State history and economic development: evidence from six millennia

Abstract

The presence of a state is one of the most reliable historical predictors of social and economic development. In this article, we complete the coding of an extant indicator of state presence from 3500 BCE forward for almost all but the smallest countries of the world today. We outline a theoretical framework where accumulated state experience increases aggregate productivity in individual countries but where newer or relatively inexperienced states can reach a higher productivity maximum by learning from the experience of older states. The predicted pattern of comparative development is tested in an empirical analysis where we introduce our extended state history variable. Our key finding is that the current level of economic development across countries has a hump-shaped relationship with accumulated state history.

1 Introduction

History has shown that economic development often thrives in states where governments guarantee the rule of law and provide public goods for their citizens. In order to reach a deeper understanding of why some countries have good government and others do not, social scientists have become increasingly interested in studying the long-run patterns of institutional development within states. The roots of countries’ contemporary failures or successes have often been traced back to “critical junctures” far back in history.¹

In this paper, we analyze how state development has interacted with economic development. More specifically, we attempt to make two distinct contributions to the literature. First, we provide a complete state history index from its first origin around 3500 BCE up until the present day. Initially developed by Bockstette et al. (2002) for 159 countries, the index covered the period 1–1950 CE. We extend the index from 1 CE backwards in time to the first origins of states around 3500 BCE and also code the 1951–2000 CE period, which was previously missing from the time series.

Second, we investigate how our extended state history index is related to indicators of long-run economic development. The key hypothesis from our theoretical framework is that modern levels of productivity and population density should have a hump-shaped relationship with the extent of state history. We expect non-linear effects of state experience in the pre-industrial era, and we predict it develops into a hump-shaped relationship by 2000 CE. In the empirical section, we then confirm that the relationship between our state history index and current levels of economic development has the shape of an inverted u, implying that countries with very much or very little state experience have the least developed economies whereas the richest countries have intermediate state history scores.

For the first of these objectives—the creation of a state history index for the BCE-period, we follow the methodology in the original effort by Bockstette et al. (2002). This combines three dimensions of state development: (1) The existence of a state above tribal level; (2) Whether rule was internally or externally based (i.e. whether a country’s territory had an autonomous government or was ruled partly or fully by an authority outside of its borders); (3) How much of its territory was under the control of a government (as opposed to multiple competing governments and regions still lacking state presence). The three indicators were coded for each of the 159 countries in our sample and for each 50-year period from the origin of the first states around 3500 BCE, yielding a panel data set with 17,490 country-period observations. The details of the sources for and construction of the index are described further below.

Our second objective hinged crucially on extending the state history data initially compiled by Bockstette et al. (2002). Their study was the first to show a significant correlation between state history and recent growth rate and between state history and income level. The numerous studies that followed strengthened the evidence that current development is positively related to state experience. Although subsequent versions of the index used in these papers expanded the set of countries, none coded the history of states BCE.

With these developments in mind and with the new data on the extended state history index, we revisit the relationship between the degree of exposure to state institutions and current output. We show that the relationship between state history and current income per capita across countries is hump-shaped rather than linear, and that this is due to the inclusion of state experience before the Common Era. Thus, in addition to young, inexperienced states, very old states also incur economic disadvantages relative to states with around 2000 years of state experience.

Our inquiry is supported by the empirical observation that countries having long state experience like Iraq, Turkey and China are poorer today than younger states like Britain, Denmark and Japan, a fact that remained unexplained in previous work. The early experience of the former was uncoded in the previous data, which effectively forced countries having much more state experience to take similar values to ones having intermediate levels, such as England (the UK).

Building on previous literature, we contribute additional knowledge about the influence of early political and societal development on modern economic development. In a stylized theoretical framework, we argue the earliest states developed the fiscal capacity and coordination needed to achieve increases in productivity, but ultimately limited that productivity due to overcentralization. Although earlier states became stagnant, younger states were able to learn from them and surpass their productivity before they reached stagnation themselves. By contrast, very young states early in the process of building fiscal and institutional capacity, are at a relative disadvantage. Thus, along with young states, a very long state experience also comes with economic disadvantages relative to countries with intermediate state experience. We show that this more complex relationship starts to be visible with respect to economic development indicators (population density and urbanization) and technology adoption in 1500 CE, but it clearly reveals itself in 2000 CE economic performance. Moreover, the relationship for current outcomes is further strengthened when adjusting the index for the ancestral lines of post-1500 migrant populations.²

The work clearly involves several methodological challenges. For instance, how should a state be defined? In this regard, we follow the tradition of Service (1962), Carneiro (1981), Johnson and Earle (2000) and others, distinguishing between bands, chiefdoms, and full-fledged states. Unlike the other forms of governments, states are further characterized by a centralized government with the ability to collect taxes, enforce laws, and mobilize forces for war. Using this definition, the literature seems to be in rough agreement about the time when states arise in different countries. Accompanying this paper is an extensive online data appendix where we motivate the coding for each country-period observation.

Another issue concerns the unit of analysis, which is the territory delimited by the modern-day country borders of 159 contemporary countries in the sample. It is a well-known fact that the borders of current countries sometimes have very little resemblance with the geopolitical logic in ancient times. As discussed in Sect. 5.4, several factors have played decisive roles in the reconfiguration of borders through history, including military conquest and colonization by foreign powers. We discuss the potential risks associated with shifting borders, in terms of retroactive measurement error for historical outcomes and bias in the estimates for contemporaneous outcomes.

A potential alternative to using country borders could have been to divide the world into equal-sized grid cells (or “virtual countries”) and then study the history of each such cell. This would however entail a separate set of coding challenges that we leave for future work. The problem of endogenous borders is arguably somewhat mitigated in regressions with contemporaneous outcome variables since most countries actually have changed their borders throughout history. We would further argue that to the extent that researchers are interested in tracking the histories of countries in order to understand contemporary levels of development, the modern configuration of countries is still a natural point of departure.

The paper is organized as follows: In Sect. 2, we provide an overview of the literature on the definition of a state and the relationship of state history to economic development. In Sect. 3, we present our theoretical framework. In Sect. 4 we present the new data and the principles guiding its construction. In Sect. 5, we carry out an econometric analysis of the relationship between economic development and state history. Section 6 concludes.

2 Literature review

2.1 State history and economic development

It is a well established empirical fact that history has shaped the contemporary economic development of nations in numerous ways. Whether initial biogeographic endowment and transition to agriculture (e.g. Hibbs and Olsson 2004; Olsson and Hibbs 2005; Galor and Moav 2007) or past technology adoption (Comin et al. 2010), early and productive starts have been typically shown to translate into better income and institutions in present times.

The experience with state institutions has been put forth as one of the important correlates of the current wealth distribution in the world. Specifically, from its original development, the State antiquity index of Bockstette et al. (2002) was shown to be positively associated with 1995 income and with the 1960–1995 GDP growth rate. Bockstette et al.’s aim was to use presence and duration of experience with macro polities as one of several potential indicators of societal complexity and level of technological advancement. The authors were interested in investigating the effect of early social and technological development on post-WW2 economic growth rates, and they assumed that the impact of very early experience would decay over time, so they did not attempt to code information on state presence before 1 CE or after 1950. They coded all countries with substantial populations for which relevant economic growth and other indicators were available, resulting in a sample of 104 countries, of which their analysis focused especially on 70 non-OECD member countries.

Roughly the same data set was also used by Chanda and Putterman (2005), and Chanda and Putterman (2007). Bockstette et al.’s data were subsequently expanded to include more ex-Communist countries (Iliev and Putterman 2007), more African countries (Cinyabuguma and Putterman 2011), and a few other countries for which complementary income or other required data had initially been viewed as unreliable. Based on this extended dataset, Putterman and Weil (2010) demonstrated that the ability of state history to predict current levels of development is greatly strengthened by replacing the state history that transpired on a given country’s territory by the weighted average state history of the places in which current residents’ ancestors lived in the past. This adjustment was motivated by the large movements of populations especially from “Old World” continents to the Americas, Australia and New Zealand after 1500. Chanda et al. (2014) apply the same procedure to demonstrate “persistence of fortune” of ancestral lines in former colonies that display a “reversal of fortune” (Acemoglu et al. 2002) in the absence of such ancestry and migration accounting.³

In short, previous work has largely agreed on a positive association between long-run state history and current development. However, as scholars have acknowledged, the present shares complex links with the past. For instance, pre-1500 economic advantages seem to have become relative disadvantages among colonized countries during the colonial era (Acemoglu et al. 2001, 2002). As of late, this idea of reversal has been revisited in two studies that are particularly relevant to our paper: Hariri (2012) presents compelling evidence that early (precolonial) experience of state institutions in countries outside Europe prevented them from transplanting democratic institutions brought by European colonizers, leaving instead an “autocratic legacy” in these countries. Olsson and Paik (2013) reveal a negative association between the time from Neolithic transition and current income levels in the Western agricultural core—Europe, North Africa and Southwestern Asia.

Furthermore, the long-run persistence literature has begun to reveal nonlinearities in how events in the very distant past affect economic development. For instance, the migration out of Africa is argued to have generated a wide array of genetic diversity levels in human populations around the world. In turn, predicted genetic diversity displays a hump-shaped relationship with indicators of economic development, including per capita income in 2000 (Ashraf and Galor 2013).

Thus, in light of these recent developments, allowing for a more flexible relationship between state history and contemporaneous levels of development is a natural extension to the literature. The closest paper to ours to have done so is Lagerlöf (2016), which presented simulations of a theoretical model consistent with a hump-shaped relationship where old states with a large extractive capacity and autocratic elites have been overtaken by younger and more democratic states with a greater growth potential. The empirical evidence hints at a concave relationship between contemporary income levels and state history in a cross-country sample, based on the 1–1950 CE state history index, but only when adjusted for post-1500 migration.

In the theoretical section, we present a framework outlining how states affect and are affected by economic development over history and in the empirical section we explore systematically this relationship considering the complete state histories.

2.2 Defining the “state”

How do we know when a state has emerged? The first challenge stems from the question of how to define the state, hardly a novel dilemma in social sciences. The classical understanding of the “state ” comes from Max Weber, who defined it as an entity which “upholds the claim to the monopoly of the legitimate use of physical force in the enforcement of its order” (Weber 1968, p. 54). This implies that we should be looking for evidence of the initial monopolization of power within a certain territory.

However, there is also the question of the extent of this original jurisdiction: how large is the population or the territory subject to the power monopoly? Is, for instance, a village with 100 tribesmen, led by a chief, large enough to classify as “state”? It appears that we need to find an appropriate threshold to distinguish between small and large scale political organization. Therefore we adopt the convention that, although simple chiefdoms fall short of being states, a paramount chiefdom which incorporates multiple individually substantial chiefdoms can be understood as a form of incipient state. Hence we decided to begin according partial weight when a polity reaches this level. By this convention, for instance, the land of what is today Belgium came under large-scale political organization for the first time between 59 and 52 BCE, when it was integrated in the Roman Empire.

This agrees with established sociological and anthropological taxonomies of human societies throughout their evolution. For instance, Johnson and Earle (2000) proposed a division of societies into small-scale local group (further divided into family, village and the Big Man group) and larger-scale regional polity, which can be a chiefdom or a state. This distinction goes back even earlier, to Charles Tilly: “the term [state] includes city-states, empires, theocracies, and many other forms of government, but excludes tribes,” (Tilly 1990, p. 1) and to Service’s (1962) proposed typology of bands/tribes/chiefdoms/states.⁴

3 Theoretical framework

In this section, we outline the theoretical framework underlying the empirical analysis. We present the three basic premises of the model and then use four country case studies to illustrate the kind of dynamics that we have in mind which eventually produce a hump-shaped hypothesized relationship between state history and indicators of aggregate productivity. In the third subsection, we explain briefly how our intuitive framework fits into a more standard Malthusian growth model.

3.1 The basic premises

The key assumption of our framework is that up to a point, accumulated state history favors capacity building, taxation and the provision of public goods, which in turn spur economic growth. But beyond a certain level, state experience is conducive to the rise of extractive institutions and powerful elites that appropriate tax revenue rather than turn it into public goods. In the absence of constraining institutions, excessive centralization ultimately leads the state into stagnation. When more experienced states have reached this level, they tend to be overtaken by less experienced states still in their expansion phase (as bottom-up pressures in the latter bring about executive constraints). Of course, the state as such is not the only driver of growth, for example processes of technological improvement and trade expansion promoted in neighboring states may spill over to nearby areas. We nonetheless emphasize state experience as it is our focal empirical measure and it represents a key dimension of the development process over the very long run. We note again that by ”state experience” we refer not only to duration of presence of macro polities, but also degree of unity, territorial coverage, and locally rather than externally based rule.

The proposed relationship between state history and long-run economic development has three distinct premises: (1) In a newly established state, an increase in fiscal and institutional capacity and in central coordination from low levels have historically been conducive to productivity increases and economic growth. (2) Governments in long enduring states often tend to misuse the tax revenue at the expense of economic progress, despite having access to a solid fiscal capacity. (3) Citizens of countries having less powerfully consolidated states can exploit the experience and mistakes of countries with greater degrees of internally-imposed power centralization, to advance their levels of productivity beyond the maximum reached by the latter. We discuss the evidence supporting these premises below.

First, there is widespread agreement that a very short state history, especially if characterized by a lack of autonomy and rule instability generally implies weaker fiscal capacity. This has recently been discussed among others by Tilly (1990), Collier (2009), and Besley and Persson (2013). Increasing evidence shows that a consolidated bureaucracy enables investments in large public projects, technological innovation, and effective warfare, thus spurring economic growth. Recent empirical studies on the historical role of state capacity reveal strong correlations and potential causal links between administrative infrastructure, high taxes and economic prosperity (Besley and Persson 2013; Dincecco and Katz 2014).

Second, older and more autonomous states are more predisposed to maintain overly centralized, often abusive power structures. The first states developed naturally from the basic need to sustain collective action in large communities, particularly in response to attacks by predators (Tilly 1990 and Olson 1993 and more recently, Mayshar et al. 2015). Against “roving bandits,” it was welfare enhancing to have a member of the community become the dictator who collected rents used partly for defense, but mostly for self-gratification (“stationary bandit”). However, without mechanisms to commit to providing a certain level of public good, the autocrat became a “roving bandit,” maximizing rent-extraction. Hence, as Olson concludes, in an autocracy good economic performance is unsustainable in the long run.

Examples of political and economic collapse in ancient states due to overcentralization abound. For instance, the demise of the expansive Mesopotamian empire Ur III (c. 2000 BCE) came when local city-states sought independence, rejecting the overbearing supercentralized bureaucracy tailored to channel resources from the periphery to Ur and finance the defense of conquered regions (Yoffee 1988). Similarly, only a few centuries later Hammurabi’s great empire collapsed after disempowered local authorities rebelled against the drain on resources for the king’s glorification in Babylon, plunging the state into an irreversible economic crisis (c. 1600 BCE).

Complementary evidence comes from Hariri (2012) who shows that older indigenously formed states were more likely to develop autocratic institutions than later states. His proposed channel is that older states more successfully fended off attempts at colonization by European powers between the 15th and 20th centuries, and hence did not experience either direct transplantation of or cultural influence related to conceptions of democratic institutions (developed between the 18th and 20th centuries within the Western colonizing powers).

Non-inclusive institutions and political instability also undermine fiscal capacity (see e.g. Besley and Persson 2013 and Dincecco and Katz 2014).⁵ Early states incurred frequent regime changes due to predatory attacks and internal strife.⁶ Moreover, ancient warfare did not stimulate tax collection to the same extent modern warfare does (Gennaioli and Voth 2015). Lacking sustainable revenue generation and constraints against turning resources away from public goods and technological innovation made earlier states more vulnerable to economic stagnation and collapse.

Third, we argue that countries with less state experience tend to be able to reach a higher level of development than countries with more (and usually longer) state experience before their stagnation phase sets in. This is supported by the pattern of superior economic development in countries with an intermediate length of a civilization, like the United Kingdom and the Scandinavian countries, compared to countries like Egypt or Iraq, as discussed in recent work by Olsson and Paik (2013).⁷ Again, the main arguments here are institutional: Olsson and Paik argue that old civilizations developed autocratic, hierarchical societies that were not conducive to the emergence of democracy and innovation, which became critical factors for economic growth during the modern era, leading to a “Western reversal of fortune” since the onset of agriculture.⁸ Similarly, Lagerlöf (2016) argues the autocratic elite in old states made repeated investments in extractive capacity, while the lack of a similar extractive capacity in younger states facilitates the transition to a faster-growing democracy. Democratic political structures with constraints on executive power may have developed in initially peripheral areas like England, the Netherlands and Scandinavia, because local and commercial elites effectively countered tendencies towards power centralization during the maturation process of these younger nation-states. Executive constraints, in turn, set the premise for unhindered technological innovation and may have mitigated the economic downturns accompanying domestic struggles for political leadership (Cox and Weingast 2015).

The regions once peripheral to the more ancient civilizations, slower to develop state institutions, were also less exposed to raids by roaming armies and to incursions by migrating peoples (such as the Persian, Hellenic, Hun, Islamic, Mongol and Turkic incursions in the old civilizational core regions). All these factors contributed to their ability to eventually overtake the levels of development in the older states.⁹

3.2 The predicted pattern

The three central factors discussed above combine to form the basis of a stylized pattern of state history and economic development. Figure 1a showcases this pattern, using four current countries, all located in the Western agricultural spread zone,¹⁰ as case studies. We assume, for simplicity, a logistic, positive relationship between state history and aggregate productivity in the spirit of existing long-run models.¹¹ We do not, however, argue that every country necessarily follows this proximate pattern. Needless to say, the short account below does not have any ambition of providing a full description of the histories of Egypt, Italy, United Kingdom and Estonia. Please see the extensive Data Appendix for details about state development and our coding of historical events. Although we only use four examples, we do argue that the model below is applicable in broad outline also to the rest of the world.¹²

To the far left in the figure, we use Egypt as an example of a very old civilization where the first signs of a state-like organization above tribal level emerged around 3200 BCE. From a low pre-state level of development, total productivity in the economy started to increase when the government of a full state was able to extract taxes from the population that, in turn, were transformed into public goods that provided security and a more efficient mode of agricultural production. A few centuries after the origins of the state, development took off in the classic period of the Old Kingdom with large cities, massive public monuments, codified language and a highly stratified, complex society.

However, as outlined in detail in our Data Appendix, the centralized government of Egypt began to run into problems by 2130 BCE during the 8th dynasty, when a decades-long drought and famine set off the fragmentation of central power. Repeated foreign invasions from the Hyksos in 1650 BCE and onwards implied that Egypt gradually lost its economic and military dominance over the region, until a succession of foreign invaders such as the Assyrians, Persians, Macedonians, and Romans caused a marked slowdown in productivity growth. During the first millennium BCE, Egypt arguably started its long period of stagnation, characterized by slower productivity growth than the civilizations on the northern rim of the Mediterranean, which has by and large persisted to the present. By 1 CE, the Romans had conquered all lands around the Mediterranean and had surpassed the Egyptians technologically, militarily, and also in terms of state organization.

State origins in Italy can be traced back to the Etruscans in 850 BCE, more than two thousand years after similar developments in Egypt. As is well known, the Roman republic with its innovative political institutions and military organization, would soon become the greatest empire in Western Eurasia and forcefully dominate the lands around the Mediterranean. Economic stagnation set in with the dissolution of the Western Roman Empire circa 476 CE. While historians still argue about the main causes, it appears that political fragmentation and power struggles between the Church, senatorial aristocracy and the military led to territorial concessions to the barbarian invaders, a weakening of fiscal capacity and ultimately economic bankruptcy.¹³ However, as seen in our stylized figure, Italy long continued to be the most developed country among our four countries even after the fall of Rome. The barbarian tribes that invaded Rome soon tried to mimic Roman traditions and Latin lived on as a lingua franca for centuries due to the rising power of the Pope in the Vatican. Merchant sea powers such as Venice and Genoa would become very powerful in the early Medieval period and the Renaissance period was still a very innovative period in the history of the Italian peninsula, with progress in many fields.

However, with the emergence of Protestantism in northern Europe, the discovery of new trading routes on the Atlantic, and the increasing power of emerging centralized states in France and Britain, Italy started its period of economic and technological stagnation.¹⁴ Compared to Italy, the emergence of a centralized state in Britain came quite late, beginning with external governance by Rome around 50 BCE and internal polities from AD 401. Spurred by revenues from Atlantic trade and from institutional innovations that strengthened private property rights and accountable political institutions, Britain rose to eventually become the leading power in the world after the Napoleonic wars ended in 1815. The cluster of innovations that formed the backbone of the Industrial Revolution contributed to a dramatic rise in output per capita and to the breakdown of the Malthusian link between income levels and fertility. After World War I, the United Kingdom arguably entered a phase of relative stagnation in which it clearly lagged behind, for instance, the United States.

Fig. 1

Sources: Data presented below

Predicted pattern of accumulated state history and productivity. a Stylized model and illustrations of long-run state history and economic development. b Relationship between log GDP per capita in 2000 and state history in four illustrative countries. c Relationship between Technology Adoption and Statehist in 1 CE, 1500 CE and 2000 CE in full cross section of countries. Note: a Shows a stylized model of the long-run relationship between state history and aggregate productivity, using Estonia, England, Italy and Egypt as examples. In the left-hand side of the figure, we explain the typical development over time in the logistic development curves with the four basic states of origins, take-off, slowdown, and eventually stagnation. In the right-hand side of the figure, we map the development of the four states over time into a specific combination of state history scores $S_i$ and its corresponding productivity from the left-hand side. b Shows the actual combinations of log GDP per capita in 2000 and the main state history score, using a depreciation rate of 1%, for these four illustrative countries. c Shows the predicted Technology Adoption index from Comin et al. (2010) fitted as a polynomial of extended Statehist, for 1 CE, 1500 CE and 2000 CE for our full cross-section of countries. The shaded areas are the confidence intervals of the mean.

The territory of current Estonia was populated by tribes with no organized government until 1237, when the area was taken by Denmark and by German knights. Sweden and Russia then controlled the region until 1919 when Estonia experienced a brief period of independence, before it was once again swallowed by the Soviet Union after World War II. The country did not become fully independent until 1991. The country therefore has a relatively low state history score. Nonetheless, the country has pushed for innovation, such as e-government, and economic growth has been fast since independence with income per capita standing at about 70% of that of United Kingdom in 2014.

Focusing on the states’ age part of state experience, the highly stylized logistic development curves in Fig. 1a only differ along two dimensions: the timing of state emergence and the maximum level of attainable development, which increases with every new state. Our argument above proposes that highly centralized governments, typically in the earliest states, in the long run supported less economic dynamism than states where powers were devolved to different institutions. In addition, relative newcomers caught up with and eventually overtook older states. The Romans could learn from experience of Egypt that a state led by a divine regent was inferior to a republican and subsequently imperial order that made better use of human resources, politically and militarily. After 1500 CE, England could overtake Italy in part due to its access to trade and colonization opportunities associated with its Atlantic-facing location, but also due to far superior political institutions that strengthened the accountability of government, encouraged commerce, saw the emergence of superior naval and military capabilities, and led to a leading position in the emerging system of international credit.¹⁵ Our long-run representation implies that Estonia may eventually overtake United Kingdom in productivity when it converges to its long-run maximum level where stagnation ensues. However, like many other countries, Estonia has not yet reached this stage.

To sum up, in Fig. 1a we see that the long-term trend is that the relationship between state experience and productivity is predicted to turn negative among more mature states. This negative relationship arises since newer states reach a higher maximum level of productivity than their older neighboring states. However, due to the fact that many countries are still in a transition period to their long-run steady state level of productivity, there will be a positive-sloping part of the overall contemporary curve as well (the segment between Estonia and England in the figure). Taken together, these opposing trends imply a hump-shaped relationship between state history and contemporary measures of productivity.

In Fig. 1b, we illustrate the consistency with this hypothesis of the pattern of data on accumulated state history and level of GDP per capita in 2000 CE for the four countries discussed above. In line with the prediction, we find that the fitted non-linear curve is indeed hump-shaped with an upward-sloping part between Estonia and United Kingdom and a negative-sloping part to the right. In the empirical section below, we investigate more formally this non-linear relationship in a large cross-section of countries.

It should further be noted that our model actually implies that the relationship between state history and aggregate productivity should change over time. In the year 500 BCE, for example, Fig. 1a suggests that Egypt was ahead of Italy, which, in turn, was ahead of England and Estonia in terms of productivity, so that the relationship between state history and productivity at the time should be strictly positive.

As an initial test of our model, we show in Fig. 1c the evolution over time of the relationship between state history and our preferred proxy for productivity (a composite index of technology adoption from Comin et al. 2010) for a cross-section of all countries in the world with available data. When the relationship is estimated early in history, in year 1 CE, we find indeed that the fitted non-linear curve has a positive slope, implying that older states (like Egypt) were more technologically advanced at the time. The curve then tilts to become distinctly concave with a short negative-sloping section in 1500 CE, with a clear hump-shape emerging in 2000 CE. In Figure C1 in the Appendix, we show that a similar pattern is evident in 1 CE and 1500 CE when we use population density as our proxy for productivity. Our interpretation of this evidence is that it is consistent with the model outlined in Fig. 1a where the disadvantages of being an early state eventually manifest and lead to a gradually stronger negative relationship for older states.¹⁶

Although Fig. 1a considers the age of states only, other dimensions of state history, such as experience with home-based rule, can also help to account for some countries’ laggardness or advance. In some cases, in North Africa and the Middle East, colonization by middle-aged states (e.g British, French) may have generated negative effects on the level of stagnation of earlier developed states. Extractive institutions and other forms of economic domination fuelled by globalization may not only have helped the later comers to overtake earlier states, but may even have influenced the productivity stagnation or decline of the latter.¹⁷

At the other end of the statehist spectrum, young colonial states, especially those in Sub-Saharan Africa with no pre-colonial state institutions, have not only had short histories, but also have spent more than half of their macro-political histories under external rule. With modern state structures thrust upon them from without, they could not build institutional capacity organically. This left them vulnerable in the face of post-colonial challenges, against the backdrop of ethnic divisions, over-reliance on natural resources, and a legacy of extractive institutions (Collier 2009). The post-colonial experience highlights that extractive institutions are not exclusively a feature of longer state histories. As a possible consequence, post-independence productivity levels in many countries with young states have declined and income per capita is lower today than in 1960. Thus, while Fig. 1c above suggests that the hump-shaped relationship between state history and aggregate productivity is only transient, other forces at play shed uncertainty over whether and when a globally negative relationship between the two variables will eventually emerge.

Note that if state history before year 1 CE is not counted, as was the case in the statehist index used by Bockstette et al. (2002) and subsequent papers until the present, the state history score for Egypt would be closer to that of Italy. This party explains why analyses based on the data heretofore available tended to find a positive relationship between state history and level of development around the end of the 20th Century (note again the exception of Lagerlöf 2016, discussed above).

3.3 Towards a growth model

The stylized pattern above can be translated into a growth theory framework generating specific hypotheses to be tested in the empirical section.

Let $S_{it}$ be the accumulated state history of a country i at time t, where a greater $S_{it}$ implies that the country has had a more substantial experience of unified, home based, large-scale states. For simplicity, we assume that $S_{it}$ is a positive, linear function of time from the date of state origins onwards. The first state is $i=1$. Thus, the index $i\in \left\{ 1,2,3,\ldots \right\}$ reflects the historical order of state formation.¹⁸ We define total factor productivity $A_{it}$ as a function $A_{i}(S_{it})$ with $\partial A_{i}(S_{it})/\partial S_{it}>0$, where the positive relationship with state history takes the form of a non-linear, logistic curve as in Fig. 1a. At low levels of $S_{it}$, aggregate productivity increases sharply with $S_{it}$ but as state history accumulates, stagnation sets in and $\partial A_{i}(S_{it},i)/\partial S_{it}$ approaches zero.

We also assume that for any given level of state history $S>0$, $A_{i}(S)<A_{i+1}(S)$, i.e. of two countries with identical accumulated levels of state history, the state which emerged at later t will be able to attain a higher level of productivity at every level of S and eventually reach a higher maximum productivity.¹⁹ This advantage stems mainly from the fact that new states could learn from the experiences of older states, or that countries in which the state developed later have been more successful at avoiding the overcentralization of power.

In the preindustrial era, in terms of a standard Malthusian growth model, total output in a territory is given by $Y_{it}= A_{it}X_{i}^{\alpha }L_{it}^{1-\alpha }$ where $Y_{it}$ is the total output, $X_{i}$ the fixed amount of land in the country, and $L_{it}$ the level of population (Ashraf and Galor 2011). In a long-run, Malthusian equilibrium, output per capita $Y_{it}/L_{it}=y_{it}$ is constant since any temporary increases in $Y_{it}$ are offset by higher birth rates, hence population, so that $y_{it}$ eventually returns to its steady-state level $\bar{y_i}.$

Combining the elements above, we can express the Malthusian equilibrium as

$$\begin{aligned} \bar{y_{i}}=A_{i}\left( S_{it}\right) \cdot \left( \frac{X_{i}}{L_{it}} \right) ^{\alpha }=A_{i}\left( S_{it}\right) \cdot p_{it}^{-\alpha } \end{aligned}$$

(1)

where $p_{it}=L_{it}/X_{i}$ is the population density of country i at time t. Over the long run, as productivity $A_{it}$ grows in response to the increase in state experience $S_{it}$, population density $p_{it}$ will also increase and country i’s Malthusian equilibrium level of income per capita $\bar{y_{i}}$ will remain intact. When two countries at a given point in time have reached the stagnation level, where $A_{it}$ is unresponsive to further increases in $S_{it}$, the level of productivity $A_{it}$ (as well as the level of population density $p_{it}$) will be relatively lower in the country where states emerged early (with a higher $S_{it}$ and a lower i) compared to the country where states emerged later.

There are thus two countervailing effects of extensive state experience on $A_{it}$: On the one hand, greater state experience increases the level of productivity within a country directly, up to the steady-state (through fiscal capacity and centralized coordination). But since a higher level of state history also typically implies an early state, its level of productivity is limited compared to countries with later states. This pattern eventually gives rise to the hump-shaped cross-sectional relationship between $A_{it}$ and $S_{it}$ some time during the Malthusian era. Note that our model also implies a hump-shaped relationship between $S_{it}$ and $p_{it}$. We expect these patterns to emerge by the end of the Malthusian era and therefore, in the empirical section, we test these hypotheses using an index of technology as a proxy for $A_{it}$ and levels of population density and urbanization rates as our measures of $p_{it}$ in 1500 CE.

In the industrial era, the Malthusian link between productivity levels $A_{it}$ and levels of population density $p_{it}$ typically disappears after a period of adjustment. Income levels per capita $Y_{it}/L_{it}=y_{it}$ then tend to be strongly positively correlated with higher levels of $A_{it}$ (Hall and Jones 1999). For the contemporary era, we will therefore proxy aggregate productivity by income per capita, which is of course also the standard measure of economic prosperity, and study whether the the historical impact of $S_{it}$ on $A_{it}$ prevails even after the transition to modern economies.

4 Data

In this section, we outline how the existing index of state history has been extended to cover the BCE period. We will also briefly present some of the key tendencies in the new data series.

4.1 Constructing the index

The construction of the index for the BCE period follows the principles developed by Bockstette et al. (2002), applied here to 159 modern-day countries.²⁰ We use evidence of written records where available. Where not extant, we rely on archaeological data, following a “diagnostic traits” approach: we consider material manifestations of the monopolization of power, as an “archaeological confirmation of the process of state formation” (Jones and Kautz 1981, pp. 16–17). These can be monumental structures, such as palaces, temples or large urban settlements etc. In the case of Iraq, for instance, there is the transition from small to large urban centers with grand architectural structures such as Uruk in the middle of the 4th millennium BCE.²¹

The second task is to mark the transition from chiefdom to fully-fledged state. Following the paradigm of the evolution of pristine states from chiefdoms (see e.g. Carneiro 1981; Earle 1987; Flannery 1995; Marcus 1992; Spencer 1990; Spencer and Redmond 2004), we mark this distinction in our data by assigning the following values: Band/tribe is marked by a rule score of 0, paramount chiefdom is assigned 0.75 and fully-fledged state receives the value 1. Robert Carneiro emphasizes that the paramount chiefdom is the evolutionary link between autonomous bands or tribes and the state.²²

While it is difficult to know exactly where the chiefdom ends and where the state begins in pre-history, we have made efforts to draw a sensible line where the evidence suggests a noteworthy evolution in socio-political organization.²³ While this approach is not uncontroversial, it is the most feasible given limited documentary resources. We further detail the assignment of scores in a later section.

For each country, the time of emergence of the first state institutions on its territory is identified, as defined above. State age is defined as the time elapsed from this date until 2000 CE. The oldest state was established on the land of today’s Iraq around 3500 BCE. Hence, the time for all countries is divided into 110 periods of 50-years.

For each country i and half century t, scores are assigned to reflect three dimensions of state presence, based on the following questions²⁴:

• 1. Is there a government above the tribal level? Score component $z_{it}^{1}$ receives 1 point if yes, 0.75 if the government can at best be described as a paramount chiefdom and 0 points if no government is present.

• 2. Is this government foreign or locally based? $z_{it}^{2}$ is 1 if the rule is locally based, 0.5 if externally based, and 0.75 for local government with substantial foreign oversight.²⁵

• 3. How much of the territory of the modern country was ruled by this government? $z_{it}^{3}$ reflects the proportions of the territory under some rule: 1 (over 50%), 0.75 (25–50%), 0.5 (10–25%), 0.3 (under 10%).²⁶

Time is indexed by t and refers to a 50-year period ranging from $t=0$ for 3500–3451 BCE when the first state arose, to $t=109$ for 1951–2000 CE). For every such time interval, we compute a composite State index score by multiplying the three components by one another and by 50²⁷:

$$\begin{aligned} s_{it}=z_{it}^{1}\cdot z_{it}^{2}\cdot z_{it}^{3}\cdot 50 \end{aligned}$$

(2)

Finally, joining the BCE-with the preexisting CE-era series, we aggregate all “flow” scores $s_{it}$ into Statehist—the comprehensive index of the cumulative state history.²⁸ The index is normalized by the score of a hypothetical state with full discounted scores between 3500 BCE and the period of interest $\tau$:

$$\begin{aligned} S_{i\tau }=\frac{\sum _{t=0}^{\tau }(1+\rho )^{t-\tau }\cdot s_{it}}{ \sum _{t=0}^{\tau }(1+\rho )^{t-\tau }\cdot 50} \end{aligned}$$

(3)

This cumulative Statehist index $S_{i\tau }$ ranges from 0 to 1 and can be calculated at virtually any point in history $\tau =\left\{ 0,1,\ldots 109\right\}$. We calculate it mainly for 1500 CE (at $\tau =99$) and for 2000 CE, usually discounting the more distant past relative to the present by setting $\rho>$ 0, although usually at a modest value. For instance, the Statehist value of a state with full scores discounted at 1% between 3500 BCE and 2000 CE is 3359.79. The previous literature has employed $\rho =0.05$, in light of the reasonable assumption that the more distant past matters less today than recent history. With the additional data, however, this rate gives almost no weight to the long stream of $s_{it}$-scores before 1 CE.²⁹ While it of course remains to be seen below just how useful placing weight on the distant past will be, our convention is to employ the 1% discount factor when calculating the normalized Statehist score.³⁰

To answer the three questions (a-c) above in a manner that is consistent across periods, we relied mainly on information in the Encyclopedia Britannica Online, but also additional sources where information in Britannica was insufficient. We provide additional detail on our data sources and illustrate the coding process and further data aggregations in online Appendix B. Accompanying this paper is also an online Data Coding Appendix, which provides a comprehensive list of coding decisions for all country-period observations.

4.2 A brief look at the data

In this section we present some patterns that arise from the complete state history time series and the data used in forthcoming analyses.

Firstly, we note that the evolution of state institutions in the world follows approximately an exponential upward trend with periods of rapid growth punctuated by periods of stagnation (Fig. 2). The graph shows the log of the aggregated percentage score for all contemporary countries in our sample on the vertical axis plotted against number of years BCE or CE on the horizontal axis. The percentage score in period t is calculated as State index world $(t)=100\cdot {\sum }_{i=1}^{N}s_{it}/\left( N\cdot 50\right)$ where $N=159$ is the number of included countries and where $s_{it}\in \left[ 0,50\right]$ is the state history score for country i during 50-year interval t, as described above.³¹ A value close to 0% in this world index indicates that there is almost no sign of state presence in any of the included countries in period t whereas a score of 100 means that all 159 countries reach the maximum value $s_{it}=50$ in our state measure during that period.³²

Several periods are characterized by rapid state evolution whereas other periods are marked by a general decline. The first boom in state emergence appears already in 3500–2300 BCE, which then ends with a long period of stagnation. The other major stagnations in the figure happened around 1750 BCE, 1200 BCE, and 400 CE. A second period of rapid growth was 850 BCE–1 CE during the Iron Age. From just after the collapse of the Roman empire around 450 CE, aggregate state emergence shows a steady upward trend.

This pattern is also visible in Fig. 3, which in addition shows the regional aggregated percentage score for all contemporary countries in our sample (this figure displays world and regional averages of the State index, rather than the natural log of that average as in Fig. 2, allowing the reader to get a sense of the trend in a form some may find more intuitive). We disaggregate the evolution of state history into the four main agricultural core areas: Western, Eastern Asia, Sub-Saharan Africa, and the Americas.³³ These four areas are created on the basis of how Neolithic agriculture and civilization spread during early historical times.

Fig. 2

Emergence of states in the world 3500 BCE–2000 CE. Note: The graph shows the logged value of the aggregate State index for 159 countries identified during 110 50-year intervals between 3500 BCE and 2000 CE. The value 100 would signify all 159 countries in our sample are full states, as defined in our text, including high geographical extent and unity and being entirely locally based. On the horizontal axis, negative values imply years BCE whereas positive values show the CE period. A linear fitted regression line has been included. The State index is calculated as described in the text

Fig. 3

Emergence of states in four agricultural core areas and in the world as a whole 3500 BCE–2000 CE. Note: The figure shows the development of the aggregated State index in the Western agricultural zone (including 62 current countries in Europe, North Africa, the Middle East, as well as Afghanistan, Armenia, Azerbaijan, Georgia, India, Iran, Kazakhstan, Kyrgyzstan, Pakistan, Russia, Tajikistan, Turkmenistan, Uzbekistan), Eastern Asia (20 countries), Americas (including 27 countries in North and South America and in the Caribbean), and Sub-Saharan Africa (47 countries). Oceania (only 3 countries in our sample) is omitted. It also shows the aggregate index for the 159 countries in the world as a whole (solid black line). On the horizontal axis, negative values imply years BCE whereas positive values show the CE period. Particular years with trend breaks are marked

When we divide up the world in this way, some striking historical differences between the regions appear: State evolution started earliest in the Western area, with Eastern Asia lagging behind until rough convergence (indeed, initially overtaking) around 500 CE, with the other regions gaining steam later and all converging only toward the end of the era of European colonialism. State emergence was earliest in Eastern Asia and in the Western region. Interestingly, both of these early civilizations took off on a more rapid path after 850 BCE. By the time of the Western Roman collapse after 450 CE, Asian state development overtakes the Western one for the first time.³⁴

The other two regions, the Americas and Sub-Saharan Africa, clearly lag behind, particularly after the Eurasian turning point 850 BCE. From about 500 CE, the pace of state emergence starts to rise in Sub-Saharan Africa. When the colonial era starts in the late 15th century CE, the lagging regions experience a dramatic increase in the State index. This increase is of course to a great extent driven by the emergence of colonial states, created by European powers (although $z^2$ is capped at 0.5 so long as colonial status continues). By the final period of observation (1951–2000), the Americas has the highest score on state presence among all regions in the world.³⁵

The Statehist index and other variables related to state experience, as well as outcomes and control variables used in all forthcoming analyses are summarized in Table 1 below. Full definitions of each variable are given in the online Appendix.

Table 1 Summary statistics

5 State history and economic development

We now proceed to analyze the relationship between state history and different indicators for historical and contemporary levels of productivity. Our proxies for aggregate productivity will be an index of technology adoption, population density, rates of urbanization, and GDP per capita.

5.1 State history and pre-industrial economic development

State history and productivity in 1500 CE

We begin with the empirical question of the relationship between state history and productivity in the Malthusian era. The key predicted pattern in Fig. 1a was a hump-shaped relationship between our state history measure and aggregate productivity in a cross-section of countries, reflecting on the one hand that newer or more inexperienced states tend to be in the process of converging to their own maximum productivity potential, whereas more experienced states already have attained a relatively low maximum level. Was such a tendency in place already in 1500 CE?

As a starting point, we proxy historical productivity with the average index of technology adoption constructed by Comin et al. (2010). Using various data sources on the presence and complexity of various technologies, the country-level index captures advances in five sectors: agriculture, transportation, communications, writing, and military. The index is computed for 1500 CE and 2000 CE, using slightly different approaches, which we describe in some detail in online Appendix A.

In order to test this prediction, we set up the following model:

$$\begin{aligned} Technology1500 _{i}=\beta _{0}+\beta _{1}\cdot Statehist1500_{i}+\beta _{2}\cdot Statehist1500_{i}^{2}+\epsilon _{i} \end{aligned}$$

(4)

On the left hand-side of Eq. (4) we have the average technology adoption index in 1500 CE. On the right-hand side we include our main independent variable, Statehist (the cumulative index shown in Eq. (3) accumulated in 1500 CE), both linear and squared, to allow for a quadratic relationship. The Statehist index is normalized with respect to 3500 BCE–1500 CE and computed using a 1% discount rate per period. This equation captures the potentially hump-shaped relationship to Statehist across countries. We also consider variants of (4) that include additional controls, represented by:

$$\begin{aligned} Technology1500 _{i}=\beta _{0}+\beta _{1}\cdot Statehist1500_{i}+\beta _{2}\cdot Statehist1500_{i}^{2}+\beta _{j}^{\prime }\cdot Z_{i}+\beta _{k}^{\prime }\cdot X_{i}+\lambda _{c}+\epsilon _{i} \end{aligned}$$

(5)

$Z_{i}$ is a vector of historical controls including: $Agyears_{i}$, the time before present since the transition to agriculture in the country in question, a variable taken from Putterman and Trainor (2006); $Origtime_{i}$—the approximate time since the first settlement on the territory of the modern-day country by anatomically modern humans, a variable introduced by Ahlerup and Olsson (2012) as a determinant of the variation in levels of ethnic diversity across the world. In a more flexible specification, we include the square of $Origtime_{i}$ in order to account for recent developments in the literature postulating that the patterns of human settlement in prehistory may have nonlinear effects on later economic development (Ashraf and Galor 2013). In the same specification we also include State age $_{i}$ in 1500 CE (the time elapsed in 1500 CE from the date of state emergence). $X_{i}$ is a vector containing geographic controls. These include: the absolute latitude of the centroid of the modern-day country i, whether the country is landlocked, its distance to coast or ocean-navigable river, average elevation, land suitability for agriculture, climatic variables for temperature and precipitation, and the risk of malaria.³⁶ $\lambda _{c}$ is a vector of continent fixed effects. The results are displayed in Table 2. Columns (1) and (2) present linear and quadratic versions of Eq. (4), while columns (3)–(7) add further controls, as in Eq. (5).

Table 2 State history and average technology adoption in 1500 CE

In column (1) we display the simple association between technology adoption and Statehist, which is positive and significant. In column (2), where we add Statehist squared, our main coefficients of interest, $\beta _{1}$ and $\beta _{2},$ display a concave pattern: $\beta _{1}$ is positive, while $\beta _{2}$ is negative, both significant at 1%. In the online Appendix Table D1, we estimate this specification using Statehist discounted by alternative factors, 0.1% and 2%, which reveal the qualitatively identical result of significant concavity. As concavity does not imply non-monotonicity, to test for the latter, we run piece-wise estimations using the linear Statehist separately in countries below and above the technology-maximizing level of Statehist implied from Eq. (4) (which is 0.44, with only 9 countries recording a higher Statehist of 1500 CE).³⁷ The results reported for different Statehist discount rates in the online Appendix Table D2, including our main case 1% rate, show positive insignificant slopes above the maximizing Statehist. Therefore, we do not find an inverse-u shape in the case of technology adoption in 1500 CE.

In column (4) we add to the model the first historical control - Agyears (shown to be positively significantly correlated with the dependent variable in column 3, for comparison purposes). Its inclusion only slightly changes the signs and the magnitude of the coefficients of the Statehist terms. Moreover, the effect of the time from transition to agriculture is reduced relative to column (3). When we also add Origtime and geographical controls in column (5), the magnitude of the estimates changes slightly, but the relationship remains concave. The continent fixed effects in the last two columns reduce the squared term’s coefficient, which becomes insignificant.³⁸

State history, population and urbanization in 1500 CE

We now inquire whether this concave pattern is reflected in other commonly used indicators of historical productivity: population density (Table 3, panel A) and urbanization rate in 1500 CE (Table 3, panel B). All specifications are analogous to those in Table 2.

The extended Statehist is positively and significantly correlated with past population density and urbanization (column 1). Column 2 shows, however, that a quadratic relationship fits the data even better, with both linear and square terms obtaining highly significant estimated coefficients (the same holds when using discounts of 0.1% of 2%, as Appendix Table D3 shows). The population density maximizing level of Statehist is that of Greece (around 0.42). For population density, the below/above maximum regressions including only linear Statehist (Appendix Table D4) display negative slopes above the maximizing value of Statehist , significant when we use a 2% discount rate for Statehist. Hence, some evidence indicates an inverse-u shape already forming in 1500 CE (albeit not when we look at urbanization, which is maximized by a Statehist value of 0.64, which is above the range represented in the data). Both for population density and urbanization, quadratic yet monotonic patterns emerge in the specifications which introduce controls and continent fixed effects. In conclusion, while there are clear signs of diminishing benefits of additional state experience as of 1500 CE, there are few indications that added state experience was a net liability for more experienced states as of that year.³⁹

5.2 State history and current economic development

In our theoretical framework, the downward sloping portion of the cross-country Fig. 1a was assumed to result from less experienced states overtaking more experienced ones. The other portion may be explained by the many states in several parts of the world which emerged only in recent centuries. Thus, we expect contemporary levels of development to correlate in a non-monotonic fashion with accumulated Statehist. To investigate this, we estimate the model with technology adoption and GDP per capita in 2000 CE as a quadratic function of state history. The results are displayed in Tables 4, 5 and 6 below.

However, when analyzing the current levels of technological sophistication or output, using the raw Statehist data means that we only account for the history within the territories of present-day countries. This ignores the state history of other territories from which people migrated in recent centuries to settle in new territories. Population flows after 1500, when the era of colonization began, are instrumental in mapping the impact of historical events to today’s economic performance. This is because the ancestors of today’s population have evidently brought with them the history, the know-how and the experience with state institutions from their places of origin (Putterman and Weil 2010; Comin et al. 2010; Ashraf and Galor 2013).

Table 3 State history, log population density and urbanization in 1500 CE

Table 4 State history and average technology adoption 2000 CE

Table 5 Statehist versus Statehist 1–1950 CE and Log GDP pc 2000. Nonlinear relationship

Table 6 Adjusted Statehist and Log GDP pc 2000. Nonlinear relationship

We therefore also use an alternative measure of state history which is obtained by adjusting the 1500 CE Statehist index with the migration matrix developed by Putterman and Weil (2010). We then re-estimate our model using this new measure—the ancestry-adjusted Statehist—which, for each country, represents the average pre-industrial Statehist of its year 2000 population’s ancestors, with the weights for each source country being the share of then-living ancestors estimated to have lived on its present-day territory. These alternative results are displayed in Tables 4, panel B and 6.

Technology adoption in 2000 CE displays a similar concave relationship with year 2000 Statehist as did the technology index of year 1500 CE (Table 4, panel A). Furthermore, using the ancestry-adjusted Statehist in 1500 CE to explain the differences in average technology adoption in 2000 yields robustly significant estimates across all specifications, with larger magnitudes and higher R-squared statistics than when using the Statehist in 2000 CE. This result is consistent with our theoretical expectation that the relationship between technology and state experience was concave in the late Malthusian era, and that it was transmitted all the way into modern-day levels of technology adoption. The unconditional relationship shown in column (2) of Table 4, panel B, withstands using different discount rates for Statehist, and the relationship exhibits the downward sloping portion of an inverted-u shape in regressions with a linear Statehist above and below its implied maximizing value (panel B in Tables D5–D8).⁴⁰

Given the central role of aggregate productivity in the standard production functions for total output, we argue the hump-shape should also emerge when we look at GDP. Figure 4 illustrates the essence of our findings. On the Y-axis we have the logarithm of GDP per capita in 2000 and on the X-axis we have the extended Statehist (normalized with respect to 3500 BCE–2000 CE and computed using a 1% discount rate per period).

Fig. 4

Non-linear relationship between Log GDP per capita in 2000 and Statehist index. Note: The figure shows a fitted quadratic regression line corresponding to the estimates in Table 5, panel A, column 2, with 154 country observations distinguished by 3-letter country isocodes. On the Y-axis we have the logarithm of GDP per capita in 2000 and on the X-axis we have the extended Statehist (see Eq. (3) above). An increase in Statehist by 0.1 is interpreted approximately as an additional 300 years of effective fully autonomous statehood

The figure displays a scatter plot of all countries in the sample, while also allowing for a quadratic fit of the relationship between output and Statehist. The hump-shaped relationship emerges when using the extended Statehist. ⁴¹ In the online Appendix Figure C3, we show that the state history index based on 1–1950 CE data does not display the downward sloping portion seen in Fig. 4.

We estimate the quadratic relationship to the logarithm of GDP per capita in 2000 CE in Table 5, panel A. In panel B, the Statehist 1–1950 CE data are used for purposes of comparison. All specifications are analogous to the ones in the previous tables.

As before, the simple correlation between per capita income and Statehist (column 1) is positive and similar in magnitude across the two panels, but slightly less precisely estimated when the independent variable is the extended Statehist. In column (2) we add the squared Statehist, and the results mirror the pattern in Fig. 4: In panel A, both coefficients are significant at 1%, and their signs confirm the concave relationship between log per capita GDP and state history. By contrast, in panel B, the counterpart of this specification using Statehist 1–1950 CE displays coefficients with the same signs but much smaller and insignificant (the coefficient of the quadratic term becomes positive when controls are included).⁴²

While Agyears is significantly positively correlated with modern-day GDP (column 3), when we control for it alongside the linear and quadratic Statehist, its inclusion hardly changes the signs and the magnitudes of the coefficients of the Statehist terms. Moreover, the effect of the time from transition to agriculture is insignificant when the Statehist terms are added (column 4), indicating that although early states may only have arisen where agriculture had long been practiced, a country’s subsequent experience with states eclipses its experience of agriculture as a predictor of current productivity. As with technology, the concavity is robust to using alternative discount rates (0.1 and 2%) to calculate Statehist (Appendix Tables D5–D6). Moreover, separate linear estimates below and above the maximizing Statehist display a large, significant downward slope for countries in the upper range, consistent with Fig. 4 (Table D6, in the online Appendix).⁴³ $^{,}$ ⁴⁴ This confirms that a very limited or very extensive experience with state institutions can become a relative disadvantage across nations.

The concavity results are robust to the inclusion of Origtime, as well as geographical controls and continent fixed effects.⁴⁵ However, the coefficients on Statehist squared are smaller than the linear terms’ coefficients as more controls are introduced in columns (5)-(7). This implies that the optimum level of Statehist falls very close to, or outside the top of its range. Does this mean that the hump-shaped relationship between state history and income suggested by the earlier estimates (e.g., columns 2 and 4) is mistaken? We think not. Our framework posits such a relationship between state history and productivity as emerging across countries, whereas each individual country’s trajectory is described by a logistic curve, with no downward sloping portion (see the left side of Fig. 1a). As the number of factors controlled by our regressions grows, we may be approaching a situation in which the estimated coefficients on the focal Statehist variables will reflect only differences between otherwise nearly identical countries within a very narrow sub-region (countries not only in the same continent but sharing almost identical geographic coordinates, climate, etc.), as well as a very similar date of state origins, as indicated by the inclusion in column (7) of the variable state age. The resulting estimates of concavity without evidence of a downward sloped portion is consistent with the logistic curve pattern expected for any single country (left side of Fig. 1a) rather than the hump shape predicted for the full cross-section of countries over which state histories show wider variation.

Is our finding of concavity of per capita GDP with respect to state history in fact attributable to having included coding of state presence in the BCE era in our analysis, unlike previous studies? To see that this is the case, compare panel B of Table 5, which shows estimates of similar specifications but using the old state history variable covering years 1–1950 CE only. The main estimates are neither significant nor similar in terms of signs with the estimates in panel A.

Lastly, from column (2) in Table 5, based on the estimates of our coefficients of interest, we can infer that the predicted income-maximizing level of Statehist is reached at 0.355, which is very close to that of the United Kingdom and most countries in Western Europe.

The effects’ magnitudes are not straightforward to assess from the tables. However, some numerical examples may show more clearly how the impact upon per capita GDP of an increase in Statehist depends on the level of state experience at which the change occurs. Take for instance the case of Indonesia, which has 1350 years of state existence and a Statehist score of 0.254. If we could hypothetically increase the Statehist score by 0.1 (which is 58% of Statehist’s standard deviation, but enough to add 335 full-state years making it reach the level of the UK score), the implied approximate effect on per capita GDP in 2000 would be roughly a 20% increase, from USD 773 to USD 944 in 2000.⁴⁶ The opposite would happen if we were to increase the value of the Statehist score by 0.1 for China, which starts off with a value of 0.582 (a value exceeded by only five countries in the sample): the approximate effect would be a drop in per capita GDP in 2000 by 44.4%.

Taken together, our estimation results so far are consistent with our predicted pattern. Moreover, this becomes evident only when we employ the new extended Statehist index. Are the estimates improved by accounting for the state histories of the ancestors of present-day populations, instead of the state histories of places? To investigate this, we estimate the model for per capita GDP above using the ancestry-adjusted Statehist index. The results are displayed in Table 6, where we use the Statehist index in 1500 CE adjusted by the migration matrix (as in previous studies, but for the first time including full state history before 1 CE).⁴⁷

We find that the concave relationship between per capita income and the ancestry-adjusted Statehist is robust to all specifications and that the coefficients of interest are significant at 1% level in all columns in panel A. Moving from a linear (column 1) to a quadratic function (column 2) in Statehist as in all other tables, greatly improves the goodness of fit, strengthening the case for a nonlinear specification. Moreover, the explanatory power of the model (as measured by the regression R-squared) when we introduce only the ancestry-adjusted Statehist terms (column 2) is now 23.4% versus 5.2% for unadjusted Statehist.⁴⁸

In the online Appendix Table D12 we look at how sensitive the results from specifications where we include controls and county fixed effects are to excluding various countries from the sample. The estimates describing a concave function are significant when we exclude in turn the Middle-East and Sub-Saharan Africa, but the standard errors increase and the Statehist squared is insignificant when we exclude both these and North Africa (column 4). This is because this exclusion takes away a large part of the variation in Statehist that is due to very short and very long state age. Nevertheless, the signs and magnitudes of the coefficients point to the same concave relationship. There is no evidence of a concave relationship in the sample of European colonies, which are all very young and with little experience with home-based rule relative to the rest of the sample (median age 550 years and average 1180 years).⁴⁹

5.3 Statehist mechanisms

So far we have looked at the accumulated Statehist, which summarizes via their interaction (i) the variation in state age (the time elapsed since the first occurrence of $z^1>0$),⁵⁰ (ii) the degree to which the state was home-based ($z^2$), and (iii) the state’s territorial completeness and unity ($z^3$). In this sub-section, we briefly investigate the distinct effects of those three components and their respective contributions to the inverse-u shape. In Appendix Table D13, we estimate variants of columns (1), (2), (5) and (7) from Table 2, replacing Statehist with each of its components, in turn. Without controls (column 2), a quadratic relationship is found for each component (panels A–C), although for $z^3$ (territorial completeness), the coefficients in quadratic specification (2) are not statistically significant, whereas when included alone, the level term attains a positive coefficient (in column 1) that is significant at the 5% level. When all three separate component terms are included simultaneously, in panel D, the results are similar. Two of the three components (state age, and territorial completeness/unity) also obtain highly significant positive coefficients in a strictly linear specification (column 1). Taken together, these results support that each component, not only state age, has some importance in its own right, a finding not explored in previous studies. When additional controls are added in column (3), and continent fixed effects as well in column (4), only the coefficient on level of $z^2$ (home rule) remains statistically significant. Its positive sign suggests that having been independent rather than part of an externally based empire is most robustly and significantly associated with favorable outcomes, among the three components.⁵¹

5.4 Borders endogeneity and spatial dependence

We now turn to the issue of whether endogeneity of borders raises concerns about the reliability of our findings. Throughout history, borders shifted as states consolidated or weakened their administrative or military capacities, or incurred political regime changes. For instance, Alesina and Spolaore (1997) showed that democratization can lead to secession, an example of which is the case of the Balkan countries. Gennaioli and Voth (2015) showed that the post-1500 military innovations in Europe led to more efficient warfare, stronger state capacity to finance warfare, and imminent territorial conquests or redistribution, which have redefined country borders. Their examples of Silesia, Alsace-Lorraine, and the Duchy of Milan are joined by numerous others, such as the regions split in the Peace of Westphalia in 1648 amongst Sweden, France, The Dutch Republic, Spain and the Holy Roman Empire. European colonisation is a leading example where late comers, having overtaken older states, proceeded to redraw the borders of these states in Latin America, North Africa, the Middle East and the Indian subcontinent. Against this backdrop, defining the state experience and economic indicators on the basis of present-day country borders may induce several risks.

First, there is a risk of retroactive measurement error, due to the fact that the current countries’ territories used to compute historical values of Statehist often bear little resemblance with the geopolitical logic in ancient times (with a few exceptions like Norway, Sweden, and Japan). We should note that our definition of Statehist requires keeping track of the changing boundaries of states-of-the-time within and across half-centuries, and the scores account for multiple polities, the internal or external basis of their rule and the percentage of territory they occupied within what may be thought of as an arbitrarily defined territory, from the standpoint of early periods. Although the state presence in some polities or sub-territories may be measured with error, if this error is random, at worst our estimates are biased towards zero. It would be more difficult to predict the bias if measurement error was correlated with the outcome variables, e.g. if borders shifted through conquest or colonization to include territories with more/less state history, productivity and technological sophistication, potentially also associated with the accuracy of data sources. Even so, the amalgamation of low information areas together into countries with better information and more state history is likely to lead to a reduction of error for the low Statehist parts of those countries, since the histories of such areas are typically better documented than are those of otherwise similar areas more distant from record-keeping societies. A second concern is that, as discussed earlier, productivity and technological sophistication sometimes drive conquest, moving borders and influencing state history. While we cannot fully dismiss the problem of endogenous borders, the issue of selection into state history is partly mitigated in cross-country regressions with contemporaneous economic development by the fact that whatever state-level polities existed at this and that period within the boundaries of what are today countries have changed their borders on many occasions and for a wide variety of reasons.

An alternative to using country borders could have been to divide the world randomly into equal-sized grid cells and then study the history of states and economic development in each such cell (e.g. Michalopoulos 2012). State history has been coded at the grid-cell level for sub-Saharan Africa after 1000 CE by Depetris-Chauvin (2014). For the present study, however, this would require constructing disaggregated data on a global scale for nearly six thousand years. The challenge with this approach, in addition to the sheer magnitude of the exercise, would be the precarious state history information for many grid cells. Average quality of data may well be higher with the countries-of-today approach than would be achievable with grid cells, unless a research effort several orders-of-magnitude larger were undertaken. With these caveats in mind, to the extent that researchers are interested in tracking the histories of countries in order to understand contemporary levels of development, the modern configuration of countries is still a natural point of departure.

A third concern, which is particularly salient in the context of jointly determined borders, is that the histories and outcomes of neighbouring countries transcend national boundaries. Countries in the same geographical or geopolitical regions tend to have correlated productivity levels and experience similar productivity shocks, for instance through contemporary diffusion of new technologies. In the presence of such spatial spillovers, the standard assumptions on the independence of observations are violated, and OLS regressions might yield biased and inefficient estimates. In this case, modelling spatial dependence is a more suitable approach. To account for potential spatial autocorrelation in the disturbances, we estimated alternative regressions using the Conley (1999) correction of the standard errors.⁵² The results, displayed in Appendix Table D17 (panel A for Statehist and panel B for Ancestry-Adjusted Statehist) are very similar to the counterparts in Tables 5 (panel A) and 6, despite the fact that standard errors are, if anything, slightly larger than in the OLS regressions. We also estimate models where we allow the dependent variable to be a function of neighbouring countries’ outcomes (the spatial autoregressive model—SAR) and where we allow both the dependent variable and the errors terms to follow spatial autoregressive processes (SARAR).⁵³ In both models we input a matrix of weights given by the inverse great-circle distances between geodesic centroids.⁵⁴ Appendix Tables D18 and D19 show that the results with the Ancestry-Adjusted Statehist are robust across models, and that the SARAR results for Statehist generate qualitatively similar but insignificant results. We also report the SARAR estimates using a contiguity matrix defined not by geographical proximity, but by whether countries share the same legal origins (Appendix Table D20).⁵⁵ All results hold, with the linear and squared terms of both extended and Ancestry-Adjusted indices remaining significant. In sum, accounting for spatial spillovers does little to affect the results with extended state history and leaves the results with Ancestry-Adjusted indices unchanged.⁵⁶

6 Discussion and conclusions

To sum up, we have presented a model of the role of state history in economic development wherein growing state experience is associated with increases in productivity in the individual country, but where countries with less history of state presence may have a productivity advantage compared to ones with more experience of state institutions. We have coded and assembled a comprehensive data series on state history from state emergence (which often occurred before the Common Era) to 2000 CE for a sample of 159 countries, building on the previously constructed State antiquity index of Bockstette et al. (2002). The resulting empirical analysis revealed consistent reduced-form regressions, where a hump-shaped relationship is confirmed between extended Statehist and technology and economic development in 2000 CE. This relationship is robust to using a fairly wide range of discounts of the past, to controlling for duration of reliance on agriculture, and to assuming that migrants beginning with the age of European colonization carried the state experience of their home countries with them in the form of a portable and transmissible cultural heritage. Although the estimated coefficients cease to indicate an absolute decline in income at the highest levels of state experience after the addition of enough controls, this finding is consistent with the relationship being concave and monotonic within the individual country, the hump-shaped pattern being predicted only for cross-country and evidently cross-regional comparison at a particular historical juncture.

Although an extensive analysis of the causal mechanisms is beyond the scope of this paper, we believe we have offered support from the literature that improvements in the quality of institutions that are more easily achieved by states of intermediate age provide a plausible explanation for this pattern. Our finding appears to be consistent with the fact that while there is indeed a great deal of persistence of early societal advantages, it is also the case that the technological and institutional know-how of societies can slowly diffuse to neighbouring societies through migration or trade. These societies with younger states and/or a higher degree of autonomy can then pick the best practices of the older societies and potentially avoid some of the pitfalls that might have become a drag for the old civilizations.

State capacity might be one example of such institutional transfer across state borders. The ability to levy taxes and to consolidate an administrative infrastructure has recently been shown to produce regional spill-overs to neighbouring areas’ economic performance (Acemoglu et al. 2015). Such spill-overs might occur more easily, however, among countries sharing similar histories, including depth of experience with centralized authority.

There are also other factors that have been proposed for explaining the reversal in the Western core, including environmental degradation in the Fertile Crescent and in parts of the Mediterranean region. Once agriculture spread out of the Fertile Crescent, the more robust loess soils of northern Europe, combined with a reliance on rain rather than irrigation for cultivation, proved to be an advantage in the long run (Jones 1981). It has also been suggested that the rise and fall of dominant empires of the Western core followed cycles of expansion, over-extension, and eventually decline, with a gradual shift of power towards the northwest (Kennedy 1989). Acemoglu et al. (2005) show that the emergence of Atlantic trade after 1500 CE had a major impact on the rise of for instance Spain and the United Kingdom.⁵⁷

A similar process can potentially explain comparative development in East Asia. Japan’s less powerful central court and greater perceived vulnerability to potential Western colonizers led it to undertake decisive modernization measures almost a century before China. This development had spillover effects on Korea and Taiwan, all young states in comparison with China.

In summary, the new pattern uncovered by the extended Statehist shows that although greater depth of state experience is associated with better economic outcomes when the measure used emphasizes mainly the millennium before 1500, the relationship of state history to income and technology includes a range of excessive age or senescence, if we account for the BCE millennia and give them sufficient weight. We leave it for future work to attempt to identify the exact mechanisms behind this pattern. However, we believe caution is recommended against the interpretation of these disadvantages as fully automatic and insurmountable consequences of long state histories. Our view is not that a long uninterrupted state history is always bad for economic development and as such undesirable. We believe this is a story of moderation in the exercise of centralized power and adaptability of the state institutions to the ever-changing economic realities. While those in the middle range of state history have thus far exhibited such moderation and adaptability more effectively, on average, there remains a considerable space of indeterminacy within which political actors may still exert influence over their countries’ fates.