🤖 AI Summary
This study addresses real-time forecasting of U.S. quarterly GDP growth, with particular emphasis on model robustness under the COVID-19 shock. Leveraging the FRED-MD dataset of high-frequency monthly macroeconomic indicators, it systematically compares dynamic factor models (DFMs) against four neural network architectures—multilayer perceptrons (MLPs), one-dimensional convolutional neural networks (1D CNNs), long short-term memory (LSTM) networks, and gated recurrent units (GRUs)—while rigorously examining the impact of input sequence length on forecast accuracy. Key contributions include: (i) the first demonstration that 1D CNNs achieve superior cross-cycle robustness and generalization performance in real-time macroeconomic forecasting, outperforming both alternative deep learning models and DFMs; and (ii) empirical evidence that excessively long input sequences (>6 quarters) degrade generalization—especially during pandemic-induced volatility—thereby challenging the conventional “more historical data is better” assumption.
📝 Abstract
In our study, we apply different statistical models to nowcast quarterly GDP growth for the US economy. Using the monthly FRED-MD database, we compare the nowcasting performance of the dynamic factor model (DFM) and four artificial neural networks (ANNs): the multilayer perceptron (MLP), the one-dimensional convolutional neural network (1D CNN), the long short-term memory network (LSTM), and the gated recurrent unit (GRU). The empirical analysis presents the results from two distinctively different evaluation periods. The first (2010:Q1 -- 2019:Q4) is characterized by balanced economic growth, while the second (2010:Q1 -- 2022:Q3) also includes periods of the COVID-19 recession. According to our results, longer input sequences result in more accurate nowcasts in periods of balanced economic growth. However, this effect ceases above a relatively low threshold value of around six quarters (eighteen months). During periods of economic turbulence (e.g., during the COVID-19 recession), longer training sequences do not help the models' predictive performance; instead, they seem to weaken their generalization capability. Our results show that 1D CNN, with the same parameters, generates accurate nowcasts in both of our evaluation periods. Consequently, first in the literature, we propose the use of this specific neural network architecture for economic nowcasting.