Nowcasting consumer sentiment in Turkey: A mixed frequency data regression approach

This study proposes a novel framework to establish the relationship between consumer sentiment and four important high-frequency variables in an emerging economy. In this regard, autoregressive distributed lag mixed data sampling (ADL-MIDAS) regression models are used to obtain monthly forecasts for two consumer confidence indices (CCIs) in Turkey. This approach yields an adjustable setting by considering the contribution of daily data to forecasting of a monthly variable. Specifically, by exploiting information from highly frequency variables like the exchange rate, stock exchange, oil price and gold price now casting consumer sentiment by the flexible data-driven weighting scheme is achieved for two different time periods and with two different estimation schemes. The data used in this study is publicly available and covers the time interval between 2002 and 2017. The data sources include the Bloomberg HT Consumer Confidence Index, the Central Bank Republic of Turkey Consumer Confidence Index, the exchange rates of the U.S. dollar and Euro (http://evds2.tcmb.gov.tr/), gold price from the world gold council, oil price from the U.S. Energy Information System, and stock price. Empirical results show that models with exchange rate variables outperform others for both time intervals. Due to Turkey’s economic structure, the exchange rate is the most important channel affecting consumer confidence. Additionally, subsample including the global financial crisis is also used to understand the influence of turbulent times on consumer sentiment. It is shown that, for crisis periods, out-of-sample results are very promising leading one to the finding that consumer confidence index adapts quickly to this turbulent period. Accordingly, both CCIs receive similar signals from these high-frequency variables and help us to understand the current stance of the economy even in turbulent times. By using pre-crisis period data, we are able to understand economic conditions and household behavior during global economic and financial crises establishing the forecasting power of sentiment indices for the future, even during volatile time intervals.