Content uploaded by John Bradley
Author content
All content in this area was uploaded by John Bradley on Apr 03, 2014
Content may be subject to copyright.
I
ASSESSING THE IMPACT
ASSESSING THE IMPACTASSESSING THE IMPACT
ASSESSING THE IMPACT
OF EU COHESION
OF EU COHESION OF EU COHESION
OF EU COHESION
POLICY:
POLICY:POLICY:
POLICY:
WHAT CAN ECONOMIC MO
WHAT CAN ECONOMIC MOWHAT CAN ECONOMIC MO
WHAT CAN ECONOMIC MODELS TELL US?
DELS TELL US?DELS TELL US?
DELS TELL US?
JOHN BRADLEY
GERHARD UNTIEDT
Presentation made at the Bruegel Workshop
Brussels, 15th May 2012
HERMIN Economic Paper 2-2012
ISSN: 2194-7708
II
HERMIN WORKING PAPER 2-2012
Web: http://www.herminonline.net
ISSN: 2194-7708
COMMUNICATING AUTHOR:
Dr. John Bradley
EMDS - Economic Modelling and Development Strategies
14 Bloomfield Avenue, Portobello, Dublin 8, Ireland
Email: john.bradley@iol.ie
RePEc: http://ideas.repec.org/e/pbr138.html
EDITOR:
Prof. Dr. Gerhard Untiedt
GEFRA - Gesellschaft für Finanz- und Regionalanalysen,
Ludgeristr. 56, 48143 Münster, Germany
Tel: (+49-251) 263 9311 Fax: (+49-251) 263 9319
Email: gerhard.untiedt@gefra-muenster.de
Web: http://www.gefra-muenster.de
ReEPc: http://ideas.repec.org/e/pun15.html
1
“The consequences for human welfare involved in
questions (of growth) are simply staggering:
Once one starts to think about them, it is hard
to think about anything else.”
Robert Lucas, 1988
2
1 INTRODUCTORY REMARKS
The challenge of evaluating the impacts of cohesion policy lies in the complexity of the
public policy instruments being used in terms of individual projects, wider measures,
operational programmes and the entire investment package taken as a whole. The goal of
cohesion policy – to promote accelerated growth and development in lagging EU member
states and regions, i.e. development at the aggregate macroeconomic level – is ambitious
and the evaluation of its likely impacts draws on economic and other research that is still
at an early stage of evolution. The context within which cohesion policy is designed,
implemented and evaluated is also complex and this should serve as a warning against
simplistic evaluations and premature judgements.
In the course of cohesion policy impact evaluation there are really only two crucial
decisions to be taken. First, do you need to construct an explicit policy counterfactual?
Second, if the answer is “yes”, how does one define the counterfactual? If one wishes to
identify the specific contribution of a policy action, it would be difficult to answer other than
“yes” to the first question. But there are a range of possible answers to the second
question. On the one hand there is what we shall term a “soft” counterfactual, defined in
the Barca report as follows:
Counterfactual impact evaluation (here termed simply impact evaluation) allows
the assessment of policy effects without the use of complex econometric models, in
which strong hypotheses need to be made which are often hard to appreciate by the
layman, and to act as a leverage for policy improvement. It focuses on using data of
good quality and on the robustness of the method through which a population “similar”
to the target population is identified. (Barca, 2009, p.47).
On the other hand there is what we shall term a “hard” counterfactual which can only be
constructed using an explicit model that attempts to articulate how policy affects the
economy, thus permitting direct analysis of “with policy” compared to “without policy”
scenarios.
1
1
One could argue about the allocation of the terms “hard” and “soft”. In our use of the terms,
“hard” signals that the counterfactual scenario can be defined fairly precisely, but depends on
the acceptability of the model as a true and accurate portrait of how economies function and
how policy instruments affect economic processes. “Soft”, on the other hand, signals that it is
very difficult to identify with any degree of precision “a population “similar” to the target
population”.
3
Our background to the preparation of this paper was a period of more than twenty years
developing and using a specific model framework to identify the cohesion policy
counterfactual. This framework is the HERMIN model which had its origins in the EU
HERMES model and is now the model used within DG-REGIO’s Cohesion System of
HERMIN Models (or CSHM).
2
However, we do not feel that it is useful simply to present
in isolation the results of policy impact analysis based on the HERMIN model since this
would represent just the application of one model (HERMIN) and could not claim to be
unambiguously preferable to results based on any alternative model. Furthermore,
presentation of the technical details of the HERMIN model structure and its operational
use would risk confusing the layman (to use Barca’s language), and would simply uncover
areas of macroeconomic theory and knowledge where researchers and policy makers can
and do legitimately hold different and often opposing views. The ambiguity of the Barca
quest for “a population ‘similar’ to the target population” finds a parallel in the quest for an
“appropriate model” for use in in model-based cohesion policy analysis. Neither approach
to defining a policy counterfactual – hard versus soft - can claim unambiguous superiority.
Open discourse on these issues is vital if we are to move towards a better understanding
of policy impact analysis.
We set ourselves the following objectives in this paper. First, we stand back from the
technical aspects of the analysis of cohesion policy impacts in order to identify and
describe what we see as the essential stages of model-based evaluation with specific
attention to areas where evaluators may legitimately differ from each other. Second, we
briefly examine two model-based evaluations of cohesion policy impacts that were
produced using different models: the CSHM of DG-REGIO and the QUEST model of DG-
ECFIN.
3
Third, in light of the different policy impact results obtained from these two
models, we initiate a discussion of possible explanations for these differences.
2
HERMES (d’Alcantara and Italianer, 1982) was an ambitious effort to model EU supply-side
responses to the second oil price shock. Early documentation of HERMIN includes Bradley et
al, 1985 and Bradley et al, 1995. For recent HERMIN documentation, see Bradley and Untiedt,
2010.
3
Our inclusion of QUEST-based simulation results from the Fifth Cohesion Report is merely for
the purposes of identifying differences in policy impacts that arise from differences in model
properties. We make no assertion that any one analysis is better than any other.
4
2 THE LOGIC OF “HARD” COHESION POLICY
IMPACT ANALYSIS
We can identify ten separate logical steps in any model-based analysis of cohesion policy
impacts. These can be collected into the two main stages: methodology (stage 1) and
presentation and interpretation of results (stage 2).
Stage 1: Evaluating cohesion policy interventions: methodology
Step 1: Economic theory and public investment: Recent theoretical advances in trade
theory, growth theory and economic geography provide insights that can be drawn on for
the planning and analysis of cohesion policy.
4
These theoretical advances tell us
something about the role of investment in physical infrastructure, human capital, R&D and
innovation. In particular, they suggest ways in which these policies could promote growth
and wealth.
Step 2: Empirics of investment impacts: Given the theoretical insights that are provided in
the trade, growth and spatial literatures, we can then seek to establish what the
international empirical literature tells us about the strength of these drivers of growth and
development in different circumstances. This literature is still at an early stage and it is
easy to become agnostic!
5
What is important is to draw lessons from empirical studies
that provide guidance as to how these driving forces can be related to model mechanisms
and equations that trace through the consequences for changes in sectoral output and
productivity.
Step 3: Why models are needed: The complexity of an economy, with all its internal and
external interactions, and the complexity of cohesion policies means that explicit and
often complex models must be used to evaluate their structural impacts. Without models
it is very difficult to isolate the influences of cohesion policy from all the other factors that
drive growth in a small open economy.
6
In addition, the cohesion policy financial
injections are usually so large that there will be macroeconomic consequences that will
4
Helpman and Krugman (1985), Lucas (1988), Grossman and Helpman (1991), Romer (1994),
Krugman (1995), Fujita, Krugman and Venables (1999), Aghion and Howitt (2005).
5
Some early references include Aschauer (1989), Munnell (1993), Bajo-Rubio and Sosvilla-
Rivero (1993), Herve and Holtzman, 1998, Schalk and Untiedt (2000), Acemoglu and Angrist
(2000), Sianesi and van Reenen (2003), Congressional Budget Office (2005), Romp and de
Haan (2007). More recent research includes de la Fuente (2010 and 2011; Sveikauskas
(2007); Pessoa (2010); Eid (2012).
6
Monitoring should be clearly distinguished from impact evaluation. Monitoring indicators can be
used to show (for example) how much motorway has been constructed, but cannot identify the
role of roadway improvements in boosting output and/or productivity.
5
affect all structural aspects of the economy, and not just the areas that are directly
influenced by the investments (e.g., output and productivity).
Step 4: What kind of macro model: One has to ask the important question of what kind of
model is appropriate for the evaluation of cohesion policy impacts. This will be influenced
by insights into what are the key characteristics of the recipient countries. What kind of
paradigm best captures these characteristics and gives an appropriate description of the
supported country? What level of sectoral disaggregation is required? But it is important
to stress a methodological point here. Economic models are imperfect representations of
the real world. Modern modelling practice has tended to assign high status to frameworks
that incorporate complete rational optimising behaviour and perfect foresight.
7
Such
models are elegant but may trap policy analysts into interpreting policy impacts on the
basis of models that may not represent the realistic behaviour of agents in the real world
(Akerlof, 2005 and 2007). The price of realism may be a lack of complete optimising
elegance!
Step 5: Demand versus supply impacts: Cohesion policy investments have
implementation impacts during programme execution and supply-side impacts both during
but mainly long after the programmes have terminated. One must be careful with how this
distinction is captured in the models. A wide range of other questions also becomes
important. In particular, how are we to handle demand and supply impacts that arise
during implementation and after termination? The recipient states sometimes have rather
specific structural characteristics. Given the known characteristics of the recipient states,
what could be expected in terms of impacts? Total crowding out of private sector activity
as a consequence of the rise in public sector activity? Partial crowding out? Crowding
in? Ricardian equivalence? The answers to these questions are heavily influenced by
the known facts about the economies being aided.
Step 6: Sectoral issues in modelling: A specific and very important issue arises with all
macromodels concerning the level of disaggregation of sectoral production. Cohesion
policy is likely to impact on the various sectors of the economy in very different ways, so
one needs to be aware of how each different model addresses questions of sectoral
disaggregation on the production side of the economy. Can these differences be
subjected to empirical testing? Which approach is more plausible?
7
Bayoumi (2004) describes the IMF DSGE model, GEM; Ratto et al. (2005) describe DG-
ECFIN’s new DSGE implementation of QUEST III.
6
Stage 2: Evaluating cohesion policy interventions: results
Step 7: The “no cohesion policy” counterfactual: The creation of a “no-cohesion policy”
baseline is not trivial. In using a macro model to quantify the impacts of cohesion policy
shocks, all models must go through the following stages:
8
a) Project all non-cohesion policy (CP) exogenous variables out to the terminal year
of the simulation (i.e., world, domestic policy instruments, etc.). For example, in
the case of the analysis of the 2007-2013 cohesion programme, this year might be
taken to be 2020.
b) Set all CP instruments to the appropriate counterfactual values (see below)
c) Simulate the model out to 2020
d) Re-set the cohesion policy instruments to the appropriate actual values
e) Re-simulate the model to 2020
f) Compare results obtained from stage (e) to results from stage (c), to evaluate CP
impacts
However, a range of different “no-cohesion policy” counterfactuals are possible. We can
distinguish three main cases: the “zero” substitution case; the “full” substitution case; and
the “partial” substitution case.
(a) The “zero substitution” case:
Here, domestic authorities do not substitute with domestic finance and cancel the entire
investment programme (usually selected as the default case). In some cases the fiscal
imbalances in a recipient economy would preclude any expansions of public investment.
However, in other cases the national authorities could step in and fund the cohesion
policy investment programme purely out of local resources.
9
Of course, in the latter
situation, there would be more severe fiscal consequences for the public sector budget
balance compared to the case of EU-funded cohesion policy.
8
Hanging over step 7 is the spectre of the so-called “Lucas critique” (Lucas, 1976). The Lucas
critique argues that it is naïve to try to predict the effects of a change in economic policy
entirely on the basis of relationships observed in historical data, especially highly
aggregated historical data. However, models such as HERMIN and QUEST are based
carefully on micro-foundations, albeit in different ways and to different degrees. Model research
has come a long way from the reduced form, time-series models, the use of which Lucas
convincingly destroyed in the 1970s!
9
Countries are expected eventually to grow out of the need for EU development aid. For
example, the Irish cohesion policy funding effectively ended in 2006, having run from 1989 to
2006.
7
(b) The “full substitution” case:
Here, domestic authorities fully implement original CP investments, but finance them
entirely out of their own resources (see discussion above). This could be a mixture of
public expenditure re-allocation to the kinds of investments involved in cohesion policy,
borrowing and tax increases.
(c) The “partial substitution” case:
Here, domestic authorities implement only part of the original CP investments, but
financed out of their own resources.
Very different implications arise from these counterfactuals. For example, in the “zero”
substitution case, impact analysis would attribute to cohesion policy the entire economic
benefits of the CP investments, treating the funding as a grant. In the “full” substitution
case, impact analysis would be identical to the “zero substitution” case, except for the
negative impacts (such as higher tax rates, offsetting cuts in expenditure, higher interest
rates, exchange rate effects, etc.) of the need to finance domestically. Finally, the “partial”
substitution case is difficult to evaluate. If the cancelled cohesion policy investments were
genuine barriers to growth, the outcome might fall well below “full” substitution. If the
cancelled investments were poorly designed (high deadweight/crowding out), then this
case might be actually better than the case of “full” substitution.
Step 8: Policy Impacts for a single country: It is useful to present the empirical results,
initially for a single country so that the presentation can refer to country specifics. The
analysis should then provide a wide range of information aimed at interpreting the
analysis, such as;
a) Present stylised facts about the country model.
b) Present the no-cohesion policy baseline under different assumptions, e.g., zero
substitution and full substitution.
c) Present sensitivity analysis with respect to important model parameters, e.g., the
so-called externality parameters that link changes in stocks of infrastructure,
human capital and R&D to changes in sectoral output and productivity. Discuss
the consequences in terms of what micro-scoring might indicate about the “quality”
of the cohesion policy planning and implementation.
10
10
Bradley, et al, 2006 examines the close, but as yet under-researched, relationship between
macro and micro economic techniques used in policy impact analysis.
8
d) Design the presentation of the results for a given country in a way that facilitates
comparisons with other countries. The concept of a cumulative cohesion policy
multiplier is particularly useful here, defined as:
Change in GDP
Normal policy multiplier = ----------------------------------
Change in public investment
The cumulative policy multiplier (between time t and time t+n) is defined as:
Cumulative percentage change in GDP
Cumulative CP multiplier = -------------------------------------------------------
Cumulative percentage share of CP in GDP
Step 9: Policy Impacts for many countries: In a multi-country evaluation, we need to
present summary results for all the countries, and explore international comparisons,
spillovers and differences.
Step 10: Drawing conclusions. We need to discuss what the evaluations tell us about how
policy can alter the initial structure and characteristics of the economies. Why do different
models produce different results? What can be done about it? This is, of course, an
important question, but it comes at the end of a long list of other issues that also influence
the answers. Only when the question of model-based impact comparisons is placed in the
above wider context can we isolate and rationally explore these differences.
It is useful to enquire into whether there are likely to be strong differences of approach to
these steps as between different modelling groups. To that end, we suggest that the ten
points can be subdivided into two distinct groups. In the first, we suggest that there ought
to be no significant differences of approach between different modelling frameworks. In
the second, unfortunately, strong differences of approach can and do legitimately arise.
9
3 AREAS OF POSSIBLE BROAD AGREEMENT
Within the range of different impact evaluation studies, there are likely to be areas within
the above 10 steps where there is broad agreement. The most obvious cases for
agreement might include the following:
Step 1: Economic theory and public investment: Faced with the challenge of analysing
cohesion policy impacts, all modelling groups dip into new growth theory and economic
geography in order to articulate the theoretical roles of physical infrastructure, human
capital and R&D in promoting faster growth and catch-up. There is likely to be a lot of
common ground here.
Step 3: Why are models needed: All modelling groups tend to accept that the role of
models is to generate policy counterfactual scenarios and place the cohesion policy
interventions in a wider macro context, where macro and other spillover impacts can be
examined.
Step 7: The “no cohesion policy” counterfactual: There should be little or no differences
between modelling groups on the definition of the counterfactuals. However, the
counterfactuals are seldom discussed explicitly and there may be differences of opinion
as to the most appropriate counterfactual to adopt as a standard.
10
4 AREAS OF POSSIBLE DISAGREEMENT
Step 2: Empirics of investment impacts: It is possible that all modelling groups have a
common interpretation of the role of theory in exploring the drivers of growth and catch-
up. However, there may be differences between the groups as to the strength of these
relationships. Here we are focusing on the immediate relationship between (say)
improved physical infrastructure and (say) manufacturing output or manufacturing
productivity. We are not referring to the wider macro-economic outcome that is obtained
when the immediate relationships are embedded in large-scale models. The literature
presents a wide range of options from empirical studies, and is fraught with
methodological and conceptual difficulties. However, even if there were agreement on
what to take from the rather confused empirical literature, there could still be problems.
The structures of the different models often impose differences in the underlying cohesion
mechanisms.
Step 4: What kind of macro model: The most important difference between modelling
groups probably lies in their choice of the modelling framework. This is not to say that
there are any deep, fundamental paradigmatic differences between the models, such as
exist between, say, central planning and market-based economics. All models draw in
varying degrees from recent advances in modelling within the neo-Keynesian and CGE
traditions. All tend to have a significant degree of micro underpinnings and are probably
reasonably robust to the so-called Lucas critique.
Step 5: Demand versus supply impacts: Although the need to distinguish demand
(implementation) effects from long-lasting supply (post-implementation) effects is
accepted by all groups, the empirical analysis can lead to dramatically different outcomes,
mainly due to the issues mentioned in Step 4 (choice of model structure).
Step 6: Sectoral issues in modelling: Under this heading we emphasise the fact that any
detailed examination of cohesion policy impacts needs to be performed at a level of
sectoral disaggregation that permits – at the very least – the separate analysis of key
production sectors such as manufacturing, market services, agriculture and government.
With few exceptions, the main sectoral driver of growth has been manufacturing, or sub-
sectors of manufacturing. The rise of market services from a very low base has also been
a common characteristic of the post-Communist transition of the new EU member states
of the CEE area. Also, the agricultural sector has very specific characteristics that may
serve to distort cohesion policy analysis unless the sector is isolated. Differences of the
degree of sectoral disaggregation may also distort the comparisons of their results.
11
Steps 8-10: Policy impacts: Any different analyses of cohesion policy impacts derived
from models such as QUEST or HERMIN are simply the results of all the divergences in
modelling that are outlined above.
12
5 TWO MODEL-BASED EVALUATIONS OF
COHESION POLICY: 2007-2013
Two model-based ex-ante evaluations of EU cohesion policy were commissioned by DG
Regional Policy in 2009 and formed an input to the Fifth Cohesion Report published in
2010.
11
These evaluations explored the likely impact of the investments funded during
the 2000-2006 and 2007-2013 budgetary programmes. A common set of cohesion policy
financial data was used by both modelling groups: the QUEST model of DG ECFIN and
the HERMIN models of the CSHM developed for DG-REGIO.
Although all models have the potential to examine the impacts of cohesion policy on many
different aspects of economic performance, the impacts on aggregate GDP tends to be
emphasised. Such analysis is usually presented in terms of the comparison of a “with
cohesion policy” scenario relative to a “without cohesion policy” scenario. This distinction
is not without its complications, as we discussed above, and there are a range of
alternative counterfactuals. Using the terminology set out above, both models
implemented the “zero” substitution counterfactual (i.e., domestic authorities do not
substitute with domestic finance, and cancel the entire investment programme).
We reproduce a series of six figures that were published in the Fifth Cohesion Report,
which contain summary comparisions of HERMIN and QUEST-based cohesion policy
impact analysis. These are not ideal as examples of the kind of detailed, country-specific
output that can be derived from such models, but they have the virtue of being in the
public domain.
Figure 1 shows the average annual contribution of the EU element of cohesion funding,
expressed as a percentage of national GDP. The “old” member states were still the main
beneficiaries during the budgetary period 2000-2006, but this situation changed
dramatically for the 2007-2013 budgetary period after the 2004 enlargement (Figure 2).
11
See European Commission (2010), Chapter 4, Section 6 (pages 201-257) for the model
results. See European Commission (2007) for earlier results based on HERMIN and QUEST II.
13
Figure 1: Cohesion policy expenditure relative to GDP, average 2000-2006
Figure 2: Cohesion policy expenditure relative to GDP, average 2007-2013
Figure 3 summarises the impacts on the level of total GDP during what we may term the
policy “implementation” years of programme 2000-2006.
12
It is seen that the impacts
derived from HERMIN are larger in all cases, and significantly larger in the cases of the
Italian Mezzogiorno, East Germany and Ireland. In the cases of the two regional
economies, the differences are probably accounted for by the fact that the HERMIN
models were specifically regional models of the respective national economies (Italy and
Germany). In the case of Ireland, a national economy, the differences remain puzzling.
In the cases of the new member states the comparison is mixed, with QUEST suggesting
a much larger impact on Latvia than HERMIN does, but a much smaller impact for
Hungary, Estonia, Malta and the Czech Republic.
12
Recall that under the so-called “n+2” rule recipients were permitted two extra years, 2007-
2008, to draw down all funding.
14
Figure 3: Estimated impact of cohesion policy expenditure on GDP, average 2000-2009
Figure 4 shows the impacts on GDP for the year 2014, i.e., six years after expenditure
from the 2000-2006 budgetary period ceased.
13
Here there is a dramatic difference
between the HERMIN and QUEST results, with QUEST suggesting much bigger post-
implementation impacts.
13
Of course by the year 2014 the cohesion expenditures from the 2007-2013 budgetary
programme will be flowing. The analysis being described in Figure 4 is focused purely on the
2000-2006 budgetary programme. The models permit us to carry out this kind of
counterfactual analysis. The Commission accounting regulations require separate impact
analysis of each budgetary programme. From an economic point of view this is meaningless!
15
Figure 4: Estimated impact of cohesion policy expenditure on GDP in 2014
In the 5th Cohesion Report, employment impacts were only presented in the case of
HERMIN. This was not because there were no QUEST-based results. Rather it was
because the QUEST analysis suggested that there were only very small employment
impacts during the 2000-2006 budgetary programme implementation years. An
interpretation of this result is that QUEST suggests that the during the implementation
years increased public investment expenditures funded by cohesion policy effectively
crowds out private sector employment by a broadly similar amount to the employment
creation generated by cohesion policy actions.
14
Figure 5: Estimated employment creation induced by cohesion policy expenditure, 2000-2009
14
It was unfortunate that the employment impacts were not presented for QUEST in the 5th
Cohesion Report since there are real issues to be addressed concerning the radically different
policy impacts on employment produced by QUEST and HERMIN.
16
Finally, Figure 6 shows the impacts on GDP of the implementation years of the 2007-
2013 budgetary programme. It is seen that the HERMIN-based impacts are bigger for the
new member states, but equal or smaller for the “old” member states. Although the post-
implementation impacts (i.e., after 2015) were not presented in the report, the pattern of
behaviour is broadly similar to that shown in Figure 4 (i.e., the post implementation
QUEST-based impacts on GDP are larger than the HERMIN-based results).
Figure 6: Estimated impact of cohesion policy expenditure on GDP, 2007-2016
17
6 INTERPRETING DIFFERENCES IN MODEL-
BASED IMPACT ANALYSES
Drawing on the above comparisons of analysis carried out using QUEST and HERMIN,
plus other unpublished results, one is driven to the conclusion that these two models are
based on rather different views of how the economies of the recipient countries behave.
Our diagnosis is that the first key difference between QUEST and HERMIN-based
cohesion impact analysis concerns the manner in which spillover, or externality effects
from the Structural Funds are incorporated into the two models. These were briefly
discussed in Steps 1 and 2 above, but are now examined in more detail.
6.1 MODELLING COHESION POLICY SUPPLY-SIDE
SPILLOVER EFFECTS
In the analysis reported in the 5th Cohesion Report, QUEST and HERMIN used similar
financial data for the cohesion policy shock. The demand-side impact mechanisms are
handled in a similar way, with elements of public expenditure being boosted during the
programmes implementation phase (i.e., 2000-2006(+2) and 2007-2013(+2)). Of course,
both models differ in the modelling of expenditure (private consumption and investment, in
particular), but we will return to this point. There are some differences in the manner in
which the financial data were transformed into changed stocks of physical infrastructure,
human capital and R&D, but these are likely to be minor. The biggest difference is in the
manner in which the improved stocks influence sectoral output and productivity in the
models.
QUEST is essentially a one-production-sector model, with modelling at the level of
aggregate private sector output. Improved stocks of infrastructure and human capital
feed into capacity output in QUEST, through a Cobb-Douglas (CD) production function
that has constant returns to private factor inputs (labour and private capital) and
increasing returns to public capital. Consequently, it is mainly through the consequences
of capacity utilization that QUEST reacts on the supply-side. During the implementation
phase, capacity utilization is driven up, as demand impacts outstrip the more gradual
build-up of new capacity. This seems to generate large crowding-out mechanisms, which
may be further increased by assumptions made on the expenditure side of QUEST.
Production modelling in the HERMIN models of the CSHM is on the basis of five sectors:
manufacturing, market services, building and construction, agriculture and non-market
services (Bradley and Untiedt, 2010). Factor demands in the first three are determined on
18
the basis of cost minimization (using a CES production function constraint). A simpler
approach is used in agriculture, and output in non-market services is policy driven through
employment and wages.
For the important manufacturing sector (and also for the market service and building &
construction sector), HERMIN draws on small open economy modelling research, where
country (capacity) output is not determined directly by a national production function
constraint (as in QUEST). Rather, the national production function appears in the
determination of the national technology (via national factor demand equations), and the
national output equation originated from a higher level "global" production function
(Bradley and FitzGerald (1988)). This approach attempts to capture the essential notion
that integrating within the EU Single Market, and particularly the integration of its
peripheral and weaker economies, is best modelled directly through the
internationalisation of production than indirectly, through trade flows.
Consequently, output determination in manufacturing in HERMIN can be directly
influenced by improved infrastructure, human capital and R&D, through making the
recipient economies more attractive as hosts to inward investment and by strengthening
the internal attractiveness of the competitive environment for locally owned firms. The
international empirical literature is used to provide plausible values for the externality
parameters.
National productivity can also be influenced directly by improved stocks of infrastructure,
human capital and R&D, and these effects are incorporated into the national CES
production functions. In other words, while the output effects are mainly international in
their consequences (affecting the international allocation of production), the productivity
effects are local and serve to modify the local production technology.
15
6.2 OTHER DIFFERENCES BETWEEN QUEST AND HERMIN
Another important difference between QUEST and HERMIN-based analysis lies in the
different nature and strength of crowding out mechanisms, through the labour market
(Philips curve), through fiscal tightening and through monetary tightening. The material
placed in the public domain does not permit a thorough analysis of these issues, so they
need to be explored further. For example, the assumption is made in QUEST that all
increases in productivity are passed on to labour. Consequently, none of the productivity
increases caused by cohesion policy will have any effect in increasing cost
competitiveness in the recipient countries. In HERMIN, on the other hand, empirical
analysis suggests that there can sometimes be a less than full pass-through of
15
See Bradley, Petrakos and Traistaru (2004) for further details.
19
productivity changes to wages. This is quite striking in some countries, such as Poland
during the years immediately prior to and after EU accession. Where there is significant
foreign ownership of firms, this also affects the role of productivity pass-through. Our
judgement is that the strong employment crowding-out features that are hard-wired into
the QUEST DSGE model may not be appropriate to the lagging economies of many of the
new member states. In particular, the “grant” nature of cohesion policy funding, with a
“weakened” concept of additionality, may not be reflected in the QUEST analysis.
Another difference between QUEST and HERMIN is that the former imposes model-
consistent expectations, while the latter uses static (or auto-regressive) expectations.
What this means is that in QUEST agents have perfect (model consistent) information
about the exact future consequences of cohesion policy impacts and consequences, and
can react today in light of tomorrow’s impacts. HERMIN makes no such assertion. Rather,
it takes a pragmatic view that for the analysis of extremely long-tailed structural
investment policies in rapidly transforming economies, the incorporation of model
consistent expectations (MCEs) is probably not justifiable in terms of the context of these
economies. Furthermore, if the basic model set-up is inappropriate, the incorporation of
MCEs simply compounds the initial error and increases the possibility of misinterpretation
of the policy analysis. MCEs are perhaps more appropriate for the analysis of short-term
demand and monetary shocks, where the underlying economic structure is fairly stable
and well understood. There is less justification for their use for long-term supply-side
shocks administered mainly through public investment in productive infrastructure, human
capital and R&D, in a situation where the underlying structure is not well understood and
may be rapidly changing.
On a more technical issue, in QUEST the degree of liquidity constrained consumption
behaviour assumed for the new member state models is 40% compared with 30% in the
“old” member states.
16
Is there strong empirical evidence that the liquidity constraint in
the new member states of the CEE area is so low? With such a low degree of liquidity
constraint, and the assumption of MCEs, it is not surprising that there is so much
crowding out of employment in the QUEST-based analysis.
Another technical issue concerns the nature of the production technology used in QUEST.
It should also be noted that a property of the Cobb-Douglas (CD) production function is
that all factor inputs are substitutes.
17
In a more generalised production function (e.g.,
nested CES, Generalised Leontief, etc.), the possibility arises that public and private
capital might actually be complements. This CD-based restriction may be a factor in the
high crowding out mechanisms that appear to operate within QUEST.
16
In other words, 40 per cent of households are assumed to be liquidity constrained, and the
remaining 60% can be modelled in terms of (forward-looking) permanent income.
17
The assumption is also made in QUEST that the marginal product of public capital stock (K
pub
)
is the same as the marginal product of private capital stock (K
priv
).
20
7 CONCLUDING REMARKS
It is possible to attempt to pinpoint more accurately those aspects of the QUEST and
HERMIN model frameworks that may be driving the rather significant differences in their
implications for the analysis of the impacts of cohesion policy. We stress that we offer
merely initial insights for the purposes of stimulating further discussion. Macro-models
are very complex tools and are intrinsically difficult to compare. In addition, their
application can go far beyond measuring growth effects of cohesion policy. The most
active area of design and analysis of cohesion policy are currently the former centrally
planned states of Eastern Europe, where the relevant modelling culture is weak and one
has access to time series data only from the mid-1990s.
We conclude with the observation that the impact analysis of cohesion policy
interventions is very complex and the final results published are often determined by a
series of hidden decisions taken by the modellers which may not be completely
transparent. To some extent, the criticism of the Barca report noted in our introduction is
correct, but nevertheless the use of a complex tool for policy diagnosis and impact
evaluation should not be excluded or replaced by a “softer” method simply because the
“layman” does not understand it. But to be able to use models and judge the results, it is
absolutely necessary to be fully transparent concerning the exact set-up of the models.
Otherwise, cohesion policy impact analysis using macro-economic models will continue to
be an impenetrable “black-box” and the theoretical advantage claimed for macro-models,
i.e., to be able to look at cohesion policy impacts in a way that takes into account the
specific and realistic economic relations within the recipient countries and their linkage to
the rest of the world, will not be realised in practice.
However, even the 5th Cohesion Report, which made limited use of HERMIN and QUEST
results, tends to sell the benefits of macro model-based analysis short. For example, in
the report it was stated that:
Like any evaluation method, macroeconomic models have their strengths but need to be used
with other evaluation methods to complete the picture. This especially so, since Cohesion Policy
has goals which go much further than only GDP growth (5th Cohesion Report, page 248)
The impression is left that models can only quantify impacts on GDP. However, models
can be used to examine the deep structural impacts on economies, including GDP,
employment, unemployment, productivity, wage rates, labour force participation, sectoral
structure of production, household income, competitiveness, etc. The fact that these
kinds of insights have not found their way into public policy discourse is as much the fault
of the policy makers as it is of the model-using policy analysts and researchers.
21
So, on the basis of the above limited output from HERMIN and QUEST, what can we
conclude about the question posed in the paper’s title? First, both models suggest that
the investment programmes financed by cohesion policy do benefit the recipient
economies in terms of raising the level of GDP, both during and after programme
implementation. However, the boost to GDP growth is rather modest. In the case of
HERMIN, the boost to growth is transitory and growth reverts to what one might term a
“normal” rate of growth determined by the underlying structure of the recipient economy,
its investment and trading relationships with the rest of the EU and world, and other non-
cohesion policy issues. In the case of QUEST, the endogenous growth mechanisms
manifest themselves more strongly, promising a more enduring boost to growth.
Second, it has been suggested that models like HERMIN and QUEST do not prove that
cohesion policy is beneficial, but that the beneficial effects are simply “hard-wired” into the
model structures. However, as we discussed earlier, the crucial stage of research in this
area takes place off-model, in the growing economic literature that seeks to explain more
precisely how improved physical infrastructure, human capital and innovation/R&D can
accelerate growth and development in lagging economies. If this literature were to point
to cohesion policy ineffectiveness, at the detailed level of individual measures and
operational programmes, then that ineffectiveness would also appear at the more
aggregate level in the macro-models. However, such research tends to be available in
the more advanced EU economies, but is singularly lacking in the new EU member states
who are the main targets of cohesion policy. This research gap, rather than differences
between model structures, is the current weakness of model-based cohesion policy
analysis.
The fact that models are complex and that the answers they generate are also complex is
a fact of life that we would be unwise to wish away. As the Irish playwright Oscar Wilde
said: “The truth is rarely pure and never simple”.
22
BIBLIOGRAPHY
Acemoglu, D., Angrist, J. (2000), “How Large are Human Capital Externalities? Evidence
from Compulsory Schooling Laws,” Bernanke, B., Rogoff, K. (eds.), NBER
Macoeconomics Annual, 2000. Cambridge (MA): MIT Press.
Aghion, P., Howitt, P. (2005), “Appropriate Growth Policy: A Unifying Framework”,
http://www.economics.harvard.edu/faculty/aghion/papers.html
Akerlof, G. E. (2007), “The missing motivation in Macroeconomics”, American Economic
Review 97(1), pp. 5-36.
Akerlof, G. E. (ed.) (2005), “Explorations in Pragmatic Economics: Selected Papers of
George A. Akerlof (and Co-Authors)”, Oxford and New York: Oxford University Press.
Aschauer, D. (1989), “Is Public Expenditure Productive?”, Journal of Monetary
Economics, Vol. 3, pp. 177-200.
Bajo-Rubio, O., Sosvilla-Rivero, S. (1993), “Does public capital affect private sector
performance? An analysis of the Spanish case, 1964-1988”, Economic Modelling, 10(3),
pp. 179-185.
Barca, F. (2009), An Agenda for a Reformed cohesion policy. A place-based approach
to meeting European Union challenges and expectations, Independent Report prepared
at the request of Danuta Hübner, Commissioner for Regional Policy.
(http://ec.europa.eu/regional_policy/policy/future/seminars/barca_sem_220609_en.htm)
Bayoumi, T. (2004). A New International macroeconomic Model, Occasional Paper 239,
International Monetary Fund, Washington.
Bradley, J., C. Fanning, C. Prendergast and M. Wynne (1985). Medium-Term Analysis of
Fiscal Policy in Ireland : A Macroeconomic Study of the Period 1967-1980, Research
Paper No. 122, Dublin: The Economic & Social Research Institute.
Bradley, J., FitzGerald, J. (1988), "Industrial output and factor input determination in an
econometric model of a small open economy," European Economic Review 32(6), pp.
1227-1241.
Bradley, J., J-A. Herce and L. Modesto (Guest Editors) (1995). Special Issue: The
HERMIN Project, Economic Modelling, Vol. 12, No. 3, July.
Bradley, J., Mitze, T., Morgenroth, E., Untiedt, G. (2006), “How can we know if EU
cohesion policy is successful? Integrating micro and macro approaches to the
evaluation of Structural Funds”, GEFRA Working Paper, Muenster, March.
(http://www.gefra-muenster.org/deutsch/publikationen/abstract.php?pub_id=96)
Bradley, J., Petrakos, G., Traistaru, I. (eds.) (2004), Integration, Growth and cohesion in an
Enlarged European Union, New York: Springer.
Bradley, J., Untiedt, G. (2010), The COHESION System of Country and Regional
HERMIN Models: Description and User Manual, Report prepared for the European
Commission, DG-Regional Policy, Brussels.
23
Congressional Budget Office (2005), R&D and Productivity Growth, Congressional
Budget Office, The Congress of the United States, June.
D’Alcantara, G. and A. Italianer (1982). European Project for a Multinational
Macrosectoral Model, Commission of the European Communities, Brussels.
De la Fuente, Angel (2010), Infrastructures and Productivity: An updated survey. BBVA
Working Paper Number 10/18.
www.bbvaresearch.com
De la Fuente, Angel (2011), Human Capital and Productivity. Barcelona GSE (Graduate
School of Economics) Working Paper No. 530
http://research.barcelonagse.eu/Working_Papers.html
Eid, A. (2012), Higher education R&D and productivity growth: an empirical study on high-
income OECD countries. Education Economics, Vol. 20, p. 53-68.
European Commission (2007), Fourth Progress Report on Economic and Social
cohesion.
(
http://ec.europa.eu/regional_policy/sources/docoffic/official/reports/interim4_en.htm
).
European Commission (2010). Investing in Europe’s Future: Fifth Report on Economic,
Social and Territorial Cohesion, Brussels.
(http://ec.europa.eu/regional_policy/sources/docoffic/official/reports/cohesion5/index_en.cfm)
Fujita, M., Krugman, P., Venables, A. (1999). The Spatial Economy: Cities, Regions, and
International Trade, Massachusetts: The MIT Press.
Grossman, G., Helpman, E. (1991), Innovation and Growth, Cambridge: The MIT Press.
Helpman, E., Krugman, P. (1985), Market Structure and Foreign Trade: Increasing Returns,
Imperfect Competition and the International Economy, Cambridge, Massachusetts: MIT
Press.
Herve, Y. and R. Holtzmann. (1998). Fiscal Transfers and Economic Convergence in the
EU: An Analysis of Absorption Problems and an Evaluation of the Literature, Baden-
Baden: NOMOS Verlagsgesellschaft.
Krugman, P. (1995), Development, Geography, and Economic Theory, Cambridge,
Massachusetts: The MIT Press.
Lucas, Robert (1976), "Econometric Policy Evaluation: A Critique", in Brunner, K.,
Meltzer,
A., The Phillips Curve and Labor Markets, Carnegie-Rochester Conference Series on
Public Policy, 1, New York: American Elsevier, pp. 19–46
Lucas, R. E., (1988), “On the Mechanics of Economic Development,” Journal of Monetary
Economics 22(1), pp. 3-42.
Munnell, A. (1993), “An assessment of trends in and economic impacts of infrastructure
investment”, in Infrastructure policies for the 1990s, Paris, OECD.
Pessoa, A. (2010), R&D and Economic Growth: How strong is the link?, Economics
Letters, Vol. 102/2, p. 152-154.
Ratto, M., Roeger, W., in’t Veld, J., Girardi, R. (2005), An Estimated New Keynesian
Dynamic Stochastic General Equilibrium Model for the Euro Area, European Economy
Economic Papers, No. 220, Brussels, January.
24
Romer, P. (1994), “The Origins of Endogenous Growth”, The Journal of Economic
Perspectives, 8(1), pp. 3-22.
Romp, W., de Haan, J. (2007), “Public Capital and Economic Growth: A Critical Survey”,
Perspektiven der Wirtschaftspolitik, Volume 8 (Special Issue), pp. 6-52.
Schalk, H.J., Untiedt, G. (2000), “Regional Investment Incentives in Germany: Impacts on
Factor Demand and Growth”, The Annals of Regional Science 34, pp. 173-195.
Sianesi, B., van Reenen, J. (2003), “The Returns of Education: Macroeconomics,” Journal
of Economic Surveys 17, pp. 157-200.
Sveikauskas, Leo (2007), R&D and Productivity Growth: A Review of the Literature, U.S.
Buerau of Labor Statistics Working Paper No. 408,
www.cbo.gov/publication/16635
