Nowcasting of Service Sector by Using Traffic Counting Data in Iran

Predictable and unpredictable delays in the national accounts data dissemination in Iran highlight the nowcasting of the economy’s state with using timely and high-frequency data. The large share of service sectors in GDP make forecasting of this sector more important. This paper seeks to answer the question of whether the status of the service and transportation sector can be predicted by using the vehicle traffic count dataset. In this regard, daily data on 2590 points of the country's roads from 2015 to September 2021 is used. In addition to using a simple aggregation method to construct the index, Artificial Neural Network model (ANNs) and Bayesian Model Averaging (BMA) are also used. The results show that the estimation indices extracted from these data have less forecast error than the benchmark models (ARMA) and can represent changes in both services and transportation sectors. The comparison of different methods of index construction shows the index extracted from Neural Network model has less error than other methods.