Arij Ouweneel and Catrien C. J. H. Bijleveld question “the usefulness of the tithe data of colonial Mexican bishoprics as an index of agrarian production in New Spain.” After reading their article, one is persuaded of the dangers of using tithes as a source. Tithes promise the key to information on agricultural production and fulfill their promise, but they mix valuable information with unwanted clutter. Tithes are like velcro, every change in the economy adheres to them. Each tithe figure is a blend of information on agricultural production, variations in relative prices, inflation, changes in the product mix, and changes in the capacity of the church to collect the tithes. Tithes say plenty about the economy, too much. As Ouweneel and Bijleveld make clear, before drawing conclusions from tithe data, it is necessary to find a methodology to separate the effects of all the variables that affect them. To follow our anachronistic metaphor, it is necessary to clean the velcro.

Ouweneel and Bijleveld are more successful in analyzing the problems of tithes than in developing a methodology to unscramble the puzzle. In the authors’ view, two key problems deserve attention: the impact of inflation, and what they call the “bureaucratic component” (changes in the ability of the church to collect tithes). Since the data are incomplete, neither of these problems can be solved with standard methodologies. The orthodox way of dealing with inflation is to use a price index to deflate the data. The bureaucratic component cannot be measured so easily. With better data it would be possible to compare agricultural production and tithe collection; the difference would be the measure of the church’s ability to collect tithes. Unfortunately, the very reason why tithe data are so interesting is that we do not have information on total agricultural production.

Since conventional methods are not feasible, Ouweneel and Bijleveld enlist the services of modern statistical analysis. In my view, there are two reasons to adopt a new methodology. The first one is to learn something that we would not otherwise know. The second is to know better, bring more clarity, or offer more solid support to something that we already know. I will argue that in the first part of their article, the more conventional one, Ouweneel and Bijleveld succeeded in establishing the main shortcomings of tithe data at the bishopric level, but that their methodological exercise is flawed by the selection of variables.

Their methodology involves the analysis of 17 different variables with three different approaches. They try to find the correlation between a dependent variable, tithes, and three sets of independent variables: inflationary, bureaucratic, and purchasing power. The criteria used to select variables were overly catholic. As the authors acknowledge in the Appendix, statistical techniques are useful to the social sciences when they are a tool of theory. Without a solid theoretical foundation, one may paraphrase Gertrude Stein and say that a statistical correlation is a statistical correlation is a statistical correlation. Variables ought to be selected only when there is a very good theoretical reason to do so. The correlation between liberal ideas and ownership of Volvos is well established, but no serious political scientist gathered data on Volvo sales to predict the outcome of the last presidential election.

Thus, the selection of variables deserves careful scrutiny. Three of the inflationary variables chosen are beyond reproach since they are related to silver production. The correlation between inflation and the money supply is widely accepted, and there is strong theoretical and empirical work supporting it. One can question, however, the decision to use all three variables, since they say basically the same thing. The use of tribute payers as a proxy for population is more debatable. In standard economic literature, population growth affects inflation only insofar as it affects the rate of unemployment.1 By itself, the tributarios series does not say anything about unemployment. Is it fair to refer only to standard economic literature? In the context of this article the answer is yes. The purpose of the article is to elucidate the usefulness of tithes as an historical source, and not to develop a new theory of inflation.

Notwithstanding these misgivings, the emphasis on the importance of inflation is well taken. Cecilia Rabell’s case study of tithes in San Luis de la Paz, a town in the Michoacán region, has enough information to shed light on the discussion.2 The monetary veil can indeed hide the true nature of changes in the tithe series. Each year a variety of products were tithed: chicken, horses, corn, wheat, grapes, etc. The figure for a year was the aggregation of the quantities of all the products multiplied by their price. But the price for each product changed every year, and the amount of each product was also different. Moreover, price changes were the result not only of changes in relative prices (changes in supply and demand) but also of changes in the general price level. The problem is not new and has a standard solution: a price index can be used to deflate the series, but a good price index is difficult to construct. Ideally, the tithe price index should be a weighted average of the prices of the products included in the tithes. This approach, the authors argue, is impractical. This is true at the bishopric level, but in a case study the problem is of manageable proportions. Rabell constructed a tithe price index for San Luis de la Paz and used it to deflate her original data. A comparison between the series in current pesos and the deflated series gives an easy-to-grasp illustration of the importance of the monetary veil. As could be expected, the two series were correlated, the R square was 0.80.3 Another way of interpreting the coefficient is that prices introduced an overall distortion of about 20 percent. Ouweneel and Bijleveld confirm this point, and their analysis indicates that silver production was a source of inflation, but they do not provide an intuitive measure of the problem. As we have seen in the case of San Luis de la Paz, inflation was important, but its importance can be exaggerated. The thick monetary veil was not completely opaque. I do not think that they have succeeded in adding population growth to the standard explanation for eighteenth-century inflation. Both silver production and population increased throughout the century, and both of them are strongly correlated to tithes in pesos. That does not mean that population growth played a role in the price increase.

There is a compelling logic behind the argument that changes in the bureaucratic capacity of the church also affect the usefulness of tithe data. Again, the basic argument is eminently sound and is confirmed by the cases of San Luis de la Paz and San Miguel el Grande.4 Rabell describes a bureaucracy that was constantly evolving. “La burocracia eclesiástica,” she writes, “renovaba cuando era necesario sus procedimientos y se adaptaba con gran flexibilidad a las situaciones peculiares de cada parroquia.”5 I would argue that the authors’ solid argument was not strengthened by the statistical exercise. The variables selected as proxies for the bureaucratic factor (receipts of pulque taxes, alcabala, novenos reales, and tributos reales of the Caja de México) are inadequate. The direction of the correlation between these variables and tithe collection is not clear. One can think of plausible scenarios under which the correlation could be positive, negative, or close to zero. An increase in the population of Mexico City, for example, could increase the collections of alcabala. The number of transactions would increase, and it would be easier to collect taxes in the city, where the population is concentrated, than in the countryside. Urban tax collection can increase even if nothing changes in rural areas. In general, I see no reason to expect a close link between the capacity of the civil authorities to collect taxes in the city and the capacity of the church to raise tithes in rural areas. In fact, during the late eighteenth century, the colonial authorities, in their effort to consolidate their authority at the expense of the church, tried to take control of tithe collection and ultimately failed.6

As to the “purchasing-power” variables, there are two that are particularly problematic. The correlation between revolts and the purchasing power of the poor rests on shaky theoretical foundations. One of the main models that deal with this issue is linked to the “relative deprivation” model whose main exponent is Ted Gurr.7 According to this model, revolts tend to occur when there is a difference between what people expect their economic situation to be and what it really is. In this context, both a depression and a smaller-than-expected rate of growth could increase the likelihood of a revolt. On the other hand, an expected depression would not create instability. There may be good reasons to disagree with the “relative deprivation” model, but the point is that the correlation between peasant revolts and decreasing purchasing power is not well established.8 In fact, there are examples of revolts taking place as a result of social dislocations created by economic growth. It can be argued that there is evidence in eighteenth-century Mexico for the relationship between revolts and decreasing purchasing power, but in that case the variable is selected on an ad hoc basis, and all it does is to reinforce a preconceived notion.

Monument construction, another variable measuring purchasing power, is also questionable. Van Oss and Slicher found that the construction of churches and monasteries reflected the economic environment of the time. However, that is not a solid reason to include monument construction as a proxy for the purchasing power of the poor. It could be just the opposite. In periods of economic bonanza in the market sector there was more pressure on labor, and therefore the Indian poor had less time to devote to their subsistence plots. This may be the reason why the coefficient of the “monument construction” variable is negative in all the correlations.

Because of the weak theoretical basis for the selection of some variables, it is impossible to interpret the meaning of the coefficients obtained through the various statistical exercises. A statistical correlation does not mean causality (this argument has been abused by the Tobacco Institute, but is nonetheless true); its role is to strengthen a solid argument. The meaning of the correlation is blurred when solid variables chosen on the basis of sound theory are combined with more questionable ones. In such cases, one may wonder, is this the “Volvo effect,” or is it causality?

Ouweneel and Bijleveld’s main points remain valid. The usefulness of tithes as an index of agrarian production is weakened by the presence of inflation and of the bureaucratic factor. On the other hand, tithe studies at the parish level contain all the data that are missing at the bishopric level and deserve more attention.


For a standard treatment of this problem, see Rudiger Dornbusch and Stanley Fischer, Macroeconomics (New York, 1981), chap. 13.


Cecilia Rabell, Los diezmos de San Luis de la Paz (Mexico City, 1986). For another case study, see Silvia Galicia, “Precios y producción en San Miguel el Grande, 1661-1803,” Cuadernos de Trabajo del Departamento de Investigaciones Históricas, INAH, Nov. 1975, mimeo.


In standard correlation analysis, the R square (or coefficient of determination) is the percentage of the variance of the dependent variable explained by the independent variable.


Rabell, Los diezmos, 40–50 and Galicia, “Precios y producción,” 97.


Rabell, Los diezmos, 50.


N. M. Farris, Crown and Clergy in Colonial Mexico, 1759-1822 (London, 1968), 152-155.


Ted Gurr, Why Men Rebel (Princeton, 1970).


For critiques of Gurr’s model, see Theda Skocpol, States and Social Revolutions (Cambridge, 1979) and Charles Tilly, From Modernization to Revolution (Reading, MA, 1978).