How frequent a BEER? Assessing the impact of data frequency on real exchange rate misalignment estimation
| Published date | 01 July 2021 |
| Author | Claire Giordano |
| Date | 01 July 2021 |
| DOI | http://doi.org/10.1111/sjpe.12274 |
Scott J Polit Econ . 2021;68:365–404. wileyonlinelibrary.com/journal/sjpe
|
365
© 2021 Scottish Ec onomic Society
1 | INTRODUCTION
The price compet itiveness of a countr y is generally proxied by i ts real effecti ve exchange rate (REER), a weighte d
geometric aver age of its bilatera l real exchange rat es (RERs) vis- à- vis the currency of eac h of its main trad ing
partners . The REER captures the develop ments in domestic prices o r costs relative to the weighted aver age of
those of its comp etitors, all expresse d in a common currency. Prices or co sts are generally expr essed as indices:
they hence provid e information solely on pri ce- competitiveness dyn amics, and not on how competi tive a coun-
try is at a given po int in time relative t o a benchmark , “equilibrium” level . Various empiric al models have been
Accepted: 25 Jan uary 2021
DOI: 10 .1111/sjpe.1 2274
ORIGINAL ARTICLE
How frequent a BEER? Assessing the impact
of data frequency on real exchange rate
misalignment estimation
Claire Giordano
Economics, St atistics and Res earch DG,
Balance of Paym ents Analysis Div ision,
Bank of Italy, Rome , Italy
Correspondence
Claire Giordano, Economics, Statistics and
Research DG , Balance of Payment s Analysis
Division, Ba nk of Italy, Rome, Italy.
Email: claire.giordano@bancaditalia.it
Abstract
This article exp lores the robust ness of Behavioura l
Equilibrium Exchange R ate models, employe d to estimate
real effecti ve exchange rate misalignm ents, to the fre-
quency of the unde rlying data. It comp ares misalignment s
stemming from an ann ual model, estimated since 1980, and
a comparable quar terly model, e stimated since 1999. The
two sets of estim ates are similar. Moreover, the in- sample
power of quarterly R EER misalignments in explaining su bse-
quent REER development s is higher than that of the annual
estimates. This ar ticle therefore suggests tha t the “optimal”
frequency of a BEER mod el depends on whet her its result-
ing estimates are emp loyed for research pur poses or for
policy- making activities.
KEYWORDS
real effect ive exchange rate, equilibr ium exchange rate, BEER
model, data frequency
JEL CLASSIFI CATION
F00; F31
366
|
GIORDANO
employed to oper ationalise the t heoretica l concept of this unob servable e quilibrium va lue and to derive t he
resulting REER mis alignment, yet no approach h as been found to achieve a superio r performance to the other s
(Ajevskis et al ., 2014; Driver & Westaway, 2004; MacDonal d, 2000). The most comm on methodologies are the
following.
The Behaviour al Equilibrium Exchange Rate (B EER) methodology is labe lled this way due to the fact tha t it is
based on the assum ption that the “behaviour” of a R EER is determined by the “behaviou r” of its macroeconomic
drivers in the lo ng run (Clark & MacDonald , 1998). It involves direct estimat ion of the reduced- form coin tegrating
relationship be tween the REER and a set of relev ant economic fundament als, which leads to the def inition of the
REER equilibriu m value. This modelling tech nique does not generally have a ny comprehensive theoreti cal under-
pinning and is a posi tive approach, since it is not b ased on any normative assump tion.
The Permanent Eq uilibrium Exchange Rate (PEER) a pproach, described in Cl ark and MacDonald (200 4), can
be seen as an evolut ion of the BEER model in which cointegr ation techniques are employe d in order to extract
the permanen t, as opposed to transitor y, components of both the REER an d the economic fundament als, which
are then compare d to each other. This exercise is sim ilar in spirit to a BEER model w hen filtered (i.e. tren d) values
of the economic fu ndamentals a re employed in lieu of t heir actual v alues in order to co mpute the equil ibrium
REERs.
Another appr oach is the Natur al Real Exchange R ate (NATREX) meth odology, origin ally formulate d by
Stein (1990). It is theoretic ally grounded on a dynami c stock- flow model. In particul ar, it defines the “natural”
REER as the REER that ens ures both the internal and th e external equilibrium si multaneously in the long r un.
The internal eq uilibrium is achi eved when the out put gap is zero; the external equili brium is obtained when
the current acco unt balance is “sust ainable” given a co untry's de sired net foreign a sset position. A lthough
there have been some attempts to measure the structural model underlying the NATREX (e.g. Gandolfo &
Felettigh, 1998; Sir egar & Rajan, 20 06), this approach often boils dow n to estimating a r educed- form equa-
tion. In this cas e, the main dif ference betwe en the BEER and th e NATREX is only that th e latter is more
explicitly th eory- based (Stein, 2006). A gain, similarly to the BEER m ethodology, the NATREX model ad opts a
positive approach.
The last appro ach is the Fundame ntal Equilibri um Exchange Rate (F EER) methodol ogy, advocated by Wren -
Lewis (1992) and Williamson (1994). Simila r to the NATREX approach, the FEER is the R EER that simultaneously
attains inter nal and external balance . In its most popular appl ications (Cline & Williamso n, 2010; Isard, 2007; Lee
et al., 2008), t he FEER method is based on a partial e quilibrium model and, in par ticular, on the computation of
the REER adjustm ent required to close the gap b etween the cyclically a djusted current account and t he “current
account norm,” whic h represents an optimal v alue of the current account over a m edium- term horizon. The no rm
is either set in a nor mative manner o r is derived from “b ehavioural” r educed- form regressio ns that estimate a n
equilibrium re lationship between the cu rrent account and a set of plausibl e fundamentals. The cal ibration of the
required REER adj ustment is, however, highly sensiti ve to the assumptions made conce rning both exchange- rate
pass- through coeff icients and price elasti cities of trade (Schnatz, 2011). The FE ER model can be reconciled wit h
the NATREX approa ch by estimating a target level for t he net foreign asset position rath er than for the current
account balance.
These approa ches, far from being op posed to each other, are comple mentary, both becaus e they are all based
on restrict ive (albeit diff erent) assumpt ions and becau se they may be inter preted as assess ing equilibriu m ex-
change rates over dif ferent time horizons (Bénassy- Queré et al., 2010). Indeed , the FEER, for ins tance, may be
considered as cor responding to a medium- run concept of equilibriu m, whereas the (filtere d) BEER, and even more
so the PEER, to a long- run concept.
This paper focu ses solely on one methodolog y, namely the BEER approach. Yet, even wh en focusing on one
method only, many mod elling choices arise, concernin g the selection of the countr y sample, the e stimation pe-
riod, the data f requency, the weights and th e deflator employed to constr uct the REERs, the numer aire currency
against which to e xpress the bilateral R ERs underlying the REER s, the relevant set of econ omic fundamentals , the
|
367
GIORDANO
appropriate pr oxies of these economic fund amentals, as well as several e conometric aspects , in particular linked
to accounting for country heterogeneity and tackling cross- sectional dependence.
Few studies amo ngst the many th at estimate BEER mo dels systemat ically addr ess these manif old method-
ological issue s. Recent excepti ons are the followi ng. Bussière et a l. (2010) focuses lar gely on the appro priate
econometric te chniques to be e mployed in estim ating BEER model s. Bénassy- Queré et al. (2010) discusses t he
correct measurement of the relevant economic fundamentals, namely the Balassa- Samuelson effect, and the
choice of the numer aire country. Adler and G risse (2017) is centred on the se lection of fundament als, the country
sample and the e stimation methodo logy. Fidora et al. (2020) ad opts various price an d cost indicators to const ruct
the REERs, as well a s employing alte rnative Balass a- Samuelson proxies, car efully selec ting the set of rel evant
explanatory variables and using different estimation techniques.
Adding to these s tudies, this article is con cerned with the robustne ss of equilibrium REERs and of the co rre-
sponding misa lignments alo ng one dimension only, namely the choic e of the data frequ ency (i.e. ann ual versus.
quarterly), a d imension which thus fa r has not been investigate d by the existing litera ture. Various methodol ogical
aspects de pend on this choice, such as the sel ection of the sample perio d and of the relevant set of fundamen-
tals. All othe r modelling choices are hel d fixed between the ann ual and the quarterly m odels, namely the count ry
sample, the r ange of price and cost indicators e mployed to compute the REER, the nu meraire currency, and the
econometric technique used for estimation.
A priori, quar terly models have two main ad vantages. First, the hig her number of observatio ns relative to an
annual datase t estimated over t he same time spa n implies that qua rterly mod els may be estima ted more effi-
cien tly. 1 Second, est imating a BEER mod el at quarterl y frequenc y is of paramount im portance f or surveillan ce,
early- warning and polic y- making purposes, sin ce it allows track ing infra- annual imbalan ces in a timely man ner
(Giordano, 2018, 202 0a). Conversely, the main drawbac k is the restriction of the e stimation window to a shor ter
time span, coveri ng only the most recent years , for which quarterly dat a are available. This limit ation, in turn, ex-
acerbates two si gnificant shortcomi ngs underlying the BEER m ethodology.
The first sho rtcoming concerns the nee d to include country fixed e ffects amongst th e regressors included in
a panel BEER model , due to the fact that the de pendent variable, the R EER is generally an index numb er (Adler &
Grisse, 2017; Cubeddu e t al., 2019; Fischer & Hossfeld, 2014; Mano et al . 2019).2 T he inclusion of country fixe d
effects is i ndeed a means to account for count ry- specific price levels in th e base year. Yet, by including fixed ef-
fects, the p redicted (i.e . equilibrium) R EERs are by constr uction on aver age equal to the lon g- run actual REER
means, or in othe r terms each countr y's regression resid uals— and thus REER mis alignments— are forced to average
to zero over the sampl e period. The fixed- effect estimation s hence make estimates les s reliable for countrie s with
short time dat a spans; moreover, “persis tent” deviations of REER s from their equilibr ium value are by constr uction
not envisaged in f ixed- effects pan el BEER models.
The second draw back of the BEER methodo logy is that it implicit ly assumes that the eco nomic fundamenta ls—
against which t he actual REER is a ppraised— are at their e quilibrium val ues. This is les s likely to be the cas e for
short time sp ans. Moving to annual dat a implies being able to s ignificantly ex tend the sample peri od and therefore
making both the a ssumptions of zero misalignm ents and of economic fund amentals at their equili brium values on
average over the est imation horizon more palat able.
In order to assess t he impact of data frequ ency on REER misalignme nt estimates, this ar ticle first describ es an
annual BEER mode l since 1980 for a vast sample of advance d and emerging countries, u sing recently developed
techniques bot h to select the relevant ec onomic fundamentals (B ayesian model averaging, B MA) and to estimate
1As well as enha ncing effici ency, the higher n umber of obser vations under lying a quart erly model all ows more flexibi lity in modell ing the mean
reversion to e quilibrium, f or example lead ing to the possib ility of addre ssing the exist ence of potential n onlinear mea n reversion, whi ch has been
found to be pre sent in the case of r eal exchange rat es referred to so me currencies (e .g. Chorta reas, Kapeta nios and Shin, 20 02).
2An alternat ive would be to focu s on time- series analyses of ind ividual count ries, yet these a nalyses suf fer from a limited n umber of obser vations;
panel esti mation is hence ge nerally empl oyed in this liter ature, as discus sed in Section 2 .
Get this document and AI-powered insights with a free trial of vLex and Vincent AI
Get Started for FreeStart Your 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 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 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 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 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
