 754 2 349 8.0
us5 California 141 411049 25 622 5.7
US6 Colorado 142 269 595 3 178 7.5
us7 Connecticut 143 12997 3 154 3.3
US8 Delaware 144 5 295 613 3.0
us9 District of Columbia 145 178 623 6.2
us10 Florida 146 151 939 10976 5.2
us11 Georgia 147 152576 5 637 5.4
us12 Hawaii 148 I6 759 1039 3.8
US13 Idaho 149 216 432 1001 7.9
us14 Illinois 150
~. I45 934 I 1 511 7.2

120
 Table 3.A. 1 (Continued). Area, population and unemployment rates for regions of OECD countries
Regional Regional name Number
Population Unemployment
eodc Area (km')" (thousands)b ratec

United States US15 Indiana 151 93 720 5 498 6.3

(continued) US16 Iowa 152 145 753 2910 5.4
US17 Kansas 153 213098 2438 4.8
US18 Kentucky 154 104660 3 723 8.7
US19 Louisiana 155 123677 4462 11.8
us20 Maine 156 86 156 1156 4.4
us21 Maryland 157 27 092 4 349 4.2
us22 Massachusetts 158 21 456 5 198 3.2
US23 Michigan 159 151 586 9 075 8.0
us24 Minnesota 160 218601 4 162 5.3
US25 Mississippi 161 123515 2 598 10.0
US26 Missouri 162 180516 5 008 6.2
US27 Montana 163 380 848 828 7.3
US28 Nebraska 164 200 350 1606 4.9
US29 Nevada 165 286 352 911 6.2
US30 New Hampshire 166 24 032 977 2.5
US31 New Jersey 167 20 169 7715 4.0
US32 New Mexico 168 3 14 925 1424 8.8
us33 New York 169 127 189 17 735 4.8
us34 North Carolina 170 136413 3 300 4.4
us35 North Dakota 171 183 119 686 5.1
US36 Ohio 172 107044 10752 6.9
us37 Oklahoma 173 181 186 3 298 7.2
US38 Oregon 174 251 419 2 674 6.1
us39 Pennsylvania 175 117348 11901 5.6
US40 Rhode Island 176 3 140 962 3.8
US41 South Carolina 177 80 582 3 300 5.5
US42 South Dakota 178 L 99 130 706 4.2
us43 Tennessee 179 109 152 4717 6.5
us44 Texas 180 691 030 15 989 8.3
us45 Utah 181 219889 1652 6.2
US46 Vermont 182 24900 530 3.7
us47 Virginia 183 105 586 5 636 4.2
US48 Washington 184 176479 4 349 7.5
us49 West Virginia 185 62 759 1952 10.7
US50 Wisconsin 186 145 436 4 766 6.0
us51 Wyoming 187 253 326 511 8.6
a) Areas reported include lakcs in some countries and exdude them in others: they are usually from statistical yearbooks.
b) Population in 1985 or closest available war.
r) Annual average 1987.

121
 Table 3.A.2. Definition of regions used for Chart 3.1
Regions Number hdministralivc units (national namcs)

Norway 5 [Groups of Counties]

North Finnrnark, Nordland, Trorns
Middle Nord-Trbndelag, Slpr-Trgndelag
West Hordaland, Mbre og Romsdal, Sogn og Fjordane
East Akershus, Buskerud, Hedmark, Oppland, Oslo, Telemark,
Vestfold, @stfold
South Aust-Agder, Rogaland, Vest-Agder

Sweden 5 [Groups of Counties]

Forest Counties Gavleborgs, Varmlands, Kopparbergs, Vasternorrlands,
Jhtlands, Vasterbattens, Norrbottens
Middle Uppsala, Sodermanlands, Ostergotlands, Orebro,
VLtmanlands
Stockholm Stockholms
West Hallands, Goteborgs och Bohus, Alvsborgs, Skaraborgs
South Jonkopings, Kronobergs, Kalmar, Gotlands, Blekinge,
Kristianstads, Malmohus

Finland 4 [Groups of Provinces]

North Lapin, Oulun, Pohjois-Karjalan
Central Mikkelin, Kuopion, Keski-Suomen, Vaasan
South Turun ja Porin, h a n d , Hameen, Kymen
Uudenmaan Uudenmaan

Ireland 4 (Groups of Planning Regions]

North Donegal, North West, West
Central Midlands, Midwest
East North East, East
South South West, South East

United Kingdom 11 [Standards Regions]

Denmark 3 [Groups of Amter]

Capital Region Hovedstadsregionen
East of Storebaelt @st for Storebaelt
West of Storebaelt Vest for Storebaelt

Germany 11 [Lander]

Netherlands 4 [Groups of Provinces]

North Groningen, Friesland. Drenthe
East Overijssel, Gelderland
West Utrecht, Noord-Holland. Zuid-Holland, Zeeland
South Noord-Brabant, Limburg

Belgium 3 [Regions]
North Vlaams Gewest
South Rkgion wallone
Bruxelles Bruxelles

Luxembourg 1 (Whole Country]

France 11 [Groups of Regions]

Nord-Pas-de-Calais Nord-Pas-de-Calais
Normandie-Picardie Normandie, Picardie
Ile-de-France Ile-de-France
West Bretagne, Pays de la Loire, Poitou-Charentes
Central Centre, Bourgogne, Champagne-Ardennes
East Lorraine, Alsace, Franche-ComtC
Auvergne-Limousin Auvergne, Limousin
South West Aquitaine, Midi-PyrCnCes
RhBne-Alpes R hBne-AIpes
MCditerran6e Languedoc-Roussillon, Provence-Alpes-CBte d’Azur
Corse Corse

122
 Table 3.A.2 (Continued). Definition of regions used for Chart 3.1
Regions Number Administrative units (national names)

Switzerland 4 [Groups of Cantons]

North Aargau, Appenzell A.Rh., Appenzell I.Rh., Basel-Stadt,
Basel-Land, St .Gallen, Schaffhausen, Zurich, Solothurn,
Thurgau
West Fribourg, Gentve. Jura, Neuchltel, Vaud
Centre Schwyz, Uri, Zug, Bern, Luzern, Obwalden, Nidwalden,
Glarus
South Ticino, Valais, Graubiinden

Austria 3 [Groups of Bundeslander]

Danubiana Niederosterreich, Oberosterreich. Wien
West Salzburg, Tirol, Vorarlberg
South East Burgenland, Karnten, Steiermark

Italy 12 [Groups of Regions]

North West Piemonte, Val d’Aosta, Liguria
Lombardia Lombardia
North East Trentino-Alto Adige, Veneto, Friuli-Venezia Giulia
Emilia-Romagna Emilia-Romagna
Central Toscana, Umbria, Marche
Lazio LAO
Campania Campania
Abruzzi-Molise Abruui, Molise
Puglia Puglia
South West Basilicata, Calabria
Sicilia Sicilia
Sardegna Sardegna
//. \

Spain 11 [Groups of Autonomous Communities) 1‘

Galicia Galicia \
Asturias-Cantabria Asturias, Cantabria
North West Pais Vasco, Navarra, Rioja
Aragon Aragon
East CataluEa, Baleares, Comunidad Valenciana
Castilla-Leon Castilla-Leon
Madrid Madrid
Castilla-la Alancha Castilla-la Mancha
Extremadura Extremadura
South Murcia, Andalucia
Canarias Canarias
Portugal 5 [Groups of Regional Coordination Committees
and Autonomous Regions]
North Norte
Central Centro
Lisboa Lisboa e Vale to Tejo
South Alemejo, Algarve
Islands AFores, Madeira
Greece 3 [Groups of Development Regions]
North Voreia Ellada
Central Kentriki Ellada
South East Anatolika Kai Notia Nisia

Turkey 5 [Groups of Provinces]

West Aydin, Balikesir, Bursa, Canakkale, Denizli, Edirne,
Istanbul, lzmir, Kirklareli, Kocaeli, Manisa, Sakarya,
Tekirdag
South Adana, Antalya, Burdur, Gaziantep, Hatay, Isparta,
Iqel, Mugla
Central Afyon, Ankara, Bilecik, Bolu, Cankiri, Corum, Eskisehir,
Kayseri, Kirsehir, Konya, Kutahya, Nevsehir, Nigde, Tokat,
Usak, Yozgat
North Amasya, Artvin, Giresun, Kastamonu, Ordu, Rize Samsun,
Sinop, Trabzon, Zonguldak
East Adiyaman, Agri, Bingo], Bitlis, Diyarbakir, Elazig,
Erzincan, Erzurum, Gumushane, Hakkari, Kars, Malatya,
K. Maras, Mardin, Mus, Siirt, Sivas, Tunceli, urfa, Van

123
 Table 3.A.2 (Continued). Definition of regions used for Chart 3.1
Regions Number Administrative units (national names)

Japan 10 [Groups of Prefectures]

Hokkaido Hokkaido
Tohoku Aomori, Iwate, Miyagi, Akita, Yamagata, Bukushima
South Kanto Saitama, Chiba, Tokyo, Kanagawa
North Kanto Ibaraki, Tochigi, Gunma, Yamanashi, Nagano
Hokuriku Niigata, Toyama, Ishikawa, Fukui
Tokai Gifu, Shizuoka, Aichi, Mie
Kinki Shiga, Kyoto, Osaka, Hyogo, Nara, Wakayama
Chugoku Tottori, Shimane, Okayama, Hiroshima, Yarnaguchi
Shikoku Tokushima, Kagawa, Ehirne, Kochi
KYUSYU Fukuoka, Saga, Nagasaki, Kumamoto, Oita,
Miyazaki, Kagoshima, Okinawa

Australia 7 [States and Territories]

New Zealand 2 [Statistical Areas]

North Northland, Central Auckland, South Auckland, East Coast,
Hawke’s Bay, Taranaki, Wellington
South Marlborough, Nelson, Westland, Otago, Canterbury,
Southland

Canada 12 [Provinces]

United States 51 [States]

Finland Sweden
Data are from the monthly Labour Force Survey, sup- SCB (1987), AKU&vnedeltal, Tab. 37.A. Averages of
plied by the Central Statistical Office of Finland. Averages monthly estimates.
of monthly estimates.
Turkey
Japan
Data are from the 1985 Household Labour Force Survey,
Statistics Bureau (1987), Annual Report on the Lahour extrapolated to 1987 using national rates of growth of the
Force Survey, Table 31. Averages of monthly estimates. labour force and of persons unemployed, taken from
Tables D and I of the Statistical Annex in this volume.
New Zealand Persons classified by the 1985 Survey as “Inactivc unem-
Data are from the quarterly Household Labuur Force ployed” were excluded from the estimates of the civilian
labour force and of pcrsons unemployed. Data on the
Survey, and werc supplied by the Department of Statistics.
civilian Labour force wcre adjusted to a total labour force
Averages of quarterly estimates. Data for the civilian labour
basis using estimates of the armed forces for 1985, obtained
force were adjustcd using a 1986 estimate of persons in the
from OECD (1988), Labour Force Statistics, Part 2.
armed forces.

Norway United States

Data are from the quarterly labour force survey and were Current Population Survey, averages of monthly esti-
supplied by Statistisk Sentralbureau. Averages of quarterly mates. Data were supplied by the Bureau of Labor Statis-
estimates. tics.

124
 Annex 3.B

UNEMPLOYMENT RATE DATA FOR TIME-SERIES ANALYSIS

This annex describes data on regional unemployment and the national total for unemployment, as measured in the
labour force for dates between 1960 and 1987, used in survey, has been allocated to regions on the basis of the
Tables 3.2 to 3.5 and Chart 3.2 of this chapter. registered Unemployment at the end of month. Data from
population censuses in 1954, 1962, 1968, 1975 and 1982
Australia were supplied by INSEE.
Both sets of data relate to the 22 regions shown in
Sources used were ABS (1978), The Labour Force 1978, Table 3.A.2.
which gives estimates consistent with the questionnaire
introduced in February 1978 and with the 1976 census; JWn
ABS (1986), The Labour Force, Australia, Historical
Summary 1966-1984, and the monthly bulletin ABS, The Labour force survey data, annual averages, were taken
Labour Force. Data refer to August. from Statistical Bureau, Management and Coordination
Data were adjusted by the OECD Secretariat for minor Agency (1980), AnnualReport of the Labour Force Survey,
statistical discrepancies at 1978. Unemployment rates pro- and Labour Force Survey, Regional Tables, various issues.
portional to the national rate were allocated to the Aus- They relate to the 10 regions shown in Table 3.A.1.
tralian Capital Territory before 1976and Northern Territory Population census data for 1960, 1970, 1980 and 1985
before 1979. The eight regions used arc those listed in were taken from Statistical Yearbook of Japan and data for
Table 3.A.1. 1950, 1955, 1965 and 1975 were supplied directly by the
Japanese Statistics Bureau, Managcmcnt and Coordination
Canada Agency. Data refer to the population 14 years and over in
1950, 15 years and over in latcr ycars. Analysis was limited
The source used was Statistics Canada (1987), Historical to 46 prefectures, as listed on the right hand side of
Lahour Force Statistics. Data arc annual averages. Tahk 3. A.2, excluding Okinawa, which was first included
For ycars for which no estimates were published, total in the census in 1975.
uncmployment for Prince Edward Island, which accounts
for about one half of a per cent of the Canadian labour Germany
force, was estimated as the difference between the Canada
total and all other regions. The ten regions used in time- The source was Bundesminister fiir Arbeit und Sozialord-
series analysis are those shown in Table 3.A.1. nung (19871, Arbeits- und sozialstathtik, Hauptergebnisse
and Statistisches Bundesamt Wiesbaden (1988), Stand und
Finland Entwicklung der Erwerbstatigkeit - 1986.
Unemployment rates are on the basis of unemployment
Labour force survey data for 12 provinces, for the pop- registered with employment services divided by the sum of
ulation 15-74 years of age, were supplied directly by the this unemployment with employees as estimated from the
Central Statistical Office of Finland. The data for the Mikrozensus. The eight regions retained for time-series
period 1971-1975 are not fully comparable with those of analysis are : Schleswig-Holstein-Hamburg ; Niedersach-
following years, due to the introduction of a new question- sen-Bremen ; Nordrhein-Westfalen ;Hessen ;Baden-Wiirt-
naire in January 1976. temberg ; Bayern ; and West Berlin.
Statistics for Ahvenanmaa, not available in certain years,
were aggregated with those for the neighbouring mainland I@Y
province. The eleven regions retained for time-series ana-
lysis were : Uudenmaan ; Turun-Porin-Ahvenanmaa ; Regional data based on the labour force survey, for the
Hameen ;Kymen ; Mikkelin ;Pohjois-Karjalan ;Kuopion ; civilian population 14 years old and over, are available since
Keski-Suomen ; Vaasan ; Oulun ; Lapin. 1960 for 21 regions.
National data have been adjusted by the OECD Secre-
France tariat for several discontinuities : the element of estimation
needed to reconstruct a continuous series was relatively
Data linked with the labour force survey, annual averages large. 20 regions were used for analysis; data for the region
for 1974-87,were directly supplied by INSEE. In these data of Val d’Aosta, unreported in some years, were aggregated
- ” _

I25
with data for neighbouring Piemonte ; the other 19 regions United States
-
are thZ’Xsfe3 Tn TibIF3.-A.2: - ”’
Sources used were :
Sweden - The Current Population Survey (data directly supplied
by the Bureau of Labor Statistics) : this provides esti-
Data derived from the labour force survey were taken mates relating to all states for 1976-87 and for 27 states
from AKU (1986), Arbetskraftsundersiikningarna 1976- (accounting for 86 per cent of the population) for 1970-
1985, Table 24 and 25, and Arbetskrafasundersokningen, 87. National totals for employment and unemployment
Arsmedeltal, 1986 and 1987 respectively. No adjustments on the CPS basis were taken from Employment and
were needed. The 24 counties used in time-series analysis Earnings, January 1988 ;
are listed in Table 3.A.2. - Census data for 1960, 1970 and 1980 : from Census of
the Population. Part I, U.S. Summary, in each case ;
United Kingdom - The Manpower Report of the President 1971 and
Employment and Training Report of the President, 1976
For unemployment, early series relating to registered
and 1978 : these sources provided annual data 1960-
unemployment (CSO, Abstract of RegionaI StatiMcs and
70, 1970-75, and 1973-78 respectively. 1960-70 data
Regional Trends, various issues) were used to extend-back
relate to the unemployed on a state administrative
the series for 1976-87, which relates to icdividuals claiming
(often a registration) bas&andto_-_e_pJoyeeson non-
unemployment benefits. For employment, a continuous
agricultural payrolls. 1973-78 data are on an estimated
data series was obtained by means of interpolations of the
CPS basis.
data published in Regional Trenak, in Department of
Employment (1976), “New Estimates of employment on a See Note 13 of this chapter for a comparison of the CPS
continuous basis”, Department of Employment Gazette, and census sources.
August, and in various issues of Ministry ofLabour Gazette, Estimates on a CPS basis for employment and unem-
always taking thc most recently published data as bench- ployment at the state level for 1960, and where necessary
marks. Other adjustments have been used in respect of 1970, were constructed using census employment and unem-
discontinuities for specific regions. The resulting employ- ployment totals. Where necessary estimates for 1973-75
ment data are on the basis of employees in employment in were taken from the 1973-78 series referred to above. Tak-
June, while the Unemployment data are on the basis of the ing these-estimates as benchmarks, annual data for 197l-(j 2
annual average number of unemployment benefit claimants. 72 were where necessary estimated by interpolation using
The eleven regions used in time-series analysis are those the 1970-75 series, while data for 1961-69 were estimated
shown in Table 3.A.1. by interpolation using the 1960-70 series.

126
 Annex 3.C

THE CYCLICAL BEHAVIOUR OF REGIONAL UNEMPLOYMENT RATES

As illustrated in this chapter, simple statistics of regional movements risk being distorted by genuine structural shifts
differentials are strongly dependent upon the national in regional differentials, while estimates that use short-run
unemployment rate. Construction of a statistic whose move- cyclical movements could be distorted if high unemploy-
ments are more sensitive to specifically regional changes ment regions enter troughs and peaks before low unem-
than to the general level of unemployment requires an initial ployment regions, or vice versa. In the data in Chart 3.2,
assessment of the empirical relationship between regional the contemporaneous correlation coefficient, averaged
and national unemployment rates. across 10 countries, between the top and bottom quarter
Different types of behaviour that might theoretically unemployment rates is 0.99, which falls to 0.89 if the top
arise at the regional level are illustrated in Table 3.C.1 in is lagged one year relative to the bottom or vice versa. This
terms of a linearity coefficient, p. In the proportional case indicates that short-run cyclical behaviour is not greatly
(p = 0 ) , an increase of one percentage point in the central affected by lagging or leading behaviour, so that the second
rate from 9 to 10 per cent is accompanied by a similar type of distortion does not appear to be important.
percentage increase in each regional rate : in the linear case Estimates of the linearity coefficient p inTable 3.C.2vary \

(p = l), there is a 1 point increase in each regional rate. by country and time-period. There is one estimate below
For obvious reasons, measures of regional differentials zero, and one above I (high estimates for Canada and
move differently in the two cases. Australia may reflect the crudeness of the regional disag-
Estimates of the degree of linearity that exists cmpirically gregations, see Note 18 of this chapter). A typical value in
are given in Table 3.C.2. The estimator, described exactly the table is about 0.4 (thc avcrage is 0.48 estimated from
in Note u ) to this table, is based upon the principle that if 1-year changes and 0.27 estimated from 5-year changes).
behaviour i s linear, the level difference between high and Taking into account the fact that some of the extremes in
low-unemployment regional rates will be on average con- the estimates probably reflect estimation errors, the
stant when unemployment changes, while if behaviour is valuc 0.4 was retained for all countries as the basis for the
proportional, the ratio difference between rates will be on measure of differentials shown in Table 3.3. It is evident
average constant, Estimates were made using both short- from the discussion here, and the dcfinitions in Tables 3.C.2
term (1-year) and medium-term (5-year) changes in the and 3.3, that this measure will, by construction, not show
diffcrenccs. large movements in differentials associated with the short-
It was recognised that estimates that use longer-term run economic cycle.

D=O R = 0.5 0- I
Initial regional (proportional) (intermediate) (linear)
unemployment rate Typical pattern of regional unemployment rates
when a recession increases the 9 per cent rate to LO

2.0 2.22 2.49 3.0

5.0 5.55 5.75 6.0
9.0 10.00 10.00 10.0
14.0 15.55 15.24 15.0
20.0 22.22 21.48 21.0
Measura of differentials between highest (initial 20 per cent)
and lowest (initial 2 oer cent) unemlonnent rates

Difference 18.0 20.0 18.99 18.0

Ratio 10.0 10.0 8.63 1.0

127
 Table 3 .C.2. Estimated degree of linearity In regional-level responses to cyclical changes in unemployment”
Auslralia Canada Finland FranEeb Germany Italy Japanh Sweden United
Kingdom United
States

1 -year changes 1.36 0.95 0.31 031 0.32 0.39 0.61 -0.13 0.45 0,OO
5-year changes 0.34 0.44 0.28 0.31 0.20 0.10 0.33 0.13

u) Figures shown are values of D such that the correlation coefficient between changes in (U$- U,n) and UAo is zero, where U,, U, and U, are top, bottom and all-
economy unempioyment rates respectively (see Table 3.1 for details). Eftimation used the maximum sample available, e.g. for those countries with 28 years of data
(1960-1987) the correlation coefficients were calculated with a sample of 21 abservations for I-year changu and 23 observations for 5-year changes. Estimates on
the 5-year basis are reponed only where data were available before 1970.
b) For France and Japan, I-year changes are based on labour force survey data For 19741987, and 5-year changes (Tor France, 7 or $-year changes were used) are based
on census data.
Source: See Annex 3.8.

I28
 Annex 3. D

CHANGE IN REGIONAL PATTERNS

OF UNEMPLOYMENT RATES THROUGH TIME

The correlations of regional unemployment rates with If correlations below 0.7 are taken as indicating significant
the patterns in 1975 and 1987 shown in Table 3.4 give only change, the tables suggest that the regional pattern has
a partial picture of regional changes through time. altered significantly several times in Australia and in the
Table 3.D. 1 below shows cross-correlation matrices for United States (for example, in the United States correla-
regional patterns of two-yeZii5averagE-unemplo~ment rates, tions between 1960-61 and 1964-65, between 1964-65 and
omitting countries for which the entries never fell below 1974-75, between 1974-75 and 1980-81, and between 1980-
0.7, which were Finland (where the lowest entry was 0.89), 81 and 1986-87, are all below 0.7), twice in Canada and
Japan (0.91), and Sweden (0.77). In quinquennial census Germany, and only once in France, Italy and the United
data for Japan (not tabulated here) the lowest correlation Kingdom, over the periods for which data were obtained.
coefficient after 1960was 0.84, also indicatinggreat stability.

129
 . . . .

Table 3.D. 1 . Correlation matrices between regional unemployment rates at two-year intervals
1962-63 1961-65 1966-67 1968-69 1910-71 1Y72-13 1974-75 1976-77 1978-79 1980-81 1982-83 1984-85 1986-87

AUSTRALIA
1966-67 1.00 0.63 0.74 0.48 0.41 0.54 0.51 0.57 0.83 0.69 0.71
1968-69 1.00 0.94 0.29 0.69 0.62 0.43 0.51 0.43 0.59 0.72
1970-71 1.00 0.48 0.48 0.58 0.64 0.63 0.66 0.73 0.81
1972-73 1.00 0.16 -0.22 0.92 0.92 0.51 0.23 0.22
1974-75 1.00 0.36 0.08 0.39 0.06 0.16 0.29
1976-77 1.00 0.02 -0.05 0.40 0.65 0.73
1978-79 1.oo 0.91 0.57 0.34 0.33
1980-81 1.00 0.45 0.19 0.22
1982-83 I .OO 0.88 0.81
1984-85 1.OO 0.97

CANADA
1966-67 1.00 0.93 0.90 0.92 0.91 0.91 0.85 0.76 0.95 0.91 0.85
1968-69 1.00 0.95 0.97 0.89 0.91 0.89 0.84 0.97 0.88 0.85
1970-71 1.00 0.96 0.78 0.77 0.73 0.66 0.90 0.86 0.82
1972-73 1.00 0.85 0.84 0.81 0.74 0.93 0.94 0.93
1974-75 1 .00 0.97 0.94 0.87 0.95 0.84 0.76
1976-77 1.00 0.98 0.94 0.96 0.79 0.74
1978-79 1.oo 0.98 0.94 0.73 0.68
1980-81 1.00 0.89 0.61 0.57
1982-83 1.OO 0.86 0.80
1984-85 I .oo 0.98

FRANCE
1974-75 1 .MI 0.96 0.80 0.62 0.52 0.43 0.48
1976-77 1.03 0.89 0.71 0.65 0.56 0.59
1978-79 1.oo 0.95 0.90 0.84 0.87
1980-81 1.00 0.96 0.93 0.95
1982-83 1.00 0.99 0.98
1984-85 1.00 0.99

GERMANY"
1960-61 0.93 0.91 0.70 0.70 0.78 0.66 0.50 0.51 0.44 0.43 0.45 0.36 0.32
1962-63 1.00 1.00 0.87 0.85 0.92 0.74 0.68 0.63 0.47 0.45 0.43 0.34 0.27
1964-65 1.00 0.88 0.86 0.93 0.74 0.70 0.64 0.48 0.44 0.42 0.32 0.25
1966-67 1.00 0.99 0.97 0.90 0.94 0.89 0.74 0.70 0.64 0.57 0.49
1968-69 1.00 0.96 0.93 0.95 0.93 0.80 0.76 0.71 0.65 0.58
1970-71 1 .00 0.88 0.87 0.81 0.62 0.59 0.55 0.49 0.41
1972-73 1.00 0.93 0.96 0.88 0.88 0.87 0.83 0.76
1974-75 I .00 0.97 0.85 0.82 0.76 0.72 0.65
1976-77 1.00 0.94 0.92 0.88 0.85 0.80
1978-79 1.00 0.99 0.96 0.93 0.91
1980-81 1.00 0.98 0.96 0.93
1982-83 1.00 0.98 0.97
1984-85 1.OO 0.99

ITALY
19Mr61 0.93 0.72 0.85 0.90 0.87 0.88 0.87 0.80 0.83 0.82 0.77 0.73 0.76
1962-63 1.00 0.84 0.89 0.86 0.80 0.82 0.83 0.76 0.76 0.76 0.12 0.67 0.66
1964-65 1.00 0.83 0.65 0.58 0.58 0.55 0.45 0.50 0.50 0.45 0.42 0.41
1966-67 1.00 0.87 0.77 0.80 0.81 0.76 0.73 0.71 0.64 0.65 0.60
1968-69 1.00 0.97 0.97 0.95 0.92 0.95 0.92 0.85 0.80 0.83
1970-71 1 .oo 0.99 0.96 0.90 0.94 0.94 0.88 0.80 0.88
1972-73 1.00 0.97 0.91 0.94 0.93 0.88 0.81 0.88
1974-75 1.oo 0.96 0.94 0.94 0.91 0.87 0.91
1976-77 1.00 0.96 0.95 0.91 0.89 0.88
1978-79 1.OO 0.98 0.94 0.89 0.90
1980-81 1.00 0.98 0.94 0.95
1982-83 1.00 0.97 0.97
1984-85 1.00 0.94

130
 Table 3.D. 1 (Continued). Correlation matrices between regional unemployment rates at two-year intervals
1962-63 1964-65 1966-67 1968-69 1970-71 1972-73 197675 1976-77 1978-79 1980-81 1982-83 1984-85 1986-87

UNITED KINGDOM
1960-61 0.97 0.99 0.98 0.92 0.90 0.84 0.82 0.83 0.84 0.68 0.61 0.63 0.70
1962-63 1.00 0.99 0.97 0.96 0.95 0.93 0.89 0.89 0.91 0.78 0.72 0.75 0.80
196465 1.00 0.98 0.95 0.95 0.90 0.87 0.87 0.88 0.73 0.65 0.69 0.75
1966-67 1.00 0.97 0.93 0.85 0.86 0.86 0.88 0.72 0.65 0.67 0.73
1968-69 1.00 0.97 0.90 0.91 0.89 0.92 0.81 0.74 0.77 0.82
1970-71 1.00 0.97 0.95 0.93 0.96 0.87 0.80 0.83 0.88
1972-73 1.00 0.96 0.94 0.97 0.92 0.87 0.90 0.93
1974-75 1.oo 0.99 1.00 0.94 0.89 0.91 0.93
i976-77 1.00 0.99 0.93 0.88 0.90 0.92
1978-79 1.oo 0.95 0.90 0.92 0.94
1980-81 1.00 0.99 0.99 0.99
1982-83 1.00 0.99 0.98
1984-85 1.00 0.99
1986-87 1.00

UNiTED STATES
1960-61 0.91 0.64 0.61 0.66 0.69 0.60 0.58 0.46 0.61 0.67 0.70 0.67 0.42
1962-63 1.00 0.81 0.74 0.76 0.70 0.62 0.50 0.45 0.56 0.44 0.55 0.55 0.43
1964-65 1.00 0.94 0.91 0.83 0.75 0.49 0.48 0.43 0.13 0.23 0.25 0.23
1966-67 1.00 0.96 0.82 0.73 0.53 0.49 0.48 0.22 0.29 0.32 0.25
1968-69 1.OO 0.86 0.74 0.54 0.46 0.48 0.28 0.35 0.38 0.32
1970-71 1.00 0.89 0.67 0.56 0.55 0.41 0.40 0.37 0.20
1973-73 1.00 0.76 0.74 0.68 0.33 0.27 0.23 0.04
1974-75 1 .oo 0.89 0.74 0.47 0.35 0.15 -0.14
1976-77 1.00 0.84 0.33 0.18 0.04 - 0.24
1978-79 1.oo 0.62 0.48 0.40 0.07
1980-81 1.00 0.92 0.78 0.37
1982-83 1.OO 0.91 0.56
1984-85 1.00 0.78

FRANCE (census data)

1962 1968 1975 1982

1954 0.83 0.88 0.80 0.18

1962 1.00 0.87 0.92 0.42
1968 1.00 0.89 0.38
1975 1.00 0.53

a) Data for Germany in 1986-87 refer to 1986 only.