Sei sulla pagina 1di 13
A Century of Purchasing-Power Parity Author(s): Alan M. Taylor Reviewed work(s): Source: The Review of

A Century of Purchasing-Power Parity Author(s): Alan M. Taylor Reviewed work(s):

Source: The Review of Economics and Statistics, Vol. 84, No. 1 (Feb., 2002), pp. 139-150 Published by: The MIT Press

Accessed: 08/02/2012 19:04

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

information about JSTOR, please contact support@jstor.org. The MIT Press is collaborating with JSTOR to digitize,

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to The Review of Economics and Statistics.

http://www.jstor.org

A CENTURY OF PURCHASING-POWERPARITY

Alan M. Taylor*

Abstract-This paperinvestigatespurchasing-powerparity(PPP) since

twenty countries

over 100

of annualdatathanhaseverbeen

studied.The evidencefor

variateand univariatetests of

been

generally as expected; thisresultis not

attributableto significantlygreaterpersistence(longerhalf-lives) of de-

shocks to the real

exchange rate process

but is due to the

viations in such

century, therewas relatively little

market integration to smooth out real exchange rate shocks. Instead,

changes in the size of shocks depended on the political economy of

choiceundertheconstraints imposed

by

monetary and exchange rate regime the trilemma.

in such episodes. In the course of the twentieth

thelatenineteenth century. I collecteddatafora

years,

a larger historical panel

group of

long-run PPPis favorable using

higherpower.

PPP,

recentmulti-

Residualvariance analysis

larger

shows that episodes of floating exchange rates have

associatedwith larger deviationsfrom

regimes,

change in the capacity of international

I.

Introduction

WI

HATnew findings canthis paper claimto offer given the wealth of researchon

purchasing-powerparity

years. From

(PPP) in the past? It first should be noted

that empirical

support forPPPhaswaxedandwanedoverthe

a historical standpoint, therehavebeennumerousstudiesof PPPfor variouscountriesoverthe period in question, some

covering a particular era or monetaryregime.McCloskey andZecher (1984) argued thatPPPworked very well under the Anglo-Americangold standardbefore 1914. Diebold,

Husted, and Rush (1991) explored a very long

nineteenth-century datafor six countriesandfound support

for PPPbasedon the

short-sample studies. Abauf and Jorion (1990) studied a

century of dollar-franc-sterlingexchange ratedataandver-

ified PPP;

for two centuries of

(1990) also found evidence thatreal exchange rates were

stationary for Japan, theUnited

and Francefor the period 1875-1986,

change rates exhibited only trend stationarity-an oft-

repeatedfinding thatthe

ciatedoverthe long run against all currencies.In

monographs, both Lee (1978) and Officer (1982) found strong evidencein favorof PPPbasedon analysis of long

run of

low-frequency information lacking

in

LothianandM. P. Taylor(1996) foundthe same

dollar-franc-sterling data. Lothian

States, theUnited Kingdom,

althoughyen

ex-

real yen exchange

ratehas appre-

full-length

Received for

*

publicationFebruary26, 1998. Revision accepted for

2000.

on the historicalevolutionof international capital mo-

I am indebtedto

publication October 19,

University of California-Davisand NationalBureauof Economic

Research.

Thisworkis relatedto a largerproject withMauriceObstfeld (Obstfeld

& Taylor,2002)

bility. This study of purchasing-powerparity(PPP)began withan earlier

Ronald Albers, Michael

Bordo,StephenHaber, JanTore Klovland, Matthew Jones, James Lothian,

Maurice Obstfeld,Hugh Rockoff, and PierreSicsic for help with data

enquiries,

advice on econometric practice. For their

thankMichael Knetter, James Lothian, Christian Murray, David Papell,

and seminar participants at the NBER Conferenceon Exchange

(May 1996) and workshopparticipants at Indiana University. The paper

hasbeen

greatlyimprovedby All errorsaremine.

thecommentsof theeditorandthereferees.

workingpaper(Taylor,1996).

and to Graham Elliott, Peter Pedroni, and MarkWatsonfor

helpfulcomments,

I wish to

Rates

time series running fromthe pre-1914gold standardto the managed floatof the 1970s.1 Of late, new studieshave appeared in abundance.In their recent comprehensive review of the PPP literature,Froot and Rogoff(1995) coulddeclarethatwhatwas a "fairly dull research topic" only a decade ago has recently been the focus of substantial controversy andthe subject of a grow- ing body of literature.Recent empiricalresearch,mostly basedon the time series analysis of short spans of datafor the floating-rate(post-BrettonWoods) era, led many to concludethatPPPfailedto hold,andthatthereal exchange ratefolloweda random walk, withno mean-reversion prop- erty.However,a newly emerging literature exploits more dataand higher-poweredtechniques, andclaims that, in the long run, PPP does indeed hold: it appears from these studiesthatreal exchange ratesexhibitmeanreversionwith

half-lifeof deviationsof fourto five years(M. P. Taylor, 1995; Froot & Rogoff, 1995). The newer findings various steps to expand thesize of samples usedto testPPP.

a

use

As noted, it has been possible to use much longer-run time seriesforcertainindividual countries,spanning a century or more;typically suchexerciseshave concentratedon more-

developed countrieswith good historicaldata

(forexample, theUnited States,Britain, and France). Alter-

natively, researchershave expanded the datafor the recent float or postwar periods cross-sectionally to exploit the

additionalinformationin panel data (Wei &

Frankel& Rose, 1996;Pedroni,1995;Higgins & Zakrajsek,

availability

Parsley,1995;

1999).

It is still too early to say whetherthe revisionistPPP

findings will proverobust, and challenges to this interpre-

tationhave alreadyemerged. One may

ways in whichcross-sectionalinformationand panel meth-

odologies have been applied (O'Connell, 1996, 1998). Some have notedthatthe inferencesbasedon panel meth-

selection, and many results

country, for example,

the United States versus Germany(Papell, 1997; Wei &

Parsley,1995; Edison, Gagnon, & Melick, 1994). Others

caution that detecting a unit root in time series may be

complicatedby the fact that price indicescan be viewed as

the

and a

sum of a stationary tradable relative-pricecomponent

nonstationary nontradable relative-pricecomponent

appear sensitiveto the choice of base

ods are sensitive to sample

find fault with the

(Engel,2000; Ng & Perron,1999). This

venerableBalassa-Samuelson objection to the pure PPP

finding echoes the

hypothesis basedon differentialratesof

in traded and nontraded goods sectors (Balassa, 1964;

productivitygrowth

strong foundationof historical

work by a numberof scholarsand covering variouscountriesin different

time periods. Otherstudiesof

Edison,1987;Johnson,1990;Glen, 1992;Kim, 1990).

1 Obviously, this paper buildson a very

long-run dataarenumerous (Frankel,1986;

The Review of Economics and Statistics,

February2002,

84(1): 139-150

? 2002 by thePresidentandFellowsof Harvard College andtheMassachusettsInstituteof Technology

140

THE REVIEW OF ECONOMICSAND STATISTICS

Samuelson, 1964). Of course, such long-run trends may be purely deterministic (Obstfeld, 1993). The distinction of the present study is to bring very recent empirical innovations to a longer span of historical data, both to investigate the robustness of the recent findings and to explore the historical evolution of PPP. I proceed as follows. Section II introduces the real exchange rate data,

and preliminary analysis shows that old-style univariate tests cannot reject the unit root null. Section III examines a multivariate test of PPP by M. P. Taylor and Sarno (1998). In the search of a more powerful test, section IV applies the univariate efficient tests of Elliott, Rothenberg, and Stock (1996). We find that long-run PPP can be supported in all cases with allowance for deterministic trends. The importance of the long-run trends is explained in section V wherein I model the dynamics of real exchange rates at different times in the twentieth century. Four re-

gimes are investigated: the gold standard 1870-1914,

interwar period 1914-1945, the Bretton-Woods era 1946- 1971, and the recent float 1971-1996. The importantquan- titative differences found are in residual variance, and the floating regimes exhibit much larger shocks to the real

exchange rate process, accounting for the much larger deviations from PPP during these eras. Thus, the history of

PPP in the twentieth

century shows, surprisingly, that there

was relatively little change in the ability of international market integration to smooth out real exchange rate shocks. Instead, I argue, the changes in the variance of the shocks reveal a great deal about the differing degrees to which monetary policy was kept in check or not by commitment mechanisms (under fixed rates) or their absence (under floating). In light of this, I end with a discussion that relates

these findings to the question of the political economy of monetary and exchange rate regime choice under the con- straints imposed by the macroeconomic policy trilemma.

the

II. Data and Preliminary Analysis

rates, Eit, measured

as domestic currency units per U.S. dollar, and price indices,

Pit, measuredas consumer price deflators (or, when they are

not available, GDP deflators). We will refer to the log levels

of these variables, denoted eit =

The index i = 1,

Argentina, Australia, Belgium, Brazil, Canada, Denmark, Finland, France, Germany, Italy, Japan, Mexico, Nether- lands, Norway, Portugal, Spain, Sweden, Switzerland, the United Kingdom, and the United States. The index t runs over the set of years from 1850 to 1996, but a complete cross section of twenty exchange rates does not exist before

, 20 covers the set of countries

The data consist of

annual exchange

log Eit and Pit = log Pi,.

1892, the starting date of the Swiss series.2

2 In constructing the data set, I have reliedon standardsources.After

1948theseriesaretakenfromtheIMF'sInternationalFinancialStatistics

sources are the statistical

volumesof BrianMitchell.Forthe

price and exchange ratedatafromtheseandothersources,I am grateful

on CD-ROM.The

principalpre-1948 price

provision of electronicallycompiled

Given these data, some preliminary transformationsand tests were performed. Let the U.S. dollar-denominated price

level of country i at time t be denoted by Rit = Pi/Ei,, with

rit =

were filled in for each series. In all cases, this amounted to

imputing a value to a few wartime years for certain coun-

tries, using linear interpolation on rit. This

20 X 105 panel of data from 1892 to 1996. Such an interpolationproceduremay be ad hoc, but it was

deemed necessary to give any stationarity test a fair chance on this data, because, in several cases, the missing data

appear after explosive

exchange rate often depreciated. Without interpolation in

these periods, any subsequent reversion back toward the mean (or trend) in this variable would be missed by any

estimation procedure, and a bias against stationarity would result. An importantexample would be the wide divergence in real exchange rates in the 1930s following the collapse of

the gold standard; this episode was

to many missing observations in the data, and thus much of the reversion of these divergent real exchange rates toward

PPP during and after the war would be omitted from the sample absent any interpolation. With interpolations complete, the real exchange rate se- ries was generated two ways: first, relative to the U.S. -

dollar, as qit =

"world" (N = 20) basket of currencies, as qw =

where riw = lN 1 ji rt.3 The second definition follows

O'Connell (1996) and may help us avoid problems associ- ated with the choice of the United States as a base country.4 The complete series qit and qwi for all twenty countries are shown in figure 1. One way to test the PPP hypothesis is to ask: are these real exchange rates stationary, that is, mean reverting? A cursory inspection suggests that for many countries real exchange rates have been fairly stable over the long run, and we might expect to easily support the hypothesis of stationarity.Nonetheless, our eyes are drawn to certaincases in which there appears to be a long-run trend or random walk. Here, the most obvious and well-known problem would be the case of Japan, but similar symptoms

of

Switzerland, Brazil, and in some other countries' experience in specific periods, such as interwar Germany and Italy. Clearly, a powerful statistical test will be needed to resolve this question. We can begin analysis using more-traditional unit root tests. Table 1 shows the results of applying the augmented

log Rit =

Pit =

eit. As an initial step, missing data

yields a balanced

inflations during which the real

followed by war, leading

rit

rus,t, and second, relative to the -

it

rW

drift or nonstationarity might also be perceived for

to Michael Bordo. An appendix containing the data and full documenta- tion is available from the author upon request.

might prefer to use trade-weighted real exchange rates, but

such data do not exist in the form of annual time series for the entire

countries. Future research would

twentieth century for a wide sample

need to be directed to original sources to collate the necessary bilateral

trade volumes, and this would be a

3 Ideally,

one

of

4 This discussion was omitted in O'Connell (1998).

significant undertaking.

1870

1890

1910

1.5

1.0

0.5

0.0

-1.0

-1.5

-2.0

-2.5

Argentina

1870

1890

0.4

0.2

0.2

Canada

0.0

-0.2

-0.4 I

-0.6

1910

1870

0.6

1890

1910

0.4

0.2

0.0

-0.2

-0.4

04

Germany

1870

1890

1910

0.8

0.6

0.4

0.2

0.0

-0.2

-0.4

-0.6

Netherlands

R

1870

1890

0.6

0.4

0.2

0.0

-0.2

- 0.4

Sweden

1910

1930

0

1930

1930

A

1930

1930

1950

1950

1950

1950

1950

1970

1970

1970

A

1970

1970

1990

1990

1990

1990

1990

A CENTURYOF PURCHASING-POWERPARITY

FIGURE1.-A

1870

1890

1910

0.6

0.4

0.2

Australia

1930

1950

-0.6

-0.6

.8

1870

1890

1910

0.8

0.6

0.4

0.2

0.0

-0.2

-0.4

-0.6

Denmark

-

1870

1.5

--

1890

1.0 Italy

0.5

0.0

-0.5

-1.0

1910

1870

1.0

0.5

0.0

-0.5

-1.0

.---

1890

-

Norway

1910

18 190 1910 - Swi tzerland

1870

1.0

0.8

0.6

0.4

0.2

0.0

-0.2

-0.4

-0.6

1930

1930

1930

.

1930

L0.6

1950

1950

1950

1950

CENTURYOF REAL EXCHANGERATES

1970

1970

1970

1970

1970

1990

1990

1990

1990

1990

1870

2.0

1.5

1.0

0.5

0.0

-0.5

-1.0

1890

1910

1930

1950

1970 1990

Belgium

_wL^v

1870

- -

1890

Finland

2.0

1.5

1.0

0.5

0.0

-0.5

-1.0

1910

1930

1950

0

-

1970

1990

1870

1.5

1.0

0.5

0.0

-0.5

-1.0

-1.5

Ja

1890

pan

1910

1930

1950

1970

1990

1870

1890

1.0

0.5

0.0

-0.5

-1.0

Portugal

1910

1930

1950

-~ u~- -

1870

-

----

1890

-.-

.

1910

1930

0.4

0.2

0.0

United Kingdom

-0.2

-0.4

-0.6 i

1950

1970

1970

1990

-005

1990

1870

1.5

1.0

0.5

0.0

-0.5

-1.0

-1.5 I

1890

1910

Brazil

X

1930

1870

.6

0.4

0.2

0.0

-0.2

-0.4

-0.6

F

1890

ce

1910

1930

1870

1890

1910

1930

1.5

1.0

0.5

0.0

-0.5

-1.0

-1.5

Mexico

y^'W^A

1870

1.0

1890

Spain

1910

1930

-0.5

~t-1.0-I

1870

1890

1910

0.4

0.2

0.0

-0.2

-0.4

-

United States

1

i

1930

1950

N

1950

1950

1950

1950

1970

1970

1970

1970

1970

141

1990

1990

1990

1990

1990

See text and appendix. The thickerline shows qi,, the real exchange rate relative to the U.S. dollar.The thinnerline shows qw,, the real exchange rate relative to the "world" (N = 20) basket of currencies.

Dickey-Fuller(ADF) unit root tests to the univariatereal

exchange rate series, andtheresultsare

findings in the

root null cannotbe

present does not seemto

rejected in most cases. However, a simple OLS regression on a constantanda trendseemsto indicate that, for at least

some of the series,

this trend component is a sizeable 1.5%

case of Japan, 0.74% per annumfor Switzerland,but is

small (no more thanhalf a

cases. All in all, we are left with the conclusion that,

although someof the series

addition, have some deterministicdrift.This im-

most, in

pression is given whetherone uses the United Statesas a base country, oronemeasuresreal exchange ratesrelativeto the "world"basket.

percentper year) in all other

expectedgiven

the

previous literature.5In manycases, the unit

rejected. Even allowing a trendto be

helpverymuch, andthenullis not

a deterministictrend might be present;

per annumin the

may be I(1), many are I(0) and

5 Lag lengths in the ADF tests were chosen by the Lagrange multiplier

(LM) criterion for residual serial correlation, allowing up to a maximum

of six lags.

Onetraditional response to

such findings has always been

power, andto point to the

to faultthesetestsfortheirlackof

fact that, with slow convergencespeeds, the autoregressive

parametermight be very

a very long

With more

sufficient

are still unableto find broadevidenceof

recentliterature suggests two possible

multivariateor panelmethods, andtheuse of more-efficient

univariatetests.We pursue both routes, to

might lend support to the PPP hypothesis.

close to unity, andone wouldneed

null (Frankel,1990).

data, we might just have

span of datato reject the

than a century of

span

to have a

reasonablypowerfultest, but we

stationarity. The

directions:theuse of

see which, if any,

III. A Multivariate Test

If PPPholds among a set of N + 1 countries, thiswould

imply that every single log real exchange rate, for all

N(N

axiomatic property of well-definedPPP measuresis base

countryinvariance; that is, the concept of PPP must be

invariantto the choice of base

bilateral pairs, would be stationary. One

+

1)/2

country.Thus, we may,

142

Base: UnitedStates

Argentina

Australia

Belgium

Brazil

Canada

Denmark

Finland

France

Germany

Italy

Japan

Mexico

Netherlands

Norway

Portugal

Spain

Sweden

Switzerland

United Kingdom

Base: "World"Basket

Argentina

Australia

Belgium

Brazil

Canada

Denmark

Finland

France

Germany

Italy

Japan

Mexico

Netherlands

Norway

Portugal

Spain

Sweden

Switzerland

United Kingdom

United States

THE REVIEW OF ECONOMICSAND STATISTICS

 

TABLE 1.-PRELIMINARY

DATA ANALYSIS

Demeaned

Detrended

T

ADF

LM

ADF

LM

OLS Trend

113

-4.85***

0

-4.81***

0

-0.0013

127

-2.44

0

-3.04

0

-0.0030***

117

-3.25**

0

-3.85**

0

0.0053***

108

-2.74*

1

-2.70

1

0.0000

127

-2.59*

0

-3.76**

0

-0.0011***

117

-2.16

0

-2.77

0

0.0033***

116

-4.50***

0

-4.63***

0

0.0016**

117

-3.54***

1

-4.18***

1

-0.0030***

117

-2.96**

1

-3.28*

1

0.0029***

117

-3.27**

0

-3.27*

0

0.0005

112

-0.29

0

-1.99

0

0.0151***

111

-2.96**

0

-3.91**

0

-0.0065***

127

-2.00

0

-2.27

0

0.0022***

127

-2.41

0

-2.61

0

0.0025***

107

-2.60*

0

-2.46

0

-0.0037***

117

-2.34

0

-2.25

0

-0.0019***

117

-2.78*

0

-3.44**

0

0.0032***

105

-1.93

1

-3.43**

1

0.0083***

127

-3.19**

0

-3.41*

0

-0.0014***

105

-5.19***

0

-5.32***

0

-0.0029**

105

-2.38

0

-3.56**

0

-0.0045***

105

-3.44**

0

-4.07***

0

0.0051***

105

-2.32

0

-2.26

0

-0.0017

105

-1.24

0

-2.04

0

-0.0033***

105

-2.84*

0

-3.35*

0

0.0028***

105

-4.70***

0

-4.69***

0

-0.0002

105

-2.59*

0

-3.96**

0

-0.0038***

105

-1.80

0

-1.80

0

0.0017***

105

-3.20**

0

-3.19*

0

-0.0004

105

-0.97

0

-2.35

0

0.0150***

105

-1.45

2

-4.44***

1

-0.0084***

105

-2.13

0

-2.54

0

0.0027***

105

-2.50

0

-2.51

0

0.0012**

105

-2.93**

0

-3.71**

0

-0.0051***

105

-2.03

0

-2.30

0

-0.0030***

105

-2.67*

0

-2.54

0

0.0017***

105

-1.05

0

-2.75

0

0.0074***

105

-2.00

0

-2.88

0

-0.0032***

105

-2.34

0

-2.53

0

-0.0013**

See text and appendix. T is the sample size. ADF is the augmentedDickey-Fuller statisticwith LM the lag length selected by the Lagrangemultiplier criterion.Demeanedis the case in which each seriesis replaced by the residualsfroma regression on a constant.Detrendedis the case in which the regression is on a constantanda lineartrend.Trendis the OLS estimateof the lineartrend. Finite-sample criticalvalues areshown based on 4,000 simulationsof the null.

* denotes significance at the 10%level.

** denotes

significance *** Denotes significance at the 1%level.

at the 5% level.

without loss of generality, take a particular choice of a base country, and then consider the N bilateral log rates relative

to that country.6 An elegant test for the stationarity of these, and hence all,

log rates was proposed by M. P. Taylor and Sarno (1998). They note that a necessary and sufficient condition that all

N series be stationary would be the existence of N indepen-

dent cointegrating vectors among the series (Engle & Granger, 1987). Conversely, if one takes the null to be the absence of such a condition, that is, the existence of any nonstationary real exchange rates, the null would corre- spond to a situation in which the N series had fewer than N

cointegrating vectors.

6 All log real exchange rates between bilateral pairs can then be derived,

all cross-rates in the exchange market, as

assuming arbitrage amongst

linear combinations of the set of N log rates for the given base country.

These hypotheses permit some simple tests based on the cointegration methods of Johansen (1988, 1991). To briefly

review this approach, let qt = (qlt,

vector of real exchange rates at time t. Under the PPP

hypothesis, all N components of q are I(0).

correction representation for the dynamics of q is

, qNt) be the NX 1

The error-

Aqt =

FrAqt-1

+

** rk-lAqt-k+l

+

rkqt-k

+

JL +

Ot.

The NXN matrix rk has rank equal to the number of

cointegrating vectors, so the null hypothesis of one or more

nonstationary series is

alternative hypothesis where all series are stationary is H :

rank (rk) = N. Because full rankof rk would imply that all of the eigenvalues are nonzero, a test of H1 against a null of

Ho amounts to a test of the restriction that the smallest eigenvalue XN of the estimated rk matrix is zero. The

Ho

:

rank (rk)

<

N,

and the

A CENTURY OF PURCHASING-POWERPARITY

TABLE2.-JOHANSENLIKELIHOODRATIOTEST

Base: World

Group 1: Europe

Belgium, France,Germany,Italy, U.K. Netherlands

excluding Netherlands excluding Netherlandsand Germany Group 2: Scandinavia Denmark,Finland,Sweden, Norway excluding Norway Group 3: Iberia and LatinAmerica Argentina,Brazil, Mexico, Portugal,Spain Group 4: Other Australia,Canada,Switzerland,Japan excluding Switzerlandand Japan

Demeaned

SIM

JLR

p

4,032

1.77

[.37]

3,968

2.99

[.18]

3,840

3.80

[.13]

3,840

3.11

[.22]

3,584

7.41

[.02]

3,968

4.06

[.08]

3,840

0.65

[.59]

3,072

6.53

[.03]

Power

JLR

0.25

5.04

0.31

4.77

0.43

8.09

0.54

4.09

0.60

7.25

0.33

4.86

0.12

4.34

0.23

7.64

Detrended

p

[.01]

[.02]

[.00]

[.14]

[.03]

[.01]

[.03]

[.01]

143

Power

0.02

0.02

0.02

0.29

0.40

0.02

0.01

0.07

See text and appendix. JLR is the Johansenlikelihood ratio test statistic.The finite-samplesignificance level is based on SIM simulationsestimatedunderthe assumptions of the null. The power of the 5% test is based on 4,096 simulationsestimatedunderthe assumptions of the alternative.

Johansen likelihood ratio (JLR) test statistic for this case is JLR = -T ln(l - XN), where JLR has an asymptotic distribution that is X2(1). The chief merit of this test may be the clean specification of null and alternative hypotheses. Other multivariate tests based on panel methods, such as the multivariateform of the ADF test, may reject a unit root null when only some of the

multivariate approach

also places furtherrestrictions on any empirical framework.

Given the base-country-invariance postulate, the structure imposes k lags of the Aq, in all equations of the system. Such a restriction is commonly not a feature of univariate

tests of PPP using single-country exchange rate series, yet it ought to be present if we are really thinking in terms of a joint hypothesis test involving the stationarity of all the series taken together. But what lag choice should one make? In the previous section, the tests performed in table 1 report a variety of lag lengths selected by the Lagrange multiplier criterion. By inspection, we note thatthe