岭南观点·08 基于实时信息流的中国宏观经济不确定性测度

Abstract  Since the great recession of 2008,many researchers have shown interest in understanding economic uncertainty and its effects on the macro-economy. Economic uncertainty reflects the unpredictability of an economic system and is of great theoretical and empirical value for optimizing the asset allocation of households, developing reasonable economic policies, and predicting economic and financial risk. As macroeconomic uncertainty is not directly observable, many scholars focus on ways to measure it. Studies have adopted the perspectives of financial market volatility, forecast error, and text data of news information, among others. However, existing indices have drawbacks for measuring China's macroeconomic uncertainty. First, the commonly used uncertainty indices represented by VIX and EPU cannot fully measure macroeconomic uncertainty, as they mix macroeconomic uncertainty with other kinds of uncertainty. Second, prior studies either construct a quarterly uncertainty index based on quarterly GDP directly or give up the quarterly GDP and construct a monthly uncertainty index to improve timeliness. Obviously, the quarterly uncertainty index cannot track the economic condition in a timely manner, and ignoring the GDP data will result in the loss of valuable information. Third, existing macroeconomic uncertainty indices update monthly or quarterly, and cannot reflect the economic condition in a timely manner. Finally, the related literature constructs the uncertainty indices based on final data and does not consider the data revision problem. The data also suffer from other problems, such as periodic missing, ragged-edge data. If we ignore these problems, then the final data used by scholars will be different from the real data used by policy-makers. In this paper, we follow Scotti(2016)to construct China's macroeconomic surprise index and uncertainty index. We choose seven economic variables:the year-on-year growth rates of quarterly GDP, quarterly urban disposable income, monthly industrial value-added, monthly investment in fixed assets, retail of social consumer goods, total imports and exports, and tax revenue. We evaluate the contributions of each economic variable on the economic condition using Zheng & Wang's(2013) year-on-year mixed frequency dynamic factor model. We then use it to calculate the weighted sum of the forecast error, and we further construct the surprise and uncertainty indices. Compared with the existing uncertainty index,  our surprise and uncertainty indices have the following advantages. First, we consider the important role of GDP and extract common factors based on the mixed frequency dynamic factor model. Second, the index is daily, as it updates every time new information about the economic condition releases. We not only consider the data revision problem but also allow the various variables to release at different times. Third, we compute the weights representing the contribution of the economic indicators to the business condition. These weights are time-varying and depend on the time elapsed. Finally, the constructed uncertainty index is the weighted accumulation of historical forecast errors rather than single-period forecast errors. We apply the constructed indices to study the effects of economic uncertainty on the interest rate market and the business cycle. Specifically, our application to the business cycle shows that when the economy is highly volatile, the deterioration of the economic situation will increase macroeconomic uncertainty. When the economy is stable, macroeconomic uncertainty will affect the economic situation.The data on the macroeconomic surprise and uncertainty indices are daily. The indices update once new information becomes available due to the newly released data. These indices can therefore reflect the latest information on economic activities and enable timely and reliable evaluation of the economic situation, which is valuable for the determination of macroeconomic policies and the implementation of enterprises' development strategy. Timely measurement of economic uncertainty is of great significance for macro control at the national level and micro decision-making at the market level. With the collection of more real-time data and the development of the high-dimensional mixed-frequency dynamic factor model, we will be able to extend our project to big data by incorporating more economic variables.