this pdf is a selection from an out-of-print volume from the … · 2020. 3. 20. · t the author...
TRANSCRIPT
This PDF is a selection from an out-of-print volume from the NationalBureau of Economic Research
Volume Title: The Role of the Computer in Economic and SocialResearch in Latin America
Volume Author/Editor: Nancy D. Ruggles
Volume Publisher: NBER
Volume URL: http://www.nber.org/books/rugg74-1
Publication Date: 1974
Chapter Title: Economic Alternatives for Mexico: A Quantitative Analysis
Chapter Author: Alan S. Manne
Chapter URL: http://www.nber.org/chapters/c6623
Chapter pages in book: (p. 211 - 230)
IECONOMIIC ALTERNATIVES FOR MEXICO:A QUANTITATIVE
ALAN S.Stanford University
11. INTRODUCTION
This work is intended as a step toward multilevel planning—identifying theinformation flows necessary for consistency between project, sectoral, and macro-economic decisiqns. (Parallel process-analysis models of Mexico's agriculturaland energy sectors have been constructed.) To build a macroeconomic model thatwill yield shadow prices relevant to specific project decisions, it appears essentialto allow for labor inputs. Otherwise, it is implied that the marginal productivityof labor is zero. With a constant-returns technology, this, in turn, implies that themarginal productivity of capital will coincide with the economy-wide averageoutput-capital ratio. As noted by Harberger (1967, p. 142), a zero shadow pricefor labor (hence a high shadow price for capital) is "virtually a kiss of death forprojects with long gestation periods or long economic lives.'9
For a macroeconomic model to generate meaningful criteria for project deci-sions, it is not sufficient that "labor' have a positive shadow price. Since theskill-mix differs substantially among alternative investment projects, and sincewage differentials between skill groups appear quite wide in Mexico, it is essentialto separate labor by skill categories. Skill disaggregation is easier said than done.
In our initial experiments, we adopted a manpower requirements formulation—hoping to avoid the data difficulties inherent in a human-capital-formationapproach. The manpower availability in each skill category was projected exoge-
though the supply of labor skills were completely inelastic. Employingthis formulation, plus an activity-analysis technology that turned out to be virtuallyas rigid as a Leontief system, we ran into major difficulties with respect to theshadow price of labor. The efficiency price differentials between skills were eitherzero—or with a minor perturbation in labor availability, these differentials becameunbelievably large.t Under the influence of Marshall's dictum ("natura non fad:saliwn"), we have therefore searched for additional elements of substitution and
°A revised version of this paper will form a chapter in a forthcoming book titled Multi-LevelPlanning: Case Studies in Mexico, North-Holland Publishing Co. Data-gathering and computationswere supported by the Development Research Center of the International Bank for Reconstructionand Development. The results were written during a year in which the author held a Ford Foundationfellowship at the Center for Advanced Study in the Behavioral Sciences.
t The author is indebted to Leopoldo Solis for having provided access to the resources of theDepartamento de lEstudios Económicos, Banco de Mexico. Helpful comments on successive draftswere received from: Bela Balassa, Gerardo Bueno, Yves Franchet, Louis Goreux, Donald Keesing.Janos Kornai, Mordecai Kurz, Saul Trejo R., and Thomas Vietorisz. The drafts were typed carefullyand cheerfully by Maureen Seymour. All computational aspects—including programs for matrixgeneration—were handled by Richard Inman with the assistance of Enrique Novelo Berrón. Thespecific facts, methods of analysis, and conclusions are the sole responsibility of the author.
'Much the same experience is reported in Bruno (1966, pp. 343—45). Apparently, in both thesenumerical models, there was insufficient indirect substitution via international trade to avoid knife-edge behavior of the shadow price with respect to the exogenously specified availability of labor skills.
2112 Economic and Social Research in Latin America
have turned away from regarding labor skills as a demographically given primaryfactor of production. Much like Correa and Tinbergen (1962); Spiegelman, Baum,and Talbert (1965); Adelman (1966); and IBowles (1967); the current version of
includes endogenous time-phased activities for upgrading unskilledinto skilled manpower. Also included are activities for capital-labor substitutionin agriculture and for short-run substitution between skills. Our calculationssuggest that Mexico is approaching the end of the labor-surplus phase of herdevelopment, and that capital-labor substitution could become increasinglyimportant.
With the exception of labor, the dynamic multisector linear programmingmodel (for short, IDJ[NAMIICO) follows along familiar lines. Among the standardingredients are: a 115-sector current-account interindustry matrix, capital coeffi-cients linking investment demands to capacity expansion in each of six future timeperiods, and alternative activities for trade-balance improvement.2 part, foreignexchange is viewed as an exogenously given primary resource—and in part, as anitem for which there exist substitution possibilities. un addition to foreign-exchangeearnings .through traditional exports and tourism, the model allows for the possi-bility of exporting manufactures from high-cost "infant" industries, lit is supposedthat foreign exchange is also available through capital inflows—one portionon concessional terms and another portion through direct private foreigninvestment.
Altogether, the programming matrix contains some 300 constraint rows, 400activity columns, and 4,000 nonzero coefficients. At this size, numerical optimiza-tiion did not prove to be a bottleneck. The model evolved during a two-yearperiod of experimentation. With a continuing series of improvements in the basicdata, there were six successive versions of the "basic case." Eventually, a special-purpose program was written ito generate the matrix and to facilitate datarevisions.
2. ALTERNATIVES TO THE BASIC CASE
For ease in future reference, we define the "basic case" to be the one describedin memoranda 71-9, 71-12, and Table 1 contains a list of the eight alter-natives to be evaluated here. These eight cases refer to alternative formulationsof constraints on the primary factors of production: foreignexchange and labor. Also considered are alternatives to the maximand adoptedfor the basic case: aggregate consumption, subject to a "gradualist" constraint onthe time path and a target annual growth rate g = 7 percent. The eight alternativesare examined one at a time—neglecting interactions between them.
The individual alternatives stem not only from different value judgements asto what is desirable, but also from different practical judgements as to what is
It is assumed that the reader is already familiar with standard references on dynamic numericalplanning models, e.g., Adelman (1966); Chenery and MacEwan (1966); Bruno (1967); Eckaus andll'arikh (I 968); Bruno, Dougherty, and Fraenkel (1970); and Murakami. Tokoyama, and Tsukui (1970).
These reports are available upon, request to the Development Research Center. Internationall3ank for Reconstruction and Development. 1818 H Street, N.W,, Washington, D.C. 20433. Eventually.they will appear as chapters in Multi-Level Planning: Case Studies in Mexico.
TAB
LE I
ALT
ERN
ATI
VES
TO
TH
E B
Asi
c C
AS
E
.
Ass
umpt
ions
Und
erly
ing
the
Bas
ic C
ase
Alte
rnat
ive
Ass
umpt
ions
Iden
tifIc
atio
n N
umbe
rof
Alte
rnat
ive
Cas
e
Max
iman
d: a
ggre
gate
con
sum
ptio
n, su
bjec
t to
grad
ualis
tre
stric
tion
and
annu
al g
row
th ta
rget
g7%
Max
iman
d: sa
me
exce
pt g
row
th ta
rget
g =
6%M
axim
and:
sam
e ex
cept
gro
wth
targ
et g
=8%
Max
iman
d: d
isco
unte
d co
nsum
ptio
nM
axim
and:
term
inal
con
sum
ptio
n
1 2 3 4
Initi
al d
iffer
entia
l of 3
0% b
etw
een
dom
estic
cos
ts a
nd fo
reig
n-ex
chan
ge e
arni
ngs f
rom
hig
h-co
st m
anuf
actu
red
expo
rtsD
irect
priv
ate
capi
tal i
nflo
ws (
FDF'
) res
trict
ed to
an
aver
age
of5.
0 an
d a
max
imum
of 5
.5 b
illio
n pe
sos i
n an
y on
e ye
ar
Initi
al e
xpor
t cos
t diff
eren
tial =
50%
Sam
e av
erag
e ra
te, b
ut—
exce
pt fo
r the
initi
al a
nd te
rmin
al y
ears
—no
lim
it on
inflo
ws (
FDP)
in a
ny o
ne y
ear
5 6
Incl
udes
con
stra
ints
on
the
supp
ly a
nd d
eman
d fo
r lab
or sk
ills
Allo
ws f
or c
apita
l-lab
or su
bstit
utio
n in
agr
icul
ture
thro
ugh
activ
ities
1(4
'
Labo
r con
stra
ints
elim
inat
edD
oubl
ed m
argi
nal p
rodu
ctiv
ity o
f lab
or in
agr
icul
ture
, the
reby
doub
ling
cost
of a
ctiv
ities
1(4
'
7 8
0 0 0 C)
CD
b-s
C.) 0
Economic and SocialL Research in Latin America
politically or technically feasible—e.g., one policymaker will say that the basiccase is altogether too pessimistic in projecting the subsidy required for promotingmanufactured exports; another will say that we have been too optimistic on thisscore. Case 5 permits us to check for the indirect implications of these alternativeviews. Similarly, through case 6, we may examine another aspect of theexchange constraints—the year-by-year limits on the inflows, of foreign privatecapital.
A priori, it might be supposed that alternatives 5 and 6 would have a significanteffect upon the GDP growth rate. Similarly, one might have anticipated sizablemacroeconomic effects from such alternatives as: (1) and (2), changing the annualgrowth target to 6 or 8 percent; or (3), changing the maximand to discountedconsumption; or (4), changing the maximand to terminal consumption.
Under each of these alternatives—when taken one at a time—it turns outthat the 1968—1980 optimal annual ODIP growth rate varies only between 6.8 and7.11 percent. The output growth rates for individual sectors also tend to be insensi-tive to the variations considered under cases 1—6. The effects are concentrated upona comparatively small number of primal variables: the amount of capital-laborsubstitution within agriculture, the inflows of foreign capital, and the marginalexport activities for trade-balance improvement.
11t is not until we turn to cases 7 and 8—those involving the labor constraints—that the alternatives become radically different. Case 7 is calculated as thoughthe marginal productivity of labor in Mexican agriculture were zero, as thoughthere were no social costs of rural-urban labor transfer, and as though the socialproduct foregone by creating human capital were also zero. With the labor-surplus hypothesis carried to this extreme, all labor constraints may be neglected.The marginal productivity of physical capital would rise to 30—33 percent per year—virtually identical to the incremental ratio of aggregate output to physicalcapital. This also means that the 1968—1980 optimal annual GIDP growth ratewould be 7.6 percent, and that each sector's output requirements would be increasedcorrespondingly.
From the viewpoint of income distribution, perhaps the most significantalternative is case 8. the labor constraints are reintroduced. ilt is supposedthat long-term credit is made available to the agricultural sector on more favorableterms than heretofore, and that this policy is pushed far enough so as to doublethe marginal rate of substitution of capital for unskilled agricultural workers.The macroeconomic and foreign-trade effects are not sizable, but the incomedistribution then shifts significantly in favor of unskilled labor.
For our quantitative comparison of alternatives, the results are summarizedin Tables 2 through 7. (Further details are to be found in the computer listings.These are available for inspection in the author's office.) Tables 2 through 7 arearranged as follows:
Table 2. Macroeconomic resultsTable 3. Resource gap—financial flowsTable 4. Foreign-exchange projections, 1980Table 5. Efficiency prices of foreign exchange and foreign aidTable 6. Gross production levels, 1980Table 7. Employment, efficiency wages, and labor income, 1980
'Per
iod
1 =
2..
2Per
iod
S=
4.
rTl 0 0 I
TAB
LE 2
MA
CR
OEC
ON
OM
ICR
ES
ULT
S
[bill
ions
of19
60pe
sos]
Ann
ual
.
CO
N'
SA
W'
INV
'G
DP'
mps
, 196
8-80
SAV
4—SA
P°—
GD
Ø—
GD
P°
GD
PG
row
thC
ase
Iden
tific
atio
n19
741
1980
219
74'
1980
219
74'
1974
'19
802
1968
—80
0B
asic
cas
e30
3.6
445.
881
.913
6.8
81.9
128.
238
5.5
582.
626
.3%
6.9%
Ig
=6%
308.
144
8.8
78.3
123.
978
.311
5.2
386.
457
2.7
23.0
6.8
2 g
=8%
298.
944
1.9
89.0
150.
289
.014
2.6
387.
859
2.1
29.6
7.1
3D
isco
unte
d co
nsum
ptio
n29
8.9
442.
690
.715
2.3
91.8
144.
738
9.7
594.
930
.07.
14
Term
inal
con
sum
ptio
n29
5.8
434.
489
.414
8.8
86.5
145.
438
5.2
583.
230
.06.
95
50%
exp
ort c
ost d
iffer
entia
l6
No
annu
al li
mits
on
FDPt
303.
130
3.7
444.
644
6.0
81.3
83.2
136.
314
3.2
81.3
87.6
127.
612
5.5
384.
438
6.8
580.
8S8
9.2
26.3
27.7
6.9
7.0
7La
borc
onst
rain
tse!
imin
ated
316.
647
8.4
89.5
146.
689
.513
7.9
406.
162
4.9
26.0
7.6
8D
oubl
ed c
apita
l-lab
or su
bstit
utio
n ra
te30
0.2
437.
386
.113
2.3
86.1
123.
638
6.3
569.
626
.06.
7
rn 0 0 0 0 p
TAB
LE 3
RES
OU
RC
E G
AP—
FIN
AN
CIA
LF
LOW
S'
[bill
ions
of 1
960
peso
s)
t'J 0\
Fore
ign
Dire
ct P
rivat
e C
apita
l Inf
low
s, FD
P''7
1'7
4'7
7'8
0'8
3'8
6C
ase
Iden
tific
atio
nI
23
45
6
Less
: Int
eres
t and
Pro
fit R
emitt
ance
s on
Dire
ct P
rivat
eC
apita
l Inf
low
s of P
rior Y
ears
'71
'74
'77
'80
'83
'86
12
34
56
0B
asic
case
5.0
5.5
5.5
5.5
3.5
5.0
Ig
=6%
5.0
5.5
5.5
5.5
3.5
5.0
2 g=
8%5.
05.
53.
55.
55.
55.
0D
isco
unte
dcon
sum
ptio
n3.
05.
55.
55.
55.
55.
04
Term
inal
cons
umpt
ion
00
02.
45.
55.
05
50%
expo
rtcos
tdiff
eren
tial
5.0
5.5
5.5
5.5
3.5
5.0
6N
oann
ualli
mits
onFD
P5.
09.
99.
30.
80
5.0
7La
borc
onst
rain
tsel
imin
ated
5.0
5.5
5.5
5.5
3.5
5.0
8D
oubl
ed c
apita
l-lab
or su
bstit
utio
n ra
te5.
05.
51.
33.
25.
55.
0
—4.9
—7.5
—10
.4—
13.2
—16
.1—
17.9
—4.
9—
7.5
—10
.4—
13.2
—16
.1—
17.9
—4.
9—
7.5
—10
.4—
12.2
—15
.1—
17.9
—4.
9—
6.5
—9.
3—
12.2
—15
.1—
17.9
—4.
9—
4.9
—4.
94.9
—6.
1—
9.0
—4.
9—
7.5
—10
.4—
13.2
—16
.1—
17.9
—4.
9—
7.5
—12
.7—
17.6
—17
.9—
17.9
—4.
9—
7.5
—10
.4—
13.2
—16
.1—
17.9
—4.
9—
7.5
—10
.4—
11.0
—12
.7—
15.6
Plus
: Con
cess
iona
l Cap
ital_
Inflo
ws L
ess I
nter
est,
FC' —
INF
C'7
1'74
'77
'80
'83
'86
Case
Iden
tific
atio
n1
23
45
6
= R
esou
rce
Gap
, RG
AF'
'71
'74
'77
'80
'83
'86
12
34
56
Bas
ic c
ase
2.4
2.0
.9—.9
—2.9
—4.
7I
g =
6%2.
42.
0.9
—.9
—2.9
—4.
7g=
8%2.
42.
0.9
—.9
—2.
9—
4.7
Dis
coun
ted
cons
umpt
ion
2.4
2.0
.9—
.9—2.9
—4.
7Te
rmin
alconsumption
2.4
2.0
.9
—.9
—2.9
—4.
750
%ex
portc
ostd
iffer
entia
l2.4
2.0
.9
—.9
—2.
9—4.7
No
annu
al li
mits
on
FDP'
2.4
2.0
.9—
.9—
2.9
—4.
77
Labo
rcon
stra
ints
elim
inat
ed2.
42.
0.9
—.9
—2.
9—
4.7
Dou
bled
cap
izal
-labo
rsub
stitu
tion
rate
2.4
2.0
.9—
.9—
2.9
—4.
7
2.5
0—
4.0
—8.
6—
15.5
—17
.62.
50
—4.
0—
8.6
—15
.5—
17.6
2.5
0—
6.0
—7.
6—
12.5
—17
.60.
51.
0—
2.9
—7.
6—
12.5
—17
.6—
2.5
—2.
9—
4.0
—3.
4—
3.5
—8.
72.
50
—4.0
—8.6
—15.5
—17.6
2.5
4.4
—2.
5—
17.7
—20
.8—
17.6
2.5
0—
4.0
—8.
6—
15.5
—17
.62.
50
—8.
2—
8.7
—10
.1—
15.3
0 2 3 4 5 6 8
'By year and
perio
d.Periodt=1:1971.
t=3:
1977
.1=
5:19
83.
t=2:1974.
1=4:1980.
t=6:1986.
CD
CD p I
TAB
LE 4
FOR
EIG
NE
XC
HA
NG
E P
RO
JEC
TIO
NS
. 198
0[b
illio
ns o
f 196
0 pe
sosj
Impo
rts
0B
asic
case
41.1
3I
g=6%
38.9
52
g=8%
43.2
73
Dis
coun
ted
cons
umpt
ion
43.8
74
Term
inal
con
sum
ptio
n43
.11
550
% e
xpor
t cos
t diff
eren
tial
40.3
76
Noa
nnua
llim
itsor
iFD
P'41
.15
7La
bor c
onst
rain
ts e
limin
ated
43.2
58
Dou
bled
cap
ital-l
abor
subs
titut
ion
rate
39.7
1
Exog
enou
sly
spec
ified
upp
er b
ound
. Z?(
l -1-
e,)'
2Ex
ogen
ousl
y sp
ecifi
ed lo
wer
bou
nd,
+
Cas
e Id
entif
icat
ion
0 0 0 C)
Mer
chan
dise
Exp
orts
, at P
rodu
cers
' Pric
es.
i'=I
23
45
67
89
10
item
.
4.90
3.78
.54
5.80
.98
.62
2.07
.29
.72
2.18
4.90
3.78
.54
5.80
.98
.62
2.07
.29
.72
2.18
4.90
3.78
.54
5.80
.98
.62
2.07
.29
.72
1.80
4.90
3.78
.54
5.80
.98
.62
2.07
.29
.72
2.18
4.90
3.78
.54
5.80
.98
.62
2.07
.29
.72
2.18
7.77
3.78
.54
5.80
.98
.62
2.07
.29
.72
2.18
4.90
3.78
.54
5.80
.98
.62
2.07
.29
.72
2.18
7.77
3.78
.54
5.80
.98
.62
2.07
.29
.23
2.18
4.90
3.78
.54
5.80
.98
.62
2.07
.29
.72
2.18
7.77
3.78
.54
5.80
.98
.62
2.07
.29
.72
2.18
4.90
2.36
.45
2.88
.58
.31
0.66
.12
.23
0.56
00
TA
BLE
4 (c
oncl
uded
)
C) 0 0 S C) 0 C)
CD
CD 5.
2 3 4 5 6 7 8
Expo
rts o
f Hig
h-C
ost
Man
ufac
ture
s, at
Prod
ucer
s' Pr
ices
,
Cas
e Id
entif
icat
ion
Fore
ign
Exch
ange
Ear
ning
s. at
Mar
ket P
rices
Res
ourc
e G
ap.
RG
AP4
=Im
port
s—
ZA
4—
Zr4
Mer
chan
dise
Exp
orts
,2To
uris
m.
ZA'
Zr4
Bas
icca
se12
.01
g =
6%10
.13
g =
8%13
.34
Dis
coun
ted
cons
umpt
ion
13.4
8Te
rmin
al c
onsu
mpt
ion
9.21
50%
exp
ort c
ost d
iffer
entia
l8.
26N
o an
nual
lim
its o
n FD
P'19
.84
Labo
r con
stra
ints
elim
inat
ed4.
84D
oubl
ed c
apita
l-lab
or su
bstit
utio
n ra
te10
.84
39.1
410
.63
36.9
610
.63
40.2
310
.63
40.8
310
.63
35.9
010
.63
38.1
210
.89
48.1
910
.63
33.6
018
,29
37.7
810
.63
—8.
64—
8.64
—7.
59—
7.59
—3.
42—
8.64
—17
.67
—8.
64—
8.70
Exog
enou
sly
spec
ified
upp
er b
ound
. Z?(
l + c
)'2—
Exo
geno
usly
spec
ified
low
er b
ound
. Z?(
l + e
,)12
0—
-•18
.29
—10
.63
2 Z
A4
= m
erch
andi
seex
ports
, at m
arke
t pric
es.
.865
8ZA
4 =
mer
chan
dise
expo
rts, a
t pro
duce
rs' p
rices
=+
ZM
'..1
125
Z44
= c
omm
erce
mar
gins
on
mer
chan
dise
exp
orts
..0
217
ZA' =
serv
ice
mar
gins
on
mer
chan
dise
exp
orts
.
rn C) 0 0 C) -t CD
TAB
LE S
EFFI
cIEN
cYP
RIC
ES
OF
FO
RE
IGN
EX
CH
AN
GE
AN
DF
OR
EIG
N A
ID
1960
Pes
os W
orth
of M
axim
andt
per
Tho
usan
d 19
60 P
esos
Wor
thof
Item
Fore
ign
Exch
ange
(row
s [F'
})Fo
reig
n A
id (r
ows [
FGA
P'}
)
Cas
e Id
entif
icat
ion
1974
219742
Effic
ienc
y Pr
ice
of F
orei
gnEx
chan
ge R
elat
ive
to P
rice
ofTr
adab
le M
anuf
actu
res
"Ow
n" R
ate
ofIn
tere
st o
nFo
reig
n
1974
—80
.Annual
Rat
e19
742
0B
asic
cas
e72
.40
24.7
272
.40
24.7
2I
g6%
77.0
726
.17
77.0
726
.17
2g
=8%
74.4
120
.55
61.8
525
.98
3Discounted consumption
547.92
238.20
759.36
333.96
4Terminalconsumption
373.61
290.17
1,284.79
892.21
550%exportcostclifferential
82.08
29.92
78.49
29.9
26
No annual limits on FDP'
71.3
127.22
62.97
27.22
7Laborconstraintseliminated
130.91
.24.61
130.91
24.61
8Doubledcapital-Iaborsubstitutionrate
53.59
22.96
53.59
22.96
1.15
1.15
1.15
1.15
1.16
1.13
1.13
1.13
1.02
1.20
1.28
1.35
1.05
1.21
1.17
1.17
1.13
1.17
20%
20 24 IS 4
18 17 32
IS
Exce
pt fo
r cas
es 3
and
4. i
he m
axim
and
is C
ON
', an
d th
e un
its o
f mea
sure
men
t of e
ffic
ienc
y pr
ices
are
ther
efor
e id
entic
al. I
n ca
se 3
. the
unit
of m
easu
rem
ent
refe
rs to
dis
coun
ted
cons
umpt
ion;
in c
ase
4. to
term
inal
con
sum
ptio
n.2
Perio
d I
= 2
.Pe
riod
1=
4.
I
TAB
LE 6
GR
OSS
PR
oDuc
rioN
LE
VE
LS. 1
980.
BY
SE
CT
OR
[bill
ions
of l
960
peso
s]
Cas
etde
ntifi
catio
nA
gric
ultu
re(1
)M
inin
gP
(2)
etro
leum
Foo
d(3
)(4
)T
extil
es(5
)W
ood
(6)
Che
mic
als
(7)
Non
met
allic
(8)
0B
asic
cas
e74
.68.
639
.011
2.8
54.8
22.6
54.8
14.9
Ig
= 6
%73
.38.
438
.3l1
2.0
54.6
2l.9
53.7
14.0
'2
g =
8%
76.3
8.8
39.5
112.
954
.723
.155
.616
.0*
3D
isco
unte
dcon
sum
ptio
n76
.28.
939
.711
3.3
54.8
23.3
55.9
16.1
*
4T
erm
inal
con
sum
ptio
n74
.48.
838
.910
8.5
53.4
22.6
53.6
15.9
'5
50%
exp
ort c
ost d
iffer
entia
l77
.48.
538
.811
0.9
54.4
22.3
53.8
14.9
6N
oann
ualli
mits
onF
DP
76.5
8.7
39.5
118.
155
.823
.357
.415
.27
Labo
rcon
stra
ints
elim
inat
ed80
.0*
8.8
41.3
*11
5.7
58.7
'23
.656
.115
.9*
8D
oubl
edca
pita
l-lab
orsu
bstit
utio
nrat
e74
.08.
438
.011
0.1
53.6
21.9
53.2
14.4
Bas
ic M
etal
sM
achi
nery
Con
stru
ctio
nE
lect
ricity
Com
mer
ceT
rans
port
atio
nS
ervi
ces
Cas
e Id
entif
icat
ion
(9)
(10)
(II)
(12)
(13)
(14)
(15)
0B
asic
case
37.0
100.
159
.!19
.819
6.7
24.2
161.
0I
g =
6%
34.9
95.2
53.9
'19
.419
3.9
24.1
159.
1
2g
= 8
%39
.l'10
4.4
64.9
'20
.219
9.1
24.2
162.
1
3D
isco
unte
d co
nsum
ptio
n39
.6'
106.
l'65
.4'
20.3
200.
224
.316
3.5
4T
erm
inal
cons
umpt
ion
38.6
103.
865
.9'
19.9
196.
323
.816
1.3
550
%ex
port
cost
diffe
rent
ial
36.3
98.4
59.9
19.7
195.
524
.116
0.4
6N
o an
nual
lim
its o
n F
DP
38.1
101.
858
.520
.019
8.2
24.3
161.
57
Labo
rcon
stra
ints
elim
inat
ed37
.910
5.5'
65.4
'21
.l'21
0.2'
26.2
'17
0.3'
8D
oubl
ed c
apita
l-lab
or s
ubst
itutio
n ra
te35
.796
.756
.819
.3l9
2.0
23.6
156.
6
NoT
E: *
Diff
ers
by m
ore
than
5%
from
Icve
l sho
wn
for
basi
cca
se.
tTJ 0 0 0 ci.
C,, 0 0 C
D
CD pa CD I-t 0 p
0 0 ()TA
BLE
7EM
PLO
YM
ENT,
WA
GE
S,
AN
D L
AB
OR
INC
OM
E,19
80,
BY
LA
BO
R S
KIL
L C
ATE
GO
RY
'
Empl
oym
ent.
(mill
ions
of p
erso
ns)
-S
kill
Cat
egor
ySu
btot
al.
Tota
l.C
ase
Iden
tific
atio
n1
23
45
1—4
1—5
Effi
cien
cy W
ages
. Ann
ual A
vera
ge C
ente
red
on19
802
(thou
sand
s of 1
960
peso
s per
man
-yea
r of s
kill
s)
Skill
Cat
egor
y
I2
34
5
0B
asic
case
.183
1.31
02.
673
9.52
16.
765
13.6
8720
.452
Sg
=6%
.176
1.28
62.
619
9.26
57.
350
13.3
4620
.696
2 g
=8%
.189
1.32
82.
720
9.75
96.
359
13.9
9620
.355
3D
isco
unte
d co
nsum
ptio
n.1
911.
339
2.74
19.
836
5.99
614
.107
20.1
034
Term
inal
cons
umpt
ion
.187
1.31
92.
693
9.66
75.
734
13.8
8619
.600
550
%ex
portc
ostd
iffer
entia
l.1
821.
305
2.66
19.
467
6.88
413
.615
20.4
996
No
annu
al li
mits
on
.186
1.31
82.
698
9.62
36.
901
13.8
2520
.726
7La
borc
onst
rain
tsel
imjn
ated
.193
1.38
72.
839
10.1
349.
895
14.5
5324
.448
8D
oubl
edca
pita
l-Iab
orsu
bstit
utio
nrat
e.1
781.
274
2.60
29.
254
7.23
913
.308
20.5
47
112.
153
.423
.07.
53.
910
6.8
53.7
23.4
7.5
3.9
129.
459
.924
.17.
94.
275
.941
.817
.78.
24.
841
.821
.112
.78.
75.
511
1.6
52.0
22.9
7.6
4.0
117.
953
.824
.27.
94.
2—
——
——
90.2
52.8
24.2
-11
.37.
9
'Ski
ll ca
tego
ry s
=I
:en
gine
ers
and
scie
ntis
ts.
s =
2:
s =3:
$ =
4:s =
5:2A
vera
ge o
f eff
icie
ncy
othe
r pro
fess
iona
l and
tech
nica
l wor
kers
.ad
min
istra
tive
and
cler
ical
wor
kers
.m
anua
l and
sale
s wor
kers
out
side
agr
icul
ture
.un
skill
ed a
gric
ultu
ral w
orke
rs.
wag
es fo
r 197
7, 1
980,
and
198
3, n
orm
aliz
ed b
y du
alva
riabl
efo
r con
sum
ptio
n go
ods.
TAB
LE 7
(con
dude
d)
Labo
r Inc
ome
=E
mpl
oym
ent
x Ef
ficie
ncy
Wag
es(b
illio
ns o
f 196
0 pe
sos)
Skill
Cat
egor
yTo
tal.
Cas
e Id
entif
icat
ion
I2
34
SI—
S
0B
asic
cas
e20
.570
.061
.571
.426
.424
9.8
Ig
=6%
18.8
69.1
61.3
69.5
28.7
247.
42
g =
8%24
.579
.565
.677
.126
.727
3.4
3D
isco
unte
d co
nsum
ptio
n14
.556
.048
.580
.728
.822
8.5
4 Te
rmin
alco
nsum
ptio
n7.
827
.834
.284
.131
.518
5.4
550
%ex
port
cost
diffe
rent
ial
20.3
67.0
60.9
71.9
27.5
248.
56
Noa
nnua
llim
itson
FDP
21.9
70.9
65.3
76.0
29.0
263.
17
Labo
r con
stra
ints
elim
inat
ed—
——
——
—
8D
oubl
edca
pita
l-Iab
orsu
bstit
utió
nrat
e16
.167
.363
.010
4.6
57.2
308.
2
445.
856
%44
8.8
5544
1.9
6244
2.6
5243
4.4
4344
4.6
5644
6.0
5947
8.4
—
437.
370
rTl
C,
Agg
rega
teC
onsu
mpt
ion
(bill
ions
of 1
960
peso
s)La
bor I
ncom
e as
Fra
ctio
n0
CO
N4
ofA
ggre
gate
Con
sum
ptio
nC
)pa C
I) 0 C)
CD
CD pa pa I
Economic Alternatives for Mexico 223
3. ALTERNATIVE OBJECTIVE FUNCTIONS
Cases I through 4 all deal with the problem of welfare distribution betweensuccessive generations—near future versus distant future increases in aggregateconsumption. Somewhat surprisingly, these different objective functions do notlead to great differences in the absolute levels of consumption during the earlytime periods. (See Figure 1 and Table 8.)
Aqgregote consumption, CONt(billion 1960 pesos)900
Ratio 5c01e
800—
700 —
600— / -
500— —
400 - -Bostc case;
onnuol growth target g=7% ,
/4erminal Consumplion moximand;/ 30% limit on marginal propensity
300 —to save —
200 I
Year 1968 '71 '74 '77 '80 '83 '86 '89Period! 0 1 2 3 4 5 6 7
FIGURE I ALTERNATIVE TIME PATHS OF AGGREGATE CONSUMPTION
TABLE 8AGGREGATE CONSUMPTION, CONt: Emci OF ALTERNATIVE OBJEcrIvEs1
[billions of 1960 pesosj
Case Identification1968
01971
1
19742
19773
19804
19835
19866
19897
0
234
Basic caseg = 6%g = 8%Discounted consumptionTerminalconsumption
208.8208.8208.8208.8208.8
251.4254.1248.7245.3245.9
303.6308.1298.9298.9295.8
367.5372.3362.2363.5359.5
445.8448.8441.9442.6434.4
541.8540.0542.2542.5535.9
659.3648.6668.7692.7700.1
803.3777.9828.0848.6857.6
1.By year and period.Period 1=0:1968. t=3:1977. t=6:l986.
1=1:1971. (=4:1980. t7:1989.t=2:1974. 1=5:1983.
224 Economic and Sociall Research in Latin America
For cases I and 2, the tradeoffs are examined through variations in theasymptotic target growth rate—still retaining the restriction that the time pathbe of the gradualist form with increments of consumption growing geometricallyat the annual rate of g.4 lln case II, for example, the target rate is reduced from 7to 6 percent. This policy would make it possible to increase CON' from 251.4 to254.1 billion pesos. Higher consumption levels would also be achieved duringperiods t = 2, 3, and 4, but lower levels during all subsequent time periods.Symmetrical effects are obtained when the target growth rate is raised to 8 percent.That is, case 2 provides lower consumption levels than the basic case duringperiods 11 through 4, but higher levels thereafter.
Although these alternatives differ from each other by less than 1 percent inthe absolute levels of near-term consumption, they do have different implicationsfor near-term fiscal policy. By lowering the annual growth target from 7 to 6cent, the 1968—1980 marginal savings propensity is reduced from 26.3 to 23.0percent. (Table 2.) This means lower taxes and lower prices on the products ofpublic-sector enterprises. Over the long term, of course, this "soft" option leadsto a lower aggregate growth rate and lower employment opportunities. Conversely,if the growth target is raised to 8 percent, the required 1968—1980 marginal savingspropensity is 29.6 percent. The fiscal policy tasks would become correspondinglymore onerous, but the 1980 employment opportunities would then increase from113.7 to 114.0 million jobs within the four highest-paying skill categories. (SeeTable 7.)
lln all cases, we have supposed that there are political constraints uponMexico's fiscal policies, and that it would be infeasible to push the marginalsavings propensity above 30 percent.5 With the gradualist objective function(cases 0, 11, and 2), typically it turns out that the 30 percent savings limit is aredundant constraint.6 When the gradhalist objective function is replaced by thatof maximizing discounted consumption (case 3) or terminal consumption (case 4),the savings constraints become critically important. Without them, the time pathof aggregate consumption is exceedingly erratic from one period to the next. Witha 30 percent upper bound upon the marginal propensity to save, then—even undera growth-maximizing criterion such as terminal consumption—there are no sharpdiscontinuities in the optimal consumption level between two successive periods.When the savings constraint is operative, the specific form of the objective functionhas little effect upon the optimal values of the primal variables.7
For numerical evaluation in case 3, we have taken 15 percent as the annualldiscount rate on future consumption,° and have retained 7 percent as the annuall
For the properties of the gradualist consumption maximand—and its relation to postterminalgrowth—see Hopkins (1969) and Manne (1970).
For the statistical base period 1960—1968, the marginal savings propensity was 23.2 percent.6 In case 2. the 30 percent limit on the domestic savings propensity is binding only during periods
2,3, and 4. Even during these periods, the shadow prices are quite low for the savings constraints, rows(GSAV').
The optimal dual solution is sensitive to differences in the objective functions. For example, itturns out that the 1974—1980 "own" rate of interest on foreign exchange is 15 percent per year in case 3.and 4 percent in case 4. (See Table 5.)
In principle, the consumption discount rate is a subjectively determined parameter—one thatexpresses the social rate of time preference. It is sheer coincidence that the same numerical value hasbeen employed as for the cost of foreign private capital.
Economic Alternatives for Mexico 225
target for postterminal growth. Case 4 has the identical constraints as case 3, buta different objective function. Like a "turnpike" growth model, the maximand isCON7, the level of aggregate consumption reached during the terminal year.Despite the difference in appearance, it can be shown that there is a close relation-ship between the objective functions considered in cases 3 and 4. The latter maybe viewed as a special case of the former—a limiting case obtained by loweringthe discount rate sufficiently. With this interpretation, the objective functions ofcases 3 and 4 have the identical structure—that of discounted consumption. Theonly difference is that the consumption discount rate is 15 percent in case 3, andthat it approaches 7 percent in case 4. The numerical results (Table 8) imply thatthe optimal solution is insensitive to variations in the discount rate within a ratherwide range—7 to 15 percent per year. With this form of model, apparently theupper bound on the marginal savings rate has far more influence upon the optimalconsumption path than does the subjectively determined time discount rate.
4. EXCHANGE
un the basic case, it turns out that the efficiency price of foreign exchange isII. 15—taking the domestic producers' price of tradable manufactures as thenuméraire. 8efore advocating that 15 percent is therefore an appropriate tariffrate or export subsidy or adjustment in foreign-exchange parity, it is essential tocheck through the reasoning that led to this numerical result. un DINAMICO,one of the least reliable econometric components is the marginal cost of exportearnings. The basic case was set up as though there were no limits upon the exportof high-cost manufactures. It is supposed that the domestic cost of theseproducts will exceed the foreign exchange earnings by 30 percent at the time thatthe manufactured item is first exported, and that this cost differential will drop tozero through the experience acquired during an 18-year period.9 As an alternative,case S is calculated as though these marginal export items will have an initial costdisadvantage of 50 percent rather than 30 percent. We continue to suppose thatwill require 18 years of experience before the cost differential can be eliminated.
Case 5 makes little or no difference in the macroeconomic results (Table 2),the pattern of foreign capital inflows (Table 3), the aggregate requirements forimports (Table 4), the gross production levels (Table 6), or the employment levels(Table 7). From the viewpoint of overall reliability of IDINAMICO, it is fortunatethat the principal effects of this parameter change are concentrated upon a smallnumber of variables: the commodity composition of exports (Table 4) and theshadow price of foreign exchange (Table 5). lIt becomes optimal to reduce the 1980exports of high-cost manufactures from 12.01 to 8.26 billion pesos, and to offsetthis Loss of foreign exchange through an increase in agricultural exports. Associatedwith this shift in the direction of comparative advantage, there is a shift in the
e Presumably, in order to provide a financial inceVtive [or the export of high-cost manufacturesfrom individual enterprises, it would be necessary to provide export subsidies. Such subsidies might takea number of forms—either outright or in the form of permission to import raw materials and equipmentduty free. (Mexico is not a member country of GATT [General Agreement on Tariffs and Trade].)With any of these measures, there are administrative difficulties—but no more so than in the case ofexisting import restrictions. The principal difference between an import tariff and an export subsidyis that one provides an inflow of pesos to the treasury, and the other generates an outflow.
226 Economic and Research in Latin Aitnerica
shadow price of foreign exchange. It no longer remains 11.115, but increases to 11.35in 11980—still taking the price of tradable manufactures as the numéraire. Case 5also means that foreign aid would have a higher marginal productivity than in thebasic case.1°
Case 6 provides an instance of interdependence between the optimall timepath of capital inflows, the shadow price of foreign exchange, and the compositionof exports. IHiere we have eliminated the limits on the private capitãilinflows (t = 2. . . 5), but have retained the overall constraint that theaverage inflows are not to increase above the 19711 rate. Case 6 has little effect uponthe macroeconomic iresults or the gross production or the employment levels. Theeffects are concentrated upon the sources and uses of foreign exchange. litbecomes optimal to allow exports to lag and to incur a sizable foreign-exchangedeficit during periods 2 and 3, then to push up the level of manufacturedexports to 19.84 billion pesos in period 4 (1980), and to irun a sizableexchange surplus (17.67 billion pesos) in that year. Associated with this shift inthe time pattern of exports, the shadow price of foreign exchange would irise overtime—from 1.05 to 11.21 between periods 2 and 4 (11974 and 19809 respectively).This also means that the marginal productivity of foreign aid would be lowerduring period 2 and higher during period 4 than in basic case. (See Table 5.)
From case 6, we cannot conclude that year-by-year constraints on foreigncapital inflows are essential in a model of this type. Other devices may also beemployed to avoid sharp discontinuities in the time pattern of exports. Amongsuch devices are: an upward-sloping supply curve of foreign capital, asloping demand curve for exports, or a recursive programming constraint on therate of growth of exports.1 Among these alternative devices, the simplest toestimate is the limit on private capital inflows. For the basic case,this has the disadvantage that the upper bound is an effective constraint duringthe early time periods, and that the unknowns appear to be predetermined.It does not happen, however, that the upper bound is always an effective constraintduring the initial time periods. For examples, see cases 2,4, and 8 in Table 3.
5. EMPLOYMENT AND IINCOME
For case 7. the labor constraints have been eliminated. This makes it possibleto achieve a significant'y higher aggregate growth rate, higher output rates in
10lExcept for cases 3 and 4, the 30 percent limit upon the marginal savings propensity is notcritical constraint. Except for these cases, therefore, it makes little or no difference whether the impactof additional foreign aid is measured through the shadow price of the foreign exchange rows (F') orthrough the resource gap rows Cases 3 and 4 are typical of the "two-gap" phenomenon.There, foreign aid has considerably more leverage than foreign-exchange earnings alone. For furtherdiscussion of the two-gap model, see the interchange between Chenery and Strout (1968) and Fei andRanis (1968).
The shadow price of the domestic savings rows (GSA V') may be read off from Table 5 here. In allcases, this equals the difference between the shadow price of rows (F') and (FGAIP'). To prove thisproposition, refer back to the coefficients of the resource gap activity RGAP and of the concessionallforeign aid activity in Table 4, memorandum 71—13.
Let denote the exports of item I during period t, and let denote the maximum feasible rateof growth of this item between periods: — I and t. Then a recursive programming constraint on exportswould be written: � (I + - See Day (1963); and Dougherty, and Fraenkel (1970).
Note that a recursive programming constraint refers to pairs of unknowns, but that the exportbounds in DINAMICO are imposed upon individual unknowns.
Economic Alternatives for Mexico 227
individual sectors, and greater employment opportunities outside traditionalagriculture. (See Tables 2, 6, and 7.) 11t has already been noted that a zero shadowprice for labor implies a high shadow price for capital. Almost equally dramaticare the implications for comparative advantage in foreign trade. With a zeroshadow price for labor, it becomes optimal to set agricultural exports and tourismearnings at their upper rather than their lower bounds, and to place less dependenceupon high-cost manufactured exports. (See Table 5•)12 Case 7 is included herefor the sake of completeness—not because we believe that it provides a irealisticbasis for projections.
lincreasingly, Mexican policymakers are directing research toward the sourcesinequities in income distribution. There is strong evidence that, along with the
increase in average incomes, the extent of inequality worsened between 1950 and11963. (During those years, the Gini coefficient of inequality increased from 0.50to 0.55.) ltn part, this problem arises from interregionall differences in develop-ment—and from imperfect labor mobility between regions. un 11965, for example,the 8 highest-income regions accounted for 30.3 percent of the total population,and produced 11,075 pesos per capita. The 17 lowest-income regions, with 43.8percent of the population, produced only 2,417 pesos per capita. (Source : Navarrete[11970, p. 411 and Table 8].)
Not only regional but also occupational differences are a significant sourceof inequalities. During each of the years between 1950 and 1967, in industry andservices the per capita output exceeded that in agriculture by a factor of approxi-mately 5:1. Even within agriculture—despite the post-Revolution policies ofagrarian reform and land redistribution—the pattern of development has been
un the irrigated districts of the north, farming is commercialized andproduces high economic returns. Elsewhere—in the central plateau and in thesouth—subsistence agriculture is hard pressed to keep up with the demographicpressures. (See Solis [1970, pp. 148 and 291].)
DINAMXCO is too highly aggregated to be helpful in analyzing the detailedregional aspects of income inequalities and labor mobility. Case 8 is focused ononly one dimension of the tradeoff between aggregate growth versus equity inincome distribution. As proxy measures for these concepts, we have taken aggregateconsumption and labor's income share, both as of 11980 (t = 4). 13y measuringlabor's income at efficiency wages, we have, in effect, adopted the view that moneytransfer payments could not be large enough to achieve a significant redistributionof real income between social classes. (if money transfers could be made sufficientlylarge—and transfers entailed no loss in productive efficiency—the question of agrowth-equity tradeoff would not arise.)
Case 8 has been constructed as follows: Suppose that public infrastructureinvestments are made available for the benefit of the smaliholders in the denselypopulated central and southern centers of Mexican agriculture. These investmentsmight take the form of extension services, roads, tractors, irrigation, and land
12 Rather than suppose that the marginal productivity of labor is zero in all skill classes (case 7),it might have been worth exploring a less extreme version of the labor-surplus hypothesis. Suppose, forexample, that the only change in the basic case had been to assume that unskilled agricultural labor hasa zero or a low marginal product. It would then be optimal to shift the composition of foreign-exchangeearnings—setting agricultural exports at their, upper bounds and reducing the exports of high-costmanufactures.
228 Economic and Social Research in Latin America
leveling. To the extent that the aggregate demand for agricultural products isinelastic, these investments might have to be made at the expense of northernagriculture. It is possible that this would lead to a loss in economic efficiency, butthat nonetheless—from the viewpoint of income distribution—this would proveto be a desirable shift in investment policy.
For case 8, it is supposed that this reorientation of agricultural policy ispushed far enough so as to double the marginal rate of substitution of capital forunskilled agricultural workers.'3 It turns out that this would have a significant
Labor income (billion 1960 pesos)325
.8300 —
275— .2Key:1 g6%2 g8% .63 Discounted consumption4 Terminal consumption5 50% export cost differential
250 — 6 No annual limits on FDPI •osic case_8 Doubled capital-labor
substitution rate
225 —
200 —
175 —
425 430 435 440 445 450Aggregate consumption, CON4 (bLilion 1960 pesos)
FIGURE 2 INcOME DISTRIBUTION VERSUS AGGREGATE GROwTH, 1980
13 For the basic case, at a point of time y years after 1960. it is assumed that it would require15(1 thousand pesos of capital to replace one unskilled agricultural worker. For case 8. thismarginal rate of substitution is instead taken to be 30(1.02)'.
Economic Alternatives for Mexico 229
effect not only upon the efficiency wage of agricultural workers, but also uponthat of unskilled urban workers. (See Table 7 and Figure 2.) Other policy measureswould also affect labor's share, e.g., case 2, where the target annual growth rateis raised to 8 percent. None of the alternatives, however, would exert as pronouncedan effect upon income distribution as would this shift in the direction of agricul-tural investments. In 1980, this would make it possible to double the income ofunskilled agricultural workers—from 26.4 to 57.2 billion pesos at 1960 prices.Total labor income would increase from 249.8 to 308.2 billion. The loss in aggregateconsumption (from 445.8 to 437.3 billion pesos) might well be worth the gain inequity from such a policy.
REFERENCES
A.delman, I., "A Linear Programming Model of Educational Planning: A Case Study of Argentina,"ch. 14 in I. Adelman and E. Thorbecke, eds., The Theory and Design of Economic Development(Baltimore: Johns Hopkins Press, 1966).
Bowles, S., "Efficient Allocation of Resources in Education," Quarterly Journal of Economics, May1967.
Bruno, M., "A Programming Model for Israel," ch. 12 in Adelman and Thorbecke, eds., op. cit.— . "Optimal Patterns of Trade and Development," Reviewof Economics and Statistics, November
1967.Bruno. M., Dougherty. C.. and Fraenkel, M.. "Dynamic Input-Output, Trade and Development,"
ch. 3 in A. Carter and A. Brody, eds., Applications of Inpza-Ouzpu: Analysis (Amsterdam: North-Holland. 1970).
Bueno, G., "The Structure of Protection in Mexico" (Colegio de Mexico, 1970).Chenery. H. B. and MacEwan, A., "Optimal Patterns of Growth and Aid: The Case of Pakistan." ch. 6
in Adelman and Thorbecke, eds., op. cit.Chenery. H. B. and Strout. A. M., "'Reply' to a Comment by Fei and Ranis" American Economic
Review. September 1968.Correa, H. and Tinbergen, "Quantitative Adaptation of Education to Accelerated Growth." Kykios,
fasc. 4, 1962.Day, R.. Recursive Programming and Production Response (Amsterdam: North-Holland, 1963).Eckaus, R. S. and Parikh, K. S., Planning for Growth (Cambridge, Mass.: M.I.T. Press, 1968).Fei, J. C. H. and Ranis, G., "Foreign Assistance and Economic Development: Comment," American
Economic Review, September 1968.Harberger. A., "Techniques of Project Appraisal." in M. Millikan, ed., National Economic Planning
(New York: National Bureau of Economic Research, 1967).Hopkins, D. S. P., "Sufficient Conditions for Optimality in InfInite Horizon Linear Economic Models,"
Technical Report No. 69-3 (Operations Research House, Stanford University, March 1969).Manne, A. S., "Sufficient Conditions for Optimality in an Infinite Horizon Development Plan,"
Econonietrica, January 1970.Murakami, Y.. Tokoyama, K., and Tsukui, .1, "Efficient Paths of Accumulation and the Turnpike of
the Japanese Economy," ch. 2 in Carter and Brédy, eds., op. cit.Navarrete, I. M. de, "La Distribución del Ingreso en Mexico: Tendencias y Perspectivas," El Perfil de
Mexico en 1980 (Mexico: Siglo XXI Editores, 1970).SoIls, L., La Realidad Económica Mexicana: RezrovisiOn y Perspeclivas, (Mexico: Siglo XXI Editores,
1970).R. G.. Baum, E. L., and Talbert, L. E., Application of Activity Analysis to Regional
Development Planning, Technical Bulletin 1339 (Washington. D.C.: U.S. Department of Agricul-ture, March 1965).