Spatial variation of (un)reliability in migration data as Schrödinger’s cat of demographic statistics in Croatia

Abstract

Numerous empirical findings suggest that migration data published by the Croatian Bureau of Statistics may lack reliability, particularly in underestimating the scale of emigration following Croatia’s accession to the European Union. Most evidence is based on comparisons between Croatian official data and those from receiving countries, indicating that Croatian statistics may systematically underestimate emigration trends. This paper considers the spatial dimension of (un)reliability in Croatian migration statistics. The analysis examines the differences in net migration rates between official statistics and vital-statistical method data across regional and local levels from 2011 to 2021. For both datasets, the average annual net migration rate was calculated, and the difference between them – used as the observed variable – was analysed using descriptive statistics and spatial analysis methods. The methodological framework enabled an assessment of the reliability of both migration and broader demographic data at lower spatial scales. Results show considerable spatial variation in data quality. Overestimation of net migration in official data, previously identified at the national level, is also observable at the county level. At the local level, it remains a dominant pattern, though its intensity varies across space. Around half of local units show acceptable levels of data reliability, while in one-quarter to one-third of cases, data quality remains questionable. Coastal areas and large cities stand out for lower reliability, largely due to intense migration dynamics, a high share of unregistered movements, and discrepancies between registered and actual residence – influenced by population behaviour. In some units, unreliable data reflect underestimated migration flows, pointing to potential weaknesses in census data as well. Findings offer a tool for identifying sources of data unreliability and determining whether local population figures are overestimated or underestimated. They also provide institutions with a framework for improving data collection systems and correcting records in areas of low reliability.