Unemployment rate hysteresis and the great recession: exploring the metropolitan evidence

Analysis of the hysteresis hypothesis is not confined to unit-root testing. Alternative approaches include, among others, fractional integration models (Gil-Alana and Henry 2003), Markov-switching models (Bianchi and Zoega 1998), and threshold models (Coakley et al. 2001).

We note that the mainstream macroeconomic literature differs somewhat from the recent econometric literature on the use of the term “persistence.” In the former case, persistence refers to the speed of adjustment of a macroeconomic process to economic shocks and is generally measured by the sum of the coefficients in an autoregressive process, which is assumed to be I(0). In contrast, the literature on breaks in persistence concerns switches in the order of integration of the process.

A few analyses use state-level data in conjunction with panel unit-root tests. Song and Wu (1997), using the Levin et al. (2002) panel unit-root test, find that hysteresis does not characterize the unemployment rate dynamics of the US states. León-Ledesma (2002) reaches similar conclusions, using the Im et al. (2003) panel unit-root test. Cheng et al. (2012), on the contrary, employing the PANIC method that permits cross-sectional dependence between the US states find strong evidence of hysteresis in state-level data, especially when the tests include the new data from the recent Great Recession. Clemente et al. (2005) use national, regional, and state-level data to construct panels for the nine divisions and four regions considered by the US Census. They provide evidence against a unit root for the US economy and most of the US states. The evidence against a unit root weakens when considering the Census nine divisions, and even weaker when considering the four Census regions. They conclude, therefore, suggesting that the time-series properties of the unemployment rate may depend, among other things, on the assumed level of disaggregation. García-Cintado et al. (2015) find strong support for the hysteresis hypothesis in Spanish regional unemployment rates. Lanzafame (2012) rejects the hysteresis hypothesis in Italian regional unemployment rates. Fallahi and Rodriguez (2011) investigate the degree of persistence in the Canadian provinces allowing for structural breaks and find evidence against the hysteresis hypothesis.

Although the issue of seasonal adjustment does not appear to have received attention in the context of tests of change in persistence, the issue may prove even more important in this context, since the tests relate to the long-run properties of the data. Halunga et al. (2009) also use both seasonally adjusted and seasonally unadjusted monthly data to analyze changes in inflation persistence in the UK.