Social Network Analysis of Peasant Agriculture: Cooperative Labor as Gendered Relational Networks

Abstract

Social network analysis is increasingly used by geographers to better understand network phenomena. This article employs a promising analytic approach for the study of relational networks in human geography—dyadic regression analysis (DRA)—to examine labor sharing networks among peasant households. Data from a floodplain community in the Peruvian Amazon are incorporated into a relational data set that represents all reported cooperative labor flows for agriculture over one year. Matrices capturing both link and node (household) attributes are analyzed to explain the formation of cooperative labor links and the intensity (and direction) of labor flows. Results reveal that cooperative labor, which entails work predominantly done by males, is strongly shaped by relations among women, as well as by kin networks and household structure. Our findings suggest the utility of DRA for the study of other relational networks of interest to geographers.

社会网络分析，逐渐被地理学者运用来更佳地理解网络现象。本文将一个前景看好的分析方法，运用于人文地理学中的关係性网络研究——对偶迴归分析（DRA）——以检视小农家户间的劳动力共享网络。本研究将秘鲁亚马逊流域中的一个洪泛平原社群搜集而来的数据，整合进一组代表一年中所有农务协力劳动力流动的关係性数据集之中。本研究分析同时捕捉连结与节点（家户）特徵的矩阵，以解释协力劳动连结的形成和劳动力流动的强度（与方向）。研究结果显示，招致主要由男性进行工作的协力劳动，强烈地由女性之间的关係、以及亲属网络和家户结构所形塑。我们的研究发现，为地理学者建议以对偶迴归分析研究其他关係性网络的效用。

El análisis de redes sociales es de uso creciente por geógrafos interesados en una mejor comprensión de fenómenos encadenados. Este artículo utiliza un enfoque analítico prometedor para el estudio de redes relacionales en geografía humana—el análisis de regresión diádica (DRA)—para examinar redes de trabajo compartido entre hogares campesinos. Los datos de una comunidad de las planicies inundables de la Amazonia peruana se incorporaron en un conjunto de datos relacionales que representan todos los flujos de trabajo cooperativo reportados para la agricultura durante un año. Se analizan las matrices que captan atributos de enlace y nódulo (hogar) para explicar la formación de lazos de trabajo cooperativo y la intensidad (y dirección) de los flujos laborales. Los resultados revelan que el trabajo cooperativo, referido a trabajo ejecutado predominantemente por varones, está fuertemente configurado por las relaciones entre las mujeres, lo mismo que por redes de parentesco y por la estructura hogareña. Nuestros descubrimientos apuntan a la utilidad del DRA para el estudio de otras redes relacionales de interés para los geógrafos.

Key Words:

• cooperative labor
• dyadic regression analysis
• gender relations
• peasant agriculture
• social network analysis

关键词:

• 协力劳动
• 对偶迴归分析
• 性别关係
• 小农农业
• 社会网络分析。

Palabras clave:

• trabajo cooperativo
• análisis de regresión diádica
• relaciones de género
• agricultura campesina
• análisis de redes sociales

Acknowledgments

We are grateful to the residents of Nuevo Miraflores who kindly participated in this study. Rinda Mozombite and Carlos Rengifo Upiachihua provided valuable assistance in the field. Sébastien Breau and Ben Cecil provided insightful comments on earlier versions on this work. We thank Barney Warf, the Editor, and three anonymous referees for their most helpful comments and suggestions.

Funding

Funding was provided by the Social Sciences and Humanities Research Council of Canada, the Fonds Quebeçois pour la Formation de Chercheurs et l’Aide à la Recherche (FCAR), McGill University, the University of Toronto, and the Japan Society for the Promotion of Science.

Notes

1 Actor network theory (ANT) is another important network approach used in geography. ANT, however, focuses less on structure and network connectivity than SNA (see Bosco [2006] for a review; see also Rocheleau and Roth [2007]).

2 Proximity has been a major focal point of research in economic geography (e.g., Bathelt 2005; Boschma 2005; Boschma and Frenken 2010; Mould and Joel 2010; Balland 2012; Broekel and Boschma 2012; Li, Bathelt, and Wang 2012), often recognizing different types of proximity and its positive and constraining effects.

3 Two households were excluded from the census; they were headed by a school teacher and a young fisherman.

4 Applying ordinary least squares (OLS) to Equation 1—incorrectly assuming an independent and identically distributed (i.i.d.) error term—yields inconsistent estimates, thus leading to incorrect inference.

5 The total number of possible ij pairs is N², but diagonals (self-relations) are omitted from the analysis. The corresponding conventional regression model would be Y_i = α + βX_i + u_i, where Y_i is a column vector and X_i is an (N × K) matrix.

6 This is the equivalent of using OLS to explain the likelihood of engaging in cooperative labor (linear probability model) and the amount of cooperative labor received in the conventional approach.

7 A similar approach has been employed by D’Exelle and Holvoet (2011) to study gender and network formation in a rural community in Nicaragua.

8 Due to data constraints, godparents (compadrazgo) and cross-gender relations are not considered.

9 Based on weighted data: mañaneo = 0.43 person * days, ayuda = 1 person * days.

10 To reduce the weight of outliers while avoiding losing all zero observations, we take the natural logarithm of (1 + Y_ij) as the dependent variable in Equation 3(3) , measuring effects on percentage change in the amount of cooperative labor received.

11 How continuous node-attribute variables can be constructed in other ways while maintaining the symmetry requirement is an empirical question (e.g., Broekel and Boschma 2012).

12 Although the estimated β₁ for male head born in the village is statistically significant in some specifications, the estimated combined effect is close to zero (β₁ + β₂ = –0.071 + 0.074 = –0.003, column 2).

13 Composite households give 8 and 21 percent more labor to other composite and nuclear households, vis-à-vis exchanges among nuclear households (column 5). Nuclear households give 12 percent more labor to composite households than they do to nuclear households (i.e., composite-to-composite: β₂/2 = 0.166/2 = 0.083; nuclear-to-composite: β₁ + β₂ = –0.042 + 0.166 = 0.124; and composite-to-nuclear: –β₁ + β₂ = – (–0.042) + 0.166 = 0.208).

14 The most marginal position in the cooperative labor network is occupied by household 21, which did not clear any fields during the period under study.

Additional information

Notes on contributors

Christian Abizaid

CHRISTIAN ABIZAID is an Assistant Professor in the Department of Geography & Program in Planning and in the School of the Environment at the University of Toronto, Toronto, ON, Canada M5S 3G3. E-mail: [email protected]. His research interests focus on peasant agriculture, rural livelihoods, social networks, and environmental change adaptation in the Amazon and Mexico.

Oliver T. Coomes

OLIVER T. COOMES is a Professor in the Department of Geography at McGill University, Montreal, QC, Canada H3A 0B9. E-mail: [email protected]. His research examines peasant livelihoods, poverty, and environmental change among traditional peoples of Amazonia and elsewhere in the Neotropics.

Yoshito Takasaki

YOSHITO TAKASAKI is a Professor in the Faculty of Humanities and Social Sciences, University of Tsukuba, Tsukuba, Ibaraki, Japan 305-8571. E-mail: [email protected]. His research interests include poverty, institution, social networks, rural livelihoods, and shifting cultivation.

Stéphanie Brisson

STÉPHANIE BRISSON is a former master's student in the Department of Geography at McGill University, Montreal, Canada H3A 0B9. E-mail: [email protected]. Her research addressed land and labor inequalities in the Peruvian Amazon. She now works as a social science advisor for a private consulting company.