download - quarterly journal of economics
Post on 09-Feb-2022
1 Views
Preview:
TRANSCRIPT
THE POTATO’S CONTRIBUTION TOPOPULATION AND URBANIZATION: EVIDENCE FROM
A HISTORICAL EXPERIMENT∗
NATHAN NUNN AND NANCY QIAN
We exploit regional variation in suitability for cultivating potatoes, togetherwith time variation arising from their introduction to the Old World from theAmericas, to estimate the impact of potatoes on Old World population andurbanization. Our results show that the introduction of the potato was re-sponsible for a significant portion of the increase in population and urbaniza-tion observed during the eighteenth and nineteenth centuries. According to ourmost conservative estimates, the introduction of the potato accounts for approx-imately one-quarter of the growth in Old World population and urbanizationbetween 1700 and 1900. Additional evidence from within-country comparisonsof city populations and adult heights also confirms the cross-country findings.JEL Codes: J1, N1N5, O14.
I. INTRODUCTION
Between 1000 and 1900, world population grew from under300 million to1.6 billion, and the share of population living in ur-banareas morethanquadrupled, increasingfromtwotooverninepercent. As shown in Figure I, the increase accelerated dramati-callyovertimeandoccurredalmost entirelytowards theendoftheperiod. The determinants of these phenomena have been of muchinterest toeconomists, demographers, and historians alike.1 Thisstudy uses country-level historical data on population and urban-izationtoempiricallyinvestigatetheextenttowhichthishistorical
∗We thank the editor Larry Katz and five anonymous referees for commentsthat significantly improved the paper. We also thank David Canning, GregoryClark, Angus Deaton, Azim Essaji, Andrew Foster, Oded Galor, Claudia Goldin,Mikhail Golosov, Jonathan Hersh, Wolfgang Keller, Joel Mokyr, AlejandroMolnar, Petra Moser, Jean-Laurent Rosenthal, Yona Rubinstein, Andrei Shleifer,Peter Temin, Hans-Joachim Voth, and David Weil for valuable feedback anddiscussions. We also thank seminar participants at Bocconi University, BostonUniversity, Brown University, Harvard University, MIT, Ohio State Univer-sity, Princeton University, Stanford University, Tufts University, University ofChicago School of Business, BREAD, CEA Annual Meetings, ISNIE Annual Con-ference, NBER SI, and NEUDC Annual Conference. We also thank Sayon Deb forexcellent research assistance.
1. For studies in the growth literature that have focused on the link betweenpopulation increase and factors such as per capita incomes, see Galor and Weil(2000), Jones (2003), and Voigtlander and Voth (2006). For micro-level studies ofthe determinants of increased life expectancy, see the literature review providedby Cutler, Deaton, and Lleras-Muney (2006).
c© The Author(s) 2011. Published by Oxford University Press on behalf of President andFellows of Harvard College. All rights reserved. For Permissions, please email: journals.permissions@oup.com.The Quarterly Journal of Economics (2011) 126, 593–650. doi:10.1093/qje/qjr009.Advance Access publication on June 27, 2011.
593
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
594 QUARTERLY JOURNAL OF ECONOMICS
FIGURE IGrowth in World Population and Urbanization, 1000–1900
increase is due to the introduction of potatoes from the NewWorld to the Old World, by which we mean the entire EasternHemisphere.
Potatoes provide more calories, vitamins, and nutrients perarea of land sown than other staple crops. The potato’s role inincreasing population and promoting economic development hasbeen a subject of much discussion amongst scholars across a vari-ety of disciplines (e.g., Langer 1963; McNeill 1948, 1999; Salaman1949). Forexample, historianWilliam Langer (1963, p. 14) arguesthat within Europe, “the spread of the potato culture everywherecorrespondedwiththerapidincreaseof population.”Potatoes dra-matically improved agricultural productivity and provided morecalories and nutrients relative to preexisting Old World staples.In The Wealth of Nations, Adam Smith extols the advantages ofpotatoes over preexisting staples in Europe, writing that “the foodproduced by a field of potatoes is . . . much superior towhat is pro-duced by a field of wheat. . . . No food can afford a more decisiveproof of its nourishing quality, orof its being peculiarly suitable tothe health of the human constitution” (Smith 1776, pp. 67–68).2
2. Other historians have attributed even greater significance to the potato.McNeill (1999, p. 67), inanarticle titled“HowthePotatoChangedWorldHistory,”argues that “potatoes, byfeedingrapidlygrowingpopulations, permittedahandfulof Europeannations toassert dominationovermost of theworldbetween1750 and
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 595
Similar observations have been made outside of Europe. A partic-ularly interesting example comes from anthropologist ChristophvonFurer-Haimendorf (1964), whoargues that theintroductionofthe potato into Nepal significantly increased food production andagricultural surplus. He writes that “the population of Khumbuwas a fractionof its present sizeuntil themiddleof thenineteenthcentury and there can be no doubt that the great increase of thelast hundred years coincided with the introduction and spreadof the potato” (pp. 9–10).3 Another example is from the famousJapanese scholar Takano Choei, who wrote of the benefits of thepotato in his 1836 treatise Ni butsu ko. He argued that extensivecultivation of potatoes would cure many social ills of the empireby alleviating food demands from a growing population (Laufer1938, p. 83).
Despite qualitative accounts of the benefits of the potato tothe Old World, empirical evidence quantifying the overall impactofthepotatois scarce. This is nodoubt partlyduetotheestimationdifficulties caused by the endogeneity of potato adoption. First,there is an issue of reverse causality. The adoption of potatoesmight have caused population growth, but alternatively, latentpopulation pressure and the associated demand for food mighthavecausedtheadoptionof potatoes. A numberof historians haveargued for the latter relationship (e.g., Salaman 1949; Connell1962; Cullen 1968). The second problem is joint determination.Bothpopulationgrowthandtheadoptionofnewagricultural tech-nologies can be the outcome of a third unobserved factor.
Tothebest ofourknowledge, theonlyexistingempirical studyattempting to estimate the causal effect of potatoes is a study byJoel Mokyr(1981), inwhichheexamines therelationshipbetweenpotato adoption and population growth across Irish counties in1845. He addresses endogeneity issues by estimating a system oftwo equations using 2SLS, instrumenting for potato cultivation
1950.” He continues, writing that “an essential—but by nomeans the only—factorexplaining the surprising rise of the west . . .was the extra food that potato fieldsmade available to the peoples of northern Europe. It is certain that without pota-toes, Germany could not have become the leading industrial and military powerof Europe after 1848, and no less certain that Russia could not have loomed sothreateningly on Germany’s eastern border after 1891” (p. 82).
3. Von Furer-Haimendorf also argues that the increase in agriculturalsurplus from the potato was responsible for the sophisticated Buddhist civiliza-tion that developed in the Sherpa Khumbu region of Eastern Nepal (von Furer-Haimendorf 1964, pp. 10–11).
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
596 QUARTERLY JOURNAL OF ECONOMICS
with per capita income (intended to capture the demand forpotatoes), thecapital tolaborratio(intendedtocapturethesupplyof manure), the standarddeviation of altitude, andthe proportionof land classified as “improvable” for tillage but not currently un-der cultivation (both of which are intended to capture geographicfeatures that increased potato adoption). Mokyr finds that potatocultivation resulted in a statistically significant increase in popu-lation growth. He finds no evidence of the reverse causalrelationship—that the potato was adopted in response to rapidpopulation growth.4
Theprincipal contributionofourstudyis toprovidearigorousquantitative analysis of the historical role of potatoes in increas-ing population and urbanization. We expand the scope of Mokyr’sanalysis by examining the effect of potatoes on the entire OldWorld during the eighteenth and nineteenth centuries and by ex-aminingurbanization, citygrowth, andadult height inadditiontopopulation. Our estimation strategy exploits two sources of vari-ation. The first is time variation arising from the introduction ofpotatoes as a field crop in the Old World. Potatoes did not exist inthe Old World for most of history. They are native toSouth Amer-ica and were first discovered by Europeans during the voyages ofColumbus. Potatoes were widely adoptedas a fieldcropin Europetowards the end of the seventeenth and beginning of the eigh-teenth centuries. Their cultivation then spread to the rest of theOld World, mainly through European sailors and missionaries.The second source of variation is cross-sectional and arises fromdifferences in countries’ suitability for cultivating potatoes, as de-terminedbytime-invariant geoclimaticconditions. Conditional onaccess to potatoes, regions that are more suitable for potato cul-tivation will be able to increase food production more. Our iden-tification strategy relies on the interaction of the two sources ofvariation, and only the interaction can be interpreted as plausi-bly exogenous. Our strategy, similar in spirit to a differences-in-differences (DD) strategy, compares population and urbanizationlevels between Old World countries that were more suitable forpotato cultivation to regions that were less suitable, before andafter potatoes were adopted in the Old World.
4. Additional evidence about the effects of the potato, although not causallyidentified, comes from Baten and Murray (2000). The study examines the deter-minants of the heights of 4,100 male and female prisoners from Bavaria between1856 and 1908. One of their findings is that prisoners from regions with greaterpotato production were taller.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 597
Our strategy shares most of the advantages and disadvan-tages of a standard DD strategy. On the one hand, it allows usto control for both country and time-period fixed effects so thatall time-invariant differences across countries—such as geogra-phy, food preferences, or institutions (to the extent that theychange slowly over time)—and secular changes over time—suchas global improvements in health, sanitation, and technologicaladvancements—are controlled for. On the other hand, the strat-egy relies on there being no other shocks occurring around thesame time that potatoes were adopted that are correlated withcountries’ suitability for potato cultivation. We address this iden-tification concern by directly controlling for time- and country-varying factors that might bias our estimates. In the baselineestimates, we control for potentially important characteristics,eachinteractedwiththefull set of time-periodindicatorvariables.This allows the effect of each factor tovary flexibly over time. Thecharacteristics include a country’s suitability for cultivating theOldWorldstaplecrops wheat andrice, as well as threegeographiccharacteristics that are correlated with potato suitability: terrainruggedness, elevation, and the presence of a tropical climate. Wealso control for a number of characteristics that have been identi-fied as being important for historical growth and development.
We find that Old World regions that were suitable for potatocultivation experienced larger increases in population and urban-izationaftertheintroductionofpotatoes. Theestimates arerobustto a number of sensitivity checks, which include controlling for alarge number of alternative determinants of population and eco-nomicgrowth. These include legal origin, identity of the colonizer,the prevalence of disease (measured as distance from the equatorand potential prevalence of malaria), distance from the coast, ahistory of Roman rule, the prevalence of Protestantism, being anAtlantic trader, and the historical volume of the slave exports. Inaddition, weshowthat weobtainsimilarresults whenweexaminevariationacross countries withincontinents. Together, theresultssuggest that it is unlikely that our findings are simply capturingother historical determinants of economicdevelopment, includingthose factors that led to the divergence of Europe from the rest ofthe world.
To assess the magnitudes of our estimated effects, we calcu-late how much of the increase from 1700 to 1900 in Old Worldpopulation and urbanization can be explained by the introduc-tion of the potato. Our baseline estimates suggest that the potato
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
598 QUARTERLY JOURNAL OF ECONOMICS
accounts for approximately 25–26 percent of the increase in totalpopulation and 27–34 percent of the increase in urbanization.
To verify our cross-country estimates, we use our estimationstrategy to examine within-country variation in city populationgrowth and adult heights (a conventional measure of the nutri-tional investments made during an individual’s years of growth).Looking within countries, we find that the adoption of potatoesspurred city growth: more suitable locations experienced fasterrelative population growth after the diffusion of the potato. Usinginformation from military records of French soldiers born in theseventeenth andeighteenth centuries, we showthat potatoes alsoincreased adult height. Our estimates suggest that for villagesthat were fully suitable for potato cultivation, the introductionof the potato increased average adult heights by approximatelyone-half inch.
Thesefindings contributetoseveral existingliteratures. First,our findings add to the debate about the importance of improve-ments in nutrition in explaining the historical decline in mortal-ity and increase in life expectancy (e.g., Fogel 1984, 1994, 2004;McKeown 1976; Livi-Bacci 1991). Our results provide evidenceof a causal effect of access to improved nutrition on populationgrowth. Second, our findings alsocontribute tothe understandingof the relationship between agricultural productivity and aggre-gate economic growth. Because of reverse causality and omittedvariables bias, identifying the causal link between agriculturalproductivity and economic growth is difficult.5 Our findings pro-vide evidence that increased agricultural productivity can lead toincreased economic growth, as measured by urbanization rates.
Finally, our findings contribute tothe historical debate aboutthe importance of the Columbian Exchange for Old World livingstandards. The traditional view has been that the period of glob-alization following the discovery of the New World did not havesignificant effects on Old World living standards until the earlynineteenth century (e.g., O’Rourke and Williamson 2002). Thishas been challenged by recent studies which, for example, havefound that the growth of Atlantic traders such as Great Britainaccelerated relative tothe rest of Europe starting in the sixteenthcentury (Acemoglu, Johnson, and Robinson 2005), or that the in-creased availability of certain commodities such as sugar, tea,
5. Existing empirical studies, such as Tiffin andIrz (2006), have attemptedtoinfer the effect of agriculture on economic development by examining panel dataand employing lags and conducting Granger causality tests.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 599
coffee, and tobacco increased the welfare of the English popula-tion prior to the nineteenth century (Hersh and Voth 2009). Ourfinding that potatoes positively affectedpopulation starting in theeighteenth century is consistent with the more recent view.
The paper is organized as follows. Section II provides a dis-cussion of the nutritional and historical background of potatoes,while Section III outlines the conceptual framework. Section IVdescribes the data used in the analysis. Section V presents theempirical strategy and baseline results. Section VI reports addi-tional robustness checks, including our within-country estimatesexamining city populations and adult heights. Section VII offersconcluding remarks.
II. BACKGROUND
II.A. Virtues of the Potato
From a nutritional standpoint, potatoes were superior topre-existing staple crops because they provided more vitamins andnutrients and they provided a greater supply of calories. Becausepotatoes containnearlyall important vitamins andnutrients, theysupport life better than any other crop when eaten as the solearticle of diet (Davidson et al. 1975; Reader 2008). Humans canhave healthy diets from consuming potatoes, supplemented withonly dairy, which contain the two vitamins not provided for bypotatoes, vitamins A and D (Connell 1962; Davidson et al. 1975).6
Historically, this was the typical Irish diet, which althoughmonotonous, was able toprovide sufficient calories, vitamins, andnutrients (Connell 1962; Burton 1948, p. 189).
According to the U.S. Department of Agriculture (2007), amedium potato (150 grams/5.3 ounces) with the skin provides29.55 milligrams of vitamin C (45 percent of the daily value [DV]).This is important since other staple crops such as wheat, oats,barley, rice, and maize do not contain any vitamin C, a necessarydeterrent for scurvy. For much of the Old World, the potato pro-videdthe only source of vitamin C andprotection against scurvy.7
6. Dairy is not actually necessary for vitamin D because humans produce itafter absorbing sunlight.
7. As an example, the average Irish diet of 4.5 to 6.5 kilograms of potatoesper day provided forty tosixty times the quantity of vitamin C required topreventscurvy (Hughes 2000). An alternative source of vitamin C in the Old World wasturnips(althoughpotatoesprovidemorevitaminCthanturnips). Turnipswerealsorelatively hardy in coldweather. Some have arguedthat they playedan important
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
600 QUARTERLY JOURNAL OF ECONOMICS
TABLE IAVERAGE CROP YIELDS OF ENGLISH FARMS IN THE EIGHTEENTH CENTURY
Energy value Acres of land neededAverage yield per acre of crop to provide 42 megajoules perBushels Kilograms Megajoules day for one year
Wheat 23 650 8,900 1.70
Barley 32 820 11,400 1.40
Oats 38 690 9,300 1.60
Potatoes 427 10,900 31,900 0.50
Notes. Data are from eighteenth-century England, recorded in Young’s (1771, p. 20) The Farmer’s Tourthrough the East of England, Volume 4; reproduced in Davidson et al. (1975).
A medium potato also contains 632 milligrams of potassium (18percent of DV), 0.44 milligrams of vitamin B6 (20 percent of DV),as well as significant amounts ofthiamin, riboflavin, folate, niacin,magnesium, phosphorus, iron, and zinc. Moreover, the fiber con-tent of a potato with skin (3.5 grams) is similar to that of manyother cereals such as wheat.
The second benefit of potatoes is that relative to Old Worldstaples, theyrequireless landtoproducethesameamount of calo-ries (Connell 1951, p. 391; Langer 1963, pp. 11–12). Historical ev-idence of the caloricsuperiority of the potatoover preexisting OldWorld crops is shown in Table I, which reports data collected inArthur Young’s (1771) surveyof farmingcommunities throughoutEngland in the 1760s. The first twocolumns compare the averageyields of oats, wheat, barley (three Old World staple crops), andpotatoes. It shows that yields (measured in either bushels or kilo-grams) are well over ten times higher for potatoes relative to theother crops. To adjust for the fact that potatoes are 75–80 per-cent water and therefore naturally heavier and more bulky thanthe other crops, the third column compares the energy value ofthe yields reported in the first two columns. It shows that an acreof potatoes yields approximately three times more energy thanan acre of each of the other crops. The final column shows thenumber of acres required to provide the total energy needs for afamily of twoadults and three young children, which is estimatedto be 42 megajoules (or approximately 10,000 calories) per day.
role in providing nutrition for places such as England (Timmer 1969). However,relative to a potato, the turnip provides fewer nutrients, and more importantly, itprovides less than one-quarter the calories (U.S. Department of Agriculture 2007).
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 601
While this family couldsubsist by cultivating a plot of only half anacre of potatoes, it would need to cultivate about 1.5 acres—threetimes as much land—if it were togrowwheat, oats, or barley. Thedata from Table I confirm historical reports that a single acre ofland cultivated with potatoes and one milk cow was nutritionallysufficient for feeding a large family of six to eight (Langer 1963;McNeill 1999).
Although there is a consensus among historians that pota-toes required less land to produce the same amount of calories,it is less clear whether potatoes required more or less labor thanOld World cereals. Turner (1996, Ch. 6) reports historical laborrequirements data from an Irish tenant farm collected between1837 and 1885. In the sample, potatoes required approximately2.5 times more labor input per acre cultivated than wheat, oats,or barley. However, the figure does not account for the fact thatpotatoes yieldthreetimes morecalories peracre(recall thefiguresfrom Table I). Therefore, in terms of labor per calorie harvested,potatoes appear to be comparable or even better than cereals.
An additional benefit of potatoes was that their cultivationdid not require a complete switch away from the cultivation ofOld World staples; it was possible to plant potatoes between thegrowing seasons of other crops. Potatoes could be planted on theland that was otherwise left fallow between the periods of graincultivation (Mokyr 1981; McNeill 1999). McNeill (1999, p. 79) de-scribes this benefit of potatoes, writing that “by planting potatoeson the fallow, and using hoes to eliminate weeds, there was ab-solutely no need to decrease the grain supply! What a bonanza!”One shortcoming of the potatowas that unlike rotation crops suchas clover and legumes, it did not increase the supply of nitrogenin the system, which was a constraining factor for agriculturalproductivity at the time (Allen 2008; Chorley 1981).
Potatoes alsoprovidedindirect benefits. Being relatively easyto store, potatoes provided excellent fodder for livestock (primar-ily pigs and cattle), especially through the winter. Often, a signif-icant proportion of the potato crop would be used as fodder.8 Thismeant that potatoes also increased meat consumption, as well asmanure, which was a valuable input for cropproduction.9 Finally,potatoes, as well as other crops introduced over time, provided
8. See O Grada (2006, p. 8) for evidence from nineteenth-century Ireland.9. For example, according to nineteenth-century data presented in Burton
(1948, pp. 84–85), manuring increased the yield of potatoes per acre by over 150percent.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
602 QUARTERLY JOURNAL OF ECONOMICS
additional means of crop diversification, which reduced vulnera-bility to famine (O Grada 2009, p. 35).
II.B. The Potato’s Diffusion from the New World to the Old World
Archeological evidence suggests that the potato was first cul-tivated in the Andes between seven thousand and ten thousandyears ago (Messer 2000b). Pre-Columbian cultivation occurredwithin modern-day Colombia, Ecuador, Peru, Bolivia, Chile, andNorthern Argentina (Glendinning 1983, pp. 79–80). Althoughparts ofMexicoandtheEasternportionofNorthAmericaaresuit-able for potato cultivation, the historical evidence suggests thatthepotatonevermigratedtotheseareas until afterEuropeancon-tact (Ugent 1968).
For the Old World, the potato was discovered along with thediscovery of the Americas. The first evidence of potatoes beingconsumedinEuropeis fromSeville, Spain, wheretherearerecordsof potatoes being purchased by a hospital in 1573 (Salaman 1949,pp. 68–69). After Spain, the potato next appeared in Italy in1586, then England in 1596 and Germany in 1601 (Brown 1993,pp. 363–364).
Despite the benefits of potatoes, their widespread adoptiondid not follow immediately after their first appearance at the endof the sixteenth century. For most of the seventeenth century,potatoes remained a botanical curiosity in Europe. Several char-acteristics of the potato prevented its immediate adoption. First,because the potato belongs to the poisonous nightshade family,it was initially presumed to be poisonous. Second, the similaritybetween the appearance of the potato—discolored and lumpy—and the flesh of those afflicted with leprosy led many to believethat potatoes causedthedisease(Brown1993; Langer1975).10 Fi-nally, because many characteristics of the potatowere unique andpreviously unobserved prior to its arrival, these also caused sus-picion and apprehension. Salaman (1949, p. 112) describes these,writing that “the potato was the first edible plant in Europe tobe grown from tubers and not from seed, and till then, no similarplant . . . was grown which bore on underground stems numerouswhite or flesh coloured nodules; both the cultivation, behaviourand the habits of the plant were unusual.”
10. The tendency tolink plants todiseases based on their outward appearancewas a common form of reasoning, called the “Doctrine of Signatures,” at the time(Salaman 1949, pp. 109–121).
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 603
The eventual diffusion of potato cultivation was gradual anduneven. Adoption first began in the late seventeenth century byIrish peasants. Also in the late seventeenth century, potatoes be-gan to be cultivated on the European mainland in East and WestFlanders and in Alsace (Vandenbroeke 1971, pp. 17, 20–21). Bytheearlyeighteenthcentury, potatocultivationhadalsospreadtothe Scottish Highlands, as well as toparts of England and France(Langer 1963, pp. 12–13; Laufer 1938, pp. 60–61).11 In Scandi-navia, the cultivation of the potato lagged behind the rest of Eu-rope. Cultivation in Sweden and Norway did not begin until themid-eighteenth century (Laufer 1938, pp. 67–68).
FromEurope, thepotatowas spreadacross therest of theOldWorld by mariners who carried potatoes to ports in Asia, Africa,and Oceania. The potato was probably introduced to China onseveral different occasions during the seventeenth century. It wascultivatedas earlyas 1603 byDutchsettlers ofthePenghuIslands,andlater in Taiwan after the Dutch occupiedthe islandfrom 1624to1662. Given the Dutch initiation of trade links between Taiwanand the coastal province of Fujian, it is likely that the potato wasalso introduced to mainland China during this time. According toLee (1982, p. 738), by 1800 the farmers in Southwest China hadreplacedthetraditional loweryieldcrops ofbarley, oats, andbuck-wheat with either potatoes or maize, another New World crop.12
The potato first reached India not long after it arrived inEurope, introduced by either the British or the Portuguese. Theearliest known reference topotatocultivation in India is a writtenaccount fromthe1670s byJohnFryer(Laufer, 1938, p. 91). Bythelate eighteenth century there are various accounts of widespreadcultivation in many parts of India (Pandey and Kaushik 2003;Stuart 1923, p. 381). The introduction of potatoes to Africa is notparticularly well documented. Existing accounts suggest thatpotatoes arrived later than in other parts of the world, aroundthe end of the nineteenth century. In Ethiopia, the potato wasintroduced in 1858 by a German immigrant named Wilhelm
11. We are unable to provide an exhaustive summary of the historical adop-tion in different parts of the Old World here. We have summarized the existingevidence in an unpublished Online Appendix that accompanies the paper. Thisis available from the authors’ web pages. Solar (1997, pp. 114–117) also providesa comprehensive overview of the pattern of adoption within Europe by the mid-nineteenth century.
12. In our analysis, we explicitly control for the effects of maize and other NewWorld food crops.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
604 QUARTERLY JOURNAL OF ECONOMICS
Schimper. Subsequent adoption by native farmers occurred grad-ually over a period of several decades.
Thehistorical evidencesuggests that otherthantheconsump-tion of the potatoin Seville, Spain in 1573, widespreadcultivationdid not begin until the late seventeenth and early eighteenth cen-turies. Over the next 150 years, cultivation gradually increased,both in geographic coverage and in intensity. Our analysis usesthe beginning of the widespread cultivation of potatoes as ourpostadoption cutoff period. Based on this, we take our post-treatment periods to be observations after 1700, which include1750, 1800, 1850, and 1900. We do not take this cutoff for gran-ted. In the empirical analysis, we check for patterns in the datadirectly. We find that consistent with the historical record, potatosuitable regions begin to experience systematic increases in pop-ulation and urbanization after 1700. We also find that consistentwith the gradual diffusion of cultivation, the strength of the rela-tionship monotonically increases between 1750 and 1900.13
II.C. Other New World Staple Crops
After the discovery of the Americas, other New World cropswere also introduced along with potatoes. These include maize,cassava, tomatoes, chili and bell peppers, cacao, sunflowers, andthe sweet potato (Grennes 2007; Nunn and Qian 2010). Of these,the three crops that became high-caloric staples in the Old Worldare maize, cassava, and the sweet potato.
Maize is unable to rival potatoes in terms of nutrients orcalories. It produces significantly fewer calories per acre of land.Moreover, humans are unable to subsist on a diet that is too con-centratedinmaize. Significant consumptionof maizeis associatedwith pellagra, a disease caused by niacin deficiency. The effectsof pellagra include skin, digestion, mental disorders, and, if un-treated, eventual death. The disease was first observed in the1730s in Italy and even today continues toaffect poor populationswith diets that rely too heavily on maize. A second adverse effectof a corn-heavy diet is protein deficiency (Messer 2000a).
Sweet potatoes are also nutritious and produce similaramounts of calories per acre of land as potatoes, but they differfrom potatoes in two important ways. First, the archaeological
13. For urbanization, the monotonic increase is delayed by fifty years, and theincrease is weaker in the early periods. See Section V.B. for a full discussion.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 605
evidence suggests that sweet potatoes, transported by Polynes-ians, reached the Old World long before the European discoveryof the NewWorld. For many countries in our sample, their impactwould have been felt as early as 1000 (Hather and Kirch 1991).Second, a close substitute to the sweet potato, the yam, alreadyexisted in the Old World (O’Brien 2000). Yams are broadly simi-lar to sweet potatoes in terms of both nutritional content and therequirements for cultivation. Many regions that were suitable forcultivating sweet potatoes had already cultivated yams when theformer were introduced.
The New World staple, cassava, which is also called maniocor yuca, alsoprovides abundant calories. But its deficiency in pro-teinandotherimportant nutrients causes it tobealess “complete”food than potatoes (Cock 1982). In addition, because cassava con-tains toxiccyanogenicglycosides (e.g., cyanide), failuretoproperlyprepare cassava causes konzo, a neurological disease that causesparalysis.
In Section V.D., we showthat the estimates are robust tocon-trolling for the potential effects of these other New World staplecrops on population and urbanization.
III. CONCEPTUAL FRAMEWORK
Our analysis examines the effect of the positive agriculturalproductivity shock from the introduction of the potato on popu-lation and urbanization levels. The way an increase in agricul-tural productivity can affect a population is straightforward: anincrease in food productivity increases living standards, causingfertility and life expectancy to increase, both of which result inincreased population.
The effect of potatoes on urbanization might occur through anumber of different channels. The first potential channel arisesbecause a shock to agricultural productivity changes the relativereturns to agriculture and industry, affecting the number of peo-ple that work in the agricultural sector and live outside cities. Toillustrate the conditions under which a shock to agricultural pro-ductivity can increase urbanization, we develop a simple modelwhere the introduction of potatoes is interpreted as an increasein the productivity of agricultural production.
Individuals choose to work in one of two sectors: agriculture,which takes place in the countryside, and manufacturing, whichoccurs in cities. Let LF and LW denote the number of farmers
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
606 QUARTERLY JOURNAL OF ECONOMICS
employed in agriculture and the number of workers employed inmanufacturing. Total labor is normalized to 1. Each worker inthe manufacturing sector produces one unit of the manufacturinggood, and each agricultural farmer produces e units of the agri-cultural good. Preferences over the twocommodities are the samefor both types of workers and are given by Ui = 1
1−σ
(ci
A
)1−σ+ ci
M,where i denotes farmers or workers and ci
A and ciM denote con-
sumption of agricultural and manufactured goods. The price ofthe agricultural good relative to the manufacturing good is givenby p.
Farmers maximize utility subject to their budget constraint:
max1
1− σ
(cF
A
)1−σ+ cF
M subject to pcFA + cF
M = pe.
Similarly, the worker’s problem is
max1
1− σ
(cW
A
)1−σ+ cW
M subject to pcWA + cW
M = 1.
The first-order conditions from the farmer and worker’s problemsyield
(1) cFA = cW
A = p−1/σ.
Free mobility and labor market clearing require agents to be in-different between working in either sector, and therefore, pe = 1.Market clearing for the agricultural product requires
(2) cFA + cW
A = eLF.
Substituting (1) into (2) and using the fact that pe = 1 givestheequilibriumlaborforce(andlaborshare) employedintheagri-cultural sector:
LF = 2e1/σ−1.
Therefore, the share of labor employed in agriculture is decreas-ingintheproductivityof agricultural production e iffσ > 1, whereσ measures the inverse of the price elasticity of demand for theagricultural good: 1/σ = −∂cA/cA
∂p/p .14 A value of σ > 1 means that a
14. To see this, note that total consumption of the agricultural good is cA =2p−1/σ (where cA ≡ cF
A + cWA ). Therefore, the price elasticity of demand is
∂cA/cA∂p/p =−1/σ.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 607
1 percent decrease in the price of the agricultural good increasesdemand for that good by less than 1 percent (i.e., the demand foragricultural good is price inelastic).
Therefore, as long as the demand for agricultural goods isinelastic, an increase in agricultural productivity decreases theshare of the population that are farmers living in rural areas.This occurs because an increase in agricultural productivity e de-creases the price of agricultural goods p (since p = 1/e). When thedemand for the agricultural good is inelastic(σ > 1), its increasedconsumption does not fully offset the price decrease, and this de-creases therelativereturns toagricultureandcauses a movementof labor from the countryside into cities. The existing empiricalevidence suggests that the demand for agricultural goods is in-deed price inelastic. Studies typically estimate the price elasticityof food demand to be between −0.80 and−0.20 (e.g., Tobin 1950;Tolley et al. 1969; Van Driel et al. 1997).
The second channel through which potatoes might affect ur-banization arises from the close relationship between urbaniza-tionandpercapita income.15 Thepositiveproductivityshockfromtheadoptionof potatoes canalsoincreasepercapita income. Galorand Weil (2000) provide a unified framework that illustrates theconditions under which this might occur. Their analysis modelsthree periods of economic growth: a Malthusian period, a post-Malthusian regime, and a modern regime. In the Malthusianregime, a positive productivity shock is fully offset by fertility in-creases. Although living standards might increase temporarily, inthe long run this is fully offset by increased fertility, which leavesper capita income constant and population levels higher. In thisregime, which is dated to have been in effect until the nineteenthcentury, potatoes arepredictedtoaffect thelevel of populationbutto have no effect on steady-state per capita income.16
In the post-Malthusian regime, which is dated to have begunin the nineteenth and early twentieth centuries, a positive pro-ductivity shock is not fully offset by a fertility increase, andthere-fore it will have a positive effect on economic growth, increasing
15. See DeLong and Shleifer (1993); Acemoglu, Johnson, and Robinson (2002);and Acemoglu, Johnson, and Robinson (2005); as well as our discussion inSection IV.B. for details.
16. Interestingly, Malthus (1798, pp. 451–457) wrote of the relationship be-tween potatoes, population, and income. He was concerned that the adoption ofthe potato would simply increase population, leaving the population similarly im-poverished in the long run.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
608 QUARTERLY JOURNAL OF ECONOMICS
living standards even in the long run.17 Therefore, in this regime,the introduction of the potato might affect long-term per capitaincome (and therefore urbanization). This mechanism is featuredmore generally in growth models in which income growth is in-creasing in population size (i.e., there are scale effects) and stan-dard Malthusian mechanisms—that cause population increasesto fully erode income gains—are weak or absent.18
Thefinal regimeof theiranalysis is themodernregime, whichis dated to have begun at the end of the nineteenth century inEurope, andthelatetwentiethcenturyforLatinAmericaandAsia(Galor2005, pp. 195–198).19 Theregimeis characterizedbyacom-plete absence of Malthusian mechanisms. Increased productivityhas a negative effect on fertility and population, not a positiveone as in the first two regimes. Therefore, once one begins to mixobservations from the modern regime with those from the ear-lier Malthusian and post-Malthusian regimes, there is no longera clear theoretical prediction about the effect of a positive produc-tivity shock on population. For this reason, we endour analysis in1900.
Our baseline regressions estimate the impact of the introduc-tion of the potato on either total population or urbanization. Aswe show in Section V.B., we find evidence that the potato causedsustained increases in population and urbanization levels overtime. These findings are consistent with potatoes having an ef-fect on the growth rate of population andurbanization. This couldoccur in a world where the rate of innovation is increasing inpopulation size (i.e., there are scale effects), and therefore a one-time shock to population levels causes an increase in the growthrate. However, the finding is also consistent with potatoes onlyhaving long-term level effects (and no growth effects) if conver-gence to the new steady-state levels of population and urbaniza-tion occurred gradually. The slowconvergence could occur in partbecause the adoption of the potato was gradual. Because we donot have a sense for precisely how long it takes to transition fromone steady state to another or data on the adoption process in
17. The beginning of the post-Malthusian regime is believed to have variedacross regions. It is typically dated as beginning in the early nineteenth centuryin Europe and the late nineteenth century for most of Asia, Latin America, andAfrica (e.g., Galor 2005, pp. 180–195).
18. See Jones (2005) for a comprehensive reviewof growth models that featurescale effects.
19. Africa is identified as not yet having entered the modern regime.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 609
different parts of the world, our empirical results are silent aboutwhetherpotatoes onlyaffectedsteadystate levels orwhethertheyalso affected steady state growth rates.
IV. DATA
IV.A. Crop Suitability
Our empirical analysis relies on measures of a region’s suit-ability for growing potatoes, as well as suitability measures forother staple crops. Toconstruct these variables, we use data fromthe Food and Agriculture Organization (FAO)’s Global Agro-Ecological Zones (GAEZ) 2002 database. These data measure thesuitability for cultivating individual crops at a very disaggregatedgeographic level.
The construction of the FAO’s GAEZ database occurred intwo stages. The FAO first collected information on the character-istics of 154 different crops. These data were used to determinewhat environmental conditions are required for the cultivationof each crop. The FAO then compiled data on the physical envi-ronment of 2.2 million grid cells, spanning the entire globe. Eachcell is 0.5 degrees by 0.5 degrees, which is approximately fifty-six kilometers by fifty-six kilometers (measured at the equator).The primary characteristics used are climatic and are taken froma global climatic database that has been compiled by the ClimateResearchUnit at theUniversityof East Anglia. Intotal, ninevari-ables from the global climaticdatabase are used by the FAO: pre-cipitation, frequency of wet days, mean temperature, diurnal (i.e.,daily) temperature range, vapor pressure, cloud cover, sunshine,ground-frost frequency, and wind speed. The second set of char-acteristics are land characteristics and are taken from the FAO’sDigital Soil Map of the World. The final characteristic is the slopeof soils, which is from the GTOPO30 Database, developed at theU.S. Geological Survey (USGS) EROS Data Center.
Combining the information on the constraints for the cultiva-tionofeachcropwiththedataonthephysical environment ofeachgrid-cell, the FAO calculated an estimate of the potential yield ofeachcropineachgridcell, givenanassumedlevel of cropmanage-ment and input use. This process involved a number of detailedsteps, which we briefly summarize here.
First, for each grid cell and crop, the FAO identified the daysof the year when the moisture and thermal (i.e., temperature) re-quirements of the crop are met. With this information, the FAO
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
610 QUARTERLY JOURNAL OF ECONOMICS
determined the exact starting and ending dates of the length ofgrowing period(LGP) for each cropandgridcell.20 An initial clas-sificationof eachgrid-cell andcroppairwas thenperformed. If theminimumrequirements forcultivationwerenot satisfied, thenthecell was determined to be unsuitable for cultivation of the crop.21
If the minimum requirements were met, then a second stage wasperformed where potential yields are determined. For each crop,constraint-free crop yields were determined, and the yield in eachgrid-cell was measured as a percentage of this benchmark. Next,theFAO identifiedadditional constraints that exist ineachcell foreach crop. The procedure quantified the agroclimatic constraints(i.e., variability in water supply and existence of pests and weeds)as well as the agroedaphic suitability (i.e., soil erosion) of eachgrid cell.
Theendproduct oftheprocedureis, foreachcrop, aGIS rasterfilewithglobal coveragethat contains informationonthesuitabil-ity of each grid cell for the cultivation of each crop in question.The FAO also constructed a country-level version of the databasethat reports the proportion of each country’s land that is classi-fied under five mutually exclusive categories describing how suit-able the environment is for growing each crop. The categories arebased on the calculated percentage of the maximum yield thatcan be attained in each grid cell. The five categories and theircorresponding yields are: (i) very suitable land (80–100 percent),(ii) suitable land (60–80 percent), (iii) moderately suitable land(40–60 percent), (iv)marginallysuitableland(20–40 percent), and(v) unsuitable land (0–20 percent). Toapproximate historical con-ditions as closely as possible, we use variables constructed underthe assumption that cultivation occurs under rain-fed conditionsand under medium input intensity.
Wedefinelandtobesuitableforcultivationif it is classifiedinthe database as being either “very suitable,” “suitable,” or “mod-erately suitable.” Put differently, our measure defines land to besuitable if it yields at least 40 percent of the maximum possibleyield.22
20. The growing period is defined as the period of time for which the minimumtemperature and moisture requirements of the crop are satisfied.
21. This is done by comparing each crop’s requirements with the grid cell’scalculated (i) length of the growing period (LGP), (ii) temperature profile, (iii) andthe accumulated temperature.
22. As we discuss in Section VI, the results are very similar when we use 20percent or 60 percent of the maximum yield as alternative cutoffs.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 611
Figure II illustrates the FAO’s suitability measures at thegrid cell level for the Old World countries of our sample. The mapreports an FAO-created suitability index, which is a summarymeasure of suitability based on the underlying categories. Thesuitability index is constructed by assigning the value of 0.90 to“very suitable” land, 0.70 to “suitable” land, 0.50 to “moderatelysuitable” land, and 0.30 to “marginally suitable” land. The suit-ability index then reports the average value of each grid cell,weighted by the proportion of land in each category. A higher in-dex value and a darker shade corresponds to greater suitability.
Figure II shows that much of the landarea suitable for potatocultivationis concentratedinEurope. This fact is apotential causeof concern because it is well known that Western Europe divergedfrom the rest of the world after 1700 for reasons beyond the factthat it was moresuitableforpotatocultivation. Theotherunderly-ingcauses of thedivergencecouldbias ourestimatedimpact of theintroduction of potatoes on population and urbanization if theyare positively correlatedwith cultivation for spurious reasons. Weview this as the greatest potential threat to our estimation strat-egy and take a number of measures toaddress it. We include con-trols for potential underlying determinants of Western Europe’sdivergence. We also check the robustness of our results to the in-clusion of continent-year fixed effects, which capture any time-varying differences between Europe and the rest of the world. Wealsoexaminetheeffect of thepotatooncitypopulations usingonlywithin-continent, and even within-country, variation.
An additional concern is whether the measure calculated inthe 1990s by the FAO is an accurate indicator of suitability twohundred years earlier. The construction of the suitability mea-sures do not give any obvious cause for concern. In fact, the con-struction suggests that the suitability measures should be goodproxies for historical conditions because they are primarily basedon climaticcharacteristics such as temperature, humidity, lengthof days, sunlight, and rainfall that have not changed significantlyover the period of our study. Land characteristics, such as soil pHor slope, which can be affected by human intervention, only af-fect the yield calculations after a grid cell has been classified assuitable or not based on the climatic characteristics. Moreover,in constructing our measure, we also deliberately used the FAOmeasures that assume rain-fed conditions to avoid measurementerror from changes over time in irrigation intensity andtechnologies.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
612 QUARTERLY JOURNAL OF ECONOMICS
FIG
UR
EII
Ave
rage
Pot
ato
Su
itab
ilit
yin
the
Old
Wor
ld
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 613
An additional potential source of concern arises from the factthat potato varieties evolved over time. In particular, the concernis that potatoes might have been bred to be cultivated in loca-tions experiencing fast population growth. If this were the case,then our estimates would reflect the effect of historical popula-tion growth on potatosuitability. The historical evidence suggeststhat while new varieties were developed, the focus was not ondeveloping varieties that could be grown in locations with highpopulation growth. There are two well-documented epochs in thedevelopment of potato varieties. In the nineteenth century, afterthe potatoblights of 1845 and1846, there was, for the first time, aconcertedeffort todevelopnewvarieties withincreasedresistanceto the disease (Salaman 1949, pp. 159–166). Redcliffe Salaman(1949, p. 165) writes that “no spectacular development took placein variety raising until after the crisis caused by the pandemicBlight (Phytophthora infestans).” To the best of our knowledge,the focus was not on developing varieties that could be grown inclimates with rapid population growth. In the twentieth century,commercial cultivars focusedon developing varieties that were vi-sually appealing to consumers. To this end, varieties were devel-oped that were oval or kidney-shaped, had a uniform skin color,and were without sprouts in their pits. In addition, there havebeen attempts to develop varieties that are resistant to dry rot,which can occur during transit (Salaman 1949, pp. 169–171). Wehave not found evidence of deliberate breeding of new varietiesthat couldbegrowninnewlocations withhighpopulationgrowth.
It is also important to recognize that the FAO’s potato suit-ability measures are not based on the most commonly producedvariety today or on varieties only grown in specificclimates. Suit-ability is calculated using four varieties that intentionally span awide range of geographic environments. The varieties chosen canbe cultivated in a range of climates including boreal, temperate,subtropical, and tropical climate zones. See Tables 4.1, 5.10, and5.11 of Fischer et al. (2002) for further details.
Although we have found no evidence that modern varietieswere developed to be grown in locations with fast-growing popu-lations, we cannot rule out this possibility. Therefore, in our anal-ysis we perform a number of checks to ensure that our estimatesare not being driven by the selective breeding of potato varieties.These are reported in Section V.D.
We are also able to conduct a direct check of how well ourex post suitability measure captures historical conditions by
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
614 QUARTERLY JOURNAL OF ECONOMICS
FIGURE IIIBivariate Relationship between Potato Suitability and Potato Production
in 1900
examining the correlation between a country’s suitability forpotato cultivation and historical potato production. The earliestperiod for which production data are available for a cross-sectionofcountries is 1900. ThesedataarefromMitchell (1998, 2003). Wenormalize both the land suitable for potatocultivation and the to-tal production of potatoes (measured in tons) in 1900 by the totallandarea of a country (measuredin thousands of hectares).23 Thebivariate relationshipbetween the share of landsuitable for pota-toes and the natural log of potatoproduction per hectare is shownin Figure III. There is a strong positive correlation between thetwovariables. The standardizedbeta coefficient is 0.52 andis sta-tistically significant at the 1 percent level.24 The strong positive
23. We take the natural log of production per hectare because the distributionis highly skewed otherwise.
24. The strong positive relationship between suitability and production is sim-ilar if alternative specifications are used. For example, taking the natural log ofshare of land suitable for potatoes yields a beta coefficient of 0.52. Alternatively,examining the relationshipbetween the natural log of the total production of pota-toes and the natural log of land suitable for potatoes, while controlling for eitherthe natural log of land suitable for any crop, land suitable for Old World crops, or
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 615
relationship provides reassurance that the ex post suitabilitymeasure provides a reasonable proxy for historical conditions.
Inouranalysis, wealsocontrol forthesuitabilityofotherNewWorldstaplecrops (grainmaize, silagemaize, sweet potatoes, andcassava) and for Old World staple crops (wetland rice, drylandrice, andwheat). Inaddition, somespecifications control foramea-sure of overall agricultural suitability, which is a composite mea-sure of the suitability of an area for growing any agricultural cropforhumanconsumption(i.e., not includingcrops forfodder). Theseadditional suitability measures are also constructed using datafrom the FAO’s GAEZ database and using the same 40 percentsuitability threshold as for potatoes.
IV.B. Population and Urbanization Data
Our analysis examines two outcomes at the country level:total populationandurbanization. Total population, whichis mea-sured as the number of individuals living on land that is a mod-ern country today, are taken from McEvedy and Jones (1978).Urbanization is measured as a country’s total urban population,defined as people living in locations with forty thousand or moreinhabitants, divided by the total population. Data on thepopulations andlocations ofcities usedtoconstruct thenumeratorare from Chandler (1987), Bairoch (1988), and Modelski (2003).Finerdetails oftheprocedurearereportedintheDataAppendix.25
Since historical measures of income per capita are unavail-able prior to 1500, and even in 1500 they are only available fortwenty-two Old World countries, studies examining historical in-come, suchasDeLongandShleifer(1993); Acemoglu, Johnson, andRobinson (2002); and Acemoglu, Johnson, and Robinson (2005);use urbanization as a measure of historical per capita GDP. Thestrong correlation between urbanization and per capita incomeis documented in Acemoglu, Johnson, and Robinson (2002).26
total land yields a beta coefficient of 0.46, 0.47, or 0.45 (respectively), all of whichare statistically significant.
25. Our decision to measure population and urbanization at the country levelis driven by the fact that the historical population estimates from McEvedy andJones (1978) are only available at the country level. As we describe below, we alsoexamine the effect of the potato on city populations.
26. Using the most extensive historical income data available, which are fromMaddison (2003), we have also examined the relationship between urbanizationand income back to 1500. In a panel setting with either country fixed effects oryear fixed effects, we find that the correlation between urbanization and income isextremely strong and highly significant.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
616 QUARTERLY JOURNAL OF ECONOMICS
Following previous studies, we interpret urbanization as a proxyfor income.
Our study includes twelve time periods between 1000 and1900: eight time periods spaced at one hundred-year intervals(1000, 1100, 1200, 1300, 1400, 1500, 1600, and1700)andfourtimeperiods spacedat fifty-yearintervals (1750, 1800, 1850, and1900).
In auxiliary regressions, we examine variation across Euro-pean cities with populations of one thousand or more. These dataare taken from Bairoch, Batou, and Chevre (1988) and DeVries(1984). We alsoexamine variation in adult heights within France.These data are from Komlos (2005).
Accuracy is an obvious concern for historical data that spansuch a long time horizon andbroadcross-section. However, classi-cal measurement error in our outcome variables will not bias ourregression estimates. Similarly, any systematic measurement er-ror that varies by time-period or by country is captured bythe country and year fixed effects, which are included in allspecifications.
V. EMPIRICAL STRATEGY AND MAIN RESULTS
V.A. Empirical Strategy
Ourmainestimationstrategyfollows thesamelogicas astan-dard differences-in-differences (DD) strategy. We compare therelative change in population and urbanization in the postadop-tion period relative to the preadoption period between locationsthat wereabletoadopt thepotatoandthosethat werenot. Thedif-ference between our estimates anda standardDD strategy is thatwe use a continuous measure of the intensity of treatment (i.e.,potatosuitability) andtherebycapturemorevariationinthedata.Because the date of actual adoption in a country could be partlydriven by endogenous factors such as latent population demand,we use the same date of initial adoption for all countries. The his-torical evidence summarized in Section II.B. suggests that exten-siveadoptionas a fieldcropbeganinthelateseventeenthcentury,and by the early eighteenth century, adoption was well underwaythroughout much of Europe. This is also the period of initial con-tact and early adoption in many parts of the world like India, In-donesia, andJapan. Therefore, inourestimates, weusethis as thecutoff period and define the periods prior to and including 1700as the preadoption periods, and the periods after 1700 (i.e., 1750,
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 617
1800, 1850, and 1900) as the postadoption periods. Note that wedonot take this cutoffdate for granted. In the next section, we usea number of procedures to check that our chosen cutoff is consis-tent with the data. We also show that our results are very robustto alternative cutoffs close to the chosen one (see Section VI).
Our main estimating equation assumes that a country’s totalpopulation is proportional to the total amount of land suitable forpotato cultivation. Therefore, the natural log of total populationis proportional tothe natural log of the amount of suitable land.27
This is written as
yit = β lnPotato Areai ∙ IPostt(3)
+1900∑
j=1100
X′iIjtΦj +
∑
c
γc Ici +
1900∑
j=1100
ρj Ijt + εit,
where i indexes countries and t indexes time periods, which are1000, 1100, 1200, 1300, 1400, 1500, 1600, 1700, 1750, 1800, 1850,and 1900. The variable lnPotato Areai is the natural log of the to-tal amount of landthat is suitable forpotatoes, and IPost
t is anindi-catorvariablethat equals onefortheperiods after1700 (i.e., 1750,1800, 1850, and 1900). The outcome of interest, denoted yit, is ei-ther the natural log of total population or the urbanization rate.The equation also includes country and year fixed effects,
∑c Ic
i
and∑
j Ijt, and country-specific characteristics interacted with
time-period fixed effects,∑1900
j=1100 X′iIjtΦj. The country controls in-
clude a large set of geographic and historical variables. They aredescribed when introduced in the analysis.
The coefficient of interest in equation (3) is β, which is theestimated impact of potato suitable land on either population orurbanization. For concreteness, consider the case with total popu-lation as the dependent variable. The estimatedcoefficient βmea-sures the additional growth in population levels experienced bycountries that are suitable for potatoes (relative to those that arenot) after potatoes were introduced in 1700 (relative to before). Apositive coefficient indicates that countries with a geographic en-vironment more suitable for growing potatoes witnesseda greaterincrease in population growth after 1700 relative to before 1700.
27. We take the natural log of the variables toremove the skewness that existsin their distributions otherwise.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
618 QUARTERLY JOURNAL OF ECONOMICS
The estimation strategy has all of the advantages and po-tential pitfalls of standard DD estimators. Country fixed effectscontrol for all time invariant factors that differ between coun-tries. Time period fixed effects control for any secular patternsof population growth or urbanization that affect all regions sim-ilarly. Our identification relies on the assumption that there areno other events beyond those we have controlled for that also oc-curred around 1700 and affected population or urbanization. Thisassumptionshouldnot betakenforgrantedbecausetheOldWorldexperienced many changes during the eighteenth and nineteenthcenturies. We address this in detail below.
V.B. Flexible Estimates
Equation (3) examines the average effect of the potatoon pop-ulation and urbanization after its introduction. Estimation requ-ires that we choose a date in which the potato was adopted inthe Old World. The historical evidence suggests that the adop-tion of the potato began in a few locations in the late seventeenthcentury and spread significantly by the early eighteenth century.Given the evidence, reasonable cutoff dates range between 1700and 1750, and therefore 1750 is the first postadoption time pe-riod. Before taking this cutoff as given, we use a number of dif-ferent strategies to examine whether the patterns in the data areconsistent with this assumption. The first strategy is to estimatea fully flexible estimating equation that takes the followingform:
yit =1900∑
j=1100
βj lnPotato Areai ∙ Ijt(4)
+1900∑
j=1100
X′iIjtΦj +
∑
c
γc Ici +
1900∑
j=1100
ρj Ijt + εit,
where all variables are defined as in equation (3). The onlydifference from equation (3) is that in equation (4), rather thaninteracting lnPotato Areai witha postadoptionindicatorvariable,we interact the suitability measure with each of the time-periodfixed effects. The estimated vectors of βjs reveal the correlationbetween potato suitability and the outcomes of interest in eachtime-period. If, forexample, theintroductionof potatoes increasedpopulation, then we would expect the estimated βjs to be con-stant over time for the years before potatoes were adopted (e.g.,
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 619
β1100 ≈ β1200 ≈ β1300 ≈ . . . ). We also expect the coefficients tobe larger in magnitude for the years after adoption, and becausediffusion occurred gradually, the effects might be increasing overtime (e.g., . . . ≈ β1600 ≈ β1700 ≤ β1750 ≤ β1800 ≤ . . . ).
It is important tonote that we are not particularly interestedin the individual magnitudes of the point estimates. Becauseln Potato Areai is time invariant and equation (4) includes coun-try and time-period fixed effects, the estimated βjs must be mea-suredrelative toa baseline time-period, which we take tobe 1000.Therefore, the absolute level simply tells us the difference in therelationship relative to an arbitrarily chosen baseline. Choosingan alternative baseline period changes the point estimates andstandard errors of all coefficients. Instead, we are interested inthe pattern over time; specifically, whether we observe a discon-tinuity in the pattern around the time potatoes were adopted inthe Eastern Hemisphere.
Estimates of equation (4) are reported in Table II. Columns(1)–(3) report estimates for total population and columns (4)–(6)report estimates for urbanization. The first specification, reportedin columns (1) and(4), includes time-periodfixedeffects andcoun-try fixed effects only, without additional controls. In columns (2)and (5), we include a control for natural log of the amount of landsuitable for Old World crops interacted with the time-period fixedeffects.28 This is done to ensure that the effect of introducingpotatoes is not confounded by other changes in the importanceof Old World staple crops over the same time-period. The finalestimates, in columns (3) and (6), report estimates for our base-line specification. In addition to the control for land suitable forOld World crops (interacted with the time-period fixed effects),we also control for three geographic characteristics, which we ex-pect a priori might be correlated with potato suitability. Becauseone of the attributes of potatoes is that they can be successfullycultivated on rugged terrain at high altitudes, we control for thenatural log of a country’s average elevation and the natural log
28. We define Old World staple crops to be wet-land rice, dry-land rice, andwheat. As with our potato suitability measure, we define land to be suitable forcultivation if it is classified as either “very suitable,” “suitable,” or “moderatelysuitable.” The measure is the natural log of the amount of land that can grow themost suitable Old World crop. An alternative strategy is to measure the union ofthe land that is suitable for each Old World crop (i.e., the fraction of land that cangrowany OldWorldcrop). This alternative measure yields very similar estimates,which is unsurprising since the two measures are highly correlated.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
620 QUARTERLY JOURNAL OF ECONOMICST
AB
LE
IIF
LE
XIB
LE
ES
TIM
AT
ES:T
HE
RE
LA
TIO
NS
HIP
BE
TW
EE
NP
OT
AT
O-S
UIT
AB
LE
LA
ND
AR
EA
AN
DP
OP
UL
AT
ION
OR
CIT
YP
OP
UL
AT
ION
SH
AR
EB
YT
IME
PE
RIO
D
Dep
end
ent
Var
iabl
e
lnto
tal
pop
ula
tion
Cit
yp
opu
lati
onsh
are
(1)
(2)
(3)
(4)
(5)
(6)
lnP
ota
to-S
uit
abl
eA
rea×
1100
0.01
30.
011
0.01
2–0
.001
8–0
.001
3–0
.000
6(0
.003
)(0
.003
)(0
.004
)(0
.001
4)(0
.000
9)(0
.001
3)ln
Pot
ato
-Su
ita
ble
Are
a×
1200
0.02
90.
024
0.02
4–0
.001
1–0
.001
3–0
.001
2(0
.005
)(0
.005
)(0
.007
)(0
.000
9)(0
.000
9)(0
.001
2)ln
Pot
ato
-Su
ita
ble
Are
a×
1300
0.03
90.
031
0.03
00.
0002
–0.0
005
0.00
14(0
.007
)(0
.007
)(0
.010
)(0
.000
8)(0
.001
1)(0
.001
4)ln
Pot
ato
-Su
ita
ble
Are
a×
1400
0.01
90.
004
0.02
10.
0008
0.00
020.
0010
(0.0
08)
(0.0
08)
(0.0
12)
(0.0
012)
(0.0
015)
(0.0
014)
lnP
ota
to-S
uit
abl
eA
rea×
1500
0.03
40.
014
0.02
70.
0003
–0.0
002
0.00
08(0
.009
)(0
.010
)(0
.014
)(0
.000
9)(0
.001
2)(0
.001
3)ln
Pot
ato
-Su
ita
ble
Are
a×
1600
0.04
10.
021
0.02
60.
0002
–0.0
010
–0.0
000
(0.0
09)
(0.0
11)
(0.0
15)
(0.0
014)
(0.0
025)
(0.0
029)
lnP
ota
to-S
uit
abl
eA
rea×
1700
0.04
30.
018
0.02
40.
0020
0.00
170.
0022
(0.0
12)
(0.0
13)
(0.0
18)
(0.0
010)
(0.0
013)
(0.0
015)
lnP
ota
to-S
uit
abl
eA
rea×
1750
0.05
50.
030
0.03
10.
0015
0.00
110.
0013
(0.0
12)
(0.0
14)
(0.0
20)
(0.0
009)
(0.0
013)
(0.0
018)
lnP
ota
to-S
uit
abl
eA
rea×
1800
0.07
30.
048
0.04
10.
0020
0.00
160.
0018
(0.0
14)
(0.0
15)
(0.0
22)
(0.0
009)
(0.0
013)
(0.0
017)
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 621
TA
BL
EII
(CO
NT
INU
ED
)
Dep
end
ent
Var
iabl
e
lnto
tal
pop
ula
tion
Cit
yp
opu
lati
onsh
are
(1)
(2)
(3)
(4)
(5)
(6)
lnP
ota
to-S
uit
abl
eA
rea×
1850
0.09
50.
069
0.06
00.
0024
0.00
220.
0031
(0.0
15)
(0.0
17)
(0.0
20)
(0.0
011)
(0.0
014)
(0.0
017)
lnP
ota
to-S
uit
abl
eA
rea×
1900
0.12
10.
092
0.08
00.
0118
0.01
230.
0100
(0.0
17)
(0.0
21)
(0.0
24)
(0.0
023)
(0.0
024)
(0.0
032)
Bas
elin
eC
ontr
ols
(×Y
ear
fixe
def
fect
s):
lnO
ldW
orld
Cro
ps
Are
aN
YY
NY
Yln
Ele
vati
onN
NY
NN
Yln
Ru
gged
nes
sN
NY
NN
Yln
Tro
pic
al
Are
aN
NY
NN
Y
Obs
erva
tion
s15
5215
5215
5215
5215
5215
52R
-squ
ared
0.99
0.99
0.99
0.42
0.42
0.46
FS
tat
for
Join
tS
ign
fica
nce
1750
–190
017
.88
13.6
04.
208.
028.
824.
89
Not
es.
Obs
erva
tion
sar
eat
the
cou
ntr
y-ye
arle
vel.
All
regr
essi
ons
use
aba
seli
ne
sam
ple
of13
0O
ldW
orld
cou
ntr
ies.
Cou
ntr
ies
inN
orth
and
Sou
thA
mer
ica
are
excl
ud
ed.
Th
ep
erio
ds
are
1000
,110
0,12
00,1
300,
1400
,150
0,16
00,1
700,
1750
,180
0,18
50,a
nd
1900
.Th
ed
epen
den
tva
riab
leis
eith
erth
en
atu
rall
ogof
the
tota
lpop
ula
tion
ofth
eco
un
try
mea
sure
din
per
son
s(l
nto
talp
opu
lati
on)or
the
shar
eof
the
pop
ula
tion
livi
ng
inci
ties
wit
hfo
rty
thou
san
din
hab
itan
tsor
mor
e(C
ity
pop
ula
tion
shar
e).l
nP
ota
to-S
uit
abl
eA
rea
isth
en
atu
rall
ogof
lan
dth
atis
defi
ned
assu
itab
lefo
rth
ecu
ltiv
atio
nof
pot
atoe
s.T
his
isin
tera
cted
wit
han
ind
icat
orfo
rea
chp
erio
dof
the
sam
ple
.All
regr
essi
ons
incl
ud
eye
arfi
xed
effe
cts
and
cou
ntr
yfi
xed
effe
cts.
Eac
hco
ntr
olva
riab
leli
sted
isin
tera
cted
wit
ha
full
set
ofti
me-
per
iod
fixe
def
fect
s.F
ull
det
ails
ofea
chco
ntr
olva
riab
lear
ep
rovi
ded
inth
ete
xtan
dD
ata
Ap
pen
dix
.Th
ein
clu
sion
ofa
con
trol
ofa
vari
able
inte
ract
edw
ith
the
full
set
ofti
me-
per
iod
fixe
def
fect
sis
ind
icat
edby
aY
;Nin
dic
ates
that
the
con
trol
isn
otin
clu
ded
inth
esp
ecifi
cati
on.C
oeffi
cien
tsar
ere
por
ted
wit
hst
and
ard
erro
rs,c
lust
ered
atth
eco
un
try
leve
l,in
par
enth
eses
.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
622 QUARTERLY JOURNAL OF ECONOMICS
of its ruggedness, both interacted with the time-period fixedeffects.29 Furthermore, since potatoes are relatively less suitablefor cultivation in humid tropical climates, we also control for thenatural log of a country’s land that is defined as being tropical.30
The tropical climate control also accounts for the fact that duringour sample period, locations with temperate climates prosperedrelative to locations with more tropical climates. We address thisissue in more detail in Section VI.
A clear pattern emerges from the table. The relationshipbetween potato-suitable land and population is constant overtimeandsmall inmagnitudeduring1000 to1700 andthensteadilyincreases in magnitude from 1750 to 1900. The patterns inthe data can be seen most clearly by plotting the coefficients ofthe interaction terms over time. Figures IVa and IVb plot thepoint estimates from columns (3) and (6) in Table II and their95 percent confidence intervals.
We learn several important facts from the fully flexible esti-mates. First, we do not observe any clear trends of the estimatedinteraction effects during the time periods immediately prior tothe adoption of potatoes. We confirm this impression with a moreformal analysis in the next section. The second insight we gain isthat after1700, thepopulationandurbanizationofpotato-suitablelocations begin to increase relative to locations that are notsuitable. The effect on population appears to begin immediatelyafter 1700, while the effect on urbanization appears to lag behindthe effect on population by approximately fifty to one hundredyears. The coefficients for the interaction between lnPotato Areai
and the time-period fixed effects do not begin to increaseuntil after1750, andeventhentheincreaseis moderateuntil after1850. There are many possible explanations for this. One stemsfrom the long-run growth theories discussed in Section III—inparticulartheunifiedgrowthframeworkof GalorandWeil (2000).The delayed effect of potatoes on urbanization could reflect thefact that a weakening of Malthusian links did not occur until thepost-Malthusian regime, which is dated by Galor (2005) tohave begun in the nineteenth century. Prior to this regime, thebenefits from potatoes might have manifested themselves aspopulation increases, rather than increase in per capita income
29. Similarly, Mokyr (1981) uses the standard deviation of ruggedness as oneof the instruments for potato cultivation in his empirical analysis.
30. Details of the data sources and construction of the measures are providedin the Data Appendix.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 623
FIGURE IVFlexible Estimates of the Relationship between Potato-Suitable Land and Either
Total Population or City Population Share
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
624 QUARTERLY JOURNAL OF ECONOMICS
and urbanization. Throughout our analysis, we conservativelyuse 1700 as the cutoff data for both population and urbanization.Using a later date for urbanization would strengthen the results.
The final insight we gain from the flexible estimates is thatthe positive correlation between potato suitability and the out-comes of interest persistently increase in magnitude during theyears after 1700. As we discussed in Section III, this is consistentwith potatoes having effects on growth, but it is also consistentwith potatoes only having effects on levels, and with the gradualadoption of potato cultivation over the period.
V.C. Rolling Estimates
The second strategy we employ as a check for our chosen cut-off date is similar in spirit to tests for structural breaks.31 Wesystematically examine four-hundred-year segments of our fullpanel.32 For each window, we estimate the baseline specificationfrom equation (3), defining the later two centuries as the post-adoption period. The estimated coefficient for the interaction be-tween potato suitability and the postadoption indicator variablereveals the average increase in population and urbanization be-tween the pre and post periods for suitable countries relative tounsuitable countries. We expect the estimates to be close to zerountil the cutoff begins to coincide with the historical descriptionof the approximate date of potato adoption.33 Prior to the adop-tion date, there is noreason toexpect potato-suitable countries tohave differential growth in population or urbanization.
The estimates are reported in Table III. Columns (1) and (2)report estimated effects for population and urbanization usinga sample that includes four hundred-year periods, ranging from1200 to 1500 (i.e., 1200, 1300, 1400, and 1500). For these regres-sions, the post indicator variable IPost
t takes on the value of zeroin1200 and 1300 and the value of one in 1400 and 1500. Since theOldWorldadoption of potatoes as a staple cropdidnot begin until
31. Because of the limited number of time periods in our sample, we are notable to conduct the standard statistical tests for detecting structural breaks. SeeHansen (2001) for a review of the literature on testing for structural breaks.
32. The results are robust to the choice of different window lengths.33. An alternative strategy is to choose alternative cutoffs including all time
periods and conduct a Chow test. The difficulty with this exercise is that for allspecifications, a significant portion of the post period coincides with the “true”post-potato adoption period. The results are qualitatively similar if this alterna-tive strategy is employed. They are not reported in the paper for brevity but areavailable upon request.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 625
TA
BL
EII
IT
HE
IMP
AC
TO
FT
HE
PO
TA
TO
WIT
HA
LT
ER
NA
TIV
EC
UT-O
FF
S
Pla
cebo
Tre
atm
ent
Per
iod
s16
00–1
900;
1600
–190
0;12
00–1
500;
1300
–160
0;14
00–1
700;
1500
–180
0;P
ost
=18
00,1
850,
Pos
t=
1750
,180
0,P
ost
=14
00,1
500
Pos
t=
1500
,160
0P
ost
=16
00,1
700
Pos
t=
1700
,180
019
0018
50,1
900
Cit
yC
ity
Cit
yC
ity
Cit
yC
ity
lnp
op.
shar
eln
pop
.sh
are
lnp
op.
shar
eln
pop
.sh
are
lnp
op.
shar
eln
pop
.sh
are
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12
)
lnP
ota
to–0
.002
0.00
080.
001
–0.0
008
0.00
10.
0002
0.00
60.
0014
0.03
30.
0038
0.02
80.
0030
Are
a×
Pos
t(0
.006
)(0
.001
2)(0
.005
)(0
.001
7)(0
.006
)(0
.001
6)(0
.007
)(0
.001
7)(0
.011
)(0
.001
5)(0
.009
)(0
.001
7)
Obs
erva
tion
s51
651
651
851
852
052
065
065
078
078
078
078
0R
-squ
ared
0.99
0.56
0.99
0.56
0.99
0.59
0.99
0.59
0.99
0.62
0.99
0.6
2
Not
es.
Obs
erva
tion
sar
eat
the
cou
ntr
y-ye
arle
vel.
All
regr
essi
ons
use
aba
seli
ne
sam
ple
of13
0O
ldW
orld
cou
ntr
ies.
Cou
ntr
ies
inN
orth
and
Sou
thA
mer
ica
are
excl
ud
ed.
Th
ep
erio
ds
vary
bysp
ecifi
cati
onan
dar
ere
por
ted
inth
eco
lum
nh
ead
ings
.Th
ed
epen
den
tva
riab
leis
eith
erth
en
atu
ral
log
ofth
eto
tal
pop
ula
tion
ofth
eco
un
try
mea
sure
din
per
son
s(l
np
op.)
orth
esh
are
ofth
ep
opu
lati
onli
vin
gin
citi
esw
ith
fort
yth
ousa
nd
inh
abit
ants
orm
ore
(Cit
ysh
are)
.ln
Pot
ato
Are
ais
the
nat
ura
llo
gof
lan
dth
atis
defi
ned
assu
itab
lefo
rth
ecu
ltiv
atio
nof
pot
atoe
s.T
he
Pos
tin
dic
ator
vari
able
vari
esby
spec
ifica
tion
.It
sd
efin
itio
nis
rep
orte
din
the
colu
mn
hea
din
gs.
All
regr
essi
ons
incl
ud
eye
arfi
xed
effe
cts,
cou
ntr
yfi
xed
effe
cts,
and
each
ofth
efo
llow
ing
base
lin
eco
ntr
ols
inte
ract
edw
ith
the
full
set
ofti
me-
per
iod
fixe
def
fect
s:ln
Old
Wor
ldC
rop
Su
ita
ble
Are
a,
lnE
leva
tion
,ln
Ru
gged
nes
s,ln
Tro
pic
al
Are
a.F
ull
det
ails
ofea
chco
ntr
olva
riab
lear
ep
rovi
ded
inth
ete
xtan
dD
ata
Ap
pen
dix
.Coe
ffici
ents
are
rep
orte
dw
ith
stan
dar
der
rors
,clu
ster
edat
the
cou
ntr
yle
vel,
inp
aren
thes
es.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
626 QUARTERLY JOURNAL OF ECONOMICS
the endof the seventeenth century, the results from this specifica-tion can be interpreted as a placebo experiment. The same is truefor the estimates reported in columns (3)–(6). Columns (3) and (4)examine the 1300–1600 periods and use a post indicator variablethat equals one in 1500 and1600, while columns (5) and(6) exam-ines 1400–1700 and use an indicator variable that equals one in1600 and1700. Thecoefficient estimates for lnPotato Areai ∙IPost
t inall six specifications are close tozeroandinsignificant. We findnoevidence of a differential relationship between potato suitabilityon population and urbanization in these early preadoption timeperiods.
Columns (7) and (8) report estimates for the window from1500 to 1800, with a post indicator variable that equals one in1700, 1750, and 1800. In this specification, two of the three yearsin the post periodcoincide with the postadoption period(1750 and1800). Therefore, we expect the estimates to capture some of theeffects of the introduction of potatoes. Consistent with this, wefind a positive but small and statistically insignificant coefficientfor lnPotato Areai ∙ IPost
t .Columns (9) and (10) report estimates for 1600–1900, with a
post indicator variable that equals one in 1800, 1850, and 1900.Here all three of the post years coincide with the postadoptionperiod (1800, 1850, and 1900), although one of our postadoptionperiods (1750) is classified as a preadoption year. We now expectthe coefficients to more fully capture the introduction of potatoes.Indeed, we find much larger positive coefficients that are highlysignificant.
As a final check, in columns (11) and (12) we use the 1600–1900 windowand alter the indicator variable totake on the valueof one in 1750, 1800, 1850, and 1900. The post indicator vari-able now coincides exactly with the postadoption period. Again,the results yield positive and statistically significant coefficientestimates.
Taken together, the results confirm the finding from theflexible estimates: the relationship between suitability for potatocultivation and population or urbanization changes after 1700.Therefore, we use 1700 as the last preadoption period and 1750as the first postadoption period in the DD estimations.34
34. The results do not depend on the exact time period chosen. For example,qualitatively similar results are obtained if 1750 is included in the preadoptionperiod.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 627
V.D. Baseline Estimates
We now turn to the estimates from our main estimatingequation (3), which are reported in Table IV. The table reportsresults from five specifications for each outcome of interest. Thefirst three specifications, reported in columns (1)–(3) and (6)–(8),include the same control variables as the specifications for theflexible estimates reported in Table II. The first specificationincludes country and time-period fixed effects only; the secondincludes the additional controls for land suitable for Old Worldstaple crops interacted with time-period fixed effects; and thethird, which is our baseline specification, further adds controls forruggedness, elevation, and tropics, each interacted with the time-period fixed effects.
The estimates of equation (3) confirm the earlier findingsthat suggest that the introduction of potatoes increased total pop-ulation and urbanization. The estimated coefficient of thepotato suitability interaction term, lnPotato Areai ∙ IPost
t , revealsthe average increase in our outcomes of interest arising from in-creased access to the potato after 1700.35 According to the esti-mate in column (3), increasing the amount of land suitable forpotatoes by 1 percent increases population by 0.032 percent onaverage. The estimated coefficient for the city population sharefrom column (8) suggests that a 1 percent increase in land that issuitable for cultivating potatoes increases the urban populationshare by 0.36 percentage points.
To illustrate the magnitudes of our estimates, we perform asimple calculation that measures how much of the observed in-crease in log population and urbanization (of the Old World) be-tween 1700 and 1900 can be explained by the introduction of thepotato. Based on our data, the natural logarithm of Old Worldpopulation increased by 0.90, from 20.21 in 1700 to21.11 in 1900.Using the baseline estimate reported in column (3) of Table IV,we can calculate the counterfactual population in 1900 for eachcountry if potatoes had not been introduced. This is equal to theobserved log population in 1900 minus the estimated impact ofpotatoes, β, multiplied by the natural log of the country’s landarea suitable for potato cultivation, i.e., lnTotal Populationi,1900—β ∙ lnPotato Areai. We then aggregate all countries’ counterfactual
35. We report standard errors clustered at the country level. We have alsocal-culatedConley (1999) standarderrors that correct for possible spatial autocorrela-tion. These are virtually identical to the clustered standard errors reported here.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
628 QUARTERLY JOURNAL OF ECONOMICST
AB
LE
IVT
HE
I MP
AC
TO
FT
HE
PO
TA
TO
:BA
SE
LIN
EE
ST
IMA
TE
S
Dep
end
ent
Var
iabl
e
lnto
tal
pop
ula
tion
Cit
yp
opu
lati
onsh
are
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10
)
lnP
ota
toA
rea×
Pos
t0.
059
0.04
40.
032
0.03
40.
043
0.00
440.
0046
0.00
360.
0039
0.00
39(0
.009
)(0
.011
)(0
.012
)(0
.011
)(0
.014
)(0
.000
9)(0
.000
9)(0
.001
2)(0
.001
1)(0
.001
1)B
asel
ine
Con
trol
s(×
Yea
rfi
xed
effe
cts)
:ln
Old
Wor
ldC
rop
sA
rea
NY
YN
YN
YY
NY
lnE
leva
tion
NN
YY
YN
NY
YY
lnR
ugg
edn
ess
NN
YY
YN
NY
YY
lnT
rop
ica
lA
rea
NN
YY
YN
NY
YY
Oth
erC
ontr
ols
(×Y
ear
fixe
def
fect
s):
lnA
llC
rop
sA
rea
NN
NY
NN
NN
YN
lnM
aiz
eA
rea
NN
NN
YN
NN
NY
lnS
ila
geM
aiz
eA
rea
NN
NN
YN
NN
NY
lnS
wee
tP
ota
toes
Are
aN
NN
NY
NN
NN
Yln
Ca
ssa
vaA
rea
NN
NN
YN
NN
NY
Obs
erva
tion
s15
5215
5215
5215
5215
5215
5215
5215
5215
5215
52R
-squ
ared
0.99
0.99
0.99
0.99
0.99
0.38
0.39
0.44
0.44
0.4
8
Not
es.
Obs
erva
tion
sar
eat
the
cou
ntr
y-ye
arle
vel.
All
regr
essi
ons
use
aba
seli
ne
sam
ple
of13
0O
ldW
orld
cou
ntr
ies.
Cou
ntr
ies
inN
orth
and
Sou
thA
mer
ica
are
excl
ud
ed.
Th
ep
erio
ds
are
1000
,110
0,12
00,1
300,
1400
,150
0,16
00,1
700,
1750
,180
0,18
50,a
nd
1900
.Th
ed
epen
den
tva
riab
leis
eith
erth
en
atu
rall
ogof
the
tota
lpop
ula
tion
ofth
eco
un
try
mea
sure
din
per
son
s(l
nto
talp
opu
lati
on),
orth
esh
are
ofth
ep
opu
lati
onli
vin
gin
citi
esw
ith
fort
yth
ousa
nd
inh
abit
ants
orm
ore
(Cit
yp
opu
lati
onsh
are)
.ln
Pot
ato
Are
ais
the
nat
ura
llo
gof
lan
dth
atis
defi
ned
assu
itab
lefo
rth
ecu
ltiv
atio
nof
pot
atoe
s.T
he
Pos
tin
dic
ator
vari
able
equ
als
zero
for
the
per
iod
s10
00–1
700
and
one
for
the
per
iod
s17
50–1
900.
All
regr
essi
ons
incl
ud
eye
arfi
xed
effe
cts
and
cou
ntr
yfi
xed
effe
cts.
Eac
hco
ntr
olva
riab
leli
sted
isin
tera
cted
wit
ha
full
set
ofti
me-
per
iod
fixe
def
fect
s.F
ull
det
ails
ofea
chco
ntr
olva
riab
lear
ep
rovi
ded
inth
ete
xtan
dD
ata
Ap
pen
dix
.T
he
incl
usi
onof
aco
ntr
olva
riab
lein
tera
cted
wit
hth
efu
llse
tof
tim
e-p
erio
dfi
xed
effe
cts
isin
dic
ated
bya
Y;N
ind
icat
esth
atth
eco
ntr
olis
not
incl
ud
edin
the
spec
ifica
tion
.Coe
ffici
ents
are
rep
orte
dw
ith
stan
dar
der
rors
,clu
ster
edat
the
cou
ntr
yle
vel,
inp
aren
thes
es.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 629
populations and calculate a counterfactual measure of Old Worldlog population. According to the calculations, the counterfactuallog population in 1900 wouldhave been 20.87 (rather than 21.11),and the increase would have been 0.66 (rather than 0.90). There-fore, the increase would have only been 74 percent (0.66/0.90) oftheobservedincreaseifpotatoes hadnot beenintroduced. Inotherwords, according to the estimates, the introduction of the potatoexplains 26 percent of the observed increase in Old World popula-tion between 1700 and 1900.
The average urbanization rate across Old World countriesincreased by 0.0402 (4.02 percent), from 0.0179 (i.e., 1.79 percent)in 1700 to 0.0581 (5.81 percent) in 1900.36 Based on our baselineestimates (from column [8] of Table IV), the increase would havebeen 0.0265 without potatoes. Hence, the increase would haveonly been 66 percent (0.0265/0.0402) of the observed increase ifpotatoes had not been introduced. Therefore, the potato explains34 percent of the observed increase in Old World urbanization be-tween 1700 and 1900.
Columns (4), (5), (9), and (10) of the Table IV report the ro-bustness of our results to the use of other baseline control vari-ables. An alternative to controlling for Old World crop suitabilityis to control for the existence of land suitable for producing anycrop, rather than just Old World staple crops. These estimates,which are reported in columns (4) and (10), are nearly identical tothe baseline estimates reported in columns (3) and (8). The esti-mates in columns (5) and (12) showthat results are unaffected bycontrolling for other New World crops. The specifications includethe natural log of land suitable for cultivating grain maize, silagemaize, sweet potatoes, andcassava, eachinteractedwiththetime-period fixed effects. The results are robust to controlling for theseother New World crops.37
Sensitivity to Variable Definitions. The first sensitivity checkthat we perform tests the robustness of our estimates to the useof alternative postadoption dates and suitability measures. Es-timates of equation (3) using alternative cutoff dates and suit-ability thresholds are reported in Table V. Each cell of the tablereports the coefficient and standard error for β (the coefficient
36. The average is an unweighted country average. The results are similar ifwe calculate country averages weighted by city populations or total populations.
37. The results are similar if we instead control for the crops one at a time orif we control for a composite measure of New World crop suitability.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
630 QUARTERLY JOURNAL OF ECONOMICS
for lnPotato Areai ∙ IPostt ) from one regression. Each row reports
results using an alternative definition of the postadoption period.The first row uses the baseline definition of the postadoption pe-riod, which is 1750 and later. The other two rows use 1700 andlater, and 1800 and later, as similar but alternative definitions.Each column uses a different suitability measure. Columns (1)and(5) useourbaselinethresholdof 40 percent assumingmediuminput intensity. Columns (2) and (6) use a lower threshold of20 percent.
The estimates show that the results remain robust to thechoice of slightly different adoption dates anddifferent suitabilitythresholds. All combinations of the alternative measures produceestimates that are positive, similar in magnitude, and highlysignificant.
The remaining columns of Table V report results usingalternative suitability measures that address the potential endo-geneity of our measure of potatosuitability due toselective breed-ing. As we discussed in Section IV.A., one potential concern isthat potato varieties in locations with higher historical popula-tion growth might have been more intensively bred to develophigh yielding varieties. In other locations, without fast popula-tion growth, varieties were not as intensively bred and thereforethey provide lower yields today. To address this possibility, weconstruct an alternative measure of suitability that measures theamount of landin each country that can growany amount of pota-toes, even if the land is only able to yield a small fraction of themaximum obtainable yield. In other words, we create a measurethat does not include the intensity of yield, which might havebeen affected by the evolution of varieties. In practice, this is theamount of land that can yield 20 percent or more under low in-put intensity, which is the lowest threshold one can choose. Theresults reported in columns (3) and (7) show that the estimatesremain robust to the use of this alternative measure.
A second related concern is that areas with historically highpopulation growth bred varieties to not only yield more, but to begrown more extensively on the land within the country. There-fore, breeding might have not only affected the intensive marginof cultivation, but also the extensive margin as well. To addressthis potential endogeneity, we construct an alternative suitabilitymeasurethat equals thetotal amount ofarablelandifanyamountof land within the country can be used to cultivate potatoes (nomatter what the yield or how small the amount of land that can
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 631
TA
BL
EV
RO
BU
ST
NE
SS
TO
TH
EU
SE
OF
AL
TE
RN
AT
IVE
DE
FIN
ITIO
NS
FO
RP
OT
AT
OS
UIT
AB
ILIT
YA
ND
FO
RT
HE
PO
ST
AD
OP
TIO
NT
IME
PE
RIO
D
Dep
end
ent
Var
iabl
e
lnto
tal
pop
ula
tion
Cit
yp
opu
lati
onsh
are
Com
ple
teC
omp
lete
suit
abil
ity
ifan
ysu
itab
ilit
yif
any
amou
nt
ofla
nd
amou
nt
ofla
nd
Mod
erat
ein
pu
tM
oder
ate
inp
ut
Low
inp
ut
yiel
ds>
20%
atM
oder
ate
inp
ut
Mod
erat
ein
pu
tL
owin
pu
tyi
eld
s>
20%
atin
ten
sity
&40
%in
ten
sity
&20
%in
ten
sity
&20
%lo
win
pu
tin
ten
sity
&40
%in
ten
sity
&20
%in
ten
sity
&20
%lo
win
pu
tsu
itab
ilit
ycu
toff
suit
abil
ity
cuto
ffsu
itab
ilit
ycu
toff
inte
nsi
tysu
itab
ilit
ycu
toff
suit
abil
ity
cuto
ffsu
itab
ilit
ycu
toff
inte
nsi
ty
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Pos
t=
1750
,180
0,0.
032
0.03
90.
038
0.02
60.
0036
0.00
350.
0035
0.00
2018
50,1
900
(Bas
elin
e)(0
.012
)(0
.012
)(0
.011
)(0
.009
)(0
.001
2)(0
.001
1)(0
.001
1)(0
.000
9)
Pos
t=
1800
,185
0,19
000.
039
0.04
50.
045
0.02
90.
0044
0.00
400.
0040
0.00
20(0
.012
)(0
.013
)(0
.012
)(0
.010
)(0
.001
2)(0
.001
3)(0
.001
3)(0
.001
0)
Pos
t=
1700
,175
0,18
00,
0.02
70.
033
0.03
30.
025
0.00
350.
0035
0.00
360.
0021
1850
,190
0(0
.012
)(0
.012
)(0
.011
)(0
.009
)(0
.001
1)(0
.001
0)(0
.001
0)(0
.000
9)
Not
es.E
ach
cell
ofth
eta
ble
rep
orts
the
coef
ficie
nt
for
lnP
ota
toA
rea×
Pos
tfr
omon
ere
gres
sion
.Th
ere
gres
sion
su
sed
iffe
ren
tp
osta
dop
tion
cuto
ffs
and
dif
fere
nt
pot
ato
suit
abil
ity
mea
sure
s.T
he
per
iod
sfo
rw
hic
hth
eP
ost
ind
icat
orva
riab
leeq
ual
son
ear
ere
por
ted
inth
ele
ftm
ost
colu
mn
ofth
eta
ble.
Th
ep
otat
osu
itab
ilit
ym
easu
reu
sed
isre
por
ted
inth
eco
lum
nh
ead
ings
.In
allr
egre
ssio
ns,
obse
rvat
ion
sar
eat
the
cou
ntr
y-ye
arle
velf
ora
sam
ple
that
incl
ud
esal
lOld
Wor
ldco
un
trie
san
dth
efo
llow
ing
tim
ep
erio
ds:
1000
,110
0,12
00,1
300,
1400
,15
00,16
00,17
00,17
50,18
00,18
50,an
d19
00.A
llre
gres
sion
sin
clu
de
year
fixe
def
fect
s,co
un
try
fixe
def
fect
s,an
dth
efo
llow
ing
base
lin
eco
ntr
ols,
each
inte
ract
edw
ith
the
full
set
ofti
me-
per
iod
fixe
def
fect
s:ln
Old
Wor
ldC
rop
Su
ita
ble
Are
a,l
nE
leva
tion
,ln
Ru
gged
nes
s,ln
Tro
pic
al
Are
a.C
oeffi
cien
tsar
ere
por
ted
wit
hst
and
ard
erro
rs,c
lust
ered
atth
eco
un
try
leve
l,in
par
enth
eses
.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
632 QUARTERLY JOURNAL OF ECONOMICS
deliver this yield). Intuitively, the variable assumes full (andequal) treatment to all countries for which there is any evidencethat potatoes can be cultivated on the land.38 Variation in yieldor the expanse of land that can cultivate potatoes, both of whichmight have been impacted by historical breeding, does not en-ter into the measure. The results reported in columns (4) and (8)show that the estimates are robust to this alternative variableas well.
VI. ROBUSTNESS
VI.A. Controlling for Additional Determinants of Population andUrbanization
The results thus far show that after the introduction of pota-toes, locations with greater suitability for potato cultivationexperienced faster increases in population and urbanization. Inour analysis, we control for a baseline set of country character-istics, each interacted with the full set of time-period fixed ef-fects. Wenowconsidera host of additional factors (eachinteractedwith time-period fixed effects) that might have affected historicalpopulation and urbanization. Our choice of controls is guided bythe determinants of long-term economic development that havebeen emphasized in the literature.
Ourfirst set ofcontrols captures differences ingeographythatmight not be captured by the geography controls in our baselinespecification. Thefirst is motivatedbystudies that haveidentifieda strong relationship between a country’s distance from the equa-tor and its economic development (e.g., Hall and Jones 1999). Wethereforecontrol forthenatural logarithmofthedistancefromtheequator (measured in degrees). We also control for the country-average of Kiszewski et al.’s (2004) malaria stability index. Thisis included to capture the historical impact that malaria (and thediseaseenvironment moregenerally) hadondomesticinstitutionsbecause of colonial rule (Acemoglu, Johnson, and Robinson 2001).
Wealsoincludeadditional covariates that capturedifferencesin the evolution of countries’ legal institutions. We control forindicator variables that identify each country’s legal origin, asdefined by La Porta et al. (1998): British, French, German, and
38. An alternative, but similar strategy, is toconstruct a “treatment” indicatorvariable that equals one if any amount of landin the country can yieldany amountof potatoes. Using this measure also produces robust estimates.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 633
Socialist (the omitted category is Scandinavian). We also controlfor the identity of the colonizer among former colonies. Thecolonizer identities, which are taken from Nunn and Puga (forth-coming), are: British, Portuguese, French, Spanish, and otherEuropean (the omitted category is for countries that were nevercolonized).
Our next set of controls captures factors that have been citedas causing the rise of Western Europe over the past three cen-turies. If European geoclimatic conditions are on average moresuitable for potatoes, then other factors that led to the rise ofEurope could cause our estimates to overstate the true effect ofpotatoes. We capture these determinants with indicator variablesthat identify countries that were Atlantic traders, part of the Ro-man Empire, and majority Protestant in 1600. All three factorshave been argued to have been the key determinant of the riseof Europe: Acemoglu, Johnson, and Robinson (2005) argue for theimportance of being an Atlantic trader; Jones (1981) and Landes(1998) argue for the importance of a history of Roman rule; whileWeber (1993) identifies the importance of the Protestant religion.
Our final two variables account for the fact that the spreadof potatoes occurred during a period of increased globalizationand overseas trade. We account for differential impacts of global-ization by controlling for countries’ natural openness to overseastrade measured by a country’s average distance from an ice-freecoast. We also explicitly control for the trade in slaves from theAfrican continent. The slave trade peaked during the eighteenthcentury, at approximately the same time that potatoes were beingadopted. If the countries that were least able to adopt potatoes—e.g., parts of sub-Saharan Africa—were also countries that weredepopulated because of the slave trade, then this might accountfor part of the estimated impact of potatoes. To account for this,we control for the number of slaves taken between periods t − 1and t.
The estimates are reported in Table VI. In Panel A, the de-pendent variable is total population, and in Panel B it is urban-ization. Column(1)shows ourbaselineestimates ofequation(3)asa benchmark for comparison. Columns (2)–(10) report estimatesthat separately control for each alternative determinant inter-acted with time-period fixed effects. Column (11) controls for allalternative determinants simultaneously.
In all specifications, the point estimates for lnPotato Areai ∙IPost
t remain stable, suggesting that estimated impacts of potatoes
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
634 QUARTERLY JOURNAL OF ECONOMICS
TA
BL
EV
IR
OB
US
TN
ES
ST
OC
ON
TR
OL
LIN
GF
OR
AD
DIT
ION
AL
TIM
E-
AN
DC
OU
NT
RY-V
AR
YIN
GF
AC
TO
RS
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11
)
Pan
elA
.Dep
end
ent
Var
iabl
e:ln
tota
lp
opu
lati
o n
lnP
ota
toA
rea×
Pos
t0.
032
0.03
00.
025
0.02
90.
033
0.03
30.
032
0.02
80.
034
0.03
40.
031
(0.0
12)
(0.0
12)
(0.0
16)
(0.0
11)
(0.0
11)
(0.0
12)
(0.0
12)
(0.0
12)
(0.0
12)
(0.0
11)
(0.0
16)
Con
trol
s(×
Yea
rfi
xed
effe
cts)
:B
asel
ine
con
trol
sY
YY
YY
YY
YY
YY
lnD
ista
nce
from
Equ
ato
rN
YN
NN
NN
NN
NY
Mal
aria
Ind
exN
NY
NN
NN
NN
NY
Leg
alO
rigi
nIn
dic
ator
sN
NN
YN
NN
NN
NY
Col
onia
lO
rigi
nIn
dic
ator
sN
NN
NY
NN
NN
NY
Atl
anti
cT
rad
erIn
dic
ator
NN
NN
NY
NN
NN
YR
oman
Her
itag
eIn
dic
ator
NN
NN
NN
YN
NN
YP
rote
stan
tIn
dic
ator
NN
NN
NN
NY
NN
Yln
Dis
tan
cefr
omC
oast
NN
NN
NN
NN
YN
Yln
Sla
veE
xpor
tsN
NN
NN
NN
NN
YY
Obs
erva
tion
s15
5215
5215
5215
5215
5215
5215
5215
5215
5215
5215
52R
-squ
ared
0.99
0.99
0.99
0.99
0.99
0.99
0.99
0.99
0.99
0.99
0.9
9
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 635
TA
BL
EV
I(C
ON
TIN
UE
D)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
Pan
elB
.Dep
end
ent
Var
iabl
e:ci
typ
opu
lati
onsh
are
lnP
ota
toA
rea×
Pos
t0.
0036
0.00
370.
0034
0.00
350.
0036
0.00
350.
0033
0.00
290.
0030
0.00
370.
0029
(0.0
012)
(0.0
012)
(0.0
014)
(0.0
011)
(0.0
013)
(0.0
012)
(0.0
011)
(0.0
012)
(0.0
011)
(0.0
011)
(0.0
013)
Con
trol
s(×
Yea
rfi
xed
effe
cts)
:B
asel
ine
con
trol
sY
YY
YY
YY
YY
YY
lnD
ista
nce
from
Equ
ato
rN
YN
NN
NN
NN
NY
Mal
aria
Ind
exN
NY
NN
NN
NN
NY
Leg
alO
rigi
nIn
dic
ator
sN
NN
YN
NN
NN
NY
Col
onia
lO
rigi
nIn
dic
ator
sN
NN
NY
NN
NN
NY
Atl
anti
cT
rad
erIn
dic
ator
NN
NN
NY
NN
NN
YR
oman
Her
itag
eIn
dic
ator
NN
NN
NN
YN
NN
YP
rote
stan
tIn
dic
ator
NN
NN
NN
NY
NN
Yln
Dis
tan
cefr
omC
oast
NN
NN
NN
NN
YN
Yln
Sla
veE
xpor
tsN
NN
NN
NN
NN
YY
Obs
erva
tion
s15
5215
5215
5215
5215
5215
5215
5215
5215
5215
5215
52R
-squ
ared
0.44
0.45
0.45
0.49
0.46
0.45
0.47
0.46
0.47
0.45
0.5
5
Not
es.
Obs
erva
tion
sar
eat
the
cou
ntr
y-ye
arle
vel.
All
regr
essi
ons
use
aba
seli
ne
sam
ple
of13
0O
ldW
orld
cou
ntr
ies.
Cou
ntr
ies
inN
orth
and
Sou
thA
mer
ica
are
excl
ud
ed.
Th
ep
erio
ds
are
1000
,110
0,12
00,1
300,
1400
,150
0,16
00,1
700,
1750
,180
0,18
50,a
nd
1900
.Th
ed
epen
den
tva
riab
leis
eith
erth
en
atu
rall
ogof
the
tota
lpop
ula
tion
ofth
eco
un
try
mea
sure
din
per
son
s(l
nto
tal
pop
ula
tion
)or
the
shar
eof
the
pop
ula
tion
livi
ng
inci
ties
wit
hfo
rty
thou
san
dor
mor
ein
hab
itan
ts(C
ity
pop
ula
tion
shar
e).l
nP
ota
toA
rea
isth
en
atu
ral
log
ofla
nd
that
isd
efin
edas
suit
able
for
the
cult
ivat
ion
ofp
otat
oes.
Th
eP
ost
ind
icat
orva
riab
leeq
ual
sze
rofo
rth
ep
erio
ds
1000
–170
0an
don
efo
rth
ep
erio
ds
1750
–190
0.A
llre
gres
sion
sin
clu
de
year
fixe
def
fect
s,co
un
try
fixe
def
fect
s,an
dth
efo
llow
ing
base
lin
eco
ntr
ols,
each
inte
ract
edw
ith
the
full
set
ofti
me-
per
iod
fixe
def
fect
s:ln
Old
Wor
ldC
rop
Su
ita
ble
Are
a,l
nE
leva
tion
,ln
Ru
gged
nes
s,ln
Tro
pic
al
Are
a.
Fu
lld
etai
lsof
each
add
itio
nal
con
trol
vari
able
are
pro
vid
edin
the
text
and
Dat
aA
pp
end
ix.
Th
ein
clu
sion
ofa
con
trol
vari
able
inte
ract
edw
ith
the
full
set
ofti
me-
per
iod
fixe
def
fect
sis
ind
icat
edby
aY
;Nin
dic
ates
that
the
con
trol
isn
otin
clu
ded
inth
esp
ecifi
cati
on.C
oeffi
cien
tsar
ere
por
ted
wit
hst
and
ard
erro
rs,c
lust
ered
atth
eco
un
try
leve
l,in
par
enth
eses
.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
636 QUARTERLY JOURNAL OF ECONOMICS
are robust to controlling for other determinants of long-termgrowthanddevelopment. Usingthepoint estimates fromTable VI,wecanagaincalculatehowmuchof theincreaseinOldWorldpop-ulation and urbanization can be attributed to the introduction ofthe potato. Performing the same calculations as in Section V.D.and using the estimates from column (11), we find that the potatoaccounts for 25 percent of the increase population and 27 percentof the increase in urbanization. These effects, although slightlylower than those from our baseline estimates, are still sizeable.
Motivated by the concern that there still remain omittedgeographiccharacteristics, weperformanumberofadditional sen-sitivity tests in which we control for geographic characteristicsused by the FAO as inputs in the construction of their GAEZ cropsuitabilityvariables. As wehaveexplained, theFAO’s constructedcrop suitability measures are created using three underlyinggeographic characteristics that constrain crop growth: climateconstraints, soil constraints, and terrain slope constraints. Cli-mate constraints are determined by a host of geographic factorsincluding daily precipitation, evapotranspiration, average dailytemperatures, accumulated daily temperatures, number of frost-free days, and so on. Soil constraints are a function of depth, fer-tility, drainage, texture and chemical constraints. Terrain slopeconstraints are a function of the slope of the land.
We have tested the robustness of our results to controllingfor the three geographiccomponents, each interactedwith the fullset of time-period fixed effects. The estimates, which we report inthe Online Appendix, show that the estimated impact of potatoesremains robust to controlling for these underlying geographicconstraints. The point estimates remain positive and highlysignificant and are almost exactly the same magnitudes as thebaseline estimates.
VI.B. Examining within Continent Variation
Wenowreturntotheconcernthat ourresults might besimplycapturing the fact that Europe is, on average, more suitable forpotatoes andthat Europeanpopulationandurbanizationdivergedfrom the rest of the world for reasons other than the adoptionof potatoes. Our previous strategy was to control for factors thatmight have caused Europe divergence. Here we pursue an alter-native strategy and estimate the effect of potatoes using within-continent variation only. We do this by adding continent fixedeffects interacted with time-period fixed effects to our baseline
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 637
specification, equation (3). With the continent-year fixed effects,our coefficient of interest, β, is identified from within-continentvariationonly. Therefore, theestimates arenot identifiedfromdif-ferences between Europe and the rest of the world. An additionalbenefit of the continent-year fixed effects is that they capture anyhistorical continent-wide shocks that affected countries within acontinent similarly.39
Estimates are reported in Table VII. Columns (1) and (3)report baseline estimates for comparison, while the remainingcolumns ofthetablereport estimates withthecontinent-yearfixedeffects, with andwithout the set of baseline control variables. Theestimates show that although the point estimates are reducedslightly, theyremainpositiveandsignificant. Theslightlysmallerpoint estimates might reflect a loss of precision arising fromthe fact that there are relatively few countries within eachcontinent.
TABLE VIIROBUSTNESS TO USING WITHIN-CONTINENT VARIATION ONLY
Dependent Variable
ln total population City population share
(1) (2) (3) (4) (5) (6)
ln Potato Area × Post 0.032 0.031 0.020 0.0036 0.0018 0.0022(0.012) (0.013) (0.013) (0.0012) (0.0011) (0.0013)
Controls (× Year fixed effects):Baseline Controls Y N Y Y N YContinent Fixed Effects N Y Y N Y Y
Observations 1552 1552 1552 1552 1552 1552R-squared 0.99 0.99 0.99 0.44 0.45 0.48
Notes. Observations are at the country-year level. All regressions use a sample of Old World countriesthat does not include countries in North and South America. The periods are 1000, 1100, 1200, 1300, 1400,1500, 1600, 1700, 1750, 1800, 1850, and 1900. The dependent variable is either the natural log of the totalpopulation of the country measured in persons (ln total population), or the share of the population living incities with forty thousand inhabitants or more (City population share). ln Potato Area is the natural log ofland that is defined as suitable for the cultivation of potatoes. The Post indicator variable equals zero forthe periods 1000–1700 and one for the periods 1750–1900. All regressions include year fixed effects, countryfixed effects, and the following baseline controls, each interacted with the full set of time-period fixed effects:ln Old World Crop Suitable Area, ln Elevation, ln Ruggedness, ln Tropical Area. The continent fixed effectsare for Europe, Asia, Africa, and Oceania. The inclusion of a control variable interacted with the full set oftime-period fixed effects is indicated by a Y; N indicates that the control is not included in the specification.Coefficients are reported with standard errors, clustered at the country level, in parentheses.
39. The continents in the sample are Europe, Asia, Africa, and Oceania. Thecontinents of North America and South America are not in the sample.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
638 QUARTERLY JOURNAL OF ECONOMICS
VI.C. City Populations
As an alternative to examining variation across countrieswithin a continent, we alsoexamine variation in city populations.This has several benefits. The first is that this provides a richersource of variation for the within-continent estimations. Thesecond benefit is that historical population data are generally be-lieved to be more reliable for cities (Bairoch 1988, pp. 524–525).
The conceptual link between a city’s suitability for potatocultivation and its population is made possible because, unlikeOldWorldgraincrops, potatoes werepredominantlyasubsistencecrop, grown by those who consumed them. Trade in potatoes wasvery difficult because of their low value to weight and because oftheir tendency to spoil during transport. Therefore, historicallythere was almost nolong-distance trade in potatoes (Mokyr 1983,p. 122). However, short-distancetransport didoccurandwas morecommon. For example, Hoffman andMokyr (1981, p. 42) note thaturban residents in Ireland often rented plots of land just outsidethe city to grow potatoes for their own consumption.
Our analysis requires a measure of a city’s suitability for cul-tivatingpotatoes. Inchoosingthelandtoexaminewhenconstruct-ing this measure, we face a tradeoff. We want toinclude land thatwas just outside of the city center because this land might havegrown potatoes consumedby those in the city. However, given thelackof longer-distancetradeinpotatoes, wedonot want toincludeland that is toofar from the city center. Therefore, we consider allland that is within one-hundred-kilometers of the centroid of thecity and calculate the natural log of the land within this area thatis suitable for potato cultivation.40
Thecity-level estimates arereportedinTable VIII. All specifi-cations control for time-period and city fixed effects. We conserva-tively cluster the standard errors at the country level and reportConley (1999) standard errors in square brackets (just below theclustered standard errors).41 All regressions control for the base-line controls—Old World crop suitability, elevation, ruggedness,and prevalence of a tropical climate—but measured at the citylevel (and each interacted with time-period fixed effects).
40. We have also used fifty- or two-hundred-kilometer distances from a city’scenter and find that the results are robust.
41. Conley (1999) standard errors adjust for potential spatial interdependenceof observations. It is assumed that independence is decreasing in the distance be-tween observations and there is complete independence for cities that are morethan five degrees apart.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 639
TA
BL
EV
III
TH
EE
FF
EC
TS
OF
TH
EP
OT
AT
OO
NC
ITY
PO
PU
LA
TIO
NS
Dep
end
ent
Var
iabl
e:ln
city
pop
ula
tion
Cit
ies
wit
h40
,000
+P
opu
lati
on(B
asel
ine
Sam
ple
)C
itie
sw
ith
1000
+P
opu
lati
on
All
Old
Wor
ldC
itie
sO
mit
tin
gE
uro
pe
Eu
rop
eO
nly
Eu
rop
eO
nl y
(1)
(2)
(3)
(4)
(5)
(6)
(7)
lnP
otat
oA
rea×
Pos
t0.
050
0.04
70.
035
0.03
90.
034
0.04
20.
029
Clu
ster
eds.
e.(0
.023
)(0
.021
)(0
.022
)(0
.019
)(0
.075
)(0
.025
)(0
.015
)C
onle
ys.
e.[0
.020
][0
.020
][0
.024
][0
.021
][0
.045
][0
.023
][0
.019
]C
ontr
ols
(×ti
me-
per
iod
fixe
def
fect
s):
Bas
elin
eco
ntr
ols
YY
YY
YY
YC
onti
nen
tfi
xed
effe
cts
NY
NY
YN
NC
oun
try
fixe
def
fect
sN
NN
NN
NY
Obs
erva
tion
s16
0716
0793
393
367
493
1993
19R
-squ
ared
0.77
0.79
0.75
0.75
0.87
0.74
0.7
9
Not
es.
Obs
erva
tion
sar
eat
the
city
-yea
rle
vel.
Col
um
ns
(1)–
(5)
use
asa
mp
leof
citi
esw
ith
fort
yth
ousa
nd
orm
ore
inh
abit
ants
and
the
foll
owin
gti
me
per
iod
s:10
00,
1100
,12
00,
1300
,140
0,15
00,1
600,
1700
,175
0,18
00,1
850,
and
1900
.Col
um
ns
(6)
and
(7)
use
asa
mp
leof
Eu
rop
ean
citi
esw
ith
one
thou
san
dor
mor
ein
hab
itan
tsan
dth
efo
llow
ing
tim
ep
erio
ds:
1000
,120
0,13
00,1
400,
1500
,160
0,17
00,1
750,
1800
,an
d18
50.F
oral
lsp
ecifi
cati
ons,
the
Pos
tin
dic
ator
vari
able
equ
als
zero
for
the
per
iod
s10
00–1
700
and
one
for
per
iod
s17
50an
dla
ter.
All
regr
essi
ons
incl
ud
eye
arfi
xed
effe
cts,
city
fixe
def
fect
s,an
dth
efo
llow
ing
base
lin
eco
ntr
ols,
each
inte
ract
edw
ith
the
full
set
ofti
me-
per
iod
fixe
def
fect
s:ln
Old
Wor
ldC
rop
Su
ita
ble
Are
a,l
nE
leva
tion
,ln
Ru
gged
nes
s,ln
Tro
pic
al
Are
a.T
he
incl
usi
onof
aco
ntr
olva
riab
lein
tera
cted
wit
hth
efu
llse
tof
tim
e-p
erio
dfi
xed
effe
cts
isin
dic
ated
bya
Y;N
ind
icat
esth
atth
eco
ntr
olis
not
incl
ud
edin
the
spec
ifica
tion
.In
allsp
ecifi
cati
ons,
coef
ficie
nts
are
rep
orte
dw
ith
clu
ster
edst
and
ard
erro
rs,c
lust
ered
atth
eco
un
try
leve
l,in
par
enth
eses
.Con
ley
stan
dar
der
rors
are
rep
orte
din
squ
are
brac
ket
s.S
pat
ial
auto
corr
elat
ion
isas
sum
edto
exis
tam
ong
obse
rvat
ion
sth
atar
ew
ith
inte
nd
egre
esof
each
oth
er.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
640 QUARTERLY JOURNAL OF ECONOMICS
Column (1) of Table VIII reports estimates of the baselinespecification for all Old World cities. The estimate shows thatincreasing suitability by 1 percent increases city population by0.050 percent. This estimate is not very different from our base-line country-level population effect of 0.032 percent (see column[3] of Table IV). In column (2), we addcontinent-year fixedeffects.Theestimatedcoefficient of0.047 is verysimilarinmagnitudeandremains highly significant.
Next, we probe the robustness of our results to the omissionof Europe from the sample. Columns (3) and (4) report estimatesfor non-European cities, with andwithout continent-year fixedef-fects. The point estimates are slightly lower but similar tothe fullsample estimates from columns (1) and (2). In column (5), we re-port estimates for a sample that includes only European cities.The coefficient for the European sample remains robust, with apoint estimate that is similar to the estimates for the full sample,although the estimate is much less precise. We investigate thisfurther by examining finer city-level data that are available forEurope only. Recall that our baseline measure of city populationsonly includes cities that have a population of forty thousand ormore. Whenwerestrict oursampletoEuropeancities, thenweareable touse finer data from Bairoch, Batou, andChevre (1988) andDeVries (1984) on cities with one thousand or more inhabitants.
Estimates using the sample of European cities above thelower population threshold are reported in columns (6) and (7) ofTable VIII. Column (6) reports a point estimate that is similar tothe Europe-only estimate using the higher forty thousand-personthreshold. However, the point estimate is now much more pre-ciselyestimatedandis statisticallysignificant. Column(7)reportsestimates from a more demanding specification that includescountry-year fixed effects. The estimates, therefore, are identi-fied from variation across cities within countries. The estimatedecreases slightly but remains positive and significant.
VI.D. Adult Soldier Heights within France
In our final exercise, we move to an even finer level of vari-ation and examine the adult heights of soldiers that were bornin France between 1658 and 1770. Height provides a summarymeasure of the stock of nutritional investments made duringan individual’s growing years of life and is associated with in-creasedlivingstandards, increasedlifeexpectancy, anddecreasedmortality (Waaler 1984; Fogel et al. 1982; Micklewright and
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 641
Ismail 2001; Meng and Qian 2009). Our sample includes 13,646soldiers that were born in France between 1658 and 1770. Theseare all soldiers from the database constructed by Komlos (2005)for which we could determine the location of their town or villageof birth.
Weusethesameestimationstructureas inourbaselineequa-tion (3) to test whether the potato had an effect on adult height:
Heightivt = η lnPotato Areav ∙ IPostt + X′iΓ(5)
+∑
d
X′vIdt Φd +
∑
v
γv Ivi +
1770∑
j=1658
ρj Ijt + εivt.
The outcome of interest, adult height, Heightivt, is measured ininches. It varies across individuals, indexedby i, that wereborninvillagev inyear t. Thespecificationincludes town-of-birthfixedef-fects,
∑v γv Iv
i , as well as year-of-birth fixedeffects,∑
j ρj Ijt. X′i de-
notes controls fortheageandage-squaredof individual i whentheheight was measured, and
∑d X′vId
t Φd denotes a vector ofvillage-level control variables interacted with decade (indexedby d) fixed effects.
We continue to define years after 1700 (i.e., 1701–1770) asthe postadoption period. Like much of Europe, adoption inFrancebeganinthelateseventeenthcenturyandearlyeighteenthcentury, during which time potato cultivation spread in Franche-Comte, Lorraine, Burgundy, Lyonnais, and Alsace (Laufer, 1938,pp. 60–61).42 Tomeasurepotatosuitabilityacross Frenchvillages,we use a more localized twenty-five-kilometer radius, rather thanthe one-hundred-kilometer radius we used to examine variationacross cities globally.
Estimates of equation (5) are reported in Table IX. The esti-mates in column (1) control for town-of-birth fixed effects, year-of-birth fixed effects, the age and age-squared of the soldier whentheheight measurement was taken, andourset ofbaselinecontrolvariables, measured at the village-of-birth level, each interactedwith decade-of-birth fixed effects. In column (2) we also includeregion-of-birth fixed effects, also interacted with the decade-of-birth fixed effects. In columns (3) and (4), we also include the
42. The estimates are robust to the choice of alternative but similar cutoffdates. Specifically, choosing any cutoff date between 1680 and 1730 yields simi-lar results.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
642 QUARTERLY JOURNAL OF ECONOMICS
TABLE IXTHE EFFECTS OF THE POTATO ON SOLDIER HEIGHTS WITHIN FRANCE
Dependent Variable: Adult height (inches)
(1) (2) (3) (4)
ln Potato Area × Post 0.102 0.062 0.103 0.054Clustered s.e. (0.025) (0.018) (0.025) (0.018)Conley s.e. [0.014] [0.013] [0.017] [0.015]Controls (× decade fixed effects):
Baseline controls Y Y Y YFrench region fixed effects N Y N YAdditional controls N N Y Y
Observations 13646 13646 13646 13646R-squared 0.23 0.23 0.23 0.24
Notes. The dependent variable is the height of a French soldier (measured in inches), born in Francebetween 1658 and 1770. Observations are at the town-of-birth and year-of-birth levels. The sample includes1467 French towns and all years from 1658 to1770. The Post indicator variable equals zerofor years-of-birthfrom 1658–1700 andone for years 1701–1770. All regressions include year-of-birth fixedeffects, town-of-birthfixedeffects, the soldier’s age andage squaredat the time of height measurement, andour set of baseline con-trols, each interactedwith decade fixedeffects (ln Old World Crop Suitable Area, ln Elevation, ln Ruggedness,ln Tropical Area). The additional controls include: ln Distance from Equator, malaria index, and ln Distancefrom the Coast. For all specifications, the inclusion of a control variable interacted with decade fixed effectsis indicated by a Y; N indicates that the control is not included in the specification. In all specifications, co-efficients are reported with standard errors, clustered at the province level, in parentheses. Conley standarderrors are reported in square brackets. Spatial autocorrelation is assumed to exist among observations thatare within five degrees of each other.
additional controls from Table VI, each interacted with thedecade-of-birth fixed effects. Because many of the controls fromTable VI do not vary within France (e.g., legal origin), the set ofadditional controls is reducedto: lnDistance from the Equator, themalaria index, and ln Distance from the Coast. The table reportsstandard errors clustered at the province level as well as Conleystandard errors.
The estimates provide evidence of a positive effect of potatoeson adult height. All estimates of η are positive and significant.The coefficients suggest that for towns that were fully suitable forpotato cultivation, the introduction of the potato increased aver-age adult height by 0.41–0.78 inches.43
In addition to providing estimates based on variation withinfinely defined French villages where much is held constant, theresults also provide evidence for the mechanisms underlying ourpopulation estimates. They suggest that potatoes had a positive
43. For towns with full suitability, the introduction of the potatoincreased thepotato interaction, lnPotato Areav ∙ IPost
t , from 0 to 7.58. Therefore, the calculatedeffects range from 7.58× 0.054 = 0.41 to 7.58× 0.103 = 0.78.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 643
impact on long-term health. And because previous studies findthat withinFranceduringthis time-period, height correlates withwealth, education, and literacy (Komlos 2003, pp. 159–160), theestimates alsocomplement our urbanization findings andsuggestthat potatoes also increased living standards and income.
Overall, the estimates presented in Tables VII–IX providereassurance that our baseline estimates of the effect of potatoesare not biased by other factors responsible for the rise of Europeduring the period. Table VII shows that the country-level esti-mates are still positive and similar in magnitude when we controlfor continent-year fixed effects. The estimates from Tables VIIIand IX show that the same is true when we examine variationin city populations within continents and countries and when weexamine variation in heights across individuals within France.
VII. CONCLUSIONS
This study examines the effect of the introduction of potatoeson population and urbanization in the Old World during theeighteenth and nineteenth centuries. The nutritional and caloricsuperiority of potatoes relative to existing Old World staplecrops, together with their introduction from the New World tothe Old World, allows us to use a strategy similar to difference-in-differences to estimate the causal effect of the availability ofpotatoes on population and urbanization. According to our esti-mates, the introduction of the potato explains 25–26 percent ofthe increase in Old World population between 1700 and 1900 and27–34 percent of the increase in urbanization.
Additional evidence from within-country comparisons of citypopulations and adult heights confirmed our cross-country find-ings. We found that the adoption of potatoes spurred city growth.Within a country, more suitable locations experienced fasterrelative population growth after the diffusion of the potato. Usinganthropometricmeasurements of French soldiers born in the sev-enteenth and eighteenth centuries, we also found that potatoesincreased adult height. According to our estimates, for villagesthat were completely suitable for potatocultivation, the introduc-tion of the potato increased average adult heights by about halfan inch.
Our findings contribute tothe historical debate about the im-portance of nutritional improvements in explaining part of the
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
644 QUARTERLY JOURNAL OF ECONOMICS
rapid population increase over the past three centuries. Theyprovide evidence that nutrition matters. Furthermore, becauseurbanization rates and adult heights provide reasonable proxiesfor economic development and overall standards of living, our re-sults suggest that the availability of potatoes also played an im-portant role in spurring economic growth in the eighteenth andnineteenth centuries.
DATA APPENDIX
A. Crop Suitability Measures
Data on the suitability of climates for growing different cropsare from the FAO’s Global Agro-Ecological Zones (GAEZ) 2002database (http://fao.org/Ag/AGL/agll/gaez/index.htm). We use thecountry-level measures constructedbytheFAO inourregressionsat the country level.
The FAO’s measure of overall agricultural suitability is acomposite measure of the suitability of an area for growing anyagricultural crop for human consumption (not including crops forfodder).
The suitability variables are constructed as the natural logof one plus the amount of suitable land. In the country-levelregressions, land is measured in thousands of hectares. For thecity or village level regressions, land is measured in squarekilometers.
B. Population and Urbanization
Country-level population data are from McEvedy and Jones(1978). Data on the populations of cities with more than fortythousand inhabitants are from Chandler (1987), Bairoch (1988),andModelski (2003). Among the three sources, Chandler (1987) isthe most complete, followed by Bairoch (1988) and then Modelski(2003). We therefore use the city measures from Chandler (1987),supplemented by Bairoch (1988) and Modelski (2003) if a city’spopulation is missing from Chandler (1987) but exists in eitherBairoch (1988) or Modelski (2003).
C. Baseline Geographic Control Variables
The variables measuring each country’s average elevation,average ruggedness, and amount of land in a tropical climate areall from Nunn and Puga (forthcoming). Average elevation is
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 645
measured in meters; ruggedness is the terrain ruggedness indexmeasured in hundreds of meters. A country’s total tropical landarea, measured in thousands of hectares, is calculated by multi-plying the fraction of a country’s land that is tropical from Nunnand Puga (forthcoming) by a country’s total land area from theFAO’s GAEZ database. Each variable enters the regressionequation log-linearly.
D. All Other Control Variables
A country’s average distance from the equator (measured indegrees) andits averagedistancefromanice-freecoast (measuredin thousands of kilometers) are both taken from Nunn and Puga(forthcoming). Indicator variables for a country’s legal origin andfor the identity of the colonizer are also from Nunn and Puga(forthcoming). The indicator variable for a country’s history of Ro-man rule equals one if a country was a part of the Roman Empireand not subsequently part of the Ottoman empire. The variablewas definedandconstructedby Acemoglu, Johnson, andRobinson(2005), who rely on information from Langer (1972). The Atlantictrader and Protestant religion indicator variables are also takenfrom Acemoglu, Johnson, andRobinson (2005). The Protestant in-dicator variable equals one if a country had a majority Protestantreligion in 1600. The index of the stability of malaria transmis-sion is taken from Nunn and Puga (forthcoming), who calculatecountry-level averages usingunderlyingdatafromKiszewskiet al.(2004). The number of slaves taken from a country duringthe Indian Ocean, Red Sea, trans-Saharan, and trans-Atlanticslave trades in the years between each time-period(i.e., from t− 1to t) is taken from Nunn (2008). Because of data limitations,estimates areonlyavailablefrom1400. Weassumedthat noslaveswere taken prior to 1400. Although this is true for the trans-Atlantic trade, it is not true, strictly speaking, for the otherthree slave trades, in which a modest number of slaves wereexported.
E. City-Level Variables
Data on the populations of cities with more than forty thou-sand inhabitants are from Chandler (1987), Bairoch (1988),and Modelski (2003). For the Europe-only analysis of cities withmore than one thousand inhabitants data on city sizes are fromBairoch, Batou, and Chevre (1988) and DeVries (1984); becausethe former source has wider geographic coverage and a longer
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
646 QUARTERLY JOURNAL OF ECONOMICS
timeseries, thesefigures areusedwithmissingobservations filledin with data from DeVries (1984) where possible. Because of lim-ited temporal coverage in both sources, the analysis using thisdata only includes the following time periods: 1000, 1200, 1300,1400, 1500, 1600, 1700, 1750, 1800, and 1850. The geographiclocations of cities were identified using the global gazetteerGeonames, which is accessible at www.geonames.org.
Data on the suitability of the climate of a city for growingvarious crops are from the FAO’s Global Agro-Ecological Zones(GAEZ) 2002 database. The data are publicly available and canbe downloaded from www.iiasa.ac.at/Research/LUC/SAEZ/index.html. We use the underlying grid cell data, which are availableas GIS rasterfiles. Thedataontropical climates arefromthesamesource (plate 3). We define the tropics and subtropics as having atropical climate. City level suitability is measuredby the suitabil-ity of land within a one-hundred-kilometer radius of the city.
The underlying data used to construct city-level measures ofruggedness are from the FAO’s Terrastat 2002 data compilation.The data are originally from the USGS GPOTO 30 elevation grid,which is the same source usedby Nunn andPuga (forthcoming) toconstruct a country-level measure of terrain ruggedness. Rugged-ness is measured as the average uphill slope of land within aone-hundred-kilometer radius of the city. Data used to constructeach city’s average elevation (measured in meters) is from GlobalMappingInternational’s Seamless Digital Chart oftheWorldBaseMap (DCW), version 3.2.
F. Soldier Heights Variables
Data on the heights of French soldiers and their dates andtowns of birth are from Komlos (2005). Soldier heights are mea-suredininches. Village-level potatosuitability is measuredbytheamount of suitable land within a twenty-five-kilometer radius ofthe soldier’s village of birth. The construction of the village-levelcontrols also uses a twenty-five-kilometer radius.
HARVARD UNIVERSITY, NBER AND BREADYALE UNIVERSITY, NBER AND BREAD
REFERENCES
Acemoglu, Daron, Simon Johnson, and James A. Robinson, “The Colonial Originsof Comparative Development: An Empirical Investigation,” American Eco-nomic Review, 91 (2001), 1369–1401.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 647
, “Reversal of Fortune: GeographyandInstitutions intheMakingof theMod-ern World Income Distribution,” Quarterly Journal of Economics, 117 (2002),1231–1294., “The Rise of Europe: Atlantic Trade, Institutional Change and Economic
Growth,” American Economic Review, 95 (2005), 546–579.Allen, Robert C., “The Nitrogen Fixing Hypothesis and the English Agricultural
Revolution: A Biological Analysis,” Journal of Economic History, 68 (2008),182–210.
Bairoch, Paul, Cities and Economic Development: From the Dawn of History to thePresent (Chicago, IL: University of Chicago Press, 1988).
Bairoch, Paul, Jean Batou, and Pierre Chevre, The Population of European Citiesfrom 800 to 1850 (Geneva: Librairie Droz, 1988).
Baten, Jorg, and John E. Murray, “Heights of Men and Women in nineteenth-Century Bavaria: Economic, Nutritional, and Disease Influences,” Explorat-ions in Economic History, 37 (2000), 351–369.
Brown, Charles R., “Origin and History of the Potato,” American Journal of PotatoResearch, 70 (1993), 363–373.
Burton, W.G., The Potato: A Survey of its History and of Factors Influencing itsYield, Nutritive Value and Storage (London: Chapman & Hall Ltd., 1948).
Chandler, Tertius, Four Thousand Years of Urban Growth: An Historical Census(Lewiston, NY: The Edwin Mellen Press, 1987).
Chorley, G.P.H., “The Agricultural Revolution in Northern Europe, 1750–1880:Nitrogen, Legumes, and Crop Productivity,” Economic History Review, 34(1981), 78–93.
Cock, James H., “Cassava: A Basic Energy Source in the Tropics,” Science, 218(1982), 755–762.
Conley, Timothy, “GMM Estimation with Cross-Sectional Dependence,” Journalof Econometrics, 92 (1999), 1–45.
Connell, K.H., “The History of the Potato,” Economic History Review, 3 (1951),388–395., “The Potato in Ireland,” Past and Present, 23 (1962), 57–71.
Cullen, L.M., “Irish History Without the Potato,” Past and Present, 40 (1968),72–83.
Cutler, David M., Angus Deaton, and Adriana Lleras-Muney, “The Determinantsof Mortality,” Journal of Economic Perspectives, 20 (2006), 97–120.
Davidson, Stanley, R. Passmore, J.F. Brock, and A.S. Truswell, Human Nutritionand Dietetics (New York, NY: Churchill Livingstone, 1975).
DeLong, J. Bradford, and Andrei Shleifer, “Princes and Merchants: City GrowthBefore the Industrial Revolution,” Journal of Law and Economics, 36 (1993),671–702.
DeVries, Jan, European Urbanization, 1500–1800 (Cambridge, MA: Harvard Uni-versity Press, 1984).
Fischer, Gunther, Harrij van Nelthuizen, Mahendra Shah, and Freddy Nachter-gaele, Global Agro-Ecological Assessment for Agriculture in the 21st Century:Methodology and Results (Rome: Food and Agriculture Organization of theUnited Nations, 2002).
Fogel, Robert W., “Nutrition and the Decline in Mortality since 1700: SomePreliminary Findings,” NBER Working Paper No. 1402, 1984., “Economic Growth, Population Theory, and Physiology: The Bearing of
Long-Term Processes on the Making of EconomicPolicy,” American EconomicReview, 84 (1994), 369–395., The Escape from Hunger and Premature Death, 1700-2100 (New York, NY:
Cambridge University Press, 2004).Fogel, Robert W., Stanley L. Engerman, and James Trussell, “Exploring the Uses
of Data on Height,” Social Science History, 6 (1982), 401–421.Galor, Oded, “From Stagnation to Growth: Unified Growth Theory, Volume 1,” in
Handbook of Economic Growth, Philippe Aghion and Steven N. Durlauf, eds.(Amsterdam: North-Holland, 2005), 171–293.
Galor, Oded, and David N. Weil, “Population, Technology and Growth: FromMalthusian Stagnation to the Demographic Transition and Beyond,” Amer-ican Economic Review, 90 (2000), 806–828.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
648 QUARTERLY JOURNAL OF ECONOMICS
Glendinning, D.R., “Potato Introductions and Breeding up to the Early twentiethCentury,” New Phytologist, 94 (1983), 479–505.
Grennes, Thomas, “The Columbian Exchange and the Reversal of Fortune,” CatoJournal, 27 (2007), 91–107.
Hall, Robert E., and Charles I. Jones, “Why Do Some Countries Produce So MuchMore Output Per Worker than Others?” Quarterly Journal of Economics, 114(1999), 83–116.
Hansen, Bruce, “The New Econometrics of Structural Change: Dating Breaks inU.S. Labor Productivity,” Journal of Economic Perspectives, 15 (2001),117–128.
Hather, Jon, and Patrick V. Kirch, “Prehistoric Sweet Potato (Ipomoea Batatas)from Mangaia Island, Central Polynesia,” Antiquity, 65 (1991), 887–893.
Hersh, Jonathan, and Hans-Joachim Voth, “Sweet Diversity: Colonial Goods andthe Rise of European Living Standards after 1492,” Mimeo, Universitat Pom-peu Fabra, 2009.
Hoffman, Elizabeth, and Joel Mokyr, “Peasants, Potatoes, and Poverty: Transac-tions Costs in Prefamine Ireland,” Discussion Paper 474, NorthwesternUniversity, 1981.
Hughes, R.E., “Scurvy,” in The Cambridge World History of Food, Volume II,Kenneth F. Kiple, ed. (New York, NY: Cambridge University Press, 2000),988–1000.
Jones, Charles, “Population and Ideas: A Theory of Endogenous Growth,” inKnowledge, Information, and Expectations in Modern Macroeconomics: InHonor of Edmund S. Phelps, Philippe Aghion, Roman Frydman, JosephStiglitz, and Michael Woodford, eds. (Princeton, NJ: Princeton UniversityPress, 2003), 498–521.
Jones, Charles I., “GrowthandIdeas,”inHandbook of Economic Growth, Volume 1,Philippe Aghion and Steven N. Durlauf, eds. (Amsterdam: North Holland,2005), 1063–1111.
Jones, Eric L., The European Miracle: Environments, Economies, and Geopoliticsin the History of Europe and Asia (New York, NY: Cambridge UniversityPress, 1981).
Kiszewski, Anthony, Andrew Mellinger, Andrew Spielman, Pia Malaney, SoniaEhrlich Sachs, and Jeffrey Sachs, “A Global Index of the Stability of MalariaTransmission,” American Journal of Tropical Medicine and Hygiene, 70(2004), 486–498.
Komlos, John, “An Anthropometric History of Early Modern France,” EuropeanReview of Economic History, 7 (2003), 159–189., “Height of French Soldiers, 1716–1784,” (2005), [Computer file]. ICPSR
04363-v1.La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny,
“Law and Finance,” Journal of Political Economy, 106 (1998), 1113–1155.Landes, David S., The Wealth and Poverty of Nations: Why Some are so Rich and
Some so Poor (New York, NY: W.W. Norton & Company, 1998).Langer, William L., “Europe’s Initial Population Explosion,” American Historical
Review, 69 (1963), 1–17., An Encyclopedia of World History, Fifth Edition (Boston, MA: Houghton
Mifflin Company, 1972)., “American Foods and Europe’s Population Growth,” Journal of Social
History, 8 (1975), 51–66.Laufer, Berthold, “The American Plant Migration. Part I: The Potato,” Field
Museum of Natural History: Anthropological Series, 28 (1938), 9–132.Lee, James, “FoodSupplyandPopulationGrowthinSouthwest China, 1250–1850,”
Journal of Asian Studies, 41 (1982), 711–746.Livi-Bacci, Massimo, Population and Nutrition: An Essay on European Demo-
graphic History (Cambridge, MA: Cambridge University Press, 1991).Maddison, Angus, The World Economy: Historical Statistics (Paris: Organisation
for Economic Co-operation and Development, 2003).Malthus, Thomas R., An Essay on the Principle of Population (London: Reeves and
Turner (Reprint 1872), 1798).McEvedy, Colin, andRichardJones, Atlas of World Population History (NewYork,
NY: Penguin Books, 1978).
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
POTATO’S CONTRIBUTION TO POPULATION AND URBANIZATION 649
McKeown, Thomas, The Modern Rise of Population (New York, NY: AcademicPress, 1976).
McNeill, WilliamH., “TheIntroductionof thePotatointoIreland,” Journal of Mod-ern History, 21., “How the Potato Changed the World’s History,” Social Research, 66 (1999),
67–83.Meng, Xin, and Nancy Qian, “The Long Term Consequences of Famine on
Survivors: Evidence from a Unique Natural Experiment using China’s GreatFamine,” NBER Working Paper No. 14917, 2009.
Messer, Ellen, “Maize,” in The Cambridge History of World Food, Volume I,Kenneth F. Kiple and Kriemhild Conee Ornelas, eds. (New York, NY:Cambridge University Press, 2000a), 97–112., “Potatoes,” in The Cambridge History of World Food, Volume I, Kenneth F.
Kiple and Kriemhild Conee Ornelas, eds. (New York, NY: Cambridge Univer-sity Press, 2000b), 187–201.
Micklewright, John, and Suraiya Ismail, “What can Child Anthropometry Revealabout Living Standards andPublicPolicy? An illustration from Central Asia,”Review of Income and Wealth, 47 (2001), 65–80.
Mitchell, B.R., International Historical Statistics: Africa, Asia and Oceania, 1750-1993 (London: Palgrave MacMillan, 1998)., International Historical Statistics: Europe (London: Palgrave MacMillan,
2003).Modelski, George, World Cities: 3000 to 2000 (Washington, D.C.: Faros, 2003).Mokyr, Joel, “Irish History with the Potato,” Irish Economic and Social History, 8
(1981), 8–29., Why Ireland Starved: A Quantitative and Analytical History of the Irish
Economy, 1800–1850 (London: George Allen & Unwin., 1983).Nunn, Nathan, “The Long-Term Effects of Africa’s Slave Trades,” Quarterly
Journal of Economics, 123 (2008), 139–176.Nunn, Nathan, and Diego Puga, “Ruggedness: The Blessing of Bad Geography in
Africa,” Review of Economics and Statistics, forthcoming.Nunn, Nathan, and Nancy Qian, “The Columbian Exchange: A History of
Disease, Food, and Ideas,” Journal of Economic Perspectives, 24 (2010),163–188.
O’Brien, Patricia J., “Sweet Potatoes and Yams,” in The Cambridge History ofWorld Food, Volume I, Kenneth F. Kiple and Kriemhild Conee Ornelas, eds.(New York, NY: Cambridge University Press, 2000), 207–218.
O Grada, Cormac, Ireland’s Great Famine (Dublin: University College DublinPress, 2006)., Famine: A Short History (Princeton, NJ: Princeton University Press, 2009).
O’Rourke, Kevin, and Jeffrey Williamson, “When Did Globalization Begin?”European Review of Economic History, 6 (2002), 23–50.
Pandey, S.K., andS.K. Kaushik, “Origin, Evolution, HistoryandSpreadof Potato,”inThe Potato: Production and Utilization in Sub-Tropics, S. M. Paul Khurana,J. S. Minhas, and S. K. Pandey, eds. (New Delhi: Mehta, 2003).
Reader, John, Propitious Esculent: The Potato in World History (London: WilliamHeinemann, 2008).
Salaman, Redcliffe, The History and Social Influence of the Potato (Cambridge,MA: Cambridge University Press, 1949).
Smith, Adam, An Inquiry into the Nature and Causes of the Wealth of Nations(London: T. Nelson and Sons, Paternoster Row, 1776).
Solar, Peter M., “The Potato Famine in Europe,” in Famine 150: CommemorativeLecture Series, Cormac O’Grada, ed. (Dublin: Leinster Leader, 1997),113–127.
Stuart, William, Potato: Its Culture, Uses, History and Classification (Philadel-phia, PA: J.B. Lippincott and Company, 1923).
Tiffin, Richard, andXavierIrz, “Is AgriculturetheEngineofGrowth?”AgriculturalEconomics, 35 (2006), 79–89.
Timmer, Peter, “The Turnip, the New Husbandry, and the English AgriculturalRevolution,” Quarterly Journal of Economics, 83 (1969), 375–395.
Tobin, James, “A Statistical Demand Function for Food in the U.S.A.” Journal ofthe Royal Statistical Society, Series A, 113 (1950), 113–141.
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
650 QUARTERLY JOURNAL OF ECONOMICS
Tolley, George S., Yi Wang, and R. G. Fletcher, “Reexamination of the Time SeriesEvidence on Food Demand,” Econometrica, 37 (1969), 695–705.
Turner, Michael, After the Famine: Irish Agriculture, 1850–1914 (Cambridge, MA:Cambridge University Press, 1996).
Ugent, Donald, “The Potato in Mexico: Geography and Primitive Culture,” Eco-nomic Botany, 22 (1968), 108–123.
U.S. Department of Agriculture, USDA National Nutrient Database for StandardReference, Release 20 (Washington, D.C.: U.S. Department of Agriculture,Agricultural Research Service, 2007).
Van Driel, Hans, Venuta Nadall, and Kees Zeelenberg, “The Demand for Food inthe United States and the Netherlands: A Systems Approach with the CBSModel,” Journal of Applied Econometrics, 12 (1997), 509–523.
Vandenbroeke, Chr. “Cultivation and Consumption of the Potato in the seven-teenth and eighteenth Century,” Acta Historiae Neerlandica, 5 (1971), 15–39.
Voigtlander, Nico, andHans-Joachim Voth, “Why England? DemographicFactors,Structural Change and Physical Capital Accumulation during the IndustrialRevolution,” Journal of Economic Growth, 11 (2006), 319–361.
von Furer-Haimendorf, Christoph, The Sherpas of Nepal: Buddhist Highlanders(Berkeley, CA: University of California Press, 1964).
Waaler, Hans, “Height, Weight and Mortality: The Norweigan Experience,” ActaMedica Scandinavia, 215 (1984), 1–51.
Weber, Max, The Protestant Ethic and the Spirit of Capitalism (London:Routledge, 1993).
Young, Arthur, The Farmer’s Tour through the East of England, Volume 4 (London:W. Strahan, 1771).
Dow
nloaded from https://academ
ic.oup.com/qje/article-abstract/126/2/593/1868756 by guest on 10 April 2019
top related