Colonial agricultural estates and rural development in twentieth-century Mexico

ABSTRACT

This study documents that municipalities in central Mexico closer in the past to an agricultural estate (hacienda) are associated with higher literacy and lower poverty throughout the twentieth century than municipalities similar in other respects but farther from a hacienda. The results are robust to various specifications, neighbour matching analyses, and a placebo-type test. The complementarities between late-colonial haciendas in central Mexico and mining and trade appear to have set municipalities close to a hacienda on a distinct development path. The evidence points to local-scale economies in hacienda locations that coordinated new investments away from agriculture and towards the new industrial and commercial sectors. The twentieth-century land reform and the railroad play a small role in explaining hacienda legacy. Our findings highlight the role of landed estates as centres linking rural economic activity to the main colonial economic activities, mining and trade.

KEYWORDS:

• colonial institutions
• agricultural estates
• economic development
• agglomeration effects
• economic geography

Acknowledgements

We thank Lucy Hackett and Mariano Cepeda for excellent research assistance. We benefited from helpful suggestions and comments by two anonymous referees, Martha J. Bailey, Natalie Bau, Dora Costa, Michela Giorcelli, Steve Haber, Daniel Haanwinckel, Herbert Klein, Joaquín Serrano, Prerna Singh, Tuan Hwee Sng, and John Tutino. We also thank participants at the UCLA Applied Proseminar, the Workshop on the Historical Political Economy of Mexico at Georgetown University, the Economic History of Developing Regions Virtual Seminar, the Economics and Public Policy Congress SobreMéxico, the LANE HOPE virtual seminar, the CEPR Symposium, the V AMHE Jornadas, and Agricliometrics IV. Part of the research was undertaken while Luz Marina Arias was a CASBS Fellow at Stanford University.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 See Bleakley and Lin (2012); Colmenares (1969); Faguet, Matajira, and Sánchez (2017); Fergusson, Larreguy, and Riaño (2015); Fujiwara, Laudares, and Valencia Caicedo (2019); Mahoney (2010); Valencia Caicedo (2019); Waldinger (2017), among others.

2 The Spaniards had access to native labour and their produce through the encomienda and the repartimiento, early institutions that had declined by the late sixteenth century (Gibson 1967, 66–68; Knight 2002).

3 Colonial settlement and early labour institutions were influenced by the pre-colonial political organization of the native societies encountered by the Europeans in the Americas (Arias and Girod 2014).

4 Haciendas emerged as private estates yet the mendicant orders also acquired and managed landed estates. The Jesuits acquired large tracts of land in central Mexico and became skilled agriculturalists and cattle farmers: see Konrad (1981) and Riley (1973). The Dominicans owned land in the Oaxaca valley while Augustinians owned some estates in the Bajío. See Chevalier (1999, chapter VII).

5 The Bajío encompasses the contemporary states of Guanajuato, Querétaro, San Luis Potosí, and parts of Michoacán. Nueva Galicia, to the west of the Bajío, includes the contemporary states of Jalisco, Nayarit, Aguascalientes, and parts of Zacatecas.

6 From the ‘Relación goeográfica’ about Tlajomulco in Jalisco, cited in Van Young (2006, 107).

7 Mexico is composed of 32 autonomous states that are divided into municipalities. Municipalities have changed but Mexican public records allow us to identify movements across time.

8 The number of municipalities is based on the 1970 census. The number of municipalities with haciendas changes depending on the municipalities for which there is information in each census.

9 In contrast to current measures of soil suitability, altitude and land gradient are likely to not have changed much since the colonial period.

10 We obtained the geographic data from the Global Agro-Ecological Zones (GAEZ) provided by the Food and Agriculture Organization of the United Nations (FAO) – see http://www.fao.org/nr/gaez. Altitude is measured in kilometres.

11 We thank Alberto Díaz-Cayeros for sharing the data.

12 Ranchos also shared the rural environs and were typically smaller agricultural units than haciendas. However, there is no systematic way to differentiate between haciendas and ranchos.

13 We choose 100 km because (i) it includes all municipalities within one standard deviation of the mean of distance to hacienda, and (ii) after a distance of 100 km, the Moran statistic is very close to 0 once we account for spatial autocorrelation in the residuals. See Appendix D.

14 For the 26 haciendas whose locality we are unable to identify, we use the centroid of the municipality as the coordinates of the hacienda. The results are robust to not including the 26 haciendas.

15 Tables A2, A3, and A4 show the results in table form.

16 Below we also show the predicted literacy rates for nonhacienda municipalities (Figure 8a).

17 We observe differences in distance to nearest colonial city and median altitude (standard deviation ≥ 0.25). Yet our average treatment effects on the treated (ATT) do not vary by $\delta$, and they are similar to those using the neighbours sample (Tables C1 and C2).

18 Additionally, we replicate all previous analyses using the same modification. The results do not change significantly. For parsimony, we do not include the results but they are available upon request.

19 As before, the only exception is distance to nearest colonial city.

20 While the magnitude of the average treatment effect on the treated (ATT) varies by specification, this obtains the most restrictive estimates. Table C2 shows three other specifications using the mean and median distance in each sample: non-parametric Malahanobis with 2 and 3 neighbours – respectively, and bias-adjustment, and a propensity score matching (PSM) from a logit model with at least 1 neighbour. For the restricted neighbours sample, we estimate the ATT using the modified model with interactions.

21 The 1970 Population Census does not have complete information for Oaxaca at the municipality level. Therefore, we are not able to calculate the marginalization index in 1970 for this region. The results for 1980 and 1990 are presented in Figures B7 and B8.

22 In the first stage, a mediator model is estimated as a function of the treatment and the covariates. Two predictions for the mediator are obtained, one under the treatment and the other under the control. In our case, these correspond to the predicted proportion of urban localities, say, for municipalities within 30 km of hacienda and for those farther away. The second stage fits a regression model of the outcome as a function of the mediator, the treatment, and the covariates. The causal mediation effect corresponds to the average difference in the predicted outcome using the two different predicted values of the mediator.

23 Our results are similar, but smaller in magnitude compared to a standard OLS mediation analysis (See Table C8).

24 New minerals were extracted in the late nineteenth century, yet most new mines were located to the north of our area of study (Velasco Ávila et al. 1988).

25 There are only 14 municipalities with more than one railway station; the maximum is three.

26 We thank Sánchez-Talanquer (2017, 145) for sharing his data. Results are robust to using land petitions approved by the president between 1916 and 1976 from Sanderson (2013); not all approved petitions were executed. The RAN data documents grants executed.

27 See Wolf (2017) for an example in the Bajío.