This article questions the usefulness of the tithe data of colonial Mexican bishoprics as an index of agrarian production in New Spain. It will present evidence of the close relationship that exists between the Mexican tithe data and those of Mexican silver production and Royal Treasury income, a relationship that might indicate inflation. In doing so, the main objective is to highlight the background of the inflationary development. For this purpose, quantitative material on Bourbon Mexico, which has appeared over the past few years, is brought together and investigated exploratively using modern methods of statistical analysis. Investigating the validity of the data, some of which have been controversial, it will be shown that several series are indeed masked by what we have named a bureaucratic component. This essay thus joins a current and central debate in Mexican historiography regarding the assessment of economic performance in the late colonial period.

Since the publication of the initial volume of the Mexican colonial treasury data by John TePaske and José Hernández Palomo (and the subsequent publication of Andean and additional Mexican treasury data), there has been a lively debate on their relevance. Several authors rely on these data for generalizations about the economic development of colonial Mexico. Some stress the strong evidence of growth and increased wealth in the colony in the second half of the eighteenth century; others speculate on a late colonial mining industry in deep trouble.1 Currently, the tithe series of Mexican bishoprics are being drawn into this debate. Because colonial Mexico was an agrarian society—according to the argument— the tithe receipts could be introduced as indicators of economic development, complementing the treasury data. Morin, for instance, discusses the receipts of the bishopric of Michoacán as part of an overview of its economic development; Medina Rubio published a series of the receipts of the bishopric of Puebla with the same intention; and Garavaglia and Grosso interpreted the series as an index of economic development in their review of the eighteenth-century agrarian situation in colonial Mexico. A group of researchers under the leadership of Rodolfo Pastor published the tithe series of the bishopric of Oaxaca with some sagacious comments, but nevertheless presented them as a reflection of agrarian development.

Héctor Lindo Fuentes, discussing the findings of Pastor et al., refers to one of their conclusions: “De acuerdo con ellos la forma de la curva del volumen de plata acuñada en la Nueva España durante el siglo XVIII es similar a la de la curva de los diezmos, y esto indica que hubo una relación simbiótica entre la producción agrícola y la minería.” But he adds, “Cuando una cantidad mayor de dinero compra el mismo número de productos, los precios tienen que subir. Al comparar los datos de acuñación con la serie de diezmos a precios corrientes lo que los autores ven, al menos en parte, es la relación entre el crecimiento de la masa monetaria y la inflación.”2 Lindo Fuentes correctly discovers the weak point of the tithe data, for in all cases what is used is the tithe receipts of the bishoprics in cash—la recaudación del diezmo líquido en pesos—rather than in kind.

As stated above, our first aim is to demonstrate the limits of this source as a “luminous” index of agrarian, and indeed economic, development.3 We agree with Lindo Fuentes that the tithe in cash reflected above all the monetary situation of the economy of New Spain: the late eighteenth-century boom in tithe receipts was caused by an explicit price-inflationary pressure. This was a consequence of growing money output, decreasing value of silver coins, and growing demand in combination with lagging output in agrarian and commodity production. Two further aims are to explore the possibilities of shedding additional light on this inflationary pressure by analyzing different series of data and to reveal the bureaucratic component that distorts the data, especially those concerning the Royal Treasury. For these efforts—critique and exploration—we will use correlation coefficients, canonical correlation analysis, and the technique of linear dynamic systems analysis (the latter by means of the program DYNAMALS) in order to examine the relations among the tithe receipts and economic indicators classified as inflationary, bureaucratic, and purchasing-power variables.

Scope and Context

The reliance on tithe receipts in cash by historians working on Latin America is curious in the light of European agrarian history. From the first application by French historians, the tithe receipts have been severely criticized as an index of agrarian and economic development.4 It has been recognized that the tithe in cash primarily reflected church income, and therefore church politics to increase its revenues. The tithe as a biblical institution dates back over two thousand years. It became a traditional means by which the church could collect each year a percentage of the “fruits of the earth,” e.g., crops and livestock, but there were impressive local and regional variations in the level of the tithe in Europe, depending on the power of the church to effect its policies. The rate at which it was collected hovered around 10 percent, but it could be as high as 14 or as low as 2 percent. There were five ways of collecting:

It proved most secure (and most profitable) to farm the tithes out (i.e., lease them) for several harvests in money. Farming out to one or more persons, the tithe farmers, was also a practice in the bishoprics of the Spanish colonial empire. The tithe farmer was responsible for collecting the tithes and for handing over to the lessor the sum agreed in the tithe contract. Under normal circumstances, he further had to make profit, which was his remuneration. Usually, the lease covered a longer period than one particular harvest and included all the crops and newborn animals within a certain tithe jurisdiction. But the tithes paid in money give an indication of the tithe income in the currency of the period. They can only begin to yield interesting data, according to the French historian Joseph Goy, if one takes the trouble to compare them with price curves.

Here complications arise, for such a comparison really amounts to “deflating” the cash tithe by expressing it in terms of its equivalent. Deflation can be achieved by dividing the money tithe series by a price index. For a tithe series based on a combination of products, this means using a composite index derived from the prices of the main products of each tithe jurisdiction, according to the proportions of each of these products. In New Spain these would include, among others, maize, wheat, barley, beans, sheep, and cattle. Goy, in general, is pessimistic about this method, which he labels an impossible undertaking. As he points out,

it presupposes a knowledge of crop proportions of a given tithe, whereas in fact the documents—which refer to a tithe ‘translated’ into cash—do not tell us so much. We must therefore give up the attempt to work out a deflating index of this kind—it is an attractive idea but simply not practicable.6

In the Mexican case, there are other points which make such an undertaking impractical. In the bishopric of Puebla, for instance, the main sources of church income shifted during the eighteenth century from wheat-producing areas in the center of the bishopric to the cotton- and tobacco-producing areas in the eastern provinces of Córdoba, Orizaba, and Jalapa. Furthermore, in the bishopric of Michoacán the economy revolved around silver production rather than agriculture, and, for this reason alone, could not constitute an adequate indication for estimating agricultural trends. Seemingly most important of all, the small peasants in the villages, being exempted by law as “Indians,” did not pay any tithe at all. Nevertheless, Garner—referring to the work of Morin—sees no major difficulties: “Tithes provide a rough measure of the colony’s agricultural output. Landowners paid the tithes in kind, on the basis of assessments against their operations, to special collectors who sold the produce for cash.. . . Although many technical problems are associated with the preparation of tithe series, they can often be minimized or solved.”7 Garner and others have further pointed out the synchronic trend of the Poblano and Michoacán tithe curves (see Figure 1); the correlation between both series is indeed reasonably large: R = .89.

All of the known tithe series of Mexican bishoprics show a trend of increased church income, especially during the late eighteenth century. Florescano published a series of tithe data of five main bishoprics in New Spain for the period 1771-89, precisely the period of rapid acceleration (see Table 1a)8; and, in general, the correlations between series are high (see Table Ib). There is, though, one intriguing exception: the Poblano series of Florescano is basically different from one published by Medina Rubio. The correlation between both Poblano series is .53; the squared correlation R2 (0.29) amounts to 29 percent of common variation in the two series. At the same time, a correlation coefficient of R = .54 is obtained between Florescano’s Poblano and Michoacán series. The discrepancy between the data of Florescano and Medina Rubio might be clarified by a detailed examination of the original sources, but that task is beyond the scope of our study. We have chosen, in any case, to use the figures of Medina Rubio, not only because they cover almost the entire eighteenth century but because they were collected on the basis of meticulous research.

Florescano’s data could bring us to support Garner’s conclusion: this synchronous development of church income seems to be the result of the existence of a colonywide market for foodstuffs like maize, beans, wheat, sheep, and cattle.9 It seems to be confirmed by the data of the novenos reales, the state tax on the tithes, that boomed during the 1780s (see Figure 1). But, as has been stated, the development of the tithe series of the different bishoprics was also similar to that of the silver production in the viceroyalty, a phenomenon that can be explained by the process of inflation. Therefore, our main hypothesis concerns the inflationary bias of Bourbon Mexican economic development. We will try to show that New Spain suffered from a price inflation during the late eighteenth century that distorted the series of tithe receipts as well as other series—including most of the treasury data—based on the trade in basic commodity products.

The assumption of the existence of a colonywide maize market rests on a presupposed correlation of tithe series and maize prices in several bishoprics of New Spain. It led some authors to work with a computed colonywide maize price. This assumption is, however, yet to be verified.10 In the case of Central Mexico, the correlation was caused by the effects of its generally severe climate: harvest failures.11 But that is not the only point. One of the main characteristics of an integrated market is an intensive basic commodity exchange between different regions, leveling off price differences. Such a market existed in early modern Europe: wheat shortages in Barcelona, Valencia, or the Languedoc could easily be offset by wheat imports from Sicily, northern France, Sweden, Prussia, or even Russia. This exchange generally passed through Amsterdam, and correlations between the price movements of wheat in Europe are high because of the influence of Amsterdam as a central emporium. Such an integrated market could exist because of Europe’s relatively well-developed infrastructure and its favorable geographic conditions: fast and cheap water transport on rivers, canals, and by sea and good cart, car, or coach transport on well-paved, mostly level roads. The provisioning of regions in years of shortages proved to be extremely difficult only in mountainous areas like southern Germany, Switzerland, or Spain.12

All evidence shows that such a system for fast and relatively inexpensive exchange of basic commodity products did not exist in New Spain. Morin, for example, notes a wide variety of price movements inside the bishopric of Michoacán during the eighteenth century; there were numerous local markets in which prices were anarchic. Lindo Fuentes observes that “las diferencias ecológicas y los costos de transporte hacen que haya fuertes divergencias entre las diferentes regiones.” Van Young’s work on the Guadalajara region indicates a relatively late market development, which was almost totally intraregional in scope. There was virtually no introduction of basic foodstuffs from outside, even in times of severe crisis. The same was true for all highland regions in the colony, as evidenced in the published documents of the agrarian crises of 1785-86 and 1809-10: almost every cathedral chapter or main city government in highland Mexico had problems with the transport of maize from the tierra caliente. Numerous mule loads were lost in crossing the mountainous landscape. In general, even maize transport between Mexico City and the nearby Bajío was so irregular that correlations between Mexico City prices and those in the Bajío town of Silao are not higher than .50 (R2 = 0.25). The severe climate, the difficult topography, the poorly paved roads, and the lack of good waterways between the population centers impeded the formation of an integrated market system. Only through a system of forced and illegal trade at fixed prices—the repartimiento de efectos—was there a profitable interregional exchange of luxury products. The transport costs were too high for an integrated market system of commodity products to develop.13

These regional differences were mapped by Van Oss and Slicher van Bath. Although they used different criteria—from monument building to economic diversification—they were able to distinguish four regional economies of New Spain, called complexes by Slicher:

  • Central Mexico, the hinterland of the cities of Mexico, Puebla, Toluca, Cholula, Tepeaca, Cuernavaca, and Tlaxcala on the central altiplano;

  • Michoacán, the mining region of New Spain including the cities of Valladolid, Guanajuato, San Miguel el Grande, San Luis Potosí, Acámbaro, and Querétaro;

  • Oaxaca, including the Mixteca highlands;

  • Guadalajara and its hinterland, including the northern mining enclaves of Parral and Zacatecas.14

Every regional economy (or region) was characterized by concentric circles, in which the inner circles were marked by high levels of population concentration, intensive agriculture, intensive religious and architectural activity, and the like. If one moved out toward the periphery, population levels and economic activities declined. These four regional economic entities thus had variegated internal economies, and were separated by a dry and mountainous landscape that was difficult to traverse. Again, luxury products and high-valued goods could be transported and sold colony wide, but foodstuffs could not because of the high transport costs involved. A true mercado novohispano could not develop.

Any detailed, in-depth study of the economic cycle of colonial Mexico must limit itself to one of these four economic regions. The evidence now published confirms the impressive differences between the four in time, scope, and character. Central Mexico and Oaxaca were ancient indígena regions, with many small villages and intensive local market systems. These villages were geographically and economically complemented by relatively small haciendas that produced—with the use of intensive methods—basic nutriments and industrial raw materials (wood, tallow, fat, wool) for the cities. Central Mexico was the most populated region and contained a high number of non-Indians as well. Michoacán and Guadalajara were generally much more Hispanicized and did not truly develop until the mideighteenth century. In both regions, there were few pueblos de indios, but many rancheros. The haciendas were generally larger and produced cattle, food, and industrial raw materials for the cities as well as for the nearby mining enclaves. The only integrating element of the economy of New Spain, cutting across all four major regions, was the long-distance trade of silver, sheep, cattle, and repartimiento products (basically textiles like cotton cloth, jícaras, cochineal, and cattle and mules for the Indians). This trade system involved finished manufactured goods but not luxury articles imported from Europe. As described by several visitors and state officials of New Spain in the late colonial period,15 it resembled more the trans-Saharan caravan system than the European basic commodity trade.

For the discussion that follows, we limit our scope to Central Mexico, a complex of 51 provinces (see Map 1), which had the highest population density, the largest cities, and a lively internal basic commodity trade. We will give a detailed discussion of the available quantitative variables concerning the cyclical movements of the late colonial economy of this region. Because the silver was minted in the colony’s capital and caused a growth of coinage, the silver production data (of the main industry of the region of Michoacán) will be included in the analysis. In the last sections of this article, we will discuss the statistical methods used in analyzing the relationships between the variables and interpret the results.

The Time Series: A Hypothetical Discussion of Trends and Cycles

Gathering quantitative material on eighteenth-century Central Mexico is not an easy task; it is especially difficult to obtain data from different and independent sources. The general problem is that much of the empirical data is considerably less elaborated than for European economic history, and some important data are missing altogether. However, the variables of our analysis consist of 17 time series:16

  1. tithe receipts in cash of the bishopric of Puebla;

  2. tithe receipts in cash of the bishopric of Michoacán;

  3. number of tributarios (tribute payers) of the region of Central Mexico;

  4. silver production in marcos according to Von Humboldt;

  5. minting of silver coins in Mexico City;

  6. taxes on silver production;

  7. receipts of the pulque taxes of the Caja de México;

  8. receipts of the alcabala of the Caja de México;

  9. receipts of the novenos reales of the Caja de México;

  10. receipts of the tributos reales of the Caja de México;

  11. maize prices of the capital’s municipal granary;

  12. flour prices of the capital’s bakeries;

  13. quantity of flour shipments to these bakeries;

  14. number of land disputes in the Central Mexican provinces;

  15. number of registered revolts in the Central Mexican countryside;

  16. years of crisis due to harvest failures or epidemics in the Central Mexican provinces;

  17. number of monuments under construction in the Central Mexican provinces.

In order to evaluate the results of our statistical analyses properly, these variables must be subjected to careful critical evaluation. However, as we rely mainly on published sources, our criticism will be general, and we cannot go into the details of the relevance of, for example, all of the data of the Caja de México.

The first two variables have already been discussed in the previous section. The third variable, tributarios, is based on the government registers of the inhabitants of the pueblos de indios. Each pueblo de indios had its own list of tributarios and was a juridical entity with special rights. Officially, the Indians were considered the descendants of the pre-Spanish peoples of Mexico, although (as is shown by Chance, Seed, McCaa, and others) in the eighteenth century the word Indian already had, above all, a socioeconomic connotation. Some 80 percent of the inhabitants of Central Mexico were registered as tributarios living in pueblos de indios. If these Indians left their own towns and moved into Hispanic towns or cities without being transferred from one tribute list to the other, they could easily be considered mestizos or mulattos.17 Throughout the eighteenth century, great numbers of rural residents indeed moved into the cities looking for jobs as artisans or wage workers. However, it is interesting to note that the number of tributarios in the 51 provinces of Central Mexico remained relatively stable between 1740 and about 1790 (see Figure 2), while in the same period the number of mestizos and other non-Indians recorded in the cities was rapidly growing.18

The advent of rural stagnation had followed a century of immense population growth in the villages, which was brought to an end by the great matlazahuatl epidemic of 1736-39. After about 1770, emigration was slowly matched by a new population increase in the villages. Exactly how much growth occurred is still not known. Like all series, population figures for colonial Central Mexico suffer from imperfections; the tribute counts were not always made with the utmost care. By 1770, the procedure was improved, however, and it probably caused some increase in the registered number of tributarios. At the same time, a new set of regulations had included more children, aged people, and mulattos in the registered number of tribute payers than before. It is clear, even so, that especially during the last decades of the colonial period the population of Central Mexico was growing impressively, especially in villages of the Valleys of Mexico and Toluca and in mining areas like Zimapán, Temascaltepec, and Taxco. This trend is illustrated by our tributarios graph. If we combine trends, we must conclude that the increase in the number of tributarios in the countryside and the increase in the number of non-Indians in the urban world must have led to a substantial increase in demand for food, housing, and work. Around 1800, population growth was the major force behind agricultural and other economic change.19

At the same time, demographic expansion might be viewed as one of the causes of the inflationary bias of late colonial Central Mexico. In Europe, periods of population growth have been periods of rising agricultural prices. The long-term trend in cereal prices in Europe is known: upward in the thirteenth century, slowly downward in the later fourteenth century, rising remarkably in the sixteenth, but from the midseventeenth century to the mideighteenth again falling or stagnant. Prices again rose from the later eighteenth century until the midnineteenth century. These major phases have been related to periods of expanding population, in which demand outran supply.20 Unfortunately, there are no such long-term data on cereal prices in the Spanish empire in America. For Central Mexico, in fact, we could only use the maize prices of the capital’s granary, published by Florescano, and the flour prices of the capital’s bakeries, published by Suárez Argüello and García Acosta. Both series are difficult to interpret. Florescano’s series presents yearly averages, but, as he himself pointed out, only about 27 percent of these averages are based on data for all 12 months, and about 40 percent are too incomplete to be really trustworthy (since they cover less than 6 months). A closer look at the series reveals that the incomplete years give measurements mainly for preharvest months, thereby stressing the highest seasonal prices. Another problem is the free selling of maize in the capital’s street markets and shops. The general public only appealed en masse to the granary in years of harvest failures—which occurred frequently after about 1770, and not only caused boom selling but drove up prices enormously. The data thus show exaggerated fluctuations. Nevertheless, we could agree with Florescano that the series broadly reflects the consumer price of Mexico City, because many shopkeepers bought their stock in the granary. The influence of the granary’s prices outside the city remains obscure and no doubt must have been limited because of severe transport problems.

Our series of Mexico City’s flour prices is deduced from the figures of Suárez Argüello and García Acosta, whose research made clear that these were producer prices.” The flour market in Mexico City was a monopoly of about 14 wheat mills around the city, which in turn were dominated by 1 or 2 great mills. According to Artís Espríu, the Molinos Santa Mónica and Río Hondo controlled about half of the amount of flour shipped to the city’s bakeries.21 Both mills were owned by elite families who also controlled the majority of the bakeries. One of the discoveries of the “wheat project” of Suárez Argüello, Artís Espríu, and García Acosta is the overproduction of wheat by the haciendas of the Central Mexican altiplano, a finding that is confirmed by numerous hacienda sources. Since wheat was the main product of a large number of these altiplano haciendas, prices had to be kept high, which was only possible by storing the harvested wheat and by speculation with wheat stocks.22 Artís Espríu shows that in 1807, according to her a typical year, the Molino de Santa Mónica sold only 15 to 20 percent out of stock. The smallest mills, which had to sell more, nevertheless sold only about 35 to 45 percent out of stock. Wheat production was not a sound investment for the Central Mexican hacendados if they could not effectively control the wheat market. And there were few who could.

Wheat flour and bread were in general three to four times as expensive as maize. Moreover, as is shown in Figure 3, flour prices fluctuated much more than maize prices during the eighteenth century. According to Suarez Arguello and García Acosta, flour prices declined enormously in the late 1750s because of overproduction and decreasing export markets (Cuba was lost to sellers from Louisiana). However, by reducing the flour shipments between about 1740 and 1790, the millers were able to profit from the population increase and, for the most part, to maintain high prices. After about 1790, the population increase in Mexico City was so impressive that more flour could be shipped to the city’s bakeries. The maize prices show a much more stable curve, broken only by sharp increases in periods of crisis (the late 1740s, the late 1750s, and especially the 1780s, 1790s, and after 1800). The trend was slowly moving upward after the 1770s, and increasingly so around 1800 as a consequence of several harvest failures. The molinero monopoly meant an almost total control on the flour trade; there was no substantial correlation between prices and supply of flour as might be expected on a free market (flour shipments correlate R = .36 with flour Prices).

In general, if excessive demand because of population growth is not matched by expanding production, it raises the price level. For this reason, demographic expansion in Central Mexico could have been at the roots of an inflationary development. It is perhaps worth noting, too, that according to recent research this trend occurred in a period of nascent industrial activity: textiles in Puebla, Mexico City, and Querétaro; tobacco in New Spain’s major cities; and a rise of other industry supplying the repartimiento de efectos. How important the repartimiento was in the last decades of the eighteenth century is shown by Horst Pietschmann for the Valley of Puebla. He concludes that to “indios y demás clases inferiores se les repartían bienes por valor de al menos 250,000 pesos anuales y se les compraba al menos por 65 a 70,000 pesos, lo que hace un monto total de capital circulante de unos 320,000 pesos.. . .”23 Other research has suggested a similar amount of capital circulating within the Valleys of Mexico and Toluca. This was a period of what we would like to call emerging protoindustrialization, in a way comparable to that studied in David Levine’s monograph on eighteenth-century England, Slicher’s study of the Dutch province of Overijssel, or Fischer’s overview of early modern rural Germany. These and other studies show how expansion of the labor market by domestic industries broke down the traditional social control that had maintained a demographic equilibrium, in which population size was more or less kept in line with resources. They suggest that employment created population growth because of earlier marriages. It seems reasonable to assume that the population pattern in Central Mexico was likewise partly created by the growth of domestic industry.24

Protoindustrialization must have led to an increase in employment for the poor, and certainly also to an increase in the turnover of money. If it went hand in hand with a rise in the stock of money—which it did: MV increased, the amount spent on output according to the classical version of Irving Fisher’s equation of exchange—and if commodity trade (T, according to Fisher’s equation) was in the hands of relatively few people— as was also the case—we have another reason for inflationary bias. Even though it is a truism, and a simple one, the equation of exchange could be a useful approach in this respect to guide our inquiry,25 because in Mexican historiography inflation was traditionally seen as a consequence of silver production only (M up = P up). This purely monetary point of view is understandable in light of the unprecedented growth of mining production in the eighteenth century, but it does not necessarily give a full explanation for what happened.26

Little research has been done on the character and development of commodity trade and (proto)industrial output in colonial Central Mexico, and neither has the exact relationship between rural industrialization and population changes been established by research, even though we know that there is an interrelation between the two. It is of interest to note further that the mining boom accompanied population growth in the beginning of the century, and later again in the 1770s and 1780s. As has been stated by Jacobsen and Puhle, this expansion was unthinkable without demographic growth: either the necessary labor for new mining operations would not have been available or high wage levels would have made many enterprises unprofitable. The industry was in trouble in the last decades of the century because of declining exports during the Napoleonic Wars in Europe,27 yet it formed part of a pattern of economic development in much of the colony that started in the early eighteenth century and led to an inflationary bias by the end of the century. Since silver production was a powerful economic force, it must naturally be included in our analysis. As indicated above, the impressive growth of Mexican silver production (see Figure 4) is broadly discussed elsewhere and needs no further ex planation here. There is, however, some disagreement on the output of this industry in the second half of the eighteenth century. We relied on Von Humboldt’s data in marcos, the real physical output brought to the minting houses in Mexico City. Apart from this we also used archival data on the mintage of coins in Mexico City. These data refer to physical production: a certain quantity of silver marcos was received by the minting houses, recorded in their files, and later minted at about 65 to 70 reales to the marco. By royal decree alone, the value of money in Mexico declined in the eighteenth century some 4 percent. In the beginning of the century, the alloy of one silver marco was about 11 dineros and 4 granos, or 930.551 milésimos. The devaluation of 1728 reduced the value somewhat to 11 dineros or 916.666 milésimos. There was a secret devaluation in 1772 (discovered shortly thereafter) to 10 dineros and 20 granos, or 902.782 milésimos. Finally, in 1786, the value of the marco was devaluated to 10 dineros and 18 granos, or 895.837 milésimos. These devaluations must have contributed to the inflationary bias of the economy.28

If our hypothesis about inflation (prices) as a consequence of population growth (demand), expanding silver production (stock of money), and nascent protoindustrialization (industrial activity and turnover of money) is correct, other data measuring sales of commodity products must follow trends identical to those of the tithes discussed in the preceding section. The main problem for the contemporary historian is that such data do not exist. We may be brief on the relevance of the revenues of Mexico’s Real Caja, which are supposed to reflect taxation of trade and production. Using these data is hazardous, for the returns are in no sense a record of overall revenue, not even of Central Mexican revenue, but of receipts entering the treasury in Mexico City. Three main studies of the Real Caja data point to the difficulties of relying on them uncritically. In several articles, TePaske asserts that this expansion in treasury income occurring during the eighteenth century (see, for instance, Figures 5 and 10) was distorted by the imposition of forced loans on various ramos of the treasury, by the imposition of new taxes and higher tax levies, and by intensified collection methods.29 The practice of borrowing from its own ramos especially distorts the Real Caja data; by the end of the eighteenth century, the royal treasury in effect owed millions of pesos to itself. TePaske’s caution is supported by David Brading, who points to the fact that we still do not have any clear idea of which taxes, and in what amounts or for what reasons, were paid into the central treasury.30 He also warns against the danger of double counting of revenue and of confusing capital debt with annual income. It would seem, then, that the central treasury acted as both a regional receipt house and a colonywide final deposit. It could be that not all taxes found their way into its coffers, and, indeed, the establishment of provincial treasuries at Puebla, Valladolid, and Oaxaca must have influenced the receipts of the Caja Matriz after the 1780s. This blur ring of the real trade and production figures is what we propose to call the bureaucratic component of the data. It renders the treasury data almost useless for short-run analysis of economic development, unless it can be “unmasked.”

An excellent critical work was written by the Spanish historian Hernández Palomo, concentrating on the pulque revenues in Central Mexico and Oaxaca. His study stresses both the bureaucratic component of the data and their usefulness as an indicator of real growth in physical trans port of tuns of pulque to the urban markets.31 He provides case evidence of the almost single-minded interest of the Bourbon government in taxes as a source of state revenue and of bureaucratic maneuvering to increase them. Next to mining, pulque taxes (at their peak in the 1770s) stood with the alcabala as the most important of these tax revenues. The dramatic growth in revenues from pulque began in the 1760s with the change from the asiento system (a kind of farming out) to collection by royal officials, in combination with advances of the levies (see Figure 5). At the same time, real pulque output increased following new demands of the growing urban population. The decline of pulque revenue after the mid-1780s is ex plained as the result of the agrarian crises in the 1780s, 1790s, and 1800s, as well as of progressively higher taxes on pulque, which made its consumption less attractive. Of equal importance, however, was the decline in purchasing power of the urban poor after the 1770s, a consequence of the steadily rising prices.

Can this conclusion be extended to a discussion of the alcabala figures? As can be seen in Figure 5, the alcabala curve is not much different from the pulque curve; the alcabala revenues have R = .874 with those from pulque. They correlate moderately with other ramos: R = .681 with novenos reales (the government taxes on the tithes) and R = .660 with the tributos reales (the revenues of the royal tribute paid by the inhabitants of the pueblos de indios). Hernández Palomo’s findings suggest, on the one hand, that the common element in these correlation coefficients is the bureaucratic component, but at the same time that some growth of production must also be taken into account.

Just what was taxed de facto by the alcabala is not entirely clear. The alcabala was a sales tax, collected by the aduana on the roads and in the cities, and the variety of products subject to it was too large for its yield to be of much use in detailed short-run economic analysis. These products included wheat, beans, chiles, maize, and cattle produced by haciendas or village communities and sold in towns or cities, but also retail commodities (artesanía) and industrial products like textiles. Later on, a large proportion of receipts could have been derived from excise duties on European goods as well. However, a series of documents from around 1787 indicates that the alcabala may have taxed mainly the illegal repartimiento de efectos. In practice, the rule that only “Spanish” trade was to be taxed was widely broken. In the province of Chalco, even the Indian sellers on the Friday tianguiz were being taxed. (The revenues of the tianguiz stallage provided approximately 30 percent of the income from the alcabala in Chalco in the 1760s!) Besides this, there was no consistency in levies; regional variations compete with differences over time. The Mexican historian Bahena Pérez discovered that most alcabala posts were charging the 6 percent levy as late as 1791, while in years of war the levy was formally raised to 8 percent.32

According to Hernández Palomo’s experience with the pulque data, the alcabala data must contain trade information and reflect real trade development in some way. For that reason it is noteworthy to observe a decrease in the curve of the alcabala after about 1780, while the expansion of silver production was quite explosive in the same period. If we may believe Von Humboldt, the alcabala revenues would have risen, not declined, until they reached 3.2 million pesos in 1803. But, as expressed before, there are other problems. Indeed, the downturn after the 1780s peak could have resulted from the establishment of the new provincial treasuries, into which money once destined for the Caja Matriz was now paid directly. Of course, some of the funds went eventually to Mexico City, but it is not clear under what heading. The message in any case is clear: until we know more about the criteria, rules, or procedures by which this kind of revenue was deposited in the capital, it is best to be careful in interpreting its figures. We chose to use the data published by TePaske and Hernández Palomo mainly for two reasons: they are complete series covering the entire century, and we lack data of the provincial treasury of Puebla.

The growth pattern of the alcabala, tributos reales, and pulque revenues administrated by the Caja Matriz manifested itself during the 1740s and 1750s (see Figure 5). A decline began during the 1760s, but a rock bot tom was not reached: a new increase began in the same decade. Although this revival is influenced strongly by the bureaucratic component, the development mirrors the growth of population. The 1740s, 3750s, and 1760s must be considered a time of prosperity and economic expansion. We notice rising alcabala revenues, rising tributos reales revenues, and a rapid growth of silver production, accompanied by low maize prices and low flour prices. But during the 1780s, the situation must have changed impressively. Prices rose again, and government revenues like the pulque tax and tributos reales seem to have been in decline. (The lower tributary revenue, “evidence” of a massive shortfall, could also be explained as a result of the establishment of provincial treasuries. In that case we could look at Von Humboldt’s data which show a relative decline.) These were the taxes paid by the urban and rural poor. Agrarian crises due to over population, epidemics, deviant weather conditions, and harvest failures limited the potential of the poor to pay tribute to the crown, or to consume much pulque. This occurred in years of expanding silver production and war.33

There is now growing support for the conclusion that the late eighteenth century was a period of “agrarian compression” (Tutino), causing “dearth and dislocation” (Hamnett), in which “the rich got richer and the poor got skewed” (Van Young). The decreasing purchasing power of the poor after the 1770s-80s is confirmed by the other variables of our analysis. These also reveal a period of crises in the first decades of the eighteenth century, culminating in the great epidemic of 1736-39. At first sight, two variables stand out: the number of peasant revolts and the number of land disputes (see Figure 6). The number of peasant revolts in Central Mexico was low in the middle of the century, but it increased rapidly during the 1780s. Taylor stresses the economic and political causes of these revolts: they followed a period of subsistence crises, as well as changes in the socioeconomic and political organization of the countryside, whereby the relationship between the rulers and the ruled had been changed. This is a topic for separate research and must be left aside here, but all research on the moral economy of the poor shows a similar combination of declining living standards and cultural and political change leading to popular protest.34 The number of revolts thus indicates, indirectly, changes in the economic cycle.

In an area like Central Mexico, the scarcity of suitable land for cultivation was an important aspect of agriculture. Tutino had the brilliant idea of using modern archival catalogues to make an index of agrarian tensions by counting the number of land-related disputes originating in the Valley of Mexico and Toluca brought before the Audiencia in Mexico City.35 We copied the method, but added the disputes in the provinces now belonging to the states of Puebla and Hidalgo. These disputes almost without exception involved pueblos de indios: suits between estates and villages and between villages and villages. The suits between residents of single villages were left out of our analysis, because they involved mainly local power politics (these suits usually included caciques).36 Our series gives an indication of the increasing scarcity of cultivated lands in the pueblos de indios relative to their growing populations. As Figure 6 illustrates, the frequency of land-related suits was high during the 1710s (a period of intensive composiciones de tierras, or legalizations of land titles by hacendados and villagers alike), low in the “prosperous” midcentury decades, and high again from the 1760s onward. Figure 7 also shows that the late eighteenth-century increase ran parallel to population growth, whereas the number of land disputes was low in the midcentury period of stagnating tributario numbers. After the epidemics of 1736-39 there was apparently no shortage of land in the pueblos de indios until the 1760s.

Apart from the tributos reales, the number of rural revolts, and the land disputes (we shall call these variables “agrarian”), a fourth indicator of the purchasing power of the poor may be introduced: the number of monuments under construction in Central Mexico. As has been shown by van Oss and Slicher van Bath, this variable can be a powerful indicator of economic development.37 The variable contains a selection of collected data on the chronology and distribution of architectural activity compiled by van Oss.38 The construction of monuments like chapels, churches, and monasteries can tell the economic historian much about economic cycles of the period in which their construction began, for many of them represent huge material investments, and they therefore reflect the economic climate of the time. The great majority of van Oss’s data was located in our Central Mexican region of 51 provinces (see the differences between the number of monuments of van Oss’s inventory in the entire colony of New Spain and those in our region of Central Mexico in Figure 8). Until the 1770s, monument building followed the pattern of population growth, disrupted by a slight decrease only during the matlazahuatl epidemic.

The level of monument-building activity fell steeply during the 1770s, and it hit bottom around 1800. This coincides remarkably with the “agrarian” variables mentioned above, suggesting that monument building may have been financed by the generosity of the majority of the population, including the poor.

The conclusion just put forward can be illustrated by comparing the curves of monument-building activity and land disputes (see Figure 9), as well as the curves of pulque revenues and tributes reales (see Figure 10). The number of land disputes has a moderate negative correlation with the number of monuments under construction: increased monument-building activity coincided with a decreased number of land disputes. Interesting is the high building activity in the 1750s following the epidemics. The number of land disputes fell after the peak of the early 1800s, when the villagers could simply no longer afford to press them. Indeed, the period of greatest crisis must have been around 1800, although the turning point in purchasing power of the masses probably came in the early 1780s. The year 1786 was the one and only year of starvation in the eighteenth century. However, there were about 15 maize harvest failures between 1770 and 1800 because of droughts and night frosts, underscoring the climatic component of our series.39 A process of pauperization during the 1790s and 1800s is expressed, then, by the variables tributos reales, monument building in Central Mexico, and partly by pulque revenues.

To summarize this section, there are five provisional conclusions:

  1. population growth can be identified by combining increased mestizaje (growing number of non-Indians) and a growing number of tributarios in the second half of the eighteenth century, following a setback during the great epidemics of 1736-39;

  2. this expansion caused growing demand for foodstuffs and retail commodities, thereby stimulating agriculture, protoindustrial activity, and silver production;

  3. price movements were stable until the 1780s—namely during the advent of protoindustrialization;

  4. continuing population growth and expanding silver production in combination with harvest failures, agrarian crises, and subsequent demand outgrowing supply brought an end to the midcentury prosperity;

  5. a process of pauperization finally set in during the 1790s.

None of the 17 variables by itself clearly illustrates this sequence of developments. We need other methods to unmask the bureaucratic component and to prove the inflationary bias of the late colonial economy. This brings us back to the tithe data and to our hypothesis that these do not reflect agrarian production but inflation.

We have classified the variables themselves into:

  • inflationary variables (tributarios, silver production in marcos according to Von Humboldt, the occurrence of crisis years due to harvest failures or epidemics, and in a way also the tithe receipts of the bishoprics of Michoacán and Puebla, although we prefer to keep the tithe data apart as a separate group of variables; variables 1, 2, 3, 4, and 16),

  • bureaucratic variables (the Real Caja data; variables 7, 8, 9, and 10), and

  • purchasing-power variables (maize prices, flour prices, number of land disputes, number of registered revolts, and number of monuments under construction in the Central Mexican provinces; variables 11, 12, 14, 15. 17).

Variables 5 and 6 (minting of silver coins and taxes on silver production) are exchangeable with variable 4 (silver production in marcos); variable 13 (flour shipments to Mexico City) can, as indicated before, be related to variable 12 (flour prices). So far, however, our reasoning has been speculative, and as a next step it must be checked by the use of appropriate statistical analyses. The following section will describe how two of these

Statistical Analysis of the Data

In the foregoing discussion we contested the common interpretation and use of the tithe receipts as an indicator of economic development in agrarian colonial Mexico. We suggested that the tithe receipts reflected on the contrary inflationary and bureaucratic developments. It is remarkable that amid all the elaborate theoretical reasoning that has been applied in favor of the first assumption no effort at empirical verification or falsification has been made as yet. We have therefore examined the relationship between the tithe variables and inflationary, bureaucratic, and purchasing power variables using canonical correlation analysis as well as one- and two-dimensional linear dynamic systems analyses. The interpretation of the latter will be done using a graphic representation.

First, we shall discuss the results from the canonical correlation analysis.40 Canonical correlation analysis is a linear multivariate analysis technique that computes the strength and nature of the relationship between two sets of variables. In these so-called first and second sets, the respective variables are by weighted addition combined into one new, “communal” variable, the so-called canonical variable. The correlation coefficient between the canonical variables of both sets is computed just as if they were ordinary variables; this coefficient is called the canonical correlation, indicating the strength of the relationship between the two sets. The weighted addition of the respective variables into canonical variables is done in such a way that the canonical correlation between the two sets is maximized. The contribution of the respective variables in relating the two sets is reflected in the correlations that the variables have with the canonical variable of their set. These latter correlations give the nature of the relationship between the two sets.

A setup to test for the relation of the tithe variables with the “bureaucratic,” “inflationary,” and “agrarian” variables is presented in Figure 11. If it were the case that the tithe receipts in cash did reflect agrarian development, we would find high correlations of “agrarian” variables such as the maize or flour prices, the number of land disputes, and the number of revolts in the Central Mexican countryside with the canonical variable of their set. If it were the case, however, that the tithe receipts did not reflect agrarian but rather inflationary and bureaucratic development, we would find high correlations of the inflationary and bureaucratic variables and low correlations of the “agrarian” variables with the canonical variable. These solutions are the hypothetical extremes.

Table II shows the results of this analysis. The canonical correlation between the two sets is quite high: .98. Both tithe series correlate strongly with the canonical variable, as do the variables tributarios, pulque revenues, silver production, and two of the bureaucratic variables, alcabala and novenos reales. Tributes reales and years of crisis perform in an un expectedly meager way, but as far as the tributes are concerned this is in line with their divergent course observed earlier in Figure 5: tributes reales decreased after 1770, while other bureaucratic variables increased sharply. The purchasing-power variables all have correlations around .5, which are markedly lower. This supports our postulation about the relationship between the tithe data in cash and the inflationary and bureaucratic variables of our data. It brings us to the first conclusion: the tithe series appear to be indicators of inflationary and bureaucratic rather than of agrarian development.

As we are dealing with time-series data, it makes sense to suppose that the variables years of crisis and tributos reales might perform better in an analysis in which time is modeled. For that purpose, we continue our examination of the relationships between the same groups of variables with linear dynamic systems analysis.41 Incorporation of the time dependency between measurements enables us to make inferences about the causal relationships with more confidence. Unlike canonical correlation analysis, where the two sets of variables are treated symmetrically, linear dynamic systems analysis distinguishes between exogenous and endogenous variables; in the linear dynamic model, the exogenous and endogenous variables are related through time. The variables and the relations between them constitute the linear dynamic system. The influence of past occurrences on future occurrences passes through what is best thought of as the “memory” of the system, often called the “state” of the system. Like canonical correlation analysis, linear dynamic systems analysis re turns correlations of the respective variables with the memory or state of the system that indicate the nature of the relationship between the exogenous and endogenous variables. Also, a measure comparable to the canonical correlation is presented—the fit—that reflects the strength of the relation between exogenous and endogenous variables.

In Figure 12 a schematic representation is given of linear dynamic systems analysis of the tithe, inflationary, bureaucratic, and purchasing power variables. Table III gives the correlations of all variables with the state; the fıt of the solution was .989. In comparison with the canonical correlation analysis solution, the correlations have not changed markedly. Thus, the correlations of the purchasing-power variables are still among the lowest, and also the performances of the tributes reales and years of crisis have hardly changed. The interpretation of the linear dynamic systems analysis solution is in fact identical to that of the canonical correlation analysis solution. Incorporation of the time dependency thus supports the conclusion made on the basis of the simple cross-sectional case.

In contrast to the other variables, the correlation of tributes reales is rather low. This suggests the desirability of a further step, based on the possibility of relating the exogenous and endogenous variables in more than one way. Variables can relate in several, distinctive manners; thus an explanation for the low correlation of the tributes reales might be that the other variables are related to one aspect of the tithe data, while the tributes reales are related to another aspect of these same tithe data. Such a differential relationship can be revealed by letting the state have more than one dimension; each dimension, independent of the former, brings out another qualitatively different relationship between the variables. By finding the common aspect of the variables that have high correlations with the respective dimensions of the state, these dimensions can be interpreted. We therefore ran linear dynamic systems analysis allowing for a second dimension of the state; the results of this analysis, which has a fit of .992, are given in Table IV. From the correlations we conclude that neither of the tithe variables has been related satisfactorily to the second dimension of the state.

The two-dimensional linear dynamic systems analysis solution is best explored using a graphic representation of the relationships between the variables, an image of the solution. The graphic representation is made by taking the first two dimensions of the state as the x-axis and y-axis, and the correlations of the variables with these dimensions of the state as scores on both axes. The variables are then represented as arrows or vectors in a two-dimensional space. The importance of a variable can be inferred from the length of its vector in the figure; the higher the correlations of a variable, the longer its vector. Because of our interest in their interrelatedness, we will focus on the relative positioning of the variables. In case the vectors of two variables point in opposite directions, a high score on one variable implies a low score on the other. If they stand at right angles, the variables are thought to be independent of each other: in the latter case, there is no connection between the scores on one variable and the scores on another. The clustering of vectors in approximately the same spot indicates that they measure the same phenomenon, but variables whose vectors point in the opposite direction are said to measure the opposite.

As expected, in Figure 13 the tithe vectors are almost covering each other, which underscores the identical role of both tithe variables in our analysis. Of all other arrows, the one nearest to the tithe vectors is silver production; in fact, silver production and the tithes form a cluster of highly correlating, interrelated variables. The only agrarian variable, or purchasing-power variable, that points in the same direction is the one of registered revolts in the Central Mexican countryside, but then its arrow is rather short (coordinates .519 and −.066). The same applies to the “years of crisis” variable. In the bottom right of the figure, the agrarian variables land disputes, maize prices, and flour prices form a second cluster of highly interrelated variables (together with the inflationary variable tributarios). Contrary to the tithes/silver cluster, this cluster correlates just as strongly with the first as with the second dimension of the state.

Two variables stand out from these clusters. As would be expected, we find the tributos reales at an odd-one-out position in the figure, at nearly a right angle from both clusters. This right angle indicates that the tributos reales variable follows an independent course from the variables united in both clusters. The different orientation with respect to other bureaucratic variables like pulque revenues or novenos reales revenues indicates that the series of the tributos reales was influenced more by other factors than by the bureaucratic component. Tributos reales is the variable that “loads” most strongly on the second dimension. The other striking divergence is the position of the monument-building variable, which is reversely related to the agrarian/tributarios cluster. This means that in periods of decreasing maize and flour prices, waning land disputes, and decreasing number of tributarios, the number of monuments under construction increased. In other words, the inflationary development of rising prices, rising tensions in the countryside, and a clear overpopulation in the pueblos de indios of colonial Mexico reduced the possibility of, in this case, the rural poor to contribute to church building. This is a remarkable conclusion for a period of strong religious fervor. The bureaucratic variable alcabala also has some positive correlation with the second dimension, giving evidence that a part of it reflects the pauperization posited by our thesis. Flour prices and tributarios, however, have negative correlations with the second dimension.

The overpopulation, epidemics, and harvest failures after about 1770-80 reduced the contributions of the poor not only to church building, but to tribute paying as well. It can be documented that the state could not or did not want to force the inhabitants of the pueblos de indios to deliver their tribute, fearing to lose the legitimacy of its authority in the communities.

The inflationary bias of the late eighteenth-century economy was further explored through a linear dynamic systems analysis with the inflationary variables as exogenous and the bureaucratic and purchasing-power variables as endogenous; the bureaucratic component was reduced by excluding not only the tithe variables but the alcabala and the novenos reales as well. The fit of the solution was .795 (see Table V). The structure of our solution—that is, the relative positioning of the variables’ arrows—has changed minimally; compare Figure 14. (In fact, for reasons of comparison, we have rotated the structure approximately 65 degrees clockwise; there is no difference in interpretation between the solutions in both figures.) The “purchasing-power” aspect of the tributos reales is now stressed. Four variables—land disputes, tributarios, flour prices, and maize prices—again form a distinct cluster. The variable tributarios is most strongly related to the purchasing-power variables in the cluster and opposed to monument building. The monetary, inflationary variable silver production is nearest to the bureaucratic variable pulque revenues. At this point, it underscores the bureaucratic component in the gathering of taxes. Still remarkable is the absence of high correlations of the number of years of crisis with either of the state dimensions. What accounts for this anomaly is not clear.

As indicated above, the interpretation of the dimensions of the state space analysis is given by the common aspect of variables that “load” on the dimensions. If the variables that have high correlations with a dimension all have one aspect in common or measure the same phenomenon, that dimension can then be said to represent that aspect. In Figure 14, the first dimension has a strong relationship with the pulque revenues and silver production, but a much weaker relationship with the number of revolts in the Central Mexican countryside and of the years of crisis. There is a weaker relationship with land disputes, tributarios, flour prices and maize prices, but these variables correlate also—and in some cases more highly—with the second dimension. We identified the production side of the economy as the common element here, and it is tempting to label the x-axis, a bit adventurously, as the dimension of production. The second dimension correlates most strongly with the tributos reales and monument-building variables; flour prices and tributaries play an opposite and less prominent role. The common element seems to be consumption; the y-axis could thus be named—in an equally adventuresome way—the dimension of consumption.

From the foregoing analysis, a tentative picture of the dynamics of the economy of Central Mexico between 1710 and 1809 emerges. Years of high levels of pulque revenues and silver production were usually also years of peasant revolts and land disputes. Often these were years of crisis, harvest failures, and higher flour and maize prices, and generally speaking, they occurred in periods of expanding population pressure—of relative deprivation of the poor. The years of sharply declining purchasing power of the poor are situated in the lower right part of the figure of our linear dynamic systems analysis solutions. Years of relative prosperity score in the upper left corner of the figure. These years are characterized by intensive monument building, relatively few land disputes, substantial yield of the tributos reales, and low maize prices. The years of prosperity all occurred in the middle period of the eighteenth century, when the number of tributarios had not yet risen to its late eighteenth century peak.

Summary and Prospects

In this article, only some aspects of the Central Mexican economy in the eighteenth century could be discussed. Working in a science of con text, our main effort was directed toward relating the meaning of series of tithe data to a set of other series. We try to underscore one of the basic virtues of our discipline: critical evaluation. In colonial Mexican historiography an undertaking of quantitative research is not an easy task. Too much remains hidden in the archives: the precise population development; the rhythm of the prices of basic commodity products; the criteria or procedures by which state revenues were deposited; the development of wages in different cities, towns, and villages; etc. Above all, the agrarian development remains largely obscure, despite some excellent monographs on the social morphology of the region. We lack information on changes in resource management, in cultivation and the crop system, or in the results of the harvests during the colonial period. At the same time, the spread of protoindustrialization is not known, although some tentative articles have appeared (by Pietschmann, Dehouve, and Thomson). What we have shown is that some of the published data contain a previously almost unrecognized bureaucratic component; apart from expressing treasury in come they also represent government policy to borrow from its own cajas, open up new cajas in the provinces, deposit money from elsewhere, and the like.

Economic historians never cease to thirst for data, as Lindo Fuentes wrote in introducing his review of the study of tithe data by Pastor et al. Serial data are hard to obtain, especially on agriculture. In European historiography, historians can, to a certain extent, rely on tithe data in kind. The Mexican tithe data, however, are generally in cash, and we have demonstrated that these are unsuitable for use in “agrarian conjuncture.” What alternatives exist? The analysis of changes in yield ratios, so profitable in European agrarian history, is also not without problems in the Mexican case, because intensification of agricultural systems during the colonial period would hardly have been possible, as a consequence of thin soils, a dry climate, and the overproduction of wheat. Besides this, as Wilken showed, the agricultural system of colonial Mexico already must have had a high level of intensification. Thus, even a method such as that used by Slicher van Bath in analyzing European yield ratios will probably not be very fruitful in Mexican agrarian history. Mark Overton suggested—for East Anglia, in Britain—another method: computer analysis of probate in ventories. But this method would mean a very time-consuming gathering of data in several thousands of files of hacendado and cacique properties in the Ramo de Tierras of Mexico’s national archive.42

However, the 17 variables of our critique can be used for some generalizations on eighteenth-century Central Mexican developments; for instance, on the pauperization of the urban and rural poor at the turn of the century, a pauperization that must have followed several decades of midcentury prosperity. The evidence generally points toward serious problems in the economy of Central Mexico in the late eighteenth century, and we were able to explore the complex relationship that exists between some variables using multivariate analyses. On the one hand, this revealed the relative importance of silver production in the Michoacán region for the Central Mexican economy, related to its inflationary bias. But, on the other hand, one of our main conclusions must be that silver production was not the only variable related to this bias: population growth pushed prices up as commodity production in the end could not match demand. This was, no doubt, due to the reaching of a ceiling on the kind of expansion the region had been experiencing, causing a demand-driven inflation, of which the tithe series are thermometer-like indicators. Although much research must still be done, this conclusion seems to be supported by the results of our multivariate analyses. So, at a time of growing disillusionment with quantitative data like those of the Real Caja of Mexico or the tithe receipts in cash, historians, known for their critical reading of sources, can use modern statistical techniques for a critical evaluation of quantitative data. The use of these techniques promises to give new perspectives for quantitative, exploratory historical research.

Seventeen Indicators of Economic Development in the 51 Provinces of Central Mexico; 1710-1809* (five-year moving averages on the basis of the data in the sources; missing values were estiamted)


For description or these variables see text.

Canonical Correlation Analysis and Linear Dynamic Systems Analysis

It seems appropriate to set out clearly the possibilities and limitations of any statistical analysis of historical data. Scientists have always been interested to uncover the causes of the phenomena they observe. They formulate physical laws that explain their observations, or, if that is impossible, they conduct experiments to find relationships between phenomena, gathering empirical data to use in theory building. In an experiment, the researcher controls and manipulates the (usually random) distribution of subjects or observation units to achieve experimental conditions. On that ground, the researcher can prove that a treatment here caused a result elsewhere. The big stumbling block in the analysis of historical data is that it is, of course, impossible to conduct experiments. As in other observational research, the historian has to be satisfied with the registration of events and the computation of correlation coefficients between phenomena. As correlation coefficients are an index of simultaneous occurring, and of nothing more than that, they do not justify conclusions other than: “The occurrence of phenomenon A is related to the occurrence of phenomenon B.” In a correlation coefficient, the relationship between two events has no direction. In other words, the correlation between two events may be the result of event A causing B, event B causing A, or a third event C causing both A and B. All three mechanisms are equally likely, and it is impossible to uncover statistically which of the three is true. Of course, when there is no correlation between two events A and B, one may confidently infer that A does not cause B, and vice versa. Thus, for testing hypotheses about causality, simple correlation coefficients are an inadequate tool.

Also, we are not interested in numerous specific relations (17 variables produce 136 correlation coefficients!); we would rather investigate the “general” relations in our data. For instance, the inflationary variables tributarios, years of crisis, and silver production are not of interest to us by themselves, but for their relation to something we could not measure directly: inflation. The inflationary variables are considered indicators of this phenomenon, all three reflecting various aspects of inflation. Thus, it would not be informative to know the correlation between, for example, one of the tithe variables and the tributarios variable. Instead, what we do want to know is the strength of the relationship between, e.g., both tithe variables and the group of inflation variables, or the strength of the relationships between both tithe variables and the group of bureaucratic variables. For this we need a technique that investigates the relations between groups or sets of variables.

As set out in the text, canonical correlation analysis is a linear multivariate analysis technique that gives the strength (reflected by the canonical correlation) and nature (reflected by the correlations of the variables with the canonical variable of their set) of the relationship between two sets of variables. By weighted addition the original variables have been combined into one communal canonical variable for each set; the canonical correlation is the ordinary correlation coefficient between the canonical variables. The weighted addition of the respective variables into canonical variables is done in such a way that the canonical correlation between the two sets is maximized. If, for instance, of the variables in the first set one particular variable is strongly related to the variables in the second set, it is advantageous to let this particular variable play an ample role in the weighted construction of the canonical variable of the first set. The contribution of the respective variables in relating the two sets is thus reflected in the correlations that the variables have with the canonical variable of their set. In Figure 11 a schematic representation was given of canonical correlation analysis of the tithe receipts and the inflationary, bureaucratic, and purchasing-power variables. Note that, as with the correlation coefficient, in canonical correlation analysis the relation between the two sets has no direction; switching the first and second sets has no effect on the solution.

However, simple correlation coefficients, as well as canonical correlation analysis, do not incorporate the time character of our time-series data. Like most statistical analysis tools, the former two have been designed for the analysis of observation units that are assumed to be independent replications of the same basic structure. This assumption is not tenable for our data; it is in fact quite likely that, for instance, the flour price of a certain year is related to the flour price of the former year. In the case of time-related data, the observation points are dependent in time and cannot be assumed to be independently randomly distributed. The fact that two time series T and U are correlated is, as in the independent case, in itself no indication of a causal relationship between them. In addition, even when there is no correlation between the two time series T and U, there could still exist a causal relationship between them. If, for instance, event T causes event U after one month, we will never uncover this causal relationship using data that are on a yearly basis only. In short, if we wish to examine the causal relationship of population pressure, inflation, or other forces acting on Central Mexican economic development, we will have to use a technique that accommodates the time structure of the data. The technique we propose to use for this purpose is called state space or linear dynamic systems analysis. Unlike canonical correlation analysis, where the two sets of variables are treated symmetrically, linear dynamic systems analysis distinguishes between exogenous and endogenous variables (also sometimes called the input and output variables) that are related through time. The variables and the relations between them constitute the linear dynamic system.

The linear dynamic system is specified more formally as follows. If we have measurements from time 1 until time T, then the linear dynamic model states that the exogenous variables at time t influence the endogenous at time t. Furthermore, past occurrences influence future occurrences through a postulated latent variable, named state or latent state. Given T measurements on a number of exogenous variables X (X = {x1, . . ., xT}) and on a number of endogenous variables Y (Y = {1,. . ., yT}), and given the latent states Z (Z = {z1,. . ., zT}), the linear dynamic model is specified as:

zt=Fzt-1+Gxt(system equation)yt=Hzt(measurement equation),

with F, G, and H transition matrices specifying the respective influences of past states on current states, of exogenous variables on the states, and of the states on the endogenous variables. The latent states Z function as a memory of the system, as all influence of past occurrences on the present state zt is contained in the past state zt−1. Also, the states function as a filter, because all influence of the exogenous variables on the endogenous variables “passes” through the latent states. A schematic representation of linear dynamic systems analysis of the inflationary, bureaucratic, and purchasing-power variables versus the tithe variables was given in Figure 12.

Contrary to ordinary correlation coefficients and canonical correlation analysis, linear dynamic systems analysis treats exogenous and endogenous variables differently. This can also be seen by comparing Figures 11 and 12. In Figure 11, the variables in the upper part of the diagram may be exchanged with those in the bottom part; this has no effect on the solution. But if we switch the variables in Figure 12 in the same way, we analyze a fundamentally different model: the direction of the arrows specifies in that case a model in which the tithe variables are postulated to cause the number of tributarios, years of crisis, etc.!

The linear dynamic systems analysis solution is interpreted using the correlations of the respective variables with the latent states; as in canonical correlation analysis, not the weights (in this case the transition matrices F, G, and H) but the correlations of the original variables with the constructed variable serve to interpret the solution. A measure of the strength of the relation between exogenous and endogenous variables is given by the so-called fıt of the solution. Several latent states z may be modeled to accommodate differential relations between variables; each so-called dimension of the state embodies another qualitatively different relationship between the variables. By finding the common aspect of the variables that have high correlations with the respective dimensions of the state, these dimensions may be labeled and interpreted. However, while canonical correlation analysis is performed by a straightforward computational procedure that is available in software packages like SPSS-X and SAS, linear dynamic systems analysis is a newly developed technique which requires a more complicated estimation procedure, for which the computer program DYNAMALS has been developed recently.43

To sum up, linear dynamic systems analysis has two advantages over canonical correlation analysis: it incorporates the time relation between measurements; and it provides imbalanced treatment of exogenous and endogenous variables. Although a strong relationship between two phenomena appearing from linear dynamic systems analysis is still not irrefutable proof of a causal relationship, causality is, for both reasons, more likely than in the case of canonical correlation analyses.


John J. TePaske et al., La Real Hacienda de Nueva España. La Real Caja de México (1576-1816) (Mexico City, 1976); TePaske and Herbert S. Klein, Royal Treasuries of the Spanish Empire in America, 3 vols. (Durham, 1982) and Ingresos y egresos de la Real Hacienda en México, 3 vols., forthcoming. Other works include TePaske, “General Tendencies and Secular Trends in the Economies of Mexico and Peru, 1750-1810: The View from the Cajas of Mexico and Lima,” in The Economies of Mexico and Peru During the Late Colonial Period, 1760-1810, Nils Jacobsen and Hans-Jürgen Puhle, eds. (Berlin, 1986), 316-339; David A. Brading, “Facts and Figments in Bourbon Mexico,” Bulletin of Latin American Research (hereafter BLAR), 4:1 (1985), 61-64; Richard Garner, “Silver Production and Entrepreneurial Structure in Eighteenth-Century Mexico,” Jahrbuch für Geschichte von Staat, Wirtschaft und Gesellschaft Lateinamerikas (hereafter JbLA), 17 (1980), 157-186, “Exportaciones de circulante en el siglo XVIII (1750-1810),” Historia Mexicana (hereafter HMex), 31:124 (1982), 544-598, “Further Consideration of ‘Facts and Figments in Bourbon Mexico’,” BLAR, 6:1 (1987), 55-63; John Coatsworth, “The Limits of Colonial Absolutism: The State in Eighteenth-Century Mexico,” in Essays in the Political, Economic and Social History of Colonial Latin America, Karen Spalding, ed. (Newark, 1982), 25-51 and “The Mexican Mining Industry in the Eighteenth Century,” Economies of Mexico and Peru, 26-45; Klein, “La economía de la Nueva España, 1680-1809: Un análisis a partir de las Cajas Reales,” HMex, 34:136 (1985), 561-609; B. H. Slicher van Bath, Real Hacienda y economía en Hispanoamérica, 1541-1820 (Amsterdam, 1989).


Héctor Lindo Fuentes, “La utilidad de los diezmos como fuente para la historia económica,” HMex, 30 (1980/118), 273-289 (quoted passage pp. 287-288). See also Claude Morin, Michoacán en la Nueva España del siglo XVIII: Crecimiento y desigualdad en una economía colonial (Mexico City, 1979); Arístides Medina Rubio, La iglesia y la producción agrícola en Puebla, 1540-1795 (Mexico City, 1983); Juan Carlos Garavaglia and Juan Carlos Grosso, “La región de Puebla-Tlaxcala y la economía novohispana (1670-1821),” HMex, 35:140 (1986), 549-600, “De Veracruz a Durango: Un análisis regional de la Nueva España borbónica,” Siglo XIX, 2:4 (1987), 9-52, and “Estado borbónico y presión fiscal en la Nueva España, 1750-1821,” in America Latina dallo stato coloniale allo stato nazione, Antonio Annino et al., eds. (Milan, 1987), 78-97; Rodolfo Pastor et al., Fluctuaciones económicas en Oaxaca durante el siglo XVIII (Mexico City, 1979); Cecilia Rabell Romero, Los diezmos de San Luis de la Paz: Economía de una región del Bajío en el siglo XVIII (Mexico City, 1986); Alexander Von Humboldt, Ensayo político sobre el reino de la Nueva España (Mexico City, 1822; reprinted 1978), 316-317.


Emmanuel LeRoy Ladurie, The Peasants of Languedoc (Urbana, 1974), passim.


This section follows Joseph Goy’s excellent introduction, “Methodology,” LeRoy Ladurie and Goy, Tithe and Agrarian History from the Fourteenth to the Nineteenth Centuries. An Essay in Comparative History (Cambridge, 1982), 3-70, based on the articles in Prestations paysannes, dîmes, rente foncière et mouvement de la production agricole à l’époque préindustrielle, 2 vols., LeRoy Ladurie and Goy, eds. (Paris, 1982).


Goy, “Methodology,” 15.


Ibid., 44.


See Garner, “Price Trends in Eighteenth-Century Mexico,” HAHR, 65:2 (May 1985), 279-325 (quoted passage p. 319); Morin, Michoacán, 102-121, esp. 116; and Medina Rubio, Iglesia, passim. Although Morin does acknowledge that an increase in the tithe receipts in cash resulted from more rigorous collection in combination with demographic and economic growth, it may not account for the major part of the increase. In his view, later backed by Garner, economic, agricultural, and population growth in the bishopric helped to generate the major part of the revenue increase. Garner, on his part, expressed some doubts on the growth rates. Comparing the Michoacán tithe figures with Bajío maize prices (also Michoacán) in the eighteenth century, he found a similar rhythm of growth of both series until the 1780s but noticed a strange price development afterward (without establishing a clear trend), while the tithe series continued to grow. It is our view that the tithe series and the price series are strongly influenced by the same factor: inflation. This questions not the data for the late colonial development, but the similar rhythm until the 1780s and therefore the tithe data themselves. The diezmo files in the Archivo General de Indias (hereafter AGI) show how difficult it is to transform these tithe data into workable agrarian data (see AGI, México, leg. 2576, bishopric of Puebla). The accounts are impressively heterogeneous. Some of the lessors send their receipts of different contracts all at once to Mexico City or Puebla; others promised to pay but died before they could do so, leaving their family in debt to the church. There were different contracts for wheat, seeds, chiles from Tepeaca, chiles and menudencias from Cholula and Texmelucan (two different tithe jurisdictions!), etc. Cholula paid tithes on the income from its propios. Some of the great hacendados agreed to pay a fixed amount only por convenio antiguo. The Indians contributed mainly with money. At the same time, there were lessors and government agents collecting the tithes; the latter had special barns at their disposal to store the products (e.g., Texmelucan and Huamantla). The collector of Texmelucan visited the haciendas in his jurisdiction personally but irregularly, while the one in Huamantla mainly collected the tithes of the former Jesuit haciendas in the provinces of San Juan de los Llanos, Tepeaca, and Tlaxcala.


Enrique Florescano, Origen y desarrollo de los problemas agrarios de México, 1500-1821 (Mexico City, 1976), 69.


Lindo Fuentes, “Utilidad,” 276-277.


Eric Van Young, “The Rich Get Richer and the Poor Get Skewed: Real Wages and Popular Living Standards in Late Colonial Mexico,” paper given at the meeting of the All-UC Group in Economic History, Los Angeles, 1987, p. 51, n. 30. See also the works of Garner and Coatsworth, n. 1 above.


On the climate of Central Mexico, see Florescano, Precios del maíz y crisis agrícolas en México (1708-1810) (Mexico City, 1969); John M. Tutino, “Creole Mexico: Spanish Elites, Haciendas and Indian Towns, 1750-1810” (Ph.D. diss., University of Texas, Austin, 1976) and From Insurrection to Revolution in Mexico: Social Bases of Agrarian Violence, 1750-1940 (Princeton, 1986); and Arij Ouweneel, Onderbroken groei in Anáhuac: De ecologische achtergrond van ontwikkeling en armoede op het platteland van Centraal-Mexico, 1730-1810 (Amsterdam, 1989), 47-69.


Slicher van Bath, De agrarische geschiedenis van West-Europa, 500-1850 (Utrecht, 1960; reprinted 1977); Jan de Vries, The Economy of Europe in an Age of Crisis, 1600-1750 (Cambridge, 1976); Pierre Vilar, La Catalogue dans l’Espagne moderne (Paris, 1962); David R. Ringrose, “Transportation and Economic Stagnation in Eighteenth-Century Castile,” Journal of Economic History (hereafter JEH), 28 (1968), 51-79, and “Madrid and the Castilian Economy,” Journal of European Economic History (hereafter JEEH), 10 (1981), 481-490; David E. Vassberg, Land and Society in Golden Age Castile (Cambridge, 1984).


Lindo Fuentes, “Utilidad,” 276. Brading, Haciendas and Ranchos in the Mexican Bajío: León 1700-1860 (Cambridge, 1978); Van Young, “Urban Market and Hinterland: Guadalajara and Its Region in the Eighteenth Century,” HAHR, 59:4 (Nov. 1979), 593-635; Fuentes para la historia de la crisis agrícola de 1785-1786, 2 vols., Florescano and Pastor, eds. (Mexico City, 1981); Morin, Michoacán, 195-200; Fuentes para la historia de la crisis agrícola (1809-1811), Florescano and Victoria San Vicente, eds. (Mexico City, 1985). On the repartimiento de efectos in New Spain, see Ouweneel, Onderbroken groei, 122-149; Brian Hamnett, Politics and Trade in Southern Mexico, 1750-1821 (Cambridge, 1971) and Roots of Insurgency: Mexican Regions, 1750-1824 (Cambridge, 1986), 26, 78, 149, 152, 170; Horst Pietschmann, “Der Repartimiento-Handel der Distriktsbeamten im Raum Puebla im 18. Jahrhundert,” JbLA, 10 (1973), 236-250 and “Agricultura e industria rural indígena en el México de la segunda mitad del siglo XVIII,” in Empresarios, indios y estado: Perfil de la economía mexicana (siglo XVIII), Ouweneel and María Cristina Torales Pacheco, eds. (Amsterdam, 1988), 72-87; Danièle Dehouve, “El pueblo de indios y el mercado: Tlapa en el siglo XVIII,” Empresarios, indios y estado, 88-104; Thomas Gerst, Die wirtschaftliche Entwicklung Mexikos und das Problem der Proto-Industrialisierung am Ausgang der Kolonialzeit (Munich, 1988), 39-54; Torales Pacheco et al., La compañía de comercio de Francisco Ignacio de Yraeta (1767-1797): Cinco ensayos, 2 vols. (Mexico City, 1985). For an argument similar to that expressed here, see Richard J. Salvucci and Linda K. Salvucci, “Crecimiento económico y cambio de la productividad en México, 1750-1895,” HISLA Revista Latinoamericana de Historia Económica y Social, 10 (1987), 67-89; and the concluding essay by Van Young, “A modo de conclusión: El siglo paradójico,” Empresarios, indios y estado, 206-231, which might be considered an elaboration of his The Age of Paradox: Mexican Agriculture at the End of the Colonial Period, 1750-1810, Economies of Mexico and Peru, 64-90.


These concentric circles resembled a pattern of contour lines, broken and distorted at points by the road system. Each of the four regions had its own road network; the most elaborate systems were in the most extensive and differentiated regions. Within these regional entities, the road systems were densest in the central parts, but thinned out toward the peripheral landscape separating the regions from each other. Between the four regions were few overland communication routes, usually only one or two. See A. C. van Oss and Slicher van Bath, “An Experiment in the History of Economy and Culture,” JEEH 7 (1978). 407-427 and Geschiedenis van maatschappij en cultuur (Baarn, 1978); van Oss, “Architectural Activity, Demography and Economic Diversification: Regional Economies of Colonial Mexico,” JbLA, 16 (1979), 97-145; Slicher van Bath, Spaans Amerika omstreeks 1600 (Utrecht, 1979), “Economic Diversification in Spanish America Around 1600: Centres, Intermediate Zones and Peripheries,” JbLA, 16 (1979), 53-95, Bevolking en economie in Nteuw Spanje (ca. 1570-1800) (Amsterdam, 1981), “Dos modelos referidos a la relación entre población y economía en Nueva España y Perú durante la época colonial,” Empresarios, indios y estado, 15-44. These four regions are also discussed in Ouweneel, Onderbroken groei, 23-30. On transport and the road system, see Ross Hassig, Trade, Tribute, and Transportation. The Sixteenth-Century Political Economy of the Valley of Mexico (Norman, 1985); Wolfgang Trautmann, Las transformaciones en el paisaje cultural de Tlaxcala durante la época colonial (Wiesbaden, 1981), 199-233. According to the concepts developed by location theorists and applied to colonial Mexico by Van Young in a recently published article, these regional economies should be seen as solar regions and not as dendritic regions; Van Young, “Haciendo historia regional: Consideraciones metodológicas y teóricas,” Anuario del IEHS (Tandil), 2 (1987) and “A modo de conclusión,” 214-221. A solar region had a Von Thünen-like structure, with its economic center within it, and it consumed more of its products than it exported. It was characterized further by a complex internal economic structure and social stratification. On the contrary, the dendritic region was dependent on outside economic centers. This holds true for export-oriented regions such as northern Yucatan in the late nineteenth century, but not for the four regional economies of New Spain, although some subregions like sugar-exporting Cuernavaca-Cuautla did have dendritic characteristics within the solar system.


Diario del viaje que hizo a la América en el siglo XVIII el P. Fray Francisco de Ajofrín, 1763-1766, 2 vols. (Mexico City, 1964); G. F. Gemelli Careri, Viaje a la Nueva España, Francisca Perujo, ed. (Mexico City, 1976); William Bullock, Seis meses de residencia y viajes en México (Mexico City, 1983; orig. pub. in English, London, 1825); Mark Beaufoy, Mexican Illustrations, Founded Upon Facts (London, 1828; reprinted Washington, 1987); Von Humboldt, Ensayo político, passim. See also William B. Taylor, Landlord and Peasant in Colonial Oaxaca (Stanford, 1972) and Town and Country in the Valley of Oaxaca, 1750-1812,” in Provinces of Early Mexico. Variants of Spanish American Regional Evolution, Ida Altman and James Lockhart, eds. (Los Angeles, 1976) 63-96; Van Young, Hacienda and Market in Eighteenth-Century Mexico: The Rural Economy of the Guadalajara Region, 1675-1820 (Berkeley, 1981); Rabell Romero, “San Luis de la Paz: Estudio de economía y demografía histórica (1645-1810)” (Tesis de doctorado, Universidad Nacional Autónoma de México, 1983); Catalina Rodríguez, Comunidades, haciendas y mano de obra en Tlalmanalco (siglo XVIII) (Mexico City, 1982); F. Hurtado, “Dolores Hidalgo en el siglo XVIII: Una aproximación cuantitativa,” HMex, 27:4 (1978), 507-540; Charles Gibson, The Aztecs under Spanish Rule: A History of the Indians of the Valley of Mexico, 1519-1810 (Stanford, 1964); Stephanie G. Wood, “Corporate Adjustments in Colonial Mexican Indian Towns: Toluca Region, 1550-1810” (Ph.D. diss., University of California, Los Angeles, 1984); Cheryl E. Martin, Rural Society in Colonial Morelos (Albuquerque, 1985); Morin, Michoacán, passim; Hassig, Trade, passim; Brading, Haciendas and Ranchos, passim.


The sources: 1) Medina Rubio, Iglesia, passim; 2) Morin, Michoacán, passim; 3) AGI, Audiencia México, legs. 798 and 2106; Mexico, Archivo General de la Nación (hereafter AGN), Ramo de Tributos, vol. 2, exps. 1 and 2, and vol. 36, exp. 17; Peter Gerhard, A Guide to the Historical Geography of New Spain (Cambridge, 1972); 4) Von Humboldt, Ensayo politico, 386-387; 5) Mexico, Archivo Histórico de Hacienda (hereafter AHH), leg. 26-2; 6), 8), 9), and 10) TePaske, Real Hacienda de Nueva España, passim; 7) Hernández Palomo, La renta del pulque en Nueva España, 1663-1810 (Seville, 1979); 11) Florescano, Precios del maíz, app. IV; 12) and 13) Clara Elena Suárez Argüello, La política cerealera en la economía novo-hispana: El caso del trigo (Mexico City, 1985); Virginia García Acosta, “Manufactura y colonia: Las panaderías de la ciudad de México en el siglo XVIII” (Tesis de maestría, Universidad Iberoamericana, 1985); 14) Fuentes para la historia de la tenencia de la tierra en el estado de Hidalgo, J. M. Menes Llaguno, ed. (Pachuca, 1976); Índice de documentos relativos a los pueblos del estado de México, Mario Colín, ed. (Mexico City, 1966); Índice del ramo de tierras del estado de Puebla, E. Méndez Martínez, ed. (Mexico City, 1979); 15) Taylor, Drinking, passim; 16) various sources; 17) van Oss, Inventory of 861 Monuments of Mexican Colonial Architecture (Amsterdam, 1978). Not all variables were complete. The missing values of the incomplete variables had to be replaced by estimated values in order to run the analyses. Missing values were estimated: for each incomplete series, estimates of the missing points were obtained using the coefficient of multiple regression of all complete variables on the nonmissing parts. Secondly, not all variables were comparable: most of them had been measured on a yearly basis, some variables had been measured every five years only. Others had been measured on an irregular basis; for instance, taxes were sometimes collected once in three years, giving a high score in the year of collection, but very low scores in both other years; double counting has the same effect. In order to be able to relate such variables and to level off unreal fluctuations, all variables were replaced by their five-year moving averages. All collected series ran from 1710 to 1809 (n= 100; see  Appendix I). Because we assumed, on a rather ad hoc basis, that the tributarios variable would be slower than the other variables to have an impact on the Central Mexican economy (approximately two years), we used the scores of the tributario variable from 1710 to 1807 and the scores of all other variables from 1712 to 1809.


John K. Chance, Race and Class in Colonial Oaxaca (Stanford, 1978), “On the Mexican Mestizo,” Latin American Research Review, 14:3 (1979), 153-168, “The Ecology of Race and Class in Late Colonial Oaxaca, in Studies in Spanish American Population History, David J. Robinson, ed. (Boulder, 1981), 93-117; Chance and Taylor, “Estate and Class in a Colonial City: Oaxaca in 1792,” Comparative Studies in Society and History (hereafter CSSH), 19 (1977), 454-487; Leon Yacher, Marriage Migration and Racial Mixing in Colonial Tlazazalca (Michoacán), 1750-1800 (Syracuse, 1977); Robert McCaa and Michael M. Swann, “Social Theory and the Log-Linear Approach: The Question of Race and Class in Colonial Spanish America” (Dept. of Geography discussion paper 76, Syracuse University, 1982); Patricia Seed, “Social Dimensions of Race: Mexico City, 1753,” HAHR, 62:4 (Nov. 1982), 569-606; and the debate between Seed and Philip F. Rust, on the one hand, and, on the other, McCaa, Stuart B. Schwartz, and Arturo Grubessich in CSSH, 25 (1983), 703-724. Most revealing is McCaa, “Calidad, Clase, and Marriage in Colonial Mexico: The Case of Parral, 1788-1790,” HAHR, 64:3 (Aug. 1984), 477-501; Gibson, Aztecs, 136-150.


Morin, Santa Inés Zacatelco (1646-1812). Contribución a la demografía histórica del México Central (Mexico City, 1973); Thomas Calvo, Acatzingo: Demografía de una parroquia mexicana (Mexico City, 1973); Wayne Osborn, A Community Study of Metztitlán, New Spain, 1520-1810” (Ph.D. diss., University of Iowa, 1970); Martin, Rural Society, passim; Ouweneel, Onderbroken groei, 77-97.


See the excellent introduction of the topic by David Grigg, Population Growth and Agrarian Change: An Historical Perspective (Cambridge, 1980) and, by the same author, The Dynamics of Agricultural Change. The Historical Experience (London, 1982), 37-43.


Wilhelm Abel, Agricultural Fluctuations in Europe. From the Thirteenth to the Twentieth Centuries (London, 1980), 2; Slicher van Bath, Agrarische geschiedenis, passim; Grigg, Dynamics, 54-56.


Gloria Artís Espríu, “Economía administrada y estrategias de regatones y maquileros: El mercado de trigo en la ciudad de México (siglo XVIII)” (Tesis de lic., Universidad Iberoamericana, 1984).


The findings of Suárez Argüello and García Acosta are confirmed by data from hacienda bookkeeping; see Ouweneel, Onderbroken groei, 208-262. On hacienda agriculture, see Ouweneel, “Eighteenth-Century Tlaxcalan Agriculture: Diary 9 of the Hacienda San Antonio Palula, 1765-1766,” in Haciendas in Central-Mexico from Late Colonial Times to the Revolution, R. Th. J. Buve, ed. (Amsterdam, 1984), 21-83, “Schedules in Hacienda Agriculture: The Case of Santa Ana Aragón (1765-1768) and San Nicolás de los Pilares (1793_1795), Valley of Mexico,” Boletín de Estudios Latinoamericanos y del Caribe, 40 (1986), 63-97, and “Don Claudio Pesero y la administración de la hacienda de Xaltipán (Tlaxcala, 1734-1737),” Empresarios, indios y estado, 168-188; Trautmann, Transformaciones, 121-198.


Pietschmann, “Agricultura,” 79.


Guy Thomson, “The Cotton Textile Industry in Puebla during the Eighteenth and Early Nineteenth Centuries,” Economies of Mexico and Peru, 169-239; Manuel Miño Grijalva, “Espacio económico e industria textil: Los trabajadores de Nueva España; 1780-1810,” HMex, 32:4 (1983), 524-553 and “El camino hacia la fábrica en Nueva España: El caso de la ‘Fábrica de Indianillas’ de Francisco de Iglesias, 1801-1810,” HMex, 34:1 (1984), 135-148. A considerable body of literature now exists on what scholars mean by the term protoindustrial.” Usually, it signifies home or small workshop-based manufacture of goods for a nonlocal market, on the basis of credit by merchants; the “putting out system,” see F. F. Mendels, “Proto-Industrialization: The First Phase of the Industrialization Process,” JEH, 32 (1972), 241-261; Wolfram Fischer, “Rural Industrialization and Population Change,” CSSH, 15 (1973), 158-170, esp. 161; Hans Medick, “The Proto-Industrial Family Economy: The Structural Function of Household and Family During the Transition from Peasant Society to Industrial Capitalism,” Social History, 3 (1976), 291-315; David Levine, Family Formation in an Age of Nascent Capitalism (New York, 1977); Industrialization before Industrialization: Rural Industry in the Genesis of Capitalism, Peter Kriedte, Hans Medick, and Jürgen Schlumbohm, eds. (Cambridge, 1981); Kriedte, Peasants, Landlords and Merchant Capitalists: Europe and the World Economy, 1500-1800 (Cambridge, 1983), 70-91, 115-145; Richard Wall, “Leaving Home and the Process of Household Formation in Preindustrial England, ” Continuity and Change, 2 (1987), 77-101, esp. 82-90, and 99, n. 16. Interesting as well is Slicher van Bath, Een samenleving onder spanning: Geschiedenis van het platteland van Overijssel (Assen, 1957; reprinted Arnhem, 1977). Mendels defines protoindustrialization as a distinct stage in European economic development in which large-scale cottage industry was typical. Merchants put work out into the countryside and collected it to be sold in distant markets, thus not locally. His model suggests that protoindustries were commonly found in areas of subsistence farming with large populations accessible to urban cloth merchants. The consequences most often were demographic: whole families were drawn into protoindustrial work; more children were born to couples, meaning rapid population growth; and regional crises occurred when the industry contracted. Kriedte called proto-industrialization “one of the most important engines of population growth” (see his Peasants, 103). Although most authors would agree, a different view is expressed by Gay L. Gullickson, Spinners and Weavers of Auffay: Rural Industry and the Sexual Division of Labor in a French Village, 1750-1850 (New York, 1986). She found typical protoindustrial production in the Pays de Caux of Normandy, but without the occurrence of rapid population growth. On protoindustry in New Spain, see Ouweneel, Onderbroken groei, 123-131; Pietschmann, “Agricultural” Dehouve, “Pueblo de indios”; and Gerst, Wirtschaftliche Entwicklung Mexikos, 39-79. Even credit was given to the peasants in the villages by merchants, as part of the repartimiento de efectos. Richard Salvucci recently published a study on the obrajes in Mexican history, in fact showing that the obraje was not a protoindustrial industry, according to the definition introduced by Mendels and Medick, and refined by authors like Kriedte, Fischer, and Wall; see his Textiles and Capitalism in Mexico: An Economic History of the Obrajes, 1539-1840 (Princeton, 1987).


On this point, see Ronald A. Wykstra, Introductory Economics (New York, 1971), 249-255 and Slicher van Bath, Agrarische geschiedenis, 217.


On the unprecedented mining boom, see the literature on silver production mentioned in n. 1 above. On commodity trade in the hands of few people, see Amartya Sen, Poverty and Famines: An Essay on Entitlement and Deprivation (Oxford, 1982) and Slicher van Bath, Agrarische geschiedenis, 217.


Jacobsen and Puhle, “Introduction,” Economies of Mexico and Peru, 1-25, esp. 10; Garner, “Silver Production,” passim; Javier Cuenca Esteban, “Statistics of Spain’s Colonial Trade, 1792-1820: Consular Duties, Cargo Inventories, and Balances of Trade,” HAHR, 61:3 (Aug. 1981), 381-428; Coatsworth, “Limits of Colonial Absolutism,” passim and “Mexican Mining Industry,” passim.


Florescano, Precios, 77-81. Curiously, however, a simple calculation of the ratio of the number of coins minted to the number of marcos brought into the mint results in a value of about eleven to twelve pesos. There occurs a deviation in this ratio after 1800 which might have been due to irregular administration of the growing stock in Mexico City, caused by increased transportation to the colony’s capital and stagnating exports to Europe. The mean of these post-1800 values, however, is again a little over eleven pesos per marco. This points to the irregular peak of the production figures for the 1790s and their subsequent decline. Because we are dealing with the shipments of silver to Mexico City, both the decline of the 1780s and the peak of the 1790s can be explained by the scarcity of mules during the 1780s. Serrera Contreras shows that the droughts of the 1780s caused the death of large numbers of mules. Also, transporters were reluctant to send their mules on the way in that period. This impeded silver transport but not silver production, and when the interregional trade revived after the droughts were over and the lost mules could be replaced, the production of the 1780s was brought into the capital. Future research must reveal in what way production figures must be corrected for transport difficulties. After all, what is important to the silver debate is how much was actually produced; how much was shipped to Mexico City is less important. Perhaps there was no real decline in silver production until the next period of droughts and the subsequent peasant revolts after 1808; see Ramón Serrera Contreras, Guadalajara ganadera: Estudio regional novohispano, 1760-1805 (Seville, 1977). Quite revealing are the reports of the intendants on this problem in 1786 and 1787, see AGI, México, leg. 1675. On the connection between bullion and the economy, see M. Morineau, Incroyables gazettes et fabuleux métaux: Les retours des trésors américains d’après les gazettes hollandaises (XVIe-XVIIIe siècles) (New York and Paris, 1985) and Slicher van Bath’s essay “Het zilver en goud: produktie en spreiding,” in his Indianen en Spanjaarden: Een ontmoeting tussen twee werelden. Latijns Amerika 1500-1800 (Amsterdam, 1989), 139-174.


TePaske, “General Tendencies,” 322-325.


Brading, “Facts and Figments” and personal communication.


Hernández Palomo, Renta del pulque, passim; Michael C. Scardaville, “Alcohol Abuse and Tavern Reform in Late Colonial Mexico City,” HAHR, 60:4 (Nov. 1980), 643-671.


AGI, México, legs. 1675 and 2096; M. Bahena Pérez, “La transportación de carga en el comercio de la Nueva España: La arriería (1789-1810)” (Tesis de maestría, Universidad Iberoamericana, 1985); Miquel Izard, “Metropolitanos, criollos y reformistas: La Nueva España de Revillagigedo (1789-1794),” Boletín Americanista, 22:30 (1980), 181-222.


Susan Linda Swan, “Climate, Crops and Livestock: Some Aspects of Colonial Mexi can Agriculture” (Ph.D. diss., Washington State University, 1977); Gibson, Aztecs, passim; Tutino, “Creole Mexico,” passim; Ouweneel, Onderbroken groei, passim; Brading, “Facts and Figments”; and Garner, “Further Consideration.” On pauperization, see Van Young, “A modo de conclusión” and “Rich Get Richer” and Von Humboldt, Ensayo político, 541-544.


Tutino, From Insurrection to Revolution, 41-182; Hamnett, Roots of Insurgency, 102-124; Van Young, “Rich Get Richer”; Jacobsen and Puhle, “Introduction.” On the moral economy, see Taylor, Drinking, 113-151; David W. Sabean, Power in the Blood: Popular Culture and Village Discourse in Early Modern Germany (New York, 1984); George Rudé, The Crowd in History (New York, 1964) and Ideology and Popular Protest (New York, 1980); E. P. Thompson, “The Moral Economy of the English Crowd in the Eighteenth Century,” Past and Present, 50 (1971), 76-136; James Scott, The Moral Economy of the Peasant: Rebellion and Subsistence in South East Asia (New Haven, 1976); José Miguel Palop, Hambre y lucha antifeudal: Las crisis de subsistencias en Valencia (siglo XVIII) (Madrid, 1977); Anthony McFarlane, “Civil Disorders and Popular Protest in Late Colonial New Granada,” HAHR, 64:1 (Feb. 1984), 17-54; Erick D. Langer, “Labor Strikes and Reciprocity on Chuquisaca Haciendas,” HAHR, 65:2 (May 1985), 255-277.


Tutino, “Creole Mexico,” 242-246.


The situation in the villages in the eighteenth century must have resembled that during the revolution. Compare Tutino, “Provincial Spaniards, Indian Towns and Haciendas: Interrelated Sectors of Agrarian Society in the Valleys of Mexico and Toluca, 1750-1810,” Provinces of Early Mexico, 182-187, with Frans J. Schryer, “A Ranchero Elite in the Region of Huejutla. The Career of General Juvencio Nochebuena of Atlapexco,” paper given at the CEDLA workshop, Amsterdam, 1987.


Van Oss and Slicher van Bath, “Experiment,” passim; Slicher van Bath, Real Hacienda.


Van Oss, Inventory of 861 Monuments, passim.


Ouweneel, Onderbroken groei. 48-65; Swan, “Climate,” passim; Tutino, “Creole Mexico,” passim; Florescano, Precios, passim.


More technical explanation and discussion of our statistical techniques appear in  Appendix II. For a brief and illustrative description of canonical correlation analysis, see M. G. Kendall, A Course in Multivariate Analysis (London, 1972). A more mathematical description is in J. P. van der Geer, Introduction to Linear Multivariate Data Analysis, 2 vols. (Leiden, 1986). Canonical correlation analysis is widely used in many fields of social and other scientific research as a powerful tool for exploratory multivariate analysis. Applications can be found in psychology, sociology, anthropology, econometrics, biology, etc. To mention a few: C. C. J. H. Bijleveld, “The Effect of Education on Sudanese Women’s Attitudes Towards Female Circumcision” (Leiden, 1986); P. M. Bentler and G. J. Huba, “Symmetric and Asymmetric Rotation in Canonical Correlation Analysis: New Methods with Drug Variable Examples,” in Multivariate Applications in the Social Sciences, N. Hirschberg and L. G. Humphreys, eds. (Hillsdale, NJ, 1982); H. Hoteling, “Relations Between Two Sets of Variates,” Biometrika, 28 (1936), 321-377.


On state-space or linear dynamic systems analysis see Bijleveld, Exploratory Linear Dynamic Systems Analysis (Leiden, forthcoming) and Bijleveld, DYNAMALS Progress Report (Leiden, 1987; Dept. of Data Theory, RR 87-16). On the relation between canonical correlation analysis and linear dynamic systems analysis, see J. de Leeuw and Bijleveld, Fit ting Longitudinal Reduced Rank Regression Models by Alternating Least Squares (Leiden, 1988; Dept. of Data Theory, RR 88-03).


Slicher van Bath, “Yield Ratios, 810-1820,” AAG Bijdragen, 10 (1963), 1-264; Mark Overton, “Computer Analysis of an Inconsistent Data Source: The Case of Probate Inventories,” Journal of Historical Geography, 3 (1977), 317-326 and “Estimating Crop Yields for Probate Inventories: An Example from East Anglia, 1580-1740,” JEH, 39 (1979), 363-378; Gene C. Wilken, Good Farmers, Traditional Agriculture and Resource Management in Mexico and Central America (Berkeley, 1987).


Bijleveld, DYNAMALS Progress Report, passim.

Author notes


The authors wish to thank Pitou van Dijck, Chad Gundy, Robert Johnson, Paul Klein, Jan de Leeuw, and Frank Schenk for their comments on the manuscript. We are also indebted to Ursula Ewald and her colleagues of the University of Heidelberg, Germany, for sending us their map of colonial New Spain. Part of the research was carried out with funding from the Netherlands Foundation for the Advancement of Tropical Research (WOTRO) and from the Institute for Road Safety Research (SWOV).