Recessions and Potential Output: Disentangling Measurement Errors, Supply Shocks, and Hysteresis Effects*

• Author: Jonas Dovern,Christopher Zuber
• DOI: http://doi.org/10.1111/sjoe.12385
• Date: 01 October 2020
• Published date: 01 October 2020

Scand. J. of Economics 122(4), 1431–1466, 2020

DOI: 10.1111/sjoe.12385

Recessions and Potential Output:

Disentangling Measurement Errors, Supply

Shocks, and Hysteresis Effects*

Jonas Dovern

Friedrich-Alexander University Erlangen-N¨urnberg (FAU), DE-90403 N¨urnberg, Germany

jonas.dovern@fau.de

Christopher Zuber

Heidelberg University,DE-69115 Heidelberg, Germany

christopher.zuber@awi.uni-heidelberg.de

Abstract

This paper investigates expert revisions of potential output (PO) estimates followingrecessions.

Using data from the Organisation for Economic Co-operation and Development (OECD), we

show that downward revisions are substantial, permanent, and mostly driven by supply shocks.

In contrast, PO estimates do not significantly react to demand shocks. Revisions are also partly

caused by mismeasurement of PO before recessions. In particular, weshow that the length of the

preceding boom and pre-recession values of the current account balance and credit volumes are

correlated with post-recession PO revisions. Our results call for improvedmethods for estimating

PO and provide evidence against the existence of substantial hysteresis following demand

shocks.

Keywords: Hysteresis; OECD; output gap; potential output; trend

JEL classification:C52; E32

I. Introduction

Estimates of potential output (PO) are important for making decisions about

monetary and fiscal policy. Although PO estimates are meant to proxy the

level of economic output that is sustainable in the long run and independent

of cyclical (demand-driven) fluctuations, for many countries they have been

revised downwards in response to the Great Recession (see Benati, 2012;

Ball, 2014). This has led to renewed interest in the question of how sensitive

*We would like to thank Pia Pinger; the participants of seminars at Heidelberg University, the

University of Birmingham, and the ifo Institute; and the participants of the 13th Conference

on Real-Time Data Analysis hosted by the Banco de Espa˜na for their helpful comments. We

also thank two anonymous referees for their valuablecomments on an earlier draft of our paper.

J. Doverngratefully acknowledges financial support from the Ministry of Science, Research and

theArts of Baden-W ¨urttemberg. Part of the research was conducted while J.Dovern was affiliated

with the AlfredWeber Institute for Economics at Heidelberg University.He is also affiliated with

the CESifo research network.

C

The editors of The Scandinavian Journal of Economics 2019.

1432 Recessions and potential output

PO estimates are to severe economic downturns and which factors might

help to anticipate and avoid post-recession revisions of such estimates.

Against this background, this paper addresses the following research

questions. Do recessions have permanent effects on PO estimates? What are

the reasons behind downward revisions of PO estimates in the aftermath

of recessions? To empirically investigate these questions, we use a newly

compiled dataset with real-time vintages of the Economic Outlook (EO)

provided by the Organisation for Economic Co-operation and Development

(OECD), which allows us to trace the development of revisions to the level

of PO after recessions.

Avoiding PO revisions (and policy mistakes caused by them) in the

future requires a better understanding of the underlying reasons for PO

revisions. We distinguish between four main explanations as to why PO

estimates are revised downwards following a recession. If recessions are

caused by permanent supply shocks, the revisions are a reflection of

the lower-than-previously-expected long-run output path of the economy

(explanation 1). If recessions are caused by demand shocks, revisions of PO

estimates would indicate that the analysts believe that hysteresis effects lead

to permanent output effects (explanation 2). If PO is overestimated before

a recession, subsequent revisions of PO estimates are “merely” a correction

of previous measurement errors (explanation 3). Finally, causation could run

in the opposite direction, as suggested by Blanchard et al. (2015, 2017). In

this case, declining income expectations induced by the PO revision lead

to a fall in aggregate demand and a recession (explanation 4). Note that,

in this case, PO revisions should precede recessions.

Our paper presents three main findings regarding the revision of PO

estimates. First, we document substantial post-recession revisions of the

level of PO estimates that are predictable and partly result from pre-

recession estimation errors. We also show that revisions happen very

gradually over a period of up to five years. Second, we find little evidence

that PO revisions precede recessions. Finally, supply shocks systematically

lead to PO revisions while demand shocks prove to be unimportant for

revisions of the level of PO.

We contribute to a number of strands of the literature. First, our paper

adds to the literature that investigates the reliability of PO estimates.

Starting with Orphanides and van Norden (2002) and Orphanides (2003),

the unreliability of such estimates in real time is documented in many

studies (see Orphanides et al., 2000; Camba-Mendez and Rodriguez-

Palenzuela, 2003; Marcellino and Musso, 2011; Jacobs and van Norden,

2016), with Edge and Rudd (2016) presenting more optimistic findings for

output gap estimates published by the Federal Reserve Board since the

1990s. A number of contributions suggest broadening the information set

that is used to estimate PO to obtain more stable estimates. Garratt et al.

C

The editors of The Scandinavian Journal of Economics 2019.

J. Dovern and C. Zuber 1433

(2008) show how one can use information from different data vintages

and model data revisions explicitly to obtain more reliable PO estimates.

More recently, Borio et al. (2014, 2017) suggested using information

about the financial cycle to improve estimates. Our paper expands on this

body of literature as being the first study that systematically documents

the instability of estimates of PO levels after recessions as well as the

underlying reasons.

Second, our paper relates to other studies that analyze whether

recessions or financial crises affect potential output (estimates) or the

corresponding growth rates. Based on OECD real-time data for 23

countries, Ball (2014) shows that PO estimates remain permanently below

pre-recession trends after the Great Recession of 2008–2009.1Haltmaier

(2012) (using the Hodrick–Prescott (HP) filter) and Martin et al. (2015)

(using exponential trends) use PO estimates obtained ex post by filtering

the most recent data vintage. Both papers find that PO growth decreases

permanently following recessions. Using the production function approach

to estimate PO, Furceri and Mourougane (2012) document a similar effect

for the times after financial crises for a sample of 30 OECD countries.

Finally, Benati (2012) uses a structural vector autoregressive (VAR) model

to provide evidence that PO growth slowed down after the Great Recession

in the United States (US), the euro area, and the United Kingdom. So far,

none of these papers have used comprehensive real-time data on actual

estimates of PO levels. Such real-time data are however necessary to

understand the exact timings and causes of PO revisions.

The lack of use of real-time data is also a shortcoming of the third

strand of literature to which our paper contributes. This body of literature

analyzes whether recessions or financial crises affect actual output or its

growth rate. The most notable study in this context is by Cerra and Saxena

(2008), who show that output losses following financial or currency crises

were very persistent in the period 1960–2001. Other studies, such as Papell

and Prodan (2012) and Abiad et al. (2009), confirm these findings. Based

on data for 100 financial crises over the last 150 years, Reinhart and Rogoff

(2009, 2014) broaden the view and show that financial crises have negative

impacts on a wide range of variables, such as asset prices, employment,

and government debt. Finally, a number of studies (see Hosseinkouchack

and Wolters, 2013; Blanchard et al., 2015) provide evidence that regular

recessions also tend to permanently reduce the level of output.

In a wider context, our paper is related to the literature on

macroeconomic hysteresis effects, i.e., the notion that temporary shocks,

1Using PO estimates from real-time vintages of the International Monetary Fund (IMF) World

Economic Outlook, Fat´as and Summers (2018) provide evidence that fiscal consolidations

contributed to the decline of PO during this period.

C

The editors of The Scandinavian Journal of Economics 2019.