Assessing the Impact of Non-Consideration of Stationarity on Regression Analysis: Evidence from Labor Market Data
Abstract
This study examines the empirical errors that may arise from ignoring stationarity properties when assessing the relationship between labor market indicators using regression analysis. The main objective of the research is to demonstrate, through empirical examples, the impact of the data preprocessing stage on regression results when working with time series and panel data. The analysis is based on open statistical data on the labor market of Uzbekistan, where wage and employment indicators are considered as panel data across regions for the period 2017–2024. Within the empirical framework, regression results obtained without accounting for stationarity are compared with those estimated using first-differenced variables. The findings indicate that constructing regressions directly on trending and non-stationary series may yield statistically significant results; however, such results can lead to substantively incorrect and unreliable conclusions.