Anda di halaman 1dari 57

Democracy and Economic Growth:

A Historical Perspective

John Gerring
Boston University
Department of Political Science
232 Bay State Road
Boston MA 02215
jgerring@bu.edu

Philip Bond
University of Pennsylvania
Wharton Finance Department
Steinberg Hall-Dietrich Hall
3620 Locust Walk
Philadelphia PA 19104-6367
pbond@wharton.upenn.edu

William T. Barndt
Princeton University
Department of Politics
Corwin Hall
Princeton NJ 08540-1012
wbarndt@princeton.edu

Draft: June 14, 2004


Democracy and Economic Growth:

A Historical Perspective

ABSTRACT

Recent studies appear to show that democracy has no robust association with economic growth.

Yet, all such work assumes that the causal effect of democracy can be measured by a country’s

regime status in a particular year (T), which is correlated with its growth performance in a

subsequent period (T+1). We argue that democracy must be understood as a stock, rather than level,

measure. That is, a country’s growth performance is affected by the number of years it has been

democratic, in addition to the degree of democracy experienced during that period (running from

1900-2001). In this fashion, democracy is re-conceptualized as a historical, rather than

contemporary, variable -- with the assumption that long-run historical patterns may help us to

understand present trends. We speculate that these secular-historical influences operate through

four causal pathways, each of which may be understood as a type of capital: physical capital, human

capital, social capital, and political capital. This argument is tested in a cross-country analysis and is

shown to be robust in a wide variety of specifications and formats.

2
Does regime type affect economic performance?1 The predominant view is that democracy has

either a negative effect on GDP growth or no overall effect. Countries with authoritarian political

systems are thus predicted to grow as rapidly as democracies, perhaps even faster. To be sure,

democracy may have some positive indirect effects – e.g., greater stability or more extensive property

rights. However, the econometric evidence suggests that these positives are balanced by negatives

such that the net effect of democracy on growth performance cross-nationally over the last five

decades is negative or null (e.g., Barro 1996; Feng 1997, 2003; Krieckhaus 2004; Kurzman, Werum,

Burkhart 2002; Przeworski and Limongi 1993, 1997; Przeworski et al. 2000).3 For the most part,

case study approaches to this question confirm the results of crossnational growth empirics (e.g.,

Chan 2002; Rodrik 2003; Woo-Cumings 1999). Thus, although most of the rich countries in the

world are democratic, the direction of causality is unclear. One must keep in mind that many rich

countries become rich under authoritarian auspices.

If this conventional conclusion is correct, one might be justified in concluding that

democracy is a luxury to be enjoyed only by countries rich enough to afford it. This, indeed, is a

common argument among authoritarian leaders in the developing world.4 Others might counter

that democracy is worth the price of economic development because civic equality, political

participation, and civil liberties are intrinsically important (Shapiro and Hacker-Cordon 1999).

However, this argument is unlikely to inspire enthusiasm among those currently forced to live on

one dollar a day, roughly one fifth of the world’s population. In summarizing the results of a recent

poll in eighteen Latin American countries the authors note that “the preference of citizens for

democracy is relatively low; many Latin Americans value development above democracy and would

even stop supporting a democratic government if it proved incapable to resolving their economic

problems” (UNDP 2004: 6). Indeed, the demand for “food first” sounds more sensible than the call

for “democracy first.”

3
In this paper, we show that the skeptical view of democracy and development rests on a set

of econometric findings in which democracy is treated as a more or less immediate cause. This

year’s level of democracy is thought to influence growth performance in the following period

(usually a decade or two). We argue that this is an unrealistic assumption. If democracy matters to

growth it is more reasonable to assume that this effect stems from a country’s regime history, not

whether it currently happens to be democratic or authoritarian at some particular point in time.

Democracy is thus best considered as a stock, rather than level, variable. The distant past has

contemporary effects.

We begin by constructing a prima facie case for a historical understanding of democracy and

its relationship to economic growth. Second, we discuss the definition and measurement of the key

concept, democracy. Third, we discuss the method that will be employed to test our proposition,

the time-series cross-section (TSCS) regression. Fourth, we discuss the results of these regression

tests. Fifth, we discuss the “democratic growth effect” – its magnitude, its policy significance, and

directions for future research. We conclude with a discussion of the possible merits of a historical

approach to the quantitative analysis of political-institutional variables.

WHY HISTORY MIGHT MATTER

Work on democracy and growth has usually focused on the possible causal effect of a country’s

contemporary level of democracy on its subsequent growth performance. The subsequent period at

issue might be a single year or some multi-year period (e.g., a decade); in any case, scholars have

conceptualized the problem as the effect of regime type at time T on growth rates at time T plus

some specified period. These studies look forward, but not backward, in time.

4
We propose to consider regimes as historically informed phenomena rather than as

contemporary variables. This means looking backward, as well as forward (via lagged predictors), in

time. In particular, it means measuring a country’s democratic accumulated stock rather than its level

of democracy at a particular point in time. The core insight is that institutional effects unfold over

time, sometimes a great deal of time.

Nowhere does this seem more likely than when one is considering the causal effects of a

regime type. Democracy and authoritarianism construct deep legacies, extending back several

decades -- perhaps even centuries. Therefore, it is these historical legacies rather than their

contemporary status (as democratic or authoritarian), that ought to be of central concern if we wish

to understand the causal effect of a regime type on a variety of current outcomes – social, cultural,

political, or economic.

Our interest here of course is in one specific outcome -- economic performance, as

measured by GDP per capita growth. Thus, we are impelled to think carefully about the causal

mechanisms that might link a country’s regime history with its current growth performance. Our

explanation rests on the key concept of capital.

Capital, in common usage, implies a fungible resource that accumulates over time (given

favorable circumstances), and promises increased returns in the future. An investment today should

bring a higher yield at some later date. If a democratic regime endures, we argue that it is likely to

foster four types of capital: physical capital, human capital, social capital, and political capital. Extant

studies indicate that all four of these forms of capital have positive impacts on growth performance.

Thus, we anticipate that the longer a country remains democratic, the greater its physical, human,

social, and political capital, and the better its growth performance. These relationships are depicted

in Figure 1. Since political capital is the newest concept in this panoply of capital-isms, our

theoretical discussion focuses primarily on this causal pathway.

5
[Figure 1 about here]

PHYSICAL, HUMAN, AND SOCIAL CAPITAL

The extent to which physical capital is equally distributed throughout a society, is now commonly

regarded as an important component of long-run economic growth (Besley and Burgess 2000;

Sokoloff and Engerman 2000; but see Forbes 2000). Thus, insofar as democracy is successful in

redistributing wealth – either through progressive taxation, social policies, land redistribution, or

simply by opening up markets and institutions in civil society to previously excluded groups (Lenski

1966; Lipset 1959; Meltzer and Richards 1981; Muller 1988) – it seems reasonable to expect that the

longer this regime type is in existence, the greater its aggregate effect will be on the achievement of

social equality, and hence on growth. With the exception of a few dramatic instances of land reform

under the auspices of authoritarian/communist rule or wartime expropriation, patterns of wealth

and income distribution in a society change slowly, if they change at all (Banerjee 1999; de Janvry et

al. 2001). Thus, while democracy per se may have little effect on equality, a prolonged experience

with democratic rule may have important consequences (Gradstein and Milanovic 2000; Tavares and

Wacziarg 2001).

Consider, as well, the role of human capital, which endogenous (“new growth”) theory

identifies as a strong influence on growth performance (Barro 1997; Sala-i-Martin 2002). It seems

plausible to regard democracy as an important institutional factor in the development of human

capital, as measured by declining fertility rates and improvements in education, public health and life

expectancy. Political elites in a democracy have electoral incentives to avoid famine and improve the

quality of life for the least advantaged, incentives which are not present, or are present to a much

smaller degree, in authoritarian systems. Studies have shown, almost uniformly, that democratic rule

translates into improvements in a society’s human capital resources (Baum and Lake 2003; Dreze

6
and Sen 1989; Feng 2003; McGuire 2005). Because of the (literally) long-lived nature of human

capital, the longer a democracy is in place the more pronounced its impact on the level of human

capital.

Finally, many commentators have argued that social capital also contributes to a society’s

economic well-being. For example, in a recent careful study, Guiso et al. (2004) present evidence

that social capital enhances financial development (which is, in turn, generally regarded as growth-

enhancing). Since democracy has been considered a precondition for the development of social

capital (Putnam 2002; Warren 1999), it stands to reason that a long-term democracy would have

long-term effects on the development of social capital. Again, these observations suggest that

democracy’s effect on growth may become more pronounced over time, as a country’s stock of

social capital grows.

POLITICAL CAPITAL

In addition to physical, human, and social capital, we argue that established democracies create a

species of capital that is explicitly political. Just as deferred consumption generates physical capital,

which in turn contributes towards output, so should a country’s political experience today affect

tomorrow’s political capital, and its economic output.

Before introducing a new capital-ism to the social science lexicon it is important to

acknowledge the intrinsic ambiguity of the core concept. “Capital” poses severe problems of

conceptualization and measurement by virtue of its theoretical abstraction. It is, after all, a capability

-- a potentiality. As such, it cannot be directly measured. Even so, the ambiguities of capital-based

theories may be redeemed by the theoretical leverage that they offer. In particular, they allow us to

think about the role of social institutions through time. This, we believe, is a significant theoretical

advantage and is not well captured by other terms.

7
Politics, too, may be capital-ized. Recall that physical capital is usually measured in monetary

terms (e.g., accumulated investments), human capital by the spread of education and other job-

related skills, and social capital by patterns of social interaction (although there are a variety of other

possible measures [Knack and Keefer 1997; Woolcock 1998]). We propose that political capital may

be operationalized by various measures of the relative health of a polity – e.g., trust in political

institutions, low levels of corruption, bureaucratic capacity, political consensus, political stability, the

wisdom and far-sightedness of political leaders, and so forth (Englehart and Gerring 2005). These

resources, like other sorts of capital, accumulate through time and may be drawn upon for a range of

purposes.

In some respects, our use of the term political capital is an extension of the colloquial usage,

where it is usually understood as a resource attached to a particular figure -- someone who is said to

have accumulated (or lost) political capital by some action or set of actions. Granted, the parallel is

not exact. An individual’s political capital may be nothing more than a quid pro quo, a favor done

for a friend that is “banked” for some period of time. However, an individual’s political capital may

also refer to something more complex: a pattern of behavior that establishes a reputation for fidelity

and competence which translates, in turn, into a relationship of extended trust. (It is at this point

that a representative becomes a trustee rather than simply a delegate and the transactional costs

usual to politics are lowered -- since explicit quid pro quo agreements between actors are no longer

required.) The reputational quality of political capital plausibly operates for institutions as well as for

persons. But this is not a key point of the argument, for there are obvious differences between

personal and institutional political capital.

How, then, might democratic political arrangements foster political capital in a country and,

ultimately, enhanced economic growth? To begin with, we surmise that democracy sets in place a

learning process in which policies and institutions are refined. Politicians learn how to govern,

8
electors learn how to vote, and politicians learn how to respond. Second, a sustained experience of

democracy enhances stability and consensus. Both of these processes, which we discuss in much

greater detail below, should have a strong influence on economic performance over the long haul.

To be sure, our notion of political capital overlaps somewhat with those of human and social

capital. Learning about policymaking is in a sense simply a special form of human capital

appreciation; while the emergence of consensus is reminiscent of some conceptions of social capital.

What we have in mind, however, is knowledge and consensus that can only be acquired through

political exposure. It is the nature of the “investment” required to generate political capital that

most clearly distinguishes it from human and social capital.

BETTER POLICIES AND INSTITUTIONS. Consider, first, the realm of economic policy and good

governance more generally. There is no question that public policies have a strong impact on the

growth performance of nation-states, although there remain serious disagreements over which

policies, precisely, are most growth-enhancing. For purposes of discussion we shall assume that

large budget deficits, high inflation, high tariff barriers, and heavy regulatory burdens usually have

negative consequences over the long-term. This does not mean that the state should withdraw from

the marketplace; it means, rather, that economic policies should be “market-augmenting” (Azfar and

Cadwell 2002). The security of property rights depends upon the intelligent engagement of the state

(North 1990, 1994). Similarly, good institutions – as measured by low corruption, high bureaucratic

capacity, and other indicators – should foster stronger growth performance, all other things being

equal (Ades and Di Tella 1997; Keefer and Knack 1997; Mauro 1995). With these assumptions

granted, we may now proceed to a discussion of the role of regime type in economic policymaking.

There is good reason to suppose that the length of a country’s experience with democratic

rule is relevant for the intricacies of economic policymaking. First, the advent of democracy is

9
often associated with the installation of a new set of political elites, perhaps even a wholesale

replacement of politicians and bureaucrats that previously staffed the central government (Gifford

and Louis 1982; Jens-Hesse 1993). Let us take as given that political experience matters for effective

policymaking and that political experience is accumulated over time in a stable polity. (We imagine

that this is particularly the case with respect to economic policy, which tends to be more complex

and less amenable to intuition than other areas of public policy.) Politicians and bureaucrats in an

established democracy are likely to have served a long period of apprenticeship, either in an

established party or a government agency; they will have moved up within some institution’s ranks.

This does not guarantee that merit will govern appointment and promotion, but it does guarantee

that whoever occupies the commanding heights of government will be reasonably seasoned. It

follows that the longer a democratic polity is in place, the more experience its elites will have.

By itself, learning by policy-making elites suggests only that long-lived regimes will tend to

produce better policies than short-lived ones -- regardless of whether the regime in question is

democratic or autocratic. However, the very existence of an electorate may dramatically increase the

importance of learning in democratic regimes. That is, not only must governing politicians learn

what constitutes good policy. Voters must also learn to recognize good policy. There may even be

a third stage of learning in which politicians learn that voters have learned to distinguish good

policies from bad.

The most important lesson to be learned by democratically elected elites is that growth

performance matters for their political future and their historical legacy, as well as for the future of

their party or chosen successors. In new democracies, politicians frequently adopt short-term

policies intended to pay off political supporters and stimulate the economy during the electoral

season (Dornbusch and Edwards 1991). However, in established democracies, if elites and voters

have experienced a series of electoral and economic cycles and if political power is channeled into

10
well-established organizations (parties, bureaucracies, and so forth), longer time-horizons may

prevail. Voters may have directly experienced the effects of populist economic policies; if so, they

are likely to be skeptical of claims that soaking the rich, inflating the economy, abrogating debt

agreements, or massive expropriation of property will enhance their livelihoods (Remmer 1991;

Weyland 2002). This certainly seemed the case during the disastrous presidency of Fernando Collor

de Mello in Brazil (Rosenn & Downes 1999). Political elites, by the same token, will have had

occasion to familiarize themselves with the bottom-line of contemporary electoral politics.

Economic growth, so long as it is not monopolized by a few, is a salient political issue in most

national campaigns (Lewis-Beck and Stegmaier 2000). Thus, leaders may be inclined to impose

sacrifices over the short-term to facilitate stronger growth performance over the course of their

administration (Stokes 2001, 2002) and in the long run. Political leaders in a highly institutionalized

polity must be concerned with their party’s economic record in subsequent elections. The more

institutionalized the polity, under conditions of democratic accountability, the longer the time-

horizons of its leaders.

More generally, democracies institute a learning process among elites and masses in which

economic performance and electoral contests offer periodic corrections. Of course, this learning

process is driven not only by actual rewards and punishments meted out by the electorate. It is also

driven by anticipated rewards and punishments. In Sartori’s (1987: 152) well-chosen words: “Elected

officials seeking reelection (in a competitive setting) are conditioned, in their deciding, by the

anticipation (expectation) of how electorates will react to what they decide. The rule of anticipated

reactions thus provides the linkage between input and output, between the procedure (as stated by

Schumpeter) and its consequences.” Sartori (Ibid.) refers to this as a “feedback theory of

democracy.”

11
The longer a democracy has been in place, the further this learning curve will extend, all

other things being equal. We expect a slow transition away from a populist style of politics and

policymaking as democratic experience accumulates. As a result, countries with extensive

democratic histories should institute better (i.e., market-augmenting) policies than transitional

democracies or authoritarian regimes. These policies, in turn, should have strong effects on

aggregate growth performance. Empirical grist for this theoretical argument can be found in a series

of studies indicating that a country’s democratic history is a strong predictor of low corruption,

bureaucratic quality, property rights, and other good-governance outcomes (Gerring and Thacker

2004; Keefer 2003; Persson and Tabellini 2003; Treisman 2000).

POLITICAL CONSENSUS AND STABILITY. Thus far, we have outlined how a long-lived democratic

regime might enhance growth through its effect on capital allocation and formation, and through

society’s learning about what constitutes good policies. However, a large literature on democratic

overload posits that democracy engenders costly and destabilizing power struggles among various

subgroups (e.g., Crozier et al. 1975). As we argue below, this cost of democracy is one that may well

decrease over time.

The literature on democratization is replete with examples of the difficulties encountered by

newly democratizing countries – particularly when those countries are poor, ethnically divided, or

where the question of nationality is open to question (Chua 2003; Fein 1995; Mousseau 2001; Snyder

2000). Such countries are burdened with a surfeit of expectations, accumulated over many years.

Citizens have been told to expect great achievements from self-government, and they generally

expect these goods to materialize in a hurry. It is the fashion of political leaders during the long and

dangerous struggle for democracy to over-promise. Thus, when the transition finally occurs it is

greeted with enormous hopes and extravagant expectations. Transitions offer little preparation for

12
the hum-drum nature of everyday politics. Democracy experienced is never quite the same as

democracy envisioned (see, e.g., O’Donnell and Schmitter [1986] on desencanto). The process of give-

and-take among competing priorities may seem to barter away what had initially been gained, a

corruption of the pure democratic ideal into brokerage politics. Needless to say, such

disillusionment does not augur well for political stability and consensus.

In addition, democratization frequently stimulates a surge of demands on the part of

previously quiescent, and perhaps actively repressed, groups. These might be lower classes, excluded

ethnic or racial groups, or some other category of out-group (Eckstein 1989; Escobar and Alvarez

1992; Stepan 1989; Tarrow 1998). While beneficial in the long-run, the short-run effects of such

mobilizations from below may be destabilizing and may have adverse effects on the investment

climate, as in Bolivia during the early 1980s (Haggard & Kaufman 1995: 184-86).

If democracy survives its tumultuous youth, however, we anticipate that the extreme nature

of political conflict will moderate over time. A democratic political system is, by definition, open to

the inclusion of all sizeable social groups and interests. Once granted a taste of political power,

elites at the head of radical social movements may find it in their interest to join an existing party or

coalition. Moreover, the relatively open nature of deliberation in an established democracy may

diminish the appeal of conspiracy theories, which tend to flourish in the fog of authoritarian rule.

Generally speaking, social and political consensus should become more likely as democracies

become more open to the participation of citizens and as citizens accumulate experience under

democratic rule. Whatever centripetal tendencies are inherent in democracy are more likely to be in

evidence when those democratic arrangements have been in operation for some time. A regime is

unlikely to elicit much legitimacy among out-groups until they have been included in an elected

government. This requires at least one transfer of power, and probably awaits the outcome of

several electoral contests (Przeworki 1991). For these reasons, the thesis of democratic overload is

13
much more compelling when applied to new democracies than when applied to old. New

democracies tend to be boisterous, obstreperous affairs. Established democracies, by contrast, tend

to be more restrained. In particular, the norm of incremental change is more likely to be accepted.

Although the greatest threats to political stability presumably arise within segmented

societies where distrust among various social groups is high it is not necessary to presuppose a

conflict-ridden political environment to appreciate the possible long-term benefits of maintaining a

democratic framework. Politics, we suppose, involves formidable problems of coordination (Hardin

1999). Somehow, people must agree upon, or at least agree to respect, the imposition of society-

wide policy solutions that impose uneven costs and benefits. In an authoritarian setting, these

coordination problems are solved by the sovereign, Hobbes’ Leviathan. Rule by coercion, insofar as

it is successful, can be imposed without loss of time and without negotiation. The threat of force is

immediate. With democracies, things are different. To resolve problems of coordination, a

democratic polity must institutionalize procedures for negotiation (formal or informal) among rival

groups. Rule by consent, in order to become effective, takes time.

Thus, the problem of overload, insofar as it exists, probably does not arise from institutional

sclerosis (Olson 1982), but rather from insufficient institutionalization (Huntington 1968). When

political institutions are new and few, one rightly anticipates political conflict. Over time, however,

democracies institutionalize political conflict. This process of institutionalization – in Rustow’s

(1970) words, “habituation” – occurs through multiple arenas including political parties, legislatures,

committees, and electoral contests. It is an iterated process, and can only be conceptualized through

time. Institutionalization, therefore, should coincide with the achievement of greater political

stability, consensus, and social trust. Each of these factors presumably has positive implications for

investment and for long-term growth.

14
EMPIRICS

DEMOCRACY: CONCEPT AND MEASUREMENT

Our theoretical expectation is that the length of time a country is democratic – and, secondarily, the

degree of democracy found in that country – is likely to influence its economic performance. In order

to transform this expectation into a testable hypothesis we must face the difficult question of how to

conceptualize and measure the key theoretical concept, democracy. Evidently, democracy is a broad

concept, encompassing a wealth of possible attributes (Sartori 1987; Collier and Levitsky 1997). Any

plausible definition must not violate patterns of ordinary usage and must therefore integrate the core

meaning of democracy -- rule by the people. At the same time, in order to be useful a definition

must strive to capture the essentials of whatever theoretical argument the concept is intended to test.

Concepts are active tools, in addition to serving as (neutral) containers of language. Thus, our

definition, while resonating with ordinary usage, is not intended to capture all nuances of

“democracy” and should not be viewed as a general definition of that term.

Since our theoretical question is whether democracy has a positive effect on economic

performance, our definition of this difficult concept must attend to those aspects of democracy that

seem most likely to affect economic performance. So viewed, democracy is perhaps best

understood, following Joseph Schumpeter (1942/1950: 269), as “that institutional arrangement for

arriving at political decisions in which individuals acquire the power to decide by means of a

competitive struggle for the people’s vote.” This feature, rather than the closeness of the vote

(Vanhanen 1990) or actual turnover among ruling elites (Przeworski et al. 2000), seems mostly likely

to affect economic performance. The key question is whether the current in-group stands a realistic

chance of becoming an out-group in the near future. Competition may remain a potentiality; it need

not be actualized over the short-term. Indeed, members of a polity may freely choose to re-appoint

15
a ruling group over many years, and the differentials between winning and losing parties may be

quite large. Botswana, where the Botswana Democratic Party (BDP) has ruled since independence,

exhibits both of these features but is still, according to our reckoning, a country in which the ruling

elites face potential rejection at each election. Thus, we shy away from simple, mechanistic

indicators of democracy.

Instead, the concept of potential political competition compels us to consider a raft of

matters that are sometimes difficult to measure – they are, indeed, analogous to the equally fuzzy

notion of an “open” market. (How does one know for sure that a market is unconstrained?) With

respect to political markets, the following issues seem critical, but are by no means comprehensive.

Elected officials should be sovereign; non-elective bodies such as a military junta, monarchy, or

caste must not exercise real power behind the scenes. There should be regular elections, and these

elections must be open to all citizens (both as candidates and electors), subject to minor restrictions

having to do with age or residency. Suffrage should be broad, though not necessarily universal.

Political liberties should be extensive, particularly as pertains to the task of political organization and

campaigning. Election resources, including access to money, the media, and voters, should be fairly

allocated. Each of these is perhaps best understood as a matter of degrees (can one envision a

perfectly fair allocation of election resources?). This inclines us towards a continuous, rather than

dichotomous, empirical measure of democracy.5

Our central hypothesis concerns the temporally-dependent role of democracy, its long-term

potential to foster economic growth. Thus, we wish to capture not only differences in degree of

democracy-authoritarianism but also differences of duration across secular time. Therefore we

propose to measure democracy as a stock, rather than level, variable; that is, to measure the

accumulation of democratic experience, rather than a snapshot view of democracy.6 Since the

16
returns to democracy are presumably not linear – at some point the value of an additional year of

democracy should level off – we transform the democracy (stock) variable by its natural logarithm.

We employ the Polity IV “Polity2” variable as our principal measure of democracy (Marshall

and Jaggers 2000). The variable is constructed by measuring the extent to which democratic or

authoritarian “authority patterns” are institutionalized in a given country. It takes into account how

the executive is selected, the degree of checks on executive power, and the form of political

competition. This indicator is highly sensitive (it employs a 21-point scale), and offers extensive

country coverage (all sovereign polities excepting micro-states) and good historical coverage.

Moreover, it allows us to consider both the degree and duration of democracy in any given country-

year. The Polity dataset, however, imposes two serious costs. First, the rules used to create the key

variable, Polity2, are dizzyingly complex. The Polity User’s Manual makes a valiant effort to

explicate coding procedures, but the methods remain rather difficult to unpack. Second, there are

serious questions regarding measurement error in the index (Bollen and Paxton 2000; Bowman et al.

2004; Munck and Verkuilen 2002; Treier and Jackman 2003). Granted, questions might be raised

with respect to all extant, and all conceivable, democracy indices (see previous citations). Polity2 is

no worse than the rest, and probably better than the average. It is, indeed, the industry standard,

owing largely to the strengths noted above. Reassuringly, the Polity2 variable correlates highly with

other existing measures of democracy.7 There is no reason to suspect systematic errors in this index

that might affect the substantive findings of this study.

To create a stock measurement of democracy from this variable we simply add up each

country’s Polity2 score from 1900 to the present year. The year 1900 is chosen as a threshold year

ushering in a period a) in which mass democracy becomes a world-historical phenomenon (no

longer restricted to the US and a few European states), b) in which it is not unreasonable to assume

17
a causal relationship between democracy and growth, and c) in which the data exists to test such a

relationship.

Because the historical component of this index weighs heavily on our understanding of the

concept and because the Polity dataset ignores non-sovereign states in its coding procedures, we

supplement the Polity2 coding with our own coding of several nation-states that were previously

part of contiguous empires. The procedure is as follows. For each year that a nation-state belonged

to an imperial power it receives the same Polity2 score as its imperial ruler; e.g., Estonia receives the

same score as the Soviet Union from 1941 through 1990. We use this procedure only for nation-

states that were contiguous with the empire to which they belonged. We assume that contiguous

colonies are likely to be governed in the same manner as the imperial power itself, a dynamic less

likely to be true for overseas colonies.8

To correct for Polity2’s exclusion of micro-states, an exclusion that might also bias our

sample, we impute democracy scores for these excluded cases using other democracy indices that are

conceptually and empirically close to the Polity2 measure: a) the Freedom House Political Rights

indicator,9 b) Ken Bollen’s Liberal Democracy variable (Bollen 1993), c) Tatu Vanhanen’s

Competition variable (Vanhanen 1990), d) Arthur Banks’s Legislative Effectiveness variables (I and

II), and e) Banks’s Party Legitimacy variable (Banks 1994). These various measures of democracy

take into account the degree to which citizens can participate freely in the political process, the

extent of suffrage, the competitiveness of national-level elections, the degree of party

competitiveness, and the degree to which the legislature effects public policy.

These various additions to the core Polity2 variable increase the number of potential cases in

our analysis only slightly – by about 500 cases (compare models 1 and 2 in Table 2).

ANALYSIS

18
The relationship between democracy and growth may be tested in many ways. We take as our point

of departure a time-series cross-sectional format, with countries as our units of analysis. Even so,

this scarcely limits the methodological field since any of the multitudinous approaches employed in

current growth regressions might also be applied to this particular question (see Temple 1999 for a

recent review). The researcher faces choices about what time-intervals to consider, how to correct

for serial and spatial autocorrelation, and how to resolve issues of specification, simultaneity, and

endogeneity, among other matters.

Fortunately, there is general agreement about how to measure the dependent variable,

economic growth, which is usually understood as the percentage change in GDP per capita. We

employ the World Development Indicators (World Bank 2003) growth variable, measured in

constant dollars. Data for the 1950s is imputed using the Penn World Tables (PWT [Summers and

Heston 1991) 6.1 dataset (Chain index, constant dollars). Our choice of the WDI dataset as the

primary data source for indicators of country growth is motivated by two concerns. First, WDI

country coverage is considerably larger than offered by the PWT dataset. Second, for various

reasons explored by Nuxoll (1994) and Temple (1999: 118-19), the WDI indicator is probably the

best measure of growth performance.

Since data comprising the dependent and independent variables of interest are compiled at

annual intervals, and since significant changes in both right- and left-hand variables occur from year

to year, it makes sense to employ annual data in this analysis. Not to do so constitutes a waste of

information and is prone to the aggregation problem (how long should intervals of aggregated data

be?). So long as patterns of temporal autocorrelation and period-specific effects can be corrected,

we see no justification for aggregating data over five- or ten-year periods (or longer), as is sometimes

done. The unit of analysis in all the following data tables is therefore the country-year. (In the

19
appendix we offer a series of additional tests that employ data aggregated over 5- and 10-year

periods.)

Serial autocorrelation is a serious concern, as it is in most time-series formats. Our approach

assumes a one-period (AR1) temporal disturbance whose value is calculated on the basis of the

Durbin-Watson statistic. (Other methods of calculating rho do not have an appreciable effect on the

results of the regressions.) An alternative approach, with much to recommend it in certain situations,

employs a lagged dependent variable in order to de-trend the data (Beck 2001; Beck and Katz 1995).

We find little difference in substantive results when this alternative method is employed (see

Appendix). Since the interpretation of results is considerably less intuitive when a lagged dependent

variable is introduced on the right side (coefficients represent the change in trend of the dependent

variable), we prefer the more straightforward AR1 correction.

The most serious challenge to cross-country analyses of distal causal relationships is the

problem of simultaneity. Suppose, for instance, that a growth regression identifies Factor X as

positively contributing to growth. Almost invariably the skeptic can plausibly argue that this finding

is a consequence of the existence of some unmeasured Factor Z that impacts both growth rates and

Factor X. In short, other (unmeasured) factors that are correlated with democracy may account for

superior growth performance, rendering our results spurious. Indeed, it is plausible to suppose that

countries that are able to maintain a high level of democracy over a long period of time are also

blessed with other advantages – good institutions, good resources, and so forth. Needless to say, it

is difficult, perhaps even impossible, to measure all such country-specific (unit) effects. (This

supposition is borne out by various tests of spatial autocorrelation comparing the results of fixed

and random effects regressions -- e.g., Breusch and Pagan Lagrangian multiplier tests and Hausman

tests.)

20
To control for possibly severe unit effects (spatial autocorrelation) we employ a fixed-effect

format in all analyses. Although this precludes examining variation across countries, a loss of

explanatory leverage, it effectively removes many of the specification problems that plague cross-

country studies. Note that the fixed-effect format imposes a unique intercept for each country.

Factors that are for the most part constant across the time-period of interest (1950-2000), such as

geography, culture, and ethnicity, are eliminated by this research design. Thus, our results will suffer

from omitted variable bias only if the change in growth rate and the change in democracy stock are

both driven by some other (unmeasured) factor. We attempt to control for such factors in Table 1

(see below).

Endogeneity between democracy and growth is a potential problem wherever democracy is

considered as a level (rather than stock) variable. Democratic/authoritarian changes from year to

year may be contingent upon contemporaneous growth performance, creating circularity in the

argument. However, endogeneity is considerably less problematic when the measure of democracy

is itself longitudinal, extending 50-100 years back in time. As a further precaution, we lag democracy

(as well as all control variables) one year. Longer lags -- of 10 and 20 years respectively – are

introduced for our democracy variable in later tests (see Table 2).

Specification problems pervade all cross-country growth regressions (Levine and Renelt

1992; Sala-i-Martin 1997; Temple 1999). While the fixed-effect format handles the problem of

invariant controls (e.g., dummy variables registering a country’s location, its culture, its degree of

ethnic heterogeneity, its colonial history, and so forth), it does nothing to control for factors that

might vary through time. Our approach is to include in our benchmark model any factor that is

statistically significant in most robustness tests (conducted in other studies) and agreed-upon by

most writers on the subject. As it happens, this reduces the field to exactly one variable: GDP per

capita (natural logarithm), intended to capture the convergence hypothesis -- richer countries grow

21
slower, ceteris paribus.10 Recall that we are interested in the effect of democracy on a country’s

growth rate given its current level of economic wealth; so it makes sense to include this important

control, assuming that there are, in fact, convergence effects. This variable is drawn from the WDI

dataset (World Bank 2003), with missing cases from the 1950s imputed from the PWT 6.1 dataset

(Summers and Heston 1991).

Other controls are less obvious by virtue of their possibly endogenous relationship to

democracy, their lack of robustness, or their theoretical status. At the same time, it is vital that we

test as comprehensive a set of alternate controls as possible. These controls must encompass not

only those identified by the prodigious literature on economic growth but also those factors that

might affect the simultaneity problem discussed above. These control variables are introduced

seriatim and then in a series of “full” regressions, so as to test their individual and collective effects

on the variable of interest.

To summarize, our main estimation model is


g i ,t = α i + β ln( POLITYSTOCK i ,t −1 ) + γZ i ,t −1 + ε i ,t (1)

where gi,t is the growth rate in country i in year t, αi is a country-specific intercept term,
POLITYSTOCK i,t is the sum of Polity2 scores since 1900 (as described above) for country i as of
year t, Zi,t is a vector of control variables for country i in year t, and εi,t is the error term. As
discussed, we use an AR1 specification, so that εi,t = ρεi,t-1+υ i,t for some constant ρ, with υ i,t
independently and identically distributed over both countries i and dates t. The main object of
interest is the parameter β, which measures the relationship between the stock of democracy and the
growth rate.
It may be helpful to observe that equation (1) can be equivalently expressed in first-
difference form as
POLITYLEVELi ,t −1
g i ,t − g i ,t −1 = β ln(1 + ) + γ ( Z i ,t −1 − Z i ,t − 2 ) + (ε i ,t − ε i ,t −1 ) (2)
POLITYSTOCK i ,t − 2

22
That is, the change in a country’s growth rate is related to the level of democracy in the proceeding

year. In other words, democracy may only have a detectable effect on growth rates if it persists for a

significant period of time.

RESULTS
Table 1 presents a series of specification tests probing the performance of the main independent

variable of interest, democracy. We begin with a level measurement of democracy drawn from the

Polity IV dataset, as discussed above. We include only one all-purpose control in this reduced-form

equation, GDP per capita. The results confirm the standard finding: democracy has no statistically

significant effect on economic growth. This non-relationship is robust across a wide range of

democracy indicators (not shown). It matters not how one measures the level of democracy in a

given year; it still has no effect on subsequent economic performance. There is some variation in

effect depending upon the time-period of the analysis (annual, 5-year, or greater). However, none of

these results is particularly robust, as demonstrated in the Appendix.

[Table 1 about here]

We proceed in the remaining models (and in the remaining tables), to investigate democracy

as a stock variable, as described above. Model 2 presents what we regard as our benchmark model,

with only one control variable (replicating the specification of model 1). Model 3 drops that control

variable, demonstrating that Polity stock enhances growth performance even when convergence

effects are ignored. Model 4 adds a variable measuring trade-weighted growth per capita (each

country is assigned the mean value of the growth rate of all other countries in the world in that year,

weighted by their bilateral trade with the country in question). Model 5 adds dummy variables for

each year in the dataset (1950-2000), minus one. Because of the loss of degrees of freedom that this

23
imposes, and because it has only minor effects on the Polity stock coefficient, these controls are not

employed in later models.

Model 6 includes a whole series of time-varying controls that are a) common in the growth

literature, b) available for a large number of country-years, and c) might help to control for the

simultaneity problem that we have discussed. These include: Inflation, understood as annual

percent change in consumer prices (natural logarithm; World Bank 2003); Investment, understood as

the share of real GDP comprised by investment (PWT 6.1 [Summers and Heston 1991]); Foreign

direct investment, understood as net inflows as a share of GDP (World Bank 2003); Openness,

understood as imports and exports as a share of GDP (PWT 6.1 [Summers and Heston 1991]);

Government consumption, understood as the government share of real GDP per capita (PWT 6.1

[Summers and Heston 1991]); Population growth (World Bank 2003); Years independent (based on

data compiled by Robert Jackman [personal communication]; see also Jackman 1993); Regime

durability, understood as the number of years since the last three-point change in the composite

Polity2 score (Polity IV); Instability, including assassinations, general strikes, guerilla warfare,

government crises, purges, riots, revolutions, and anti-government demonstrations (these variables,

drawn from the Banks [1994] dataset, are added together to form a composite index [construction of

index by the authors]); Social conflict, including civil violence, civil war, ethnic violence and ethnic

war (these variables, drawn from Marshall et al. [2003], are added together to form a composite

index); Life expectancy (World Bank 2003), Illiteracy, as a logged function (World Bank 2003); a

trend variable (1950=1, 1960=2, . . .); an oil shock dummy (1950-1973=0, 1974-2000=1); and

decade dummies (not shown).11 In additional analyses (not shown), we introduced these control

variables seriatim into the benchmark equation so as to make sure that their individual effects did

not impair the performance of Polity stock.

24
Models 7-9 add a series of variables that are highly correlated with one another and reduce

the sample size considerably. For this reason, these controls are tested seriatim, rather than all

together. Expropriation measures the risk that the government will seize private property (Political

Risk Services 2004). Repudiation measures the risk that the government will disclaim its contracts

with the private sector (Political Risk Services 2004). Corruption measures the overall spread of

corruption in government (Political Risk Services 2004). All these variables are constructed so that

high scores indicate better performance (i.e., lower levels of expropriation, repudiation, or

corruption).

Although various specification tests affect the coefficient and standard error of Polity stock,

as one might expect, the variable retains statistical significance at the .01 level. When measured as a

stock variable, democracy appears to have a strong positive relationship to growth performance

regardless of the specification of the growth equation. While it is tempting to speculate on the

causal stories contained in various specifications of this growth regression, such speculation is not

immediately germane to our argument. Nor are these stories immediately apparent. It is by no

means clear, for example, which of the control variables should be regarded as exogenous, and

which endogenous, relative to democracy. Thus, we feel it is safer, all things considered, to conduct

further tests with a reduced-form model including only GDP per capita, as in model 2. This is the

only control which can claim some degree of theoretical consensus, is empirically robust, and – with

respect to the causal question at hand -- exogenous.

Having tested a series of control variables we turn now to alternate conceptualizations of our

key variable, Polity stock, and alternative samples. Table 2 begins with the benchmark equation

(model 1), which replicates model 2 in Table 1. This will provide the basis for easy comparisons

across alternative measures and samples. Model 2 excludes all cases for which we added or imputed

additional data for the Polity stock variable, and thus represents the stock variable drawn directly

25
from the Polity IV dataset (Polity2). Not surprisingly – since we have lost only 507 observations –

the Polity stock coefficient and standard error are stable.

[Table 2 about here]

It is still possible, however, that Polity stock is measuring something other than democracy

per se. One possibility is that this variable is acting as a proxy for the length of time a country has

been autonomous, i.e., the duration of a nation-state. It could be, in other words, that a high Polity

stock score is indicative of a long historical experience of sovereignty, rather than (or in addition to)

a long experience with democratic elections. We offered an initial test of this hypothesis in Table 1,

where we entered a control variable measuring the duration of sovereignty (the number of years a

country has been independent). Even so, our Polity stock variable is suspect since countries enter

the Polity IV dataset only when they become sovereign nations. Except in the case of contiguous

colonies (discussed above), we have no obvious way to treat non-nations in periods prior to

nationhood. This pertains especially to Africa, where most nations came into existence only in the

postwar period. In model 3, therefore, we re-calculate the benchmark regression including only

those countries for which complete data is available for the entire twentieth century (1900-), without any

imputations or additions. Although the number of countries in the dataset as well as the

corresponding number of observations drops to less than one third, the coefficient and standard

errors are fairly stable, suggesting that our results are not the product of peculiarities introduced by

the construction of the Polity stock variable.

Models 4 and 5 introduce longer lags than are employed in other regressions. Recall that all

regressions reported in Table 1 systematically lag independent variables by one time-period, i.e., one

year. Model 4 shows a ten-year lag on Polity stock and model 5 a twenty-year lag. Remarkably, the

value of Polity stock’s coefficients diminish only slightly, while the standard errors decrease,

26
indicating that the relationship between democracy (stock) and growth is not an instance of reverse

causality.

In model 6, we test the disaggregated effects of the two composite regime-characteristic

indicators, Democracy and Autocracy, that comprise the Polity2 index. Both are defined on a scale

ranging from 0 to 10, with higher scores representing more Democracy/Autocracy

respectively. Recall that in order to allow for the possibility of diminishing returns in the growth

benefits accruing from a country’s stock of democracy we use the logarithm of the sum of Polity

stock scores (from 1900 to the present) in all regressions. This specification is consistent with the

increasing costs of increases in the Autocracy stock. To maintain consistency with our standard

regression, in the disaggregated regression reported in model 6 we include the log of the sum of

Democracy scores as well as the log of one thousand minus the sum of Autocracy scores. (One

thousand is the maximum attainable Autocracy score over the course of the twentieth century.)

From model 6 it is apparent that Polity stock’s effect on growth is not simply an artifact of non-

autocracy, as suggested by some writers (Barro 1996, 1997). Rather, both the absence of autocracy

and the presence of democracy matters to growth performance.

Models 7-11 test the effects of democracy stock on growth when the concept of democracy

is understood as a dichotomous concept. In order to re-code the Polity2 variable as a dichotomous

concept we code a country-year democratic if it falls above 4 on the Polity2 scale (from -10 to +10).

The first variable, Continuous stock, is created by adding up all years of continuous democracy

experienced by a country in a given year. In 2000, Chile had experienced only 12 years of

continuous democratic rule, and thus receives a score of 12. The US, by contrast, had experienced

100 years of democracy since 1900, and therefore receives a score of 100. These sums are then

logged, as with our other stock measures, in conformance with the expectation of decreasing returns

(declining marginal utility). Results are depicted in model 7.

27
Democracy might also be considered as a cumulative phenomenon such that a country’s

economic performance depends upon the total number of democratic years since 1900 –

disregarding any authoritarian breaks. In this rendition, Chile receives credit for its previous

democratic history, prior to the Pinochet coup. Again, the summed score is logged. Results for this

variable are depicted in model 8.

Model 9 includes both dichotomous measures of democracy – continuous stock and

cumulative stock – together. Models 10 and 11 reprise models 7 and 8 with the addition of our

preferred measure of democracy, Polity stock, where democracy is understood as a scalar concept

(on the twenty-one point Polity2 scale).

Two important conclusions can be drawn from these five regressions (7-11). First, the

relationship to growth is stronger when democracy is considered as a cumulative, rather than

continuous, concept. Prior democratic experience matters, regardless of whether or not such

experience was interrupted by an authoritarian interlude. Second, the relationship of democracy to

growth is much stronger when democracy is measured as a scalar, rather than dichotomous, concept.

Polity stock swamps the effects of both dichotomous measures. Apparently, both the degree and the

duration of democratic experience matter when one considers the effect of democracy on growth.

Finally, in Table 3, we offer a series of split-sample tests of our major hypothesis. Evidently,

the results shown here are unlikely to be affected by the results of any single country. With a sample

of 178 countries across the world – virtually the entire universe of independent nation-states -- it is

virtually inconceivable that an individual country would constitute an influential case. However, it is

possible that particular regions of the world might affect the results reported here. In this series of

regressions, we exclude regional or socioeconomic blocs that might serve as an influential set of

cases in the benchmark equation: the Middle East (model 1), sub-Saharan Africa (model 2), Asia

28
(model 3), Latin America and the Caribbean (model 4), and the OECD (model 5). None of these

split-sample tests threaten the statistical significance of the Polity stock variable.

[Table 3 about here]

Note that the final model restricts the sample to the developing world. This offers a test of

the possibility that the effects measured in our aggregated results might be driven by peculiarities of

the – mostly Anglo-European – early democratizers. Model 5 indicates that the connection between

democracy and growth captured in the Polity stock variable is as strong among late-democratizers as

it is among early democratizers. Thus, it seems safe to conclude that our results are not the product

of idiosyncratic regional effects. This is not to say that the relationship between democracy and

growth is identical in all regions; most assuredly, it is not. It is merely to point out that the aggregate

results are not being driven by regional peculiarities.

THE DEMOCRATIC GROWTH EFFECT

We have shown thus far that the relationship between democratic stock and growth is robust in a

variety of plausible specifications and operationalizations. (Further tests are conducted in the

appendix.) We turn now to the question of its practical significance. Is the democratic growth effect

significant in real (policy) terms?

Let us consider the results of model 6 in Table 1, in which we control for a range of other

possible causal factors. We regard this model as offering a conservative estimate of causal effects

since many of the variables introduced as controls in this model may be endogenous to democracy,

and hence might be suppressing democracy’s true causal effect on growth.

Recall that in any given year a country’s Polity score can range from -10 to 10. As such, a

country lying in the middle of the distribution -- neither decisively democratic nor autocratic – will

have a Polity stock approximately equal to zero. If such a country subsequently experiences a full

29
decade of high-quality democracy, its Polity stock increases to approximately 100 points. To

estimate the predicted effect of this change on growth, note that because of the possibility of

negative values Polity stock enters the regression as ln(1000 + Polity stock). So from model 6 in

Table 1, the predicted growth impact of a decade of high quality democracy is approximately

7.7 * ( ln(1000+100) – ln(1000+0)) ≈7.7*0.1 = 0.77%.

The effect for a country that is exiting from a long period of autocracy is greater, translating into an

increase in the growth rate of 1.54% a year.12

Given the well-known cumulative effects of small increases in the growth rate, these changes

are significant. For instance, an increase in the annual growth rate from 2% to 2.77% reduces the

time needed to achieve a doubling of incomes from 35 to 25.4 years; an increase to 3.54% further

reduces the doubling period to 19.9 years.

Of course, many important questions remain to be considered. Among them, we shall

mention two. First, our econometric specification implies that the benefits of a year of democratic

experience last forever. This is unlikely to be the case. One would expect “investments” in

democratic stock to depreciate just as is the case for investments in physical (or for that matter,

human) capital. The Czech Republic’s current growth rate is surely more affected by its democratic

experience in 1998 than 1938. We have not attempted to model this, rather different, variety of

time-dependence. Another important issue concerns the pathways by which democratic stock might

influence a country’s current growth rate. In our theoretical discussion we suggested several

mechanisms by which such a causal relationship might be realized. However, we did not attempt to

assign a causal weight to any of these causal pathways, a task we must also defer to future work.

INSTITUTIONS IN TIME

30
This paper has demonstrated that the effect of regime type on growth is mediated by a country’s

secular-historical experience of democracy and authoritarianism. This claim stands in sharp contrast

to the conventional wisdom that there is at best no detectable relationship between growth and

democracy or, at worst, a negative relationship. Our somewhat surprising findings stem from a

conceptual insight that democracy/growth effects materialize from a country’s democratic stock

rather than its contemporaneous regime type. Thus, while a country’s level of democracy in a single

year has no measurable impact on its growth rate, its democratic experience over the course of

twentieth century is positively associated with growth. Long-term democracy leads to stronger

economic performance.

This finding, while important in its own right, may also prompt us to re-conceptualize the

causal role of other institutional variables common in the social sciences. Let us begin by

contrasting institutional explanations with other sorts of causal frameworks. When a causal

explanation is rooted in opinions, interests, or coercion (which may be considered a special variety

of interest-based argument), the relationship between a cause and its putative effect is generally

immediate and monotonic. A person’s interests (or opinions or coercive incentives) today should be

acted upon today and will remain the same tomorrow, unless some other factor changes the nature

of that interest (or value). Thus, temporal delays and non-monotonic causal relationships are

possible only if one moves beyond the realm of interests, opinions, and coercion tout court. For

example, if a person does not fully realize his or her true interests then some delay between cause

and effect might be anticipated. Similarly, if the nature of that interest changes over time, we can

imagine non-monotonic results. But this moves us beyond the realm of a purely interest-based

explanation.

With institutional explanations, however, the relationship between cause and effect is often

quite different. We might not expect a rule or norm to have strong immediate effects on the

31
patterns of behavior and expectations it is thought to generate. Instead, these causal relationships

may be conceptualized over a period of time, perhaps quite a long period of time. Moreover, these

slowly developing causal relationships may change over this period in non-monotonic ways. Thus,

we commonly speak of threshold effects, increasing returns (Pierson 2000), reactive sequences

(Mahoney 2000), and a wide variety of non-monotonic relationships. Arguably, the most important

and far-reaching institutional causes have the fewest immediate effects and the greatest non-

linearities. Constitutional frameworks, electoral laws, and property laws are all examples of

structural causes with distal and non-linear effects.

These observations are commonplace among historical-institutionalists, where time is

reckoned in decades, centuries, or millennia and where non-linear causal relationships are more or

less assumed (Mahoney and Rueschemeyer 2003; Pierson 2004; Steinmo et al. 1992; Waldner 1999).

However, the long-distance, non-linear approach to time is only beginning to take hold among

scholars who work with large-N datasets. Here, explanatory factors are usually considered as

contemporaneous events, or lagged (but still recent) events. The essentially contemporaneous

treatment of variables is entirely appropriate for proximate variables, whose effects are indeed

immediate. But it may not be an appropriate way to consider political institutions whose effects are

“long-memoried” (Pierson 2004). The causal effects of federalism, an electoral system, or a system

of executive organization (parliamentary or presidential) is likely to be mis-specified if one measures

only the current status of these constitutional factors or if one takes no account of their changing

relationship through time (Gerring and Thacker 2005).

To be sure, cross-country regression analyses have begun to introduce secular-historical

variables into the mix. It is now common to employ variables that measure a country’s colonial

history, climate, and geography (Acemoglu et al. 2001; La Porta et al. 1999). By usual construction,

these are static variables. But it is not self-evident that their effects are static through time. That is,

32
it may be important how long a country was colonized, or how long ago. While this is probably not

a meaningful question in the context of geographic variables, it is certainly a meaningful question

with respect to most humanly-created institutions (e.g., colonialism, democracy, electoral systems,

constitutions, property law regulations).

In this paper, we have proposed a particular way of thinking about the causal effect of social

institutions. The key insight is that institutions sometimes have cumulative effects, effects that are

noticeable only if an institution’s history is brought into view. Our analysis of the relationship

between democracy and growth suggests that the significant feature of this relationship is the

accumulated history of the variable in question. In this initial study we have employed the simple

summation of a leading measure of democracy to capture a country’s democratic history, and a

logarithmic transformation of that sum to represent non-linearities in the expected causal

relationship. More nuanced views of the ways in which political experience aggregates over time

clearly exist, and warrant investigation. Nonetheless, we hope our results are sufficient to establish

the viability of a cumulative approach to institutional causes. Since institutions are by definition

enduring, and since the effects of institutions are, almost by definition, constitutional – in the sense

of altering the perceptions and behavior of many actors at a systemic level – a wide array of causal

factors might be reconceptualized in this fashion.

Whatever the method of operationalization, the key concept of stock, or capital, may be useful

as a general term for this species of causal relationship since it allows one to conceptualize the

cumulative effects of an ongoing causal relationship. It is worth reflecting upon the fact that while

the concepts of investment capital and human capital are well known among economists, and the

concept of social capital increasingly common in all the social sciences, we lack an equivalent

concept pertaining to the political realm. Arguably, a country’s stock of democracy is one way of

measuring its political capital.

33
One concrete implication of this approach concerns the effect of institutional reforms,

reforms that currently occupy the attention of academics and policymakers in the developed and

developing world. If institutions contribute to a community’s political capital, and thereby to

secular-historical changes in policy outcomes, then it is inappropriate to judge the results of such

reforms on the basis of immediate policy gains. Although this renders the testing of hypotheses

more complicated – since we must await long-run results – it may temper the impatience of those

who jump to premature conclusions about the success or failure of regime change, neoliberal reform,

electoral reform, and other reforms of basic institutions. It is unrealistic to expect that the sweeping

changes brought about by such reforms would show instantaneous results. Indeed, while immediate

effects of institutional change are often negative – since such change introduces uncertainties and

information costs in the short run – positive changes are likely to take longer to materialize, since

they depend upon the establishment of a new equilibrium. The concept of political capital may help

policymakers to conceptualize, and thereby accurately account for, the true causal effects of

institutional reform.

34
REFERENCES

Acemoglu, Daron, Simon Johnson, and James A. Robinson. 2001. “Colonial Origins of

Comparative Development: An Empirical Investigation.” American Economic Review 91:5

(December), 1369-1401.

Ades, Alberto and Rafael Di Tella. 1997. “The New Economics of Corruption: A Survey and Some

New Results.” In Paul Heywood (ed), Political Corruption (Oxford: Basil Blackwell), 80-99.

Azfar, Omar and Charles Cadwell (eds). 2002. Market-Augmenting Government: The Institutional

Foundations for Prosperity. Ann Arbor: University of Michigan Press.

Banerjee, Abhijit V. 1999. “Land Reforms: Prospects and Strategies.” Conference Paper, Annual

World Bank Conference on Development Economics, Washington D.C.; and MIT Department

of Economics Working Paper No. 99-24.

Banks, Arthur S. 1994. “Cross-National Time-Series Data Archive.” Center for Social Analysis,

State University of New York at Binghamton. Binghamton, New York.

Barro, Robert J. 1996. “Democracy and Growth.” Journal of Economic Growth 1 (March), 1-27.

Barro, Robert J. 1997. Determinants of Economic Growth. Cambridge: MIT Press.

Baum, Matthew A., and David A. Lake. 2003. “The Political Economy of Growth: Democracy and

Human Capital.” American Journal of Political Science 47:2 (April), 333-47.

Beck, Nathaniel. 2001. “Time-Series Cross-Section Data: What Have We Learned in the Last Few

Years?” Annual Review of Political Science 4 (June), 271-93.

Beck, Nathaniel and Jonathan Katz. 1995. “What to Do (and Not to Do) with Time-Series-Cross-

Section Data in Comparative Politics.” American Political Science Review 89:3 (March), 634-47.

Besley, Timothy and Robin Burgess. 2000. “Land Reform, Poverty Reduction and Growth:

Evidence from India.” Quarterly Journal of Economics 115:2 (May), 389-430.

35
Bhalla, Surjit S. 1997. “Freedom and Economic Growth: A Virtuous Cycle?” In Axel Hadenius

(ed), Democracy’s Victory and Crisis (Cambridge: Cambridge University Press), 195-241.

Bleaney, Michael and Akira Nishiyama. 2002. “Explaining Growth: A Contest between Models.”

Journal of Economic Growth 7:1 (March), 43-56.

Bollen, Kenneth A. 1993. “Liberal Democracy: Validity and Method Factors in Cross-National

Measures.” American Journal of Political Science 37:4 (November), 1207-30.

Bollen, Kenneth A. and Pamela Paxton. 2000. “Subjective Measures of Liberal Democracy.”

Comparative Political Studies 33:1 (February), 58-86.

Bowman, Kirk, Fabrice Lehoucq, and James Mahoney. 2004. “Measuring Political Democracy:

Data Adequacy, Measurement Error, and Central America.” Working paper, Department of

Sociology, Brown University.

Chan, Sylvia. 2002. Liberalism, Democracy and Development. Cambridge: Cambridge University Press.

Chua, Amy. 2003. World on Fire: How Exporting Free Market Democracy Breeds Ethnic Hatred and Global

Instability. New York: Random House.

Collier, David and Robert Adcock. 1999. “Democracy and Dichotomies: A Pragmatic Approach to

Choices About Concepts.” Annual Review of Political Science 2:1 (June), 537-66.

Collier, David and Steven Levitsky. 1997. “Democracy with Adjectives: Conceptual Innovation in

Comparative Research.” World Politics 49 (April) 430-51.

Crozier, Michael J., Samuel P. Huntington, Joji Watanuki. 1975. The Crisis of Democracy. New York:

New York University Press.

de Janvry, Alain, Gustavo Gordillo, Jean-Philippe Platteau, and Elisabeth Sadoulet (eds). 2001.

Access to Land, Rural Poverty and Public Action. New York: Oxford University Press.

Dornbusch, Rudiger and Sebastian Edwards (eds). 1991. The Macroeconomics of Populism in Latin

America. Chicago: University of Chicago Press.

36
Dreze, Jean and Amartya Sen. 1989. Hunger and Public Action. Oxford: Clarendon Press.

Eckstein, Susan (ed). 1989. Power and Popular Protest: Latin American Social Movements. Durham, NC:

University of North Carolina Press.

Englehart, Neil and John Gerring. 2005. “The State: From Concept to Measurement.”

Forthcoming.

Escobar, Arturo and Sonia Alvarez (eds). 1992. The Making of Social Movements in Latin America.

Boulder, CO: Westview Press.

Fein, Helen. 1995. “More Murder in the Middle: Life-Integrity Violations and Democracy in the

World, 1987.” Human Rights Quarterly 17:1, 170-91.

Feng, Yi. 1997. “Democracy, Political Stability and Economic Growth.” British Journal of Political

Science 27:3 (July), 391-418.

Feng, Yi. 2003. Democracy, Governance, and Economic Performance: Theory and Evidence. Cambridge: MIT

Press.

Forbes, Kristin. 2000. “A Reassessment of the Relationship between Inequality and Growth.”

American Economic Review 90:4 (September), 869-87.

Gerring, John and Strom Thacker. 2004. “Political Institutions and Corruption: The Role of

Unitarism and Parliamentarism.” British Journal of Political Science 34:2 (April), 295-330.

Gerring, John and Strom Thacker. 2005. Good Government: A Centripetal Theory of Democratic

Governance. Forthcoming.

Gifford, Prosser and William Roger Louis (eds). 1982. The Transfer of Power in Africa: Decolonization,

1940-1960. New Haven: Yale University Press.

Gradstein, Mark and Branko Milanovic. 2000. “Does Liberte=Egalite? A Survey of the Empirical

Links Between Democracy and Inequality with some Evidence on the Transition Economies.”

CESifo Working Paper Series No. 261.

37
Guiso, Luigi, Paola Sapienza, and Luigi Zingales. 2004. “The Role of Social Capital in Financial

Development.” American Economic Review 94:3, 526-556.

Haggard, Stephan and Robert R. Kaufman. 1995. The Political Economy of Democratic Transitions.

Princeton: Princeton University Press.

Hardin, Russell. 1999. Liberalism, Constitutionalism, and Democracy. Oxford: Oxford University Press.

Huntington, Samuel P. 1968. Political Order in Changing Societies. New Haven: Yale University Press.

Jackman, Robert W. 1993. “Rationality and Political Participation.” American Journal of Political

Science 37:1 (February), 279-90.

Jens-Hesse, Joachim (ed.). 1993. Administrative Transformation in Central and Eastern Europe. London:

Blackwell.

Keefer, Philip. 2003. “Democratization and Clientelism: Why are Young Democracies Badly

Governed?” Presented at the annual meeting of the American Political Science Association,

Philadelphia (August).

Keefer, Philip and Stephen Knack. 1997. “Why Don’t Poor Countries Catch Up? A Cross-

National Test of an Institutional Explanation.” Economic Inquiry 35, 590-602.

Knack, Stephen and Philip Keefer. 1997. “Does Social Capital Have an Economic Payoff?: A Cross-Country

Investigation.” Quarterly Journal of Economics 112, 1251-88.

Krieckhaus, Jonathan. 2004. “The Regime Debate Revisited: A Sensitivity Analysis of Democracy’s

Economic Effect.” British Journal of Political Science (forthcoming).

Kurzman, Charles K., Regina W. Werum, and Ross E. Burkhart. 2002. “Democracy’s Effect on

Economic Growth: A Pooled Time-Series Analysis, 1951-1980.” Studies in Comparative

International Development 37:1 (January) 3-33.

La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert W. Vishny. 1999. “The

Quality of Government.” Journal of Economics, Law and Organization 15:1 (March) 222-79.

38
Leblang, David A. 1997. “Political Democracy and Economic Growth: Pooled Cross-Sectional and

Time-Series Evidence.” British Journal of Political Science 27:3 (July) 453-72.

Lenski, Gerhard. 1966. Power and Privilege: A Theory of Social Stratification. Chapel Hill: University of

North Carolina Press.

Levine, Ross and David Renelt. 1992. “A Sensitivity Analysis of Cross-Country Growth

Regressions.” American Economic Review 82:4 (September), 942-63.

Lewis-Beck, Michael and Mary Stegmaier. 2000. “Economic Determinants of Electoral

Outcomes.” Annual Review of Political Science 3 (June), 183-219.

Lipset, Seymour Martin. 1959. “Some Social Requisites of Democracy: Economic Development

and Political Legitimacy.” American Political Science Review 53:1 (March), 69-105.

Mahoney, James. 2000. “Path Dependence in Historical Sociology.” History and Theory 29:4 (August)

507-48.

Mahoney, James and Dietrich Rueschemeyer (eds). 2003. Comparative Historical Analysis in the Social

Sciences. Cambridge: Cambridge University Press.

Mainwaring, Scott, Daniel Brinks, and Anibal Perez-Linan. 2001. “Classifying Political Regimes in

Latin America, 1945-1999.” Studies in Comparative and International Development 36:1 (October) 7-65.

Marshall, Monty G. et al. 2003. “State Failure Problem Set, 1955-2001”

(http://www.cidcm.umd.edu/inscr/stfail/).

Marshall, Monty G. and Keith Jaggers. 2000. “Polity IV Dataset Project: Political Regime

Characteristics and Transitions, 1800-1999.” (http://www.bsos.umd.edu/cidcm/polity). Page

numbers cited refer to Dataset User’s Manual.

Mauro, Paolo. 1995. “Corruption and Growth.” Quarterly Journal of Economics 110:3 (August) 681-

712.

39
McGuire, James W. 2005. Politics, Policy, and Mortality Decline in East Asia and Latin America.

Forthcoming.

Meltzer, A.H. and S.F. Richards. 1981. “A Rational Theory of the Size of Government.” Journal of

Political Economy 89:5 (October), 914-27.

Mousseau, Demet Yalcin. 2001. “Democratizing with Ethnic Divisions: A Source of Conflict?”

Journal of Peace Research 38:5 (September), 547-67.

Muller, Edward N. 1988. “Democracy, Economic Development, and Income Inequality.”

American Sociological Review 53:1 (February), 50-68.

Munck, Gerardo L. and Jay Verkuilen. 2002. “Conceptualizing and Measuring Democracy:

Alternative Indices.” Comparative Political Studies 35:1 (February) 5-34.

North, Douglass C. 1990. Institutions, Institutional Change and Economic Performance. Cambridge:

Cambridge University Press.

North, Douglass C. 1994. “The Historical Evolution of Polities.” Working Paper, Department of

Economics, Washington University, St. Louis, MO.

Nuxoll, Daniel A. 1994. “Differences in Relative Prices and International Differences in Growth

Rates.” American Economic Review 84:5 (December), 1423-36.

O’Donnell, Guillermo and Philippe Schmitter. 1986. Transitions From Authoritarian Rule: Tentative

Conclusions about Uncertain Democracies. Baltimore: Johns Hopkins University Press.

Olson, Mancur. 1982. The Rise and Decline of Nations. New Haven: Yale University Press.

Persson, Torsten and Guido Tabellini. 2003. The Economic Effects of Constitutions. Cambridge: MIT

Press.

Pierson, Paul. 2000. “Increasing Returns, Path Dependence, and the Study of Politics.” American

Political Science Review 94:2 (June) 251-67.

40
Pierson, Paul. 2004. Politics in Time: History, Institutions, and Social Analysis. Princeton: Princeton

University Press.

Political Risk Services. 2004. “International Country Risk Guide” (accessed May 10, 2004).

http://www.prsgroup.com/icrg/icrg.html

Przeworski, Adam. 1991. Democracy and the Market. Cambridge: Cambridge University Press.

Przeworski, Adam and Fernando Limongi. 1993. “Political Regimes and Economic Growth.”

Journal of Economic Perspectives 7:3 (Summer), 51-69.

Przeworski, Adam and Fernando Limongi. 1997. “Modernization: Theories and Facts.” World

Politics 49:2 (January), 155-83.

Przeworski, Adam, Michael Alvarez, Jose Antonio Cheibub and Fernando Limongi. 2000.

Democracy and Development: Political Institutions and Material Well-Being in the World, 1950-1990.

Cambridge: Cambridge University Press.

Putnam, Robert D. (ed). 2002. Democracies in Flux. Oxford: Oxford University Press

Remmer, Karren L. 1991. “The Political Impact of Economic Crisis in Latin America in the

1980s.” American Political Science Review 85:3 (September), 777-800.

Rodrik, Dani (ed). 2003. In Search of Prosperity: Analytic Narratives on Economic Growth. Princeton:

Princeton University Press.

Rosenn, Keith S. and Richard Downes (eds). 1999. Corruption and Political Reform in Brazil: The Impact

of Collor’s Impeachment. Miami: North-South Center Press.

Rustow, Dankwart A. 1970. “Transitions to Democracy: Toward a Dynamic Model.” Comparative

Politics 2:3 (January), 337-363.

Sala-i-Martin, Xavier X. 1997. “I Just Ran Two Million Regressions.” American Economic Review 87:2

(May), 178-83.

41
Sala-i-Martin, Xavier X. 2002. “15 Years of New Growth Economics: What Have We Learnt?”

Unpublished manuscript, Columbia University, Department of Economics.

Sartori, Giovanni. 1987. The Theory of Democracy Revisited. Chatham: Chatham House.

Schumpeter, Joseph A. 1942/1950. Capitalism, Socialism and Democracy. New York: Harper & Bros.

Shapiro, Ian and Casiano Hacker-Cordon (eds). 1999. Democracy’s Value. Cambridge: Cambridge

University Press.

Snyder, Jack L. 2000. From Voting to Violence: Democratization and Nationalist Conflict. New York:

W.W. Norton.

Sokoloff, Kenneth L. and Stanley L. Engerman. 2000. “Institutions, Factor Endowments, and

Paths of Development in the New World.” Journal of Economic Perspectives 14:3 (Summer) 217-32.

Steinmo, Sven, Kathleen Thelen, and Frank Longstreth (eds). 1992. Structuring Politics: Historical

Institutionalism in Comparative Analysis. Cambridge: Cambridge University Press.

Stepan, Alfred P. 1989. Democratizing Brazil: Problems of Transition and Consolidation. Oxford: Oxford

University Press.

Stokes, Susan Carol. 2001. Markets, Mandates, and Democracy: Neoliberalism by Surprise in Latin America.

Cambridge: Cambridge University Press.

Stokes, Susan Carol (ed). 2002. Public Support for Market Reforms in New Democracies. Cambridge:

Cambridge University Press.

Summers, Robert and Alan Heston. 1991. “The Penn World Table (Mark 5): An Expanded Set of

International Comparisons, 1950-1988.” Quarterly Journal of Economics 106:2 (May) 327-68.

Tarrow, Sidney. 1998. Power in Movement: Social Movements and Contentious Politics. Cambridge:

Cambridge University Press.

Tavares, Jose and Romain Wacziarg. 2001. “How Democracy Affects Growth.” European Economic

Review 45:8 (August), 1341-78.

42
Temple, Jonathan. 1999. “The New Growth Evidence.” Journal of Economic Literature (March) 112-

56.

Treier, Shawn and Simon Jackman. 2003. “Democracy as a Latent Variable.” Presented to the

Political Methodology Meetings, University of Minnesota, Minneapolis-St.Paul (July).

Treisman, Daniel S. 2000. “The Causes of Corruption: A Cross-National Study.” Journal of Public

Economics 76:3 (June) 399-457.

United Nations Development Programme (UNDP). 2004. Overview: Democracy in Latin America:

Towards a Citizens’ Democracy. New York: United Nations Development Programme. Accessed

on May 16, 2004 at http://www.undp.org/democracy_report_latin_america/

Vanhanen, Tatu. 1990. The Process of Democratization: A Comparative Study of 147 States, 1980-88. New

York: Crane Russak.

Waldner, David. 1999. State Building and Late Development. Ithaca: Cornell University Press.

Warren, Mark E. 1999. Democracy and Trust. Cambridge: Cambridge University Press.

Weede, Erich. 1996. “Legitimacy, Democracy, and Comparative Economic Growth

Reconsidered.” European Sociological Review 12:3 (December), 217-25.

Weyland, Kurt. 2002. The Politics of Market Reform in Fragile Democracies: Argentina, Brazil, Peru, and

Venezuela. Princeton: Princeton University Press.

Woo-Cumings, Meredith (ed). 1999. The Developmental State. Ithaca: Cornell University Press.

Woolcock, Michael. 1998. “Social Capital and Economic Development: Toward a Theoretical

Synthesis and Policy Framework.” Theory and Society 27:2 (April), 151-208.

World Bank. 2003. World Development Indicators [data file]. Washington: World Bank.

43
APPENDIX

44
Appendix A:

Review of Alternative Approaches

In the body of the paper we defend a particular method for testing the democracy/growth relationship in a

time-series crossnational format. Our approach relies on annual data, fixed effects, an AR1 correction for

serial autocorrelation, and a century-long measurement of regime history (1900-). In this appendix we show

that our results are robust even when various elements of this methodology are altered.

Table A1 displays the results of simple regressions of growth rates over the 1990s on

democracy, democracy stock, and an assortment of standard control variables. As is clear, even in

this simplest version of the analysis, democracy stock is positively and significantly correlated with

growth rates. As usual, the level of democracy is uncorrelated with growth unless almost all controls

are omitted. Its effect disappears once the democracy stock variable is introduced into the equation

(model 3). Thus, the conclusion that democracy stock significantly impacts growth is not an artifact

of our panel data approach. Of course, the correlations of Table A1 may be a result of the omitted

variable bias --- which is precisely why we concentrate on estimates of fixed-effect regressions in the

main text.

In Table A2, we turn to a more complex set of methodological issues: fixed-effects versus random-

effects, annual versus five-year increments of data, the possible influence of OECD cases (tested this time in

a random-effects format), a lagged-dependent variable approach to modeling serial autocorrelation, the

possible peculiarities of the Polity stock variable, and a wide variety of static (time-invariant) control variables.

For each model with annual data we have followed our usual approach of measuring all independent

variables in the year prior to the dependent variable. For each model with five-year increments we have

maintained the same approach, this time measuring the independent variables in the first year of the period

under study. The dependent variable in this case is a five-year average of growth performance during the

subsequent period. So, in both cases the dependent variable is forward-lagged one time-period.

45
The standard method of correcting for autocorrelation is employed where annual data is used (AR1

error correction), but not when 5-year increments are used. (By virtue of the fact that we are dealing with 5-

year increments it should be less of a problem.) No correction for serial autocorrelation is usually necessary

when a lagged dependent variable is included, as in models 8 and 9, so none is employed.

Model 1 is a fixed-effect model with one control (GDPpc) and growth data aggregated across five-

year periods. Model 2 is a random-effects model with annual data and the same all-purpose control. Model 3

is a random-effects model with annual data that includes all large-N controls employed previously (see Table

2). (The inclusion of small-N controls such Capital expenditures, Expropriation, Repudiation, and

Corruption would significantly erode the sample size of the analysis.) Model 4 is a random-effects model

with five-year data increments and the same set of controls. Model 5 is a fixed-effects model with five-year

increments and all relevant (varying) controls. Models 6 and 7 replicate models 3 and 4, this time excluding

OECD cases. Models 8 and 9 test the Beck/Katz approach to TSCS analysis (Beck and Katz 1995), with a

lagged dependent variable included as a way of de-trending the data. Models 10 and 11 test another version

of the political capital thesis. Instead of measuring democracy stock since 1900, we instead employ a moving

sum variable that adds up democracy scores (from the same PolityIV dataset) over 20- and 50-year intervals.

In each of these various tests we find that the Polity stock variable retains statistical significance.

46
Table A1:
Cross-sections, 1990s
Model: 1 2 3 4 5 6
Polity level (ln) .561** .055 .311 -.024
(.260) (.337) (.382) (.397)
Polity stock (ln) 2.463*** 2.430*** 2.009** 2.023**
(.641) (.739) (.892) (.940)
GDPpc (ln) .062 -.099 -.108 -.467 -.915** -.915**
(.143) (.176) (.162) (.392) (.353) (.353)
Investment .056 .069* .069*
(.041) (.038) (.038)
Government .034 .028 .028
consumption (.043) (.042) (.042)
Openness .007* .007* .007*
(.004) (.004) (.004)
Population growth -17.527 -29.835 -29.845
(32.572) (29.255) (29.358)
Life expectancy .074 .100* .101
(.063) (.059) (.063)
Constant -.725 -14.887*** -14.714*** -2.208 -13.403** -13.499**
(1.298) (3.949) (4.325) (2.532) (5.889) (6.247)
Observations 167 167 167 126 126 126
Countries 167 167 167 126 126 126
R-squared .02 .11 .11 .14 .18 .18
Prob>F .0804 .0003 .0006 .0010 .0000 .0001

Ordinary least squares analysis. Units of analysis: country-year. Period of analysis: 1990-2000. Dependent variable: average
growth rate, 1990-2000. Independent variables all measured in 1990. Robust standard errors in parentheses. *** p<.01 ** p<.05
*p<.10 (two-tailed tests) Variables and procedures defined in the text.

47
Table A2:
Alternative Estimators and Models
Model: 1 2 3 4 5 6 7 8 9 10 11
Estimator: fe re re re fe re re fe fe fe fe
Periods of analysis: 5 years Annual Annual 5 years 5 years Annual 5 years Annual Annual Annual Annual
Non-
Sample: All All All All All Non-OECD All All All All
OECD
Polity stock (1900-) (ln) 5.584*** 3.339*** 2.281*** 2.034*** 10.192*** 2.316*** 2.286** 3.946*** 7.679***
(1.171) (0.476) (0.625) (0.686) (1.966) (0.834) (0.913) (0.774) (1.877)
Polity stock (20-year) 0.008**
(0.003)
Polity stock (50-year) 0.441***
(0.140)
GDPpc (ln) -4.319*** -0.910*** -1.078*** -1.211*** -5.483*** -1.113*** -1.200*** -2.766*** -6.152*** -4.267*** -5.723***
(0.326) (0.132) (0.264) (0.285) (0.668) (0.308) (0.327) (0.210) (0.622) (0.399) (0.540)
Growthpc (lagged dep var) 0.259*** 0.150***
(0.012) (0.020)
Inflation (ln) -0.131 0.064 -0.125 -0.077 0.096 -0.259***
(0.095) (0.100) (0.110) (0.110) (0.115) (0.095)
Investment -0.061*** -0.049** -0.051* -0.058** -0.037 -0.090***
(0.023) (0.022) (0.027) (0.027) (0.026) (0.026)
Foreign investment 0.166*** 0.027 0.093* 0.181*** 0.024 0.180***
(0.054) (0.051) (0.053) (0.062) (0.059) (0.052)
Openness 0.013*** 0.012** 0.016* 0.010* 0.009 0.040***
(0.004) (0.005) (0.009) (0.005) (0.006) (0.008)
Government consumption -0.029* -0.033** -0.014 -0.018 -0.023 0.002
(0.016) (0.017) (0.022) (0.019) (0.020) (0.021)
Population growth 4.209 -45.746*** -44.257** 5.913 -43.724** 23.920
(13.368) (17.745) (21.849) (15.293) (21.117) (14.647)
Years independent 0.003 0.002 0.131 0.003 0.003 0.094
(0.003) (0.003) (0.238) (0.004) (0.004) (0.282)
Regime durability -0.014* -0.012 -0.011 -0.017 -0.025* 0.003
(0.008) (0.009) (0.014) (0.012) (0.013) (0.013)
Instability -0.076*** -0.003 -0.002 -0.092*** -0.004 -0.095***
(0.025) (0.023) (0.023) (0.031) (0.029) (0.024)
Social conflict -0.233 0.338 0.057 -0.152 0.215 -1.045*
(0.556) (0.552) (0.634) (0.644) (0.641) (0.606)
Life expectancy 0.161*** 0.152*** 0.215*** 0.167*** 0.159*** 0.184***
(0.031) (0.034) (0.059) (0.036) (0.039) (0.050)
Illiteracy (ln) 0.130 0.024 0.637* 0.171 -0.013 0.712**
(0.178) (0.188) (0.342) (0.238) (0.247) (0.334)
Trend 0.032 0.098** -0.028 0.010 0.055 -0.025
(0.043) (0.048) (0.243) (0.053) (0.061) (0.284)
Oil shock -1.178** -0.948* -0.407 -1.068* -0.659 -0.831*
(0.490) (0.509) (0.505) (0.596) (0.620) (0.430)
Latitude (ln) -0.083 -0.117 0.014 -0.205
(0.391) (0.438) (0.496) (0.542)
English legal origin -0.012 -0.007 -0.009 -0.003
(0.008) (0.009) (0.009) (0.010)
Muslim -1.062 -0.755 -0.943 -0.855
(0.783) (0.889) (0.974) (1.085)
Ethnic fractionalize (Alesina) 3.273*** 3.493*** 3.259*** 3.354***
(0.695) (0.784) (0.926) (1.019)
East Asia 0.849 0.840 0.876 0.700
(0.812) (0.922) (0.930) (1.031)
Middle East -0.948* -0.929 -0.923 -1.362*
(0.523) (0.586) (0.727) (0.794)
Latin America 0.403* 0.219 0.247 0.042
(0.221) (0.253) (0.254) (0.283)
Decade dummies Yes Yes Yes Yes Yes Yes
Constant -4.342 -14.296*** -14.702*** -15.045*** -52.568*** -15.120** -15.711** -5.015 -25.225 34.507*** 47.665***
(8.309) (3.095) (4.634) (5.039) (19.489) (5.956) (6.392) (5.356) (20.770) (1.882) (2.296)
Observations 1084 6107 2459 492 509 1892 378 5980 2516 4309 2660
Countries 180 181 124 117 119 98 91 180 126 172 95
R-squared 0.18 0.00 0.08 0.27 .34 0.08 0.28 0.10 0.14 0.03 0.04

fe = fixed-effects. re = random effects. Period of analysis: 1990-2000. Dependent variable: average growth rate, 1950-2000. All
predictors lagged one time-period. Standard errors in parentheses. *** p<.01 ** p<.05 *p<.10 (two-tailed tests) Variables and
procedures defined in the text.

48
Descriptive Statistics for Key Variables

Countries Obs Mean Std. Dev. Min Max


Polity Stock 200 8468 -63.75 344.54 -930.10 1010.00
Polity Stock (ln) 200 8468 6.77 .39 4.25 7.61
GDPpc (ln) 190 6903 7.45 1.52 3.88 10.94
Investment 171 5860 16.82 9.89 -3.60 60.07
Government consumption 171 5860 19.00 12.12 1.29 116.89
Openness 171 5863 65.72 45.31 3.15 473.86
Population growth 201 8363 .02 .02 -.44 .22
Life expectancy 195 7130 60.92 11.88 31.22 81.07
Growth pc (trade-weighted) 189 7052 2.19 1.75 -10.56 10.05
Inflation 161 4629 1.94 1.44 -5.21 10.08
Foreign direct Investment 170 4024 1.93 4.89 -82.81 145.21
Instability 195 6751 -.08 4.25 -2.21 68.50
Social conflict 188 6206 2.09 .22 2.00 3.45
Years independent 222 11024 39.52 63.90 .00 453.00
Regime durability 164 6493 20.95 23.41 .00 100.00
Illiteracy (ln) 176 7076 2.59 1.71 -1.61 4.61
Primary education 185 2294 93.59 24.10 3.01 174.69
Corruption 124 1575 3.30 1.46 .00 6.00
Repudiation 124 1543 5.98 2.27 .50 10.00
Expropiation 124 1543 6.72 2.23 .50 10.00
Capital Expenditure 159 2885 18.31 12.34 .00 68.01
Sample period: 1950-2000

49
Bivariate Correlations of Key Variables

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
1. Polity Stock (ln) 1.00
2. GDPpc (ln) .41 1.00
3. Investment .31 .57 1.00
4. Governm. Consumption -.19 -.19 -.24 1.00
5. Openness .08 .29 .22 .20 1.00
6. Population growth -.22 -.49 -.24 .06 -.16 1.00
7. Life expectancy .15 .78 .57 -.12 .28 -.50 1.00
8. Growth pc global mean .04 -.07 .05 -.16 -.13 .04 -.19 1.00
9. Inflation (ln) -.17 -.12 -.11 .15 -.07 .00 -.04 -.13 1.00
10. Foreign direct invest. .06 .10 .18 .14 .38 -.12 .12 .03 -.07 1.00
11. Instability .00 -.03 -.01 -.05 -.24 .03 .00 .02 .12 -.09 1.00
12. Social conflict -.07 -.22 -.11 .07 -.20 .09 -.13 -.09 .14 -.10 .35 1.00
13. Years independent .16 .34 .20 -.14 -.23 -.21 .28 -.06 .05 -.08 .17 .06 1.00
14. Regime durability .35 .56 .34 -.19 .12 -.28 .46 -.07 -.17 .07 -.12 -.09 .27 1.00
15. Illiteracy .24 -.52 -.32 .13 -.26 .59 -.67 .05 -.11 -.12 .00 .13 .01 -.03 1.00
16. Primary education .17 .43 .34 -.06 .19 -.26 .56 -.04 -.04 .18 -.04 -.08 .18 .18 -.37 1.00
17. Corruption .45 .64 .51 -.19 .21 -.45 .53 .00 -.10 .12 -.14 -.27 .29 .54 -.36 .29 1.00
18. Repudiation .39 .69 .60 -.34 .27 -.51 .60 .02 -.27 .20 -.11 -.28 .30 .42 -.42 .39 .64 1.00
19. Expropriation .33 .62 .52 -.32 .28 -.49 .55 .05 -.22 .23 -.10 -.30 .23 .38 -.45 .37 .62 .86 1.00

50
TABLES AND FIGURES

51
Figure 1:
Hypothesized Causal Pathways

Physical, Human,
and Social Capital

Democracy Economic
stock growth

Political Capital:
a) Better Policies &
Institutions
b) Stability, Consensus

52
Table 1:
Specification Tests

Model: 1 2 3 4 5 6 7 8 9
Polity level -.030
(.024)
Polity stock (ln) 6.188*** 4.704*** 4.843*** 6.336*** 7.677*** 9.938*** 10.282*** 11.022***
(1.031) (1.058) (1.005) (1.053) (1.859) (3.717) (3.726) (3.614)
GDPpc (ln) -3.606*** -3.920*** -3.648*** -5.268*** -8.351*** -11.317*** -11.562*** -11.594***
(.302) (.303) (.296) (.393) (0.784) (1.607) (1.624) (1.597)
Growth pc .536***
(trade-weighted) (.049)
Inflation (ln) -0.205* -0.217 -0.231 -0.248*
(0.109) (0.141) (0.141) (0.138)
Investment -0.087*** -0.207*** -0.198*** -0.193***
(0.031) (0.055) (0.054) (0.053)
Foreign investment 0.166*** 0.221** 0.217** 0.209**
(0.058) (0.096) (0.097) (0.095)
Openness 0.049*** 0.086*** 0.088*** 0.089***
(0.009) (0.015) (0.014) (0.014)
Government 0.000 0.016 0.013 0.017
consumption (0.025) (0.037) (0.037) (0.036)
Population growth 27.915* 43.085* 40.206 36.348
(15.207) (25.168) (25.106) (24.724)
Years independent 0.032 0.082 0.110 0.090
(0.117) (0.091) (0.089) (0.084)
Regime durability 0.018 0.015 0.015 0.014
(0.015) (0.023) (0.023) (0.023)
Instability -0.101*** -0.151*** -0.150*** -0.148***
(0.026) (0.039) (0.039) (0.038)
Social conflict -1.966*** -2.923*** -3.061*** -3.347***
(0.714) (1.064) (1.060) (1.056)
Life expectancy 0.193*** 0.060 0.068 0.128
(0.063) (0.138) (0.139) (0.136)
Illiteracy (ln) 0.902** -1.008 -1.609 -2.884
(0.418) (2.334) (2.295) (2.143)
Trend 0.041 0.647 0.668 0.637
(0.123) (0.466) (0.467) (0.461)
Oil shock -1.260**
(0.516)
Expropriation 0.271*
(PRS) (0.163)
Repudiation 0.163
(PRS) (0.167)
Corruption -0.213
(PRS) (0.286)
Annual dummies Yes
Decade dummies Yes Yes Yes Yes
Constant 28.753*** -11.204** -30.353*** -5.249 -4.734 -5.018 9.919 8.722 7.589
(1.538) (4.811) (5.020) (4.858) (4.827) (8.365) (11.852) (11.874) (11.447)
Observations 5926 5926 6082 5720 5926 2390 1010 1010 1037
Countries 180 180 186 175 180 121 91 91 91
Rsq (within) .03 .03 .00 .05 .07 0.10 0.15 0.15 0.14
Prob>F .0000 .0000 .0000 .0000 .0000 0.0000 0.0000 0.0000 0.0000

Fixed effect regressions with AR(1) disturbance. Units of analysis: country-year. Period of analysis: 1950-2000. DV: annual per capita growth
rate. All predictors lagged one year. Standard errors in parentheses. *** p<.01 ** p<.05 *p<.10 (two-tailed tests) Variables and procedures
defined in the text.

53
Table 2:
Polity Stock, Variously Operationalized
Model: 1 2a 3b 4 5 6 7 8 9 10 11
GDPpc (ln) -3.920*** -1.554*** -1.292*** -4.013*** -3.840*** -3.710*** -3.829*** -4.039*** -4.012*** -3.990*** -4.188***
(.303) (.190) (.245) (.343) (.365) (.335) (.318) (.322) (.324) (.317) (.321)
Polity stock (ln) 6.188*** 7.327*** 4.753*** 6.089*** 5.805***
(1.031) (1.030) (1.682) (1.049) (1.049)
Polity stock (ln), 10-year lag 3.639***
(.827)
Polity stock (ln), 20-year lag 3.952***
(.692)
POLITY COMPONENTS:
Democracy stock (ln) 1.173***
(.221)
Autocracy stock (reverse scale, ln) 5.494***
(1.239)
DICHOTOMOUS MEASURES:
Continuous stock (ln, Polity>4) .274* -.160 .102
(.164) (.221) (.165)
Cumulative stock (ln, Polity>4) .734*** .877*** .520**
(.221) (.296) (.222)
Constant -11.204** -37.009*** -20.248*** 7.826** 5.505** -11.910* 30.042*** 30.694*** 30.449*** -10.135** -7.481
(4.811) (4.650) (6.854) (3.699) (2.746) (6.157) (1.574) (1.559) (1.575) (4.950) (4.920)
Observations 5926 5419 6.854 5229 4396 5285 5926 5926 5926 5926 5926
Countries 180 156 53 179 173 156 180 180 180 180 180
Rsq (within) .03 .02 .01 .03 .03 .03 .03 .03 .03 .03 .03
Prob>F .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000 .0000

Fixed effect regressions with AR(1) disturbance. Units of analysis: country-year. Period of analysis: 1950-2000. DV: annual per capita growth rate. All predictors lagged one year.
Standard errors in parentheses. *** p<.01 ** p<.05 *p<.10 (two-tailed tests) Variables and procedures defined in the text. a: Includes no imputed scores for Polity2. b: Includes only
countries for which data for Polity stock is available for the non-imputed core variable (Polity2) over the course of the entire twentieth century (Polity2_complete==1).

54
Table 3:
Split-Sample Tests
Model: 1 2 3 4 5
Latin Am/
Excluding: Middle East Africa Asia OECD
Caribbean
Polity stock (ln) 3.616*** 5.474*** 9.626*** 6.496*** 6.291***
(1.312) (1.042) (1.134) (1.137) (1.298)
GDPpc (ln) -3.639*** -3.290*** -5.831*** -3.681*** -4.598***
(.320) (.316) (.375) (.333) (.407)
Constant 4.051 -9.076* -20.344*** -15.226*** -9.064
(5.876) (4.676) (5.153) (5.260) (6.244)
Observations 5292 4292 5192 4685 4576
Countries 161 131 157 145 147
Rsq (within) .02 .03 .05 .03 .03
Prob>F .0000 .0000 .0000 .0000 .0000

Fixed effect regressions with AR(1) disturbance. Units of analysis: country-year. Period of analysis: 1950-2000. DV: annual per
capita growth rate. All predictors lagged one year. Standard errors in parentheses. *** p<.01 ** p<.05 *p<.10 (two-tailed tests)

55
NOTES

1
This research was funded by a generous grant from the Frederick Pardee Center for the

Study of the Longer-Range Future at Boston University. We are grateful to Carola Moreno for her

research assistance in this project.


3
Barro (1997: 58) finds that “growth is increasing in democracy at low levels of democracy,

but the relation turns negative once a moderate amount of political freedom is attained.” A few

recent studies find a positive overall relationship between democracy and growth (e.g., Bhalla 1997;

Leblang 1997), but this is not the standard finding.


4
See Lee Kuan Yew, quoted in The Economist (August 27, 1994: 15).
5
On conceptualizing and measuring democracy, see, Collier and Adcock (1999), Mainwaring

et al. (2001), Munck and Verkuilen (2002), Przeworski et al (2000).


6
Among extant studies of democracy and growth we have found only a few that approach

the concept of democracy over time (e.g., Weede 1996), and none that stretches back to encompass

the whole twentieth century. Persson and Tabellini (2003) look at the effect of democracy --

conceptualized as a dichotomous concept over the twentieth century -- on various economic policies

and outcomes, but not on growth per se.

7 Correlations between Polity2 and other democracy indices (introduced below in the text)

are as follows: “Political Rights” (Freedom House) = -.85; “Liberal Democracy” (Bollen) =.92;

“Democracy index” (Vanhanen) = .85.

8 This re-coding affects the following countries: Albania (1900-1912, Ottoman Empire),

Andorra (1900-present, France), Armenia (1900-1990, Russia/USSR), Azerbaijan (Russia/USSR

1900-1990), Belarus (Russia/USSR, 1900-1990), Bosnia-Herzegovina (1908-1917, Austria-Hungary;


Yugoslavia 1929-1991), Croatia (1900-1917, Austria-Hungary; Yugoslavia 1929-1991), Czech

Republic (1900-1917, Austria-Hungary), Slovakia (1900-1917, Austria-Hungary), Estonia (1900-1916

and 1941-1990, Russia/USSR), Finland (1900-1916, Russia), Georgia (1900-1990, Russia/USSR),

Iraq (1900-1917, Ottoman Empire), Israel (1900-1917, Ottoman Empire), Kazakhstan (1900-1990,

Russia/USSR), Kyrgyzstan (1900-1990, Russia/USSR), Latvia (1900-1917 and 1941-1990,

Russia/USSR), Lithuania (1900-1917 and 1941-1990, Russia/USSR), Macedonia (1922-1990,

Yugoslavia), Moldova (1900-1945, Romania; 1946-1990, USSR), Mongolia (1900-1920, China),

Bangladesh (1947-1971, Pakistan), Slovenia (1900-1917, Austria-Hungary; Yugoslavia 1929-1991),

Syria (1900-1917, Ottoman Empire), Tajikistan (Russia/USSR, 1900-1990), Turkmenistan (1900-

1990, Russia/USSR), Ukraine (1900-1917 and 1920-1990, Russia/USSR), Uzbekistan (1900-1990,

Russia/USSR), and East Timor (1976-1999, Indonesia).


9
Freedom in the World, survey methodology, on the Freedom House web site:

www.freedomhouse.org/research/freeworld/2000/methodology.htm
10
A recent reevaluation of cross-country growth empirics concludes that the log of GDP per

capita is the only variable that is robust across all models (Bleaney and Nishiyama 2002: 45).
11
We interpolate missing values for illiteracy, life expectancy, and population growth in

order to maintain a consistent sample (failing to do so would have significantly reduced the sample

size and perhaps biased our results).

12 Consider a country that has experienced fifty years of severe dictatorship. The effect of a

decade of full democracy on the log of Polity stock would be ln(600)-ln(500) ≈ 0.2.

57