Title: Applying Individual Growth Models to Analyze Regional Development Trends: A Multilevel Longitudinal Approach
Authors: Salih, Sufian Munther
Volume: 9
Issue: 4
Pages: 401-414
Publication Date: 2025/04/28
Abstract:
Individual growth models are increasingly applied in urban and regional planning as powerful statistical tools to analyze dynamic changes over time. This study aims to present and evaluate individual growth models in their various forms and estimation techniques, and to apply them in assessing spatial development trends across multiple years. The analysis is based on data collected from a random sample of 829 geographic units, where development indicators (such as infrastructure investment or population growth) were tracked over a four-year period (2010/2011 to 2013/2014). Additional contextual variables-such as population density and urban/rural classification-were also included. The dataset represents a two-level longitudinal structure: level one reflects within-unit variation over time, while level two captures differences between units (e.g., neighborhoods, districts, or regions). To explore these dynamics, four growth models were fitted and their parameters estimated: 1. The unconditional means model. 2. The unconditional linear growth model. 3. The conditional linear growth model. 4. The nonlinear growth model. Results from the unconditional means model showed that 28% of the variance in development indicators was attributable to differences between geographic units, highlighting the necessity of considering multilevel data structures in regional planning analysis. Moreover, incorporating time (years) alongside second-level predictors (such as population density and classification) in the conditional linear growth model enhanced model fit, explaining 30% of the variance in average development and 24% of the variance in growth/change rates over time. When comparing the quality of fit of the four growth models using the Model Deviance (D), BIC, and AIC measures, it became clear that the non-linear growth model had the lowest quality of fit, achieving the highest values for all three measures. The quality of fit of the growth model improves with the addition of explanatory variables, whether at the first or second level, that have a significant impact on the dependent variable.