l. ., are ilev0..mi. as was -uggtiiId above.
The hvp,,heLi- tests for market inieyratlion
then involve part .ular elements til the F and


B, paraimeter matrices. PjirllluinllLi the r
matrix with nm ..llnulllinullICe in group I and n
commodities in group II.


I YL2 -
Y l 1 ...

mry~


7,...' :Yscllrmrll Ya.3.3..'-~


The corresponding partitioned 1, marix is (dropping the third subscript since we are s-
 suiing a %ingle la l:




 U., 0. Pi P ... P. =.,. "- Pl...
 13:

 m. 13..: ... .- .t- P; *. i *l -, n.



 P...l.l PI -. a. ... .. ..2.ar. .... '. 1 .* P -. -.. .


 Reci-iin- the null hlIpIthlies that the ele-
ments ilb the upper rihln-h.ind hilK.ks and lower
left-hand blocks of and H, are jointly zero
implies th;i the markets between lommlllodliie.
in group I are inltgratld ith the commodities
in gnip IIt. The products in grunip I cnlJd be
considered lke products w ilI the owni iunldies
in group I,. Conversely, failuie to reject ihe null
hypothesis ,uget IhaJL thei rei is mn.uljicnl ev -
idence to .n'*,lu'hde that the commodities inthe
two groups ae Iike prOuctl; their prices move
independently. so they are egmnented markets,
Additional tests could be done to confirm that
onu'lniiiitieie' uwitihin each group are integrated
with each other, foir example. re.icLin.i the null
hvwlphe'.is that the upper lf blocks 4t I and
II. are joiinll% zero would confirm that the mar-


kets for commodities u lhin r.ip I are inte-
gr:L[rel with each other. However, the primary
interest is presumably the first test between the
cl.nlc.'std tonmiruldiily groupp Il versus thc ques-
tioned like li-idic' (group I),
 Es'tmation of the above system can be
done by any simultaneous equaiimon4 estimator.
GCten the cross-equation re"lr.iiionsl' to be
tested, ihrec- a~gc least squares is a reasonable
choice I'm an estimator. By inspection, each of
the equations in the system is idennuied via
exclusion restrictions on the lagged quantity
variables The hypotheses are all linear and


 Because tljh test is .',li'.iiril n ll on the result of
the first test, iL!'niLdniLe levels of the se"A d amn sub-
sequCet icsi should be .dNIICCJ accordingly.


17, r=-


Ytm'2~


i :


31


~1


~Ym+~ x


I~t~L. ~III


Jorrrimfol ond Appfird 40CF1V YAW


I