Andorra high-frequency indicators and GDP real-time forecast

High-FrequencyIndicators and GDP Mariarosaria Comunale SIP/2026/039 IMF Selected Issues Papers are prepared by IMF staff asbackground documentation for periodic consultations withmember countries.It is based on the information available at 2026MAY IMF Selected Issues Paper European Department High-Frequency Indicators and GDP Nowcasting in Andorra, Principality of AndorraPrepared by Mariarosaria Comunale Authorized for distribution by Jeff Danforth IMF Selected Issues Papersare prepared by IMF staff as background documentation for periodicconsultations with member countries.It is based on the information available at the time it was ABSTRACT:High-frequency signals for Andorra were combined into a timely real-time estimate of quarterlyGDP growth using a mixed-frequency nowcasting model. The resulting framework delivers a transparent, RECOMMENDED CITATION:Comunale, Mariarosaria (2026). High- Frequency Indicators and GDPNowcasting in Andorra. IMF Selected Issues Paper, SIP/2026/039, International Monetary Fund, Washington SELECTED ISSUES PAPERS High-Frequency Indicators and Principality of Andorra Prepared by Mariarosaria Comunale HIGH-FREQUENCY INDICATORS AND GDP High-frequency signals for Andorra were combined into a timely real-time estimate of quarterly GDPgrowth using a mixed-frequency nowcasting model. The resulting framework delivers a transparent, A.Introduction 1.Timely information on economic activity is essential for monitoring the business cycleand informing policy and risk assessments.However, national accounts are released with a delayimplying that the most recent, or current, quarter of GDP is typically observed only after asubstantial lag. By contrast, there is a wide range of higher-frequency indicators (e.g., tourism flows, 2.A survey of available high-frequency indicators suggests they can be effective inunderstanding short-term trends.Main drivers of the economy on a gross value-added (GVA)basis include tourism, construction, real estate, and retail. An illustrative example of few keyindicators is shown in Figure 1. For tourism and retail, the number of overnight tourists hasincreased in 2025 and retail sales in 2025 to date are broadly in line with past data. Imports relatedto retail trade and hospitality point to continued strength in GVA through the end of 2025. Imports 3.Selected high frequency indicators are combined with historical national accounts datato develop a nowcasting model of quarterly GDP growth.The focus of the nowcasting exerciseis on estimating GDP growth in the current quarter, which is partially observed through B.Methodology 4.We combine these heterogeneous signals into a timely real-time measure of quarterlyGDP growth through a mixed-frequency nowcasting framework.We adopt a mixed-frequencydynamic factor model (DFM) estimated in state-space form (See Appendix and de Resende, 2024and Linzenich and Meunier, 2024). The key idea is that a small number of latent common factors cansummarize co-movements across a large panel of indicators, while idiosyncratic componentscapture series-specific variation. To avoid overfitting and improve interpretability, the initial indicator 5.We correct the model for large volatilities in the data.The COVID-19 episode representsa particularly important challenge for nowcasting models because it introduced large, transitorydisruptions and sharp changes in volatility that can dominate parameter estimates and inflateforecast uncertainty if not properly treated. We therefore consider a COVID-aware strategy that 6.The nowcasting dataset combines monthly and quarterly indicators.The estimationsample spans 2015m01 to 2025m11, and the data are retrieved from Haver Analytics or directly fromthe Andorran Department of Statistics. This window is chosen to ensure both the largest timedimension possible and a fully populated estimation panel, i.e. all series used in the baselinespecification are available without internal missing observations over the estimation period (no C.Selection of Leading Indicators 7.The results from the BMA selection are broadly consistent with prior expectationsregarding the key drivers of short-term activity.The indicator selection step prioritizes variablescapturing external demand and tourism-related dynamics, notably Spain GDP growth and measures for health services and electricity consumption of hotel and restaurants enter among the selectedseries, suggesting that consumption- and tourism-intensive components are central to the currentnowcast signal. Variables directly related to construction activity are not selected in the baselinescreening stage. As an additional accuracy check, we therefore augment the information set with 8.It is important to note that indicator selection in a BMA-based nowcasting frameworkis inherently data- and vintage-dependent.As new information becomes available, or when theexercise is conducted for different quarters (either in real time or retrospectively), the se