Economic Complexity and Environmental Impact using a Neural-Network Embedded Semiparametric Mixture of Experts Model

Sphiwe Skhosana (), Abeeb Olaniran (), Najmeh Nakhaei Rad () and Rangan Gupta ()
Additional contact information
Sphiwe Skhosana: Department of Statistics, University of Pretoria, Pretoria, South Africa. National Institute for Theoretical and Computational Sciences (NITheCS), Gauteng node, University of Pretoria, South Africa
Abeeb Olaniran: Department of Economics, University of Pretoria, Pretoria, South Africa
Najmeh Nakhaei Rad: Department of Statistics, University of Pretoria, Pretoria, South Africa. National Institute for Theoretical and Computational Sciences (NITheCS), Gauteng node, University of Pretoria, South Africa
Rangan Gupta: Department of Economics, University of Pretoria, Pretoria, South Africa

No 202618, Working Papers from University of Pretoria, Department of Economics

Abstract: The environmental Kuznets curve (EKC) postulates an inverted-U relationship between environmental quality and economic performance. Recent studies examine the EKC hypothesis across groups of countries (e.g., developing and OECD) by considering the impact of economic complexity (a measure of economic performance) on CO2 emissions (a measure of environmental quality). However, most of the studies impose, possibly, restrictive assumptions on the data: (1) all countries are assumed to follow the same developmental path and hence EKC; (2) the functional relationship between CO2 emissions and economic complexity is assumed to have a known parametric form. In this paper, we propose to relax these assumptions. First, we allow for the possibility that countries may follow different developmental paths by using the mixture of experts (MoE) approach. This gives rise to multiple EKCs. The MoE approach further allows us to endogenously identify different developmental paths and thus accounts for unobserved heterogeneity. Moreover, this approach allows us to probabilistically classify each country into any one of the obtained developmental paths. Second, we assume that the functional relationship between CO2 emissions and economic complexity is represented by a nonparametric (unknown) function of economic complexity. This allows the EKC the flexibility to take any form based on the data. Therefore, each expert regression model is a partially linear model (PLM). In contrast to existing PLMs, we allow the nonparametric part to be a multivariate function of any dimension. For model estimation, we propose a maximum likelihood estimation procedure through a hybrid algorithm that combines feed-forward neural networks (NN) with the classical expectation-conditional-maximization (ECM) algorithm, termed the ECM-NN algorithm. We use a simulation study to demonstrate the performance of this estimation procedure across various scenarios with different dimensions of the nonparametric function. Finally, we apply the proposed methods to a cross-country panel dataset of 64 countries spanning 2003 to 2022 and find two developmental paths in which the EKC holds.

Keywords: Environmental Kuznets Curve (EKC); Economic complexity; Expectation-conditional-maximization; Mixture of experts; Neural networks; Partially linear models (search for similar items in EconPapers)
JEL-codes: C14 C33 C45 Q56 (search for similar items in EconPapers)
Pages: 29 pages
Date: 2026-06
References: Add references at CitEc
Citations:

Downloads: (external link)
https://drupalwebprod-files.up.ac.za/Public/2026-0 ... PMY2AftCdP8q21B.RQHN (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:pre:wpaper:202618

Access Statistics for this paper

More papers in Working Papers from University of Pretoria, Department of Economics Contact information at EDIRC.
Bibliographic data for series maintained by Rangan Gupta ().