This project analyses and predicts Scottish GDP growth using a range of macroeconomic indicators, including labour market conditions, productivity, inflation, oil prices, and exchange rates. Multiple modelling approaches were applied. This included linear regression, lagged models, decision trees, and a simple ARIMA model to compare both explanatory power and forecasting performance.
The results showed me that productivity (output per job growth) is the most consistent and significant driver of GDP growth, with retail activity and external factors such as oil prices also playing a role in certain models. While regression models provided strong interpretability and reasonable in-sample performance, the decision tree model struggled out-of-sample, highlighting the risks of overfitting in small macroeconomic datasets.
Overall, the analysis demonstrates that while statistical models can provide valuable insights into economic drivers, forecasting GDP remains challenging due to limited data, structural shocks (black swan events), and economic complexity. The project emphasises the importance of combining data-driven modelling with critical economic judgement when supporting policy and decision-making.