ECONOMIC FREEDOM, INSTITUTIONALQUALITY, AND CROSS-COUNTRY DIFFERENCES
IN INCOME AND GROWTHJames D. Gwartney, Randall G. Holcombe,
and Robert A. Lawson
In the past few decades, the issues in the literature on economicgrowth have broadened from the development of general theories ofgrowth, largely based on Solow (1956), toward an examination of whythere are differences in growth rates across countries, and why somecountries continue to grow while others stagnate at low levels ofincome. This study takes an institutional approach and uses a recentlydeveloped measure of institutional quality, the Economic Freedom ofthe World (EFW) index (Gwartney and Lawson 2003) to examine theissue of cross-country differences in income levels and growth rates.The emphasis on the importance of institutions to economic prosper-ity goes back at least to Adam Smith (1776), and has been found in themore recent work of Olson (1982), Scully (1988), North (1990), Barro(1996), Barro and Sala-i-Martin (1995), Landes (1998), Hall andJones (1999), and Acemoglu, Johnson, and Robinson (2001). Despitethis interest in institutions, much work on economic growth treatsinstitutions peripherally if at all.
One challenge to the institutional approach is to find a way toquantify the quality of institutions. The EFW index used here is ameasure of institutional quality and, to the extent that higher EFWratings lead to more rapid growth and higher income levels, it pro-vides insight into the characteristics of an environment conducive toprosperity. The results show that better institutions lead to higher
Cato Journal, Vol. 24, No. 3 (Fall 2004). Copyright Cato Institute. All rightsreserved.
James Gwartney is Professor of Economics and Gus A. Stavros Eminent Scholar Chair atFlorida State University. Randall Holcombe is DeVoe L. Moore Professor of Economics atFlorida State University. Robert Lawson is Professor of Economics and George H. Moor Chairat Capital University.
205
income, and that institutional improvements result in higher rates ofeconomic growth.
Three Explanations of Cross-Country Differences inEconomic Performance
Over the past decade, the economics literature has offered threedifferent types of explanations for the differences in income levelsand growth rates across countries. The most well-established expla-nation in the literature takes a production function approach based onthe work of Solow (1956). The second approach explains differencesin income and growth across countries as a function of institutions, andis represented by the work of North (1990) and Landes (1998). A thirdtype of explanation, promoted by Sachs (2003), points to the effects ofgeography and location as determinants of growth and income.
The production function approach views output (Q) as a function ofcapital (K) and labor (L), such that Q = f(K,L). Within this frame-work, output is increased by increasing the amount of inputs (K andL), and by technological improvements that alter the productionfunction so that more output can be produced with the same amountof inputs. This approach focuses on increasing human and physicalcapital, and on technological progress through, for example, researchand development. This explanation suggests that higher growth ratescan be generated by increasing inputs into the production function,and by discovering ways to employ those inputs more productively.
The institutional approach to growth is based on the idea that boththe availability and productivity of resources will be influenced by theinstitutional and policy environment. While there is some debateabout the exact characteristics of the institutions that are most ap-propriate for economic growth and prosperity, there is considerableagreement that secure property rights are crucial, and that the im-pediments to exchange must be minimal. Institutions and policies arereflective of government actions. To promote economic growth, govern-ments must not only follow actions that are supportive of secure propertyrights and freedom of exchange, they must also make a convincing andcredible commitment that the policies will be maintained in the future.Public policy must be designed to implement what Mancur Olson (2000)has referred to as market-augmenting government.1
A third approach to identifying factors that lead to prosperity looksat geographical factors. During the last several years, Jeffrey Sachshas promoted the idea that geography and location are major deter-
1Olsons ideas are further developed in Azfar and Cadwell (2003) and Knack (2003).
CATO JOURNAL
206
minants of cross-country differences in income levels and growth.Sachs has stressed the importance of three major geographic-locational factors: a tropical climate, access to an ocean port, anddistance of country from the worlds major trading centers (Rotter-dam, New York, and Tokyo). According to this view, a tropical climateinhibits economic growth because of the increased threat posed bydiseases such as malaria, and because of the negative impact of a hotand humid climate on the energy level and productivity of labor. Thelack of access to an ocean port will mean higher transactions costs andless trade with a sizable portion of the worlds population. A distantlocation from the major markets of the world will also retard trade. Inturn, less trade will reduce the gains from division of labor, specializa-tion, and economies of scale. Furthermore, each of these geographic-locational factors will reduce the attractiveness of a country as a base forproduction, and thereby retard its ability to attract investment.
These three alternative theories of cross-country differences in in-come levels and growth are not necessarily inconsistent with eachother, and may even be mutually reinforcing. For example, if theinstitutional and geographic-locational factors influence capital for-mation and the productivity of capital and other inputs, this has im-plications for the production function approach to growth. However,the policy implications of the three models have substantial differ-ences. The production function approach naturally focuses on policiesthat will increase the quantity and improve the productivity of capitaland labor. The institutional approach focuses on economic, legal, andpolitical institutions, reasoning that if appropriate institutions are inplace, the market system provides an incentive for market partici-pants to invest in human and physical capital, and to improve theirmethods of production through innovation. The geographic-locationalapproach suggests that greater attention should be paid to the controlof tropical diseases and an analysis of how technology can be appliedto affect the productivity of resources in tropical regions.
Measurement of Cross-Country Differences inInstitutional Quality
Institutional quality will be measured using the Economic Free-dom of the World index published in Gwartney and Lawson (2003).2
2Strictly speaking, the EFW index measures both longer term institutional variables such asthe quality of the legal system and shorter term public policies such as marginal tax rates.Throughout this study we used the term institutional quality to refer to both.
ECONOMIC FREEDOM
207
The EFW index has been used in a number of previous studies, anda review of both the index and other studies in which it has been usedis found in Berggren (2003). The EFW index measures institutionalquality in five major areas: (1) size of government, (2) legal structureand security of property rights, (3) access to sound money, (4) ex-change with foreigners, and (5) regulation of capital, labor, and busi-ness. The index provides current ratings for 123 countries, but dataare available for only about 100 countries continuously (at five-yearintervals) throughout 19802000. These countries make up the dataset for the empirical analysis that follows.
The EFW index reflects the key elements of the new institutionaleconomics. For many years, Douglass C. North (1990), FriedrichHayek (1945, 1960), Peter Bauer (1957, 1972), Hernando de Soto(1989), Gerald Scully (1988, 1992), and Scully and Slottje 1991) havestressed the importance of institutions and related policy variables.Following this same path, the new growth theory argues that soundinstitutions and policies are the keys to economic progress (e.g., seeTorstensson 1994; Knack and Keefer 1995; Barro 1995, 1996; Olson2000; Knack 2003; and Azfar and Cadwell 2003). The EFW indexis also closely related to what Hall and Jones (1999) call social in-frastructure. Using the language of Hall and Jones, a quality infra-structure is present when the institutions and government policies ofa country encourage productive behavior (e.g., accumulation of skillsor the development of new goods and production techniques) anddiscourage predatory activities (e.g., rent seeking, corruption, andtheft.)
The EFW measure is available for a large number of countries overa lengthy period of time. This is a major advantage because it allowsthe study of how changes in institutional quality affect economicgrowth.3 While there are advantages to using proxies for institutionalquality, such as is done by Hall and Jones (1999) and Acemoglu,Johnson, and Robinson (2001), that approach precludes looking at theeffect of decade-by-decade changes in institutional quality. In con-trast, the approach employed here makes it possible to investigatedirectly the impact of changes in institutional quality on economicperformance.
3Hanson (2003) criticizes the empirical literature using various economic freedom indexes,arguing that different indexes measure different things even though they come up withsimilar results. Because this study uses only the EFW index, Hansons criticisms of thebroader literature are only peripherally related to the work undertaken here.
CATO JOURNAL
208
Measurement of Geographic and Locational FactorsJeffery Sachs has popularized the view that a countrys level of
economic activity will be adversely affected by a tropical climate anda location that is distant from the worlds major market centers whileaccess to an ocean coastline will exert a favorable impact. We measurethese factors in the same manner as Sachs and his fellow researchers.The proportion of a countrys geographic area located in a tropicalregion (Tropics) will be used to measure the tropical location vari-able.4 The distance from core markets (Air Distance) variable is theminimum air distance (in kilometers) of a country from any one of theworlds major trading centers (Rotterdam, New York, or Tokyo). Fi-nally, the coastal variable is a percentage of a countrys populationliving within 100 kilometers of an ocean coastline.5
Measurement of Physical and Human CapitalThe data for physical capital per worker (Kpw) and human capital
per worker (Hpw) are from Baier, Dwyer, and Tamura (2003). Thephysical capital stock was derived from annual investment data in theusual manner. A 7 percent depreciation rate was used to convert theannual investment data into capital stock estimates. The human capi-tal estimates reflect cross-country differences in both years of school-ing and demographic (age) factors that can be expected to influencethe years of work experience. The years of schooling were also ad-justed for differences in returns across schooling categories (elemen-tary, secondary, and higher education). We believe that these data arethe most comprehensive cross-country human capital estimates cur-rently available. The physical and human capital data are available for91 of the 99 countries in our core data set. All of the eight omittedcountries have a population of less than one million.
Cross-Country Differences in Income Levels andGrowth Rates: Empirical Results
Because the economic performance of an economy will generallyreflect the quality of its institutional arrangements and policies over asubstantial time period, empirical analysis should also employ a mea-sure that reflects institutional quality over a substantial time period.The core database for this study comprises 99 countries for which
4Tropical regions are defined as areas located between 23.5 degrees of latitude North(Tropic of Cancer) and 23.5 degrees of latitude South (Tropic of Capricorn).5See Gallup, Sachs, and Mellinger (1999) for additional details on these data.
ECONOMIC FREEDOM
209
the Economic Freedom of the World data are available in 1980, 1985,1990, 1995, and 2000.6 The EFW rating used throughout this paperis the mean summary rating for these five years during the 19802000period, which reflects the quality of a nations institutions and policiesover a period of two decades.
The Determinants of Cross-Country Differences inper Capita GDP
Earlier, the study identified three different explanations in therecent economics literature for cross-country differences in economicperformance. Table 1 looks at each of these explanations separately tosee how much of the cross-country differences in per capita GDP inthe year 2000 each can explain. Because the level of income is thedependent variable, this analysis will reflect the cumulative long-runincome-enhancing effects of the independent variables. The first re-gression takes the EFW index as the sole measure of institutionalquality, and using that variable alone finds that differences in insti-tutions explain 63.2 percent of the cross-country variation in percapita GDP. The square of EFW rather than the linear form is usedhere because it gives a slightly better fit. This reflects the fact that aone-unit increase in EFW has a larger impact on per capita GDP forcountries with higher EFW ratings than for those with lower ratings.7
The three key variables suggested by Sachs are incorporated into thegeographic-locational model of Equation 2. This model explainsslightly more than half of the variation in per capita income. Theproduction function approach, using measures of human and physicalcapital in the third regression, explains 92.8 percent of the cross-country variation in income.
Clearly, each of the three explanations of cross-country differencesin per capita GDP has considerable explanatory power. The t-statistics on the coefficients show that all of the independent variablesin all three regressions are statistically significant as well, providingempirical support for all three explanations. The R-squares indicate
6These data were available for 103 countries. Because their per capita GDP figures andgrowth rates were dominated by conditions in the world market for crude oil, four of thecountries (Bahrain, Kuwait, Oman, and the United Arab Emirates) were omitted from thecore database.7The $651 coefficient for the square of EFW in Equation 1 indicates, for example, that atypical country with a long-term EFW rating of 6.0 had an income level in 2000 that was$7,161 ($651 times 11, the difference between the square of 6 and the square of 5) higherthan one with a long-term EFW rating of 5.0.
CATO JOURNAL
210
that the production function approach explains the greatest percent-age of the variation in incomes across countries, but while this showsthe importance of human and physical capital to the generation ofincome, it does not explain why the stock of human and physicalcapital varies across countries.
The regressions in Table 2 examine the impact of institutions andgeography on the 1999 levels of physical and human capital. The firstthree regressions use the stock of physical capital as the dependentvariable while the stock of human capital is the dependent variablein the last three equations. Because both of the dependent variablesare stock measures, the coefficients for the independent variableswill reflect their estimated cumulative effects over lengthy time pe-riods. Regressions 1 and 4 show that the 19802000 mean EFWrating by itself explains a substantial amount of the variation acrosscountries in the levels of both physical and human capital. Regression1 shows that a one-unit increase in the square of the EFW rating is
TABLE 1THE DETERMINANTS OF CROSS-COUNTRY DIFFERENCES IN
GDP PER CAPITA: THREE ALTERNATIVE MODELS
Dependent Variable: GDP Per Capita, 2000(t-ratio in parentheses)
Independent Variables
I-PModel
(1)
G-LModel
(2)
P-FModel
(3)
EFW Rating2, 651.0019802000 (13.00)
Tropics 10,590.00(9.33)
Coastal 4,554.00(2.20)
Air Distance (1,000s km.) 1.22(3.51)
Kpw, 1999 0.25(15.93)
Hpw, 1999 954.00(4.27)
Intercept 11,183.00 18,831.00 3,900.00
R2 (adjusted) 63.20 50.80 92.80Number of Countries 99 99 91
ECONOMIC FREEDOM
211
TA
BL
E2
EC
ON
OM
ICF
RE
ED
OM
, GE
OG
RA
PHY,
AN
DT
HE
STO
CK
OF
PHYS
ICA
LA
ND
HU
MA
NC
API
TA
LPE
RW
OR
KE
R
Dep
ende
ntV
aria
ble:
Kpw
,199
9(t
-rat
ioin
pare
nthe
ses)
Dep
ende
ntV
aria
ble:
Hpw
,199
9(t
-rat
ioin
pare
nthe
ses)
Inde
pend
ent
Var
iabl
es(1
)(2
)(3
)(4
)(5
)(6
)
EF
WR
atin
g2,1
980
2000
1,89
7.00
1,61
0.00
1,52
7.00
0.11
0.09
0.07
(13.
41)
(12.
52)
(11.
11)
(8.7
5)(7
.38)
(6.0
0)T
ropi
cs2
1,50
2.00
22,
203.
001
.90
2.0
2(6
.06)
(6.2
7)(5
.82)
(6.4
7)C
oast
al7,
275.
001.
27(1
.62)
(3.2
0)In
terc
ept
32,
704.
001
1,54
3.00
12,
243.
002.
664.
524.
40
R2
(adj
uste
d)66
.50
76.1
076
.50
45.7
060
.30
64.1
0N
umbe
rof
Cou
ntri
es91
9191
9191
91N
OT
E:W
hen
Air
Dis
tanc
ew
asad
ded
toE
quat
ions
(3)
and
(6),
itex
erte
dan
insi
gnifi
cant
impa
cton
the
depe
nden
tva
riab
le.
CATO JOURNAL
212
associated with an increase of $1,897 in the 1999 stock of physicalcapital per worker, and regression 4 shows that a one-unit increase inthe EFW rating increases the stock of human capital per worker by0.112 years.8
Equations 2 and 3 add the tropical location and share of populationnear a coastline variables to the physical capital model, while Equa-tions 5 and 6 add them to the human capital model. The addition ofthese variables reduces the size of the EFW coefficient somewhat,but it remains sizable and statistically significant. The tropical locationvariable is negative and significant, and it adds substantially to theexplanatory power. This is consistent with the view articulated bySachs that a tropical location adversely affects capital formation. Thecoastal variable is positive and significant in the human capital equa-tion, but insignificant in the physical capital equation. The size of thet-ratio for the coastal variable and its additional contribution to R2
indicate that it is substantially less potent than the institutional envi-ronment (EFW) and tropical location as a determinant of both physi-cal and human capital. The distance from the major markets variablewas omitted from Table 2 because it was insignificant in both thephysical and human capital stock regressions. The findings of Table 2indicate that the institutional (EFW) and tropical variables are im-portant determinants of cross-country differences in the stock of bothphysical and human capital. The impact of the coastal variable issmaller, particularly as a determinant of the physical capital stock.
The impact of institutional factors on the levels of physical andhuman capital across countries shows that the production functionapproach to economic growth leaves out an important factor if it doesnot account for institutional differences across countries. Levels ofphysical and human capital do have a substantial impact on a coun-trys income, but a countrys institutional quality has a major effect ona countrys level of human and physical capital. Better institutionsprovide a greater incentive for individuals to invest in their humanand physical resources. The regressions show that the tropical andcoastal locational variables also have an impact on the levels of physi-cal and human capital.
8These coefficients indicate that, measured in 1995 U.S. dollars, the capital stock perworker of countries with long-time EFW ratings of 6.0, for example, were more than$20,000 greater in 1999 than those with long-time EFW ratings of 5.0. Correspondingly,the EFW coefficient in the human capital equation indicates that countries with long-termratings of 6.0 had approximately 1.2 additional years of human capital than those withlong-term ratings of 5.0. Basically, the human capital variable is a countrys mean years ofschooling adjusted for its age composition and the diminishing returns associated withadditional schooling.
ECONOMIC FREEDOM
213
Table 3 looks at the impact of institutional and geographical factorsalong with levels of human and physical capital as determinants of percapita GDP. The first regression shows that institutional differencesalong with location in the tropics explain 75 percent of the cross-country differences in the 2000 per capita GDP. Equation 2 illus-trates that once the effects of EFW and tropical location are takeninto account, the coastal variable is insignificant and fails to add to theexplanatory power of the model. Equation 3 adds the levels of bothphysical and human capital per worker (Kpw and Hpw) to the model.The four independent variables of Equation 3 together explain about94 percent of the cross-country variation in per capita GDP.
Because the EFW rating and the tropics variable exert a majorimpact on the levels of both human and physical capital, as shown inTable 2, those variables have both a direct and an indirect impact onper capita GDP. The direct effect reflects their impact on per capitaGDP via the productivity of human and physical capital. The indirecteffect reflects their impact through the level of capital formationthe fact that the levels of human and physical capital per worker area function of the EFW rating and tropical location.
Because the levels of physical and human capital are included inregression 3, the indirect effect of those variables on per capita GDPis concealed. In order to measure the indirect effectsthrough capi-tal formationas well as the direct effects of EFW and tropicallocation on per capita GDP, only the portion of the human andphysical capital variables that is independent of EFW and tropicsshould be included in these measures. The residuals from Equations2 and 4 of Table 2 provide this information. The residuals measurethe amount of physical and human capital that are not explained byEFW and Tropics. When these residuals are substituted for the physi-cal and human capital variables, the coefficients for the EFW andtropical variables will register both their direct and indirect effects.Equation 4 of Table 3 presents these results.
Note how the inclusion of the indirect effects through Kpw andHpw substantially increase the coefficients and t-ratios of both theinstitutional quality and tropical variables. Once the indirect effectsare included, a one-unit increase in the square of EFW enhances percapita GDP by slightly more than $500. This implies, for example,that an increase in the mean 19802000 EFW rating from 5.0 (ap-proximately the levels of Argentina and Columbia) to 6.0 (approxi-mately the level of South Korea) enhances 2000 per capita GDP byabout $5,500. Similarly, an EFW increase from 6.0 to 7.0 enhances2000 per capita GDP by approximately $6,500 ($500 multiplied bythe square of 7 minus the square of 6). Equation 4 of Table 3 also
CATO JOURNAL
214
TA
BL
E3
EC
ON
OM
ICF
RE
ED
OM
, GE
OG
RA
PHY,
AN
DPH
YSIC
AL
AN
DH
UM
AN
CA
PIT
AL
AS
DE
TE
RM
INA
NT
SO
FG
DP
PER
CA
PIT
A
Dep
ende
ntV
aria
ble:
GD
PPe
rC
apita
,200
0(t
-rat
ioin
pare
nthe
ses)
Inde
pend
ent
Var
iabl
es(1
)(2
)(3
)(4
)(5
)(6
)
EF
WR
atin
g2,1
980
2000
529.
4251
6.73
107.
5550
8.94
109.
4148
5.86
(11.
91)
(10.
91)
(2.9
9)(2
3.75
)(3
.03)
(20.
96)
Tro
pics
8,4
72.0
08
,566
.00
2,0
61.0
07
,982
.00
1,9
27.0
08
,176
.00
(7.0
3)(7
.06)
(2.8
1)(1
3.50
)(2
.52)
(13.
68)
Coa
stal
1,17
0.00
516
.90
2,02
1.00
(0.7
9)(0
.64)
(2.6
6)K
pw,1
999
0.21
0.21
(10.
31)
(10.
28)
Hpw
,199
976
2.90
804.
50(3
.47)
(3.5
0)K
pw,1
999
(res
idua
ls)
0.21
0.21
(10.
31)
(10.
28)
Hpw
,199
9(r
esid
uals
)76
2.90
804.
50(3
.47)
(3.5
0)In
terc
ept
2,5
75.0
02
,733
.00
3,7
58.0
02
,709
.00
3,8
93.0
02
,904
.00
R2
(adj
uste
d)75
.10
75.3
093
.70
93.7
093
.60
93.6
0N
umbe
rof
Cou
ntri
es99
9991
9191
91N
OT
E:T
heva
lues
for
Kpw
(res
idua
ls)a
ndH
pw(r
esid
uals
)ofE
quat
ion
4w
ere
deri
ved
from
Tab
le2,
Equ
atio
ns2
and
5,re
spec
tivel
y.T
heK
pw(r
esid
uals
)an
dH
pw(r
esid
uals
)of
Equ
atio
n6
wer
ede
rive
dfr
omT
able
2,E
quat
ions
3an
d6,
resp
ectiv
ely.
ECONOMIC FREEDOM
215
highlights the importance of tropical location. The coefficient for thisvariable indicates that, other things constant, a tropical location ad-versely impacted 2000 per capita GDP by almost $8,000 once theindirect as well as the direct effects were taken into account.
Equation 5 of Table 3 adds the coastal variable to the model ofEquation 3. When only the direct effects are taken into account, thecoastal variable is insignificant. Equation 6 of Table 3 incorporates themethodology of Equation 4; the residuals from regressions 3 and 6 inTable 2 are substituted for the Kpw and Hpw variables, respectively.Thus, Equation 6 will capture the indirect, as well as the direct,effects of the institutional and geographic-locational variables. Whenthe indirect effects are taken into consideration, the coastal variableexerts a positive and significant impact on 2000 per capita GDP. Thecoefficients and t-statistics on the EFW and tropical variables areapproximately the same size in Equation 6 as in Equation 4.
Tables 1 through 3 show that institutional differences across coun-tries, as measured by differences in their EFW ratings, have a majorimpact on cross-country differences in income levels. This is espe-cially true when one considers the impact that institutional differ-ences have on the levels of physical and human capital across coun-tries.
The Determinants of Cross-Country Differences inthe Growth of GDP
Higher income levels are the result of higher past rates of growth.If there is a causal relationship between institutional quality (or anyother independent variable) and per capita GDP, differences ingrowth rates should also reflect this relationship. This section willexamine the importance of cross-country differences in the quality ofinstitutions, as measured by countries EFW ratings, and other vari-ables as determinants of differences in long-term growth rates amongcountries. In order to measure long-term growth more accurately andminimize the impact of business cycles and other factors that willtemporarily influence growth rates, the analysis will focus on differ-ential growth rates over the entire 1980 to 2000 time period.
Table 4 analyzes the separate contributions of institutions, geogra-phy, and physical and human capital as determinants of growth in percapita GDP during 19802000. The first regression shows that cross-country differences in the EFW rating explain 23.6 percent of thevariation in the annual rates of growth during the two decades, andthe coefficient on EFW has a t-statistic of 5.59. A one-unit change in
CATO JOURNAL
216
the EFW rating is associated with an increase in long-term annualgrowth of a little more than nine-tenths of a percent. The meangrowth rate of per capita GDP for the 99 countries of our basicdatabase was only 1.3 percent during 19802000, so a 0.9 percentagepoint increase in growth is a substantial impact.
Equations 2 and 3 of Table 4 examine the impact of the geo-graphic-locational and production function models as sources ofgrowth. As Equation 2 shows, the three variables of the geographicmodel explain approximately 22 percent of the cross-country variationin per capita growth. The tropical location and coastal populationvariables of the geographic model are statistically significant, but thedistance from major markets does not appear to have a significantimpact on a countrys growth rate. The growth rate of physical andhuman capital taken together explain slightly more than 42 percent ofthe cross-country variation in the growth of per capita real GDP, withthe growth rate of physical capital significant and the growth rate ofhuman capital insignificant at generally accepted confidence levels.
TABLE 4THE DETERMINANTS OF CROSS-COUNTRY DIFFERENCES IN
GROWTH OF GDP PER CAPITA: THREE ALTERNATIVE MODELS
Dependent Variable: Average Annual Growth Rate of GDP PerCapita, 19802000 (t-ratio in parentheses)
Independent Variables
I-PModel
(1)
G-LModel
(2)
P-FModel
(3)
EFW Rating, 19802000 0.94(5.59)
Tropics 1.78(3.97)
Coastal 1.48(2.91)
Air Distance (1000s km.) 0.02(0.19)
Growth of Kpw, 19801999 0.49(7.85)
Growth of Hpw, 19801999 0.34(1.42)
Intercept 3.99 1.40 0.15
R2 (adjusted) 23.60 21.90 42.20Number of Countries 99 99 91
ECONOMIC FREEDOM
217
As with the regressions in Table 1 that looked at determinants ofthe levels of income across countries, there is statistical evidencesupportive of each of the three major explanations for differences ineconomic growth across countries. Once again, the production func-tion explanation using physical and human capital as explanatory vari-ables produces the highest R2, but all three of the models have someexplanatory power.
As illustrated in Table 2, institutions and geography have an effecton the stock of both human and physical capital. They also exert animpact on the rate of capital formation. Using several alternativemeasures of capital formation, Table 5 addresses this issue. In Equa-tion 1 of Table 5, real annual investment per worker (measured in1995 U.S. dollars) during 19802000 is the dependent variable. As theequation indicates, EFW exerted a strong impact of $1,281 on aver-age annual rate of investment per worker during the two decades.This was true even after the effects of the initial (1980) per capitaincome level and the tropical and coastal variables were taken intoaccount. In contrast, neither the tropical nor coastal variables exerteda significant impact on cross-country differences in real investmentper worker.
The investment per worker figures of Equation 1 include bothprivate-sector and public-sector investment. Foreign direct invest-ment (FDI) per worker provides an alternative measure that will bealmost entirely reflective of private investment flows. Furthermore,the FDI figures will reflect the attractiveness of a countrys invest-ment climate to those residing outside of the country. As Equation 2illustrates, the EFW measure of institutional quality also exerted astrong impact on FDI per worker during 19802000. The impact ofthe other variables in Equation 2 was similar to that of Equation 1. Ahigher initial income level was associated with more foreign invest-ment per worker, but neither tropical location nor coastal populationshare exerted a significant impact on FDI.
In Equation 3 of Table 5, investment as a share of GDP (I/GDP)is the dependent variable. Once again, the EFW rating is positive andstatistically significant, indicating that the quality of a countrys insti-tutions exert a strong impact on the rate of investment. Even thoughcountries with a lower initial GDP invested a smaller dollar amountper worker (Equations 1 and 2), the negative sign on the 1980 percapita GDP variable in Equation 3 indicates these countries actuallyinvested a larger share of their GDP during the two subsequentdecades. A tropical location exerts a negative and significant impacton investment as a share of GDP, while a larger coastal popula-tion enhances the I/GDP ratio. Thus, Equation 3 indicates that both
CATO JOURNAL
218
TA
BL
E5
EC
ON
OM
ICF
RE
ED
OM
, GE
OG
RA
PHY,
AN
DL
OC
AT
ION
AS
DE
TE
RM
INA
NT
SO
FIN
VE
STM
EN
T
Dep
ende
ntV
aria
ble
(t-r
atio
inpa
rent
hese
s)
Inde
pend
ent
Var
iabl
es
Inve
stm
ent
Per
Wor
ker
(US$
),19
802
000
(1)
FD
IPe
rW
orke
r(U
S$),
1980
200
0(2
)
I/G
DP,
1980
200
0(3
)
Gro
wth
ofK
pw,1
980
1999 (4)
Gro
wth
ofH
pw,1
980
1999 (5)
EF
WR
atin
g,19
802
000
1,28
1.20
545.
702.
161.
240
.08
(4.1
2)(4
.00)
(3.0
9)(3
.76)
(0.7
5)G
DP
Per
Cap
ita,1
980
837.
5912
1.80
0.6
00
.51
0.0
0(in
1,00
0sU
S$)
(8.4
6)(2
.92)
(2.7
0)(4
.76)
(0.0
9)T
ropi
cs5
62.8
01
5.70
3.7
62
.36
0.1
4(0
.92)
(0.0
6)(2
.74)
(3.6
9)(0
.73)
Coa
stal
535
.40
42.
503.
000.
530.
16(0
.83)
(0.1
6)(2
.06)
(0.7
7)(0
.75)
Inte
rcep
t6
,457
.00
2,8
83.0
012
.28
2.0
42.
03
R2
(adj
uste
d)79
.20
51.2
018
.50
22.0
00.
00N
umbe
rof
Cou
ntri
es99
9799
9191
NO
TE
:Hon
gK
ong
and
Tai
wan
wer
eom
itted
from
Equ
atio
n2
beca
use
the
FD
Ida
taw
ere
unav
aila
ble.
ECONOMIC FREEDOM
219
institutional and geographical factors have an impact on investment asa share of GDP.
The last two regressions of Table 5 focus on the growth of physicaland human capital per worker. The fourth regression shows thatbetter institutions enhance the growth rate of physical capital perworker. Again, the level of GDP in 1980 has a negative impact on thegrowth rate of physical capital, so countries that are poorer initiallytend to have higher investment growth. The growth of physical capitalfor countries in the tropics tends to be slower, but the coastal variablefails to exert a statistically significant impact on the growth of Kpw. Inthe final regression with the growth rate of human capital as thedependent variable, none of the independent variables are statisticallysignificant.
Taken as a group, the regressions in Table 5, like those of Table 3,indicate that a countrys institutional environment exerts a strongimpact on capital formation. Investment tends to flow toward coun-tries with institutions and policies that are more consistent with eco-nomic freedom. While the results are mixed for the geographic andlocational variables, there is some evidence that a tropical locationadversely affects the investment rate of physical capital.
As Tables 2 and 5 have shown, institutional quality affects both thestock of capital and rate of investment. Institutional factors may alsoinfluence the productivity of investment. Table 6 analyzes this issue.The dependent variable in Table 6 is the annual growth rate of percapita GDP from 1980 to 2000. The first regression uses investmentas a fraction of GDP for 19802000 as the independent variable,measured as the average of each years ratio of investment to GDP,and shows that the level of investment explains a substantial share(43.5 percent) of the variation in GDP growth across countries. Equa-tion 2 of Table 6 partitions the 99 countries in the database into threegroups based on their EFW ratings. The first independent variablemultiplies I/GDP by one if a nations EFW rating is 7 or above, andzero otherwise. The second independent variable does the same fornations with an EFW rating between 5 and 6.99, and the third inde-pendent variable separates out countries with an EFW rating below5. All of the independent variables are significant, and together ex-plain nearly half of the variation in GDP growth across countries.
The key feature of this regression is the magnitude of the coeffi-cients. For countries with EFW ratings of 7 or above, the coefficientis .275, which is greater than the .236 coefficient for the countrieswith ratings from 5 to 6.99, which in turn is greater than the .197coefficient for the countries with ratings below 5. Equation 3 of Table6 adds the tropical and coastal variables to the model. The tropics
CATO JOURNAL
220
TA
BL
E6
EC
ON
OM
ICF
RE
ED
OM
AN
DT
HE
PRO
DU
CT
IVIT
YO
FIN
VE
STM
EN
T
Dep
ende
ntV
aria
ble:
Ave
rage
Ann
ualG
row
thR
ate
ofG
DP
Per
Cap
ita,1
980
2000
(t-r
atio
inpa
rent
hese
s)
Inde
pend
ent
Var
iabl
es(1
)(2
)(3
)(4
)(5
)
I/G
DP,
1980
200
00.
244
(8.7
4)I/
GD
P,19
802
000
E
FW
>7.
00.
275
0.24
2(9
.40)
(7.8
1)I/
GD
P,19
802
000
5.
0