European Historical Economics Society ! EHES!WORKING!PAPERS!IN!ECONOMIC!HISTORY!!|!!!NO.!65! New crops, local soils and urbanization: Clover, potatoes and the growth of Danish market towns, 1672-1901 Torben Dall Schmidt University of Southern Denmark, Sønderborg Peter Sandholt Jensen University of Southern Denmark, Odense Amber Naz University of Southern Denmark, Sønderborg OCTOBER!2014! ! EHES!Working!Paper!|!No.!65!|!October!2014! New crops, local soils and urbanization: Clover, potatoes and the growth of Danish market towns, 1672-1901 Torben Dall Schmidt* University of Southern Denmark, Sønderborg Peter Sandholt Jensen*** University of Southern Denmark, Odense Amber Naz*** University of Southern Denmark, Sønderborg Abstract This research evaluates the impact of the introduction of clover and potatoes on urbanization using a panel of Danish market towns from 1672 to 1901. We find evidence that both clover and potatoes contributed to urbanization using a difference-in-difference type estimation strategy which exploits that the breakthrough of clover and potatoes should have differential local effects because of soil suitability. To take into account the endogeneity of clover adoption, we instrument by suitability for growing alfalfa, which like clover is a legume. Importantly, alfalfa did not have its breakthrough in the period studied. Our IV estimates suggest that clover accounted for about 8 percent of market town population growth between 1672 and 1901, whereas roughly 6 percent can be attributed to potatoes. The analysis also indicates that the potato had its breakthrough later in Denmark than in many other countries as suggested by the historical narrative. JEL classification: N1, N9, O1, R4, R11 Keywords: clover, potatoes, agricultural productivity, urbanization Acknowledgements: We thank Philipp Ager, Steve Broadberry, Jouke van Dijk, Casper Worm Hansen, Nicolai Kaarsen, Paul Sharp, and Christians Skovsgaard as well as participants in seminars at the University of Southern Denmark, and the 2014 E EAESEM Congress in Toulouse, Fr ance for useful comments and suggestions. We thank Christian Skovsgaard for helping with digitizing the clover maps. We also thank Troels Kjærgaard Christensen for help with collecting geo- information on market towns. Any errors remain ours. Jensen is corresponding author (phone +45 65504472) * Torben Dall Schmidt, University of Southern Denmark, E- mail: [email protected] ** Peter Sandholt Jensen, University of Southern Denmark, E-mail: [email protected] *** Amber Naz, University of Soutern Denmark, E- mail: [email protected] Notice The material presented in the EHES Working Paper Series is property of the author(s) and should be quoted as such. The views expressed in this Paper are those of the author(s) and do not necessarily represent the views of the EHES or its members 1. Introduction From 1672 to 1901, average Danish market town populations1 increased from 1,826 to 15,493 inhabitants corresponding to an increase by a factor of nearly 8.5.2 The evolution of urban populations is arguably one of the best proxies for economic development historically (Acemoglu, Johnson and Robinson, 2005b; Cantoni, 2013). Therefore, understanding the causes of this evolution provides insights into the causes of economic growth in periods in which it is difficult to measure growth. In addition, urban populations capture the geographical distribution of economic activity as pointed out by Fernihough  and  O’Rourke   (2014). Further, the historical rise of towns and cities arguably matters for present-day urban agglomerations as suggested by Percoco (2014) who stresses physical geography along with institutions as important sources of path dependence in urbanization through economies of agglomeration. In this paper, we zoom in on the interaction between new crops and physical geography as a source of higher agricultural productivity which cause urbanization. We do so by investigating whether the widespread adoption of two new crops—clover and potatoes— affected market town populations differentially across Denmark because of soil conditions. There are several reasons why higher agricultural productivity could affect town development.3 First, it is arguably the case that only societies with a certain level of agricultural productivity can sustain urban centers; Acemoglu et al. (2005b).4 Second, higher agricultural productivity may spur rural-urban migration if it serves to lower demand for labor in the agricultural sector as stressed by for example Nunn and Qian (2011). Third, Gollin, Parente and Rogerson (2007) propose a model in which agricultural productivity is important for the development of societies in terms of income per capita. They argue that countries  initially  suffer  from  the  “food  problem”—a term due to Schultz (1953)—which refers to the fact that many poor countries devote most of their resources to agricultural production in order to meet subsistence needs due to low productivity. The present research investigates  how  two  potential  solutions  to  the  “food  problem”  used  in  Denmark5 and other North European countries affected urbanization. Finally, historically it was the case that 1Based on 56 market towns for which data exist for the population over the whole period 1672 to 1901. 2This development runs parallels to what happened elsewhere in Europe (Acemoglu, Robinson and Johnson, 2005a; Van Lottum, 2011; Percoco, 2013). 3Other channels such as the impact on nutrition and health may also be important. 4See also Glaser (2014) who stresses agricultural productivity as historically important for the development of cities. 5Kjærgaard  (2003)  observes  that  Denmark  was  one  “among the many European countries that experienced more or  less  identical  problems.”  This  suggests  that  the  lessons  from  Denmark  may  apply  more broadly. 2 Danish market town dwellers were directly involved in the agriculture just outside their town, and therefore they would experience productivity increases directly.6 Clover and potatoes contribute to agricultural productivity via a number of principal advantages. The first advantage of growing clover is that crop yields increase since clover is a nitrogen-fixing plant. In fact, nitrogen governs the yields of crops that have enough water (Kjærgaard, 1995:p.3). As stated by Taylor (2008), clover was “the   chief   provider   of   atmospheric nitrogen for cereals.” The second advantage is that clover provides excellent animal fodder, which allows for a larger cattle population and increased production of milk and butter. Fixing nitrogen may also be important for potato adoption as suggested by Kjærgaard (1995). He argues that the slow expansion of potatoes was caused by an absence of “biological  opportunity”  which was eventually provided by clover. Unlike clover, potatoes do not increase the supply of nitrogen in the soil (Nunn and Qian, 2011), but they do have a number of other advantages. These advantages include that (1) they produce higher yields than many old world crops; (2) they can be used as animal fodder; (3) they are excellent human food since they are superior in nutritional terms, (4) they allow for crop diversification, and (5) they can be planted on land otherwise left fallow (Nunn and Qian, 2011). We discuss these advantages in more detail below. Both clover and potatoes may have been crucial for economic development. For example, Taylor (1983) proclaims that  clover  had  “a  greater  influence  on  civilization  than  the  potato.” Clover is also emphasized by Chorley (1981) for northwest Europe and by Allen (2008) for England. Kjærgaard (1991, 1995, and 2003) highlights the importance of clover for the Danish case and notes that contemporary observers of Danish agriculture also praised clover. One example is the founder of the Royal Danish Agricultural Society, who argued that clover would  transform  Denmark  “into  a  new  Canaan  with  milk  and  honey”. Similarly, 19th century agronomist  Karl  A.  Jørgensen  proclaimed  that  clover  was  “a  blessed  gift,  one  of  Denmark’s   great  benefactors”  (Kjærgaard,  1995:p.11).  Regarding the potato, Langer (1963) associated larger populations with the spread of potato agriculture in Europe, and Adam Smith emphasized that the food produced on a field of potatoes is “much  superior  to  what  is   produced  by  a  field  of  wheat” in the Wealth of Nations. In line with this, Nunn and Qian 6For example, Elkjær (2001) stresses that Danish market town dwellers were simultaneously craftsmen, merchants and farmers as they grew crops on the fields outside their town. She also documents that three of the market towns in Jutland in the western part of Denmark all had clover and potatoes production at the time of the first agricultural census of 1837. For the 18th century, Mikkelsen (1993) demonstrates substantial production in the vicinity of market towns on Zealand in the eastern part of Denmark, but at this point in time neither potato or clover production is reported. 3 (2011) find that the introduction of the potato can account for about a quarter of population growth from 1700 to 1900. The potato has also been stressed as important by Ax (2008) for the case of Denmark. The Danish case has at least four features that make it appealing for testing the effects of widespread adoption of clover and potatoes. First, the Danish data show considerable regional variation in soil suitability for growing potatoes and legumes such as clover. They also show variation in the places which adopted clover early. Suitability measures arguably capture exogenous variation, and we therefore worry less about reverse causality bias regarding potatoes. For clover, we test the effect of adoption using data from Kjærgaard (1991) on the distribution in 1805.7 By this year, clover had spread to many places in Denmark. Since these data indicate adoption, reverse causality concerns naturally arises as early adoption may be caused by increased demand from rising urban populations. To address this, we instrument clover adoption by using suitability for growing alfalfa—another legume—which according to the historical narrative had its breakthrough after the period studied. This allows us to gauge the relative importance of clover and potatoes, which has not been done in previous literature. Second, detailed data for clover adoption are not available for other countries, and the Danish data therefore provide a unique opportunity for investigating the role of clover for urbanization. In related research, Allen (2008) carried out simulations, which attributes more than 50 percent of the rise in crop yields in England between 1300 and 1800 to nitrogen fixing legumes such as clover. Yet, Allen (2008) had no measures of local clover adoption and local economic activity needed to evaluate the effect econometrically.8 The breakthrough of clover in Denmark has been placed from the early 19th century to the 1830s. In some countries, the breakthrough seems to have happened earlier, whereas in countries such as Denmark and Sweden, this happened later (see Kjærgaard, 1995: p.5).9 The relatively late 7The adoption of clover spread gradually from 1775 to 1805, according to Kjærgaard (1991), where adoption in 1775 was very limited. We also explore this variation. 8Chorley (1981) provides a calculation suggesting that legumes were important in expanding the nitrogen supply from 1750 to 1880. Yet, like Allen (2008), he provides no econometric analysis. 9He gives the following years as indication of some adoption for the following places: Andalusia (before 1270), Brescia (1550), Flanders (1563), France (1583), England (1620), Mainz (1645), Fehmarn (1710), Würzburg (1720), Schleswig-Holstein (1725), Berlin (1735), Moravia (1740). For Sweden, Mats Morell indicated in personal communication that clover had early introductions in Sweden in the early 18th century, but that it was only about 100 years later that it had its breakthrough. 4 breakthrough,  combined  with  Kjærgaard’s  data,  allows us to investigate the effect of clover on urbanization. Third, some confounding factors can be ruled out. Denmark does (and did) not have coal deposits, and this rules out that proximity to coal confound the results. Fernihough and O’Rourke  (2014)  demonstrate  a  relationship  between  proximity  to  coal  and  city growth after 1750. Their maps indicate that the Danish cities in the Bairoch dataset were all far from coal reserves.10 Falbe-Hansen (1887, pp.7-9) emphasized that agriculture was the most important sector in the Danish economy as late as 1880. He noted that more than 50 percent of the population earned its living from agriculture according to census data from that year. Also, while the potato was introduced in Denmark in the period studied, this was not the case for other new world crops such as maize.11 Finally, Cook (2013) argues that countries with higher shares of lactase persistent population, i.e. the share of the population that can digest milk, benefit more from the potatoes. In the Danish case, lactase persistence is almost 100 percent, Ingram et al. (2009). This suggests that the benefits can be enjoyed in full. Focusing on Denmark also allows us to control for the effect of concurrent institutional changes which are difficult to control for in studies that exploit variations in city size across borders as done by Nunn and Qian (2011).12 During the period studied the kingdom of Denmark was subject to serfdom from 1733 to 1800, and we can control for this by including common time effects. The common time effects arguably also capture other common institutional changes that were taking place in this period. Further, there were even some (early) institutional differences within Denmark, and we demonstrate that these do not drive our results. Fourth, the data for Danish market towns have a number of advantages relative to the Bairoch et al. (1988) and de Vries (1984) city population data used by Nunn and Qian (2011) and others in analyses of European cities. The first advantage is that we can evaluate more tightly when potatoes (and clover) had their breakthrough in Denmark. This cannot be done with the 10We note below that deposits of brown coal were mined on the island of Bornholm. Since we cannot obtain data on this, we opt to leave out this remote island for this and a number of other reasons in our baseline results. 11Thaarup (1820) discusses some attempts to grow maize in Denmark from the late 18th century to the 1810s. While, the experimenters report some success, later sources argue that maize could not be grown for human food in Denmark due to the growing season being too short (Madsen-Mygdal, 1912; Encyclopedia Denmark). The same source discusses some experiments with using maize as a fodder crop, but this use was recent and not widespread around 1912 when Madsen-Mygdal published his handbook. According to FAO soil suitability measures, sweet potatoes cannot be grown successfully in Denmark though this crop reached Europe very early (Nunn and Qian, 2011). 12They also consider the effect of potato introduction on the heights of French soldiers, and in this setup they control for country-specific institutions as well as changes to these. We note, however, that the outcome of these regressions is heigh, and not urbanization. 5 Bairoch et al. (1988) dataset, which only has data for 10 Danish towns from 1600 to 1850.13 By contrast, the Danish data include 56 market towns with data available from 1672. Nunn and Qian (2011) assumed that the potato had its breakthrough from 1750 onwards in Europe as a whole. However, the historical accounts for Denmark and Sweden14 suggest that this year is too early for these countries. For Denmark, the timing has been argued to be from the 1820-30s onwards. Thus, the Danish data allow us to relax the assumption of a widespread breakthrough from 1750 onwards and investigate whether the historical narrative fits the quantitative evidence. The second advantage is that the data are not selected on the basis of a size criterion. By contrast, the Bairoch et al. data include towns that reached 5,000 inhabitants at some point in the period 1600-1800. In this way, towns that historically reached a certain size are selected into the dataset. This increases the risk of using the wrong counterfactual when carrying out empirical analyses. Using a larger set of market towns, which also include those that did not reach 5,000 inhabitants, but might have similar soil conditions for growing potatoes and clover reduces this risk.15 The third advantage is that the Danish data have higher frequency which increases over the 19th century. This allows for a tighter assessment of the impact of the potato on the growth trajectory of market towns. The final advantage of the Danish data is that they are actual census data. By contrast, Bairoch et al. (1988) construct data at regular intervals—every 100th and later every 50th year—by interpolating between census years.16 Based on a difference-in-difference type estimation strategy which exploit that the two crops should have differential local effects because of variation in soil condtions, we find that both clover and potatoes had an effect on urbanization. Our instrumental variables estimates suggest that clover had its effect mainly in the 19th century, whereas the OLS estimates suggest that towns which had adopted clover by 1805 had high populations before adoption. This may plausibly be attributed to endogeneity bias, which is further corroborated when we exploit time variation in the adoption of clover since we find little evidence of an effect 13Bairoch et al. (p. x) note that they have included all cities that reached 5,000 inhabitants at least once between 800 and 1800. In the Danish case, only Copenhagen and Ribe have coverage before 1600. Nunn and Qian (2011, p. 640) indicate that they use European cities with at least 1,000 inhabitants, but the Danish cities that they include all come from the Bairoch et al. database. This means that 10 cities which had 1,000 inhabitants in 1672 are missing from the dataset used by Nunn and Qian (2011). 14Hecksher (1945) argues that widespread adoption happened in Sweden in the early 19th century. 15As explained below, the Danish part of the Bairoch et al. data for 1700 and 1750 were constructed using data for 1672 and 1769. Yet, for two out of ten towns, data are not included in the Bairoch database for these years. 16Bairoch et al. (1988, p.2) indicate that they have used either 1) the probable evolution as indicated by other sources or 2) an interpolation using the surrounding periods. The data used for 1700 seem to be based on interpolation between 1672 and 1769, as the sources used give data for these periods, and not for 1700 and 1750. 6 before 1805. For the potato, the effects also appear in the 19th century around the time that the historical narrative suggests. Moreover, the combined effect from the IV estimation of the two crops corresponds to roughly 14 percent of growth in market town populations from 1672 to 1901 with a little more than half being attributable to clover. The rest of the paper is organized as follows. Section 2 gives historical background and discusses the advantages of clover and potatoes in more detail. Section 3 describes the empirical strategy and the data. Section 4 presents the results. Section 5 concludes. 2. Historical background This section provides background on the main advantages of clover and potatoes and their historical breakthrough in Denmark. We need to explain in more detail what benefits arise from these crops in order to understand why they would be adopted and how they might impact agricultural productivity. The timing of breakthroughs informs the empirical analysis on when to expect effects from clover and potatoes. Advantages of clover As mentioned above, the first advantage of introducing clover as a crop is that it serves to increase nitrogen supply in the soil. According to Cooke (1967, p.3), nitrogen is in a class of its own, for in most agriculture its supply governs the yields of crops that have enough water. For the historical period under consideration, Kjærgaard (1995, p.4) concludes that the main way to increase the supply of nitrogen was to increase the cultivation of leguminous crops. In northern Europe, this was done mainly by introducing clover, which is considerably better at fixing nitrogen than many other crops such as peas which were already grown, see Kjærgaard (1991, p. 111) who shows that clover adds about 3 times as much nitrogen as e.g. peas. Allen (2008, p.186) notes that experimental data posits a proportional relationship between crop yields and nitrogen input, and his simulations suggest that clover contributed more than half of the increase of nitrogen for England.17 The second advantage is that clover served as fodder for cattle, whereby milk and butter production could be increased. Moreover, as 17Allen (1992) ran regressions using a dataset of 28-35 Oxford shire villages and found no relationship between the share of land used for leguminous crops such as beans and pulses and yields for wheat and  barley.  Allen’s   data do not include clover or yields of clover (which indicate suitability) and are drawn from a very small database. By contrast, when we use data from the agricultural census for Germany of 1886 from the Prussian Economic History Database (Becker et al., 2012) with 518 observations, we do in fact find strong, positive relationships between clover yields and yields of wheat and barley. [Results are available upon request]. 7 pointed out by Mokyr (2009, p.180), better-fed animals produce more fertilizer.18 He also points out that the empirical relationship between clover and soil fertility was known at the time. This is corroborated by Marshall (1929, p. 46) who notes that the relationship was known by 18th century writers such as Stephen Switzer and Robert Maxwell, who posited that clover  draws  nitrogen  from  the  air  and  “gives it  to  the  land.” Finally, we note that alfalfa has similar advantages but as described below, it did not have its breakthrough in Denmark in the period studied. Advantages of the potato The potato has certain advantages compared to other crops. First, they produce higher yields than old world crops. Evidence on this is given in Figure 1 from early 19th century farmer Søren Pedersen from western Zealand, who kept a diary in which he recorded seed and harvest. The yield measure is given per tønde of seed. The unit tønde corresponds to 100 kilograms (plural: tønder). On average, potato yields are higher than the other crops with around 12 tønder per tønde of seed in comparison to 7.05 for wheat and 6.92 for rye. Second, they can be used as fodder for animals, as also emphasized by contemporary writers (Ax, 2009). Third, they are superior to other staple crops in nutritional terms. This was acknowledged by some writers at the time, e.g. Bendix (1807, p.22), but others did not believe that potatoes were proper food for humans due to their role as fodder crops (Ax, 2009). Fourth, they allow for crop diversification and as noted by Ax (2008), this was one of the main arguments that writers of the time  used  to  promote  the  potato.  Finally,  “potatoes could be planted on the land that was otherwise left fallow between the periods of grain cultivation”  (Nunn  and  Qian,  2011: p.601), and in this way one did not need to switch completely away from old staple crops. The case of Søren Pedersen also illustrates this. 18He also argues that the above mentioned prediction that clover introduction would transform Denmark into a “new  Canaan  of  milk  and  honey”  should  be  understood  literally.  Clover  contributed  both  to  increased  milk   production and increased the biological niche for the honeybee. 8 Figure 1: Yield for Potato, Wheat and Rye of Søren Pedersen, 1811-1838 35.00 d e e 30.00 S e d 25.00 n ø T r 20.00 e p er 15.00 d n ø 10.00 T s, d 5.00 el Yi 0.00 1822 1824 1826 1828 1830 1832 1834 1836 1838 Year Potato_yield Wheat_yield Rye_yield Source: Schousboe (1983). Breakthrough of clover According to Falbe-Hansen (1887, p.136), it was only from the 1830s that clover was introduced on many farms in Denmark. Kjærgaard (1991, 1995) provides a short historiography of clover in Denmark. Importantly, he provides maps of its diffusion covering 1775, 1785, 1795, and 1805. Clover seems to have been grown in a single town in 1732, but Kjærgaard (1995, p.6) stresses that it was still regarded as a new crop and only grown experimentally. His maps suggest that only few places had adopted clover by 1775 (see panel A of figure 2) and it was only by 1805 that it had spread to a significant proportion of the country (see panel B of figure 2). The map for 1805 shows limited diffusion in the western part of the country, whereas adoption in the eastern part was much more widespread. It is important  to  note  that  Kjærgaard’s  maps only indicate clover adoption and not the intensity. This would be consistent with a later breakthrough as suggested by Falbe-Hansen (1887). 9