The Profitability of Integrated Pest Management: Case Studies for

The Profitability of Integrated Pest Management:
Case Studies for Cotton and Citrus in the
San Joaquin Valley
1
By
C.
DARWIN
HALL
of reducing both pesticide use and food production may
be a better alternative than either cancellation or continuing pes.ticide use at the present greater than the social
optimal quantity. Even more attractive would be the
alternative of reducing pesticide use while maintaining
current levels of production to profit by growers. Fisher
and Peterson (1976) refer to policies which accomplish
such changes as dominant policies. This latter alternative
obviously requires a shift to a new technology.
Synthetic chemicals now occupy a crucial role in the
production of food and fiber. Headley (1968) estimated
that in 1963 the ratio of marginal product to marginal
cost of pesticides was 4 to 1, predicting the subsequent
increase in pesticide use. Pesticide production for the
United States increased from 877,197,000 pounds to
1,417,158,000 pounds during the decade 1965 to 1974-a
62% increase. Given projected increases in world population and income, demands upon the agricultural sector will
likely lead to a continuation of past trends toward increased use of pesticides.
Integrated Pest Management (IPM)
is a new technology which substitutes knowledge and information
(labor) for pesticides (materials and capital). This substitution occurs by optimally choosing from a wider set
of available actions; considering interactions between
pests, natural enemies, weather patterns, crop growth, and
utilizing more accurate knowledge of such interactions;
monitoring insect and mite populations in a propitious and
precise fashion; and utilizing more accurate monitoring
methods and devices. Originally termed integrated control, the concept of integrated pest management was formulated by a small group of entomologists in the 1950's."
Pesticides have negative external effects to human
health, wildlife, and industry. The persistence of chlorinated hydrocarbons has resulted in residues throughout
the biosphere, posing the risk of carcinogenesis, mutagenesis and teratogenesis. Acute toxic effects of carbamates and organophosphorus compounds are manifest in
cholinesterase inhibition in farm workers. The existence
of certain wildlife species has been threatened. Unestimated losses to the fishing industry have occurred. Agriculture suffers in 2 ways: pesticide drift disrupts biological controls in neighboring fields and orchards as well as
killing bees and reducing pollination and honey production; amI, increasing resistance to pesticides is a common
pool resource problem described by Rueth and Regev
(1974) and Taylor and Headley (1975).
The 1st serious attempt at measuring the externalities
in pesticide use was presented by Langham and Edwards
(1969). For technical reasons, some noted by these authors, an optimal tax or optimal standards cannot be derived. No relationship has been established between pesticide use and residue. Even more difficult would be to
establish an agreed-upon relationship between exposure
and residue level with acute and chronic health effects for
humans. The number of compounds and synergistic effects possible make such a task prohibitively expensive.
Finally, an estimate of the value of life would be controversial at best. Current legislation and regulation have
centered upon (1) reducing accidents and (2) cancelling
the use of the most dangerous pesticides.
Five major results are presented below. First, IPM
reduces pesticide use by between % and %. Second, the
difference in profit between IPM and conventional pest
control is negligible. Third, IPM maintains yield at
present levels. Fourth, IPM reduces total pest management expenditures, including a fee per acre charged by
the commercial entomologist practicing IPM. Fifth, IPM
reduces risk by substituting knowledge and information.
Hence, policies which require IPM, if no more costly to
implement than present policies, are dominant.
Conflict of Interest
Current state regulations in California (§ 11702) provide for licensing of all advisors who make recommendations regarding pest management. Most licensed advisors
are employed by chemical companies with a direct interest in the promotion and sales of pesticides. These advisors practice "conventional control (CC)." A small but
growing cadre of independent (from chemical companies)
consultants are also licensed advisors. These advisors
practice "supervised control (SC) ,"
The policy option of pesticide cancellation only permits
the 2 corner solutions in the trade-off between food and
fiber with external effects of pesticides. The compromise
1 Presented at
tbe Joint Meetings of tbe American and
Western Agricultural Economics Associations, Symposium on
Economics of Integrated Pest Management, San Diego, Calif.,
Aug. I, 1977 (Giannini Foundation Paper no. 481). Tbis research
was supported by the EPA (WA 74·R345, NSF GB·34718), the
National Science .Foundation (GB-34718), the Ford Foundation
(720·0325, 739-0003),and the USDA (W-120) through the Univ.
of Calif, The findings, opinions, and recommendationsexpressed
herein are those of the author and not necessarily those of the
above organizations, Received for Iublication Oct. 7, 1977.
"l;ormerly Asst. Specialist an Projector Director, Dept. of
Agricultural and Resource Economics and the Giannini Foundation of Agricultural Economics, Vniv. of Calif., Berkeley. The
author thanks the growers who contributed to tbis Iroject with
lengthy personal interviews. The advice, insight, an SUllportof
Dr. R. B. Norgard was of inestimable value. Carl Friberg led
the team of interviewers, and R. L. }'arnsworth led the team of
data recorders. Max Leavitt undertook the majority of the programming burden. Dr. W. R. Z. Willey provided initial direction.
The results presented below indicate that the unit isoquants of the 2 technologies, SC and CC, cross as shown
in Diagram A. The option exists for growers to switch
to a supervised control program but no significant incentive exists if the growers maximize profit.
The major component of SC is the quantity and quality
of monitoring activity. Fields and groves are monitored
to estimate more accurately the pest population. In addition, a more accurate estimate of the economic threshold,'
due to a finer partitioning of the state space, can be combined with the monitoring activity to avoid pesticide applications for insurance purposes. Pesticide salesmen have
267
"See Stern et al. (1959).
• ];01' a definition,see !fall (1973).
268
ESA
Vol. 23, no. 4 1977
BULLETIN
Y is an Nxl vector which measures
expenditures. {J' = (fJo, fJ,).
Labor
yield or pesticide
X is an Nx2 matrix where the 1st column is a column
of ones and the second column %. is given by
S
%.
1 for users of SC
of CC
= l 0 for users
e is an N x 1 error term.
Supervised
Two sets of assumptions can be made concerning the error term if statistical inference is desired. The STRONG
ASSUMPTIONS
are that (i) e is an independently
normally distributed random vector with mean equal to
o and variance equal to u"I, and (ii) X is non-stochastic.
Control
Assumption
Conventional
Control
Materials
Diagram
e""
and Capital
The WEAK
a)
Salesmen have an incentive to learn and practice IPM
techniques only to the point where they can remain competitive with independent consultants. Contrarily, independent consultants can charge a per acre fee up to the
savings in pesticides obtained. Hence, they have an incentive to remain current in all new IPM techniques developed by the university and by the consultants themselves. One may argue whether supervised control and
integrated pest management should be used as synonyms.
Clearly an economic incentive exists which tends to support the interchangeability of the 2 terms for crops where
integrated pest management programs have been developed.
The large reduction in pesticide use which results from
the use of independent consultants demonstrates that salesmen have a conflict of interest between profit maximization and the social goal of reduced pesticide use.
Statistical Analysis
In 1972, a stratified random sample of ca. 100 cotton
and citrus growers was obtained. The sample was stratified between users of SC and users of CC, as well as by
county and crop. The optimal number of observations in
each stratum was obtained by minimizing a quadratic
form (weighted variance-covariance matrix) subject to a
budget constraint and the cost of interviewing. A Mahalanobis (1936) distance test was used to test the hypothesis
that there was non-response bias. The hypothesis was rejected. For details see Willey (1974)."
In 1975, ca. 75 cotton and citrus growers were reinterviewed. The 2 interviews provided 5 years of yield and
pesticide data, 1970 through 1974. In addition, some data
for 1969 and 1975 was obtained. An analysis of variance
was performed for each crop to estimate the difference in
yield between users of SC and users of CC, and the difference in expenditures for pesticide materials between users
of SC and users of Cc. Estimates were obtained for each
year separately in addition to estimates for the entire
cross-section, time series sample.
The analysis of variance model can be written:
" Willey, W. R. Z.
information
technology.
Cniv. of Calif., Berkeley.
=
1974.
Ph.D.
N(O, erI).
ASSUMPTIONS
(2)
are as follows;
A
no incentive to monitor as closely since they would prefer
to substitute pesticides for monitoring activity at the same
total cost, yield, and profit to the grower.
Y
(i) can be written
X{1
+
Eo
(1)
The diffusion of pest management
dissertation,
Dept. of Economics,
X is stochastic and independent of e,
b) E(.IX)
= 0,
= 0-"1,
c)
E(e"IX)
d)
Plim(~
e' e )
e)
plim (~
X' X )
f)
plim (~
X' e )
= u',
= 1:, a finite matrix,
= O.
The weak assumptions are only applicable if a large sample size is available.
Independent consultants have claimed that potential
clients can at least maintain yield and may increase yield.
Others claim SC reduces yield. Further, the consultants
claim that savings from reduced pesticide use are greater
than or equal to the fee per acre charged for their service.
The opposite has also been charged. These claims and
counter-claims can be phrased as hypotheses which can
be tested statistically and either rejected or verified.
Let Y = yield. The null hypothesis that no difference
in yield occurs can be written
= O.
Ho: {1,
Yield was measured in pounds of lint and number of field
boxes for cotton and citrus respectively. From Table I
it can be seen that for the 5-yr period 1970-1974, a
2-tailed test does not reject the null hypothesis, with a
significance level at 0.05.
This result holds for both crops. Hence, the null hypothesis is accepted. The number of observations is large
enough for both crops to base this result upon the weak
assumptions listed above.
Whenever the null hypothesis is accepted, the risk is
run of making a Type II error. The probability of a
Type II error equals 1-,8 where ,8 is the power of the
test. Since hypothesis testing was not conclusive to determine the effect of SC on yield, interval estimates are
provided in Table II. The interval +2% to -8% includes the true change in cotton yield with a probability
of 0.95. The interval +15% to -7% includes the true
change in citrus yield with a probability of 0.95. The
large sample size permits the conclusion that even if SC
reduces or increases yield, the change in yield is not appreciably large.
Vol. 23, no. 4 1977
269
ESA BULLETIN
=
Crop quality is a crucial marketing characteristic for
citrus. Cosmetic standards insure an attractive product
exterior. Van den Bosch et al. (1977) hypothesize that
citrus cosmetic standards force unnecessary additional use
of pesticides. Moreover, they assert that the effectiveness
of IPM is jeopardized by cosmetic standards. For example, if pesticides are applied for citrus thrips to control for
scars on the rind, predacious mite populations can be reduced, upsetting an effective biological control. The imbalance created can lead to increased mite pest populations, requiring additional pesticide applications. An upset of natural enemies leading to secondary outbreaks of
pests has been referred to as the "pesticide treadmill."
Van den Bosch et al. (1977) argue that cosmetic damage
from thrips is indistinguishable from wind damage. They
argue that marketing co-ops set cosmetic standards in
order to regulate the supply of oranges in the fresh fruit
market. As supportive evidence, it may be noted that
during frost years when the quantity produced drops, special grades are permitted which allow slightly damaged
fruit to be marketed as fresh fruit. During abundant
years that same quality fruit is culled for much less
valuable juice, by-products, and waste. In addition, they
argue that standards for all grades of fresh fruit are
lowered during seasons when weather adversely affects
the quantity of yield. This marketing procedure acts as
an incentive for each grower to demand additional pesticide applications in order to avoid culling on the basis of
cosmetic damage. Evidence that SC programs accept a
reduction in fruit quality to avoid the pesticide treadmill
would support the claims by van den Bosch et al.
to SC where P
the avg price obtained by users of CC
over the period 1970-1974.
and ~'2 are given in Table
1. P
$2.21/field box. Hence, ~Q
-$54.57/acre.
An
interval estimate for ~Q is given by (+51.9, -161.0)
with a significance level of 0.05. The hypothesis that ~Q
= 0 cannot be rejected at the 0.05 significance level. The
point estimates suggest that yield quality may be slightly
reduced by IPM, with % of the losses recouped by increases in yield. This supports the contention that cosmetic standards hamper the effectiveness of IPM.
=
Statistical inference for a particular year requires the
strong assumptions because of the smaller sample size
for each year. Any results requiring the strong assumptions should therefore be discounted accordingly. First,
it can be noted from Table 1 that there was a statistically
significant decrease in cotton yield during 1972 and a statistically significant increase in citrus yield and revenue
during 1973 for users of Sc. For all other years, no statistical significance is noted. Yield may decrease from
IPM due to pest damage. Yield may increase from IPM
due to phytotoxic effects of pesticides and also due to pest
damage caused by pest resurgence and secondary outbreaks associated with conventional control. Second, point
estimates indicate that the change in yield due to SC
varies from year to year and from crop to crop. Third, a
time trend can be detected suggesting that SC is an improving technology with respect to yield and revenue.
This trend exists in both crops for the 3 years from 1972
through 1974.
Earlier results by Hall et al. (1975) estimated a decrease in pesticide use of 50% between users of SC and
CC in both cotton and citrus for 1970-1971. Here, pesticides are measured in dollars. Those results are confirmed
with the larger sample size afforded by 3 more years of
observations from 1972 through 1974. Using the weak
assumptions, the null hypothesis that pesticide use is not
reduced can be written
H. : fh ;;. O.
In order to test the hypothesis that SC reduces the
quality of citrus yield, let Y = revenue from citrus. That
is, Y is a weighted avg of yield quality where the weights
(prices) reflect yield quality. Let ~u represent the estimate of fJ, when Y = yield in field boxes. Let {J,. represent the estimate of {j, when Y
revenue. Then ~Q
p~- P·~n is an estimate of the reduction in quality due
=
Pn
=
=
Table 1. Effect of SC on per acre yield and per acre revenue.
1
~._Jll
Yl'II[''L-..-_--+
l:.t>tll1l,ltt'J.
T
•
f
B,
•
Lint
911.4&
45.tH!
1
1912
1971
1970
J::stimdt~d
T
Ratto
... ~~[J!~~~ill'l.-...l_coefflC:lent
959.5)
11:1.19
1:15&.22
IH.48
.6558
b4,11
1.0664
1121.2J
,
-,,::t1.62
JI,
-.987J
"l.U
I
~)9Jl
8'
~'ldd
"
~Ii
:t-.
__
-:'.9~
65
331.0.
JJ.38
332. ~2
1J,ijl
.6271
-16.25
I
",
39U•• '
Rt!Vl.'IlUl'
8.
-11.4 .•
~-","-~I~~
L-
.SI1Uf'::,.l'.1
4_2
__
45.H7
-.!4.0"
l
dlhllyail>
I
219.83
1971 ••
of
I
18.19
i
.6558
1.0.44
57
lJ.18
333.84
14.14
33.03
.5314
LJ.!.65
2. bib.!'"
90.0J
1.3.SS
35
.0085
Jb
.1676
3.1.86
21.44
-S •• 71
-2.5705"
.90.80
13.40
-51.10
-1.2491
I
.9
.01.7
51
.1l/Hi
}4
10.46
I"
~13. 76
10.10
891.97
9.29
4t15.77
I
-41,97
5•• 41
.1938
':'50.~H
20
.0048
18
from ""illey
-.1540
.0169
I
1971.-1914
datd
35
Survey.
mCd:.ured in pounds.
Citrus
by multiplyinS
cotton
lint
estimates
J.LIlJ
~/lb.;
1973" 71.S4 dlb.:
1914
•..ln<:c of
cltruR
J(l:.t,\ti~:
N •• numb~r
of observa.tions:
.- ••• sign if iCdnt
at the • OS level.
-. 46Y~
37
(rlJc
Hall
.00&)
survey
___
39__
•• multiple
is ceasured in field
tilDes
apprupriate
correlation;
do
All
<'lndel
"_"._U48U
430.7S
26.14
-17.16
-."/059
.U291
".l6
l.3796*
.26ll.
171.1S
.6j]S
21
_~J!.~
prices
Years"
boxes.
r~portcd
For illustration
by U.S.D.A.
as
Citru!'l
revenue
39.97.
purpost!s
only, cottoo revcnu~ C'Sl1i,1<.ltl'8
follo\ols:
1970·
22.91
c/lb.
j
1971 •• JO.6U
estimates
were formed
by pcrformLnu an
revenue.
R'
.009v_
•
yield
•• 42,53 <:/ib.:
-. 70~Y
.0291
.OU.8
-1)2.70
26.24
-40.34
54
.5
I
1012.I:H
231.J7
487.75
i
vdrl
18.48
19.b7
23.40
-1.2491
16.48
.0131
.!62.69
T
Ratio
.1181:1
L4.J'!
~_~
l'.Ilcul.1ted
41
5l
j----)67.10
15.46
965.61
-71.:'7
-1..S71,1S*
.00J3
-.2'14J
197U-l~71 data
C/lb.;
.0167
-.7194
.00t.J
.S~.I!.!.~n.ltlons: Cottun lLnt Is
Wt'rl'
JJ5.5.
-.417U
-.987J
1jU.l~
e
b9
15.JJ
t
---~! - _._Ot!~
c.:utlllll
ItL'Yenut'
.0068
21.44
-l7b.18
I
1
1__
~-",-_R_~
1973
E.Stlr.1olted
COi:lfHcient
T
RatIo
t:sLlliloltcd
Coefficient
T
Ratio
E"tlmolted
Coefficient
are
estimate!:>
of
a;
a.ndal
given
in
equ<.ltion
I):
SC"
supervised
controlj
I
270
ESA
Table 2.-Confidence
All
\'cdr6
;'.:Ji
CC
t:otton
5
---t-
-8 ••
-76.JO
lint
S
50.04
nell.!
III
IJ.tH
"
1
0••• 11
.
U.I;lO
- 8"
•• ~.
I;X
35
I
I
[>.
164.3':1
_t :..
lJ",
I
7.
-36.17
'. o[
LC
oJr
LC
s,~
47.lJ
1974
1912
I ':Ill
I
107.11}
-J.,
= Yield.
intervals for {Jl when Y
-1-
, of
CC
I
19.0b
(U.s)
c.;itruJ;
1970
Vol. 23, no. 4 1977
BULLETIN
"
-bl,17
-5.
2 ••• 4.J
-17o.ld
-tb';
-71.57
;.
of
55 ••. 0
3.'~
-7:
)"
-';",).j4
~_-l91.19
-~--~-
I
Jl.
u(
----~-
CC
lJb,u6
Jb ..
j
-lb.
-7.
-2.2 .4~
-tU.85
S
J.b/.
Revenue
II
-11. ••",
-J ~
-30.50
-b.
kevcnUl'
ii
17X
-11.1u
-lb.70
77.1)2
164
-24.05
-62.7v
-lJ ..
1970-1971
J.1t..1
;:111l')' Survey.
frur:"
11,.40
7.
lY.u7
1)2.7';
IUd.71
-4l~
Table 3.-Effect
1970
Estimat~d
Coe.fficit:nt
All Years
Eltlrlated
T
Coefficient
Ratio
Coltvn
Uateri,ds
,·
N
j
Citrus
B'I
Hat~rials
aI :
!L..1L._._
CottLln
M
+
A
,
Citrus
tl
+
A
·
I
.1276
48
19.:t7
24.05
27.18
-14.14
l77
I
-II. 93
-6, IoJ5l:l*
457
t
Cotton
P
rPMC
B,
I
t
23.75
57.01
-27.16
91
--- 25.91
I
I
;: I
-11.51
-5.3731*
.09;0
lb.7I
I
·i
!LJL
15.36
-10.50
257
I
B, '
·
~--:-t
16.71
-6. lOsa*
.127&
22.85
277
.0950
9.90
-17.71
i
48
I
I
-5.37.3l.1
23.75
·
S
TPHC,;
P,
N
R
AFPA
·
Cotton.
Citru~
Source:
-7.19
257
o..:itrus
·
-2b~
387.54
4J';'
fr.:.;':1
control
5'.40
-24.19
I,
57.01
-3.U79l.1
.1n&
I
T
Ruio
7.70
I
I
-).I:iU02* :
.23~9
'7.16
1971
Estimated
Coefficient
13.29
1l.6;
I
i
-38:
-1.74.72
percentdge
of
Percellt.Jg~s
5.
.lr~ olculatl'd
by dividing
1971
Lst1mated
T
Cucfficil"nt
Ratio
T
Ratio
Il.Ou
9.06
5.44
I
48
13.28
44.95
7.70
I
-13.2l::1
17.1&
!
I
!
I
:
!
-1.7909:11
.2389
l:l.9lJ
!
-1.
7b7~'"
I
277
4.74
1970-1971 data
-1.4746'
-I. 5~
-. '602
.0950
91
.1305
19.79
from :/illey
4.70
Survey.
.t ••t1&t1{:
20.61
1971•.l974 data
~.tl lllat1!d
(.;u~~f
~s!.('nt
-14.79
68
54
55.68
-l8.51
-3. b9O):II
• IHl
9.0&
-3.9994'
.23;'
-IJ.l8
•.1. Jlhrll
51
.099;
~f'
10.47
4; .4'J
-l.0700*
-l; • J6
8. Jl
I
-1.96J7*
i
-U,dJ
l•
16.57
5.44
I
-7.9U
-I.7b7!.*
;
4"
.0636
I
d.Jl
I
I
I
I
13.26
I
-3.6905·
1
,
44.h
.14':H
I
68
.1711
22.43
537
-8.77
-2.6391*
I---i 14.b~
,
I
H
I
-Iv. LiO
51
,
47.50
-1.9637*'
l4
.h91
HI. 57
J.U
-2.86
-.6395
!
.8
.0636
!
55.68
6.89
i
44.1':J
••• 31
-8.56
-1.1077
q.H
.1711
19.1j,j
-4
.9.
_~_b___
.1l.14
-- --
7.29
!:I./.'.'
-l.N~~rIl
____ .0)(;1
7.4~
-1. J5.•It
.y':'d_'_
II!_
l5 .8b
-"
-2. J.,!74*
-~
• jt.
t1. ~Ij
-1.79bO*
- -
10.47
07.10
7.49
I
-l5.60
.2351
68
T
Ratio
,; .J,
7.29
':"J
1),
54
!
--~---
T
katio
14.60
-t).
i
i
i
i
.0636
I
-J.dOO:l:ll 1
-3.6542·
-~. 6d
.235l
ll.43
I
•.).999 .•11
54
:
.lJl::I9
'973
Lst lmated
Coe£fic1ent
-44.7J
1_____
-7.6
l4
5.U4
-.58n
.141)1
18.00
-48
-2.0700·
J!___ ·~I~
24.&4
-6.11
-1. 3371'
5'
.O9!15
•• 7,Su
-.5J5&
33
.1214
18.33
.l::I'i
-1.h4~t
-
-
,; .M
-J
• jJ
4.rlb
-.
7l.~;
--
5.4J
-6.!tlJ
4.59
-
4. J2
I
-7.47
the currt."Vondin~
in Table l).
2d.88
i
q.56
52.40
167X
I
-7.87
.1305
.'305
48
77LJ.4lJ
<1.3.11survey.
(Le.
-3.6542*
91
I
-11':i.60
of SC on per acre pest management expenditures.
25.92
-13.01
iJ.5u
JX
50.41
-19;.67
dolta
• I
!h-R
-YJ.73
I
·1
!!..a.-.!-I___
14.07
-7.07
-ltl ••
41"
Explan.cions:
All cvnf1dt:!t~":t'intcrv.:Jl •• have a significance
level equal to .u5.
next to the pt;:rcent by the .l;lpNpriate estioate
of a,given in Tablt! 1.
Notation:
S. Buprecn..•.•
ll uf the l.:r1tl.cal
Region
--,-.
infimuM of the Critical
RC£,ion
H. mean. % of CC • perc~ntage of conventio.lIil
-56.71
-41. 9;"
1972-l974
Y.
-7.!.llu
19.
50.':'",
lu.8C,
-20J.2l
~:
JJ ••..d
-27.
-':IJ.71
3b.30
Cotton
t..:ltrus
-;:'7.:'j
2.5
"7 .}IJ
I
J.tilJ
-Jv
i
;
from Hall survey.
Explanations:
All 1:IeaSUreIllent6in dollars.
~1at~rial plu6 Application
statistics
w~re calculated
as iollow!/. - for COtton. IIIatl.!ridl expenditurt>s Wl'r~
multiplied
by 1. 6878j for citrus.
material expenditures
were multiplied
by 1.91.7!l1(se~ text for dhcussion).
Total pe.t mana"ement Loats w~re calcuLItf'd
as follows - for ~Il the average fee per acre was subtrac.ted from the matl;:!r1l1 plus application
est!.llate of BI j for T-rati.ol,
the l:fitimated coefficil'nt
W'asdivIded by the adjusted standard devIation;
the adjusted standard deviation cqudls the standard deviatLoo of B1lor materials
Cimrlo 1.6~7Cj ur 1.1Ji./.;
for cotton and citrus respectively.
a,
~:
N. number of observations.
R
multiple conelationi
and BI are estimates
of a. and 8. given in equation
Materials·
expenditures
[or pesticide
materials
only; H + A • expenditures
for materials
plus application;
tPHC • total
inc1udu expendituuus
for materials
plue application
plus th~ independent consultant's
fee per acre where appropriate;
charged by the consultant;
•• dgnificant
at the .05 level; t • s1.goificaot at the .10 level.
Z
•
1). l:iC· Io"perviaed controJl;
pelt maoa"au:nt coat anu
AFpA • ay~r."e tee per d~r~'
lilttt:o
Vol. 23, no. 4 1977
ESA
This hypothesis is rejected for both crops at the 0.01
significance level using a one-tailed test. The alternative
hypothesis is accepted with a great deal of confidence that
pesticide use is reduced.
and per acre fees charged by independent consultants, as
well as expenditures for materials noted above. Application costs are not presently available from the sample. A
reasonable assumption could be made that a linear relationship occurs between material costs and application
costs. Ferguson (1975) reports that in 1971 cotton growers spent $51,324,000 on pesticide materials applied by
commercial applicators and $35,300,000 for application of
the material. For citrus, he reports that $9,197,000 was
spent on pesticide materials applied by commercial applicators and $8,534,000 for application of the material.
Hence, cotton growers spent 1.6878 times their expenditures on materials for application and materials while
citrus growers spent 1.9279 times their expenditures on
materials for application and materials. Point estimates
of the reduction in per acre expenditures due to SC for
application and materials, given in Table 3, equal $11.93
and $27.26 for cotton and citrus, respectively. Interval estimates equal (-$8.73, -$15.14) and (-$18.91, -$35.61)
for cotton and citrus. Unweighted avg fees per acre
charged by independent consultants for 1970-1974 are
given in Table 3 at $4.74 and $19.79 for cotton and citrus,
respectively. Point estimates of per acre savings in total
pest management expenditures due to SC are $7.19 and
$7.47 for cotton and citrus. The hypothesis that savings
in total pesticide expenditures due to SC equal zero is
rejected at the 0.1 significance level. It can be concluded
that supervised control reduces total pest management
expenses, taking into account the fee per acre charged
by the independent consulant.
Interval estimates are provided in Table 4. The probability is 0.95 that the interval (-37%,
-64%)
includes
the true reduction of pesticide use in cotton. The probability is 0.95 that the interval (-37%, -62%)
includes
the true reduction of pesticide use in citrus. It can be
concluded that SC reduces pesticides by at least If.J and
by as great as % in both crops. The huge expenditures
for IPM research seem to have produced a social benefit.
Using the strong assumptions, the null hypothesis that
pesticide use is not reduced is rejected at the 0.10 significance level for both crops in each year.
The point estimates in Table 4 indicate that pesticide
reduction due to SC varies from year to year and crop to
crop. There is evidence of a time trend in the percentage
of reduction in pesticide use. It appears as though the
percentage reduction may not be as great as it once was.
For cotton, the time trend of the point estimates of {3o
have an interesting interpretation.
Dollar expenditures
were not adjusted for the slight increases in pesticide
prices which occurred from 1972-1973 and the sharp rise
from 1973-1974 due to the oil embargo. However, estimates of {3o remained relatively stable, indicating a reduction in per acre pesticide use by conventional control over
time. In order to remain competitive with SC, pesticide
salesmen may have adopted some IPM techniques. Salesmen have claimed that they check fields more often and
more uniformly for pest populations than in the past.
Total pest management
costs include application
Table 4.-Confidence
--"IHtun
"
1J.tE'rhlll
M
---
--
1t['ulI\
"t",['ta18
--
1 ___
-5.1$!>
-38~
-4.51
-)<.;:.
-1.01
-ill:;
-10.50
-b8%
-1.ijl
-5~"
-99%
-11.11
-6.26
II
-14.14
-48.;
II
A'Dulil-atiun
1
___-18.47
___-b2:
-8.73
-J7,
-11.9J
M
-~O:
-M.t
_~~4
" I
of
CC
-Jl.
-15.14
-1 •• 91
I
I
I
I
-Jll
-23,
I
-2M,'
-11.51
-42;';:
-14.79
-51 _
-lb.76
-b2X
-21.48
-7w,~
-9.89
:
-8':-
-8.10
Cutt,on
I
-.J" ..•
-lI.1l
-b8t
-13. ,8
-5; ..
-2~.55
-99"
-18.l:S5
-cl':'.,
-1.2..1.17
-n.
-15.bl
-~;';';
-42~
-lij.51
-;1,1;
-J5.bl
-bl%
-32.JI
-bl%
-41.41
-1t+l.
"
~~_l-rTutdl
'1
-7.19
-3O;
-1Il.4Q
-44.
1
t:ltrus
S
f
Tntdl
M
I
I:.Xpll'lII'Io'1I
I
~:
I
-10;':
-3.l0
-144
-lJ.Ol
-502
-8.17
-~~.•.
I
-'0.ij5
-8O.
-14.J4
-b4~
,
.81
-7.47
-1,).81
2%
-131
I
-l.l
\
U.53
Ib2
-1. 58
-J:";
-11.b9
Survey.
i
I
-- ------
LL
-1.1<j
-12::
-. j;
-L4
-4.bij
-43.
-b.~0
-!t41
-4.1)4
-Ja.
-9.LJ
-1:13"
-74:4
-Y.!')
-1.10
-n
-13.lU
-58%
-10.8)
-,. J'
-1004
-.1.9U
-150.41
-25. bO
-12.d;
-
- 9'
-_-?.~----=~-
I
-H
-l;4
I
83.
-71.
-,)1:1%
-47.91
-IDOl
4.b5
lSI
-l •• b
-I ~2
-,14
-3.1.1_
-10.8u
-44~
-71.1:,
- ... ,~
-2.4.;
J
-41,.1J&
I
-L)~
332
13•• '
-15,.
!
-7.b
-17 ••
-6.40.1
-31"
I
-lJ.1~
~.j
-56",
-IU~.l:
-.SIJ
-6. J4
-Jl.l.
-15.b5
11.97
-b. ,1
14. Il
, ..-.::?()_ ..
-48.tltJ
-9,"
IJi.
-l5.;b
-9.:
b.
I
-56"
b.ll
I
-.:d ~
1.;;
-10.37
1
-ll",
-ld.31
i
4.34
-8.
data
&1Knlf1cance
8~
Sob
-21.4b
-loll
1971-1974
All confidt!lIct! intlL!.rv.lls ha\le a
n"xt to tlle J.lcrcent by thl;! apprupriate estloate
--~~~-~
4.4l
111,
-J. JJ
-29.91
level
equal
to
.05.
Z'cn:cnta~cs
arl!
~f 0. b1vcn in Iable Ill.
S· 5upremum
of the I.:ritical
Region
infimLlll of the Critical Rl"gion
M • mun. % of CC • percentage of conventional
'"
control
-bts4
-lb.td
,91
-lO.tl4
3". J/
-In
-3U.4 •••
-5b';
-1J1.Jb
-J~1.
from Hall sunrey.
Nnt_t!lln:
--1-·
"ccof
oC
-2.
-J.lb
I
-.1~74
. -.; o{--
1973
•cc
I
191U-1971 datd from \/illey
EXplanation.:
72
-, :.!j
i
-ll.19
-~.19
•
-.39
-7.71
-~tl.
-lfJ"
I
__ -14.1$!>
-J81
-17.lb
-J.99
Pest Management
I
and
~.~
of Integrated
Management Expenditures.
-.
•
-5.17
-3J:
:1,Hltr1ah
no!
1911
oC
CC
-b4~
= Pest
when Y
(3.
:
uf
CC
-9.ijl
s
K,dtt'ri ••b.
intervals for
.-ij.91
C.Jtt,1I1
l:!trub
~
The Profitability
There is no statistically significant difference in profit
between supervised control and conventional control for
costs
1970
S
--_._~.!.
.
All Ye;us
271
BULLETIN
(i.e.
percentag~
o,)f
at
ir. Table
lII).
calculatt:d
by
dividing
the correapund1nt;
bt,llLII_
lib,lcd
272
ESA
both crops when considering the entire s-yr period 19701974. Results are presented in Table 5. The large sampie affords a stronger conclusion for cotton: from the
interval estimate (4%, -6%)
one can be almost certain
that the difference in profit is negligible. For citrus, the
interval estimate (19%, -24%)
is somewhat larger although not alarmingly so. These conclusions may be
based upon the Weak Assumptions.
The net profitability of SC does vary from year to year,
given the Strong Assumptions. In 1970 and 1971, profit
was greater for SC than CC in cotton, with a significance
level equal to 0.1. In 1972, profit was less for SC than CC
in cotton, with a significance level equal to 0.05. No statistically significant difference is noted for cotton in 1973
or 1974. For citrus, a statistically significant increase in
profit of SC occurred in 1973. All other years had no
statistically significant difference for citrus.
Limits of the Empirical
It would be difficult to accept the assumption that X is
non-stochastic.
If X is stochastic, assumption (i') must
include independence of X and e. In addition, assumption
(ii') must be altered as follows:
elX ,.., N (0, d'I) .
Johnston (1972) has proved that the unconditional t-distribution is equivalent to the conditional t-distribution,
given a fixed sample size. As noted above, the number of
observations in each stratum was fixed before interviewing. However, the point estimates are only conditionally
optimal under the strong assumptions if X is stochastic.
Assumption (i) states that e is normally distributed.
This assumption is questionable at best since Y ~ O.
A truncated normal distribution would be more appropriate.
Y
= {3. +
Sun[~mum
All
,,,
'lcars
%u
1970
An
where Y = cotton yield or pesticide materials and applif
dX
S 1 if the grower uses consulant i
ca lon an
\ = l 0 otherwise
of fh through flu are given in Table 6.
Applying the strong assumptions, it can be concluded
that 3 consultants reduced yield, although 1 of the 3
compensated by large enough savings in pest management
expenses. The hypothesis that no reduction in pesticide
use was obtained is rej ected for over half of the consultants with a significance level equal to 0.05. The hypothesis that no change in profit occurred is rej ected for
3 consultants, one of whom increased profit while the
other 2 decreased profit.
Assumptions (i') and (a) require stochastic independence of X and e. This means that no omitted variable
which affects Y may be correlated with the explanatory
variable.
If Y = yield, a number of omitted variables may be
correlated with X. It is expected that the amount of
pesticides applied depends upon whether the grower uses
1 ~71
T
Ratio
11.75
-.3617
ull,
23.82
111.
;3.85
2b •.• 4
Ii
-4.91
17.11
1.3922t
-2%
i
-2.97
]0.75
1.5106t
i
1972
197)
A.
A.
22%
11.
-1%
-16.06
-5).8;
-91. 64
-4%
-lSi;
-25%
24.19
-44.99
-114.17
4%
1974
An
27.58
6,
~:
1970-1971 data
-7%
-J).8)
-n
SUllr~UJum
Mean
Infimum
1)9.15
-l6.65
-1/2.45;
gross
I
-55.24
I
-J)%
Survey.
19~
I
II
I,
79.78
-61.1l
16%
-J)t
-)].41
ll7.80
22.51
-2. )923*
41.25
-I. 091
I
I
24.75
-.5So8
supervised
controli
Ij
signifi.cant
*.
;
I
T
Ratio
I
-.1758
94.71
~
--JI
--l
-20~~:~ ~~~J7 I ~~;:-_.II
- •• 1",
_
I -134.00
I
-267.0~
I
-.37llJ
395.9
04.01
_.-
i
44~
7%
.3J17
192.40
I
!
Uo.ov
,i
109. )2
-2b4.5L.
I
270. Mo
I
7MS.94
I
I
-3u~ i
457. Jl
169%
I
9M~
201. 64
667.82
I
16%
fran Hall survey.
standard
deviations
it was aS6um~d
pest management expenditures.
All
the Critical
Region.
6"11' • change in profit
or net;. profit
of
revenues
are estimated
by So in Table
Stannard
j)~vhtion
-7,;'
I
55Z
1972-1974 data
I
,I
."
-L,~
I
-In
from Willey
Explanations:"
In Rrder
to estimate
revenue
and BI,,:::t 6) when Y • total
The !;iupremum ~nd infimum determine
~:
where
I
!
24%
(3)
profitability of integrated pest management.
Standard
Deviation
\5.08
52.55
+e
Citrus
Infimum
Mean
11
2: (3,X,
i=l
Cotton
I
1977
Assumptions (b) and (d) through (f) require little
in the way of restrictions upon reality. However, assumption (c) rules out the possibility that the effect of SC
varies from grower to grower or from consultant to consultant. This assumption could be relaxed if a production
function and a demand for pesticides relation could be
specified.
Empirical evidence indicates that yield and pest management costs do vary from consultant to consultant. Consider the model
Estimates
Analysis
Table s.-Net
Vol. 23, no. 4
BULLETIN
eli
that
confidence
l6
at
at,. 81 when Y equdl
to .05.
and
aft:"!
uncorrelated,
'Where
intervals
have a significance
level
~III
•• the pt:"!rcentage of 6ro!:iS revenues
the .05 level;
t • significant
at
which
the .10
Off
r~prl!scntSt
level.
ESA
Vol. 23, no. 4 1977
Table 6.-Comparison
St ..mdard
Estimated
£Q..~..!(h·il'nt ~~lon
~,
17.6l
19.51
-7.l6
33.49
_-,ll...
lL.J
~.
-51.
",
1.••• 34
35.41
0'_
_~J.L...J_~.-I.
T
-'18.98
~.
5B.1J
56.90
I
.-.1._
~".. -87.15
69.47
I
Source:
1971J-1911 data
J=;xulanation ••:
~:
All
Pee
Per
Acre
N
Ratio
30
-ll.94
1.37
16.89
3.42
-3.4954'
25
54
-11.61
-2.1000'
I
18
-10.44
.690]
!
7
-1. 755d*
I
-15.45
9
-.2~
5.39
-2.8645'
9
Prof it of
I
--
SC
I
Estimated
Coefficient
T
Rat!o
Standard
Deviation
3.20
29.56
19.Bl
1.49221
2.11
8.19
33.92
.2415
51
-2.2936'
13
1.19
-15.56
15.69
.9917
3.28
-40.67
24.76
-1.64261
7.13
.3391
5
1. 71
23.63
]B.40
5.13
-3.605B'
10
7.36
__-B2.18
35.78
I
3
-22.56
9.14
-2.4690'
3
7.50
-55.21
57.63
.95BO
i
7
-10.14
-1.4230
5
3.00
-28.84
3B.40
.7510
·1.2212
1
2.67
-]B.95
59.0J
70.36
I
i
7.13
I
-19.18
3
2
15.70
:
5.56
-2.64
4
I
from \:tlley
ml'9surer:lt.mts
Survey.
are
N - number of observations
11.15
I
..4992
2
3.25
-92.71
9.14
I
-.2BBB
3
1.6]
-.22
1n dollars
••
L972-1974
per
data
acre.
- slgnflcant
at
from halt
the .05 level;
A number of other omitted variables could be correlated with X, such as managerial quality, scale of opera·
tion and pest population. If better or large-scale growers
are more likely to use SC, this would tend to bias the
estimates in Tables 1 and 2 since managerial quality and
scale of operation are e.xpected to influence yield. In ad·
dition, adopters of SC are likely to have larger beneficial
insect populations, leading to reduced pest populations.
Pest populations presumably reduce yield. Again, these
difficulties could be overcome by specifying a production
function with measures of managerial quality, scale and
pest populations as e.xplanatory variables in the production
relationship.
=
If Y
pesticides, a number of omitted variables may
be correlated with X, leading to bias in the estimates pre·
sented in Tables 3 and 4. These include managerial quality, visits with chemical salesmen, the pest population, and
pesticide prices. The solution to this problem is to specify
a demand for pesticides relation which include as explanatory variables these variables as well as whether the
grower uses Sc.
Pest Management
An important factor in the adoption of a new technology is the risk involved. Until a potential adopter collects information about a new technology, he will likely
perceive risk even if the technology is less risky than the
conventional technology. Unless it is believed that the 2nd
and higher moments of a distribution are accurately assessed, perceived risk and actual risk should be distin-
I
_-:1.
296B' •
.6598
-1.3177
50.27
.0044
survey.
EBtl;:l:d.t~d coefficients
SC and it is expected that yield depends upon the amount
of pesticides applied. Hence, the estimates in Tables 1
and 2 are affected by simultaneous equations, although
the magnitude and direction of bias are unknown. The
problem would be alleviated by specifying a production
function with pesticides as an explanatory variable.
Risk and Integrated
.6154
I
I -1. 2546
-.0579
!
-IB.51
!
1.0215
-4.4910*
4.55
I
8
-2.84]8'
-1.0331
2.58
!
2.42
I
I
,
I
49.44
Net
I
.23.17
I
_._-2~B6
leo t
(S)
T
i
I
3667
J7.73
Coec fie
N
ADol1cation
Standard
Deviat ion
!
37.73
i6.U'
+
Es timated
I
.9031
13.47
II
_-:IUU.69
I
T
Ratio
.-----.- -49.72
---- 1.94
--------
_)JI!&~~_
of individual cotton consultants with Cc.
t1aterials
Revenuli'
273
BULLETIN
t",
correspond
significant
at
to equation
the
J)
in
the
text.
.10 level.
guished. In addition, there is a distinction between risk
preferences and risk perceptions.
Since, for large samples, random variables converge in
distribution to normal distributions, the 2nd moment captures the element of uncertainty. rPM is an information
technology. Since information tends to reduce uncertainty,
it is expected that the variance of yield is greater for
growers using conventional control than for growers under a supervised control program.
=
Let UIO= variance in yield for SC and uf
variance
in yield for Cc. Estimates for UIOand u,o are given by
1
_
1
_
SIO= - ~(Ym - YI)O and S22 = - ~(Yn - Y2)' where
M
N
M and N are the number of observations in each stratum.
For cotton yield during the period 1970-1975, SI2
=
=
=
42524.84 and Sf
83462.21 with M = 160 and N
136. For citrus yield, S12
12273.58 and Sf
17538.44
with M = 43 and N = 169. The null hypothesis may be
stated
=
=
Noting that
U12
p, { --
uf
(N-1)M
SI'}
(M-1)N
Sf
~ 1.35 ------
= .95
J
the null hypothesis is rej ected for both crops, The conclusion is that conventional control is riskier than supervised control.
Policy Implications
The availability of IPM provides the Environmental
Protection Agency (EPA) with an additional policy option. In accordance with the amended Federal Insecticide,
Fungicide, and Rodenticide Act as amended, EPA promulgated regulations
(§ 162.11 (C) (5»
which state:
274
ESA BULLETIN
"The Administrator may additionally or alternatively impose other restrictions by regulation. Such regulatory restrictions may include, but are not limited to, '" , Iimitation of use to approved pest management programs.'"
Pesticides which pose a serious hazard to human health
should be restricted to "approved pest management programs."
If a state, such as California, adopted regulations and
procedures which insured the use of integrated pest management by licensed independent (of pesticide distributors) pest management advisors, then EPA could consider that state's pest management program for approval.
The inducement to states to develop such programs would
be the possibility of using pesticides cancelled otherwise.
A compromise would be possible instead of either cancelling the pesticide or allowing its continued use at present greater than socially optimal quantities. If the cost
of such regulation is less than the cost of current regulation, this policy is dominant.
Hall, D. C, and R. B. Norgaard.
1973. On the timing and application of pesticides. Am. J. Agric.
Econ. 55: 198-201.
Hall, D. C, R. B. Norgaard, and P. K. True.
1975.
The performance of independent pest management
consultants in San Joaquin cotton and citrus. Calif.
Agric. 29: 12-14.
Headley, J. C 1968. Estimating the productivity of
agricultural
pesticides. Am. J. Agric. Econ. 50:
13-23.
Hueth, D., and U. Regev.
1974. Optimal agricultural pest management with increasing pest resistance. Ibid. 56: 543-52.
Johnston,
J. 1972. Econometric
Methods.
2nd ed.
McGraw Hill, New York.
Langham, M. R., and W. F. Edwards.
1969. Externalities in pesticide use. Am. J. Agric. Econ.
51: 1195-201.
Mahalanobis, P. C 1936. On the generalized distance
in statistics. Proc. Nat. Instit. Science India. 12.
ill See the Federal
Register, Thursday, July 3, 1975, Vol. 40,
No. 129, Pt. II, p. 28284.
REFERENCES
Vol. 23, no. 4 1977
CITED
Ferguson, W. 1. Farmers'
expenditures for custom
pesticide services, 1971. Economic Research Service, USDA, Agric. Econ. Rep. No. 314, 31 pp.
Fisher, A. C, and F. M. Peterson.
1976. The environment in economics: a survey. J. Econ. Lit.
14: 1-33.
Stern, V. M., et at.
1959. The integration of chemical
and biological control of the spotted alfalfa aphid:
the integrated control concept. Hilgardia. 29: 81101.
Taylor, R. C, and J. C Headley.
1975. Insecticide
resistance and the evaluation of control strategies
for an insect population. Can. Entomol. 107: 237-42.
van den Bosch, et al.
Investigation of the effects of
food standards on pesticide use. EPA Contract no.
68-01-2602.
New Approved Common Names for Insecticides
A Consolidated List of Approved Common Names of Insecticides, compiled by the Committee on Insecticide
Terminology, was published in the 27th edition of Pesticide Handbook-Entoma,
1977-78. Since then, the Committee has approved 7 new common names. These plus 3 approvals from the American National Standards Institute may be added to the consolidated list.
Chemical Name
Common Name
azocyclotin
1- (tricyclohexylstannyl)
bisazir
P,P-bis (l-aziridinyl)
epofenonane
Other Designation
-IH -1,2,4-triazole
Class
BAY BUE 1452
I,M
2-ethyl-3- [3-ethyl-5- ( 4-ethylphenoxy) pentyl] -2-methyloxirane
Ro 10-3108
C
G
*etrimfos
0- (6-ethoxy-2-ethyl-4-pyrimidinyl)
phosphorothioate
SAN 197 I
EKAMET@
I
*nitrilacarb
4,4-dimethyl-5- [ [[ (methylamino) carbonyl] oxy] imino]pentanenitrile
AC 72,613
AC 85,258 (complex
with zinc chloride (1 :1»
M,I
penfluron
*pirimicarb
-N -methylphosphinothioic
2,6-difluoro-N - [ [[ 4- (trifluoromethyl)
carbonyl] benzamide
2- (dimethylamino)
amide
O,O,-dimethyl
C
phenyl] amino]-
-5,6-dimethyl-4-pyrimidinyl
dimethylcarbamate
PP062
PIRIMOR<!i
I
CGA 15324
CURACRON@
I
profenofos
0- (4-bromo-2-chlorophenyl)
sulprofos
O-ethyl 0- [4- (methylthio) phenyl] S -propyl phosphorodithioate
BAY NTN 9306
BOLSTARTH
I,M
thiocyclam
N,N -dimethy I-I ,2,3-trithian-5-amine
SAN ISS I
EVISECTTH
I
O-ethyl S -propyl phosphorothioate
{hYdrOgen
oxalate
salt (1 :1)
• ANSI common name.
E. M. OSBORNE,Chairman
Committee on Insecticide Terminology