This study investigates whether the integration of indicators derived from Google Trends can en- hance the predictive performance of traditional econometric models for Italian inflation, during the 2004–2024 period. Nine thematic groups of search queries were constructed, and through principal component analy- sis, two complementary indices were developed: one capturing direct attention to inflation (Indice Inflazione GT), and another reflecting interest in broader, economically related topics (Indice Te- matico GT). These indicators were incorporated into a SARIMAX(1,0,1)(0,0,0,12) model specified on the first difference of the seasonally adjusted NIC index. The analysis includes Granger causality tests, structural break detection via Chow tests, and an eva- luation of in-sample and out-of-sample predictive performance using standard metrics and dedicated statistical tests (Diebold-Mariano, Clark-West). The findings reveal bidirectional causality between Google Trends indicators and inflation, with the inflation-focused GT index leading the NIC by 4 to 16 months. Contrary to expectations, the war-related shock of March 2022 does not constitute a permanent structural break; rather, the most significant events are the outliers observed in January and October 2022. Clark-West tests indicate statistically significant improvements in forecasting accuracy at 6- and 12-month horizons (p-value < 0.05), accompanied by a marginal reduction in out-of-sample RMSE. Unlike prior studies comparing alternative indicators, this research assesses the marginal contribu- tion of Google Trends indicators within a structured econometric framework. The results suggest that information derived from online search behavior can provide useful leading signals for inflation forecasting over horizons relevant to monetary policy.
Questo lavoro valuta se l’integrazione di indicatori derivati da Google Trends possa migliorare la capacità predittiva dei modelli econometrici tradizionali per l’inflazione italiana, con particolare riferimento al periodo 2004-2024. Sono stati costruiti nove gruppi tematici di query di ricerca, e attraverso l’analisi delle componenti principali, sono stati costruiti due indicatori complementari: un indice focalizzato sull’attenzione diretta all’inflazione e un indice che cattura l’interesse per tematiche economiche correlate. Questi indicatori sono stati integrati in un modello SARIMAX(1,0,1)(0,0,0,12), specificato sulla differenza prima dell’indice NIC destagionalizzato. L’analisi include test di causalità di Granger, identifi- cazione di rotture strutturali mediante test di Chow, e valutazione delle performance predittive in-sample e out-of-sample attraverso metriche standard e test statistici specifici (Diebold-Mariano, Clark-West). L’analisi ha fatto emergere una causalità bidirezionale tra gli indicatori Google Trends e l’inflazione, con l’indice di inflazione GT che anticipa il NIC di 4-16 mesi. Contrariamente alle aspettative, lo shock bellico di marzo 2022 non costituisce una rottura strutturale permanente; gli eventi significa- tivi sono invece gli outlier di gennaio e ottobre 2022. I test di Clark-West dimostrano miglioramenti statisticamente significativi della capacità predittiva per orizzonti di 6 e 12 mesi (p-value < 0.05), con una riduzione marginale del RMSE out-of-sample. A differenza di studi precedenti che confrontano indicatori alternativi, questa ricerca valuta il con- tributo marginale degli indicatori Google Trends all’interno di un modello econometrico strutturato. I risultati suggeriscono che l’informazione derivata dai comportamenti di ricerca online può forni- re segnali anticipatori utili per la previsione dell’inflazione su orizzonti temporali rilevanti per la politica monetaria.
Predire l’Inflazione Italiana con Google Trends: Una Nuova Metodologia
MERICI, TOMMASO
2024/2025
Abstract
This study investigates whether the integration of indicators derived from Google Trends can en- hance the predictive performance of traditional econometric models for Italian inflation, during the 2004–2024 period. Nine thematic groups of search queries were constructed, and through principal component analy- sis, two complementary indices were developed: one capturing direct attention to inflation (Indice Inflazione GT), and another reflecting interest in broader, economically related topics (Indice Te- matico GT). These indicators were incorporated into a SARIMAX(1,0,1)(0,0,0,12) model specified on the first difference of the seasonally adjusted NIC index. The analysis includes Granger causality tests, structural break detection via Chow tests, and an eva- luation of in-sample and out-of-sample predictive performance using standard metrics and dedicated statistical tests (Diebold-Mariano, Clark-West). The findings reveal bidirectional causality between Google Trends indicators and inflation, with the inflation-focused GT index leading the NIC by 4 to 16 months. Contrary to expectations, the war-related shock of March 2022 does not constitute a permanent structural break; rather, the most significant events are the outliers observed in January and October 2022. Clark-West tests indicate statistically significant improvements in forecasting accuracy at 6- and 12-month horizons (p-value < 0.05), accompanied by a marginal reduction in out-of-sample RMSE. Unlike prior studies comparing alternative indicators, this research assesses the marginal contribu- tion of Google Trends indicators within a structured econometric framework. The results suggest that information derived from online search behavior can provide useful leading signals for inflation forecasting over horizons relevant to monetary policy.File | Dimensione | Formato | |
---|---|---|---|
Thesis_Final_A.pdf
accesso aperto
Dimensione
4.61 MB
Formato
Adobe PDF
|
4.61 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14247/25668