Portal de Periódicos da CAPES

Wealth Holding in Southeastern Brazil, 1815–60

2005; Duke University Press; Volume: 85; Issue: 2 Linguagem: Inglês

10.1215/00182168-85-2-223

1527-1900

Zephyr Frank,

Historical Economic and Social Studies

Southeastern Brazil underwent a series of profound transformations over the course of the first half of the nineteenth century. Rio de Janeiro, the capital-in-exile of the Portuguese Empire, became the capital of a newly independent but also monarchical Brazil. The city was transformed by urban growth, a massive influx of African slaves, and the rise of wealthy merchants and coffee planters to the head of local society.1 Beyond the capital, other important zones in Southeastern Brazil also prospered as the result of domestic market forces and the pull of the Atlantic economy.This portrait of growth contrasts with the conventional image of Brazil as a "sleeping giant" during the first half of the nineteenth century. Most accounts of nineteenth-century Brazilian economic history begin their story of growth and transformation with the rise of São Paulo as a coffee producer, then industrial center, sometime in the latter part of the nineteenth century.2 Sometimes the province of Rio de Janeiro is singled out as the locus of growth, especially from the 1840s through the 1880s. For the rest of Brazil, scholars have reported little or no real per-capita economic growth over the nineteenth century. Nathaniel Leff's influential account estimated per-capita income growth in Brazil from 1822 to 1913 at just 0.1 percent—largely due to the lackluster productivity of Brazil's extensive subsistence agriculture sector. Even Southeastern Brazil, Leff calculated, experienced real growth rates in the range of only 0.2 to 0.4 percent—still quite low.3 The enormous weight of low-productivity subsistence agriculture, antiquated financial laws and credit instruments, low schooling rates, high transport costs, and declining terms of trade have all been adduced as factors retarding economic growth in Brazil during the first three-quarters of the nineteenth century.4Yet, over the past two decades, the literature has begun to portray parts of Brazil, particularly the Southeast, as much more economically dynamic during the first half of the nineteenth century.5 Inspired by this revisionist trend, my research shows that areas of Southeastern Brazil were relatively wealthy at the dawn of independence and that mean wealth grew substantially over the next four decades. In addition, I suggest that as wealth grew, wealth distributions followed varying trajectories. Rising wealth did not always mean rising inequality. Finally, I seek to reveal the sources of growth in wealth by focusing on institutions (property rights, slavery, commercial law) and markets (internal and external). Important changes in institutions and markets occurred in Brazil prior to 1860, including the end of the Atlantic slave trade, the creation of modern capital markets, and the rise of coffee exports.The analysis focuses on two distinct periods: 1815–25 and 1850–60. The first period straddles the end of Portuguese rule and the first years after Brazil won her independence in 1822. The second period, roughly two generations after independence, begins after the definitive suppression of the slave trade in 1850 and the implementation, in 1850 and 1854, of new laws in land and commerce in response to the growing crisis in slavery. This periodization allows for analysis of the role of institutional change in patterning the growth and distribution of wealth. It also allows for an examination of the power of export growth to transform wealth holding: exports were low and stagnant through the late 1840s, after which they more than doubled by the mid-1850s.6My data were drawn from samples of postmortem estate inventories (inventários) covering each of the two periods and drawn from three locations: the city of Rio de Janeiro, the counties (municípios) of São João del-Rei and São José in Minas Gerais, and a more limited sample from the city of São Paulo. I will refer to the two counties in Minas as São João–São José unless I specifically refer to only one of them. I use Rio de Janeiro to refer only to the city, and not the entire province, unless otherwise indicated. I took care to collect sufficiently large samples of wealth holders in each location. In Rio, I selected one out of every three inventories in the National Archive for the two periods. The São João–São José sample represents one out of every two inventories. In São Paulo, owing to the paucity of inventories for the earlier period, I was forced to use every available inventory. In all, I collected a total of 1,220 inventories containing quantifiable information for the primary dataset. Additional inventory samples were drawn for São José and Rio de Janeiro in order to clarify the relationship between the wealth of the deceased and that of living wealth holders.Although researchers have long used estate records to study wealth holding in Brazil and elsewhere, methodology has been unsystematic and rough.7 There is as much art to selecting, deciphering, and coding estate inventories as there is science. Comparisons between different studies are difficult due to differing sets of categories, methods of presenting size distributions, sampling rates, and periodizations. Moreover, the estate inventory literature on Brazil rarely engages in careful sensitivity testing with regard to the robustness of its claims.8Using estate inventories to measure wealth and change over time entails at least three basic problems. First, the sample size in any given year is likely to be too small to be statistically significant and reliable. The very wealthy, in particular, do not die on a regular annual basis, and so samples must cover multiple years in order to capture both a significant number of decedents and a fair number of the wealthiest decedents in a locale.9 Second, the inventoried population is probably older and wealthier than the population of living wealth holders. Poorer wealth holders are less likely to have left an inventory.10 We must account for potential age bias and the underrepresentation of poorer estates if we are to make reliable comparisons regarding wealth holding and distribution over time. In short, when comparing 1815–25 with 1850–60, we need to be reasonably sure that the age structure of the population had not radically altered and that the rate of underrepresentation of poorer estates had not changed. If these factors can be held relatively constant, then we can proceed with our comparisons of wealth holding over time.There are good reasons to believe that ceteris paribus is a reasonable assumption for comparisons between the periods under review. From 1815 to 1860, Brazil did not absorb large numbers of free immigrants, although it did receive a steady but small flow of Portuguese immigrants and a massive flow of African slaves up to 1850. Age structures (among the free) probably did not change dramatically in any of the locations under review.11 Moreover, there were no major institutional changes that would have made declaring wealth in estate inventories more or less likely, and thus it is reasonable to assume that poorer estates were equally underrepresented in both periods.12 The fact that the lowest observed values do not change significantly over the periods is also strong evidence against censoring owing to changing thresholds at which wealth was declared—that is, there does not appear to be a tendency for very small values to drop out of the sample over time. I will present further evidence regarding levels of underrepresentation later on.13Third and last, even if we can be reasonably certain that our samples are large enough and reasonably representative (after accounting for age bias and other distortions) of wealth holders at all levels of wealth, we still need to know what proportion of all households held wealth. Unfortunately, we can only address this last question obliquely. In Southeastern Brazil, 30–40 percent of all households owned slaves and 30–70 percent owned homes (with the lower percentages found in the larger cities). As a preliminary measure of fit between the estate inventories and the living population of wealth holders, it is possible to compare the mean size of slave holdings in the two groups. In the city of Rio de Janeiro, for example, about 33 percent of households owned slaves and held, on average, about 11 slaves per household, circa 1849.14 In the inventory sample for 1845–49, the average holding among slave owners was 12.35 (although this includes slaves held on rural estates outside the city).15 Thus, in Rio, decedents and living wealth holders owned similar numbers of slaves. The same holds true for earlier and later sample periods.16 Manuscript census returns from 1831–32 for São João–São José (graciously provided by Clotilde de Paiva) also allow us to calculate the distribution of slaves among the living population and to compare that distribution with inventory samples. Living slave owners held a mean of 7.82 slaves, according to the census.17 Estate inventories list an average of 11.6 slaves per owner—higher than among the living, as we would expect, but not outlandishly so. Ideally, one could simply divide the mean slaveholding of the living by the mean holding in the inventory sample to derive a conversion factor to correct for differences in wealth between living wealth holders and decedents. Unfortunately, slave wealth does not predict levels of nonhuman wealth very accurately.18The best way to judge the relationship between the estate samples and the living is to link estate inventories to heads of households listed in the 1831–32 São João–São José census. To maximize the chances of making positive matches, I examined all inventories for São José for 1833–39. Several factors complicated the matching process, including repeated names and the omission of full names for many household members in the census. In the end, 58 strong matches were made.19 This relatively small number still provides critical insight into the fit between the samples and the living.It is reassuring to note that, in many ways, the inventories correspond well with the living. The matched sample averaged 8.74 slaves per decedent versus 7.42 slaves per living owner. Major professional categories are also similarly represented among decedents and living household heads, with the exception of merchants, who turn up roughly twice as often in the inventories. The most important information contained in table 1, however, corresponds to the question of age bias. I expected the inventoried population to be older and wealthier than the living population of wealth holders. Based on the matched sample analyzed here (which is too small to draw any final conclusions), it appears that correcting for age bias would lower mean wealth by about one-third. Put another way, the inventory sample appears to overstate wealth by about 33 percent relative to a hypothetical age-balanced sample. All in all, estate inventories do not appear to be greatly at odds with patterns observable among the living population of household heads, at least in the case of São José. The inventoried population is, however, older (by about 11 years, on average), and therefore wealthier, than living wealth holders.I performed a similar analysis of 49 cases for which I had age data from Rio de Janeiro in the 1880s (out of a total of 143 cases). As expected, the youngest age group (15–39) held less wealth on average than the middle and upper groups. The consequence of multiplying each of these age groups in the sample by their approximate weight in the total living population resulted in a 23 percent reduction in mean probate wealth after adjustment.20 This figure is calculated based on a small sample of data from a period outside of our present study, so it must be seen as a very rough approximation to be taken into consideration along with the results reported above for São José.The proper interpretation of the estate samples depends upon the questions being asked. If we want to use the inventories to estimate the amount of wealth held by living household heads, we will need to correct for the age bias and other limitations of the inventory samples. If, however, we want to make comparisons over time, we are less concerned with age bias (so long as it remains constant) than with the comparability of the samples in terms of geographic representation, marital status, professional status, and so on. Finally, if we want to assess levels of inequality, we need to know how age bias and the presence of zero-wealth households affect the distribution of wealth based on our samples. A central question, then, is: How representative and stable are the estate samples? How many households among the living held measurable wealth that was likely to be inventoried?Among the living, 30–40 percent of households owned at least one slave; among the inventory samples, 80–90 percent of decedents reported owning slaves. After adjusting for age and other sources of bias, it seems reasonable to suggest that the inventory samples shed light on the third or so of households known to own slaves. Another 10–20 percent of the inventories in the samples, however, represent households that held wealth but owned no slaves. Clearly, the incidence of wealth holding among the living extended beyond slaves. A promising alternative to slaveholding as a measure of the incidence of wealth holding among the living is found in property tax lists. In the case of São João, I obtained the urban property tax roll for 1826. According to this illuminating document, 68 percent of households with identifiable occupants were owner-occupied (including houses occupied by heirs and widows), and 32 percent were rented.21 In the case of Rio de Janeiro, I analyzed the complete run of the urban property tax for 1849. As expected, the ownership rate was lower, at approximately 37 percent, and the implicit rental rate higher, at about 63 percent, in Brazil's capital.22 It is likely, based on property tax and slaveholding data, that more than half of all households held at least some wealth (at the very least, a house or a slave) in all three of the locations covered in this study. As long as our samples are broadly representative of the living, they are not limited to a small elite.23Finally, even if the samples are consistent and we can correct for bias, treating human beings as a form of property is problematic. On the one hand, nineteenth-century Brazilians certainly counted slaves as property, and wealth in slaves accounted for 14–40 percent of all wealth listed in the inventory samples analyzed. On the other hand, when we consider questions of economic inequality, we must remember that slaves were themselves persons. This does not mean that we can or should retroactively "abolish" slavery by omitting wealth in slaves from our analyses. It does mean, however, that we must problematize the discussion of human beings as property, lest we fall into the trap of accepting social constructions for more than they are. Slaves were wealth only insofar as the society in which they lived made them so. Paradoxically, as this essay will show, the institution of slavery drastically limited the freedom and potential of the enslaved population, while at the same time it engendered a more even distribution of wealth among free wealth holders.According to every measure, real wealth grew in Rio de Janeiro and São João– São José at a substantial rate between 1815 and 1860. São Paulo shows an even higher rate of growth, although this may be overstated owing to the smaller size of the sample and low level of mean wealth found for the earlier period. Table 2 presents six different measures of mean wealth holding for Rio de Janeiro and São João–São José and two measures for the smaller São Paulo sample. Although there is significant variation in the implicit annual rate of growth, depending on the measure, all of the results point to rates in excess of 1 percent per annum.Following Alice Hanson Jones, I prefer physical wealth (measure 3) over net figures (measure 2) that include debts and credits. As Jones points out, debts and credits should cancel out over the whole of a society.24 Moreover, we cannot be certain of the quality of debts and credits. There is some evidence to indicate that at least some debts could not be collected.25 Finally, although it is likely that cities and towns were in a position of being net creditors to the countryside, thereby explaining the fact that credits outweighed debts in nearly every time period (see table 3), we cannot corroborate this with data at this point. Measures 2 and 3, representing net and physical real wealth respectively, both point to growth in wealth deflated by a combined index of internal prices and the exchange rate.26Although I view the estimate of real mean physical wealth—measure 3—as the most accurate, it is also possible to deflate the wealth estimates by rendering them in pounds sterling, as reported in measure 4. This procedure has advantages as well as disadvantages. On the positive side, it skirts the issue of whether or not the price indexes utilized in measures 2 and 3 are accurate and comparable across regions. More specifically, the rate of inflation reported by Lobo for Rio de Janeiro is much higher than any other measure, even when using her most reasonable series. This injects a downward bias on the estimates of growth for Brazil's capital and renders comparison with São João–São José slightly problematic.27 Unlike Lobo, Graça Filho provides convincing data for prices in São João, including a wider range of products and a component for imported goods.28 On this basis, it is fair to conclude that the net and physical wealth estimates for São João–São José are the most accurate of all. On the negative side, the pounds sterling estimates fail to account fully for what we know was a high rate of internal inflation in Brazil over the period in question. In the last analysis, the pound-denominated values represent an upper bound and the mil-réis-denominated values the lower bound for real wealth in Southeastern Brazil.A fifth measure attempts to address the underrepresentation problem in a blunt way by omitting all estates below the level of 100 pounds sterling. It does little, however, to change the annual rate of growth compared with the entire sample. Even more to the point, the percentage of estates falling below this level remains steady over the two periods: 86 percent in period 1 and 93 percent in period 2 for Rio de Janeiro; and 86 and 83 percent respectively in São João–São José. The underrepresentation of smaller estates did not, by this measure, change significantly over the time period in question.29Even if we have discounted the potential problem of changing levels of underrepresentation, it remains conceivable that the samples somehow differ in their composition. Measure 6 provides further evidence of the coherence and stability of the samples over time by isolating married decedents. Not only do married decedents make up similar proportions of all decedents in both periods, the mean value of their estates and rate of growth over time is also similar to the full sample. This finding is, in fact, significant, given that inheritance law stipulated the equal division of an estate between the surviving spouse and necessary heirs in most cases.30 As such, the estates of married decedents represented total family wealth, whereas the estates of widows and widowers represented, in most cases, just half of the total estate. The consistency of the proportion of married decedents and the consonance of their mean wealth with that of the whole sample lend further support to our claim that our samples are consistent over time.If the reader wishes to convert the values reported in table 2 to wealth per living household head, the following rule of thumb will suffice (although the results are necessarily rough): wealth per living household head should equal approximately 0.7 times mean inventoried wealth (derived from table 1 and the discussion in the text) divided by mean household size, then multiplied by the estimated wealth-holding household ratio. For São João–São José, I suggest a ratio of 0.7, and for Rio, 0.5. The lower ratio for Rio reflects the fact that fewer residents of Brazil's capital owned their own home, and therefore the implicit number of households holding zero or no wealth must have been higher. Using these multipliers, nominal mean wealth per head among the living population of Rio de Janeiro circa 1820 was 328 mil-réis.As I have noted before, assuming that 50 percent of Rio's households held no wealth may be extreme. Employing Jones's concept of nonprobate-type wealth, we might assume that the population is divided into two groups: those likely to be inventoried and those not. Working with probate records from colonial North America, Jones suggested a nonprobate ratio of 0.25 for wealth in this class—that is, the wealth of uninventoried individuals was estimated at one quarter, on average, of the inventoried population's average wealth holding.31Jones's technique was complicated, and I do not intend to repeat it here. However, a very simplified application of this ratio to the 50 percent of households we have hitherto assumed to hold no wealth raises our estimate of mean wealth per living head to 367 mil-réis in nominal values circa 1820.Neither Rio de Janeiro nor São João–São José can be characterized as poor at the start of the national period; even the smaller sample for São Paulo indicates a fairly high level of wealth holding. Moreover, all three locations saw substantial increases in wealth over the first four decades of the empire. Were these rates of growth common throughout Brazil? In other parts of Southeastern Brazil affected by the rise of marketing and trade, and especially by the development of the coffee industry, the answer is likely yes. Renato Marcondes's study of the town of Lorena, in São Paulo's Paraíba Valley, indicates extremely rapid growth in wealth holding over roughly the same period. Although his sample size is small, the estate inventory data show a rise in mean wealth from 656 pounds sterling in 1830–39 to 2,614 pounds sterling for 1855–64.32 As was the case in São Paulo, this growth was fueled by coffee.Moving outside the Southeast, Katia Mattoso's data for Salvador, capital of the important Northeastern state of Bahia, show that, as expected, rates of growth in other parts of Brazil were much lower. According to Mattoso's figures (if my calculations are correct), the mean nominal value of an estate in Salvador—roughly 2,000 pounds sterling for 1801–21—was slightly higher than in Rio de Janeiro between 1815 and 1825.33 Because I have not been able to obtain Mattoso's raw data, it is probably safest to assert that the level of wealth in Rio was similar to that in Salvador. Again, I argue that roughly equal values in these two major cities on the cusp of Brazilian independence are consistent with expectations. It is important to note that whereas Rio de Janeiro showed rapid growth through the 1850s, Salvador did not. Mattoso's data indicate that the mean size of an estate remained substantively unchanged for 1845–60, at 2,083 pounds sterling.34In sum, as expected, wealth in large coastal cities was greater than wealth in the interior of the country. Wealth holdings in São João–São José averaged about two-thirds the level seen in Rio and ended up about half as much by the 1850s. The low level of wealth found in São Paulo circa 1820 is surprising. However, the sample size for this period is too small (N = 62) to draw definite conclusions about poverty in the Paulista capital. By the 1850s, according to a larger sample, São Paulo had converged on São João–São José. Higher rates of growth in the Southeast moved Rio de Janeiro well ahead of Salvador by the 1850s, while the smaller cities and towns of the interior were rapidly approaching the levels of wealth found in Bahia's capital.Essentially, there are three possible explanations for change in wealth-holding patterns during this period. These explanations are not mutually exclusive and, indeed, may influence one another. First, it is possible the accumulation of wealth during this period resulted, in part, from the rise in international trade associated with the transport revolution and increasingly integrated world markets. This transformation would presumably have had the greatest effect beginning with the first major upsurge in exports during the 1850s. Trade may have had substantial consequences both for aggregate rates of economic growth and for social structure. Rising exports signified an expansion of what Leff termed the "advanced" sector of the economy.35 And Jeffrey Williamson has argued that the nineteenth-century Atlantic economy engendered powerful transformations in social structures in both labor- and land-rich societies. Inequality dropped in labor-rich societies due to increased trade and immigration. In contrast, inequality tended to increase in land-rich societies, as the privileged stratum of landowners saw their assets appreciate much more rapidly than other forms of wealth.36Institutional changes may have also altered patterns of wealth holding and led to growth. The institutional approach, associated with Douglass North, posits institutions as primary determinants of economic structure and growth.37Institutional changes following Brazilian independence might have influenced wealth after 1822.38 The power and effectiveness of the Brazilian state increased during the late 1830s and 1840s, culminating in a strong government headed by an active monarch, Dom Pedro II.39 Increased state power set the stage for a series of important changes in Brazil's legal framework and economic institutions. Legislators reacted to the suppression of the Atlantic slave trade in 1850 by modifying property-rights institutions in favor of (large) landowners and enacting a "modern" commercial code that encouraged the formation of joint-stock companies.40The effects of these institutional changes were felt immediately in Rio de Janeiro. During the first year of the new code, just three companies were traded ten times on the exchange. By 1855, near the peak of the first cycle of company formation and trading, there were 16 companies involved in 113 trades.41 The total capital of all registered companies traded between 1850 and 1865 came to 237,448 contos.42 According to the standard literature, there was a direct connection between the suppression of the slave trade and the rise of more sophisticated financial markets, as capital sought alternative avenues of investment.43 Also, the maturation of the Brazilian nation-state in the 1830s and beyond created an expanded market for public debt. As of 1827, roughly at the end of the first period analyzed here, the total internal public debt stood at 5,007 contos; by 1860, this had grown to 61,500 contos worth of bonds (apólices).44 As we shall see, this expansion of stocks and bonds is reproduced in the disaggregated instruments of wealth holding in the samples.A third factor, perhaps closely related to the first two, was internal economic growth associated with the development of markets, regional specialization, and credit networks. Although a few scholars have suggested cases of such growth, this variable has received the least amount of attention in the literature.45 In a series of recent publications, João Fragoso issued a series of revisionist claims regarding the timing and scope of economic growth in Southeastern Brazil. Fragoso argues, in particular, that the first half of the nineteenth century saw the rise of a merchant capitalist elite in Rio de Janeiro that, most importantly for our purposes, he largely associates with "endogenous accumulation." By this, Fragoso means an elite that engages in what Marxists call extraction of surplus within a context that remained largely precapitalist. His underlying argument is that the rise of a local merchant elite and the expansion of the domestic market led to economic growth over the first half of the nineteenth century.46 Fragoso's research into trade up and down the coast (cabotagem) revises our conception of Brazil as a series of enclaves oriented toward foreign export, within a sea of subsistence production. Indeed, he found that trade with other coastal cities constituted the greater part of the commerce conducted by major merchants in Rio's marketplace.47 This revelation makes it immediately clear that the size of the foreign export "motor" is less important than previously thought in priming the pump for domestic growth.48As markets form or expand, producers modify their behavior and resource allocation—even in cases such as São João–São José that seem to lack the horsepower (such as export growth) to respond. In the interior of the country, especially where coffee was of limited importance, such growth is critical to explaining rates of wealth accumulation. Indeed, because so much of the "stagnant Brazil" story has centered on the tardiness of Brazil's export boom, it is the internal side of the growth equation that most needs elaboration. One of my central claims, then, is that parts of Southeastern Brazil had substantial capacity for internal economic growth and wealth accumulation in the first half of the nineteenth century, without recourse to technological innovation or major increases in capital. I stake this claim, in large part, on the empirical and theoretical evidence for domestic economic growth in the United States and Europe between 1700 and 1850.49 Jan de Vries sums up the essentials of this idea in his definition of the "Industrious Revolution," which took place from the mid– seventeenth century to the early nineteenth and which "consisted of two transformations: the reduction of leisure time as the marginal utility of income rose, and the reallocation of