«Stephen Broadberry and Bishnupriya Gupta Department of Economics, University of Warwick, Coventry CV4 7AL, United Kingdom ...»
INDIAN GDP BEFORE 1870: SOME PRELIMINARY ESTIMATES AND A
COMPARISON WITH BRITAIN
Stephen Broadberry and Bishnupriya Gupta
Department of Economics, University of Warwick, Coventry CV4 7AL, United
18 October 2009
Abstract: This paper provides estimates of Indian GDP constructed from the output
side for the pre-1871 period, and combines them with population estimates to track changes in living standards. Indian per capita GDP declined steadily between 1600 and 1871. As British living standards increased from the mid-seventeenth century, India fell increasingly behind. Whereas in 1600, Indian per capita GDP was more than half the British level, by 1871 it had fallen to less than 15 per cent. As well as placing the origins of the Great Divergence firmly in the early modern period, the estimates suggest a relatively prosperous India at the height of the Mughal Empire, with living standards well above bare bones subsistence.
JEL classification: N10, N30, N35, O10, O57 Key words: Indian GDP, comparison, Britain
Álvarez-Nogal and Prados de la Escosura, 2007; Broadberry et al., 2009; van Leeuwen and van Zanden, 2009). This paper applies similar methods to Asia, providing estimates of Indian GDP for the period before 1870. There is a strong need for estimates of Indian GDP during the early colonial period, to assess the strong revisionist claims about Indian economic performance made recently in the context of the Great Divergence debate. Parthasarathi (1998) has made the most striking claims for South India during the eighteenth century, arguing that living standards were just as high as in Britain, while Bayly (1983) has painted a picture of a thriving north Indian economy during the eighteenth century.
This paper presents estimates of GDP constructed from the output side for the pre-1871 period, and combines them with population data. We find that Indian per capita GDP declined steadily between 1600 and 1871. As British living standards increased from the mid-seventeenth century, India fell increasingly behind. Whereas in 1600, Indian per capita GDP was more than half the British level, by 1871 it had fallen to less than 15 per cent. These estimates support the claims of Broadberry and Gupta (2006), based on wage and price data, that the Great Divergence had already begun during the early modern period. They are also consistent with a relatively prosperous India at the height of the Mughal Empire, although much of this prosperity had disappeared by the eighteenth century. Projecting back from Maddison‟s (2003) widely accepted estimates of GDP per capita for the late nineteenth century in 1990 international dollars, we arrive at a per capita income in 1600 of $833, well above the bare bones subsistence level of $400, or a little over a dollar a day. This is more in line with the recent revisionist work on Europe, which suggests that Maddison (2003) has substantially underestimated living standards in the pre-modern world (Broadberry et al., 2009).
The paper proceeds as follows. Section II provides an overview of methods, drawing on previous work reconstructing national income in Britain and Europe before 1800. Section III then describes the procedures for estimating output in Indian agriculture, industry and services, before aggregating the sectoral outputs into real GDP for India during the period 1600-1871. In Section IV, these series are then combined with data on population to derive estimates of Indian GDP per capita, and used to compare living standards in India and Britain. Section V concludes.
II. AN OVERVIEW OF METHODSThe first phase of historical national accounting focused on reconstructing national income for a small number of relatively rich countries in Western Europe and North America, and starting around 1870, at the beginning of the modern statistical age (Kuznets, 1946; Clark, 1957; Maddison, 1982). A natural development was the application of this approach to other parts of the globe, and many non-western countries now have historical national accounts reaching back to around 1870 (Maddison, 1995). For the period before 1870, there has now been a substantial period of experimentation, beginning with the study of British economic growth back to 1688 by Deane and Cole (1967).
Deane and Cole‟s (1967) study was remarkable for the way in which the authors made efficient use of the limited range of processed data series that were available at the time. Subsequent research by many authors has dramatically extended the range of data now available, with the revised estimates of Crafts and Harley (1992) proving an important staging post. Broadberry and van Leeuwen (2008) have now succeeded in producing annual estimates of GDP for Great Britain over the period 1700-1850. Furthermore, Broadberry et al. (2009) have extended the approach back to 1300 for the territory of England.
Deane and Cole‟s (1967) approach now seems remarkably simple in the light of the vast amount of subsequent research. Nevertheless, its simplicity and modest demands on data makes it particularly suitable as a starting point for Asian historical national accounting in the period before the wide availability of official statistics at a national level. We focus here on Deane and Cole‟s (1967) method for the eighteenth century, where they constructed an index of total real output, based on industry, agriculture and services. Their estimates are reproduced here in Table 1. The sector that was most firmly grounded in the data was industry. For the export industries, such as cotton, output was assumed to grow in line with exports, for which abundant data were available. For home industry, production was assumed to move in line with the physical quantities of output of leather, beer, candles and soap. Finally, since Deane and Cole had no independent data on commerce, the index of industrial output was assumed to apply also to the commercial sector. It is not much of an exaggeration, therefore, to say that the whole of the industrial and commercial sector was dependent on the export data.
For agriculture and services, by contrast, the key data series was population.
For agriculture, an index of production was derived by assuming that agricultural demand grew in line with population, which amounted to assuming constant per capita corn consumption. An adjustment was then made for known imports and exports of grain, to convert demand to domestic production. For services, even in modern national accounts it is not uncommon to assume that real output moves in line with employment. Since for the eighteenth century Deane and Cole had only fragmentary evidence on employment, they assumed that service output grew in line with population. For the government sector, however, it was possible to obtain direct estimates of output from government budget sources. Finally, to combine the individual series into an index of GDP, it was necessary to find appropriate weights for agriculture, industry and services. These were taken from Gregory King‟s  social tables, and are given at the top of each column in Table 1.
It should by now be clear that Deane and Cole‟s (1967) estimates of British GDP in the eighteenth century are overwhelmingly dependent on the path of population and exports, with a minor role for government expenditure and a restricted set of volume indicators for home industry. It would not be difficult to assemble a similar data set for India between 1600 and 1871, and that is what we proceed to do in the next section. However, we will not stop there, because work conducted since Deane and Cole‟s (1967) study suggests a number of ways of improving upon this approach, and again in ways which can be replicated with the data available for India.
First, subsequent work on the agricultural sector has allowed for a more sophisticated treatment of demand. Crafts (1976) criticised Deane and Cole‟s assumption of constant per capita corn consumption while real incomes were rising and the relative price of corn was changing, and Crafts (1985) recalculated the path of agricultural output in Britain with income and price elasticities derived from the experience of later developing countries. The approach was developed further by
Allen (2000) using consumer theory. Allen (2000: 13-14) starts with the identity:
where qa is agricultural output, r is the ratio of production to consumption, c is consumption per head and N is population. Agricultural consumption per head is assumed to be a function of its own price (pa), the general consumer price level (pc),
and income (y). Assuming a log-linear specification, we have:
where α1 and α2 are the own-price and cross-price elasticities of demand, β is the income elasticity of demand and α0 is a constant. Consumer theory requires that the own-price, cross-price and income elasticities should sum to zero, which sets tight constraints on the plausible values, particularly given the accumulated evidence on elasticities in developing countries (Deaton and Muellbauer, 1980: 15-16, 60-82). For early modern Europe, Allen (2000: 14) works with an own-price elasticity of -0.6 and a cross-price elasticity of 0.1, which constrains the income elasticity to be 0.5.
Second, a number of authors have used the share of the population living in towns as a measure of the growth of the non-agricultural sector. This approach began with Wrigley (1985), and has recently been combined with the demand approach to agriculture to provide indirect estimates of GDP in a number of European countries during the early modern period (Malanima, 2003; Álvarez-Nogal and Prados de la Escosura, 2007; Pfister, 2008). With the path of agricultural output (qa) derived using
equations (1) and (2), overall output (q) is derived as:
where the share of non-agricultural output in total output (qna/q) is proxied by the urbanisation rate. The approach can be made less crude by adjusting the urbanisation rate to deal with rural industry or agricultural workers living in towns.
III. ESTIMATING INDIAN NATIONAL INCOME
1. Population The first full census of India was conducted non-synchronously between 1867 and 1872, but is usually presented as the first decennial census for 1871. For the period 1801-1871, we use the decadal estimates of Mahalanobis and Bhattacharya (1976), who assembled information collected by the British for the three Presidencies of Bengal, Madras and Bombay, and supplemented this with assumptions about the rate of population growth in the non-enumerated regions. For earlier years, we have drawn on the estimates collected together by Visaria and Visaria (1983: 466), based on a 50year frequency. We use the Bhattacharya estimates for 1751-1801, the mean Datta estimates to link 1600 and 1750, the Wilcox estimates to link 1600 with 1650, and log-linear interpolation for 1700.
Given the hybrid nature of the series projected back from the 1871 benchmark, it is worth noting that Habib (1982: 164-166) provides a useful cross-check for the absolute population level in 1600, on the basis of three alternative methods of estimation. One approach, based on the cultivated area, yields an estimate of 142 million, while an alternative approach based on land revenue suggests a population of
144.3 million. A third method, based on the size of armies, suggests a population of 140 to 150 million. All three estimates are broadly consistent with our population figure of 142 million in 1600.
2. Agricultural output The simplest procedure for estimating an index of agricultural output is to follow Deane and Cole‟s (1967) assumption of constant per capita grain consumption in deriving domestic demand. This is also the approach used by Wrigley (1985) for preindustrial Europe. However, Deane and Cole also made an allowance for net exports of grain, and in the case of India, we shall need to allow for net exports of agricultural crops, particularly during the nineteenth century as exports of cotton cloth declined.
Focusing initially on domestic demand, our first index of domestic agricultural production is simply the index of the population level. Following Crafts (1985) and Allen (2000), however, it is desirable to allow for consumer response to changing real incomes. Table 3 thus sets out an index of real wages for unskilled labourers in India, derived from Broadberry and Gupta (2006) for the seventeenth and eighteenth centuries, supplemented by additional information for the nineteenth century from Mukerjee (1967). Although the precise magnitude of the fall in the real wage from its high level in the early seventeenth century is a matter of controversy, most scholars have acknowledged the downward trend (Desai, 1972; 1978; Moosvi, 1973; 1977;
Indices of domestic agricultural production are provided in Table 4A. The first index is based on the assumption of constant per capita grain consumption, while the second series is derived from the demand model with an income elasticity of demand of 0.5. Whereas the constant per capita grain consumption model suggests a substantial growth of agricultural output with the expansion of the population, the demand model suggests an agricultural sector that was struggling to maintain output at its Mughal peak until well into the nineteenth century.
Turning to the impact of foreign trade, Table 4b provides an index of agricultural exports. This is derived by obtaining the value of total exports in current prices and the share of agricultural crops from Chaudhuri (1983), and deflating the resulting series of agricultural exports in current prices by an agricultural price index from Mukerjee (1967). For the seventeenth and eighteenth centuries, we have assumed that agricultural exports grew in line with domestic agricultural production.