Dynamic Modeling and Testing of OPEC Behavior

No. 8917 DYNAMIC MODELING AND TESTING OF OPEC BEHAVIOR Carol Dahl* Visiting Scholar, MIT and Associate Professor of Economics Louisiana State University and

Mine Yucel* Federal Reserve Bank of Dallas

Research Paper Bank of Dallas Federal Reserve Bank.

This publication was digitized and made available by the Federal Reserve Bank of Dallas' Historical Library ([email protected])

No. 8917 DYNAMIC MODELING AND TESTING OF OPEC BEHAVIOR Carol Dah1* Visiting Scholar, MIT and Associate Professor of Economics Louisiana State University and Mine Yuce1* Federal Reserve Bank of Dallas December 1989

•

*The views expressed in this article are solely those of the authors and should not be attributed to the Federal Reserve Bank of Dallas, the Federal Reserve System, MIT, or Louisiana State University.

.'

Dynam i c: Mod" II n9 and T"st i n9 of OPEC B"hav i Or

*

by Carol Dahl Visitin9 Sc:holar MIT-Ener9Y Lab Bld9 E40-411 One Amherst St. Cambrid9'" MA 02139 and Assoc:iat" Prof"ssor of Ec:onomic:s D"partment of Ec:onomic:s Louisiana State University

Baton ROUge, LA 70B03-6306 and Mine Yuc:el Senior Economist

Federa I Res"rv" Bank of Da I I as Station K Dallas, TX 75222 Dec:"mber 3, 19B9 Abstrac:t Althou9h c:onv"ntional wisdom sU99"sts that OPEC is a c:artel, many studies sinc:e 1973 hav" c:onsider"d other underlyin9 forc:es in order to understand and forec:ast OPEC b"havior Usin9 the most gen"ral model to dat" on quart"rly data fr~m 1971:1 to 1986:IV we ec:onometric:ally test a vari"ty of h"poth"ses. We find that th" various OPEC c:ountries behav" in quite dissimi lar ways SU99"stin9 that a c:artel hypothesis is not appropriate. Under our specification there was no evidence tor dynamic optimization Or a stron9 target revenue model. There was some "videnc:" that a form of tar9"t r"v"nue may be inc:luded in th" 90als fOr Iran, Libya, Saudi Arabia, and the UAE. Iraqi b"havior was most consistent with a static competitive market structure, whi Ie a static noncompetitive market structure was not rejected tor

AJgeria, Nigeria, Saudi Arabia, Kuwait, and V"nezuela. 9iven their divergence

in behavior we do not conclude

How"v"r In

favor of

a w"ak c:art,,1 but that ther" is a nonc:omp"titive c:ore of SWin9 produc:ers that eac:h swin9 to their own rhythm.

OPEC market

stru~ture

debate sin~e 1973/74 when into pubJ ic prl~e

tailing

ba~k

SOur~e

~ontinued

in 1978/79 but subsided

~Ioser

to historical

levels

wisdom suggests that OPEC is a cartel functIoning cartel,

~onsiderabie

ot

large prl~e In~reases ~atapulted OPEC

The debate

attention.

In~reaSes

has been a

gropIng

In

with even

urgen~y

In 1986.

Or at

larger

with

pri~es

Conventional

least a weakly

towards an optimal

level

with recent price decreases signaling that the cartel

ot

revenue

is

losing

its grip On the market. The cartel

argument, however,

IS not universally held. 1

A

variety ot arguments have been put tOrth to try to explain OPEC behavior.

Since simulations ot OPEC as a cartel

did not simulate the high explained

changing

pri~es

ot the 1980's, some modelers

continuing hi9h prIces OPEL~ehavior,

with political

arguments,

Or changing OPEC perceptions.

Competitive arguments suggest that market torces prices and then to

or a monopoly

lower prices.

led to high

Property rights arguments

suggest that they resulted trom shitting property rights

trom the

companies with a higher discount rate to OPEC countries with a lower discount rate.

A ccmpet

j

t

j

ve target revenue mode I,

wh i ch

yields backward bending supply curves Once target revenue has *The authors wou I d like to thank without imp I i cat ing Dermot Gately and Clitton Jones tor comments on an earlier dratt ot th,s paper and Professor James Griftin tor encouragement and generOuSly providing uS with his data.

1.For more complete surveys ot the I iterature see Fischer, and Gately, and Kyle (1975), Hammoudeh (1979), Gately (19B4 Dahl and Ywcel (1988). 1

been attained,

sU9gests

that

higher

prices

lead

to

lower

OPEC

systematically

test

output. Griffin OPEC market

(1985)

structure

quarterly data econometric

behavior

from

models

and

a

market

sharin9

price

Ne i ther dynamic

to

behavior

been

or the

a

We

dynam i c

favor

than

each country

use

behav i Or

discount

rates

tor

companies are appropriate is evidence for

a

target

both

whether OPEC

in

the i r a

wou I d

(1987) with

have ·on

or

these

property ri9hts tor

the

to

better

dynamiC

as

has

first by

tests

a

test

static

to

models

tor explaining OPEC behavior,

z

allows

in one

multinational

noncompetitive behavior,

and

0+ our tests

separately

characterized

than

and

base

hypotheses are nested

of

of

the

static

increase the power

results

a

model

framework

model

a

their

theoretical

more general

ot

dynamic model

whether

OPEC

revenue'

favor

revenue

tested

each hypothesis

the

target

of

imp I i cat ions

the

stron9

assume

IS

a

expectations

considerin9

a I

In OPEC

hypotheses:

of

upon

We

theories

concludes

for

bu i I d

a

static

Sa I eh i-Isfahan i

and

this

We use

and

tests

than

testing We

simple

competin9

cons i dered

Providin9

mode l i n wh i ch

rather

ear I i er

there

in

-

allows

is mOre appropriate.

dynamic model.

lower

model

directly rather

a

~

estimates

OPEC.

by explicitly

done earl ier.

whether

for

on their

that

hypotheses.

tour

model

stud i es

hypotheses.

by bu i I d i n9 equation

1982 he

model

one at a time.

test

competing

ri9hts

these

to

a competitive model,

concludes

their··--trork

competin9 uS

ot

paper

represent

data and

imp I i cat ions

the

to

model,

optimization

hypotheses extend

that

cartel

and

first

across

property

usin9 Griffin's

the

1971

cartel

--l3

model,

la9ged

is

test

with 0,1 whet~e"

and whether

t~~~~:

revenue appears to be the primary goal We also do tormal

countries. countries same

to see

I ags

In

a~rOSS

the various OPEC

it they have the same economic soals

and the

behav i Or.

-,Data I imitations in~luding ~ost

pre~luded

in their _model.

intormation allowed uS to more complete model testing tor serial appropriate,

testing

for any of the OPEC

either ot these studies trom Re~ently

in~lude

this

specification.

~ost

available

important variable tor a

Econometr i c

advances

correlation and correcting tor

I nc I ude

it where

testing tOr simultaneity using a Sims'

test and

estimating using 2 stage seemingly unrelated regressions where appropriate,

and paying somewhat more attention to creating

quarter I y trom annual

data.

We,

0+

course,

include mOre

recent

data than the""or I gina I stud i es, wh i ~h a I lows est I mat I on over periods at dramatic: price

increases as wei I as mere recent

price

decl"'eases.

I. Model For a producer of a nonrenewable reSOurce economic theory dynami~

sugest a

optimization model.

Hotel I ing type of behavior does not

Since assuming such a pre~lude stati~

behavior,

hypothes i zing such a mode I a I lows uS to test both stat dynaml~

behavior within this single tramework.

produce~s ov~~

maximjzjn9 the

p~~s~nt

j ~

and

We start with

value of profits from exports

a finite time horizon. T

MaxJ [f(y,q)q - C1 (R)q - CZ(wl]e- rt dt

o

3

subjec:t

to

where

is

c:ost

f of

the

demand

produc:t i On,

exploration,

w the

disc:overies

to

Hamiltonian

tor

Hw

=

-

C

R the level

date, this

D~der

The first

Zw e

func:tion,

q

I eve I of reserveS,

C

=

A I though

z

the

exploratory effort,

and G the p'roblem

disc:overies

output,

-rt

+

G w (¢l

-H

R

=

C

1R qe-

of

Sum of

func:tion.

all

The

IS

+ ¢z) =

D

rt

we cannot obta in an

of

wh i c:h

approx i mate with

we

q

exported J

C 1 the

conditions are:

func:tion

oi

c:ost

X the

ex~

lie i t

express i on

tor q,

t rom the above t i rst order cend l t ions that an

see

is c:alc:ulated

the

I eve I

demand minus

to be a

ot

reserveS equa I

non

OPEC

supply,

wi I I represent the world demand heav i I

y

imp lie i t

dependent

on

pI'" ice,

The quantity

fo I low i n'3 mode I . func:tion

of

the costs at extraction and exploration, the

ane can

would be

discDveries minus extraction.

is

is

....-

¢1

and

is

.~

•

011

Inc:ome,

of

Y

to

in i t i a I

OPEC

for

1n

mode I,

tunction. pI'" tee

4

and

the

demand

0

p I uS

iii s

PI'" ice

Sinc:e

for

interest rate·

reserves

Demand Our

the

or

and

nOn

income may

wOr

nelJJ

Id

Income

OPEC

s~~~

represE'1""

~

'3

reduced

form

for OPEC demand.

We wi I I

production directly alons with price

adds any

3nd

information to the estimation.

dr I I I ed represents effect

a I so add nonOPEC

ot

the

development,

exp I orat i On.

Initial

and

level

of

I ittlns costs

The

Income to

test

The number

if

"

ot we I Is

Intercept picks up the

ExplorationJ

reser"ves.

Since ,he

are entered directly.

inclusion ot extra variables does not bias parameter estimates we also

include

inves,ment

In tixed capital

target revenue hypothesis.

Ln OOIL = So

+ =

Where OOIL

0

I I

+ Sr

Our model

Sp Ln POlL r

t

+

tormatlon to

Sl Ln

+

Sw Ln WELLS lnv t

Sc Ln COST

(1)

exported.

POlL = current and/or

I assed rea I pr ices ot

0

i I .

= CUrrent and/or I assed non-OPEC tree wor I d

Ow

the

is:

Sq Ln Ow

Sy Ln GOP t

test

0

i I

production.

= current and/or I assed we I Is dr i I led.

WELLS r

= current and/or

lassed the

GOP = current and/or

lassed

In,erest rate.

i'ndi:::es of

sross domestic'

product of buyers ot OPEC 01 I. lnv

= current and/or I assed Investment in sross t I xed cap I ,a I

COST

=a

used as

five year

the

tar set

revenue.

runnlns averase ot extraction and

exploration costs per barrel. We wi I I

besln with curren,

also conduct a wide array ot

values

ot

al

var i ab I es but

lag testin9 to determine what

1 Ii 't. a j" e s already

In

percentage torm are entered directly,

1:.

j"

All

d

t

E::'

~ ,

ocher

w' I I 50rtS

variabl~s

ar~

~ntered

as

~ence

logs and

th~lr

cD~ttlCtents ar~

elasticities. II This

model

1986'IV Inr

Indonesia,

IS

the

fnr

Iran;

86' IV

Inr

UAE.

estimated

tn 82'IV

Gabnn,

Qatar,

nut

supplied

by Griffin,

Bulletin

and

the is

all

data

consumption,

Arabia,

fnr

Iraq

is

fnr

with

tn Algeria,

71:

and 72'1

tnn

The price at

updated by

-

and Venezuela;

and Libya;

and Ecuadnr

analysIs.

1971'

to

tn

much misSing

oi I

the OPEC Annual

in dollars)

Statistlcai

While Griffin used OPEC

we use export data because domestic pricing and

which are

countr.j~.s,

data

are available

the Mnnthly Energy Review.

production data,

data

nf

Technique

quarterly

Saudi

71,1

left

expnrt

an

where

Nigeria,

data are

these

and Estimation

cnuntries

Kuwait,

85'IV

Data

isolated

tram world markets

may be respond i ng to po lit i ca I

acquired by

adjusting Griffin's

in many of

goa j s. data

Our

thrnugh

1982

and updating tram monthly observations at oi I exports from the

Q.cJ... and Gas Jnurnal. ~uarterly

Wells

dri lied

basis but are created by

and real

the proxy GOP

for

numbers, made

Is

0

tor the

rea I GOP

On

A

one

experimenting with annual prnvldes

the

best

fit.

avai lable an

proxy tor

return

world an

data by

revenues.

nf

buyers

industrial

~uarterly

i I

rate

fnr

only available

into

lag on

the

nnt

a

i.nterpolations using quarterly

The

exploration data on rig counts. tac I ng OPEC

are

of

annual

nn U.S.

OPEC crude

Irnm

the

the

baSIS

0

interest rate

treasury

b I I : s'

I lis

'I

IMF.

an

ndex

n

t

Investment

trom the

IMF,

are

Interpolations based on a one year year

lag

is used since

regreSs)c~S

lags uP to three years suggests that ~r'lce

1r1

U.S.

dollars

is detlatea '-

t

the U.S. U.

S.

GOP deflatoc base yeac 1982.

Investment deflator base vear

Adelman and Shahi occurS

averagE' of It

canvert.ec

IS

do I I acs by the exchan'3e cat,. and th,.n def I ated by the U.S.

fixed

that

Investment

trom

(1989). year

to

the preVIOUS

1982.

Cost

taken

IS

Given the random vartatlQn year

for

each

year

five years costs and

we

took

from

In costs

a

moving

Interpolated to make

quarterly.

We estimate and test using seemingly unrelated regressions unless otherwise specified.

Given the difficulty

with differing sample sizes,

we estimate on 4 sample sizes 71

In

AI I

with

sufficient

eire done

In the

on

the

initial

data

longest

included saml=lle

In

each

In

which

of

the

the

COuntr

runS.

country

IS

i

es

The

"tests

IncluCleCl.

estimates the Durbin Watson statistic suggested

.-

tirst order serial which would

are

programml~g

correlation was a problem

lead to biased and

varIance covariance

mat~ix.

data was adjusted by a cho.

to~

al I but Iraq,

Inconsistent estimates of the

To Obtain consistent estimates

t~e

which was estimated uSln'3 a

Hi I dreth-Lu search proceduce for each el:fuation.

FurtherJ

some

the OPEC exportecs have si'3nificant market shaces. exports

i nf I uenc:e

bias estimates.

the pr i c:e ot

0

ii,

0

i I pI'" i ce endogene i ty

A Sims'(19721 exo'3eneitv t,.st on ,.ach ,.quatc-

rejected the hypothesis that the pr,c" of IranI

Kuwait,

the UAE,

and \Jene'zue)a. 1

I.Thls test was conducted by the

where dl tferent

The

trOm zero,

s

was

01

For

Including a

nul I h~pothes's the coeffIcient on the tutwre

equation.

WI

these countrIes

"'::, ~~

Jag on prlC~ for each c:o~~· is significantly future

rejected

ia9

!?x0genOus

instrumental

variable was substituted tor

regressing the price of oi I on

the price of oi

by

1

lags at the other variables.

III HypothesIs Testing Table 1 shows a wide variation

In reserves,

production,

costs, and absorption capacity tor OPEC economies. iable 1:

Variables representing OPECJs productIon capacity, costs, absorptive capacity, and export varIance. EXPOrts WELL PRODUCT! ON Average Range OIL RESERVES RIP DEPTH NUMBER 1000 I WELL ARTIF POP. PER CAP. 1000 1000 LIFT 1000 mb in yr teet WELLS bid bid GOP 101 d !of d

COLNTRY 1987

1987

1986

1987

1987

1987

1986

1986 1986 US$

71'1-86'~

===================================================================================

Algeria 8500 35.9~ 856S Ecuador 1615 28.18 78~8 Indonesia 84.00 19.4.0 4.501 1ran 92850 108.62 774.3 Iraq 100000 130.71 NA Kuwait 91920 229.78 3717 Libya 21000 56.4.1 1230 Nigeria 15980 35.34 NA Qatar 3150 30.39 7750 5audi A 166980-'12.85 5870 UAE 96605 194.47 86BB VenezueI a56300 96.89 NA

8~0

922 577 361 615 363 661 125 174. 5BB

680 979

771.~3 205 21720 2566 157 170.28 802 9650 1153 4.51 205.4.0 514. 16694.0 1186 234.2 64.87.53 0 ~4.21o 3362 NA 11120 1225 2096 34.08.13 20 1790 9556 1096 3019.28 NA 3600 5514. 1020 154.3.12 374 NA 98520 1239 988.83 284 1632.18 2 330 15000 NA 1154.0 64.46 4054 6894.56 1380 1545 1361 2001.4.7 223 162.50 851 17320 2808 1592

631

6~8

959 3113 1618 1538 1587 1593 420

6347 145Ll 1402

958 799

5355 3036 2702

2574. 1719 388 7769 1351 1809

===================================================================================

Data Detinitions' bid

= barrels

per day, mb

= ml I I ions

ot barrels, yr

= years.

There are approximately 7.4 barrels per metric ton at Sources:

01

i.

Qll and Gas Journal,

Dec. 1987 Opec Annua I Stat 1st ica I Bu I Iet In, International Financial Statistic

Despite these dltterences,

it OPEC is strictly a cartel Wit'"

some sort of market shar i ng scheme Or

it countr i es have

S I

m I! 21r

market structures,

we might expect simi lar estimated coefticients

acrOSS countries.

Our first

three hypotheses are to forma I: v

test thiS conjecture for OPEC and tor

two cores of producers

using equation 1 and current values of all

Hypothesis l '

variables.

All OPEC Countr les share a Similar market

structure and have similar ::oettlc entS.

a

Let

and j

repri?5e n

t

a I I OPEC countries,

th .. n H , I3 pi = I3 p j 0 I3 w1 = I3 wj 13'li = 13'lJ 13,; = l3,j l3 yi = l3 y j 13 j i = I3jj I3 c i = I3 cJ H1 '

Hypoth .. sis 2.

I3 p ; I3 w ; 13'li l3'i l3 y i 13 j i Sci

# # # # # # ,~

Sam .. as 1 but

I3 p j I3 wj 13'lj l3,j I3 YJ I3jJ I3 c j

I .. t

to, a I I

''''J

to, a I I

iloj

and J =

Iran,

lracl'

Kuwait)

Saudi A,ab;a, and UAE.

Hypoth .. s i s 3·-...· Sam .. as 1 but Saudi

I .. t

A,abia, UAE,

and j

=

I t'" an,

and test statistics vary, we report significance

t .. stin9

significance

null

•

level

hypothesis .

t

J

S'lua,e tests.

levels ot thes .. t .. sts and al I subse'lu .. nt

hypoth .. ses t .. sts a, .. 9iven

Sine .. all

j

and V.. n .. :u .. la.

These th'ee hypoth .. ses a, .. t .. st .. d usin9 Chi Th .. sl9niticane ..

1 r aq, Kuwa

In Tabl .. 2.

Since de9rees ot

tor economy and clarity of expositionl

levels rather than the test statistic.

is don .. at th .. 57. Si9nitieane .. less

tr~edom

than 5". results

in a

lev .. l, any

rejection

of

the

Table 2: HI

Signi t icance

I nd 07.

~

0%

Levels tor a i

lrn It:.s LJ% 3%

Kuw 0%

Lib 07.

I

Hypothesis Tests

lli..s Sau UAE Ven 0%

07.

07.

H2

07.

1r.

07.

0%

0%

H3

07.

0%

07.

07.

0%

0%

H4 H5

997. 100% 0%

267.

Nu I I Hypothes I s All OPEC Same

0%

(5) MidEast Same Core (6) Same

217. 99% 1007.

887. 1007.

94% 100%

947.

NonDynamic

47% 377.

40%

717.

647.

No Property rights

15%

587.

247.

H6 1007.

887. 100%

1%

8% 100%

OPEC Non Compl;ltitlve

H7

2"'.

21%

197.

3%

0%

6"0

07.

0%

89%

0%

OPEC Non Monopoly

H8

07.

07.

0%

07.

0%

0%

07.

0%

oro

0%

Stron9 Target Rev

H9

07.

17.

46%

07.

07.

16%

07.

267.

0%

0%

Weak Target Rev

17% 1007. 100% 100%

A I I hypotheses that a I I countr i es Or a cOre ot

-

countrie5 are simi lar

at all

variables.

testIng we

are strongly rejected using current

Before proceeding to complete Our hypothesis

investigate whether current values or some

structure better explains these countries)

separate regressions are run with

lags

behavior.

lag

To do so

from 0 to 20 tor each

The 10.9 len9th is chosen that minimizes

variable except cost.

2~ )/T,

the SchwarZ (1978) CriteriOn = (RSS + K 109(T) is

values

the regression Sum of squares,

where RSS

K is the number of regressorS,

r is the estimated standard errOr ot the re9ression and T is number

of

because oui It

it

into

observations. IS

it.

a

five Table

Cost

year 3

not

IS

running

containS

including

average the

lag

0+ the variables usin9 this procedure.

to

and

in

the

already

len9th

chosen

the

testln9

has tor

lags eac~

Table 3,

Lag Lengths Chosen by the Schwarz Criteria. .E..Q.i..!.. 1

Algeria

We I Is 20

!.

0

Inv 0

Ow 0

Indonesia

1

0

0

1

0

0

Iran

3

0

0

0

8

16

Iraq

8

9

0

0

9

8

KuwaIt

0

0

0

17

0

0

Libya

0

0

0

0

0

8

Nigeria

1

0

0

4

0

3

Saudi

0

0

20

0

0

20

UAE

0

0

0

4

0

0

Venezuela

3

12

8

12

4

0

Arabia

The

lag testing most often suggests that the

fits better than a distributed are

appropr

I

a.tow"

behavior the

lags added any

lags that added

information.

Interest rate tel""

Only

Significantly

lags

In

Included

Only those

is retained

lags whose sum

in the model

IS

The

In the final

income for Kuwait and the

Iraq.

the coefficient on Non-OPEC free world praductlun

IS examIned to determine

model

investigate

information and were retained

estimation results are those on

Finally,

countries but are

is reestimated to determine

significantly different from zero only

value

Where lags

lags resulting from the above testing are

in each equation and the model

whether

acrOSS

To further

less.

I:urrent

lag of the variable.

they vary cons j·der-ab I y

generally 8 quarters or

•

GOP 1

tor

if

Nigeria and

dittere'nt

from

it

added any

Iraq

IS

:erO

leadIng

: I

Its

information to the coefficient to

its

inclUSion

11""\

~ .....

prete~~ed results.

These results,

subsequent hypothesis testing,

whIch are used tor al I

are given

In Table 4.

The +Irs:

row of numbers next to the sample years are the estimated caeft i c i ents wh i I e the second row of numbers are the t statistiCS,

The R2 s variation fOrmal

Impl~ that between 36 and 94 percent ot the

In exports

explained by these variables.

IS

AlthOugh

tests did not find countrIes to have the same

coefficients,

there are a number of qualitative similarities

acrOSS countries.

The coefficient on wells

and most often positive. faster

Thus,

always

IS

inelastiC

dr I I I I ng has tended to ta I I much

than exports suggesting excess capacity.

The

Significantly negative coefficient for Nigeria and Iraq may Suggest difficulty increased as

in maintaining exports since wells drilled

~~¢orts

decreased.

As would be expected, negative,

the coefficient on cost _is most etten

The coetticlent on the price ot

negat I ve,

that On the

while that

On

f

nterest rate

investment

signifIcant variables.

IS

The

01

is a I most a I ways negat I ve,

implications of these coeffic:cnts;:;r formally,

The negative coefficient on the prIce of conSistent with dynamic optimizatiOn With price riSlns and exports

exports pass I b

I

I""

I

I se to make uP tor

I ties

W I

I I be conS

most otten

always positive and One at the most

market structure are nOw examined

noncompetitive static world,

IS

+aill~g

Or

With

ta I

leer"

'?d.

01

could be

in a Hoteiling type 0+ over

tim~,

WO~

with a.

a target revenue goal

Each ot these

Wh~~~

•

Table 4'

Econometric Estimates ot Cf'EC Exports

AIgeria

C

1971 198b •

Po I I

t stat Indonesia

Po i I

C

1971 198b Iran

Po I I

C

1971 1985 t stat

Po i I

(

1973 1982 -21.06 -5.15

t stat

13.56 14.39

Libya t stat

1971 1986 t stat

10.15 4.31

UAE

r

1972 198b Venezuela

Pei I

(

1971 1986

A key

test

ot

coefficient On the nterest in

decreasing

rate the

it

r

-0.03 -2.77 r

-0.01 -1.56

0.05 0.90

r

-0.02 -2.75

Wells

6.13 -0.36 7.87 -5.23

t stat

-0.01 -1.17

Wells

3.23 -0.17 3.13 -3.52

t stat

0.02 0.47

0.17 3.68

r

-0.01 -1.59

whether

interest WOU

Id

ground

Y Inv Cost -1.62 0.62 -0.74 -2.39 12.79 -1.66

OW

Y -0.48 -1. 28

QW

Inv

the

Y Inv Cost 1.79 0.59 -2.42 1. 31 11.55 -3.39

OW

0.78 1.89 8.00 rho

OW

0.61 2.29 5.65 rho

QW

OW

0.68 1.82 5.84

Y Inv (ost OW -3.77 0.70 -0.20 2.54 -2.40 17.49 -0.87 3.97

Y 2.56 1.96

I nv

Cost

rho

OW

(ost

QW

rho

0.47 -0.43 7.86 -1.23

OW

'lnv (cst Y 0.07 0.34 -0.31 0.15 8.25 -2.36

OW

Y -0.84 -3.20

lnv

(cst

0.18 3.90

0.37 1.01

0.94

OW

R2 0.Q3

OW

model _is

static

rho

rho

~

!;J-",,"

0.55 1.68 4.18 OW

0.41 1.80 3.08

or

dynamic

R2 0.61

0.54

OW

rhe

R2 0.40

.."'"7

OW

0.78 1. 91 8.80

OW

R2 0.80

1.72

0.76 1. 75 6.38

(ost Y lnv -2.07 0.51 -0.16 -4.12 10.00 -0.81

R2 0.36

P:

0.45 1.88 1.89

0.25 3.16

R2 0.74

OW

Y Inv rho Cost QW OW 1.54 0.19 -0.30 -1.01 0.35 2.02 3.17 8.93 -2.37 -3.39 2.59

:5

0.6'9

0: is

C

':-'0:>

rate.

increasE' and

Cost

rho

0.20 -0.18 3.67 -2.09

Y(Ll7) Inv r -0.04 -4.84 0.55 -4.39 -13.90 14.88

Wells

Po I I

(

011

0.01 0.31

0.06 -0.13 0.96 -3.53

3.96 -0.40 7.51 -3.99

t stat

over

r (L9)

We II s

Po i I

(

1971 1986

I

Wells

-"15 0 I I Wells

(

Saudi Arab

the

r

0.01 0.82

-1.35 -0.62 -0.15 -0.98 -5.19 -1.43

Nigeria

r

-0.02 -2.79

0.11 1.05

We I Is

0.09 1.43 Po i I

(

1971 1982

•

Po I I

(

1975 1986

0.11 3.70

0.40 -0.54 -0.35 2.80 -7.70 -14.28

t stat Kuwait

-0.02 -2.33

Well s

-1.18 -0.49 -0.54 -4.37

Iraq

0.33 7.48 Wells

3.28 -0.07 5.21 -1. 17

t stat

r

Wells

2.85 -0.27 4.06 -3.38

the

should

va I we

inCrease

should decrease output.

13

Thus

at

a i I

output,

Br ) 0

j

n

the

bar""

while !S

cons

s:~-:

In a static model

with a dynamic model.

only a cost at production. in~rease

An

the

i ncr*eaS i n9

lntere'st rate

r*ate at

IS

i ntere'st wou I d

costs at production decreasing exports and yielding a HypothesIs 4

IS that COuntries do not behave

,n a

dynamic manner verSUS the alternative that they do Or: Hypothesis 4;

Countries are dynamic or

tor each OPEC

~ountry

for each for each SurprisinglYJ

in no case can we reject

the nul I hypothesIs 1

In favor of the alternative that countries behave dynamically.' However,

both the countries and the companies have produced aver

the sample period with the control

0+

exports transferred over

time from the multinationals to OPEC.

Property ownership

•

ar9uments su~est that they mi9ht have different discount rates The

COmpan j es

r j sk i n9

nat i ona Ii zat i On

may

interest rate than the countries social

have had a higher

rate of

interest.

Alternatively Adelman (1986) ar9ueS that countries whose ecOnom i es are very dependent on an u'nstab leo i I market shou I d have had hi9her discount rates than the companies. event J the use of One

j

In either

nterest rate m j ght reSu I t i n a

fa i I ure to

detect dynamic behavior. To

test th i 5

conjecture we a I low separate discount rates

the companies and countries and retest to determine whether

":::Jr

t~ey

1.These results supersede Our preliminary results in Dahl and (ucel(l988). After updatin9' checkln9> and correctin9 al dat~ and transtormatlons and uSing a mOre complete and correct spe~I+lcation, we no longer tlnd much evidence conSistent w!: . . . dynamic oPtimization. 14

•

each display static

dynamic: behavior and

01""

it

We hypothes i ze that the soc: i a I rate

ot the private rate .

•

IS

their behavIor ~l""

I

5

r = [r(l-G)

+

some percent

Under this hYPothesis the rate ot

a weighted average ot the private and the social ~rG]

IS

interest

rate Or'

(2)

• Where Gis the percent ot OPEC output contro I I ed by the OPEC countries Or the country participation rate and (l-G) share controlled by the multinationals. expression

into

(1)

Substituting this

gives us Our testing equatIon:

+ Sp Ln POlL + Sw Ln WELLS + Sr [r(l-G) +

Ln QOIL = So

is the

+ Sy Ln GOP

+

+

Sl Ln Inv

~rG]

Sc Ln COST (3)

This equation al lows uS to test property rights arguments or

whether the countries and the multinationals behave the same

• against the alternative that they behave di tterently or'

Hypothesis 5'

Multinationals and OPEC countries have the same

discount rate Or tor each OPEC country for each tor each Only for Algeria, where the coetticient for Algeria negat i ve and

5

Igo it i cant wh i I e that tor the c:aml=!an i es

IS

is I=!OS I

and significant, do the multinationals and the country behave

a dissimilar manner. signIficantly ne9ativel dynamically ol=!timize.

Since the coettlcient tor Algeria there

IS

No!"" does

sti I I

Given nO evidence analyze the

for

I VI?

,n

is

no evidence that countries

this test support prOl=!erty r:9 h tS

arguments.

•

t

dynamic optimization we p!""oceed to

implications of behaVior

15

in a static:

f!""C1mework.

For static behavior,

we can further

test market structure.

competitive world we know that price equals mar9inal

1n

cost and

would thus expect price and quantity to be directly related. nul I hypothesis af no competitive behavior

a

is tested against

Tne thiS

alternative that the countries behave competitively' Hypothesis 6: country

OPEC countries are competitive Or

for each OPEC

I:

Only for behavior.

the case

= 0

for each

> 0

for each

Iraq do we reject

in favor of competitive

An upward sloping supply curve would be neCessary but

not necessarily sufficient tor competitive behavior. test

for

•

market structure we

investi9ate the

To further

impl ications of

•

monopoly beha¥jor. 1 t OPEC countr i es are behav j n9 wou I d eXl=lect that

income

In

ina monopo Ii s-t i c manner,

we

i ndustr i a I j zed CDuntr i es and perhaps

NonOPEC supply would affect export patterns.

Althou9h a supply

function may not exist tor the monopoly case,

comparative statics

shows what the s i 9nS on P and Y m i 9ht be expected to be.

Tota i I

differentiatin9 the first order condition MR - MC = 0 9;ves'

( ~M R I [jQ - [jM C I [jQ l (0 t rom sec 0 n d

0 r

de r

con d i t ion s,

wh i I e

aMR/~L, eqUa Is:

= [jCP zero, a sufficient condition

for dP/dQ to be negative

Thus,

= gp/aQ

+ gP/dy dy Or dP/dQ

dQ

is a negative dQ/dy.

is positive,

In the mOre

the sign of dP/dq

ikely

ambiguous.

IS

a negative significant coefficient on Y wi I I require a

negat i ve coe"f""·j c i ent

on pr ice

tor

uS

cone I ude

to

in

favor

of

monopoly.

The nul

hypothesis

is

monopoly behavior against the

nO

alternative at monopoly behavior.

Hypothesis 7'

Monopoly behavior Or for each OPEC country

Ho·.

= 0

for each

HI'

t. 0

for each

We can see trom Table 2 that the candidates for monopoly behaVior

Saudi 1$

by

testing

Arabia,

are Algeria,

and Venezuela.

However

negative and Si9nificant and

SignIficant which

•

Income

Further,

SInce

is

export

IS

in

r raq)

Kuwait, the

the sign on price

inconSIstent

supply

Iraq,

with

gn

on

I

nc::Jme

is positive

a~d

monopoly behavior.

upward slopins and

are POSitively COrrelated With mOre com~etitive

17

5 j

Nigeria .

Iraqi

exPOr:s

NonOPEC sUPP'

con~lude

that

IraqI

behavior

than monopoly behavior. changes

Iraqi

is more consistent with competitive

Whether the ending of

the war with

behavior to be more consistent with other Middle

Eastern Countries remains

to be seen.

problems suggests movements

Investment

is the

Recent resolution of

last variable to be discussed.

in Our sample

are a major source of

leadin9 to the strict

form at

in 1986 was over

total

~wota

in that directiDn.

The

correlation between Dt I revenues and GOP per capita across cDuntries

Iran

.90.

income as wel

last hypothesis, this hypothesis

0

i I

; nvestment

revenues income

In the

the target revenue model. let

Inv* be the target revenue.

Then Inv* = QOIL-POIL or QOIL = Inv*/POIL. tormulation ot this hypothesis

as

Thus,

the

IS for

A 109 I i near

the coetticient on POlL to

be -1 and that on Inv* to be + 1. Hypothesis 8:

•

Target revenue stron9 Or for each OPEC country Ho ' I3 p i=-1 and 13li" 1, HI'

I3 p i#-1 and 131;# 1

As can b .. s .. en

in Tab I .. 2,

r .. jected for al I countri .. s the hypothesis which

th i s .hypothes i SIS stron91

leadin9 uS to test a weak .. r

form 0+

is tor the coettici .. nts on POlL to be

ne9ative and that on Inv* to be equal Hypothesis 9:

y

Tar9 .. t

and opposite

in Si9n.

revenue weak Or for each OPEC country

Ho : I3 p i=- 13 11 I3 p i#-l3j

I

Three countries do not hypothesis tests

of

Iran,

the target

Libya,

reject

the weaker

and Saud!

revenue model

form of

the

Arabia.

are most otten rejected,

1d

•

,

.

investment

is posItive and significant

most equations

•

e~uation.

In every

In

is the most significant variable sU9gesting

it

that a significant portion of oj I export revenueS are earmarked tor sroSS domesti~ ~apital

tormation, onl~

Indonesia and the UAE are the with at

least one ot the hypotheses

least percentage variation

over

production

ratio,

ot the hishest,

wells

less

over 190

•

a I thoush

I

n~nes

coefficient on

I

0+

than

20

~ears.

tar

lowest reserve

whi Ie

th5' UAE

statisti~s

is

a I most

as

One

implYing

i mpOt'tant,

I east S i sn it i ~ant

a had the sma I I est and

the countries tested.

investment of al

has

sussest that

Indonesian eXl=lorts,

Investment

the

al I countries tested but

years,

The t

~onslstent

not

Both have shown

Indonesia has the

is the best forecaster

I tttle excess capacity .

tested,

in exports

are extremes otherwise since

~ountries

Indones.3

• a I so had

the

sma I I est

percentage vat"' i at

government I='artic:ij:'ation variation

in

exports

For the UAE,

r"ate,

and

b~

is

far

eXPorts with price running second. and the

fa~t

that the

~oetti~lent

Investments are opposite Important

but not

In as

on

I

n exports,

smallest

percent

these economic

explained by

investment

the

j

the best

This on the

lowest

of

variables. p~edlctor 01

impol"'tanc:e of pri~e

thE?

InveStment

ot 0; I and

in sign suggest target revenues are strong

a

torm as either hypotheSIS testec

here. Our testing results have

•

Saudi

ImplicatIons on the hYPotheSIS

Arabia or some cOre of countries act as swing produce-;

Since a swing producer would be noncompetitive, +01'"

':-~':

sWing

producer

the

candldat~~

t,..om the ab8vl; analysis are Algeria,

19

NI~eria

·- .

Saudi

ArabiaJ

Kuwait,

V~nezuela.

and

that a swing producer would show

the sample.

larger SWIngs

Examinln9 the eVidence

t i ve have had

a I I

large sw i ngs

in

lcan and lcaq have had

exports.

Further,

we would expect In

production Over

In Table 1 we can see that

exports

as

a

percent

at

average

lacge swin9s as wei I but they

may have been more war and revolution related since the testing target revenues may be motivating their

sU9gested competitive Or

behavioc. As with any study of this natuce data quality and mw I t i co I I i near i ty

course 0+

present

We have found ovec the

prob I ems.

the study that Our

specification and urge the

results are somewhat sensitive to

reader

to view

fcom Our most·1:!"tjmplete specification

the present conclUSIons

In that

•

li9ht.

lV Conclusions So

is OPEC a cartel?

Althou9h a

surrounds OPEC decision making,

lot of uncertainty sti I I

econometric model

Our

developed

out of dynamic optimization sU9gests that various OPEC countcles seem

to behave

sharin9 cartel

in quite dissimi lsI""

hypothesis

is not appropriate.

any core of countries that had ike

countries did not behave ot noncom~etitive behavior Arabia, large to

and Vene:uela.

swings

qua Ii ty

strict

as

Nor did we find

coefficients.

cartel

tal"" Algeria,

ThiS

countr i es

identical a

there

is

evidence

Kuwait, Nigeria, Saudi

noncom~etitive

in production but

these

Hence a strict market

ways.

diSSimilar

behavior coupled

coefficients

sw i ng produce'rs

rather

leads

than

as

With

uS

a

car te! .

We were disappointed to optimizatIon using Either

find

current

I ittJe

or

20

evidence at dynamic

distributed

lags

11n

inter~s:

., Althou9h such myopic behavior ml9ht wei I be qUite

rates. rational

work

in a highly uncertain environment.

This

econometric

is consistent with the disappointin9 results obtained by

dynamic optimization models and SUPPorts efforts of modelers to

•

move away

•

from them.

dynamic behavior! adjustment. only buyer

In addition to not tindin9 eVidence tor

la9 teStin9 sU9gested rather short

la9s

In

For many variables current values were preferred and

in two cases Income

(la9ged

interest rate

in the Iraqi

in the Kuwa It i equat i on) did

equation and

I a9S add any

information to the estimation. the short term nature of Not

surprisingly,

the decision process.

there

IS

little evidence

that

companies

dynamically optimized either' since the majority of the mu It i nat i ona I product i on

•

nationalized. These

• 90al

for

; n these nat Ions was 9radua I I y

Nor was the property ri9hts ar9ument supported.

IS evidence that some form of target revenues may be a

Iran, Libya,

and Saud; Arabia.

However,

tests of the target revenue model Were rejected of cases, impOrtant,

forecasters should note that usually the most

important,

• 21

although tormal

in the majority

investment explanatory

is sti I I an variable .

.Retl:lrences

Adelman, M.A.

(1986).

UOi I Producing Countries 8'309-329.

Resources and Energy.

•

Adelman, M.A. and Shahi, M. (1989). uOil development-operating cost estimates, 1955-85,u Energy Economics. 11(1) '2-10. Dahl, Carol A. and Yucel, Mine (1988) uOPEC Market Behavior,u Compet i t i on, Prot i tab i I I ty, and Po Ii Cy ~ the Energy 1ndustries, Proceedings ot the 10th Annual North American Meeting ot the International Association tor Energy Economics, Houston, TX, Oct. 31-Nov 2, 1988, p. 8-19.

In

Fischer, D.; Gately, O. and Kyle. J. F. (19751. uThe Prospects tor OPEC' A Critical Survey ot Models ot the World Oi I Market. u Journal ~ Development Economics 2 (4), 363-386. Gately, Dermot (1984). UA Ten-Year Retrospective' OPEC and the World Oi I Market. u Journal ~ Economic Literature XXII (September)' 1100-1114. Grittin, James M. Hypothesis.

JJ

(1985). uOPEC Behavior' American Economic Review

A Test ot Alternative 75 (5), 954-963.

Hammoudeh, Shawkat (1979). uThe Future ot Oi I Price Behavior ot OPEC and·-~audi Arabia' A Survey ot Optimization Models. U Energy Economics (July)' 156-199. International

Monetary Fund,

Washin9tcn,

Oil

D.C.,

International

varIOuS

and Gas Journal, Pennwell

Financi-al

•

StatistiCS,

Issues.

Pub'

Tulsa, OK., various

Issues.

Organization ot Petroleum Exporting Countries, Annual Stat,st'cal Bulletin.,

Vienna,

Austria,

various

issues.

Salehi-Istahani. D. (1987). uTesting OPEC Behavior' A Comment Mimeo. Virginia Polytechnic Institute, Blacksburg, VA. Schwarz. G. (1978). UEstimating the Dimension at a Model,u Annals Ql Statistics. 6'461-464. Sims, C.A. (1972). UMoney. Income, Economic Review, 62:540-552. U.S.

and Causality,U

American

Department of Ener9y-Energy Intormation Administration, Monthly Energy Review, Washington, D.C., various Issues.

•

RESEARCH PAPERS OF THE RESEARCH DEPARTMENT FEDERAL RESERVE BANK OF DALLAS Available, at no charge, from the Research Department Federal Reserve Bank of Dallas, Station K Dallas, Texas 75222

• 8801

Estimating the Impact of Monetary Policy on Short-Term Interest Rates in a Rational Expectations--Efficient Markets Model: Further Evidence (Kenneth J. Robinson and Eugenie D. Short)

8802

Exchange and Interest Rate Management and the International Transmission of Disturbances (W. Michael Cox and Douglas McTaggart)

8803

Theoretical Macroeconomic Modelling and Qualitative Specifications of the Bond Market (William R. Russell and Joseph H. Haslag)

8804

Augmented Information in a Theory of Ambiguity, Credibility and Inflation (Nathan Balke and Joseph H. Haslag)

8805

Investment and the Nominal Interest Rate The Variable Velocity Case (Evan F. Koenig)

8806

Tax Policy and Texas Economic Development (Stephen P.A. Brown)

8807

Unionization and Unemployment Rates: A Re-Examination of Olson's Labor Cartelization Hypothesis (William C. Gruben)

8808

The Development and Uses of Regional Indexes of Leading Economic Indicators (Keith R. Phillips)

8809

The Contribution of Nonhomothetic Preferences to Trade (Linda Hunter)

8810

Evidence on the Two Monetary Base Measures and Economic Activity (Joseph H. Haslag and Scott E. Hein)

8811

The Incidence of Sanctions Against U.S. Employers of Illegal Aliens (John K. Hill and James E. Pearce)

8901

An Econometric Analysis of U.S. Oil Demand (Stephen P.A. Brown and Keith R. Phillips)

8902

Further Evidence on the Liquidity Effect Using an Efficient-Markets Approach (Kenneth J. Robinson and Eugenie D. Short)

8903

Asymmetric Information and the Role of Fed Watching (Nathan Balke and Joseph H. Haslag)

8904

Federal Reserve System Reserve Requirements: H. Haslag and Scott E. Hein)

8905

Stock Returns and Inflation: Further Tests of the Proxy and DebtMonetization Hypothesis (David Ely and Kenneth J. Robinson)

•

•

,

1959-88--A Note (Joseph

8906

Real Money Balances and the Timing of Consumption: Investigation (Evan F. Koenig)

8907

The Effects of Financial Deregulation on Inflation, Velocity Growth, and Monetary Targeting (W. Michael Cox and Joseph H. Haslag) Who Really Bears the Risk?

An Empirical

8908

Daylight Overdrafts:

8909

Macroeconomic Policy and Income Inequality: An Error-Correction Representation (Joseph H. Haslag and Daniel J. Slottje)

8910

The Clearing House Interbank Payments System: A Description of Its Operations and Risk Management (Robert T. Clair)

8911

Are Reserve Requirement Changes Really Exogenous? An Example of Regulatory Accommodation of Industry Goals (Cara S. Lown and John H. Wood)

8912

A Dynamic Comparison of an Oil Tariff, a Producer Subsidy, and a Gasoline Tax (Mine Yucel and Carol Dahl)

8913

Do Maquiladoras Take American Jobs? Results (William C. Gruben)

8914

Nominal GNP Growth and Adjusted Reserve Growth: Nonnested Tests of the St. Louis and Board Measures (Joseph H. Haslag and Scott E. Hein)

8915

Are the Permanent-Income Model of Consumption and the Accelerator Model of Investment Compatible? (Evan F. Koenig)

8916

The Location Quotient and Central Place Theory (R. W. Gilmer, S. R. Keil, and R. S. Mack)

8917

Dynamic Modeling and Testing of OPEC Behavior (Carol Dahl and Mine Yucel)

•

(Robert T. Clair)

•

Some Tentative Econometric

•

•