Working Paper: CEPR ID: DP10575
Authors: Margherita Comola; Marcel Fafchamps
Abstract: Many studies have used self-reported dyadic data without exploiting the pattern of discordant answers. In this paper we propose a maximum likelihood estimator that deals with mis-reporting in a systematic way. We illustrate the methodology using dyadic data on inter-household transfers from the village of Nyakatoke in Tanzania, investigating the role of wealth in link formation. Our results suggest that observed transfers are grounded in mutual self-interest, and we show that not taking reporting bias into account leads to incorrect inference and serious underestimation of the total amount of transfers between villagers. The method introduced here is applicable whenever the researcher has two discordant measurements of the same dependent variable.
Keywords: dyadic data; informal transfer; reporting bias; social networks
JEL Codes: C13; C51; D85
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Wealth (D31) | Observed Transfers (F16) |
| Misreporting (C59) | Underestimations of Total Transfers (H29) |
| Mutual Self-Interest (D71) | Transfer Dynamics (F16) |
| Wealth (D31) | Likelihood of Reporting a Transfer (F16) |
| Misreporting (C59) | Reporting Behavior (C92) |
| Wealth Interaction (D31) | Transfer Likelihood (F16) |