How efficient are China's macroeconomic forecasts? Evidences from a new forecasting evaluation approach

Surveys, which are designed to collect data and include professional forecasts of macroeconomic variables, are of great interest to central banks and various institutions. The formal evaluation of such forecasts has attracted considerable attention in literature. However, empirical studies focusing on China are limited. This paper firstly proposes a new approach to test forecasts' accuracy and efficiency under asymmetric loss function by using a unique dataset from surveys conducted by the State Administration of Foreign Exchange in China. It is found that forecasters of the four macroeconomic variables (i.e., GDP growth rate, CPI, exports and imports) partially utilize new information and publicly available information under LINEX loss function when they update the forecasts, which are similar to those under quadratic loss function. Besides, a new finding in this paper is that over-smoothing hardly exists in the forecasts of GDP growth rate and CPI in China, which is different from the results in developed countries. Our findings suggest clear areas of opportunity to improve the accuracy of the forecasts, such as considering the negative autocorrelation found in forecast revisions of CPI.