an assessment of the 1994-99 Objective 1 CSF The effect of ... file1 The effect of Structural Fund spending on the Spanish regions: an assessment of the 1994-99 Objective 1 CSF Angel

1

Th

e effect of Stru

ctural Fu

nd

spen

din

g on th

e Sp

anish

regions:

an assessm

ent of th

e 1994-99 Ob

jective 1 CS

F

An

gel de la Fu

ente*

Institu

to de A

nálisis E

conóm

ico (CS

IC)

First version, 1999

Th

is version, S

eptem

ber 2002

Ab

stractT

his pap

er analyzes the growth and

emp

loyment effects of the 1994-99 C

omm

unity

Support Framew

ork (CSF) for the O

bjective 1 Spanish regions using a simple supply-

side m

odel estim

ated w

ith a panel of regional d

ata. The resu

lts suggest that the

imp

act of the Structu

ral Fund

s in Spain has been qu

ite sizable, add

ing around

ap

ercentage point to annu

al outp

ut grow

th in the average Objective 1 region and

0.4p

oints to emp

loyment grow

th. Over the p

eriod 1994-2000, the Fram

ework has

resulted in the creation of over 300,000 new

jobs and has elim

inated 20%

of the initialgap in incom

e per capita between the assisted

regions and the rest of the country.

________________________________ * T

his pap

er is part of a research p

roject cofinanced by the E

urop

ean Regional D

evelopm

ent Fund

and Fu

ndación C

aixa Galicia. A

dd

itional financial support from

the Spanish Ministry of Science and

Technology u

nder grant SE

C99-1189 is also gratefu

lly acknowled

ged. I w

ould

like to thank Juan

Varela, T

eresa Dabán and

Antonio D

íaz (Spanish M

inistry of Finance), Antonio A

vila (Junta d

eA

ndalu

cía), Juan A

res and M

elchor Fernández (U

niversidad

de Santiago d

e Com

postela) and A

liciaA

vilés (Universid

ad d

e Málaga) for their com

ments and

suggestions and

for their help in gathering

the data used

in this study. T

he two researchers cited

last collaborated actively in preparing A

ppendix

2.Correspondence to Institu

to de A

nálisis Económ

ico (CSIC

), Cam

pu

s de la U

niversidad

Au

tónoma d

eB

arcelona, 08193 Bellaterra, B

arcelona, Spain. T

el: 34-93-580-6612. Fax: 34-93-580-1452. E-M

ail address:A

ngel.DeL

[email protected].

2

1. Introd

uction

The Stru

ctural Fu

nds are the m

ost imp

ortant instrum

ent of the Eu

ropean U

nion'sregional cohesion p

olicy. They channel a large volu

me of resou

rces aimed

atp

romoting the d

evelopm

ent of the poorest regions of the U

nion through the

correction of existing deficiencies in end

owm

ents of strategic prod

uction factors,

such as infrastructures and hum

an capital, and through aid

to private enterprises.

Given the im

portance of the Stru

ctural Fu

nds, the evalu

ation of their imp

act isnecessary, not only in ord

er to satisfy the control requirem

ents of the Eu

ropean

Com

mission, bu

t also as an imp

ortant ingredient in p

olicy planning and

design. A

tthe m

acroeconomic level, the aim

of such evalu

ation mu

st be to estimate the joint

impact of the d

ifferent projects and program

mes co-financed

by the EU

on aggregateeconom

ic indicators su

ch as regional outp

ut, em

ploym

ent and p

rivate investment,

and to analyze the relative effectiveness of d

ifferent types of structural expenditure.

Most previous attem

pts to quantify the impact of the Structural Fund

s have relied on

conventional country-level m

acroeconometric m

odels. 1 T

hese mod

els are probably

the best available tool for the analysis of the short- and m

ediu

m-term

effects ofC

omm

un

ity policies th

rough

their im

pact on

aggregate dem

and

. In gen

eral,how

ever, they cannot be used

to prod

uce regional-level estim

ates and are not

especially w

ell suited

for the analysis of the sup

ply-sid

e effects that are sought by

structu

ral interventions because their prod

uction blocks are not d

esigned to captu

resuch effects. 2

1 See for instance Brad

ley, Whelan and

Wright (1995), M

odesto and

Neves (1995), H

erce and Sosvilla-

Rivero (1995), B

radley, H

erce and M

odesto (1995), and

Christod

oulakis and K

alyvitis (2000) for impact

evaluations that m

ake use of the H

ER

MIN

family of m

odels, and

Roeger (1996) for an exercise based

on the European C

omm

ission's QU

EST

II mod

el.2 For instance, in the H

ER

MIN

mod

els the original production function includ

es only physical capitaland

labour as inpu

ts. To captu

re the effects of infrastructu

res and hu

man capital, the scale param

eterin the prod

uction fu

nction is re-specified as a fu

nction of the stocks of these factors and "reasonable"

values of the relevant elasticities are chosen on the basis of existing results in the literature. In some of

the QU

EST

simu

lations (Roeger, 1996) all C

SF expend

iture is treated

as having the same effects as

investment in physical capital.

3

In this paper I will prod

uce regional estim

ates of the impact of the Stru

ctural Fu

nds

using a m

odel that is sp

ecifically designed

and estim

ated to cap

ture the relevant

sup

ply effects. T

he mod

el has two basic ingred

ients. The first one is an aggregate

prod

uction fu

nction which relates regional ou

tpu

t to the level of emp

loyment, the

stocks of prod

uctive factors (in

frastructu

res, other p

hysical cap

ital and

the

edu

cational attainment of the w

orkforce) and to the level of technical efficiency. T

hesecond

comp

onent of the mod

el is an emp

loyment equ

ation which d

escribes theevolution of this variable as a function of changes in factor stocks and

wage rates.

One shortcom

ing of this app

roach is that the mod

el does not take into accou

ntd

emand

effects that can be quite im

portant in the short ru

n and fails to cap

ture

indu

ced changes in p

rices and w

ages that may p

artially offset the direct su

pp

lyeffects of stru

ctural interventions. I w

ill try to partially overcom

e this limitation by

making use of an investm

ent equation to estimate the response of private investm

entto the relevant policy shock.

The m

odel w

ill be estimated

using a panel of Spanish regional data, and

will be used

to prod

uce an estim

ate of the imp

act of the Structu

ral and C

ohesion Fund

s on thegrow

th of output and em

ployment in the regions of Spain that are currently includ

edin O

bjective 1 du

e to their low incom

e levels. I will focu

s in particu

lar on them

acroeconomic effects of the 1994-99 O

bjective 1 Com

mu

nity Sup

port Fram

ework

(CSF) w

hich encompasses m

ost of the regional developm

ent projects that have beencofinanced

by the Eu

ropean U

nion du

ring this period

. I will also com

pu

te "social"rates of retu

rn that sum

marize the m

arginal contribution to the grow

th of regionalou

tpu

t of each of the four broad

pu

blic expend

iture item

s that make u

p the bu

lk ofthe Fram

ework: investm

ent in prod

uctive infrastru

ctures and

in other types of

physical capital, subsidies to private firm

s, and training expend

iture.

The analysis is repeated

under tw

o different scenarios. In the first one, I w

ill take theC

SF at face value and assum

e that it adequately d

escribes all the relevant investment

flows. T

his amou

nts to the assum

ptions that i) none of the p

ublic or p

rivateinvestm

ent projects includ

ed in the C

SF wou

ld have been u

ndertaken in its absence,

and ii) that C

SF-related exp

enditu

res have had no ad

ditional effect on p

rivateinvestm

ent behaviour. In the case of public investment, assum

ption i) is probably thenatu

ral one to make if the objective is to m

easure the im

pact of these resou

rcesind

epend

ently of their true "ad

ditionality". In the case of p

rivate investment,

however, it seem

s preferable to estimate the m

arginal increase indu

ced by stru

ctural

program

mes. Id

eally, this should

be done by estim

ating a private investm

ent

4

function w

ith regional level data. U

nfortunately, this is not feasible d

ue to d

atalim

itations and, in particu

lar, to the lack of regionally desaggregated

information on

subsidies and

other aids to enterprises. T

o get around this d

ifficulty, I will rely on an

investment fu

nction estimated

with national d

ata for a samp

le of OE

CD

countries,

and extrap

olate the results to the regional case. A

lthough the exercise is certainly

risky, it should

provide an ed

ucated

guess on the im

pact of the Structu

ral Fund

s onp

rivate investment, and

it does serve as a w

arning that taking CSF d

ata on private

co-financing as estimates of ind

uced private investm

ent is probably not a good id

ea.

The p

aper is organized

as follows. Section 2 ou

tlines the econometric m

odel and

presents the results of its estim

ation. Section 3 quantifies the contribu

tion of the CSF

to the accum

ulation of d

ifferent prod

uctive factors in the sam

ple. Fu

rther details on

both issues are p

rovided

in the Ap

pend

ices. Imp

act estimates are p

resented in

Sections 4 and 5. Section 4 focuses on short-run effects, and

Section 5 contains med

imand

long-run im

pact estim

ates that take into account d

epreciation and

the sluggish

response of employm

ent to positive supply shocks. Section 6 concludes.

2. Meth

odology an

d d

ata

The im

pact estim

ates I will p

resent below are based

on an aggregate prod

uction

function and

on an employm

ent equation that allow

s for the existence of adju

stment

costs in an ad-hoc fashion. T

he production function is assum

ed to be of the form

(1) yit =

θl ait +

θk kit +

θp p

it + θ

h hit +

θl lit

where y is (the logarithm

of) aggregate regional outpu

t, l (the log) of employm

ent, k,p and

h are the logs of the stocks of physical cap

ital, infrastructu

res and hu

man

capital and a is an ind

icator of technical efficiency or total factor produ

ctivity (TFP

).T

he parameters θ

i (with i =

l, k, h and p) m

easure the elasticity of output with respect

to the stocks of the different p

rodu

ctive factors. A 1%

increase in the stock ofinfrastru

ctures, for instance, w

ould

increase regional outp

ut by θ

p %, hold

ing

constant the stocks of the other factors and the level of technical efficiency.

Setting the marginal p

rodu

ct of labour equ

al to the real wage and

rearranging, we

obtain a labour dem

and sched

ule of the form

(2) lt * =

11-θl (ln θ

l + θ

l ait + θ

k kit + θ

h hit +

θp p

it - wit )

5

where w

is the log of the real wage. T

his function w

ould

describe aggregate labou

rd

emand

under perfectly com

petitive conditions in prod

uct and factor m

arkets in theabsence of em

ploym

ent adju

stment costs. Since this last assu

mp

tion is clearlyinap

prop

riate, I will interp

ret (2) as a long-term d

emand

schedu

l and assu

me that

emp

loymen

t adju

sts gradu

ally toward

s the level given

in th

is expression

. Inparticu

lar, I will assu

me that the grow

th rate of employm

ent, ∆lt , is a fu

nction of thegrow

th of the long-term d

emand

for labour (∆

lt *) and of the existing gap

between

actual em

ploym

ent and its op

timal long-term

level (lt * - lt ), as described

by thefollow

ing equation:

(3) ∆lt =

- d + γ1 ∆

lt * + γ2 (lt * - lt )

where d d

enotes the exogenous rate of em

ployment d

estruction. C

ombining (2) w

ith(3), the short-term

elasticity of emp

loyment w

ith respect to the stock of factor i w

illbe given by

(4) λi =

γ1 θi

1-θl .

My estim

ates of the CSF's short-term

impact on em

ployment w

ill be obtained as the

produ

ct of the increases in (log) factor stocks attributable to the Fram

ework and

therelevant elasticities given in (4). N

otice that this proced

ure assu

mes im

plicity that

the imp

lementation of the C

SF has no imp

act on the evolution of real w

ages.O

therwise, the net grow

th of emp

loyment w

ould

be the difference betw

een therep

orted estim

ates and the loss of em

ploym

ent du

e to the increase in real wages

induced

by EU

structural expenditure.

Tab

le 1: Estim

ated valu

es of the m

ain p

arameters of in

terest________________________________________________________________________

parameter

coeff.(t)

parameter

coeff.(t)

θk0.297

(5.73)θl

[0.597]

θp0.106

(2.14)γ1

0.181(6.47)

θh0.286

(7.30)γ2

0.040(5.21)

________________________________________________________________________

- Note: t statistics in p

arentheses next to each coefficient. The coefficient of em

ploym

ent, θl , is not

estimated

directly but recovered

from the assum

ption of constant returns to scale and the estim

ates ofthe other param

eters, with θl =

1 - θk - θ

p .

6

Ap

pend

ix 1 describes in greater d

etail the joint estimation of equ

ation (3) and a

dynam

ic version of equation (1) w

hich allows for regional fixed

effects and for

technological diffusion. T

he main results are sum

marized

in Table 1. M

y estimates of

the prod

uction fu

nction param

eters are generally consistent with those obtained

inother stu

dies w

ith Spanish regional d

ata. 3 This is also tru

e for the coefficient ofinfrastructure (θ

p ), which is a priori the m

ost problematic param

eter, given its crucial

relevance for the comp

utations that follow

and the lack of consensu

s in the recentliterature on the subject. In the case of Spain, how

ever, most existing stud

ies (both atth

e nation

al and

at the region

al level) tend

to confirm

the sign

ificance of

infrastructu

re variables even with p

anel specifications w

hich allow for u

nobservedregional effects -- w

hich is often not the case for the U.S. and

other samp

les. One

possible exp

lanation for this difference is that the Sp

anish data on regional cap

italstocks are p

robably of better quality and

cover a longer period

than those availablefor other cou

ntries. A second

possibility, for w

hich there is some circu

mstancial

evidence, is that there m

ay be some sort of "satu

ration" effect in connection with

infrastructu

re, so that its contribution to p

rodu

ctivity may be greater in the case of

Spain than in other countries with m

ore adequate stocks of this factor.

The m

odel is estim

ated using regional panel d

ata for the period 1964-93. T

he data on

regional emp

loyment (nu

mber of jobs), ou

tpu

t (gross value ad

ded

, GV

A) and

wage

costs are taken from the p

ublication of Fu

ndación B

BV

Renta nacional de E

spaña y su

distribución provincial, and

come at in

tervals of generally tw

o years (with

one

exception w

here it is three). 4 The d

eflator for regional outp

ut is constru

cted u

singnational p

rice indices for fou

r large sectors to account for d

ifferences across regionsin the sectoral com

position of outpu

t. The series on regional factor stocks have been

constru

cted by th

e Institu

to Valen

ciano d

e Investigacion

es Econ

omicas an

dp

ublished

by Fund

ación BB

V (1998) and

Fund

acion Bancaja (M

as et al, 1998). As a

proxy for the stock of hu

man cap

ital, I use the fraction of the em

ployed

pop

ulation

with at least som

e secondary schooling. T

he (net) stock of physical cap

ital, which is

measu

red in m

illions of 1990 pesetas, is broken d

own into tw

o comp

onents. The

infrastructu

re comp

onent (p) includ

es pu

blicly financed transp

ortation networks

(roads and

highways, p

orts, airports and

railways), w

ater works, sew

age, urban

3 See for instance Mas, M

aud

os, Pérez and

Uriel (1995), d

e la Fuente and

Vives (1995), G

onzález-Páram

o and A

rgimon (1997) and

Dabán and

Lam

o (1999).4 G

VA

data are p

rovided

at factor cost. In the case of the agricultu

ral sector, I have ded

ucted

fromrep

orted ou

tpu

t an estimate of the volu

me of E

U agricu

ltural su

bsidies w

hich is taken from C

orreaand

Malu

quer (1998). W

ithout this correction, the ap

parent p

rodu

ctivity of agricultu

re disp

lays anextrem

ely sharp increase following Sp

ain's accession into the EU

which continu

es to be noticeable atthe aggregate level in som

e regions.

7

structu

res and p

rivately-financed toll highw

ays. The stock of non-infrastru

cture

capital (k) inclu

des p

rivate capital, exclu

ding resid

ential housing, and

the stock ofp

ublic cap

ital associated w

ith the provision of ed

ucation, health and

generalad

ministrative services. T

hese last three items are aggregated

with the cap

ital stockof the p

rivate sector because m

y outp

ut m

easure inclu

des governm

ent-provid

edservices and

the available information d

oes not allow a consistent segregation of this

sector. 5

3. Th

e CS

F's contrib

ution

to factor accum

ulation

Given the estim

ated m

odel, the calculation of the im

pact of the Structural Funds only

requires an estimate of the contribution of the C

SF to the accumulation of prod

uctivefactors in each region. C

onstructing su

ch an estimate w

ould

be a simp

le matter if a

detailed

breakdow

n of CSF exp

enditu

re by region and by fu

nctional category were

available but, u

nfortunately, the existing d

ata on Structu

ral Fund

disbu

rsements is

far from ad

equate. 6

After exp

loring the available sources, I have chosen to base m

y calculations on a

Provisional Financial P

lan (PFP

) for the Objective 1 C

SF which com

bines data on

disbu

rsements u

ntil 1997 and on p

lanned exp

enditu

res for the rest of the relevantp

eriod to p

rovide overall com

mitm

ent targets for the entire 1994-99 period

. These

totals are broken dow

n by Fund

, by functional head

ing and su

bheading and

bysou

rce of financing (Structu

ral Fund

grants and national p

ublic and

private co-

financing). 7

On

e imp

ortant lim

itation of th

is source is th

at a significan

t fraction of C

SFexp

enditu

re is not allocated am

ong regions. The P

FP breaks d

own the C

SF into a"M

ultiregion

al Su

bframew

ork," w

hich

is

executed

by

the

Span

ish

Cen

tralG

overnment, and

a set of "Regional Su

bframew

orks," one for each autonom

ous

5 In add

ition, I am not su

re that focusing only on p

rivate sector outp

ut w

ould

be a good id

ea as thisw

ould

leave out su

bstantial benefits from investm

ent in pu

blic edu

cation and health care. T

hep

rocedu

re I have chosen, however, im

plicitly assu

mes that the p

rivate and p

ublic sectors have a

similar prod

uction fu

nction. My gu

ess is that this is probably not a bad assu

mption, at least if pu

blicservices could

be somehow

valued at m

arket prices.6 T

he Ministry of Finance d

oes provid

e relatively detailed

data on E

RD

F grant disbu

rsements by

region and by type of expend

iture, but there is little systematic inform

ation in this or other sources onregional and

private co-financing rates, and on the expend

itures of other Fu

nds (especially the Social

and A

gricultural Funds).

7 The relevant Fu

nds are the E

urop

ean Regional D

evelopm

ent Fund

(ER

DF), the E

urop

ean SocialFu

nd (E

SF), the Gu

idance Section of the E

urop

ean Agricu

ltural G

uid

ance and G

uarantee Fu

nd(E

AG

GF), the Financial Instrum

ent for Fisheries Guid

ance (FIFG) and

the Cohesion Fund

.

8

region or city, which are carried

out ty the regional adm

inistrations. Since the first ofthese Subfram

eworks finances projects in all O

bjective 1 regions and no geographical

breakdow

n is provid

ed, I have had

to construct it u

sing information from

various

souces to estimate regional expend

iture shares for each of the European Fund

s. Sincethis inform

ation is available for each Fund

only at the aggregate level (and not by

heading and

subhead

ing), I have had to assu

me that the fu

nctional comp

osition ofth

at part of each

Fun

d's exp

end

iture th

at is inclu

ded

in th

e Mu

ltiregional

Subframew

ork is the same for all regions.

A second

problem

is that the PFP

does not p

rovide any inform

ation about the

"physical ou

tpu

t" (in man-years of training) of the hu

man resou

rce program

mes

financed by E

U grants. Since these figures are need

ed for the im

pact calculations andI cou

ld find

no other sources that p

rovided

reliable and reasonably com

plete

information, I have had

construct w

hat is und

oubted

ly a very rough estim

ate of thetotal nu

mber of m

an-years of training financed by the C

SF in each region. This

estimate is obtained

by divid

ing regional expend

iture on d

ifferent types of training

program

mes by an estim

ate of their unit costs (p

er man year of training) that has

been constructed using d

ata from tw

o intermed

iate evaluation reports for the human

resources p

rogramm

es includ

ed in the regional Su

bframew

orks for And

alucía and

Galicia.

The d

etails of the calculations I have ju

st sketched can be fou

nd in A

pp

endix 2.

Tables 2-4 su

mm

arize the results. A

fter exclud

ing some m

inor items (those that

finance technical assistance, evaluation programm

es and em

ployment subsid

ies), thevariou

s headings and

subhead

ings in the CSF are grou

ped

into five expend

iture

categories or programm

es according to their econom

ic nature: 8 p

ublic investm

ent inprod

uctive infrastructures (infraest), public investment other types of physical capital

(pubinv), subsid

ies to the private sector (subs), p

ublic exp

enditu

re in training anded

ucation (training), and

the private co-financing of investm

ent projects su

bsidized

by the EU

(private). 9 This breakd

own w

ill be used

below to ap

proxim

ate the CSF's

contribution to the accu

mu

lation of the stocks of the inpu

ts that enter the regionalprod

uction function (physical and hum

an capital and infrastructures).

8 See Append

ix 2 for more d

etails.9 In

frastructu

re expen

ditu

re inclu

des p

ublic in

vestmen

t in tran

sportation

, water su

pp

ly and

environmental p

rotection, as well as the C

ohesion Fund

. Pu

blic investment in non-infrastru

cture

capital

inclu

des

expen

ditu

re in

ed

ucation

an

d

health

-care facilities

and

on

en

ergy an

dtelecom

mu

nications, all of which are inclu

ded

in the stock of non-infrastructu

re capital (k) in the

production function.

9

Tab

le 2: Fun

ctional an

d region

al comp

ositionof th

e expen

ditu

re chan

neled

throu

gh th

e Ob

jective 1 CS

FA

nn

ual average

__________________________________________________________________________________________infraest

pubinvsubs

trainingtot. pub.

privatetotal

Andalucía

98,28127,175

44,58329,346

199,38448,203

247,588A

sturias25,375

5,97310,205

7,91549,469

9,53159,000

Canarias

23,4505,720

9,3647,522

46,0569,233

55,288C

antabria10,365

4,1737,422

3,53025,491

6,83732,328

Castilla y L

eón55,474

16,61528,519

17,072117,679

23,915141,594

Castilla la M

ancha31,571

5,75918,879

9,70165,911

15,24281,152

Valencia

45,89111,427

16,04115,690

89,04917,102

106,151E

xtremadura

15,1834,660

12,02610,793

42,6638,743

51,406G

alicia60,055

13,22036,494

11,549121,319

36,523157,841

Murcia

17,5695,393

7,8296,789

37,5798,390

45,969total O

bj. 1383,214

100,115191,364

119,907794,599

183,717978,317

% of total

39.17%10.23%

19.56%12.26%

81.22%18.78%

100.00%__________________________________________________________________________________________ -N

otes: Average annual expend

iture over the period 1994-2000 in m

illions of 1990 pesetas. All figures

are deflated

using the Spanish GD

P d

eflator. tot pub is total public expenditure per year, calculated

asthe su

m of the previou

s colum

ns. Total annu

al expenditu

re is shown in the last colu

mn and

includ

esalso private cofinancing (priv).

Table 2 su

mm

arizes the functional and

regional composition of the O

bjective 1 CSF.

For each functional category or exp

enditu

re program

me, the table show

s averageannu

al expend

iture in each region m

easured

in millions of 1990 p

esetas and the

weight of the item

in total aggregate expenditu

re (which is show

n in the last row of

the table). Average annu

al expend

iture is calcu

lated u

nder the assu

mp

tion that allthe resou

rces allocated to the C

SF are disbu

rsed over the p

eriod 1994-2000, that is,

add

ing one year to the theoretical du

ration of the Framew

ork to correct for theobserved

delay in its execu

tion. 10 The figu

res shown in the table refer to total C

SFexp

enditu

re rather than to EU

subsid

ies. In particu

lar, pu

blic expend

iture inclu

des

the contributions of the variou

s levels of the Spanish ad

ministration in ad

dition to

grants from the E

U, and

private exp

enditu

re refers to the (declared

) private sector

contribution to the financing of CSF-supported

projects.

The volu

me of resou

rces chanelled throu

gh the Framew

ork is quite su

bstantial.T

ranslated

in

to eu

ros of

2001, average

ann

ual

CSF

expen

ditu

re cam

e to

app

roximately 9 billion. 11 Investm

ent in prod

uctive infrastru

ctures accou

nts forabout 40%

of total expenditure and

half of the available public financing. Subsidies to

10 See Append

ix 2.11 T

o convert 1990 pesetas into 2001 euros, the figures in Table 2 m

ust be multiplied

by 0.009341.

10

private activities are the next largest item. P

ublic expend

iture accou

nts for over 80%of the overall bu

dget. T

he largest recipients of regional aid

in absolute term

s areA

ndalu

cía (which absorbs 25.5%

of total expenditu

re), Galicia (16.1%

) and C

astilla yL

eón (14.5%).

Tab

le 3: Average an

nu

al expen

ditu

re per cap

ita(average for th

e Ob

jetive 1 regions =

100)__________________________________________________________________________________________

infraestpubinv

substraining

tot. pub.private

totalA

ndalucía84.1

89.076.4

80.382.3

86.183.0

Asturias

141.8127.8

114.2141.4

133.4111.1

129.2C

anarias92.4

86.273.9

94.787.5

75.985.3

Cantabria

119.0183.4

170.7129.5

141.2163.8

145.4C

astilla y León

133.0152.5

136.9130.8

136.1119.6

133.0C

astilla la Mancha

113.779.4

136.1111.6

114.4114.5

114.4V

alencia71.3

67.949.9

77.966.7

55.464.6

Extrem

adura85.9

100.9136.2

195.0116.3

103.1113.9

Galicia

133.0112.1

161.981.8

129.6168.8

137.0M

urcia99.2

116.688.5

122.5102.3

98.8101.7

average Obj. 1

100.0100.0

100.0100.0

100.0100.0

100.0avge. in ptas. per cap.

16,5374,320

8,2585,174

34,2897,928

42,217__________________________________________________________________________________________- N

ote: Average annu

al expenditu

re divid

ed by the popu

lation of each region in 1994 and norm

alizedby average exp

enditu

re per cap

ita in the entire Objective 1 territory. P

opu

lation data are taken from

the Tem

pu

s database of the N

ational Statistical Institute (IN

E). T

he last row show

s averageexpend

iture per capita in 1990 ptas.

Figure 1: P

ub

lic grants p

er capita fin

anced

by th

e CS

Fvs. relative in

come p

er capita

60 70 80 90

100

110

120

130

140

150

5060

7080

90

ExtA

nd

Val

Cana

Cant

CyL

Mur C

-M

Gal

Ast

grants per capita

relative income per capita

- Note: B

oth variables are normalized

; grants per capita by their average value in the entire Objective 1

territory, which is set equ

al to 100, and incom

e per capita by its average value in the Spanish regions

not included

in Objective 1.

11

Table 3 show

s average annual exp

enditu

re per cap

ita in each region, broken dow

nby p

rogramm

e and norm

alized by average p

er capita exp

enditu

re in the entireO

bjective 1 territory. In terms of public grants per capita (tot. pub.) the m

ost favouredregions w

ere Cantabria, C

astilla y León, A

sturias and

Galicia, and

the least favourd

ones Valencia, A

ndalucía and

Canarias. Figure 1 show

s that, contrary to what m

ay beexp

ected, there d

oes not seem to be a system

atic relationship betw

een grants per

capita and incom

e per capita in 1993 (which is norm

alized by average incom

e in theSp

anish regions not includ

ed in O

bjective 1). The lim

itations of the available data,

how

ever, suggest th

at some p

recaution

may be n

ecessary before extracting

conclusions in this regard.

Tab

le 4: Exp

end

iture b

y fun

ction as a fraction

of 1994 GV

A___________________________________________________________________________

infraest.pubinv.

subs.private

trainingtotal pub.

totalA

ndalucia1.51%

0.42%0.69%

0.74%0.45%

3.07%3.81%

Asturias

1.97%0.46%

0.79%0.74%

0.62%3.85%

4.59%C

anarias1.18%

0.29%0.47%

0.46%0.38%

2.32%2.78%

Cantabria

1.60%0.65%

1.15%1.06%

0.55%3.94%

5.00%C

astilla y León

1.79%0.54%

0.92%0.77%

0.55%3.80%

4.57%C

astilla la Mancha

1.73%0.32%

1.04%0.84%

0.53%3.62%

4.45%V

alencia0.88%

0.22%0.31%

0.33%0.30%

1.71%2.03%

Extrem

adura1.60%

0.49%1.27%

0.92%1.14%

4.50%5.42%

Galicia

1.96%0.43%

1.19%1.19%

0.38%3.96%

5.15%M

urcia1.54%

0.47%0.69%

0.74%0.60%

3.30%4.04%

total/GV

A O

bj. 11.49%

0.39%0.74%

0.71%0.47%

3.09%3.80%

total/GV

A Spain

0.74%0.19%

0.37%0.36%

0.23%1.54%

1.90%___________________________________________________________________________- N

ote: Average annu

al expenditu

re financed by the C

SF as a fraction of regional Gross V

alue A

dd

ed(G

VA

) in 1994. Both variables are m

easured

in millions of 1990 p

esetas using the Sp

anish GD

Pd

eflator. GV

A figures for 1994 are taken from

Fundación FIE

S.

Tables 4 and

5 relate CSF expend

iture to various regional macroeconom

ic aggregatesu

sing data for 1994. T

able 4 shows average annu

al expend

iture in each fu

nctionalcategory as a fraction of regional ou

tpu

t in 1994 (measu

red as G

ross Valu

e Ad

ded

,G

VA

). In the last row of the table, total expend

iture in O

bjective 1 regions is divid

edby aggregate Spanish G

VA

. In Table 5, C

SF infrastructu

re expenditu

re is divid

ed by

total infrastructu

re investment (Iinf), w

hile the rest of the CSF cap

ital expend

iture

programm

es (pubinv, subs and private) are show

n as a fraction of total investment in

non-infrastructu

re physical cap

ital (Iother). In the case of training expend

iture, the

table shows the result of d

ividing the total num

ber of man-years of training financed

12

annually by the C

SF by the observed increase in the nu

mber of years of form

alschooling of the w

orking age population between 1993 and

1994. 12

Tab

le 5: Sh

are of regional in

vestmen

t finan

ced b

y the C

SF

___________________________________________________________________________infraest.

invpub.subs.

privatek

training%

Iinf%

Iother%

Iother%

Iother%

Iother%

∆years

[1][2]

[3][4]

[5] = 2+

3+4

[6]A

ndalucia47.18%

3.36%5.52%

5.96%14.84%

5.88%A

sturias52.22%

3.91%6.67%

6.23%16.81%

19.02%C

anarias49.75%

1.98%3.25%

3.20%8.43%

5.75%C

antabria34.94%

5.14%9.14%

8.42%22.69%

10.01%C

astilla y León

62.10%3.62%

6.22%5.21%

15.05%78.75%

Castilla la M

ancha46.74%

2.40%7.87%

6.35%16.61%

16.79%V

alencia37.56%

1.63%2.28%

2.43%6.34%

3.76%E

xtremadura

32.87%3.78%

9.76%7.09%

20.64%62.98%

Galicia

55.39%3.02%

8.34%8.35%

19.71%11.04%

Murcia

51.69%3.47%

5.04%5.40%

13.91%7.23%

total/Inv. Obj. 1

47.82%2.90%

5.55%5.33%

13.78%8.21%

total/Inv. Spain29.03%

1.48%2.84%

2.72%7.04%

4.38%___________________________________________________________________________

Notes:

- Colum

ns [1]-[5] = average annu

al CSF-financed

expenditu

re as a fraction of the relevant investment

aggregate for 1994 (data from

Fundación B

BV

). All variables are m

easured in m

illions of 1990 pesetas.- C

olumn [6] =

annual average nu

mber of m

an-years of training financed by the C

SF/increase in the

total stock of years of education of the ad

ult population between 1993 and

1994.- T

he stock of years of schooling of the adult population is calculated

using the attainment d

ata in Mas

et al (1998). I assign 0 years of schooling to those classified as illiterates, 4 years to those w

ith some

prim

ary schooling, 10 to those with second

ary schooling and 15 (17) to those w

ith some (com

plete)

higher education.

Th

e figures sh

own

in th

ese tables show

that th

e CSF is qu

ite significan

t inm

acroeconomic term

s. 13 Total C

SF expend

iture rep

resents 3.8% of the aggregate

output of the Objective 1 regions (1.9%

of Spanish output). At the regional level, this

figure ran

ges between

2% in

the case of V

alencia an

d 5.4%

in E

xtremad

ura.

Structu

ral Fun

d exp

end

itures accou

nt for a con

siderable fraction

of regional

investment, representing alm

ost 50% of total expend

iture in infrastructure and 13.8%

of other investment in p

hysical capital in O

bjective 1 regions. The effect on hu

man

capital stocks is smaller. T

he Framew

ork's contribution represents approximately 8%

12 This figu

re can be rather mislead

ing in some regions becau

se it is very sensitive to the evolution of

the pop

ulation. T

he increase in the stock of years of schooling will be low

in those regions that losepopulation, and

this can make the C

SF's contribution appear to be quite large.13 R

ecall that our expenditure figu

res includ

e private and pu

blic national contributions in add

ition toE

U grants. T

his last item rep

resents app

roximately 70%

of pu

blic expend

iture and

a bit over 50% of

the total volume of resources channelled

through the CSF.

13

of the increase in the stock of total years of schooling of the working-age popu

lationbetw

een 1993 and 1994

4. Th

e imp

act on grow

th an

d em

ploym

ent: i) sh

ort-run

analysis

In this section and the next one I w

ill present an estim

ate of the contribution of the

Structural Funds to the grow

th of output and em

ployment in the O

bjective 1 Spanishregions. T

o facilitate the exposition, this section w

ill focus on the Fram

ework's

imp

act du

ring its first year of operation (1994), w

hile the next one will d

eal with its

cum

ulative m

ediu

m and

long-term effects taking into accou

nt dep

reciation and the

sluggish ad

justm

ent of employm

ent to a positive supply shock. In su

bsections a to c,I w

ill use the case of G

alicia as an examp

le to illustrate the estim

ation proced

ure

und

er different assu

mp

tions about the behaviou

r of private investm

ent and the

calculation of w

hat I will call the "social" rate of retu

rn on the different exp

enditu

reprogram

mes d

iscussed

above. Resu

lts for the remaining regions w

ill be presented in

subsection d. All calculations w

ill be mad

e under the assum

ption that the Framew

orkis executed

at a uniform pace, w

ith a similar volum

e of real expenditure in each year

between 1994 and

2000.

a. Scen

ario 1: imp

act of the C

SF w

ithou

t ind

uced

investm

ent effects

Using the figu

res reported in the previou

s section and the estim

ated prod

uction and

employm

ent functions, it is easy to obtain an estim

ate of the imm

ediate contribu

tionof the C

SF to the growth of output and

employm

ent in each region. As anticipated

inthe introd

uction, I w

ill carry out the requ

ired calcu

lations und

er two alternative

scenarios. The first and

simp

ler one is based on the assu

mp

tion that the figures that

appear in the Provisional Financial Plan for the CSF fully d

escribe its effects. That is, I

take the CSF at face valu

e and assu

me that there are no ad

ditional effects w

orkingthrough ind

uced changes in private investm

ent (aside from

those already includ

ed in

the Framew

ork as private cofinancing). Und

er this assumption, the calculation of the

short-run effects of the Fu

nds is very sim

ple: w

e need only p

lug the Fram

ework's

contribution to the stocks of the d

ifferent produ

ctive factors (calculated

in Section 3)into the m

odel estim

ated in Section 2 to obtain the ind

uced

increase in outp

ut and

emp

loyment relative to the observed

values of these variables in 1993. It shou

ld be

noted that the calcu

lation is somew

hat mislead

ing because it im

plicitly assum

es thatthere are no lags betw

een investment and

the resulting increase in ou

tpu

t. This is

particu

larly unrealistic in the case of large infrastru

cture p

rojects, where p

ayments

14

are typically sp

read over several years bu

t prod

uctivity effects w

ill only start tobecom

e app

arent after comp

letion. Hence, the resu

lts presented

in this sectionshou

ld be interp

reted as a first estim

ate of the average annual d

irect imp

act of theC

SF over the programm

ing period that d

oes not take into account depreciation or the

dynam

ics of employm

ent.

Table 6 sum

marizes the relevant com

putations as well as the und

erlying data and

theestim

ated valu

es of the relevant elasticities. The first colu

mn (∆

log stock) shows the

increase in the logarithm of the stocks of the d

ifferent produ

ctive factors that can beattribu

ted to the C

SF. Und

er this first scenario, the increase in the stock of physical

capital is calcu

lated as the su

m of p

ublic investm

ent in non-infrastructu

re capital,

subsid

ies to firms and

declared

private co-financing. In the case of hu

man cap

ital,there are tw

o different figures. T

he first one (0.21%) represents the C

SF's contributionto the average level of ed

ucation of the G

alician working-age p

opu

lation (WA

P)

while the second

one (0.16%) refers to the ind

uced

change in the average attainment

of emp

loyed w

orkers, which is the variable that enters the p

rodu

ction function. T

oobtain this second

figure, I have estim

ated the relationship

between the attainm

entlevels

of th

e w

orking-age

and

em

ployed

p

opu

lations

(controllin

g for

the

emp

loyment ratio, d

efined as the fraction of the w

orking age pop

ulation that is

emp

loyed), obtaining an elasticity of 0.743 that I ap

ply to the second

variable toestim

ate the first. This yield

s an estimate of 0.11%

for the CSF's contribu

tion to theaverage level of schooling of em

ployed w

orkers.

Colu

mn (2) show

s the estimated

values of the elasticity of ou

tput w

ith respect to thestocks of the d

ifferent prod

uctive factors (θ

i ). Mu

ltiplying these coefficients by the

increase in the correspond

ing stocks, we obtain the d

irect contribution of C

SFinvestm

ent in capital, infrastru

ctures and

training to aggregate value ad

ded

(∆Y

1)

which is show

n in colum

n (3). Colu

mn (4) show

s the short-term em

ploym

entelasticities (λ

i ) of the different factors, w

hich are multiplied

by ∆log stock to obtain the

indu

ced (log) increase in em

ploym

ent (colum

n 5). Finally, we have to take into

account the fact that the increase in em

ploym

ent will in tu

rn raise outp

ut by an

amou

nt equal to the p

rodu

ct of the log increase in emp

loyment and

the elasticity ofou

tpu

t with resp

ect to this factor (which is 0.597). T

he result of this com

pu

tation,d

enoted by ∆

Y2, is show

n in colum

n (6). Ad

ding this figure to ∆

Y1, w

e finally arriveto the C

SF's total contribution to ou

tpu

t growth (∆

Y3), w

hich is reported

in colum

n

(7).

15

Tab

le 6: Imp

act of the C

SF on

outp

ut an

d em

ploym

ent grow

thG

alicia, 1994S

cenario 1: n

o ind

uced

effects__________________________________________________________________________________________

(1)(2)

(3)(4)

(5)(6)

(7)∆

log stockoutputelast. θ

i

direct∆

Y1

employm

.elast. λ

i

∆ em

ploy.induced

∆Y

2total∆

Y3

physical capital1.97%

0.2970.59%

0.1330.26%

0.16%0.74%

infraestructures6.23%

0.1060.66%

0.0480.30%

0.18%0.84%

h. cap. wkng. age.

0.21%h. cap. em

ployed0.16%

0.2860.05%

0.1280.02%

0.01%0.06%

employm

ent (jobs)0.597

total1.29%

0.58%0.35%

1.64%__________________________________________________________________________________________

Notes:

- Totals d

o not add

up exactly due to round

ing error.- T

he variable ∆ log stock is calculated

as follows. L

et K93 be the observed

stock of (non-infrastructure)physical capital at the end

of 1993 and K

MA

C the estim

ated contribu

tion of the CSF to investm

ent inp

hysical capital d

uring 1994. T

hen, ∆log K

= ln(K

93 + K

MA

C) - ln (K

93) where ln d

enotes natural

logarithms. T

he procedure is id

entical for the rest of the productive factors.

Ad

ding up the effects of the d

ifferent expenditure item

s, the total increase in Galician

outp

ut d

ue to the C

SF du

ring 1994 was of 1.64 p

ercentage points. A

bit over half ofth

is total

(0.84%)

comes

from

infrastru

cture

investm

ent,

followed

by

the

accum

ulation of other p

hysical capital (0.74%

) and by the increase in ed

ucational

attainment (0.06%

). The d

irect effects of these three types of investm

ent amou

nt toapproxim

ately 1.3 percentage points of outpu

t growth and

the remaining 0.35 com

efrom

induced

job creation, which represents a 0.6%

increase in employm

ent.

b. S

cenario 2: in

du

ced effects th

rough

private in

vestmen

t

The analysis in the p

revious section assu

mes that the actions inclu

ded

in the CSF

affect private investment only through subsid

ies to enterprises, and that the increase

in private investm

ent indu

ced by the Fram

ework is given by the su

m of these

subsid

ies and the d

eclared p

rivate contributions to the financing of the assisted

projects. In practice, both assumptions are probably inad

equate and the net im

pact ofthe C

SF on private investment cou

ld be either larger or sm

aller than I have assum

edin the p

revious section d

epend

ing on the relative imp

ortance of three effects which

pull in different d

irections.

16

First, it seems reasonable to expect that at least part of the investm

ent projects which

benefit from E

U grants w

ould

have been und

ertaken even withou

t such su

pport. Inthis case, p

art of the subsid

ies will only rep

lace private financing and

the net effecton investm

ent will be low

er than the Framew

ork's projections. Second

, we have to

take into account a crowd

ing-out effect that would

work in the sam

e direction. T

o theextent that public expend

iture must be financed

through taxes or debt (w

hich detract

resources from

the private sector and m

ay generate various d

istortions), it will tend

to redu

ce private saving and

investment. In the cu

rrent context, this effect will be

mitigated

by the fact that an imp

ortant part of stru

ctural aid

is financed by E

Utransfers w

hich (if we take as given Sp

ain's contribution to the U

nion's bud

get) do

not imp

ly an increase in taxes or debt. Finally, there exists a p

ositive "crowd

ing-in"effect w

hich has not been taken into account in m

y previou

s calculations. Since

Structu

ral Fund

grants finance the accum

ulation of p

rodu

ctive inpu

ts that can beexpected

to be complem

ents of private capital, one of their effects will be to raise the

rate of return on this factor, thereby increasing the incentive for p

rivate investment.

A sim

ilar effect may w

ork through d

emand

channels if pu

blic spend

ing "pu

lls-in"p

rivate investment throu

gh an increase in pu

rchases of goods and

services fromprivate suppliers.

The net effect of these three factors is u

ncertain and m

ust be estim

ated em

pirically.

The natu

ral way to constru

ct such an estim

ate wou

ld be throu

gh the estimation of a

private investm

ent function u

sing regional data. U

nfortunately, the exercise is not

feasible du

e to the lack of regionalized d

ata on investment su

bsidies and

othervariables of interest. A

s an imp

erfect substitu

te, I will u

se an investment fu

nctionestim

ated w

ith national data to ap

proxim

ate the reaction of private investm

ent tod

ifferent types of CSF expend

iture.

In de la Fu

ente (1997) I have used

OE

CD

data to estim

ate a private investm

entfu

nction which inclu

des variou

s fiscal indicators as exp

lanatory variables. This

function is of the form

(4) skit = Γ

ο + Γ

g GT

OT

it + Γ

p sGit +

Γs sub

it + Γ

tr transfit + Γ

x xit

where skit is private investm

ent in country i at time t, G

TO

T total public expend

iture,sG p

ublic investm

ent (both in infrastructu

re and in other typ

es of physical cap

ital),subs subsid

ies to enterprises, transf transfers to households (all m

easured as a fraction

of GD

P) and

x a vector of non-fiscal variables which inclu

des the relative p

rice ofcapital good

s, income per capita and

dem

ographic variables among other things. T

he

17

first two regressors attem

pt to capture, respectively, the crowd

ing-out and crow

ding-

in effects of public expenditure and

public investment. T

he equation allows transfers

to household

s to have a different im

pact than, say, public consu

mption, becau

se theform

er comp

onent of pu

blic expend

iture d

oes not redu

ce the disp

osable income of

the p

rivate sector and

this m

ay mitigate its ad

verse imp

act on savin

gs and

investment.

Tab

le 7: Estim

ated p

arameters of th

e private in

vestmen

t fun

ctionS

ensitivity to variou

s fiscal variables

______________________________________________coeff.

(t)total public expenditure

-0.319(8.28)

public investment

0.533(3.75)

transfers to households0.144

(2.61)

subsidies to firms

0.854(3.86)

______________________________________________ N

otes:- t statistics in parentheses next to each coefficient.- T

he equation includes as regressors other variables not includ

ed in the table. See d

e la Fuente (1997)for d

etails.

The resu

lts of the estimation (see T

able 7) suggest that the crow

ding-ou

t effect issizable: each eu

ro of pu

blic expend

iture (financed

either through taxes or throu

ghd

ebt) redu

ces private investm

ent by 32 cents. There is also evid

ence of a positive

crowd

ing-in effect of pu

blic investment on p

rivate capital accu

mu

lation. Since thiseffect is stronger than the p

revious one, the net im

pact of p

ublic investm

ent isp

ositive and rather consid

erable: each euro of p

ublic investm

ent seems to increase

private investm

ent by around

twenty cents. Finally, m

y estimates su

ggest that,althou

gh subsid

ies to firms d

o tend to increase total private investm

ent, the indu

cedincrease is sm

aller than the subsidy. E

ven without taking into account the crow

ding-

out effect, each

dollar of su

bsidies in

creases total private in

vestmen

t (wh

ichp

resum

ably does inclu

de su

bsidies) by only 85 cents -- im

plying that p

rivatefinancing falls by 15 cents per euro of subsid

ies.

Using the param

eter estimates reported

in Table 7, I w

ill calculate a "net m

ultiplier"

coefficient for each type of p

ublic exp

enditu

re contemp

lated in the C

SF (subsid

iesand

investment in infrastru

ctures, other cap

ital and training). T

his coefficient will

then

be used

to estimate th

e amou

nt of p

rivate investm

ent in

du

ced by each

programm

e. Since the size of the crowd

ing-out effect will d

epend on the share of E

Ufinancing, I first calcu

late a national co-financing coefficient for each expend

iture

program

me by d

ividing the contribu

tion of the various Sp

anish adm

inistrations by

18

the total pu

blic expend

iture of the sam

e type (inclu

ding E

U grants) record

ed in the

CSF.

Tab

le 8: Crow

din

g-out an

d m

ultip

lier coefficients

for various p

ub

lic expen

ditu

re items

_____________________________________________________________________________(1)

(2)(3)

(4)national

co-financingcrow

ding-outcoefficient

crowding-in

coefficientnet

multiplier

infraestructures0.342

-0.1090.533

0.424direct investm

ent0.467

-0.1490.533

0.384subsidies

0.288-0.092

-0.146-0.238

training0.251

-0.0800.144

0.064_____________________________________________________________________________

- Note: the crow

ding-ou

t coefficient for subsid

ies is calculated

as the coefficient of subsid

ies toenterprises in the investm

ent equation (0.854) minus one.

Mu

ltiplying this coefficient, w

hich is shown in colu

mn (1) of T

able 8, by thecrow

ding-ou

t coefficient for total government exp

enditu

re (-0.319), I obtain ad

ifferent crowd

ing-out coefficient for each typ

e of pu

blic expend

iture (colu

mn 2).

Colu

mn (3) d

isplays the crow

ding-in coefficient im

plied

by the estimates in T

able 7u

nder the assu

mp

tion that this coefficient is the same for p

ublic investm

ent ininfrastru

ctures and

in other capital (since these tw

o items are not sep

arated in the

estimated

private investm

ent equation) and

treating training expend

iture as an in-

kind transfer to hou

seholds. Finally, the su

m of colu

mns (2) and

(3) gives the netm

ultip

lier coefficient for each expend

iture p

rogramm

e, which is show

n in colum

n(4).

The prod

uct of the net m

ultiplier and

the corresponding exp

enditu

re item yield

s anestim

ate of the increase in private investm

ent indu

ced by each p

ublic sp

ending

programm

e. The resu

lt of this compu

tation for Galicia is show

n in Table 9, together

with the correspond

ing figures for Scenario 1 and the im

plied net m

ultipliers. As can

be seen in the table, my second

scenario is considerably less op

timistic abou

t theam

ount of ind

uced

private investm

ent. This is p

articularly so in connection w

ithsu

bsidies to p

rivate firms, w

hose contribution to p

rivate capital accu

mu

lation goesfrom

+36.523 to -8.686 m

illion ptas. A

large fraction of this decrease, how

ever, iscom

pensated

by the positive crow

ding-in effects associated

with the rest of the

public expenditure item

s.

19

Tab

le 9: Exp

end

iture attrib

uted

to the C

SF, G

aliciaC

omp

arison of th

e two scen

arios_________________________________________________________________________________

scenario 1 scenario 2 _____________________ ____________________

publicexpenditure

netm

ultiplierinduced

private invest.net

multiplier

inducedprivate invest.

infraestructures60,055

00

0.42425,463

direct investment

13,2200

00.384

5,076subsidies

36,4941.001

36,523-0.238

-8,686training

11,5490

00.064

739total expenditure

121,31936,523

22,593_________________________________________________________________________________- N

ote: induced

investment is m

easured in m

illions of 1990 pesetas.

Proceed

ing as in the previous section, I calcu

lated the contribu

tion of the CSF to the

growth of G

alician outp

ut and

emp

loyment u

nder the assu

mp

tions of Scenario 2.T

he results, d

isaggregated into the contribu

tions of the three types of capital we are

considering, are rep

orted in T

able 10, together with those of Scenario 1. Since the

indu

ced increase in p

rivate investment is sm

aller in the second scenario, the

estimated

effects on growth and

emp

loyment are now

somew

hat smaller (arou

nd7%

).

Tab

le 10: CS

F's imp

act in G

alicia, disaggregated

by p

rodu

ctive factorC

omp

arison of scen

arios 1 and

2__________________________________________________________________________________________

scenario 1 scenario 2 __________________________________ ________________________________

(1)(2)

(3)(4)

(5)(6)

(7)(8)

infraest.capital

trainingtotal

infraest.capital

trainingtotal

∆ log stock

6.23%1.97%

0.16%6.23%

1.66%0.16%

∆Y

10.66%

0.59%0.05%

1.29%0.66%

0.49%0.05%

1.20%

∆ em

ployment

0.30%0.26%

0.02%0.58%

0.30%0.22%

0.02%0.54%

∆Y

20.18%

0.16%0.01%

0.35%0.18%

0.13%0.01%

0.32%

∆Y

3, total0.84%

0.74%0.06%

1.64%0.84%

0.62%0.06%

1.52%__________________________________________________________________________________________

c. Th

e social return

on p

ub

lic expen

ditu

re

A reasonable criterion for the allocation of p

ublic resou

rces among alternative

develop

ment p

rogramm

es within a given region is the m

aximization of their

aggregate imp

act. If this allocation

is optim

al, the m

arginal retu

rn to p

ublic

20

expenditure, m

easured by its contribution to regional incom

e, should be the sam

e forall program

mes. If this cond

ition does not hold

, it will be possible to increase ou

tput

with a given volum

e of expenditure by shifting resources tow

ards those program

mes

with the highest returns.

In this section I will constru

ct an indicator of w

hat I will call the social rate of return

for each of the four expenditure program

mes contem

plated in the C

SF (investment in

infrastructu

res and in other p

hysical capital, su

bsidies to p

rivate firms and

trainingp

rogramm

es) und

er each of the two scenarios d

iscussed

above. This ind

icator isd

efined as the d

iscount rate that makes the present value of the flow

of increments of

regional income generated

by each type of investm

ent (which falls over tim

e as aresu

lt of dep

reciation) equal to the relevant public exp

enditu

re und

ertaken in theinitial year. 14 N

otice that, since I do not take into accou

nt the relevant private costs,

this indicator d

oes not measu

re the rate of return on the projects in the proper sense

of the term, bu

t it does p

rovide a u

seful su

mm

ary measu

re of the imp

act of eachpu

blic expenditu

re programm

e on the growth of overall regional ou

tput, taking into

account both its d

irect effects and those that op

erate through ind

uced

private

investment and

employm

ent. This inform

ation is likely to be of considerable interest

for policym

akers, both for the evaluation of the cu

rrent Framew

ork and for the

design of future program

mes.

In order to com

pute the rate of return of the different program

mes, I have to estim

atetheir resp

ective contributions to the grow

th of regional outp

ut. Since the resu

lts ofthe p

revious tw

o sections are disaggregated

by prod

uctive factor (infrastru

ctures,

other physical cap

ital and hu

man cap

ital) rather than by program

me, this requ

iressom

e calculations. In p

articular, the contribu

tion of physical cap

ital (k) to outp

ut

growth m

ust be broken d

own into three com

ponents that reflect the im

pact of,

respectively, d

irect pu

blic investm

ent in

non

-infrastru

cture p

hysical cap

ital,su

bsidies to enterp

rises and ind

uced

private investm

ent. Then, the last one of these

items m

ust be allocated

to the different p

rogramm

es in proportion to the volum

e ofinvestm

ent indu

ced by each of them

(a calculation that m

ust be d

one differently in

each of the scenarios). 15 Finally, the resulting (ind

irect) gains in outp

ut m

ust be

add

ed to the d

irect effects of each program

me to obtain its total contribu

tion toregional grow

th.

14 See Section 4 of Append

ix 1 for the details of the calculation of the social rate of return.

15 For this calculation, ind

uced

private investm

ent is attributed

only to subsid

ies to enterprises in

Scenario 1, and is allocated

among all the public expend

iture programm

es in Scenario 2. The necessary

data are in T

able 9.

21

Tab

le 11: Tab

le 9: Imp

act of the C

SF b

y pu

blic exp

end

iture p

rogramm

ean

d social rate of retu

rn on

pu

blic fu

nd

s. Galicia.

Com

parison

of Scen

arios 1 and

2__________________________________________________________________________________________

(1)(2)

(3)(4)

(5)(6)

Scenario 1public

expend.%

∆Y

direct%

∆Y

induced%

∆Y

total∆

Ytot. m

ptas.

return onpublic exp.

infraestructures60,055

0.84%0.00%

0.84%25,577

38.5%direct investm

ent13,220

0.11%0.00%

0.11%3,479

18.5%subsidies

36,4940.31%

0.31%0.63%

19,21444.8%

training11,549

0.06%0.00%

0.06%1,745

15.0%total public expendit.

121,3191.32%

0.31%1.64%

49,68435.4%

Scenario 2infraestructures

60,0550.84%

0.22%1.06%

32,26848.9%

direct investment

13,2200.11%

0.04%0.16%

4,81928.7%

subsidies36,494

0.31%-0.07%

0.24%7,325

12.3%training

11,5490.06%

0.01%0.06%

1,93716.2%

total public expendit.121,319

1.32%0.19%

1.52%46,046

32.6%__________________________________________________________________________________________

Notes:

- Colum

ns (1) and (5) in m

illions of 1990 pesetas.- I have assu

med

a depreciation rate of 4.1%

for infrastructu

res and of 7.8%

for other physical capital,and

a usefu

l life of 34.13 years for hum

an capital. T

he first two figu

res are recovered from

theinvestm

ent and cap

ital stock series used

in the estimation of the em

pirical m

odel and

correspond

tothe last year of the sam

ple. See footnote 17 for the assu

mp

tions used

to estimate the u

seful life of

human capital.

Table 11 sh

ows th

e estimated

rates of return

on th

e differen

t expen

ditu

rep

rogramm

es in Galicia together w

ith the information requ

ired for their calcu

lation.C

olumn (1) show

s average annual public expenditure in each program

me in m

illionsof 1990 p

esetas. Colu

mn (2) show

s the direct contribu

tion of each item of p

ublic

expenditure to the grow

th of regional output, and colum

n (3) its indirect contribution

through ind

uced

private investm

ent (taking into account in both cases the gain in

outpu

t brought abou

t by the indu

ced increase in em

ployment). N

otice that colum

ns(1) and

(2) are identical for both scenarios. C

olum

n (3), by contrast, varies acrossscenarios reflecting d

ifferences in the assum

ptions abou

t the response of p

rivateinvestm

ent. Ad

ding colu

mns (2) and

(3), we obtain the total contribu

tion of eachp

rogramm

e to the growth of regional ou

tpu

t in percentage (logarithm

ic) terms

(colum

n (4)), and recover the ind

uced

increase in outp

ut m

easured

in millions of

1990 ptas. (column (5)). 16

16 The proced

ure used to recover the contribution of each program

me to regional incom

e measured

inm

illions of pesetas is as follows. L

et Y93 be the output of a given region in 1993, m

easured in m

illionsof 1990 pesetas and

∆yj the logarithm

ic increase in output induced

by programm

e j in the same region.

22

The social rates of return on the d

ifferent public expenditure program

mes are show

nin colu

mn (6) of T

able 11. Their calcu

lation requires som

e assum

ptions abou

t therelevant rates of d

epreciation. For investm

ent in infrastructu

res and other p

hysicalcapital, I have used

the depreciation rates im

plicit in the capital stock and investm

entseries u

sed to estim

ate the emp

irical mod

el of Section 2 (4.1 and 7.8%

respectively).

In the case of human capital, I have assum

ed that the increase in the stock of years of

training financed by the C

SF disap

pears all at once w

ith the retirement of the

beneficiaries of the relevant program

mes after a "u

seful life" that I estim

ate in 34.13years. 17 H

ence, it is assumed

that the flow of output gains generated

by CSF training

expenditure rem

ains constant over this period (w

hich amounts to ignoring d

eath andm

igration) and d

rops to zero thereafter.

Inspection of colu

mn (6) of T

able 11 shows that the estim

ated rates of retu

rn arequ

ite respectable. In both scenarios, the aggregate social rate of retu

rn on CSF

expend

iture in G

alicia exceeds 30%

. Looking at the d

ifferent program

mes, the rates

of return range from

12 to 49% d

epend

ing on the type of exp

enditu

re and on the

scenario under consid

eration.

As m

ay be expected

, the main d

ifference between the tw

o scenarios has to do w

iththe social retu

rn on subsid

ies to enterprises. If w

e accept the (extremely favou

rable)assu

mp

tions imp

licit in Scenario 1 about the crow

ding-in effects of su

bsidies, this

item is by far the one w

ith the highest social rate of return. Und

er the probably more

realistic assum

ptions of Scenario 2, the social retu

rn on subsid

ies drop

s by 75% and

this instrument falls to the last position in term

s of its capacity to create employm

entand

increase output per euro of public expenditure.

d. R

esults for th

e remain

ing O

bjective 1 region

s

Following the sam

e procedure as in the previous sections, I have calculated

the short-ru

n contribution of the d

ifferent pu

blic expend

iture p

rogramm

es to the growth of

Sum

ming over the d

ifferent program

mes, j, w

e obtain the total increase in the logarithm of regional

output, ∆y. T

he "final" value of log output is then yf = ln Y

93 + ∆

y, from w

here we recover the level of

outp

ut Y

f = E

xp (yf) and the increase in the level of ou

tput m

easured

in millions of pesetas generated

by the entire Framew

ork, ∆Y

= Y

f - Y93. Finally, this increase is allocated

among the d

ifferentprogram

mes in proportion to their contributions to the grow

th of log output, (∆yj /∆

y).17 T

o arrive at this figure, I assu

me that the u

seful life of d

ifferent training programm

es is as follows:

40 years for formal vocational training (w

ithin the secondary schooling system

), 35 years for thetraining of researchers and

25 in the case of training programm

es for adult (em

ployed or unem

ployed)

workers. T

hese figures are w

eighted by the share of each typ

e of program

me in the total increase in

the stock of years of training induced

by the CSF for the entire set of O

bjective 1 regions.

23

regional ou

tpu

t and

emp

loymen

t in each

of the O

bjective 1 regions an

d th

ecorrespond

ing social rates of return.

The resu

lts for the two scenarios are show

n in Tables 12 to 16. T

he penultim

ate rowof each table su

mm

arizes the imp

act of the CSF on the entire set of O

bjective 1regions. T

otal job creation and the total increase in regional ou

tpu

t measu

red in

millions of 1990 pesetas (w

hich is used

to calculate the social rate of retu

rn shown in

the last column) are obtained

by add

ing up the analogous figures for all the Objective

1 regions. The resu

lt of this calculation is then d

ivided

by total emp

loyment or by

aggregate outp

ut in this sam

ple in 1993 to obtain the p

ercentage increases of GV

A(%

∆Y

total) and em

ploym

ent (%∆

employ.). 18 T

he last row show

s the contribution of

the Objective 1 C

SF to the growth of ou

tput and

employm

ent in the whole of Spain.

These resu

lts are obtained in the sam

e way as the p

revious ones, bu

t taking as areference aggregate output and

employm

ent in the entire country (with the exception

of Ceuta and

Melilla) rather than in the set of regions eligible for O

bjective 1 support.

Tab

le 12: Imp

act of pu

blic in

vestmen

t in p

rodu

ctive infrastru

ctures

___________________________________________________________________________ scenario 1 scenario 2 ___________________________________ ____________________________________

%∆

Y total

%∆

employ.

no. ofjobs

rate ofreturn

%∆

Y total

%∆

employ.

no. ofjobs

rate ofreturn

Andalucía

0.50%0.18%

3,25928.4%

0.68%0.24%

4,42639.1%

Asturias

0.65%0.23%

80528.9%

0.84%0.30%

1,03837.6%

Canarias

0.56%0.20%

95441.8%

0.73%0.26%

1,23754.6%

Cantabria

0.54%0.19%

32029.0%

0.69%0.25%

41237.8%

Castilla y L

eón0.52%

0.18%1,516

24.7%0.70%

0.25%2,045

33.8%C

ast. la Man.

0.42%0.15%

76820.2%

0.59%0.21%

1,06628.7%

Valencia

0.43%0.15%

1,99944.2%

0.54%0.19%

2,52056.0%

Extrem

adura0.37%

0.13%393

18.9%0.51%

0.18%539

26.6%G

alicia0.84%

0.30%2,806

38.5%1.06%

0.37%3,542

48.9%M

urcia0.60%

0.21%685

34.7%0.79%

0.28%898

45.9%total O

bj. 10.54%

0.19%13,506

31.4%0.71%

0.25%17,724

41.7%total/Spain

0.27%0.11%

0.35%0.14%

___________________________________________________________________________N

otes:- T

he increase in the num

ber of jobs is calculated

in the same w

ay as the increase in the level ofregional incom

e (see footnote 16).-%

∆Y

total (%∆

employ.) =

percentage or logarithmic increase of output (em

ployment) in each region or

in the set of all Objective 1 regions, excep

t for the last row, w

here it refers to the contribution of the

CSF to the grow

th of the relevant variable in the whole of Spain (exclu

ding C

euta and

Melilla). In all

cases, the figures refer to the increase over the observed value of the relevant variable in 1993.

18 Notice that the nu

mber obtained

in this manner w

ill be a percentage in the strict sense of the term,

and not a logarithm

ic change as in the preceding row

s of the table.

24

Tab

le 13: Imp

act of pu

blic in

vestmen

t in oth

er ph

ysical capital


%∆

Y total

%∆

employ.

no. ofjobs

rate ofreturn

%∆

Y total

%∆

employ.

no. ofjobs

rate ofreturn

Andalucía

0.12%0.04%

76119.6%

0.16%0.06%

1,05430.1%

Asturias

0.10%0.04%

12914.7%

0.14%0.05%

17923.3%

Canarias

0.10%0.03%

16324.3%

0.13%0.05%

22536.6%

Cantabria

0.15%0.05%

8814.9%

0.20%0.07%

12223.6%

Castilla y L

eón0.13%

0.05%374

15.9%0.18%

0.06%518

25.0%C

ast. la Man.

0.07%0.02%

12814.5%

0.10%0.03%

17823.0%

Valencia

0.07%0.02%

30621.9%

0.09%0.03%

42333.2%

Extrem

adura0.10%

0.04%105

12.3%0.14%

0.05%146

20.1%G

alicia0.11%

0.04%382

18.5%0.16%

0.06%529

28.7%M

urcia0.14%

0.05%154

20.7%0.19%

0.07%214

31.6%total O

bj. 10.10%

0.04%2,591

18.3%0.14%

0.05%3,586

28.3%total/Spain

0.05%0.02%

0.07%0.03%

___________________________________________________________________________

Tab

le 14: Imp

act of sub

sidies to th

e private sector


%∆

Y total

%∆

employ.

no. ofjobs

rate ofreturn

%∆

Y total

%∆

employ.

no. ofjobs

rate ofreturn

Andalucía

0.40%0.14%

2,59949.3%

0.15%0.05%

95213.1%

Asturias

0.35%0.12%

42635.7%

0.14%0.05%

1689.3%

Canarias

0.31%0.11%

53056.0%

0.12%0.04%

20316.7%

Cantabria

0.51%0.18%

30135.8%

0.20%0.07%

1199.5%

Castilla y L

eón0.40%

0.14%1,180

35.8%0.17%

0.06%489

10.3%C

ast. la Man.

0.42%0.15%

76032.4%

0.18%0.06%

3209.2%

Valencia

0.19%0.07%

88753.5%

0.07%0.03%

32714.8%

Extrem

adura0.45%

0.16%470

27.0%0.20%

0.07%207

7.5%G

alicia0.63%

0.22%2,108

44.8%0.24%

0.08%804

12.3%M

urcia0.41%

0.14%464

51.2%0.15%

0.05%171

13.9%total O

bj. 10.39%

0.14%9,725

42.9%0.15%

0.05%3,761

11.8%total/Spain

0.19%0.08%

0.07%0.03%

___________________________________________________________________________

25

Tab

le 15: Imp

act of trainin

g expen

ditu

re___________________________________________________________________________ scenario 1 scenario 2 ___________________________________ ____________________________________

%∆

Y total

%∆

employ.

no. ofjobs

rate ofreturn

%∆

Y total

%∆

employ.

no. ofjobs

rate ofreturn

Andalucía

0.06%0.02%

36311.9%

0.06%0.02%

41613.0%

Asturias

0.08%0.03%

9512.2%

0.09%0.03%

10513.1%

Canarias

0.07%0.03%

12118.0%

0.08%0.03%

13519.5%

Cantabria

0.07%0.03%

4413.1%

0.08%0.03%

4814.1%

Castilla y L

eón0.07%

0.03%214

13.0%0.08%

0.03%238

14.0%C

ast. la Man.

0.07%0.03%

13513.7%

0.08%0.03%

14914.7%

Valencia

0.05%0.02%

25117.6%

0.06%0.02%

27819.0%

Extrem

adura0.12%

0.04%131

10.5%0.14%

0.05%147

11.3%G

alicia0.06%

0.02%192

15.0%0.064%

0.02%214

16.2%M

urcia0.08%

0.03%95

13.7%0.09%

0.03%107

15.0%total O

bj. 10.07%

0.02%1,641

13.7%0.07%

0.03%1,837

14.8%total/Spain

0.03%0.01%

0.04%0.01%

___________________________________________________________________________

Tab

le 16: Overall im

pact of C

SF exp

end

iture


%∆

Y total

%∆

employ.

no. ofjobs

rate ofreturn

%∆

Y total

%∆

employ.

no. ofjobs

rate ofreturn

Andalucía

1.07%0.38%

6,96828.6%

1.05%0.37%

6,83428.0%

Asturias

1.18%0.42%

1,45225.3%

1.21%0.43%

1,48726.0%

Canarias

1.04%0.37%

1,76437.9%

1.06%0.38%

1,79638.6%

Cantabria

1.26%0.45%

75125.9%

1.18%0.42%

70124.0%

Castilla y L

eón1.12%

0.40%3,277

23.8%1.12%

0.40%3,284

23.9%C

ast. la Man.

0.98%0.35%

1,78721.6%

0.94%0.33%

1,70920.5%

Valencia

0.74%0.26%

3,43637.6%

0.77%0.27%

3,54038.8%

Extrem

adura1.04%

0.37%1,098

17.6%0.99%

0.35%1,037

16.5%G

alicia1.64%

0.58%5,475

35.4%1.52%

0.54%5,076

32.6%M

urcia1.23%

0.44%1,396

31.5%1.22%

0.43%1,387

31.3%total O

bj. 11.09%

0.39%27,404

29.2%1.07%

0.38%26,853

28.6%total/Spain

0.54%0.21%

0.53%0.21%

___________________________________________________________________________

26

The rem

ainder of this section analyzes the im

plications of the rate of return estimates

for the different p

rogramm

es and regions, leaving for a later section a d

iscussion of

the macroeconom

ic imp

act of the Framew

ork. Figure 2 show

s the average rates ofretu

rn in Objective 1 territory of the fou

r pu

blic expend

iture p

rogramm

es I haveconsid

ered u

nder each of the tw

o scenarios. As anticip

ated in the p

revious section,

the social rate of return to su

bsidies to p

rivate enterprises is m

uch low

er und

erScenario 2 than u

nder Scenario 1, w

here it is assum

ed that all p

rivate cofinancingconstitu

tes new investm

ent. Und

er the more realistic assu

mptions of Scenario 2, the

expen

ditu

re program

mes w

ith th

e high

est rates of return

are investm

ent in

infrastructu

res and in other typ

es of physical cap

ital, followed

at a considerable

distance by training expend

iture and by subsid

ies.

Figure 2: A

verage rate of return

on d

ifferent p

ub

lic expen

ditu

re program

mes in

Ob

jective 1 regions

0% 5%

10%

15%

20%

25%

30%

35%

40%

45%

infraestructures

scenario 1scenario 2

other investment

trainingsubsidies

My estim

ates of the social rate of return on training exp

enditu

re are especially

uncertain d

ue to the p

articularly p

oor quality of the d

ata on the outp

ut of training

program

mes and

to the large num

ber of auxiliary assu

mp

tions required

to estimate

the growth effects of this expend

iture item

. In any event, it should

be noted that the

relatively low rates of retu

rn estimated

for this program

me are d

riven by the highcost of E

U-sp

onsored training schem

es and have nothing to d

o with the qu

ality of

27

these courses. 19 W

hile the cost of a man-year of form

al secondary schooling w

as of230,000 1990 p

tas. (in And

alucía in 1994), I estim

ate that the average cost of a man-

year of training financed by the C

SF was 404,000 p

tas. of the same year. T

his figure

rises to 678,000 ptas. if w

e restrict ourselves to training p

rogramm

es aimed

at(em

ployed and

unemployed

) adult w

orkers. If the unit cost of CSF-financed

traininghad

been the same as that of form

al secondary schooling, the social rate of retu

rn ontraining expend

iture w

ould

have been 26%, w

hich is roughly the sam

e as the return

estimated

on non-infrastructure public investment.

There are also good

reasons to susp

ect that my estim

ates und

erestimate the retu

rnsto training expend

iture. In particular, the mod

el used in this paper only picks up the

direct effects of hu

man cap

ital on the level on prod

uctivity and

does not allow

forind

irect effects that wou

ld op

erate through the contribu

tion of this factor to fastertechnical p

rogress. The evid

ence available in the literature su

ggests that this secondeffect is im

portant and

can raise the return to these p

rogramm

es by somew

herebetw

een 30 and 50%

. 20

My resu

lts should

also be considered

tentative, and not only in relation to training

program

mes, becau

se they are partly based

on a private investm

ent function w

hichis estim

ated w

ith a different d

ata set, and becau

se there are few com

parable stud

iesin the literatu

re that may be u

sed to check m

y findings. W

ith the caution this

requires, the exercise d

oes suggest that a reallocation of Stru

ctural Fu

nd resou

rcescou

ld resu

lt in a significant increase in their imp

act on outp

ut and

emp

loyment.

Accord

ing to my estim

ates, in the case of Spain it wou

ld be d

esirable to invest more

in infrastructu

res and other capital and

to redu

ce the amou

nt of subsid

ies. As noted

,there is greater uncertainty concerning the returns to training expend

iture but it does

seem likely that there is room

for cost reductions in this area.

Figure 3 show

s the average rate of return on C

SF pu

blic expend

iture in each of the

Objective 1 regions. T

his variable ranges between 16.5%

in Extrem

adu

ra and a bit

over 38% in V

alencia and C

anarias. Cross-regional d

ifferences in rates of return are

19 All the calculations have been m

ade und

er the assumption that the effects of a year of schooling are

the same for all types of training. H

ence, I am not controlling for quality and

this may bias the results

against CSF-financed

training if these programm

es have a higher impact on prod

uctivity than form

alschooling. T

his is not necessarily imp

lausible, as the E

SF generally finances app

lied vocational

training programm

es that are supposed to supply qualifications that are in d

emand

in the job market.

Bu

t this differential p

rodu

ctivity effect wou

ld have to be very large for the rate of retu

rn on CSF-

financed training expend

iture to be comparable to those of other E

U-fund

ed program

mes.

20 See de la Fu

ente and C

iccone (2002) for a detailed

discu

ssion of these issues and

a review of the

available empirical evid

ence.

28

therefore su

bstantial, an

d retu

rns are gen

erally high

er in th

e most ad

vanced

Objective 1 regions (V

alencia and C

anarias) and in those that have the low

est stocksof capital per job (G

alicia and M

urcia).

Figure 3: A

verage social rate of return

on C

SF p

ub

lic expen

ditu

re by region

0%

10%

20%

30%

40%

Val

Cana

Gal

Mur

And

Ast

Cant

CyL

C-M

Ext

scenario 1scenario 2

avge.

The w

ide d

ispersion of returns across regions suggests that the current criteria for theallocation of E

uropean cohesion expend

iture generate an im

portant efficiency cost --or equ

ivalently, that the overall imp

act on the Spanish econom

y could

be mu

chgreater if efficiency consid

erations were given greater w

eight in the allocation ofthese fund

s. This w

ould certainly entail an im

portant change in the orientation of EU

cohesion policy as structu

ral assistance wou

ld shift tow

ards the richer regions of the

cohesion countries. T

his wou

ld p

robably favour faster convergence am

ong mem

berstates at the cost of som

e increase in internal inequality. But since there are im

portantred

istribution m

echanisms in op

eration within m

ember cou

ntries, a significant part

of the income gains w

ould

be redirected

toward

s the poorer regions. For the case ofSpain, I have estim

ated elsew

here that a policy shift in this direction w

ould generate

a net welfare gain. 21

5. Th

e imp

act on grow

th an

d em

ploym

ent: i) m

ediu

m an

d lon

g-term effects

In this section I will present estim

ates of the cum

ulative effects of the Fram

ework on

outp

ut an

d em

ploym

ent in

the m

ediu

m an

d lon

g run

. Th

ese estimates are

21 See de la Fuente (2002a).

29

constructed

und

er the assum

ptions of Scenario 2, taking as a reference the 1993

values of the relevant variables. In particular, the calculations that follow assum

e thatin the absence of the C

SF the stocks of the different p

rodu

ctive factors (and hence

regional outpu

t, in the absence of technical progress) wou

ld rem

ain constant foreverat their 1993 levels. T

o quantify the Framew

ork's contribution, I add

to these baselinefactor stocks the accu

mu

lated and

prop

erly dep

reciated flow

s of CSF-financed

investment and

calculate the resu

lting change in outp

ut and

emp

loyment u

sing them

odel of Section 2. T

he details of the com

pu

tations are discu

ssed in Section e of

Append

ix 1.

Figure 4: C

um

ulative im

pact of th

e 1994-99 CS

F on factor stock

sen

tire Ob

jective 1 territory

0% 5%

10%

15%

20%

25%

19941999

20042009

2014

infrastructuresother capital

years of training

- Note: cum

ulative logarithmic d

ifference from the value of each variable in 1993 ind

uced by the C

SF.A

ll calculations are mad

e under the assum

ptions of Scenario 2.

Figures 4 and

5 show the cu

mu

lative imp

act of the CSF on the stocks of p

rodu

ctivefactors and

on the levels of outp

ut and

emp

loyment of the entire set of O

bjective 1regions (exclu

ding C

euta and

Melilla) d

uring the p

eriod 1994-2015. Figu

re 4 shows

that the CSF can be seen as a large positive "shock" that, over a period

of seven years,raises aggregate factor stocks significantly above their starting levels (u

p to 20%

inthe case of infrastru

ctures). O

nce the Framew

ork has been executed

(and assu

ming

there are no new interventions), the stocks of physical capital and

infrastructures areallow

ed to grad

ually retu

rn to their original levels as CSF-financed

investments

dep

reciate. The im

pact on the stock of hu

man cap

ital, by contrast, remains constant

until the end

of the working life of the beneficiaries of training p

rogramm

es which,

on average, will take place after the end

of the period covered

in the figure.

30

Figure 5: C

um

ulative im

pact of th

e 1994-99 CS

F on ou

tpu

t and

emp

loymen

t

entire O

bjective 1 territory

0% 1% 2% 3% 4% 5% 6% 7% 8%

19941999

20042009

2014

outputem

ployment

- Note: cum

ulative logarithmic d

ifference from the value of each variable in 1993 ind

uced by the C

SF.A

ll calculations are mad

e under the assum

ptions of Scenario 2.

Figure 5 traces ou

t the imp

act of these shocks on the evolution of ou

tpu

t andem

ploym

ent. As m

ay be expected

, the outp

ut effect has ap

proxim

ately the same

profile as factor stocks, and

begins to decline as soon as the Fram

ework has been

completely execu

ted (that is, after 2000). T

he time path of em

ployment, on the other

hand, is very d

ifferent from the p

revious one. Since this variable ad

justs slu

ggishlyover tim

e, net job creation remains positive u

ntil about 15 years after the conclu

sionof the program

ming period

.

Table 17 su

mm

arizes the cum

ulative im

pact of the Fram

ework on the ou

tpu

t andem

ployment of each of the O

bjective 1 regions in 2000 and 2005. T

he table shows that

the growth effects of the C

SF vary significantly across regions, reflecting differences

in both the volume of investm

ent and its rate of return. For the O

bjective 1 regions asa w

hole, the Framew

ork add

s 6.9 percentage p

oints to outp

ut and

3.4 points to

emp

loyment in 2000. W

hen we take as ou

r reference the entire country, the C

SFcum

ulative contributions to Spanish growth and

employm

ent in the same year are of

3.5 and 1.85 points respectively.

31

Tab

le 17: Cu

mu

lative imp

act of the 1994-99 O

bjective 1 C

SF

________________________________________________________________ accum

ulated over 1994-2000 acumulated over 1994-2005

_____________________________ _ ______________________________

%∆

Y%

∆ em

ploy.no. of jobs

%∆

Y%

∆ em

ploy.no. of jobs

Andalucía

6.79%3.29%

60,6056.20%

4.19%77,130

Asturias

7.80%3.78%

13,1327.23%

4.86%16,897

Canarias

6.90%3.35%

15,9656.39%

4.30%20,518

Cantabria

7.65%3.71%

6,2107.00%

4.73%7,914

Castilla y León

7.28%3.53%

29,1536.66%

4.50%37,185

Castilla la M

ancha6.16%

2.99%15,298

5.66%3.82%

19,535V

alencia5.03%

2.44%31,807

4.68%3.15%

40,973E

xtremadura

6.53%3.16%

9,3706.04%

4.06%12,035

Galicia

9.67%4.68%

44,0328.91%

5.99%56,443

Murcia

7.95%3.85%

12,2847.33%

4.93%15,749

total Objective 1

6.92%3.38%

237,8566.37%

4.33%304,380

total/Spain3.44%

1.85%3.17%

2.37%________________________________________________________________- N

otes: Spain exclu

des C

euta and

Melilla. C

alculations based

on Scenario 2. Percentage (rather than

logarithmic) increm

ents over 1993 regional output and em

ployment.

Tab

le 18: Con

tribu

tion of th

e CS

F to regional grow

th an

d con

vergence

________________________________________________________________ grow

th 94-00 CSF contribution 1994-2000 convergence effect

__________________ ___________________________ __________________(1)

(2)(3)

(4)(5)

(6)(7)

netgross

total%

net%

grossypc dif. 93

conv. ratioA

ndalucía21.02%

45.49%6.79%

32.31%14.92%

-43.76%15.52%

Asturias

13.19%37.22%

7.80%59.11%

20.96%-26.96%

28.93%C

anarias36.55%

60.60%6.90%

18.88%11.39%

-22.42%30.78%

Cantabria

21.97%46.02%

7.65%34.84%

16.63%-25.05%

30.54%C

astilla y León17.09%

41.14%7.28%

42.58%17.69%

-24.44%29.79%

Castilla la M

.24.71%

48.74%6.16%

24.95%12.64%

-32.89%18.73%

Valencia

29.92%54.08%

5.03%16.80%

9.30%-18.45%

27.27%E

xtremadura

23.41%47.16%

6.53%27.90%

13.85%-45.76%

14.27%G

alicia21.99%

46.13%9.67%

43.97%20.96%

-31.17%31.02%

Murcia

28.89%52.87%

7.95%27.50%

15.03%-34.49%

23.05%total O

bj. 123.91%

48.08%6.92%

28.93%14.39%

-32.16%21.52%

EU

's contrib.4.82%

20.15%10.02%

14.99%

________________________________________________________________

Table 18 help

s pu

t the effects of the Framew

ork in persp

ective by comp

aring themw

ith observed output grow

th between 1993 and

2000 and w

ith the initial differential

in income p

er capita betw

een each region and the average of the territories that are

not includ

ed in O

bjective 1. The first tw

o colum

ns of the table show the observed

cum

ulative grow

th of regional outp

ut betw

een 1993 and 2000, d

istinguishing

32

between net and

gross growth. T

he first of these variables refers to the observedgrow

th of value ad

ded

, and the second

one is calculated

by add

ing to the first anestim

ate of the decline in regional ou

tpu

t that wou

ld have been ind

uced

du

ring thesam

e period

by the d

epreciation

of the in

itial stocks of ph

ysical capital an

dinfrastructures in the absence of any investm

ent. 22 Colum

n (3) shows the cum

ulativecontribution of the C

SF to output growth in 2000, and

columns (4) and

(5) display the

result of divid

ing this contribution by net and by gross grow

th respectively (columns

(1) and (2)).

For the Objective 1 regions taken as a w

hole, the Framew

ork's contribution accou

ntsfor alm

ost 30% of the (net) ou

tpu

t growth observed

between 1993 and

2000. This

figure, how

ever, overestimates the im

portance of the C

SF because it im

plicitly

allocates the entire burd

en of replacing w

orn out cap

ital to non-CSF investm

ent.W

hen the calculation is repeated taking as a reference gross grow

th, the Framew

ork'scontribu

tion drop

s to a bit less than 15% for the entire O

bjective 1 territory, andexceed

s 20% in A

sturias and G

alicia.

The last group of colum

ns quantifies the Framew

ork's contribution to convergence inincom

e per cap

ita between O

bjective 1 regions and the rest of the cou

ntry. Colu

mn

(6) shows the incom

e per cap

ita differential betw

een each region in the samp

le andthe average valu

e of the same variable in the rem

ainder of Sp

ain. Divid

ing theFram

ework's contribu

tion to outp

ut grow

th (colum

n (3)) by the previou

s variable,w

e obtain a convergence ratio (colum

n (7)) that measu

res the fraction of the originalincom

e gap that w

ould

have disap

peared

as a result of the execu

tion of theFram

ework (if the population of the d

ifferent regions had rem

ained constant over the

sample period

and grow

th performance had

been uniform

across regions except for

the effects of the CSF). For the w

hole of the territory covered by the Fram

ework this

coefficient is a bit over 20%, and

reaches values above 30%

for Canarias, C

antabriaand

Galicia.

Finally, the last row of the table contains an estim

ate of the contribution of EU

fund

sper se (that is, of the part of the Fram

ework that is financed

by EU

grants, exclud

ingnational cofinancing) to grow

th and convergence. T

his estimate is obtained

bym

ultiplying the total effect of the Fram

ework by the w

eight of EU

grants in the total

22 To qu

antify the imp

act of dep

reciation, I follow the sam

e proced

ure u

sed above to estim

ate thecontribu

tion of the CSF u

nder the assu

mp

tion that investment is zero d

uring the p

eriod u

nder

consideration.

33

pu

blic expend

iture channeled

by the CSF. I estim

ate a value of 69.67%

for thiscoefficient, w

hich is calculated using d

ata from the PFP. 23

6. Con

clusion

In this pap

er I have quantified

the contribution of the 1994-99 C

SF to outp

ut and

employm

ent growth in the O

bjective 1 regions of Spain using a regional prod

uction

function and an em

ployment equation estim

ated w

ith Spanish regional data.

It is imp

ortant to emp

hasize that these estimates shou

ld be interp

reted w

ith a fairam

ount of cau

tion for at least two reasons that tend

to increase the margin of error

above the level that is already inevitable in any exercise of this type. T

he first one isthe lack of consensu

s in the literature on the valu

es of some cru

cial parameters, and

in p

articular on

the coefficien

ts that m

easure th

e imp

act of investm

ent in

infrastructu

re and hu

man capital on ou

tput grow

th. Althou

gh my estim

ates of thesep

arameters seem

quite reasonable and

fall within the range of valu

es obtained in

similar stud

ies for Spain, the great diversity of results found

in the literature must be

kept in

min

d. 24 Second

ly, the analysis in this pap

er is based on the im

plicit

assum

ption that investm

ent financed by the Stru

ctural Fu

nds has exactly the sam

eim

pact as other p

rojects of the same natu

re. It is possible, how

ever, that because of

the low

margin

al cost of these resou

rces, both to th

e nation

al and

regional

adm

inistrations and to the private sector, they m

ay be used to finance projects w

hichw

ould

not survive a strict cost-benefit analysis, or that the m

anagement of these

fund

s may be less efficient. T

o investigate the validity of this hyp

othesis, which

underlies the w

idespread

criticisms of w

aste and inefficiency that are often leveled

atthe Stru

ctural Fu

nds, it w

ould

be necessary to und

ertake an analysis of theird

ifferential imp

act that wou

ld requ

ire rather detailed

data w

hich are currently not

available.

With these caveats, m

y results d

o suggest that the contribu

tion of the Structu

ral andC

ohesion Fund

s to the growth of Sp

anish outp

ut and

emp

loyment and

to theconvergence of assisted

regions with the rest of the cou

ntry has been considerable.

For the Objective 1 regions as a w

hole, the CSF has ad

ded

around

one percentage

23 This sou

rce does not give a breakd

own of the C

ohesion Fund

by source of financing. For this

instrument, I have assum

ed that E

U grants am

ount to 80% of public expend

iture.24 See for instance E

vans and K

arras (1994), Holtz-E

akin (1994), García-M

ilà, McG

uire and

Porter

(1996) and G

orostiaga (1999) for largely negative results on the grow

th effects of infrastructu

reinvestm

ent. De la Fuente (2002c) contains a survey of this literature.

34

point p

er year to outp

ut grow

th, and 0.4 p

oints per year to em

ploym

ent growth (or

27,000 new jobs). In the m

ediu

m ru

n, the accum

ulated

imp

act on emp

loyment

exceeds 300,000 new

jobs, and the contribu

tion to the growth of ou

tpu

t in the lessfavou

red regions exceed

s six percentage p

oints. This am

ounts to 20%

of the initialgap in incom

e per capita between the O

bjective 1 regions and the rest of Spain.

My estim

ates also suggest that the retu

rn on pu

blic CSF exp

enditu

re has been quite

high. What I have called

, perhap

s mislead

ingly, the "social" rate of return on this

expend

iture has been arou

nd 30%

. Althou

gh this figure d

oes not take private costs

into account and

should

therefore not be comp

ared w

ith more stand

ard rates of

return, it does suggest that prod

uctive public spending has been an im

portant sourceof prod

uctivity gains in assisted regions.

As for the relative retu

rns on the different typ

es of CSF exp

enditu

re, the results are

extremely sensitive to the crow

ding-in assu

mptions em

bodied

in the two alternative

scenarios I have analyzed. If w

e take the CSF at face valu

e and assu

me that the

private investment contem

plated in it has been ind

uced

by, and is ad

ditional to, E

Ugrants, then aid

to private enterp

rises is the program

me that generates the greatest

increase in outp

ut and

emp

loyment p

er unit of p

ublic exp

enditu

re. On the other

hand, if w

e rely on more d

irect estimates of the im

pact of the various programm

es onp

rivate investment, exp

enditu

re in infrastructu

re is the alternative with the highest

rate of return. Since the second of these scenarios is based

on what I believe are m

orerealistic assum

ptions, I interpret these results as a clear indication that infrastructu

reinvestm

ent should

continue to be given p

riority until the d

eficits in this area thatpersist in Spain have been substantially red

uced.

Finally, I have also found that there are very im

portant differences in rates of return

on Structu

ral Fund

investment across regions. T

his suggests that the im

pact of

European grants on the Spanish econom

y as a whole could

be significantly increasedby assigning som

e weight to efficiency criteria in the regional allocation of these

fund

s. This w

ould

of course have a certain cost in the form

of slower convergence in

prod

uctivity across regions, and

wou

ld rep

resent a significant dep

arture from

thep

rinciples that cu

rrently guid

e EU

cohesion policies. B

ut, to the extent that the

existing m

echan

isms for red

istribution

at the p

ersonal level gu

arantee a fair

distribu

tion of the resulting efficiency gains, the net effect of su

ch a policy change

could be a significant w

elfare gain.

35

Ap

pen

dix 1: T

heoretical fram

ework

and

estimation

1. Th

eoretical framew

ork

This section d

evelops the descriptive m

odel of regional grow

th and em

ployment that

has been used

to prod

uce the estim

ates reported

in the body of the p

aper. T

he firstcom

ponent of the m

odel is a p

rodu

ctivity equation that com

bines an aggregateprod

uction function with a technical progress relation w

hich allows for technological

diffu

sion across regions. The sp

ecification is the one prop

osed in d

e la Fuente

(2002b), expand

ed to inclu

de the stock of infrastru

ctures as an argu

ment of the

produ

ction function. T

he second equ

ation describes the evolu

tion of employm

ent asa fu

nction of the behaviour of factor stocks and

wages and

is informally m

otivatedby com

bining a competitive labou

r dem

and sched

ule w

ith a story about ad

justm

entcosts.

a. Prod

uctivity

I will assum

e the aggregate production function is a a C

obb-Douglas of the form

25

(1) Yit =

Kit θ

k Pit θ

p (Ait R

i Lit H

it ) θh (A

it Ri L

it ) λ = K

it θk P

it θp H

it θh (A

it Ri L

it ) θl

where the coefficient of labour in the second

expression on the right-hand sid

e, θl =

λ+

θh is the su

m of the elasticities of ou

tput w

ith respect to employm

ent per se and to

the stock of human capital. In this expression Y

denotes aggregate regional output, K

the stock of (non-infrastructu

re) physical cap

ital, P the stock of infrastru

cture, L

isem

ployment and

H an ind

icator of the stock of hum

an capital per worker. T

he main

difference w

ith standard

specifications is that I assu

me that the ind

ex of regionaltechnical efficiency has tw

o distinct com

ponents, Ait and

Ri . I interp

ret the first one,A

it , as an index of "transferable" technical know

ledge, and

the second one, R

i , as aterm

which cap

tures sp

ecific and non-transferable regional characteristics that m

ayhave an im

pact on p

rodu

ctivity (e.g. geographic location, clim

ate, endow

ments of

natural resources and other unobserved

regional characteristics).

25 Notice that equation (1) d

iffers from the prod

uction function shown in Section 2 of the text in that it

includ

es a time-invariant regional effect, R

i . This is im

portant in the estimation, bu

t I have omitted

itin the text to sim

plify a bit the exposition.

36

Taking logarithm

s of this expression (denoted

by lower-case letters),

(2) yit =

θl ri +

θl ait +

θk kit +

θp pit +

θh hit +

θl lit ,

differencing the resu

lt and ad

ding a rand

om d

isturbance (ω

it ), the equation to be

estimated

is of the form:

(3) ∆yit =

θl ∆

ait + θ

k ∆kit +

θp ∆

pit + θ

h ∆hit +

θl ∆

lit + ω

it .

At this stage, the stand

ard p

ractice in the literature involves treating the level of

technical efficiency (ri +ait ) and

/or its grow

th rate (∆ait ) as exogeneou

s unobservable

variables and introd

ucing fixed

or random

effects to control for possible d

ifferencesin them

across regions and period

s. It seems preferable, how

ever, to control directly

for these factors to the extent that it is possible. W

ith this pu

rpose, I w

ill partially

endogenize the rate of technical progress, allow

ing for technological diffusion across

regions. 26

I will start by w

riting the (log of the) level of transferable technical efficiency ofregion i at tim

e t in the form

(4) ait = at +

dit

where at =

(1/N) ∑

i ait is the "national average" of ait and dit =

ait - at the "tecnological

distance" betw

een region i and the national average. In w

hat follows, I w

ill treat theaverage level of (transferable) technical efficiency, a

t , as an exogenous variable

(possibly d

etermined

by the technological gap betw

een Spain and

other countries

and the level of R

&D

effort) and focu

s on the determ

inants of the evolution of the

relative technical efficiency of each region.

In particular, I will assum

e that

(5) ∆at =

g + ct,

i.e. that the average rate of technical progress is equal to an exogenous constant plus,possibly, a trend

, and that the technological d

ifferential of region i evolves following

the equation

26 The original sp

ecification in de la Fu

ente (2002b) also allows the rate of technical p

rogress to be afunction of the relative ed

ucational attainment of each region. Since this rate effect from

human capital

turns out not to be significant when regional fixed

effects are included

in the mod

el, I have excluded

itfrom

the start.

37

(6) ∆dit =

- εdit .

That is, the technical progress d

ifferential in favour of a given region is an increasingfunction of its technological gap relative to the sam

ple average. If technology diffuses

across regions, the coefficient of dit should

be negative -- that is, other things equal,

the rate of technical progress should be higher in the less d

eveloped regions. T

he signof the coefficient ε w

ill therefore allow us to test the hypothesis that there is a process

of technological convergence across regions. Since the fixed effects, ri , m

ay differ

across territories, convergence in TFP

levels will only be cond

itional, with each

region gradually approaching a stable level of relative technical efficiency w

hich will

be determ

ined by the characteristics sum

marized

by ri and by the speed

of diffusion,

ε.Ad

ding up (5) and

(6), the rate of techical progress in region i during period

t will be

given by:

(7) ∆ait =

∆at +

∆dit =

g + ct - εdit .

Substitu

ting this expression into (3) w

e obtain a specification of the p

rodu

ctionfu

nction in first differences in w

hich the rate of technical progress in each region is

expressed as a function of its technological gap w

ith respect to the national average.

In order to estim

ate this equation w

e have to find som

e way of m

easuring the

transferable tech

nological gap

, dit . This variable is not d

irectly observable in

principle but, since we have d

ata on factor stocks and regional output, w

e can invertthe p

rodu

ction function and

write d

it as a function of observable variables and

coefficients to be estimated

. In particu

lar, solving for ait in (2) and ignoring the

disturbance w

e have:

(8) ait = yit - θ

k kit - θp pit - θ

h hit - θl lit - θ

l riθ

l .

Since the equation is linear in logs, m

oreover, the same relation w

ill hold am

ong theaverages of the relevant variables. T

his allows us to com

pute at using

(9) at = yt - θ

k kt - θp pt - θ

h ht - θl lt

θl

- r

38

where the absence of the subind

ex i indicates that w

e are working w

ith interregionalaverages (of th

e variables in logs). Su

btracting (9) from

(8), the tran

sferabletechnological gap of region i relative to the sam

ple mean at tim

e t will be given by:

(10) dit = a ∼it =

ait - at = y ∼it - θ

k k ∼it - θp p ∼it - θ

h h ∼it - θl l ∼it

θl

- r ∼ i

where the tild

es denote d

eviations from the average and

, in particular, r ∼ i = ri - r, w

ith

r = (1/N

) ∑i ri .

Com

bining (7) and (10) w

e can finally write the rate of technical progress of region i

in the form

(11) ∆ait =

g + ε r ∼ i +

ct - ε y ∼it - θk k ∼it - θ

p p ∼it - θh h ∼it - θ

l l ∼it

θl

Substitu

ting this expression into (3) and

introdu

cing du

mm

y variables (DR

EG

i ) tocap

ture the fixed

regional effects, ri , we finally arrive at a sp

ecification written

entirely in terms of observable variables and

coefficients to be estimated

:

(12) ∆yit =

θl (g +

εr ∼ v + ct) +

θk ∆

kit + θ

p ∆pit +

θh ∆

hit + θ

l ∆lit

- ε

y ∼it - θk k ∼it - θ

p p ∼it - θh h ∼it - θ

l l ∼it - Γ

ii≠

v∑

DR

EG

i + ω

it

where the su

bindex v d

enotes a reference region and the coefficient of the i-th

regional dum

my is of the form

Γi =

θl r ∼

i - θl r ∼

v .

b. E

mp

loymen

t

Und

er conditions of perfect com

petition and absence of ad

justm

ent costs, firms w

illchoose em

ployment so that its m

arginal produ

ct is equal to the real w

age. Om

ittingall subind

ices, this condition can be w

ritten

∂Y∂L =

Kθ

k Pθ

p Hθ

h (RA

) θl θ

l Lθ

l -1 = W

,

which im

plicitly d

efines a regional labour d

emand

function. Solving for L

, theoptim

al employm

ent level will be given by

39

L* =

θ

l Kθ

k Pθ

p Hθ

h Aθ

l

W 1/(1−θ

l )

and taking logs (d

enoted as usual by low

er case letters), we obtain

(13) l* =

11-θl [ln θ

l + θ

k k + θ

p p + θ

h h + θ

l (a+r) - w

].

Taking first d

ifferences of this expression, w

e can write the grow

th rate of labour

dem

and as a function of the grow

th rates of factor stocks and real w

ages:

(14) ∆l* =

11-θ

l (θk ∆

k + θ

p ∆p +

θh ∆

h + θ

l ∆a - ∆

w) .

If we are w

illing to assum

e that emp

loyment levels in the Sp

anish regions ared

emand

-determ

ined (w

hich seems reasonable enough at least in the last tw

o decad

esin view

of the extremely high rates of u

nemp

loyment observed

in all of them), w

ecan u

se any of the equations w

e have just d

erived to analyze the evolu

tion ofem

ployment in our sam

ple (being careful to allow in the estim

ation for the more than

likely endogeneity of the average w

age). This labou

r dem

and sched

ule, how

ever,assu

mes that em

ploym

ent adju

sts imm

ediately to changes in its d

eterminants -- an

assum

ption w

hich is probably far from

reasonable, as suggested

also by some

preliminary attem

pts to estimate (13) or (14) d

irectly.

In view of the existence of consid

erable adjustm

ent costs (induced

in part by Spanishlabou

r legislation), a more reasonable assu

mp

tion is that emp

loyment tend

s toap

proach the long-term

level described

by equation (13) only grad

ually. L

etting dd

enote the exogenous rate of job d

estruction and

γ the emp

loyment ad

justm

ent

coefficient, a simp

le stock adju

stment m

odel w

ould

be given by the following

equation

lt+1 = lt - d +

γ(lt+1 * - lt )

which can be rew

ritten in the form

∆lt =

lt+1 - lt = - d +

γ[(lt+1 * - lt *) + (lt * - lt )]

or

(15) ∆lt =

-d + γ∆

lt * + γ(lt * - lt ).

40

After som

e attempts to estim

ate an equation of this form

, I have opted for a slightly

more general specification w

hich allows the coefficients on the last tw

o terms on the

right-hand sid

e to differ from

each other. The em

ploym

ent equation I estim

ate is ofthe form(16) ∆

lt = -d +

γ1 ∆lt * +

γ2 (lt * - lt ).

Some ad

ditional m

anipu

lation is required

before this equation is in a form

suitable

for estimation. U

sing the preceding expressions, the last term

of (16) is of the form

(17) lt * - lt =

11-θl [ln θ

l + θ

k k + θ

p p + θ

h h + θ

l (a+r) - w

- (1-θl )l] .

Notice that this equ

ation includ

es the term a+

r, which is not d

irectly observable. To

eliminate it, w

e use the production function in levels given in equation (2) to get

θk k +

θp p +

θh h +

θl (a+

r) = y - θ

l l,

and substitute this expression into (17) to arrive at

(18) lt * - lt =

11-θl (ln θ

l + y - l - w

) .

This exp

ression says that the gap betw

een observed and

long-term em

ploym

ent isp

roportional to u

nit labour costs (i.e. to the ratio betw

een the real wage and

outp

ut

per worker).

Using (18) in (15), the em

ployment equation can be w

ritten in the form:

(19) ∆lt =

-d + γ1 ∆

lt * + γ2 (lt * - lt )

=

ln θ

l

1-θl - d

+ γ1 1-θ

l (θk ∆

k + θ

p ∆p +

θh ∆

h + θ

l ∆a - ∆

w) +

γ21-θ

l (ln θl +

y- l - w) .

Notice that this equ

ation also includ

es an unobservable term

( ∆a). W

e can, however,

use equ

ation (11) to write ∆

a as a function of observable variables and

coefficients to

be estimated

.

41

2. Sp

ecification an

d em

pirical resu

lts

I estimate equations (12) and

(19) jointly using a panel of data for the Spanish regions

covering the period

1964-93 at intervals of generally two years. T

he system form

edby these tw

o equations is estim

ated by non-linear 3SL

S imposing constant retu

rns toscale in prod

uction (that is, θk +

θp +

θl =

1) and all the cross-equation restrictions on

the coefficients implied

by the theoretical mod

el.

The choice of an instru

mental variables techniqu

e seems ap

prop

riate given thesu

spected

endogeneity of (at lest) several of the regressors. In p

articular, I treat as

endogenou

s variables the level and grow

th rate of average wages and

the growth

rate of the stock of infrastructures. This last variable is instrum

ented because there is

evidence that the investm

ent behaviour of the p

ublic ad

ministration in Sp

ain issensitive both to efficiency and

to equity considerations. 27

The instru

ments chosen are (the logs of) the initial stock of infrastru

cture (kinf), the

level of employm

ent (le), aggregate regional output (q), the average level of schoolingof the w

orking-age population (hpet) and the grow

th rates of this last variable (ghpet)and

of the working-age p

opu

lation (gpet). The first three variables are intend

ed as

instrum

ents for the growth rate of the stock of infrastru

ctures, as the average

prod

uctivity of this factor (q - kinf) and

its stock per w

orker (kinf - le) may be

reasonable indicators of infrastructure need

s and expected

returns, the two variables

that seem to d

rive pu

blic investment d

ecisions. The rem

aining variables should

capture factors that affect wages through labour supply.

The equ

ations I estimate also inclu

de tw

o ad-hoc term

s that do not com

e out of the

derivation in the preced

ing section. To pick up cyclical d

isturbances, I have included

as a regressor in the prod

uction equ

ation the average annual increase in the rate of

unem

ploym

ent. In the emp

loyment equ

ation, I control for the growth rate of non-

salaried em

ploym

ent, as my d

erivation ignores self-emp

loyment, w

hich is quite

significant in the data. Finally, I introd

uce a trend

which allow

s the rate of jobd

estruction to increase over time (that is, d =

do +

d1 t).

Table A

.1 summ

arizes the results of the estimation.

27 See de la Fuente and

Vives (1995) and

de la Fuente (1996).

42

Tab

le A.1: E

mp

irical results

________________________________________________________________________param

etercoeff.

(t)param

etercoeff.

(t)θk

0.297(5.73)

θl g+εrv0.025

(3.64)

θp0.106

(2.14)θl c

-0.0003(1.93)

θh0.286

(7.30)ε

0.206(7.20)

θl0.597

do-0.008

(2.51)

γ10.181

(6.47)d1

-0.00036(2.88)

γ20.040

(5.21)gnoasal

0.247(9.21)

dU-0.060

(1.01)R

2 (12)0.588

N238

R2 (19)

0.484________________________________________________________________________

Notes

- t statistics in parentheses.- T

he coefficient of emp

loyment, θl , is not estim

ated d

irectly but recovered

using the assu

mp

tion ofconstant returns to scale in capital, infrastructures and

labour, i.e. θl = 1 - θk - θ

p .- N

is the number of observations. T

he D-W

statistics for equations (12) and (19) are, respectively, 2.13

and 1.65. T

he prod

uction fu

nction includ

es fixed regional effects, w

hich enter as shown in equ

ation(12). T

he reference region is Valencia.

3. Com

pu

ting "social" rates of retu

rn

The "social" rates of retu

rn reported

in Section 4 of the text are comp

uted

und

er theassu

mp

tion that the marginal p

rodu

ct of capital rem

ains constant over time. I

imagine a regional econom

y in a steady-state p

osition, with a constant stock of

capital K

o and other p

rodu

ctive factors, and a level of incom

e Yo w

hich, in the

absence of shocks, wou

ld rem

ain constant forever. Given this initial situ

ation, Iassum

e that at a given point in time (t =

0) an investment project is und

ertaken which

increases the initial capital stock by I = ∆

Ko units. T

his investment is then allow

ed to

depreciate (at a constant rate δ) u

ntil the regional capital stock returns to its original

level.

New

investm

ent gen

erates an in

come stream

, ∆Y

t , wh

ich at tim

e t

can

be

approximated

by the expression

(20) ∆Y

t = M

Pk ∆

Kt =

MP

k ∆K

o e -δt = ∆

Yo e -δt

where ∆

Kt =

∆K

o e -δt is the increase in the capital stock induced

by the project at time t

and M

Pk is the m

arginal produ

ct of capital which (for relatively low

values of I) can

be assumed

constant since, except for the investment und

ertaken at time 0, the stocks

of productive factors rem

ain fixed by assum

ption.

43

The social rate of retu

rn on pu

blic investment is d

efined as the d

iscount rate ρ

thatm

akes the net present valu

e of the investment p

roject equal to zero. T

hat is, ρ is the

solution to the following equation

(21) NP

V =

- I + ∫

o ∞ ∆Y

t e -ρt dt = 0.

Substituting (20) into (21),

I = ∫

o ∞ ∆Y

o e -δt e -ρt dt,

and solving the integral, w

e have

I = ∆

Yo

δ+ρ ,

where w

e can solve for ρ:

(22) ρ = ∆

Yo

I - δ.

In the calculations summ

arized in Sections 4c and

4d of the text, I is public investm

entd

urin

g 1994 and

∆Y

o den

otes its total estimated

contribu

tion to 1994 ou

tpu

t

(includ

ing indirect effects throu

gh indu

ced em

ployment), both m

easured

in millions

of 1990 pesetas. When there are no ind

uced investm

ent effects, the results reported in

the text are obtained d

irectly from equ

ation (22) using the d

epreciation rate implicit

in the last year of the data.

When there are ind

uced

investment effects, or w

hen we consid

er the return on the

CSF as a w

hole, the comp

utation is slightly m

ore comp

licated becau

se the stocks ofseveral d

ifferent prod

uction factors m

ay be affected at once. In this case, p

ublic

investment can generate d

ifferent income flow

s (say ∆Y

1t and ∆

Y2t ) w

hich decrease

over time at p

ossibly different rates that reflect the rates of d

epreciation of the

relevant capital stocks (say δ1 and

δ2 ). In this case, the sam

e argument as above lead

s

to the equation

(23) I = ∆

Y1o

δ1 +ρ + ∆

Y2o

δ2 +ρ

which is solved

numerically for ρ.

44

Finally, in the case of training expend

iture I have assu

med

that the increase in thestock of hum

an capital financed by the C

SF disappears all at once after T

periods (the

estimated

usefu

l life of training programm

es). In this case, the rate of depreciation is

zero, but the increm

ental stream of ou

tpu

t lasts only for a finite period

. The rate of

return is then the solution to the equation

(24) NP

V =

- I + ∫o T ∆

Yo e -ρt dt =

- I + ∆

Yo 1 - e -ρT

ρ =

0

when there is no ind

uced

private investm

ent. In more com

plicated

cases, I solve anextension of equ

ation (23) in which the term

that measu

res the present valu

e of thed

irect contribution of training exp

enditu

re to outp

ut has the sam

e form as the last

term on the right-hand

side of (24).

4. Calcu

lation of th

e med

ium

and

long-term

effects

The cu

mu

lative increase in the log of outpu

t and em

ployment ind

uced

by the CSF is

calculated

by sum

min

g the con

tribution

s to these variables of in

vestmen

t ininfrastru

ctures, other physical capital and

hum

an capital financed or ind

uced

by theC

SF. These contribu

tions are calculated

using the p

rocedu

re that is described

ind

etail below for the case of infrastructures, keeping in m

ind that in the case of hum

ancap

ital dep

reciation takes place all at once at the end

of the estimated

usefu

l life.O

nce we have calculated

the total increase in the logs of output and em

ployment, the

changes in the levels of these variables (measu

red in m

illions of 1990 pesetas and in

jobs created) are recovered

in the way ind

icated in footnote 16 to the text. A

llestim

ates of cum

ulative effects are p

rodu

ced u

nder the assu

mp

tions of Scenario 2.H

ence, total investment in physical capital (k) is obtained

as the sum

of direct pu

blicinvestm

ent in this factor, subsidies to private sectors and

the private investment that

is indu

ced by the p

revious tw

o items and

by investment in infrastru

ctures and

intraining.

We w

ill now

work th

rough

the d

etails of the calcu

lations for th

e case ofinfrastru

cture investm

ent. Let K

INF

io be the stock of this factor in region i at the end

of 1993. First, we accu

mu

late the flow of infrastru

cture investm

ent financed by the

CSF (m

easured in m

illions of 1990 pesetas) using the same d

epreciation rate as in thecalcu

lation of the social rate of return for this factor. In this w

ay we obtain for each

region i and each year t an estim

ate of the Framew

ork's contribution to the stock of

infrastructu

res (KM

AC

it ). This variable is extend

ed to 2015 by letting the stock of

45

accum

ulated

Structu

ral Fund

investment d

epreciate with the passage of tim

e. In thisw

ay, we take into accou

nt the fact that CSF-financed

projects w

ill continue to affect

output in the future until they are fully depreciated

.

Next, w

e calculate the cumulative contribution of the C

SF to the increase in the log ofthe stock of infrastru

ctures in each region (D

KIN

Fit ) and its annu

al contribution to

the same variable (dK

INFit ),

(A.25) D

KIN

Fit =

ln (KIN

Fio +

KM

AC

it ) - ln (KIN

Fio ) and

(A.26) dK

INF

it = D

KIN

Fit - D

KIN

Fit-1 .

We can now

estimate the im

pact of the C

SF on regional outp

ut and

emp

loyment.

Notice that there are several effects to consid

er. First, an increae in the stock ofinfrastru

ctures has a d

irect effect on outp

ut (Y

) through the p

rodu

ction function

given in equation (1) of the text. T

o calculate this effect (w

hich will be d

enoted by

DY

1 or dY1), w

e multiply D

KIN

Fit or dK

INF

it by the elasticity of outpu

t with respect

to the stock of infrastructures, that is

(A.27) D

Y1

it = θ

inf DK

INF

it and dY

1it =

θinf dK

INF

it .

Second, an increase in the stock of infrastru

ctures also raises the d

emand

forem

ploym

ent, although only grad

ually. T

o quantify this effect, w

e need to start by

calculating the increase in the long-term labour d

emand

, which is given by

(A

.28) Dlt * =

11-θ

l θinf D

KIN

Fit and

dlt * =

11-θl θ

inf dKIN

Fit

where, as before, w

e use D

to denote cu

mu

lative differences (i.e. the total d

ifferencebetw

een the value of the variable of interest in period t and

its baseline or 1993 value)and

d to refer to annual increases. A

ccording to the equ

ation that describes the

evolution of em

ploym

ent, lit , (equation (3) in the text), an increase in long-term

labour d

emand

has two effects on em

ploym

ent. The first one (d

enoted by dl1) is

imm

ediate and

is given by

(A.29) dl1

it = γ1 dlit *

while the second

one (dl2it ) cap

tures the p

artial redu

ction in the initial gap betw

een

employm

ent and long-term

labour dem

and,

(A.30) dl2

it = γ2 (D

lit-1 * - Dlit-1 ).

46

Ad

ding u

p dl1 and

dl2 we obtain the total change in em

ploym

ent observed d

uring

the current year (dlit ) and

, sum

ming it to the increm

ent accum

ulated

in previou

speriod

s, we can recursively construct the variable D

lit that measures the accum

ulated

employm

ent effect,

(A.31) D

lit+1 =

Dlit +

dlit = D

lit + dl1

it + dl2

it .

Finally, we have to take into accou

nt the fact that an increase in emp

loyment also

raises outp

ut throu

gh the prod

uction fu

nction. Calling dY

2, this indu

ced effect,

which is given by

(A.32) dY

2it =

θl dlit ,

the total increase in output over the period is given by

(A.33) dY

it = dY

1it +

dY2

it .

Analogous expressions w

ill hold for the cum

ulative output gains (DY

and D

Y2).

47

Ap

pen

dix 2: T

he Fram

ework

's contrib

ution

to factor accum

ulation

One of the m

ain difficu

lties I have found

du

ring the preparation of this paper is thescarcity of clear and

detailed

information of the com

position and

financing ofStru

ctural Fu

nd exp

enditu

res and on the "p

hysical" outp

ut of the hu

man resou

rceprogram

mes financed

by these Funds.

The m

ain source of the d

ata I have used

is a Provisional Financial P

lan (PFP

) for the1994-99 O

bjective 1 Framew

ork that was put together using the available inform

ationon the execu

tion of the CSF u

ntil 1997 and u

pdated

projections for the remaind

er ofthe p

rogramm

ing period

. This P

lan disaggregates C

SF expend

iture by Fu

nd and

byfu

nction

al category (head

ings an

d su

bhead

ings) an

d p

rovides fairly d

etailedinform

ation on the sources of financing, d

istinguishing betw

een EU

grants, thecontribu

tions of the national and regional Sp

anish adm

inistrations and p

rivatecofinancing for certain p

rojects. The Fram

ework is d

ivided

into a Mu

ltiregionalSubfram

ework, w

hich includes those projects to be executed

by the Spanish nationalgovernm

ent, and a set of R

egional Framew

orks (one for each Objective 1 region) that

fall und

er the purview

of the regional adm

inistrations. The expend

iture inclu

ded

inthe M

ultiregional Subframew

ork is not disaggregated

by region in the PFP.

Using this inform

ation and som

e add

itional sources that w

ill be discu

ssed below

, Ih

ave estimated

the region

al allocation of C

SF expen

ditu

re and

its fun

ctional

breakdow

n in each region. This task can be d

ivided

into four p

arts. First, it was

necessary to elaborate a functional classification of expenditure that could

be used to

app

roximate the Fram

ework's contribu

tion to the stocks of prod

uctive inp

uts u

singthe available inform

ation on the comp

osition of comm

itments by head

ing andsu

bheading. Second

, I had to estim

ate the regional and fu

nctional breakdow

n of theM

ultiregional Su

bframew

ork. Third

, I had to constru

ct an estimate of the ou

tpu

t ofthe C

SF-financed hu

man resou

rces programm

es measu

red in m

an-years of training.A

nd fou

rth, it was necessary to m

ake a correction for the observed d

elay in theFram

ework's execu

tion. The rem

ainder of this A

pp

endix d

iscusses in d

etail theproced

ure followed

in each case.

48

1. Th

e fun

ctional com

position

of CS

F expen

ditu

re

The P

rovisional Financial Plan (P

FP) contains a breakd

own by fu

nctional categories(head

ings and su

bheadings in E

U term

inology) of CSF sp

ending com

mitm

ents forthe period

1994-99 measu

red in 1997 ecu

s. These d

ata are converted into m

illions of1990 p

esetas using the average p

eseta-ecu exchange rate for 1997 and

the Spanish

GD

P d

eflator. The figu

res obtained in this w

ay are divid

ed by the d

uration of the

planning period (in principle six years, from

1994 to 1999) to obtain annual averages.

Tab

le A2.1: P

lann

ed C

SF exp

end

iture

An

nu

al totals for all the O

bjective 1 region

s______________________________________________________________________ regional fram

eworks m

ultregional framew

. total CSF

functional heading:public exp.

private exppublic exp.

private exppublic exp.

private exp1. T

erritorial articulation60,360

236,068296,428

2. Develop. of productive fabric

36,44361,274

93,825121,060

130,268182,333

3. Tourism

11,54412,088

4,083621

15,62812,709

4. Agricult. and rural developm

ent57,427

3,1014,672

62,0993,101

5. Fishing118

29,08716,764

29,20516,764

6. Other infrastructure

46,737199,834

246,5707. H

uman resources

49,878128,539

178,4178. T

echnical assistance2,098

3,5845,681

total264,605

76,463699,692

138,445964,297

214,908______________________________________________________________________ - N

ote millions of 1990 ptas. per year betw

een 1994 and 1999.

The results of these calculations for the set of all O

bjective 1 regions are sum

marized

in Table A

2.1, which show

s average annual p

lanned C

SF expend

iture in m

illions of1990 p

esetas, disaggregated

by functional head

ing and by sou

rce of the fund

s. Inparticular, I d

istinguish between public expend

iture, which is the sum

of grants fromthe E

U and

spend

ing by Spanish p

ublic ad

ministrations, and

private exp

enditu

re,w

hich correspond

s to the private co-financing for som

e of the projects inclu

ded

inthe Fram

ework. T

he table also shows the breakd

own of total exp

enditu

re between

the Multiregional Subfram

ework and

the sum of the R

egional Subframew

orks. 28

Using the available inform

ation on the breakdow

n of comm

itments by head

ing andsubhead

ing, I have classified the bulk of planned

CSF expend

iture into the five largeitem

s or

programm

es discu

ssed in

the text: p

ublic in

vestmen

t in p

rodu

ctiveinfrastru

ctures (infraest), p

ublic investm

ent other types of p

hysical capital (pubinv),

subsidies to the private sector (subs), public expend

iture in training and ed

ucation

28 I exclud

e expenditure in the N

orth-African autonom

ous cities of Ceuta and

Melilla.

49

Tab

le A2.2: C

orrespon

den

ce betw

een fu

nction

al sub

head

ings

and

expen

ditu

re program

mes

______________________________________________________________________

a. Investm

ent in

prod

uctive in

frastructu

res =

transport infrastructures (subheadings 1.1-1.6, road

s, railroads, ports, airports, channels and

other transport infrastructures)

+ w

ater works (subhead

ing 6.1)+

environmental protection and

improvem

ent (6.3)+

Cohesion Fund

(*)

b. T

rainin

g expen

ditu

re=

strengthening of technical and professional ed

ucation (7.2)+

ongoing worker training (7.3)

+ 74%

of expenditure on em

ployability (helping the unemployed

gain or regain employm

ent) (7.4**)+

50% of expend

iture on the labour market integration of persons w

ith special dificulties (7.5**)

+ specific training need

s in R&

D (6.4)


s (2.4) in relation to heading 2, w

hich includes aid

to various industries and

local d

evelopment)


s in tourism (3.1b)


s in agriculture and fishing (approxim

ated by Social Fund

expenditure

included

in headings 4 and

5).

c. Pu

blic in

vestmen

t in oth

er ph

ysical capital (exclu

din

g prod

uctive in

frastructu

res)=

telecomm

unications investment (1.7)

+ cultural resources of touristic interest (3.2)

+ energy (6.2)

+ aid

to R&

D (6.4.a) (***)

+ health-related

infrastructures (6.5)+

information society (6.6)

+ ed

ucational infrastructures (7.1)

d. S

ub

sidies to th

e private sector =

public expenditure on

subsidies to food

processing and other ind

ustries and to the crafts (2.1a and

2.1b)+

local developm

ent and services (2.2)

+ ind

ustrial zones (2.3)+

subsidies to investm

ent in tourism (3.1a)

+ agriculture and

rural developm

ent (heading 4, except for Social Fund

expenditure)

+ fishing (head

ing 5, except for Social Fund expend

iture)

e. Private co-fin

ancin

g of investm

ent =

expected private expend

iture insubsid

ies to food processing and

other industries and

to the crafts (2.1a and 2.1b)

+ local d

evelopment and

services (2.2)+

subsidies to investm

ent in tourism (3.1a)

+ agricu

lture an

d ru

ral develop

men

t (head

ing 4, excep

t for the cofin

ancin

g of Social Fun

dexpend

iture)+

fishing (heading 5, except for the cofinancing of Social Fund

expenditure)

______________________________________________________________________ N

otes:(*) T

he Cohesion Fu

nd finances investm

ent projects inclu

ded

in headings 1 and

6, but I cou

ld not find

abreakd

own of this expend

iture.(**) Su

bheadings 7.4 and

7.5 finance both training courses and

emp

loyment su

bsidies. T

he share of trainingexpend

iture in these su

bheadings I u

se correspond to A

ndalu

cia and have been su

pplied by the E

conomics and

Finance Dep

artment of the regional governm

ent. For lack of other data, I have u

sed these coefficients for all the

regions in the sample.

(***) R&

D grants are includ

ed in group c (rather than d

) because most of these fund

s go to universities.

50

(trainin

g), and

the p

rivate co-finan

cing of in

vestmen

t projects su

bsidized

byC

omm

unity fu

nds (private). In ad

dition to these five item

s, the Framew

ork alsofin

ances som

e emp

loymen

t subsid

ies and

techn

ical assistance an

d evalu

ationp

rogramm

es. I have exclud

ed these exp

enditu

res from the analysis becau

se they do

not correspond to the inputs of the regional prod

uction function. 29

Table A

2.2 shows the corresp

ondence betw

een the classification of expend

iture into

subhead

ings and the five exp

enditu

re program

mes. T

able A2.3 su

mm

arizes thefunctional com

position of the different Subfram

eworks.

Tab

le A2.3: Fu

nction

al comp

osition of p

lann

ed C

SF exp

end

iture

(total for all the O

bjective 1 region

s)___________________________________________________________________

regional sub-fram

eworks

multiregional

subframew

orkC

SFtotal

a. productive infrastructures88,318

358,765447,083

b. public investment in non-infraest. capital

35,03381,767

116,800c. subsidies to private sectors

100,381122,876

223,257d. training

31,904107,988

139,892total public expenditure

255,636671,396

927,032

e. private co-financing75,892

138,445214,337

total private and public expenditure331,528

809,8411,141,369

___________________________________________________________________ - N

ote: millions of 1990 ptas. per year betw

een 1994 and 1999.

2. Th

e regional allocation

of the M

ultiregion

al Su

bfram

ework

To

estimate

the

regional

and

fu

nction

al allocation

of

the

Mu

ltiregional

Subfram

ework, I have proceed

in two steps. First, I estim

ated the d

istribution across

regions of each of the Eu

ropean Fu

nds. T

hen, I tried to ap

proxim

ate the functional

distribution of expend

iture within each region.

For the first calculation, I have u

sed a nu

mber of sou

rces that provid

e a breakdow

nby region (or enou

gh information to ap

proxim

ate it) of the total public expend

iture

chan

neled

by

each

of th

e E

urop

ean

Fun

ds

inclu

ded

in

th

e M

ultirregion

alSu

bfram

ewo

rk30. I calcu

late the share of each region in the relevant total andm

ultip

ly this coefficien

t by the total com

mitm

ents of each

Fun

d w

ithin

the

29 T

hat's why the totals of T

ables A2.1 and

A2.3 are d

ifferent.30 T

he relevant Funds are the R

egional Developm

ent and Social Fund

s (ER

DF and

ESF), the G

uidance

section of the Agricultural Fund

(EA

GG

F), the Fisheries Instrument (FIFG

) and the C

ohesion Fund.

51

Mu

ltirregional Subfram

ework to estim

ate its total spend

ing in each region. The

regionalization of private expenditure is d

iscussed below

.

Tab

le A2.4: R

egional sh

ares in P

lurirregion

al Su

bfram

ework

expen

ditu

re___________________________________________________________________

ER

DF

ESF

EA

GG

FF

IFG

Cohesion F

undA

ndalucía25.54%

25.18%19.60%

15.78%23.68%

Asturias

7.06%8.02%

4.67%6.85%

6.69%C

anarias5.43%

4.12%4.54%

5.67%6.22%

Cantabria

4.57%3.58%

3.04%8.27%

0.83%C

ast. y León

17.00%15.86%

19.91%1.22%

12.44%C

ast. la Mancha

7.53%9.94%

15.45%0.07%

7.77%V

alencia10.10%

10.26%6.19%

9.37%15.21%

Extrem

adura4.38%

8.51%8.62%

0.18%1.92%

Galicia

12.85%8.02%

15.11%50.14%

21.01%M

urcia5.54%

6.51%2.87%

2.45%4.24%

Fuente:

CE

S Gal

MT

yAS

Marcos R

egsC

ES G

alN

avarro et al___________________________________________________________________

Notes and sources:

- ER

DF:

share of each Objective 1 region in total com

mitm

ents for 1994-97 according to the

Multiregional Subfram

ework for the O

bjective 1 regions. Data from

CE

S Galicia (1999).

- ESF: share of each region in total E

SF planned expend

iture included

in the Multiregional O

bjective 1Su

bframew

ork calculated

using d

ata on disbu

rsed expend

iture for 1994-98 and

expected expend

iture

in 1999. This inform

ation was supplied

by the Ad

ministrative U

nit for the ESF of the Spanish M

inistryof L

abour and Social A

ffairs.- E

AG

GF-G

uidance section: I use the w

eight of each region in the total planned

expend

iture for this

Fund includ

ed in the R

egional Subframew

orks according to the PFP.

- FIFG: share of each region in regionalized

subsid

ies for 1994-97. Part of the exp

enditu

re is notregionalized

. This item

correspond

s to the first year of the program

me. I im

plicitly assu

me that this

amount w

as distributed

in the same w

ay as the remaining expend

iture. Data from

CE

S Galicia (1999).

- Cohesion Fund: D

ata from N

avarro et al (2000), who in tu

rn take if from the Sp

anish Ministry of

Econom

ics and Finance. I u

se the share of each region in total Cohesion Fu

nd grants to O

bjective 1regions d

uring 1994-99. T

he entire Cohesion Fu

nd is inclu

ded

in the Mu

ltiregional Subfram

ework

according to the PFP.

Table A

2.4 shows the regional shares I have u

sed and

their sources. It shou

ld be

noted that in som

e cases these coefficients have been obtained u

sing information for

the period 1994-97 rather than for the entire program

ming period

. Du

e to the lack ofother inform

ation, in the case of the Guid

ance section of the Agricultural Fund

I haveassu

med

that the Mu

ltiregional Framew

ork is distribu

ted across regions in the sam

ew

ay as the Regional Fram

ework (for w

hich the PFP

does p

rovide a regional

breakdow

n).

52

Tab

le A2.5: Fu

nction

al comp

osition of p

ub

lic expen

ditu

re by d

ifferent E

urop

eanFu

nd

s inclu

ded

in th

e Plu

rirregional Fram

ework

______________________________________________________________________E

RD

FE

SFE

AG

GF

FIF

GC

ohesionF

unda. productive infrastructures

57.86%100.00%

b. public investment in other capital

23.93%c. subsidies to the private sector

16.35%100.00%

100.00%d. training

1.86%100.00%

______________________________________________________________________- Source: PFP, M

ultiregional Objective 1 Fram

ework, 1994-99.

For the second calcu

lation, I have had to assu

me that the fu

nctional comp

osition ofexp

enditu

re is the same across regions for any given Fu

nd. T

he weights of the

different p

rogramm

es in the Mu

ltiregional Framew

ork are obtained from

the PFP

and are show

n in Table A

2.5.

At this p

oint, we have a regional and

functional d

isaggregation of the pu

blicexp

enditu

re financed by the M

ultiregional Fram

ework that can be ad

ded

to thecorrespond

ing figures for the Regional Fram

eworks, w

hich are directly available.

Tu

rnin

g to private exp

end

iture, th

e situation

is similar. W

hile th

e Region

alFram

eworks contain regionally d

isaggregated d

ata, the Mu

ltiregional Framew

orkonly gives a total that m

ust be allocated

among the d

ifferent territories. To d

o this, Icalcu

late the ratio between the am

ount of p

rivate cofinancing (line e in Table A

2.3)and

the total volum

e of subsid

ies to enterprises (line c in the sam

e table) using

aggregate data for the M

ultiregional Su

bframew

ork. This ratio (w

hich is equal to

1.127) is then mu

ltiplied

by the estimated

volum

e of subsid

ies in each region und

erthe M

ultiregional Subframew

ork to obtain the desired

estimate.

Table A

2.6 (which com

es at the end of the p

aper) su

mm

arizes the results of the

calculations described

in this section.

3. Th

e outp

ut of h

um

an resou

rces program

mes

Most of the exp

enditu

re items w

e have estimated

in the previou

s sections financeinvestm

ent in infrastructures and other types of physical capital and

can therefore beused

directly in our im

pact calculations because they are measured

in the same units

as the corresponding factor stocks that appear in the prod

uction function. In the caseof ed

ucational and

training program

mes, how

ever, it is necessary to "translate"expend

iture figu

res into physical units that w

ill be at least roughly com

parable with

53

our p

roxy for the stock of hum

an capital. H

ence, I have calculated

the CSF's

contribution to the educational stock m

easured in years of training by com

bining theexpend

iture data given in T

able A2.6 w

ith an estimate of the average cost of a year of

training in various types of human resources program

mes.

The unit cost estim

ate is based on tw

o intermed

iate evaluation reports for the human

resources p

rogramm

es includ

ed in the R

egional Subfram

eworks for A

ndalu

cía andG

alicia. These rep

orts contain information on the nu

mber of beneficiaries of the

relevant training program

mes, the average nu

mber of hou

rs of training received by

them and

the total cost of each program

me. T

he information is d

isaggregated by

types of p

rogramm

es, distin

gush

ing betw

een su

pp

ort for formal vocation

aled

ucation

, the train

ing of research

ers, and

ongoin

g trainin

g program

mes for

unem

ployed

and em

ployed

workers (w

ith a partial sectoral breakd

own for the last

group

in one of the regions). Table A

2.7 shows the average u

nit cost of eachprogram

me (in m

illions of 1990 pesetas per year of training) that have been obtainedu

sing the data in these reports. For these calcu

lations, I have assum

ed that a year of

training is comp

rised of forty 30-hou

r weeks, excep

t for the case of researchertraining, w

here to each beneficiary (presumably a grad

uate student) w

e attribute oneyear of training.

Tab

le A2.7: A

verage un

it costs of trainin

g______________________________________________________

Andalucía

Galicia

averagesupport to form

al vocational training0.193

0.2330.213

voc. tr. in agriculturena

0.7650.765

voc. tr. in fishingna

0.7540.754

training of researchers1.026

1.0071.017

training of employed w

orkers0.454

0.6450.549

training of unemployed w

orkers0.755

0.6650.710

______________________________________________________ - N

ote: millions of 1990 pesetas per year of training; n.a. =

not available.

The u

nit costs shown in the table are com

bined w

ith my p

revious estim

ates of therelevant expend

iture to approximate the num

ber of years of training financed by the

Framew

ork in each region. For each region, I divid

e total expenditu

re in each of therelevant su

bheadings by the average u

nit cost (last colum

n of Table A

2.7) of thetraining activity that seem

s to correspond m

ost closely to the subheading. T

able A2.8

shows the correspond

ence between the expend

iture breakdow

n by subheadings and

54

the classification of training activities used

in Table A

2.7, as well as the u

nit costattributed

to each subheadings (in m

illions of 1990 pesetas per year of training):

Tab

le A2.8: C

orrespon

den

ce betw

een su

bh

eadin

gs and

the classification

oftrain

ing activities in

Tab

le A2.7, an

d u

nit costs assu

med

for each su

bh

eadin

g______________________________________________________________________

subheadings:classification in T

able A2.7

unit cost2.4 specific training need

s, heading 2


orkers0.5493781

3.1.B. specific training need

s, tourismtraining of em

ployed workers

0.54937814. agriculture and

rural developm

entvocational training in agriculture

0.764646935. fisheries

vocational training in fishing0.75428922

7.2. strengthening of technical and professional

education

support to formal voc. training

0.21308065

7.3. ongoing worker training


orkers0.5493781

7.4. employability


orkers0.70986097

7.5. labour market integration


orkers0.70986097

6.4.B. pecific training need

s, R&

Dtraining of researchers

1.01654633______________________________________________________________________

Table A

2.9 (enclosed at the end

of the paper) shows C

SF training expenditure broken

dow

n by subhead

ing and the estim

ated nu

mber of years of training financed

by theFram

ework in each region.

4. Ad

justm

ent for th

e delay in

the execu

tion of th

e CS

F

All the estim

ates presented in the previous sections of this A

ppendix refer to planned

expend

iture for the p

eriod 1994-99. A

ctual C

SF disbu

rsements can in p

ractice fallbelow

planned

comm

itments (if the Sp

anish adm

inistrations fail to present enou

ghaccep

table projects to fu

lly exhaust the available resou

rces) and m

ay be partially

executed

after the end of the p

rogramm

ing period

, as Structu

ral Fund

regulations

allow for d

elays of up two years in the execution of the paym

ents.

The inform

ation I have found

on the execution of the 1994-99 C

SF is rather lessd

etailed than the one p

rovided

by the PFP

(except in the case of E

RD

F) but it d

oessu

ggest that the resources assigned

to the CSF have been p

ractically exhausted

,although w

ith a certain delay. In the case of E

RD

F, for instance, the overall degree of

execution of the O

bjective 1 CSF w

as of 82.11% at the end

of 1999 and of 95.83%

in

55

Decem

ber 2000. 31 Althou

gh I do not have d

etailed inform

ation for all the relevantprogram

mes, the available d

ata suggest that a reasonable correction for the observedd

elay in the execution of the Fram

ework m

ay be to assum

e that the availableresources w

ere spent over a period of seven rather than six years (i.e. assum

e that theC

SF was com

pletely execu

ted bu

t with a d

elay of one year). Hence, the annu

alexpend

iture figu

res I have used

in the impact calcu

lations discu

ssed in the text w

ereobtained

by mu

ltiplying by 6/

7 the estimates d

iscussed

in the previou

s sections ofthis A

ppendix.

31 The available d

ata also suggests that the differences across regions in the d

egree of execution of theC

SF are not significant, with the p

ossible exception of E

xtremad

ura, w

hich seems to be lagging

somew

hat behind.

56

Referen

ces

Brad

ley, J., J. A. H

erce and L

. Mod

esto (1995). "The M

acroeconomic E

ffects of theC

SF 1994-99 in the EU

Perip

hery. An A

nalysis Based

on the HE

RM

IN M

odel."

Econom

ic Modelling 12(3), July.

Brad

ley, J., K. W

helan and J. W

right (1995). "HE

RM

IN Ireland

." Econom

ic Modelling

12(3), July.

Christod

oulakis, N

. and S. K

alyvitis (2000), “The E

ffects of the Second C

omm

unity

Sup

port Fram

ework 1994-99 on the G

reek Econom

y.”Journal of Policy M

odelling,22(5), Septem

ber.

Consello E

conómico e Social (C

ES) d

e Galicia (1999). E

stado de execución dos FondosE

structurais na Com

unidade Autónom

a de Galicia no ano 1997. C

olección Informes

1/99, Santiago d

e Com

postela.

Correa, M

. D. and

S. Malu

quer (1998). E

fectos regionales del presupuesto europeo enE

spaña (1986-1996). Institut d

'Estu

dis A

utonòm

ics, Generalitat d

e Catalu

nya,B

arcelona.

Dabán, T

. and A

. Lam

o (1999). "Convergence and

public investment allocation, Spain

1980-93." Docu

mento d

e Trabajo D

-99001, Dir. G

ral. de A

nálisis y Program

aciónPresupuestaria, M

inisterio de E

conomía y H

acienda, M

adrid

.

de la Fuente, A

. (1996)."Inversión pública y redistribución regional: el caso d

e España

en la décad

a de los ochenta."P

apeles de Econom

ía Española 67, pp. 238-56.

de la Fu

ente, A. (1997). "Fiscal p

olicy and grow

th in the OE

CD

." CE

PR

Discu

ssionPaper no. 1755.

de la Fuente, A

. (2002a). "Is the allocation of public capital across the Spanish regionstoo red

istributive?" CE

PR D

iscussion Paper no. 3138.

de la Fu

ente, A. (2002b). "O

n the sources of convergence: A

close look at the Spanishregions." E

uropean Econom

ic Review

46(3), 2002, pp. 569-99.

de la Fu

ente, A. (2002c). "Infrastru

ctures and

prod

uctivity: a su

rvey." Mim

eo,Instituto d

e Análisis E

conómico, B

arcelona.

de la Fu

ente, A. and

A. C

iccone (2002)."Hu

man cap

ital in a global and know

ledge-

based econom

y." Rep

ort for the Eu

ropean C

omm

ission, DG

for Em

ploym

ent andSocial A

ffairs.

de la Fu

ente, A. and

X. V

ives (1995)."Infrastructu

re and E

du

cation as Instrum

ents ofR

egional Policy: Evid

ence from Spain." E

conomic P

olicy, 20, April, pp. 11-54.

Evans, P

. and G

. Karras (1994). "A

re government activities p

rodu

ctive? Evid

encefrom

a panel of US states", R

eview of E

conomics and Statistics L

XX

VI(1), pp. 1-11.

57

Fund

ación BB

V (1998). E

l stock de capital en España y sus com

unidades autónomas.

Bilbao.

Fund

ación BB

V (variou

s years). Renta nacional de E

spaña y su distribución provincial.B

ilbao.

García-M

ilà, T., T

. McG

uire, and

R. P

orter (1996). "The effects of p

ublic cap

ital instate-level p

rodu

ction functions reconsid

ered." R

eview of E

conomics and Statistics

78(1), pp. 177-80.

Gon

zález-Páram

o, J. M. an

d I. A

rgimón

(1997). "Efectos d

e la inversión

eninfraestru

cturas sobre la p

rodu

ctividad

y la renta de las C

C.A

A.," in E

. Pérez

Tou

riño, ed

itor, Infraestructuras y desarrollo regional: Efectos económ

icos de laA

utopista del Atlántico. E

ditorial C

ivitas, Colección E

conomía, M

adrid

.

Gorostiaga, A

. (1999). "¿Cóm

o afectan el capital p

úblico y el cap

ital hum

ano alcrecim

iento? Un análisis p

ara las regiones españolas en el m

arco neoclásico.Investigaciones E

conómicas 23(1), pp. 95-114.

Herce J. A

., and S. Sosvilla-R

ivero (1995). "HE

RM

IN Spain." E

conomic M

odelling 12(3),July.

Holtz-E

akin, D. (1994). "P

ublic sector capital and

the produ

ctivity puzzle", R

eview of

Econom

ics and Statistics 76(1), pp. 12-21.

Junta d

e And

alucía (u

ndated

). Program

a Operativo de V

alorización de los Recursos

Hum

anos. Fondo Social Europeo, M

arco de Apoyo C

omunitario 1994-99. C

onsejería de

Trabajo y A

suntos Sociales, Sevilla.

Mas, M

., F. Pérez, E

. Uriel and

L. Serrano (1998). C

apital hu

mano, series históricas.

Unpublished

update.

Mas, M

., J. Maud

os, F. Pérez and

E. U

riel (1995). "Infrastructures and prod

uctivity inthe Spanish regions: a long-run perspective." M

imeo, IV

IE, V

alencia.

Ministerio d

e Haciend

a (varios años). La programación regional y sus instrum

entos.D

irección General d

e Fondos C

omunitarios y Financiación T

erritorial. Mad

rid.

Mod

esto L., and

P. Neves (1995). "H

ER

MIN

Portugal." Econom

ic Modelling 12(3), July.

Navarro, L

., A. C

lara Pastor, I. A

lonso y L. G

onzalez (2000?). Estudio financiero de los

Fondos Com

unitarios en Aragón en el período 1993-98. C

olección Tem

as Eu

ropeos 5,

Gobierno d

e Aragón, Z

aragoza.

Roeger, W

. (1996). "Macroecon

omic evalu

ation of th

e effects of Com

mu

nity

Structural Funds w

ith QU

EST

II." Mim

eo, European C

omm

ission, DG

II.

Xu

nta de G

alicia (und

ated). E

valuación Intermedia FSE. C

onselleria de E

conomía e

Facenda, Santiago d

e Com

postela.

an assessment of the 1994-99 Objective 1 CSF The effect of ... file1 The effect of Structural Fund spending on the Spanish regions: an assessment of the 1994-99 Objective 1 CSF Angel

Documents