The impact of monetary policy on unemployment hysteresis

Abstract

This article investigates the hypothesis that the extent to which hysteresis occurs in the aftermath of recessions depends on monetary policy reactions. The degree of hysteresis is explained econometrically by the extent of monetary easing during a recession and by standard variables for labour market institutions in a pooled cross country analysis using quarterly data. The sample includes 40 recessions in 19 Organization for Economic Co-operation and Development (OECD) countries for which the required data is available. The time period lasts for the period from 1980 to 2007. This article builds on Ball (1999) and extends the sample of countries, the time period under investigation and the set of control variables.

Keywords:

• monetary policy
• NAIRU
• structural unemployment
• hysteresis
• endogenous NAIRU

JEL Classification::

• E24
• E39
• E50

Acknowledgements

This article was in part written while Engelbert Stockhammer was visiting the Macroeconomic Policy Institute (IMK). The hospitality of the IMK and support from the Theodor Körner Fonds and FWF Project P18419-G05 are gratefully acknowledged. We thank Eckhard Hein, Erik Klär, Camille Logeay, Alfonso Palacio-Vera and Klara Zwickl for helpful comments. Remaining errors are ours.

Notes

¹ The near-rationality approach of Akerlof et al. (2000) and Akerlof (2007) leads to similar policy conclusions regarding the influence of monetary policy on the long-run Phillips curve.

² Additional explanations for hysteresis are summarized in Røed (1997).

³ For a critical discussion of OECD (1994), see OECD (2006), especially Chapter 6, and Blanchard and Katz (1997). For a critical discussion of IMF (2003), see Baker et al. (2004), Freeman (2005) and Baccaro and Rei (2007).

⁴ Theoretically, these results are based on the financial accelerator model of Bernanke and Gertler (1989) or on convex short-run aggregate supply curves (Peersman and Smets, 2001; Kuzin and Tober, 2004). Additionally, some authors focus on the psychological effects of the behaviour of monetary policy authorities in recessions (Blanchard, 2003).

⁵ The NBER Business Cycle Dating Committee ‘gives relatively little weight to real GDP because it is only measured quarterly and it is subject to continuing, large revisions’ (Business Cycle Dating Committee, 2001, p. 1). Unfortunately, their classification is only available for the USA.

⁶ In five cases (Belgium, Switzerland, Denmark, Spain and Portugal), only annual NAWRU data are available. Here, the starting point was chosen according to the quarter of the year in which the recession started. If the recession started in the first or second quarter of a year, the NAWRU from the year before to 5 years later was measured. If the recession started in the third or fourth quarter of the year, the NAWRU from this year to 5 years later was measured. As the NAWRU obtained by a Kalman filter is by design very smooth, for countries where quarterly data are available, the results with both procedures are virtually identical.

⁷ And in several cases output growth becomes positive after a recession for one, two or more quarters, and then returns to negative growth again. Therefore, the quarters after the recession also seem to be crucial.

⁸ Product market regulation is, strictly speaking, not a labour market institution. It is supposed to impact upon the NAIRU because it has an effect on firms’ mark up and is routinely included in unemployment regressions.

⁹ Detail descriptions of these data can be found in Appendix 2 of Bassanini and Duval (2006). The time series start with 1982 and are ending with 2003. In several cases, the ALMP series starts some years later and ends some years earlier. Where the observed recessions started before 1982 and in the case of ALMP, the closest available value is taken.

¹⁰ In the case of Ireland, the OECD data were complemented with data form the IMF International Financial Statistics for the quarterly short-term interest rate until 1983:Q4.

¹¹ The quarterly GDP data for Greece are unreliable according to the OECD Help Desk and no labour market institutions data are available. No quarterly GDP data is available for Austria. No data on labour market institutions is available for Luxembourg and Iceland. No quarterly data are available for Germany prior to 1991, as a consequence the German recession 1991/92 had to be excluded.

¹² Two observations had to be excluded. The degree of hysteresis is designed such that the variable lies between 0 and 1. Indeed, this is the case for most countries. However, there are some exceptions. The Finnish recession in 2001 shows a degree of hysteresis of −17.2. This is because the NAIRU was decreasing by 1.3 percentage points, while the unemployment rate rose slightly by 0.07 percentage points. Also, the Norwegian recession of 1980 is a statistical outlier as the monetary easing indicator lies more than three-and-a-half SDs under the mean. The Finnish case result from definitions used and thus illustrates the limitations of this approach.

¹³ As critical value for not rejecting the null hypothesis of no heteroscedasticity, we choose a probability value of 0.15.

¹⁴ In this, specifications consist of 41 observations, as Norway is no outlier according to this definition of real interest rates.