Uncertainty and Labour Force Participation*
| Date | 01 April 2021 |
| DOI | http://doi.org/10.1111/obes.12392 |
| Published date | 01 April 2021 |
| Author | Idriss Fontaine |
437
©2020 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 83, 2 (2021) 0305–9049
doi: 10.1111/obes.12392
Uncertainty and Labour Force Participation*
Idriss Fontaine†
†Department of Economics (CEMOI), Universit´edeLaR´eunion, 15 Avenue Ren´eCassin,
Saint-Denis 97400, Reunion (e-mail: idriss.fontaine@univ-reunion.fr.)
Abstract
Can an increase in uncertainty generate a fall in labour force participation? This paper
supports an affirmative answer to this question. Using a structural vector autoregressive
model, I show that: (i) a surge in uncertainty leads to a decrease in participation; (ii)
uncertainty shocks account for a considerable proportion of participation fluctuations; and
(iii) uncertainty shocks have in recent years contributed non-marginally to participation
variations. To explain these facts, I build a New Keynesian dynamic stochastic general
equilibrium model with a frictional labour market, endogenous labour force and stochastic
volatility.A model with flexible prices induces the expansionary effects of uncertainty on
labour force participation. By contrast, a model with price stickiness can reproduce negative
empirical comovements.As firms postpone hiring, the household follows a discouragement
effect and responds by reducing the size of the labour force. This paper suggests that in a
context of variable labour supply, labour market frictions alone do not suffice to reproduce
the negative effects of uncertainty shocks. It also confirms the importance of rigid prices
as an amplification mechanism for uncertainty shocks and suggests that the modelling of
labour force participation has some implications on the transmission channel of uncertainty.
I. Introduction
In comparison to other post-WWII recessions, the Great Recession of 2008 and the sub-
sequent slow recovery were accompanied by a substantial increase in uncertainty. Policy
makers and economic journalists often argue that uncertainty was a key factor behind the
sluggish rebounds of the years 2009–10. This macroeconomic behaviour lies at the origin
of a growing academic interest in uncertainty. The seminal contribution of Bloom (2009)
JEL Classification numbers: E32, E24, J21.
*I am deeply grateful to Alexis Parmentier for his invaluable guidance and encouragement. I am also thankful for
comments from Fran¸cois Langot, Fabien Tripier,Thepthida Sopraseuth, Antoine Lepetit and participants to the CE-
MOI seminar,the October 2016 TEPP conferencein La Reunion, the 21st Theories and Methods in Macroeconomics
conference in Lisbon, the 2017 Asian Meeting of the Econometric Society in Hong Kong, the 67th Congress of the
FrenchEconomic Association (AFSE) in Paris and the 2018 North American Meeting of the Econometric Society in
Davis. I also address a special thank to David Nortes-Martinez for a proofreading of the paper. Finally, I thank the
associate editor of Oxford Bulletin of Economics and Statistics and three anonymous referees for their enlightening
suggestions that greatly improved the quality of this article. All errors are myresponsibility.
438 Bulletin
and the subsequent publications of Fern´andez-Villaverde et al. (2011), Bachmann, Elstner
and Sims (2013), and Basu and Bundick (2017), among others, indicate that heightened
uncertainty impedes economic activity. A strand of this growing body of literature draws
attention to the consequences of higher uncertainty on the labour market. Empirically, the
work of Caggiano, Castelnuovo and Groshenny (2014) suggests that uncertainty shocks
lead to a significant increase in unemployment during recessions. Theoretically, the con-
tribution of Leduc and Liu (2016) indicates that a surge in uncertainty increases the firms’
option value of ‘wait and see’. As firms react by posting fewer vacancies, unemployment
unambiguously increases. To the best of my knowledge,none of these papers considers the
participation margin, instead remaining silent about a possible influence of uncertainty on
labour force participation (LFP).
However, the modelling and consequences of uncertainty shocks in a medium-scale
Dynamic Stochastic General Equilibrium (DSGE) model are still being debated. Basu and
Bundick (2017) argue that a model with nominal rigidities and variable labour supply but
without search and matching frictions in the labour market is sufficient for theory to repli-
cate empirical facts. However, the contemporaneous work of Leduc and Liu (2016) shows
that price stickiness, while remaining an important amplification element, is inessential in
a model with a frictional labour market and fixed labour supply. The present paper shows
that the conclusion of Leduc and Liu (2016) is sensitive to the assumption of fixed labour
supply.In par ticular, in my model with search and matching frictions and endogenous par-
ticipation to the labour market, both wealth and substitution effects drivethe LFP response
in opposite directions so that labour market frictions alone are not sufficient to replicate
the empirical facts.
Traditionally, the macroeconomic literature has not modelled the participation margin,
because empirical evidence points to an acyclical pattern of LFP over the business cycle.
Such a view is now challenged by at least two recent facts. First, Elsby, Hobijn and S¸ ahin
(2015) demonstrate that entries and exits from non-participation account for one-third of
cyclical fluctuations in unemployment. Second, the downward trend in the US LFP rate
accelerated after the Great Recession, a period characterized by a high levelof uncer tainty.1
The purpose of this paper is to connect these two concepts. In other words, what do the data
tell us about the LFP response to an uncertainty shock? To what extent does uncertainty
impact the participation rate in the US? And what is the transmission channel of uncertainty
shocks when a modelled economy explicitly includes an endogenous LFP margin?
To provide answers to these questions, I focus on the empirical link between LFP and
uncertainty. In particular, along the lines of Bloom (2009) and Jurado, Ludvigson and Ng
(2015), I estimate the joint dynamics of uncertainty and LFP within a structural vector
autoregression (VAR) model with eight macroeconomic variables. As a proxy for uncer-
tainty, I use the macroeconomic uncertainty index constructed by Jurado et al. (2015). I
follow the bulk of the literature by isolating an uncertainty shock with a Cholesky-type
decomposition. The empirical evidence is as follows. An unexpected increase in uncer-
tainty leads to a u-shaped response of the LFP rate. The LFP response is non-significant
during the first year after the shock but is thereafter significantly negative. The Forecast
1BLS measures indicate that the participation rate fell by about 2% points during the first quarter of 2008 and the
second quarter of 2011. Furthermore, Erceg and Levin (2014) found that the bulk of LFP variations during the Great
Recession can be explained by cyclical factors.
©2020 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
Uncertainty and LFP 439
Error Variance Decomposition (FEVD) of participation reveals that uncertainty shocks
explain almost 40% of its variance. Interestingly, the historical decomposition shows that
uncertainty shocks play a large role in explaining LFP variations during the 2003–12
period. A battery of checks, namely by changing the measure of uncertainty, participation
variable, or identification scheme or by including additional variables in the VAR, confirms
the robustness of the empirical evidence.
After presenting the empirical evidence, I adopt a theoretical perspective to rationalize
the transmission channel of uncertainty. On the basis of Leduc and Liu (2016), I build a
New Keynesian DSGE model enriched with search and matching frictions in the labour
market and stochastic uncertainty volatility regarding the level of aggregate productivity.
However, I depart from Leduc and Liu (2016) by explicitly modelling the LFP margin.Then,
the DSGE model is simulated and the related impulse response functions are computed
by employing a third-order perturbation method.2Based on this framework, I gradually
incorporate different elements into the model. To verify whether the option-value channel
is sufficient to obtain the recessionary effects of uncertainty shocks, as suggested by Leduc
and Liu (2016), I first consider a model with flexible prices. Under this scenario, higher
uncertainty triggers precautionary saving, leading the household to supply more labour.
The behaviour of the representative household is thus dictated by an added-worker effect
with overall expansionary effects for the model economy. Given that a framework with
exogenous LFP provides the opposite results, my finding suggests that the conclusion of
Leduc and Liu (2016) about the role of labour market frictions as a sufficient mechanism
to recover the recessionary effects of uncertainty shocks is not robust to the inclusion of a
variable participation margin.
I then consider a model with price stickiness. This framework allows theoretical co-
movements, whichare in line with those obser vedin the data. Following Basu and Bundick
(2017), I find that the demand channel helps to greatly amplify the transmission of un-
certainty shocks. Most prominently, heightened uncertainty prevents firms from resetting
prices, while they have to cut-off production to meet the depressed demand. This mech-
anism clearly triggers the recessionary effects of uncertainty shocks in the economy and
leads to a decline in future firms’ profits, with the consequence of fewer vacancies. As job
search activities have a lower probability of success, the household participation’s value
decreases, ultimately leading to a fall in participation. In such a framework, household
behaviour is driven by a discouragement effect.
The sensitivity analysis of the theoretical results indicates that LFP response is closely
linked to: (i) the internal degree of habit persistence, (ii) the amount of search frictions,
(iii) the presence of wage rigidity and (iv) the value of elasticity parameters. In particular,
habit persistence magnifies the negative effect of uncertainty shocks, because it implies a
further decline in the match value. Regarding the labour market specification, mysensitivity
analysis reveals that when search frictions are low, household behaviour is less linked to
the real option value, which is at the root of the fall in LFP. Consequently, in a context of
low search frictions, the added-worker effect dominates any discouragement effect. I also
confirm the intuition of Cacciatore and Ravenna (2015) by showing that in addition to the
2As discussed below, a third-order perturbation of the model is required to evaluatethe effects of volatility shocks.
©2020 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
To continue reading
Request your trial