Leading macroeconomic indicators of India

Abstract

Understanding leading indicator properties of macroeconomic variables has become essential newlinein wake of the Great recession. In particular, role of credit and impact of excessive credit newlinegrowth on real output has come to limelight for purpose of monitoring financial vulnerability newlineand understanding interaction between real and financial aspects of the economy. newlineMeasuring imbalance in credit and key macroeconomic variables requires an estimate of newlineunderlying trend and since these trend estimates are used for the purpose of leading indicators, newlineour concern is with the most recent values, computed in a reliable way using only the newlinedata available at the time. Since Hodrick-Prescott filter has been known to be unreliable at newlinethe end of sample, we test alternative methods of trend estimation. For this purpose, our newlinestudy compares two widely used filters in macroeconomics: symmetric HP filter and their newlinegeneralization to penalized splines versus asymmetric Henderson smoothers, classical and newlinethose based on embedding in RKHS (Reproducing Kernel Hilbert Spaces). These two set newlineof filters are evaluated in terms of mean square error, statistical revision error, forecast of newlinedirectional change and as indicators to issue early warning in signalling approach based on newlinecritical threshold. newlineThe thesis is organized as follows: We first study the inherent limitations of filtering methods newlinein terms of end of sample bias and also evaluate alternatives to HP filter on metrics of newlinequantitative, directional and revision error which are valuable in context of leading indicators. newlineThen signalling approach is used to find critical thresholds and horizons at which gaps newlinecalculated using key macroeconomic variables issue early warning for build up of financial newlineimbalances in India at 6, 12 and 18 months. Indicators and method of filtering are then newlineevaluated in terms of their usefulness, noise and signal properties. Analysis in frequency newlinedomain is then conducted which characterizes duration of credit and output cycles and investigates newlinedegree of synchronization