Transcript “How democracy and dictatorship affect economic growth: Evidence

“How democracy and dictatorship affect economic
growth: Evidence from James Monroe and the
Quing Dynasty to George W. Bush and the
Communist Party.”
Oslo seminar in positive political science
BI, Oslo, 12/2 2009
Carl Henrik Knutsen
Research question
• Is there an effect from democracy on economic
growth?
• Average effect
• The categorization of political regimes. Democracydictatorship continuum and conceptual unidimensionality
• A bit crude, but nevertheless an interesting and
much debated question
– Academic debates
– Policy debates
The important conclusions
• There is a likely positive effect on economic
growth from democracy.
– The overall evaluation of theoretical arguments
pointing to democracy’s growth advantage
– Different econometric analyses find a relatively
robust effect on a very extensive data sample.
Estimated 1 percentage point extra annual growth
in GDP per capita from “full democratization”.
– No support for Barro’s inverse U-curve. The more
democracy, the better for growth
Structure paper
• Introduction with short review of answers/positions taken on the research
question
• Theoretical arguments
– 4 Przeworski and Limongi arguments + 1 extra
– Multiple mechanisms and no single, dominating coherent theoretical
framework
– Scorecard
• Data
• Empirical analysis,
–
–
–
–
OLS with PCSE
FE and RE (answer to: what if unobserved country-specific effects?)
2SLS (answer to: what if substantial endogeneity bias?)
Matching (non-linear effects)
• Short empirical section on Barro’s claim that semi-democracy is better for
growth
• Conclusion
The (very!) general framework
Aspect of political regime related to degree of democracy (ch&bal, free speech, free
and fair elections…)
Policy
Change in inputs (physical and human capital, labor), efficiency, technological change
Economic growth
Summary theoretical scorecard
Argument
Przeworski and Limongi's conclusions
My conclusions
1) Democracy and property rights
Either way
For democracy
Either way (for democracy when adding
2) Democracy and investment
Against democracy
human capital)
particularistic pressure
Against democracy
Against democracy, but with qualifications
4) Autonomous rulers are predatory
For democracy
For democracy
5) Democracy and technological change
-
For democracy
3) Dictatorship and insulation against
Data
• Maddison’s data
– Historical data on GDP and population
• 1945 or 1960 as “year 0” in quantitative analysis
– Draws on different sources, uncertainty
– Includes many countries often left out of other
databases: e.g. North Korea.
• Reduces systematic selection The effect from democracy
has been underestimated earlier
• Polity, from 1800..
• First year used in analysis is 1820
OLS with PCSE
• GRi,t = β0+β1polityi,t-2 + β2ln(dur+1)i,t-2 + β3ln(gdp/cap)i,t-2 +
β4ln(pop)i,t-2 + Σβjreligi + Σβkregi + Σβlperiodl
• Four models: small (n=8933) and interpolated
samples (n=9438), five period vs decade time
controls
• All significant at 1%-level
• Estimated polity-coefficients between 0.047 and
0.049.
• RE: relatively similar to OLS with PCSE
• FE: significant at 5%-level when five-period
contrls, insignificant when decade-dummies
2SLS (G2SLS: random effects IV)
• Instrument
Dem
Growth
• Huntington’s waves and exogenous variation in regime type
(global trends and contagion from neighbours exogenous to
national politics)
• 2SLS as consistent, but relatively inefficient
• Model with decade dummies: sig at 10%-level with
interpolated sample, t=1.51 for small sample
• However, larger point estimates than OLS with PCSE, RE
and FEHausman tests find no significant difference with
RE modelgo with the more efficient RE?
• Extra instrument to strengthen efficiency (absolute latitude
from Hall and Jones) Both models sig at 5%-level
Matching
Sample
Polity cut-off
Number of matches
Average treatment
effect
T-value
Number of
observations
Small
0
1
2.19
3.75***
8933
Small
0
3
1.29
3.55***
8933
Small
0
5
0.98
3.49***
8933
Small
0
10
0.68
3.22***
8933
Small
4
1
1.32
2.89***
8933
Small
4
3
0.78
2.50**
8933
Small
4
5
0.67
2.73***
8933
Small
4
10
0.58
3.10***
8933
Interpolated
0
1
2.21
4.01***
Interpolated
0
3
1.37
3.99***
9438
9438
Interpolated
0
5
1.05
3.98***
Interpolated
0
10
0.74
3.73***
Interpolated
4
1
1.39
3.23***
Interpolated
4
3
0.88
3.00***
Interpolated
4
5
0.77
3.26***
Interpolated
4
10
0.65
3.58***
9438
9438
9438
9438
9438
9438
Barro’s inverse U?
Figure 2: Estimated marginal effect of democracy on economic growth
Interpolated sample
Marginal
effect
Small
sample
Polity-score
Conclusion
• Different methods, based on different
assumptions, indicate a positive effect from
democracy on growth when using the largest
sample of data available.
• No democracy-growth trade-off, rather to the
contrary.