Climate, Geography, and the Evolution of Economic and Political Institutions Roy Elis and Stephen Haber Stanford University April 25, 2015

We are grateful for financial support from the UPS Foundation. Nicholas Baldo, Andrew Brooks, Roger Cain, Ian Claras, Kevin Cook, Kara Downey, Lauren Felice, Adriane Fresh, Anne Given, Jordan Horrillo, Scott Khamphoune, Dorothy Kronick, Michelle Lee, Ian Lewis, Cole Lupoli, Annamaria Prati, Cara Reichard, Jeffrey Tran, and Tracy Williams provided invaluable research assistance.

 

1  

Why are some world regions rich and democratic, while others are poor and autocratic? The geographic clustering of development and democracy has fascinated scholars going back as far as Aristotle, and emerged as a central question of political science and economics in the decades after World War II. The founding fathers of modern political science, Lipset (1959, 1963), Huntington (1968), and Dahl (1971), were of the view that causality ran from “structural preconditions,” including high per capita GDP, to an increase in the odds that a society would become democratic. A later generation of scholars, using more sophisticated statistical methods, largely concurred (e.g., Przeworski 2000). An equally large and distinguished literature holds, however, that causality runs the other way: representative political institutions gave rise to sustained economic growth (e.g., North and Weingast 1989; Barro 1998). More recent work tends to conceptualize the co-occurrence of high levels of economic development and consolidated democracy as part of a general equilibrium (e.g., Weingast 1997; Acemoglu et. al., 2008; Persson and Tabellini, 2009; North, Weingast, and Wallis 2009). What is not yet understood is how and why societies wound up in particular equilibria, and why those equilibria tend to cluster geographically. What pushed countries in Africa and Central Asia toward an equilibrium of low per capita incomes and autocracy, countries in North America and Western Europe toward an equilibrium of high per capita incomes and consolidated democracy, and countries in Eastern Europe, Latin America, South Asia, and East Asia to an equilibrium between these two extremes? Some scholars have pointed to geographic or climatologic factors as possible drivers of these general equilibria, and have focused in particular on continent orientation, disease environments, soil types, and biomes (Diamond 1997; Engerman and Sokoloff 1997; Hibbs

 

2  

and Olsson 2004; Acemoglu et. al, 2001, 2005; 2008; Easterly and Levine 2003; Putterman 2007). These scholars generally agree that geographic and climatologic factors have neither a direct nor contemporary effect. Instead, they work indirectly and historically, by shaping a society’s fundamental institutions.1 Beyond that, consensus breaks down. There is no agreement about the identity of those fundamental institutions, the geographic-climatologic factors that structured those institutions, and mechanisms by which those institutions shaped long-run paths of development. This paper offers a contribution to this growing literature. We identify two exogenous geographic-climatologic factors. 1) The potential amount of storable food energy that could be produced in the economic hinterland of the largest city in each country in 1800 (where the size of the hinterland varies across those cities, as a function of the distance one metric ton could be moved with a given amount of energy, using pre-steam technologies); and 2) The frequency with which that hinterland was subject to weather shocks severe enough to wipe out all producers simultaneously. We show that these two factors account for a surprising amount of the variance in regime types, per capita GDP, and education levels across the world today. These results are robust to controls for possible exogenous confounders, such as climates that favor malaria, natural resource income, and colonial heritage. We advance a theory to explain these results that emanates from three first principles. 1) Human societies had to mitigate the fundamental problems of food scarcity and uncertainty—or they disappeared. 2) At any point in time, the extent of the market for food is a function of the available technology for growing, storing, and transporting crops. 3) The                                                                                                                         1

Other scholars, most notably Sachs (1999) have pointed to geography as an explanation for variance in GDP (leaving aside institutions), and posit that it has a direct effect in so far as it works through the disease environment to affect productivity levels. For a critique showing that geography actually works through institutions, see Easterly and Levine (2003).

 

3  

extent of the market determines the degree of economic specialization and, hence, investments in human capital and productivity growth. We hypothesize that at the beginning of the modern era, when the nation-state was emerging as the dominant form of political organization around the world, climatologic and geographic factors determined what a society could grow, the extent to which it could be traded, and the scale at which it had to be stored. Climate determined whether a society could grow crops that could be easily stored, and hence traded. Geography determined that society’s cost of transportation. Climate and geography combined determined whether that society could mitigate food uncertainty caused by weather shocks through local and regional trade, a centralized food storage system, or calorie storage in animal or human body fat. The resulting economic organization of agriculture shaped the incentives facing agents toward investments in human capital, the protection of contract and property rights, and the creation of institutions to limit the authority and discretion of the governments necessary to enforce those rights—which is to say that the economic organization of agriculture sent societies down quite different paths of institutional development. Only one of those paths, which we have named the Transactional State, was likely to result in a wealthy society and a democratic polity. Wealthy democracies therefore tend to cluster in some world areas, while poor autocracies cluster in others, because climates and terrain features also tend to cluster. The countries of North America, for example, emerged as wealthy democracies because they had climates and terrains that produced large economic hinterlands that were not subject to aggregate weather shocks and that were well suited to growing crops that were easy to store. Weather shocks occurred, of course, but those shocks were temporally and spatially

 

4  

idiosyncratic. Thus, scarcity, when it did occur, could be addressed by local and regional trade: Farmer A, whose wheat crop had been wiped out by a heavy rain just before harvest, could buy wheat from Farmer B, 50 miles away, whose crop had not been struck by the storm; and Farmer A could generate the income to do so by selling his stores of a crop that was unaffected by that rain because it had a different growing cycle (e.g, beans) or by selling a non-agricultural good or service, such as the repair of Farmer B’s wagon. Populations in this type of high storability, low-transport cost, environment could therefore trade their way out of food scarcity and uncertainty. Local and regional trade, in turn generated incentives to specialize, invest in human capital, and create institutions that protected transactions. Importantly, protecting transactions required a state that was powerful enough to arbitrate any dispute, but that was also characterized by checks and balances so that it could not prey on the very property rights that it supposed to be protecting. Over the long run, a society characterized by a high level and broad distribution of human capital that had well-developed institutions to protect property rights while constraining the authority and discretion of the government was more likely than other societies to generate a dynamic economy and a consolidated democracy. The countries of Sub-Saharan Africa, to cite another example, remained poor and did not consolidate fledgling democracies because they had climates and terrain features that produced small economic hinterlands that were not well suited to growing storable crops (or storing what could be grown). They were also subject to aggregate weather shocks. Their populations therefore tended not to mitigate food scarcity and uncertainty by building dense local and regional trade networks. They therefore had fewer opportunities to engage in economic specialization, invest in human capital, and build institutions that both protected

 

5  

property rights and limited the authority of the State necessary to enforce those rights. Over the long run, such societies would be unlikely to create dynamic economies, and democracy, once introduced, would be unlikely to survive. Let us be clear. We are not suggesting that nothing besides climate and geography mattered; that contingent historical events played no roles in the long run evolution of societies. None of the R2’s from our regressions using climatologic and geographic variables are equal to 1. We are saying, however, that if contingent events explain the distribution of wealth and democracy around the planet, then our regressions should not account for onequarter of the variance in these outcomes across the planet. We are also not suggesting that climate and geography determined economic development and regime types in some mechanistic sense: temperate zones with rich soils and easily navigable waterways did not lead inexorably to democracy and high living standards; rainforests with rugged terrains did not lead inexorably to autocracy and poverty. We are saying, however, that democracy and economic development were more likely when there was a high level and broad distribution of human capital, and when contract and property rights were already strongly enforced by a State that could not itself prey on them. We are also saying that those characteristics of a society were more likely to exist when large numbers of citizens could accumulate tradable surpluses and were incentivized, over the course of generations, to make investments in human capital and to erect institutions designed to encourage the state to serve as an arbiter of property rights. Finally, we are saying that those conditions were more likely to exist in societies where there were inherent pressures and tensions that favored their emergence—and a powerful source of such pressures and tensions

 

6  

was the organization of the agricultural economy, the single most important economic sector in most societies until the late nineteenth century. This paper continues as follows. Section Two provides a literature review. Section Three advances a theoretical framework that builds upon the existing literature. Section Four explains how we built the geo-coded data sets necessary to conduct our analyses. Section Five assesses the theory using cross-country regressions. Section Six concludes. Section 2: Literature Review We do not claim to be the first scholars to have noticed that climate and geography matter. There is a long literature about the effects of climate on human behavior that stretches all the way back to Aristotle. More recently, Hibbs and Olsson (2004) and Putterman (2007) have hypothesized that climate, working through the timing of the transition to agriculture from hunting and gathering, played an important role in the long run process of economic growth. Engerman and Sokoloff (1997) famously argued that variance in natural environments (differences in climate, soil quality, mineral endowments, and the size of the native population) across the Americas at the time of European colonization conditioned longrun paths of economic and political institutions, and hence explain present-day differences in per capita incomes between the United States and Canada, on the one hand, Latin America and the Caribbean on the other. Acemoglu, Johnson, and Robinson (2001) also exploit variance in climate to explain present-day differences in per capita incomes in former European colonies, but for them the mechanism linking climate and economic growth is the disease environment: where Europeans faced high mortality risk because heat and humidity favored the species of mosquitos that carry Malaria and Yellow Fever, they set up “extractive” institutions designed to maximize their economic return over the short run; but  

7  

where they faced low mortality risk, because the climate was drier and cooler, they set up “settler” institutions designed to create a political and economic environment that resembled Western Europe. These initial colonial institutions then conditioned countries’ long-run paths of development: places with extractive institutions (e.g., the Congo) culminated in autocracy and underdevelopment; places with settler institutions (e.g., the United States) culminated in democracy and high per capita incomes. These seminal contributions have pushed social scientists to take geography and history seriously. They are, however, incomplete explanations of why we observe systematic relationships between climate and geography on the one hand, and economic development and democracy on the other. Hibbs and Olsson (2004) and Putterman (2007) focus on economic growth, not political regimes. Engerman and Sokoloff (1997) see political regimes as a mediating factor between geography and economic development. They also focus solely on the Americas, and do not use statistical analyses to test their claims. Acemoglu, Johnson, and Robinson (2001) also see regime types as mediating factors between geography and economic growth. In addition, they restrict their analysis to Western European colonies. Places that were skipped over by European powers because they were inhospitable (e.g., the Arabian Peninsula), were integrated into another country only to emerge as sovereign states more recently (e.g., the countries of Central Asia, which were conquered and made part of Russia proper), which remained outside of the colonial system because they were buffer states (e.g., Thailand), or that remained sovereign states because they had the potential to resist colonization (e.g., Japan), lie outside of their framework. The countries of Europe itself—both those that were colonizers and those that were not—also fall outside of the Engerman-Sokoloff and Acemoglu, Johnson, and Robinson purviews.

 

8  

We also do not claim to be the first researchers to realize that regime types and levels of economic development are outcomes of long-run historical processes. Moore (1966) made a seminal contribution to the study of comparative democratization by looking for its social roots in centuries past. He was followed by Luebbert (1991), Rueschemeyer, Stephens and Stephens (1992), Collier (1999), and Capoccia and Ziblatt (2010). North, Wallis, and Weingast (2009) make a similarly important contribution to comparative economic development by looking for its long-run political roots. We also do not claim to be the first to have noticed that the structure of agriculture has implications for the political organization of states. Carneiro (1970) argued that centralized states first emerged in areas where agricultural production was geographically circumscribed by mountains, seas, or deserts that limited the area that could be cultivated. His theory, however, focused on the competition for scarce land, not the characteristics of the crops that could be grown on that land or the frequency with which all of those crops were lost to a weather shock. Carneiro’s framework was also about explaining the rise of the first states, not about explaining variance in regime types or levels of economic development across states. We are also not the first scholars to appreciate the fact that people cannot grow food without water, and that there is therefore a relationship between control of water and a range of political and economic outcomes. Wittfogel (1957) famously argued that control of great rivers was the driving force behind authoritarianism in the ancient and modern worlds. A generation of historians has shown that Wittfogel’s theory of hydraulic society was a fanciful set of assertions: irrigation does not inexorably lead to corvee labor, centralized bureaucracies, and totalitarian leaders (see, for example, Manning 2010).

 

9  

We would suggest, however, that Wittfogel was on to something: there does seem to be a correlation between agricultural systems based on controlling the waters of a great river, authoritarian political institutions, and economic backwardness—to wit the societies that have continually emerged over the past four millennia on the banks of the Nile, Euphrates, and Yangtze—but Wittfogel got the mechanism wrong. Irrigating crops by capturing the flow of a great river does effect fundamental institutions, but not because large-scale irrigation requires corvee labor. Rather, the need to capture the flow of a great river means that if upstream precipitation varies temporally, all producers downstream can be simultaneously wiped out by drought or flood. For example, prior to the construction of the Aswan Dam in the 1960s, wide swings in the extent and intensity of the West African Monsoon determined how much rain fell at the Nile’s headwaters in Central and East Africa. The resulting swings in the Nile’s level determined whether harvests across Egypt would succeed or fail. The fact that Egyptian agriculture was subject to shocks that were both spatially and temporally correlated meant that farmers could not trade their way out of scarcity and uncertainty. The solution instead was to create a centralized state that taxed grain at high levels during good years and distributed that grain during lean years. Determining the appropriate rate of taxation was easy: Egyptian rulers only had to measure the height of the Nile at the Aswan cataract. The existence of this system, however, fundamentally shaped Egypt’s basic political institutions. Farmers had an incentive to support a State powerful enough to insure against uncertainty by forcing producers to transfer their surpluses and that could protect those surpluses from any potential coalition or State that wanted to seize them. But, at the same time, the ability of farmers to constrain the authority and discretion of that State was weak.

 

10  

What countervailing forces would give rise to a system of checks and balances on State authority in this system? Section 3: Theory Let us build up a theory that starts with first principles. First and foremost, human societies must create stable sources of food energy, or they die out. Second, at any point in time, the extent of the market for food is a function of the available technology for growing, storing, and transporting crops.    Third, the size of the market determines the degree of economic specialization, and economic specialization produces gains in productivity. Brute facts of nature influence the size of the market for food, by affecting what can be grown, whether it can be stored, and the distance at which it can be traded. There are biological differences across the types of crops that thrive in temperate zones versus the tropics, and these influence the storability, and hence tradability, of those crops. Cereal grains and pulses thrive in temperate zones. They are in high in calories and can be stored for years on end because they have low moisture contents and go dormant after harvest. They also have definite growing seasons, which means that they are subject to weather shocks: if it is too wet or too dry at a particular time in the growing cycle, an entire year’s crop can be lost. Cereal grains and pulses are also highly sensitive to the conditions under which they are stored: the same grain that might remain edible for several years when stored under cool, dry conditions will decay within weeks if stored in a hot, humid environment. Thus, even those cereal grains that will grow under tropical conditions (e.g., maize) are much less easy to store, and hence much more difficult to trade, than those same grains grown in a temperate environment. The crops that thrive in the tropics, such as roots, tubers, and fruits can be stored for much shorter periods of time because they have high moisture contents, high  

11  

respiratory rates, are subject to fungi and microbial attack, and typically do not go dormant after harvest. Some tropical crops decay within hours of harvest (e.g., Sugarcane), some last but a few days (e.g., Bananas), some a few weeks (e.g., Taro), and some a few months (e.g. Yams).2 These crops tend not to be seasonal (they are produced year round), which means that a single weather event is less likely to wipe out an entire year’s crop. In short, prior to the invention of refrigeration and innovations in transportation technologies in the late 19th century, temperate areas were more likely to be able to support large populations than tropical areas, because of the high calorie content of the storable cereals and pulses that could grow there and because those crops’ high storability made them easy to trade. The point to keep in mind is that climates that are well suited to growing storable crops are not randomly distributed around the planet. Tropical environments, which are poorly suited to growing storable crops (and are poorly suited to storing whatever can be grown) cluster around the Equator. Temperate environments cluster in middle latitudes. Whether a society can actually trade its storable surpluses is also affected by brute facts of nature. Transportation technologies have changed dramatically across the past two centuries, but regardless of the state of technology at any point in time, the forces of gravity and friction continue to operate. All other things being equal, it requires far less energy to move goods by water, where friction is dramatically reduced, than it does over land. All other things being equal, it requires far more energy to move goods uphill than it does to move them over flat terrain. As we will show later, navigable bodies of water and flat terrains are not randomly distributed across the planet. Indeed, two areas of the globe are particularly                                                                                                                         2

Sugarcane is the quintessential example: If cut cane is left unprocessed for more than 12 hours, the sugar is lost to fermentation (Binswanger and Rosenzweig 1986; Dye 1998). Bananas and Cassava can be stored for only five to seven days; Taro for two to six weeks; Sweet Potatoes two to four months; Yams 12 to 18 weeks (Ravi, et. al., 1996; Diop and Calverley1998; Abu-Goukh 1986).

 

12  

blessed by flat terrains and slow moving rivers that remain ice-free year-round; North America and Western Europe. The extent and frequency of weather shocks is also not randomly distributed across the planet. While there have been periods of cooling and warming (e.g, the Roman Warm of 250 BC to 400 AD, the Mini Ice Age of 1350-1750, and global warming over the past four decades), the basic climate patterns of the planet have been stable for several millennia. For our purposes, the notable fact is that some world areas are affected by monsoons (shifting winds that produce dry and wet seasons which alternate dramatically and extend over huge areas), while others do not. Because of variance over time in solar radiation, monsoons can vary in intensity from year to year, or cycle over multi-year periods, producing persistent droughts or flooding over broad areas. Broadly speaking, South Asia, South East Asia, East Asia north of Australia, Mexico and the U.S. Southwest, and West, Central, and East Africa all have monsoonal climates. The El Niño/ La Niña oscillation in the Eastern Pacific that drives the East Asian Monsoon also affects, albeit in an irregular, inter-annual fashion, the west coast of South America, Central America, Mexico, and California. The dry phases of the West African and South Asian Monsoons give rise to seasonal winds that also affect a broad swathe of the planet that extends from the Gulf of Guinea, across the Sahara Desert and the African Sahel, into the Middle East. There are two central points to keep in mind. The first is that there are world areas that are not affected by monsoonal systems, and thus are less prone to spatially and temporally correlated weather shocks. These regions include Canada, the U.S. East Coast and Midwest, Western and Eastern Europe, Scandinavia, Australia and New Zealand, South Africa, and the East Coast of South America. There are certainly weather fluctuations in

 

13  

these areas, but they are idiosyncratic events. The second is that his brute fact of nature has implications for how societies can mitigate shocks to production caused by the weather. In a region characterized by temporally and spatially correlated weather shocks it is difficult for a society to trade its way out of weather-related crop losses: everyone has been wiped out simultaneously—and may be so for multiple years on end, a situation perhaps captured most evocatively in the Bible’s story of Joseph and the Seven Years of Feast and Seven Years of Famine in Egypt and Canaan. In a region characterized by idiosyncratic weather shocks, the crop can fail in Locality A but the harvest can be plentiful 50 kilometers way in Locality B, and thus A and B can trade with one another to mitigate scarcity. Moreover, in the following year, Locality B may be hit by a weather shock, while A has a bumper crop. The effects of aggregate weather shocks are magnified when agriculture is dependent on irrigation from a single great river. A weak rainy season at a river’s source means drought for all farmers downriver. An unusually wet rainy season at a river’s source means flooding for all farmers downriver. Finally, let us add one fact that is not a brute fact of nature, but that has been shown to be true on the basis of evidence from archaeology, history, and agricultural economics—and hence is not debated: nobody owns the rain as it falls, but once it is channeled into a river or stream it becomes appropriable. The implication is that when a crop must be produced using irrigation from a river or stream whoever has property rights to the water can, literally, siphon off the farmer’s surplus. Farmers who rely on the rain do not face this problem of appropriation. Let us now put these facts together in order to create a framework with which to understand how basic features of climate and geography influence the incentives of agents,

 

14  

and how differences in incentives give rise to very different institutions and social structures, some of which are conducive to long-run economic development and fertile ground for democracy to take root, and some of which are not. Transactional States In those areas of the world where transportation costs were low because they bordered navigable bodies of water, had soils and climates that were well suited to growing and storing crops that did not rapidly decay (e.g., cereal grains), and were not subject to aggregate weather shocks, economic agents only had to contend with local, idiosyncratic shocks to production. Those local failures could be addressed by trading with a nearby locality that did not experience that idiosyncratic shock. Over time, regular local and regional trade encouraged specialization, which is to say investments in human capital. Local and regional trade in storable crops also generated a problem: any store of value creates incentives for theft. Hence local and regional trade in storable crops also gave agents incentives to create institutions that protected property and contract rights, both from other members of the society and from the government that had to be erected to enforce those rights. The resulting social structures and institutions that emerged in these “Transactional States”—high levels and broad distributions of human capital, and a government that enforced property and contract rights, but that was itself limited in its authority to reduce those rights for its own ends—pushed them down a path of development that was characterized by increasing returns, and that increased the likelihood of high population density, urbanization, industrialization, and consolidated democracy. The United States, which was founded with indentured and convict labor as a backwater colony of the British Empire, is a quintessential example. Britain did not set out to  

15  

create a colony that would reject monarchism, embrace broad rights of suffrage, and eclipse it economically. Those things were able to happen because the natural environment, particularly in New England and the Mid-Atlantic, were conducive to the emergence of a population of free, commercially-oriented farmers who were evenly matched in terms of their human capital, and who had been erecting institutions to limit the authority of their British overlords for more than a century prior to the American Revolution. Insurance States In societies that could not trade their way out of food scarcity and uncertainty because frequent aggregate weather shocks wiped out all agricultural producers simultaneously, economic agents had weak incentives to invest in institutions that limited the government’s authority and discretion vis a vis contract and property rights. But if the food crops grown there could be easily stored, and if the costs of transport were low, they did have incentives to erect an insurance institution—a powerful state that could store crops centrally and distribute those stores during times of scarcity. Such “Insurance States” could sustain large populations, but the existence of a centralized state powerful enough to tax away farmers’ surpluses and then defend them against any coalition that wanted to seize them served as a constraint over the long run on economic development and the consolidation of democracy. Without any institutionalized power that could serve as a check and balance, what would stop the central state from preying on all sources of wealth or stamping out any source of dissent? China is a quintessential example. Indeed, China is unusual in world history in that it has been a recognizable political entity with a highly centralized government since 221 BC: dynasties rose and fell, Mongols and Manchurians invaded and set themselves up as rulers;

 

16  

but, unlike Western Europeans, who never succeeded in re-creating the political unity of the Roman Empire, Chinese society continually re-produced a highly centralized State with broad authority and discretion to reduce the rights of citizens. We would suggest that the re-creation of basic political institutions over the course of two millennia is not an accident. Rather, it is the outcome of two facts of climate and geography: 1) the two powerful rivers that simultaneously allowed for low cost transport, nourished the country’s agricultural lands, and threatened to wash those lands away when they spilled their banks required investments in water management that could not easily be coordinated across a system of fragmented states; and 2) dependence on riverine irrigation meant that weather shocks at river headwaters could produce droughts and floods that could wipe out all downstream agriculturalists at the same time. The response to this uncertain environment was a centralized state that, from the rise of the Han Dynasty (206 BC) to the fall of the Qing Dynasty (1912), spent enormous sums on infrastructure to control rivers and that taxed and stored prodigious amounts of grain so as to insure against harvest failures. Under the Qing Dynasty (1644-1912) this granary system became incredibly sophisticated; the central government gathered information on grain prices and weather in order to predict regional food shortages, and responded by selling rice from its local and regional granaries into the market. Nothing on this scale was ever imagined in Europe, let alone achieved (Wong 1997). This insurance system could mitigate the effects of persistent, large scale, droughts and floods, and thus permitted population to expand rapidly, but it came at a cost: a powerful centralized state could head off any threat to its persistence, whether that threat came in the form of anti-monarchical political ideas or the emergence of a merchant class that could have coupled capital to scientific discoveries in order to create an

 

17  

innovative economy. The result was a country that has been persistently autocratic and that has only recently begun to make the transition to modern economic growth.

Stateless Societies and Predatory Kingdoms In tropical environments, where high rainfall makes it difficult to grow crops that are easily stored, and where high temperatures and humidity frustrate attempts to store whatever can be grown, food scarcity and uncertainty is somewhat alleviated by the fact that the crops that thrive under those conditions do not have seasons; they ripen year round. There is a tradeoff, however: these non-seasonal crops tend to have high moisture contents and do not go dormant after harvesting, and hence decay rapidly after harvesting. In the era before refrigeration and rapid transport, those crops could therefore not be easily traded. Societies in the tropics therefore were less likely to create dense networks of local and regional exchange in order to trade their way out of weather shocks. The incentives facing economic agents to specialize, which is to say invest in human capital, or create institutions to protect contract and property rights were weak. One common outcome in such an environment was a stateless society. Examples include Brazil, Venezuela, Central America, and the Caribbean prior to the arrival of the Spanish and Portuguese in the 16th century, as well as Papua New Guinea before it was colonized by Germany and Great Britain in the late 19th century. Another common outcome was kingdoms whose method of public finance was to tax human muscle power. They therefore focused more on developing property rights in people than they did on property rights in land. Hence to the degree that States emerged, they tended

 

18  

to be predatory. Even these Predatory States tended to be rudimentary: coerced muscle-power is not particularly valuable if it cannot be used to produce something that cannot, itself, be accumulated or traded. The Kingdom of the Kongo, which stretched across modern day Northern Angola, the Republic of the Congo, the western portion of the Democratic Republic of the Congo, and Southern Gabon, is an example. Prior to the arrival of Europeans in the sixteenth century, the Kingdom of the Kongo, was characterized by a low population density, miniscule cities, and nobility that relied on slave labor to produce an agricultural surplus so modest that Europeans were struck by their impoverishment (Broadhead 1979; Thornton 2001). This predatory state was then mobilized by Europeans to provide human beings for the Atlantic slave trade. Indeed, the areas that comprised the Kingdom of the Kongo, with the exception of some thin strips of land along the coast, were not colonized by Europeans until the 1880s. Even then, the natural environment was such that Europeans looked at them as sources of natural resource wealth, particularly ivory and wild rubber. At independence, the fundamental institutions of predatory states, as well as stateless societies, were not well suited to democratic consolidation or the creation of dynamic economies. Even prior to colonization, they had not been on paths of development associated with economic specialization, investments in human capital, or the construction of a capacious state that was limited in its authority and discretion. To the degree that European colonization had an effect, it was to either to reinforce existing predatory institutions or to introduce new ones.

 

19  

Desert and Grassland Kingdoms In arid environments, where low levels of rainfall made it difficult if not impossible to grow food crops, energy had to be harnessed from animals that could turn grass or scrub into milk and meat. When weather shocks occurred, they were not mitigated by trade, but by moving the animals to a better pasture. The people that lived on these steppes specialized, but not in producing tradable goods; they specialized in making war off of the backs of horses or camels. Until the perfection of firearms, steppe peoples were potent fighting forces that could conquer vast territories, but the inherently mobile nature of these societies gave rise to personal kingdoms or empires, not to States. At the dawn of modernity, therefore, they were either conquered and subsumed into neighboring States, as happened to the nomadic populations of Central Asia and the U.S. West, or they continued to function as impoverished kingdoms, as was the case with the Arabian Peninsula until oil was discovered there in the 1930s. Section 4: Data and Measures In order to subject our theory to tests against evidence we develop estimates of a series of dependent and independent variables. Estimating the dependent variables is straightforward, while estimating the independent variables is considerably more involved. Dependent Variables For the level of economic development, we use GDP per capita in 2010 from the Penn World Tables 7.1, (chained series, at 2005 constant prices). For the level of democracy we draw on the Polity 2 measure of the Polity IV dataset, which is the standard measure of the

 

20  

degree of political democracy/autocracy in the comparative politics literature.3 It captures the competitiveness of political participation, the openness and competitiveness of executive recruitment, and the constraints on the chief executive for every country in the world with a population above 1 million, beginning at political independence, or in 1800 for countries that are not former colonies. In order to create a measure that captures both a country’s level of democracy and its long-run stability—its equilibrium level of democratic consolidation—we average each country’s’ Polity 2 score over the period 1975-2012.4 For ease of interpretation, we rescale Average Polity 2 so that it runs from 0 to 100 (instead of -10 to +10). We use two measures to capture the level and distribution of human capital; the average number of years of schooling of the adult population in 1950, which we draw from Barro and Lee (2010); and the Whipple Index, a proxy measure for numeracy, for the year 1910 from Crayen and Baten (2010).5                                                                                                                         3

http://www.systemicpeace.org/inscrdata.html This means that we do not include countries that became independent after 1975 in our regressions. Drawing averages over a shorter period of time (say 1991 to 2012) increases the number of observations, but it flattens the distribution of average Polity Scores. For example, the distance between Brazil and Switzerland is 26 points (out of 100) when measured over the period 1975 to 2012, but is only 10 points when measured over the period 1991 to 2012, because of Brazil’s democratic transition beginning in the late 1980s. It is difficult to believe, however, that Brazil’s fledgling democracy is 90 percent as consolidated as that of Switzerland, one of the longest standing democracies in the world. Even longer windows of time—for example, from 1950 to 2012—produce even wider distributions of average Polity Scores (the gap between Brazil and Switzerland grows to 37 points out of 100), but this increase in the precision of the estimate comes at the cost of the number of observations because big swathes of the planet were still under colonial rule in 1950, and hence are not scored by the Polity IV Project until they gained independence. We note that our basic results are robust to measuring average polity 2 over the period 1960 to 2012, 1965 to 2012, and 1991 to 2012. 5 The Whipple index relies on a phenomenon known as age-heaping, in which census respondents sometimes inaccurately report their ages, systematically rounding or preferring ages ending in 0 or 5. The Whipple index is a measure of the concentration of reported ages ending in 0 or 5 relative to one fifth of the full sample. An index score of 100 shows no heaping (low innumeracy) and a score of 500 indicates heaping by all respondents (high innumeracy). We expand the Crayen and Baten (2010 estimates as follows. We estimate the value for Ethiopia and Niger in 1910 by assigning their recorded value in 1940, based on the reasonable assumption that the value in 1910 had to be at least as high as for 1940. Thus, the observation we fit for 1910 is biased in favor of Ethiopia and Niger having a lower (better) score than they did in reality. We estimated the values for Syria, Lebanon, Saudi Arabia, Kuwait, Yemen, Bahrain, Qatar, Oman, and the United Arab Emirates by assigning them the lowest (best) value observed for any North African or Middle Eastern country in 1910. This means that are 4

 

21  

Independent Variables Generating the independent variables is considerably more difficult because our task is to approximate the storable agricultural capacity of the hinterland of each country’s largest city circa 1800 and to then estimate the frequency with which that hinterland would be affected by a severe weather shock. That means that we must estimate the size and shape of each city’s hinterland as a function of the state of transportation technology before the widespread dissemination of steam power (as applied to that region’s terrain features and navigable waterways at that time). It also means that we must estimate the storable agricultural capacity of that hinterland using the technology available at the end of the 18th century. Finally, it means that we must measure the frequency of weather shocks to that hinterland. We pick the hinterland around the largest city in each country in 1800 because the fundamental institutions of modern nation states tended to originate in a core area, and were then transplanted to other areas by assimilation, conquest, and colonization. Examples of such core areas include the Ile de France (the area around Paris, from which the modern French state emerged beginning in the 10th century), the Valley of Mexico (the central plateau around Mexico City, that was the center of the Teotihuacán, Toltec, and Aztec Empires, the Spanish colony of New Spain, and the nation state of Mexico after independence), and the Tokyo region (from which emerged the modern nation of Japan beginning with the Tokugawa Shogunate in 1603). The United States provides an example of how institutions spread from a core area: The 13 original colonies were all located on the Eastern Seaboard, and thus their                                                                                                                                                                                                                                                                                                                                                                                     assuming that countries that likely had very high levels of innumeracy had the same (low) value as Tunisia (which Crayen and Baten estimate at 156). This procedure bias the results against our hypotheses.

 

22  

institutions were endogenous to the natural environment there. The U.S. Constitution not only harmonized the institutions of those 13 states, but it required that any additional state that joined the union had to have its constitution approved by Congress. Thus, Arizona’s institutions are not endogenous to Arizona’s natural environment; they reflect the institutions of the core area of the United States. Hinterlands For each country in the dataset, we identify the largest city in 1800 and calculate the size of its hinterland as a function of the distance one metric ton could be moved with a given amount of energy, using pre-steam technologies. Holding transportation technology fixed to pre-steam levels, friction and gravity are the fundamental drivers of hinterland size. The uneven spatial distribution of flat terrain and navigable waterways therefore drive variation in hinterland size across the globe, as will become apparent below. We calculate hinterlands using the Path Distance Allocation tool in ArcMap 10.1. We use the tool to calculate the accumulated costs of moving a metric ton of goods from any location on the map to each city in our dataset, over a cost surface, along the least cost path. For each city, the tool begins calculating the accumulated transport costs at the city itself and works its way outwards until the accumulated transport costs from the next cell to the city reach the budget constraint. Only cells on the map that fall within the budget constraint are coded as belonging to each city's hinterland. In the contexts of our analysis, costs and budget constraints are set in terms of physical parameters (e.g. friction, energy, work) rather than economic costs (e.g. dollars per ton-mile). We provide as inputs for the tool a map of the world composed of the following layers:  

23  

•

A vector dataset that geocodes the largest city in 1800 for each country in our dataset.

•

A raster Digital Elevation Map (DEM) that encodes the average elevation of each .225 by .225 km cell on the map (resampled to 1km by 1km for computational efficiency). The DEM enables the tool both to calculate transportation costs due to elevation changes (i.e. fighting gravity) and also allows the tool to calculate the true transportation distance along a 3-dimensional surface (i.e. the hypotenuse rather than as the crow flies when traveling up or down hill).

•

A raster cost surface that encodes the per meter physical costs incurred by transporting one metric ton of goods through the cell. For the sake of simplicity we specify two types of costs based on overland and waterborne transportation. We provide three additional inputs for the tool: a physical model that translates

elevation gains to quantities of physical work expended, a set of assumptions about the physical limits of pre-steam transportation technology (e.g. a maximum grade beyond which a team of horses cannot pull a load), and a budget constraint for the total amount of energy available. We explain each of these elements in detail in the following sections. Overland Travel Determining the physical cost of transportation requires that we fix a pre-steam transportation technology from among the available options, including human porters, mule trains, and horse-drawn wagons. We chose horse-drawn wagons, the Conestoga Wagon in particular, given its historical use for transporting bulk goods across long distances circa 1800. Though porters and mule-trains enjoy some advantages, such as the ability to traverse steep grades, the overall energy efficiency of a wagon far outweighs any disadvantages.

 

24  

Wagon transport therefore provides the most optimistic estimate of energy costs for travel by land. Our strategy is to estimate the amount of energy required for a loaded Conestoga wagon to maintain a constant speed on flat land. That is, we assume away the question of acceleration and simply calculate the energy required to overcome friction, primarily rolling resistance in the case of a wagon. We rely on historical data to determine the physical characteristics of a typical Conestoga wagon, a typical load size, and the total resistance based on those factors plus road surface and road conditions. We estimate 492 newtons as the force required to overcome the total resistance (per metric ton of cargo) of a loaded Conestoga wagon on a flat, earthen road in good condition. Applying that force over 1 meter implies 492 newton-meters, or joules, as the amount of energy expended per metric ton in hauling cargo via Conestoga Wagon over a flat, earthen road in good condition, at constant speed. This is the value encoded for overland travel in the friction surface used by the Path Distance Allocation tool in ArcMap 10.1. Below we explain the data and calculations behind this number. Diameter and width of the wagon wheels are two of the factors that determine the rolling resistance of a wagon. We estimate the typical diameter and widths of a Conestoga wheel based on data in Shumway (1964). Rolling resistance is directly proportional to the gross weight of the vehicle (vehicle plus cargo), since the road surface deforms in direct proportion to gross weight. We assume a wagon weight of 3250 lbs and a cargo of 7000 lbs based on data in Shumway (1964) and the Pennsylvania State Historical Commission website.6                                                                                                                         6

 

http/www.portal.state.pa.us/portal/server.pt/community/things/4280/conestoga_wagon/478210

25  

Road surface and conditions can impact the difficulty of travel by wagon tremendously, with poor conditions capable of increasing the total resistance by an order of magnitude or more, relative to good conditions. Within the constraint of choosing a natural road surface, we choose the best possible conditions so that we do not overestimate the difficulty of land-based transport relative to water. Based on data in Baker (1965: 18) we assume a resistance of 68.5 lbs per ton based on the average value for freight wagons on a nearly dry earth road in very good condition. We also accounted for energy required to overcome elevation changes based on the gross weight of the wagon and the gradient of the path. Based on Baker (1918) we limited wagon transport to a 10% grade, at which point a horse can barely generate enough power to transport itself over any appreciable distance. Waterborne Transport In dealing with water-based modes of transportation, we make several simplifying assumptions. First, we do not distinguish river travel from travel along major lakes, seas, and oceans. Second, we largely assume away the effect of river currents in the physical model, with the exception that we code rivers with sufficiently strong currents as non-navigable using pre-steam technologies (based on historical sources). Given that coastal and ocean transport are cheaper than riverine transport in terms of energy expended per ton of cargo, the first assumption leads us to overestimate the energy costs of oceanic transport. Since most rivers flow from the hinterland towards the city, the second assumption leads us to overestimate the energy costs of river transport as well. Our estimates are therefore biased against cities that could take advantage of water transport. Since North America and Western Europe are particularly well endowed with navigable rivers and have long coastlines relative to their  

26  

overall size, our procedures bias against the size of North American and Western European hinterlands, thereby biasing against our hypotheses. As with land-based transport, determining the physical cost of transportation over water requires that we fix a pre-steam transportation technology from among the available options. In this case we chose two of the boats used by Lewis and Clark in their westward expedition between 1804 and 1806. These boats were typical of the smaller vessels used for riverine transport in the United States circa 1800, less capacious than river barges or flat boats, but better designed to navigate smaller rivers. These boats were significantly smaller that ocean-going vessels, and therefore lead us to overestimate the energy costs of ocean transport. Fixing the transport technology of Lewis and Clark's boats therefore provides an overly pessimistic estimate of energy costs for travel by water. Parallel to our treatment of land-based transport, our strategy is to estimate the amount of energy required to pull a loaded boat, at a constant speed, through still water. Again, we assume away the question of acceleration and calculate only the energy required to overcome friction, primarily drag resistance in the case of water travel. We rely on historical data to determine the physical characteristics of the boats, a typical load size, the typical draught for a loaded boat, and the drag coefficients given the shape of the boats. We estimate 53 newtons as the force required to overcome the drag resistance (per metric ton of cargo) of a loaded boat moving through still water. Applying that force over 1 meter implies 53 newton-meters, or joules, as the amount of energy expended per metric ton in hauling cargo via Lewis and Clark's boats through still water at a constant rate of 3 miles per hour. This is the value encoded for water travel in the friction surface used by the Path

 

27  

Distance Allocation tool in ArcMap 10.1. Below we explain the data and calculations behind this number. Drag is the main source of resistance encountered by an object moving through water at low speeds, and is determined by a simple physical formula: 1 𝐹! = 𝜌𝑣 ! 𝐶! 𝐴 2 where FD is the drag force, ρ is the density of fluid (in this case water), v is the relative velocity of the object, CD is the drag coefficient, and A is the cross-sectional area of the object, in this case the area submerged in water. At low speeds we can safely ignore drag forces due to air resistance. The value of ρ is given for water (approximately 1000 kg/m3 at 4°C), we assume the velocity to be 3 miles per hour, we calculate the cross-sectional area based on the shape and draft of the boats based on data about boats at that time, and use a drag coefficient of 0.295 based on the shape the boat, which approximates a bullet form (rounded in front, squared at back). 7 Converting the historical data to metric units, we arrive at a drag force of 56.6 newtons (per metric ton of cargo) for Lewis and Clark's keelboat, and a drag force of 49.1 newtons (per metric ton of cargo) for the white pirogue. Averaging the two, we arrive at the estimate of just under 53 newtons of drag force (per metric ton of cargo) for waterborne transport.

                                                                                                                        7

See http://www.lewis-clark.org/article/496; http://www.lewis-clark.org/article/3072; and http://www.grc.nasa.gov/WWW/k-12/airplane/shaped.html

 

28  

Navigable Waterways In constructing the friction surface, we assume that all oceans, seas, and lakes are navigable using pre-steam technology. River navigability, however, is a complicated matter that we address here briefly and at length in the appendix. There are at least three factors that complicate our attempts to determine the navigability of rivers circa 1800 with pre-steam technology. The first challenge is data availability. To the best of our knowledge, geo-coded historical maps of navigable rivers around the world simply do not exist. Second, whether a river is navigable as a matter of principle is a different question from whether it was in fact navigated. Ideally, we want to capture the former. At least in theory, the possibility exists that rivers that were navigable in principle were not actually used for purposes of trade due for economic reasons. Third, societies have been busy altering the natural environment since pre-historic times. One might worry, then, that navigability in 1800 is not purely a matter of natural circumstances, but is also a matter of human intervention (e.g. widening and deepening of channels, dredging, etc.) Our strategy is to rely on historical sources that describe navigation and attempted navigation with pre-steam technology to help us code the world's major rivers as navigable or non-navigable circa 1800. In the instance of the United States, we were able to find a single map detailing the principal waterways circa 1890 [Statistical Atlas of the United States (1898) retrieved from the David Rumsey map collection (www.davidrumsey.com)]. In this case, we overlaid the historical map on a modern georeferenced map using ArcGIS, matched the images of rivers from the historical map to a vector dataset of rivers and streams on the georeferenced map, and then coded each on the latter for navigability manually.

 

29  

For the remainder of the world, we started with a universe of cases consisting of major perennial rivers from the Natural Earth dataset8, as well as the CIA World Databank II9. Using country- and region-specific historical sources, we found historical references to navigation attempts circa 1800 for each and every river in the dataset. We then coded as navigable any river that was navigable at least six months of the year circa 1800 using presteam technology. The Budget Constraint The final input used for calculating hinterlands is the budget constraint, or the total amount of energy available for transporting a metric ton of cargo to each city in the dataset. We set the energy budget based on historical evidence on the distance of typical wagon hauls for agricultural goods in the 19th century United States contained in Fogel (1964, pp. 7579).10 Based on Fogel's estimates, we set our energy budget at 40 megajoules, the total amount of energy it would take to transport a metric ton of goods 50 miles on perfectly flat land using a Conestoga wagon.11 Hinterlands for the Largest City in 1800 The mean area for the hinterlands in our dataset is 53,058 km2 with a standard deviation of 38,465 km2. In Figures 1 and 2, we show the hinterlands for New York City and                                                                                                                         8

 http://www.naturalearthdata.com    Now incorporated into the Global Self-Consistent, Hierarchical, High-resolution Geography Database http://www.soest.hawaii.edu/pwessel/gshhg/   10 See Fogel (1964): "... Texas, for example, reported that the distance of wagon haulage from some counties to shipping points was 110 miles for cotton, 47 miles for corn, 47 miles for hay, 42 miles for oats, 46 miles for potatoes, and 60 miles for wheat." Further, "...data on the North Atlantic region indicate that, on average, the boundary of feasible agricultural production would have been located between 40 and 50 straight line miles from a navigable waterway" (pp 78-79). 9

11

Fifty miles is approximately 80.46 km. At 492 joules per meter (or 492,000 joules per km) transporting a metric ton of goods 50 miles by wagon requires 39,586,320 joules, which we round to 40 megajoules.

 

30  

Mexico City as calculated by the Path Distance Allocation tool to give a sense of the how terrain and access to navigable waterways jointly determine hinterland extent. New York's expansive hinterland is a function of coastal access, proximity to navigable rivers, and a relatively flat coastal plain. Mexico City's hinterland is severely limited by its reliance on land transportation and a mountainous terrain. The effect of the mountains clearly limits and shapes Mexico's hinterland, which extends nearly 80 kilometers along flat land, but only 40 km through more rugged terrain, e.g. due southwest of the city.

  Figure  1  Hinterland  for  New  York  City  at  a  40  megajoule  energy  budget  is  approximately  93,300  square   kilometers.    Scale  1:7,000,000.    Source:  authors'  calculations.  

 

  Figure  2  for  Mexico  City  at  a  40  megajoule  energy  budget  is  approximately  13,320  square  kilometers.    Scale   1:7,000,000.    Source:  authors'  calculations  

 

31  

 

Potential Caloric Yield of Storable Cereals For each hinterland, we calculate the total potential kilocalorie yield of storable crops, averaging across eleven major crops. We calculate cereals potential using data from The Global Agro Ecological Assessment for Agriculture (GAEZ 3.0), which is a joint project of the FAO and the International Institute for Applied Systems Analysis. The GAEZ 3.0 project provides a collection of raster datasets covering the globe that estimate, among other things, the potential yield per hectare of growing a number of crops under rain-fed, intermediateinput conditions. GAEZ calculates the potential yield for each crop based on crop characteristics, soil characteristics, and climate.12 We make the reasonable assumption that cross-sectional differences across countries in soil characteristics and climate for the period on which GAEZ makes its estimates (1960-1990) have not changed appreciably since 1800. Indeed, while it is possible for soils to degrade over the course of decades, the creation of soils operates on geologic time scales. Similarly, while there have been within-region changes in climate in recent decades, those within-region differences over time are dwarfed by the vast cross-sectional differences in climate across regions that have been persistent for several millennia. In terms of process, we superimpose the hinterlands map on the crop potential maps from GAEZ for eleven storable cereal crops: wheat, sorghum, rye, wetland rice, dryland rice, pearl millet, foxtail millet, oats, maize, barley and buckwheat. For each raster cell, we calculate the potential yield of each cereal in terms of metric tons. We then convert metric tons of potential yield per hectare to kilocalories of potential yield per hectare for each of the

                                                                                                                        12

 

Technical details at http://www.fao.org/fileadmin/user_upload/gaez/docs/GAEZ_Model_Documentation.pdf

32  

eleven crops based on data from the U.S. Department of Agriculture.13 For each cell we then calculate the average kilocalorie of potential yield per hectare across the eleven crops, multiply by the number of hectares per cell, and sum across all cells in each hinterland. Storability Grain storability is severely impacted by temperature and humidity. In the absence of global raster data on humidity (temperature is readily available), we proxy for grain storability based on the potential for cultivating bananas, a crop that tends to grow almost exclusively in hot, humid climates. We derive this data from GAEZ 3.0 as well, summing the total potential yield for bananas in each hinterland. We code as "Tropical" (poor storability) any hinterland with a non-negative potential banana yield (77 of 158 hinterlands, 49% ). All other hinterlands, those having zero potential banana yield, are coded as having high storability. Weather Shocks Conceptually, we think of severe shocks to agriculture as weather conditions leading to crop failure, either due to severe drought or, at the other extreme, due to extreme moisture. We start off with a measure called the Palmer Drought Severity Index (PDSI), devised by Palmer (1965).14 The Palmer index ranges from roughly -6.0 to +6.0, where -4.0 or less is considered extreme drought conditions and +4.0 or higher is considered extreme wet conditions. We retrieved monthly gridded PDSI data from 1850 to 2010 at a resolution of 2.5° by 2.5° from the NOAA15,16.                                                                                                                         13

http://ndb.nal.usda.gov/ndb/ Palmer, W. C, 1965: Meteorological Drought. Res. Paper No.45, 58pp., Dept. of Commerce, Washington, D.C. 15 http://www.esrl.noaa.gov/psd/data/gridded/data.pdsi.html 16 Dai, A., K. E. Trenberth, and T. Qian, 2004: A global data set of Palmer Drought Severity Index for 18702002: Relationship with soil moisture and effects of surface warming. J. Hydrometeorology, 5, 1117-1130. 14

 

33  

For each country in our dataset, we consider the 9-cell block surrounding the largest city in 1800, and operationalize severe shocks to agriculture as the percentage of years between 1850 and 2010 that the average PDSI across all nine cells experiences either extreme drought (PDSI at or below -4.0) or extreme wetness (PDSI at or above +4.0). A Measure of Institutions Our measure for institutions that serve to uphold contracts broadly is called "contractintensive money" or CIM. It is based on citizens' decisions regarding the form in which they choose to hold their financial assets. We follow the basic idea developed in Clague et al (1999)17, but depart from their measure by comparing the value of M2 minus M1, over M2. This gives us the percentage of the money supply held in time deposit accounts and Certificates of Deposit (excluding jumbo CDs). 18  

Section 5: Empirical Results If our theory is correct—that the productive capacity in storable crops of the hinterland of the largest city in the world in 1800, and the frequency with which that hinterland was hit by aggregate shocks, pushed societies down quite different paths of institutional development—then we should be able to detect robust relationships between those characteristics and a range of development outcomes today, such as per capita GDP and the level of democratic consolidation. Further, if we are correct that the level and distribution of                                                                                                                         17

Christopher Clague, Philip Keefer, Stephen Knack, Mancur Olson (1999) "Contract-Intensive Money: Contract Enforcement, Property Rights, and Economic Performance" Journal of Economic Growth 18 We take the raw data from: World Bank, World Development Indicators; International Monetary Fund, International Finance Statistics; World Bank, World Tables 1995; and "Euro area, Outstanding amounts at the end of the period (stocks), Central Government and MFIs (with ECB) reporting sector - Monetary aggregate M2, All currencies combined - Euro area counterpart, Other residents and other general govnt. (2120 & 2200) sector, denominated in Euro, data Working day and seasonally adjusted: Billions of euro, retrieved from http://www.economagic.com/em-cgi/data.exe/ecb/BSI-M-U2-Y-V-M20-X-1-U2-2300-Z01-E-m

 

34  

human capital and institutions that protect property rights mediated between those exogenous characteristics and developmental outcomes today, then it should be the case that: 1) there is a relationship between those characteristics and variables that proxy for human capital and property rights in the past; and 2) there is a relationship between those proxies for human capital and property rights in the past and development outcomes today. We therefore estimate a series of OLS regressions that draw on the variables detailed in Section IV. We can reject the null hypothesis that the distribution of democracy and income across countries is a product of contingent historical events. In point of fact, we find that the climate and geography variables that we detail in Section IV predict 21 percent of the variance in democratic consolidation and 31 percent of the variance in the level of per capita income across the world. These results are robust to potential confounders, such as whether a country has an environment suited to malaria or its colonial heritage. Moreover, the introduction of variables that proxy for human capital and property rights in the past cause our baseline results to attenuate.  

Political Institutions (Democracy) Our theory implies that large hinterlands with the capacity to grow and store cereal crops, and which were not subject to severe, aggregate weather shocks, were more likely to go down paths of institutional development conducive to democratic consolidation. Table 1 displays the relationship between contemporary democratic institutions (as proxied by the average Polity2 Score of a country over the period 1975-2012) against the potential to grow and store cereals as well as propensity for severe weather shocks in the hinterland of that country in 1800. We find a remarkably consistent set of results, which are robust to the inclusion of potential confounders.  

35  

In column 1 we include only two independent variables: the potential yield of storable grains of the hinterland of each countries’ largest city in 1800; a dummy variable for whether the environment is tropical (where storability of grains would be low, even if they were grown); and the interaction of the two variables. In this model, the constant term provides a predicted value for the average Polity 2 score of hinterlands that are non-tropical and where it is virtually impossible to grow grains, which is to say hinterlands that are located in deserts or grasslands. Thus, the constant term in column 1 implies that the countries in which such hinterlands are located would have an average Polity 2 score of 31, which is to say that they are highly autocratic. In this model, the coefficient on cereals potential provides a point estimate for the increase above the constant associated with being in a non-tropical environment where it is possible to grow grains. We find that a 10 percent increase in the potential to grow cereals within the hinterland of the largest city in 1800 is associated with a 1.6 point increase in a state's average Polity score. Moving from the 25th percentile to the 75th percentile of the cereals potential distribution is associated with a 10.0 percentage point increase in the Polity score (Polity is on a 0-100 scale, with a mean of 55 and standard deviation of 30.4). Under conditions of poor storability, the cereals effect is entirely negated. Note also that the adjusted R2 of this regression implies that just two geographic-climatologic variables account for 17 percent of the variance in Polity 2 scores across the globe. Column 2 adds the propensity of a hinterland to be hit by a severe weather shock. The base results remain materially the same as in column 1. We do find, however, that regardless of the average carrying capacity of the land, large scale weather shocks reduce average Polity 2 scores. The coefficient on severe shocks implies that a 10 percent increase in the percentage

 

36  

of years in shock due to either droughts or floods is associated with a 0.75 percentage-point decrease in the average Polity score. Moving from the 25th to the 75th percentile of the severe shocks distribution is associated with a 3.1 percentage point reduction in the average Polity score. The adjusted R2 climbs to 0.21. These results are consistent with our theory: weather shocks that are serially and temporally correlated create an incentive for an Insurance State to emerge, a basic feature of which is a layer of centralized, hierarchical institutions that will tend to concentrate property rights and political power. These results are robust to the inclusion of covariates that proxy for the “resource curse” (column 3, which includes crude oil production from 1945 to 2006), environments that are conducive to malaria (column 4, which includes malaria ecology from Sachs), and the identity of colonizing powers (column 5). These covariates all enter with the expected signs and statistical significance, and they raise the adjusted R2 to 0.34, but they result in only a modest attenuation of the variable on cereals potential and cereals potential interacted with tropics (cereal potential in low storability conditions).                      

 

37  

Table  1  Average  Polity  score  (1975-­‐2012)  associated  with  storable  crop  potential  and  shocks  to  agriculture  (flooding  or   drought).    Polity  rescaled  from  0-­‐100.  OLS  regressions.       Cereals  potential  (log  trillions  kcal)       Tropics  (banana  potential  >  0)       Cereals  X  Tropics  (log  trillions  kcal)       Severe  shocks  (log  %  years  in  shock)       Crude  oil  production  (log  $  per  capita)       Malaria  ecology       Constant     Colonizer  fixed  effects   Observations   2 R   2 Adjusted  R   Standard  errors  in  parentheses   * **  p  <  0.10,    p  <  0.05  

(1)   ** 16.833   (3.283)    

**

26.938   (10.437)  

 

**

   

-­‐18.976   (4.838)    

     

 

     

 

     

 

**

31.097   (6.331)   NO   137   0.172   0.153  

 

Polity  Score  (average  1975  -­‐  2012)   (2)   (3)   (4)   ** ** ** 14.065   13.689   12.994   (3.303)   (3.309)   (3.172)       ** * ** 22.615   20.535   22.340   (10.507)   (10.607)   (10.167)       ** ** ** -­‐15.586   -­‐15.156   -­‐14.447   (4.818)   (4.818)   (4.602)       ** ** ** -­‐7.859   -­‐7.592   -­‐7.052   (2.594)   (2.596)   (2.485)       *   -­‐1.110   -­‐1.637     (0.869)   (0.846)       **     -­‐1.282       (0.329)       ** 8.331   13.208   21.466   (9.692)   (10.395)   (10.196)   NO   NO   NO   134   134   133   0.231   0.241   0.310   0.207   0.211   0.277  

(5)   ** 12.460   (3.061)      

**

-­‐15.120   (4.470)    

**

-­‐7.771   (2.411)    

**

-­‐1.756   (0.817)    

*

-­‐0.658   (0.368)    

*

19.980   (10.297)   YES   133   0.403   0.344  

 

In Table 2 we assess the extent to which the effect of the potential to grow and store cereals on Polity is mediated by human capital and institutions that provide for the broad enforcement of contracts. Column 1 repeats the full specification from Table 3, column 5 as a reference point. We then run the same specification with and without the inclusion of the hypothesized mediating variables, restricting the sample to those observations for which the mediating variables are available to ensure that the results are not driven by a change in the sample. Column 2 shows the results from a specification in which we include the Whipple index in 1910, a measure of human capital based on innumeracy. The coefficient on the Whipple index is negative and significant as expected (greater innumeracy in 1910 is associated with lower average Polity scores, 1975 - 2012) and the coefficient on cereals is mildly attenuated (10.3 in column 2 compared to 12.2 in column 3), indicating that a portion of the cereals effect on democratic institutions is mediated by human capital. The coefficient  

**

24.380   (10.006)  

38  

on severe shocks is basically unchanged by the inclusion of the Whipple index, indicating that human capital is not necessarily mediating the relationship between the volatility of agricultural production and political institutions. In columns 4 and 5 we assess the extent to which property rights act as a mediating mechanism between using contract intensive money in 1973 as a proxy for institutions that uphold contract enforcement. We find that the coefficient on contract intensive money is positive and significant and again attenuates the cereals coefficient (11.7 in column 4 compared to 13.4 in column 5), indicating that a portion of the cereals effect on democratic institutions in mediated by contract enforcing institutions. The coefficient on severe shocks is again unchanged by the inclusion of contract intensive money, indicating that contract enforcement is not necessarily mediating the relationship between the volatility of agricultural production and political institutions. In columns 6 and 7 we report the results from a specification that included both the Whipple index and contract intensive money as potential mediators. The results are qualitatively similar, though the coefficient on the Whipple index is attenuated and does not enter as statistically significant (though this is not surprising given that the Whipple index is measured in 1910 and contract intensive money in 1973).                    

 

39  

Table  2  The  long-­‐run  effect  of  storable  cereals  potential  on  political  institutions  is  partially  mediated  by  human  capital  and   contract  enforcement.    The  inclusion  of  human  capital  in  1910  and  contract  intensive  money  in  1973  attenuates  the   coeffiient  on  Cereals  potential  but  not  severe  shocks  to  agricultural  production.             Cereals  potential  (log  trillions  kcal)       Tropics  (banana  potential  >  0)       Cereals  X  Tropics  (log  trillions  kcal)       Severe  shocks  (log  %  years  in  shock)       Crude  oil  production  (log  $  per  capita)       Malaria  ecology       Innumeracy  1910  (log  Whipple  index)       Contract  intensive  money  1973       Constant     Colonize  fixed  effects   Observations   2 R   2 Adjusted  R  

  No   mediator   (1)   ** 12.460   (3.061)          

**

24.380   (10.006)   **

-­‐15.120   (4.470)   **

-­‐7.771   (2.411)   **

-­‐1.756   (0.817)  

 

*

         

-­‐0.658   (0.368)    

     

 

     

 

*

19.980   (10.297)   YES   133   0.403   0.344  

 

Polity  score  (average  1975-­‐2012)     Innumeracy   Contract  intensive  money   Mediator   Restricted   Mediator   Restricted   sample   sample   (2)   (3)   (4)   (5)   ** ** ** ** 10.328   12.180   11.698   13.374   (3.296)   (3.304)   (2.856)   (3.157)         ** ** ** ** 31.141   29.923   21.950   22.352   (11.069)   (11.366)   (9.012)   (10.027)         ** ** ** ** -­‐15.248   -­‐15.346   -­‐14.253   -­‐14.250   (4.715)   (4.846)   (4.076)   (4.535)         ** ** ** ** -­‐10.539   -­‐11.178   -­‐6.036   -­‐6.656   (2.590)   (2.650)   (2.254)   (2.504)         ** ** -­‐1.304   -­‐1.361   -­‐1.778   -­‐2.034   (0.896)   (0.921)   (0.758)   (0.842)         ** -­‐0.291   -­‐0.545   -­‐0.229   -­‐0.794   (0.412)   (0.411)   (0.348)   (0.367)         ** -­‐19.601         (7.675)               **     68.498         (13.436)           ** ** 107.774   9.039   -­‐3.900   25.435   (40.234)   (11.452)   (11.394)   (10.942)   YES   YES   YES   YES   111   111   118   118   0.475   0.440   0.556   0.445   0.405   0.371   0.500   0.381  

               

Innumeracy  and  CIM   Mediator   Restricted   sample   (6)   (7)   ** ** 10.973   13.870   (3.070)   (3.250)     ** ** 26.290   26.441   (10.214)   (11.108)     ** ** -­‐14.648   -­‐14.761   (4.382)   (4.771)     ** ** -­‐8.260   -­‐9.004   (2.488)   (2.700)     * * -­‐1.525   -­‐1.573   (0.842)   (0.916)     -­‐0.072   -­‐0.545   (0.384)   (0.400)     -­‐12.253     (7.754)       ** 49.108     (16.105)       57.593   16.032   (43.424)   (11.556)   YES   YES   103   103   0.573   0.482   0.505   0.413  

   

Economic Development In this section, we explore the implications of a hinterland’s potential cereal production, its ability to store what was grown, and the effects of severe shocks on levels of economic development. We hypothesize that hinterlands in which cereal crops could be grown and stored, and which were not subject to severe, aggregate weather shocks, tended to develop institutions designed to promote local and regional trade. In these hinterlands, as well, economic agents had incentives to specialize in order to capture gains from trade, which is to say that they had incentives to invest in human capital. We hypothesize that over the long  

40  

run these institutions and incentives would encourage much more rapid economic development. Table 3 shows the basic relationship between GDP per capita in 2010 and the potential to grow and store cereals in the hinterland of the largest city in 1800, as well as the propensity for experiencing aggregate weather shocks to agriculture. We include crude oil production from 1945 to 2006 in order to control for the fact that there are countries, such as Kuwait, Libya, and Saudi Arabia, that have high incomes simply because they produce huge amounts of oil and have very small populations. The empirical evidence provided by column 1 is consistent with our theory in some respects but not all. The coefficient on cereals potential implies that a 10 percent increase in the ability to produce cereals in a non-tropical (easy storage) environment is associated with a 3.7 percent increase in per capita GDP in 2010. Consistent with the theory, the coefficient on the interaction term, cereals potential in low storability conditions in the tropics, is negative. It is not, however, statistically significant. That is, the empirical evidence suggests that cereal potential is associated with long-run economic development, but that lack of storability of cereals in tropical environments does not reduce its impact.19 The coefficient for severe shocks is indistinguishable from zero. The coefficients on all variables of interest are robust to the inclusion of a measure for malaria ecology (column 2) and fixed effects for the identity of colonizers (column 3).

                                                                                                                        19

One explanation for this finding is that our proxy for storability (positive potential for banana cultivation) is only approximately right, and that a cleaner measure based on temperature and humidity would allow for more precise estimation. We are in the process of constructing an alternative storability measure to test this hypothesis.

 

41  

    Table  3  Log  per  capita  GDP  (2010)  associated  with  storable  crop  potential  and  weather  shocks  to  agriculture.    OLS   regressions       Cereals  potential  (log  trillions  kcal)       Tropics  (banana  potential  >  0)       Cereals  X  Tropics  (log  trillions  kcal)       Severe  shocks  (log  %  years  in  shock)       Crude  oil  production  (log  $  per  capita)       Malaria  ecology       Constant     Colonizer  fixed  effects   Observations   2 R   2 Adjusted  R   Standard  errors  in  parentheses   * **  p  <  0.10,    p  <  0.05  

       

(1)   ** 0.367   (0.132)   -­‐0.651   (0.430)   -­‐0.127   (0.197)   0.007   (0.109)   **

       

0.225   (0.035)      

**

7.912   (0.421)   NO   146   0.337   0.313  

-­‐0.463   (0.371)   -­‐0.082   (0.169)   0.033   (0.094)   **

       

0.189   (0.031)    

   

Log  per  capita  GDP  2010     (2)   ** 0.336   (0.114)  

 

**

-­‐0.090   (0.012)   **

8.350   (0.371)   NO   145   0.517   0.496  

(3)   ** 0.310   (0.110)   -­‐0.095   (0.368)   -­‐0.118   (0.165)   0.003   (0.092)   **

0.185   (0.030)      

**

-­‐0.074   (0.014)   **

8.435   (0.368)   YES   145   0.581   0.542  

 

In Table 4 we assess the extent to which the effect of the potential to grow and store cereals on long-run economic development is mediated by human capital and institutions that provide for the broad enforcement of contracts. Column 1 repeats the full specification from Table 3, column 3 as a reference point. We then run the same specification with and without the inclusion of the hypothesized mediating variables, restricting the sample to those observations for which the mediating variables are available to ensure that the results are not driven by a change in the sample. In column 2 we include the Whipple index in 1910; its coefficient is negative and significant as expected (greater innumeracy in 1910 is associated with lower per capita GDP in 2010) and the coefficient on cereals is largely attenuated (0.123

 

42  

in column 2 compared to 0.219 in column 3), indicating that much of the cereals effect on economic development is mediated by human capital. The coefficient on severe shocks is also attenuated by the inclusion of the Whipple index (-0.039 in column 2 compared to -0.082 in column 3) but neither is statistically significant at conventional levels. In columns 4 and 5 we conduct a similar exercise using contract intensive money in 1973 as a proxy for institutions that uphold contract enforcement broadly. We find that the coefficient on contract intensive money is positive and significant, and the cereals coefficient is slightly attenuated (0.257 in column 4 compared to 0.330 in column 5), indicating that a portion of the cereals effect on economic development is mediated by contract enforcing institutions. The coefficient on severe shocks is not attenuated by the inclusion of contract intensive money. In columns 6 and 7 we report the results from a specification that includes both the Whipple Index and contract intensive money as potential mediators. The results are qualitatively similar and suggest that human capital and contract enforcement are independent mediators.

 

43  

Table  4    The  long-­‐run  effect  of  storable  cereals  potential  per  capita  GDP  in  2010  is  mediated  by  human  capital  and  partially   by  institutions  for  contract  enforcement.    The  inclusion  of  human  capital  in  1910  greatly  attenuates  the  coeffiient  on   Cereals  potential  and  the  inclusion  of    contract  intensive  money  in  1973  attenuates  the  coefficient  on  Cereals  potential  to   some  extent.           Cereals  potential  (log  trillions  kcal)       Tropics  (banana  potential  >  0)       Cereals  X  Tropics  (log  trillions  kcal)       Severe  shocks  (log  %  years  in  shock)       Crude  oil  production  (log  $  per  capita)       Malaria  ecology       Innumeracy  1910  (log  Whipple  index)       Contract  intensive  money  1973       Constant     Colonize  fixed  effects   Observations   2 R   2 Adjusted  R   Standard  errors  in  parentheses   * **  p  <  0.10,    p  <  0.05  

  No   mediator   (1)   ** 0.310   (0.110)      

-­‐0.095   (0.368)   -­‐0.118   (0.165)  

 

     

0.003   (0.092)    

**

 

0.185   (0.030)    

**

 

-­‐0.074   (0.014)    

     

 

     

 

**

8.435   (0.368)   YES   145   0.581   0.542  

 

Log  per  capita  GDP  2010       Innumeracy   Contract  intensive  money   Mediator   Restricted   Mediator   Restricted   sample   sample   (2)   (3)   (4)   (5)   * ** ** 0.123   0.219   0.257   0.330   (0.114)   (0.122)   (0.111)   (0.125)         0.033   -­‐0.103   -­‐0.170   -­‐0.196   (0.395)   (0.428)   (0.345)   (0.392)         -­‐0.081   -­‐0.063   -­‐0.127   -­‐0.103   (0.168)   (0.183)   (0.157)   (0.179)         -­‐0.039   -­‐0.082   0.081   0.056   (0.097)   (0.105)   (0.087)   (0.099)         ** ** ** ** 0.167   0.159   0.179   0.168   (0.032)   (0.035)   (0.029)   (0.033)         ** ** ** ** -­‐0.057   -­‐0.073   -­‐0.049   -­‐0.073   (0.015)   (0.016)   (0.013)   (0.014)         ** -­‐1.229         (0.280)               **     2.966         (0.528)           ** ** ** ** 14.745   8.625   7.571   8.842   (1.451)   (0.432)   (0.443)   (0.433)   YES   YES   YES   YES   115   115   118   118   0.664   0.600   0.704   0.614   0.621   0.553   0.667   0.570  

               

Innumeracy  and  CIM   Mediator   Restricted   sample   (6)   (7)   ** 0.136   0.294   (0.116)   (0.132)     -­‐0.048   -­‐0.104   (0.378)   (0.440)     -­‐0.125   -­‐0.113   (0.164)   (0.191)     0.026   -­‐0.013   (0.094)   (0.109)     ** ** 0.154   0.152   (0.032)   (0.037)     ** ** -­‐0.048   -­‐0.071   (0.014)   (0.016)     ** -­‐0.871     (0.295)       ** 2.103     (0.625)       ** ** 12.367   8.843   (1.658)   (0.467)   YES   YES   103   103   0.722   0.613   0.677   0.561  

 

Human Capital In this section we explore the empirical implications of our theory for the investment in broad-based human capital. In Table 5, we measure human capital in 1910 as measured by the Whipple Index, which reflects age-heaping in national census counts. Higher values of the Whipple Index indicate a less numerate population. Column 1 shows the baseline specification, in which the coefficient on cereals potential implies an elasticity of -0.1, such

 

44  

that for a 10 percent increase in cereals potential under conditions of high storability reduces the Whipple Index (innumeracy) by 1 percent. We do not see a reduction of this effect under conditions of low storability as the interaction term of cereals potential and tropics is indistinguishable from zero. In column 2 we add severe shocks, but find no effect on the Whipple index. Column 3 adds a measure for malaria ecology and column 4 adds fixed effects for colonizer identity. The coefficient on cereals potential is robust to both.     Table  5      Innumeracy  associated  with  storable  crop  potential  and  shocks  to  agriculture  (flooding  or  drought).    The  Whipple   index  reflects  "age  heaping"  in  national  censuses:  higher  values  reflect  a  less  numerate  population.  OLS  regressions.       Cereals  potential  (log  trillions  kcal)       Tropics  (banana  potential  >  0)       Cereals  X  Tropics  (log  trillions  kcal)       Severe  shocks  (log  %  years  in  shock)       Malaria  ecology         Constant     Colonizer  fixed  effects   Observations   2 R   2 Adjusted  R   Standard  errors  in  parentheses   * **  p  <  0.10,    p  <  0.05  

   

(1)   ** -­‐0.101   (0.041)   *

0.253   (0.147)  

   

0.013   (0.065)          

 

   

 

    **

5.072   (0.084)   NO   121   0.159   0.137  

   

Whipple  Index  1910  (log)   (2)   (3)   ** ** -­‐0.098   -­‐0.092   (0.043)   (0.042)     * 0.265   0.228   (0.151)   (0.147)     0.005   -­‐0.002   (0.066)   (0.064)     0.008   0.015   (0.037)   (0.036)     **   0.016     (0.005)       ** ** 5.096   5.074   (0.135)   (0.134)   NO   NO   119   118   0.162   0.242   0.132   0.208  

   

(4)   ** -­‐0.082   (0.038)   0.104   (0.136)   -­‐0.013   (0.058)  

       

0.033   (0.033)   **

0.012   (0.005)   **

5.001   (0.126)   YES   118   0.416   0.355  

               

 

45  

Table  6    Amount  of  schooling  (1950)  associated  with  storable  crop  potential  and  weather  shocks  to  agriculture.    OLS   regressions.       Cereals  potential  (log  trillions  kcal)       Tropics  (banana  potential  >  0)       Cereals  X  Tropics  (log  trillions  kcal)       Severe  shocks  (log  %  years  in  shock)       Malaria  ecology       Constant     Colonizer  fixed  effects   Observations   2 R   2 Adjusted  R   Standard  errors  in  parentheses   * **  p  <  0.10,    p  <  0.05  

(1)   ** 0.454   (0.105)    

-­‐0.326   (0.351)  

 

   

-­‐0.236   (0.161)    

     

 

     

 

  0.067   (0.206)   NO   158   0.196   0.180  

Average  years  schooling  1950  (log)   (2)   (3)   ** ** 0.443   0.406   (0.109)   (0.097)     -­‐0.372   -­‐0.203   (0.365)   (0.327)     -­‐0.213   -­‐0.164   (0.165)   (0.147)     -­‐0.034   -­‐0.025   (0.090)   (0.081)     **   -­‐0.072     (0.011)     -­‐0.032   0.219   (0.335)   (0.307)   NO   NO   155   153   0.201   0.383   0.180   0.362  

(4)   ** 0.397   (0.095)    

0.056   (0.329)  

  -­‐0.219   (0.145)     -­‐0.047   (0.080)    

**

-­‐0.057   (0.012)     0.219   (0.308)   YES   153   0.449   0.406  

 

 

In Table 6, we repeat the exercise using average years of schooling in 1950 as our measure of broad human capital accumulation. Column 1 shows the baseline specification in which the elasticity on cereals potential is 0.454, so that a 10 percent increase in cereals potential under conditions of high storability is associated with a 4.5 percent increase in average years of schooling in 1950. Under conditions of low storability the elasticity is reduced by 0.24, cutting the effect in half, though the coefficient is not statistically significant. In column 2 we add severe shocks, but find no effect on average years of schooling in 1950. Column 3 adds a measure for malaria ecology and column 4 adds fixed effects for colonizer identity. The coefficient on cereals potential is robust in both. Finally, in Table 7 we conduct a parallel analysis using average years of schooling in 2010 as our measure of accumulated human capital. Column 1 shows the baseline specification in which the elasticity on cereals potential is 0.12, so that a 10 percent increase  

46  

in cereals potential under conditions of high storability is associated with a 1.2 percent increase in average years of schooling in 2010. Under conditions of low storability the elasticity is reduced by 0.08, cutting the effect to just one third, though the coefficient is not statistically significant. In column 2 we add Severe Shocks, but find no effect on average years of schooling in 2010. Column 3 adds a measure for malaria ecology and column 4 adds fixed effects for colonizer identity. The coefficient on cereals potential under conditions of high storability is robust in the first and slightly attenuated in the second.

Table  7  Amount  of  schooling  (2010)  associated  with  storable  crop  potential  and  shocks  to  agriculture  (flooding  or  drought).     OLS  regressions.       Cereals  potential  (log  trillions  kcal)       Tropics  (banana  potential  >  0)       Cereals  X  Tropics  (log  trillions  kcal)       Severe  shocks  (log  %  years  in  shock)       Malaria  ecology       Constant     Colonizer  fixed  effects   Observations   2 R   2 Adjusted  R   Standard  errors  in  parentheses   * **  p  <  0.10,    p  <  0.05  

(1)   ** 0.115   (0.051)      

-­‐0.197   (0.172)   -­‐0.078   (0.077)  

 

     

     

     

 

**

1.911   (0.103)   NO   133   0.150   0.130  

 

Average  years  schooling  2010  (log)   (2)   (3)   ** ** 0.114   0.104   (0.055)   (0.044)     -­‐0.219   -­‐0.049   (0.177)   (0.144)     -­‐0.070   -­‐0.064   (0.079)   (0.064)     -­‐0.004   -­‐0.010   (0.045)   (0.036)     **   -­‐0.045     (0.005)     ** ** 1.902   1.955   (0.160)   (0.132)   NO   NO   132   131   0.152   0.461   0.126   0.440  

(4)   * 0.080   (0.041)      

0.084   (0.137)   -­‐0.059   (0.060)  

  -­‐0.034   (0.034)    

**

-­‐0.037   (0.006)    

**

1.973   (0.126)   YES   131   0.576   0.537  

 

 

 

47  

Section 6: Conclusion In this paper, we identify two features of climate and geography that shaped the fundamental political, economic, and social institutions of modern nation-states; these natural forces therefore shaped the divergence in their paths of development throughout history and the wide disparities of wealth, democracy and education we observe around the world today. The first of these factors is the potential amount of food energy that could be produced and stored in the economic hinterland of the largest city in each country in 1800, and the second is the frequency with which that hinterland was subject to weather shocks severe enough to wipe out all producers simultaneously. We develop a theory that links both factors to modern-day outcomes through first principles of economics, politics, and basic human survival, and show that these two factors account for a surprising amount of the variance in regime type, per capita GDP, and education levels across the world today. We introduce a measure of hinterland size based on a model of physical transportation costs in which friction and gravity jointly determine variation in the extent of a city’s economic hinterland. We then estimate the potential caloric yield of storable cereal crops for each hinterland based on soil and climate characteristics, and estimate high/low storability conditions based on these characteristics as well. We show that in high storability environments, the potential caloric yield of storable cereal crops in a country’s hinterland in 1800 is moderately correlated with the country’s current levels of development, democracy and education. The evidence is consistent with our framework, in which we argue that expansive hinterlands with high potential for growing and storing rain-fed grains were unique in that they tended to support societies of independent agricultural producers whose basic political economy revolved around specialization and trade.  

48  

We also introduce a measure of potential agriculture volatility based on climate conditions characterized by frequent and severe droughts and floods. We show that such climates also are moderately correlated with a reduction in current levels of development, democracy, and education. This evidence is also consistent with our framework, in which we emphasize that basic survival needs in such climates pushed societies to develop states characterized by robust and centralized states that served as insurance mechanisms in the face of volatility in agricultural production. The initial empirical evidence presented in this paper stands to improve as we continue to refine our measures in several respects. First, we are in the process of completing the historical coding of inland water navigability for large swaths of South Asia, East Asia, New Zealand and Australia. Completing this dataset stands to improve the precision of our measure hinterland size for future analyses. Second, we are developing a measure for grain storability based directly on temperature and moisture conditions rather than a proxy based on potential banana cultivation. Third, we are refining our measure of agricultural volatility to better capture variation in the temporal extent of droughts and floods. Our current measure is well-suited to capturing spatial correlation and frequency, and we are amending it so that it also captures the fact that seven consecutive years of plenty followed by seven consecutive years of drought is quite distinct from a climate that alternates bumper crops and crop failures on shorter cycles. Finally, since this is a project that takes history seriously, we are also in the process of gathering and synthesizing historical evidence for the processes implied by the theoretical framework in order to build on the illustrative examples provided in this paper.  

 

49