12.1 Ten Thousand Years of Prehistory

No one disputes that the city-states of southern Mesopotamia were states, or that dynastic Egypt was a state. Both arose shortly before 5000 years ago. At about the same time cuneiform writing developed in Mesopotamia and hieroglyphic writing developed in Egypt. Additional pristine states emerged in other parts of the world over subsequent millennia, and most had their own writing systems. However, from a global perspective the line between prehistory and history was crossed 5000 years before the present.

Some of the developments we have discussed in this book have unclear starting points. For example, there may have been occasional sedentism in the Upper Paleolithic at times and places where resource patches happened to be concentrated and predictable. Similarly, raiding among foraging groups may extend into the indefinite past, although warfare aimed at gaining permanent control of valuable land appears to be a more recent development. But in general, substantial and lasting transitions to sedentism, agriculture, inequality, warfare, cities, and states began between 15,000 BP and 5000 BP.

From the standpoint of an individual human life, 10,000 years is a very long time, about 500 generations. But in relation to the time Homo sapiens has been on the planet, perhaps 300,000 years, it is only 3% of the total. Of the time in which the biological genus Homo has been on the planet, perhaps 2.3 million years, it is only 0.4% of the total. On such time scales the 10,000 years of central concern in this book were a blink of an eye, and the six transitions we have studied were revolutionary in their speed and scope.

One cannot convincingly explain a set of transitions this dramatic and widespread by appealing to random local conditions or idiosyncratic cultures. Global revolutions are best explained by global causes. We believe the prime mover for the revolutionary social transformations of 15,000–5000 BP was climate change. We argued in Chapters 4 and 5 that the recovery from the Last Glacial Maximum, followed by the abrupt shock of the Younger Dryas and the arrival of the Holocene, were the triggers for sedentary lifestyles among foragers and later for village agriculture.

Domestication of plants and animals set off a process of technological innovation that has never ended. Over time this trajectory became increasingly decoupled from the natural environment and no longer required an external stimulus from climate shocks. For Malthusian reasons, rising agricultural productivities resulted in rising population densities. We argued in Chapters 6–8 that population growth led to class stratification between elites and commoners at particularly attractive sites, and that stratification laid the foundations for chronic warfare over land among rival elites.

Finally, we showed in Chapters 9–11 how these processes ultimately led to cities and states. We argued that there were at least three causal pathways in the intertwined development of cities and states: one based on property rights, involving gradual closure of the commons and falling living standards for commoners on Malthusian time scales; a second based on warfare among elites, leading to agglomeration in cities by the defenders and formation of large territorial states by aggressors; and a third based on environmental shifts during the Holocene, which sometimes pushed rural populations into refuge areas, often in river valleys, that evolved into city-states by way of urban manufacturing.

Although the resulting societies displayed kaleidoscopic institutional and cultural variation, we believe the underlying economic processes were relatively simple. The key exogenous variables were geography and climate. Geographical variations across regions of the world, and across the individual sites within regions, were largely permanent, but played a crucial role by determining the spatial distribution of natural resources. Climate varied from one decade, century, or millennium to another, as well as across regions, and accounted for the timing of several transitions through its interaction with geography.

Two other key variables, technology and population, were endogenous on long time scales. Technology largely evolved through learning by doing. This created both stagnation traps, where societies continued using known techniques for long periods of time, and intervals of rapid technological innovation when external conditions such as climate shocks pushed societies into the use of new resources or production techniques. Population largely evolved according to Malthusian principles, where regional densities rose when climate improved or technology became more productive. Throughout the 10,000-year period with which we have been concerned, the gains from technological innovation went almost entirely into population growth rather than improvements in the average standard of living.

In short-run settings where technology and population were largely fixed, social responses to environmental shocks depended on other factors. The most prominent were migration and warfare. We have argued in a number of cases that migratory responses to environmental shocks triggered cultivation and manufacturing. We have also argued that large shocks to climate or technology changed the relative productivities of sites, which sometimes led to warfare when individual migratory responses were precluded by social barriers to physical mobility.

We have offered a number of hypotheses that are open to archaeological testing. We do not expect every hypothesis to hold up under empirical scrutiny. Science is about taking risks, especially the risk of being wrong. But even if the reader is skeptical about some (or all!) of our hypotheses, we hope she or he has been persuaded that economics can be a fruitful source of research ideas about prehistory.

The great advantage of economic reasoning is that it can explain many seemingly disparate phenomena using a fairly small set of concepts and assumptions. Accordingly, it facilitates theoretical unification. The price, of course, is that many important details of individual cases must be suppressed. For those who want to stress the unique features of each society, this price will be too high. But for those who believe that the evolution of human societies displays recurrent patterns that are open to causal explanation, perhaps the price will be acceptable.

12.2 Five Thousand Years of History

This has largely been a book about “firsts”: the first sedentary foragers, the first cultivators, the first states, and so on. If one wants to understand major transitions of this kind, it makes sense to look for the earliest instances of the phenomenon in question and explore the forces that could have generated that phenomenon in its pristine form. But by definition, pristine examples are not representative. While the residents of Abu Hureyra were cultivating cereal crops, almost everyone else was gathering wild plants and hunting wild animals. While the Mesopotamian city-states were emerging, almost everyone else was living in societies where state-level institutions were absent.

A similar perspective is needed in thinking about the last 5,000 years. Until quite recently, foraging societies persisted in large parts of the Americas, Africa, and Asia, as well as all of Australia and the Arctic. Elsewhere agricultural villages of a few hundred people were the norm. In some places these villages were politically autonomous, while in other places they were components of two-tier or three-tier settlement hierarchies that anthropologists might call simple or complex chiefdoms.

As we have seen, over the last 5,000 years cities and states emerged in the Indus River region, China, Mesoamerica, and the Andes. Similar processes led to cities and states in the Sahel region of West Africa, Ethiopia, Zimbabwe, and on the east coast of Africa. Additional precocious developments occurred at Cahokia in North America and Angkor Wat in Southeast Asia. Within Europe, city-states emerged in Greece, while the village of Rome became a city, a state, and eventually the capital of an empire.

Much of world history over the last few millennia is a dreary and repetitive story about kings (and a few queens) who fought wars and built empires. Often these empires remained intact for a few generations or a few centuries, collapsed, and were replaced by new empires assembled by new rulers. Despite these micro-level instabilities, at a macro level this system of rising and falling empires superficially exhibited a kind of structural stasis. But gradually the fraction of the globe subject to state control grew (Borcan et al., Reference Borcan, Olsson and Putterman2018) and the number of autonomous political units declined (Bowles, Reference Bowles2012).

This process culminated with the conquest of the Americas and Australia by the European powers, which also colonized large areas of Africa and Asia. Perhaps 50–80% of the indigenous population of the Americas was lost to virulent diseases like smallpox and measles. Others were lost to genocide, enslavement, and forced assimilation. The Atlantic slave trade added further horrors, as did other slave trades in the Arab world and around the Indian Ocean. The settler populations of the Americas gained independence from Europe by about 1820, and decolonization movements in Africa and Asia led to the formation of many independent nations after World War II.

Three more developments had massive effects on the modern world: the Industrial Revolution, the Demographic Transition, and political democracy. Together these events provided billions of people with rising incomes and protection from abusive elites. They represent a seventh transition in world history, and we will discuss them in this chapter.

In the process we will undertake a deeper exploration of the causal connections between prehistory and contemporary global civilization. It would be natural to assume that prehistory is dead and buried, but this belief is incorrect. The ghosts of prehistory continue to stalk the modern world. Three provocative lines of research show how.

The first, pioneered by Louis Putterman, Ola Olsson, and their co-authors, argues that early agriculture and early state formation conferred persistent advantages on some modern nations. Specifically, nations that had earlier agricultural transitions and earlier state institutions on average enjoy higher incomes or rates of economic growth today. A large number of caveats must be attached to this summary statement, but the correlations are striking. The possibility that prehistoric events exerted causal influences extending over millennia, and continue to shape economic growth today, is highly intriguing. We discuss this research in Section 12.3.

Oded Galor and his co-authors approach related questions from a more theoretical point of view. Their goal is to develop a causal framework that can explain the complete sweep of economic development from Neolithic agriculture to modern industrialization. This ambitious research program, called “unified growth theory,” incorporates technology, population, and other variables that have been central to this book. We will describe this theoretical framework and the facts it can explain in Section 12.4.

Carles Boix is a political economist who adopts an equally long time frame, ranging from prehistoric foraging to modern democracies. He emphasizes the development of military technology, a subject we have not highlighted, and argues that the emergence of political democracy was tied to an unusual constellation of economic and military conditions that prevailed in early modern Europe. We review his arguments in Section 12.5, along with related arguments about the origins of political democracy from Douglass North and his co-authors, as well as Daron Acemoglu and James Robinson.

These ambitious efforts to understand social, political, and economic change in the very long run provide foundations for a further inquiry. What, if anything, can one say about trends in human welfare over the last 15,000 years? Has life gotten better for the majority of people, has it gotten worse, or is the picture more complicated? Given data limitations, we focus on basic indicators such as diet, health, and life expectancy. For the majority of the world’s population, we believe the trajectory of human welfare has followed a U shape: reasonably good for mobile hunter-gatherers, worse for commoners in agricultural societies and ancient cities, and improving drastically since the Industrial Revolution. The theory and evidence behind these claims are described in Section 12.6.

We close this chapter by returning to our key exogenous variable: climate. We have argued that climate change from the late Pleistocene through the Holocene was the principal driving force, either directly or indirectly, behind the six transitions examined in this book. Given the existential threat our own civilization faces from global warming, it is appropriate to conclude by asking whether prehistory offers any lessons for our current predicament. We hazard a few remarks on this topic in Section 12.7, followed by a brief postscript in Section 12.8.

12.3 The Shadow of the Past

Economists interested in very-long-run economic growth have been intrigued by the arguments of Diamond (Reference Diamond1997), who asserts that differing biogeographic endowments across regions of the world (in particular, the supply of domesticable plants and animals) led to differences in the timing of pristine agriculture. In turn, regions that developed an early agricultural economy tended to have a high early population density as well as an early transition to state institutions. According to Diamond these developments led to permanent technological, military, and institutional advantages that explain why Spain conquered the Incan Empire rather than vice versa, and more generally why European nations were able to colonize large parts of the world during the last 500 years.

Diamond’s arguments inspired Louis Putterman, Ola Olsson, and their co-authors to study a related question: Do countries with a longer history of agriculture or state-level institutions have advantages with respect to modern economic performance? If so, why? This agenda was pioneered by Bockstette et al. (Reference Bockstette, Chanda and Putterman2002) and Hibbs and Olsson (Reference Hibbs and Olsson2004). The literature has grown enormously in recent years and we cannot provide a comprehensive review, but we will discuss several key findings and interpretive issues.

Economists have long understood that “good” institutions contribute to economic growth while “bad” institutions restrain it (for a highly readable series of case studies, see Acemoglu and Robinson, Reference Acemoglu and Robinson2012). Institutional quality has been measured in a number of ways, including indexes for social development, social infrastructure, social capital, and the rule of law. Bockstette et al. (Reference Bockstette, Chanda and Putterman2002) added something new: the sheer antiquity of state institutions has apparently had a powerful influence on modern economic growth.

Bockstette et al. carried out the following empirical exercise. For each of the 119 modern countries having the necessary data, they divided the interval from 1 CE to 1950 CE into 39 periods of 50 years each. For each half-century they determined whether the country had a government above the tribal level, whether this government was local or foreign, and how much of the territory of the modern country it controlled. After giving points on each criterion, they calculated a measure of state antiquity for the country. To put less weight on events in the distant past, they discounted the scores by 5% for each half century, although the results were not sensitive to the details of this procedure.

The resulting index, called “statehist5,” was positively correlated with measures of political and institutional quality such as stability, lack of corruption, lack of government contract repudiation, lack of expropriative risk, the rule of law, and bureaucratic quality; negatively correlated with ethnic fragmentation; positively correlated with the population density in 1960, social development, and civic norms; and positively correlated with GDP per capita in 1960, 1970, 1980, 1990, and 1995, as well as GDP growth during the period 1960–1995. Of course, simple correlations do not imply causality, an issue to which we will return below.

The next step was to include statehist5 in conventional cross-country regression equations used to explain economic growth. The dependent variable was GDP growth during 1960–1995 and the control variables included the level of GDP in 1960, schooling, population growth, and the investment rate. When statehist5 was added, it had a positive sign and was consistently significant at the 0.01 level. This continued to be true, and the magnitude of the coefficient remained stable, when more controls were added, including measures of institutional quality, population density in 1960, ethnic fragmentation, and dummy variables for region of the world (East-Asia Pacific, Latin America, and so on). Furthermore, the effect was large: the difference in statehist5 between China (the oldest state) and Mauritania (one of the newest) accounted for about half of the total difference in growth rates between the two countries. Similar results were obtained for a subsample excluding the OECD (Organization for Economic Co-operation and Development) countries, so the findings were not driven by the presence of richer countries in the sample.

We now return to the question of causality. Given the way in which statehist5 is defined, reverse causality can be ruled out: high economic growth today does not cause state formation centuries ago. However, it is possible that this variable simply proxies for institutional quality without having any further effects of its own. Bockstette et al. tested this possibility by including both statehist5 and a commonly used measure of institutional quality in their regressions. Both variables were significant, showing that statehist5 adds predictive power. The same was true when population density was included. Therefore statehist5 is not merely proxying for these variables. Moreover, ethnic fragmentation was no longer significant when statehist5 was also included. Statehist5 remained significant and quantitatively strong when regional dummy variables were added, showing that it helps explain differences in economic growth even among countries in the same region.

These results suggest that processes of state formation over the last two millennia have had powerful effects on modern economic growth. Bockstette et al. suggest that the effects of the Industrial Revolution diffused slowly at first, but that in the second half of the twentieth century countries having greater antiquity of state institutions were better able to catch up with the industrial leaders, and that this linkage became stronger as the effects of colonialism waned. The nature of the causal linkage is not entirely clear, but Bockstette et al. propose several possibilities: (a) countries with a longer history of state institutions have had more time to benefit from learning by doing in public administration; (b) the populations in such countries may have more positive attitudes toward state bureaucracy; and (c) state antiquity promotes linguistic unity, which may encourage a sense of shared identity and social trust, while discouraging civil wars and political instability.

Deeper causal issues remain. For example, one might argue that some geographic factor X (such as having many domesticable species, or fertile soil for agriculture) led to both early state formation and high economic growth, so the correlation of statehist5 with modern growth rates is spurious. Bockstette et al. (Reference Bockstette, Chanda and Putterman2002) were aware of this issue but did not address it directly. However, research on migration discussed below tends to counter these concerns by suggesting that the causal links involve human capital rather than fixed characteristics of geographic locations.

Hibbs and Olsson (Reference Hibbs and Olsson2004) and Olsson and Hibbs (Reference Olsson and Hibbs2005) address the influence of an early pristine transition to agriculture on modern incomes. We base our discussion on the first of these articles. The authors build upon Diamond’s (Reference Diamond1997) insight that some regions of the world had more wild plant and animal species that could be domesticated, and that an early agricultural transition gave such regions a persistent head start with respect to the broader process of economic development. Causality for Hibbs and Olsson runs from the geographic features of a region (climate, latitude, and the degree of east–west orientation of the continental axes) and its biological endowment of domesticable plants and animals, to its predicted date for a pristine transition to agriculture, and finally to the level of GDP per capita in 1997. Modern countries were assigned to eight regional groups based upon the distinct biological endowments of their regions (the sample includes 112 countries).

Hibbs and Olsson (Reference Hibbs and Olsson2004) use six pristine agricultural transitions (southwest Asia, China, west Africa, the Andes, Mesoamerica, and the eastern United States) to estimate a relationship between biological endowments and the timing of the agricultural transition. The predicted timing of this transition is then used in cross-country regressions to explain modern incomes. They find that geography, predicted agricultural timing, and the square of predicted agricultural timing together account for 57% of the variance across countries in the log of 1997 per capita income. All three variables are significant at the 0.01 level and the effects are quantitatively substantial. When the authors add a standard index of institutional quality (an average of bureaucratic quality, rule of law, low corruption, low expropriation risk, and low risk of government contract repudiation), this increases the explanatory power of the equation to 80%.

A large part of the variance across countries in modern institutional quality (43%) is explained by the time since a pristine agricultural transition in the region. Furthermore, geography and time since the agricultural transition remain significant when institutional quality is included. Thus an early agricultural transition probably confers some benefit through the presence of good institutions, but it also makes an independent contribution.

A byproduct of this research is the finding that regions with better geographical and biological characteristics had earlier pristine agricultural transitions. The sample size is small, but this finding provides support for the perspective of Diamond (Reference Diamond1997). As we argued in Section 5.12, Diamond proposes a theory about the supply side of the Neolithic transition, and Hibbs and Olsson provide evidence that these supply-side factors matter. However, the timing of a given pristine transition also depends on demand-side factors, including climate change. The geographic index used by Hibbs and Olsson incorporates a static climate variable associated with the suitability of a region for agriculture, but not a variable describing changes in climate over time. Thus they do not test the hypothesis we developed in Chapter 5.

We are impressed that a large fraction of the variation in incomes across modern countries can be explained by variables that are undeniably exogenous and prehistoric in nature (geography, biology, and years since the origin of pristine agriculture in a region). The main causal issue is that these variables could be proxies for other factors that affect incomes today (such as latitude or climatic suitability of a region for agriculture), so that it is not the elapsed time since the advent of agriculture that is driving the results.

Putterman (Reference Putterman2008) builds upon the work of Hibbs and Olsson (Reference Hibbs and Olsson2004) by gathering country-specific information on the timing of initial agriculture in 112 nations. He shows that nations having early Neolithic transitions have higher income levels in the year 1997 after controlling for geography, and that this effect is already visible in income data as far back as 1500 AD. When “years since transition” is measured using the methods of Hibbs and Olsson, agricultural timing remains significant as a determinant of modern income after controlling for institutional quality, but is insignificant when agricultural timing is measured using the Putterman country-specific approach.

One important difference between Hibbs and Olsson (Reference Hibbs and Olsson2004) and Putterman (Reference Putterman2008) is that the former use information about pristine agricultural transitions in eight regions of the world, while the latter mainly relies on information about the diffusion of agriculture. Putterman’s results should be interpreted as showing the effects from an early arrival of farming in a particular nation, not the effects of an early pristine transition (although the sample does include a few pristine cases). Nevertheless, Putterman confirms the finding from Hibbs and Olsson (Reference Hibbs and Olsson2004) that geographic factors (climate, latitude, east–west versus north–south continental axes) are a powerful determinant of the timing of the agricultural transition. For more on the diffusion process, see Ashraf and Michalopoulos (Reference Ashraf and Michalopoulos2015).

Probably the most important finding of Putterman (Reference Putterman2008) is that the advantages from lengthy experience with agriculture and states can be transferred from one region of the world to another through migration. To show this, Putterman calculates a measure of “years since transition” that is a weighted average based on the years since transition for the source countries of immigrants, with weights based on the fraction of migrants from each source country. This calculation is based on migration flows over the last 500 years (including, but not limited to, the migration of Europeans to the USA, Canada, Australia, and New Zealand). Putterman shows that the fit of the model for the 1997 income data is substantially better when “years since transition” is adjusted for migration in this way. For further results along these lines, see Putterman and Weil (Reference Putterman and Weil2010).

This tends to blunt any critique that correlations between prehistoric events and modern growth are driven by some geographic factor that influences both. Apparently individuals can carry the benefits of lengthy experience with agriculture and states when they move, suggesting that the crucial causal links involve human capabilities rather than geographic conditions. These capabilities may include technical knowledge, institutional experience, cultural norms, or similar factors.

Putterman finds that the antiquity of agricultural technology is highly correlated with the antiquity of state institutions (a bivariate correlation of 0.649, significant at the .001 level), where state antiquity is measured as in Bockstette et al. (Reference Bockstette, Chanda and Putterman2002). Moreover, the state antiquity variable again helps to predict 1997 incomes, and the fit of the model improves substantially after correcting the state antiquity variable for post-1500 CE migration flows. With this adjustment, state history explains about 25% of the variance in national income per capita for 1997. Borcan et al. (Reference Borcan, Olsson and Putterman2021) have confirmed the close connection between early agriculture and early states.

The data set for the history of state institutions used by Bockstette et al. (Reference Bockstette, Chanda and Putterman2002) has been extended back to 3500 BCE by Borcan et al. (Reference Borcan, Olsson and Putterman2018), so it now covers all of the episodes of pristine state formation discussed in this book, including Mesopotamia and Egypt. Data are available for 159 modern nations. Borcan et al. discount the past using a lower rate than Bockstette et al. (1% per 50-year period rather than 5%) but their results are not highly sensitive to small changes in the discount rate.

Borcan et al. report two major empirical findings. First, a relationship between a long history of state institutions and productivity is already visible in 1500 CE. A linkage of this kind is found for various measures of productivity including technology adoption, population density, and urbanization. Interestingly, quadratic terms for state history are large and significant even with numerous control variables for geography and climate (but not if continental fixed effects are included). This yields an increasing and concave relationship between the antiquity of the state and productivity in the year 1500. Also, a variable for the antiquity of agriculture adds explanatory power for technology adoption and population density (but not urbanization).

Second, there is a similar quadratic relationship between state history and GDP per capita in the year 2000, again with numerous geographic controls. The coefficient on the quadratic term is negative enough that the overall relationship is hump-shaped, with states of intermediate age having higher income per capita than those that are either older or younger. The antiquity of agriculture is insignificant, suggesting that in modern data it makes no additional contribution to income levels beyond what state institutions provide. The hump-shaped relationship is only visible for regressions using the entire state history going back to 3500 BCE (or 5500 BP), and disappears when the state history variable is defined using the data since 1 CE. Thus the early millennia in the dataset are crucial for this result. Explanatory power is greatly improved by the use of state history measures adjusted for migration flows (the latter explain 23% of the variance in modern incomes rather than just 5%). The qualitative results described in this paragraph are also obtained when the dependent variable of GDP per capita is replaced by technology adoption.

Borcan et al. suggest that a simple theoretical model can account for these results. Suppose (a) individual states tend to have rising productivity over time due to increasing fiscal and institutional capacities; (b) this process has diminishing returns because elites eventually divert tax revenue to unproductive activities; and (c) newer states learn from the experience of older states, so their peak productivity level exceeds that of the earlier states. At a given moment in time there will be a mix of old states that have become too centralized and have relatively low productivity, states of middling age that have learned from their predecessors and achieved higher productivity, and young states that have low but increasing productivity. This pattern gives a hump-shaped relationship between state antiquity and economic performance in modern cross-sectional data.

We close this section with some fascinating recent findings about the relationship between early agricultural transitions and modern economic performance. Although at a global level, early transitions are positively correlated with contemporary incomes across nations, this relationship appears to be driven by differences across regions with distinct pristine agricultural transitions (specifically western Eurasia, East Asia, and sub-Saharan Africa). Within each of these regions, Olsson and Paik (Reference Olsson and Paik2020) find that there is actually a negative relationship between an early agricultural transition and GDP per capita in 2005. For example, in western Eurasia countries like Iraq and Syria are currently poor, while countries like the Netherlands and Sweden are currently rich, despite the late arrival of agriculture in northwestern Europe from the core area of southwestern Asia. It appears that similar reversals occurred in East Asia and sub-Saharan Africa, and that the western reversal was visible by 1500 CE, prior to European colonization and industrialization.

Olsson and Paik suggest two possible explanations. First, there is the institutional explanation of Borcan et al. (Reference Borcan, Olsson and Putterman2018), where early agriculture led to early states. The elites in early states tended to restrain economic growth, while elites in later states were able to learn from earlier states and have not (yet?) reached the point where they limit growth to the same degree. A second explanation is cultural. In this view early farmers in the core area of southwestern Asia had a strongly collectivist orientation for various reasons, and agriculture was carried into Europe through the migration of more individualistic farmers. Repeated self-selection of this kind led to more individualistic cultures in northwestern Europe, which were conducive to economic growth. Olsson and Paik (Reference Olsson and Paik2016) document the existence of a cultural gradient in collectivism versus individualism that is correlated with the timing of agricultural adoption in western Eurasia.

12.4 Unified Growth Theory

Most readers have no doubt heard of the Industrial Revolution, have a general awareness that it occurred in the last few centuries, and believe it contributed to rising standards of living for people around the world. A less well-known revolution, but as important for living standards, is the so-called Demographic Transition, which followed in the wake of the Industrial Revolution. We begin with an explanation of this concept.

Throughout this book we have worked with Malthusian population dynamics. Briefly put, the Malthusian framework says that if living standards rise in the short run (perhaps due to climate improvement, technological advance, or the availability of new land), population will rise in the long run, and the resulting population growth takes the standard of living per person back to its previous level. In the long run any productivity benefits from technological progress are absorbed through population growth rather than higher income per capita. We argued in Chapter 2 that this was an appropriate way to think about population dynamics in prehistory, and drove the point home theoretically and empirically throughout the book.

The Demographic Transition of the last century and a half turned all of this on its head. As industrialization spread around the world, societies shifted from a regime where people have more children when they become richer to a regime where people have fewer children when they become richer. This trend is visible in several ways: (a) in individual countries, the rate of population growth tends to slow down as GDP per person increases; (b) in cross-sectional samples, the richer countries tend to have slower population growth than the poorer countries; and (c) within a specific country, the richer people tend to have fewer children than the poorer people.

The consequences have been enormous. For the first time in history or prehistory, productivity growth has persistently outpaced population growth, generating a sustained rise in living standards for the majority of the world’s population. These gains have been very uneven. Some countries had an earlier demographic transition and an earlier start to modern growth, while others had a later transition (or no transition) and a later start to modern growth. This has created a “great divergence” of income levels across countries, which development economists have tried to understand and ameliorate.

A key question is how the modern world broke out of the Malthusian trap. Many economists have worked on this question, and the related question of why the Industrial Revolution happened at all. Space limitations preclude a comprehensive treatment here. Instead we focus on the contributions of Oded Galor, an economist who has attempted to create a unified theory that accounts for the trajectory of economic growth from Neolithic agriculture to the present day. Some of Galor’s ideas are controversial and we indicate a number of subjects for debate. But Galor’s ambitious attempt to bridge the gap between prehistory and modernity deserves attention.

In this section we treat Galor (Reference Ashraf and Galor2011) as the definitive statement of unified growth theory, but Galor (Reference Galor, Aghion and Durlauf2005) supplies another useful summary. Interested readers can explore the reference lists in these two sources, as well as Galor’s journal articles with various co-authors. Galor (Reference Ashraf and Galor2011, 4–5) wants to explain the following features of the development process: (a) Malthusian stagnation, (b) the escape from the Malthusian trap, (c) human capital formation, (d) the Demographic Transition, (e) contemporary sustained growth, and (f) the modern divergence of income per capita across countries.

For this purpose Galor divides economic development into a Malthusian epoch, a post-Malthusian regime, and a modern growth regime. The Malthusian epoch covers the entire period from early agriculture until recent centuries. It is characterized by gradual technological advance, gradual population growth, and stagnation of income per capita. Growth theorists frequently emphasize a positive feedback loop between technology and population under such conditions, where better technology leads to higher population for the normal Malthusian reasons, and higher population leads to more rapid technological advance due to a larger number of innovators, more learning by doing, greater gains from specialization, and the like (see Kremer, Reference Kremer1993, for an early model). When Galor refers to “stagnation” in the Malthusian epoch, he is referring to income per capita, while accepting the underlying dynamism of technology and population.

The post-Malthusian regime exhibits simultaneous growth in all three variables: technology, population, and income. The Malthusian causal channel going from income to population still exists. However, productivity growth from technological innovation and capital accumulation begins to outpace population growth so that income per capita starts to rise. There is not yet a demographic transition, and income growth leads to an acceleration of population growth.

Finally, development moves into the modern growth regime, where technology and income continue to grow, but the population growth rate slows and the Demographic Transition becomes visible. Productivity continues to rise due to technical innovation and investment in both physical and human capital. This and the declining population growth rate lead to sustained expansion of income per capita.

Most growth theorists, development economists, and economic historians would probably agree that this sequence of events is a reasonable description of reality. Galor documents the sequence using an array of facts, statistics, graphs, and regressions, which we need not review here. The transitions from one growth regime to another unfolded at different times in different parts of the world, with the earliest transitions generally taking place in the UK, western Europe, North America, Australia, and New Zealand, and later transitions occurring in Asia and Latin America. The modern growth regime does not yet prevail everywhere, and is most notably absent in some parts of Africa. The key issues of interest here are the reasons for the transitions from one growth regime to another.

First, however, we comment briefly on some empirical findings. Galor (Reference Ashraf and Galor2011, ch. 3) goes to considerable lengths to show that the Malthusian epoch was in fact Malthusian. He estimates cross-country regressions for the pre-industrial world where technology and land productivity have large significant effects on population, but insignificant effects on income per capita (see also Ashraf and Galor, Reference Ashraf and Galor2011). These results are obtained whether population and income are measured in 1 ce, 1000 ce, or 1500 ce. All of the results are robust to the inclusion of numerous geographic controls.

For most regressions, technology is proxied by the elapsed years since the start of agriculture and instrumented by endowments of domesticable plants and animals. Galor’s results are consistent with other findings that the Neolithic transition has had long-lasting consequences for subsequent economic development (see Section 12.3). In particular, the time elapsed since the beginning of agriculture in a country is strongly associated with that country’s non-agricultural technology (communication, transportation, and industry) in the years 1 ce and 1000 ce. The Neolithic transition clearly touched off a process of productivity growth that played out for many millennia afterward.

We turn next to Galor’s (Reference Ashraf and Galor2011, ch. 4) explanation for the Demographic Transition. The main question is why fertility and population growth rates dropped in the wake of the Industrial Revolution. Galor argues on theoretical and empirical grounds that fertility fell mainly because the second phase of the Industrial Revolution, starting around 1870 in the advanced countries, increased the demand for human capital. This demand resulted from the increasing complexity of industrial production and associated organizational activities like management, accounting, and marketing. One consequence was a major expansion in public education, promoted to a large degree by industrialists.

At the level of the individual household the increasing demand for human capital caused substitution from the quantity of children to the quality of children, especially the resources families invested in education. Galor argues that education was beneficial in an economic environment where technology was changing rapidly and existing skills tended to become obsolete. Due to this shift from quantity to quality, fertility declined.

A supporting factor was a shrinking gender wage gap due to greater demand for cognitive work in relation to physical labor. Because most child-raising burdens were placed on women, the increasing opportunity cost for women of having children tended to reduce fertility. Another factor was a growing wage gap between adults and children, which reduced the incentive for parents to treat their children as sources of income.

Galor rejects or minimizes the importance of various alternative hypotheses. He considers the idea that higher incomes led directly to reduced fertility, but argues that the data do not support this idea. Although the northwestern European countries had roughly simultaneous demographic transitions around the start of the twentieth century, levels of income per capita were quite different across countries. He argues that the fertility drop was more closely linked to rates of income growth than to levels of income, supporting the view that the process was driven by rapid technological innovation.

He also rejects the claim that the reduction in fertility was a response to a decline in infant and child mortality, arguing that mortality had been declining in western Europe for about a century before the reduction in fertility, and that in some countries lower child mortality was initially associated with rising fertility. Finally, Galor sees improvement in old-age security as a minor component of the Demographic Transition because the timing of the relevant institutional changes does not closely coincide with fertility trends.

Without going into the mathematics, we sketch the framework of Galor’s formal model in order to show how he explains the transition from the Malthusian epoch to the post-Malthusian regime, and from the latter to modern economic growth. There are four key endogenous variables: the rate of technological growth, the level of population, the level of parental investment in education (broadly defined), and resources per worker (in effect, income per capita).

Output is obtained from a Cobb–Douglas production function with inputs of labor and land (the latter does not change over time), with a coefficient indicating the level of technology. Consumers have Cobb–Douglas utility functions depending on consumption and the aggregate human capital of their surviving children. A crucial wrinkle is that the consumers have a positive lower bound on consumption due to subsistence requirements. Parents choose the quantity and quality of their children, as well as their consumption, where quality depends on educational investments. The human capital of a child is an increasing function of education and a decreasing function of the rate of technological change (due to the obsolescence of skills). There are two key thresholds in the model: the rate of technological progress at which parents start to invest in education, and the income level at which the subsistence consumption constraint no longer binds.

At low levels of population, the difference equations describing the model have a unique stable equilibrium where parents invest nothing in education. There is a feedback loop between technology and population but both grow slowly. Due to Malthusian forces growth in per capita income is zero or tiny. But eventually population growth generates a qualitative shift in the dynamics, where two stable equilibria are separated by an unstable equilibrium. The first stable equilibrium has Malthusian features while the second has high levels of education and technological innovation. The Malthusian equilibrium is locally stable so the system remains in this vicinity.

Another qualitative shift occurs at higher population levels, where the Malthusian equilibrium vanishes and the only stable equilibrium is the one having high education and rapid innovation. At this point the economy moves into the post-Malthusian regime. The subsistence constraint on household consumption continues to bind, but incomes increase and some resources are devoted to education. However, the demand for human capital is initially limited, so the relaxation of the budget constraint leads to increased family sizes.

Eventually a virtuous circle involving a positive feedback loop between education and innovation takes the economy into the modern growth regime. Parental substitution from quantity to quality of children gives declining fertility and a demographic transition where the rate of population growth drops. A larger share of the productivity gains from technological progress is therefore devoted to growth in income per capita.

We pause at this point to compare Galor’s agenda with our own. There are areas of overlap with respect to key variables (especially technology and population) and time frame (a shared interest in events during the Neolithic). However, the two projects have different goals. Galor’s ambition is to construct a mathematically unified theory in which his four endogenous variables mimic the transition from Malthusian stagnation to modern growth. His formal model is not designed to explain why the Industrial Revolution began in England rather than France or China, or why it began around 1750–1820 rather than a few centuries earlier or later.

Our ambition has been to explain a series of prehistoric transitions using a family of related models, which are unified by consistent assumptions about certain core factors, especially time allocation, population dynamics, and technological dynamics. We have frequently extended our framework in order to accommodate the facts about a particular transition, for example by including migration, warfare, or manufacturing. Our quest has been to explain the details of place and time: why did a transition occur within a specific region at a specific date?

Economic historians have an enormous literature that attempts to answer precisely these questions in the case of the Industrial Revolution. Galor’s theoretical approach has been criticized for operating at a level of abstraction that is too high to address questions of this kind (Temin, Reference Temin2012). We also note that Crafts and Mills (Reference Crafts and Mills2009) found empirical difficulties with the causal channel running from population to innovation, which plays a central role not only in Galor’s approach but in many other models of economic growth. Alternative hypotheses about the sources of technological innovation in the Industrial Revolution emphasize knowledge networks (Mokyr, Reference Mokyr2002), mechanistic science (Lipsey et al. Reference Lipsey, Carlaw and Bekar2005), institutional innovation (North, Reference North1981; Acemoglu et al. Reference Acemoglu, Johnson, Robinson, Aghion and Durlauf2005; Greif, Reference Greif2006), and reproductive success (Clark, Reference Clark2007). For a review of what is currently known and unknown about the Industrial Revolution, see Clark (Reference Clark, Aghion and Durlauf2014).

We note a few other important differences between unified growth theory and our own approach. Galor’s theory has no explicit role for climate, or the natural environment more broadly, and he does not explain the transitions from mobile to sedentary foraging, or from sedentary foraging to agriculture. We have argued that climate change drove the evolution of technology and population during the Upper Paleolithic (Chapter 3) and that it triggered the transitions to sedentism and agriculture (Chapters 4 and 5). In effect, we add a climate-based economic regime at the beginning of Galor’s sequence.

The amount of progress generated via endogenous growth mechanisms depends on the degree to which production is isolated from environmental shocks. In a foraging society technological advance is almost entirely driven by shocks from nature, if it occurs at all. Feedbacks from population or human capital to technology have little quantitative impact except when climate shifts prompt experimentation with latent resources, as we discussed in Chapter 3.

The spread of agriculture during the Holocene was conducive to technological progress in several ways. First, agriculture created large opportunities for productivity growth through social learning and domestication. Second, the accumulation of technical knowledge was less often disrupted by large climate shocks. Third, agricultural societies were less directly dependent upon nature than foraging societies, although they were still vulnerable to droughts and floods. Eventually manufacturing and service sectors arose that were even further removed from the influence of nature. All of this strengthened the feedback loop between population and technology emphasized in long-run growth theory.

In our view climate change remained important in shaping agricultural societies during the Holocene. For example, it led to city-state formation in certain regions of the world (Chapters 9–11). Galor’s theory avoids reliance on exogeneous shocks of this sort in favor of dynamics that are entirely internal to the system. As a result, Galor sees the transition to modern economic growth as inevitable. We do not regard our prehistoric transitions as having been inevitable, because they were often triggered either directly by exogenous climate shifts or indirectly by the technological and demographic fallout from these shifts. With a different climate history, modern civilization might not exist at all.

Relatedly, Galor has a fixed set of stages through which an economy develops. This is reminiscent of the unilinear theories of social evolution that were once popular among anthropologists and archaeologists (see Chapter 1). Our approach is multilinear because developments in a particular region are conditional on the details of climate and geography for that region, as well as the technological, demographic, and institutional history of the region.

Our agenda differs from that of Galor in other ways. For example, the transitions we want to explain are often institutional in nature. Unified growth theory is not intended to explain property rights, stratification, warfare, urbanization, or taxation. Galor devotes much attention to the sources of inequality across countries in the modern world (Reference Galor2011, ch. 6), and in that context he stresses that country-specific growth trajectories have been shaped by differences in institutions and culture. However, he does not treat institutions as endogenous in his formal modeling.

Galor has a sophisticated treatment of human capital, which in his theory results from educational investments by parents. These investments need not involve formal schooling and could occur through direct parental teaching, where parents face tradeoffs involving the quality and quantity of children as well as parental consumption levels. By contrast, we model “education” as a process in which children learn by imitating all of the adults in a community. This learning process does not have any opportunity cost.

We do not doubt that education played an increasing role in later stages of the Industrial Revolution, or that the substitution from quantity to quality contributed to the Demographic Transition. But we emphasize that foraging societies also make very large investments in human capital as practical knowledge is passed on from one generation to the next (Robson and Kaplan, Reference Robson and Kaplan2006). Even so, technological progress is largely absent in these societies, in contrast to societies following a modern growth path. Our concept of stagnation traps in foraging societies provides an explanation (see Chapter 3).

A final point of contrast is the treatment of consumption. Galor assumes that a parent’s consumption cannot fall below a biological subsistence constraint. His analysis implies that this constraint binds not only throughout the Malthusian epoch but also in the post-Malthusian regime, and only becomes non-binding with modern growth (roughly the last century among the advanced economies). We impose no such subsistence minimum, so we are free to address questions about whether living standards fell in the transitions to sedentism and agriculture, whether commoners became worse off due to stratification, or whether commoners became better off or worse off in early states. This matters because archaeologists have evidence about diet, health, life expectancy, and other indicators of living standards, and a theory of prehistoric economic development should address this evidence. We will return to this topic in Section 12.6 when we discuss the trajectory of human welfare.

12.5 Democratic Institutions

The empirical research reviewed in Section 12.3 indicates that early transitions to state-level organization have had persistent effects on modern economic growth. Borcan et al. (Reference Borcan, Olsson and Putterman2018) suggest that early states encountered diminishing returns because elites used central power to maintain their own privileges at the expense of wider economic growth. This raises a key question: why have some contemporary states evolved in a democratic direction, encouraging sustained economic growth and spreading the gains across much of the population?

Carles Boix (Reference Boix2015) offers one approach to this question. We previously discussed Boix’s views on inequality, warfare, and state formation (see Chapters 6–9 and 11). Here we consider the explanation he proposes for the rise of political democracy in Europe over the last millennium.

From a broad perspective, Boix advances arguments that are consistent with the ideas of other researchers in economic prehistory. He asserts that favorable climate and geography, along with good biological endowments of domesticable plants and animals, caused some regions to have unusually high agricultural productivity. This supported unusually high population densities. Regions of this kind had more towns and cities, which were centers of technological innovation, especially in textiles and metallurgy.

The rise of these proto-industrial centers, although necessary, was not sufficient for sustained growth. Many such places arose in non-European regions. Indeed we note that this description could be applied to some of the early cities described in Chapters 9–11, including Uruk, Mohenjo-Daro, Erlitou, and Teotihuacan. It could also be applied to cities around the year 1500 CE in China (Pomeranz, Reference Pomeranz2000). Hence Boix argues for the importance of other necessary conditions. Our discussion in the next several paragraphs follows Boix (Reference Boix2015, ch. 6).

Boix agrees with Borcan et al. (Reference Borcan, Olsson and Putterman2018) that pre-industrial elites used state power in ways that preserved high levels of inequality and thwarted economic growth. Although republican polities could have encouraged growth, they suffered military disadvantages relative to their larger monarchical neighbors and tended to be short-lived. He attributes the escape from this situation in western Europe to an unprecedented confluence of three factors: emergence of proto-industrial cities, political fragmentation of the continent, and military innovations favoring urban and commercial interests. This enabled commercial centers to capture or partially control the state, yielding parliamentary institutions (204). In Britain and elsewhere, the industrial class was eventually able to buy off the landed elite and implement a liberal political order.

Boix begins his story with the Carolingian dynasty, which relied upon a highly decentralized military system where local overlords were given territory in exchange for rents and military assistance to the emperor. European feudalism continued to evolve in a fragmented way, with many political units exercising varying degrees of sovereignty. Noble and religious authorities often allowed towns to attract population around castles and cathedrals. By the thirteenth century these towns began to assert their autonomy, demolishing castles and establishing institutions of urban governance (222–223).

This trajectory was made possible by the rising military power of urban centers. Urban militias marched in closed columns and used long pikes to defend against cavalry charges. Prosperous cities could support armies of a few thousand foot-soldiers armed with expensive military gear, while the landed aristocracy could not match this power through cavalry. “By the turn of the fourteenth century, town dwellers had succeeded in defeating the aristocracy everywhere in the urban core of Europe” (223).

Boix grants that other parts of the world had proto-industrial cities with clusters of merchants and artisans. However, he argues that these cities were generally seats of state power governed by elite rent-seekers, who were mainly concerned with the extraction of surpluses from agricultural hinterlands. He provides data to show that during 1000–1500 CE the fraction of the total population living in cities grew rapidly in western Europe but did not change in non-European regions.

Within Europe, in areas where agriculture was unusually productive the urban population in 1200 CE was unusually high. Moreover, in areas where urban population was predicted to be high in 1200 CE, there was more proto-industrialization in the form of textile manufacturing (wool, linen, silk) and metallurgy (iron forges) during the period 1200–1500 CE (213–218). The resulting distribution of city sizes in Europe differed from other regions of the world. On the eve of the Industrial Revolution, Europe had a roughly log normal distribution while East Asia, South Asia, and the Middle East tended to have bimodal distributions with a few very large cities and a collection of lesser centers with substantially smaller populations (222–228).

Subsequent military innovations reinforced the power of urban centers in Europe. While the horse had supported mounted knights and the pike helped defend urban areas, firearms were more capital intensive and the side with more economic resources enjoyed an advantage. This forced the existing monarchies to grow in scale or disappear, leading to a drop in the number of independent states. At the same time prosperous commercial areas could survive militarily and resist monarchies. The Dutch cities formed a military alliance that defeated Spain. In England parliamentary forces defeated royal forces in 1640 and 1688. Urban centers could not only finance firearms but also benefited from the close complementarity between trading activities and naval power (228–229). Boix argues that these developments established political institutions in northwestern Europe that were hospitable to economic growth and set the stage for the Industrial Revolution. Ultimately the landed aristocracy began investing in the rising industrial sector and no longer opposed political liberalization.

Like the authors discussed in Sections 12.3 and 12.4, Boix has a highly ambitious intellectual agenda. He seeks to explain the evolution of political institutions from small foraging bands to contemporary states, and provides extensive supporting evidence from archaeology, anthropology, and history. While he is often persuasive empirically, there are important gaps in his theoretical framework.

One limitation is that Boix lacks a clear story about how agriculture began. We provide such a story in Chapter 5. Boix also has an incomplete explanation of why rising productivity in agriculture would lead to an urban non-agricultural sector. There is more to urbanization than having a high population density or a food surplus that can support manufacturing. As we stressed in Chapters 9–11, even if we grant that favorable regions had higher densities for Malthusian reasons, it is necessary to explain why the people in a largely agricultural society would start to agglomerate in urban centers. Possible answers include endogenous property rights that caused wages to fall as agricultural technology improved, and thus made manufacturing profitable; warfare that pushed people to easily defensible locations; or environmental shifts that pushed people toward natural refuges. Nevertheless, Boix contributes valuable insights about the role of innovations in military technology as a determinant of political institutions. In particular he provides a coherent explanation for the triumph of industrial and trading interests over landed interests.

There are, of course, other theories connecting political institutions with economic growth. Many economists argue that the main factor behind economic growth in Europe was the quality of political institutions. This included constitutional checks on the state, the rule of law, and protection of private property, which supported investment and trade. We limit our remarks on this topic to two influential schools of thought: that of Douglass North and his co-authors, and that of Daron Acemoglu and James Robinson.

Douglass North has written extensively on the relationship between institutions and economic performance. This culminated in the book Violence and Social Orders by Douglass C. North, John Joseph Wallis, and Barry R. Weingast (Reference North, Wallis and Weingast2009; hereafter NWW). These writers distinguish three types of political/economic institutions.

The first type is the “foraging order” exemplified by egalitarian hunter-gatherers, although NWW also include “big man” societies and chiefdoms. We have already said a good deal about this subject and will not linger on it here.

The second category is the “limited access order” or “natural state.” Such states are governed by a dominant coalition consisting of elites who agree to respect each other’s privileges, including property rights over resources and activities. Limiting access to these privileges creates rents (returns to holding economic assets that are greater than what the assets could earn in their next best alternative use), which would otherwise be dissipated through competition. The rents accrue to the elite and are allocated in a way that restrains violence among elite groups or factions.

Mature natural states specify legal procedures that constrain interactions between elite groups. Perpetually lived organizations, able to commit credibly to future behavior, expand the possibilities for impersonal relationships and exchanges. Although mature natural states attempt to acquire a monopoly over violence through consolidated control of military resources, they rarely succeed completely. The threat of violence motivates the allocation of rents among elite groups in proportion to their capacities for violence. While violence is therefore limited, so is the potential for economic growth.

The third broad category of political/economic institutions consists of “open access orders.” These states are perpetually lived, combine a monopoly over violence with democratic control over the military, and impartially enforce the rule of law. All citizens can participate in political, economic, educational, and religious organizations, fostering competition in politics and the economy. There are widely held beliefs relating to inclusion and equality for all citizens. Impersonal exchange is common, and political and economic competition discipline the use of state power. Such institutions, in contrast to natural states, promote economic growth through efficient resource allocation, flexible responses to external shocks, and incentives to develop new technologies.

According to NWW, the first natural states emerged around 5000 years ago, with inequality and stratification predating state formation by thousands of years (see Chapters 6 and 9–11 in this volume). They see the transition from a “foraging order” to a “limited access order” as being driven by a number of factors: the development of agriculture, the rise of elites, the increased need to control violence in larger societies, and the integration of economic, military, and ideological sources of power. However, they do not provide a detailed analysis of these processes.

The main focus of NWW is to explain the transition, within the last few hundred years, from natural states to a small number of states characterized by open-access orders. They propose three necessary conditions (called “doorstep conditions”) for the transition: the rule of law for elites, perpetually lived organizations, and consolidated control of the military. However, they do not regard these as jointly sufficient. The principal historical examples they use to illuminate this transition are Britain, France, and the United States.

NWW depart from our treatment of elites in early states as unitary actors. Their emphasis on the potential for violence among elite groups resembles Boix’s emphasis on the conflict between landed and urban/commercial interests. However, Boix regards the transition to political democracy as resulting from a clear victory by urban interests rather than from a balanced accommodation among elite factions.

Another influential approach to the relationship between political institutions and economic growth is that of Daron Acemoglu and James A. Robinson. Here we focus on Acemoglu and Robinson (Reference Acemoglu, Robinson, Eloranta, Golson, Markevich and Wolf2016, Reference Grosman, Munro, Enzel and Bar-Yosef2017, Reference Feder2019; hereafter AR). These authors are concerned with the origins of “the inclusive state,” or more dramatically, “the shackled Leviathan.” For both NWW and AR, the fundamental question is how one can have a powerful state where power is ultimately in the hands of citizens rather than an elite.

AR consider a polity consisting of a smaller elite group and a larger non-elite group (civil society or citizens). These two groups compete for power. The distribution of power is determined by competitive investments made by both groups in determining the capacity of the state to enforce laws, provide public services, and tax and regulate economic activities. Depending on initial conditions, and assuming scale economies in political investments, three steady-state outcomes can arise: a weak state (Absent Leviathan), in which the state has little power to regulate economic or social activities; a despotic state (Despotic Leviathan), in which elites impose order through oppression; or an inclusive state (Shackled Leviathan), in which states have immense powers but are constrained by their citizens through a broad distribution of political power, as in modern democracies. The concept of a despotic state overlaps with a natural state in NWW, and the concept of an inclusive state overlaps with open access orders in NWW.

The pathway to an inclusive state requires a rough balance of power between elites and non-elites that must be maintained through continuous investment on each side. If power is approximately balanced, citizens will allow the state greater scope because they expect to retain some control over it. But if both groups have only modest power, even small initial differences can result in convergence to one of the other steady states. This leads to the imagery of a “narrow corridor” in navigating to an inclusive state.

While initial conditions are exogenous, the outcome is not predetermined. AR (2019, 435–438) provide examples where changes in initial conditions redirected state trajectories. A particularly clear case is that of Japan after World War II, where the previous despotic state was disbanded after its military defeat. To alter initial conditions that gave rise to despotism, the United States imposed total demilitarization, demanded that the emperor publicly renounce claims to divinity, and worked with senior members of the military and bureaucracy to transform the structure of the government and society. This allowed an inclusive state to emerge.

Perhaps the most fundamental difference between NWW and AR is that NWW start with a society characterized by a non-homogeneous elite, where elite factions are continuously interacting and often in conflict. AR assume the elite is a unified wealth-maximizing group.

What can prehistory add to this discussion? First, it provides the building blocks underlying these analyses: e.g., our explanations for the transitions to sedentism, agriculture, inequality, warfare, cities, and states. Second, we (and many others) view the process of early state formation through the lens of public finance. Our key question is how elites gained the power to tax, which is central to the formation of powerful states like those described by Boix, NWW, and AR.

How such states came under democratic control is a different but related question. We are sympathetic to the view of Boix that specific military technologies, along with other background conditions, played a key role in strengthening urban and commercial interests. However, it is not entirely obvious why the economic elites in industrializing societies would grant broader control over state decisions to citizens in general, and there are examples like Japan where this did not occur. We are also sympathetic to the view of NWW that elites are often internally factionalized, and the view of AR that democracy results from investments by non-elites in the development of political power. However, these arguments strike us as incomplete.

What appears to be needed is an explanation of how non-elites can (sometimes) overcome the collective action problems they face in contests against entrenched elites. As we observed in Chapter 11, elites typically have advantages of organization, wealth, leadership, and coercive power that more than compensate for their smaller numbers. In such regimes, incipient rebellions or subversive movements are easily suppressed and the ringleaders usually pay a heavy price. Under these conditions few leaders step forward.

However, technological and institutional innovations may present commoners with opportunities to organize while avoiding immediate suppression. This can include novel developments with respect to communication, transportation, weaponry, urbanization, or education, for example. As collective action problems dissipate and commoners become more capable of making demands on elites, backed up by threats to do serious physical or economic damage to elite interests, the elite (or a subset of it) may find that compromise is cheaper than violence. When this occurs, institutions must be negotiated to ensure the credibility of commitments to share power. Political democracy is one such institution (Acemoglu and Robinson, Reference Acemoglu and Robinson2006).

In a heavily agricultural economy, taxes on land and output may be sufficient to finance the state. Taxes are collected through confiscation and policies are imposed by edict. There is no need to include peasant farmers in the governing process (e.g., ancient Egypt). Economic growth, however, can result in the accumulation of wealth in urban areas through trade, manufacturing, and services. At the same time large administrative and military expenses may be required for state survival, especially in competitive state environments. Under such conditions, offering property rights, open access to markets, and voting rights in return for tax revenue and acceptance of state policies can become a preferred option for elites (see North and Thomas, Reference North and Thomas1973, for an early analysis of the transactional state). In our view factors of this kind help explain why industrialization was accompanied by democratization in Western Europe.

Voting rights are particularly important when a credible commitment on the part of the state is required. This is widely accepted for zero-sum games involving income redistribution. But NWW and AR move beyond a zero-sum setting by pointing out that democratic constraints shift state priorities toward the provision of public goods. NWW (2009, 142–143) emphasize that open-access orders encourage state investments in mass education, infrastructure, and social insurance programs that facilitate economic growth while lowering individual risks. AR (2019, 72–73) make the related point that when citizens gain some control over the state, the state can be given a long leash and can become an instrument for the political, economic, and social development of the society as a whole. These are lofty hopes. On the other hand, NWW and AR agree that “open-access orders” and “inclusive states” are rare and fragile.

12.6 The Trajectory of Human Welfare

People in rich countries today have enjoyed a century or two of exponential growth in per capita income. This trend has occasionally been interrupted by wars, pandemics, or recessions, and the benefits have been distributed very unequally. But averaging over the decades, economic growth has prevailed.

For this reason it is tempting to extrapolate backward and assume that those who lived hundreds or thousands of years ago were much worse off. In this way of thinking, the mobile foragers of the Upper Paleolithic must have been the least well-off people of all, because they are the furthest in the past. One needs to avoid such thinking. The last two centuries have no precedent in human history or prehistory, and the contours of human welfare were very different in the pre-industrial world.

However, if one wants to say something more nuanced about the path of human welfare over thousands of years, there are some obvious pitfalls. First, it is unclear how one should even think about welfare comparisons over such long time spans. Second, if we can agree on what welfare means, where can we find data about it? And third, how should we deal with inequalities across the individuals or classes making up a society?

We will address each of these issues below. Welfare comparisons across spans of millennia, and across enormously different societies, are quite difficult and must be made cautiously. Nevertheless, relevant data do exist, and the theory we have developed in this book provides some guidance. To cut to the chase, we think human welfare has generally followed a U-shaped curve for the last 15,000 years. Taking mobile hunter-gatherers as a starting point, life appears to have become worse in the transition to sedentism and again in the transition to agriculture. For the majority of the population, the standard of living probably bottomed out in early states (e.g., ancient Egypt), and did not begin to recover in any substantial way until the last two centuries.

The first question involves what we mean by the concept of human welfare. It is clear that if we define welfare in terms of subjective happiness, we will not get very far. Harari (Reference Harari2014, 376) asks, “Was the late Neil Armstrong, whose footprint remains intact on the windless moon, happier than the nameless hunter-gatherer who 30,000 years ago left her handprint on a wall in Chauvet Cave? If not, what was the point of developing agriculture, cities, writing, coinage, empires, science and industry?” This is a rhetorical question and we cannot know the answer.

Modern populations are frequently asked questions like whether they agree that “life is good” on a scale from zero to ten. The main findings are that happiness is linked with income or wealth only for people at the low end of the economic scale. Illness can cause short-term distress, but this does not generally extend to the long term unless the person is in constant pain. Family and community matter more, and happiness may be affected mostly by conditions relative to one’s expectations (Harari, Reference Harari2014, 380–384).

Another way to think about happiness involves biology and chemistry. From this perspective a sense of happiness is generated by neurotransmitters: serotonin, dopamine, and oxytocin. The average levels of these brain chemicals have probably changed little in recent millennia because they are biologically programmed. Levels may differ across individuals, but it seems unlikely that happiness in this sense would rise or fall for entire societies (Harari, Reference Harari2014, 385–390).

Perhaps what matters the most is a long-term sense of meaning or significance in one’s life, rather than the day-to-day ups and downs of happiness (Harari, Reference Harari2014, 390–396). Harari also points to the view taken by Stoics and Buddhists that unhappiness, suffering, and pain are caused by the fact that people want things, and that if they stopped wanting things, this pain would disappear. Maybe so, but physical and mental constraints appear to be important for the vast majority of people.

Whatever one believes about these proposals for conceptualizing human welfare, they all pose a fundamental problem for the economic prehistorian: lack of data. We do not have questionnaire results, neurotransmitter readings, or estimates of meaningfulness for people who lived thousands of years ago. We cannot know whether a laborer under the hot sun in the agricultural fields around Uruk had a deeply meaningful life due to the proximity of a temple, or was deeply depressed by the authoritarianism of the local elite.

Archaeologists do have data on factors that almost everyone today would consider relevant for human welfare, including nutrition, health, security, and life expectancy. The United Nations and similar bodies often base international comparisons of the quality of life on measures of this kind, although people can and do disagree about the weight that should be attached to each factor. Even if these welfare measures only reflect the preferences of contemporary people, we might want to use archaeological data to assess the quality of life in the past from the standpoint of our present values.

Furthermore, although we cannot read the minds of people from the distant past, it would be strange to argue that people 15,000 years ago did not care about such things. Evolutionary logic suggests that nature would long ago have programmed the members of our species to care about exactly these things. People who are indifferent toward nutrition, health, and security are unlikely to live very long and are unlikely to raise children to adulthood.

In principle, criteria like these can be used for welfare comparisons across diverse societies and over long time intervals. But one might still ask, what is the point? Why make welfare comparisons over 15,000 years? One answer is simple curiosity. Another is that this exercise may lead us to rethink casual assumptions about the inevitability of human progress. A third motive is to better understand social evolution. For example, if important prehistoric transitions made people worse off, how is this consistent with the usual view (at least in economics) that people try to make themselves better off whenever they can?

A complicating factor involves the role of inequality. As we discussed in earlier chapters, mobile foragers are usually quite egalitarian, apart from minor distinctions involving gender and age. Sedentary foragers often have significant inequality, and this increases with the transition to agriculture. When there are both privileged elites and impoverished commoners, how can we aggregate individual welfare to obtain an index of social welfare? Should we even try?

The political philosopher John Rawls (Reference Rawls1971) suggested that social welfare should be evaluated based on the condition of the least well off members of society. We will not give any detailed philosophical justification for this view, but it is a useful way to address the issue. When a society has elite and commoner classes, we will concern ourselves with the welfare of the commoners. This is partly because commoners will typically be the overwhelming majority of the population (probably 80–90%, even if one defines the elite to include skilled artisans and warriors). Also, commoner living standards are more problematic (no one seems very concerned about whether living standards for the elite rose or fell). If suitable data existed, we might want to study heterogeneity within the commoner class, but drawing fine distinctions of this sort is usually infeasible. For agricultural societies, in effect we will be tracking the welfare of agricultural laborers.

Of course, if there are both free commoners and slaves, the Rawlsian framework implies a focus on the welfare of the slaves. However, for many societies in prehistory it is unclear whether or not slavery existed. Even when it clearly did, we cannot necessarily obtain data that quantify the differences in living standards between these two classes.

Another complicating factor is population. In a Malthusian world, technological progress is channeled toward population growth rather than improved standards of living per person. One could conceivably argue that population growth is desirable for its own sake. However, our concern is with the living standards of individuals (both the average and the distribution). We are not Benthamites who want to maximize the sum total of human happiness. Thus we do not award a society more points for having a larger population.

In the rest of this section we review what our theory says about human welfare and then discuss some evidence. We start with the simplest case: an egalitarian society using a fixed technology and fixed natural resources to produce food. Due to diminishing returns the average product of labor (y) decreases as aggregate population (N) increases. The average product is equivalent to food per capita, an obvious candidate for a welfare index. Alternatively one could use surviving children per adult as a welfare index. This is correlated with food per capita and can be regarded as a biological objective function.

If we compare (y, N) points along the average product curve, our unwillingness to assign social value to population implies a preference for points having higher levels of y and lower levels of N. Perhaps counter-intuitively, this implies that N should ideally be the smallest number exceeding zero. But in reality there may be reasons to want a larger group size that are omitted from simple versions of the model. Starting from a low initial level, higher population may result in higher food per person due to scale economies with respect to labor specialization or insurance, the supply of public goods such as defense, or other positive externalities among group members.

Malthusian theory says that over multiple generations food per capita will settle at a constant level y* (see Chapter 2). This long-run standard of living will not normally be affected by changes in climate or technology. Recall that we do not interpret this level of food as a biological subsistence minimum, or a point below which starvation occurs. It is simply the level at which population is stationary, neither rising nor falling, because fertility and mortality rates are in balance. Whether the standard of living happens to be high or low in such a steady state is a separate question.

Our principle that aggregate population is irrelevant for social welfare means that we are not concerned with the particular population level occurring in a steady state. But such a population could be supported either by high fertility and high mortality or by low fertility and low mortality. Other things being equal, most people would probably prefer the latter. For the moment we ignore this issue and focus on another theoretical problem.

If we embrace food per capita as a welfare measure, if the world was Malthusian before the Industrial Revolution, and if food per capita (y*) was therefore constant in the long run, this implies that social welfare was likewise constant. A similar problem arises if we treat surviving children per adult as a welfare measure. In long-run equilibrium this must be unity (on average, one adult is replaced by one surviving child). Again we get a constant.

However, y* can change when large transformations in technology alter the fertility and mortality rates associated with a given level of food per person. In Chapter 4 we argued that the transition from mobile to sedentary foraging decreased y*. We constructed a simple model in Section 4.10 where parents maximized their expected number of surviving offspring, a reasonable objective from the standpoint of biological evolution, and confronted tradeoffs between the quantity and quality of children. We defined “quality” to be the probability that a child survived to adulthood.

There is a consensus among anthropologists that sedentism lowered the cost of a child because mothers had to spend less time carrying young children. We showed that as a result parents would substitute toward quantity rather than quality among children. This decreases y* in equilibrium. Higher quantity implies more births over a woman’s reproductive career while lower quality reduces childhood survival rates. The resulting equilibrium is supported by higher fertility and higher mortality.

Our model assumed that adults live for one period but a more sophisticated model could have mortality increasing for adults due to the lower food per capita. Even so, it is likely that in reality most of the increased mortality would fall upon children. One must therefore be cautious about estimates of life expectancy. Sedentism could cause a lower life expectancy at birth due to higher child mortality, while having little effect on adult life expectancy (for example, conditional on survival to the age of 15).

We used the same general reasoning in Chapter 5 to argue that the transition from sedentary foraging to agriculture likely decreased y* further. It is generally believed that agricultural technology makes children more productive relative to foraging technology. This reduces the economic cost of children and again tends to lower the equilibrium level of y* in the long run, as well as childhood survival and life expectancy from birth. Thus, our theory leads us to believe that both sedentism and agriculture had negative effects on social welfare, even when societies remained egalitarian.

As these arguments show, it is overly simplistic to assume that Malthusian forces always imply a constant standard of living in the long run. This is true, or at least a good first approximation, when climate fluctuations or technological innovations do not disturb the underlying relationship between food and demography. However, large technological transformations can shift this relationship through the quantity/quality channel, much as the Industrial Revolution led to the Demographic Transition (see Section 12.4).

Another complication for the Malthusian approach involves inequality. The idea that aggregate population must be stationary in the long run implies some equilibrium y*. But this is only the food per capita for the population as a whole. In stratified societies the elite’s food consumption will be above y* and the commoners’ food consumption will be below it. Because we do not employ the notion of a biological subsistence minimum, this need not imply that the commoners will starve to death. However, it does imply that commoners fail to replace themselves demographically (on average, each commoner has less than one surviving adult child). This means that in each generation the commoner population must be replenished by downward mobility where some children of the elite become commoners. This flow of surplus elite children into the commoner class is just enough to compensate for the rate of natural decrease among commoners themselves.

Chapter 6 showed that when property rights to land are endogenous, technological progress tends to impoverish commoners by raising the regional population and enabling insiders to close more sites to outsiders. At the same time elites at the best sites become better off because they enjoy larger land rents. In our model those who remain behind in the (shrinking) commons and those hired by elites have the same food consumption, and are worse off than insiders or elites. By our Rawlsian criterion, social welfare continues to decline as agricultural technology improves.

Chapter 10 complicated the story by adding an urban manufacturing sector, which became active when the commoner standard of living became low enough. We showed that with a fixed regional population, commoners became worse off in the early stages of city-state formation. However, it was uncertain whether this trend would continue. For example, commoners might eventually benefit from consumer surplus associated with manufacturing, from higher wages as manufacturing expanded, or from public goods provided by elites.

Further complications arise from the endogeneity of population in the long run. In previous settings this led to an equilibrium food per capita y*. When both food production and craft production occur, the model has to be generalized to yield a long-run equilibrium utility u*. As with y*, this is only an equilibrium in an aggregate sense, and inequality implies that elites get more utility per person than u* while commoners get less. We therefore need to know how the evolution of manufacturing and taxation affect elite–commoner inequality before reaching any judgment about commoner welfare and (via Rawls) a judgment about social welfare. The issue is complex, but commoner standards of living may have finally stabilized as the early city-states matured.

Much has been made of the possibility that commoners can benefit from state provision of security, insurance, infrastructure, and other public goods. However, in the long run average utility is likely to remain close to the level u* consistent with a stationary aggregate population. Malthusian dynamics will guarantee a zero-sum game where the commoner class only becomes better off when the elite class becomes worse off. Thus, commoner welfare in early states depends on whatever social, economic, political, and military forces determine the overall degree of inequality within the society.

Given the difficulty of measuring utility, we suggest using the average number of surviving offspring per commoner as an index of commoner welfare. Surviving children per adult is only required to equal unity at an aggregate level when elites and commoners are averaged together. This figure will be below unity among the commoners, and when it is closer to unity commoners have a higher standard of living. Such a welfare measure is attractive because it captures various factors such as child mortality, nutrition for adults and children, health status, physical security, and so on, insofar as these contribute to the reproductive success of commoners. One need not worry about how to assign weights to each factor because nature weights them automatically. Of course it may be difficult to obtain demographic data of this kind through archaeological methods. If so, it will be necessary to rely on narrower measures like nutrition, health, and life expectancy.

In the rest of this section we briefly review evidence about the welfare trajectory suggested by our theory. Systematic archaeological research on prehistoric standards of living developed in the 1970s and 1980s, motivated by a then-popular hypothesis about the origins of agriculture. According to this hypothesis, exogenous population pressure among Upper Paleolithic foraging societies led to declining living standards, triggering a transition to agriculture (Cohen, Reference Cohen1977). We hasten to add that this view is not consistent with the Malthusian framework used in this book. Instead we would expect a decline in living standards to generate feedback effects that lower fertility and raise mortality until foraging societies reach a stationary demographic equilibrium (see Chapters 2 and 3).

It soon became clear that there was little or no evidence for a food crisis among foragers in the Upper Paleolithic or Epi-Paleolithic, but there was strong evidence that nutrition and health declined among early farmers (Cohen and Armelagos, 1984, Reference Cohen and Armelagos2013). There is now a consensus that relative to their foraging ancestors, early farmers had less varied and nutritious diets, shorter adult stature (a common proxy for a poor diet early in life), more infectious disease, more repetitive work, and shorter life spans (see Cohen, Reference Cohen1991, Reference Bowles2009; Lambert, Reference Lambert2009; Scott, Reference Scott2017, 96–113; and the references cited there).

We do not want to romanticize the small foraging bands of the Upper Paleolithic. People in these bands probably had relatively short lives, faced occasional starvation, and had frequent individual homicides. However, it seems clear that early farmers had living standards that were measurably worse according to standard archaeological indicators.

The archaeological literature does not always distinguish between effects of the transition from mobile to sedentary foraging (Chapter 4) and effects of the transition from sedentary foraging to agriculture (Chapter 5). There is good evidence that infectious disease was more prevalent in larger settlements, a problem that would have affected both sedentary foragers and farmers, although with greater impact on the latter. There is also strong evidence that proximity to domesticated animals promoted infectious disease (Diamond, Reference Diamond1997), a problem limited to farmers and pastoralists as opposed to foragers.

Another area of ambiguity involves the causal channels leading to disease. A poorer diet tends to reduce immune function, weakening the body’s defenses against infectious disease. At the same time sedentary foragers and early farmers had greater exposure to pathogens from living in larger communities and (in the case of farmers) having domestic animals. It is frequently unclear to what extent the skeletal evidence for disease reflects poor diet and to what extent it reflects high pathogen exposure.

We have already documented the increasing inequality that arose in agricultural societies with the onset of stratification between elites and commoners (see Chapter 6). In a Malthusian world this almost certainly implied a further decline in living standards for commoners (Cohen, Reference Cohen1991, ch. 7). Although raiding among foraging bands probably has deep roots, we provided evidence in Chapters 7 and 8 that warfare over land became more common in sedentary foraging societies and chronic in agricultural societies with stratification. Our formal model in Chapter 8 omitted any costs of elite warfare for the commoner population, but in practice it appears likely that increasing stratification and increasing warfare would both have made commoners increasingly miserable.

Early cities like Uruk and Teotihuacan had high mortality rates (Cohen, Reference Cohen1991, ch. 7; Algaze, Reference Algaze2018, 26), and this likely remained true for most cities until clean water supplies and modern sanitation systems became available in the nineteenth century. On the other hand, urban populations were continually replenished through migration from the countryside, suggesting that rural life may have been even worse, or at least that migrants to the city were less risk averse than those who stayed behind.

We have discussed the possible effects of state formation on commoner welfare in Chapters 9–11 and will not repeat those remarks here. Empirical research by Ashraf and Galor (Reference Ashraf and Galor2011) confirms the Malthusian expectation that technological innovation and land productivity had no significant effect on per capita incomes in the years 1 CE, 1000 CE, and 1500 CE (see Section 12.4), so commoners would probably not have benefited from aggregate economic growth in the state societies of the time.

Commoners did occasionally benefit from changes in economic conditions. For example, real wages rose for a century after the Black Death epidemic of 1347–1351 in England (Clark, Reference Clark2005). But putting this exception aside, economic historians find that inequality of income and wealth tended to grow monotonically from 1450/1500 CE until 1800 CE in almost all of Europe (Alfani, Reference Alfani2021). Possible explanations include economic growth, population growth, proletarianization, urbanization, and inheritance systems. However, Alfani (2021) puts special emphasis on state building, warfare, the resulting heavy fiscal burden per capita, and regressive taxation.

The Industrial Revolution and the Demographic Transition changed the picture drastically. For the advanced economies, Boix (Reference Boix2015, 202) cites evidence that between 1820 and 2008, per capita income rose by a factor of 20, average height increased by 10–12 centimeters, and life expectancy almost doubled due largely to reduced infant mortality and the elimination of infectious diseases. Rates of childhood mortality (death by the age of 15) dropped from nearly 50% in pre-industrial societies to 1% or less in modern developed countries (Volk and Atkinson, Reference Volk and Atkinson2013).

This has also been a period of rapidly expanding education and, at least during 1900–1970, declining inequality. Using data on sovereign countries at 10-year intervals since 1820, Boix (Reference Boix2015, 232–242) finds that the growth of per capita income was strongly correlated with greater democracy, and that the causality ran from income to democracy. He also argues that greater democracy led to greater equality through public education, health expenditures, public pensions, and unemployment benefits.

We believe that the Industrial Revolution, the Demographic Transition, and the expansion of political democracy were each necessary for welfare to improve. Without the Industrial Revolution, there would have been no technological basis for a substantial welfare gain. Without the Demographic Transition, the productivity boost derived from industrialization would have been absorbed by Malthusian population growth rather than translating into higher income per capita. Without political democracy, elites would have captured a disproportionate share of the higher average income.

These three necessary conditions were almost certainly linked. For example, one could argue that although uneducated workers were employed in the early stages of the Industrial Revolution, subsequent stages stimulated a demand for education (see Section 12.4), which led to a demographic transition as well as pressure for political democracy. Alternatively, one could attribute the spread of political democracy to other features of industrialization such as the decreased cost of communication and transportation, which facilitated the formation of labor unions, civic organizations, and mass political parties. But in any event these three factors together were sufficient to bring about an enormous gain in the quality of life for most people in developed nations.

In sum, the evidence supports our view that material welfare for the majority of the population has followed a long U-shaped curve, with living standards deteriorating in the transition from foraging to agriculture, staying low for millennia due to stratification and Malthusian population dynamics, and then exhibiting an extraordinary jump during the last two centuries due to the combined effects of the Industrial Revolution, the Demographic Transition, and the expansion of political democracy.

We close this section where we began: it is very difficult to make meaningful welfare comparisons between societies separated by vast gulfs of time and space. We don’t really know much about the quality of life for the foragers of the Upper Paleolithic. Even within the modern world it is difficult to compare the welfare of the few remaining foragers with the welfare of people who live in cities and work in offices. Mobile hunter-gatherers who make peaceful contact with the modern economy seem to like metal tools, radios, firearms, running water, modern clothing, and modern medicine. On the other hand, very few people living in industrial societies have the technical or social skills to join hunter-gatherer societies (only rare anthropologists can do this), and we don’t know how many modern people would want to become foragers if they could. Perhaps many among us would be happier and healthier as foragers. In the opinion of the authors, the two best places to be born were probably (a) in a foraging group that had just discovered a new continent or (b) in a middle- or upper-class community located in North America in the mid-twentieth century. Most of the other options seem less attractive.

12.7 Climate: Past, Present, and Future

Whether the exponential income growth of the last two centuries will prove to be sustainable over the next two is an open question. Technological advances continue, but there are warning signs. Some poor countries remain very poor, inequality is growing in the rich countries, and democracy looks more fragile than it did a few decades ago. There are justified worries about overpopulation, resource depletion, and pandemics. But almost certainly the greatest threat to our global civilization is climate change.

Most readers are probably well aware of the climate change problem, but we will summarize a few basic points. Several gases in the Earth’s atmosphere, including carbon dioxide, methane, and water vapor, allow light from the sun to pass through but trap heat that would otherwise radiate away into space. These so-called greenhouse gases have natural sources such as volcanoes and are needed to keep the planet warm enough for life to exist. But human activities since the Industrial Revolution, especially the burning of fossil fuels (coal, oil, and natural gas), have increased the concentration of carbon dioxide and other greenhouse gases in the atmosphere. Deforestation has also played a role.

Carbon dioxide (CO₂) emissions are the largest single driver of global warming. The pre-industrial atmospheric concentration of carbon dioxide in the year 1820 was 280 parts per million (ppm). At this writing it has risen above 410 ppm for a 46% increase in two centuries. The last time the CO₂ concentration exceeded 400 ppm was during the Pliocene, 3–5 million years ago (NASA, Reference Allen, Berazzini and Heldring2020). At that time global temperatures were 3–4 °C warmer and the sea level was 5–40 meters higher. No member of the genus Homo had yet walked the Earth.

The planet is now far from an equilibrium state, and substantial temperature rises will occur as oceanic and atmospheric systems catch up with the new CO₂ concentration. We lack a technology for removing large quantities of CO₂ from the atmosphere, and the natural processes that do so operate on time scales of thousands of years. Therefore any human-produced increment in the CO₂ concentration is essentially permanent.

The Intergovernmental Panel on Climate Change (IPCC) produces the most authoritative forecasts of the effects of climate change. At this writing the most recent synthesis report is IPCC (Reference Pachauri and Meyer2014) and the next synthesis report is scheduled for late 2022 or early 2023. More specialized reports are published at more frequent intervals. The current and foreseeable consequences of global warming include more intense heat waves, droughts, floods, and storms, and higher sea levels. The severity of these consequences will depend on the future emissions path for greenhouse gases, which can be affected by technological innovation and public policy. For example, innovations that bring down the cost of renewable energy sources (such as solar, wind, and hydroelectric) would help, as would economic incentives for people to substitute away from fossil fuels (such as carbon taxes and cap-and-trade systems). It would also help to have policies that prevent and reverse deforestation, because trees are an important mechanism for removing carbon from the atmosphere. But these measures will only limit the scale of the damage. Past emissions guarantee that future warming is inevitable, and large investments will be needed to adapt to new climate conditions.

Global mean temperature has already increased by about 1 °C relative to the level prevailing prior to the Industrial Revolution. The Paris climate accord (UNFCCC, 2015) set the goal of keeping the global temperature rise during this century to no more than 2 °C (that is, one additional degree) relative to the pre-industrial level. This limit was set in part because many scientists believe that the world climate system could pass a series of irreversible tipping points beyond that. The accord also commits signatories to make an attempt to keep the rise to 1.5° (one half of a degree beyond what has already occurred). Because CO₂ emissions are permanent, this implies a “carbon budget” for the world that puts an upper bound on the additional emissions that can be tolerated without exceeding these temperature targets.

Most climate researchers believe that the existing national commitments to limit greenhouse gas emissions under the Paris accord are inadequate to hold the line at 2 °C even if they are honored, and there is much room for doubt about whether they will be honored. Accordingly, modelers often study scenarios involving increases of 3–4 °C by the end of this century relative to the pre-industrial baseline. For example, Raftery et al. (Reference Raftery, Zimmer, Frierson, Startz and Liu2017) suggest that the likely range is 2.0–4.9 °C with a median of 3.2 °C. They find a 5% chance of remaining below 2 °C and a 1% chance of remaining below 1.5 °C, taking into account uncertainties in future population growth, economic growth, carbon intensity per dollar of GDP, and climate sensitivity to carbon emissions. They do not account for the possibilities of dramatic technological breakthroughs or unforeseen natural disasters. Although the baseline is determined by pre-industrial conditions, most of the temperature increase will occur in the twenty-first century.

Aside from uncertainty about the effect of a given emissions trajectory on global mean temperature, it is difficult to forecast the effects of global mean temperature on sea level, agricultural productivity, and other crucial outcomes. It is also difficult to predict how carbon emissions will influence temperature, precipitation, and similar variables at the regional level (except for very large-scale generalizations, such as the fact that polar regions are warming much more rapidly than lower latitudes and will continue to do so). The regional effects will almost certainly be distributed in a very uneven manner.

Climate history provides a sense of scale. During the Last Glacial Maximum of 21,000 years ago, ice sheets spread as far south as London and New York, and sea level was 125 meters lower than at present. People could walk from Siberia to Alaska. The rise in global temperature to Holocene levels shifted the planet from an intense Ice Age with very cold, very dry, and highly variable conditions to the warm, wet, and relatively stable conditions of the last 11,600 years.

According to Antarctic ice core data from Petit et al. (Reference Petit1999), at the Last Glacial Maximum the temperature was about 8 °C below modern levels (see Figure 3.1). But the polar regions were subject to greater cooling, and one estimate of global mean surface air temperature places the LGM at about 6 °C below the pre-industrial level (Schneider von Deimling et al., Reference Schneider von Deimling, Ganopolski, Held and Rahmstorf2006). More recent work has produced a similar estimate of 6.1 °C with a 95% confidence interval of 5.7–6.5 (Tierney et al., Reference Tierney, Zhu, King, Malevich, Hakim and Poulsen2020).

If we regard an increase of 2 °C by the year 2100 relative to pre-industrial levels as the best reasonably plausible scenario, and we believe that the recovery from the LGM involved an increase of 6 °C, then we are looking at an increase roughly 33% of the size of the jump from the Last Glacial Maximum to the present. The situation could easily be worse. If the true extent of global warming will be 4 °C, the change we are confronting will be about 66% of the transition from the worst conditions of the last glacial period to our present interglacial period. The jump from the LGM to the Holocene required 10,000 years while we are facing a substantial fraction of such a jump within a single century.

Much effort has been devoted to examining the effects of weather on economic variables (for a thorough survey, see Dell et al., Reference Dell, Jones and Olken2014). This literature studies impacts on agricultural output, industrial output, labor productivity, energy demand, health, conflict, and economic growth. Others have investigated the effects of climate on individual and group violence (Burke et al., Reference Burke, Hsiang and Miguel2015). Although the effects in such studies are sometimes statistically significant and quantitatively large, the datasets (with a few exceptions) are mostly based on information since 1900 or 1950. We are currently facing rapid climate changes far outside this historical range.

Other researchers have explored the social effects of weather on millennial time scales. We mentioned a few examples in Chapter 7. Zhang et al. (Reference Zhang, Jim, Lin, He, Wang and Lee2006, Reference Garfinkel, Skaperdas, Hartley and Sandler2007) found a correlation between abnormal cold and domestic rebellions in China during 1000–1911 CE. Bai and Kung (Reference Bai and Kung2011) found a correlation between droughts and nomadic invasions of China during 220 BCE–1839 CE. Tol and Wagner (Reference Tol and Wagner2010) discovered a correlation between cold periods and violent conflict in Europe during 1000–1990 CE. In each case, the violence almost certainly resulted from large reductions in food supplies. At a more impressionistic level it has been argued that the Little Ice Age of 1600–1700 CE, which may have involved a decline in global mean temperature of about 1.0 °C, led to warfare, rebellions, and social collapse in many parts of the world (Parker, Reference Parker2013). But again the climate shifts of our century will be substantially larger.

If we want to grasp the social effects of truly massive climate change, we need to study prehistory. We have argued that climate change was a driving force behind several major economic transformations. The recovery from the Last Glacial Maximum and the onset of the Holocene led to a transition from mobile to sedentary foraging in many parts of the world (Chapter 4). The large negative shock of the Younger Dryas resulted in a transition from sedentary foraging to cultivation in southwest Asia and perhaps in China, with later climate shifts probably playing a role in Africa and elsewhere (Chapter 5). A shift toward increased aridity around 6000–5000 BP is thought to have triggered the rise of cities and states in Mesopotamia and Egypt, and similar mechanisms may have led to later city-states in the Indus Valley and northern China (Chapters 9–11).

Some of these events may not seem very dramatic from our own vantage point. For example, communities of sedentary foragers or early farmers typically had at most a few hundred people. But village life was a radical change from life in mobile foraging bands. It involved a commitment to use resources within walking distance of a central place, a vulnerability to natural shocks that could temporarily take away these resources, investments in stationary tools and housing, more reliance on storage to manage risks, a greater tendency to marry within one’s own community rather than outside it, stronger barriers to migration across communities, and new types of property rights. Inequality and warfare over land were not far behind.

Most of us would probably recognize the formation of early city-states as a more dramatic transition. Such cities, often with as many as 50–100,000 residents, had diverse craft manufacturing activities, monumental architecture, and elites wielding state power. The rulers taxed their citizens, frequently fought wars against rivals, and eventually built regional empires. Most early states developed systems of writing.

We are not trying to portray these developments as part of a rising tide of human progress. As we argued in Section 12.6, these transitions generally had negative impacts on human welfare, at least for the vast majority. Our point here is simply that exogenous climate shifts have had revolutionary consequences for social, economic, and political life. It is reasonable to ask whether equally large consequences should be expected now that climate has become endogenous.

First, however, we need to recognize some key differences between the world of prehistory and the world of today. Most obviously the modern world has vast technological capabilities, can generate new knowledge rapidly, and can easily transfer knowledge across space and time. Another contrast is the absence of a Malthusian population trap. In the modern world, when people become better off they generally have fewer children rather than more. Thus our long-run models of technology and population have little relevance.

On the other hand, one can argue that these differences do not matter much. Our models of technological innovation and population dynamics were designed for societies evolving over many human generations. For most of the applications we had in mind, a century was a short interval, often within the margin of error for archaeological dating. But modern climate changes will have large impacts over decades. From our modeling perspective this is the short run.

In our view the two main prehistoric analogies to the present are the rapid onset of the Younger Dryas, which may have triggered cultivation in southwest Asia on a scale of decades, and the emergence of early city-states, which in some regions may have taken less than a century (although for southern Mesopotamia and other cases, dates are highly uncertain). For these transitions we argued that the short-run effects of climate change largely involved migration.

In the Younger Dryas some previously sedentary foragers reverted to a nomadic lifestyle while others moved to a few refuge locations like Abu Hureya. The resulting local population spikes at these climate sanctuaries depressed the marginal product of labor in foraging and made cultivation attractive. We do not see cultivation as a technological breakthrough. Instead it involved a resort to a known backstop technology, previously regarded as unattractive, under difficult natural conditions. Thousands of years later, cultivation led to domestication and sustained technological progress, but this provided no comfort to the inhabitants of Abu Hureyra during the centuries of the Younger Dryas.

Similarly we believe increasing aridity during 6000–5000 BP caused migration away from areas of southwest Asia and North Africa dependent on rainfall agriculture. Again people looked for environmental refuges and found them near river valleys. The short-run effect of this climate shift, according to the model in Chapter 10, was to drive down wages and re-allocate commoner labor toward elite agricultural estates in southern Mesopotamia and the Nile Valley. Once wages had declined enough, manufacturing and taxation became profitable for the elites. These developments had enormous technological and institutional consequences in later millennia, but if the commoners of the time had somehow known this, it would have offered them little solace.

Another short-run effect from climate change involves warfare. In Chapter 7 we argued that even in societies that are internally egalitarian, climate shocks can destabilize relationships between population and productivity across sites and trigger group conflict. This is particularly likely when social barriers make it difficult for individuals to move in response to these shocks. Similar mechanisms are likely to trigger warfare in stratified societies where elites compete over the rents from land and other natural resources. Such societies are prone to military conflict, and climate shocks could shift resources in ways that make aggression more tempting, despite defensive technology or institutional restraints on war.

This, then, may be our future. Climate change will redistribute natural resources around the world, making many places poorer and a few places richer. Individuals will flee from areas that become hotter, drier, stormier, or flooded by seawater. They will look for refuge in places with milder temperatures, reliable rainfall, usable farmland, and higher elevations. Insiders who already occupy these places will try to exclude the refugees, put them to work for low wages, charge high entry fees, or screen out those who lack special skills. Inequality and warfare will increase and elites who can keep control over valuable territories will prosper. People will adopt available technologies that were not previously attractive but become necessary for survival. Political and social institutions, and technological systems requiring the support of these institutions, could crumble under the weight of migration, poverty, and violence.

People who have not read this book might make similar forecasts. We do not claim any special ability to predict the future. But economic prehistory has value in thinking about the fate of the modern world. It provides perspective on the scale and urgency of the problem, helps to identify relevant precedents, and corroborates intuitions about how people respond to environmental upheavals.

Most fundamentally, economic prehistory reveals that the technological and institutional consequences of climate change are likely to play out in ways we cannot really imagine, any more than the residents of Abu Hureyra could have imagined agricultural societies, or the residents of Mesopotamian villages could have imagined the conquests of Sargon. The only certainty is that the way of life many of us presently enjoy will eventually become a distant memory.

Meanwhile, there is always hope. Perhaps the governments of the world will impose high prices on carbon emissions. Perhaps renewable energy will become so cheap that fossil fuels will be abandoned. Perhaps researchers will find ways to remove carbon dioxide from the atmosphere more efficiently than trees do. But the clock is ticking loudly.

12.8 Postscript

We thank Carles Boix, Richard Lipsey, Ola Olsson, Louis Putterman, Robert Springborg, and John Wallis for comments on earlier drafts of this chapter. They are not responsible for our errors, omissions, or opinions.