Online Appendix of Mexico S Real Wages in the Age of the Great Divergence, 1730-1930

Online Appendix of Mexico’s Real Wages in the Age of the Great Divergence, 1730-1930.

Amílcar E. Challú and Aurora Gómez-Galvarriato

Appendix A: Construction of Price, Wage and Silver Exchange Series

These sections provide additional information on the procedures involved in creating the wages and price series.

Wages

We weighted the daily pay rates by the number of days and laborers (peones). All sources were weekly reports of labor costs in construction work for the Vizcaínas school, as well as other public and clerical institutions. Because wages tended to be stable, interpolation was used to fill in missing observations.

Prices

The Excel file reports annual series after the use of imputation for missing cases, as well as the individual spreadsheets used in their construction.

For each product we summarized local prices in three categories (institutional, retail and wholesale), ancillary data used for imputation (e.g. prices in other regions, prices from other products), notes and the source. For each product we averaged local prices of the product from retail, wholesale and institutional prices. Wholesale prices were multiplied by 1.052 to account for their lower level, as explained below. In the case of industrialized products with scant observations but that depended heavily on one commodity (tortillas, bread, candles, soap), we used the prices of the commodity scaled by a factor obtained in bivariate regressions. If the coefficient is not significantly different from 1,no correction was applied. Finally, we imputed missing observations using series from highly-correlated locations in Mexico, or from highly-correlated products. The imputation relies either on correlations orchaining observations from the ancillary series using a common base year. For the missing observations, we estimate minimum and maximum prices using the 25th and 75th percentiles of the imputed series in 10- to 25-year moving windows. These boundaries translate into an interval of uncertainty in the CPI and welfare ratio.

Systematic Bias of Institutional and Wholesale Prices

In preparing Series 1 of each product, we averaged all available retail, institutional and wholesale prices from Mexico City, adding an upward correction of 5.2 percent to the latter, and equating institutional and retail prices. In this section the rationale behind this procedure is explained.

Retail prices refer to prices paid by typical urban households, our reference population in this analysis. They include a) posturas from the colonial period fixed by the duration of one day, month or year by a regulatory body, usually in connection with varying market conditions; b) tasas de efectos and tarifas, which were prices fixed for short periods of time in the post-independence period; c) surveys of retail prices conducted by public organizations from 1890 on.

Institutional prices refer to prices paid by institutions such as hospitals, jails, and schools with a resident population. The prices refer to the cost of the product; they do not include the cost of a carrier (usually one or two realesfor large loads). In most cases we calculated these prices as the sum of the value, divided by the total quantity, and these transactions took place over more than one month.

Wholesale prices refer to current market prices quoted in newspapers. The first observations date from the early nineteenth century and often include information on the dealer or dealers that provided the information. After the 1840s they were obtained from reports of the Colegio de Corredores (the board of wholesale dealers) and are more systematic in the kind of information they convey. After 1885, newspapers such as La Semana Mercantil, El Economista Mexicano and El Boletín Financiero y Minero published weekly reports of wholesale purchase and sale prices, only interrupted from late 1913 to 1918. When provided with a range of prices, we averaged them.

No single series spanned the whole period. Moreover, we lacked a clear-cut superior series to chain the two others. We pooled all available information summarized in three types (institutional, retail and wholesale), averaging the information with weights to account for systematic biases. In 280 year-product panels we had observations of more than one type. We ran a regression on those panels using log prices as the dependent variable and dummy variables to account for fixed effects of the year, product and the type of observation (institutional purchase, retail or wholesale transaction prices). The coefficient for the institutional dummy variable was 0.040 (t score: 1.26), and -0.051 (t score: 2.13) for wholesale.

The coefficient on the wholesale variable implies that wholesale prices were 5.2% more expensive, and the result is statistically significant. However, the result for institutional purchases is not statistically significant. We then used weights for institutional, retail and wholesale prices of 1, 1.0522 and 1, respectively.

Seasonality

While prices had wide oscillations from month to month, any seasonal correction did not differ from the annual average by more than a couple percent points. Hence, we preferred the simpler approach of averaging monthly observations. We tested seasonal effects using consistent monthly information for all products available in La Semana Mercantil and Boletín Financiero y Minero (1885-1929); for the colonial period, we also had monthly and quarterly data for maize and wheat, respectively. In all products, variations followed expected patterns but their differences relative to the annual average were below 3 percent and were not statistically significant. In the case of maize, the differences were up to 5 percent in the two peak months (January and June). Applying these corrections to the final result did not change the final series given that we often had observations from multiple months, and the single observations were rarely from the months with higher seasonality. Moreover, the scarcer the data, the more likely it is that the price reflects an anomalous condition worth reporting. For instance, anecdotal reports or fixed price regulations tended to show high prices in low-price months. Instead of applying corrections, we state whether the annual value is based on an average across months or is the result of a single observation.

Explanations by products

Maize. We found observations of maize prices to build Series 1 for most years. Some observations were discriminated by geographic origin (but still sold in Mexico City), but we did not find that the differences justified applying corrections. In a small number of cases, gaps were filled by chaining prices from nearby locations in the Valley of Mexico and Toluca.

Tortillas. There were practically no direct prices for tortillas. Anecdotal evidence from the eighteenth and nineteenth centuries shows that maize represented 80 percent of the cost of making tortillas (La Gaceta de México, 25 July 1786; El Siglo Diez y Nueve, 10 October 1878). This would translate into a 12 percent higher price per kilogram, because water is added to the tortillas. In the late nineteenth and early twentieth century, price ratios of tortillas and maize per kilogram point to tortillas being 25 to 50 percent more expensive than corn (El Siglo Diez y Nueve, 23 September 1892; AGN, Departamento de Estadística Nacional, surveys of 1924 and 1930; Estadísticas Históricas de México, table 19.5). In none of the figures does mechanization seem to have reduced the final price paid by the consumer; if anything, the ratio seems to have increased in the 1910s and 1920s. We found no strong correlation to wages. We multiplied maize prices by a constant factor of 1.422 (the average of all observed price ratios) to obtain tortilla prices.

Bread. We found 41 annual observations of bread prices. They were, however, inconsistent and volatile, often depending on criticism of high prices. While we chose white and floreado bread, the quality was difficult to ascertain as reflected in varying ratios between different qualities. We preferred, instead, to use Allen’s (2001) approach and estimate a regression line for bread prices based on wheat prices and wages. The wholesale wheat series was more stable and reliable; missingobservations were imputed with prices of high-quality flour. The equation to estimate bread prices was (T ratios in parentheses):

BREAD = 0.003 + 0.92 * WHEAT+ 0.18 * WAGE

All variables were in pesos and kilograms; WAGE was the daily laborer wage (payment in bakeries ranged from wages equivalent to laborers and masons in construction work). The Pearson coefficient was 0.95. Adjusting for time periods did not result in a better fit, which is surprising, but agrees with the reporting of systematic high prices for bread in the late nineteenth century (El Siglo Diez y Nueve, 23 September 1892) and our findings for tortillas.

We were able to find local frijol beans prices from 1777 on. For the previous period we imputed prices from the sale price of tithed beans in the Valladolid (Michoacán) district using common years to build the regression line. We smoothed prices because low sales volumes made the tithe series highly volatile. The Pearson correlation of Valladolid and Mexico City prices was 0.75.

Beef and Lamb. Local price observations were relatively frequent for most of the period with the exception of the first half of the nineteenth century. We imputed missing observations of beef prices with lamb, and viceversa. We used the price of beef for the ‘barebones’ basket, but combined the beef and lamb prices for the respectable basket to account for the fact that lamb was the most popular type of meat for most of the period. The meat price index weighted lamb at 70 percent until 1813, 43 from 1814 to 1874 and 16 percent after 1875. The first and the last figures were obtained from statistics of food supply in the city in Miño Grijalva (2004), and Penafiel (1892); the middle period was interpolated but it is also consistent with a greater importance of cattle entries in the city in the 1820s and 1830s as reported in Azcárate (1839)andMoncada (2013).

It should be noted that the prices under the abasto, the regulated meat supply system in place until 1812, were very low by any standard. Quiroz (2005) discussed the functioning of this colonial institution in detail and created the authoritative price series in use by Allen et al. (2012), Arroyo Abad et al (2012), and more prominently by Dobado and García (2013). Yet, the volume of supply after the end of the abasto in 1814 and the changes in relative prices suggest that the abasto price is biased downwards. It is true that the price of beef in silver grams is well below other countries, but the price in Mexico City wasonly 17 percent above the estimated cost per kilogram of live cattle in supplying areas as shown in Brading (1978, p. 86) and Van Young (1981, p. 47), and 20 percent more expensive than the nearby city of Cuernavaca, which was closer to supplying areas in the south (Bartlet 1974). By contrast, grain prices were much lower in those areas relative to Mexico City. More interestingly, relative prices of beef in the capital increased substantially after the abolition of the abasto. Lamb was three times more expensive than beef during the abasto, even though the live weight of the slaughtered sheep outnumbered cattle by a factor of 2.25. After the end of the abasto, the price of beef doubled and the price of lamb remained constant; yet, consumption of lamb declined relative to beef. Similarly, tallow (a derivative of the slaughtering of cattle) was five times more expensive than beef during the abasto, and 2.6 times afterward. While contemporary sources mentioned the increase in the price of beef, it is apparent that the meat supplier bid a low postura for beef, and capped the entries of cattle at a low level to maintain the profitability of the operation.In fact, the available figures for the eighteenth century indicate that per-capita supply of beef remained virtually constant and never surpassed 18 kilograms per capita.

Pork and lard. Observations for regular pork meat are mostly inexistent before the mid nineteenth century. Most wholesale prices were for live animals; the correlation with retail prices was very high (R=0.95), and the equivalence was, expectedly, higher (1.26 x wholesale). We imputed missing observations in Series 1 by using lard and ham prices; both had equivalent prices and were a derivative of the pork slaughtering processed by the pork butcher houses, tocinerías (Quiroz 2005). The Pearson correlation of the lard and ham average to pork meat was 0.86. The vast majority of lard and ham prices in the colonial period was from the regulated prices (posturas) established at the beginning of the year; the posturas were for both retail and wholesale prices. (The 1730 observation was obtained from the 1729 postura of pork loin, chained using common observations in the mid century.) To impute missing values in Series 1 of lard, we used ham prices and chained the 1849 value from the wholesale pork price.

Sugar has the most complete coverage in our database by virtue of the extensive series published by Barrett (1970) and Crespo (1988). For most years we have sales prices in Mexico City, and gaps were filledby chaining sale prices from the Hacienda of Atlacomulco (100 km south of Mexico City).

Pulque prices are among the most scattered and anecdotal information we have found. More research is certainly needed to improve this series, yet the existing observations point to a pattern: increase of prices toward the end of the eighteenth century, remaining at high levels in the first half of the nineteenth century, a reduction of prices in the late 1880s, and an increase with the outbreak of the Revolution. We did not use sales prices of bottled pulque in fondas; the price was certainly higher than in the regular outlets, but in all they confirm the general high prices of pulque in the nineteenth century through the 1880s.

Soapprices were much less frequent in the early period and in the mid nineteenth century. As a byproduct of pork slaughtering, we used Lard (Series 2) for imputing missing values (R=0.89), using a higher multiplier for the period before 1880. Given the small number of cases in the eighteenth century, the 1730-1780 period was used to create maximum and minimums. Afterward we used 11- and 5-year moving windows depending on data availability.

Candles.We found 42 annual prices of candles beginning in the 1780s. Before the beginningof the twentieth century, the lowest priced candle was made of tallow, hence missing observations were imputed by multiplying the (much more frequently found) price of tallow by a constant factor of 1.39 (the Pearson correlation between these variables was 0.6). While we found references to paraffin candles since the 1870s, it was only in the twentieth century when it became less expensive than tallow candles. Wax and other expensive types of candles were in use in temples, but we did not find convincing evidence that the working classes used them.

Charcoal.The coverage of charcoal prices is sparse. For the colonial period, Castro (2008) provides prices from Real Casa de Moneda’s supply contracts. Afterwards we relied on institutional purchases, advertisements, and government reports. The most common unit of measure for charcoal was the carga, but the carga weighed from 4 to 7 arrobas, much less than the standard carga of 12 arrobas. The big or mule carga is consistent with the capacity of a standard carga considering the lower density of charcoal, while the smaller carga was probably the kind that Indian carbon peddlers hauled on their backs. Four annual price observations stated the weight equivalence clearly; for all others we used an equivalence of 5.5 arrobas. The costal, saca and saco were more problematic. Both the descriptions and relative prices of yute costales and sacas imply a capacity from 3 to 4 arrobas. We estimated tercios as the capacity of a regular twelve-arroba carga adjusted to the lower density of charcoal (0.50 kg to l, compared to 0.75-0.80 of grains). An interval regression analysis using minimum and maximum feasible ranges and controlling for dummy variables of different units of measure helped define average equivalences within these ranges and confirmed our choices.

Overall, the price of charcoal increased moderately in the mid nineteenth century, and sharply from the 1880s to the 1920s. By the 1890s prices were more than twice those of the colonial period, and by the 1920s more than five times. Observations of firewood prices in the native zontle unit (400 pieces of lumber) confirm the sharp increase of prices from the mid nineteenth century to the early twentieth century, although concerns with the unit of measure and wide disparities in firewood quality led us to discard this information before 1925, when the number of lumber pieces is stated exactly.

Manta cloth.We had 80 annual observations of local, annual prices of manta pieces, the coarse cotton cloth used in Mexico. Prices stated in newspaper market reports were retail, in that wholesale discounts were applied to the nominal retail price. However, some scant reports of factory sales in Mexico City had a discounted price. In years with missing observations, we used the price of yarn in Mexico City and manta prices from Puebla and Zacatecas, using the ratios from years with overlapping information.