The Diverging Economic Dynamics of Dictatorships:
Enlightened Rule and Tyranny?1
Draft Working Paper, 2008. Department of Political
Science, University of Oslo
Carl Henrik Knutsen
Abstract
This paper analyzes the question of why dictatorships diverge so much in their economic
outcomes. First, the hypothesis of large economic divergence among dictatorships is
investigated empirically, by looking at GDP per capita growth rates, and is found to be
strongly supported by the data. Then, a theoretical model is constructed and analyzed, and the
model points to some crucial mechanisms that might explain the result. Rational autocrats
motivated by personal power engage in widely different strategies and conduct very different
economic policies in different contexts. The central explanatory variable is the nature of the
security threat facing the dictator and his regime. If the security threat is mainly external, for
example a neighboring country and its army, the dictator will have incentives to conduct
growth- and development enhancing policies. If the nature of the security threat is mainly
internal, for example a rebel movement or a democracy movement, the dictator can have
strong incentives to hinder economic development. The model also makes other predictions,
like for example that a vast amount of natural resources in a country reduces the dictator’s
incentives to develop the economy. Some statistical evidence and historical examples are
presented to illuminate the relevance of the model. In particular, the history of how the
Kuomintang switched from a predatory strategy when fighting the Communists in the Chinese
civil war to a developmentalist strategy once it had established itself on Taiwan is treated.
This historical “quasi-experiment”, involving the same actors in two different specified
contexts, gives strong support to the validity of the model presented in the paper.
1
Thanks to Gyda Marås Sindre, Kristian Helland-Hansen and Øivind Bratberg for comments to an earlier draft.
A special thanks to Nils Olve Gillund for comments, and for rewriting my equations to a more aesthetic and
readable format.
1
1. Introduction
In “Politics”, Aristotle wrote that the best possible form of government under ideal conditions
was enlightened Monarchy, with the benevolent and knowledgeable monarch at the helm
(Aristotle, 2000). Aristotle also describes how monarchy easily slides into a form of
government he labeled “Tyranny”, for example when the Monarchy faces the unavoidable
event of monarchic succession, with the death of the old king. Therefore, Aristotle concluded
that other and more “balanced” forms of government like his “Politeia” would be more robust
in the sense that they provide decent rule under different contexts. Aristotle based his analysis
on observations of ancient Greek city states, but his analysis could perhaps also be used
fruitfully in order to explain trajectories in the more recent history of dictatorial nation states.
One key insight is how concentration of power in dictatorships brings about very divergent
outcomes in different empirical settings. Aristotle however focused on the virtues of rulers. In
this paper, I will show that even if all rulers are self-interested and motivated by power, some
autocratic states will prosper economically, while others will face developmental disasters.
The key claim is that rulers choose very different policies in different contexts, and what
seems like “Enlightened Monarchs” to the outside observer might very well be self-interested
dictators, who due to accident do things that have “good” consequences for their subjects.
Dictatorships exhibit far more variation in their empirical performances than democratic
regimes (Rodrik, 2000). You have the present-day Chinas and the Singapores; the growth
miracles that make the Lee-thesis (on the necessity of authoritarian rule for development)
resonate with many academics and politicians. Then you have the North Koreas and the
Zaïres; primary examples for the hypothetical course: “How to destroy your economy 101”.
Adam Przeworski (2007) claims that “[W]e currently do not do a good job distinguishing one
dictatorship from the next” and that “[D]ictatorships are by far the most understudied area in
comparative politics. We need to start thinking about them”. As Robert Barro puts it, it is
clear that we have two types of dictatorships when it comes to how they produce economic
outcomes. However, the necessary theory for explaining why this is so is not yet fully
developed (Barro, 1997:50). This paper seeks to provide such a theory and give explanations
to the puzzle of how and why dictatorship produces so wildly diverging economic results.
Why do some autocratic regimes become growth miracles whereas others turn out as growth
disasters? The underlying assumption is that the actions and policies promoted by dictators
explain a large share of the variation in growth and development performances. Social
scientists studying growth have turned their focus towards institutions as a primary mover of
developmental outcomes, and this is obviously important also for understanding how
dictatorships work (Przeworski, 2007). However, institutions are endogenous, and can be
affected by the actions taken by rulers and other important actors in a country, and this insight
is crucial in the following paper. In this paper, rulers are portrayed as motivated mainly by the
concern of staying in power. This motivation leads to very different strategies in different
contexts, and thereby affects the overall economy in extremely different ways.
2
2. The modern empirical record: Dictatorship, democracy and
variation in outcomes.
Before explaining the phenomenon of large economic variation among dictatorships, we have
to set the record straight in terms of examining whether relatively autocratic regimes actually
have had more disparate economic outcomes than more democratic regimes in general.
Anecdotal evidence can be provided in abundance: Think of modern-day China compared
with China under the Great Leap forward. Think of the Taiwanese and South Korean
experiences and compare them with that of Pol Pot’s Cambodia or North Korea under Kim
Jong Il. According to Mueller, “[T]he greatest advantage of democracy over dictatorship may
not be that democracies outperform dictatorships on average, but that democracies seldom
sink to the depth of misery that one too often observes under dictators” (Mueller, 2003: 425).
Data on growth rates in GDP per capita (1970-2000) from the Penn World Tables (PWT)
show that dictatorial countries figured heavily both among the best and worst performances.
Table 1 shows that 11 out of the 15 best performances had an average score on the Freedom
House Index (FHI) above 3,5, and the same was true for 14 of the 15 worst performers.
Dictatorships dominate both among the best and the worst performing countries.
Table 1: Top and bottom fifteen growth performers.
Best performers
Country
Taiwan
Singapore
Botswana
South Korea
China
Thailand
Cyprus
Ireland
Mauritius
Haiti
Indonesia
Malaysia
Cape Verde
Seychelles
Romania
Average
annual growth
1970-2000
6,7
6,4
6,3
6,1
5,2
4,7
4,4
4,4
4,3
4,3
4,2
4,2
4,0
3,7
3,7
Worst performers
Average FHI
1972-2000
4,1
4,7
2,3
3,7
6,6
3,6
1,9
1,1
2,0
5,9
5,3
4,0
4,1
4,7
5,3
Country
Zaïre/Congo
Angola
Central Afr Rep
Nicaragua
Sierra Leonne
Niger
Mozambique
Comoros
Venezuela
Zambia
Togo
Madagascar
Mauritania
Nigeria
Cote d´ Ivoire
Average
annual growth
1970-2000
-4,8
-2,6
-2,6
-2,4
-1,8
-1,5
-1,5
-1,4
-1,4
-1,3
-1,2
-1,1
-0,6
-0,6
-0,6
Average FHI
1972-2000
6,4
6,6
5,6
4,3
5,3
5,8
5,6
4,6
2,1
4,7
6,0
4,4
6,2
4,9
5,4
If we now look into the variances in growth performances among all recorded autocratic and
democratic regimes on a decade-wise basis, there is reasonable evidence for the hypothesis
that there is more variation among dictatorships than among democracies. Tables 3, 4 and 5,
which are taken from Knutsen (2006), show the variance in national growth rates in different
categories of regimes along the democracy-dictatorship continuum, based on the FHI. When
using the dichotomous classification (classification-criteria are given in the table), F-tests
(two-sided) show that the discrepancies in variances between the groups are significantly
3
different from zero at the 5% level for all the decades, and it is significant even at the 0,1%level in the 1970’s and 1990’s. When comparing the two extreme cases from the trichotomous
classification, the p-value for the 1980’s statistic is 0,051, just above the 5% significance
level, whereas the divergence in variance between the groups is significantly different from
zero at the 1%-level for the 1970’s and 0,1%-level for the 1990’s. These tests provide
convincing support for the hypothesis that dictatorships vary more in their economic growth
performances than do democracies.
Table 2: Degree of democracy and variation in PWT growth rates in the 1970’s
Average
FHI
[1, 4)
[4, 7]
Number
45
68
Variance
growth
3,40
8,92
Average
FHI
[1, 3)
[3, 5]
(5, 7]
Number
35
31
47
Variance
growth
3,95
6,45
8,81
Table 3: Degree of democracy and variation in PWT-growth rates in the 1980’s
Average
FHI
[1, 4)
[4, 7]
Number
61
61
Variance
growth
4,81
8,09
Average
FHI
[1, 3)
[3, 5]
(5, 7]
Number
47
29
46
Variance
growth
4,41
6,98
7,15
Table 4: Degree of democracy and variation in PWT-growth rates in the 1990’s
Average
FHI
[1, 4)
[4, 7]
Number
88
56
Variance
growth
4,38
15,00
Average
FHI
[1, 3)
[3, 5]
(5, 7]
Number
66
46
32
Variance
growth
3,42
5,92
22,34
However, the results above might be driven by omitted variables that impact both the
probability of having a specific regime type and the variation in growth performances. One
such factor could be for example the prior level of development in a country. I therefore
propose a better test methodologically for investigating the hypothesis related to variation; a
so-called Goldfeld-Quandt test of heteroskedasticity (Greene, 2003:223). I want to look at
how observations are spread along a regression line, derived from a regression model
proposed in Knutsen (2008). The regression equation is given below, and the independent
variables are PPP-adjusted GDP per capita, log of population, dummies for plurality religion,
geographic region and colonizer, as well as the log of regime duration (+1), share of
population in urban areas, energy production/GDP and a linear time trend:
10
16
s 4
t 11
groi    1demi   2 gdci  3 ln  popi     s colsi  t regti
24
  u regui   25enei   26 ln  duri  1   27timi   28urbi   i
u 17
4
The dependent variable, namely growth in GDP per capita, is now operationalized by using
data from the World Development Indicators, and the data are from 1972 to 2005. I run OLS
with Panel Corrected Standard Errors, with country-year as unit of analysis. The calculation
procedure of the standard errors takes into account autocorrelation (AR1), contemporaneous
correlation and heteroskedasticity between panels. I then subtract the predicted values from
the actual values of economic growth per capita, and square this difference, to obtain the
squared residuals. I then perform the Goldfeld-Quandt test to see if there is reason to reject a
H0 of homoskedasticity when comparing the subgroups of observations classified after regime
type. First I check the dichotomous classification used above, categorizing all observations
with FHI larger than or equal to four as dictatorships, and the rest as democracies. The Fvalue is 2,54 (1390, 1572 degrees of freedom), and the hypothesis of equal variation can be
rejected even at the 0,01%-level. I then compared the group with FHI above or equal to 5,5 on
the FHI with those observations that had FHI less than or equal to 2,5 on the FHI. The F-value
(789,1214 df) is 3,52, which is even higher than the F-value for the value based on the
dichotomous classification, and the test shows that the hypothesis of equal variation can be
rejected at the 0,01%-level2.
From this analysis, there is good reason to reject the claim that there is similar spread along
the regression line for both relatively democratic and relatively autocratic regimes. The
evidence points in favor of the hypothesis presented above, namely that there is systematically
higher variation among dictatorships, and this result holds even if we control for a host of
other factors. The analysis points out that not only do dictatorships grow significantly less on
average than democracies according to the model (Knutsen, 2008), but autocratic regimes are
also scattered more widely around their model-predicted values; they show much more
variation. This is the empirical fact that will be sought explained in this paper.
3. Concentration of power in dictatorships
It is hard to argue with Overland et al. in their statement that “[G]ood policy analysis should
acknowledge the realities of dictatorships” (Overland et al., 2000:2). What is it about
dictatorship that generates such vastly divergent economic outcomes? Some would
automatically respond that it is the broadness of the concept and the lack of nuances in
describing all non-democratic regimes as dictatorial, and to a certain extent this insight is a
valid one. Dictatorships come in several forms or “types”, from absolute monarchies to oneparty states to military regimes (Hadenius and Teorell, 2007). One interesting classification
comes from Linz and Stepan (1996), who distinguishes between authoritarian, totalitarian,
post-totalitarian and sultanistic regimes. These regimes might be institutionalized in widely
different ways, and their political processes are driven by relatively different “logics”.
2
Such a test, when based on country-years as unit of analysis, takes into account both between-country and
within-country (temporal) variation. There is reason to believe that dictatorships both have a larger betweencountry variance in growth performances and also larger variance in national growth performances over time
(Rodrik, 2000).
5
Nevertheless, there are common denominators, and lack of general popular control over
public decision making and lack of political equality are two of them (Beetham, 1999).
Dictatorships are then often associated with a concentration of political power, when
compared to democracies, even if the degree of power concentration varies also within the
group of autocratic regimes. Often, in practice, dictatorship is associated with a lack of
horizontal power dispersing institutional structures, like checks and balances. Additionally,
the vertical accountability links, in the form of free and fair elections and freedom of speech
and media, between populace and political elites are lacking. Political rights and civil liberties
are often either weakly or selectively protected under autocratic rule.
It is among others in the areas of power concentration and lacking political accountability we
have to look when seeking to understand the diverging economic dynamics in different
dictatorships. When power is concentrated and accountability is lacking, the scope of possible
policies the political elites can follow, if they wish, is much larger than in democracies, where
power dispersion and institutional structures like free and fair elections put constraints on the
possible policy choices of government. An autocrat bent on producing a high rate of economic
growth, can with fewer political constraints than in a democracy, drive through tough
economic reforms without fearing popular backlash in the next election and push for a high
national savings rate through restricting private and public consumption. However, autocrats
can also rampage the economy by stealing, looting and even killing off the human capital (as
in Cambodia), without being thrown out of office or the palace. This very general point of
power concentration increasing the policy scope is one of the key explanations for the high
variation in autocratic economic performance. I will elaborate on the argument and specify
potential mechanisms in the rest of the paper.
Institutions matter, but so do actors
I will in this study focus less on institutional structure and divergence in autocratic “types”
than what is perhaps appropriate. There is no doubt that institutional structures have important
real world effects, and that a complete understanding of the functioning and economic effects
of dictatorships would need to entail this insight. However, I offer a very general theory on
dictatorship and economic effects, where power concentration and survival strategies are core
concepts. There are differences between different types of autocratic regimes, especially when
it comes to the first. Some dictatorships are more personalized and some are more
institutionalized with for example the party playing a crucial role. In the latter case, it is
perhaps most appropriate to view the party, rather than the leading individual, as the actor in
the model that will be presented. Intra-party dealings and their effects will then not be
explained by the theory. However, it is wrong to believe in complete institutional
determinism. The central actors in power can build, reshape or restructure institutions, at least
in the long run, and my theory, as I will explain later, is best viewed as a theory of the long
run. Haber (2006) claims that the “logic of organizational proliferation” is one of the three
main survival strategies of autocrats, and he shows how building organizations that counter
the influence and powers of existing ones can enhance the survival of the autocrat. It is
however probably easier to crush the influence of a bureaucratic apparatus than to build a
6
well-functioning one, and the “privatization” of states in Africa has been recognized in the
Africanist literature, even though the bureaucratic structures that existed from the colonial-era
were often little to cheer about. As Evans and Rauch (1999) show, a well-functioning and
autonomous bureaucracy is an important factor in promoting economic development, but
bureaucratic structures are not exogenously given (think about the differences between
present day South and North Korea and how they diverge on this point), they can be
restructured and strengthened or weakened (for example by refusing to pay high wages or
assigning positions to the dictator’s closest friends) by the political elite in charge. My claim
is that due to power concentration in dictatorships, institutional structures are not as difficult
to change for dictators as they are for democratic government, but this does of course not
mean that institutional structures are plastic in all dictatorships. In the model presented below,
we will see that rational dictators engage in different survival strategies, including the shaping
and reshaping of institutions as well as in more concrete promotion of specific economic
policies. My argument is to a certain extent therefore classifiable as what Paul Pierson (2004)
has labeled “actor-centered functionalism”: An actor sets up or alters (or destroys) an
institution because the effects of the institutional change are viewed as beneficial for the actor.
As Pierson notes, this requires rational behavior and a good understanding of how the
institution works, and also a relatively long time horizon (which fits well with my claim that
the model presented below is a long run model).
4. Previous research on rational dictators and the economy
Political economic research has generally been focused towards explaining the interrelations
between politics and the economy in democracies. There are however credible exceptions,
both among less formalized studies in political science and among more formalized studies
from political scientists and economists. I will focus on the more formalized literature here.
Mancur Olson (2003) asks what would make a dictator provide a relatively stable and
functioning property rights framework, thereby producing economic development in a
country. A stable autocratic ruler, or a “Stationary Bandit”, can be viewed as the “owner” of a
country, and such an owner, when compared to a group of “Roving Bandits” not ruling the
society but nevertheless able to extract resources, would have an “incentive to make his
property productive” (Olson, 2003:115). Stealing, looting and extracting from citizens will
strongly affect the subjects’ incentives to produce, and therefore ultimately reduce the
resource-base from which the owner can extract: “The rational stationary bandit will take only
a part of income in taxes, because he will be able to exact a larger total amount of income
from his subject if he leaves them with an incentive to generate income that he can tax”
(Olson, 2003:118). One of the main insights from Olson’s study is the importance of the time
horizon of an autocratic ruler when it comes to incentives for protecting a property rights
framework in an economy. The main conclusion is that dictators will be much more likely to
refrain from activities such as confiscation, repudiation of contracts and hyperinflation when
they expect to be in power for a long time. The same reasoning applies to why dynasties that
are expected to last after the present ruler dies or abdicates may abstain from these activities.
7
Long term investments of different varieties are necessary for an economy to prosper.
Therefore “an autocrat who is taking a long view will try to convince his subjects that their
assets will be permanently protected not only from theft by others but also from expropriation
by the autocrat himself. If his subjects fear expropriation, they will invest less, and in the long
run his tax collections will be reduced” (Olson, 2003:124). According to Olson, dictators that
expect to be ousted from power in the near future face the same incentives as the “Roving
Bandit”, and they will loot, steal and expropriate accordingly.
Let us now take a look at some research that endogenizes the survival probability of autocrats,
by linking it to their political strategies and actions. Overland et al. (2000) constructs a model
where “[T]he determination of dictatorial type is endogenous” (Overland et al., 2000:3). This
model incorporates an underlying assumption that breakdown of a dictatorship is more likely
in times of economic crisis; that is, when for example economic growth rates have declined
over the last couple of years. A relatively unstable dictator would therefore generate a high
level of economic growth in order to survive. Internal opposition is bought off with good
economic policies. The main argument is that “since growth stabilizes the dictator’s position,
a strong fear of losing power is a positive factor for growth” (Overland et al., 2000:2). The
rational dictator will then choose to push economic growth-conducive policies in order to
promote growth. There are however some conditions that need to be satisfied in the model for
this to be true. The dictator in this model is maximizing consumption, and very unstable
(relatively certain that they will be thrown out anyway) or very shortsighted dictators will loot
and plunder until they are thrown out of office. Only dictators with a decent probability of
survival will choose to promote growth. But in this case, they will according to the model
push growth rates even higher than what is socially optimal (remember that a 100%
investment rate might skyrocket growth, but lead to zero consumption). The argument that
dictators would like to promote growth in order to stabilize their regime, and avoid coups or
revolutions, run contrary to the argument presented in my model and also that of Robinson
(2001). Mueller (2003:422) explicitly sets up Overland et al’s model and Robinson’s model as
opposite to each other in terms of their predictions for how a dictator would act in order to
stabilize his position. These diverging dynamics, as I will come back to and explain in section
8, might actually be the result of lack of specification of the time dimension.
The model that comes closest to the one presented later in this article is the one presented by
James Robinson (2001), and a related article from three years earlier (Robinson, 1998). Like
in my model, Robinson focuses on the “fact that development and changes in political power
are closely related”. This is the one of the key reasons for why “bad policies” are often
selected by dictators. Robinson claims that “while [capital] accumulation may increase total
income, it may induce institutional transition which is unfavorable to the autocrat. If a dictator
loses political power then he does not gain from development and will oppose it. Thus a
dictator may wish to slow accumulation” (Robinson, 1998:24). In Robinson’s model, dictators
want to accumulate wealth also in the future, and therefore do not push short term policies
that might strongly decrease their probabilities of being rulers also in the future. The survival
8
probability of an autocrat is endogenous to the policies selected. A rational autocrat
recognizes this fact, and acts according to his self-interest. Robinson’s model is relatively
advanced technically with dictators optimizing over an infinite time horizon (whereas I focus
on two periods), and he conducts a splendid and rigorous analysis of the implications of his
model. Some of the main results from his model are that “inefficient underinvestment
(predatory behavior) tends to arise in societies where, (1) there are large [material] benefits to
holding political power, and which are, (2) well endowed with natural resources, (3) badly
endowed with factors which are complementary to public investment, such as human capital,
and (4) intrinsically unstable” (Robinson, 2001: 1).
The autocrats in the models above are maximizing a utility function where consumption is the
argument, and this is the case also for most of the other earlier formal analyses on selfinterested autocrats. The issues of power and tenure in office do however often come sneaking
in the backdoor. Consider the analysis conducted by James Robinson (2001): In order to
continue gaining wealth from occupying office, the rational dictator wants to increase the
probability of staying in power, and takes this into account when formulating strategies. I
agree with Robinson that the endogeneity of regime survival is extremely important to
recognize in order to get an understanding of why some autocrats have consistently chosen
policies that are obviously detrimental to economic development. However, my model,
presented below, diverges from Robinson on the objective function of the autocrat, since I
focus on power rather than consumption as the main objective, and since I add that not only
the internal opposition might prove to be a security threat to the autocrat, but that foreign
countries might also pose a threat. This has bearings for the optimal policy, as we will see
later3. As Wintrobe (1990) argues, the objectives of dictators are different empirically, and
there are several historical indications that power in itself is important to many dictators and
autocratic elites. This is why I believe the argument of this paper and the model presented in
section 7, contribute to a fuller understanding of what drives autocratic economic policies, and
subsequent economic outcomes. Let me elaborate further on the psychological motivation of
dictators.
3
Nevertheless, the mechanisms of the model presented below often resemble those found in Robinson’s
model, but some interesting nuances might be made when using the power-oriented specification. Let me
conduct some premature analysis: One possible divergence stemming from modeling assumptions might come
to the forefront in the case of extremely shortsighted autocrats or autocrats who are certain to lose power. In
the case of a “Robinsonian” consumption-oriented autocrat, he might be interested in conducting actions that
are conducive to short term monetary gain, but that further reduces the chance of staying in power. A poweroriented autocrat will still stick to the strategy that increases the probability of staying in office in the next
period, even if he is most likely to be ousted in any case.
9
5. The psychological motivation of the dictator
What motivates politicians? Unlike what is the case in producer- or consumer theory in
microeconomics, there is no consensually agreed upon single objective in the theorizing on
the objectives of political elites. Persson and Tabellini’s (2003) sum up three potential
motivational aspects: Politicians are stylistically assumed to be either motivated by holding
office, by seeking economic rent or alternatively by pushing specific policies they hold dear.
Translated, politicians can be motivated by power, money or promoting their ideologies. We
might suggest some additional objectives: fortune and fame, sex and the promotion of the
interests of a specific group in society. Many political scientists and political psychologists
like to claim that power in itself is a strong motivational factor for many people in general and
political leaders in particular. When I say in particular, one mechanism that might back up this
assumption is that people motivated by power might self-select into politics, one of the areas
where power over others can be exercised. Empirical studies by political psychologists have
indicated that politicians in democracies are often motivated by “power for its own sake”
(Winter, 2003:153). Winter (2003) classifies American presidents after their main
psychological motivation, dividing presidents into three classes after their main motivation:
power, achievement and affiliation. He finds that 20 out of the 36 former presidents he
investigates have to be considered as mainly “power-motivated”. There is no reason to believe
that power is less of a motivational factor for autocratic rulers than for democratic, rather to
the contrary. Another main reason for focusing on power as a motivational force in
autocracies is that all of the other potential objectives sketched out above (money (m), fame
(f), sex (s), interest group promotion (i) and realization of ideological vision (v)) are
themselves functions of power (p), and in most practical situations they depend positively on
the amount of power held. This assumption pinned down in a ruler’s utility function takes the
following form:
U ruler  U  p, m  p  , f  p  , s  p  , i  p  , v  p  
U '  x   0  x :  p, m, f , s, i, v
z '  p   0  z : m, f , s, i, v
Of course, different rulers have different “utility functions”, and the degree to which power
promote the other motivational aspects depend on context, but I will in this paper focus on a
simplified version of the utility function above, namely: Uruler = U(p) with U’(p)>0.
Wintrobe (1990, 1998) has done some excellent work on classifying different dictators after
their motivational structure, and has arrived at several labels. One good distinction is the one
between consumption-maximizing autocrats, whom he labels “tinpot dictators”, and those
who want to maximize their influence and power over society, which are the “totalitarians”.
Wintrobe’s point is clearly well taken, and assuming that power is the general main objective
is a simplification. However, I contend that power-motivation is central enough in many
instances to legitimize my choice of utility function. Additionally, we saw in section 4 that the
10
consumption-motivated autocrat has already been analyzed to a greater extent in the literature
than the power-motivated.
By “keeping power” in this paper, I often really refer to “staying in office”. This is surely not
a precise usage of terms, but an underlying assumption in this paper is that continuing in
office enables the dictator to exercise “power” over the populace. There is a large debate on
how to conceptualize power, and I will not go deep into the issue here. The classical account
of power was given by Robert Dahl (1957), who defined one actor’s power over another as an
actor A’s capacity to make another actor, B, do something that is against B’s initial will.
Steven Lukes (1986) discusses several conceptualizations of power, and Dahl’s understanding
of power is often referred to as the first dimension of power. The second dimension deals with
the ability of an actor to set the agenda when it comes to which topics are discussed and dealt
with, and which are not. The third dimension of power is conceived by Lukes and others as
the ability to form the preferences and ideas of other actors. I will claim that holding office is
closely correlated with all these three aspects of power, and it excuses my sloppy use of term.
To sum up: In my analysis below, I will assume that dictators are motivated by gaining power
and keeping power by staying in office, and that they thereby design their policies and
political strategies in a way that enhances their probability of survival in office. Continuation
in office is assumed to be the main general power-tool. Whether continuation in office or
power more broadly is the objective entering the utility function is therefore not important to
the following analysis, but it is important to note that they are analytically separable.
6. The qualitative argument: Why dictatorships diverge in
economic outcomes
In sections 3 and 5, I established two crucial assumptions for the analysis to come:
1) There are fewer political restraints on ruling elites and government in autocratic states
than in democratic.
2) Political leaders are motivated by power, through staying in office, and design their
policies and political strategies in order to attain this goal.
The third crucial moment of the analysis is to link these assumptions to the underlying
empirical observation of large variation in outcomes among dictatorships. I will now state the
third crucial assumption of the argument:
3) There are often several relatively different strategies that can promote regime survival and
the character of the optimal strategy or strategies depend highly on social, political and
economic context.
For explanatory purposes, is not sufficient to say that there are a multitude of different
strategies available for power promotion and preservation and that the optimal strategy
depends on context. We need to link particular contextual settings to particular strategies in a
systematic fashion. I will look particularly at the importance of the political context, among
11
others the type of security threat to the regime. The main claim made from the analysis related
to the model is that dictators who mainly face external security threats conduct
developmentalist strategies, because their perceived effects are to bolster the regime against
these threats. Predatory, or non-developmentalist, strategies will be followed by leaders who
face mainly internal security threats, because development is believed to increase these
threats, at least in the long run. Rulers who face both internal and external security threats will
have to balance these threats by following an “intermediary” economic strategy.
The final link in our explanation of the empirical traits in section 2 is that different policies
and strategies from the ruler’s side indeed have widely different effects on the level and
growth of development in a country (as perceived by the rulers). These effects from political
strategies and policies on the economy also run through how economic institutions like for
example property rights systems are shaped. “Economic institutions, and their enforcement
and refinement, do not emerge from thin air… Rather they are the result of both interest group
demands and the specific features of decision making in the polity” (Haber, 2000:10).
4) Different policies and political strategies have very different economic effects.
7. A formal model
“The dictator … must live in constant fear that some general, or corporal, or alienated
academician will set a train of events into motion that will result in the dictator’s downfall.
This uncertainty leads the dictator to maximize an objective function that is dependent on the
utilities of the citizens. It differs most from that of a party in a democracy in that the weights
it places on the utilities of some group may effectively be negative.” (Mueller, 2003: 416).
The model presented here takes this insight by Mueller into account. In certain contexts,
dictators might want to harm the internal opposing forces, for example by reducing their
resources. This could be made through reducing for example public investments, or engage in
other strategies that decrease the overall production in the country. However, dictators might
also have to keep an eye on foreign security threats. If these are imminent, the dictator would
like to have a modern and capable army to fight them off, should there be an invasion. These
two effects work in different directions, in terms of the incentives they place for the rational
and power-motivated dictator to engage in development-conducive policies. The model below
tries to capture both these effects into one framework, and the subsequent analysis gives one
explanation to why the economic development experiences of different autocratic countries
have been so diverging.
I propose a model with two time periods. In my set-up, I will have three actors: The dictator
(and his inner clique), D, the opposition in the country, O, and a government of a foreign
country, S. We will not at this point specify whether the opposition will impose democracy or
autocratic rule if gaining power (we are not specifying whether we are dealing with a broad
popular movement or another contending political elite or rebel group). The dictator has the
utility function described in section 5, which I simplified to U(p), with U’(p)>0. The dictator
is only motivated by keeping power through holding office (in the second period). The
12
opposition and the foreign power are also motivated by taking over power in the country, but
they are sensitive to the costs of fighting for power, c, (for example in monetary terms) with
the current domestic regime:
dU
 0 and
dp
dU
with
 0 and
dp
U s  U  p, c  with
U o  U  p, c 
dU
0
dc
dU
0
dc
It is reasonable to assume that the utility increase of gaining power in the country is higher for
the domestic rebellion than for the foreign government. Although conquests of foreign
territories have often been cherished by occupants over the course of history, the desire to
control one’s own country is probably stronger. As Machiavelli (1999) noted, holding on to
foreign territory is costly and difficult business. Machiavelli’s advice to the Prince was that he
should either move to the occupied territory or demolish its cities to the ground. Let us build
this into the model through c, and for simplicity assume that the cost of contending power for
the internal opposition is 0, so that we can drop this argument from O’s utility function. This
is clearly a simplification, but we now assume that O, as D, is only maximizing the
probability of holding office in period 2 as well.
Let us now model the economy. We assume that industrial output, Y, is a function of public
capital, g. We have:
Y  Y  g  with Y'  g   0 and Y ''  g   0
It is D who invests in public capital. D can lend at an interest free rate to invest in as much
capital as he likes, given that he can repay it after his revenue has been realized. The
economy does also posit some natural resources, R, and these can be tapped directly by D
without any further investments. The opposition, O, cannot obtain any income from the
natural resources. The total output of the economy is therefore:
W Y R
Let us now look at the distribution of output. Let us say that the rulers appropriate a fixed, and
exogenous, share of industrial output, equal to α. The net income for the ruling elite is then:
Id  Y  R  g
The net income for the opposition is:
I o  1    Y
The foreign government has an income which is equal to a fixed share, β, of the assumed
fixed foreign output Y*:
13
Is  Y *
All the actors can convert their resources into fighting capabilities, given specified
technologies of transforming income to arms and fighting power.
Fd  Fd  I d 
Fo  Fo  I o 
Fs  Fs   I s 
How much of the income that will be generated into fighting capacity depends on the utility
function of actors, more specifically their relative evaluation of consumption and having or
gaining power. In accordance with the discussion of the utility functions, we assume that the
internal opposition and the dictator transform all their income into fighting capacity. This is
not necessarily the case for the foreign government. Again, for the sake of simplicity, we
assume that the foreign government allocates a constant share of its income, µ, to military
expenditures.
As discussed above, power considerations are related to relative resource- and income
distribution. The probability of any of the groups fighting for and winning power depends on
the relative distribution of fighting power among the actors. This again depends on relative
income between D and the other actors and on the F-functions. The probabilities of the
opposition winning, Po, and the probability of the foreign government winning, Ps, are given
by the two equations below:
  where 0  P  1 P'    0
P  P   where 0  P  1 P'    0
Po  Po
s
s
Fo
Fd
o
Fo
Fd
Fs
Fd
s
Fs
Fd
We assume that the two contending actors cannot form a military coalition, and further that
the probability of one of the contending actors winning is independent of the other actors’
actions. This is again a clear simplification (ask Hitler about the difficulties of two-front war).
R, Y* and the F-functions are exogenously given. We then have a game between the three
actors, with D moving first, setting his economic policies, g. Y is then realized. We then get
the three Is, which are subsequently transformed into fighting power, the Fs. O and S then
decide on whether they should contend power or not. The domestic opposition, O, will always
contend power, but the foreign government, S, will not always wage war, since it takes into
account the cost of war. S weighs the expected utility of war (probability of victory*utility of
holding power in the invaded country) against the costs. The outcome related to who is
holding office in the domestic country in the second period is modeled as a probability
function, where the probability of winning power depends in the above specified way on the
economic resources of the actors. We make some restrictions on the game: Maximally one of
the actors grab power in the second stage. D can then not lose his office to both O and S, even
if he loses to an internal revolt and an external invasion at the same time. In the case of both
14
parties winning power in the last stage, we assume that the parties randomly decide who is
going to realize the victory, for example by a coin-toss.
Figure 1: The timing of events in the model
R, Y*, α, β, µ given. D sets g. Y realized. Is realized. Fs realized. Ps realized. S decides on war. Final outcome
Let us now look at how the probabilities of D losing power to the two different actors depend
on the size of g, and thereby the size of the subsequently realized Y, when we take all the
other variables as given. First, let us have a look at Ps, the probability that the neighboring
country wins power through war. Since we assume that Y* is given, and thereby also Fs, it is
the military capacity of the domestic dictator that determines this probability. An increase in g
increases Y and further Id, and thereby the military capacity of D. This makes it less likely
that D will lose a war, and when the probability becomes sufficiently small, Ps≤ P*, the utility
maximizing foreign government will not wage war since the cost exceeds expected utility.
Figure 2: Probability of losing power to an external threat
Ps
P*
g
When it comes to Po, the specification of the nature of dPo/dg is not straightforward. It partly
relies on the specification of the military capability transformation functions. If the Fs are
concave, that is an increase in military capability is higher for the first dollars used (on stinger
rockets, AK47s and paying soldiers basic wages), and O receives a small fraction of the
national income, we could argue that a higher g gives a higher Po. If the amount of income (1α) received by O increases with Y (industrialization or more general economic diversification
makes it more difficult for the regime to appropriate a large share of the income), then an
increase in g will also lead to an increase in Po, but α is assumed to be exogenous here. A
more technical argument is that Y(g) is concave, whereas the cost of providing more g is
linear. At a high level of g, the returns from an increase in g becomes smaller for D (and also
for O), but the cost for D of providing the extra unit is the same as at low levels of g.
Additionally, if the regime controls a large base of natural resources, an increasing Y will tend
to even out the disparities in welfare between D and O, relatively. All these considerations
15
would lead us to believe that dPo/dg>0: An increase in public investment, with a subsequent
increase in Y will lead to an increased probability of O gaining power in period 2. If O is a
democracy-movement, modernization theory would also point out that an increase in g,
leading to a higher Y, will broaden the movement’s base, and maybe also deepen the
commitment of those involved. Some of the arguments presented, like the concavity of Y(g),
could perhaps also lead us to believe that d2Po/dg2>0.
Figure 3: Probability of losing power to an internal threat
Po
g
One threat at a time; the developmental implications from different security
threats
External threat
If D, faces only an outside threat, D would according to the logic of this model increase g in
order to increase Y, and thereby his fighting capability, to fend of the external threat. D would
at least increase g until he reached P*, given full information, and thereafter the model does
not imply anything about the dictator’s behavior. The subgame perfect Nash-equilibrium of
this game is then that the dictator invests an amount of g that is higher than or equal to g* (the
g that is needed to obtain P*) in the first round4, and that the strategy of the foreign power, S,
is to invade if g<g* and not invade if g ≥g*. S will then not invade in the second stage. If S is
very strong, P* might not be reached. D will then maximize his income in order to later obtain
the maximal amount of fighting power. D’s income is given by:
I d  Y  g   R  g
This gives us the optimality condition:
4
If we introduce some uncertainty into the model, for example related to the military capacity of the foreign
power or on where the exact position of P* is, then the rational and power-motivated dictator would certainly
invest in a higher amount of g than his best estimate of g*, in order to be totally certain that he will stay in
power in the second period.
16
dI d
1
 Y '  g   1  0  Y '  g  
dg

We see that D will set a g such that the marginal increase in production from a unit increase
in g multiplied by the share of the production he gets is equal to the marginal cost of investing
in g, which is here set to 1. This would be the solution to the optimization problem even if
P≤P*, and the dictator’s utility increases in the amount of consumption, even though
consumption is lexically ranked below power as a motivational force. D would then invest his
income in fighting power until P=P*, and consume the rest.
The historical experiences of Japan in the Meiji period provide one interesting illumination of
the argument sketched out above. As will be discussed in depth below, the g-variable in the
model can be given a relatively wide interpretation as a political survival strategy from the
dictator’s side that also influences economic development, and need not strictly relate to
public investment. The so-called “Meiji-restoration” started in 1868, with the removal of the
feudal-like Tokugawa rule. Even though the new regime conducted a political liberalization in
Japan, the rule still has to be considered an oligarchic form of government, semi-authoritarian
at best according to today’s standards. The biggest threat to the new government came not
from within Japan, but from the colonializing West. Japanese leaders feared that Japan would
go the same way as several other Asian states, and become colonialized, or at least subject to
strong political influence from foreign powers. The so-called “unequal treaties” imposed on
Japan by the Western powers were a great cause of concern, but they were feared to be only
the first step of Western domination over Japan. How could Japan avoid such a fate, and the
regime guarantee its continued existence? The answer to the Japanese leaders, including
Prime Minister Ito Hirobumi (Murphey, 2000:308-9), was modernization and economic
development. These leaders perceived that only through economic and technological
development could Japan gain the strength, prestige and military capacity to avoid
colonialization. The leaders “were quick to realize that if Japan was not to become a colony or
semicolony like the rest of Asia, it would have to adopt Western technology… Japan also saw
that military technology could not be separated from overall industrialization” (Murphey,
2000:304-5). Even though historians point to national pride also being a motivational factor, I
would in the light of the model presented above emphasize the nature of the security threat
facing the Meiji-rulers of relatively homogenous Japan: It was mainly external. According to
the logic of my model, this laid the incentives for rulers to modernize in order to scare off
possible foreign colonizers or intruders in general. The different concrete strategies followed
by the Japanese political elite were complex, and ranged from promoting industrialization
through different means to enhancing technological diffusion through openness to building
competent and well-functioning political and military institutions. The strategies succeeded in
achieving their aim as Japan remained independent.
Internal threat
What now about the case where there is no foreign power threatening the dictator, D, and
there is only a domestic opposition, O? As mentioned above, the important factor here is the
17
nature of dPo/dg. If it is non-monotonic, we would have to make a differentiation and find the
minimum point by setting dPo/dg=0, and check for appropriate second order conditions. D
will find the amount of g that minimizes his probability of losing an internal conflict with the
opposition, given the public investments’ effects on incomes and thereby fighting capabilities
of the two actors. O will then choose to fight, and “nature” determines whether O or D will
have power in period 2. If the Po(g) function is monotonic and increasing, the dictator will set
g equal to zero, since this is the amount of g that optimizes his probability of survival. Mobutu
Sese Seko infamously claimed when addressing then Rwandan president Habyiarama: “I’ve
been in power in Zaire for thirty years, and I never built one road. Now they are driving down
your roads to get you” (Sundstøl Eriksen, 2004:4). Mobutu was referring to the Tutsi RPFguerrilla flowing in from Uganda across Rwandas northern border. It seems that Mobutu, a
relatively rational dictator with a desire to stay in power, was well aware of the threats to his
power stemming from different guerrilla groups, and how these threats could be enlarged by
investing in public goods. If the specific prediction of g=0 seems to strong, the qualitative
interpretation of the result is that dictators might intentionally set public investment, and other
economic policies, such that they produce lower growth and development than what could
have been achieved. If you take a good look at Table 2, macroeconomic disasters have mainly
come in countries were dictators have largely faced internal security threats. Countries such as
Angola, Cote d’Ivoire, Congo, Nigeria, Mozambique, Sierra Leone and Nicaragua have all
experienced civil war, and several ousted or at least highly threatened dictatorial regimes.
Civil war in itself might of course also have contributed strongly to detrimental growth
performances, but my model indicates that strategies conducted by rational dictators being
aware of internal threats might have been at least equally important.
In the above examples, it is not necessarily that the dictators have not made public
investments, but the model interpreted widely will point to g representing any kind of public
policy that has implications both for probability of staying in power and for development. G is
then not only about building roads, as in the case of Mobutu. It is also about restricting
economic openness, as for example in the case of Kim Jong Il’s North Korea, and several
other policies and strategies designed by the leader to stay in power, with adverse effects for
the economy. Take also the case of general repression, which might have adverse economic
effects: “Repression is an important tool in the hands of the dictator; but just as a monopolist
in the economic marketplace may try to impede (block, repress) entry by alternative sources
of supply for the purpose of raising profits from exchange, a political dictator does not repress
for the sake of repression alone but in order to increase power” (Wintrobe, 1990:851).
I would like to mention two particular dictatorial strategies that is often used in the real world
by dictators to stay in power, and which could be represented by g. The first is clientilistic
practices and divide and rule strategies, and the second is violation of property rights.
Clientilistic strategies are existent in many dictatorships and in some democracies as well.
Clientilism in democracies is often related to the exchange of votes from a segment of the
population for allocation of resources from a particular politician or political elite through
vertical networks. Here, and more in general, we are talking about the exchange of political
backing from a particular segment of society, an ethnic group, a family, an organizational
18
group or the army, for the allocation of resources or other favors related to the formulation of
specific economic policies. The exchange need not be explicit and contractual, but might very
well relate to implicit exchange of political backing for economic favors. Certain groups
might be considered as essential backers for a dictator in his quest to maintain power, and he
needs to secure the continued support from these groups through what Haber calls the logic of
co-optation (Haber, 2006). This co-optation could be achieved through distribution of
economic rents, passing of specific law or transfer of property to one group at the cost of
other groups, or society more in general. The point is here that the dictator wants to please a
specific group in order to maintain his power, and that the policies then conducted are harmful
to the overall economy.
When it comes to property rights, I have earlier made the argument that dictators who want to
secure continuation in office might have incentives to not enforce property rights in order to
hinder economic development and thereby the rise of alternative rival power groups. The
leaders might also have incentives to enforce property rights selectively, and not protect the
property of opposition groups or potential adversaries. They might even have direct incentives
to destroy or appropriate the property of these groups (Knutsen, 2007). Such actions will have
a clear, negative impact on the protection of property rights in the country, and further on the
economic development in the country. Take for example the actions of Robert Mugabe in
present day Zimbabwe: His “2005 “clean-up” in Harare, the Zimbabwean capital, where the
government demolished homes and shelters for about 700 000 poor and sought to move many
of them to the country-side (BBC, 2005). Even though the government claimed there were
aesthetic reasons behind the operation, as well as an intention to reduce crime and
overcrowding, it is common knowledge that most revolutionary attempts dangerous to
political leaders originate in the large cities, and particularly the capital. Mugabe now had
fewer disgruntled subjects in the capital to worry about” (Knutsen, 2007:28).
In general therefore, the g-variable presented in this model, need not represent public
investment only, but rather a vast amount of policies and strategies undertaken by the dictator
and his backers. These strategies are aimed at securing the continued rule of the regime, and
the economic consequences of these strategies are negative. In certain instances, as implied in
this model, the negative economic consequences are intended, since they reduce the
oppositions potential for gaining power, but in other circumstances, the negative economic
effects might be more of an unintentional byproduct, as might be the case for clientilistic
strategies. Therefore, this aspect of the model needs to be interpreted relatively widely, when
understanding how the actions of power-motivated rulers might harm the economy of the
country. One particularly interesting case comes forward if the ruler perceives modernization
theory to be correct, and assumes that the general chance of survival in the long term is
negatively affected by the level of development in a country. In this case, my model would
predict that the rational dictator does everything in his power to hinder the economic
development in his country. As Peter Evans puts it “Extracting a larger share from a shrinking
pie is not the optimal way to maximize revenues, but it may be the only way consistent with
the survival of predatory states ... Generating an entrepreneurial class with an interest in
industrial transformation would be almost as dangerous as promoting the political
19
organization of civil society. For predatory states, "low-level equilibrium traps" are not
something to be escaped; they are something to be cherished” (Evans, 1995:248).
*
Let us now sum up the results from the analysis so far, and compare the implications for
development, stemming from a power-motivated, rational ruler choosing his policies in order
to stay in power in different contexts. Let us assume that the dictator sets g=g* in the first
context of an external security threat and g=0 in the second context of an internal security
threat. What is the overall output of these two hypothetical economies? In the first case we
will have Ws= Y(g*) + R, whereas we in the second case will have Wo= Y(0) + R. Since
dY/dg>0 and g*>0, we will have that Ws>Wo. That is, the economy will have a larger per
capita income due to higher public investment in the case where the dictator faces an external
security threat. If we assume “essentiality” in the industrial sector, that is that a small amount
of g is needed to get any industrial output at all, we will have that Wo=R. The model
therefore predicts that dictators facing an external security threat will generate much stronger
macroeconomic results through chosen policies than dictators facing internal threats. We will
now investigate how the Kuomintang’s behaviour changed in line with the model’s
predictions with a change in the surrounding security context.
A quasi-experiment: Kuomintang and strategy-change on the boat
The behavior of Kuomintang before and after defeat in the Chinese civil war serves as a very
illustrative case when it comes to highlighting the mechanisms proposed in my theoretical
model. The Kuomintang moved from a situation in the mid-40s where they largely faced an
internal security threat to a situation in the early 50s, where the main security threat was
external. This constitutes a very interesting “quasi-experiment”, since we have the same actor,
the central elite in the Kuomintang, operating in two different contexts. We can therefore
exclude explanatory factors related to differences in actor-capabilities, motivation and
preferences, and we can isolate the variation that is due to the change in context. My model
would imply that given their motivation for power (remaining in power in Taiwan, and
eventually capturing it again in mainland China), the Kuomintang would engage in totally
different strategies and conduct different economic policies in the two different contexts.
Denny Roy’s assessment of the motivation of the Kuomintang backs up the assumption of a
power motivated political elite: “Like any group of politicians, the most fundamental
objective of KMT leaders was to maintain their own power and status” (Roy, 2003: 103).
The Kuomintang, led by General Chiang Kai Shek, was engaged in a bloody civil war with
the Communists, led by Mao Tse Tung, (before and) after the Japanese occupation of China.
As the years passed, it became clear that the Kuomintang was losing ground on the mainland
to the Communists, and they eventually retreated to Taiwan, where they established their new
base. About 2 million Kuomintang supporters moved over the Taiwan Strait in the late
1940’s, among them 600 000 soldiers (Roy, 2003:76). When the Kuomintang were fighting in
the civil war on the mainland, they were actively engaging with an “internal security threat”
according to the terminology of the model. Even if it is far-fetched to claim that the
20
Kuomintang constituted a sitting regime, they did control relatively large parts of China, and
treating them as the model’s D-player is not too far-fetched. When the Kuomintang retreated
to Taiwan after defeat and established their base there, it can be interpreted as a de facto state
formation and transformation of the Communist threat from an internal to an external threat.
The Kuomintang faced a possible security threat from Taiwanese inhabitants as well, which is
illustrated by the fact that there were approximately 500 000 people working part-time as
government informers at one point in time (Roy, 2003:91). Nevertheless, the main security
threat was perceived to be Communist China, with its armies on the mainland. This was
especially true after 1951 since during purges on the island in the preceding years “the KMT
crushed most of the potential leadership of Taiwanese nationalist movements; most of those
activists not killed were co-opted by the party or forced to flee the island” (Roy, 2003:72).
Behavior during the civil war
How did the Kuomintang historically conduct its policies in the two different contexts? When
the Kuomintang was based in mainland China it “had been largely predatory, based on rentseeking” (Cho and Kim, 1998:137). The Kuomintang, its officers and officials, were
notoriously known for their corruption and stealing and looting ways, which were seen as the
best way to provide resources to fight the Communists in the civil war. It was a strategy for
the short term, but it also made sense since it did not build economic capacity that could be
later utilized by the Communists if they could get control over it. The Kuomintang controlled
Taiwan at this point in time as well, and their policies on the island were characterized by the
same looting and corruption as on the mainland. After the arrival of Kuomintang troops on
Taiwan in 1945 they “were expected to make their living by scrounging and plundering
among the local civilian community, following the pattern of Chiang[ Kai Shek]’s armies on
the mainland” (Roy, 2003: 59). And so they did, as even doorknobs were stripped off
buildings! The looting was in some instances also very organized, and in some cases, whole
factory plants were dismantled and sent to mainland China. Not only lack of protection of
property rights was characteristic of the Kuomintang rule of the island and the mainland
during the latter half of the 40’s. Chiang Kai Shek assigned the rule of Taiwan to one of his
commanders, Chen Yi, and “his administration in Taiwan was infamous for its widespread
corruption and nepotism” (Roy, 2003:61). The economy was also organized in a very statist
and concentrated way, with state-run monopolies in several sectors, allowing the Kuomintang
to extract the maximum share of the generated income. In these early days, the internal
security threat in Taiwan itself was also much larger than it would be a few years later,
illustrated by the takeover of the nine largest cities in Taiwan in 1947 by more or less
organized crowds of Taiwanese inhabitants (Roy, 2003: 68). Kuomintang armies had to be
sent from the mainland to eventually crush the uprising.
Behavior after the civil war
What happened with Kuomintang’s economic strategies some years later might then look as a
puzzle to many observers of the regime’s earlier economic policies, but is perfectly
understandable within the logic of the model presented in this paper. “[B]y late 1948 it was
apparent that the ROC regime would have to evacuate to Taiwan and make the island its new
21
base. This involved a reversal of policy from stripping down Taiwan to building it up” (Roy,
2003:76). Reversal is a modest use of words, as the change in strategies is one of the largest
turnabouts viewed in modern history. The “KMT enforced a series of reform measures with
the purpose to reconstruct Taiwan as a base for the eventual recovery of the mainland. These
measures included not only registration of the members of the KMT, strict application of
discipline, consolidation of the central leadership for its renovation, but also economic
reforms,” (Cho and Kim, 1998:137). Policies on land reform, industrialization, education
policies and general economic strategies aimed at development produced a spectacular
economic growth on the island in the 1950’s and 60’s, earning Taiwan the label of an
economic miracle. “During the decades of 1950s and 1960s, Taiwan’s primarily agricultural
economy developed into a semi-industrialized economy, domestic consumption demand
greatly increased, and unemployment faded as a serious problem. Through the 1950s,
Taiwan’s annual gross national product (GNP) grew at a rate greater than 8 percent. In the
1960s the growth rate rose to nearly 10 percent, while inflation was kept below 5 percent”
(Roy, 2003: 103). Not only good economic policies were a result of intentional Kuomintang
efforts, but also a program of “institution building”, underscoring my point made earlier that
in the long run, institutions and institutional qualities are endogenous to the strategies of
autocratic political elites. Chiang Kai Shek founded an institute for training party cadres in
1949, threw out incompetent or corrupt statesmen, and refurnished party structures and the
bureaucracy. “These years of reform and reorganization saw growth in discipline, efficiency
and morale” (Roy, 2003:81). Formal education of party officials was an important part of the
strategy. The role of the competent Taiwanese bureaucracy became vital for the island’s
economic modernization.
Interpretation: Evidence in line with the model
The fact that the same actor changed its economic strategy from a predatory one to a
developmentalist with change in circumstances, lends credibility to the argument developed
from the model: Self-interested and power motivated autocratic elites will greatly diverge in
their economic strategies under different contexts. Here we have somewhat of a quasiexperiment: the same actors combined with a clearly specified difference on the “treatment
variable” (main security threat). More specifically, it was claimed that elites with an internal
security threat have incentives to follow predatory strategies, as Robinson labels them (1998,
2001), whereas elites mainly facing an external security threat will have incentives to follow
developmentalist strategies. Kuomintang faced what could be labeled an internal threat both
in Taiwan and on the mainland before the end of the civil war, and the major security threat
after the move to Taiwan was external, namely the PRC’s forces based on the other side of the
Taiwan Strait. After the Korea-war broke out and ended, the PRC’s lust for an adventure in
Taiwan was temporarily low after taking losses in that war, and the US was at the time eager
to guarantee the security of Taiwan’s regime, fearing the “Communist threat” In the Asian
theatre. Kuomintang officials could however not with full certainty rely on the US’ security
guarantee in the long run (Roy, 2003), and needed to develop a modern and well-functioning
fighting force of its own. This was gradually achieved, thanks in large part to the booming
economy. In conclusion, the greatly diverging economic policies followed by the Kuomintang
22
before and after the exit from mainland China are well in line with the predictions generated
from the model, which were described under the headings of internal and external security
threats, respectively.
Two simultaneous threats
Often however, a dictator faces two relatively strong security threats at the same time, and
needs to take into account both of these threats when conducting policies that are meant to
enhance his survival probability. The model presented above can also be used to analyze this
type of situation. D in the model does not care whether he is overthrown by an internal
opposition or an external invading army; he just wants to minimize the probability of being
overthrown. Let us call the event of overthrow by foreign forces s, and overthrow by domestic
opposition o. The dictator then wants to minimize P(s˅o) = P(s) + P(o) – P(s˄o). P(s) is
known as Ps and P(o) as Po. Since we assumed that these probabilities were independent, we
have that the dictator will minimize the expression Ps + Po – Ps*Po, by setting g.
F
Ps  Po  Ps  Po  Ps  s
 Fd

 Fo
  Po 

 Fd

 Fs
  Ps 

 Fd
  Fo 
  Po  
  Fd 
By inserting the income functions as arguments into the F-functions, we obtain the nasty
expression:
Ps
 F  Y *   P
 Y  g   R  g 
s
Fd
o
Fo  1    Y  g  

 Fs  Y *   P Fo  1    Y  g   
  Ps

o
Fd Y  g   R  g   Fd Y  g   R  g 
Fd Y  g   R  g  
If we want to solve and find the optimal amount of g, we have to differentiate the expression
with respect to g and set the resulting expression equal to zero5.
F
Ps '  s
 Fd

dFd
  Fs  dg
 
Fd2
 
   P '  F    F     F  
d
 
 Fd 
o


o


 Fs  dFdgd


F
s
  Po Ps '     

Fd2
 Fd  

dFo
dg
o
dFd
dg
Fd2
   P P '  F    F     F     0
o
 
 Fd 
o


s o
dFo
dg
o
dFd
dg
Fd2


If we now simplify somewhat we get:
F 
F 
dFd
o
 1  Po  Ps '  s  Fs  dFdgd  1  Ps  Po '  o  Fd  dF
dg  Fo  dg
 Fd 
 Fd 

5 dFd
dg

dFd
dI d

dI d
dg

dFd
dI d
  Y '  g   1 ,


dFo
dg

dFo
dI o
23

dIo
dg

dFo
dIo
 
  1    Y '  g  
  0
This implies that:
 Fo 
dFo
dFd
 Fd  dg  Fo  dg
 Fd 
1  Ps  Po ' 

 


  1  P  P '  FF   F  
o
s

s
d

s
dFd
dg
What does this expression tell us? The right-hand side shows the marginal effect on the
probability of being ousted from an external actor when the dictator increases g with a small
amount. This effect is in optimum equal to the marginal effect on probability of being ousted
by the domestic opposition when g increases. The rational dictator therefore balances these
threats at the margin by setting an “intermediate” value of g. An increase in g from this point
would increase the severity of the internal security threat more than it reduced that of the
external threat. A decrease in g from this point would increase the gravity of the external
threat more than it reduced that of the internal.
There are some important assumptions that need to be made for this condition to represent the
optimal solution. We might very well have so-called corner solutions, which are better for D
to aim at. If the threat from O increases more with the first unit increase in g than the external
threat drops, the dictator will still go for g=0. The second option is to go for g=g*, the point
where S decides not to go to war. This happens if the external threat is grave, and the increase
in domestic threat is not much affected by the increase in g. This might be the case in a
homogenous population with no rebel movements and with low aspirations for
democratization. Another important aspect that could lead to this result is in cases where the
relative strength of D versus O is not much changed when g increases, for example by the fact
that the dictator is able to appropriate most of the resulting industrial outcome, Y, or where
the derivative of O’s fighting power transformation function with respect to income is small.
In less technical language, this is the case where a rich opposition is not able to organize any
decent effort to overthrow the existing regime.
A helpful condition, but not a necessary (one of the functions could have very strong
concavity or convexity properties) or sufficient one (corner solutions), for the first order
condition above to be the optimum, is that d2Po/dg2>0 and that d2Ps/dg2<0. This means that
the positive marginal effect on the threat from O when increasing g, increases with a higher
level of g, and that the marginal effect on threat from S diminishes faster at lower levels of g.
Public spending increases your readiness to fight off the neighboring country’s army to a
large degree when you initially have a poorly developed industrial economy, but an increase
in g is particularly bad for your ability to defeat your internal opposition when your industrial
economy is well developed. I will not go deeper into the more technical properties of the
second order conditions. Figure 4 shows how the optimality condition is determined in the
case where the “helpful” conditions presented above are satisfied. Clearly, the optimal amount
of g, g^, in this case is somewhere in between of the two extreme cases from the less general
models where the dictator faced only an internal or only an external threat.
Figure 4: Optimal solution when facing two security threats
24
Po
Ps+Po
Po, Ps,
Ps+Po
Ps
g^
dPo/d
g
g
dPo/dg
dPs/dg
dPs/dg
g^
g
In Figure 4, the dictator sets an intermediate level of g, g^, in order to minimize the
probability of being thrown out of power by either O or S. In the second stage of the game, O
contends power, and S goes to war. The resulting outcome of the game is then determined “by
nature” in accordance with the specified probabilities. Has any dictator faced the type of
context here described? Probably, the “corner solutions” with D caring overwhelmingly about
either an external or internal threat are often as relevant when it comes to giving parsimonious
descriptions of real world situations. Nevertheless, Stalin in the 1930’s might be one example
of a dictator facing two simultaneous threats (internal contenders for power within the party
and the army and Hitler), and Stalin was walking a tightrope. He purged the military, and
killed a large share of his officer-corpse. This weakened his military, but at the same time
cleared him of (perceived) internal threats. According to the logic of this model, Stalin would
have gone even further in cleansing out perceived opposition if he had not faced external
threats. The Soviet military’s fighting power was definitely weakened by the purges, but he
managed to rebuild the military over time, and eventually won the war against Nazi-Germany.
What about present-day Iran? Surely, a large fighting force is perceived as necessary to deter
a possible attack by for example Israel and the US. Iran’s economic policy has however not
been very successful. The regime does however have an interest in not strengthening the
already relatively strong middle classes in Tehran and other major cities. If the regime
perceives modernization theory as correct, a strengthened middle class could trigger political
dynamics leading to democratization. North Korea also has both internal and external threats
to fear. Kim Jong Il has vested a lot of effort in sealing off the borders to international
influence, and both foreign direct investment and technology diffusion have suffered.
However, North Korea needs a strong military fighting force to prevent an attack by its
perceived enemies. Luckily for the North Korean regime, the production of the atomic bomb
functions as a “cheap deterrent”. It is possible to obtain a deterring force, without industrial
development and the subsequent modernization of its already large standing army. According
25
to this model, a North Korea without nuclear weapons (a worse Fd-function), would probably
have needed to conduct better economic policies (higher g) in order to fight off foreign
enemies, but it would still have needed to think about its domestic concerns and not open up
the economy and modernize totally.6
Some statistical evidence
The model presented clearly predicts that the perceived nature of the main security threat
from a dictator’s point of view will be very important for the policies and strategies chosen,
and for the subsequent macroeconomic income and production of that country. The model
therefore provides one potential answer to why the observed empirical regularity in section 2
occurs. Some more specific empirical evidence also back up the theoretical predictions.
Compare the incentives facing dictators in fragmented post-colonial African states, with those
facing Asian autocratic rulers after de-colonization, in terms of the key variable of this model:
type of security threat. African dictators did not to a great extent have to fear their neighbours,
but they rather had to keep an eye on internal groups willing to take up arms to grab political
power. Africanist scholars have noted that the main threat to most of Africa’s post-colonial
strongmen have come from within the juridical borders of their own states (Clapham, 1996).
On a war-torn continent, only a small number of wars have been traditional inter-state wars.
Only in the Ogaden-war between Somalia and Ethiopia (1977-1978), in the war between
Julius Nyerere’s Tanzania and Idi Amin’s Uganda (1978-1979), and the Ethiopian-Eritrean
conflict (1998-2000) do we find historical examples of traditional wars between states, and
“[I]n only one case did the direct invasion of one African state by another lead to the
overthrow of the regime in power and its replacement by a government acceptable to the
invading state. This was the Tanzanian invasion of Uganda” (Clapham, 1996:123). The
Organization for African Unity (OAU) contributed very strongly to the non-intervention
policies followed by rulers of African states by establishing principles and norms of nonintervention on the continent, leading governments to not intervene because among others
they feared that a breach of these norms and principles would backfire by making it easier for
other actors to later intervene in their domestic affairs (Knutsen, 2005)7. Predominantly,
African conflicts have been civil wars, “and the African dictator’s worst threat to being ousted
from power has come from within. Rebel groups, militias and separatist movements were, and
still is, a characteristic of a large part of the continent’s countries” (Knutsen, 2007: 28). In
Asia, however, several of the small Asian Tiger states did not so much face the prospect of
internal rebellion as that of external invasion. According to Carles Boix “[T]wo forces
generally impel political elites to industrialize: first, a natural desire to catch up with more
modern economies; second, and more importantly, security concerns, that is, the need to
6
Note that we assume that nuclear weapons will not be used against internal opponents! The F d is then no
longer one function, but one should separate between internal and external fighting power. This distinction
would in any case be a realistic nuance to make, but it would complicate the model.
7African
governments have however on several occasions more subtly intervened in other countries’ affairs by supporting
different groups and power centres inside these countries. The phenomenon is often referred to as the fighting of “proxywars”.
26
modernize to prevent neighbours from amassing resources to defeat them in future wars”
(Boix, 2003:218). Even though the first force probably has been important and existent for
leaders of Asian states like Lee Kuan Yew and Deng Xiaoping, several of the best economic
performers in Asia have throughout their independent history faced severe threats from other
larger military powers with an eye to occupying or sub-ordering them. “Taiwan had its less
than friendly neighbour China, South Korea had North Korea and Singapore had Malaysia.
Taiwan, South Korea and Singapore were all inferior when compared to their neighbours in
military power, and all have had the fear of intervention constantly hanging over them. For all
of the regimes, the importance of building military capacity (and for example for South Korea
drawing on the US’ capacity) was of utmost importance. A strong, “modern” military
apparatus craves a relatively developed and sophisticated economy, and the outside threat
might through this channel have provided these Asian regimes with incentives to make their
economies develop.” (Knutsen, 2007:26-7).
I have earlier, using regression analysis, shown that dictatorship, when compared to
democracy, was relatively conducive to economic growth in Asia and detrimental to
economic growth in Africa south of the Sahara in the 1990s (Knutsen, 2006a:314-23). Asia
was the only continent with a positive estimated effect from dictatorial rule, and the effect
was significant at the 5%-level in some models. When it came to Africa, dictatorship had a
negative and significant effect at least at the 5%-level, and the effect was stronger than on any
other continent (Knutsen, 2006a). In order to claim that this is evidence for the argument of
my model, we have to claim that a relatively large number of Asian autocracies have faced
external security threats, or a combination of external and internal threats, and that African
dictatorships have largely faced internal security threats. There are of course also other
differences between Asian and African style dictatorships that might explain divergence, but
some of these might themselves be endogenous to the type of threat in the longer run.
We could triangulate the evidence further by looking at countries that have experienced civil
wars or anarchy due to infighting between various groups inside the borders of a juridical
state. The Polity-project registers countries that go through a period of ‘interregnum’, “during
which there is a complete collapse of central political authority. This is most likely to occur
during periods of internal war.” (Marshall and Jaggers, 2005:17). We could perhaps assume
that countries that experienced interregnum at a point in time are also the countries where the
ruling dictator faces especially grave internal security threats. There are of course serious
validity problems with making this assumption. A more appropriate test would be to go
through the historical experiences of all dictatorships and classify their dominant security
threat at a particular point in time. This would of course incur a large degree of subjectivity,
but it would be a more valid operationalization. The amount of time required for this datacollection would however be enormous.
According to the model there is reason to believe that dictatorships would be particularly bad
at producing economic growth and development in countries that experienced “interregnum”.
In Knutsen (2007), I constructed a dummy and assigned a 1-score to countries that
experienced at least one year with interregnum from 1950 to 2004. I then created an
27
interaction variable by multiplying the aggregated FHI with the dummy. The empirical
prediction from the model is that the interaction variable would be negative and significant. I
then tested the hypothesis, using Pooled Cross Section - Times Series analysis and a
regression model relatively similar to that presented in section 2, with the degree of property
rights protection as the dependent variable. The interaction variable was significant at the
0,1%-level with a t-value of 7,8, indicating that the negative effect on property rights
protection from autocratic regimes were significantly more sizeable in countries that were
assumed to face grave internal security threats.
Comparative statics: Further implications from the model
How do other contextual variables interact with the nature of the security threat to produce
economic outcomes in my model? Let us engage in some simple exercises of comparative
statics. Throughout the exercise, we will assume that the amount of g before the shift is
determined by the optimality condition derived above when D faced two simultaneous
security threats. I will not perform calculations here, but only provide the intuition and
describe the workings of the model qualitatively.
Natural resources
What happens according to the model if we get an exogenous change in the amount of natural
resources in the autocratic country? Since we assume that D receives all income from the
natural resource, R, he will get extra revenue “for free”, which he can then use to invest in
fighting capability. It is easy to see from the model that D will then choose to reduce the
amount of g. The logic behind the argument is that since D is better suited to fight S without
creating the conditions for industrial development, he can afford to reduce g in order to
diminish the threat from O. Besides the alteration in relative strength between the D and O
stemming from the reduction in g, he will of course also benefit here from the extra R he
controls. The model predicts greater survival probability for dictators with access to natural
resources, a hypothesis that can be tested at a later stage, and it additionally contributes to
explain the so-called resource curse when limited to dictatorships. Rational dictators do
simply not need to invest in industrial development or general development of the economy to
fight off foreign enemies. Natural resources give them easy access to income that can be
transferred into fighting power, and they do not have to risk the by-products of modernization:
a well-endowed opposition fighting for control over the country’s political life, for example
through establishing a democracy. Robinson’s model (2001) has relatively similar dynamics,
when dictators are assumed to be wealth maximizers.
Military build-up in foreign country
What if any of the parameters β and µ, or the foreign GDP, Y*, increases? This would lead to
a military build-up in the foreign country, thereby increasing the external security threat. It is
obvious from the model that this would increase the probability of losing power to S for D,
thereby providing an incentive for D to increase g in order to increase his own fighting
capabilities. This however will come at the cost of increased probability of losing power to O,
28
and this effect will restrain the increase in g, even though there will evidently be an increase
in g. We can conclude that dg/dY*, dg/dβ and dg/dµ >0. A more dangerous external threat
will introduce a need to take it more seriously and D will have to find a new balance between
the threats by increasing the level of development. Both Po and Ps will increase relative to the
old situation, and the dictator’s probability of survival will be lower. Y, income from the
industry sector, will be higher in the new situation.
Organizational strength of the opposition increases
Another relatively clear implication from the model comes when we look at the strength of
the opposition, O. Let us assume that the transformation function Fo improves. This implies
that O can generate more fighting power from the same amount of income. In a wider
interpretation of the function, we could perhaps say that due to some exogenous shock, the
willingness of a large amount of the citizenry to fight for democracy increases, perhaps
spurred by a successful democratization in a neighboring country. We could also interpret it
as an improvement in the organizational capability of the opposition, for example through the
rise of an undisputed charismatic leader, or external help from another government in
coordinating action against the regime. Since Fnewo(Io)> Foldo(Io), Ps increases. This would,
ceteris paribus, lead the dictator to lend more concern to the internal opposition, thereby
seeking to reduce its newly won strength. In this model, D does this by reducing g. This
comes at a cost however, since the reduction in g increases the opportunity of S successfully
invading in the second stage of the game. The probability of survival for D decreases and the
level of industrial development is also lower in the new situation.
The dictator is able to obtain a higher share of the industrial outcome
What happens if D is able to increase α, namely the share of income he can grab from the
industrial sector? Let us assume some exogenous event that enables D to increase α. D then
increases his own income and reduces that of O. The effect from this event on D’s survival
probability is unambiguous: Po will decrease, and D now sits safer in his office or mansion.
What will the response be in terms of the chosen g? This turns out to be more ambiguous
since we will have two different effects operating at the same time. First off all, D can now
turn to reducing the risk of being overthrown by S, since the internal threat is reduced, and D
will according to this mechanism set a higher g. However, there is also another effect, which
is probably not as obvious to recognize. Since D has become richer, because a rise in α
implies a rise in Id, there will be a decrease in Ps, even if D does not change g from the
previous situation. This effect by itself would actually encourage D to reduce g, since he faces
a less dangerous external threat! Which one of these two effects that are dominating is not
possible to answer a priori, but depends on the specifications of functions and parameters. It is
therefore ambiguous from the model whether D will actually increase or decrease g when α
increases. It seems intuitive however that for many specifications the first effect will
dominate. This is because when it comes to the internal threat, an increase in α both increases
D’s fighting strength and decreases that of O. When measured against S however, D’s
strength increases, whereas that of S stays constant: It might be plausible to assume that the
29
relative fighting power increases more when measured against O than against S. However, if
we assume that the increase in income tilts Ps below P*, the dictator can now forget about the
foreign actor and reduce g, as long as his income is not reduced sufficiently to bring P s above
the threshold again. This is maybe one of the least intuitive results from the model: An
increase in the share of the industrial income for the dictator might in some cases lead him to
set policies so as to reduce the overall industrial outcome.
8. A critique of the model: Economic development and legitimacy
The claim that dPo/dg is positive, could be contested on both theoretical and empirical
grounds, especially if it is viewed as a universal and non-contingent description. It could be
that an increase in the level of economic development could bolster the security of a
dictatorship, even if democratization is the relevant internal threat to the regime. It is not
uncommon to argue that economic growth and other positive economic developments bolster
the legitimacy of the regime. An alternative argument, in line with the concepts of the model
above, could be that the fighting capacity of the regime increases disproportionally more than
that of the opposition with an increase in Y.
Empirically, a high level of development is conducive to democratization, or at least the
probability of being democratic, but as Przeworski and Limongi (1997) show, the probability
of breakdown for dictatorships depend negatively on economic growth in the short term. In
this sense, my model of autocratic strategies for regime survival when facing internal
opposition is best suited for the longer run. It is not well suited to capture short term
dynamics. A richer model could have incorporated more than two periods and specified nonlinear dynamics. If the call for democratization increases, a dictator might in the real world be
wise not to send the economy into recession, since this could trigger a revolution. Davies
(1962) theory on revolution, built around the so-called J-curve is illuminating in this respect.
Revolution breaks out when expectations are not met by material conditions, and a sharp
recession could trigger widespread dissent and subsequent political action. The question of
whether revolution originates in times of crisis or because of development is an old one, and
the positions in this debate can be traced back to Tocqueville and Marx respectively. Note
however the distinction between the increase in probability of overthrow in the short run, and
the dangers to the dictator of promoting development in the long run, captured by
modernization theory and several empirical studies (Diamond, 1992 and Inglehart and Welzel,
2006). Przeworski et al. (2000) launched an empirical critique against modernization theory,
as they claimed with a basis in empirical evidence that the correlation between development
and democracy was not due to regime-breakdown in rich autocracies, but the lack of
breakdown in rich democracies. This evidence has been strongly disputed by later empirical
research, for example by Boix and Stokes (2003) and Hadenius and Teorell (2005).
Who said being a dictator is easy? There are plausible arguments that both economic
development and the opposite could throw an autocrat out in danger. Mueller (2003:411),
referring to Tullock (1987) claims that “job security is one “obvious goal that a dictator is
likely to pursue, and one of the most difficult to achieve”. Wintrobe captures this point neatly
30
by outlining the possible need for dictators to be able to draw on both “repression” and
“loyalty”-creating strategies. It adds complexity to the world of the dictator that an increase in
loyalty and a certain amount of legitimacy might depend positively on economic growth,
whereas repression of opposition might be easier if possible groups contending for power lack
resources. Based on the empirical material, it might seem that losing loyalty and legitimacy
might be a severe problem in the short run, and might trigger revolution, but modernization
theory indicates that a very long period of growth strengthens the forces that demand
democracy, and the dictator might then want to resort to other long term tactics than
generating development. My model does not capture loyalty-strategies neatly, but it could
surely have been built into it, by establishing another group that would be more prone to
contend power if the economy faced poor economic development. Another alternative could
be to yield to the claim that legitimacy due to growth is the most important survival strategy
also in the long run, and reverse the sign of dPo/dg to negative. The implication of the model
would then be that it is optimal for D to increase g as much as he can. The model would then
give similar implications as that of the threatened dictator in Overland et al’s (2000) model.
The model would then not be in line with the empirical facts: Some dictatorships, and indeed
most of the ones observed, generated policies that were not good for the overall economy, to
say the least, and this empirical trait is captured by my presented framework.
9. Conclusion
Aristotle’s analysis on the diverging natures and consequences of one-man rule served as an
entry point for this analysis. Aristotle laid out what has been one of the basic claims of this
paper, namely that one-man rule in different empirical contexts can have dramatically
divergent effects for the citizens. I have focused here on the economy. However, Aristotle
focused on the issue of how personal characteristics are important to whether a ruler can be
described as an enlightened monarch or a tyrant. This point has been underappreciated in my
analysis, and divergence in personal characteristics and how they affect an autocratically
governed state is an important topic. It is not because I do not believe that dictators differ in
their capabilities, motivation and the degree to which they are “virtuous”, I have here left it
out. I wanted to show that even if we were facing dictators with similar motivational
characteristics, the economic outcomes in two countries could still differ dramatically, if
certain contextual factors diverge. I have focused on the less virtuous rulers, the ones driven
by self-interest and who are motivated by gaining and maintaining power. Even Aristotle
noted that enlightened rulers were far and few in between, and a cursory reading of political
history should lend credibility to this claim. Summed up, my theoretical exercise has shown
that even if all rulers were motivated by self-interest and power, economic experiences in
different dictatorships will diverge big-time. Some dictators will produce developmentconducive policies and help their nations grow, some will produce policies that are
economically disastrous, and some will choose “intermediate policies”. The nature of the
security threat, but also the existence of natural resources, the ability of the dictator to
appropriate a large share of national income and the character of the internal opposition in
terms of organizational capability, were identified as key contextual variables. This, together
31
with the insight that democratic policymakers are more constrained in their choices of
policies, could contribute to explaining why we observe the empirical regularity in section 2:
Dictatorships show widely diverging economic growth rates, much more so than democracies,
and this holds true even if we control for several relevant variables.
The Aristotelian insight that different leaders are differently motivated and have different
capabilities, probably adds to the divergence in outcomes, and this explanation is therefore
complementary to mine. James Robinson has also showed that one could expect at least
relatively similar patterns of divergence in outcomes if the ruler was motivated by wealth
instead of power (Robinson, 1998 and 2001), and Mancur Olson has shown that the time
horizon of rational dictator’s matter for divergence in economic outcomes as well. What sets
the different rational choice based theories apart from Aristotle is that economic results that
might at a first glance seem to be produced by an enlightened and altruistic ruler, coming
close to the hypothetical social planner in economics, is not really the result of such a figure
doing the right thing for the right reasons. It is more the result of a ruler doing the right thing
for the wrong reasons. We can perhaps talk about a tyrant with the accidental appearance of
an enlightened ruler. According to Robert Barro “history suggests that dictators come in two
types: one whose personal objectives often conflict with growth promotion and another whose
interests dictate a preoccupation with economic development” (1997:50). It is hard to dispute
this claim when facing the evidence of section 2. Barro then goes on to claim that “the theory
that determines which kind of dictatorship will prevail seems to be missing”. I believe that
this article has helped sketch out one plausible theory, which added together with the insights
of Aristotle, Robinson and Olson, among others, can help us understand the puzzle Barro
identifies.
32
10. Literature
Aristotle (2000). Politics. New York: Dover Publications.
Barro, Robert J. (1997). Determinants of Economic Growth. A Cross-Country Empirical Study.
Cambridge, MA.: The MIT Press.
BBC News (2005). “Harare homeless in ‘very bad’ way”. To be found at:
http://news.bbc.co.uk/2/hi/africa/4495438.stm
Beetham, David (1999). Democracy and Human Rights. Cambridge: Polity Press.
Boix, Carles (2003). Democracy and Redistribution. New York: Cambridge University Press.
Boix, Carles and Susan C. Stokes (2003). “Endogenous Democratization”. World Politics 55 (4): 517549.
Cho, Hee-Yeon and Eun Mee Kim (1998). “State Autonomy and Its Social Conditions for Economic
Development in South Korea and Taiwan” in Eun Mee Kim [Ed.] The Four Asian Tigers – Economic
Development and the Global Political Economy. San Diego: Academic Press.
Clapham, Christopher (1996). Africa and the International System –The Politics of State Survival.
Cambridge: Cambridge University Press.
Dahl, Robert. 1957. “The Concept of Power.” Behavioral Science 2: 201-15.
Davies, James C. (1962). “Toward a Theory of Revolution”. American Sociological Review 27: 5-19.
Diamond, Larry (1992). “Economic Development and Democracy Reconsidered”. American
Behavorial Scientist 35: 450-499.
Evans, Peter B. (1995). Embedded Autonomy. States & Industrial Transformation. Princeton:
Princeton University Press.
Evans Peter B. and James E. Rauch (1999). “Bureaucracy and Growth: A Cross-National Analysis of
the Effects of “Weberian” State Structures on Economic Growth”. American Sociological Review 64
(5): 748-765.
Freedom House. “Freedom in the World, Country Ratings”. Data to be found at:
http://www.freedomhouse.org/
Greene, William H. (2003). Econometric Analysis. Fifth Edition. Upper Saddle River: Prentice Hall.
Haber, Stephen (2000). “Introduction: Institutional Change, Economic Growth and Economic History”
in Stephen Haber [Ed.] Political Institutions and Economic Growth in Latin America – Essays in
Policy, History and Political Economy”. Stanford: Hoover Institution Press.
Haber, Stephen (2006). “Authoritarian Government” in Barry R. Weingast and Donald A. Wittman
[Eds.] The Oxford Handbook of Political Economy. Oxford: Oxford University Press.
Hadenius, Axel and Jan Teorell (2006). “Authoritarian Regimes: Stability, Change, and Pathways to
Democracy, 1972-2003”. Lund University: Working Paper #331.
33
Heston, Alan, Robert Summers and Bettina Aten, (2002). “Penn World Table Version 6.1”
Philadelphia: Center for International Comparisons at the University of Pennsylvania (CICUP),
October 2002. Data to be found at: http://pwt.econ.upenn.edu/php_site/pwt61_form.php
Inglehart, Ronald and Christian Welzel (2006). Modernization, Cultural Change and Democracy –
The Human Development Sequence. Cambridge: Cambridge University Press.
Lipset, Seymor Martin (1959). “Some Social Requisites of Democracy: Economic Development and
Political Legitimacy”. The American Political Science Review: 69-105.
Knutsen, Carl Henrik (2005). OAUs betydning for ekstern innblanding i staters affærer i det
postkoloniale Afrika: Hva kan en modifisert versjon av Robert Keohanes teori om internasjonale
regimer forklare? [“The relevance of OAU for external interference in states’ affairs in Post-Colonial
Africa: What can a modified version of Robert Keohane’s theory of international regimes explain”]
Oslo: University of Oslo, Department of Political Science. Paper for the course STV 4344.
Knutsen, Carl Henrik (2006). Political regime types and economic growth – Are democracies better at
increasing prosperity? Oslo: University of Oslo, Department of Political Science. Master Thesis.
Knutsen, Carl Henrik (2007). Democracy and Property Rights - A theoretical and empirical analysis
on the effects of political regime type on property rights arrangements. Oslo: University of Oslo,
Department of Economics. Master Thesis.
Knutsen, Carl Henrik (2008). “The Economic Effects of Democracy – An Empirical Analysis”. Oslo:
University of Oslo, Department of Political Science. Working Paper.
Linz, Juan J. and Alfred Stepan (1996). Problems of Democratic Transition and Consolidation –
Southern Europe, South America and Post-Communist Europe. Baltimore: The Johns Hopkins
University Press.
Lukes, Steven (1986). “Introduction” in Steven Lukes [Ed.] Power. New York: New York University
Press.
Machiavelli, Niccolò (1999). Fyrsten. [The Prince]. Oslo: De norske Bokklubbene.
Marshall, Monthy G. and Keith Jaggers (2002). “Polity IV Project – Dataset Users’ Manual”. College
Park: Program Center for International Development and Conflict Management, University of
Maryland. Manual.
Mueller, Dennis C. (2003). Public Choice III. Cambridge: Cambridge University Press.
Murphey, Rhoads (2000). A History of Asia. New York: Addison Wesley Longman.
Olson, Mancur (2003). “Dictatorship, Democracy and Development”. Chapter 6 in Stephen Knack
[ed.] Democracy, Governance and Growth. Ann Arbor: The University of Michigan Press.
Overland, Jody, Kenneth L. Simons and Michael Spagat (2000). “Political Instability and Growth in
Dictatorships”. London: Royal Holloway College, University of London. Working Paper Number 354.
Persson, Torsten and Guido Tabellini (2003) The Economic Effects of Constitutions. Cambridge, MA.:
The MIT Press.
34
Pierson, Paul (2004). Politics in Time: History, Institutions and Social Analysis. Princeton: Princeton
University Press.
Przeworski, Adam (2007). “Capitalism, Democracy and Science” in Gerardo L. Munck and Richard
Snyder [eds.] Passion, Craft and Method in Comparative Politics. Baltimore: The Johns Hopkins
University Press.
Przeworski, Adam and Limongi, Fernando (1997). “Modernization: Theories and Facts”. World
Politics 49 (2): 155-183.
Przeworski, Adam, Michael E. Alvarez, José Antonio Cheibub and Fernando Limongi (2000).
Democracy and Development: Political Institutions and Well-being in the World, 1950-1990. New
York: Cambridge University Press.
Rigger, Shelley (2004). ”Taiwan’s Best-Case Democratization”. ORBIS 48 (2): 285-292.
Robinson, James A. (1998). “Theories of “Bad Policy””. Policy reform 1: 1-46.
Robinson, James A. (2001). “When is a State Predatory?”. Cambridge, MA: Harvard University.
Working Paper.
Rodrik, Dani (2000). Institutions for High-Quality Growth: What They Are and How to Acquire Them.
Cambridge MA.: National Bureau of Economic Research. NBER Working Paper no 7540.
Roy, Denny (2003). Taiwan: A political history. Ithaca, NY: Cornell University Press.
Sundstøl Eriksen, Stein (2003). “Statskollaps: Hva er det og hvorfor skjer det?” [State Collapse: What
is it and why does it happen?” in Hvor hender det? (Årgang 2003-2004). Oslo: NUPI. To be found at:
(http://www.nupi.no/index.php/publikasjoner/tidsskrifter/hvor_hender_det/2003_2004)
The Economist (2007). 24th – 30th November, 2007: “They think they have right on their side”
The Economist (2008). 5th – 11th Januar, 2008: “Economic and financial indicators”
Winter, David G. (2003). “Measuring the Motives of Political Actors at a Distance” in Jerold M. Post
[Ed.] The Psychological Assessment of Political Leaders. Ann Arbor: The University of Michigan
Press.
Wintrobe, Ronald (1990). “The Tinpot and the Totalitarian: An Economic Theory of Dictatorship”.
The American Political Science Review 84 (3): 849-872.
World Bank. (2006). World Development Indicators 2006. Washington D.C.: World Bank. Edited
internet version at: http://devdata.worldbank.org/wdi2006/contents/cover.htm
35
11.Appendix: List of symbols
D = Domestic dictator
g = Public investment
Fj = Fighting power of actor j
Ij = Income for actor j
O = Domestic opposition
Pj = Probability of actor j overthrowing D in the second period
R = Income from natural resources
S = Foreign government
Y= Domestic industrial income/output
Y* = Foreign industrial income/output
α = Share of Y going to D
β = Share of Y* going to S
µ = Share of Is S is willing to use on military capacity
36