Investigating the Use of
Regional Economic
Forecasting Models in
Wales
Report Prepared by
London Economics
for
the Future Skills Wales
(FSW) Partnership
January 2002
Investigating the Use of Regional
Economic Forecasting Models in Wales
Report by
London Economics
to
the Future Skills Wales (FSW)
Partnership
January 2002
While every care has been taken in preparing this report, London Economics is
not liable for any errors and omissions.
January 2002
Contents
Page
1
Introduction
1
1.1
Terms of Reference
1
1.2
Content and Structure of Report
1
1.3
Acknowledgments
2
2
Skills Forecasting: Theory and Practice
3
2.1
Introductory Observations
3
2.2
Role of Skills Forecasts
4
2.3
Methodological Approaches to Skills Forecasting
5
2.4
Quantitative Methods
5
2.5
Qualitative Methods
8
2.6
Key Issues and Pitfalls in Forecasting
10
2.7
Occupational and skills forecasting – the practice
13
2.8
Future Skills Wales 1998
19
2.9
Concluding Remarks
19
3
Regional Data Issues
21
3.1
Introductory Observations
21
3.2
Problems with the regional GDP data
21
3.3
Other Regional Data Issues
26
3.4
Concluding Remarks
29
4
Regional Economic Forecasting Models
30
4.1
Introductory Observations
30
4.2
Comparison of Regional Economy Model Structures
30
4.3
Comparison of Regional Economy Model Properties
37
4.4
Forecast Performance of Regional Forecasting Models
40
4.5
Concluding Remarks
46
5
Results of Interviews with Representatives from the
Future Skills Wales Partners
47
London Economics
January 2002
i
Contents
Page
5.1
Introduction
47
5.2
Results of Interviews with Representatives from the FSW
Partners
47
Results of interviews with representatives of other
institutions in Wales and outside Wales
49
7
Results of Interviews with Forecast Providers
51
8
Potential Use of Existing Forecasts for Skills
Projection
52
9
Key Conclusion and Recommendations
54
9.1
Key conclusion
54
9.2
Getting more value from the forecasts
54
9.3
Projecting skills needs
56
6
Bibliography
58
Annex 1 Occupation and skills forecasts: the practice in a
number of countries
63
Annex 2 Forecasts of three major regional forecast
providers
70
Annex 3
List of Persons Interviewed for the Study
77
Annex 4
Interview questions for forecast users
79
Annex 5 List of contact persons at forecast service
organisations
London Economics
January 2002
87
ii
Tables & Figures
Page
Table 2.1: Main characteristics of the forecasts
16
Table 2.2: Methodology for supply and demand forecasts
17
Table 2.3: Labour demand and supply dynamics
18
Table 3.1: Analysis of Evolution of UK and Welsh GDP Data 1990 and 1995
23
Table 3.2: Evolution of Annual Welsh GDP Growth - 19901999
24
Table 3.3: Changes Over Time in Sectoral Distribution of 1990
and 1995 Welsh GDP
25
Table 3.4: Sampling Error of Employment Survey at Local
Level
28
Table 4.1: Summary Observations from Comparison of Model
Structures
37
Table 4.2: BSL and Oxford/NIERC Model Responses to
Standard Economic Shocks after 5 years
39
Table 4.3: National Model Responses to 1 Percentage Point
Reduction in Standard Rate of Income Tax -- Impact
on Total GDP after 5 years (% Difference from
Baseline)
39
Table 4.4: GDP and Employment Forecast Errors Mean
absolute error of growth rate forecasts (in percentage
points)
42
Table 4.5: Evolution of sectoral employment forecasts by
Cambridge Econometrics and Oxford Economic
Forecasting
44
Table 7.1: Results of Interviews of Forecast Service Providers
51
Table A.1: Annual GDP growth forecast performance – BSL
70
Table A.2: Annual employment growth forecast performance –
BSL
71
Table A.3: Annual GDP growth forecast performance - CE
72
Table A.4: Annual employment growth forecast performance CE
73
Table A.5: Annual GDP growth forecast performance - OEF
74
Table A.6: Annual employment growth forecast performance OEF
75
London Economics
January 2002
iii
Tables & Figures
Page
Table A.7: Evolution of sectoral forecasts by Cambridge
Econometrics
76
Table A.8: Evolution of sectoral forecasts by Oxford Economic
Forecasting
76
Figure 1: Occupational and skills forecasting: the key elements
14
London Economics
January 2002
iv
The Future Skills Wales Partnership
In 1998, the Future Skills Wales (FSW) research project was undertaken in
order to establish the current and future skills needs throughout Wales. A
strategic partnership, representing a wide cross-section of public and private
sector organisations, was created to guide and support the research activity.
The Partnership currently includes the National Assembly for Wales, the
Welsh Development Agency, ELWa, the Welsh Local Government
Association, Careers Wales, the Secondary Heads Association, ACCAC,
Cyngor NTO Cymru, the Local Government Data Unit, Fforwm, ESTYN,
Wales Council for Voluntary Action, CBI Wales, the Employment Service, the
Federation of Small Businesses, DYSG, and Wales TUC.
The Future Skills Wales (FSW) Partnership provides a national infrastructure
through which truly collaborative research can be undertaken in support of
skills development in Wales.
http://www.futureskillswales.com
v
London Economics
January 2002
Executive Summary
Executive Summary
The Future Skills Wales (FSW) Partnership commissioned London Economics
to examine their use of regional forecasts and regional forecasting models,
and make recommendations on how they could individually and collectively
derive greater benefits from their forecasting related activities. In addition,
London Economics was asked to assess how the forecasts produced or used
by the Partners could be used in a skills needs projection exercise by the
Future Skills Wales Partners.
In response to the terms of reference of the assignment, we first surveyed the
current literature by academics and practitioners on the issue of occupational
and skills forecasting. We then examined the issue of the quality of the
regional data used in regional economic forecasts available for Wales.
Next, we focused on the structure and properties of the regional models and
the forecasting performance of the three regional forecasting services
producing regular economic forecasts for Wales and the other major regions
of the UK, namely, in alphabetical order, Business Strategies Limited (BSL),
Cambridge Econometrics (CE) and Oxford Economic Forecasting/Northern
Ireland Economic Research Centre (OEF).
We interviewed a number of representatives from the FSW Partners and
other institutions. These interviews focused mainly on the issues of regional
economic forecasting and skills projections.
Finally, on the basis of the observations and conclusions drawn from the
various research blocks described above, we offer a number of
recommendations on how FSW Partners could derive greater benefits from
their current forecast-related activities, and highlight a number of key issues
to consider in preparing for the next skills assessment in Wales.
Chapters Summary
London Economics is submitting this report to the FSW Partners which
includes the following chapters:
Skills Forecasting: Theory and Practice
The review of the literature and actual practice of skills forecasting shows that
one needs to be mindful of a number key issues and pitfalls:
First, the underlying labour market projections need to based firmly on sound
macro-economic projections, concerning how the economy as a whole is
evolving, both cyclically and structurally.
vi
London Economics
January 2002
Executive Summary
Second, skills forecasts are typically proxied by occupational projections.
These forecasts perform reasonably well in capturing occupational change,
and can be usefully extended into forecasts of qualifications and of job
opportunities. The real issue is whether occupational forecasts are successful
at picking up changing skill requirements.
·
In the past, the distinction between occupations and skills has been
less clear, and there is some evidence that roughly only half of all
changes in skill requirements are associated with occupational change.
This is due to the fact that the skills content of many occupations
changes rapidly.
·
Recent trends reveal that aggregate changes in skills requirements are
due more to changes in skills demanded within occupations than to
new, highly skilled occupations, driving out old, less skilled
occupations.
Third, net changes in occupational employment are only one indicator of
future demand for skills. Another measure, which is equally important for
assessing education and training needs, is the replacement demand needed to
offset outflows due to retirements, occupational mobility, etc.
·
In particular, a robust skills forecast needs to take account not only of
the skills needed in growth industries, but must also consider the
needs of stagnant and declining industries where high employee
turnover may result in substantial re-training requirements.
·
The bottom line is that a robust skills forecast must focus on the issue
of gross employment changes by occupation rather than net changes
in employment.
Fourth, on the labour supply side, at a minimum, demographic changes such
as death and migration need to be taken into account, in addition to changes
in activity rates (such as retirement and other exits from the workforce).
·
New entrants, migration, re-entrants to the workforce, as well as interoccupational mobility constitute other flows in the supply of skilled
workers.
·
In addition, skills projections should take account of the likely
responses of the labour force to various skill scarcity signals. If the
long-term projections were to ignore the spontaneous take up of
specific skills (through schooling, adult education or training) and
were to project the current skill mix into the future, the supply of
skills will likely be grossly miscalculated and could lead to seriously
misleading conclusions about likely future skill imbalances.
vii
London Economics
January 2002
Executive Summary
Fifth, knowledge of the changing demand for qualifications can be very
useful within an educational planning context. However, it is less clear
whether qualifications are a better measure of skills than occupations.
·
In this regard, some models have been developed to generate
projections of demand for and supply of qualifications taking into
consideration the existing level of qualifications and skills.
·
In particular, efforts have been made to identify generic and specific
skills. Generic skills are a set of skills that are not precisely defined
and often merge into personal attributes. Undoubtedly some generic
skills, such as those associated with IT, have become increasingly
important in recent years. Using a methodology similar to the one
adopted to link the demand for qualifications to occupations, it is
possible to generate projections of generic skill needs from the basic
occupational projections. The key additional input requirement is an
estimate of the importance of the various generic skills within a given
occupation.
·
Projecting the demand for specific skills is much more complex as
occupational change may be an unreliable predictor of required
specific skills due to the changing skill content of various occupations.
Sixth, surveys of households’ and employers’ views and perceptions can
provide additional information that can be used to either provide additional
information and data for the projections and/or test the outputs of modelbased skills forecasts.
·
To the extent possible, in addition to targeted surveys aimed at
generating additional data required for the skills projections, more
comprehensive surveys seeking labour market participants’ views of
likely developments with regards to future skill requirements and
availability should be undertaken in parallel with the model-based
projection exercise to provide an independent source of information
and validation to the users of skills-forecasts.
·
That being said, while such surveys can provide valuable information,
it is important to remember that responses in such surveys tend to be
heavily influenced by current circumstances or those that are expected
to prevail in the very near future. Therefore, some degree of caution is
always advisable in using such surveys over longer time horizons.
Finally, it is important to stress that uncertainty prevails at each of the many
steps involved in a skills forecasting exercise, namely economic activity
forecasting, employment forecasting, occupational projections taking into
viii
London Economics
January 2002
Executive Summary
account the dynamics of the labour market, skill requirement projections,
labour supply and skill supply projections.
It is interesting to note here that the U.S. Bureau of Labor Statistics is of the
view that any occupational forecast percentage error falling within the range
of plus or minus 10% the actual outcome is reasonable and acceptable.
Some steps -- such as complementary surveys and the use of alternative
scenarios -- can be taken to reduce somewhat the uncertainty, but cannot
totally eliminate it.
Moreover, as the forecast error tends to increase with the level of spatial,
industrial and occupational disaggregation, it is preferable to maintain the
focus of any skills projection at a relatively high level of aggregation.
Regional Data Issues
q
The quality of the regional data used in regional forecasts is often less
than desirable. Frequent revisions and substantial publication lags mean
that the true values of key economic variables at the starting point of the
forecast are often uncertain. This can have a significant impact on the
quality of the economic forecasts as the true state of the regional economy
at the starting point of the forecast is known only very imperfectly.
q
The uncertainty increases as one moves down from national, highly
aggregated data to disaggregated regional and sub-regional data.
Moreover, the degree of uncertainty is not eliminated entirely with
successive releases of a given year’s data.
q
The data improvement process launched by the National Assembly for
Wales should eventually contribute to alleviate some of the problems
faced by regional forecasters. However, the reality is that over the
foreseeable future these gains will most likely be limited. Over the short
to medium-term future, the starting point of any forecast of the Welsh
economy is likely to continue to be clouded by considerable uncertainty
about the precise state of the economy.
Regional Economic Forecasting Models
Next, we examined the structure of the regional models used by the three
forecasting services, focusing in particular on the interaction between the
regional model and the national model, the determinants of the long-run
equilibrium in the model and the availability of detailed employment and
occupational data at the regional and sub-regional level.
None of the forecasting models provide for a full and effective bottom up
approach to the regional forecasting exercise and each have different
ix
London Economics
January 2002
Executive Summary
characteristics and different strengths and weaknesses. The OEF macroeconomic model is well known in the UK and internationally for its
embedded theoretical properties, but is based on a regional top-down
approach. A richer regional texture is captured in the regional models of both
BSL and CE but this comes a cost of increased complexity and, especially in
the latter case, very large data requirements.
q
The model responses to standard economic shocks reported in this
study do differ across models. But, they all fall within a plausible
range and do not provide strong evidence in favour of one or the
other model. The main conclusion one should draw from such a
comparative model property exercise is that any precise model-based
estimate of the response of the economy to a particular policy shock
should always be viewed with some caution. As a protection against
putting too much emphasis on a specific model result, it may be
preferable to focus on scenarios rather than single projections or
simulations.
In assessing any forecasting performance one should always remember the
saying that the only certain thing about the future is that it is highly
uncertain. Forecasts will be wrong almost by definition, as it is impossible for
forecasters to capture all the unknowns that may impinge on the actual
performance of the economy in the years ahead. The special regional data
quality problems that confront regional forecasters in the UK only compound
the forecast error risk.
Limited data availability prevents the completion of a thorough statistical
assessment of the errors of the forecasts for the Welsh economy produced by
the forecasting services. Nevertheless, a few general observations can be
drawn from our analysis.
q
·
First, the forecast errors tend to be larger at the Welsh level than at the
national level.
·
Second, the forecast errors of employment growth tend to be larger
than those for GDP growth.
·
Third, while the forecasting errors appear to be relatively large,
ranging from 0.6% to 1.7% in the case of Welsh GDP growth, they are
very similar to the forecasting errors of GDP growth for the G-7
countries by the IMF and the OECD.
Does it mean that forecasts are useless? In our view, the answer is that
they can have a value as an input to a coherent framework for organising
one’s thinking about the future. However, less attention should be paid
to the precise point estimates shown in the forecasts for the various
economic indicators and more attention should be given to the intuition
behind the forecast. This would be particularly the case when the forecast
x
London Economics
January 2002
Executive Summary
changes substantially from one projection to the next.
In-depth
discussions with forecast providers would be most useful in that regard.
Results of Interviews
We conducted a number of interviews with a number of officers from the
FSW Partners, and from other institutions in Wales or outside Wales that
focus on regional economic development or skills forecasting.
The key issues raised during the interviews were the problems with the
quality of the data, the limited scope for regional information to be
incorporated into the regional forecasts, and the need for a good skills
assessment.
Officers of the FSW Partners noted as well that they wished to obtain greater
value from their current forecasting-related activities.
Representatives of the forecasting services were very happy to contribute to
the project and noted that they i) are also concerned about the quality of the
regional data, ii) strive to take account of region-specific information within
the constraints imposed by the modelling structure adopted and iii) would be
pleased to meet regularly key Welsh clients to discuss the forecast and
articulate the intuition behind the projection. All indicated that such
meetings would also be a valuable opportunity for them to gain additional
insights into regional economic developments.
Potential Use of Existing Forecasts for Skills Projection
All the off-the-shelf forecasts could be used as a building block for a skills
projection, but significantly more work would be required as they provide
only an occupational or an employment forecast. The latter would have to be
mapped into a skills projection and complemented by a comprehensive
projection of the likely labour supply response.
Key Conclusion and Recommendations
The key conclusion from our report on the use of regional economic
forecasting models is that any point estimates of expected GDP or
employment growth are affected by a high degree of uncertainty. This is not
specific to the forecasts for Wales but is a general characteristic of any
economic forecast. However, what is particular to Wales, and the other
regions of the United Kingdom, are the problems with the quality of the
regional data which tends be lower than that of the national data and which
add to the forecast uncertainty by seriously clouding the starting point of the
regional forecast. Unfortunately, the data quality problems worsen at the
sub-regional level.
These factors lead one to conclude that:
xi
London Economics
January 2002
Executive Summary
1. In the context of the skills assessment that the FSW Partnership plans
to undertake in 2003, analysts and policy-makers would be well
advised to avoid the temptation of putting too much weight on
specific forecast values of key variables of interest such as
employment, occupations, etc., at the regional or sub-regional level;
2. Moreover, maintaining the skills assessment at a relatively aggregated
level, in terms of occupations, skills and geographical space, will
likely result in much more robust conclusions that will stand the test
of time better than a very large and detailed assessment.
The key issue raised a number of times during the interviews is how the
subscribers to the off-the-shelf forecasts could get more value from their
investments. In particular, many expressed a strong desire to be able to go
behind the published numbers and better understand the dynamics driving
the forecast. In our view, the following two actions could go a long way
towards meeting this objective:
·
Regular inter-institution meetings between the partners could be held
to review and assess forecasts; and,
·
Forecast services could be invited to meet regularly (once or twice a
year) with key forecast subscribers in Wales to review and explain the
forecasts, and engage in an exchange on potential region-specific
factors that would have to be taken into account in the regional
forecast.
Some interviewees noted that it would be worthwhile to develop some local
capacity, housed either in one of the Partner institutions or in an outside,
possibly newly-created, body to produce regional forecasts that would allow
for greater use of local knowledge. In our view,
·
An assessment of potential benefits and costs (financial, human
resources, etc.) of creating and maintaining a local modelling and
forecasting capacity should be undertaken to inform the decision
makers likely to be involved in this decision. In the general
conclusions of this report we provide some tentative cost estimates of
moving to an in-house forecasting capacity.
Many noted in the interviews that the next FSW exercise should be carefully
prepared. In this regard we offer the following recommendations:
·
There should be a clear understanding of the purpose of the skills
assessment at the beginning of the exercise. Is the focus on the
demand for skills or is it on skills gaps, i.e., the difference between the
supply of and the demand for skills? The latter is a much more
demanding objective as it requires projections of both the demand for
skills and the supply of skills.
xii
London Economics
January 2002
Executive Summary
·
Despite all the pitfalls associated with forecasting, consideration
should be given to use again a regional employment and occupational
forecast as one of the inputs into the next FSW exercise.
·
However, as noted before, such a forecasting exercise should involve a
number of scenarios to reflect the key uncertainties underlying the
projections. For example, scenarios could be run for different
economic growth and productivity assumptions - two key
determinants of future employment.
·
Moreover, to ensure that they provide useful information, such
scenarios would need to be defined in close co-operation between the
FSW Partners and the forecasting service.
·
Given the uncertainty about precise occupational skill requirements
for specific employment groups, it would be preferable to focus the
projection exercise on generic skills or skills that are used in a number
of occupations.
·
While recognizing that there exists a need for sub-regional
information on future skills demands and supplies, we would argue
against using very detailed employment and occupational projections
at a sub-regional level as they may not be very reliable.
·
Careful consideration should be given to how occupations, skills and
labour supply responses are modelled and projected by either the
forecasting service or the in-house projection team.
·
Consideration will need to be given to how optimally combine localsurvey based information with the broader forecasting exercise. Such
surveys can be used to generate data that are not otherwise available
for the skills projections or as an external reality check of the modelbased projections.
xiii
London Economics
January 2002
Chapter 1
1
Introduction
1.1
Terms of Reference
Introduction
The Terms of Reference set out by the FSW Partnership for this assignment
specify that London Economics (LE) undertake the following tasks:
1. Review the use of forecasts and forecasting models by the FSW
Partners, namely the National Assembly for Wales, the Welsh
Development Agency, and ELWa, and make recommendations on
how they could individually and collectively derive greater benefits
from their forecasting related activities;
2. Assess how the various forecasts produced or used by the Partners
could be used in a skills needs projection scenario by the FSW
Partners.
1.2
Content and Structure of Report
This report covers the following areas of work undertaken for the assignment:
·
The results of our research on skills forecasting;
·
The results of our research on regional data and regional forecasting
in the UK;
·
The results of interviews held with a number of officers from the FSW
Partners;
·
The results of interviews held with representatives of a number of
other institutions involved in skills forecasting or regional
development;
·
The results of interviews with forecast service providers; and
·
Conclusions and recommendations.
The structure of the Report is as follows.
Chapter 2 provides a
comprehensive review of current thinking on skills forecasting and a
description of how a number of EU countries, Australia, Canada and the
United States actually look at this issue.
1
London Economics
January 2002
Chapter 1
Introduction
Chapter 3 looks at the issue of the quality of the regional data used in
regional forecasting and Chapter 4 reviews the services provided by the three
well-established organisations that produce regular regional forecasts for
Wales, namely in alphabetical order, Business Strategies Ltd (BSL),
Cambridge Econometrics (CE) and Oxford Economic Forecasting/Northern
Ireland Economic Research Centre (OEF). This chapter focuses in particular
on the structure of the various models used to produce regional forecasts for
Wales, the properties of these models, and the forecasting performance of the
three services.
Chapter 5 provides a synthesis of the results of the interviews with officers
from the FSW Partners, and Chapter 6 provides the highlights of interviews
with representatives of other institutions with a strong interest in Welsh
economic development and/or skills forecasting.
Chapter 7 summarizes the results of the interviews with regional forecast
service providers. Chapter 8 discusses whether forecasts currently used by
FSW’s Partners could be used in a skills projection. Finally, key conclusions1
and a number of recommendations are set out in Chapter 9.
As noted above, a detailed review of state-of-the-art academic thinking in
relation to regional skills forecasting is provided in the next chapter. Readers
interested only in the report’s findings on regional economic forecasting may
skip Chapters 2 and 3, and go directly to Chapter 4.
1.3
Acknowledgments
We would like to express special thanks to the FSW Partners, the Economic
Development Department and the Economic Advice Division of the National
Assembly, and the Welsh Development Agency and their representatives for
their inputs and advice. We would also like to thank all the other participants
in the interview process and the representatives of Business Strategies Ltd.,
Cambridge Econometrics and Oxford Economic Forecasting for their valuable
contribution.
1
More detailed, issue-specific conclusions can be found at the end of the main chapters of this document.
2
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
2
Skills Forecasting: Theory and Practice
2.1
Introductory Observations
Proper anticipation of how the skill content of occupations is evolving and
the type of new skills required raises many challenges for policymakers, as
failure to plan adequately could have potentially serious consequences for
economic and social development. In response, over the last 50 years, a vast
literature, both academic and more technical, has emerged (including work
undertaken by economists at London Economics), which has attempted to
develop and implement coherent skills planning frameworks2. In the
subsequent sections of this chapter, we will focus on the most recent thinking
on these issues.
Some economists have argued that any imbalance between the demand and
supply for a given skill would result in a corresponding real wage
adjustment, which will re-establish equilibrium between demand and supply.
In the case of upward pressure on the remuneration of a particular skill,
employers will tend to reduce their demand for this skill by relying more on
other types of skills or non-human inputs (an input substitution effect), while
the labour force will respond positively to the wage signal by acquiring more
of this particular skill. However, a significant body of research has stressed
that skills markets are unlikely to reach equilibrium instantaneously for a
number of reasons:
2
·
It is by now commonly accepted that recruiting workers takes time
and effort, and thus any disequilibrium in the marketplace is unlikely
to be resolved quickly.
·
In many instances, institutional wage setting mechanisms (e.g.,
industry-wide collective wage bargaining) are such that they may not
necessarily result in a ‘market clearing’ wage for a specific skill.
·
It may be difficult for current and future labour force participants to
distinguish between temporary labour imbalances and longer-term
“scarcity” of certain skills. As a result, employees and future
employees may not respond to the scarcity signals.
A comprehensive bibliography on issues related to skills forecasting and planning can be found on the
website of the European training village www.trainingvillage.gr/etv.
3
London Economics
January 2002
Chapter 2
·
Skills Forecasting: Theory and Practice
More generally, firms and workers may quite rationally decide not to
increase training if neither perceives a direct long-run interest in
doing so, resulting in an inefficient social outcome (Haskel and Holt,
1999). This may be the situation with imperfect or incomplete capital
markets for educational funding3.
Thus, skills shortages may arise and may contribute to increased costs by
putting pressure on the remuneration of such skills. They also can, in some
cases, result in lower product quality and/or reduced productivity levels if
employees are placed in positions that do not match their skills4. Also of
significance from the perspective of regional development agencies is that
skill shortages can result in the loss of mobile investment projects.
2.2
Role of Skills Forecasts
In response, all industrial countries have invested in some form of labour
market programming, although the precise shape of this involvement varies
substantially. Governments in some countries limit their role to providing
information to the public about the likely demand for different occupations,
while others are more proactive and develop active training and skill
development policies.
In practice, skills forecasting tends to plays a dual role, namely5:
·
A policy role, to supply information about likely future needs to
employment and training authorities; and
·
An information and signalling role, relaying details concerning the
current and likely state of the labour market by occupation to actual
and prospective labour force participants such as employers,
employees, students and school leavers.
The generation of skills forecasts typically relies on quantitative and/or
qualitative methods that project occupational changes at the sectoral level
and map these expected changes into expected skill requirements. In a
number of cases such exercises also translate these projections into
educational and training requirements. Hence, skills forecasting relies on
occupational forecasts that provide data on employment trends for a large
3
There exists a substantial body of economic literature that identifies labour market information as a
“public good” that would be underprovided if left to the private sources, whereas it is also believed
that capital market imperfections may deter young and unskilled workers from investing in education
and training.
4
See for example Nekkers et al. (2000) who investigate matching problems in the labour market.
5
Hughes (1999).
4
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
number of occupational sub-groups.
By linking these occupational
projections with the type of educational and vocational training that is
required, it is then possible to provide guidance to counsellors, school
leavers, employers and other labour market participants.
In this chapter, we review in detail the various methods adopted to-date in
anticipating future skill needs, focussing in particular on the critical issues to
be taken into account by any skills forecaster6. We then examine how skills
forecasting is actually implemented in a number of countries.
2.3
Methodological
Forecasting
Approaches
to
Skills
Both quantitative approaches to modelling the supply and demand for skills
and qualitative approaches are used by practitioners developing skills
forecasts. We discuss each of these approaches in further detail below.
2.4
Quantitative Methods
While skills forecasting models used internationally have tended to exhibit
considerable diversity in approach and complexity, they all share a common
quantitative approach. This approach has focused on the comparison of
manpower requirements with labour supply forecasts to enable an
identification of sectors and occupations that are likely to be in excess supply
or demand within a given time period. Skills needs, and educational or
vocational training forecasts, are then identified from predictors of
occupational change in specific economic sectors.
Typically, a forecast of the demand for occupations is constructed in three
sequential steps, as follows:
·
Construction of a macroeconomic projection for the economy.
·
Projection of economic activity at sectoral or industry level, and
mapping of sectoral projections into sectoral occupational
forecasts.
·
Translation of occupational projections into qualification, skill
and/or educational requirements.
On the supply side, forecasts are developed for the number of economically
active persons holding various qualifications who are likely to enter the
6
See Tessaring (1998).
5
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
specific segment of the labour market within a specified time period. The two
key drivers on the supply side are demographic change and the likely labour
force response to any wage and/or non-wage signals of labour market
imbalances.
Demand and supply projections are then compared to identify skills and
qualifications that are likely to be in excess supply or demand over the
projection period and, in the case of any excess demand, the likely
educational and vocational training requirements.
2.4.1
Modelling the demand for skills
The workhorse of the traditional occupational forecaster has been the
“manpower requirement approach”, which estimates the number of jobs
likely to become available across different job characteristics, including
industrial sector, occupation or work activity and qualifications (see Hughes,
1999, for a review of this approach, and also CEDEFOP, 1998).
Typically, utilising a large-scale macro-econometric model to generate
economy-wide projections of economic activity, projections of employment
are produced by sector over a 6-10-year time horizon. This generally entails
linking predicted economic growth, productivity and other economic factors,
to future employment and the occupational profile for an economy.
In the next step, occupational projections, disaggregated by sector, are
generated. Occupational shares are generally used to disaggregate sectoral
employment by occupation over the forecasting horizon.
Occasionally, the occupational forecasts are further broken down into skills
projections. For example, an educational shares model can be used to
disaggregate occupational employment by level of education or vocational
training. The educational share forecasts are then combined to provide
estimates of the likely demand for different types of educational attainment.
The occupational forecast shows the number of individuals that are likely to
be employed in a particular occupation over a specified time period. A
projected increase would then indicate a prospective rise in demand for the
occupation in question. However, this approach may yield a poor guide to
potential labour shortages if the dynamics of new vacancies are ignored.
New vacancies may arise because individuals choose to leave their existing
jobs to retire, to look after a family, to claim social benefits, to return to
education, or simply to take up alternative employment.
Finally,
demographic factors, such as mortality and migration, may also affect
vacancy dynamics.
Because of these factors, job vacancies can occur in all occupations, even in
those areas that are in decline. In developing projections, practitioners are
6
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
currently attempting to reflect these dynamics, as otherwise projections may
fail to pick up particular training and education needs. What matters for
skills projections is the gross employee turnover by occupation and not the
net change in employment by occupation.
In addition to broad, top-down projection exercises, more detailed studies of
skills needs for particular occupations or regions have become more frequent
over the last 10 to 15 years. This typically involves establishing a link
between specific functions, tasks, abilities, personal attributes, characteristics
and generic skills and specific occupations.
More recently, there has also been an increasing interest among planners and
policy advisors in combining economic information at a local level with the
most recently available national and regional data. Local skill surveys and
audits are used increasingly to obtain local economy specific information and
complement the information provided by the bigger top-down
macroeconomic exercise. While such surveys of vacancies and household
competencies provide a good snapshot of the skills situation in the local
economy at a given point in time, they are not always a good predictor of
future trends. In contrast, the broad top-down approach can be useful in
providing a ‘big picture’ of future trends but, by definition, cannot identify
local economy trends.
2.4.2
Modelling supply
On the supply side, one common approach employed to-date has been to
model labour market inflows in relation to school-leavers and the
unemployed, and then to project the skills profile of individuals from each
group within a given time period. This approach essentially entails the
application of a stock-flow model, which relates the total stock of
unemployment in one period to its previous period, using an accounting
identity linking the main inflows and outflows to the existing number of
unemployed persons.
Supplementary models are then used to determine the proportion of the stock
of economically active individuals who are likely to be engaged in a specific
employment role. Fixed shares or relative wage models are then used to
allocate this stock across occupations.
The main outflows are those due to death, retirement and other exits from the
workforce and emigration. The main inflows relate to the flow of new
entrants, re-entrants to the workforce and migration. If the focus is on
particular occupational categories, the issue of inter-occupational mobility
also needs to be considered.
7
London Economics
January 2002
Chapter 2
2.5
Skills Forecasting: Theory and Practice
Qualitative Methods
In addition to quantitative techniques employed in forecasting skills needs, a
range of qualitative approaches have also been developed. The main
characteristic of this survey-based approach is that it is not primarily
concerned with a quantitative measure of skill requirements, but focuses on
respondents’ views and perceptions of current and future trends.
Such an approach can provide important additional information that may be
used to complement and test results yielded from the quantitative
approaches. It is also generally viewed as the only real option in those cases
where statistical information is scarce or of insufficient quality to enable the
construction of a robust quantitative model7.
2.5.1
Local Labour Market Information
In recent years a growing number of countries have focused on occupational
and skills forecasts at a local level. This new emphasis in regional forecasts
started in the early 1980s, and has seen the emergence of regional study
groups (such as the Regional Employment and Training Observatory, OREF,
in France) that have developed a range of methodologies for forecasting
training needs at a local level.
Much of this analysis has shifted to more qualitative techniques for mapping
projected sectoral occupations and qualifications into local skill requirements.
However, so far there is no consensus on how such forecasts should be
conducted at a local level, and every country has developed its own ad-hoc
methodology according to its special needs. In the following paragraphs we
briefly describe some of the countries’ practices, and a more country
comprehensive review is provided in Annex 1 .
A good example of a regionally focused approach is the Finish model for
regional labour market forecasting.
Local employment offices and
employers, together with social partners develop a questionnaire for data
collection. The district labour office combines the local information with the
regional level to provide labour force prediction by sector and occupation. At
a local and regional level, predictions are obtained for staff requirements in
occupations, and for training needs by type of training.
This process involves the following steps (Kekkonen, 1998). First, interviews
of most relevant local employers are held every half-year to collect views on
expected changes in the labour force, the demand for labour and training
needs by occupational group. The interview results are then combined with
7
Wilson (2000).
8
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
other information at the local employment office, which analyses and
summarises the data for an expert panel comprising representatives from
local authorities, the banking sector, companies, and training institutions and
employment officials. This panel evaluates the labour forecast and the
training demand and decides collectively on a specific outlook. Based on the
agreed outlook, each local employment office produces then a local labour
demand forecast by industrial sector and a plan for labour market training.
Finally, using all this information, the district office produces a summary
labour force forecast by sector and municipality.
Another good example of provision of local labour information can be found
in France8. Although education and labour force forecasts are provided in
France at a national level, such quantitative results are not always used by the
professional branches at a regional level. For example, the metallurgical
professional organisation produces its own study on the basis of qualitative
forecasts. The process involves interviews with successful firms querying
their views on future skill needs. Similarly, for the region of Burgundy, a
local analysis is used to establish training needs of young people. For each
employment area in Burgundy, training needs are estimated taking into
account demographic, unemployment and staff turnover trends, as well as
results of surveys of young people.
In Australia, regional employment forecasts are provided by industry, taking
into account differences in industrial structures, regions-specific industry
effects, population movements and state government expenditures.
However, no skills forecasts are provided by region.
In Germany9, the determination of skill shortages is primarily done at the
local level. Committees are formed to collect information on local labour
conditions. The structure of these committees is relatively formal as union
representation is required as well as representation by employers. Training
decisions are taken primarily at the local level. Information collected at the
local level is given most weight, although information collected at the
national level may be given some consideration.
In the Netherlands, regional councils do exist but they play a smaller role
than in other countries in producing occupational and skills projections. In
fact, such projections are only generated for one (Limburg) of the twelve
provinces. However, training decisions are primarily based on consultations
with employers which, when appropriate, may be done at the local level.
8
9
Giffard et al. (1999).
See Henson and Newton (1995) for a more extensive description of the practices in Germany, the
Netherlands, the U.S. and Canada.
9
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
Skills shortages are largely viewed as a regional issue in the U.S.
Identification of skills shortages occurs primarily through a consultation
process with local employers. This consultation process is conducted by
means of a network of ad-hoc committees whose structure varies
enormously, not only between states but also within the states themselves.
Analysts at the local level are also highly innovative in their data collection
methods. Data may be collected on the basis of skill-types rather than formal
occupations. Because no coding system for detailed skills exists in the U.S.,
ad-hoc classification systems are devised to categorise skill types.
In Canada, consultations with employers plays a key role in the final
diagnosis of skill shortages. The major employment centres occasionally
undertake local surveys. The bulk of the training decisions are made at the
provincial or even local level. However, more so than other countries, skill
shortages may be diagnosed at the sectoral level thus leading to training at
that level.
2.5.2
Scenario Analysis
Given the uncertainty surrounding any economic prediction, a few countries
use scenarios in their approach to occupational and skills forecasting.
For example, in France, quantitative estimates of the labour market demand
for young people by level of education are provided for four different
economic growth scenarios. In Germany, occupational projections are
generated for three different employment paths -- low trend growth, medium
trend growth and high trend growth.
2.6
Key Issues and Pitfalls in Forecasting
The key issues and pitfalls to be mindful of in any skills forecasting exercise
are:
1. The Macroeconomic Outlook
Labour market projections need to be based firmly on sound macro-economic
projections, concerning how the economy as a whole is evolving, both
cyclically and structurally.
2. Occupational Projections versus skills forecasts -- the issue of changing
skills content of occupations
As discussed earlier, skills forecasts are typically proxied by projections at
occupational level. These forecasts perform reasonably well in capturing
occupational change, and can be usefully extended into forecasts of
qualifications and of job opportunities, as well as produced for local and
regional labour markets, and the national economy.
10
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
The real issue is whether occupational forecasts are successful at picking up
changing skill requirements. In the past, the distinction between occupations
and skills has been less clear, and there is some evidence that roughly only
half of all changes in skill requirements are associated with occupational
change (Haskel and Holt, 1999). This is due to the fact that the skills content
of many occupations changes rapidly. Recent trends show that aggregate
changes in skills requirements are due more to changes in skills demanded
within occupations10 than to new highly skilled occupations driving out old
less skilled occupations (Lloyd and Steedman, 1999).
3. Gross employee turnover versus net employment changes by occupations
As already noted, net changes in occupational employment are only one
indicator of the future demand for skills. Another measure, which is equally
important for assessing education and training needs, is the replacement
demand needed to offset outflows due to retirements, occupational mobility,
etc.
In particular, a robust skills forecast needs to take account not only of the
skills needed in growth industries, but must also consider the needs of
stagnant and declining industries where high employee turnover may result
in substantial re-training requirements.
The bottom line is that a robust skills forecast must focus on the issue of gross
employment changes by occupation rather than net changes in employment.
4. Labour supply side considerations
On the labour supply side, at a minimum, demographic changes such as
death and migration need to be taken into account, in addition to changes in
activity rates (such as retirement and other exits from the workforce). New
entrants, migration, re-entrants to the workforce, as well as inter-occupational
mobility, constitute other flows in the supply of skilled workers.
In addition, skills projections should take account of the likely responses of
the labour force to various skill scarcity signals. If the long-term projections
were to ignore the spontaneous take up of specific skills (through schooling,
adult education or training) and were to project the current skill mix into the
future, the supply of skills will likely be grossly miscalculated and could lead
to seriously misleading conclusions about likely future skill imbalances.
10
Borghans et al. (1996) argue that the relationship between skills and occupation is no longer unique and
the substitution process between different types of education is becoming more important, and plays a
role in the adjustment of the labour market. See also Plassard and Pluchard (1997) for a discussion on
the changing structure of education per occupation.
11
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
5. What type of skills to forecast? Generic skills, qualifications, etc.
Knowledge of the changing demand for qualifications can be very useful
within an educational planning context. However, it is less clear whether
qualifications are a better measure of skills than occupations.
In this regard, some models have developed projections of demand for and
supply of qualifications taking into consideration the existing level of
qualifications and skills. In particular, efforts have been made to identify
generic and specific skills.
Generic skills are a set of skills that are not precisely defined and often merge
into personal attributes. Undoubtedly some generic skills, such as those
associated with IT, have become increasingly important in recent years.
However, a range of other skills such as flexibility, ability to communicate, to
work in teams, and problem solving skills have all been emphasised in the
context of the continuing pressure to innovate and compete.
Using a methodology similar to the one adopted to link the demand for
qualifications to occupations, it is possible to generate projections of generic
skill needs from the basic occupational projections. The key additional input
requirement is an estimate of the importance of the various generic skills
within a given occupation.
One important issue is the extent to which occupational change is a good
proxy for expected skill change, more particularly in the case of specific skills
forecasts. This might be of significant importance as, in recent decades,
occupational shifts have made it hard to match occupations with certain
specific skills. This problem exists because classification systems are,
inevitably, simplifications that become out of date over time.
6. Surveys
Surveys of households’ and employers’ views and perceptions can provide
additional information that can be used to either provide additional
information and data for the projections and/or test the outputs of modelbased skills forecasts. To the extent possible, such surveys should be
undertaken in parallel with the model-based projection exercise, to provide
an independent source of information and validation to the users of skillsforecasts.
Nevertheless, while such surveys are very useful, it is important to remember
that responses in such surveys tend to reflect mainly perceptions of current
circumstances or those that are expected to prevail in the very near future.
Therefore, some degree of caution is advisable in using such survey results
over longer time horizons.
12
London Economics
January 2002
Chapter 2
2.7
Skills Forecasting: Theory and Practice
Occupational and skills forecasting – the
practice
The key theoretical building blocks of any occupational and skills projection
are summarised in Figure 1. This section, which summarises the approaches
used to develop occupational and skill forecasts in the UK, Germany,
Netherlands, France, Ireland11, Finland, US, Canada and Australia, shows
clearly that the best-case approach illustrated in Figure 1 is adopted to
varying degrees by these countries.
Our main findings are summarised in a set of tables below, and a more
comprehensive review of each country’s approach is provided in Annex 1 .
The main characteristics of the approach taken by each of the nine countries
are described in Table 2.1. The methodology underpinning the forecasts is set
out in Table 2.2. Finally, Table 2.3 provides more detailed information on the
labour supply dynamics built into the labour market models, such as
demographic factors, work replacement, and technological change.
q
Until the early 1990s, manpower forecasting focused almost entirely on
modelling and projecting future labour demand, with less attention being
paid to modelling labour supply and replacement demand. This reflected
earlier criticisms that projected labour market disequilibria were difficult
to interpret because potential substitution between different categories of
skilled labour was being ignored. Moreover, labour supply modelling
was hampered by the lack of good data on labour market inflows and
outflows at a detailed level, and because of difficulties with the modelling
of occupational mobility. Through the 1990s, however, increasing efforts
have been made to better understand and model the labour supply
dynamics.
q
Some countries focus on occupational projections while others generate
forecasts for various qualification or educational groupings.
Nevertheless, there seems to exist a common view that the links between
training and employment are complex and that it is difficult for a single
institution, national, regional or local, to act alone on this issue. Thus,
inter-institutional groups have been set up in many countries to further
the understanding of the skills issue and, in a number of cases, to develop
active policies.
q
Moreover, quantitative forecasts are often complemented by interviews
and surveys of industrial leaders and employment experts.
11
Campos et al. (1999b). A survey on education forecasting for some Western countries can be found in
National Observatory of Vocational Training and Labour Market (1999) or Henson and Newton (1995).
13
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
Figure 1: Occupational and skills forecasting: the key elements
Graduates and
school leavers.
Macroeconomic context.
Growth, Inflation and
general labour market
prospects projected.
Output projections by
sector and industry.
Changing patterns of
employment.
Demographic factors
Work replacement
Participation rates
Occupational
by sector.
projections
Demand for skills and
qualifications for
occupations.
Generic Skills.
Specific Skills
Labour force surveys
Skill prospects
New supply of skilled
workers per type of
education.
Source: London Economics
14
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
Assessments of occupational forecasts in France, Canada, the US, and the
Netherlands, have shown that the average forecast error increases as the
number of occupational groups increases. This is one reason why some
countries focus on only a relatively small number of occupational groups in
their forecasts. A similar observation for forecasts of labour supply was made
by Parnes (1962) who recommended that estimates of future supplies of
manpower should be made in terms of categories of educational
qualifications rather than in terms of specific occupations12.
q
While quantitative projections are useful, they are not the only instrument
that should be used in assessing future occupational and educational
requirements. Qualitative survey-based assessments, such as those used
for the Netherlands, Finland and France (for the metallurgy sector), may
also be very useful.
q
Scenario-based analysis is regularly used in a few countries to provide a
richer occupational or skills outlook. For example, the French regional
employment authority (OREF) carries out a “situation diagnosis”
identifying key factors of change and estimating the probability that these
factors will continue to be an influence. From such an assessment, it is
then possible to develop alternative scenarios, which can be further
enriched by qualitative observations about future labour market
developments, jobs, and professional skills requirements.
q
In many countries, problems with data availability and/or quality
prevent the construction of robust estimates and projections of labour
supply at a regional and local level. A key issue at the regional and local
level is the inter-regional mobility of workers that may not be adequately
captured by the available data. In this regard it is useful to note that local
surveys and interviews with local employment experts have often been
proven to be useful additional sources of information.
12
See quote in Willems (1996).
15
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
Table 2.1: Main characteristics of the forecasts
COUNTRY
Purpose
Main users (policy,
career planners)
Organisation(s) in
charge
Main Data used
Regional forecasts
Use of forecasts
UK
GERMANY
FRANCE
NETHERLAND
IRELAND
Improve labour market
information on skills
supply and demand.
Provision of a skills public
information database,
Skills base.
National planning,
Regional Development
Agencies, Learning and
Skills Councils, National
Training Organisations
and others.
Department for Education
and Skills (DfES)
Forecast expected demand
and supply of manpower
classified by education and
occupation.
Analysis of medium term
occupation trends and
explore the implications of
alternative scenarios for
education and training
needs.
Ministry of Employment
and Education.
Trade unions and other
organisations.
Track all the relevant flows
which determine the
supply and demand of
educational and
occupational labour
markets
Ministries, Career advisors
and school-leavers.
National Careers Guidance
Information Centre.
Provide information on the
changing pattern of
occupations and identify
possible changes in future
skill requirements.
Governmental decisionmaking -- strategy for
employment policy and
work reform. Qualitative
interpretation.
Provide occupationalbased employment
projections.
Ministry of Employment
and Education. Trade
unions.
Ministries.
Career guidance
Federal government
Ministry of Education.
Regional Employment and
Training Observatories.
Ministry of Education and
Science.
National Employment and
Training Authority (FAS)
Department of Labour.
Bureau of Labour Statistics
Labour Force Survey,
General Household
Survey, Skills Task Force
Employers Skills Survey,
others
12 Regions in UK
Microcenzus (census of
establishments).
Employment census,
employment surveys,
others
Labour Force Surveys,
Forecasts of School-leavers,
Educational Accounts, own
surveys.
Labour Force Surveys,
Census of Population.
Ministry of Labour, Social
Services and Health,
Industry and Trade,
provincial governments
and educational institutes.
Labour Force Survey
No
Not provided regularly.
There have been some
studies of employment
training diagnosis for some
areas.
No
Yes
There are also independent
sectoral models.
Every State decides
By province
Results for States, regions
and occupations are
derived by tops-down
disaggregations.
Quantitative forecasts.
No use of alternative
scenarios.
Quantitative forecasts.
Use of alternative
scenarios.
Quantitative forecasts.
Alternative scenarios used.
Limburg region only.
Region’s data on
employment are related to
the national model
estimates jointly with
regional unemployment
and vacancy data.
Point estimates of risk
indicators accompanied by
5-point scale qualitative
characterisation.
The forecast is scored
against an alternative
scenario of no growth
Quantitative forecasts.
No use of alternative
scenarios.
Qualitatively.
No use of alternative
scenarios.
Quantitative forecasts.
No use of alternative
scenarios.
Quantitative forecasts.
No use of alternative
scenarios.
Quantitative forecasts.
No use of alternative
scenarios.
State and schools
(administration, students,
parents).
FINLAND
US
Occupational Employment
Statistics, Current
Population Survey, Census
of Population.
CANADA
AUS
Provide latest information
about trends, future labour
market conditions,
information on all
occupational groups,
prospects for finding a job.
Career guidance
Development of national
policy, strategic plan for
vocational education and
training, as well as State
training profiles.
Human Resources
Development Canada.
Department of
Employment. Job Futures
2000
Census of Canada,
Australian National
Training Authority
National policy. Career
guidance.
Population censuses,
Labour Force Surveys.
16
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
Table 2.2: Methodology for supply and demand forecasts
COUNTRY
Demand
UK
GERMANY
NETHERLAND
FRANCE
IRELAND
FINLAND
Manpower Requirement.
Expert assessment.
Scenario Analysis
Expansion demand for 29
sectors.
Manpower Requirement
Manpower Requirement.
Influence from
representatives of sectors.
Short and long term trends
in labour force.
17 Sectors and 47 subsectors.
Manpower
Requirement
Manpower Requirement
Dynamic general
equilibrium model.
Industry employment
projections developed
from economic
indicators.
Projections for
employment by 158
industries and region.
Expansion demand for 29
sectors and 45
occupational sub-groups.
Employment 14
professional categories
by 36 sectors.
Forecasts of professional
categories are
disaggregated for 10
levels of education and 5
levels of diploma.
Expansion demand for 29
sectors and 45
occupational sub-groups.
(Reports presented for 14
occupations and 13
industries).
Not given. The
educational
qualifications can be
derived for broadly
defined levels of
education.
Occupational structure: 11
main groups and 48 subgroups.
Projections made for
more than 500
occupations and 236
industries.
Demand for labour force
divided by educational
fields and levels through a
special statistical key.
Not provided.
Each State makes its
own surveys.
Given expected production
levels the industrial
employment requirement is
derived assuming CobbDouglas production
functions.
Using each industry’s
employment-occupational
structure, industrial
employment is broken
down into occupational
categories.
No skills forecasts.
Projections for occupations
used for identifying skill
shortages and general
trends of the labour market.
Expected supply of new
entrants. 104 types of
education.
Outflow from additional
training courses.
Short-term
unemployment.
No
Student quotas included.
Number of students is
modified to the number of
qualifications using a ratio
of graduation and dropout,
by occupational group.
No
Enrolment rates and
graduates rates are
calculated for 6 levels of
schooling by age and
gender groupings.
Graduate rates are broken
down into major fields of
study by applying the most
recent observed shares for
each field of study.
No. Priority research
area.
Forecasting
method
Multisectoral Dynamic
Macroeconomic Model.
Manpower Requirement.
Qualitative judgement
Manpower Requirement
Social demand models
How are sectoral
employment
forecasts made
49 industries, although 17
used for presentational
purposes.
Projections for 17 sectors:
taken into consideration
different scenarios.
Expansion demand for 13
sectors
How are
occupational
forecasts made
Industrial and regional
employment projections
are disaggregated into 25
occupational categories for
each industry.
Sectoral figures broken
down by 34 occupational
tasks.
Sectoral figures broken
down by 127
occupational tasks.
How are
qualification
and/or skills
forecasts made
Employment of highly
qualified persons at 9
levels. Only national level.
Breakdown of activity
forecast by levels of
qualification within each
activity.
Forecasts of demand for
skills provided at 11 skill
levels.
Projections given for 104
types of education.
Stock flow approach
coupled with analysis of
activity rates.
9 Levels of high
qualifications and 6 broad
levels of qualifications.
Low-level qualifications
based on extrapolations of
information from the
Labour Force Survey and
covers regions.
12 categories
Forecasts of the schoolleavers for 5 diploma
categories.
Medium-term projections
of employment are used
to generate projections
for 11 industries and 29
sectors.
US
CANADA
AUS
Disaggregation to 340
occupations.
Forecasts of
occupational demand
changes.
Supply
Projections
graduates
of
17
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
Table 2.3: Labour demand and supply dynamics
COUNTRY
Demand
UK
GERMANY
NETHERLAND
Demographic
factors
Age and gender
structure, mortality
Yes. Social factors and
others
Yes. Female labour
participation.
Work
replacement
Outflows due to
retirement, interoccupation mobility,
family formation.
No consideration of
labour market flows.
Changing
patterns
The occupational model
is calibrated to modify
projections to include
changes. Professional
judgement based in
current development is
used.
Replacement demand
also included
Expert identification of
effects and impacts on
branches of activity
(technological,
economic, social and
other factors).
Extrapolation of future
impacts (qualitative
assessment).
Employment trend
modified to account for
influencing factors.
Replacement demand
forecasts for 127
occupations and 104
types of education
Substitution effects by
type of education also
considered.
Allowance for any shifts
that may be occurring in
the educational
structure of
occupational classes.
Flow of new qualifiers,
mortality and migration
rates.
Migration and potential
activity rates of married
females.
FRANCE
IRELAND
FINLAND
US
CANADA
AUS
Yes. Geographic and
mobility model provides
hiring needs for each
professional category.
Use of data concerning
demography, training,
employment turnover,
job seekers.
Female participation.
Age groups, retirement.
No
Yes. Mortality
No
No
Yes
Yes, net replacements
are considered.
Only withdrawals from
the labour force
(retirement, women’s
birth and child rearing,
re-education
programmes…).
Replacement Demand
will be incorporated in
the future.
Impact analysis of
professional movement
and replacement
demand.
Identification of
development and
changing factors, and
the probability of their
continuation or
cessation.
Forecasts are prepared
according to four
different scenarios to
allow different
promotion, employment
rate or industrial
changes.
Analysis of past patterns
of occupational change
and identification of
influencing factors and
relative importance to
different industries and
occupations.
The trend in the share of
each occupation in each
sub-sector is projected
after using judgement
and decision rules.
Outsiders of the labour
force are included:
retirements, students
and persons at home.
Education and
Industrial policy is
considered.
Outflows (retirement,
death and disablement)
based on past trends
and international
comparisons.
It is left to company
managers, training
programmes and
government to keep a
watch over changes in
the content of existing
jobs arising from new
technologies and new
methods of quality
improvement.
Occupational structure
or each industry’s
employment is broken
down using fixed
coefficients (from
Census), and variable
coefficients to allow
variations across
different census
Dynamic processes at a
disaggregated level are
incorporated.
Forecast of re-entrants
to the labour market.
104 types of education
No
Demographic data by
age group, retirement,
work at home and parttime work.
No
Allowance is made for
“non-completers”
No
Supply
Labour dynamics
(demographics
and replacement)
Use of trainee
population and career
started surveys.
18
London Economics
January 2002
Chapter 2
2.8
Skills Forecasting: Theory and Practice
Future Skills Wales 1998
The 1998 Future Skills Wales13 project provided a very thorough surveybased assessment of current generic skill requirements and supplies, and a
detailed model-based forecast of likely future occupational changes and
generic skills required by the various occupations. The latter projection
combined the forecasted occupational changes with an employer survey of
likely future important skills.
Careful attention was given to account for turnover in various occupations
(e.g., the focus was on gross changes rather net changes) and some labour
supply dynamics were incorporated.
However, the 1998 project used only one single projection and the process
could have provided more insightful information if alternative scenarios had
been produced using different economic growth assumptions for example.
Moreover, the labour supply dynamics built into the forecast appear to focus
only on flows into and out of the labour force and do not capture the uptake
of skills by the labour force and, therefore, do not provide a good perspective
on the skills that are likely to be supplied in the future. As such, the exercise
provided a better perspective on the future demand for various skills than on
future skill shortages.
Finally, projections were provided for a large number of occupations (23)
across 28 industries. But, readers of the 1998 FSW report were not warned
that, because of the small sample size of these categories and the inherent
forecasting uncertainty, the forecast errors (or uncertainty range) for specific
occupations or skills could be potentially significant and that, therefore, some
degree of prudence was required in interpreting the results.
Overall, the approach taken in 1998 strived to follow to a large extent the
best-practice approach discussed earlier. But, it can be improved on in a
number of areas – better labour supply dynamics, alternative scenarios, etc.
for the proposed 2003 FSW exercise.
2.9
q
13
Concluding Remarks
This chapter’s review of the literature on skills forecasting and the actual
practice in a number of OECD countries has shown that this is a complex
See Future Skills Wales (1999).
19
London Economics
January 2002
Chapter 2
Skills Forecasting: Theory and Practice
exercise involving many different elements: macroeconomic forecasting,
employment forecasting, occupational projections taking into account the
dynamics of the labour market, skill requirement projections, labour
supply and skill supply projections.
q
Uncertainty prevails at each step of such a forecasting exercise. It is
interesting to note here that the U.S. Bureau of Labor Statistics is of the
view that any occupational forecast percentage error falling within the
range of plus or minus 10% of the actual outcome is reasonable and
acceptable.
q
Some steps -- such as complementary surveys and alternative scenarios -can be taken to reduce somewhat the uncertainty but cannot totally
eliminate it.
q
Moreover, as noted in the discussion above, the forecast error tends to
increase with the level of disaggregation. Therefore, it is advisable to
maintain the focus of the skills assessment exercise as much as possible at
a relatively aggregated regional, industrial and skills level.
q
Finally, its is important to remember that forecast errors of skill gaps, and
hence the uncertainty around any point estimate of such gaps, will likely
be much larger than those of skills requirements, as the projected gap is
the difference between projected demand and projected supply. This is
illustrated by the following example.
§
Assume that the skills demand forecast is 100 and the
skills supply forecast is 95, and that the typical forecast
error is +/-5%. This implies that the “true” skills
demand is likely to fall in the range of 95 to 105 and the
“true” skills supply will fall in the range of 90.25 to
99.75.
§
The projected skill gap is 5, but the uncertainty
surrounding the supply and demand projections imply
that in fact the “true value” of the skill gap estimate lies
in the range of –4.75 (excess supply) to 14.75 (excess
demand).
20
London Economics
January 2002
Chapter 3
3
Regional Data Issues
Regional Data Issues
Before reviewing the regional models and their forecasting performance, it is
essential to review first the quality of the data used in the regional forecasts
as this has a direct impact on the quality of the forecasts. Therefore, in this
chapter we discuss in greater detail some of the data issues of particular
relevance to forecasting future Welsh economic developments.
3.1
Introductory Observations
As the issue of availability, reliability and timeliness of economic data for
Wales has already been extensively addressed in the recent report of the
Office of National Statistics (ONS)14, it suffices to note here that the quality
and lack of timeliness of such regional data presents a major challenge for
forecasters of regional economic activity.
Typically, forecasters have to generate not only estimates of the likely future
values of various economic indicators, but must also ’guess’ the starting point
of the forecast, e.g., the value of various economic variables in the current
period, and occasionally the period just passed, for which data will only be
released in the future.
At the national level, where quarterly data and forecasts are the norm, this
does not generally create major problems. In the case of regional forecasts,
however, the late release of data and the frequently substantial revisions
thereafter create considerable uncertainty about where regional economies
actually stand at the beginning of the forecast period.
Starting from an incorrect conjunctural assessment may have particular
implications for the robustness of the forecast. In some cases, it could result
in serious misinterpretation about where the regional economy is positioned
in relation to the business cycle and lead to misjudgements regarding
adjustment dynamics.
3.2
Problems with the regional GDP data
This issue of poor data availability is particularly acute in the case of output
data - GDP/GVA15 data - which, to compound the forecaster’s difficulties, are
14
See ONS (2000).
15
GDP = gross domestic product and GVA = gross value added. Following the introduction of the
European System of Accounts 1995 (ESA95), the latter economic indicator is now the standard output
measure. Since the introduction of the new system, GVA measures output net of taxes and subsidies
21
London Economics
January 2002
Chapter 3
Regional Data Issues
officially published only on an annual basis and only in nominal terms
(current prices). These data then need to be converted by the forecasters into
real, or volume, terms (constant prices) to generate a picture of real growth
developments at regional level.
To illustrate the problems with the quality of the GDP data, Table 3.1 below
shows how the published figures of total GDP (in current prices) in 1990 and
1995 for the UK and Wales have evolved over time.
The most striking fact to note is that, in contrast to the national data, the
regional GDP data never seem to stabilize, moving around with almost every
release of new annual regional data. As a result, the actual growth rate of
GDP is also much more frequently revised at the Welsh level than at the
national level. Moreover, the amplitude of the revisions to the annual growth
rate of GDP is also much larger at the regional level.
on products and GDP measures output at market prices, i.e., inclusive of taxes and subsidies on
products.
22
London Economics
January 2002
Chapter 3
Regional Data Issues
Table 3.1: Analysis of Evolution of UK and Welsh GDP Data - 1990 and 1995
1990
1995
Overall UK
Wales
Overall UK
Wales
Level
published in
1991=100
Annual
growth rate
of GDP in
1990
Level
published in
1991=100
Annual
growth rate
of GDP in
1990
Level
published in
1997=100
Annual
growth rate
of GDP in
1995
Level
published in
1997=100
Annual
growth rate
of GDP in
1997
Nov. 91
100.0
9.5
100.0
8.7
-
-
-
-
Dec. 92
100.4
8.7
101.1
8.5
-
-
-
-
Dec. 93
100.2
8.4
100.1
8.2
-
-
-
-
Dec. 94
100.2
8.4
99.9
7.7
-
-
-
-
Dec. 95
100.2
8.4
101.0
7.5
-
-
-
-
Jan. 97
100.2
8.4
101.1
7.8
100
4.3
100.0
3.1
Jan. 98
100.2
8.4
101.3
7.8
100.6
4.8
101.9
5.5
Mar. 99
101.7
8.5
99.4
6.8
102.4
5.0
101.9
5.8
-------------
-------------
-------------
-------------
-------------
-------------
-------------
-------------
Aug. 00
104.6
n.a.
101.5
n.a.
104.9
4.9
105.1
6.0
Mar. 01
105.0
8.6
101.6
7.2
105.2
4.9
105.6
6.2
Publication date in
Economic Trends
Shift to ESA95
Source: Economic Trends (various issues).
Notes:
1. Regional GDP data are first made available in press release and are then published in the subsequent monthly issue of Economic
Trends.
2. The shift to ESA95 in the data published in the August 2000 issue of Economic Trends explains the jump in the index in 2000.
3. The UK data include the extra-regio.
n.a. not available.
As a further illustration of the variability of the regional GDP data, Table 3.2
indicates that the annual GDP growth for each year’s GDP since 1990 has
changed with successive releases of new data.
Particularly noteworthy are the large revisions from the release of the initial
(or provisional) estimates to the first subsequent release of the data. The
difference, in absolute terms, averages 1.6 percentage point from 1990 to 1998,
23
London Economics
January 2002
Chapter 3
Regional Data Issues
reflecting a rather high degree of uncertainty about the true state of the
economy in the initial data release16.
Moreover, almost all years are characterized by on-going large volatility with
the standard deviation of the published annual growth rates for a given year
ranging from a low of 0.4 percentage point in 1996 to a high of 1.1 in 1995.
Table 3.2: Evolution of Annual Welsh GDP Growth - 1990-1999
Annual growth rate of GDP (in %)
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
Publication date in
Economic Trends
Nov. 91
8.7
Dec. 92
8.5
4.9
Dec. 93
8.2
3.8
4.6
Dec. 94
7.7
4.1
1.9
7.1
Dec. 95
7.5
3.3
1.9
4.4
8.3
Mar. 97
7.8
3.1
2.5
4.2
6.9
3.1
Jan. 98
7.8
3.0
2.4
4.2
6.4
5.7
3.6
Mar. 99
6.8
4.4
2.8
4.8
5.7
5.8
4.7
5.3
Aug. 00
n.a.
5.7
3.0
4.7
5.2
6.0
4.0
3.8
4.0
Mar. 01
7.2
5.7
2.8
4.8
5.5
6.2
4.0
3.7
5.5
3.9
Standard
deviation of the
published annual
growth rate
0.6
1.0
0.8
0.9
1.0
1.1
0.4
0.7
--
--
Source: Economic Trends (various issues)
n.a. not available.
16
In this regard it is interesting to note that, due to data source problems, the 1997 to 1999 data are still
characterized as provisional in the 2001 release of the regional data.
24
London Economics
January 2002
Chapter 3
Regional Data Issues
Table 3.3: Changes Over Time in Sectoral Distribution of 1990 and 1995 Welsh GDP
Share of sectoral GDP (in %) in total GDP in 1990
regional accounts as published in
Nov. 91
Dec. 92
Dec. 93
Dec. 94
Mar. 01
Agriculture, hunting, forestry and fishing
2.1
2.2
2.3
2.3
2.5
Mining, quarrying inc. oil and gas extraction
n.a.
n.a.
0.9
0.9
1.2
Manufacturing
27.3
26.9
29.5
29.5
30.1
Electricity, gas and water supply
n.a.
n.a.
3.1
3.1
3.0
Construction
7.3
7.4
7.2
7.2
6.8
Wholesale and retail trade
13.8
14.1
14.0
14.0
12.7
Transport and communication
6.4
6.1
7.0
7.0
7.3
Financial intermediation
17.0
16.5
15.6
15.3
15.9
Public administration and defence
n.a.
n.a.
8.0
8.0
6.9
Education, health and social work
n.a.
n.a.
10.3
10.5
12.4
Other services
5.7
5.7
5.3
5.3
3.9
Jan. 97
Jan. 98
Mar. 99
Aug. 00
Mar. 01
Agriculture, hunting, forestry and fishing
2.5
n.a.
2.1
1.8
1.8
Mining, quarrying inc. oil and gas extraction
0.8
n.a.
1.2
0.9
1.0
Manufacturing
28.2
n.a.
29.7
28.4
28.3
Electricity, gas and water supply
3.5
n.a.
3.0
3.3
3.2
Construction
5.3
n.a.
5.4
5.4
5.4
Wholesale and retail trade
13.3
n.a.
12.7
12.9
12.8
Transport and communication
6.5
n.a.
6.2
6.3
6.4
Financial intermediation
18.3
n.a.
17.0
17.9
18.0
Public administration and defence
7.4
n.a.
6.9
7.0
7.0
Education, health and social work
14.2
n.a.
14.3
14.5
14.6
Other services
3.4
n.a.
4.0
4.0
4.0
Share of sectoral GDP (in %) in total GDP in 1995
regional accounts as published in
Source: Economic Trends (various issues)
n.a. not available.
25
London Economics
January 2002
Chapter 3
Regional Data Issues
The problem is even more acute when one focuses on sectoral GDP figures.
This can be seen very clearly from Table 3.3, which compares, by year of
publication of the data, the shares of the various industries in total
GDP/GVA in 1990 and 1995.
For example, manufacturing’s share of 1990 GDP was initially estimated at
27.3%. It then fell to 26.9% in the subsequent release of regional accounts and
rebounded to 29.5% in the regional accounts released two years after the
initial release.
Another example of the volatility of the sectoral data is the evolution of
manufacturing’s 1995 share in total GDP that was initially put at 28.2%. In
the 1997 regional accounts, the 1995 share grew to 29.7%. But, in the 1998
regional accounts the same share dropped back to 28.4%. A similar pattern,
but of opposite sign, is observed for the share of financial intermediation in
the 1995 GDP.
Overall, sectoral GDP data for a given year exhibit considerable volatility and
do not appear to stabilize rapidly. That being said, it is interesting to note
that the passage to ESA95 did not result in drastic changes. It simply seemed
to extend the on-going trend of continued revisions to sectoral GDP data.
Such volatility in itself is not generally a major problem. But, in the context of
occupational and skills forecasts that depend crucially on sectoral output and
employment projections, it implies that the starting points of the sectoral
projections are uncertain and may affect the actual projected level.
3.3
Other Regional Data Issues
Obviously, the regional GDP data are not the only data to be affected by
quality problems. The quality issue of the Labour Force Survey is already
being addressed by the special funding provided by the Welsh National
Assembly for an increase in the sample size of the survey in Wales. Yet, even
this sample top-up will not eliminate all the uncertainty about the precise
state of the labour market in Wales.
For example, it is interesting to note that, at the present time, the lowest
employment sampling variability17 at the regional level, e.g. that for the South
East, is still of the order of +/- 0.9 per cent18 while the reported sampling
17
LFS data are based on statistical samples and, if many samples were drawn, each would give a different
result. The sampling variability represents the upper and lower bounds of a range within which 95 per
cent of the samples would contain the true value.
18
In practical terms, this means that National Statistics is confident that, in general, the “true” employment
figure will lie in the range of +/- 0.9% of the published employment figure. More precisely, this will
be the case for 95 out of every 100 samples used to generate the employment data. The sampling
variability figures reported in this study are from the November 2001 Labour Market Statistics.
26
London Economics
January 2002
Chapter 3
Regional Data Issues
variability of +/- 1.7% for Wales is the second highest of the eleven
government regions19.
The detailed breakdown of total regional employment (by industrial activity
or occupational classification, or at a sub-regional level) is subject to even
greater sampling variability as sampling variability is generally inversely
related to the actual sample size.
Unfortunately, information on the sampling error of employment by industry
in the Annual Business Inquiry (ABI) is not yet publicly available. But, in the
case of Wales, the data on the sampling error for employment at the local
market level shows a strong negative relationship between the size of
employment at the local level and the size of the sampling error20. In fact, the
sampling error varies between 2.2% (of total employment) in Cardiff to 8.3%
in Blaeneau Gwent and averages 5.6% over the 22 unitary counties (see Table
3.4)21.
The regional forecasts services are also very concerned about the quality of
some of the regional data. In particular, concerns were expressed about the
quality of the regional data on consumer spending, incomes, average
earnings, wages and salaries and employment. The ABI data are generally
used by regional forecasters to complement the information provided by the
LFS data.
Among the desired regional data improvements were more
detailed information on investment by sector and better regional migration
data.
19
The highest sampling variability, +/- 1.8%, is reported for the North East.
20
The correlation coefficient between the size of employment at the local level and the size of the sampling
error is -0.85.
21
Finally, it may be useful to note that, despite the lower quality of the sub-regional data, regional
forecasting services will generally produce sub-regional labour market forecasts, using a number of
labour market data sources, depending on the specific needs at hand. For a good overview of sources
of regional and sub-regional labour market data, see National Statistics (1999).
27
London Economics
January 2002
Chapter 3
Regional Data Issues
Table 3.4: Sampling Error of Employment Survey at Local Level22
Total Employees
Sampling Error
(in % of employment)
Wales (July-September 2001)
1,252,000
2.20
Blaeneau Gwent
19,840
8.34
Bridgend
47,432
4.39
Caerphilly
44,727
5.53
Cardiff
151,828
2.20
Carmarthenshire
45,258
5.32
Ceredigion
22,287
7.73
Conwy
32,441
7.53
Denbighshire
31,616
6.33
Flintshire
57,458
4.40
Gwynedd
37,776
5.21
Isle of Anglesey
16,605
7.12
Merthyr Tydfil
15,972
6.74
Monmouthshire
33,828
6.81
Neath Port Talbot
43,186
4.85
Newport
71,776
3.97
Pembrokeshire
30,797
7.02
Powys
49,015
5.96
Rhondda, Cynon, Taff
71,004
4.39
Swansea
86,889
3.19
Torfean
37,077
5.29
The Vale of Glamorgan
40,651
5.01
Wrexham
50,153
5.20
Source: National Statistics
22
The information is provided by National Statistics for the 1998 Annual Business Inquiry.
28
London Economics
January 2002
Chapter 3
3.4
Regional Data Issues
Concluding Remarks
The key messages to remember from this brief review of some of the data
problems faced by regional forecasters are that:
q
There is considerable uncertainty about the true value of key economic
variables at the starting point of the forecast.
q
The uncertainty increases as one moves down from national, highly
aggregated data to disaggregated regional data.
q
The degree of uncertainty is not eliminated entirely with successive
releases of a given year’s data.
q
The data improvement process launched by the National Assembly for
Wales should eventually contribute to alleviate some of the problems
faced by regional forecasters. However, the reality is that over the
foreseeable future these gains will most likely be limited23.
q
Over the short to medium-term future, the starting point of any forecast
of the Welsh economy is likely to continue to be clouded by considerable
uncertainty about the precise state of the economy.
23
The major exception is the Welsh components of the Labour Force Survey (LFS) and Annual Business
Inquiry (ABI) where the sample top-up financed by the National Assembly should significantly
improve the quality of the data.
29
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
4
Regional Economic Forecasting Models
4.1
Introductory Observations
The forecasting performance of the three forecasting services is reviewed in
this chapter. This will provide background information to improving the
potential use of forecasts by FSW Partners and a valuable input to the
decision process on how best to look at Wales’ future skill needs.
In section 4.2, we highlight the key characteristics of the regional economy
models used by the three leading forecasting organisations. We then examine
the properties of these models in section 4.3 and review the forecasting
performance in section 4.4. Some concluding remarks are offered in section
4.5.
4.2
Comparison of Regional Economy Model
Structures
Forecasting models can be compared along a multitude of dimensions.
However, not all aspects of model structure are of equal importance,
especially for the purpose of the current study.
In fact, we will focus only on three key dimensions critical for a long-term
skills forecasting exercise:
1. The interaction between the regional and the national model;
2. The long-run equilibrium of the model; and,
3. The availability of detailed employment and occupational data at the
regional and sub-regional level.
First, of note is the fact that none of the three major models currently
available is a pure regional model completely independent of any national
model or forecast.
Yet, the manner in which the regional model(s) interact(s) with the national
model(s) varies significantly, ranging from the top-down approach used by
Oxford Economic Forecasting/NIERC to the interactive effects between the
regional projections and the national forecast (Cambridge Econometrics).
Second, the definition of the long-run equilibrium (steady state) differs across
models. This may result in different medium and long-term output and
30
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
employment projections as the projected path followed by the economy from
the starting point of the projection to the long-run equilibrium will itself be
heavily influenced by the definition of the model’s assumed long-run
equilibrium.
Third, the level of regional projection detail varies markedly, with some
forecasting services providing considerably more detailed industrial and
employment information, particularly at the sub-regional level, than others.
The following sections review in greater detail the forecasting models of the
three forecasting organisations in terms of the three key characteristics listed
above.
4.2.1
Business Strategies Ltd. Regional Model24
A. Interaction between regional and national model
The regional model is a macro-economic style model of the UK regions where
demand is the key driver in the short run. Forecasts from the Regional Model
are usually constrained to line up with the UK numbers from the macromodel. In practice, the regional model is solved after the macro model.
Depending on the industrial sector, regional output is modelled either: 1) by
using straight share equations and treating UK output in that sector as
exogenous; 2) shift-share equations with regional output being a function of
UK exports and domestic regional demand; or 3) equations where the weight
on regional demand versus UK-wide demand is a function of the
transportability of the product or service.
Regional demand is generally modelled as a function of regional factors and
regional employment is a function of, among others, the region’s real wages
relative to those of the UK. Regional investment is typically a function of
regional output shares except in manufacturing where the region’s average
earnings relative to the UK average also play a role.
B. Long-run anchor
In the BSL national and regional models, an exogenous anchor is provided by
the imposition of a view on long-term national and regional supply
capacities. These determine the output level towards which the national and
regional economies will converge over the longer run25.
24
See Business Strategies Ltd (2000).
25
The speed at which the economy converges towards the long-run equilibrium will vary from model to
model. In the absence of a battery of simulations of various policy and economic shocks, it is
impossible to ascertain from the publicly available information how long it will take for the long-run
properties of the model to anchor a medium- or long-term projection.
31
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
The regional supply constraint is derived judgementally and in consultation
with regional analysts, and is based on a number of key variables such as the
region’s share of national manufacturing employment, the region’s share of
national private service employment, the region’s share of national mainly
public service employment and relative personal savings rates.
C. Availability of detailed employment/skills forecasts
BSL can provide forecasts at the regional level of employment for 28
industries. Sub-regional, county level, forecasts for output, employment and
incomes are also produced.
But these forecasts are essentially driven by shift-share equations of the
regional totals for output that then determines employment. Sub-regional
economic projections are constrained to add up the regional forecast.
Finally, the employment forecasts can be mapped into an occupation forecast
for 22 sub-major occupations for each of the 28 industries. The occupational
mapping is based on an extrapolation of the 1991 Census of Population
results and the post-census historical data are constrained to sum to the
figures of the Labour Force Survey.
Work is underway to relate occupational changes to evolving skill needs,
focussing not only on the demand for skills but also on the endogenous
reaction of the labour force to signals about the relative desirability of various
skills.
4.2.2
Cambridge Econometrics26
A. Interaction between regional and national model
The Cambridge Multisectoral Dynamic Model (MDM95) is a regionalised
energy-environment-economy model of the UK economy. It is a very largescale model combining time series and input-output analysis and is built
from the bottom up, distinguishing 49 industries and 51 categories of
household expenditure.
UK regions are the key geographical building blocks of the model and the
regional models are solved simultaneously. This allows for full feedback
from the regional economies to the UK economy. In practice, however, a
feedback effect is operative only in the case of the employment forecast and, if
necessary, regional projections are adjusted to sum up to the UK forecast27.
26
See Cambridge Econometrics (2000).
27
The model software, however, allows for many more feedbacks from the regional level to the national
level.
32
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
Regional output at the industry level is determined by final and intermediate
demand that comes from the region, the rest of the UK or abroad.
Intermediate demand is determined through the input-output matrix. Final
demand is the sum of regional household demand and regional investment.
Both depend mostly on a set of regional variables. Regional imports are
determined residually through the input-output table.
Finally, regional employment is derived by relating employment in a given
industry in the region to its output, to the regional wage relative to the
national wage and a few national variables (hours worked, etc).
B. Long-run anchor
Many of the time-series, behavioural equations are estimated using an
econometric technique that distinguishes the stable long-run relationship
(through a co-integrating relationship) from short-run dynamics (through an
error correction model). Some form of a long-term anchor is provided by this
approach as, over the longer run, only the stable long-run relationships are
operational.
C. Availability of detailed employment/skills forecasts
The regular forecasts produced by Cambridge Econometrics provide
information on output, employment, etc. for 30 regional industries in Wales
and 11 other regions.
Additional sub-regional information can be obtained from the Local Economy
Forecast Model (LEFM) that has been developed in collaboration with the
Institute for Employment Research at the University of Warwick.
This additional simulation and forecasting provides information on output,
employment, etc. at the county level for 49 industries. Employment
information is provided by gender, status and occupation. The key fact to
note, however, is that the sub-regional forecasts are driven and constrained
by the regional forecast.
Cambridge Econometrics expects to be soon in a position to add an additional
module to LEFM that will forecast future requirements for generic skills such
as verbal, manual, problem-solving, planning, client communication,
computing and autonomy skills.
33
London Economics
January 2002
Chapter 4
4.2.3
Regional Economic Forecasting Models
Oxford Economic Forecasting/NIERC28
A. Interaction between regional and national model
The Oxford Economic Forecasting/NIERC multi-regional model is a top
down model where the national model is solved first and the national
forecast drives and constraints the regional forecasts. There are no feedback
effects from the regional level to the national level except in the case of the
national labour supply that is simply the sum of the regionally determined
labour supplies. This has some strengths as well as implications for regional
information.
In practice, the UK macro-model drives the UK Industry model. The latter in
turn drives the multi-regional model, which generates regional GDP and
employment projections at a detailed level of industrial disaggregation. In
contrast to the previous two models, regional employment is derived first.
Regional employment at the disaggregated industrial level is mainly a
function of relative competitiveness, and regional wages are a function of
both regional labour demand and labour supply with the latter being affected
by interregional migration. Nevertheless, the sum of regional employment
projections in a given industry is constrained to add up to the national
forecast generated in the UK industry model.
Regional output for each industrial sector is projected by applying forecasted
regional employment in that sector to projected UK productivity in the sector,
with a fixed adjustment for relative regional productivity calculated from
historical data. Again, the regional projections are constrained to add up to
the national forecast.
B. Long-run anchor
The sequential forecasting approach adopted by Oxford Economics/NIERC
implies that the long-run anchor is provided by structure of the UK
macroeconomic model. Long-run equilibrium output (trend output) assumes
that the labour market is in equilibrium and is determined by population,
capital accumulation and productivity growth. In the long run, the economy
behaves like a one sector economy under Cobb-Douglas technology in
equilibrium and actual output will cycle around the deterministic trend29.
28
See Oxford Economic Forecasting (2001).
29
See Oxford Economic Forecasting (2000)
34
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
C. Availability of detailed employment/skills forecasts
The multi-regional model is used by Oxford Economics/NIERC to produce
regular forecasts of employment and output for 26 industrial sectors for
Wales.
Oxford Economic Forecasting has also developed a model of economic
activity and employment at the county level for 6 broad industrial sectors.
The approach is entirely top down and the sum of the county forecasts is
constrained to add up to the regional forecasts. The key drivers at the county
level are the local labour market dynamics (population, labour force and
employment) that reflect the county’s demographics, industrial mix, wage
rate, etc. relative to the other counties of the region.
4.2.4
A note on the Welsh Input-Output Tables
Input-output tables are generally not used to produce long-term forecasts of
economic activity. That being said, the most valuable feature of input-output
tables is the very detailed modelling of the industrial structure and the
supply relationship between various industries30. Therefore, policy-makers
often view such tables as valuable complementary simulation tools for
deriving detailed sectoral projections.
For example, one could use the still relatively aggregated regional demand
forecasts from the models reviewed in the previous sections as an input into
an input-output simulation exercise to derive highly disaggregated output
and employment projections.
The major strength of the input-output tables, especially for a medium-term
skills assessment, is their focus on highly disaggregated production (or
industrial) sectors and the rich depiction of the inter-industry links within a
given economy. At the same time, the latter is also their Achille’s heel as,
typically, the input-output coefficients are time-invariant, a major drawback
for any medium-term projection.
Therefore, before the Welsh input-output tables could be used as an
additional source of information, the input-output coefficients would need to
be adjusted for expected technological change and industrial restructuring,
the likely evolution of the region’s industrial comparative advantage, etc.
While formidable, this challenge is not insurmountable. The fact that the
Welsh input-output tables are based on detailed knowledge of the regional
economy should facilitate any updating and medium-term projection of the
input-output coefficients.
30
See for example Gregory et al. (2001) for a backward-looking application of the input-output
methodology to the issue of changing skill requirements.
35
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
It is also useful to note in the context of the Welsh skills needs exercise that
the Welsh input-output model team is working on modelling qualification
and vocational skill use31.
4.2.5
A note on Computable General Equilibrium
(CGE) models
So far, very little use has been made of CGE models in the field of
occupational and skills projections. The main advantage of the CGE models
is that they include a rich labour and goods/services supply-side whose
modelling is fully rooted in economic theory. As noted in the chapter on
skills forecasting, the lack of a well-developed model of how labour may react
to various market and non-market signals of skills scarcity is often viewed as
a major drawback in a skills projection. At the same, the richness of the
supply side introduces many complexities into the model, and to remain
manageable, the model cannot provide a great level of detail32. The
application of CGE models to skills’ projections is still in its infancy and,
while theoretically very appealing, such models are unlikely to be a useful
tool for large-scale skill assessment exercises in the near future33.
4.2.6
Conclusions from the review of the structure of
regional models
q
While none of the forecasting models provide for a full and effective
bottom up approach to the forecasting exercise, the models have different
characteristics and different strengths and weaknesses.
q
The table below also shows that the degree of regional and sub-regional
detail varies significantly across models.
31
See Welsh Economic Review (2000).
32
For example McGregor et al. (2001) have to limit themselves to two skill categories (skilled and
unskilled) in their CGE study of skill-biased demand shifts.
33
Of note is the fact that WDA has contracted Professor McGregor and his team to build a CGE model for
Wales.
36
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
Table 4.1: Summary Observations from Comparison of Model Structures
BSL
Cambridge
Econometrics
Oxford Economic
Forecasting/NIERC
Exogenously
determined regional
and national supply
constraints.
Long-run cointegrated regional
and national stochastic
equations determine
long-run regional
behaviour of the
model.
Determined by longrun equilibrium output
(trend output) in the
UK macroeconomic
model.
Largely top-down but
some scope for
regional specificity.
In theory almost
entirely bottom up, in
practice more top
down.
Essentially top-down.
Number of sectors for
output/employment at regional level
28
30
26
Number of sectors for
output/employment at sub-regional
level
28
49
6
Mapping of employment into
occupations
Yes
Yes
No
Work underway
Work underway
No
Long-run equilibrium
Regional-national interaction
Mapping of
employment/occupations into skills
4.3
4.3.1
Comparison of Regional Economy Model
Properties
General Observations
Practice has shown that comparative analysis and assessment of model
properties is often critical in understanding the reasons for differences in
forecasts, especially over the medium term. Unfortunately, while
considerable efforts have been made to understand major differences among
37
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
macro-economic models34, no parallel work has been done recently on
regional models35.
While model characteristics and properties have undoubtedly changed since
the mid-nineties, a study of two models (BSL and Oxford Economic
Forecasting) published 5 years36 ago clearly shows that there exist wide
variations across models in responses to standard shocks such as changes in
interest rates or tax rates. For example, the table below shows that a one
percentage point cut in the standard rate (income tax) is estimated by the BSL
model to increase the level of GDP after 5 years by 0.48% for the UK as a
whole but by only 0.38% in Wales (or 20% less than at the national level).
In contrast, the Oxford model yields a much more muted response at the
national level with the level of GDP increasing by only 0.26% (about 50% less
than in the case of the BSL model) but the impact on GDP in the UK and
Wales are almost identical.
In the case of a reduction in the VAT rate, both models yield similar estimates
of the impact on GDP in the UK and in Wales. But, the UK and Welsh
economies are about three times more sensitive to a change in VAT rate in the
Oxford model than in the BSL model.
Finally, in the case of a reduction in the base rate, the impact is more
pronounced in the case of the Oxford model than in the case of the BSL
model. Moreover, the impact on Welsh GDP is larger than at the national
level in the case of the Oxford model and smaller in the case of BSL model.
The three examples discussed above illustrate the point that regional model
structures and properties differ across models. However, this is not atypical.
For example, at the national level, the latest national comparative study of
model properties found that models’ GDP response to a one percentage point
reduction in the standard rate varied from practically nil to 0.4% (see table
below) after five years.
34
This work was done by recently closed ESRC Macroeconomic Modelling Bureau that had been
established in 1983. For the latest update on the Bureau’s work and assessment results see Church et
al. (2000).
35
See Bell (1993) and Hughes and Hunt (1994).
36
See Hunt, Slaymaker and Snell (1996)
38
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
Table 4.2: BSL and Oxford/NIERC Model Responses to Standard Economic Shocks after 5 years
Impact on total GDP
(% difference from baseline level)
UK
Impact on Unemployment
(absolute difference from baseline
level)
Wales
UK
Wales
BSL
Oxford
BSL
Oxford
BSL
Oxford
BSL
Oxford
One percentage point
reduction in standard rate
0.48
0.26
0.38
0.27
-0.38
-0.18
-0.31
-0.08
One percentage point
reduction in VAT rate
0.15
0.51
0.15
0.54
-0.02
-0.27
-0.02
-0.19
One percentage point
reduction in base rate
0.81
0.96
0.65
1.05
-0.61
-0.44
-0.50
-0.30
Economic shock
Source: Hunt, Slaymaker and Snell (1996).
Table 4.3: National Model Responses to 1 Percentage Point Reduction in Standard Rate of
Income Tax -- Impact on Total GDP after 5 years
(% Difference from Baseline)
Model
Cambridge
LBS
NIESR
HMT
COMPACT
CUSUM
Impact
0.14
0.31
0.18
0.32
0.02
0.39
Source: Church et al. (2000) and Cambridge Econometrics (2001). For comparability with the results reported in the previous table, the
responses shown in the present table are those reported by the various authors scaled back to a one percentage reduction assuming a
linear model response to the tax rate change. Cambridge = Cambridge Econometrics Multisector Dynamic Model, LBS = London
Business School, NIESR = National Institute of Economic and Social Research, HMT = HM Treasury, COMPACT = University of
Exeter/Professor Wren-Lewis model and CUSUM = Cambridge University Small UK Model.
4.3.2
q
Conclusions from the comparison of the properties
of regional models
Overall, all the model responses reported in Table 4.2 fall within a
plausible range and do not provide strong evidence in favour of one or
the other model. That being said, the main conclusion one should draw
from this comparative model property exercise is that any precise model39
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
based estimate of the response of the economy to a particular policy shock
should always be viewed with a degree a caution. As a protection against
putting too much emphasis on a specific model result it may be preferable
to focus on scenarios37 rather than single projections or simulations.
q
It would also be beneficial to forecast users if they were made regularly
aware of the properties of the model underlying the forecasts. This can
often shed some light on why specific economic variables are moving in a
particular direction over the projection horizon.
4.4
Forecast Performance of Regional Forecasting
Models
4.4.1
General observations
Forecast performance is typically assessed by comparing the forecast future
value of a given economic variable with the actual outcomes. There are a
variety of tests that are typically used by London Economics and others to
test the efficiency and unbiasedness of a forecast.
However, because of the limited number of observations38, the low frequency
of the data (e.g., annual) and frequent revisions, the scope to conduct
meaningful in-depth performance assessments of the regional forecasts is
rather limited.
For the purpose of this exercise, the performance of the current year forecasts
(i.e., the starting point of the forecast) and the year ahead forecast (i.e., the
first year following the starting point) of GDP growth and employment
growth were assessed.
While forecasts are released twice a year, in the following section we focus on
the forecasts released during the first half of the year.
As GDP data are released with a considerable lag and are subject to
significant revisions, the actual value has been defined as the second release
of a given year’s official GDP figure. For example, the actual GDP growth
rate in 1994 in this study is the growth rate shown in the second release of the
37
In this instance, one may wish to develop scenarios reflecting the uncertainty about key policy
transmission channels.
38
The Economic Division of the National Assembly graciously provided access to their holdings of older
forecast publications from the three forecasting services. Typically, these holdings went back to about
1992/93 and there exist occasional gaps in the collection of forecast records. These forecasts are shown
in Annex 2 .
40
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
data for 1994 in December 199639. As noted in the section on the data quality,
the annual GDP figures are frequently revised and one needs to be mindful of
this fact when examining the results below.
Moreover, as noted before, the official GDP figures are published only in
current prices and forecasters themselves adjust these nominal figures into
constant price data. As the object of the analysis of the GDP forecast
performance is annual growth in GDP in constant prices, the actual data are
those shown by the forecasters in their publications.
Because skills forecasting needs by definition to be forward-looking, we
examine also the long-run forecast performance of the forecast services. As
the period for which forecast records were made available to us is too short to
apply meaningfully standard forecast performance tests to the long-run
forecasts, we focus on the predicted level for GDP and employment a number
of years ahead, and examine whether this end-point is relatively stable. If the
economy were thought to converge towards some long-run equilibrium, one
would not expect the end points of the long-run forecasts to fluctuate
significantly from forecast to forecast.
Finally, because there may also be considerable interest to focus on more
disaggregated measures of output and employment in the context of skills
forecasts, we examine in greater detail the “stability” of GDP and
employment forecasts for sub-sectors such as manufacturing, mining and
quarrying, and construction.
4.4.2
Short-term forecast performance for GDP and
employment
As the number of observations used in calculating the mean absolute
forecasting errors of growth in GDP and employment in the current year and
the year ahead is very small, one should avoid drawing strong conclusions
from the results reported in the table below.
That being said, it appears that for all forecasters, forecasting errors are larger
at the Welsh level than at the UK level and employment forecast errors tend
to be often larger than the GDP forecast errors.
39
For the purpose of this exercise the “actual” employment data are defined similarly.
41
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
Table 4.4: GDP and Employment Forecast Errors
Mean absolute error of growth rate forecasts (in percentage points)
GDP
Current year
Year-ahead
BSL40
Cambridge
Oxford
BSL
Cambridge
Oxford
Wales
0.9
0.9
1.6
0.6
0.6
1.7
UK
0.5
0.4
0.6
0.4
0.4
0.7
Wales
0.7
1.8
2.1
1.4
1.6
1.5
UK
n.a.
0.8
1.0
n.a.
0.8
0.3
GDP
Employment
While these forecast errors may appear large to some, they are not unusually
large. In fact, even the forecasts of leading institutions such as the OECD and
the IMF are characterised by substantial forecasting errors.
For example, a recent study41 of IMF and OECD forecasts from the early 1970s
to the mid-1990s found that the mean absolute error for growth in GDP/GNP
(in constant prices) for the G-7 countries in the current year ranges from 0.8
percentage points to 1.2 percentage points in the case of the OECD and from
0.9 percentage points to 1.3 percentage points in the case of the IMF. The
mean absolute error for the year ahead GDP/GNP growth rate forecast
ranges from 0.9 percentage points to 1.6 percentage points in the case of the
OECD and from 1.2 percentage points to 1.7 percentage points in the case of
the IMF.
It is interesting to note that the GDP growth forecast errors of the three
regional forecasting services are very similar to those of the IMF and the
OECD. While there is always room for improvement, the size of the regional
forecasting errors are reasonable, at least in comparison to those of the IMF
and the OECD.
40
41
Employment projections for the UK by BSL are not available in the material we used for this exercise.
Pons (2000).
42
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
Another study42 of the forecasts by U.S. macroeconomic forecasters that are
published by the Wall Street Journal on the first business day of each year
found that, over the period of the mid-1980s to 1997, forecasters predicted 24
times correctly the direction of change in the new year of variables such as the
unemployment rate, interest rates, inflation and the exchange rate, and
forecasted incorrectly the direction of change 25 times.
4.4.3
Long-term sectoral forecast performance
As the forecasts published by the three forecasting services do not provide
identical long-term sectoral forecast information, it is impossible to compare
their performance very rigorously. However, it is possible to review how
some of their long-term sectoral forecasts are evolving over time. Ideally,
long-term forecasts should not change too drastically from forecast to forecast
although there may occasionally be good reasons for changing more radically
one’s views about the long term.
Overall, the detailed forecast data reported in Annex 2 suggest that long-term
views of forecasting services are susceptible to some volatility, especially if
one focuses on sectoral employment projections.
For example, the table below shows Cambridge Econometrics’ forecast for
employment in the manufacturing and construction sectors in 2005, and
Oxford Economic Forecasting’s forecast for employment in the same sectors,
but for the year 2000 as no information is provided consistently for 2005 in
Oxford Economic Forecasting’s publications43. In both cases, the forecast
employment level is set at 100 in the first reported forecast (e.g., the forecast
in the first half of 1994)44.
This volatility suggests that a scenario approach should be used for any skills
projection exercise as the employment levels at the end of the forecasting
horizon are critically important in assessing skill requirements.
42
Greer (1999).
43
Because the forecasts of BSL do not provide forecast information that is comparable over time for either a
mid- or end-point of the projection period, this section focuses only on the projections of CE and OEF.
44
The precise forecasted employment levels can be found in Annex 2
43
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
Table 4.5: Evolution of sectoral employment forecasts by Cambridge
Econometrics and Oxford Economic Forecasting
Cambridge Econometrics
Oxford Economic Forecasting
Level of employment in 2005
in
Level of employment in 2000
in
Date of forecast –
1st half of
Manufacturing
Construction
Manufacturing
Construction
1994
100
100
100
100
1995
125
90
112
54
1996
123
71
n.a.
n.a.
1997
113
95
100
41
1998
108
75
101
54
1999
109
73
n.a.
n.a.
2000
98
92
101
83
2001
103
99
99
81
n.a. not available.
4.4.4
Qualitative regional input into the regional
forecast
Any serious forecaster relies not only on a model to generate forecasts but
will almost always add a considerable dose of personal judgment to the
“pure” model predictions. Typically, during the generation of a national or
regional forecast there tends to be a bi-directional feedback loop between
pure model prediction and judgement until the forecaster is fully satisfied
with the resulting forecast.
In the case of the regional forecasts produced by the three forecasting
services, regional judgemental input is also being sought from appointed
experts in the region. This is to benefit from any local-specific information
that could affect the regional economy and may not have been adequately
incorporated into the forecasts by the forecasting services. However, some
observers believe that this regional input phase occurs generally late in the
forecasting cycle and the scope for strongly shaping the regional outlook is
rather limited.
44
London Economics
January 2002
Chapter 4
4.4.5
Regional Economic Forecasting Models
Sub-regional forecasting
As sub-regional forecasts are not regularly published by the forecast services,
it is not possible to evaluate their performance over a longer period. That
being said, if the primary interest is the forecast for a given aggregate at the
regional level, it is not obvious that producing independent forecasts for subregional entities and summing up these forecasts to generate a regional one
will yield better results.
In fact, there is an on-going, inconclusive, debate among economists on the
proper spatial dimension to adopt in forecasting exercises45. In the case of a
skills forecast for Wales, the problems with the quality of the data would
argue in favour of limiting forecasts to the regional level.
4.4.6
Conclusions
review
of
the
forecasting
performance
It is useful to remember the saying that the only certain thing about the future
is that it is highly uncertain. Forecasts will be wrong almost by definition, as
it is impossible for forecasters to capture all the unknowns that may impinge
on the actual performance of the economy in the years ahead.
q
The special regional data quality problems that confront regional
forecasters in the UK only compound the forecast error risk.
q
Do forecasting errors mean that forecasts are useless? In our view, the
answer is that they can have a value as an input to a coherent framework
for organising one’s thinking about the future.
q
However, less attention should be paid to the precise point estimates
shown in the forecasts for the various economic indicators and more
attention should be given to the intuition behind the forecast. This would
be particularly the case when the forecast changes substantially from one
projection to the next. In-depth discussions with forecast providers
would be most useful in that regard.
q
Helping forecast users to understand the key drivers of a forecast,
especially a long-term forecast, is probably the most valuable contribution
of such forecasts as it allows forecast users to test and validate/invalidate
their own views about the future.
q
The other potential benefit of forecasts is that they can provide a
45
For example Miller (1998) and Zellner and Tobias (2000).
45
London Economics
January 2002
Chapter 4
Regional Economic Forecasting Models
framework for running additional scenarios, which allow one to reflect on
what is likely to happen if a certain assumption/policy
instrument/judgement is changed. Scenarios can provide some form of
outer bounds around the point estimates of the main forecast, especially if
the alternative scenarios pick up the factors that are viewed as the most
uncertain by forecast users and forecast providers.
4.5
Concluding Remarks
q
The OEF macro-economic model is well-known in the UK and
internationally for its embedded theoretical properties, but is based on a
regional top-down approach. A richer regional texture is captured in the
regional models of both BSL and CE but this comes a cost of increased
complexity and, especially in the latter case, very large data requirements.
q
All models show plausible model properties and none of the three
regional forecasting services systematically outperforms the other. The
forecasting errors may appear to be relatively large, ranging from 0.6% to
1.7% in the case of Welsh GDP growth. But, they are very similar to the
forecasting errors of GDP growth for the G-7 countries by the IMF and the
OECD.
q
The three models and forecasting services have some strong points and
some weak points, and any choice among them will need to rely on a
broader set of criteria such as costs, services offered, etc.
46
London Economics
January 2002
Chapter 5 Results of Interviews with Representatives from the Future Skills Wales
Partners
5
Results of Interviews with
Representatives from the Future Skills
Wales Partners46
5.1
Introduction
To structure the interviews with forecasters/forecast users in Wales and
guide the discussion, a common questionnaire was used. It had been
discussed with the FSW Partners in the initial phase of the project and is
attached for information as Annex 4 .
5.2
Results of Interviews with Representatives
from the FSW Partners
Both the Economic Advice Division of the National Assembly and the Welsh
Development Agency subscribe to the standard, off-the-shelf forecasts, of
three regional forecast services in the UK, namely Business Strategies Ltd,
Cambridge Econometrics, and Oxford Economic Forecasting/NIERC, while
ELWa is an occasional consumer of long-term regional forecasts. No one
engages in the production of own in-house forecasts47.
The in-house “forecast cycle” is generally short, consisting of a review of the
key features of the forecasts received from the three services and a briefing of
senior officials in the respective institutions. Time and resources spent on the
“forecast process” are very limited, about one to two person weeks on a
yearly basis.
While interviewees felt that forecasting services did not inform appropriately
their clients about the quality of the underlying data, all of them were very
well aware of the special issues and problems with the regional Welsh data.
Moreover, there is unanimous consensus that the forecasts do not adequately
incorporate regional developments and region-specific information.
Interviewees, in general, were of the view that forecast services do not
provide substantive information on 1) model properties, 2) uncertainty
46
A list of the people interviewed for this study is at Annex 3 .
47
WDA also owns a version of the CE regional model but no systematic use of it appears to have been
made in recent years. This may simply have been due to personnel changes at WDA and/or lack of
dedicated resources with sufficient time to learn how to use the model and to run it regularly.
47
London Economics
January 2002
Chapter 5 Results of Interviews with Representatives from the Future Skills Wales
Partners
around the point estimates in the published forecasts and 3) judgement used
in the generation of forecasts.
The forecast performance of the three services has not been thoroughly
assessed by any of the institutions involved in this exercise.
The regular forecasts are not used as a quantitative input into policy-making
but are generally viewed as providing some broad picture of the general
economic context.
Both the National Assembly and the WDA show in their publications the
forecast services’ projections of some key economic aggregates. They were
also of the view that the contractual arrangements with the forecast services
prevented a more extensive release of the forecasts.
No regular meeting between the FSW Partners are currently taking place to
jointly review and assess the latest projections received.
While overall, relatively little use is currently made of the forecasts, there is a
general desire to explore ways of how forecast subscribers could get more
value from their investments.
Some also wondered whether it would be useful to develop in-house
forecasting capacity and were interested in reviewing in a greater detail the
costs and benefits of such an approach.
The non-Cardiff based FSW officials interviewed for this study noted that,
prior to the restructuring of the TECs (March/April 2001), they had being
using local economic and occupational forecasts as one of many instruments
to guide their overall work. The general view is that such forecasts were
useful provided that they were used properly. That is, such forecasts need to
be complemented by other local input and information, and users need to be
aware of the uncertainty surrounding the forecast point estimates.
48
London Economics
January 2002
Chapter 6 Results of interviews with representatives of other institutions in Wales
and outside Wales
6
Results of interviews with
representatives of other institutions in
Wales and outside Wales
The key issues raised in interviews with representatives of other institutions
in Wales and outside Wales can be regrouped under two headings: i) data
issues and ii) skills forecasting issues.
Data issues
All interviewees in Wales deplored the lack of availability of good, timely and
high quality economic data for Wales. While precise needs varied depending
on the specific activities undertaken by the institution whose representatives
were being interviewed, a common thread in all interviews was a desire for
better data.
Some institutions such as the Welsh Local Government Data Unit and the
Welsh Economy Research Unit at Cardiff Business School are planning, or
will be planning, to undertake a number of initiatives aimed at updating
and/or broadening their respective databanks. It would be worthwhile for
the FSW Partnership to explore with these institutions opportunities to
exploit potential synergies between these exercises and the broader skills
assessment that will be launched soon.
Skills forecasting
A number of issues related to skills forecasting were raised during the
interviews. The need to properly take account of potential labour supply
responses and the challenges this creates were mentioned in one of the
interviews.
In terms of detailed skills forecasting approaches, it is interesting to note that
the CITB uses a quantitative, model-based, approach to generate projections
of employment and occupations in the construction industry.
Such
projections then undergo a reality check through reviews and discussions
with representatives from the construction industry. This is very similar to
the idea put forward earlier in this document to use surveys as a potential
validity check of a model-based forecast.
It was also noted that the construction industry faces particularly acute
employment data problem as the Annual Business Inquiry employment data,
by now the main source of employment by industry, tends to miss selfemployed workers who represent a substantial share of total employment in
the construction industry. It is not clear to what extent this problem carries
over to other industries with significant self-employment. But, the issue of
potential underestimation of actual employment in industries with significant
49
London Economics
January 2002
Chapter 6 Results of interviews with representatives of other institutions in Wales
and outside Wales
self-employment will need to be properly addressed in the forthcoming skills
exercise to avoid serious errors in the employment, occupational and skills
projections.
Lantra also produced in 1998/99, with a team of five and in cooperation with
BSL, their own in-house economic and skills forecasts for land-based
industries. But, they have not produced any new forecasts since. The focus
of the skills forecast was on the skill demand side of the labour market for
land-based industries. This research was complemented by an extensive
analysis of likely skill supply, e.g., the provision of education and training
within the land-based sector, to get some perspective on the likely balance of
demand and supply of specific skills.
The forecast assumptions and results, and more generally the overall research
results, were systematically reviewed, discussed and validated through
consultations with industry, experts, and key partners. The purpose of the
exercise was to contribute to the development of Lantra’s Sector Workforce
Development Plans. So far, this research program has not been evaluated.
50
London Economics
January 2002
Chapter 7
7
Results of Interviews with Forecast Providers
Results of Interviews with Forecast
Providers
The three national providers of regional economic forecasts were also
interviewed48 as part of the study to get a better perspective on the services
they actually offer or could offer if requested. The key points to note are that
forecast service providers:
48
·
Were generally very happy to contribute to this FSW Partnership
project and wished to be kept informed of the outcome;
·
Recognise the legitimate wish of regional authorities to take into
account region-specific information and strive to do so within the
constraints imposed by the modelling structure adopted;
·
Make some information available about the quality of the data
although, with the exception of CE, this point is not regularly
addressed upfront in the forecast publications. Specific data topics
and issues, however, are raised occasionally in the forecast
publications and more general data-related issues are often
addressed in the post-forecast meetings organised for clients;
·
With the exception of CE, do not systematically provide
information about the model properties;
·
Do not provide information about the uncertainty surrounding the
point estimates of the forecasts. All, however, noted that in case of
particular uncertainty, alternative scenarios would be provided to
give a sense of the potential impact on the point estimates of the
base case or central scenario.
·
Do not provide information about the role played by judgement
versus the forecasting model in the formulation of the forecast;
·
Would all be willing to meet regularly key Welsh clients to discuss
the forecast and articulate the intuition behind the projection.
They all indicated that such meetings would also be a valuable
opportunity for them to gain additional insights into regional
economic developments.
The interview with Business Strategies Ltd was face to face while the other two interviews (Cambridge
Econometrics and Oxford Economic Forecasting) were done over the phone. The contact persons are
listed in Annex 5.
51
London Economics
January 2002
Chapter 8
8
Potential Use of Existing Forecasts for Skills Projection
Potential Use of Existing Forecasts for
Skills Projection
All three forecasts could be used as a building block for long-term forecasts of
skills requirement. However, as discussed in the chapter on skills forecasting,
the development of a good macroeconomic forecast of employment at the
sectoral level is only a first, albeit important, step in a skills forecast.
The next step involves the mapping of sectoral employment into occupations,
and the subsequent step consists of the translation of the occupational
projections into a skills forecast.
If it was decided to use the detailed regional employment forecasts from the
off-the-shelf forecasts, one could either request the forecast providers to
generate a special projection of occupations or develop such a projection inhouse, using the region-specific knowledge and expertise of the FSW
Partners, and the knowledge base of the Welsh input-output team.
Similarly, additional analytical and modelling work would have to be done
on the labour supply side.
The bottom line is that the off-the-shelf forecasts could be useful building
block, but much more work would be required before a fully-fledged skills
forecast could be generated for the off-the-shelf forecasts.
In the end, the choice between using a forecasting service to generate a skills
forecast and developing an in-house capacity to produce such projections will
depend on a cost-benefit assessment of each approach.
Finally, as mentioned a number of times already, point estimate forecasts are
subject to significant uncertainty.
Therefore, from a policy-making
perspective, it would be useful to consider a number of scenarios around the
baseline forecast as this would provide a better view of the range of likely
outcomes.
A key issue in that regard is the choice of the alternative scenarios to run.
While there is always a risk of generating too many scenarios, it would seem
logical that, in the context of a skills assessment, the chosen scenarios would
aim to provide information on the risk to the employment and occupational
projections in the base case forecast. Two of the most obvious risks are the
expected rate of growth of output and productivity in the base case, and it
would be possible to run alternative scenarios with high/low output growth
and/or high/low productivity growth. This would be very similar to the
French and German approaches to occupational forecasting.
52
London Economics
January 2002
Chapter 8
Potential Use of Existing Forecasts for Skills Projection
That being said, the most fruitful approach for identifying the most useful
scenarios would be to have key economists in Wales review with the
forecaster the base case projection that will underpin the skills projections
and agree on a common identification of the main risks to the base case
projection. Alternative scenarios could then be developed on the basis of this
risk assessment.
53
London Economics
January 2002
Chapter 9
Key Conclusion and Recommendations
9
Key Conclusion and Recommendations
9.1
Key conclusion
q
The key conclusion from this review of the use of regional economic
forecasting models is that any point estimates of future GDP or
employment growth are affected by a high degree of uncertainty.
q
This is not specific to the forecasts for Wales but is a general characteristic
of any economic forecast. However, what is particular to Wales, and the
other regions of the United Kingdom, are the problems with the quality of
the regional data which tends be lower than that of the national data and
which add to the forecast uncertainty by seriously clouding the starting
point of the regional forecast. Unfortunately, the data quality problems
worsen at the sub-regional level.
These factors lead one to conclude the following:
q
In the context of the skills assessment that the FSW Partnership plans to
undertake in 2003, analysts and policy-makers would be well advised to
avoid the temptation of putting too much weight on specific forecast
values of key variables of interest such as employment, occupations, etc.,
at the regional or sub-regional level;
q
Moreover, maintaining the skills assessment at a relatively aggregated
level, in terms of occupations, skills and geographical space, will likely
result in much more robust conclusions that will stand the test of time
better than a very large and detailed assessment
In the following sections, we provide recommendations for two key issues
that were raised during the interviews, namely getting more value from the
forecasts and projecting skills needs.
9.2
Getting more value from the forecasts
Issue 1: The dynamics underlying the numbers and region-specific factors
The key issue raised a number of times during the interviews is how the
subscribers to the off-the-shelf forecasts could get more value from their
investments. In particular, many expressed a strong desire to be able to go
behind the published numbers and better understand the dynamics driving
the forecast.
54
London Economics
January 2002
Chapter 9
Key Conclusion and Recommendations
Recommendation 1
q
Regular inter-institution meetings between the partners could be held to
review and assess forecasts.
Recommendation 2
q
Forecast services could be invited to meet regularly (once or twice a year)
with key forecast subscribers in Wales to review and explain the forecasts,
and engage in an exchange on potential region-specific factors that would
have to be taken into account in the regional forecast.
Issue 2: Developing in-house capacity
Some interviewees noted that it would be worthwhile to develop some local
capacity, housed either in one of the partner institutions or in an outside,
possibly newly-created, body to produce regional forecasts that would allow
for greater use of local knowledge.
Recommendation 3
q
An assessment of potential benefits and costs (financial, human resources,
etc.) of creating and maintaining a local modelling and forecasting
capacity should be undertaken to inform the decision makers likely to be
involved in this decision. There is likely to be a significant difference
between the merits of creating a local model with development costs
involved and developing capacity to maintain or use models developed
by specialist providers.
q
For example, we estimate that maintaining and running an off-the-shelf
model would require, once the model has become familiar to users, a
minimum of 10 to 15 person-days during the production of the forecast
and, on average, about 1 person-day per week during the non-forecast
period (assuming that the model provider provides some support
services, and updates of the model data bank can be obtained regularly).
About 2 to 4 weeks of training and experimenting would likely be
required before the model becomes fully familiar to the user.
q
The development and testing of a detailed regional in-house model built
from scratch would require about 1 to 2 years of work by two persons,
assuming most of the data can easily be assembled. In terms of using the
model thereafter, the resource requirements are, in general, unlikely to be
significantly different from those shown above for the off-the-shelf model.
The only area requiring perhaps more resources would be the regular
updating of the model databank.
55
London Economics
January 2002
Chapter 9
9.3
Key Conclusion and Recommendations
Projecting skills needs
Issue 3: How to best organise the next Welsh skills assessment
Many noted in the interviews that the next FSW exercise should be carefully
prepared. In particular, attention should be paid to role and use of
projections in such an exercise.
Recommendation 4
q
There should be a clear understanding of the focus of the exercise at the
beginning of any skills assessment. Is the focus on the demand for skills
or on skills gaps, i.e., the difference between the supply of and the
demand for skills? The latter is a much more demanding objective as it
requires projections of both the demand for skills and the supply of skills.
Recommendation 5
q
Despite all the pitfalls associated with forecasting, consideration should
be given to use again a Welsh employment and occupational forecast as
one of the inputs into the next FSW exercise.
q
However, as noted before, such a forecasting exercise should involve a
number of scenarios to reflect the key uncertainties underlying the
projections. For example, alternative scenarios could be run using
different output growth and productivity assumptions - two key
determinants of employment.
q
Moreover, to ensure that they provide useful information, such scenarios
would need to be defined in close co-operation between the FSW Partners
and the forecasting service.
Recommendation 6
q
Given the uncertainty about precise occupational skill requirements for
specific employment groups, it would be preferable to focus the
projection exercise on generic skills or skills that are used in a number of
occupations.
Recommendation 7
q
While recognizing that there exists a need for sub-regional information on
future skills demands and supplies, we would argue against using very
detailed employment and occupational projections at sub-regional levels
56
London Economics
January 2002
Chapter 9
Key Conclusion and Recommendations
as they may not be very reliable.
Recommendation 8
q
Careful consideration should be given to how occupations, skills and
labour supply responses are modelled and projected by either the
forecasting service or the in-house projection team. At a minimum, there
should be a clear understanding of the limitations of the model’s labour
supply modelling.
Recommendation 9
q
Consideration will need to be given to how optimally combine localsurvey based information with the broader forecasting exercise. Such
surveys can be used to generate data that are not otherwise available for
the skills projections or as an external reality check of the model-based
projections.
57
London Economics
January 2002
Bibliography
Bibliography
Bell D.N. F., 1993, “Regional Econometric Modelling in the UK: A Review”,
Regional Studies, Vol. 27, No 8, pp. 777-782.
Boothby D., W. Roth and R. Roy, The Canadian Occupational Projection
System: A Methodological Enhancement, R-95-2, Applied Research Branch,
Strategic Policy, Human Resources Development Canada.
Borghans L, A. de Griep and H. Heijke, 1996, Concepts and Methodology for
Labour market Forecasts by Occupation and Qualification in the Context of a
Flexible Labour Market, CEDEFOP Document.
Burns M. and M. Shanahan, 2000, Labour Market Models and their Use in
Projecting Vocational education and Training Requirements, National Centre
for Vocational Education Research, Australia.
Business Strategies Limited, 2000, A Guide to Business Strategies Forecasting
System, February.
Cambridge Econometrics, 2000, “Developments in the Database and in
Modelling” in Industry and the British Economy, pp. 51-64.
Cambridge Econometrics, 2000, “The Regionalised MDM and the Data” in
Regional Economic Prospects-Part I, pp. 61-71.
Campos N., G. Hughes, St. Jurajda and D. Munich, 1999a, When the Future is
not What it Used to Be: Lessons from the Western European Experience to
Forecasting Education and Training in Transition Economics, CERGE-EI
Working Paper No. 265, September.
Campos N.F., G. Hughes, S. Juradja and D. Munich, 1999b, Forecasting
Education and Training Needs in Transition Economies: Lessons from the
Western European Experience, in Forecasting education and training needs in
transition economies: Lessons from the Western European Experience
National Observatory of VET and Labour Market (1999) (O. Strietska-Ilina,
Ed.), Prague.
CEDEFOP, 1998, Vocational Education and Training – the European Research
Field, Background Report, Vols. I and II.
CEDEFOP, 1998, The Impact on Vocational Training of Studies Analysing and
Forecasting Trends in Occupations, Case Studies of Germany, the
Netherlands and Denmark.
London Economics
January 2002
58
Bibliography
Church K. B., J. E. Sault, S. Sgherri and K. F. Wallis, 2000, “Comparative
Properties of Models of the UK Economy”, National Institute Review, No 171,
January, pp. 106-122.
DfEE, 2000, Labour Market and Skill Trends.
Dostal, W., 1999, Forecasting Methodology for Qualification and Training
Needs in the Labour Market in Germany, in Forecasting education and
training needs in transition economies: Lessons from the Western European
Experience National Observatory of VET and Labour Market (1999) (O.
Strietska-Ilina, Ed.), Prague.
European Training Foundation, 1998, Linking Labour Market Analysis and
Vocational Training.
Feijen T., T. Reubsaet and L van Ootegem, 1996, Instruments, Tools and
Policies to Anticipate the Effects of Industrial Change on Employment and
Vocational Qualifications, Study for the European Commission, DG V,
Employment, Industrial Relations and Social Affairs.
Future Skills Wales, Report, 1999.
Greer M. R., 1999, “Assessing the Soothsayers: An Examination of the Track
Record of Macroeconomic Forecasting”, Journal Of Economic Issues, Vol.
XXX, No. 1, March, pp. 77-94.
Giffard and C. Guegnard, 1999, Forecasting Methodology for Qualification
and Training Needs in France, in Forecasting education and training needs in
transition economies: Lessons from the Western European Experience
National Observatory of VET and Labour Market (1999) (O. Strietska-Ilina,
Ed.), Prague.
Gregory M., B. Zissimos and C. Greenhalgh, 2001, “Jobs for the skilled: how
technology, trade and domestic demand changed the structure of UK
employment, 1979-90”, Oxford Economic Papers, 53, pp. 20-46.
De Gip and P. Marey, 1999, Country Report on Labour Market Forecasting in
the Netherlands, in Forecasting education and training needs in transition
economies: Lessons from the Western European Experience National
Observatory of VET and Labour Market (1999) (O. Strietska-Ilina, Ed.),
Prague.
Haskel J. and R. Holt, 1999, Anticipating Future Skill Needs: Can it be Done?
Does it Need to be Done?, Skills Task Force Research Paper 1, September.
Henson H. and C. Newton, 1995, Tools and Methods for Identifying Skill
Shortages: A Cross-Country Comparison, Occupational Projections and
Macroeconomic Studies, Applied Research Branch, Strategic Policy, Human
Resources Development Canada.
London Economics
January 2002
59
Bibliography
Hughes G., 1999, The FÁS/ESRI Occupational forecasting Model for Ireland,
in Forecasting education and training needs in transition economies: Lessons
from the Western European Experience National Observatory of VET and
Labour Market (1999) (O. Strietska-Ilina, Ed.), Prague.
Hughes L. H. and L. C. Hunt, 1994, “The BSL UK Regional Econometric
Model: a User’s Perspective”, Regional Studies, Vol. 28, No 8, pp. 859-866.
Human Resources Development Canada, (2000), Job Futures 2000, Labour
Market Information and Career Planning, Ottawa.
Hunt L. C., J. E. Slaymaker and Snell M.C., 1996, “Comparative Properties of
UK Regional Econometric Models, Regional Studies, Vol. 30, n. 8.
Institute for Employment Research, 2001, Projections of Occupations and
Qualifications 2000/2001, Research in Support of the National Skills Task
Force, University of Warwick.
Institute for Employment Research, 2001, Projections of Occupations and
Qualifications 2000/2001, Regional Results, University of Warwick.
Kekkonen K., 1998, Instruments, Tools and Policies to Anticipate the Effects
of Industrial Change in Employment and Vocational Qualification, ESF
Publication 20/98, Finland.
Lloyd, C. and H. Steedman, 1999 Intermediate Level Skills: How Are They
Changing? Skills Task Force Research Group Report, DfEE, Sheffield.
McNicoll I. And R. Marsh, 2001, Measuring the Effect of Changing Industrial
and Skills Structures on Skilled Employment, Paper presented at the RSAI
Annual Conference, September.
Meagher G. and B. Parmenter, 1996, Workforce Skills: Projections with the
Monash Model, General Paper No G-116, Centre of Policy Studies and the
Impact Project, Monash University.
McGregor P., E. P. McVittie, J. K. Swales and Y. P. Yin, 2000, Skill-biased
Demand Shifts: A Regional CGE Analysis, Strathclyde Papers in Economics,
2000/9.
Meltz N., 1996, Occupational Forecasting in Canada: Back to the Future,
Applied Research Branch, Technical Document T-96-4E, Human Resources
Development Canada.
Miller J. R., 1998, “Spatial Aggregation and Regional Economic Forecasting”,
The Annals of Regional Science, 32, pp. 253-266.
London Economics
January 2002
60
Bibliography
National Observatory of Vocational Training and Labour Market Prague,
1999, Forecasting Education and Training Needs in Transitions Economics:
Lessons from the Western European Experience.
Nekkers G., P. van Eijs, A. de Griep and B. Diephuis, (2000), Regional SupplyDemand Discrepancies: A Training Perspective, ROA-W-2000/5E, Maastricht
University.
ONS, 2000, ONS Report on Scoping Study Into Improvements To Economic
Statistics For Wales, Report presented by ONS to National Assembly for
Wales.
Oxford Economic Forecasting, 2000, The Oxford World Macroeconomic
Model, An Overview
Oxford Economic Forecasting, 2001, “The Regional Forecasting Model and
Data” in Regional Economic Outlook, Autumn, pp. 113-121.
Plassard J.M. and T. Plachard, 1997, “Méthodologie pour une prospective
d’emploi et de qualification”, Travail et Emploi, No. 71.
Pons J., 2000, “The Accuracy of IMF and OECD Forecasts for G7 Countries”,
Journal of Forecasting, Vol. 19, pp. 53-63.
Sexton J., G. Hughes, B. McCormick and C. Finn, 2001, Estimating Labour
Force Flows, Job Openings and Human Resource Requirements, 1990-2005,
Forecasting Studies, ESRI, Dublin.
Tessaring M., 1998, The Future of Work and Skills – Visions, Trends, and
Forecasts in Vocational education and training – the European research field,
Background Report, Volume 1, CEDEFOP
Van der Laan (1996), “A Review of Regional Labour Supply and Demand
Forecasting in the European Union”, Environment and Planning, Vol. 28, pp.
2105-2123.
Welsh Economic Review, 2000, Welsh Input-Output Project Update, Vol. 12.2,
August, pp. 16-17.
Wilson R.A., 2000, Forecasting Skills and Requirements at National and
Company Level: Final report, CEDEFOP and Institute for Employment
Research, January.
Willems E., 1996, Manpower Forecasting and Modelling Replacement
Demand: An Overview, Research Centre for Education and the Labour
Market, Faculty of Economics and Business Administration Maastricht
University, Maastricht, September.
London Economics
January 2002
61
Bibliography
Wisconsin Department of Workforce Development, 2001, Wisconsin
Projections 1998-2008.
Zellner A. and J. Tobias, “A Note on Aggregation, Disaggregation and
Forecasting Performance”, Journal of Forecasting, Vol. 19, pp. 457-469.
London Economics
January 2002
62
Annex 1
Occupation and skills forecasts: the practice in a number of countries
Annex 1 Occupation and skills forecasts:
the practice in a number of countries
UK – Qualifications Projections by Institute for Employment Research
(IER)
In the UK occupational and skill (e.g., qualifications in this case) projections
are prepared by the IER with a model of the economy developed by
Cambridge Econometrics.
Using economic projections of regional economic activity and employment at
a detailed sectoral level, projections of the occupational structure of
employment within each industry are constructed using information on
historical occupational distributions within a sector.
The projected
occupational structure is then applied to the projected levels of industrial
employment to obtain projected levels of employment by occupation49.
These occupational employment projections are in effect based on a submodel that is based on the regional/industrial projections produced by the
macroeconomic model combined with information on the age and gender
structure of occupational employment. It focuses on gross occupational
turnover as it takes account of outflows due to demographics (retirement and
mortality) and inter-occupational mobility50, estimated from the Labour Force
Survey.
The qualifications demand component focuses on the employment of highly
qualified persons, distinguishing nine subject categories and three levels
(postgraduate, first degree and equivalent, and other highly qualified). The
supply component of the model which accounts for economically active
persons holding higher qualifications uses a simple stock-flow model after
considering losses due to deaths and net migration plus new entrants.
These figures are then combined with projections of economic activity rates
(by gender and age), in order to derive an estimate of the total number of
qualified persons in the economically active category.
In September 2000, the Department for Education and Skills (DfES), formerly
the Department for Education and Employment (DfEE), has set up a Skills
Unit database called Skillsbase (website: www.skillsbase.dfee.gov.co.uk). Its
49
Projections of Occupations and Qualifications 2000/2001, Institute for Employment Research, 2001.
50
Institute for Employment Research (2001).
63
London Economics
January 2002
Annex 1
Occupation and skills forecasts: the practice in a number of countries
primary aim is to improve labour market information available to Regional
Development Agencies, Learning and Skills Councils, National Training
Organisations and other bodies with an interest in skill issues – it is intended
as a one stop source of information on the UK labour market. The database
comprises a series of linked MS Excel workbooks, each containing detailed
data and commentary on key labour market indicators. The database has
various subsections, e.g., macroeconomic indicators, generic/key skills,
qualifications, replacement demands and so on. In particular, the database
includes the detailed employment projections prepared by the Institute for
Employment Research in collaboration with Cambridge Econometrics. The
projections include both historical and forecast data by sector, occupation and
qualification at both the national and regional levels.
Germany
In Germany, forecasts for qualification and training needs are produced and
reported by the federal government51. A specific feature of these forecasts is
that they include the formal school system in the process.
Historically, forecasts for 24 disaggregated occupational tasks and sectors
were provided, in addition to four levels of qualifications (i.e. unskilled, first
and second level of school and college and university). These projections
were provided based on three different employment scenarios (low, medium
and high trend in employment). Subsequent developments have led to
disaggregation into 34 occupational tasks and 11 skill levels.
The initial stage involves an analysis of a series of factors influencing
employment, including technological, economic and social factors, and their
impact on occupational tasks. Their impact is analysed qualitatively as
positive or negative. This result is then used quantitatively in projections of
future effects by occupation. Typically, three scenarios are generated
allowing for different future employment trends.
Using the sectoral and occupational forecasts, a parallel skills demand
forecasts is generated.
France
At national level the BIPE (Institute of Economic Forecasting) runs a model
generating estimates of recruitment needs disaggregated by professional
categories. At the local level, the Regional Employment and Training
Observatory (OREF) is responsible for forecasting training needs52.
51
Dostal (1999).
52
Giffard and Guignard (1999).
64
London Economics
January 2002
Annex 1
Occupation and skills forecasts: the practice in a number of countries
Forecasts at the national level are derived from three different models. The
macroeconomic model (DIVA) generates employment projections by sector or
branch. Projected employment level for each professional category is then
obtained via the CALIFE model. Finally, hiring requirements of each
professional category are generated by the GESPER model, which gives
particular consideration to geographic and professional mobility factors
(including retirements, inter-sectoral mobility, promotion within companies,
etc.).
It is notable that to estimate future hiring and training needs different
scenarios are usually required. Hiring needs forecasts are obtained from the
DIVA and CALIFE models, and used jointly with the Employment Surveys
and the Vocational Training Qualification Survey to obtain professional
information on hiring needs. The final step entails an analysis of competition
between young people, job seekers and inactive women seeking employment.
The OREF is responsible for forecasting training needs at a local level. The
first step in this process is to establish links between jobs, sectors and the
need to improve qualifications. At the second stage, these findings are
assessed and future developments and changes are identified and analysed in
terms of the probability of their continuation or cessation.
Netherlands
The Ministry of Education and Science is responsible for increasing the
transparency of the match between education and the labour market by
tracking all the relevant flows of the labour market53. Hence, the main
objective of the forecasts is to inform and provide vocational and educational
guidance.
The methodology uses the Manpower Requirement Approach based on an
input-output structure of the economy. On the demand side forecasts
distinguish between the expansion demand that reflects movement in
employment levels in a particular occupation, and the changes within
occupations due to replacement demand, and are provided for 127 groups
and 104 types of education. The supply side includes flows due to schoolleavers, training courses outside the system and short-term unemployed
persons from 104 types of education. Supplementary data such as surveys of
school-leavers are used for education flows.
Five qualitative indicators are provided for each of the 104 types of education:
future labour market situation, risks of future labour recruitment problems,
opportunities for school-leavers to switch economic sectors, and a risk
measure of the sensitivity of employment to cyclical fluctuations. In addition
53
Grip and Marey (1999).
65
London Economics
January 2002
Annex 1
Occupation and skills forecasts: the practice in a number of countries
to point estimates, forecasts are accompanied by a qualitative assessment
ranging from “very low” to “very high”.
In addition to the forecast made at a national level, regular forecasts are
provided for Limbourg, one of the 12 provinces of the Netherlands.
Ireland
The occupational forecasting model for Ireland54 is run by Foras Aiseanna
Saothar (FÁS), The Training and Employment Authority55. The objective of
this model is to provide information on changing patterns of occupational
and sectoral occupations and their implications for future skill
requirements56.
Forecasts for the occupational structure of the Irish economy are derived from
medium-term projections of employment by sector and occupation.
Historical patterns of occupational change are one input required to develop
these forecasts. In this regard, a study of the labour market (using data from
the Census of Population and the Labour Force Survey, among others)
showed that the occupational and sectoral employment structure of the Irish
economy had changed significantly in recent years and helped to identify
factors influencing changes in occupations at the industrial level.
The method used for forecasting the structure of employment by occupation
and industry is the Manpower Requirements Approach (see Section 2.4.1).
Using the macro-model, employment forecasts are produced for industry and
sector. The next step entails developing forecasts for the occupational
composition of employment for each industry.
In the next stage, the trend in the share of each occupation in each sub-sector
is examined and projected to the target year using a set of equations and
judgement for selecting the projections that appear most reasonable. To
analyse the change in the level of employment in different occupations shiftshare analysis is used to statistically decompose employment changes into a
scale effect, industry effect, occupation effect, and interactive terms between
these three effects.
By combining the forecasts for employment in each sub-sector with each
occupational sub-group’s share in each sub-sector, forecasts are then derived
for employment in each occupational sub-group.
54
Hughes (1999).
55
See http://www.forfas.ie/futureskills/index.html.
56
Hughes (1999)
66
London Economics
January 2002
Annex 1
Occupation and skills forecasts: the practice in a number of countries
Occupational data show the quality of the labour force required in the future,
although specific skill forecasts are not provided.
Finland
The Advisory Committee on Education, which includes representatives of
main labour organisations and ministries, is the body in charge of providing
labour forecasts for planning education in Finland57.
The Ministry of Labour developed in the beginning of the 1990’s a
multisectoral model to forecast employment shifts. In essence, the process
follows the Manpower Requirement Approach and the change of
employment is given for 17 sectors and 47 sub-sectors.
Next, the
Occupational Structure Model is used to generate disaggregated occupational
forecasts for 11 main groups and 48 sub-groups.
The demand for educated labour force is disaggregated by educational field
and level. Projected job openings are generated from the expected sectoral
change in employment after accounting for outflows from the occupational
groups due mainly to retirement, sudden death and disability. These
estimates are mainly based on past trends. The projected supply of students
having completed their studies takes into account the likely dropout rate
during the study period.
Finally, job openings by occupational group are translated into corresponding
educational levels. Thus, forecasts of recruitment needs by educational field
and level are generated together with forecasts of labour demand for various
vocational qualifications and professional degrees.
US – Occupational Outlook
The US has an integrated system for providing users with direct access to
labour market information at a state level. Reflecting the view that market
forces will, in general, resolve skill shortages, the amount of active labour
market programming is minimal. The federal government’s Occupational
Outlook Service in the US Department of Labour’s Bureau of Labour Statistics
(BLS) is responsible for providing occupational-based employment
projections for career guidance purposes.
Currently, projections are provided on the size and composition of labour
force, economic growth, and detailed estimates of industrial production and
of employment by industry and by occupation. No specific skill forecast is
derived from the occupational forecast. Moreover, as the purpose is simply
to provide information on likely future occupations, this approach does not
incorporate labour supply responses.
57
Kekkonen (1998).
67
London Economics
January 2002
Annex 1
Occupation and skills forecasts: the practice in a number of countries
The information from the federal occupational projections is complemented
by information collected by the State Employment Security Agencies. It is left
to company managers, training programmes and State government to keep a
watch over specific developments (such as replacement demand) so that
training programmes can remain in tune with their needs58.
Some State Employment Agencies have developed State-specific approaches
and carry out additional occupational projection exercises to provide a guide
for future investment requirements in education and training. Typically,
state agencies produce information on industrial employment and
employment by occupation.
To avoid future shortages in jobs that cannot be filled by untrained workers,
these projections are used to plan formal education programmes, and to
develop training and re-skilling programmes for people in employment.
New technology and new management methods also mean that currently
employed skilled workers need to learn new skills.
Canada - Occupational profiles
The Human Resources Development of Canada (HRDC) provides labour
information to current and future labour market participants through a
regular overview of labour market trends and detailed information on
occupations (Boothby et al., 1995). No attempt is made to generate forecasts
of future skill requirements, but the projections of occupations are used,
among others, to anticipate potential skill shortages and bottlenecks59.
The Canadian Occupational Projection System, which uses the economic
projection from a macroeconomic model as a starting point, provides detailed
information by occupation such as employment, unemployment, earnings
and skill level60. Such indicators are derived from available hard data,
qualitative analyses and consultations with industry associations,
professional groups, unions and sector councils.
Occupational projections identify the type of employers who hire workers in
a given occupational group and describe this occupational group’s main
work activities. The information provided also includes the type and level of
education or specific training required, the necessary work experience and
any requirements for licences, certificates or registration. However, HRDC
makes clear that the indicators provided by their system describe conditions
58
59
60
See for example Wisconsin Projections, 1998-2008 from the Wisconsin Department of Workforce
Development (2001).
Meltz (1996)
Job Futures 2000, Labour Market
http://jobfutures.ca/jf-ea/.
Information
and
Career
Planning can
be
found
in
68
London Economics
January 2002
Annex 1
Occupation and skills forecasts: the practice in a number of countries
for any occupational group at the national level only, and cautions users that
conditions may vary among the different occupations within an occupational
group, and may be different in local or provincial labour markets.
Finally, comparisons of unemployment rates and earning levels in each
occupation relative to the national average also provide some information
about the chances of obtaining employment and expected incomes.
Australia – Centre of Policy Studies
Australia’s Centre of Policy Studies (CoPS) uses the Monash model for
producing detailed labour market forecasts. The Monash forecasting system61
takes as inputs macroeconomic forecasts developed externally. These
forecasts provide a background for assessing the skills likely to be required in
the Australian workforce over the next decade.
These macroeconomic projections are next combined with industry-specific
forecasts produced by expert organisations such as the Australian Bureau of
Agricultural and Resource Economics (ABARE), the Bureau of Tourism
Research, the Productivity Commission (PC), and the Centre of Policy
Studies. Projections for employment are then disaggregated at industry level,
by state, sub-state, region and occupation.
The model is initially calibrated to an up-to-date input-output database.
Hence, it first produces results for macroeconomic variables and the
industrial structure at the economy-wide level, followed by results for states,
regions and occupations, using a top-down disaggregation.
61
Meagher and Parmenter (1996).
69
London Economics
January 2002
Annex 2
Forecasts of three major regional forecast providers
Annex 2 Forecasts of three major regional
forecast providers
Table A.1: Annual GDP growth forecast performance – BSL
Business Strategies Ltd
Wales
Current year forecast
Date of forecast
Year ahead forecast
Date of actual
data
Forecast
Outcome
Forecast
Outcome
Jul. 93
Dec. 95
2.9
1.1
3.3
4.6
Jul. 94
Mar. 97
3.5
4.6
3.3
2.8
Feb. 95
Jan. 98
3.1
2.8
2.1
2.2
Feb. 96
Mar. 99
2.3
2.2
2.8
2.0
Mar. 97
Aug. 00
3.4
2.0
2.3
2.1
Mar. 98
Mar. 01
0.9
2.1
1.8
..
Mean absolute
error
0.9
0.6
UK
Jul. 93
Dec. 95
1.7
2.2
3.0
4.0
Jul. 94
Mar. 97
3.1
4.0
2.7
2.8
Feb. 95
Jan. 98
3.2
2.8
2.3
2.5
Feb. 96
Mar. 99
2.7
2.5
3.0
3.3
Mar. 97
Aug. 00
3.2
3.3
2.5
3.0
Mar. 98
Mar. 01
2.0
3.0
1.7
..
Mean absolute
error
0.5
0.4
70
London Economics
January 2002
Annex 2
Forecasts of three major regional forecast providers
Table A.2: Annual employment growth forecast performance – BSL
Business Strategies Ltd
Wales
Current year forecast
Date of forecast
Year ahead forecast
Date of actual
data
Forecast
Outcome
Forecast
Outcome
Jul. 93
Dec. 95
-0.7
0.5
2.4
0.8
Jul. 94
Mar. 97
1.2
0.8
0.9
-2.4
Feb. 95
Jan. 98
n.a.
-2.4
n.a.
0.9
Feb. 96
Mar. 99
1.7
0.9
1.3
0.6
Mar. 97
Aug. 00
1.6
0.6
0.9
0.9
Mar. 98
Mar. 01
0.8
0.9
-0.9
..
Mean absolute
error
0.7
1.4
71
London Economics
January 2002
Annex 2
Forecasts of three major regional forecast providers
Table A.3: Annual GDP growth forecast performance - CE
Cambridge Econometrics
Wales
Current year forecast
Date of forecast
Year ahead forecast
Date of actual
data
Forecast
Outcome
Forecast
Outcome
Feb. 94
Jan. 97
3.0
4.4
2.6
2.4
Feb. 95
Jan. 98
3.8
2.4
3.2
1.6
Feb. 96
Mar. 99
1.8
1.6
3.0
3.1
Feb. 97
Jun. 00
3.5
3.1
3.1
2.7
Feb. 98
Mar. 01
1.6
2.7
1.7
..
Mean absolute
error
0.9
0.6
UK
Feb. 94
Jan. 97
2.5
3.5
2.5
2.6
Feb. 95
Jan. 98
3.1
2.6
2.8
2.4
Feb. 96
Mar. 99
2.2
2.4
2.9
3.4
Feb. 97
Jun. 00
3.4
3.4
2.9
2.2
Feb. 98
Mar. 01
2.4
2.2
1.7
..
Mean absolute
error
0.4
0.4
72
London Economics
January 2002
Annex 2
Forecasts of three major regional forecast providers
Table A.4: Annual employment growth forecast performance - CE
Cambridge Econometrics
Wales
Current year forecast
Date of forecast
Release date of
actual data
Feb. 94
Year ahead forecast
Forecast
Outcome
Jan. 97
0.3
2.4
0.5
-2.5
Feb. 95
Jan. 98
0.7
-2.5
1.3
-0.3
Feb. 96
Mar. 99
-0.6
1.9
0.6
0.0
Feb. 97
Jun. 00
0.5
0.0
0.8
0.8
Feb. 98
Mar. 01
0.0
0.8
0.3
..
Mean absolute
error
Forecast
1.8
Outcome
1.6
UK
Feb. 94
Jan. 97
-0.1
0.7
0.5
1.4
Feb. 95
Jan. 98
0.4
1.4
0.1
1.2
Feb. 96
Mar. 99
0.3
1.2
0.5
1.4
Feb. 97
Jun. 00
0.8
2.1
0.9
1.2
Feb. 98
Mar. 01
1.1
1.4
0.1
..
Mean absolute
error
0.8
0.8
73
London Economics
January 2002
Annex 2
Forecasts of three major regional forecast providers
Table A.5: Annual GDP growth forecast performance - OEF
Oxford Economic Forecasting
NIERC
Wales
Current year forecast
Date of forecast
Release date of
actual data
Sep. 94
Year ahead forecast
Forecast
Outcome
Forecast
Outcome
Jan. 97
2.2
4.6
3.7
2.2
Spring 95
Jan. 98
3.3
2.2
2.7
n.a.
Update 96
Mar. 99
1.8
n.a.
3.6
1.1
Spring 97
Jun. 00
2.5
1.1
1.7
2.7
Spring 98
Mar. 01
1.4
2.7
2.1
..
Mean
error
absolute
1.6
1.7
UK
Sep. 94
Jan. 97
3.0
3.6
3.5
2.4
Spring 95
Jan. 98
3.3
2.4
2.7
n.a.
Update 96
Mar. 99
2.2
n.a.
3.2
3.4
Spring 97
Jun. 00
3.2
3.4
2.0
2.9
Spring 98
Mar. 01
2.2
2.9
1.9
..
Mean absolute
error
0.6
0.7
n.a. not available.
74
London Economics
January 2002
Annex 2
Forecasts of three major regional forecast providers
Table A.6: Annual employment growth forecast performance - OEF
Oxford Economic Forecasting
NIERC
Wales
Current year forecast
Date of forecast
Release date of
actual data
Sep. 94
Year ahead forecast
Forecast
Outcome
Forecast
Outcome
Jan. 97
-0.3
2.4
0.4
-2.5
Spring 95
Jan. 98
2.1
-2.5
1.2
-0.3
Update 96
Mar. 99
-0.1
1.9
1.3
0.0
Spring 97
Jun. 00
0.2
0.0
0.7
0.8
Spring 98
Mar. 01
-0.4
0.8
0.1
..
Mean
error
absolute
2.1
1.5
UK
Sep. 94
Jan. 97
0.1
0.7
0.4
1.4
Spring 95
Jan. 98
1.8
1.4
1.2
1.2
Update 96
Mar. 99
0.1
1.2
1.3
1.4
Spring 97
Jun. 00
1.2
2.1
1.2
1.2
Spring 98
Mar. 01
0.4
1.4
0.0
Mean absolute
error
1.0
0.3
75
London Economics
January 2002
Annex 2
Forecasts of three major regional forecast providers
Table A.7: Evolution of sectoral forecasts by Cambridge Econometrics
Per annum growth (in %) in GDP over period
2000-2005 in
Employment level (in thousands) in 2005 in
Manufacturing
Mining and
Quarrying
Construction
Manufacturing
Mining and
quarrying
Construction
Feb. 94
4.0
1.3
2.5
191
5
83
Feb. 95
2.8
-0.4
1.9
239
4
75
Feb. 96
3.4
-1.9
1.5
234
2
59
Feb. 97
2.9
0.6
1.2
215
3
79
Feb. 98
2.7
-1.1
1.2
207
2
62
Feb. 99
2.6
-1.4
1.0
209
2
61
Feb. 00
2.4
-1.1
1.6
188
4
76
Feb 01
1.8
-1.9
0.7
198
5
82
Forecast date
Table A.8: Evolution of sectoral forecasts by Oxford Economic Forecasting
Per annum growth (in %) in GDP over period
1995-2000 in
Forecast date
Employment level (in thousands) in 2000 in
Manufacturing
Mining and
Quarrying
Construction
Manufacturing
Mining and
quarrying
Construction
Sep. 94
n.a.
n.a.
n.a.
204 (Aug. 93)
2
78
Spring 95
n.a.
n.a.
n.a.
229 (Autumn)
3
42
Update 96
2.7
7.0
2.7
n.a.
n.a.
n.a.
Spring 97
2.5
-4.5
1.0
203
3
32
Spring 98
2.0
-20.2
2.6
206
2
42
Spring 99
n.a.
n.a.
n.a.
n.a.
n.a.
n.a.
Spring 00
0.6
6.6
6.0
206
6
65
Spring 01
-0.1
7.2
2.1
201.
5
63
n.a. not available.
76
London Economics
January 2002
Annex 3
List of Persons Interviewed for the Study
Annex 3
Study
1.
List of Persons Interviewed for the
FSW Partners
ELWa
Grenville Jackson, Director Corporate Policy and Strategy
Joanne McCallum, Research and Futures Manager
Geoff Owen, Head of Strategic Planning
Huw Owen, Economist
Geoff Supple, Head of Research and Evaluation – West Wales
National Assembly for Wales
Mike Phelps, Chief Economic Adviser, Economic Advice Division
Neil Paull, Head of Industrial Economics, Economic Advice Division
Steve Marshall, Head of Labour Market Statistics, Education,
Training and Economics Division
David Pritchard, Director of Economic Development
Welsh Development Agency
Gareth Hall, Executive Director, Strategy Development
James Price, Senior Economist
2.
Other Interviews in Wales
Sue Camper, Agent, Bank of England, Wales Agency
Sue Denman, Wales European Funding Office
77
London Economics
January 2002
Annex 3
List of Persons Interviewed for the Study
Kate Chamberlain, Data Unit, Welsh Local Government Association
Owen Jenkins, Operational Manager, Cardiff County Council
Steven Hill, Professor, University of Glamorgan Business School
Annette Roberts and Max Munday, Welsh Economy Research Unit,
Cardiff Business School
4. Interviews outside Wales
Martin Arnott, Research Manager, CITB
Jenny Baker, Workforce Development Assistant, Lantra
Peter McGregor, Head of Economics Department, University of
Strathclyde
78
London Economics
January 2002
Annex 4
Interview questions for forecast users
Annex 4
users
Interview questions for forecast
USE OF REGIONAL FORECASTING MODELS IN
WALES
INTERVIEW SCHEDULE/GUIDE FOR
INTERVIEWER62
62
Various sections of the questionnaire were used in face-to-face or telephone interviews depending on the
interviewees’ direct experience in forecasting and modeling.
79
London Economics
January 2002
Annex 4
Interview questions for forecast users
STRUCTURE OF QUESTIONNAIRE
It is intended that this questionnaire will provide guidance
on the interviewing process to gather information on how
the forecasts are currently being made and what their main
purpose is. The questions would be rather open and could be
evolving depending on the dynamics of each interview. In
particular we intend to gather information on the following
points:
1.
THE FORECASTING PROCESS
2.
DATA USED IN THE FORECAST
3.
FORECASTING MODEL USED IN THE
PRODUCTION OF THE FORECAST
4.
FORECASTING
5.
USE OF FORECASTS
6.
OVERALL ASSESSMENT OF
FORECASTING PROCESS
80
London Economics
January 2002
Annex 4
Interview questions for forecast users
1. FORECASTING PROCESS:
The following steps (A to E) are the standard steps required to conduct forecasting.
Please answer as appropriate and relevant to your forecasting process. If there are
any additional steps please specify:
A.
How do you obtain your forecast?
1. Do you buy a standard, off-the-shelf, forecast
from a forecasting organization; or
2. Do you buy a customized forecast from a
forecasting organization; or
3. Do you use any published forecasts, bulletin,
etc.; or
4. Do you use an in-house model (either bought or
internally developed) to produce the forecast?
Only if you buy a customized forecast
B.
If you buy a customized forecast, could you
provide details about the specific features that
are customized for you?
C.
How often do you forecast or receive forecasts
(monthly, quarterly or yearly)?
81
London Economics
January 2002
Annex 4
Interview questions for forecast users
D.
Could you briefly describe your full forecasting
activity cycle, from the first day to the last day?
If you buy a forecast, start with the day when
you receive the forecast. If you produce your
own forecast, start with the first day of the new
cycle.
The last day of the cycle would be the day when
the forecast is last discussed or reviewed
internally.
E.
Provide your best estimate of the resources
involved in your forecasting activities during a
typical forecasting cycle.
If possible, break down by junior and senior
personnel.
2.
DATA USED IN THE FORECAST
Listed below is a list of questions about data acquisition and validation:
Only if you produce your forecast:
A.
If you produce your own forecasts, how do you
acquire your data?
-
From primary sources such as
main data bases;
-
From intermediaries such as
forecast service providers –
please specify;
82
London Economics
January 2002
Annex 4
Interview questions for forecast users
B.
-
You develop your own data
from soft sources such as
surveys, etc - please specify; or
-
Some other source.
What is your opinion on the quality of the data
used in the forecast? (e.g. reliability or degree of
confidence or any other sanity checks).
Please provide a separate assessment for the
following data sets:
3.
·
National
·
Regional
C.
Do
you
think
that
special
regional
data/features/dimensions are adequately taken
into account in the regional forecast?
D.
Do you have any other data/reports that can
complete the information at a regional level?
FORECASTING MODEL USED IN THE PRODUCTION
OF THE FORECAST
A.
How well do you know the forecasting model
from which you get your forecast?
-
Do you have some information
about the underlying structure
of the macroeconomic model?
-
Do you have some information
about the model properties
(e.g. such as responses to
standard shocks)?
-
Do you have some information
about linkages between the
national and regional model?
Only if you buy customized forecasts:
83
London Economics
January 2002
Annex 4
Interview questions for forecast users
B.
Can you ask the forecast provider to change one
or more key equations in the model used to
produce your customized forecast?
Only if you use a model bought from outside sources:
4.
C.
Do you have access to the model code and are
you allowed to modify the model equations?
D.
If you do have access, do you sometime change
the model equations?
E.
If not, why not?
F.
Do you get technical support from the model
vendor?
FORECASTING
A.
Which are the key economic forecast indicators
used in your institution?
-
Output
-
Industrial production
-
Government revenue
-
Inflation
-
Employment
-
Unemployment
-
Labour force
-
Other
B.
Is your main interest in regional predictions?
C.
If yes, are there any variables, among the set of
regional variables you focus on, that could be
used to generate regional skills forecast?
D.
Do you focus uniquely on point forecast
estimates or do you also make use of scenarios?
84
London Economics
January 2002
Annex 4
Interview questions for forecast users
E.
Do you typically have information on the degree
of uncertainty surrounding the forecast
(“confidence interval” around the point forecast
estimates)?
F.
If you buy forecasts (standardized or custommade) are you being given information on the
judgment used by the forecast vendors in the
generation of the forecasts (add-factors or
exogenous adjustments to forecast from
behavioural equations, special regional effects,
etc)?
G.
If you buy customized forecasts, can you specify
specific values for exogenous forecast variables
or specify your own judgment adjustment?
H.
If you use a bought forecasting model, does the
vendor supply you regularly with a base case
scenario, and the values used for the exogenous
variables and the judgmental adjustments?
I.
How do you validate the forecasts you use or
produce and how often do you do this, if at all?
Only if you buy forecasts:
Only if you buy customized forecasts
Only if you buy a forecasting model
5.
-
Ex-post comparison of actual
outcomes with forecast values
for key variables
-
Comparison
forecasts
-
Other – please specify
with
other
USE OF FORECASTS
A.
How often are the forecasts reported to key
decision-takers in your organization?
85
London Economics
January 2002
Annex 4
Interview questions for forecast users
B.
C.
6.
-
in any standard report;
-
bulletins; etc
How are the forecasts used, if at all, in the
decision-making process in your organization?
-
only
as
information;
background
-
as qualitative input;
-
as quantitative input;
-
other.
Do you exchange/share your forecasts with other
key Partners in Wales. If yes, please provide
detailed information on how, what, frequency,
etc.
OVERALL ASSESSMENT
A.
On a scale of 1 (highly unsatisfactory) to 5 (very
satisfactory), how would you rate:
a.
The usefulness of the forecasts, as they are
currently produced, for the decision-taking
process (policy, strategic, etc), e.g., do the
forecasts meet the needs of your organization;
b.
How specific regional information is incorporated,
if at all, in the forecast?
c.
The value for money your institution is getting
from the process.
B.
If less than very satisfied, describe
improvements you would like to see?
the
86
London Economics
January 2002
Annex 5
List of contact persons at forecast service organisations
Annex 5 List of contact persons at forecast
service organisations
Business Strategies Ltd
-
Neil Blake, Research Director
Cambridge Econometrics
-
Saxon Bretell, Director
Oxford Economic Forecasting/NIERC
-
Alan Wilson, Senior Economist
87
London Economics
January 2002