No blessing, no curse? On the benets of being a resource-rich southern region of Italy

No blessing, no curse? On the benefits of being aresource-rich southern region of Italy

Research in Economics, forthcoming. DOI: 10.1016/j.rie.2015.03.003

Roberto Iacono

NTNU & HiST

Oxford, 22.08.2015

Roberto Iacono (Institute) No blessing, no curse? Oxford, 22.08.2015 1 / 32

Historical background

“Call your men back, let them return from wherever they migrated to, andtell them that finally there will be jobs for them, here.”E.Mattei, 27.10.1962.


The research question

Has intensive exploitation of oil fields and greater resource revenues inBasilicata, all else equal, led to a higher degree of regional economicdevelopment?

010

2030

4050

Barre

ls o

f oil

per c

apita

1980 1990 2000 2010Year

Source: UNMIG


Basilicata and the rest of Mezzogiorno


Contribution

1 Estimate the economic effect of oil in the region of Basilicata duringthe period 1980− 2009, compared to control regions.

Synthetic Control Method (SCM) (Abadie et al. 2014): weight controlregions to construct a counterfactual that replicates the treated regionbefore treatment.

2 Discussion about channels.

Control rights; Organized crime; Sectoral effects; Labor migration.

3 Results: null aggregate effects; significant sectoral effects.


Contribution







Contribution







Contribution







Contribution







Empirical Strategy

Identification strategy: exploit the fact that Basilicata producedthroughout the period of analysis a fraction close to unity of oilextracted in the 5+ 1 southern Italian regions.

.4.6

.81

Oil

extra

cted

(ton

s): B

asili

cata

/Tot

. DP

regi

ons

1980 1990 2000 2010Year

Source: UNMIG


Institutional agreement and royalties

Variable Location Net value based royaltiesOil production Onshore 7%

Offshore 4%Gas production Onshore 7%

Offshore 7%Revenue’s benefiter State (30%); Region (70%)

Law 140/1999: southern regions are entitled to 100% of royaltyrevenues. Q: still too low?


Empirical Literature

1 Sub-national economic effects of resource revenues:

Caselli and Michaels (2013); Borge et al. (2013); Kan et al. (2014);Papyrakis and Raveh (2014).

2 Comparative Case Studies using SCM:

Abadie et al. (2014); Pinotti (2012).

3 On the case of Basilicata:

Percoco (2012).








Percoco (2012).








Percoco (2012).








Percoco (2012).








Percoco (2012).








Percoco (2012).


Implementing the SCM: choosing the DP

1 Choosing the Donor Pool (DP).

Donor Pool, 5 southern Italian regions: Campania, Molise, Puglia,Sardegna, Calabria.Choice informed as well by a study of European regional economiesfrom the Bank of Italy (2012).

2 Generating weights for units in the DP.3 Estimating Impact on the Economy of Basilicata.




Donor Pool, 5 southern Italian regions: Campania, Molise, Puglia,Sardegna, Calabria.

Choice informed as well by a study of European regional economiesfrom the Bank of Italy (2012).











2 Generating weights for units in the DP.

3 Estimating Impact on the Economy of Basilicata.







Implementing the SCM: generating weights

SCM Algorithm: define GDP per capita as Y and its determinants X,such as:

Population, Labor Force, Gross Fixed Investment, Pop. shares byeducation level, Value Added shares of GDP by Industry.

Matching period (1980− 1998): Y pre0 and X0 in DP; Ypre1 and X1 in

treated unit.

Post-treatment (1999− 2009): Y post0 in DP; Y post1 in treated unit.

SC unit is given by the vector of weights W ∗ = (w1, ...,w5) (withw1 + ...+ w5 = 1) chosen as W that minimizes

k

∑m=1

vm(X1m − X0mW )2

in which vm are weights assigned to the m− th determinant.Roberto Iacono (Institute) No blessing, no curse? Oxford, 22.08.2015 10 / 32





treated unit.



k

∑m=1

vm(X1m − X0mW )2






treated unit.



k

∑m=1

vm(X1m − X0mW )2






treated unit.



k

∑m=1

vm(X1m − X0mW )2



Region Synthetic weights W ∗

Campania 0Molise .354Apulia .106Sardinia 0Calabria .54


Implementing the SCM: estimating impact

SCM Algorithm: generate synthetic Basilicata using assigned weights;compare actual and synthetic Basilicata.

Y1t −6∑j=2w ∗j Yjt

a) Real GDP per capita as dependent variable.

Matching period

050

0010

000

1500

020

000

GD

P p

er c

apita

, con

stan

t pric

es

1980 1990 2000 2010Year

Treated unit Synthetic control unit


Time-placebo tests - Treatment 1992 (left)

Real GDP per capita

Matching period

050

0010

000

1500

020

000

GD

P p

er c

apita

, co

nsta

nt p

rices

1980 1990 2000 2010Year


Matching period

050

0010

000

1500

020

000

GD

P p

er c

apita

, co

nsta

nt p

rices

1980 1990 2000 2010Year




b) Employment rate (total) as dependent variable.


Campania 0Molise .139Apulia .657Sardinia .204Calabria 0



b) Employment rate (total) as dependent variable.

Matching period36

3840

4244

46E

mpl

oym

ent r

ate,

tota

l

1980 1990 2000 2010Year




Employment rate

Matching period

3638

4042

4446

Em

ploy

men

t ra

te,

tota

l

1980 1990 2000 2010Year


Matching period

3638

4042

4446

Em

ploy

men

t ra

te,

tota

l1980 1990 2000 2010

Year




c) Gross Fixed Inv. as dependent variable.


Campania 0Molise .83Apulia 0Sardinia .009Calabria .161



c) Gross Fixed Inv. as dependent variable.

Matching period50

010

0015

0020

0025

0030

00G

ross

fixe

d in

vest

men

t, co

nsta

nt p

rices

1980 1990 2000 2010Year




Gross fixed investment

Matching period

500

1000

1500

2000

2500

3000

Gro

ss f

ixed

inve

stm

ent,

mill.

Eur

o

1980 1990 2000 2010Year


Matching period

500

1000

1500

2000

2500

3000

Gro

ss f

ixed

inve

stem

ent,

mill.

Eur

o1980 1990 2000 2010

Year



Robustness main results

Yi ,t = γi + δt + λTi ,t + X′i ,tβ+ εi ,t

with

Yi ,t = outcome of interest for region i , year t.

γi = region fixed effects.

δt = time fixed effects.

Ti ,t = dummy for the treated region in the post-treatment period.

X′i ,t = a set of covariates.

εi ,t = clustered error term.


Robustness main results

Table (1) (2) (3)GDP per capita, Employment Gross fixed inv.constant prices rate, total constant prices

Diff-in-diff -352.8* -0.646 -1,404**(179.1) (0.786) (651.9)

Fixed eff. YES YES YESObservations 180 180 180R-squared 0.995 0.529 0.827Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

Note: standard errors adjusted for clusters.


Discussion about channels

1 Control rights structure.2 The plague of organized crime.3 Sectoral effects.4 Labor migration.


Control rights structure

Brunnschweiler and Valente (2013): Int’l Partnership is linked tohigher GDP levels than Domestic/Foreign Control, regardless ofpolitical regime type.

Brunnschweiler and Valente (2013)’s coding of Italy: Foreign1930− 1956 and 1995− 2008, Partnership in between.

Q: would have Italy (and Basilicata)’s GDP benefited fromPartnership?


The plague of organized crime

Pinotti (2012): exposure to mafia activity (proxied by increase inmurders) after 1970s lowered GDP per capita by 16% in the treatedunit (Basilicata-Apulia), as compared to control group.

Q: can we rule out that public royalty revenues in Basilicatarepresented a profit opportunity for criminal organizations?


Sectoral effects

VAi ,t = γi + δt + λTi ,t + X′i ,tβ+ εi ,t

with

VAi ,t = value added (% of GDP) for sector (..) in region i , year t.

γi = region fixed effects.

δt = time fixed effects.

Ti ,t = dummy for the treated region in the post-treatment period.

X′i ,t = a set of covariates.

εi ,t = clustered error term.


Sectoral effects

Dependent variable Industry, % of GDP(1) (2) (3)

Diff-in-diff 2.886*** 4.730*** 5.306***(0.720) (0.620) (1.010)

Real GDP per capita 0.000246(0.000276)

Gross fixed inv. -0.000365***(7.83e-05)

Constant 15.60*** 18.51*** 18.56***(0.253) (0.610) (0.818)

Region fixed effects NO YES YESTime fixed effects NO YES YESObservations 180 180 180R-squared 0.044 0.560 0.662

Robust standard errors in parentheses.Asterisks denote significance levels: *** p<0.01, ** p<0.05, * p<0.1Roberto Iacono (Institute) No blessing, no curse? Oxford, 22.08.2015 26 / 32

Sectoral effects: SCM for industry

1416

1820

22S

hare

of

valu

e ad

ded,

Ind

ustr

y

1980 1990 2000 2010Year



Sectoral effects

Dependent variable Constructions, % of GDP(1) (2) (3)

Diff-in-diff -0.755* -1.577*** -0.524(0.428) (0.358) (0.329)

Real GDP per capita 0.000416***(0.000151)

Gross fixed inv. 0.000327***(4.29e-05)

Constant 8.397*** 12.67*** 11.38***(0.197) (0.352) (0.448)

Region fixed effects NO YES YESTime fixed effects NO YES YESObservations 180 180 180R-squared 0.005 0.883 0.919Asterisks denote significance levels: *** p<0.01, ** p<0.05, * p<0.1


Sectoral effects: SCM for constructions

68

1012

1416

Con

stru

ctio

ns, s

hare

of G

DP

1980 1990 2000 2010Year

T reated unit Synthetic control unit


Labor migration

5906

0161

2229

Bas

ilica

ta

1980 1990 2000 2010

1812

4557

1893

0968

Tot

al D

P re

gion

s

1980 1990 2000 2010Year

Source: ISTAT


Concluding remarks

Null hypothesis of aggregate positive economic effects: rejected.

Sectoral effects: positive for industry.

No blessing, no curse?


Concluding remarks





Concluding remarks





Thanks for attention!

Cite this article as: Roberto Iacono, No blessing, no curse? On thebenefits of being a resource-rich southern region of italy, Research inEconomics, http://dx.doi.org/10.1016/j.rie.2015.03.003


No blessing, no curse? On the benets of being a resource-rich southern region of Italy

Economy & Finance