Post-sampling crowdsourced data to allow reliable statistical inference: the case of food price indices in Nigeria

TL;DR

This paper proposes a post-stratification method to reweight crowdsourced data, enabling reliable statistical inference for food price indices in Nigeria, especially in areas where traditional data collection is challenging.

Contribution

It introduces a novel post-stratification approach to adjust crowdsourced data for valid statistical inference in developing countries.

Findings

Method effectively improves data reliability for food price indices.

Application in Nigeria demonstrates practical utility.

Reweighted data supports sound policy decisions.

Abstract

Sound policy and decision making in developing countries is often limited by the lack of timely and reliable data. Crowdsourced data may provide a valuable alternative for data collection and analysis, e. g. in remote and insecure areas or of poor accessibility where traditional methods are difficult or costly. However, crowdsourced data are not directly usable to draw sound statistical inference. Indeed, its use involves statistical problems because data do not obey any formal sampling design and may also suffer from various non-sampling errors. To overcome this, we propose the use of a special form of post-stratification with which crowdsourced data are reweighted prior their use in an inferential context. An example in Nigeria illustrates the applicability of the method.