How Big is Your Neighbourhood? Spatial Implications of Market Participation Among Filipino Smallholders

Abstract

We present a procedure for estimating two quantities defining the spatial externality in discrete-choice commonly referred to as 'the neighbourhood effect'. One quantity, the propensity for neighbours to make the same decision, reflects traditional preoccupations; the other quantity, the magnitude of the neighbourhood itself, is novel. Because both quantities have fundamental bearing on the magnitude of the spatial externality, it is desirable to have a robust algorithm for their estimation. Using recent advances in Bayesian estimation and model comparison, we devise such an algorithm and illustrate its application to a sample of northern-Filipino smallholders. We determine that a significant, positive, neighbourhood effect exists; that, among the 12 geographical units comprising the sample, the neighbourhood spans a three-unit radius; and that policy prescriptions are significantly altered when calculations account for the spatial externality.

Model Boosting for Spatial Weighting Matrix Selection in Spatial Lag Models

“…A common drawback of the model selection approaches is that the competing models need to be estimated, either explicitly (e.g. in Holloway and Lapar, 2007), or implicitly as a part of the testing procedure (e.g. in Kelejian, 2008).…”

Spatial Econometric Analysis: Potential Contribution to the Economic Analysis of Smallholder Development

“…Holloway and Lapar [37] published a seminal paper on neighbourhood effects, employing spatial econometric analysis. They estimated two quantities defining such a spatial externality: the propensity for neighbours to make the same decision and the magnitude of the neighbourhood.…”

“…First, there is a wide range of methods now available to conduct spatial econometric analyses suited to meeting a variety of research goals. Second, the volume of such analyses of smallholder production and rural development that have been undertaken using these methods remains small, notably in Southeast Asian countries with the prominent exceptions of [37,62] in the Philippines and [23] in Thailand. Third, panel data sets are crucial for the effective use of many of these methods.…”

“…Edirisinghe [20] examined the spatial interdependence of production choice by smallholder rubber producers in Sri Lanka. Drawing heavily on the modelling work by [37], he applied a Bayesian spatial autoregressive probit model to measure the impact of this production choice on the choice made by neighbouring smallholders.…”

The dynamic spatial multinomial probit model: analysis of land use change using parcel-level data

“…In the review of spatial econometric methods for bio-economic and land use modeling, mentioned several fresh studies involving spatial models for discrete response data. For example, Brasington and Sarama (2007) used a spatially correlated binary probit model to analyze deed restrictions in home sales, and Holloway and Lapar (2007) used similar techniques to estimate the magnitude of neighborhood effects.…”