Three essays on econometric analysis of the Chinese macroeconomy

China has consistently achieved high and stable rates of economic growth since 1970s.China is now the second biggest economy in the world, and its goods and services are highly competitive in international markets. These renewed academic interest in understanding the country's macroeconomic policy-making. Macroeconomic forecasting and monitoring are established practices that are vitally important for real-world macroeconomic policy-making. Having this in mind, the literature on macroeconomic forecasting and monitoring in China is limited compared to the research in Western economies. As such, this thesis consists of three individual chapters that are aimed towards contributing to the empirical investigation of macroeconomic forecasting in China. Chapter 2 aims to evaluate the quality of China's official macroeconomic statistics by examining three factors. The first is how political interference may affect the statistical reporting system. At local government level, the career incentives and the Chinese cadre evaluation system, alongside the geography-based governing logic, have motivated local officials to compete to influence the reported growth rate. At central government, the notion of independence of the National Bureau of Statistics of China is criticised because it has limited authority over the statistical divisions of other government institutions and provincial bureau of statistics. Second, it reviews the ongoing challenges of gathering, measuring and presenting economics, with focuses on incomplete survey data, issues with direct reporting systems and revisions of economic data. Third, it investigates the internal inconsistency by using quantitative methods to explain where discrepancies come from. Chapter 3 studies forecasting Chinese macroeconomic variables using large-scale factor models with mixed-frequency data and missing observations component. The fact1ormodels are particular compatible with potential data contamination, rapid institutional and structural change, which are prevalent in China. Using 251 monthly variables and 34quarterly variables over the December 2001 to June 2018 period, we find statistical evidence that mixed-frequency factor models, especially mixed-frequency factor-augmented vector autoregressive models, generated superior forecasts to the univariate and multivariate models for price series, nominal investment and nominal consumption, except for the CPI inflation rate and nominal consumption at one month ahead. Therefore, the results of this chapter provide clear guidance and important implications for academics, practitioners and the public who are interested in macroeconomic forecasting in China. Chapter 4 undertakes the task of nowcasting GDP for mainland China using machine learning algorithms. Using a large set of quarterly macroeconomic indicators and monthly indicators, we train eight popular machine learning algorithms and nowcast GDP growth for each quarter over the 1993Q1-2018Q2 period. We compare the predictive accuracy of these nowcasts with those of AR model and dynamic factor model in the state-space representation. We use the model confidence set to obtain a set of best model(s) with10% level of confidence. Our results show that shrinkage methods are covered by the model confidence set and therefore are in the set of best models. As such, ML algorithms proved useful for improving the accuracy of nowcasting the Chinese GDP growth rate. Overall, this thesis enriches our understanding of the quality of Chinese official macroeconomic data and guides practitioners toward selecting the appropriate forecasting and now casting models for China's economy in a data-rich environment and provides considerable scope for future research.