In Search of Better Labor-Market Indicators: Assessment of Inflation Predictability (in Korean)

Applying the Stock and Watson's (1999) Phillips curve forecasting model to the Korean quarterly data (1985Q1-2011Q4), we assess the performances of the conventional employment and alternative labor-market indicators for forecasting CPI inflation over the last 10 years (2002Q1-2011Q4). As for the alternative indicators, we select a few concepts of extended unemployment (U6 and the underemployed), main determinants of real marginal cost (real wage, unit labor cost, and labor share of income), and some labor-flow indicators (Shimers (2012) conditional steady-state unemployment, BSI (labor force), and hiring or quit/layoff rate, etc.). Main findings are as follows: the conventional employment indicators, such as the official unemployment rate, do not improve the model's forecasting performance significantly when compared with the naive AR model. Contrarily, we find that most of the alternative indicators?such as U6, the underemployed, real wage, BSI (labor force), quit/layoff rate, and the steady-state unemployment?significantly enhance the forecasting power of the otherwise traditional Phillips curve model. This improvement in forecasting performances is more striking when we forecast headline CPI rather than core CPI, and the indicators forecast inflation jointly rather than individually. The better performance of the alternative indicators is mainly due to their leading correlation with inflation and real activity as well as their theoretical superiority as a measure of resource utilization or labor market tightness. Moreover, most of those alternative indicators also exhibit more prominent pro-cyclicality than the official unemployment rate does.