Nowcasting Model with Dynamic Factors to estimate Ecuador's Quarterly Real GDP
In Ecuador the publication of the Gross Domestic Product (GDP) lags approximately by three months. For this reason, we propose the implementation of nowcasting methodology using dynamic factors that use data published with less periodicity to estimate variables with greater periodicity. In this pape...
| Autores: | , |
|---|---|
| Tipo de documento: | artigo |
| Estado: | Versão publicada |
| Data de publicação: | 2021 |
| País: | Ecuador |
| Recursos: | Banco Central del Ecuador |
| Repositório: | Revista Cuestiones Económicas |
| Idioma: | espanhol |
| OAI Identifier: | oai:estudioseconomicos.bce.fin.ec:article/291 |
| Acesso em linha: | https://estudioseconomicos.bce.fin.ec/index.php/RevistaCE/article/view/291 |
| Access Level: | Acceso aberto |
| Palavra-chave: | Nowcasting Dynamic factor models Gross domestic product Factores dinámicos Producto interno bruto |
| Resumo: | In Ecuador the publication of the Gross Domestic Product (GDP) lags approximately by three months. For this reason, we propose the implementation of nowcasting methodology using dynamic factors that use data published with less periodicity to estimate variables with greater periodicity. In this paper, two Nowcast models are presented (EMOE - 31 variables, 5 factors and 5 lags; and, Aggregate Sales - 25 variables, 4 factors and 5 lags). Additionally, the COVID-19 represents a new challenge in the estimation capacity of these models in the short term because it produced breaks or temporary changes in some economic indicators, for which adjustments were made (the inclusion of a dummy in the quarter with the sharpest fall in GDP; and the use of the GDP cycle to smooth the downturn of the economy, followed by a re-estimation of the models and adjustments of the results to the growth rate using error correction models) to counteract the temporary effects that the fall in GDP could have on subsequent estimates. The evaluation of the predictive capacity of the models suggests that the proposed Nowcast models have a better out-of-sample fit than the ARIMA(4,3) and ARIMAX models. |
|---|