Implementing Machine Learning Methods in Estimating the Size of the Non-observed Economy

Labib Shami, Teddy Lazebnik

Research output: Contribution to journalArticlepeer-review

2 Scopus citations

Abstract

Even though the literature on unregistered economic activity is growing at an increasing rate, we commonly encounter simple ordinary least squares methods and panel regressions, largely ignoring the recent rapid developments in machine learning methods. This study provides a new approach to more accurately estimate the size of the non-observed economy using machine learning methods. Compared to two currency demand-based models used to estimate the size of the non-observed economy, we show that a Random Forest algorithm can more accurately estimate the demand for currency, which is known to provide a fair estimation of the unregistered economic activity. The proposed approach shows superior forecasting capabilities compared to the current state-of-the-art linear regression-based methods dedicated to estimating non-observed economic activity.

Original languageEnglish
Pages (from-to)1459-1476
Number of pages18
JournalComputational Economics
Volume63
Issue number4
DOIs
StatePublished - Apr 2024
Externally publishedYes

Bibliographical note

Publisher Copyright:
© 2023, The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature.

Keywords

  • Demand for money
  • E26
  • E41
  • H26
  • Informal economy
  • Machine learning in economics
  • O17
  • Shadow economy
  • Tax evasion and avoidance

Fingerprint

Dive into the research topics of 'Implementing Machine Learning Methods in Estimating the Size of the Non-observed Economy'. Together they form a unique fingerprint.

Cite this