The Beveridge Curve Across US States: New Insights From a Pairwise Approach
| Published date | 01 April 2020 |
| Author | Mark J. Holmes,Jesús Otero |
| Date | 01 April 2020 |
| DOI | http://doi.org/10.1111/obes.12337 |
405
©2019 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 82, 2 (2020) 0305–9049
doi: 10.1111/obes.12337
The Beveridge Curve Across US States: New Insights
From a Pairwise Approach*
Mark J. Holmes† and Jes ´
us Otero‡
†School of Accounting, Finance and Economics, University of Waikato, Hamilton, New
Zealand (e-mail: mark.holmes@waikato.ac.nz)
‡Facultad de Econom´ıa, Universidad del Rosario, Bogot´a, Colombia
(e-mail: jesus.otero@urosario.edu.co)
Abstract
This paper offers newinsights into Beveridge curve analysis by modelling the unemployment–
vacancy rate relationship at state-level within a pairwise environment in which the unem-
ployment rate in one state is inversely related to the vacancy rate in another. We find that
Beveridge curve shifting, or matching efficiency, is driven by factors that include distance
between states, the labour force participation rate, homeownership and the relative afford-
ability of housing between states. A pairwise recursive analysis points to a decrease in
matching efficiency in the period that followed the Great Recession.
I. Introduction
Observation across US states points to a wide variation in the unemployment rate.Accord-
ing to the Bureau of Labor Statistics, the 2018Q2 unemployment rate ranged from 6.1%
experienced by Louisiana down to 2.5% in Iowa. Also, the ranking of states and spread
of unemployment rates across states has changed considerably over time. For example,
the gap between the highest and lowest state unemployment rate at the end of the Great
Recession in 2009Q2 was 10.8% points based on the difference between Michigan and
North Dakota. Wide outcomes such as these may in part represent different labour market
frictions in the sense that some states are better at matching workers to employment than
others. One approach to studying labour market frictions is Beveridge curve analysiswhich
is based on the relationship between unemployment and vacancies. The Beveridge curve
has links to the Phillips curve (via the excess demand for labour) and it has its roots in a
framework that views job creation or employment as the outcome of a matching process of
vacancy-filling bythe unemployed as well as those already in employment, and bythose not
in the labour force. An extensive Beveridge curve literature providing useful insights into
the employment-matching process already exists. This literature includes notable survey
JEL Classification numbers: J6, J61, J63, C33.
*Weare g rateful to an anonymousreviewer and participants at the 2019 meeting of the New Zealand Association
of Economists for helpful comments and suggestions. The usual disclaimer applies.
406 Bulletin
articles by Blanchard and Diamond (1989); Petrongolo and Pissarides (2001); Nickell et al.
(2003); Elsby et al. (2015). Various themes of particular interest havebeen explored in this
literature including the nature, stability and elasticity of the Beveridge curve relationship
and the associated issues of cyclical and structural influences.The ‘cyclical-structural’ issue
is of special importance given its potential policy implications. Petrongolo and Pissarides
(2001) conclude that while there might be empirical support for a range of variables causing
the Beveridge curve to shift (such as the participation rate, replacement ratio, real wages
and unemployment duration), the microeconomic support for these shifts is tentative only.
In this paper, we investigate the relationship between unemployment and vacancies at
state-level for the US with a view to addressing a number of key research questions from a
regional perspective. Given the presence of significant labour mobility between US states,
what are the implications for our understanding of the US Beveridge curve? How does the
relative efficiency of employment-matching across US states respond to drivers such as
home-ownership, housing affordability and labour force participation? How has the nature
of the Beveridge curve changed over time? The above-mentioned survey articles point to
an extensive literature that analyses the Beveridge curve. Most of these studies proceed on
the basis of analysing national-level data which can mask much of whatgoes on regionally
in terms of interaction between the unemployedand vacancies. We contribute to the limited
literature on US state-level matching where formal investigations into regional or state-
level Beveridge curves is a relatively under-researched area.1However, a constraining
factor in analysing regional Beveridge curves has been data limitation. While obtaining
state-level unemployment data can be relatively straightforward, this is perhaps not the
case for the consistent measurement of job vacancies across US states and over time.With
this in mind, we provide a first analysis using a balanced panel of US state-level vacancy
data that are recorded as help wanted advertised online. In analysing state-level matching,
we argue that it is important to take explicitly into account a matching process for a given
state that draws upon the degree of integration or cohesiveness with other states. Moreover,
employment-matching not only involves local unemployment and local vacancies,but also
vacancies in other states too. We therefore propose and employ a new pairwise approach
to analysing regional Beveridge curves based on all possible state-level combinations of
the unemployment and vacancy rates.This provides us with a fuller picture of employment
matching across US states.
A further contribution of our study to the literature is in terms of using the pairwise ap-
proach to investigatethose factors that might cause changes in pairwise matching efficiency,
or the pairwise Beveridge curves to shift. Further to the earlier country-level studies by
Jackman, Pissarides and Savouri (1990); Bleakley and Fuhrer (1997); Groenewold (2003);
Dutu, Holmes and Silverstone (2016), we consider the impact from state-levellabour force
participation rates, relative housing affordability, home-ownership as well as the distance
1Regional Beveridgecur vestudies include, inter alia, Dixon et al. (2014) for the case of Australia, Samson (1994);
McPherson and Flores (2012) for Canada, Bouvet (2012) for Europe, Borsch-Supan (1991); Kosfeld et al. (2008)
for Germany, Destefanis and Fonseca (2007) for Italy and Wall and Zoega (2002) for the UK. These studies have
investigated the direction and timing of the shifts in regionalBeveridge curves where, for example, changes in labour
market rigidities, long-term unemployment, as wellas cyclical shocks have been found to be responsible for outward
shifts in the case of European Beveridge curves. In the case of Canada, there are regional differences insofar as
the generosity of the employment insurance has been found to be an important factor in the West,while the mix of
structural variables appear to be important for central and eastern Canada.
©2019 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
To continue reading
Request your trial