Estimation of the Unemployment Rate in Moldova - A Comparison of ARIMA and Machine Learning Models Including COVID-19 Pandemic Periods

Aug 3, 2026

This study estimates the unemployment rate in the Republic of Moldova with explicit attention to the COVID-19 pandemic shock. Two forecasting approaches are compared: the traditional ARIMA model and multiple machine learning models. Model performance is assessed using prediction accuracy metrics across pre-pandemic and pandemic periods, focusing on how each method responds to sudden changes in labor market dynamics. Findings suggest ARIMA effectively tracks broad trends, while machine learning models adapt more flexibly to crisis-induced shocks.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages7

Reading Time18 min

AccessFree to download

Introduction

Literature Review

Munich Personal RePEc Archive Estimation of the Unemployment Rate in Moldova: A Comparison of ARIMA and Machine Learning Models Including COVID-19 Pandemic Periods Vîntu, Denis National institute for Economic Research August 2025 Online at [https://mpra. ub. uni-muenchen. de/125941/](https://mpra. ub. uni-muenchen. de/125941/) [MPRA Paper No. 125941](MPRA Paper No. 125941) , [posted 29 Aug 2025 02:42 UTC](posted 29 Aug 2025 02:42 UTC) Estimation of the Unemployment Rate in Moldova: A Comparison of ARIMA and Machine Learning Models Including COVID-19 Pandemic Periods Denis Vîntu National Institute for Economic Research August 28, 2025 Abstract This study investigates the estimation of the unemployment rate in the Republic of Moldova, focusing on the impact of the COVID-19 pandemic. Two forecasting approaches are compared: the traditional ARIMA model and several machine learning models. The performance of these models is evaluated based on prediction accuracy metrics over pre-pandemic and pandemic periods. Results indicate that while ARIMA captures general trends effectively, machine learning models can better adapt to sudden shocks, such as those induced by the pandemic. Keywords: Simultaneous equations model; Labor market equilibrium; Unemployment rate determination; Wage-setting equation; Price-setting equation; Beveridge curve; Job matching function; Phillips curve; Structural unemployment; Natural rate of unemployment; Labor supply and demand; Endogenous unemployment; Disequilibrium model; Employment dynamics; Wage-unemployment relationship; Aggregate labor market model; Multivariate system estimation; Identification problem; Reduced form equations; Equilibrium unemployment rate Jel Classification: C30, C31, C32, C33, C51, J64, J65, J68 . 1 Introduction Unemployment is a critical indicator of economic health. Accurate estimation of unemployment rates helps policymakers design effective labor market policies. In Moldova, the labor market experienced significant disruptions during the COVID-19 pandemic, making forecasting more challenging. Traditional time series models, such as ARIMA, have been widely used for unemployment rate prediction, while recent studies suggest that machine learning (ML) methods may provide better adaptability to non-linear patterns and shocks. This paper aims to: 1. Estimate Moldova’s unemployment rate using ARIMA and ML models. 2. Compare the performance of these models, especially during COVID-19 . 3. Provide insights for policymakers regarding labor market trends. Accurate estimation and forecasting of unemployment rates are critical for effective labor market policy and economic planning. In the Republic of Moldova, the labor market has experienced significant fluctuations in recent years, exacerbated by external shocks such as the COVID-19 pandemic. These disruptions have led to sudden increases in unemployment, underemployment, and changes in labor force participation, highlighting the need for robust forecasting models that can adapt to both gradual trends and sudden economic shocks. Traditional econometric approaches, particularly the Autoregressive Integrated Moving Average (ARIMA) model, have long been used for time series forecasting due to their simplicity and effectiveness in capturing linear temporal patterns. However, such models often struggle to incorporate complex non-linear relationships and sudden structural breaks in the data, which became especially pronounced during the pandemic period. In contrast, modern machine learning models, including Random Forests, Support Vector Machines, and Long Short-Term Memory (LSTM) networks, offer the flexibility to capture non-linear dynamics and interactions between multiple variables. By leveraging these advanced methods, it is possible to achieve more accurate and responsive unemployment forecasts, particularly during periods of economic volatility. This study aims to estimate and compare the u

FAQs

What forecasting methods does the study compare for Moldova’s unemployment rate?

The study compares ARIMA with several machine learning models, including approaches designed to capture non-linear dynamics and shocks during the COVID-19 period.

How does COVID-19 affect the unemployment-rate forecasting task in this research?

The pandemic is treated as a major external shock that increases disruptions and sudden shifts in labor market conditions, making forecasting harder and highlighting model differences.

What conclusion does the study draw about ARIMA versus machine learning models?

Results indicate ARIMA captures general trends effectively, while machine learning models better adapt to abrupt crisis-induced shocks.

Aug 3, 2026

Upload by Valentina