Nonlinearity in the Canadian and U.S. Labor Markets: Univariate and Multivariate Evidence From a Battery of Tests

Abstract

The nonlinearity of macroeconomic processes is becoming an increasingly important issue at both the theoretical and empirical levels. This trend holds for labor market variables as well. The reallocation theory of unemployment relies on nonlinearities. At the same time there is mounting empirical evidence of business cycles asymmetries. Thus the assumption of linearity/nonlinearity becomes crucial for the corroboration of labor market theories. This paper turns the microscope on the assumption of linearity and investigates the presence of asymmetries in aggregate and disaggregate labor market variables.The assumption of linearity is tested using five statistical tests for U.S. and Canadian unemployment rates and growth rates of the employment sectoral shares of construction, finance, manufacturing, and trade. An AR(p) model was used to remove any linear structure from the series. Evidence of nonlinearity is found for the sectoral shares with all five statistical tests in the U.S. case but not at the aggregate level. The results for Canada are not clear-cut. Evidence of unspecified nonlinearity is found in the unemployment rate and in the sectoral shares. Overall, important asymmetries are found in disaggregated labor market variables in the univariate setting. The linearity hypothesis was also examined in a multivariate framework. Evidence is provided that important asymmetries exist and a linear VAR cannot capture the dynamics of employment reallocation.

Regional and Sectoral Evidence of the Macroeconomic Effects of Labor Reallocation: A Panel Data Analysis

“…, where u i,t is the unemployment rate, following the suggestion by Wallis (1987) to employ the logistic transformation of the unemployment rate, a variable bounded between 0 and 1 (see also Panagiotidis and Pelloni 2007). 18 The aggregate series on the monetary variables, FR t and TB t , as well as the fiscal variables, G t , FG t , and SG t , are taken from the Federal Reserve Economic Data (FRED) database and transformed into first-differences (ΔFR t and ΔTB t ) and logarithmic first-differences (Δln(G t ), Δln(FG t ), and Δln(SG t )).…”

State Dependence in Labor Market Fluctuations

“…Our analysis relates to empirical and theoretical studies on the asymmetry of labor market fluctuations over the business cycle. On the empirical side, the works by Neftci (1984), Altissimo and Violante (2001), Panagiotidis and Pelloni (2007), Barnichon (2012), Abbritti and Fahr (2013), Barattieri et al (2014), Caggiano et al (2014), and Benigno et al (2015) show that unemployment and wages fluctuate differently across phases of the business cycles. Compared to these studies, we show that state dependence in labor market fluctuations is linked to the level of aggregate productivity, and we extend the analysis to job transition rates.…”

Non‐linear Predictability in G7 Stock Index Returns*

“…This nice property of the BDS test is further confirmed by the simulation experiments conducted by de Lima (1997). Guided by the above results, we set the window length to be 500 observations, which is relatively larger than previous studies applying the BDS test to economic time‐series data (see, for example, Panagiotidis and Pelloni, 2003, 2007; Ashley and Patterson, 2006). Moreover, Belaire‐Franch and Contreras (2002) argue that the issue of unreliability in small samples can be overcome by means of simple bootstrapping, and this is confirmed earlier by Patterson and Ashley (2000) whose Monte Carlo experiments demonstrate that the bootstrapped BDS test has the correct size and stands out in terms of overall power against various non‐linear alternatives even at sample length of 200.…”