2. THE ACCURACY OF OFFICIAL BOLIVIAN FOREIGN TRADE STATISTICS

On the eve of the 20^th century the composition of Bolivian exports experienced huge changes. On the one hand, the relative importance of silver, the oldest and most traditional export good, was declining as a consequence of a sharp drop in international prices. This decrease took place in line with an upsurge in tin exploitation, a process that took place in the same regions where silver had been previously exploited since colonial times. Meanwhile, a new commodity, rubber, was emerging in a non-traditional export region in the north-east of the country. Thus, according to official statistics, silver represented 40 per cent of the total value exported in 1902, tin 30 per cent and rubber 20 per cent. The rest of exports consisted of other minerals. Ten years later, these shares had changed to 5 per cent, 67 per cent and 17 per cent, respectively. During the First World War, the rubber boom ended abruptly and, at the end of the war, the relative importance of silver, tin and rubber in the total value exported was around 4 per cent, 71 per cent and 6 per cent, respectively. Thereafter, tin exports represented between 65 per cent and 75 per cent of Bolivian total exports; the rest consisted of other minerals and it was not until the World War II that rubber exports recovered some importance (Peres-Cajías and Carreras-Marín Reference Carreras-Marín and Badia-Miro2017).

Besides this high dependence on one specific product, Bolivian exports were also concentrated in few markets. During the first half of the 20^th century, exports to the United Kingdom and the United States constituted around 90 per cent of the total value exported according to official Bolivian foreign trade statistics (Table 1). Regarding imports, whereas intraregional trade occasionally played an important role for Bolivia (Carreras-Marín et al. 2013), it stands out that the United Kingdom, the United States and Germany supplied between 42 per cent and 60 per cent of Bolivian imports during the first half of the 20^th century.

The identification of exports’ destinations and imports’ origins allows for a mirror accuracy analysis of official statistics. As previously noted, this analysis can be done by focussing on each bilateral trade or by looking at the whole picture via an aggregated index. The first exercise has been done using the ratio between Bolivian exports to country j, according to Bolivian sources (free on board (f.o.b.) prices), and imports from Bolivia, recorded in the sources from country j (cost, insurance and freight (c.i.f.) prices). Taking into account previous work on other Latin American countries (Tena-Junguito and Willebald Reference Federico and Tena2013; Bonino-Gayoso et al. 2015) and reports from the Bolivian government Charles Mc Queen (1925), differences between f.o.b and c.i.f. prices were set around 20 per cent. Therefore, a good fit of Bolivian foreign trade statistics takes place if the ratio moves between 0.8 and 1Footnote ³ .

The aggregated index has been calculated based on (Tena-Junguito and Willebald Reference Federico and Tena2013) as follows:

The accuracy index is the ratio of the sum of Bolivian exports to all partner countries, according to Bolivian sources, with the sum of total imports from Bolivia, according to trade partners’ sources. We have taken into account information from 13 different trade partners: Argentina, Belgium, Brazil, Chile, France, Germany, Italy, the Netherlands, Peru, Spain, Switzerland, United Kingdom and United States. The sample covers on average 85 per cent of Bolivian exports and 82 per cent of Bolivian imports (see Appendix 1). Both the bilateral approach and the aggregated index are based on trade flows in current U.S. dollars, using the exchange rates from Frederico and Tena-Junguito (Reference Herranz-Loncán and Peres-Cajías2016), the MOXLAD database (Mc Queen 1925) and the Bolivian Central Bank yearbooks.

At this point, some features of Bolivian trade must be noted. First, Bolivian exports and imports were valued at official prices until 1907 and 1919, respectively (Societé des Nations Reference Agramont and Peres-Cajías1927, pp. 138-139). Thereafter, export values were declared values f.o.b. at the Bolivian frontier, including export duties. Import values were declared values c.i.f. at the Bolivian frontier, excluding import duties (Societé des Nations Reference Agramont and Peres-Cajías1927, pp. 138-139; US 1940, p. 5). Second, Bolivian trade was not always reported separately by its main trade partners. This was the case, for instance, of the official Swiss statistics until 1924Footnote ⁴ . Due to its small scale, Bolivian trade with some of its trade partners was often placed in a residual category named «Other countries».

Third, the accuracy literature has suggested that «import records are usually more reliable than geographical export assignment records» (Agramont and Peres-Cajías 2016). However, it has also been claimed that landlockedness may seriously distort the geographical assignment of trade. This fact was recognised in Societé des Nations (Reference Agramont and Peres-Cajías1927, p. 139), which claimed that part of the Bolivian trade could be systematically assigned wrongly to Chile and Peru, its most important transit countries (Agramont and Peres-Cajías 2016). In the same line, the office in charge of U. S. foreign trade statistics recurrently stressed the difficulties involved in properly assigning Bolivian trade. This is clearly stated in the following quote: «trade of Bolivia with the United States is difficult to ascertain accurately, owing to the fact that it must be transhipped through any one of four countries» (US, 1924).

The previous considerations have a strong influence on the interpretation of the mirror test results between official Bolivian foreign trade statistics and those of its main trade partners. In the case of the United Kingdom (Figure 1), the existence of four cycles stands out: large statistical discrepancies during the First World War, an improvement during the 1920s, a growing gap again in the early 1930s and a new improvement thereafter. The magnitudes of the statistical discrepancies in most years cannot be solely explained through transport costs. Thus, other factors must be at play in influencing the size of these discrepancies.

FIGURE 1 ACCURACY INDEX OF BOLIVIAN TRADE BY PAIR OF COUNTRIES: UNITED KINGDOM, 1910-1949 (%) Sources: Official Bolivian and British foreign trade statistics, various years.

Differences between Bolivian and British official records could be caused by different reasons. A first explanation could be related with an overestimation of Bolivian exports to United Kingdom in Bolivian sources; this would be the case if exports to continental European countries were assigned by Bolivia to the main port of destination. However, the Bolivian overstatement of its trade with United Kingdom, due to inaccurate geographical assignment, can be captured by using the information about re-exportation declared in British sources. In this way we found that some of the differences during the First World War diminish slightly, but the explanatory power of this correction is reduced significantly thereafter, when re-exports represented <1 per cent of U.K. imports from Bolivia.

Another cause of discrepancy can be placed on the British side, if there was an underestimation of Bolivian imports by British sources. There is some evidence to believe that this was part of the problem, at least during the 1930s. Indeed, during this period official British information detailed the sources of tin imports from South America and identified two principal suppliers, Bolivia and Chile. However, Chilean foreign trade sources, as well as mining production information from industrial censuses, show that Chile did not export tin to the United Kingdom during the 1930s, and did not even have any tin mining production during most of this periodFootnote ⁵ . This lack of correspondence could be explained by the fact that 85 per cent of Bolivian exports transited through Chile (Agramont and Peres-Cajías 2016) and, therefore, British authorities may erroneously have registered as Chilean a product that originated in Bolivia.

If this is true and the so-called Chilean tin is added to total imports from Bolivia, the accuracy index should improve significantlyFootnote ⁶ . The first column of Table 2, presents the results of the accuracy index between Bolivia and United Kingdom without any correction. The second column shows the value of U.K. imports from Bolivia, according to British sources. The third and fourth columns distinguish the value of tin imports in pounds as well as a percentage on British imports from Bolivia. The fifth column reflects the British records for tin imports from Chile. The sixth column calculates the accuracy index between Bolivia and United Kingdom, including the “presumably false” Chilean tin. The last column shows the difference between both accuracy indexes. These figures clearly prove that accuracy improves significantly if we correct for a mistaken allocation of tin imports in official British sources. It should be noted, however, that differences remain high even after this correction.

The bilateral contrast with the United States is even worse than in the British case (Figure 2)Footnote ⁷ . Statistical discrepancies were extremely high during the First World War; thereafter, whereas the imports’ ratio converged to reasonable levels, differences were still noticeable in the case of exports. Once more, however, there is some evidence which suggests that this lack of correspondence could be due to statistical assignment problems outside Bolivia. In fact, as previously noted, U.S. records are full of quotes which stress the inability of U.S. statistics to reflect fairly the amount of imports from Bolivia, such as the following:

FIGURE 2 ACCURACY INDEX OF BOLIVIAN TRADE BY PAIR OF COUNTRIES: UNITED STATES, 1910-1949 (%) Sources: Official Bolivian and USA foreign trade statistics, various years.

In contrast to the previous results, it is reasonable to expect smaller differences between Bolivian records and those of its neighbouring countries due to a reduction in the potential of erroneous geographical assignment. The bilateral contrast with Peruvian and Chilean sources shows that this is not necessarily true: regarding exports, notorious differences arise between Bolivian sources and those of both countries; as for imports, differences are particularly noticeable in the contrast with Peruvian sources (Figures 3 and 4)Footnote ⁸ .

FIGURE 3 ACCURACY INDEX OF BOLIVIAN TRADE BY PAIR OF COUNTRIES: PERU, 1910-1949 (%) Sources: Official Bolivian and Peruvian foreign trade statistics, various years.

FIGURE 4 ACCURACY INDEX OF BOLIVIAN TRADE BY PAIR OF COUNTRIES: CHILE, 1910-1949 (%) Sources: Official Bolivian and Chilean foreign trade statistics, various years.

These gaps could be explained by the small quantities that Bolivia exported, that is a scale effect. In this context, smaller differences in absolute values between Bolivian statistics and those of its trade partners could generate large ratio changes; the smaller the trade flow, the higher the potential of discrepancy in the accuracy ratio. For instance, the absolute difference in current US$ between Bolivia and United States was, on average, only one-third of the difference between Chile and United States for the period 1910-1938. However, despite the fact that the statistical discrepancy between Chile and Unitd States was higher in absolute terms, the statistical discrepancy between Chilean and U.S. sources is irrelevant in relative terms because of the higher magnitude of the trade flow (Bolivian exports to United States were, on average, 16 per cent of Chilean exports to United States). If we focus on even smaller trade flows, such as the case of Bolivian exports with neighbouring countries, it could be expected that the scale effect increases potentially.

Another explanation could be related with the existence of a porous border where smuggling activities were not rare (Langer and Conti 1991) and are still frequent to this day. We have explored this hypothesis through the comparison of product composition of Bolivian exports to Chile (using Bolivian sources) and Chilean imports from Bolivia (using Chilean sources). In 1917, Bolivian exports to Chile reached one of the highest levels throughout the period under study and the difference between Bolivian and Chilean sources is 1.75. The contrast shows the existence of problems in Bolivian sources; they declare 328 units of cows exported, but Chilean sources declare 1088 units imported from BoliviaFootnote ⁹ . As a consequence, cows represented 0.58 per cent of the total value exported from Bolivia to Chile, according to Bolivian sources, but 19 per cent of Chilean imports from Bolivia, according to Chilean sources. However, there are also problems on the Chilean side; they do not declare imports of antimony and wolfram from Bolivia and under-declare (particularly in values) imports of tin and copper. These problems are critical since these mineral products are among the top 10 exports from Bolivia to ChileFootnote ¹⁰ . The contrast of both sources was repeated in 1927, a year when the difference between Bolivian and Chilean sources is 4.27. The analysis shows that the five most important exports from Bolivia to ChileFootnote ¹¹ (which accounted for 80 per cent of the total value exported from Bolivia to Chile) are either undervalued (both in values and quantities) or not declared in Chilean sources. Thus, together with scale effects, smuggling or the lack of a precise control in custom offices, this could also explain statistical discrepancies between Bolivian sources and those of its neighbouring countriesFootnote ¹² .

It should also be stressed that statistical differences between Bolivian foreign trade statistics and those of its main trade partners could be explained by the existence of different exchange rates; the gap would be a consequence of the use of different exchange rates in official statistics. Whereas plausible, the explanatory power of this idea is restricted to very specific periods. Indeed, the Bolivian Central Bank Yearbooks show the existence of up to three different exchange rates (and sometimes with huge differences) during the period 1935-1940. The Bolivian exchange rate was unified again in June 1940 and it was not until October 1947 that an additional exchange rate appeared once more.

Therefore, the high ratios of bilateral discrepancies, although with reasonable potential explanations, could generate pessimism both in the use of Bolivian foreign trade statistics and its reconstruction through foreign statistics. This would be in line with Morgenstern’s (Reference Absell and Tena-Junguito1963) claims. However, as suggested by other scholars (Federico and Tena 1991; Tena-Junguito Reference Carreras-Marín, Badia-Miró and Peres Cajías1991; Tena Reference Gómez1991), accuracy issues should also be approached via an aggregated accuracy indexFootnote ¹³ . This is presented in Figure 5 and Table 3, which indicate a clear improvement in the accuracy of Bolivian foreign trade statistics after the First World War. Regarding imports, the ratio is within reasonable levels during the 1930s. As for exports, despite the previously mentioned improvement over time, the ratio suggests that the value of exports declared by Bolivian records was 50 per cent higher than the value of exports declared by its main trade partners.

FIGURE 5 AGGREGATED ACCURACY INDEX OF BOLIVIAN EXPORTS AND IMPORTS, 1910-1949 (%) Sources: Bolivian and its trade partners Official Foreign Trade Statistics, various years.

Summing up, both bilateral and aggregated indices reveal the lack of correspondence between Bolivian official records on exports and those of its main trade partners. It has been stressed that these differences may not necessarily have been caused by a lack of accuracy in Bolivian sources. Indeed, some of the problems can be reasonably related to inaccurate geographical assignments in British and U.S. sources, which absorbed most of Bolivian exports. This hypothesis is supported by qualitative information in U.S. sources and by the finding of an erroneous allocation of tin imports in British sources. We have also argued that the scale effect overstates statistical discrepancies that were not so extreme in absolute terms. In any case, the fact that accuracy indices are systematically above 1, still suggests that Bolivian sources overvalued exports during the first half of the 20^th century. This hypothesis is further explored in the next section.

3. WAS THERE A PROBLEM OF OVERVALUATION IN BOLIVIAN SOURCES?

Previously, it has been stressed that, after 1907, Bolivian exports were valued according to market prices and included export taxes. The fiscal nature of exports may introduce some biases which could operate in opposite directions. On the one hand, it could affect the amount declared by exporters, in order to minimise the payment of taxes. On the other hand, government agents could have the incentive to increase the value registered in order to maximise revenues or, more in the spirit of political economy, to create an artificial impression of high taxation on tin exporters. Overall, the net effect of taxes on the value of exports is ambiguous.

Table 4 shows that export taxes represented a significant share of the Bolivian central government’s total revenues (first column), but they were equivalent to a small share of Bolivian exports (fourth column). If we consider overall taxes on mining activities (second column), which constituted almost all Bolivian exports, it stands out that its relative importance over total revenues increased dramatically from the early 1920s to the 1950s. Likewise, its magnitude as a proportion of the total value exported achieved more significant shares over time (last column). Thus, whereas the impact of export taxes on the value of exports could be limited, the impact of overall mining taxes could affect the final price of exports. This would require the ability of mining producers to shift taxes to final consumers, a possibility that has been questioned elsewhere (Ingusltad et al. 2015; Peres-Cajías Reference Bértola2015).

An alternative to identifying the potential overvaluation of Bolivian exports is to compare the implicit price of Bolivian tin exports (total value exported divided by total quantities exported) with the international price of this commodity. The focus on tin prices is justified by its high relative importance in Bolivian total exports; between 65 per cent and 75 per cent, as previously mentioned. The comparison with international prices should be carried out with caution since the quality of Bolivian tin was lower than the ore from Malaysia and Cornwall (US 1940, p. 2; Gómez 1978). In relation to this, Table 5 shows that Bolivian tin exports were overwhelmingly composed of tin rods (barilla de estaño) and tin waste (escoria de estaño), two products with a lower implicit price than tin bars (barra de estaño). In spite of this, it must be considered that the ore quality of these exports was not homogeneous, a factor which determined where Bolivian ores were smelted (Hillman 1988, 1990).

Taking into account these considerations, Figure 6 compares different prices: the price of tin in the British and the U.S. markets, the price of tin straits (which were of higher quality), the price of tin ore and the implicit price of tin exports obtained from official Bolivian statisticsFootnote ¹⁴ . In order to make the comparison feasible, the implicit Bolivian price of tin exports has been increased by 20 per cent, which represented the transport cost between Bolivia and the markets previously mentioned. The first feature to underline is the high correlation of the implicit Bolivian price with prices in the United Kingdom and the United States; a positive correlation of 0.80 and 0.75, respectively, for the period 1908-1949. Regarding price levels, the heterogeneous quality of Bolivian tin exports suggests that the implicit Bolivian price of tin does not have to fit perfectly with international prices. In this context, from 1908 to 1929, the implicit Bolivian price of tin is more in line with the tin ore price and below the price of tin in the British and U.S. markets; from 1930 to 1940, the obtained implicit price is around the same levels as the international price of tin in United Kingdom and United States; from 1941 to 1949, the implicit Bolivian price is certainly above international prices.

FIGURE 6 IMPLICIT PRICE OF TIN EXPORTS AND INTERNATIONAL PRICES, 1908-1950 (US$/TON) Sources: Official Bolivian and British foreign trade statistics and Tena-Junguito and Federico (Reference Herranz-Loncán and Peres-Cajías2016).

Summing up, with the exception of the period 1941-1949, the statistical discrepancy of Bolivian exports could hardly be attributed to an overvaluation of Bolivian export prices. Moreover, given that implicit Bolivian prices are sometimes below the international price of tin, following the common approach to correcting export values with international prices (Tena-Junguito and Willebald Reference Federico and Tena2013; Absell and Tena-Junguito 2016), this would simply increase the gap between Bolivian records and those of its main trade partners, worsening the accuracy index. Likewise, it must be noted that the period with the highest differences between Bolivian and international prices, is the period with the lowest differences between official Bolivian statistics and those of its trade partners. All this suggests that, given the available information, the use of official Bolivian foreign trade records seems to be the best option to analyse the evolution of Bolivian exports. In any case, before concluding with this statement, the next section develops some additional issues regarding the accuracy of Bolivian trade statistics.

4. ADDITIONAL ISSUES CONCERNING BOLIVIAN STATISTICS

Despite the fact that, so far, the reasonable accuracy of official Bolivian foreign trade statistics has been stressed, there are some minor issues that have to be addressed. On the one hand, as a mining producer, Bolivia faced the problem of bullion; silver, gold and nickel could be exported as ore, as a manufacture or as coins. In the latter case and according to standard procedures, these flows do not have to be included in the trade balance of countries, but in the capital balance. Although the relative importance of silver exports decreased during the first half of the 20^th century, we highlight this issue as an additional cause of discrepancy with trade partners’ sources. This would be the case if Bolivia was recording minted minerals as merchandises while its trade partners were assigning them to capital inflows. On the other hand, Bolivian sources also face the problem of re-exportation. This could also be a source of overvaluation of exports due to an inaccurate record in the Bolivian side. In the following two sub-sections, we analyse the importance of both effects on the accuracy of official Bolivian statistics.

5. SERIES OF BOLIVIAN EXPORTS AND IMPORTS (1910-1950)

Until the recent reconstruction of the Bolivian GDP and GDP per capita from 1846 to 1950 (Herranz-Loncán and Peres-Cajías 2016), knowledge of the Bolivian economy, during the First Globalisation and the interwar period, relied essentially on the analysis of the official foreign trade statistics (Peñaloza Cordero Reference Albarracin Millan1985; Morales and Pacheco 1999; Bértola 2011). In this paper, we have not found a solid argument to reject the use of these figures. Indeed, we have stressed that the standard accuracy analysis, based on the mirror contrast with trade partners, presents too many flaws in the Bolivian case. Our main objection to the standard methodology is based on the systematic existence of a geographical bias on the trade partner side due to landlockedness and the effects of transit trade. This idea is supported both by quantitative and qualitative evidence, particularly in the case of the United States and United Kingdom. Furthermore, we have found no evidence of overvaluation of prices in Bolivian statistics which would have led us to distrust Bolivian data. Silver bars, wrongly assigned by trade partners’ statistics, have also been put forward as a potential explanation, although with limited importance, of statistical differences. Thus, given the available evidence, we can state that official Bolivian foreign trade statistics are the best option for the study of Bolivian trade.

Once we have stated the reasonable accuracy of official Bolivian trade statistics, we can offer the series of exports at current prices (Figure 8). The new series does not match completely with the original official figures as we have modified them according to the minor changes explained in the previous section. These consisted of the subtraction of: (a) coinage goods, either from silver, gold or nickel; and (b) re-exportation of different kind of manufactures. As Figure 8 shows, these corrections are not of great significance and neither the official nor the corrected series is coincident throughout the whole period, with some exceptions at the beginning of the period and in 1929.

FIGURE 8 BOLIVIAN EXPORTS: OFFICIAL VS. CORRECTED SERIES, 1910-1950 (US$ AT CURRENT PRICES) Sources: Official Bolivian foreign trade statistics, various years.

Figure 9 presents the export series of Bolivia in US$ at current and constant prices from 1900 to 1950Footnote ¹⁷ (see Appendix 3). This period is characterised by an increasing concentration of Bolivian exports in one specific product (tin) and three main local producers (the so-called Tin Barons). This generated an ongoing debate regarding tin exports as a potential engine of growth and their mitigation by the oligopolistic structure of the sector. Some scholars argue that the economic impact of tin exports, as a driver of structural change in Bolivia, was negligible (Peñaloza Cordero Reference Albarracin Millan1985). In this context, a widespread claim in Bolivian historiography has emphasised the Barons’ lack of interest in promoting industrialisation, based on the fact that most of tin exports were almost exclusively composed by raw mineralFootnote ¹⁸ . Moreover, taking into consideration the vertical integration and/or the registration of their companies in North America and Europe, it has been suggested that the Bolivian Tin Barons (namely Simón Patiño, José Avelino Aramayo and Mauricio Hotschild) behaved like foreign capitalists and transferred all their profits abroad (Albarracin Millan 1995)Footnote ¹⁹ .

FIGURE 9 BOLIVIAN EXPORTS, 1900-1950 (US$ CURRENT AND CONSTANT PRICES) Sources: Official Bolivian foreign trade statistics, various years.

However, before studying the use of exports’ earnings, a first basic condition in any export-led growth model is related with the analysis of export sector dynamism. Our data allow us to identify three clear periods concerning this issue: a clear upward trend from 1904 (the year when tin consolidated as the main export) to 1920; a second period from the early 1920s to the late 1930s, when Bolivian exports experience constant and significant oscillations; a third period characterised by a positive trend from the outbreak of the Second World War which lasted, in nominal terms, until 1949, but stopped abruptly at the end of the war in real terms. Therefore, our data suggest that the potential of Bolivian exports as a growth engine was highest at the beginning of the 20^th century. Further analysis (Peres-Cajías and Carreras-Marín Reference Carreras-Marín and Badia-Miro2017), which contrasts different indicators that show both the direct and indirect contribution of Bolivian exports, reaches similar conclusions. Thereafter, the impact of exports was reduced because of volatility, which was partly determined by the high concentration of Bolivian exports in one product and few markets.