Abstract / Description
Timely identification of business-cycle turning points is essential for policymakers, firms and investors, yet traditional macroeconomic indicators are often published with considerable delays. This study examines whether vehicle registrations constitute an effective Leading Economic Indicator (LEI) for the Spanish economy and proposes an integrated framework combining business-cycle analysis, monthly forecasting and daily nowcasting. The cyclical properties of registrations are analyzed using the Christiano–Fitzgerald band-pass filter, cross-correlation analysis and the Bry–Boschan dating algorithm. Monthly forecasting performance is evaluated by comparing TBATS and Prophet through rolling-origin cross-validation, while a novel machine learning-inspired distribution model (MD) is developed to transform monthly forecasts into daily estimates. The results show that passenger-car registrations anticipate industrial production by three to five months, with a ninety per cent bootstrap interval of one to six. TBATS consistently outperforms Prophet in one-step-ahead forecasting. An out-of-sample test confirms that lagged registrations reduce the forecast error of industrial production without recourse to two-sided filtering, although the reduction is modest in magnitude and is not found for quarterly GDP. The proposed MD-shallow model achieves the highest distributional accuracy, surpassing both Prophet and a persistence benchmark. The findings demonstrate that vehicle registrations are a valuable standalone Leading Economic Indicator and that integrating traditional economic indicators with advanced forecasting techniques improves real-time monitoring of automotive demand and business-cycle dynamics. The proposed framework provides a practical tool for policymakers, manufacturers, and researchers seeking earlier and more reliable assessments of economic activity.