Dynamics of unemployment insurance claims

The weekly unemployment insurance initial claims (UIIC) is a timely indicator of labor market conditions in the United States.  As such, UIIC has served as an important component of the index of leading indicators and can potentially play a valuable role in macroeconomic forecasting models.  Despite its importance, there is an absence of a critical evaluation of the nature of UIIC measurement, a rigorous exploration of its time series properties, and a careful analysis of its relation with other macro variables. 

This paper has three objectives: First, it provides a critical evaluation of UIIC measurement and the extent to which it represents the conditions in the labor market.  Second, it examines the univarite time series properties of UIIC using data over the longest available period covering 1967-2012.  Our preliminary results suggest that UIIC follows a stationary but highly seasonal pattern.  Also, the distribution of UIIC exhibits asymmetry and leptokurtosis, suggesting the possibility of volatility clustering and time-varying volatility.  Thus, we model the conditional mean and conditional volatility of UIIC, and examine its out-of-sample forecasting performance.  Modeling of the conditional mean follows the univariate ARIMA(p,d,q)x(PDQ)s procedure, which accounts for cyclical and seasonal autoregressive and moving average components.  As for its conditional variance, we consider a number of alternative volatility models including autoregressive conditional heteroskedasticity (ARCH), generalized ARCH (GARCH), exponential GARCH (EGARCH), asymmetric power ARCH (APARCH), and fractionally integrated GARCH (FIGARCH).  The results of this exercise are used to answer the following research questions: (1) Does the UIIC series exhibit time-varying volatility, and if so, does volatility respond symmetrically to positive and negative shocks?  (2) How persistent are shocks to conditional volatility? Do they dissipate an exponential rate consistent with covariance stationary GARCH-variety models, or at a slow hyperbolic rate as in the FIGARCH model?  We also examine the out-of-sample forecasting performance of these models. 

Third, the paper examines the dynamics of UIIC using a multivariate vector autoregressive (VAR) model which includes measures of conditions in the credit market (bank loans, long-term interest rates, Chicago Fed national financial conditions), monetary policy (money stock), and foreign exchange market (the exchange rate).  The results of this investigation are expected to provide additional insights on the interaction of conditions in the labor market with those of credit and foreign exchange markets.