Do Institutions Rule? October 2007 () Do Institutions Rule? October 2007 1 / 9
“Institutions Rule: The Primacy of Institutions over Geography and Integration” Rodrik, Subramanian and Trebbi (2001), (Journal of Economic Growth, 2004) Basic idea: To conduct a “horse race” between competing “fundamental” determinants of cross country per capita incomes Considers three competing hypotheses: (1) direct impact of geographical factors (2) impact of institutional quality (3) impact of globalization (increased trade) Takes advantage of previously discovered “good instruments” , ! settler mortality rates (Acemoglu, Johnson and Robinson, 2001) , ! predicted trade share using “gravity model” (Frankel and Romer, 1999) () Do Institutions Rule? October 2007 2 / 9
Figure 1: The “deep” determinants of income income level (4) (7) (8) (9) - 24 - (1) (5) integration institution s endogenous (6) (2) (3) geography exogenous
Estimation Framework (2SLS) Second stage regression: log y i = µ + α INS i + β INT i + γ GEO i + ε i where INS i = “institutional quality” (rule of law index) INT i = “integration” (trade–GDP ratio) GEO i = \ geography” (e.g. distance from equator) Firs stage regressions: INS i = λ + δ SM i + ϕ CONST i + ψ GEO i + ε INSi INT i = θ + σ CONSTi + τ SM i + ω GEOi + ε INTi where SM i = settler mortality rates CONST i = predicted trade share Note identi…cation requires at least 2 instruments () Do Institutions Rule? October 2007 3 / 9
Main Results Once institutional quality is added, geography and openness have no additional explanatory power (Table 2, Columns 4-6) , ! “geography matters only through its impact on institutions" Similar results for larger sample (including uncolonized) with alternative instruments (Table 2, Columns 7-9) Samilar pattern for physical/human capital and productivity (table 3) Robust across various speci…cations (Table 4) Robust to choice of measure of geography (Table 5) , ! “malaria prevalence” has small, statistically signi…cant direct impact , ! not truly exogenous variable: depends on eradication success () Do Institutions Rule? October 2007 4 / 9
Table 2: Determinants of Development: Core Specifications AJR sample Extended AJR sample Large sample Extended AJR sample log log log log log log log log log GDP GDP GDP GDP GDP GDP GDP GDP GDP Dependent variable per per per per per per per per per capita capita capita capita capita capita capita capita capita 1995 1995 1995 1995 1995 1995 1995 1995 1995 RULE LCOPEN (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Panel A. Ordinary least squares Geography (DISTEQ) 0.74 0.20 0.32 0.81 0.25 0.36 0.76 0.21 0.24 0.82 -0.72 (4.48) * (1.34) (1.85) ** (5.35) * (1.85) *** (2.37) ** (10.59) * (2.75) * (2.9) (5.71) * (-3.47) * Institutions ( RULE) 0.78 0.69 0.79 0.70 0.80 0.77 0.57 (7.56) * (6.07) * (8.96) * (6.86) * (12.41) * (10.71) * (4.14) * Integration (LCOPEN) 0.16 0.15 0.08 0.34 (1.48) (1.61) (1.31) (3.37) * Panel B. Two-stage least squares Geography (DISTEQ) 0.74 -0.42 -0.56 0.81 -0.44 -0.70 0.76 -0.05 -0.14 0.78 -0.86 - 31 - (4.48) * (-1.19) (-1.23) (5.35) * (-1.22) (-1.34) (10.59) * (-0.4) (-0.91) (5.64) * (-3.09) * Institutions ( RULE) 1.67 1.78 1.76 2.00 1.19 1.32 0.77 (4.29) * (3.78) * (4.4) * (3.55) * (7.91) * (6.77) * (2.33) ** Integration (LCOPEN) -0.18 -0.302 -0.17 0.23 (-1.23) (-1.07) (-1.35) (2.04) ** No. of observations 64 64 64 80 80 80 140 140 140 80 80 R-square 0.25 0.54 0.562 0.264 0.51 0.52 0.417 0.50 0.55 0.54 0.38 Test for over-identifying restrictions (p-value) (0.0071) (0.0365) Panel C: First Stage for Endogenous Variables (Institutions (RULE) and Integration (LCOPEN)) Dependent variable RULE RULE LCOPEN RULE RULE LCOPEN RULE RULE LCOPEN LCOPEN RULE Geography (DISTEQ) 0.41 0.47 -0.25 0.46 0.53 -0.19 0.65 0.64 -0.04 0.01 0.46 (2.8) * (3.21) * (-1.99) *** (3.25) * (3.76) * (-1.42) (10.35) * (10.92) * (-0.75) (0.09) (3.25) * Settler mortality (LOGEM4) -0.39 -0.40 -0.30 -0.34 -0.34 -0.27 -0.28 (-3.87) * (-4.1) * (-3.49) * (-3.63) * (-3.75) * (-3.2) * (-3.63) * Population speaking 0.19 0.18 0.17 English (ENGFRAC) (2.69) * (2.69) * (2.66) * Population speaking other 0.12 0.16 -0.11 European langages (EURFRAC) (1.74) *** (2.43) ** (-1.65) Constructed openness 0.20 0.90 0.19 0.80 0.25 0.70 0.80 (LOGFRANKROM) (1.95) ** (10.28) * (2.16) ** (9.68) * (4.37) * (12.4) * (9.10) * F-statistic n.a. 22.9 17.2 41.7 n.a. 23.3 17.8 37.2 n.a. 46.3 44 42 45.0 23.3 R-square 0.41 0.44 0.66 0.36 0.39 0.58 0.49 0.55 0.54 0.53 0.36
Table 3. Determinants of Development: Channels of Influence Extended AJR sample Larger sample Dependent variable Income per Capital per Human capital Total factor Income per Capital per Human capital Total factor worker worker per worker productivity worker worker per worker productivity Geography (DISTEQ) -0.94 -1.68 -0.25 -0.32 -0.26 -0.39 -0.05 -0.14 (-1.47) (-1.59) (-1.5) (-0.97) (-1.15) (-1.11) (-0.91) (-0.89) Institutions ( RULE) 2.22 3.41 0.57 1.06 1.36 1.95 0.35 0.72 (3.29) * (3.01) * (3.14) * (3.08) * (5.01) * (4.5) * (5.21) * (3.7) * Integration (LCOPEN) -0.41 -0.68 -0.15 -0.13 -0.36 -0.53 -0.12 -0.15 (-1.31) (-1.26) (-1.84) *** (-0.79) (-2.27) ** (-2.34) ** (-3.19) * (-1.27) R-square 0.60 0.52 0.52 0.45 0.58 0.54 0.59 0.35 No. of observations 74 74 74 74 122 122 122 122 Notes: The four dependent variables—income per worker, capital per worker, human capital per worker, and the level of total factor productivity--are expressed in natural logarithms and are from Hall and Jones (1999). IV estimates for the AJR sample use settler mortality (LOGEM4) as the instrument for institutions and EURFRAC and ENGFRAC as the instrument for the larger sample. All regressors, except RULE, are in logarithms and are scaled. Standard errors are corrected, - 33 - using the procedure described in Frankel and Romer (1999), to take into account the fact that the openness instrument is estimated. T-statistics are reported under coefficient estimates. Significance at the 1 percent, 5 percent, and 10 percent levels are denoted respectively by “*”, “**”, and “***”.
Table 4. Determinants of Development: Robustness to "Influential" Observations, Neoeuropes, Legal Systems, Origin of Colonizer, and Religion Baseline 1 (1)* (1)** (1)*** (1)**** Baseline 2 (2)* (2)** (2)*** (3) (4) (5) Two-stage least squares: Dependent variable is log GDP per capita in 1995 Geography (DISTEQ) -0.70 -1.34 -0.66 -0.90 -0.58 -0.14 -0.14 0.02 -0.36 -0.96 -0.67 -0.81 (-1.34) (-1.08) (-1.38) (-1.14) (-0.81) (-0.91) (-0.91) (0.17) (-2.12) ** (-1.45) (-0.98) (-1.27) Institutions (RULE) 2.00 2.68 1.82 2.82 1.97 1.32 1.32 0.90 1.69 2.43 2.22 2.13 (3.55) * (3.03) * (3.31) * (2.43) ** (1.67) *** (6.77) * (6.77) * (8.47) * (4.87) * (3.09) * (2.56) * (2.97) * Integration (LCOPEN) -0.302 -0.44 -0.31 -0.75 -0.42 -0.17 -0.17 0.03 -0.36 -0.41 -0.23 -0.32 (-1.07) (-1.68) (-1.23) (-1.30) (-0.81) (-1.35) (-1.35) (0.25) (-1.46) (-1.50) (-0.79) (-1.12) REGIONAL DUMMIES Latin America (LAAM) 0.44 0.17 0.25 (1.25) (0.33) (1.655) *** Sub-Saharan Africa (SAFRICA) -0.19 -0.43 -0.63 (-0.51) (-1.11) (-3.79) * East Asia (ASIAE) 0.24 0.07 0.12 (0.56) (0.14) (0.62) - 34 - Legal origin [0.133] Identity of colonizer [0.058] *** Religion [0.019] ** R-square 0.52 0.56 0.65 0.44 0.63 0.55 0.55 0.67 0.55 0.53 0.56 0.59 No. of observations 80 78 79 76 76 140 140 137 136 80 80 80 Omitted observations None Singapore Ethiopia Australia Australia None None Cuba Australia None None None Ethiopia Canada Canada Czech Rep. Canada New Zealand New Zealand Germany New Zealand USA USA USA Notes: The dependent variable is per capita GDP in 1995, PPP basis. Baseline 1 corresponds to the specification in column (6) of Table 2. Baseline 2 corresponds to the specification in column (9) of Table 2. In columns labeled with 1 and 2 asterisks, influential observations are defined according to the Belsey, Kuh, and Welsch (1980) DFITS statistic, which requires omitting those observations for which DFITS exceeds 2(k/n)^(1/2), where k is the number of regressors and n is the sample size. In columns labeled with three or four asterisks, observations for Australia, Canada, New Zealand, and Canada (Neoeuropes) are omitted. Standard errors are corrected, using the procedure described in Frankel and Romer (1999), to take into account the fact that the openness instrument is estimated. T-statistics are reported under coefficient estimates. For legal origin, identity of colonizer, and religion, p-values for joint significance of the underlying variables (LEGFR and LEGSO for legal origin, COLUK and COLFR for colonizer’s identity, and CATH, PROT, and MUSL for religion) are reported. Significance at the 1 percent, 5 percent, and 10 percent levels are denoted respectively by “*”, “**”, and “***”. All regressors are scaled as described in the notes to Table 2.
Recommend
More recommend
