Plot‐level technical efficiency accounting for farm‐level effects: Evidence from Chilean wine grape producers

Abstract

This paper applies alternative panel data models to a cross-sectional dataset that contains observations at the plot level for a sample of wine-grape farms in Central Chile. The input-output data as well as key attributes of the production system are at the plot level, at which individualized management exists. However, plots belonging to a particular farm are also subject to overall centralized (farm-level) management. Thus, this data configuration offers the possibility of analyzing technical efficiency (TE) both across plots and across farms. A Generalized True Random Effects model, which permits the separate identification of farm-level and plot-level inefficiency while controlling for unobserved farm-level heterogeneity, shows that TE varies across farms but not among plots within the same farm. Geographical location also affects grape production and agro-climatic conditions influence production levels, with grape farms located on cooler zones producing significantly less than their counterparts in warmer zones, as expected. The analysis underscores the value of using recent methodologies typically applied to panel data when cross-sectional information is available for individual plots within a farm unit or in similar settings.

Total factor productivity and technical efficiency among smallholder groundnut farmers in Northern Mozambique

“…variable elasticities of substitution), but violate global curvature properties. In addition, C-D and translog estimates typically provide similar TE estimates (Baccouche & Kouki 2003;Ogundari 2014;Bravo-Ureta et al 2020).…”

“…We consider a production model in which traditional agricultural inputs -land, labour and seedsare combined to produce groundnuts. The main approach to productivity analysis taken here assumes that firms maximise expected profits, which provides the rationale for estimating production frontier models where inputs are predetermined, thus avoiding the simultaneity bias issue (Zellner et al 1966;Karagiannis & Kellermann 2019;Bravo-Ureta et al 2020). Maximum likelihood estimation (MLE) is the preferred methodology used to fit stochastic production frontiers (SPFs) (Greene 2003).…”

Transient and persistent efficiency: an application to Portuguese wineries

“…These changes present unseen challenges to traditional wine producers, who must adapt by reinforcing their efficient use of resources. Bravo-Ureta et al ( 2020) include a literature review on technical efficiency in the wine industry The decomposition of productive efficiency is limited to few contributions (Bravo-Ureta et al, 2020;Adom &Adams, 2020), with further theoretical and empirical research being demanded, particularly benefiting from panel datasets covering highly heterogeneous and competitive markets, such as the Portuguese wine industry.…”

“…Taking the Portuguese wine industry as an example, specifically the aims are: (i) to estimate productive efficiency of Portuguese wineries, decomposing into transient and persistent components; (ii) to generate knowledge towards an efficient improvement path to stakeholders; and (iii) to provide research guidelines within the topic. Transient and persistent inefficiency have been analyzed for the wine sector by Bravo-Ureta et al (2020). However, they used cross-sectional data, where the panel structure referred to farm-level and plot-level observations".…”

Rural infrastructure for inclusive rural transformation in developing countries: case studies on roads,markets and irrigation from Bangladesh and the Philippines

“…21 The Cobb-Douglas and transcendental logarithm (or translog) have been the two most commonly used functional forms for this type of analysis (Bravo-Ureta et al, 2007;Ogundari, 2014). While TE results from these two forms have been found to be very similar in different contexts (e.g., Baccouche and Kouki, 2003;Bravo-Ureta et al, 2020c) we opt for the Cobb-Douglas because it satisfies regularity conditions imposed by microeconomic theory globally and this is not the case for the translog (Ahmad and Bravo-Ureta, 1996;O'Donnell, 2012;O'Donnell, 2016). The choice of inputs was based on the literature, contextual knowledge and data availability.…”