2. AN OVERVIEW OF BRAZIL’S HISTORICAL FOREIGN TRADE SERIES

The official series of Brazil’s exports and imports begins in 1821Footnote ¹ . It remains unclear what the sources were for the period from independence in 1822 to the 1840s, although the government’s reliance on — and thus constant preoccupation with the calculation of — the fiscal revenues derived from the taxation of commerce resulted in the reorganisation of the customs house in the 1830s. The increased dynamism of foreign trade after independence led to organisational changes and investment in maritime infrastructure (Brazil. Ministério da Fazenda 1835, pp. 57-64). Over time, these changes seemingly generated the demand for the organisation of a statistical apparatus. In 1845, the Imperial government created the first Imperial statistical commission with the explicit intention of organising and publishing the foreign trade statistics for the fiscal years spanning the period 1830-1831 to 1844-1845. The commission soon admitted that the period prior to 1840 was beyond its technical reach (or, as the 1846 report of the commission drily put it, «… the elements do not yet exist to form a complete collection»), and committed itself to elaborating annual series of exports and imports by product, origin and destination (Brazil. Ministério da Fazenda 1846, pp. 34-36; Brazil. Ministério da Fazenda 1848, pp. 27-28). To assemble the statistics on a national level, the commission relied upon the customs houses to provide the information on quantities, values, origins and destinations of imports, exports and internal trade. Officials complained repeatedly about the quality of the information supplied by customs houses; products of diverse qualities were often included in a single category, old official price lists were used, customs officials deliberately misreported quantities and values, and sometimes the paperwork containing this information (the mappas parciaes) did not arrive at all (Brazil. Ministério da Fazenda 1894; Brazil 1853). To obviate the latter problem, the commission calculated missing import and export values using the available information on fiscal revenue. The fruit of the commission’s work is compiled in the Collecção de Mappas Estatísticos do Commercio e Navegação do Império do Brasil, of which the fiscal years 1841-42, 1842-43, 1846-47, 1848-49 and 1849-50 have been digitisedFootnote ² .

During the period encompassing the fiscal years 1853-1854 to 1867-1868 the Finance Ministry reported the official value of bilateral trade in its Ministerial Reports, the data being provided by the Diretoria Geral das Rendas Publicas. The detailed work of the first commission, however, was discontinued until 1870, when the government created a second statistical commission headed by Sebastião Ferreira SoaresFootnote ³ . Much like its predecessor, the role of the second commission was to organise the country’s statistics on maritime commerce that had fallen into disrepair since the 1850s (Brazil. Ministério da Fazenda 1885, p. 39). Soares and team’s work yielded the publication of a series of volumes of Estatística do Commercio Marítimo do Brazil covering the early 1870sFootnote ⁴ . As Gustavo Henrique Barroso Franco observed, the period from the death of Soares in 1887 to the reconfiguration of the statistical apparatus around the turn of the century by British civil engineer J. P. Wileman is terra incognita as far as the origin of the available national totals of exports and imports is concerned (Franco Reference Franco1990, p. 559). However, government officials did show concern regarding the disorder of the country’s statistics. In 1895 Honorio Alonso Baptista Franco, member of the commission for the revision of the Brazilian tariff, was sent to inspect the workings of the Rio de Janeiro customs house. In an account published in the annual report of the Finance Ministry in 1896, Baptista Franco observed that «It is physically and morally impossible that work of such a scope … referring to not only the biggest customs house in Brazil, but also in all of South America … can be elaborated by only six employees, two of which were sometimes distracted by other tasks» (Brazil. Ministério da Fazenda 1896, anexo, p. 23). Customs house officials continually justified delays in the elaboration of statistics by a lack of resources and specialised personnel (Brazil. Ministério da Fazenda 1896, p. 224). In response to these calls and to cover required expenses, the government created a «statistics tax» (imposto de estatística) in 1897 that provided enough revenue to cover the salary and rent expenses of the newly created Serviço Especial de Estatística (Brazil. Ministério da Fazenda 1901, p. 47). Wileman was appointed as director of the Serviço in 1900, and from 1901 onwards, Brazil’s trade statistics were reported annually in Commercio Exterior do Brasil, which, like Wileman’s previous work, explicitly concerned itself with the mirror comparison of Brazil’s trade records with those of its principal trading partners in an effort to improve the accuracy of the data.

The results of one such mirror comparison for the period 1902-1914, incorporating the export statistics of Great Britain, the United States and Germany, and the corresponding official Brazilian import statistics, provided grounds for celebrating the success of the modernisation of Brazil’s statistical apparatus. The report concluded that «…the differences between the totals … are not important … and are sometimes even trifling» (Brazil 1915, p. LXX)Footnote ⁵ . This accords with the recent evaluation of the export series, which tended towards accuracy after 1901 (Absell and Tena-Junguito Reference Absell and Tena-Junguito2016). Nevertheless, it remains to be seen whether such conclusions apply to the 19^th century statistics.

There are two possibilities regarding the bias of the official statistics: one concerning quantities, the other prices. The accuracy of the first may be affected by several causes, the most common being smuggling, misreporting and incorrect classification of merchandise. If these were truly problems, then one would assume that the series would be biased downwards, as the official statistics under-report the true quantity imported. Officials were only too aware of these problems, however, and the use of specific rates (taxas fixas) to tax imports was justified on the grounds that they reduced the incidence of fraudulent activity (Fontoura Reference Fontoura1889, p. 47). The second potential source of bias concerns the official prices. These were defined principally for fiscal reasons, given that taxes on trade (mostly imports) were Brazil’s principal source of fiscal revenue during most of the 19^th centuryFootnote ⁶ . In the case of imports, tariff commissions defined the classification scheme and official price of each product based on an average of market prices. Specific rates were then calculated by applying an ad valorem tariff to the corresponding official pricesFootnote ⁷ . Similarly, in the case of exports, the official values were fixed by a weekly price list (the pauta), meant to reflect the average market prices of each commodity. Customs officials used this information, together with the quantities registered of incoming and outgoing merchandise, to calculate the weight of the tax and the corresponding values, which were included in the mappas sent to the statistical agency to calculate the aggregate statistics. As we shall see below in the case of imports, these official prices were liable to diverge from market prices, whether due to strategic overvaluation for fiscal reasons, or delays in updating the official price listsFootnote ⁸ . From 1901 onwards, import values were listed as declared values in cost, insurance and freight (c.i.f.) form, while export values relied principally on market prices and included export duties but not freight, insurance or landing costs (Brazil 1915, p. XXI).

3. THE IMPORT SERIES

Here we seek to examine the accuracy of Brazil’s official import series. The standard procedure for assessing the accuracy of the import series is to compare the official prices and quantities of Brazil’s import statistics with those of its principal trading partners. Unfortunately, a complete series of official import prices and quantities do not exist, so we are unable to implement such a procedure. Instead, we estimate a new import series based on the existing partner statistics and compare this to the official series. Thus, the methodology adopted here is to take the sum of the trading partner statistics and to add a «residual» that corresponds with the percentage of imports not covered by our trading partner sample. The coverage is calculated as the percentage of each trading partner in total import value according to official Brazilian sources. As previously mentioned, we possess only a broken series of bilateral observations from Brazilian sources for the 19^th century, including the information compiled in the Collecção de Mappas Estatísticos, the bilateral data published in the Relatorios of the Finance Ministry, a set of years from the Estatística do Commercio Marítimo, and the period from 1901 onwards given in Commercio Exterior do Brasil. We fill in the gaps by taking several benchmark years for imports into Rio de Janeiro from the Mappas estatísticos do commercio e navigação do Porto do Rio de Janeiro (Brazil. Alfândega do Rio de Janeiro, Various years). Furthermore, we add a benchmark for Rio de Jeneiro in 1836 from Sturz (Reference Sturz1837, p. 100). We assume that the geographical distribution of imports into Rio de Janeiro was representative for these benchmark years. For those periods in which the geographical distribution of imports is not available, we construct a time series by way of interpolation. We additionally assume that interpolation is an acceptable practice in this context given that there were no sudden changes in the geographical composition of Brazil’s imports. To this effect, we use data from eight of Brazil’s principal trading partners (Great Britain, the United States, France, Spain, Portugal, Chile, Belgium and Hamburg/Germany), adjusted for insurance and freight costs. Table 1 displays the descriptive statistics of this sample. Unfortunately, for the first 5 years of the period (1821-1826) we can only include the United Kingdom and the United States. However, these two countries alone constituted around 54 per cent of Brazil’s imports by value. Over the following few decades the sample expands to include eight countries and to represent over 90 per cent of Brazil’s imports by value.

We construct the new import series using the formula:

where Brazil’s imports (M) in year t are calculated by taking the product of the sum of the exports (X) of each trading partner i in the sample and the inverse of the sum of the percentage of trading partner ci in total import value.

When attempting to estimate the import series, an important assumption is the size of the freight factor. Like the export series, existing outgoing freight data (from Europe and the United States) either does not cover the whole period under study, is for destinations other than Brazil, or does not cover the entire composition of Brazil’s imports. For these reasons, again, like the export series, we calculate a lower and upper bound estimate of the freight factor used to adjust the partner statistics from c.i.f. to free on board (f.o.b.) values. The lower bound is the weighted average of the freight factor of 11 of Brazil’s principal import commoditiesFootnote ⁹ . A number of these freights are for different routes and are adjusted for the corresponding distance differential. Furthermore, we chain several of these series to existing freight indices to overcome the problem of temporal coverage. The lower-bound factor ranges from 6 to 15 per cent, averaging 12 per cent over the period. Although freights were generally higher than 12 per cent of the import price, especially for the period before the dominance of steam, this estimate is biased downwards by the preponderance of cotton manufactures in Brazil’s import composition and the corresponding low freight factor of this product. The upper bound estimate attempts to balance this downwards bias by taking an unweighted average of the same sample of commodities. The upper bound factor ranges from 18 to 30 per cent, averaging 23 per cent. Estimates of the freight factor given in Commercio exterior do Brasil (Brazil 1915, pp. 9-13) for the period 1901-1914 lie in between our lower- and upper-bound estimates. We also assume a 2 per cent insurance cost, in line with existing data for the period (Schöller Reference Schöller1951; Llorca-Jaña Reference Llorca-Jaña2011).

A further consideration is the inclusion of transit trade in the trading partner records. Given that Brazil, and especially Rio de Janeiro, was an important port of call for the transit trade, the geographical assignment of the partner series to Brazil may include commodities ultimately destined for the Rio de la PlataFootnote ¹⁰ . If this is the case, the geographical distribution of Brazil’s imports should appear as under-valued when compared to the trading partner statistics. On the other hand, the Argentine and Uruguayan statistics, where available, should be over-valued when compared to the corresponding partner records. We test this hypothesis by running a mirror comparison of the sums of the value of exports of the Rio de la Plata’s principal trading partners with the official values listed in Brazilian, Argentine and Uruguayan sources. We include the trading partners mentioned previously, excepting Portugal for which we lack sufficient bilateral data in the case of Uruguay. We adjust these partner records by the lower-bound freight factor and insurance estimate as described above. As previously mentioned, the construction of a continuous time series is restricted by the paucity of sources. Although this prevents us from undertaking a precise evaluation of the accuracy of the geographical distribution of Brazil’s bilateral series, it allows us to evaluate the degree and tendency of the bias in the statistics. An additional restriction is the temporal coverage of Argentine and Uruguayan official sources: these are limited until the period from 1870 onwards. We have extended the Argentine series by using import data for Buenos Aires for several years. This is justified by the fact that Buenos Aires was the principle Argentine receiving port for trade during these yearsFootnote ¹¹ .

Figure 1 displays the results of our test of geographical distribution. Although it is difficult to draw conclusions due to the paucity of data, it is not immediately obvious that trading partner statistics were overwhelmingly biased by the mis-specification of exports. To begin with, the period when Brazil’s geographical distribution tends towards under-valuation (seemingly from 1862 until the reorganisation of the statistical apparatus around the turn of the century) corresponds with the under-valuation of the Argentine and Uruguayan records. The only period that may lend weight to the argument is the 1890s, when both the Argentine and Uruguayan records were around 20 per cent over-valued with respect to partner records. Unfortunately, we only possess official Brazilian bilateral data from 1899 onwards, and thus are unable to confirm or reject this hypothesis for this period. For the purposes of our re-estimate, however, we assume that the trading partner statistics are not biased beyond remedy by the incorrect specification of the geographical distribution of exports.

FIGURE 1 ACCURACY TEST OF GEOGRAPHICAL DISTRIBUTION OF IMPORTS, BRAZIL (BRA), ARGENTINA (ARG), URUGUAY (URU), 1841-1913 (1=ACCURATE) Sources: Same as Table 1.

An additional question concerns the inclusion of foreign re-exports. Re-exports were an important component of the exports of Brazil’s principal trading partners. In the case of the United States, re-exports averaged around 17 per cent of the country’s total exports to Brazil during the period 1821-1850, declining to an average of 2 per cent during the rest of the century. Great Britain’s exports of colonial and foreign merchandise accounted for a lesser share, averaging 4 per cent of exports from 1854 to 1913. The value of the French commerce général, which included both French and foreign exports, was 38 per cent higher on average over the century than France’s exports of domestic merchandise, or commerce spécial Footnote ¹² . A key question is whether the official series reflected these additional trade flows. The period before the modernisation of Brazil’s statistical apparatus around the turn of the century is ambiguous, given the general paucity of information regarding the elaboration of the import series. The introductory notes to the Collecção de Mappas Estatísticos defined general imports as including «… all categories and merchandise of foreign production and manufacture imported directly from the country of origin» («dos propios paizes») (Brazil, Various years). A similar definition appeared in Soares’ glossary included in the Estatística do Commercio Marítimo (Brazil. Sebastião Ferreira Soares, Various years). After 1901, the Brazilian import statistics explicitly included «… both direct imports, and those that arrive in transit from other countries» (Brazil 1904). Here we seek to confirm whether re-exports were included in the official imports series. If so, then the official series should be over-valued when compared to our estimate without re-exports. Likewise, the addition of re-exports to the trading partner statistics should bring our estimate closer to parity with the official series. We include the re-exports of the above-mentioned three countries, given their primacy as trading partners and the availability of data on re-exports for most of the period under question.

Table 2 shows the ratio of the official import series, at current prices, to our re-estimate, considering the above-mentioned issues. We present the ratio adjusted for (a) the weighted, or lower bound, estimate of the freight factor, (b) the unweighted, or upper bound, estimate of the freight factor and (c) the latter as well as the inclusion of re-exports from the United States, Great Britain and France. For the whole period (1821-1913), the official series is surprisingly accurate. The results, however, indicate a certain degree of under-valuation during most of the 19^th century. This is especially evident when re-exports are included. As expected, the new estimate corresponds closely to the official series after 1901, reflecting the improved accuracy of the Brazilian statistics.

Figure 2 displays the official series alongside the sum of our trading partner series and the re-estimate including the lower-bound freight factor and re-exports, in current US dollars. Since the coverage of the sum of our sample of trading partners does not reach 100 per cent, this sum should always lie below the official series. With few exceptions, however, we observe that this is not the case. The degree of under-valuation of the official series is particularly notable during the period immediately following independence (1822-1840), and during the period spanning the Paraguayan War (1864-1870) to the fall of the Empire and founding of the First Republic (1889).

FIGURE 2 SUM OF TRADING PARTNER SAMPLE, OFFICIAL IMPORT SERIES AND RE-ESTIMATE OF TOTAL IMPORTS, CURRENT PRICES, BRAZIL, 1821-1913 Sources: Sum sample and new estimate: see Table 1; official series: IBGE (1990).

Here, we briefly explore two possible explanations for the under-valuation of the official series, one concerning quantities, the other official prices. The former may be under-counted due to the presence of contraband. While contemporary reports suggest that certain regions, such as Rio Grande do Sul, were characterised by an active contraband trade, it is difficult to calculate precisely what percentage of total trade this representedFootnote ¹³ . In the case of Rio Grande de Sul, the nature of the local economy determined that smuggling focussed predominantly on cattle rustling (Flores da Cunha Thompson Flores Reference Flores Da Cunha Thompson Flores2012; Pereira Reference Pereira2016, p. 9; Bonino-Gayoso et al. Reference Bonino-Gayoso, Tena-Junguito and Willebald2015, p. 18). While important on a local level, it is likely that the impact on aggregate statistics would be small. Moreover, the under-valuation of the import series suggests a more systematic explanation that lies in the procedure used to elaborate the statistics, rather than quantities missing due to the contraband trade. The use of partner statistics effectively obviates the problem of contraband by including what might be missing trade on the Brazilian side.

The second explanation that official prices were the principal determinant of the observed under-valuation is evident in contemporary reports on the elaboration of the country’s tariff schedule. In fact, it seems that official prices were only changed by infrequent tariff commissions. A tariff commission report in 1871 observed that «The tariff that currently regulates the collection of customs house revenues, with a few small exceptions, is that of 1860 … Not having elevated the official prices, and raising them alongside events in the market, resulted in … a percentage of fiscal revenue less than the rightful amount…» Citing the report, the Viscount of Rio Branco recommended that the situation be rectified by way of the updating of official prices (Fontoura Reference Fontouran/d, pp. 60-61). The issue was not only one of official prices, but also of the corresponding classification scheme, as Francisco Belisario Soares de Souza argued in his report to the Finance Ministry in 1887: «Effectively, since the last tariff was enacted [in 1881] … the industrial method of production of certain products has undergone important transformations … rest assured that it is not small the number of official prices … that presently diverge from the real average market price of the products to which they correspond…» (Fontoura Reference Fontouran/d, p. 84). Given that (until 1901) the values that appear in the available sources for import by product are aggregates of such official prices, we are able to test this hypothesis by comparing the trend of official import prices against that of corresponding trading partner export pricesFootnote ¹⁴ . If official import prices were fixed for long periods, then this should appear in the evolution of these prices over time when compared to trading partner export prices.

Figure 3 compares the official prices of six commodities with the corresponding prices of the principal supplier of that commodity. We include cotton manufactures (plain and printed or dyed), dried and salted codfish, and coals from Great Britain, wheat flour from the United States, and beef jerky from Uruguay. In the cases of wheat flour, coals, codfish and beef jerky it is evident that, until the turn of the century, official prices remained fixed and did not reflect the trend of international prices. What’s more, perhaps apart from coals, the official prices of these commodities remained under-valued for much of the century. From 1901 onwards, however, the prices of these commodities (reflecting the declared value instead of the official value) changed according to fluctuations in international prices, and the strong degree of under-valuation that characterised the previous century disappeared. In the case of the official prices of cotton manufactures, while these also showed fixed properties, they seem to exhibit more frequent fluctuations. However, this could simply reflect changes in the composition of this import category (which included multiple types of cotton textiles) rather than alterations in the official price. In any case, it is clear that the trend and level of the official prices used to calculate the official statistics aid us in understanding the under-valuation of the official import series.

FIGURE 3 OFFICIAL IMPORT PRICES AND TRADING PARTNER EXPORT PRICES OF SIX IMPORT COMMODITIES, 1846-1913 Sources: BRA: Brazil (Various years), Brazil (Sebastião Ferreira Soares, Various years), Brazil (1939). US: United States. Treasury Office (Various years). UK: United Kingdom (Various years), United Kingdom. Board of Trade (Various Years). RIO: Jornal de Commercio (Various years).

4. THE EXPORT SERIES

As mentioned above, the historiography of Brazil’s export economy during the 19^th century has been primarily based on the official statistics, calculated using a list of official prices (the pauta) that was based on the average of market prices. In a previous paper, we showed that the official prices, when compared to corresponding international prices, were generally under-valued. This under-valuation was found to be greater during the decades following independence. We then went on to correct the official series using international prices and official quantities, adjusting the international price by subtracting trade costs. The results revealed a dynamic post-independence period. Figure 4 displays the corrected and official series in current United States dollars.

FIGURE 4 TOTAL EXPORTS, CORRECTED AND OFFICIAL SERIES, CURRENT PRICES, BRAZIL, 1821-1913 Sources: Absell and Tena-Junguito (Reference Absell and Tena-Junguito2016) and IBGE (1990).

When undertaking these corrections, we made several assumptions regarding the weight of trade costs that are sensitive to the data being used. Specifically, we rely on estimates of freight rates and export taxes that are perhaps not totally representativeFootnote ¹⁵ . Here we confront each of these issues in turn.

For the period before 1870, there is a general scarcity of freight rate data for Brazil. Juan Oribe Stemmer’s (Reference Stemmer1989) seminal study of Latin American freights during the 19^th century included Brazilian coffee, although only provided a graphical series for the North American route and from 1845 onwards. Other authors provided other series, but with varying degrees of temporal, commodity and geographic coverage (see, e.g. Angier Reference Angier1920; Harley Reference Harley1988; Klovland Reference Klovland2009). To cover the first two decades of the post-independence period, we used Paul Schöller’s (Reference Schöller1951) index for freights from Antwerp to Rio de Janeiro, referenced by freight rate data for Brazil from various sourcesFootnote ¹⁶ . However, the use of Schöller’s index may exaggerate the average level of rates when other routes are included. Antwerp, although an important destination for Brazil’s products, was by no means the only destination. Furthermore, Schöller’s was an outgoing index, and thus may be unrepresentative of the commodity composition of homeward freights for Brazil’s commodities. As the freight rates were deducted from the international prices used to reconstruct the new exports series, the use of our previous freight series might have served to distort the correction of the official export series, and thus the calculation of the growth rates that our argument hinged upon.

To test the robustness of our original correction, we have constructed a new database of freight rates from Brazilian primary sourcesFootnote ¹⁷ . The new database consists of monthly observations of freights from Brazilian to foreign ports for Brazil’s principal export commodities. The data from the period before 1840 is less frequent although it is sufficient to construct an average that reflects the level and trend of prices. For coffee, we include quotations from Rio de Janeiro to Liverpool, London, Hamburg, Le Havre and Marseilles; the rates for sugar are from Rio de Janeiro, Pernambuco and Bahia to Liverpool, the British Channel, Hamburg and Le Havre; cotton quotations are from Rio de Janeiro, Pernambuco and Maranhão to Liverpool, Le Havre and Lisbon; and quotations for hides are from Rio de Janeiro and Pernambuco to Liverpool.

Figure 5 displays the arithmetic average of the available routes for each year for the three commodities, alongside the original series used to convert the international prices from c.i.f. to f.o.b. values. In the cases of coffee and hides, it is evident that the use of Schöller’s series exaggerates the average level of freights, especially during the first decade of independence, although the series converge with the original estimates around mid-century. The original sugar series, however, under-values the true level of average freight rates. The difference in the tendencies and levels of the two estimations for the cotton series is trifling.

FIGURE 5 ORIGINAL AND NEW FREIGHT RATE SERIES FOR SELECTED EXPORT COMMODITIES, BRAZIL, 1821-1850 Sources: Original: Absell and Tena-Junguito (Reference Absell and Tena-Junguito2016); New average: see footnote 17.

A second point is the weight of export taxes. The original estimate used the export tax average (export tax fiscal revenue/total export value in current prices) as a proxy for taxation rates. This approach, however, disregards the provincial rates that were also charged on exports, most of which were higher than the Imperial government rates. The current price list given in the Correio Mercantil of 1827, for example, listed an export tax from the port of Rio de Janeiro for coffee of 8 or 9 per cent (the dízimo), depending on origin («Serra acima» or «Serra abaixo»), plus 80 réis per arroba (equal to 2 to 3 per cent of the average market price) if the average price dropped below 4,000 réis/arroba, 2 per cent otherwise. Thus, the total weight could sum to 12 per cent of the market price, much higher than the 3.6 per cent implied by the fiscal data for that year. The same can be observed for sugar and cotton exports from Pernambuco in 1840, which were 10 and 12 per cent, respectively, higher than the 7.1 per cent given by the export tax average (Diario de Pernambuco, 15 January 1840). The ideal solution to this problem would be an average for the provincial export taxes for each commodity weighted by principal exporting regions. Data scarcity prohibits us, for the moment, from elaborating such a series. We thus add an additional 15 per cent to the export tax average, resulting in an average of 20 per cent per annum for the period from 1821 to 1850. We also consider the possibility that export taxes were included in the final prices. In the commodity chain from plantation to port, producers were subjected to export taxation before the product was sold to the export houses. This would indicate that the market prices that the official prices were based upon already included the export tax. Although the evidence indicates otherwise, it is also useful to provide an estimate without taxation to show their impact on growth.

Using these revised data and assumptions, we re-estimate the growth rates to ascertain whether the overall results and subsequent historical interpretation of Brazil’s export growth are sensitive to these assumptions. Here, we present three growth estimates: estimate A is the original as presented in Absell and Tena-Junguito (Reference Absell and Tena-Junguito2016), estimate B includes the new freight series and higher export taxes, and estimate C includes the new freight series but does not include export taxes. As can be seen in Table 3, changing the assumptions for the initial period does not radically change the story. In fact, the revised series with higher taxes and the new freight series is almost identical to the original estimate, since freights for coffee have been reduced by a larger factor than taxes have been increased. In the year 1821, for example, our export tax rate has been increased by 15 per cent, while the freight for coffee has been reduced from 8.26 to 3.88 pounds per ton, a reduction of 53 per cent. Estimate C demonstrates that growth rates were not particularly sensitive to changes in the size of the trade cost factor. Figure 6 displays the levels of the three estimates in current US dollars. In this context, estimates B and C can be interpreted as lower and upper bound estimates, respectively. While the different assumptions do not change the story in terms of constant prices, there is considerable divergence in terms of current prices during the period 1860-1890. In fact, the lower-bound estimate closely corresponds with the official series during this period. Thus, the altering of these assumptions might seriously affect the interpretation of indicators requiring the use of current values during this period (such as the trade balance). The results for the post-independence period to 1860 and the first decade of the First Republic, however, remain considerably different from the official series.

FIGURE 6 TOTAL EXPORTS, UNDER ALTERNATIVE TRADE COST ASSUMPTIONS, CURRENT PRICES, BRAZIL, 1821-1913 Sources. Same as Table 3.

Of course, any exercise in the reconstruction of historical statistics must make assumptions based on the availability and reliability of the empirical evidence at hand. Accounting of this sort should not be interpreted as a precise calculation, but rather as being merely indicative of the actual trend. It is evident, however, that the post-independence period was the most dynamic in terms of export growth in 19^th century Brazil. This dynamism extended until the mid-century but disappeared alongside the consolidation of the coffee sector. This result is robust to the assumptions made regarding the adjustment of international prices used to correct the under-valued official series. It remains the work of scholars to explore the significance of these results.

5. PRICE INDICES

Here, we review the work performed on price indices for Brazil’s export commodities and introduce a new import price index. Figure 7 displays three estimates of Brazil’s export price index. The first, presented in Absell and Tena-Junguito (Reference Absell and Tena-Junguito2016) is a Fisher index spanning the period 1821-1913 that uses the corrected prices of a sample of six of Brazil’s principal export commodities (cacao, coffee, cotton, hides, rubber and sugar) and moving weights, as outlined in the previous section. The second, constructed by Reinaldo Gonçalves (Reference Gonçalves1981), is also a Fisher index for the period 1850-1913 that uses the official unit values of eight commodities (cacao, coffee, cotton, herva mate, hides, rubber, sugar and tobacco) and moving weights. The third, an index commonly used in the literature on export growth presented by Christopher Blattman et al. (Reference Blattman, Hwang and Williamson2004), is a chained Laspeyres index for the period 1860-1913 that uses the British c.i.f. unit values and a sample of five commodities (cocoa, coffee, cotton, rubber and sugar) and fixed weights, changing every 20 years. We present each index to compare the aggregate price trends generated using differing methodologies.

FIGURE 7 EXPORT PRICE INDICES (1913=100), BRAZIL, 1821-1913 Sources: Absell and Tena-Junguito (Reference Absell and Tena-Junguito2016), Blattman et al. (Reference Blattman, Hwang and Williamson2004), Gonçalves (Reference Gonçalves1981).

Despite being constituted by a reduced commodity composition, the Absell and Tena-Junguito (Reference Absell and Tena-Junguito2016) index is unique in the sense that it covers the entire period. Furthermore, the omission of tobacco and herva mate does not change the overall tendency of the price index, due to the overbearing weight of coffee and, later, rubber in the commodity composition. Overall, Brazil’s export price index declined from a century-long high generated by the Napoleonic Wars to follow the cyclical price trend of coffee until the crisis of oversupply in the 1890s.

Regarding the import prices, we elaborate a new index using the international price data available for Brazil. This is especially constructive for the period before 1850, for which much of the previous work has relied on the general British export price index constructed by Albert Imlah (Reference Imlah1958, pp. 94-98) (see, e.g. Leff Reference Leff1982, Ch. 5; Goldsmith Reference Goldsmith1986, p. 31). The most widely used index for the period from 1850 to 1913 is that elaborated again by Reinaldo Gonçalves (Reference Gonçalves1981), who constructed a Fisher index using British exports to BrazilFootnote ¹⁸ . However, the commodity composition of Gonçalves’ index was not specified. This is important when the issue of quality is considered. The official British sources of bilateral trade include general categories such as machinery and cotton manufactures that incorporate a wide range of heterogeneous commodities. The use of such aggregate categories thus forsakes homogeneity in the interests of commodity coverage. Changes in the composition of these categories might distort the index by exaggerating or understating the weights of such categories in the overall index. Furthermore, as previously stated, Great Britain was not Brazil’s only trading partner during the 19^th century. Important commodities such as wheat flour and beef jerky were sourced almost exclusively from other markets (the United States and Uruguay, respectively).

We thus construct an index for the period 1827-1913 using three principal sources: the British and American bilateral data available from official sources, and current wholesale prices of imported goods listed in contemporary Brazilian periodicals. The index begins in 1827 as the declared values and quantities of British exports to Brazil are unavailable for most commodities before this year. We forsake commodity coverage for homogeneity to capture the long-run price movements of a select sample of homogeneous products. This is, however, only a problem after the turn of the century when the commodity composition of Brazil’s imports diversified considerably. Apart from the current wholesale prices, we then adjust for freight and insurance costs. Our final index includes a minimum of ten and a maximum of sixteen products, accounting for between 50 (in 1841) and 27 (in 1913) per cent of Brazil’s import valueFootnote ¹⁹ . As weights, we use a series of benchmarks of import values by product from the above-mentioned official sources. Table 4 displays the weighting system and the percentage of the sample in total import value. As we do not have a time series of weights, we construct a series of chained Laspeyres indices, using each of the six benchmarks as reference years. We then take the geometric mean of these indices.

Figure 8 displays the new index, alongside the Gonçalves index of British imports, and Imlah’s British export price index. The new series shows that import prices, weighted heavily by cotton manufactures in the early years, followed the general decline of British textile prices during the first half of the century, moderated slightly by the prices of foodstuffs. The second half of the century was dictated by fluctuations in the prices of raw materials (the most important of which being British coals) and foodstuffs. Inspection of the new index raises a number of interesting points. First, by ignoring the weight of foodstuffs (particularly American wheat flour), the reliance on the Imlah index prior to 1860 exaggerates the price level. Second, the price shocks of the American Civil War in the 1860’s and the rise of the price of British coals in the 1870s are accentuated in the new series due to the inclusion of American foodstuffs and the relatively greater weight of coals in the weighting system. Finally, the new series converges with the Gonçalves series towards the end of the period, indicating that the lower coverage of the later benchmarks does not affect the representativeness of the index to any great degree.

FIGURE 8 IMPORT PRICE INDICES (1913=100), BRAZIL, 1827-1913 Sources: Great Britain: 1827-1852: United Kingdom (Various years). 1853-1913: United Kingdom. Board of Trade (Various Years). United States: 1827-1913: United States. Treasury Office (Various years). Rio de Janeiro current market prices: Anglo-Brazilian Times, Correio Mercantil, Diario de Pernambuco, Diario do Rio de Janeiro, Diario Mercantil, Semanario Mercantil, Jornal do Commercio, O Cruzeiro, O Despertador, Retrospecto Commercial of the Jornal do Commercio, Rio Commercial Journal, and Rio Mercantile Journal.