Modelling of Economic and Financial Conditions for Real‐Time Prediction of Recessions*
| Published date | 01 June 2021 |
| Author | Cem Çakmakli,Hamza Dem I˙rcani,Sumru Altug |
| Date | 01 June 2021 |
| DOI | http://doi.org/10.1111/obes.12413 |
663
©2020 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 83, 3 (2021) 0305–9049
doi: 10.1111/obes.12413
Modelling of Economic and Financial Conditions for
Real-Time Prediction of Recessions*
Cem C¸ akmakli,†Hamza Dem˙
ircan‡ and Sumru Altug§,¶
†Ko¸c University, Istanbul, Turkey (e-mail: ccakmakli@ku.edu.tr)
‡Central Bank of the Republic ofTurkey, Istanbul,Turkey (e-mail: hamza.demircan@tcmb.
gov.tr)
§American University of Beirut, Beirut, Lebanon
¶CEPR, London, UK (e-mail: sa287@aub.edu.lb)
Abstract
In this paper, we propose a method for real-time prediction of recessions using large sets
of economic and financial variables with mixed frequencies.This method combines a dy-
namic factor model for the extraction of economic and financial conditions together with
a tailored Markov regime switching specification for capturing their cyclical behaviour.
Unlike conventional methods that estimate a single common cycle governing economic
and financial conditions or extract economic and financial cycles in isolation of each other,
the model allows for a common cycle which is reflected with potential phase shifts in
the financial conditions estimated alongside with other parameters. This, in turn, provides
timely recession predictions by enabling efficient modelling of the financial cycle system-
atically leading the business cycle. We examine the performance of the model using a
mixed frequency ragged-edge data set for Turkey in real time. The results show evidence
for the superior predictive power of our specification by signalling oncoming recessions
(expansions) as early as 3.6 (3.0) months ahead of the actual realization.
I. Introduction
Monitoring business activity for anticipating economic downturns in a timely manner is of
key importance for economic agents. To this end, various econometric methods have been
proposed to generate indicators of business conditions using large data sets.These typically
involve modelling the co-movement of a large number of variables using econometric
models to extract joint behaviour, namely, factor models.As is well known, factor models
yield behaviour conformable with the notion of a common cycle with distinct dynamics in
economic expansions and contractions.
The recent global recession together with its underlying financial roots have made
understanding the impact of financial conditions on real activity a key requirement for
JEL Classification numbers: C11, C32, C38, E37.
*Cem C¸ akmaklı, acknowledgesthe financial support of the AXA Research Fund. This project was supported by
TUBITAKGrant No. 109K495. We thank to Marco del Negro, Sylvia Kaufmann, Gary Koop, John Maheu and Mike
West for valuablecomments and suggestions. Any remaining errors are our own.
664 Bulletin
timely prediction of business cycle turning points. Therefore, in addition to conventional
economic variables, several financial series that are available in real time have emerged
as key indicators in factor models for measuring business conditions. Typically, there are
two polar cases for the use of economic and financial variables in econometric models
to extract the common behaviour in these series. On the one hand, conventional practice
typically merges economic variables, often released with a delay, with timely information
on financial variables to extract the single common cyclical behaviourin real time, see, for
example, Aruoba, Diebold and Scotti (2009); Doz and Petronevich (2016). On the other
hand, economic and financial indicators are measured in isolation of each other, leading to
distinct cycles for economic and financial conditions. See, for example, Chauvet and Piger
(2008); Camacho, Perez-Quirosand Poncela (2018), among others, for timely measurement
of economic conditions, and Hatzius et al. (2010); Koop and Korobilis (2014); Galati et al.
(2016), among others, for measuring financial conditions.
However, neither joint nor independent modelling of cyclical behaviour in economic
and financial variables might be appropriate for timely predictions of economic downturns.
Financial conditions typicallyare closely tied to economic conditions but they presage these
conditions due to the forward-looking behaviour of many financial variables, essentially
implying an intermediary case. This is the focus of our paper. Specifically, we propose an
econometric model for estimation of economic and financial conditions that are governed
by a common cycle or, put differently, the business cycle that is potentially reflected in
the cycle of the financial conditions with phase shifts. This implies that we allow for the
financial cycle to lead/lag the business cycle in a systematic waywhen estimating economic
and financial conditions and thereby predicting recessions in real time. This specification
combines a dynamic factor model to extract economic and financial factors/conditions
together with a Markov regime switching dynamics allowing for phase shifts between
cyclical regimes of these factors. This, in turn, facilitates the inference of economic and
financial indicators with a more precise estimation of the turning points of these indicators.
Therefore, our model enables us to exploit a rich data set of economic and financial variables
for predicting economic downturns accurately in real time.
We examine the efficacy of this approach in a key emerging economy, Turkey, where,
unlike the US, neither an official business cycle dating procedure, nor widely accepted
indicators of economic and financial conditions are available. Using our framework, we
construct probabilities of recessions together with indicators of economic and financial
conditions for Turkey.1We use a mixed frequency data set with different time spans (for
the earliest case) starting from January 1999 until November 2019, that is, the data that
are available to us as of the first week of December 2019.
The results indicate that the financial indicator enters recessions (expansions), on av-
erage, 3.6 (3.0) months earlier than the recession (expansion) for the economic indicator.
A recursive forecasting exercise in real time shows that the proposed model can predict
cyclical downturns in a more timely manner compared to a model with independent cycles
and to a model with a single common cycle. Moreover, by virtue of the joint modelling
1Earlier studies on developing leading and coincident indicators for Turkey include Atabek, Cosar and Sahinoz
(2005) who construct a composite leading indicator for the Turkisheconomy and Aruoba and Sarikaya (2013) who
develop a monthly indicator of real economic activity using multiple indicators at mixed frequencies byemploying
the dynamic factor model proposed in Aruoba et al. (2009).
©2020 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
Get this document and AI-powered insights with a free trial of vLex and Vincent AI
Get Started for FreeStart Your 3-day Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your 3-day Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your 3-day Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your 3-day Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting
Start Your 3-day Free Trial of vLex and Vincent AI, Your Precision-Engineered Legal Assistant
-
Access comprehensive legal content with no limitations across vLex's unparalleled global legal database
-
Build stronger arguments with verified citations and CERT citator that tracks case history and precedential strength
-
Transform your legal research from hours to minutes with Vincent AI's intelligent search and analysis capabilities
-
Elevate your practice by focusing your expertise where it matters most while Vincent handles the heavy lifting