FROM MAJOR POWER TO ECONOMIC BACKWATER: PORTUGAL'S GDP, 1500-1850 Jaime Reis* Conceição Andrade Martins* Leonor Freire Costa** Paper presented at the CAGE conference on Accounting for the Great Divergence at the University of Warwick, 28 May, 2013. *ICS-University of Lisbon **ISEG-University of Lisbon Preliminary version: please do not cite without authorization of authors Contact: Jaime.reis@ics.ul.pt 1 1. Introduction At the beginning of the 16th century, it may not be too much to describe Portugal as a “major power” in Europe. In economic terms it was highly placed on a rough ranking of the per capita GDP of European nations, below Italy, of course, but ahead of Spain, France, England, the Netherlands and Sweden.1 Although a small country, of only one million inhabitants, it had demonstrated a capacity to mobilize resources – finance, technology, human capital and knowhow – on such a scale and to such effect as to enable it to establish a seaborne empire which reached out to India, China and South America. At that time, few were able to replicate this feat. By the 1850s, on the other hand, it had fallen to the last position in the same league table, and what was left of its overseas empire was almost worthless. In a recent paper, Costa et al. (2013) have argued that Portugal’s imperial venture, which lasted for three centuries, was a fairly consistent economic success and by 1800 accounted for some 15 percent of GDP. Yet at this time, it was already some 35 per cent behind England and some 27 percent behind the Netherlands, suggesting that internal factors too had a significant impact in the country’s relative economic decline and divergence vis-à-vis the core countries of the pre-industrial era. The present study uses a freshly mined data set covering the years 1500-1850, as well as some preliminary medieval evidence, to examine this proposition. To do this, we assess Portugal’s overall economic performance during this period. This involves quantifying appropriate aggregates, such as GDP and GDP per capita, indicators of sectoral productivity, and measures of economic structure. The principal conclusion is that between the initial and terminal dates considered aggregate GDP increased substantially though when expressed in per capita terms, it suffered a 1 For evidence supporting this assertion, see below table 8. 2 moderate decline relative to the then two leading economies in the West - Britain and Holland. A second finding is that structural change was on a modest scale, an experience paralleled by most other nations in Europe as far as we can tell from the currently available data. Thus, despite a weak showing, the economy can hardly be said to have “failed”, as has often been thought. Rather, by the standards of Early Modern Europe its performance was creditable, and this despite having one of the fastest growing populations of this epoch and therefore being vulnerable to a likely Malthusian constraints. The present exercise is, obviously, hardly the first of its kind. Indeed, it is one of a number of recent efforts to try and lay the foundations for a new stage in an economic history of Europe which focuses on both intra and inter-continental comparisons, as well as on finding answers to the riddles of the Great and the Little Divergences. At present, the sample in question includes case studies of Britain, Holland, Germany, Sweden, Spain and Italy, and a strong likelihood exists that this oil slick will soon spread to other nations.2 The new Early Modern macroeconomic literature naturally draws on the findings and lessons from the seminal work of Angus Maddison (2001, 2003 and 2007), which preceded it, but improves on it in several respects. One of them is a stronger emphasis not only on the aggregate, but also the sectoral dimensions of the economies which are being quantified. A second one is the necessity this creates for a quantum leap in the amount of new data collected and processed, thus significantly adding robustness to historical conjecture. The third is that it has stimulated the emergence, informally, of a vast and growing common data base, the greatest merit of which lies in its potential to induce further systematic comparative analysis. Finally, the use in each case of 2 To our knowledge, there are projects currently under way for the Early Modern period which focus on Denmark and Finland. 3 analogous research designs means that comparability is greatly enhanced, thus making it easier to estimate multi-country models. The reconstruction of Portugal’s macroeconomic aggregates during the pre-industrial age contributes to our understanding of the economics of the West during the long runup to the Industrial Revolution in several ways. Most obviously, it enlarges the still far too small a sample of countries which up till now have been the subject of intensive quantification. In particular, it sheds much needed light on the less studied group of nations which Allen (2003) has described as the “unsuccessful”, as opposed to the much better understood cases of the two he rightly deemed as the “successful”. This is an for the heuristic value of comparisons between them. Secondly, Portugal was one of the two earliest cradles of European overseas expansion, yet, cannot be said to have suffered a “reversal of fortune” as a result of this. In terms of this important debate, it therefore deserves fuller consideration than it has met with to date. Finally, given the attention received recently by Spain as an unusual case, in Europe, of early high incomes and resource abundance (Alvarez-Nogal and Prados de la Escosura, 2007 and 2012), an obvious need now arises to test this challenging notion by means of a complementary analysis of the only other Iberian country. The structure of the paper is as follows. In the next section, an overview of Portuguese economic history between 1500 and 1850 is presented. Recent efforts at macroeconomic quantification are also discussed and assessed. The third section introduces the methodology and the data base which are employed in the exercise. The fourth submits the results and analyses their significance in terms both of the conventional wisdom regarding Portugal within the context of its own historiography, and the long run implications for the European economy as a whole. The fifth section concludes. 4 2. The Portuguese economy, 1500-1800 Few modern studies have focused explicitly on Portugal’s long-term economic growth before the nineteenth century and even fewer on the possibility of measuring it with standard indicators such as the real wage, GDP per capita or productivity. While the absence of a coherent quantitative research strategy has been a serious obstacle to assessing the country’s macroeconomic trends accurately, it has not prevented conclusions from being pronounced on the subject of how the Early Modern economy evolved. On the whole, the verdict has been on the pessimistic side. Three factors have been put forward to explain this poor performance.3 One is Portugal’s relegation, after 1500, to a semi-peripheral role in the international division of labour of the Modern World-System (Wallerstein, 1974; 1980). This imposed an excessive reliance on foreign capital, shipping and commercial services, and manufacturing which inevitably stunted the development of its most dynamic sectors (Magalhães, 2012). The second was an archaic, technically stagnant agricultural sector which struggled in vain to overcome the Malthusian trap (Justino, 1981; Santos, 2003) and kept the mass of the population permanently at the lowest levels of consumption. An institutional framework which fostered incomplete property rights in land and caused an excessive concentration of income in the hands of a rentier class was at the root of this. Both features seriously distorted the incentives for increasing output and productivity (Magalhães, 2010; Neto, 1997, Oliveira, 1980). The third factor of long-term economic backwardness, and often considered to be the most important, was the colonial system. This had several dimensions. One of them was the succession of colonial booms which diverted resources and entrepreneurship from home manufacturing and held back the diversification of country’s economy (Godinho, 3 For an up-to-date survey of Portuguese economic history, see Costa et al. (2011). 5 1955; Macedo, 1983). At the same time, agriculture languished because of the persistent drain on its labor force caused by the attraction of better living conditions overseas or from the major port cities which serviced the empire (Sergio, [1927] 1984). The riches which flowed from the colonies have been identified by Pedreira (2004) as yet another structural barrier. They made foreign foodstuffs easier to acquire and thus crowded out agriculture and prevented its improvement. But the inflow of colonial wealth had yet another deleterious effect: it promoted the emergence of a bloated, parasitic tertiary sector. This discouraged the rise of the influential, development-minded national bourgeoisie which might otherwise have spearheaded the thrust for structural change (Godinho, 1978). Recent revisions have suggested a somewhat rosier picture, at least as far as the 18th century is concerned. Serrão (2010) has argued that during the seventeen and eighteenth centuries the Portuguese farm sector expanded and became more market integrated,. This led to increased specialization, internationalization and technical progress in some areas. A similar view has been advanced by Santos (2003) for the large scale, commercialized farming of southern Portugal. Meanwhile, according to Pedreira (2004), there was significant expansion and technical change also in both the large scale and the proto-industrial segments of industry, particularly from the 1770s when rising colonial demand provided a significant impulse for manufactured products. 4 Thus far two attempts have been made to quantify long run Portuguese per capita GDP and their results are close. Together, they have rejected the traditionally “pessimistic” view depicted above and have come up with a moderately “optimistic” one instead. In Madison’s case (2001), the estimate points to an increase between 1500 and 1820 of 52 per cent in this indicator, or an annual rate of 0.13 per cent. In that of Valerio (2010) the 4 This manufacturing surge has led Alexandre (1986) to argue that this was Portugal’s great “missed opportunity” for industrial take-off, a leap which would have occurred had the breakdown of colonial ties with Brazil not come about in the 1820s. For a contrary view, see Lains (1991). 6 figures are 72 per cent overall for the period 1500-1800, or an annual rate of 0.18 per cent. There are several reasons for doubting either finding. Maddison’s study assumes, without any theoretical or practical justification, that an estimate of Castille’s long run GDP by Yun (1994) is not only representative of Spain as a whole, but also of Portugal. The Spanish trend is therefore used in both situations without any adjustment, a simplification which calls for re-examination. Valério’s estimate is inferred directly from a single indicator – urbanization – a procedure for obtaining GDP per capita indirectly which others have favoured in the past (de Long and Shleifer, …; Acemoglu et al., 2005). This ignores entirely, however, any impact on this variable that other contemporary changes might have had, for example, in economic structure or in sectoral productivity, particularly agriculture. Two further objections, simply on empirical grounds, can be raised regarding these two exercises. One is the contradiction implied by the fact that the per capita annual growth rate for Holland, an economy universally considered as one of the most dynamic of this era, was of the same order of magnitude, i.e. 0.17 per cent (van Zanden and van Leeuwen, 2012). The other uses a comparison between the evolution over time of Maddison’s and Valério’s estimates and the corresponding Portuguese welfare ratio. The latter is calculated according to the methodology proposed by Allen (2001). This represents real consumption standards welfare per wage earner and is securely anchored in an extensive and reliable data base.5 The three variables are displayed in figure 1. Over the three and a half centuries considered, it reveals a large and increasing discrepancy between the welfare and the output measures which is too large to be plausible. Even allowing that a certain divergence between the two concepts may to be 5 The data employed for calculation these ratios are described in section 3. below. The workers for which this measure was estimated were large city unskilled workers. 7 expected over extended historical periods (Broadberry et al. 2011; Angeles, 2008), it is difficult to conceive of a situation in which GDP per capita could by half, while the welfare ratio fell in the same proportion. The distributional implications of this would be too far fetched. [figure 1 about here] 3.Methodology and data 3.1 Methodology In a pre-statistical era such as we are considering, the choice of methodology for estimating macroeconomic aggregates is strongly conditioned by the availability and quality of the data required. In a case like Portugal’s, where quantitative history is still in its infancy, these restrictions are bound to be severe. Adopting an income side approach to GDP, as Clark (2010) has done, is out of the question in the present state of the art. In addition to wage (agricultural and industrial) and land rent series, it calls for an income from capital time series, which simply does not exist. The alternative of estimating GDP from the output side using direct information regarding the different sector outputs poses equally formidable problems for these countries. Those in which this has recently been tried successfully – Britain, Holland and Sweden6 –are precisely those where a large stock of relevant evidence has been built up over decades of focused research. In the absence of such an infrastructure, the only possibility remaining is an output approach which employs indirect sources to arrive at estimates of agricultural and non-agricultural production, the two essential building blocks for the construction of a GDP series. Not surprisingly, this has been the route followed for Spain, Italy and Germany, and is inevitably that which will be pursued also in the present instance.7 6 7 Broadberry et al. (2011), van Zanden and van Leeuwen (2012) and Schon and Krantz (2012). Alvarez-Nogal and Prados de la Escosura (2007 and 2013), Malanima (2011) and Pfister (2009). 8 The procedure entails two steps and is by now well known and plentifully discussed. We therefore present it in a summary fashion. The first one is inspired by Wrigley (1985) and Crafts (1985) and significantly revised by Allen (2000). It involves a demand-for-food function to estimate gross agricultural output, which is considered equal to food consumption. In other words, it assumes that exports and imports of food were equal and/or negligibly different. Agricultural product (Qa) is given by the expression Qa = I P M N (1) in which P is the real price of agricultural products, I is real income per capita, M is the real price of other consumer goods and N is total population. The coefficients , and are, respectively, the own price, income and cross elasticities of demand. The equivalence between output and consumption, though widely accepted by the literature, is one of the aspectsof this approach which has yet to be examined thoroughly. Not only did agriculture also produce industrial raw materials besides foodstuffs but also self-sufficiency in food is far from obvious in Europe at this time. We have estimated this differential for Portugal insofar as foreign trade in food is concerned and compared it to agricultural output. The result turns out to be very small. Appendix 1 shows the data and calculates the respective impact. The general supposition is thereby confirmed.8 Unfortunately, a possible raw material bias is beyond our reach to assess and is not considered here, as happens indeed in the remaining literature. 8 Non-marketed food production, which was important at this time, is presumed to form a part of our estimate of total consumption given that although not bought or sold, its producers and consumers were in contact with the market and adjusted their decisions to changes in the market’s parameters. Pfister (2008:22) makes this point. For a different view, see Alvarez-Nogal and Prados de la Escosura (2012: 167). For an estimate of this component of agricultural output in eighteenth century France, see Morrison and Sneider (…). 9 The food-consumption function described above takes into account the impact on food consumption of fluctuations in real income, food prices and non-food prices, as well as their respective elasticities. The principal difficulty posed by this approach lies in how to estimate the real income variable, all the others being reasonably accessible. Ideally, it should comprise all forms of factor remuneration (labour, land and capital), as well as their year-to-year changes, but this information is not easily available. To circumvent this problem, several solutions have been advanced. One is to use the real daily wage as a proxy for annual real income (Allen, 2000; Malanima, 2011). A second is to assume constancy in annual wage incomes, with the justification that labour’s effort would vary in accordance with changes in day wage levels, and thus compensate for the latter’s fluctuations.9 A third is to employ a weighted index of annual wages (0,75) and land rents (0,25), as done by Alvarez-Nogal and Prados de la Escosura (2012: 9), which is what we opt for here. To obtain the index of wage income, the standard number of days of work per year is set at 250 as is the current practice, to be multiplied by the daily rate. In a predominantly rural economy, which is subject to pronounced seasonality, this seems large, though less so for proto-industrial workers and perhaps not at all for city-dwelling labour. Likewise, it seems unreasonable to consider that these three categories all enjoy similar levels of pay. In our estimation, we have consequently experimented with three scenarios which take these objections into account. The simplest one replicates Allen (2000)’s formulation, i.e., unskilled wages times 250 days per year for the entire labour force. Greater complexity is introduced by changing this to a weighted index of unskilled labourers earning the corresponding wage and working 150 days a year; skilled urban labourers, who worked 250 days and earned skilled wages; and rural non- 9 This is debated but not implemented by Alvarez-Nogal and Prados de la Escosura (2012). 10 agricultural workers, who are presumed to have been employed 180 days a year and earned a rate equal to the mean of the unskilled and the skilled wages. The last hypothesis and our preferred option is to add to the preceding situation an annual land rent index, with an appropriate weight.10 As will be shown in the following section, the different outcomes of these procedures do not show much divergence. A fairly wide range of choices also exists with regard to the selection of demand and income elasticities. In the absence of any empirical evidence in this matter, the literature has resorted to emulating the present day elasticities of less developed economies with presumed similar traits to those of Early Modern economies. The possibilities go, for own-price elasticity, from -0.4 to –0.7, and for income elasticity, from 0.3 to 0.6. We have found the arguments advanced by Alvarez-Nogal and Prados de la Escosura (2012) convincing and have therefore opted here for the set they propose in which = -0.4 , =0.3 and =0.1. The other major step in estimating GDP is the quantification of the non-agricultural part of the economy. Various possibilities have been tried. Both Malanima (2011) and Alvarez-Nogal and Prados de la Escosura (2012) have postulated a significantly stable relation between the urban share of the population and that of the secondary and tertiary sectors in the economy as a whole. The former has retropolated this link all the way back to 1300 from a linear regression covering the years 1861- 1936 in which nonagricultural output was the dependent variable and aggregate trade and industry industrial served as the covariate. The latter has simply used change over time in the country’s “adjusted” urbanization rate to “proxy those in non-agricultural output per capita” (2013: 14). Both approaches have disadvantages, the principal one being that 10 The third scenario affords us quite a reasonable coverage vis-à-vis total income, if we accept the same distribution as Malanima (2011) has done for Early Modern Italy, namely 70 per cent for labour, 20 per cent for rents and 10 percent for capital. 11 focusing on urban production alone entails implicitly ignoring the contribution of protoindustry to non-agricultural production. A second one is that this overlooks the more than likely rise in productive efficiency which arose everywhere during the industrial era. To surmount these difficulties, we resort to the procedure proposed by Pfister (2009), which presumes that the inter-sectoral productivity gap (p) between agriculture and total output is time invariant over the period considered. The expression he uses is GDPi = Qai / (p *Lai / Li) (2) in which Qai is agricultural output and Lai and Li are agricultural and total labour, all of them at time i, and p is the productivity gap. All that is needed in order to derive the GDP series is to determine this gap at a point in time for which this is possible and all the other variables are known, and then retropolate the desired result back as far as needed. An estimate of p for Portugal exists for the mid 19th century and is obtained independently of the present exercise. Moreover, it is of a credible order of magnitude, namely in the band 0.68-0.72,11 since it is similar to the value derived for Spain (Alvarez-Nogal and Prados de la Escosura, 2007), which was 0.66. It is lower than those for Italy (Malanima, 2011) and Germany (Pfister, 2009), which in both cases were 0.81, suggesting a higher level of economic development. The problem, however, is that the underlying assumption, of constancy for p is a strong one and requires further discussion. In the estimations presented in the next section we display two versions of Portuguese GDP. One is based on the procedure described an employs a constant value for p of 0.7. 11 . The p estimates for Portugal use occupational data from Reis (2000 and 2005) and sector output shares from Lains (2003). 12 This is clearly less satisfactory, for reasons outlined above. The other follows the same course but transforms this coefficient into a variable. This requires two modifications. One is to alter the way in which the inter-sectoral productivity gap is defined. Instead of making it the ratio of the productivities of* agriculture and the whole economy, it uses for p the ratio of the productivities of the agricultural and the non-agricultural sectors. This enables us to introduce a realistic method for adjusting over time p’s value to the shifts in sector productivity. A second modification therefore involves making p vary with the ratio of the TFPs, respectively, of agriculture and industry.12 This is not ideal given, besides manufacturing, the non-agricultural sector also comprises trade and other services, for which we have no efficiency measure. It is a second best solution but allows us to gauge the impact on the final result of changing the initial assumption about the constancy of p. The expression for obtaining this coefficient now is p = Qai / Lai / (Qnai / Lnai) (2) in which Qai is agricultural output, Qnai is non-agricultural output and Lai and Lnai are, respectively, the total labour of these two sectors, at time i. P at time I is obtained by multiplying its 1850 figure by the ratio of the TFPs of agriculture and manufacturing. The change in the way in which p is now conceptualised means, of course, that its value for the mid 19th century benchmark has to be also different. Figure 2 shows the result of this calculation, alongside the inter-sectoral productivity gaps for various other countries at roughly the same time. The order of magnitude is similar and suggests that levels of development had not seriously diverged at this time yet within Europe with the exception of Britain. [Figure 2 about here] 12 Appendix 2 is a brief note on how these two indicator have been estimated. 13 3.2 Data 3.2.1 Information from the market As in other country studies, this one rests on the supposition that markets played an important role in economic activity during the Early Modern era. Market signals are therefore deemed highly relevant for understanding the production and consumption behaviour of economic agents as well as the outcomes of such decisions. Our paper makes use of a data set which comprises the prices both of products and of factors of production between 1300 and 1910. This has been under construction since 2008 and remains so to this day, with the title PWR – Prices, wages and rents in Portugal, 1300-1910. Long term annual series have been gathered for wages, commodity prices and land rents for several regions.13 We restrict ourselves for present purposes to those pertaining to Lisbon and its hinterland, following the widely accepted principle of the “national representativity” of the data from the country’ principal cities, in this case Lisbon.14 In common with other such projects, the data for this one come chiefly from the accounts of religious foundations, the University of Coimbra, hospitals, charitable institutions, municipalities and royal palaces. They refer to actual market transactions, usually on a small scale. In order to homogenize results, all monetary values have been converted into grams of silver and prices have been normalized to correspond to metric units. The existence of gaps in the series, particularly during the 16th century, forces us, in keeping with current practice, to fill them in by means of linear interpolation. Given 13 Quantitative history in Portugal has a long and respectable tradition going back to the 1950s. This has yielded a considerable and mostly published stock of price and wage statistics, though not of land rents. However, the heterogeneity of procedures followed by different scholars make it difficult to use it for long run macro-economic analysis. Hence the development of the project PWR Portugal - aiming at constructing a new data base for all three variables, which can be visited at http://pwr.dev.simplicidade.com/000000/1/index.htm. 14 Allen (2001) and many others. The adoption here of this principle is justified also by the following circumstances. Portugal is a small country (89,000 square km) and Lisbon is located more or less in the middle of it. It has long had reasonable communications by sea and river with most of the country’s regions and their markets. By European standards (Federico 2010), the integration of its markets for basic food products at least in the 18th century was already fairly high (Justino 1988; Santos, 1998). In the case of labour, restrictions on mobility were non-existent and qualitative evidence regarding internal labour migration during this period is plentiful. This too suggests a well-integrated market (Silbert, 1986; Dias, 1998, Reis, 2005b). 14 the impact this may have on the soundness of the results, the degree of coverage of the data being used is summarized in table 1.15 [table 1 about here] a) Wages Although there is a wide array of occupations for which rates of pay are known, we have chosen to use here only the daily wages of skilled and unskilled male adult workers as the measure of the factor raw labour. These wages refer always to employment in either agriculture or the building industry and to situations in which nonmonetary complementary remunerations were entirely absent. Data for “white collar”, usually salaried workers is also abundant but are left out of the present exercise. b) Prices Again, we use only a small part of the entire range of information on commodity prices available. For the agricultural sector we have selected those corresponding to the principal articles of consumption and of production. They comprise wheat and maize, meat, olive oil, wine, eggs and hens, all of which form part of the standard consumption basket used by much of the Early Modern period literature and described in Allen (2001). On the side of manufactures, we take the prices of charcoal, linen cloth, soap and candles all of which belong to this basket too. We further add those for nails, lime, paper and ink, in order to endow our data set with a greater sensitivity to the prices of capital and luxury goods. The consumer price index used in the estimates below adopts the structure of the so-called “Strasbourg basket” for 1745-1755, with the necessary adaptations arising out of the differences in consumption and production patterns between Portugal and northern Europe (Allen, 2001). c) Land rents Land rents have attracted little attention from economic historians who deal with Portugal. At this time, most land in use was not directly cultivated by its lords. Possibly a little less than half of all agricultural land was under commercial tenancy, with leases typically running from three to ten years (Monteiro, 2005). The remainder was held under long term or perpetual emphyteutic contracts, whereby the lord received a fixed fee and the tenant enjoyed an assignable right to the exclusive enjoyment of all the fruits 15 For comparable evidence on Early Modern Dutch statistics, see van Zanden and van Leeuwen (2013). 15 of the land (Costa et al. 2011; Fonseca and Reis 2011).16 For present purposes, we assume that the rent of the first category of contracts provides a reliable indicator for the market value of the services provided by agricultural land as a whole. The argument for this is that tenant turnover in the second, more rigid category, although lower, was not insignificant given the mortality of tenants and their successors. Emphyteutic rents would thus have been more flexible and responsive to market forces than might be expected and closer in their variance to commercial ones. Data for land rents are drawn from two sources. One is the aggregate rental income of a time invariant set of thirty two estates owned and regularly leased by a charitable institution in the region of Alentejo between 1595 and 1850 (Santos, 2003). The other comes from a similar set located in the region of Ribatejo and which covers the years 1560- 1700. We fill the gap left for the earlier 16th century, by interpolating from the land rent index for Spain for the same period published by Alvarez-Nogal and Prados de la Escosura (2012). 3.2.2 Population and occupational structure Long term continuous time series for population and its occupational structure are one of the hardest data problems we have to face regarding Portugal’s Early Modern period. Until recently, long run demographic trends were shrouded in uncertainty. Empirical evidence on a national scale was almost entirely drawn from hearth counts at rare and irregular intervals. The first proper census was not carried out until 1801 (Rodrigues, 2008).The problems this caused were compounded by the uncertainty regarding the average number of persons per hearth, which makes it difficult to transform hearth counts into population statistics. The common practice has therefore been to anchor any historical analysis of Early Modern demographics onto a set of six to eight benchmarks taken from these data and to interpolate linearly between them. A paper by Palma and Reis (2013) has now managed to overcome the first of these informational bottlenecks. Thanks to a fairly numerous set of recent parish level microdemographic studies ranging from the sixteenth to the nineteenth centuries, it has marshalled an array of records of births and deaths respecting most of the principal regions of the country and thus been able to construct a proxy for net national 16 While the first of these arrangements applied mostly to larger units of production, the latter corresponded to small or minuscule farms. Other things being equal, monastic and religious lords had a preference for the latter and lay lords for the former. 16 demographic variation year by year. Instead of linear interpolations linking the few available benchmarks, it interpolates this indicator from the first reliable population count – that of 1527-32 – to deduce a yearly population figure, which is then pinned to another reliable benchmark, such as that of the 1801 census. The terminal date for the series is 1850. It should be pointed out that available registers of births and deaths only start around 1570 so that, for the time being, we are using linear interpolation for the period 1500-70. The demographic profile we obtain for Portugal is presented in figure 3. [figure 3 about here] Unfortunately, a similar continuous occupational structure time series is unlikely to be available in the near future. This forces us to rely on constructing a set of evenly spaced benchmarks for this variable and linking them year to year by means of linear interpolation. We start in 1500 and continue at 50 year intervals all the way down to 1850. Grouping the working population into categories which are relevant to the present purpose is done here at the most elementary level, and yet even so poses enormous problems. To begin with, our categorisation does no more than distinguish between the two most basic economic sectors – agriculture and non-agriculture – since we lack consistent information to permit a more detailed analysis. Some benchmarks allow a breakdown of “non-agriculture” into manufacturing, administration, trade, transport and so on but not all of them. Secondly, we are unable to consider the active population separately from the rest of the population and thus get an accurate measure of labour supply, since the evidence for this does not occur in Portugal until the early 20th century. Our quantification refers entirely therefore to the aggregate population of families dependent on a particular economic activity, without regard for whether their individual members were employed full time, part time or not working at all. We should note that the historical sources we employ, in particular tax records 17, are organised also on a family basis, never on an individual one, and only mention the occupation of the heads of households. 17 Most of this information comes from the records of the décima tax, a family income tax which started in 1643 and left a voluminous though rarely explored mass of documentation. For a good overview of the records this valuable fiscal measure. see Silva (…). 17 The first step in this exercise is to separate the urban population, defined as the inhabitants of agglomerations of more than 5,000 residents. We employ for this a single source, Bairoch (1988), which has two advantages. It provides complete coverage for the period considered and uses the same methodology throughout. Nevertheless, care is needed for how these data are employed. Bairoch included in his aggregate estimate all urban centres irrespective of size. Instead we recalculate his aggregates by excluding all towns with less than 5,000 inhabitants. At the same time, we have recovered the residents of all the towns which Bairoch dropped from his count every time they were not mentioned in a given year in his sources though they are known to have continued to exist. We assume that this was due to an error, omission or simply to a gap in the historical records, and not to a contraction of the population to a figure below the stipulated minimum. We have interpolated the “missing” inhabitants at the level observed in the count of the previous benchmark, as long as this was not less than 5,000. A common assumption in such exercises, which we endorse here, is that the urban population was engaged entirely in non-agricultural activity.18 Non-urban population, on the other hand, consisted of rural dwellers, not all of whom were associated with agriculture in some way or another, and of those in small towns, villages and hamlets of whom the same can be said too. The difficulty lies in how to carry out the partition of the rural population into these two categories. For 1500, we accept (Wrigley, 1985; Allen, 2000) that in Portugal, like the rest of rural Europe, up to the early sixteenth century, agriculture occupied some 80 percent of the population, with the remaining 20 percent corresponding to non-agricultural occupations.19 At the other end of the continuum, we have reliable data from Reis (2005) for 1800 and 1850, and, for 1750, from Sá (2005). For 1700, we use unpublished material concerning personal tax rolls 18 Allen (2000). In particular, we ignore the argument by Alvarez-Nogal and Prados de la Escosura (2007 and 2013) about agro-towns, a phenomenon which was present in Portugal only in the Alentejo region, with 10 percent of the Portuguese population. 19 These proportions are confirmed by the scarce evidence available for late medieval Portugal. See Godinho (1968-71), Rodrigues (1989) and Marques, (1980). 18 from various regions.20 We derive the remaining benchmarks – 1550, 1600 and 1650 – from the shares employed by Alvarez-Nogal and Prados de la Escosura (2007) for Spain in the same years. The results are displayed in figure 4. [figure 4 about here] 4. Results and discussion 4.1. Overview and periodization Figures 5a and 5b present the results of our estimates of total real GDP between 1500 and 1850 in its two versions. As described in section 3.1 above, GDP I is obtained with a constant value for the inter-sectoral productivity gap. GDP II uses a variable instead. Interestingly and reassuringly, the two results have important similarities. They both display steady upward tendencies and strong and frequent oscillations about the trend. Their long run growth rates are within the same order of magnitude, which is small but not trivial when we consider for how long it was sustained. A relevant difference is that the former’s annual rate was 0.38 percent over the entire period, implying a cumulative rise over these three and a half centuries by a factor of 3.7. The second is roughly two thirds of this, with an annual growth rate of 0.23 percent which was equivalent to an overall increase by a factor of 2.2. For the sake of comparison, in a similar period (1510-1897) Holland, one of Europe’s most dynamic nations, achieved a four and a half fold expansion, corresponding to an annual rate of 0.5 percent (van Zanden and van Leeuwen, 2013). [figures 5a and 5 b about here] 20 Data for 1700 exist for the following counties: Avis, Castro Marim, Tavira-Cacela, Portalegre, Viana do Castelo , Montemor- o-Novo and Vila do Conde. 19 From this point of view, there can be no question that Portugal’s Early Modern economy was far from stagnant, contrary to what some have claimed (…). On the other hand, the population it was sustaining in the meantime was experiencing one of the fastest growth rates of Europe at the time, amounting to a rise of about 450 percent in the same time span. When this is taken into account, our evaluation switches from positive to negative, as can be seen from figures 6a and 6b. Real per capita product in version I falls at a rate of -0.19 percent per annum on trend, while version II’s variation falls at -0.07 percent a year. In cumulative terms this represents a long run contraction of, respectively, 30 and 22 percent. Thus, Holland now leaves Portugal behind by a differential of 90 percentage points in the first instance and 82, in the second, a sharp contrast in welfare between the two countries. [figs 6a and 6b about here] Figures 6a and 6b reveal another similarity of interest involving the two per capita GDP estimates. One regards to the long swings of this indicator which brings to light unexpected dimensions of the ebb and flow of this economy. A three sub-period division emerges. The first one covers more or less the whole of the fifteen hundreds, a time marked by very severe fluctuations relating to climatic circumstances (1521-2, ….), but also by the severest secular decline, from 1500 to 1600 (circa 40 percentage points) in per capita output of the Early Modern era. In national terms, this is unexpected given that the conventional wisdom has been to ascribe to this century a positive economic status, partly associated with the emergence of Portugal’s seaborne empire. The second swing in our periodization is an upward one, running from the early seventeenth to the mid eighteenth centuries. Again, this will cause surprise to some, for two reasons. The first is that it contradicts the traditional view of an “iron seventeenth 20 century”, of economic recession and human hardship, assumed to be in consonance with most of Europe and also with the now beleaguered fortunes of the empire. The second reason is the shift to a burst of economic growth during the first half of the eighteenth century. In the traditional view, this has been accepted as the consequence of the lustre of Brazil’s Golden Age. What now emerges as of greater interest still is the amplitude of this movement which the graphs reveal. Figure 6a indicates that, by 1760, GDP per capita had reached for the first time the level attained in 1500, thanks to a cumulative growth of 66 percent relative to the low point of the early seventeenth century. Figure 6b depicts an even more remarkable fact, namely that the level achieved in 1760 may have exceeded the 1500 benchmark by more than 40 percent. The third long swing, which runs from 1760 to the mid nineteenth century, portrays a significant retreat of the economy, which wiped out most of the gains of the previous growth period. At one point (fig 6b), this drove the Portuguese economy, in 1815, to the level of the fearful crisis of 1521. It was a time of natural (the Lisbon earthquake) and man-made disasters (the Seven Years and the Napoleonic wars) which might account, in part at least, and for part of the decline. Political turmoil, even civil war, and loss of empire after 1820 might help account, at least in part, for the absence of recovery from then until 1850. At this juncture, it would have been desirable to participate fully in the scholarly debate concerning the economic performance of European nations in the century and a half following the Black Death. How quick was the economic and demographic recovery from this great shock? What implications may this have had for subsequent economic development during the Early Modern period? In the present state of research, the PWR data base cannot support a backward extension to 1348 of our analysis along the lines we have been following here. All it allows us to do is to create a real wage index for 21 skilled labour going back to before the Plague and joining it up, after 1500, with the real wage time series we have been using to obtain our GDP figures. The result, presented in figure 7, suggests that this as yet unquantified demographic shock was large enough to produce the expected jump in real wages, which, by the end of the fourteenth century, may have tripled. This enormous gain was then eroded as the population gradually recovered on the course of the fifteenth century and finally reached, in circa 1500, its 1347 level, real wages gradually declined, and ended up back at their pre-Black Death level soon after 1500. This total loss of the gains for labour associated with the 1347 cataclysm did not occur in England (Clark, 2007), Holland or Spain (Alvarez-Nogal and Prados de la Escosura, 2012) while in Italy, it did so only much later (Malanima, 2011). We may conjecture that this probably constituted a serious comparative disadvantage for Portuguese long term development starting in the 1500s. [figure 7 about here] 4.2 The Portuguese economy in the European mirror The recent study by Alvarez-Nogal and Prados de la Escosura (2012) allows us to situate our present estimates of Portuguese per capita GDP in the broader historical and geographical context of European national growth patterns between 1550 and 1850. It thereby raises interesting comparative issues. Table 2 contains their compilation of GDP per capita series at fifty year intervals and starting in 1500, for a number of countries adjusted for different price levels in accordance with the short cut procedure pioneered by Prados de la Escosura (2000). They are expressed in US 1850 relative prices, with the UK level at that date held as the term of reference (=100). We now add to the original data, a full run of new Swedish estimates Schon and Krantz, (2012), as well as 22 our Portuguese results, both of them projected back from their respective 1850 benchmark figures. [table 2 about here] Obviously, these figures cannot be treated as the result of a particularly precise methodology, but they are of interest even when just considered as approximations. Three main finding can be drawn from them. As was mentioned in the opening statement of this paper, at the beginning of the Early Modern period, Portugal appears to have belonged to the group of economic leaders, along with Spain and Italy. Those which in later centuries were to lead Europe economically were still at the back of the pack, though the Netherlands can be seen as being by then not far from its early and well-known “explosion” of modern growth. This first impression corroborates the view propounded by Alvarez-Nogal and Prados de la Escosura (2007, 2012) that setting up, in the Early Modern era, vast seaborne empires flung around the globe could only be done by countries endowed with abundant resources and not because they were impoverished and had to seek outlets to overcome this constraint. On the other hand, what made the Iberian nations so far advanced in terms of GDP is another matter which will be taken up for consideration further on in this paper. A second conclusion suggested by table 2 is that although Portugal’s long-run macroeconomic performance was not outstanding, it was not by current standards a failure either, particularly if we leave out the 1800-1850 period out. Slow decline from the sixteenth to the nineteenth centuries was the norm in Europe, which, from time to time, was broken by a burst of growth temporarily inverting the trend. In Portugal, this burst came during the eighteenth century. In particular, if we compare it with Britain and the Netherlands, at times it was ahead of them, other times it fell behind, but until the 1750s, it was never losing ground to them permanently. This evaluation leads us to 23 the need for a definitive reconsideration of the persistent concern with national “backwardness” and “decadence” which has been the hallmark of the Portuguese historiography of the pre-industrial era. This does not appear to have been the time of Portugal’s little divergence. The last 50 year phase we are observing here, from 1800 to 1850, links the Early Modern period with the beginnings of Kuznetzian Modern Economic Growth. Two features stand out in Europe in this epoch. One is that whereas in earlier times, in every half century, some economies, but not always the same ones, progressed and others regressed, this time, all economies but one experienced significant expansion. Britain, of course, stands out in this respect and its impact on the other economies, coupled with the advent of globalization, no doubt contributed strongly to this generalized surge. The other is that Portugal, which was also involved in this process and also endured, like all the others, its share of social, political and institutional disruption, was left out of it. The reasons for this have barely been on the agenda of present day research on Portuguese contemporary Economic History. The latter has remained continually focused on the issue of national economic backwardness, which has been considered as a feature of the latter part of the nineteenth and of the twentieth centuries. Yet, looking at the very long run, as one can here, it is tempting to recognize, first of all, how much these fifty years appear to be the moment when the country evidently broke off from the rest of Western Europe and began persistently to fall behind. In the second place, it is impossible to avoid thinking of the deleterious impact of two then momentous losses for Portugal’s economy: the end of its commercial privileges in the newly liberated colony of Brazil; and the loss of its supremacy in the British wine market. The latter plunged the important wine sector into a deep crisis. The former implied, according to the recent study by Costa et al., 2013), foregoing a gain for GDP 24 in a range of 15 to 20 percent, as was estimated for the end of the seventeen hundreds. The failure to reorganize the economy in order to cope with these two shocks plagued the following century and was possibly explains why in Europe Portugal suffered the strongest divergence of all vis-à-vis the advanced core. 4.3 Iberian parallels or contrasts? The issue of possible similarities between the two Iberian Peninsula’s neighbours are a special though nevertheless interesting case of Portugal’s reflection in the European mirror. The two countries have much in common in their geography and history which is relevant to the present discussion. On the first score, they are southern European, they have important coast lines but difficult internal communications, their natural average aptitude for temperate climate agriculture is less than satisfactory, yet for this reason too they are able to specialize in tradable commodities which are sought by Northern Europe. On the second score, both polities21 emerged after a prolonged struggle of reconquest of land from the Moorish kingdoms and therefore had all the spin-offs of military activity strongly imprinted on their birth marks. As part of this, they set themselves up as nations with unusually abundant land resources, strong towns, free and mobile labour and emphatic political checks on central power. In the sixteenth century, they created impressive overseas empires, which brought amazing riches though had probably much less impact on GDP. The macroeconomic profiles of the two countries during the Early Modern period are also somewhat similar. The trend growth of GDP for Spain had an annual rate for this time span of just under 0.4 percent, that for Portugal was either 0.38 or 0.23 (versions I or II). In per capita terms, up to 1800 they were very much alike too, at which point Spain had a burst of growth while Portugal experienced a decline. 21 We are aware that speaking of “Spain” in the present historical context is a serious anachronism but one which is widely practised for reasons of simplicity and inter-temporal comparability. It is equivalent to speaking of Germany, Switzerland or even France in the same period. 25 In their recent re-interpretation of Spanish economic history between 1270 and 1850, Alvarez-Nogal and Prados de la Escosura (2012) have come up with a challenge to the conventional view on the transition of this economy from the late Middle Ages to the first century of the Early Modern period. Their claim is that this made the country quite different in important ways from others in Western Europe, and may have led to contrasts which were decisive for future developments. The initial regime, starting from the thirteenth century, was based on “a high land–labour ratio frontier economy, which [was] pastoral, trade-oriented, and led by towns. Wages and food consumption were relatively high” (p.1). This would account for Spain’s affluence prior to, as well as for a while after the American expansion. The second regime, from circa 1600 on, was “a more agricultural and densely populated low-wage economy which, although it grew at a pace similar to that of 1270–1600, ended at a lower level” (p.1). Do we find these regimes also in Portugal and with the same transitional timings? Are their respective macroeconomic consequences comparable too? Unfortunately, as noted above, data on Portugal for the period prior to 1500 are too scarce to allow us to make more than a very rough assessment of the nature of the regime which the country experienced between 1300 and 1600. For two centuries before 1500, population appears to have fluctuated considerably, with a possible peak of 1.2 million in the pre-Black Death decades, a low point in the 1420s and a recovery of the 1340s mark around the 1520s (Henriques, 2013). Throughout this time, therefore, density varied a lot but was always low compared to what happened once it started to grow rapidly and consistently over the course of the sixteenth century. In the hundred years before 1600, population almost doubled, causing the land-labour ratio to reach a figure which was under 50 percent of its level in 1500 and probably for a long time before this date too. It seems fair to conclude that in the two centuries 26 before1500, the regime in Portugal was indeed one of land and food abundance, as in Spain. The real wages then being earned were relatively high too, as shown in figure 7 above, or at least far higher than was the experience of the sixteenth century and beyond. Such findings enhance the plausibility of our two papers’ parallel claims regarding the international prominence of Iberia in terms of per capita GDP at the start of the Early Modern period. This situation was to change dramatically, however, in the course of the sixteenth century. Figures 8 and 9 depict, respectively, the profiles of the land-labour ratio and of food consumption per capita from 1500 onwards. It corroborates the preceding analysis by adding quantitative content to it. In particular, figure 9 shows that the early 1600s were the moment when food consumption not only reached its lowest ebb since as far back as we can know, but made a transition (in two successive steps: 1500-1550s and 1550s-1600s) to a new and permanent level 30 percent below the 1500 mark. It was not until the early eighteenth century that any improvement occurred in this respect, but it neither reached the bounty of the pre-1500 era, nor was long-lasting. After the 1760s, Portuguese society returned again to its food consumption amounts of the early 1600s and remained there until well past the middle of the nineteenth century. [figure 8 about here] [figure 9 about here] It seems indubitable that the outline of Spain’s first economic regime described above fits the circumstances of Portugal quite closely and in the same time frame, although more work still needs to be done on this topic. From a structural perspective, the comparison is hard to sustain, mainly owing to the weakness of available data. The role of the urban economy or of its strong trade orientation, for example, are hard to deal with comparatively. In the case of the contrast between the earlier pastoral character of 27 the primary sector and its later more agricultural and land-intensive vocation the empirical ground for assessing Portugal is not particularly firm but tentative inference is supportive of a similar conclusion. Evidence on the composition of Portuguese agricultural output in 1515 shows that animal products were then worth significantly more than arable ones, in other words, more meat than bread was being consumed (Godinho,1968).22 Another sign pointing in this direction is revealed by figure 10, namely the secular increase by a factor of three in the total land yield. This could only have happened with large increases in labour inputs, an occurrence which would have been highly unlikely on this scale in the case of animal husbandry continuing to prevail. [figure 10 about here] 4.4 Structure and agriculture Most European countries in the Early Modern period underwent only very slow, if any change in their economic structures. The exceptions were the Netherlands (van Zanden and van Leeuwen, 2012) and Britain (Broadberry et al., 2011), the only ones to have experienced long term real per capita GDP growth. In such a situation, the efficiency of the primary sector takes on a particularly important role, given that it is subject to the constraint imposed by the fixed factor of production, i.e. land. Population increase in this context becomes a problem, due to the burden of decreasing marginal returns to labour and was bound to weigh heavily on the mean result for all sectors together. Even when industry and services were raising their respective productivities, that of GDP could easily fall. These problems were clearly visible in the case of Portugal, which had one of the fastest growing populations in Europe over these 350 years and a share of agriculture in total 22 By the nineteenth century, these positions had been inverted (Reis, 2000). 28 output slightly below 50 percent.23 This combination raises the issue of how, during this period, the sector was able nevertheless to feed, on a fixed amount of land, a number of mouths which quadrupled, whilst keeping food consumption per capita steady for most of the time, with the exception of the 16th century. It also prompts concern over the issue that although at first sight lacklustre, the performance of the Portuguese primary sector may actually have been rather more impressive than one might suppose. Finally, it helps us to understand why structural change was apparently so hard to achieve in this and other similarly structured economies, where numerous non-food producers weighed heavily on low surplus cultivators. The problem was how to produce, from a fixed amount of land, the enormously increased agricultural output needed by this expanding society, without reducing the marginal product of labour to the point of hopeless impoverishment of the work force. Four circumstances combined to achieve this result. One was land clearance (arroteias), which occurred with varying degrees of intensity over time.24 The second was a gradual shift from animal to arable husbandry, as happened elsewhere throughout Europe, about which little is known quantitatively. The third was the so-called “maize revolution”, which spread throughout the northern half of Portugal and significantly displaced other grains, mainly rye and millet. The fourth was the development and expansion of a commercial wine sector from the late 17th century. All of these processes were gradual and only one – the introduction and spread of maize - might be described as a technological shock (Costa et al., 2011). Unfortunately, too little is known about any of 23 In Spain it was closer to 55 percent but similarly stable over the long run until, surprisingly, the last 70 years of our period. 24 Our view is that by the 16th century, there was little cultivable land which had never had some economic use or another, no matter how tenuous. Given the pastoral vocation of much of the national territory, it seems likely that a lot of the clearance was in order to convert rough pasture into arable. It therefore represented an intensification of land use, but not an increase in “agricultural land”. 29 them. In particular, their respective chronologies, impacts and rates of diffusion remain obscure and tracking these movements with precision is hard. Land clearances had long been the classic way to intensify the use of the fixed production factor in agriculture and thereby raise efficiency. It was a “land augmenting” improvement and unlikely therefore to have entailed gains in labour productivity. They were certainly important before 1500 (Barata and Henriques, 2011) and continued to be so during the fifteen hundreds (Gil 1965). At this time they may even have been the principal positive influence on TFP, since the other improvements had not yet made themselves felt. A glance at the land yield figures (see figure 10) suggests that it was concentrated between the 1540s and the 1600s. Throughout the next two centuries, however, there was an unavoidable decline in the amount of under-utilized land available for putting under the plough. Data from the accounts of several monasteries suggest that new leases for clearing land were becoming less common, with a clear drop from the 17th to the 18th centuries. They probably comprised land which was of increasingly poor quality too (Oliveira 1979; Maia 1991; Neto 1997; Campos 1989; Silva 1994; and Amorim 1994). The introduction to Portugal, in the early 16th century, of American corn (zea maïs) and its subsequent diffusion has been glorified by historians as a “revolution” (Ribeiro, 1963). In fact it took until the first half of the 19th century to be completed.25 Once a garden plant, the advantages of converting it into a field plant were manifold. Its yields were considerably higher than for other breadstuffs (wheat and rye), its caloric content per kilogram was 30 percent greater, and its unit costs of production were lower. In 25 Thorough studies of the Portuguese “maize revolution” are surprisingly sparse. See, however, Almeida (1992). For a world view on this plant, see Langer (1975) and Messer (2001). There is an obvious parallel with the potato but two contrasts are important. The potato revolution came later and was shorter and maize never had to face a natural catastrophe, remaining Portugal’s principal breadstuff until the twentieth century. Nunn and Qian (2011) give reasons why people resisted the widespread adoptions of the potato until late. Historical evidence indicates it did not play a major role in Portugal until the second half of the nineteenth century. 30 addition, it required more labour per hectare, was less sensitive to climate fluctuation and could be used for different purposes besides human food. Adopting maize involved a learning curve for Portuguese peasants, who despite being already familiar with millet, had now to reorganize the field system, create micro irrigation facilities and therefore undertake a certain amount of investment, both physical and informational. It was thus not a simple matter and involved a degree of social transformation and cohesion. Little is known regarding the timing of the spread of this crop. It is clear, nevertheless, that around 1600 its share in total grain output was small, though non trivial, and by 1850 had reached its long term ceiling, having attained 60 percent of total grain production. In between these benchmarks, information from local histories points to a vigorous spread all through the 1600s and the first half of the 1700s (Oliveira, 1990 and 2002). Not surprisingly, a national survey of parishes in 1758, the memórias paroquiais, shows that maize had taken on a leading role in most of the country north of the Tagus river, where conditions were best suited for it. The growing of vine and the production of wine, although hardly a novelty for the Early Modern period, developed remarkably at this time and became the second pillar of the sustained improvement in agriculture that we can observe. Much more is known about this sector, in particular the timing of its expansion, from the late 17th century to its heyday in the 1750s and subsequent deceleration down to circa 1800. Total output doubled between 1700 and 1800 (Martins, 1998). Can we speak in this case of a technological shock? There were two drivers behind this change. One was sustained market expansion, both on the national and international fronts. The latter was significantly stimulated by a change in military and diplomatic relations which led to the signing of the Methuen treaty between England and Portugal in 1703. This gave Portuguese wine a signal 31 advantage throughout the 1700s in the English market, the only one on a major scale in Europe at this time (Cardoso et al. 2003).The other was technical progress on the ground, which occurred in grape growing, the processing of wine and its marketing. As in the case of maize, the conjunction of circumstances meant deep adjustments in the regions where the process made itself felt. It required profound changes in the field system, new equipment and production techniques, and substantial investment. It led to pronounced productive specialization – some districts in northern Portugal became practically mono-cultural – and required the achievement of higher levels of marketing expertise (Serrão, 2010). Two questions are raised by this sketch of the evolution of Portuguese evolution. The first arises when we compare the vague chronology of events presented above with the curve of agricultural output per worker presented in figure 11. This reveals three stages in this partial productivity measure, of which the first, the sixteenth century, appears with a significant fall at a time when only some land enhancing improvement was counteracting the effect of population growth. The second phase covers the next century and a half. It reflects the benefits of the spread of maize and of wine, both of which had tremendous land enhancing effects, besides involving significant capital inputs as well. This happened during a time when the demographic pressure on the stock of agricultural land continued, though much abated by comparison with the sixteenth century. The result was a remarkable increase in agricultural product per worker at an annual rate of about 0.34 percent. These changes were must have been labour enhancing too. The third phase, from the 1750s to 1850, witnesses a severe reversal previous successes, with a negative growth rate in average productivity of 0.35 per cent per annum. [figure 11 about here] 32 Why did this happen? Was it simply a Malthusian downswing following the upswing of the second phase? Although Malthusian influences were certainly involved, two non – Malthusian ones played a role here as well. One was the disengagement between Portugal and England in the late 18th century which caused a considerable loss of share for wine growers and a price collapse in their markets, clearly an exogenous influence. The other was the inevitable slowdown in the spread of maize as good land with irrigation potential became increasingly scarce. The problem for Portugal’s economy at this juncture was that there were no further technological shock available any more, but with population still growing after 1800, the drop in the marginal return to labour became increasingly difficult to contain. The repercussions of this went through wages and caused the downward shift in GDP per capita. 5. Conclusion Portuguese economic historiography has viewed the country’s macroeconomic performance on the whole with a jaundiced eye. For those concerned with the Early Modern period, the majority view has been that from the end of the Middle Ages to the end of the Ancien Regime this could best be characterized with the words “decline” or decadence”.26 Some bright moments existed, such as during the Golden Age of Brazil, but essentially the experience was consistently negative. For those who sought to account for how Portugal had become a “backward” country in the second half of the twentieth century, the answer was persistent divergence from the core countries since the mid nineteenth century, when the latter began to surge in their performance.27 Interestingly, from an intellectual point of view these two strands never met. Otherwise, they would have had some difficulty in reconciling their respective perspectives, unless 26 27 See text above, section 2. Pereira (1983), Reis (1993), Justino (1988-9), Lains (2003). 33 they could accept that Portugal had had an enviable economic position at the end of the medieval period and declined ever since. This paper has attempted to reconstruct time series for the main indices of macroeconomic performance, from the dawn of the Early Modern period to the mid eighteen hundreds to work out these long term historiographic problems. We have used for this a new data base of prices, rents and wages and followed the procedures which are current in the literature that quantifies the long term economic evolution of nations. The principal finding is that overall Portugal did not fall behind most of Europe between 1500 and 1800. Indeed, it enjoyed some comparatively high moments, such as in the transition from the fifteenth to the sixteenth centuries or in the first half of the eighteenth centuries. It also experienced slow but long periods of growth thanks to the conjunction of external shocks to its agricultural sector. It fell behind the two leaders of Early Modern growth, but more or less always kept up with the remaining nations. This story could have been different if some circumstances had been different. Most important of all in this was demography, which finally eroded, in Malthusian style, the productivity gains achieved. We have not tried here to explain why the Portuguese demographic response to increases in per capita GDP was more intense than in other countries. Another was the exhaustion of opportunities for raising productivity, in part a matter of politics and therefore beyond control, in part a matter of finite natural resources, also perhaps beyond control. The Portuguese economy began to really decline during 1800-1850, just as the whole of Europe seemed to be breaking out of the Malthusian grip. This is when divergence began to occur, and not before. Once it had started, however, it became cumulative. Probably the reason was that the new domestic conditions required for benefitting from the forces of globalization which the Industrial Revolution had unleashed were not 34 sufficiently present in Portugal, and this is the second part of the narrative of “backwardness”. It lies however outside this paper. REFERENCES FIGURES AND TABLES Fig 1 GDP per capita (Maddison and Valerio) and Welfare ratio (1500=100) 200 150 100 50 0 1500 1550 1600 1650 Maddison 1700 Valerio 1750 1800 1825 1850 Welfare ratio Sources: Maddison (2001) and Valério (2010) for GDP per capita; for the welfare ratio, see text 35 Fig 2 Inter-sectoral productivity gaps (p): agriculture vs non-agriculture 1,2 1 0,8 0,6 0,4 0,2 0 France (1840) Germany (1850) Sweden (1880) Spain (1850) Britain (1840) Portugal (1850) Sources: for Spain, Prados de la Escosura (2003); for Portugal, Lains (2003) and Reis (2005); for all others, Crafts (1984) Note: for the calculation of p, see text Table 1 Data coverage of series, 15001850 (percentage of number of years for which reliable data are available) wages wheat maize meat eggs hens wine 16c 17c 18c 1800-50 38 40 65 100 62 90 96 100 na 84 94 100 30 98 99 100 33 100 100 100 Sources: see text 36 35 100 100 100 55 82 100 78 olive oil 39 100 100 100 charcoal linen 52 98 80 100 32 60 83 na land rents 40 100 100 100 Fig 3 Population of Portugal, 1500-1850 (linear interpolation 1500-1570) 4500000 4000000 3500000 3000000 2500000 2000000 1500000 1000000 500000 Sources: see text. Fig 4 Occupational shares of Population 15001850 0,8 0,6 0,4 0,2 0 1500 1550 urban 1600 1650 rural non agric Source: see text 37 1700 agric 1750 1800 total non agric 1842 1824 1806 1788 1770 1752 1734 1716 1698 1680 1662 1644 1626 1608 1590 1572 1554 1536 1518 1500 0 Fig 5a GDP I (1500=100) 400,00 350,00 300,00 250,00 200,00 150,00 100,00 50,00 Fig 5 b GDP II 1500=100 Sources: see text 38 1830 1808 1720 1698 1676 1654 1632 1610 1588 1566 1544 1522 1500 450,0 400,0 350,0 300,0 250,0 200,0 150,0 100,0 50,0 0,0 1836 1773 1786 1815 1752 1764 1794 1731 1742 1710 1689 1668 1647 1626 1605 1584 1563 1542 1521 1500 0,00 Fig 6a GDP I pc (1500=100) 120,00 100,00 80,00 60,00 40,00 20,00 1836 1815 1794 1773 1752 1731 1710 1689 1668 1647 1626 1605 1584 1563 1542 1521 1500 0,00 Sources: see text Fig 6b GDP II pc (1500=100) 180,0 160,0 140,0 120,0 100,0 80,0 60,0 40,0 20,0 Sources: see text 39 1836 1815 1794 1773 1752 1731 1710 1689 1668 1647 1626 1605 1584 1563 1542 1521 1500 0,0 Fig 7 Real wage of skilled labour, 1340-1600 14 12 10 8 6 4 2 1589 1572 1555 1538 1521 1504 1487 1470 1453 1436 1419 1402 1385 1368 1351 1334 1317 1300 0 Sources: see text Note: the deflator of the nominal wage is composed of wine and wheat prices only, the two largest consumption items. Table 2 GDP per capita in Western Europe (UK 1850= 100) Britain Netherlands Germany France Italy Spain Sweden Portugal 1500 1550 1600 1650 1700 1750 1800 1850 39 39 37 34 55 61 75 100 37 37 68 69 54 60 67 79 49 na 34 Na 40 45 42 61 50 Na 50 na 54 55 56 78 68 64 60 62 65 68 60 66 50 54 53 41 48 46 54 64 na 35 na 42 62 47 46 52 64 35 41 45 46 61 49 46 Sources: for Sweden, Schon and Krantz (2012); for Portugal, see text; for the rest, see Alvarez-Nogal and Prados de da Escosura (2013). 40 Note: the figures are ten year averages centered on each benchmark year; we have chosen version I of Portuguese per capita GDP for this table because the stronger oscillations of version II would lead to severe distortions. Figure 8 Land-population ratio (ha per inhabitant) 5,0 4,0 3,0 2,0 1,0 1794 1815 1836 1794 1815 1836 1773 1752 1731 1710 1689 1668 1647 1626 1605 1584 1563 1542 1521 1500 0,0 Sources: Figure 9 Agricultural output per capita 120 100 80 60 40 20 Sources: see text 41 1773 1752 1731 1710 1689 1668 1647 1626 1605 1584 1563 1542 1521 1500 0 Sources: 42 100 80 60 40 20 0 1800 1780 1760 1740 1720 1700 1680 1660 1640 1620 1600 1580 1560 1540 1520 1500 1840 120 1840 Average agric product per worker (1500=100) 1820 Sources: 1820 1800 1780 1760 1740 1720 1700 1680 1660 1640 1620 1600 Fig 11 1580 1560 1540 1520 1500 Figure 10 Average land yield 90 80 70 60 50 40 30 20 10 0 APPENDIX 1 A note on the estimation of Portuguese food surpluses and deficits, 1500-1850 The need to produce estimates of exports and imports of the main foodstuffs in Portugal during these 350 years in the context of estimating its gross agricultural product arises from two circumstances. One of them is the need to verify whether the widely held assumption that food output and consumption in Early Modern economies were roughly equal holds for Portugal. This premise must be clarified so that a correct estimate of food output can be derived from one for consumption. The consensus in the literature has been to accept Allen’s (2000) claim that, in the main, the international trade in foodstuffs occurred on a very small scale at this time and that such an identity may be taken for granted (Pfister, 2008; Alvarez-Nogal and Prados de la Escosura, 2007; Malanima, 2010). The exceptions were the Netherlands and England, where even so the differential never exceeded 10 percent and would consequently have had no more than a modest impact on the estimation of the gross agricultural product. In the specific case of Portugal, a second reason for raising this issue is the strongly held view that the country has always suffered from a pronounced food deficit, as a result of the inability of its arable sector to meet the nutritional requirements – mainly grain - of its population. Already a favourite topic by the early 17th century, it has been taken up in every century thereafter by the many writers who have pondered over the problems of the country’s decline and stressed the chronic need for massive imports of cereals as a sign of economic failure (Marques, 1971). More recently, it has also been taken up by historians who have mostly endorsed and reinforced this belief. Fairly typically, Godinho (1955:147), for instance, has written that “une des constantes les plus 43 indiscutables et plus importantes de l’histoire économique portuguaise c’est le déficit en grains”. The purpose of this Note is first of all to gather the available information concerning the foodstuffs consumed in Portugal and which were the object of significant international transactions. This will enable us to calculate for each one of our benchmark years the global surplus or deficit of the trade in agricultural commodities and compare this with our estimates, at current prices, of agricultural consumption, in order to determine the ratio (r) of domestic agricultural production to consumption. This exercise suffers from several shortcomings. Portuguese foreign trade data are sparse, especially for the 16th century, and those we could obtain do not always coincide with our benchmark years. Until the late 18th century, there are no Portuguese official trade statistics and most of the estimates proposed by historians are obtained in official reports, accounts by travellers and consuls, petitions and remonstrances sent to the king, and also the foreign trade statistics of Portugal’s principal commercial partners. We have focused here exclusively on the three most important items – grain, wine and olive oil. Other commodities such as almonds, figs, wool, etc. have been ignored, given their lesser importance and the tremendous effort required in order to quantify their respective trades. The results have been converted to both volume and value estimates using contemporary prices and the standard measures supplied by the PWR data base. For each benchmark year we have expressed the export or import of our three main items in their respective current values in grams of silver (cols 1, 2 and 3) and summed them up to calculate the overall food deficits or surpluses in silver too (col 4). Since we do not possess direct any quantification of national agricultural consumption at current prices, we use instead an estimate for it in grams of silver obtained using a short cut estimation method proposed by Malanima (2011: 179). This affords us a standard of 44 comparison with which to gauge the importance of the food surplus/deficit relative to the national effort at producing foodstuffs. The implementation of this procedure starts by multiplying the total wage bill (see the main text) in grams of silver by 1.4 to obtain the presumed income of all production factors (including also land and capital). In turn this is multiplied by a coefficient of 0.6, to arrive at a figure for the total expenditure on food of the recipients of national income, also in grams of silver (col 5 of table #). This in turn allows us to assess the magnitude of the food surpluses/deficits relative to food consumption (col 6 of table #) and to determine the value of the ratio r (col 7 of table #) which helps us to adjust correctly food consumption when trying to estimate agricultural output. 1500 and 1550 Agricultural exports from Portugal during the 16th century were very small compared to re-exports from the empire, mainly to northern Europe. In particular, wine exports, later to be a key item in Portuguese foreign trade, appear as insignificant in the available sources. Total agricultural exports for the years 1535-7 amounted to 8.2 million reis in monetary value or, in silver, 0.680 tons. Wine exports in 1535-7 amounted to a negligible total of 18,425 hl, worth something of the order of …grams of silver, the rest being mostly olive oil worth …(Costa et al., 2011). For want of a better solution, we have assigned these data to our 1550 benchmark. To obtain figures for 1500, we have fitted a linear trend to the first of these items, given that olive trees are a slow developing perennial, and a polynomial to the second, given that there is qualitative evidence of accelerating growth in sales abroad of wine in the second half of the 17th century. These are based on the respective volumes in 1550 (Costa et al., 2011) 45 and in 1700 (Hanson, ??) and enable us to justify, tentatively, attributing negligible exports of wine and olive oil to the start of the 16th century. National grain imports during this entire century are shrouded in mystery despite frequent references by foreign and national observers to their substantial scale and the concern of rulers to ensure that adequate supplies, particularly for Lisbon and also for lesser cities, were assured. Rebelo da Silva, cited by VMG (Descobrimentos, 279), claimed that between 1525 and 1562, the Cortes stated that yearly grain M were 500,000 cruzados, i.e. 200 contos. At the then current price of 5.53 rs/l this would have represented a volume of 43,679 moios, or 36.2 million litres, and worth 17.4 tons of silver, which we have adopted for this benchmark. This result is preferred here to VMG’s (Prix: 147) estimate of 100,000 moios of grain circa 1550 for which the evidential support is weak.28 1600 Given the absence of any export figures for olive oil and wine, we have interpolated them also, as described above, from the two benchmarks for which evidence exists. For the former, we have inferred a volume of 3.1 million litres worth 14.2 tons of silver. For the latter, we obtain a volume of 0.22 million litres worth 0,021 tons of silver. For grain and for lack of anything better, we have taken the mean of the imported volumes of grain in 1550 and 1650 – i.e. 24.5 million litres - and multiplied it by the current price of 1.0 grams of silver per litre, making a global value of 24.5 tons of silver for this benchmark. 1650 28 VMG’s first reference to this number, in 1955, was taken from Barros (1941: 115). Barros however was citing a mid seventeenth century document produced by Lisbon’s municipality, which in any case has never been traced. 46 According to trade figures inferred from Rau’s (???) 17th century shipping estimates, total Portuguese grain imports amounted to 13.0 million litres on average during the 1640s, a figure which we adopt for 1650. At a price for wheat in Lisbon of 1.1 grams of silver per litre, the total purchases abroad of this item would have come to about 14.3 tons of silver. An independent source – an estimate by the juiz do povo of Lisbon in 1632 (VMG, Descobrimentos, p. 279) - puts grain imports at 15 tons of silver, the then value of 11.4 million litres. Again as above, we are obliged to infer data for wine and olive oil exports by interpolation from the two benchmarks for which we have any evidence. As a result, we get a volume of 5.72 million litres worth 18.5 tons of silver for the latter and of 0.66 million litres worth 0.91 tons of silver for the former. 1700 Data for wine exports are only available at this date for those sent to Britain, which was anyway the overwhelming consumer of this article. We therefore assume an identity between total wine exports and those to England only. According to Martins (1998), the volume was 8.1 million litres in 1700-4. This was a period of very rapid growth in this trade - in 1690, this rubric had been 1.2 m litres (Hanson, 216 and 220). We valued it by the price of wine landed in England in 1700-4 (Fisher, 1984) after a presumed deduction of 20% for freight and other shipping expenses: This implied a price in Portugal of 1.47 grams of silver per litre. The total Portuguese value of this export in 1700 would have been therefore around 11.91 tons of silver. There is no quantitative information on national grain imports at this time but we have the figures for those from England, which was an important supplier. Two sources are available for the 1700 benchmark. Fisher (1984) is for wheat only, while Francis (1984) represents all types of grain. Both cases correspond to cereals sold in all 47 Portuguese ports. According to Francis (1960), Portugal imported 22.7 million litres of grain in the early 18th century. To make Fisher’s 1700-4 yearly averages comparable, we increase them, by a factor of 1.4, which is the ratio between the volumes respectively for total grain and wheat implicit in Francis’ (1960) import data over the 1700-8 period. This would have implied a yearly average of 16.8 million litres for all cereals landed in Portugal. Hanson’s (…) and Rau’s (19..) estimates for the late 17 th century were respectively 12.4 and 19.3 million litres and we have consequently opted for an intermediate value of 18 million litres. Given the prevalence of wheat in these imports, we value the entire amount at the PWR silver price for this item in Lisbon, which was 0.7 grams per litre, and obtain a total value for grain of tons of silver. Costa et al.(2011: 192) afford us the evidence to estimate exports of olive oil. In volume these would have been 8.2 million litres in 1690 worth 27.2 tons in silver. 1750 Total wine exports by volume are obtained from Martins (1998:..) and come to 23.5 thousands of pipas. We multiply this by the port wine price (Martins, 1990: 243 – average of the prices for 1750, 1756 and 1757), given that at this time port represented around 90 % of all wine exports. The total amounts to 532 contos, or 14.9 tons of silver. In the absence of any data on foreign grain consumption around 1750, we have adopted the mean for the years 1729 and 1776-7, i.e. 37.1 million litres. Valued at a (PWR) wheat price of 0.62 grams of silver per litre, the total grain import to Portugal is estimated at an equivalent of 23.0 tons of silver. A similar solution has had to be employed for olive oil exports, for which no data are available either, even approximately. In this case we have estimated them in terms of volume as the mean of the quantities shipped in 1700 and 1800, and valued this 48 at the PWR price for 1750. The results are 4.737 million litres, valued at 3.3 grams of silver per litre and a total export worth 15.6 tons of silver. 1800 Value and volume of wine exports are., from Martins (1998), respectively …..; .The quantity of grain imports, for the entire country is from Serrão (2005) averaged over the interval 1796-1810, i.e. 890,000 hl. This is valued by a grain price from PWR (using the price of wheat for all grains = 60.7 reis/lt). The value of grain exports comes to 151.6 tons of silver. The figure for the export of olive oil in 1800 is given by the average of the amounts drawn from the Balanças Comerciais, an annual compilation in manuscript form of Portugal’s foreign trade, for three years – 1796, 1797 and 1799. They include sales to foreign countries, the colonies (mainly Brazil) and the islands of Madeira and the Azores. In volume, this comes to 1.274 million litres, which are valued at the PWR 1800 price of 6.5 grams of silver per litre and represent a total of 8.3 tons of silver. In order to estimate the adjustment factor for international trade in food represented by the ratio of domestic consumption to domestic production, we convert all variables relevant to this exercise into grams of silver, which function here like current prices. Food consumption is derived by multiplying the total wage bill in silver by 1.4, to get total income (including from land and capital). The result of this is multiplied in turn by 0.6, to arrive at a figure for total expenditure on food (Malanima, 2011: 179) again in grams of silver. This can then be compared with the food surplus/deficit in order to obtain the value of r, which is either larger or smaller than 1 depending on whether output exceeds or not consumption. 1850 Percentages of food, grain and wine in terms of exports and imports from Lains (1995); values of total imports and exports at current values, from Valerio (2001); 49 imports of grains and wines from Lains (1995) combined with Valerio (2001). Total value of agricultural output in 1850 at current prices from Valerio (2001) combined with Lains (2003) for sector shares (Agric=48%= 120,000 contos). Olive oil exports were 2% of total exports, i.e. 160 contos. In volume this was 0.73 million litres, with a total value in silver of 4.3 tons. NB: Braudel, in o Mediterrâneo, vol.1, p. 629 estimates international wheat trade as 1% of needs. Figure 1A Agricultural output: corrected and uncorrected by foreign trade 400 350 300 250 Corrected 200 uncorrected 150 100 50 0 1500 1550 1600 1650 1700 1750 1800 1850 Appendix 2 Notes on the calculation of the TFPs of agriculture and manufacturing Total factor productivity in manufacturing The lack of suitable data precludes the possibility of obtaining an index which covers a wide range of products. It also obliges us to use the “dual” method for estimating 50 manufacturing TFP, instead of the “primal”, a choice discussed by Antràs and Voth (2003). In a similar situation, Allen (2003) estimated the TFP of one important subsector alone – woollens – and considered this as representative of manufacturing as a whole. The index was obtained by calculating the geometric mean of the price of the raw material and the skilled wage rate and then dividing this by a cloth price series. Likewise, we have adopted a cheap (and therefore not imported) variety of linen cloth, one of the principal textiles manufactured during the period in Portugal. And for the input side we have calculated the geometric mean of the skilled wage and of the price of flax.29 This is divided by the market value of linen cloth to find to get the TFP index. Total factor productivity in agriculture Analogous problems are met with agricultural TFP but the solution is slightly different. We continue to make use of the “dual” approach. The price of aggregate of output is the same as was used to estimate PAPM (see above d)). Input prices are given by the weighted geometric mean of wages and rents, using the weights, respectively 65 and 35%, employed by Rosés et al. (2007), for Early Modern Spain.30 29 For a different methodology, see Rosés (2007). We avoided Allen’s (2003) approach to the calculation of agricultural TFP because this would have required annual data for agricultural population and output, which we are unable to estimate with any degree of confidence. 30 51