Democracy, Dictatorship and Protection of Property Rights
Carl Henrik Knutsen
Department of Political Science, University of Oslo
This paper investigates how political regimes influence property rights. The paper
reviews arguments for and against the hypothesis that democracy enhances property
rights protection, and then conducts empirical tests. Democracy is likely endogenous
to property rights protection. The analysis takes this into account by utilising an
innovative instrument for democracy. The results, based on data from 1984 to 2004 for
over 120 countries, show that democracy enhances property rights protection, even
when controlling for endogeneity and country-specific characteristics.
Property rights are vital for generating economic development (for example North, 1990;
Knack and Keefer, 1995). But, which factors affect the protection of property rights? This
paper investigates whether democracy enhances property rights protection. Previous empirical
studies (e.g. Leblang, 1996; Adsera et al., 2000; Clague et al., 2003) show that democracies
generally have better property rights protection than dictatorships. However, this may be
because democracy strengthens property rights, because property rights protection enhances
democratisation or democratic stability, or that other factors affect both democracy and
property rights systematically. Although there are strong arguments indicating that democracy
affects property rights, a large literature in the social sciences suggests that property rights and
other “free market institutions” are preconditions for democracy, rather than consequences.
More generally, both modernisation theorists following Lipset (1959) and academics drawing
on Moore’s (1966) historical-sociological approach have considered democracy a
consequence and not a cause of economic factors. Also, historical, geographic or social
conditions may impact on both democracy and property rights, and generate a positive
correlation even in the absence of any causal effect between the two.
Most previous studies on democracy’s effect on property rights have failed to sufficiently
control for the endogeneity of democracy. Moreover, many studies have failed to control for
country-specific factors that may affect both democracy and property rights. This study
utilises panel data instrumental variable methods to deal with these two problems, and
contributes to the literature by developing a novel instrument for democracy, based on
Huntington’s (1991) insight that democracy has thrived globally in specific time periods. The
instrument enables the isolation of the effect from democracy on property rights from other
sources of correlation, leading to consistent estimates of the effect. The results show that
democracy indeed strengthens property rights protection, and these results are very robust.
Section 2 presents a short discussion of property rights, and their economic effects. Section 3
reviews theoretical arguments on how democracy and dictatorship may affect property rights.
Section 4 briefly discusses the methodical issues of omitted variable bias and endogeneity of
democracy, before introducing and discussing the new instrument for democracy. Section 5
presents the data and the empirical analysis.
2. Property rights and their economic effects
One straightforward way to define property rights is “[T]o say that an individual has property
rights over something simply means in a legal and practical sense – that an individual can say
that a thing belongs to him and others will act accordingly” (Leblang, 1996: 7).
distinction is that between de jure and de facto protection of property rights. Here, we are
concerned with de facto property rights protection. Several distinct policies may alter de facto
control over property. Property can be taken from an individual or firm by extra-legal actions,
but property can also be transferred by law. Heavy taxation, expropriation, nationalisation and
theft are some means for state actors to acquire property. Also subsequent property
distribution may take on different forms, from public goods provision and universal
redistributive transfers to targeted handouts (see Bueno de Mesquita et al., 2003). Other types
of property redistribution include partially compensated expropriation, debt cancellation,
hyperinflation (redistribution from lenders to borrowers), land-reform and collectivisation.
Also, the structure and enforcement characteristics of a state’s legal system affect the extent
of theft or destruction of property by non-state actors. The operational definition of property
rights in Section 5 is relatively broad, and incorporates confiscation and expropriation of
private property, contract breaches, and threats to property from theft and lacking law and
Different methods of property redistribution may have different economic effects. Land
reforms often have positive efficiency effects (Banerjee and Duflo, 2005), whereas theft and
hyperinflation have adverse economic effects. However, generally, weak protection of
property rights introduces uncertainty to economic transactions (North, 1990) and reduces
expected gains from different productive activities. Property rights “create an incentive to
channel individual economic effort into activities that bring the private rate of return close to
the social rate of return” (North and Thomas, 1973: 1). With weak property rights protection,
actors may not undertake investments and other productive activities; it may rather pay off to
steal and engage in other predatory activities (North, 1990). Profitability from different types
of investment, like investments in physical capital, specific human capital and in innovative
activities, is determined by security of property rights to the investment capital and its future
rents (North, 1990; Romer, 1990; Leblang, 1996; Hall and Jones, 1999). Physical capital,
human capital and innovation are again key determinants of income level, and even long-term
economic growth according to some growth models (Barro and Sala-i-Martin, 2004).
Comprehensive property rights protection also reduces transaction costs and the risk of
economic transactions, thereby allowing for efficiency-enhancing specialisation of labour and
subsequent trade (North, 1981, 1990). Property rights enforced by a well-functioning legal-
system are particularly vital for inter-temporal transactions, as they enable actors to make
credible promises. Various empirical studies find that property rights protection enhances
income level and economic growth (Torstensson, 1994; Knack and Keefer, 1995; Goldsmith,
1995; Leblang, 1996; Hall and Jones, 1999; Acemoglu et al., 2001).
Thereby, whether democracy enhances economic development depends in part on whether
democracy strengthens property rights (North, 1990; Przeworski and Limongi, 1993; Leblang,
1996; Przeworski et al., 2000). However, there is disagreement on whether democracy
strengthens or weakens property rights. Let us review the different arguments.
3. Review of arguments and evidence
3.1 Democracy weakens property rights: Democracy, class politics and
Contemporary observers considered democratisation a threat to existing private property
rights structures in 19th-century industrialising Western societies. Poor and uneducated voters
were expected to use their electoral powers to redistribute, or even collectivise, the private
property of wealthy industrialists and landlords. For example David Ricardo argued to
“deprive those of the elective franchise against whom it could be justly alleged that they
considered it in their interest to invade [property rights]” (cited in Bethell, 1998: 336).
Przeworski and Limongi (1993) revisit the historical debate and conclude that the right to vote
and freedom of organisation were commonly perceived to harm private property rights
protection: Democracy implies political power to the majority through voting, but also
through improved political and civil rights. The majority of citizens are assumed to be
relatively poor, and those in majority maximise income not only in the market but also
through politically induced redistribution. Democracy thus allegedly leads to redistribution of
property from landholders, capitalists and other rich elites to the poor majority.
More recently, Acemoglu and Robinson’s (2006) and Boix’ (2003) models have linked
progressive redistribution to democracy. Consider a simple median voter model in a society
with income inequality. Assume that the upper socioeconomic strata earlier had the
opportunity to choose the government. In a given year, this is changed by democratisation
after which the whole adult populace is enfranchised. If voters maximise income, widespread
redistribution to poorer population segments would follow, in accordance with the new
median voter’s preferences. More generally, any group that constitutes a democratic majority
could enforce redistribution of property from minorities to themselves. The electoral channel
is not the only democratic channel that may affect property distribution. Freedom of
organisation is a core civil liberty, and trade unions are heavily regulated organisations in
dictatorships. Many trade unions have historically aimed at comprehensive redistribution or
worker control over capital (Gramm, 1981), which together with freedom of organisation
under democracy may imply a negative effect from democracy on property rights.
The argument above rests on several problematic assumptions. One is that the counterfactual
alternative to a democracy is a dictatorship based on support from economic elites, where
“[C]onservative politicians and the military intervene to sustain the property rights of
capitalists” (Boix, 2003: 16). This description might fit Latin America in the 1960s and 70s,
where several military dictatorships came to power through coups, with explicit or tacit
support from economic elites (Smith, 2005: 82-83), and where landless agricultural workers
and industrial workers wanted to alter existing private property arrangements. However,
dictatorial regimes may be based on support from relatively poor population segments. If
dictators are responsive to their supporters’ preferences, the argument above would not imply
any difference between a democracy with a poor median voter and a dictatorship backed by
the poor. Indeed, there may be theoretical reasons to expect such dictatorships to go farther in
confiscation and redistribution than democracies (see Boix, 2003; Acemoglu and Robinson,
2006). Historically, Marxist dictatorships such as the Soviet Union, Cuba and China
proceeded with nationalisation of industries and collectivisation of agriculture on a scale
uncommon in democracies.
3.2 Democracy strengthens property rights: The unchecked dictators’ grabbing
The argument above predicted that democracy endangers property rights through politically
empowering the poor. One assumption was that property redistribution is progressive, and
that political elites do not grab the property of the general populace, rich and poor. Once we
realise that political elites can redistribute property from all citizens to themselves and their
political backers, dictatorship in general, not only Marxist dictatorship, may hurt property
rights protection. As North argues (2000: 51), socially non-optimal property rights designs,
and non-optimal redistribution of property rights, can be privately beneficial to political
leaders or their supporters. If politicians are consumption-maximising or motivated by
political survival, we can expect more property rights violations in dictatorships than in
democracies. There are several reasons why:
First, democratic leaders have broad winning coalitions, and universal property rights
protection may thus be a cheaper way for them to ensure their political survival than selective
grabbing and redistribution to all their supporters (Bueno de Mesquita et al., 2003). Dictators
with small winning coalitions often rely on socially inefficient distribution of private goods to
their supporters to stay in power. Dictators are thus likely to violate the property rights of
those outside their winning coalitions and redistribute property to those within. Second,
relatively democratic systems often have institutionalised checks and balances, with for
example independent judiciaries and separation of legislative and executive authority.
Dictatorships often lack such checks and balances (North, 1981; North and Weingast, 1989).
Thus, “democratic institutions provide a check on the arbitrary power of the state. Democratic
regimes will less likely confiscate possessions and are more likely to provide a stable
environment governed by efficient property rights structures” (Leblang, 1996: 18). Third,
there is a higher degree of political accountability in democracies. Democratic electorates can
vote out leaders they are dissatisfied with the performance of. One would expect many
citizens to value a stable, universal property rights framework. Retrospective voting should
thus discipline democratic politicians to not grab property. Only in some circumstances are
self-interested dictators disciplined into behaving similarly (Besley and Kudamatsu, 2007).
Fourth, different politicians are likely motivated by different things. If citizens to some degree
are able to discern between “good” and “bad” politicians, democratic elections provide
citizens with a tool to elect leaders who are less likely to grab property in self-interest. All
these mechanisms indicate that property grabbing by political elites is less prevalent in
democracies than in dictatorships.
However, some dictators may have incentives to restrain grabbing of property. Excessive
property violation will negatively affect subjects’ incentive to invest and produce, thereby
reducing the future resource-base from which a dictator can extract. Thus, patient dictators
and dictators who expect long tenures may refrain from confiscation in self-interest (Olson,
2003). Analogous reasoning indicates that dynastic regimes may protect property rights. If the
monarch values the welfare of his successor, he wants to give a well-functioning economy as
3.3 Previous empirical studies
The overall effect from democracy on property rights is thus theoretically ambiguous.
Accordingly, some argue that “[T]he idea that democracy protects property rights is a recent
invention, and we think a far-fetched one” (Przeworski and Limongi, 1993: 52), whereas
others argue that the regime best securing property rights “is a modern democratic society
with universal suffrage” (North, 1990: 109). The effect from democracy on property rights
therefore has to be determined empirically. Some, but not many, studies have empirically
investigated the effect from democracy on property rights:
According to Goldsmith (1995), Cheibub (1994) uses level and change in tax revenue as share
of GDP as proxies for property rights, and finds no significant effect from political regime.
However, tax revenue captures only one aspect of property rights protection. In a follow-up
article, which utilises country fixed-effects and selection models to account for endogeneity,
Cheiub (1998) finds no significant effect from democracy on taxation, but does not explicitly
link taxation to property rights. Leblang (1996) uses “exchange rate controls” and “credit
allocated to private enterprises as percentage of GDP” as proxies for property rights security,
and finds that “democratic nations tend to protect private property rights to a greater extent
than non-democratic” (p. 6). However, Leblang uses pooled cross-sectional time-series
models without incorporating country-fixed effects, and does not consider that democracy
may be endogenous. Boix (2003) finds that democracies on average have lower expropriation
risk than dictatorships, but controls only for newspaper circulation, and does not consider
endogeneity. Adsera et al. (2000) also find that democracy reduces expropriation risk, when
controlling for other relevant variables, but the results are only based on cross-section
evidence; neither country-fixed effects nor any solution for endogeneity is incorporated.
Clague et al. (2003) conduct thorough studies on the effects from regime type on property
rights, and include models with country-fixed effects. Their models also control for income,
which is argued to reduce endogeneity problems; strong property protection over time
generates high income, which again enhances probability of having a democracy. However,
this is an insufficient control for endogeneity. Nevertheless, the authors find that democracy
in general protects property rights better. The fixed effects results are not robust; but many of
these regressions rely on data with short time-series. One important nuance is that strong
protection is only likely in consolidated democracies. Furthermore, they find support for
Olson’s (2003) argument that dictators with long time horizons tend to protect property and
contracts better. On this issue, the authors use instrumental variable techniques as robustness
checks to deal with endogeneity of autocrats’ tenure length, and obtain mixed results. They
use autocrats’ age when entering office and tenure of previous autocrat as instruments.
Many of the studies above find a positive effect from democracy on property rights. However,
in addition to Cheibub, who studies effects on taxation, only Clague et al. (2003) incorporate
country-fixed effects; these models yield non-robust results. This may indicate that country-
specific omitted factors affect the positive correlation between democracy and property rights.
Moreover, as there is no proper solution to the endogeneity problem in the above studies,
except for Cheibub (1998), the correlation may also be due to property rights affecting
The analysis below deals with endogeneity and omitted variable bias more
thoroughly. Let us first look closer at these methodical problems.
4. Methodical problems and solutions
4.1 Omitted variable bias and endogeneity
We can control for observable factors, like income level, in an ordinary regression, but there
may be unobservable factors that generate omitted variable bias. As Acemoglu et al. (2008)
argue, historical, country-specific factors likely affect different institutional structures and
economic outcomes systematically. Acemoglu et al. (2001) showed that country-specific
factors related to historical colonial institutions impact on present institutional structures
related to property rights protection, but also present democracy. More generally, underlying
cultural, geographic, political-historical and socio-economic factors may generate a positive
correlation between democracy and property rights, and some of these factors are difficult to
measure. However, we can utilise panel-data techniques with country-fixed effects to mitigate
omitted variable bias.
Moreover, democracy is likely endogenous to property rights protection. Various political
theorists, social scientists and historians have argued that property rights are fundamental
determinants of political and civil liberties (Nozick, 1974; Locke, 1988; Pipes, 1999;
Friedman, 2002; Feng, 2003: 256-257). The main hypothesis from this literature is that “while
property in some form is possible without liberty, the contrary is inconceivable” (Pipes, 1999:
xiii). If this is true, statistical studies investigating democracy’s effect on property rights are
plagued by endogeneity-bias. Some quantitative studies indicate effects from capitalist
economic institutions, like property rights, on democracy directly (for example Burkhart,
2000) or indirectly via affecting income level (Wu and Davis, 1999).
A positive correlation between democracy and property rights could thus stem from property
rights affecting regime type. Property rights may enhance democracy through increasing
development level, reducing inequality and through accumulation of economic resources
outside the state’s control (Weimer, 1997: 9-10). Moreover, property rights protection
increases the middle-income groups’ share of total income (Scully, 1991; Weimer, 1997: 9).
This should widen and strengthen the middle classes economically, but also enhance their
political clout. Several political scientists have argued that a strong middle class enhances
democracy (for example Lipset, 1959). Thus, there are several channels through which
property rights protection may increase the probability of having a democracy.
4.2 Methodical solutions
As mentioned above, including country-fixed effects mitigates omitted variable bias.
However, this means throwing away important information from cross-country comparisons
(Beck and Katz, 2001). Luckily, there is substantial intra-national variation in this study’s
data, which enables usage of panel data techniques. Single-equation methods are not well-
equipped to deal with endogenous independent variables. But, if a valid instrument is utilised,
2SLS-versions of panel data methods yield consistent estimates of democracy’s effect on
property rights, even if property rights affect democracy.
The instruments used in the literature on institutions’ economic effects, such as settler
mortality (Acemoglu et al., 2001) and fraction of European language speakers (Hall and
Jones, 1999) are not valid instruments for democracy when property rights protection is
dependent variable. These variables are not only linked to democracy, but also, as argued in
the abovementioned studies, to economic institutions that influence property rights protection.
An instrument is only valid if it is correlated with the independent variable (here regime type),
but not directly linked to the dependent variable (property rights), after controlling for the
regressions’ other variables. Generally, it is difficult to find instruments for regime type. One
of few suggestions is given by Helliwel (1994), who uses historical political regime values to
instrument for present values, when investigating the effect on economic growth.
Here, I develop an instrument for political regime based on Huntington’s (1991) observation
that democratisation (globally) has historically proceeded in three temporal waves. The first
long wave lasted from 1828 to 1926, and the second from 1943 to 1962 (Huntington, 1991:
16). According to Huntington the third wave began with the military takeover in Portugal in
1974 (Huntington, 1991: 3). There are indications that the number of democracies stagnated
globally in 1998 (reference removed), and I therefore assign 1998 as the third wave’s end-
year. Between these waves, there have been reverse waves, where the number of democracies
globally either stagnated or regressed. The first reverse wave lasted from 1922 to 1942, and
the second from 1958 to 1975, according to Huntington.
Significant geo-political events have arguably contributed to the temporal clustering of
democratisation worldwide. Huntington links the first wave to the American and French
revolutions and the second wave to the allied victory in World War II. The third wave started
in Southern Europe in the mid-70’s, and soon transplanted to Latin America. However, many
of the democratisation processes in Eastern Europe, Central Asia and Africa more than a
decade later can be connected to the fall of the Berlin Wall and Soviet Empire. However, not
only important geo-political events make democratisations cluster. There are probably also
“contagion effects” or “spill-over effects” between countries, for example neighbours (see
Starr, 1991); one country’s democratisation increases the probability of neighbouring (and
other) countries democratising. It is hard to determine the relative importance of major geo-
political factors and contagion from neighbours, but this is irrelevant here since the clustering
of democratisation experiences in any case can be attributed to factors exogenous to national
political and economic factors.
If Huntington’s waves represent exogenous factors that affect the probability of regime
transition to democracy, and whether a regime transformed within these waves or not is not
directly related to current property rights protection (when controlled for other variables in the
model), a dummy that records whether the last regime transition was within a wave or reverse
wave is a valid instrument. The instrument allows us to utilise exogenous variation in degree
of democracy, thereby allowing us to filter out variation in democracy caused by domestic
property rights protection. The idea of using institutional change-date as instrument for
institutional structure comes from Persson and Tabellini (2003). By using Huntington’s
observation and Persson and Tabellini’s methodological trick, I can apply 2SLS and achieve
consistent estimates of democracy’s effect on property rights.
The instrument-dummy used here, WAVE, is scored 1 if the last regime change, according to
Polity-data, was outside one of Huntington’s reverse waves and a 0 if not. A regime change is
recorded by Polity if a country experiences a three-point movement or more on the Polity-
index (PI), in three years or less, or experiences the end of a transition period “defined by the
lack of stable political institutions” (Marshall and Jaggers, 2004: 16). Huntington’s
categorisations are somewhat ambiguous, with waves and reverse waves overlapping.
However, an instrument based on Huntington’s reverse wave-categorisation has stronger
correlation with the democracy measure (PI) than an instrument based on Huntington’s wave-
According to econometric theory (Greene, 2003), instruments that have
stronger correlation with the endogenous independent variable are preferable, since this
generates more efficient estimators. Therefore, country-years where the current regime
originated in (,1827], [1922,1942], [1958,1975] and [1998,2004] are scored 0 on WAVE,
and all other country-years are scored 1.
I further discuss and test the validity of WAVE
5. Empirical analysis
The democracy-measure used in this study is the PI from Polity IV, which goes from -10
(most dictatorial) to 10 (most democratic). The PI is additive, and the components are
competitiveness and openness of executive recruitment, constraints on the chief executive and
competitiveness and regulation of political participation (Marshall and Jaggers, 2002: 12-15).
I exclude interregnum-periods (Marshall and Jaggers, 2005: 17), mainly periods of internal
anarchy or civil war. I also exclude periods of “foreign interruption”, mainly periods of
foreign occupation. The mean PI-score after 1984 is 1.7, and the overall standard deviation is
7.3. The standard deviation is 6.6 between panels and 3.3 within panels.
The property rights data are taken from the ICRG dataset (The PRS group, 2009).
specifically, I constructed an index based on two of the ICRG data’s components. The first is
the Investment Profile index (IP). IP goes from 0 to 12 and consists of three sub-components,
which all range from 0 to 4: Contract Viability/Expropriation Risk, Profits Repatriation and
Payment Delays. The second component is the Law and Order index (LO). LO consists of a
Law-component, which assesses strength and impartiality of the legal system and an Order-
component which assesses popular observance of the law. LO ranges from 0 to 6, with the
two sub-components ranging from 0 to 3. In order to capture different aspects of property
rights protection, I construct ICRGPROP = IP+2*LO, which goes from 0 to 24.
country-year score on ICRGPROP belongs to Somalia in 1993 (1), and the highest (24) is
given to Austria, Canada, Finland, Ireland, Luxembourg, Netherlands, Singapore, Sweden and
the United Kingdom in one or more years. The highest score is thus achieved by liberal
market economy democracies, coordinated market economy democracies and one (at least)
semi-authoritarian regime (Singapore). The mean score in the sample of 3166 observations is
14.3 and the overall standard deviation is 4.8. The standard deviation between panels is 3.9,
and the standard deviation within panels is 2.7; indicating substantial variation in property
rights protection after 1984, the first year with data.
The models below incorporate several control variables. A high income per capita level
influences the probability of having a democratic regime positively (see Przeworski and
Limongi, 1997; Boix and Stokes, 2003), and a high income level is also conducive to strong
property rights protection (Clague et al., 2003: 152-153). Democracy might be easier to
establish in countries with small populations, and population level might also affect property
rights protection. Political stability influences both political regime and protection of property
rights (Clague et al., 2003). Moreover, cultural factors correlated with for example the
dominant religion or the geographical region in which a country is situated may affect both
probability of being democratic and property rights. Other historical variables could also be
of importance. La Porta et al. (1999) find that legal origin is important for quality of
government. Legal origin is correlated with colonial history and colonial history may also
affect the probability of having democracy. In all models below, I therefore control for log
GDP per capita, log population, log (regime duration+1) as a proxy for political stability,
plurality religion, geographic region and nationality of historical coloniser. The GDP and
population data are from the World Development Indicators (World Bank, n.d.), and the
regime duration data are from Polity IV (Marshal and Jaggers, 2002). Plurality religion,
region and coloniser are sets of dummy-variables taken from (reference removed), and fall out
of Fixed Effects models since they are time-invariant. In the main analysis, the independent
variables are lagged two years to further mitigate endogeneity problems, but other lag-
specifications are tested for checking the results’ robustness.
5.2 Preliminary analysis
I first run an OLS with PCSE regression that takes into account panel-specific autocorrelation
(AR1), heteroskedasticity between panels and contemporaneous correlation. This analysis
allows for both cross-country comparisons and comparisons over time as basis for inference.
There are almost 2500 observations from 122 countries, with data from 1984 to 2004 for most
countries, although some have shorter time series. There is a positive and significant effect at
the 1%-level from democracy (t=4.0), measured by PI, on property rights protection,
measured by ICRGPROP. However, this analysis may be plagued by omitted variable bias, as
it allows cross-country comparisons. Nevertheless, even if we control for country-specific
effects, by running robust Random Effects (RE) or robust Fixed Effects (FE) models, the
positive effect from PI on ICRGPROP is still positive and significant at the 1%-level. The t-
values are even higher in the RE (t=14.3) and FE (t=7.1) models. To illustrate, the estimated
increase in ICRGPROP is about 3.4 when experiencing a democratisation equal to that of the
Philippines in 1986 when Marcos was ousted (from -6 to 8 on PI), according to the RE results.
Democracy thus goes together with better property rights protection, even if we control for
country-specific factors. However, these results may be due to property rights protection
influencing regime type. Indeed, different tests based on 2SLS-FE models indicate that both
endogeneity and country-fixed effects should be controlled for: The Davidson-Mackinnon
(1993) test of exogeneity, which is particularly suitable for panel data (Baum, 2002), indicate
that democracy is likely endogenous to property rights protection: The null hypothesis that an
ordinary FE regression would yield consistent estimates is rejected at the 5%-level (p=0.02).
When it comes to the relevance of country-specific effects, an F-test on the 2SLS-FE model
shows the hypothesis that all country-dummies are equal to zero can be rejected at the 0.01%-
5.3 Main analysis
In order to check the results’ robustness, I ran both RE and FE versions of 2SLS. I also tested
models that included year-dummies, to control for time-specific effects.
Some results below
are therefore based on a very stringent inference-procedure, controlling for country-specific
factors, time-specific factors, and taking into account endogeneity of regime type. The 2SLS-
FE models also de facto controls for factors that are relatively constant nationally within a
twenty-year time frame, such as ethnic heterogeneity, the existence of natural resources and to
a certain extent also income inequality, which tends to change very slowly.
Nevertheless, the validity of 2SLS results relies on the validity of the instruments used.
WAVE was discussed above, but its validity can be checked further through various empirical
tests and statistics. First, the correlation between WAVE and the endogenous independent
variable, PI, should be sufficiently high to ensure efficient estimates. As Table 1 indicates,
there is no weak instrument problem. WAVE is highly correlated with democracy even when
controlling for the model’s other variables: WAVE’s t-values range between 9.4 and 12.2 in
the different models’ first stage regressions. More formally, F-tests for the exclusion of the
instrument in the first stage show the hypothesis that WAVE can be excluded is rejected at the
0.1%-level. The F-values are 103.9 for the FE model without year-dummies and 91.3 when
year-dummies are included.
WAVE is thus a relatively strong instrument.
**TABLE 1 HERE**
However, there may be independent effects from WAVE on ICRGPROP, which would
violate the 2SLS models’ exclusion restriction and thus generate inconsistent estimates. But,
regressions (OLS with PCSE, RE and FE) with WAVE as a regressor alongside PI and the
other control variables, did not show any significant effect from WAVE on ICRGPROP at the
5%-level level. These results indicate there is no independent effect from WAVE on
ICRGPROP. It is difficult to conduct valid overidentification tests, to further test the
exclusion restriction’s validity, because of few other good instruments for democracy. One
solution is to use a lag of WAVE as an additional instrument. I ran different 2SLS models
(RE and FE, with and without year-dummies) with WAVE and a 10-year lagged WAVE-
variable. The Sargan-Hansen statistics indicate the instruments are valid: The null hypothesis,
which assumes the instruments are uncorrelated with the error term and correctly excluded
from the second stage equation, cannot be rejected at the 10%-level, and the p-values are
generally higher than 0.50. The only exception is the test based on the RE model without
year-dummies, which yields a p-value of 0.02. Helliwell (1994) used lagged values of
democracy to instrument for present values, with economic growth as dependent variable. I
ran 2SLS models using both WAVE and 10-year lagged PI-scores as instruments. No model
yielded rejection of the null-hypothesis at the 10%-level, and most models yielded p-values
higher than 0.50.
To sum up, WAVE seems to be a valid instrument for democracy; it is
correlated with democracy and does not directly affect property rights.
We can thus proceed to the second stage regressions and investigate how democracy affects
property rights. As Table 2 shows, the estimated effect from PI on ICRGPROP is positive and
significant at least at the 1%-level in all the 2SLS models. Table 2 thus provides very strong
evidence for the hypothesis that there is a positive effect from democracy on property rights
protection. The PI-coefficients vary from 0.20 to 0.43, indicating that the estimated increase
on the 24-point ICRGPROP when going from least to most democratic on PI varies from 4.2
to 9.0. To illustrate such a difference, the 2007 ICRGPROP-score in Robert Mugabe’s
Zimbabwe was 8.5 points lower than India’s. Japan had a 4.8 higher 2007-score than Russia
and a 9.0 higher score than Sudan. Even when we control for all the abovementioned control
variables, including year- and country-specific effects, and incorporate that democracy is
endogenous, we find a statistically significant and substantially large effect from democracy
on property rights.
**TABLE 2 HERE**
There are several historical examples of dictators grabbing property to their own economic or
political benefit, which shed light on the mechanisms underlying the strong statistical results
above. Here are two: In 1973, Mobutu decided that farms, plantations and commercial
enterprises in Zaire should be turned over from foreign owners to “sons of the country”
(Wrong, 2000: 92), followed by confiscation of manufacturing plants. Mobutu not only
enriched himself on confiscated property, but also his winning coalition; people within his
own tribe and leading figures from those major ethnic groups he needed support from
(Wrong, 2000: 93). Another example “is that of the dictatorship of Rafael Trujilo in the
Dominican Republic between 1930 and 1961... Trujilo expropriated much of the land and
businesses of the country so that he eventually directly controlled about 85% of the economy
… and owned 60% of all land” (Robinson, 1998: 23). These two cases are not exceptional
(see for example Bueno de Mesquita et al., 2003; Meredith, 2005).
There is also historical evidence that democracies do not behave as the argument in 3.1
predicts. Several Western European countries experienced a decline in inequality following
franchise expansion, as taxation rates were increased and tax-revenue redistributed through
social welfare programmes and free public education (Acemoglu and Robinson, 2006; North
et al., 2009). Nevertheless, large-scale confiscation and expropriation of property seldom
followed democratisation in Western Europe. Similarly, democratic Botswana and Mauritius
pursued social programs and expanded education, but rich property holders largely kept their
properties (reference removed). One explanation is that majorities impose self-restraints
because they fear coups or that capital will be moved out of the country if expropriation is
attempted (Acemoglu and Robinson, 2006; Boix, 2003). In many democracies with high
inequality, such mechanisms may have contributed to the majority not expropriating the
wealthy minority. For example, Chile and South Africa experienced increasing rather than
decreasing ICRGPROP-scores after the Pinochet- and Apartheid regimes were substituted
with democracy. A democratic majority may also refrain from expropriation followed by
redistribution if it believes this will hurt future economic growth (Olson, 2003). These
empirical and theoretical considerations help explain the statistical results above.
5.4 Robustness checks
I ran different robustness checks to test the result’s credibility. First, the ex-Soviet and Eastern
European countries may be special cases affecting the results: These countries democratised
in Huntington’s third wave, but they also experienced a transition from planned economies to
market economies in this period, for reasons not necessarily related to democratisation. I rerun
all 2SLS models without these countries, but the effect from democracy is still significant at
least at the 1%-level. The significant results hold up when excluding other regions as well,
although some models yield insignificant results (5%-level) when Asian or Latin American
countries are excluded. Nevertheless, the point-estimates are generally stable, and t-values are
relatively large and positive.
Second, it is difficult to predict the time-lag of democracy’s effect on property rights. A two-
year lag was chosen above. However, the results are robust when using one-, three-, four-, and
five-year lags. All models, FE and RE with and without year-dummies, show significant
results at the 1%-level, except the 1-year lag models with year-dummies, which show results
significant at the 5%-level. Third, ICRGPROP was calculated on the basis of IP and LO from
ICRG. The 6-point range LO was given twice the weight of the 12-point range IP; a change
from minimum to maximum on LO and IP thus yield similar changes in ICRGPROP.
However, the results are quite robust to different specifications. When LO is dependent
variable, the PI is significant at the 1%-level for the RE and FE models without year-
dummies, but insignificant at conventional levels when including year-dummies. When IP is
used, all models show significant results at the 5%-level. I tried out various linear
combinations of IP and LO, and almost all PI-coefficients were significant at the 1%-level. I
also tested models using the component extracted from principal component analysis, on IP
and LO, as dependent variable. The PI was significant at the 1%-level in all these models.
Fourth, I checked the robustness of the results by substituting ICRGPROP with the Rule of
Law-index (RLI) from the World Governance Indicators. Many of the indicators used in the
scoring of RLI is related to property rights protection (Kaufman, Kraay and Mastruzzi, 2007:
74). The correlation between ICRGPROP and RLI is 0.85. Unfortunately we can only use
data from 1996, 1998 and 2000-2004 when using RLI. The RLI-models without year-
dummies then also had insignificant PI-coefficients (10%-level). However, despite the short
time series, the RE-version of 2SLS with year-dummies found a positive significant effect at
the 10%-level (p=0.057) and the FE-version at the 5%-level (p=0.022). The RLI covers more
countries than the ICRGPROP. 145 countries were included in the RLI-models, compared
with 122 in the ICRGPROP-models. When running the RLI-models on the original 122
country sample, the PI’s t-values decreased relative to those in the full samples. Thus, there
may be a sample selection bias on the effect from democracy in the ICRGPROP-regressions;
the actual effect may be even larger than estimated above. Examples of countries with RLI-
data and without ICRGPROP-data are Afghanistan, Cambodia, Lao, Turkmenistan and
several African countries. Table 3 shows some of the robustness test results; only FE models
without year-dummies are shown in the table.
To sum up, the positive effect from
democracy on property rights protection is quite robust.
**TABLE 3 HERE**
Although some dictatorships have provided decent property rights protection after WWII
(Przeworski et al., 2000: 211), this paper’s empirical analysis shows that democracy enhances
property rights protection. This result holds even when taking into account that property rights
protection may increase the probability of having a democracy, and that democracy and
property rights may be jointly determined by country-specific, unobservable factors.
Everything else equal, more democratic regimes protect property better than less democratic
regimes. Previous literature has identified both an enfranchised majority and a powerful
dictator interested in personal consumption or political survival as potential threats to property
rights. This study indicates that a strong dictator with excessive powers is a worse threat for
private property than an enfranchised population. Since property rights are vital for long-run
economic development, there is thus good reason also to believe that democracy is the
political system that generates most prosperity in the long run.
Acemoglu, D. and Robinson, J.A. (2006) Economic Origins of Dictatorship and Democracy
(New York: Cambridge University Press).
Acemoglu, D., Johnson, S. and Robinson, J.A. (2001) The Colonial Origins of Comparative
Development: An Empirical Investigation. American Economic Review, 91(5), pp. 1369-
Acemoglu, D., Johnson, S., Robinson, J.A. and Yared, P. (2008) Income and Democracy.
American Economic Review, 98(3), pp. 808-842.
Adsera, A., Boix, C. and Payne, M. (2000) Are You Being Served?: Political Accountability
and Quality of Government. RES Working Paper 4241, Inter-American Development Bank,
Research Department, Washington D.C.
Banerjee, A.V. and Duflo E. (2005) Growth Theory through the Lens of Development
Economics, in: P. Aghion and S.N. Durlauf (eds) Handbook of Economic Growth Volume 1,
(North Holland: Elsevier), pp. 473-552.
Baum, C.F. (2002) Facilitating Applied Economic Research with Stata, in: S.S. Nielsen (ed)
Programming Languages and Systems in Computational Economics and Finance, (Boston:
Kluwer Academic Publishers), pp. 173-197.
Besley, T. and Kudamatsu, M. (2007) Making Autocracy Work. DEDPS Working Paper No
48, London School of Economics, London.
Boix, C. (2003) Democracy and Redistribution (New York: Cambridge University Press).
Boix C. and Stokes, S.C. (2003) Endogenous Democratization. World Politics, 55(4), pp. 517-
Bueno de Mesquita, B., Smith, A., Siverson, R.M. and Morrow, J.D. (2003) The Logic of
Political Survival (Cambridge, MA: MIT Press).
Burkhart, R.E. (2000) Economic Freedom and democracy: Post-cold war tests. European
Journal of Political Research, 37(2), pp. 237-253.
Cheibub, J.A. (1994) Democracy and Property Rights: Is There a Link? Paper presented at the
Annual Meeting of the American Political Science Association, New York: 1st to 4th
Cheibub, J.A. (1998) Political Regimes and the Extractive Capacity of Governments:
Taxation in Democracies and Dictatorships. World Politics, 50(3), pp. 349-376
Clague, C., Keefer, P., Knack, S. and Olson, M. (2003) Property and Contract Rights in
Autocracies and Democracies, in: S. Knack (ed) Democracy, Governance and Growth, (Ann
Arbor: The University of Michigan Press), pp. 78-109.
Davidson, R. and MacKinnon, J.G. (1993) Estimation and Inference in Econometrics (New
York: Oxford University Press).
Feng, Y. (2003) Democracy, Governance and Economic Performance (Cambridge, MA.: The
Friedman, M. (2002) Capitalism and Freedom (Chicago: The University of Chicago Press).
Goldsmith, A.A. (1995) Democracy, Property Rights and Economic Growth. The Journal of
Development Studies, 32(2), pp. 157-174.
Gramm, W.S. (1981) Property Rights in Work: Capitalism, Industrialism, and Democracy.
Journal of Economic Issues, 15(2), pp. 363-375.
Greene, W.H. (2003) Econometric Analysis. 5th edition (Upper Saddle River: Prentice Hall).
Grugel, J. (2002) Democratization: A Critical Introduction (New York: Palgrave).
Haan, J.d. and Sturm, J.E. (2003) Does more democracy lead to greater economic freedom?
New evidence for developing countries. European Journal of Political Economy, 19(3), pp.
Hadenius, A. and Teorell, J. (2006) Authoritarian Regimes: Stability, Change, and Pathways
to Democracy, 1972-2003. Working Paper #331, Lund University, Lund.
Hall, R.E. and Jones, C.I. (1999) Why Do Some Countries Produce So Much More Output
Per Worker Than Others? Quarterly Journal of Economics, 114(1), pp. 83-116.
Helliwel, J.F. (1994) Empirical Linkages between Democracy and Economic Growth. British
Journal of Political Science, 24(2), pp. 225-248.
Huntington, S.P. (1991) The Third Wave. Democratization in the Late Twentieth Century
(Oklahoma: The University of Oklahoma Press).
Kaufmann, D., Kraay, A. and Mastruzzi, M. (2007) Governance Matters VI: Aggregate and
Individual Governance Indicators 1996-2006. Paper, The World Bank, Washington D.C.
Knack S and Keefer P. (1995) Institutions and Economic Performance: Cross-Country Tests
Using Alternative Institutional Measures. Economics and Politics, 7(3), pp. 207-227.
La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R.W. (1999) The Quality of
Government. Journal of Law, Economics, and Organization, 15(1), pp. 222–279.
Leblang, D.A. (1996) Property Rights, Democracy and Economic Growth. Political Research
Quarterly, 49(1), pp. 5-26.
Lipset, S.M. (1959) Some Social Requisites of Democracy: Economic Development and
Political Legitimacy. The American Political Science Review, 53(1), pp. 69-105.
Locke J. (1988) Two Treatises of Government. (Cambridge: Cambridge University Press).
Marshall, M.G. and Jaggers, K. (2002) Polity IV Project – Dataset Users’ Manual. Manual,
Program Center for International Development and Conflict Management, University of
Maryland. College Park.
Meredith, M. (2005) The State of Africa: A History of Fifty Years of Independence (London:
The Free Press).
Moore, B. (1966) Social Origins of Dictatorship and Democracy: Lord and Peasant in the
Making of the Modern World (Boston: Beacon Press).
North D.C. (1981) Structure and Change in Economic History (New York: W.W. Norton).
North, D.C. (1990) Institutions, Institutional Change and Economic Performance (New York:
Cambridge University Press).
North, D.C. (2000) Institutions and Economic Growth: A Historical Introduction, in: J.A.
Frieden and D. Lake (eds) International Political Economy. Perspectives on Global Power
and Wealth, (Boston: Bedford/St. Martin’s), pp. 47-59.
North, D.C. and Thomas, R.P. (1973) The Rise of the Western World: A New Economic
History (New York: Cambridge University Press).
North D.C. and Weingast B.R. (1989) Constitutions and Commitment: The Evolution of
Institutional Governing Public Choice in Seventeenth-Century England. The Journal of
Economic History, 49(4), pp. 803-832.
North, D.C., Wallis, J.J. and Weingast, B.R. (2009) Violence and Social Orders: A
Conceptual Framework for Interpreting Recorded Human History (Cambridge: Cambridge
Nozick, R. (1974) Anarchy, State and Utopia (New York: Basic Books).
Olson, M. (2003) Dictatorship, Democracy and Development, in: S. Knack (ed) Democracy,
Governance and Growth, (Ann Arbor: The University of Michigan Press), pp. 115-135.
Persson, T. and Tabellini, G. (2003) The Economic Effects of Constitutions (Cambridge, MA.:
The MIT Press).
Przeworski, A. and Limongi, F. (1993) Political Regimes and Economic Growth. The Journal
of Economic Perspectives, 7(3), pp. 51-69.
Przeworski A. and Limongi F. (1997) Modernization: Theories and Facts. World Politics,
49(2), pp. 155-183.
Przeworski A., Alvarez, M.E., Cheibub, J.A. and Limongi, F. (2000) Democracy and
Development: Political Institutions and Well-being in the World, 1950-1990 (New York:
Cambridge University Press).
Robinson, J.A. (1998) Theories of “Bad Policy”. Policy reform, 2(1), pp. 1-46.
Romer, P. (1990) Endogenous Technological Change. Journal of Political Economy, 98(5),
Scully G.W. (1991) Rights, Equity and Economic Efficiency. Public Choice, 68(1-3), pp.
Schaffer, M.E. (2010) xtivreg2: Stata module to perform extended IV/2SLS, GMM and
AC/HAC, LIML and k-class regression for panel data models.
Smith, P.H. (2005) Democracy in Latin America: Political Change in Comparative
Perspective (Oxford: Oxford university Press).
Starr, H. (1991) Democratic Dominoes: Diffusion Approaches to the Spread of Democracy in
the International System. The Journal of Conflict Resolution, 35(2), pp. 356-381.
The PRS Group (2009) International Country Risk Guide: Researcher Dataset.
Torstensson, J. (1994) Property Rights and Economic Growth: An Empirical Study. Kyklos,
47(2), pp. 231-247.
Weimer, D.L. (1997) The political economy of property rights, in: D.L. Weimer (ed) The
Political Economy of Property Rights, (Cambridge: Cambridge University Press), pp. 1-19.
World Bank (n.d.) World Development Indicators: Online Database.
Wrong, M. (2000) In the Footsteps of Mr Kurtz (London: Fourth Estate).
Wu, W. and Davis, O.A. (1999) The two freedoms, economic growth and development: An
empirical study. Public Choice, 100(1-2), pp. 39-64.
Table 1: First stage results (2SLS); Polity-index as regressand
RE, no year-
FE, no year-
RE, with year-
FE, with year-
Ln regime duration
Ln GDP pc
Eastern Eur. + Sov.
Mid. East + N.A.
The table shows b-coefficients from the 2SLS-models’ first stage regressions, with t-values in parenthesis. ***
indicates p<0.01, ** indicates p<0.05 and * indicates p<0.10. Top row indicates whether Fixed- or Random
Effects are used, and whether year-dummies are controlled for.
Table 2: Second stage results (2SLS); ICRGPROP as dependent variable
RE, no year-
FE, no year-
RE, with year-
FE, with year-
Ln regime duration
Ln GDP pc
Eastern Eur. + Sov.
Mid. East + N.A.
Wald χ2/F (Pr.> χ2/F)
The table shows b-coefficients from the 2SLS-models’ second stage regressions, with t-values in parenthesis.
*** indicates p<0.01, ** indicates p<0.05 and * indicates p<0.10. Top row indicates whether Fixed- or Random
Effects are used, and whether year-dummies are controlled for.
Table 3: Robustness checks, selected 2SLS-models (second stage).
Ln reg. dur
Ln GDP pc
The table shows b-coefficients from the 2SLS-models’ second stage regressions, with t-values in parenthesis.
*** indicates p<0.01, ** indicates p<0.05 and * indicates p<0.10. All regressions are Fixed Effects versions of
2SLS, without year-dummies. The first three models (from left) use ICRGPROP as dependent variable (DV).
The first model excludes Eastern European and ex-Soviet countries, and the two next use different independent
variable lags. The four last use different DVs, as indicated in top row.
See Weimer (1997: 3-4) for a good discussion on the definition of property rights.
One exception when it comes to dealing with endogeneity is Feng (2003), but he investigates the relationship
between democracy and “economic freedoms” more broadly. He relies on Granger-causality tests, rather than
2SLS. Feng finds that most of the partial correlation between democracy and economic freedom is due to
democracy causing economic freedom.
The correlation between PI and the reverse wave instrument (between 1984 and 2004) was .33, whereas the
wave instrument correlated .20 with PI.
The US regime, categorised to have started before 1827 by Polity, is scored a 0, given its explicit categorisation
in the first wave by Huntington (1991: 16).
See https://www.prsgroup.com/ICRG_Methodology.aspx for a description of methodology.
Also Clague et al. (2003) normalise the different ICRG-indexes before constructing a composite. However,
these authors add corruption and bureaucratic quality measures, which arguably capture other phenomena than
property rights protection.
Measuring de facto property rights protection is difficult without relying on subjective evaluation from
qualified judges, which is ICRG’s approach. Unsystematic measurement errors are however mitigated by 2SLS
Results are available on request.
F(1,2354)=5.80. This test is performed with the dmexogxt-module in STATA.
Hausman-tests on differences between FE and RE showed diverging results. For the models without year-
dummies the hypothesis that the (consistent) FE model’s coefficients are equal to the (efficient) RE’s is rejected
at the 0.01%-level, thus indicating we should use FE. However, when including year-dummies, we cannot reject
the hypothesis of equality at the 10%-level, indicating we should opt for RE.
These tests were conducted with the xtivreg2-package (Schaffer, 2010) in STATA.
When using longer lags than 10-years on WAVE and PI, the picture is somewhat more mixed. But, most
overidentification tests still indicate the exclusion restriction is valid, especially in models with year-dummies
All models using two instruments yield a positive effect from democracy on property rights, significant at least
at the 1%-level.
I also substituted the PI with an average of Freedom House’s Political Rights and Civil Liberties indexes, and
with the dichotomous ACLP-measure for democracy (Przeworski et al., 2000). The results are qualitatively
similar in models without year-dummies, with significant effects at least at the 5%-level. However, the p-values
were generally slightly higher for models with year-dummies. Although all models generated a large, positive
estimated effect from democracy, not all effects were significant at the 5%-level.
All results from the robustness checks are available on request.