Introduction
Recently, global economic historians have shown a growing interest in the effects of trade globalization on income inequality in colonies and former colonies which produced agricultural commodities for the world market during the ‘long nineteenth century’ (1815–1913). The initial spurt of historical interest in globalization and the ‘great divergence’ that occurred around 2000 focused on the modern roots of inequality between rather than within nations.Footnote 1 Over the last decade, several works have examined the effects of the late-twentieth century growth of global markets on inequality,Footnote 2 but the topic is just as relevant for the nineteenth century. Historians have now begun to systematically ask: who were the winners and losers of the ‘first globalization’? This issue is particularly controversial for regions producing agricultural exports, some of which became comparatively prosperous but also highly unequal. However, precise figures on the distributive effects of the ‘first globalization’ remain few and far between.
In this article, we contribute to this evolving field of research by quantifying the distributional impact of globalization in some of the world’s major agricultural exporters of this age. We develop a novel approach that bridges the gap between the economic history literature on inequality and the literature on price convergence. So far, the phenomenon of declining price gaps between exporting and importing countries has been neglected by scholars interested in the links between globalization and inequality. This is a missed opportunity: the analysis of price convergence allows us to precisely quantify the exposure of regional economies to globalization. The underlying idea is that when trade costs decline or when markets integrate, trade expands, causing export prices to rise and import prices to fall.Footnote 3 Price convergence gained prominence in the study of globalization thanks to a seminal article by O’Rourke and Williamson.Footnote 4 They argued that globalization began in the nineteenth century when intercontinental prices started to converge, implying that for the first time in history, local economies became fundamentally affected by global trade. Footnote 5 Following this lead, price convergence was labelled by de Vries as ‘hard globalization’, as opposed to the ‘soft’ concept of global interactions.Footnote 6 A large and still expanding body of literature has examined trends and causes of price convergence but, to date, only a handful of studies have measured the welfare effects of market integration.Footnote 7 Their focus has been on aggregate gains in Britain and the United States.Footnote 8 This article extends the analysis of price convergence to study the effects of globalization on income inequality between and within regions producing agricultural exports, including also two major European colonies.
We estimate the effects of market integration during the ‘long nineteenth century’ (1815–1913), the period of the ‘first globalization’, on income inequality in three major exporters of agricultural commodities that are blessed with comparatively abundant data on prices and income distribution: British India, Colonial Indonesia, and the United States. While we cover only three countries, the selected cases were important, in terms of their economic weight, as well as rather diverse. British India was the largest European colony of the nineteenth century, a sub-continent in its own right. Colonial Indonesia exported spices directly into Europe as early as the sixteenth century and became one of the most exploited areas when the Dutch consolidated their rule across the archipelago. The United States was a former colony with a vast open land frontier which was mainly inhabited by European settlers. Owing to its diversity, our sample lends itself to generalizations about the distributional consequences of market integration in agricultural exporting economies.
Our study covers eleven commodities imported by Europe from overseas producers. We examine the effects of price convergence on inequality among areas (between regions) and income groups (within regions). We start by estimating nationwide gains, the ‘cakes’ to be divided. Then we distribute their ‘slices’ among producing regions in all three countries, and between income groups within Java and the southern United States. We also estimate what inequality between and within regions would have looked like without global market integration and compare this counterfactual with the actual level of inequality.
We begin with a survey of the global economic history literature on globalization and inequality, with a focus on recent works on our three countries. We then explain our empirical strategy by providing a non-technical summary of our model and a presentation of our database (with details on the model and sources provided in Appendices A and C, in the Supplementary material to this article). The following three sections report our results. We first show that national gains for British India, Colonial Indonesia, and the United States were uneven but significant and were particularly large from cash crops such as American cotton and Java sugar. Next, we show that gains were concentrated in a few selected regions that were also particularly poor, with the result that market integration mitigated inequality between regions. Finally, we show that globalization substantially increased interpersonal income inequality within the southern United States and Java, where plantation owners were the main beneficiaries.
Globalization and inequality in agricultural exporters: debates and evidence
Relations between globalization and the welfare of agricultural exporters have been highly controversial since at least the nineteenth century, when Marx, quoting the Governor General’s report of 1834–1835, wrote that the bones of local weavers were ‘bleaching the plains of India’.Footnote 9 This image was as provocative then as it is today. Standard economic theory, grounded in Ricardo’s concept of comparative advantage, posits that international trade was mutually beneficial, as it delivered an optimal allocation of resources.Footnote 10 Each country would specialize according to its factor endowment: land-abundant countries had a comparative advantage in the production of primary commodities to be exchanged with industrial products from places with abundant human and physical capital. Economists hold that a fall in trade costs would cause trade to grow, increasing prices in producing countries, due to additional external demand—and decreasing prices in consuming countries, due to increased supply – thus benefitting both producers and consumers. Economic models can be used to precisely measure these static gains (assuming constant factor endowments and technology).Footnote 11 However, this opportunity has thus far been barely exploited. Chilosi and Federico’s study of the aggregate effects of intercontinental price convergence on agricultural producers such as Egypt and eastern Europe, as well as on western European and Japanese consumers, remains an isolated example. This article demonstrates that the global integration of cotton and wheat markets during the ‘long nineteenth century’ substantially contributed to the welfare of both producing and consuming nations.Footnote 12
Past and present critics of globalization challenged economic orthodoxy by stressing that specialization in the export of primary products implied dynamic losses, as it negatively affected investments and productivity growth. Losses from de-industrialization have long been a major issue in India’s ‘nationalist historiography’. At the turn of the twentieth century, DuttFootnote 13 claimed that Britain forced her newly acquired colony to maintain free trade exposing her to competition with Lancashire cotton goods, thus destroying the once-thriving Indian textile industry.Footnote 14 In an updated version of this view, Parthasarathi influentially argued that textile imports prevented India from developing a labour-intensive technology suited to its own factor endowment.Footnote 15 Free trade thus consigned the British colony to a lower developmental path than its metropole.Footnote 16 Several India specialists disputed the idea that de-industrialization is to be blamed for the country’s impoverishment.Footnote 17 However, the de-industrialization hypothesis was reframed by Clingingsmith and Williamson and then extended to the whole periphery by Williamson.Footnote 18 He argued that improvements in terms of trade pushed countries in the periphery to specialize in primary exports.Footnote 19 In Williamson’s view, short-term gains from trade were obtained at the cost of jeopardizing long-term growth prospects.Footnote 20 Skills were lost when traditional domestic manufactured products were substituted with imports. Landowners were the main winners and increased inequality weakened incentives to invest in human capital.
The two conflicting views—gains from trade versus dynamic losses from primary specialization—are echoed in emerging debates on globalization’s impact on inequality between regions. As early as 2012, Roy called for global historians to move beyond the ‘great divergence’ debate on north–south comparisons and to shift the focus to exploring how uneven relations with the global economy before 1913 led to relative affluence in exporting regions within the periphery.Footnote 21 True to his word, Roy went on to explore regional differences within British India in 1907.Footnote 22 Using fiscal revenues per square mile as a proxy for income,Footnote 23 Roy argued that coastal regions well-endowed with water, such as Bengal and Bombay, successfully cultivated export crops and thus became better off than upland and low-rainfall interior zones.Footnote 24 Commercial gains were reinvested: Bombay and Bengal were also among the leading regions in the development of modern industry in the periphery during the second half of the nineteenth century.Footnote 25 New evidence from social tables in colonial Africa painted a similar picture. In colonial Ghana, Aboagye and Bolt detected an increase in inequality between the relatively more prosperous south and the poorer northern regions after the growth of cocoa exports.Footnote 26 Similarly, de Haas found that in colonial Uganda cash-crop cultivating districts were better off than those engaged in the cultivation of subsistence crops.Footnote 27
However, questions remain as to whether gains from trade were sufficiently large to offset possible dynamic losses for exporting regions. Thus, in Bengal, on the one hand, land scarcity constrained agricultural productivity and therefore gains from globalization.Footnote 28 On the other hand, Bengal saw a particularly marked and precocious de-industrialization of its exports in the early nineteenth century.Footnote 29 Regional GDP per capita estimates show that between 1875 (the earliest available figure) and 1911, Bengal went from being the third poorest to becoming the least developed province of British India.Footnote 30 Bengal was not an isolated case. For Lindert and Williamson, the American south was much richer than New England in 1774 because it was much more involved in international trade, thanks to its high land productivity of cash crops such as tobacco and indigo.Footnote 31 However, commercial gains turned out to be fleeting. The growth of cotton exports in the nineteenth century coincided with a reversal of fortunes for the southern states, which increasingly fell behind their northern counterparts.Footnote 32 A comparatively small industrial sector, Crafts and Klein showed, accounts for the relative decline of southern states in the postbellum decades.Footnote 33 In turn, slow industrialization in the American south has long been linked to institutions associated with cotton plantation slavery, which, according to recent econometric analyses, led to reduced investments in human and physical capital.Footnote 34 Within the Indonesian archipelago, Java was the poorest island in terms of GDP per capita, as well as other developmental indicators, such as schooling and stature, in the decades around 1900.Footnote 35 Bosma and van Leeuwen speculated that Java was so poor because of, rather than despite, its integration with the world economy.Footnote 36 Sugar, its main staple export, turned out to be an ‘unlucky ticket’ in the ‘commodity lottery’: its price sharply declined in the closing decades of the nineteenth century. The dominance of plantations in Java’s economy led to particularly high inequality, with few gains for local peasants.Footnote 37
Increased inequality within regions is routinely stressed as one of the key mechanisms behind dynamic losses for agricultural exporters. Global historians have long argued that peasants in countries exporting agricultural products did not gain much because others—colonial powers, European expatriates, or local capitalists—creamed off most profits.Footnote 38 Recent research added details and nuance, stressing how the impact of globalization on income inequality was uneven. One key determining factor was access to land. Plantation agriculture, in particular, has been associated with high inequality in several places.Footnote 39 For instance, de Zwart found that the growth of plantations, rather than exports per se, explained rising inequality within colonial Java.Footnote 40 Recent research also stressed the role of indigenous institutions in determining patterns of land use, thus questioning geographical determinism: the growth of coffee exports in the 1920s was associated with high inequality in Java, where plantation land was owned by former aristocrats, while it was correlated with low inequality in western Sumatra, where small landholdings were controlled by communities of peasants.Footnote 41
Another crucial factor in determining the consequences of globalization for inequality within regions was the extent to which market imperfections allowed powerful actors to extract rents from peasants. The infamous Dutch Cultivation System in Colonial Indonesia was surely one of the most extractive institutions of the nineteenth century. At its peak, in the 1850s, it accounted for more than half of the Netherlands’ state revenue and 3.8% of its GDP.Footnote 42 The Cultivation System very effectively exploited its monopsonistic power against peasants, who were paid a fraction of the market price for the cash crops they were forced to cultivate; as little as 7% for coffee, for instance.Footnote 43 Similarly, Tadei found that European farmers in East Africa obtained a fair price (i.e. world price less freights and other trade costs) from colonial trading companies, while native producers in West Africa were paid less—66% of the market price in French colonies and 84% in British territories.Footnote 44 Such colonial institutions are perhaps the best documented cases of extraction but were hardly unique. There is widespread, if less than uncontroversial, evidence that, regardless of their nationality, traders and large landowners, as well as providers of credit and transport, used their market power to squeeze rents out of labourers and small-scale producers.Footnote 45 This is not to say that market imperfections were always the rule. The Shanghai raw silk trade offers a counterexample: competition was strong at all stages of trade and production. In consequence, profit margins were low.Footnote 46 The indigo export market in British India, too, was quite competitive: there were 32 exporting houses and the top six managed approximately two-thirds of total trade.Footnote 47
In summary, an expanding body of literature in global history has cast new light on the nexus between globalization and inequality in countries exporting agricultural products during the ‘long nineteenth century’. However, the relationship between globalization and the welfare of these countries remains controversial, not least because questions remain as to how large the effects of globalization were. This article provides new estimates of the static effects of global market integration on welfare and inequality.
Modelling a counterfactual world economy without market integration
This article estimates the effects of transoceanic price convergence on inequality between and within regions in three major primary products’ exporters: British India, Colonial Indonesia, and the United States. In the first step, we rely on a microeconomic model to compute new estimates of how different prices and therefore aggregate welfare in our three countries would have been without market integration. Moving beyond our earlier study, we add a new set of bilateral trade estimates to that study’s multilateral estimates.Footnote 48 While the multilateral model developed in that study considers interactions between all major exporters and importers, our bilateral model considers interactions between only one producing country and one consuming country, which allows us to expand the coverage from two to eleven commodities. In the second step, these aggregate gains are distributed between and within regions, enabling comparisons between actual and counterfactual (without market integration) measures of inequality. Technical details on the model are provided in Appendix A, and details on the sources are provided in Appendix C. Here, we provide a non-technical summary of the model and present the key features of our database.
British India, Colonial Indonesia, and the United States all shared in the rapid growth in world trade in the ‘long nineteenth century’, although there were substantial differences in extent and timing (see also Figures A3 and A4 and Table A3 in Appendix B). The Unites States stood out for the rapid growth of its trade: it nearly doubled its share of world trade in the ninety years before 1912, from 6.5% to nearly 13%. By the end of our period, the United States exported more than the whole of the Asian continent, whose overall share of global trade remained roughly stable. British India increased its share up to almost 7% in the late 1880s, but experienced a steep fall to below 4% in the wake of the collapse of exports of wheat and the stagnation of those of jute and cotton cloths.Footnote 49 Total Indian exports did recover and by the eve of the First World War they were about a tenth higher than in the 1880s, but India’s share of world trade never returned to the pre-crisis peak. Colonial Indonesia’s share more than doubled in the 1830s when peasants were forced to provide growing quantities of coffee, sugar, and spices under the Cultivation System.Footnote 50 Colonial Indonesia’s share of world exports peaked at 1.8% of world trade in the early 1840s but declined in the 1850s and 1860s and fluctuated around 1% thereafter. Part of the increase in world trade was determined by the growth of trading economies, but market integration was also a key driver. Thus, market integration accounted for about two thirds of the increase of Indian and American cotton exports and one fifth of American wheat exports.Footnote 51
Unfortunately, we have not been able to include any industrial products. The prices of manufactured commodities are scarce and hard to compare: their quality, unlike those of our primary products, typically differs significantly across places and over time. Anyway, as seen in Figure A5 in Appendix B, industrial products were likely a negligible share of exports from Colonial Indonesia. Despite British India’s de-industrialization and eventual re-industrialization, manufactured products never accounted for more than 20% of its exports. The only country in our sample where manufactured products mattered more than primary products was the United States, but even there the transition occurred rather late: after the Civil War when, in the space of two decades, the share of manufacturing in exports went from c. 20% to c. 60%. Moreover, even in the postbellum decades, exports of cotton grew nearly fourfold.Footnote 52 The agricultural products upon which our analysis concentrates thus continued to dominate the exports of the southern United States.
Our sample of commodities can be regarded as quite representative of the export mix of our three producers: they accounted for half or more of total exports from the United States before the Civil War and from British India and Colonial Indonesia for nearly the whole period (Table A2 in Appendix B). It includes an interesting mix of different production structures, mostly plantations (sugar and coffee from Java; indigo, jute and tea from India; and cotton from the antebellum United States and from British India), quite a few cases in which small-scale farm production played a significant role (wheat and, after the Civil War, cotton from the United States, coffee from Sumatra, cotton and wheat from British India), as well as the odd mineral (tin from Colonial Indonesia).
Columns 4 and 5 of Table 1 report estimates of the change in trade costs between ports in producing countries (e.g., Calcutta or Batavia) and in Europe (mostly London).Footnote 53 In our model, trade costs are expected to be equal to price gaps: if they are higher, traders will lose money; if they are lower, trade will generate abnormally high profits, attracting new traders. In both cases, equilibrium would be restored by changes in traded quantities. The price series show an impressive process of convergence over the ‘long nineteenth century’, with different timing between different components of trade costs. Trade costs declined faster in the early phase of globalization, from Waterloo to 1870, than in the alleged ‘heyday of globalization’ from 1870 to the First World War.Footnote 54 Early price convergence was determined mostly by policy changes.Footnote 55 Thus, England abolished the trade monopoly of the East India Company with India in 1813 and the Corn Laws in 1843, which protected domestic production from overseas wheat. Some years later, the Netherlands liberalized the market for the transport of exports from Colonial Indonesia, which as part of the Cultivation System had been previously managed by the Netherlands Trading Society (De Nederlandsche Handel-Maatschappij N.V., or NHM). The existence of a trading monopoly meant that contracts were awarded to well-connected shipowners who charged freights up to four times higher than the world market rates.Footnote 56 Transoceanic trade was broadly free of barriers by 1870; thus, further convergence was driven by technological gains in transport (the steamship) and communication (the telegraph).
Summary of the database (figures in %)
Note: to facilitate comparison, we also report changes in trade costs with Britain for estimates based on the multilateral model, neglecting, e.g., trade costs from British India to Japan.
⸸ Or the earliest available date.
§ 1902–03 for British India and 1926 for Colonial Indonesia. Computations of the Herfindahl index (equal to the sum of the squared production shares) are based on 15 regions in British India, 10 regions in Colonial Indonesia, and 47 regions in the United States.
Sources: columns 3 and 4: Chilosi and Federico, ‘Early Globalizations’ (see Appendix C for details on computations). The figures in column 5 are computed with statistics drawn from a wide range of secondary sources and period statistics, as detailed in Appendix C. The figures in column 6 are computed from statistics drawn from India Statistics Department, Area and Yield of Rice, Wheat, Cotton, Oilseeds, Jute, Indigo, Sugar Cane for Various Periods from 1891–92 to 1902–03 (Calcutta: Office of the Superintendent of Government Printing, 1903), 30–38 for British India; Centraal Kantoor voor de Statistiek, Mededeelingen van het Centraal Kantoor voor de Statistiek: De Landbouwexportgewassen van Nederlandsch-Indie (The Export crops of the Netherlands East Indies in 1926) (Batavia: G. Kolff & Co, 1926), 90 and Amarjit Kaur and Frits Diehl, ‘Tin Miners and Tin Mining in Indonesia, 1850–1950’, Asian Studies Review 20, no. 2 (1996): 95–120, for Colonial Indonesia; and the ATICS database in Federico and Sharp, ‘The Cost of Railroad Regulation’, 1017–38, for the United States.
Our model assumes that producers received the full amount of prices at the country border. This is clearly an oversimplification. Producers had to pay transport costs from farms to the port of embarkation and possibly also fees to intermediaries who might have extracted some monopsony rents.Footnote 57 The existence of these costs would not affect our aggregate estimate of gains if transport costs and intermediaries’ fees were paid to residents. However, if the residence and income group of the intermediaries differed from those of the producers, commercial costs could affect the distributions of gains across space and income groups. Domestic transport costs and, arguably, associated rents declined thanks to the building of railways in Colonial Indonesia, British India, and the United States.Footnote 58 Similarly, the demise of the Cultivation System reduced gaps between export prices and payments to farmers in Colonial Indonesia. It is possible that market imperfections also declined in other countries. Thus, domestic market integration might have increased the ‘share of the cake’ going to producers, but lack of series of farm prices prevents any estimation of the size of these effects. Nevertheless, we can be confident that biases are unlikely to be sufficiently large to alter our conclusions. In both British India and Colonial Indonesia, producing areas that benefited the most, Java and Bengal, were also major export hubs. Wealthy merchants and plantation owners both belonged to top income groups; hence, our computations of the effects of market integration on inequality within regions would be hardly affected by violations of our assumption that wealthy merchants did not share in the gains. Besides, as stressed in the previous section, the literature has shown that ‘size’ mattered. It is therefore unlikely that intermediaries could squeeze large rents from crops cultivated on plantations, which were very large economic players. As plantations accounted for the lion’s share of our estimated gains, their distribution is unlikely to be sensitive to violations of our assumption that producers rather than intermediaries benefitted from market integration.
Our bilateral trade model is a modified version of a standard neoclassical partial equilibrium analysis of the negative welfare effect of protectionism—the so called ‘Harberger triangles’ (for details and formulas, see Appendix A). These triangles measure welfare gains from the abolition of positive duties on a particular good. Our model, following Hufbauer, more realistically examines the effects of a change between positive levels of trade costs.Footnote 59 Total welfare gains for producing countries are determined by increased production and exports, as well as higher prices. Higher prices benefitted producers but also harmed local consumers. The upshot was that aggregate gains increased with the shares of each product in total GDP and of exports in production. In other words, gains depend on the share of net exports in GDP (Table 1, column 6). This share was, for example, particularly low for American wheat, large quantities of which were consumed domestically, while it was comparatively high for Indonesian sugar, which accounted for a sizeable share of GDP and was nearly all exported.
We estimate gains relative to GDP in 1913 rather than in any other year for two reasons. First, price ratios between producing and consuming countries in 1913 were close to one: trade costs were low and thus there was no room for rents from imperfections in the international transport market (see Table A10 in Appendix C). Second, our estimation requires data on agricultural output and consumption, which are available for all countries as of 1900. In our counterfactual setting we ask how different prices and the aggregate welfare of exporting countries would have been in 1913, with the technology and factor endowment of that year, if the trade costs had been as high as those in 1815, or in 1870. The first comparison estimates the gains over the whole ‘long nineteenth century’; the second, the gains during ‘late globalization’ (1870–1913); the difference between the two is a crude estimate of gains during ‘early globalization’ (1815–1870). Our estimates of aggregate welfare effects need a set of elasticities that measure how responsive demand and supply were to changes in prices. We draw these parameters from an extensive survey of the literature, matching as closely as possible the period, place, and commodity. We thus obtain separate estimates of gains by product, which we sum to obtain total gains for the country. As with all estimates of static gains, they neglect the long-term dynamic effects of globalization: in all likelihood, factor endowment and technology in 1913 would have been different if trade costs had been much higher and thus exports had been much lower. However, the implicit counterfactual is too complex to be manageable with available economic models and data.
We then distribute nationwide gains between areas and income groups, simply assuming that they were proportional to the share of each region or income group on each product’s output. Particular gains therefore tended to increase with the concentration of production. For instance, the spatial concentration was much lower for American wheat, which was cultivated in more than two-thirds of the states of the union, than for American cotton, which was produced in less than one-third of the states, all located in the south (Table 1, column 7). Following standard approaches in the literature, we measure inequality between regions with the population-weighted coefficient of variationFootnote 60 of income per capita and inequality within regions with the (agricultural) labour income share or the Gini coefficient and the top 1% income share, depending on the data availability.Footnote 61 We compute counterfactual distributions by region and income group by deducting gains from market integration from 1913 GDP.
In our baseline, we rely on the textbook ‘Marshallian’ allocation across income groups and assign gains to profit-earners.Footnote 62 However, we subsequently relax this assumption in two ways. In a first ‘Ricardian’Footnote 63 extension, we allow the reallocation of capital to erode incumbents’ profits in the long run, with the consequence that at least part of the gains end up transferred to the owners of fixed factors of production, in our context of land and specialist engineering skills.Footnote 64 In a second ‘Keynesian’ extension, we allow for frictions in the labour market, which imply that some of the gains accrued to workers who would have been unemployed had there been no market integration.Footnote 65 Having discussed our empirical approach, we are now ready to present our results.
National gains
Figure 1 reports national gains for each producer by good and period, relying on multilateral estimates when available and on bilateral estimates otherwise, as explained in the previous section. We also report the results of a sensitivity analysis, assuming demand and supply elasticities to range between 66% and 150% of their baseline values. All estimates refer to GDP rather than GNP and thus include transfers to foreign owners. Such transfers were significant for sugar but not sufficiently large to substantially change our results.Footnote 66
Welfare gains as a share of national GDP (%), 1815–1913.
Source: Table A4 in Appendix B.
Note: Figures a to c have different scales; thus, the heights of the bars are not directly comparable. The range plots use two-thirds and three halves of the elasticities in the baseline specifications.
Almost all the gains for the United States and for Colonial Indonesia came from cotton and sugar. However, the causes behind this dominance were different for the two producers. Wheat and cotton accounted for roughly the same proportion of American GDP in 1913, but the share of domestic consumption was much greater for wheat than for cotton (85% vs. 30%). The extent to which gains for wheat producers offset losses for consumers is correspondingly lower than that for cotton. Colonial Indonesia gained almost exclusively from sugar because trade costs declined much less for coffee and remained almost constant for tin (Table 1, columns 4 and 5). Furthermore, in 1913, coffee accounted for a small share of GDP (0.45% vs. 2.96% for sugar) because production levels had not fully recovered from the destruction caused by the rust disease in the 1880s.Footnote 67 Nevertheless, a simple calculation confirms that the dominance of sugar is not simply driven by the use of 1913 as the baseline year: if we let coffee account for the same share of GDP as sugar in 1870, gains from coffee in 1815–1913 can be estimated as equivalent to 0.47% of Colonial Indonesia’s GDP in 1913, compared with 1.87% gains from sugar in our baseline.Footnote 68 In contrast to the other two producers, British India gained from a wide range of goods. The most relevant products, cotton and wheat, accounted each for between one-third and one-fourth of the total gains. Next came jute, accounting for a sixth of the total surplus. Both cotton and jute accounted for 2–3% of GDP and were mostly exported (60% cotton and 70% jute). The contribution of the two major food crops was relatively small, despite their high share of GDP (20% rice, 5% wheat) because 95% of the rice and 80% of the wheat were consumed domestically. Other goods accounted for less than 1% of the GDP in 1913.
Comparing gains in the long run and in 1870–1913 offers additional insights into the timing of globalization. Both are expressed as shares of 1913 GDP; thus, the ratio between 1870–1913 and 1815–1913 is a crude measure of the proportion of gains during late globalization. The evidence of faster price convergence before 1870 (Table 1, columns 4 and 5) would suggest that the ratio was low. This is indeed the case for American exports (15%), where the price series cover the whole period since 1815, but it is also the case for indigo (6%). The ratios were greater for some other commodities, such as coffee (33%) and jute (67%), partly because the price series started later (Table 1, column 2). We are not always sure why prices were not registered, but we speculate that trade was very small or non-existent because costs were too high to make it profitable. Sugar is somehow an outlier: gains divided almost perfectly between the two periods of globalization because trade costs included hefty tariffs from consuming countries, which were aimed at protecting European sugar beet production and thus were removed only gradually.
Bilateral estimates might overstate gains as they ignore competition from other producers, which also benefitted from a fall in trade costs. Nevertheless, they also fail to consider the benefits caused by exports to third consumers. A comparison of twenty results obtained with the two methods provides reassurance (Figure A2 in Appendix A). Bilateral estimates tend to be only slightly greater than and highly correlated with multilateral estimates, with only one clear exception: the long-run estimate for Indian cotton. This was a special case, as the United States’ dominance of the world cotton market was not replicated for any other commodity. Moreover, it is likely that any positive bias in the bilateral estimates is compensated for by our neglect of general equilibrium effects arising from interactions with other products and factors’ markets: these are expected to be positive, but not large.Footnote 69
Welfare gains might at first sight look rather small. Yet the American figures for the long run compare rather favourably with Irwin’s estimate of the total welfare cost of tariffs, which between 1870 and 1913 hovered around 0.2% of American GDP in 1913.Footnote 70 Moreover, our estimate covers only a part of trade. A rough estimate of total gains from market integration can be computed under the assumption that our products were representative, dividing gains from them by the share of trade they covered. Such estimates of total gains would have been equivalent to 3% of GDP in the United States, 12% in Colonial Indonesia, and 5% in British India or expressed as a ratio of per capita economic growth since 1815, 4%, 30% and 66%, respectively.Footnote 71
A comparison with available estimates of gains from market integration in other places (Table A5 in Appendix B) shows that our countries were hardly exceptional. For instance, in the long run, the aggregate gains of Britain from goods in our sample were equivalent to 1.3% of its GDP;Footnote 72 on this measure, they were close to the total Indian gains (1.6% of GDP). British per capita GDP, however, was about seven and a half times that of British India; thus, the absolute gains were much greater in Britain than in India. This comparison highlights how the absolute size of gains could steeply increase with openness: in 1913, Britain exported 22.4% of its GDP, nearly twice as much as India (11.7%).Footnote 73 Openness was still greater in Egypt (29.2%), which gained as much as 5% of its GDP thanks to the integration of the world market for cotton alone. This gain is more than ten times greater than the corresponding figure for the United States (0.4%), which had a large domestic market and was the least open country in our sample (7.5%).
In brief, aggregate welfare gains, while unevenly distributed across products and places, were significant. Who obtained them? The next two sections answer this question, starting by looking at the distribution of gains between regions.
Distribution of gains between regions
The production of goods for export was geographically concentrated in all three producing countries (Table 1, column 7). In the United States in 1913, seven states, all in the south, each produced more than 5% of the total cotton output; their joint contribution was 90%. Wheat was grown in 32 American states, but the top seven states by output accounted for 60% of the total supply.Footnote 74 Tin in Indonesia was extracted from mines in Bangka and Belitung and other small islands off Sumatra’s eastern coast.Footnote 75 In 1926, central and eastern Java produced almost all sugar and approximately half of the coffee, while Sumatra produced nearly all the remainder.Footnote 76 The supply of export crops was also fairly concentrated in British India.Footnote 77 The share of the largest state by output ranged from one-third for wheat, rice and cotton to almost nine-tenths for jute (Bengal). The shares of the two top states ranged from one-half for cotton to 96% for jute. However, these regional pockets of specialization were scattered across the subcontinent. As a result, the overall concentration was low: eight states produced more than 5% of the total output of the export crops we cover, and the most important of them, Bengal, did not reach a quarter of the total.
Figure 2 plots the estimated gains separately for the long run and the ‘late globalization’ period (1870–1913) by geographical unit as a share of GDP in 1913. Footnote 78 Cotton-producing states in the southern United States gained handsomely from market integration, especially in the long run: without market integration, South Carolina’s GDP would have been 5.9% lower over the long run and 0.60% lower since 1870. Long-run gains close to 4% were also enjoyed by Alabama, Mississippi, and Georgia. Such large profits are consistent with stable export cotton prices in the antebellum decades, given that productivity gains imply declining cotton production costs.Footnote 79 In contrast, the gains from wheat were rather small. The maximum was 0.2% in the long run for the Dakotas in the Midwest, which accounted for an unusually large share (13%) of American wheat production in 1913.
Turning to Colonial Indonesia, the outer islands gained little, except for north Sumatra, which included the tin-producing islands (with gains of 2.3% since 1870). Within Java, too, gains were very unbalanced, ranging from 0.8% in the west to 5.6% in the east. As stated in the previous section, Indonesian gains were mainly driven by sugar. Cane sugar had to compete with the rapid growth in the supply of beet sugar in the decades around 1850.Footnote 80 Yet, thanks to market integration, the export price of Java sugar held on—though it did not rise—until the mid-1880s. By then, there was no longer room for markets to further integrate and the price started declining (Figure A6 in Appendix B). Footnote 81 Admittedly, we have not been able to consider export products such as tobacco, mostly from Java, or oil, rubber or copra from the outer islands.Footnote 82 However, these products became relevant only quite late during our period.Footnote 83 Moreover, hardly any of the profits from oil were kept in the local economies.Footnote 84 Hence, while the inclusion of more products might mitigate the difference between Java and the outer islands somewhat, it is not expected to alter the conclusion that Java benefited the most from market integration, particularly over the long run.
Consistent with the wider range of products, gains for British India were regionally more diffuse than those for the other two countries but were nevertheless rather large in the northeastern states. All the provinces gained something, ranging from 0.08% in Mysore to 4.3% in Bihar and Bengal in the long run. The benefits of late globalization were smaller in general but remained rather significant in Bengal and Bihar (2.7%) and were nearly as high in Assam (2.6%), where most Indian tea came from.
To examine the effects of market integration on inequality between regions, Figure 3 compares the standard measure of dispersion of regional GDP pc, the (population weighted) coefficient of variation, with the counterfactual estimate without market integration around 1913. We obtain these counterfactual estimates by subtracting the regional gains from market integration: the difference from the actual level of dispersion can be interpreted as the effect of market integration on regional inequality.
Inequality between regions was particularly high in British India. In 1913, the GDP per capita of the richest region, Burma, was three and a half times that of the poorest, Bengal and Bihar. As we have just seen, Bengal and Bihar also happened to be the Indian provinces that benefited the most from market integration. Overall, the population-weighted coefficient of variation of regional GDP declined from 0.401 to 0.396 between 1875, the earliest available date, and 1911. Market integration accounted for almost three quarters of this very small change. Inequality between regions was also relatively pronounced in the United States, where average per capita production in the top five states (Dakota, California, Nevada, New York, and Washington) was more than three and half times greater than that in the five bottom states (South and North Carolina, Alabama, Mississippi, and Georgia), all located in the cotton-producing south. Thus, market integration of our products reduced inequality between regions there, too. However, it is not possible to precisely quantify its contribution to overall trends in inequality between regions: the effect of market integration was significant only in the long run and in the first decades of the nineteenth century GDP data by state are unavailable. In Colonial Indonesia, as in the other two countries, the main exporting areas (especially central and eastern Java) were poorer than the rest of the country. However, in contrast to the situation in British India, the gap widened after 1870: the (population-weighted) coefficient of variation in regional GDP per capita increased from 0.234 in 1870 to 0.342 in 1904. Therefore, in this case, market integration mitigated the increase in inequality between regions rather than contributed to its fall.Footnote 85
In summary, we discovered a common pattern whereby the spatial distribution of the welfare gains from globalization was uneven within countries, with large gains for the regions producing agricultural commodities for the world market. Since the regions producing export crops tended to be poorer than the rest, inequality between regions was significantly mitigated by these static gains. What about inequality within regions?
Distribution of gains within regions
We measure the effects of market integration on income inequality within regions by comparing actual inequality outcomes with the counterfactual inequality levels in the absence of market integration. We compute these counterfactual measures after distributing gains among income groups. We rely on standard measures of income inequality: the labour income share, the Gini coefficient and the top 1% income share. Available sources make it possible to compute counterfactual distributions of income only in the cotton-producing states of the United States and Java, which nevertheless accounted for respectively 83% and 94% of the gains that we estimate for these countries. Unfortunately, there are no suitable social tables for India: the available tables are distant in time, distinguish too few groups and refer to the whole country rather than India’s export areas. Footnote 86
We distribute gains from the integration of global cotton (and wheat) markets in the southern United States with the factor shares implied by DeCanio’s Cobb–Douglass agricultural production function, using the marginal productivities of labour as estimates of actual labour income shares.Footnote 87 In so doing, we exploit a well-known result in economics. Footnote 88 The producer’s surplus and, therefore, welfare gains would accrue to landowners under both the ‘Marshallian’ and ‘Ricardian’ approaches described above, since they owned almost all of the capital invested. Most gains preceded emancipation (Figure 2e and 2f).Footnote 89 Thus, it is highly implausible that labourers shared in the gains to any significant extent.Footnote 90 It follows that the labour income share in agriculture declined because of market integration. Comparing actual and counterfactual labour income shares (Figure 4) allows us to estimate the size of these losses. Footnote 91
Changes in the labour income share of agricultural GDP in 1913 within the American southern states (%).
Source: Table A8 in Appendix B.
Note: Aggregate figures for the southern United States are simply averages weighted by the share of agricultural value added. We report only long-term results, as gains after 1870 were very small (Figure 2d).
A striking feature of the figure is how low agricultural labour income shares were in the southern United States. The standard textbook figure of the labour income share, which matches post-World War II US data, is 66%.Footnote 92 In contrast, the highest share in Georgia was just over 40% and the lowest in Texas was nearly half of that. Market integration helps to explain this anomaly: the labour income share would have been significantly higher in the absence of integration: 2.76 percentage points in the southern US as a whole and up to a maximum difference of 6.63 percentage points in Georgia. Lindert and Wiliamson detected particularly high increases in inequality and property income shares in the American south, during the antebellum decades.Footnote 93 Our results suggest that market integration was a significant contributor to these trends.
Our estimate of the distribution of gains for Java relies on the 1924 social table reconstructed by Booth.Footnote 94 We assumed these data to be representative of the distribution in 1913, as the effect of the First World War on income distribution was likely small in a colony of a neutral country.Footnote 95 On the island, coffee trees were grown in plantations and sugar was produced in modern ‘factories’.Footnote 96 Canes quickly lose their sugar content and thus must be processed within two days after harvest. In the late nineteenth century, sugar plantations, called ‘factories’ in Java and ‘centrales’ in Cuba, were integrated with modern steam-powered processing plants with internal railways to transport the cane as quickly as possible.Footnote 97 In our baseline estimation, we assumed that all gains accrued to owners of sugar factories or coffee plantations. Eighteen of the approximately 190 factories operating in 1908 were owned by Chinese capitalists, while the rest belonged to Europeans.Footnote 98 We therefore allocate 9% of the sugar’s gains to ‘Asiatic foreigners’ and 91% to ‘Europeans’.
Both counterfactual measures (Figure 5) imply that inequality in Java would have been lower in 1913 without market integration.Footnote 99 The increase in inequality is particularly pronounced if one measures income distribution with the share of the top 1% of incomes.Footnote 100 Europeans living in Java accounted for nearly half of the total income in this category and were the main beneficiaries of any increase in the profits of plantations. Europeans were very few, accounting for a mere 0.18% of the population, and benefitted greatly from market integration; in its absence, their income per capita would have been 34% lower in the long run and 17% lower since 1870. There were twice as many ‘Asiatic foreigners’ as Europeans living in Java. These individuals were the second richest group, but with a significant difference: their income per capita was seven times lower than that of Europeans. Consequently, even if Chinese capitalists owned only a small share of sugar factories, their gains were rather significant: 12% of income per capita in the long run and half of that since 1870.
This baseline ‘Marshallian’ estimate implicitly assumes that the ‘big’ sugar producers received all rents from plantation estates and that there were no other fixed factors (i.e. factors of production the availability of which cannot be changed easily). Footnote 101 Both assumptions are questionable. Planters did not own the land that they allocated to crop production. They leased about half of the land directly from peasants who had been granted collective ownership by the Agricultural Law of 1870. Alternatively, factories made agreements with indigenous rulers or obtained land from the government with 75-year-long leases.Footnote 102 In addition, to function, sugar factories relied on the engineering skills of specialized European workers. We thus add to the list of beneficiaries from market integration the European staff of ‘factories’, the government (and thus ultimately the civil servants who were paid with state revenues), and Javanese peasants. Footnote 103 We implicitly assumed that European human capital and peasants’ land had no alternative uses and thus their returns depended exclusively on sales of sugar. This assumption is highly plausible for specialized European workers and is also an acceptable simplification for Javanese land. Peasants grew rice for subsistence in rotation with cane on the rest of their farms and there was no obvious alternative crop. Furthermore, the market for land was not competitive. Peasants could only lease land to the local sugar factory and contracts were often managed by village headmen and other members of the local elite.Footnote 104 We thus allocate, somewhat crudely, 60% of the total gains to small landowners, 20% to the local elite and 20% to government/civil servants. Any violation of the assumption of asset specificity would imply a bias in favour of Javanese landowners and specialized labour in the overall distribution of the gains. Thus, the actual distribution would be closer to our baseline estimates.
These ‘Ricardian’ extensions to the baseline model imply opposite effects of the increase in production from market integration. Europeans were paid more than natives, and therefore any increase in their number augmented overall inequality. In contrast, increased rents reduced inequality, particularly because we assign most gains to relatively poor small landowners. In 1926, the year of the social table, sugar factories used 179,677 hectares of land and paid 94.3 guilders per hectare as rent.Footnote 105 This sum amounted to only approximately 10% of the total producer’s surplus, suggesting that plantations very effectively exploited their monopsonistic power vis-à-vis the peasants.Footnote 106 Over the long run, without market integration, each peasant renting her land would have lost approximately 3 guilders a year, corresponding to nearly 12% of the average income of a small landowner, according to the social table. This source of income was open only to about one peasant household out of ten, assuming that there were as many peasants renting their land as sugar cultivators.Footnote 107 We compute the no-market integration counterfactual income of European staff by assuming that their population was proportional to the production of sugar.Footnote 108 Without market integration, sugar output would have been 29% lower over the long run and 16% lower since 1870; the total income of European residents would have been 5% and 2% lower, respectively. Factoring in both effects in the counterfactual computations yields results very close to the baseline: the counterfactual 1% top income share becomes 8.8% in the long run and 10% since 1870, as compared to baseline values of 8.8% and 10.1% and to an actual share of 12.6%.
Last but not least, we drop the assumption that labour was perfectly mobile—i.e. that additional demand in cane cutting and factories was fully met by immigration, with no increase in local wages. Even if the cane harvest attracted a substantial number of workers from other islands,Footnote 109 immigration might not be sufficient. Indeed, the demand for labour was very strong: in the mid-1920s, when the sugar industry was in full bloom, it employed a total of 1 million cultivators, 250,000 cane cutters and 150,000 factory workersFootnote 110 out of 4,217,247 agricultural labourers and 1,240,296 ‘workers in Chinese and European enterprises’.Footnote 111 In this ‘Keynesian’ extension, if the market for labour were imperfect, some gains would have trickled down to local workers. Assuming, conservatively, an elasticity of wage with respect to the labour supply of −0.04,Footnote 112 the increased sugar production due to market integration would have pushed up the wages of ‘workers in Chinese and European enterprises’ by 0.38% and of ‘agricultural labourers’ by 0.28%. Footnote 113 Even in this somewhat unrealistic lower-bound hypothesis, market integration benefitted mainly ‘big sugar’ and significantly increased inequality between income groups on the island.
Conclusion
This article generated new quantitative evidence on the effects of the ‘first globalization’ on welfare and inequality within countries exporting agricultural products. We employed a novel approach, bridging gaps between the literature on inequality and that on market integration. We began by estimating the welfare gains of three major exporters of primary products, namely, British India, Colonial Indonesia, and the United States, from the increase in commodity prices caused by the integration of global markets of a representative sample of eleven different agricultural commodities. We confirmed that the gains were substantial and were particularly large for cash crops, such as Java sugar and American cotton. Our aggregate estimates omit gains from other trade flows, but their size suggests that transoceanic market integration contributed significantly to economic growth in the two Asian colonies. The effect was particularly large in Colonial Indonesia, which also grew faster than British India. The economic significance of market integration was not as marked for the United States, which was less open but experienced much higher rates of economic growth than the other two countries. These results are in line with those of our earlier study.Footnote 114
Subsequently, we analysed how the ‘cake’ was divided between and within regions. As production tended to be spatially concentrated, gains were unevenly spread, reaching as much as nearly 6% of GDP in some regions with large exports. The effects on inequality between regions conformed to a shared pattern: since the regions producing export crops tended to be poorer than the rest of their countries, the growth of global markets for our primary products, far from promoting regional divergence within their producing countries, acted as a leveller. We covered most agricultural exports; thus, it seems unlikely that extending the coverage of traded commodities would alter our main conclusion. We neglected to consider that merchants rather than producers might have seized some of the gains. However, the producing areas tended to be export hubs as well. Moreover, the lion’s share of the gains came from crops cultivated on large plantations, which likely obtained favourable conditions from intermediaries. Future research could investigate the extent to which our results are robust after considering globalization’s dynamic effects on investment and technology, which we are unable to capture with the approach we adopted here.
Our analysis of inequality within regions showed that, both in the American south and Java, plantation owners were the main winners. As a consequence, market integration significantly contributed to exacerbating the inequality of their income distributions. Without the integration of the global cotton market, in the southern United States, the agricultural labour income share would have been nearly 3 percentage points higher. In the absence of the integration of sugar and coffee global markets, in Java, the income of European expatriates would have been as much as a third lower; the concentration of income at the top, as measured by the top 1% income share, would have been 4 percentage points lower. In Java, owing to relatively abundant data, we were able to show that our results are robust to several relaxations of our baseline assumption that profit-earners were the only recipients of the producer’s surplus. We allowed other stakeholders, including peasants renting out land to plantations, specialized managers or technicians and other plantation workers to share in the gains. The baseline conclusion remained robust: the predominance of plantation agriculture in the cultivation of export crops meant that market integration significantly increased income inequality. This finding aligns well with the recent emphasis in the literature on the key role of land ownership and land use in mediating the effects of globalization on income inequality.Footnote 115
Our article also contributed methodologically to the global history literature. We developed a new microeconomic model to estimate aggregate gains from market integration. This model is easy to apply because it requires relatively little data.Footnote 116 Combining our aggregate gains with data on the distribution of production and income between and within regions, we developed a new approach to study a paradigmatic global history question: how was the income distribution of agricultural exporters affected by uneven interactions with the global economy during the ‘long nineteenth century’?
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1740022824000056.
Acknowledgements
The authors thank the editor and three anonymous referees for their valuable guidance and insightful feedback, Pim de Zwart for inviting them to the Workshop ‘Globalization and Development in Asia and Africa, 1800–2020’ (Wageningen June 2022), the participants to that workshop as well as the participants to the session ‘Long-term drivers of inequality’ at the annual conference of the Economic History Society (Warwick April 2023) for insightful comments, Alessandro Citanna for comments on the microeconomic model, Tirthankar Roy for his help with the sources, Maanik Nath and Pablo Martinelli for their help with the maps, and Justin Schwartz, Sara Pecchioli and AJE for outstanding assistance with editing the text.
Competing interests
The authors declare none.
David Chilosi is an economic historian, with a specific interest in the quantitative analysis of socio-economic change over the very long-run. Before joining King’s College London, David Chilosi taught economic history at the University of Groningen and the LSE. He studied economics and econometrics at the University of Essex and the LSE and developed an interest in comparative history during his PhD at the University of Surrey. David has published several articles on a diverse range of subjects, including structural transformation, nationalism, education, market integration, public debts and globalization, spanning both European and global history.
Giovanni Federico is an economic historian, with a strong emphasis on data-intensive quantitative work. He started his academic career studying modern Italian economic growth and he is still working on Italian issues (most recently the North–South divide). Later he developed an interest in global and comparative history. He has written books on the silk industry and world agriculture since 1800. He has published several articles on market integration, trade, industrial and agricultural policy, as well as on the development of economic history. He has co-built the Federico-Tena World Trade Historical Database, with series of trade by country from 1800 to 1938.
Open access
