Bansi Malde

Methods to Identify Linear Network Models: A Review

In many contexts we may be interested in understanding whether direct connections between agents, such as declared friendships in a classroom or family links in a rural village, affect their outcomes. In this paper, we review the literature studying econometric methods for the analysis of linear models of social effects, a class that includes the ‘linear-in-means’ local average model, the local aggregate model, and models where network statistics affect outcomes. We provide an overview of the underlying theoretical models, before discussing conditions for identification using observational and experimental/quasi-experimental data.

Credibly identifying social effects: accounting for network formation and measurement error

Understanding whether and how connections between agents (networks) such as declared friendships in classrooms, transactions between firms, and extended family connections, influence their socio-economic outcomes has been a growing area of research within economics. Early methods developed to identify these social effects assumed that networks had formed exogenously, and were perfectly observed, both of which are unlikely to hold in practice. A more recent literature, both within economics and in other disciplines, develops methods that relax these assumptions. This paper reviews that literature. It starts by providing a general econometric framework for linear models of social effects, and illustrates how network endogeneity and missing data on the network complicate identification of social effects. Thereafter, it discusses methods for overcoming the problems caused by endogenous formation of networks. Finally, it outlines the stark consequences of missing data on measures of the network, and regression parameters, before describing potential solutions.

Family networks and infant health promotion: a mixed-methods evaluation from a cluster randomised controlled trial in rural Malawi

Objective Parents may rely on information provided by extended family members when making decisions concerning the health of their children. We evaluate whether extended family members affected the success of an information intervention promoting infant health. Methods This is a secondary, sequential mixed-methods study based on a cluster randomised controlled trial of a peer-led home-education intervention conducted in Mchinji District, Malawi. We used linear multivariate regression to test whether the intervention impact on child height-for-age z-scores (HAZ) was influenced by extended family members. 12 of 24 clusters were assigned to the intervention, in which all pregnant women and new mothers were eligible to receive 5 home visits from a trained peer counsellor to discuss infant care and nutrition. We conducted focus group discussions with mothers, grandmothers and peer counsellors, and key-informant interviews with husbands, chiefs and community health workers to better understand the roles of extended family members in infant feeding. Results Exposure to the intervention increased child HAZ scores by 0.296 SD (95% CI 0.116 to 0.484). However, this effect is smaller in the presence of paternal grandmothers. Compared with an effect size of 0.441 to 0.467 SD (95% CI −0.344 to 1.050) if neither grandmother is alive, the effect size was 0.235 (95% CI −0.493 to 0.039) to 0.253 (95% CI −0.529 to 0.029) SD lower if the paternal grandmother was alive. There was no evidence of an effect of parents’ siblings. Maternal grandmothers did not affect intervention impact, but were associated with a lower HAZ score in the control group. Qualitative analysis suggested that grandmothers, who act as secondary caregivers and provide resources for infants, were slower to dismiss traditionally held practices and adopt intervention messages. Conclusion The results indicate that the intervention impacts are diminished by paternal grandmothers. Intervention success could be increased by integrating senior women.