Introduction to Modern Economic Growth
.
LUX
USA
L o g G D P p e r ca p ita , P P P , in 1 9 9 5
SGP
HKG QAT
AUS
10
ARE
ISR
BHSBHR KWT
BRB
MYS
SAU
MEX
OMN LBY
MUS
TTO
ITA
ESP
NZL
KORGRC
PRT
MLT
ARG
CHE
AUT BEL
FRA
NLD DNK
DEU
GBR
IRL
CZE
SVN
HUN
SVK
CHL
URY
ZAF
FJI
THAKNA
BRA
CRI
BWA
IRNTUN
VEN
PAN
DOM
DZA
PER LCA
PRY
GUY
NAM
SWZ
PHL
GRD
DMA
EGY
SLVGTM JAM
MAR CHN
JOR
BLZ
LKA
SYR
IDN
ECU
BTN
VCTHND
NIC
BOL
LSO
IRQ
PNG
ZWE
IND
PAK
VNM MRT
GHA
CMR
SEN
CIV
CAF
TGO
BGD NPL
CPV HTI
KEN
GIN
MMR
STP
UGA
GMB
SDN
LBR BEN BFATCD MOZ
COG
NGA
GNB
MWI NER
DJI
ZMBMLI MDG
YEM
AFG
AGO
COM
RWA TZA
ZAR BDI
ETH
SLE
SOM
GAB COL
8
JPN
NOR
CAN
ISL
SWEFIN
EST
POL
BLR LTU RUS
LVA
TUR
ROMKAZ
BGRHRV
GEO
TKM
UKR
AZE
MDA
ARM
UZB YUG
BIH
TJK
6
0
.2
.4
Latitude
.6
Figure 4.2. Relationship between latitude (distance of capital from
the equator) and income per capita in 1995.
depicted in Figure 4.1, and for that matter that shown in Figure 4.2, do not necessarily reflect causal relationships . As we pointed out in the context of the effect
of religion or social capital on economic performance, these types of scatterplots,
correlations, or their multidimensional version in ordinary least squares regressions,
cannot establish causality.
How can we overcome the challenge of establishing a causal relationship between
(economic) institutions and economic outcomes? The answer to this question is to
specify econometric approaches based on convincing identifying restrictions. This
can be done by using estimating structural econometric models or by using more
reduced-form approaches, based on instrumental-variables strategies. At the moment we do not know enough about the evolution of economic institutions and their
impact on economic outcomes to be able to specify and estimate fully-structural
econometric models. Thus as a first step, we can look at more reduced-form evidence
that might still be informative about the causal relationship between institutions
and economic growth. One way of doing so is to learn from history, in particular
180
.8