energy and the use of conservation tillage in us agriculture

8
Journal of Energy Policy 27 (1999) 299 } 306 Energy and the use of conservation tillage in US agriculture1 Noel D. Uri* Resource Inventory Division, Natural Resources Conservation Service, US Department of Agriculture, Washington, DC, USA Received 19 March 1998 Abstract The e!ect of energy on the use of conservation tillage is of special importance in addressing concerns about the impact of agricultural production on the environment in the United States. After establishing that a relationship exists between the price of energy and the use of conservation tillage via Granger causality, the relationship is quanti"ed. It is shown that while the real price of energy does not a!ect the rate of adoption of conservation tillage, it does impact the extent to which it is used. Finally, there is no structural instability in the relationship between the relative use of conservation tillage and the real price of energy over the period 1963}1997. ( 1999 Elsevier Science Ltd. All rights reserved. Introduction The use of conservation tillage has been increasing in the United States in response to growing concerns about the impact of agricultural production on the environ- ment.2 There is an identi"able upward trend until the last few years, which show no discernible change. A longer term perspective can be obtained from Figure 1. The use of conservation tillage increased from 1% of planted acreage in 1963 to 37% of planted acreage in 1997 (Schertz (1988) and Conservation Technology Informa- tion Center [annual]). Use of conservation tillage practices frequently reduces the negative impacts associated conventional tillage sys- tems which include energy use, soil erosion, leaching and runo! of agricultural chemicals, and carbon emissions. The relationship between energy and the use of conserva- *Present address: NRCS/RID (RM 1-2118A), US Department of Agriculture, 5601 Sunnyside Avenue, Beltsville, MD 20705, USA. Tel.: 301 504 2281; fax: 301 504 2230. 1The views expressed are those of the author and do not necessarily represent the policies of the US Department of Agriculture or the views of other US Department of Agriculture sta! members. 2 The de"nition of conservation tillage that is commonly used, and the one used here, is any tillage and planting system that maintains at least 30% of the soil surface covered by residue after planting to reduce soil erosion by water. Where soil erosion by wind is the primary concern, any system that maintains at least 1000 pounds (per acre) of #at, small grain residue equivalent on the surface during the critical wind erosion period. Two key factors in#uencing crop residue are (1) the type of crop, which establishes the initial residue amount and determines its fragility, and (2) the type of tillage operations prior to and including planting (Conservation Technology Information Center, 1996). tion tillage is of special importance and is the subject of what follows.3 In the context of the di!usion of conservation tillage as a new technology (production practice), a sizeable num- ber of studies are available that provide some insights into the important factors that a!ect the adoption of conservation tillage (Pagoulatos et al, 1989; Uri, 1997; Gray et al, 1996; Carter and Kunelius, 1990; Batte et al, 1993). The greater risk associated with the adoption of con- servation tillage has been shown to be a deterrent to the adoption of conservation tillage in a number of studies. Risk in these studies is typically de"ned as variability in yields or variability in net returns (Mikesell et al, 1988; Williams et al, 1989; Westra and Olson, 1997). Even though it is recognized as being important, one factor typically absent from explicit consideration in em- pirical examinations of conservation tillage adoption and use is energy. The variability in energy costs associated with conservation tillage is a consequence of the change in the number of trips across a "eld associated with conservation tillage relative to conventional tillage. Fewer tillage operations result in fewer trips across the "eld (Frye, 1995; Gri$th et al, 1977).4 In addition, 3 Energy for tillage operations accounts for approximately 3% of total farm energy use while energy use for fertilizer and pesticides applications accounts for 42 and 4%, respectively, of total farm energy use in the United States (Torgerson et al, 1987). 4 It has been estimated that energy use for tillage operations is reduced by between 3.4 and 9.8% when conservation tillage is used instead of conventional tillage (Frye, 1984). 0301-4215/99/$ - see front matter ( 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 3 0 1 - 4 2 1 5 ( 9 8 ) 0 0 0 6 8 - 8

Upload: noel-d-uri

Post on 03-Jul-2016

218 views

Category:

Documents


4 download

TRANSCRIPT

Page 1: Energy and the use of conservation tillage in US agriculture

Journal of Energy Policy 27 (1999) 299}306

Energy and the use of conservation tillage in US agriculture1

Noel D. Uri*Resource Inventory Division, Natural Resources Conservation Service, US Department of Agriculture, Washington, DC, USA

Received 19 March 1998

Abstract

The e!ect of energy on the use of conservation tillage is of special importance in addressing concerns about the impact ofagricultural production on the environment in the United States. After establishing that a relationship exists between the price ofenergy and the use of conservation tillage via Granger causality, the relationship is quanti"ed. It is shown that while the real price ofenergy does not a!ect the rate of adoption of conservation tillage, it does impact the extent to which it is used. Finally, there is nostructural instability in the relationship between the relative use of conservation tillage and the real price of energy over the period1963}1997. ( 1999 Elsevier Science Ltd. All rights reserved.

Introduction

The use of conservation tillage has been increasing inthe United States in response to growing concerns aboutthe impact of agricultural production on the environ-ment.2 There is an identi"able upward trend until the lastfew years, which show no discernible change. A longerterm perspective can be obtained from Figure 1. The useof conservation tillage increased from 1% of plantedacreage in 1963 to 37% of planted acreage in 1997(Schertz (1988) and Conservation Technology Informa-tion Center [annual]).

Use of conservation tillage practices frequently reducesthe negative impacts associated conventional tillage sys-tems which include energy use, soil erosion, leaching andruno! of agricultural chemicals, and carbon emissions.The relationship between energy and the use of conserva-

*Present address: NRCS/RID (RM 1-2118A), US Department ofAgriculture, 5601 Sunnyside Avenue, Beltsville, MD 20705, USA.Tel.: 301 504 2281; fax: 301 504 2230.

1The views expressed are those of the author and do not necessarilyrepresent the policies of the US Department of Agriculture or the viewsof other US Department of Agriculture sta! members.

2The de"nition of conservation tillage that is commonly used, and theone used here, is any tillage and planting system that maintains at least30% of the soil surface covered by residue after planting to reduce soilerosion by water. Where soil erosion by wind is the primary concern, anysystem that maintains at least 1000 pounds (per acre) of #at, small grainresidue equivalent on the surface during the critical wind erosionperiod. Two key factors in#uencing crop residue are (1) the type of crop,which establishes the initial residue amount and determines its fragility,and (2) the type of tillage operations prior to and including planting(Conservation Technology Information Center, 1996).

tion tillage is of special importance and is the subject ofwhat follows.3

In the context of the di!usion of conservation tillage asa new technology (production practice), a sizeable num-ber of studies are available that provide some insightsinto the important factors that a!ect the adoption ofconservation tillage (Pagoulatos et al, 1989; Uri, 1997;Gray et al, 1996; Carter and Kunelius, 1990; Batte et al,1993).

The greater risk associated with the adoption of con-servation tillage has been shown to be a deterrent to theadoption of conservation tillage in a number of studies.Risk in these studies is typically de"ned as variability inyields or variability in net returns (Mikesell et al, 1988;Williams et al, 1989; Westra and Olson, 1997).

Even though it is recognized as being important, onefactor typically absent from explicit consideration in em-pirical examinations of conservation tillage adoption anduse is energy. The variability in energy costs associatedwith conservation tillage is a consequence of the changein the number of trips across a "eld associated withconservation tillage relative to conventional tillage.Fewer tillage operations result in fewer trips acrossthe "eld (Frye, 1995; Gri$th et al, 1977).4 In addition,

3Energy for tillage operations accounts for approximately 3% oftotal farm energy use while energy use for fertilizer and pesticidesapplications accounts for 42 and 4%, respectively, of total farm energyuse in the United States (Torgerson et al, 1987).

4 It has been estimated that energy use for tillage operations isreduced by between 3.4 and 9.8% when conservation tillage is usedinstead of conventional tillage (Frye, 1984).

0301-4215/99/$ - see front matter ( 1999 Elsevier Science Ltd. All rights reserved.PII: S 0 3 0 1 - 4 2 1 5 ( 9 8 ) 0 0 0 6 8 - 8

Page 2: Energy and the use of conservation tillage in US agriculture

Fig. 1. Percent of planted acres on which conservation tillage is used in the United Sates

however, while the number of trips across a "eld to applyfertilizer does not vary between tillage systems,5 the num-ber of pesticide treatments and, hence, the number oftrips across the "eld to apply pesticides, is greater forcorn, soybeans, and spring wheat grown under conserva-tion tillage than under conventional tillage for someyears between 1990 and 1995 (Table 1). Associated witha larger number of treatments is an increase in energyuse. Thus, while energy use has typically been consideredto be a signi"cant factor in the tillage adoption decision,6

5Based on National Agricultural Statistical Service/Economic Re-search Service (NASS/ERS) Cropping Practices Survey data, the aver-age for corn for 1995 was 1.92 (0.03) trips per "eld for conservationtillage and 1.90 (0.02) trips per "eld for conventional tillage. (Standarderrors of the means are in parentheses.) The average for soybeans,winter wheat, spring wheat, and durum wheat for conservation tillage is0.31 (0.02), 1.32 (0.05), 1.31 (0.09), and 1.46 (0.11), respectively. Theaverage for soybeans, winter wheat, spring wheat, and durum wheat forconventional tillage is 0.32 (0.02), 1.40 (0.03), 1.29 (0.05), and 1.40 (0.09),respectively. Consequently, there are no statistically signi"cantly di!er-ences in energy costs associated with the application of fertilizer be-tween conservation tillage and conventional tillage. The same resultsalso hold for 1990}1994.

6This is illustrated by a provision in the Energy Security Act of 1980hat amended the Agricultural Conservation Program (ACP) authoriz-ing legislation to include conservation tillage and other ostensibly moreenergy e$cient conservation practices (US Congress, 1980). The ACPprovided cost-sharing assistance to farmers for implementing conserva-tion practices designed to restore and improve soil fertility and minim-ize erosion caused by wind and water.

important questions are whether energy considerationsin the aggregate de facto a!ect the use of conservationtillage and, if so, to what extent do they impact the use ofconservation tillage in the United States. Both of thesequestions will be subsequently addressed.

Energy and conservation tillage use

To address the question of whether energy consider-ations a!ect the use of conservation tillage in the UnitedStates, one approach is to assess the impact of the price ofenergy on the use of conservation tillage. This will bedone using a test for directional causality.

Methodology

Granger (1969) causality is a convenient and verygeneral approach to detecting the presence of an empiri-cal relationship between two variables consistent withthe theoretical one. Granger causality is de"ned as fol-lows: If X and > are the only two random variables ina universe and one knows that > cannot cause X but ifthe possibility of X causing > is still open to question,then an observed signi"cant correlation could be inter-preted causally. The assumption that > does not causeX gives su$cient structure to the situation for a causalinterpretation to be given. A method of giving structureto a group of economic variables is to apply the following

300 N.D. Uri / Journal of Energy Policy 27 (1999) 299}306

Page 3: Energy and the use of conservation tillage in US agriculture

Table 1Average number of pesticide treatments by tillage practice } 1995

Commodity Conventional Conservationtillage tillage

Corn

1990 1.63 (0.02)! 1.69 (0.03)1991 1.69 (0.02) 1.78 (0.04)1992 1.71 (0.02) 1.82 (0.03)1993 1.67 (0.02) 1.77 (0.02)1994 1.72 (0.02) 1.82 (0.02)1995 1.79 (0.03) 1.93 (0.03)

Soybeans

1990 1.49 (0.02) 1.51 (0.03)1991 1.54 (0.03) 1.63 (0.04)1992 1.57 (0.03) 1.63 (0.03)1993 1.47 (0.02) 1.62 (0.02)1994 1.67 (0.02) 1.77 (0.02)1995 1.64 (0.03) 1.87 (0.04)

Winter wheat

1990 0.41 (0.02) 0.55 (0.05)1991 0.39 (0.02) 0.36 (0.05)1992 0.39 (0.02) 0.32 (0.04)1993 0.50 (0.02) 0.49 (0.05)1994 0.53 (0.02) 0.55 (0.05)1995 0.76 (0.02) 0.75 (0.05)

Spring wheat

1990 1.26 (0.05) 1.42 (0.14)1991 1.17 (0.04) 1.20 (0.07)1992 1.14 (0.05) 1.03 (0.07)1993 1.19 (0.06) 1.12 (0.08)1994 1.24 (0.07) 1.18 (0.05)1995 1.07 (0.02) 1.17 (0.03)

Durum wheat

1990 1.48 (0.09) 1.52 (0.12)1991 1.38 (0.09) 1.47 (0.10)1992 1.45 (0.11) 1.27 (0.09)1993 1.67 (0.27) 1.56 (0.33)1994 1.38 (0.07) 1.42 (0.10)1995 1.41 (0.09) 1.50 (0.14)

! The values in parentheses are the standard errors.Source: Cropping practices survey } 1990}1995.

rules (following Granger and Newbold, 1977): (1) Thefuture cannot cause the past (strict causality can occuronly with the past (or present) causing the present orfuture); and (2) It is sensible to discuss causality only fora group of stochastic processes. (It is not possible todetect causality between two deterministic processes.)

Given these rules, the de"nition of causality used is interms of predictability. Speci"cally, one variable X is saidto cause another > with respect to a given universe orinformation set that includes X and> if present> can bebetter predicted by using past values of X than by notdoing so, all other information contained in the past of

the universe being used in either case. Thus, causalityruns from X to > if knowledge of X results in a smallererror variance in predicting > than would result froma prediction based solely on past observations of >.

In order to implement the de"nition of causality, as-sume that the information set consists of two variablesX and > and that there exist transformations

xt"¹

xX

t(1)

and

yt"¹

y>

t(2)

such that (xt, y

t) is a nonsingular linear covariance sta-

tionary purely nondeterministic time series where X and> are related causally in the same manner as x and y.Frequently, ¹

9and ¹

:will consist of "rst di!erences or

seasonal di!erences since it is often the case that this typeof transformation is both necessary and su$cient torender the observed series stationary. Since such trans-formations are linear and since the optimal predictors interms of the de"nition of causality used here are alsolinear, any causality event is true of (X,>) if and only if itis true of (x, y). Under these restrictions on x and y, it hasbeen shown (Harvey, 1981) that the bivariateprocess

Cxt

ytD

has the representation

Cxt

ytD"

=+j/0

(iC

at~j

bt~jD"((B)C

at

btD (3)

where ( is a sequence of 2]2 matrices;

Cat

btD

is a vector white noise sequence satisfying

ECat

btD"0 (4a)

ECat

btD (a

sbs)"G

& positive definite, t"s

0 tOs(4b)

Now the assumptions previously made imply thatxtand y

teach have representations as univariate linear

processes which can be written in autoregressive form as

CF(B)

0

0

G(B)D Cxt

ytD"C

ut

vtD (5)

Following Pierce and Haugh (1977), a joint model ofthe univariate residuals can be derived from this formwhich is given as

Ca(B)

c(B)

b(B)

d(B)DCut

vtD"C

at

btD (6)

N.D. Uri / Journal of Energy Policy 27 (1999) 299}306 301

Page 4: Energy and the use of conservation tillage in US agriculture

Fig. 2. The real price of crude oil in the United States

where all operators are one sided.7 Thus, for example,

a(B)"G=+j/0

ajBj

Finally, a0"d

0"1.

Taking b0"E(a

tbt)"0, it follows that causality fails

to exist if and only if cj"0, all j.

The second equation of relationship (6) can be writtenas

vt" +

j50

(!cj)u

t~j# +

j'0

(!dj) v

t~j#b

t(7)

Now, one can conclude that causality exists if thecurrent period observation of v

tis related to the observa-

tions of utin the current and/or previous periods. This

forms the basis of the statistical test to be employed. Thatis, it must be determined whether the c

jjointly are statist-

ically signi"cantly di!erent than zero. That is, the nullhypothesis is H

0: c

0"c

1"c

2"2"0. To implement

the test empirically, the data series must be suitablytransformed (ie as in relationship (5)) and the residualwhite noise series used in a regression analogous torelationship (7). The presence of statistically signi"cant

7Note that the series under consideration, x and y, are univariate.The discussion technically could be broadened to a multivariate analy-sis although most discussions of causality issues have focused onpairwise comparisons (Geweke, 1984).

estimates of the coe$cients cjwill lead one to conclude

that causality exists (Mills, 1990).

Empirical results

To assess whether there is an identi"able relationshipbetween the use of conservation tillage and the price ofenergy, data covering the period 1963}1997 were used.The untransformed crude oil price data, which is used asa proxy for the price of energy8 and which represent thecomposite re"ner acquisition cost, were taken fromCouncil of Economic Advisors (1979) and the EnergyInformation Administration (1997). The price data arede#ated by the gross domestic product implicit pricede#ator. These data were obtained from the Bureau ofEconomic Analysis of the US Department of Commerce.The real crude oil price data are presented in Figure 2.

8Actually, the concern is with the impact of the prices of re"nedpetroleum products, which are factors of production used in producingvarious agricultural commodities. But since there is a high degree ofcollinearity between the price of crude oil and the prices paid by farmersfor diesel fuel, gasoline, and lique"ed petroleum gas (simple correlationsare 0.93, 0.97, and 0.96 between the price of crude oil and the prices ofdiesel fuel, gasoline, and lique"ed petroleum gas, respectively, over theperiod 1970}1997), and since reliable data on the prices of re"nedpetroleum products paid by farmers do not go back to 1963, the crudeoil price (re"ner acquisition cost) is used as a suitable proxy.

302 N.D. Uri / Journal of Energy Policy 27 (1999) 299}306

Page 5: Energy and the use of conservation tillage in US agriculture

The conservation tillage use data were obtained fromSchertz (1988) and the Conservation Technology In-formation Center [annual] (Figure 1). Given the focus ofthe analysis is on the impact of the price of energy on theuse of conservation tillage practices relative to the use ofconventional tillage practices, the percentage of totalplanted acreage on which conservation tillage is prac-ticed is used as the dependent variable in the analysis.

To implement the test for causality, the "lters F (B) andG(B) of relationship (5) must be estimated. The "lterestimates, in turn, can be used to estimate the vector(u

t, v

t). Subsequently, v

tis regressed on lagged values of

v and current period and lagged values of u.Table 2 reports the "lters estimated by the approach of

Box and Jenkins (1971) for both the percent of plantedacres on which conservation tillage is used and the realprice of crude oil. The "lter selection is based on choosingthe speci"cation that maximized the Bayes InformationCriterion (BIC) (Sawa, 1978; Priestley, 1981). Up to eightautoregressive parameters and eight moving averageparameters were considered. The reported Ljung-Boxmodi"ed Q-statistics (Ljung and Box, 1978), based on 32degrees of freedom, suggest that the residuals for eachseries were reduced to white noise at the 5% level.

The vector (ut, v

t) (from relationship (5)) is obtained by

comparing "tted and actual values from the "lters withconventional backcasts used to obtain initial periodvalues. In implementing the test, the value of j for thelagged-dependent variable was set at 12 while the valueof j for the explanatory variable was set at eight. Longerlag lengths were also considered, but the conclusions arenot di!erent than those reported. The truncation of thelag polynomial for the explanatory variable at less than12 was done because it has been suggested that, in orderto maintain the power of the causality test, the length ofthe lag on the explanatory variable should be kept lessthan the length of the lag on the dependent variable(Geweke et al, 1983).

Table 2Time series "lters

Series Estimated "lters!,",# Q$

Percent of planted (1#0.3632 B) vt

12.74acres on which (0.1672)conservation tillageis usedReal price of (1!0.3395 B) u

t12.84

crude oil (0.1700)

! Standard errors of the estimates in parentheses." Note that B is general operator notation. That is Biz

t"z

t~i.

# The utand v

tare univariate residuals possessing the properties de"ned

in the text.$ The Ljung}Box modi"ed Q-statistic. The computed statistics are

based on 32 degrees of freedom. The critical value at the 5% level is46.19.

The test for the presence of unidirectional causalityrunning from the real price of crude oil to the percent oftotal planted acres on which conservation tillage is usedis a F-test comparing the unrestricted speci"cation whichcontains both lagged values of the dependent variable(relative use of conservation tillage) and current periodand lagged values of the explanatory variable (the realprice of crude oil) to the restricted speci"cation whichcontains only lagged values of the dependent variable.The computed F (12,28) is 10.83. The critical value at the5% level is 2.48. Thus, the results indicate unidirectionalcausality running from the price of crude oil to thepercent of total planted acres on which conservationtillage is used during the period. Moreover, these resultsare fairly robust when di!erent lag lengths are con-sidered.9 (Details are available upon request.)

Quantifying the e4ect of the price of energy on the use ofconservation tillage

Preliminary analyses

Since the price of energy clearly does a!ect the use ofconservation tillage in the United States, the next issuedeals with the nature of this e!ect. To address this,conventional empirical techniques will be used "rst toidentify any lags in the impacts and any anomalies in thedata and/or estimates and subsequently, its empiricalcharacter.

The functional speci"cation considered relates the useof conservation tillage and the real price of crude oil inthe current and previous periods. The lengths of the lagson the explanatory variables are determined by a zerorestrictions test (Judge et al, 1985). The results of the test(details are available upon request) indicate that the realprice of crude oil in just the current period a!ects theextent to which conservation tillage is used in productionagriculture in the United States. Thus, the impact of thereal price of crude oil on conservation tillage use isimmediate with no detectable lag e!ects.

Before turning to estimating empirically the relation-ship between the use of conservation tillage and the realprice of crude oil, one additional item needs to be ad-dressed. It involves the presence of data outliers. It is not

9To make the analysis complete, a test for unidirectional causalitywas performed with the percent of planted acres on which conservationtillage is used made the explanatory variable and the real price of crudeoil made the dependent variable. Using the same lag con"guration aspreviously (ie the crude oil price as the dependent variable lagged 12periods and the relative use of conservation tillage as the explanatoryvariable lagged eight periods), no evidence of unidirectional causalityrunning from relative conservation tillage use to the real price of crudeoil was detected. The relevant computed test statistic is F (12,28)"1.51.

N.D. Uri / Journal of Energy Policy 27 (1999) 299}306 303

Page 6: Energy and the use of conservation tillage in US agriculture

uncommon in empirical work to "nd that the results arevery much in#uenced by a subset of the total observa-tions used in the estimation. As a check on the possibilitythat coe$cient estimates were inordinately in#uenced bysuch a subset, the preliminary estimates were subjected tothe regression diagnostics of Belsley et al, (1980). The factthat a subset of the data can have a disproportionatein#uence on the estimated parameters is of concern be-cause it is quite possible that coe$cient estimates in themodel are generated primarily by this subset of the datarather than by all of the data equally. Belsley, Kuh andWelsch identify four diagnostic techniques to help inisolating in#uential data points: RSTUDENT, HATDIAGONAL, COVRATIO, and DFFITS. Each of thesediagnostics is employed here.

Regression diagnostics were performed on the basicrelationship. The regression diagnostics } RSTUDENT,HAT DIAGONAL, COVRATIO, and DFFITS } in-dicated no outliers.10 There were no observations thatare beyond the cuto! points.

Empirical estimates

One way to characterize the production innovationcycle discussed above is by a logistic curve (Waterson,1984). Beginning with Griliches (1957), this functionalrepresentation has been used repeatedly with success. Itwill be relied upon here. One adjustment will be made,however, to the standard logistic speci"cation. There isno reason to a priori suppose that the upper boundasymptotic value is static. Whether it is, testable hypothe-sis. In the current instance, since it has previously beenshown that energy considerations a!ect the use of conser-vation tillage and, consistent with the preliminary em-pirical results presented above, this upper bound will bemade as a function of the price of energy in just thecurrent period.11

Consequently, the speci"c functional representationfor the adoption and di!usion of conservation tillage inproduction agriculture in the United States is given as

Xt"(c

1>

t)/(1#c

2exp(!c

3t))#w

t, (8)

where Xtdenotes the percent of total planted acres de-

voted to conservation tillage, >tdenotes the real price of

10An observation is determined to be an outlier if two or more of thefour regression diagnostics cuto! points are exceeded. A completediscussion and accompanying "gures of the regression diagnostics areavailable upon request.

11 In preliminary analyses, the precise nature of the impact of the realprice of crude oil was examined. Its e!ect on both the extent of use (asmeasured by the upper bound asymptote) and the rate of adoption wereinvestigated. These issues were studies sequentially. While the real priceof crude oil clearly a!ects the extent of use of conservation tillage in theUnited States as indicated by the empirical results reported, there is nostatistically signi"cant e!ect on the rate of adoption.

energy, wtdenotes a vector white noise sequence with

mean zero and "nite variance, t denotes the time period,and c

1, c

2, and c

3are coe$cients to be estimated.

To estimate equation (8), classical least squares with anadjustment for serial correlation might be employed.This is not feasible, however, because of the relationship'snonlinear nature. The appropriate technique is max-imum likelihood estimation (Judge et al, 1985) thatadjusts for serial correlation. The data used in the estima-tion are those previously discussed.

The estimation results are

xt"4.1914>

t/

(1.2152)

(1#357.2011 exp (!0.1927 t))(148.4903) (0.0309) (9)

with the logarithm of the likelihood function"!93.23.The estimate of the serial correlation coe$cient was

not statistically signi"cantly di!erent from zero at the95% level in preliminary analyses. Consequently, therewas no adjustment for serial correlation in the computingthe "nal estimates.

The results are interesting. While the impact ofa change in the real crude oil price is relatively small, it isstatistically signi"cant at the 95% level. Thus, a 10% risein the real price of crude oil will lead to a increase in thepercentage of total planted acres devoted to conservationtillage of 0.4%, all other things given.

Testing for structural stability

To complete the analysis, one "nal issue needs to beinvestigated. Namely, has the underlying structural rela-tionship between the use of conservation tillage and thereal price of crude oil changed over the estimation peri-od? That is, for example, in light of the energy crisesduring the decade of the 1970s, is the use of conservationtillage more (or perhaps less) responsive to the real priceof crude oil today than it was, say, prior to 1970? Aninvestigation of this is the subject of what follows.

To analyze this, the stability of the estimated relation-ship must be studied. Stability is de"ned here in thestatistical sense of the estimated coe$cients on the ex-planatory variables remaining constant over time.A method of determining whether a regression relation-ship is constant over a given time period has been de-veloped by Brown et al (1975). Essentially, this approachnecessitates the computation of one-period predictionresiduals, which are obtained by applying the regressionestimated with r!1 observations to predict the rth ob-servation using k explanatory variables (including theconstant). The method is based on a test statistic, S(r),which equals the ratio of the sum of squared residuals ofone period prediction from the k#1 period to the rthperiod to the sum of the squared residuals of one period

304 N.D. Uri / Journal of Energy Policy 27 (1999) 299}306

Page 7: Energy and the use of conservation tillage in US agriculture

prediction from the k#1 period to the ¹th period,where ¹ denotes the sample size. The null hypothesisthat the regression relationship is constant over timeimplies that the expected value of the test statistic S(r),E(S(r)), will lie along (in a statistical sense) its mean valueline. For a more complete description of this test, theinterested reader is referred to Harvey (1981).

The results12 suggest that, for the estimated equation,the underlying structural relationship did not signi"-cantly change over the period 1963}1997. There are noyears in which the test statistic ventures above or belowthe upper or lower bound, respectively, of the 95% con"-dence interval. That is, the relationship between the useof conservation tillage and the real price of crude oil isstable over the entire sample period. The implication ofthese results is transparent. Events over the past threeand one-half decades have left virtually unchanged theimpact that the real price of crude oil has had on therelative use of conservation tillage in the United States.One must be careful, however, in inferring that the use ofconservation tillage remained constant over time.Clearly, this is not so. The estimation results show thatthe real price of crude oil impacts the extent of use ofconservation tillage. The magnitude of this, however,over the period 1963}1997 did not vary.

Conclusion

The relationship between energy and the use of conser-vation tillage is of special importance in addressing con-cerns about the impact of agricultural production on theenvironment in the United States. It has been the subjectthis paper. After establishing that a relationship existsbetween the price of energy and the use of conservationtillage using Granger causality, the relationship is quanti-"ed. It is shown that while the real price of crude oil, theproxy used for the price of energy, does not a!ect the rateof adoption of conservation tillage, it does impact theextent to which it is used. Finally, there is no structuralinstability in the relationship between the relative use ofconservation tillage and the real price of crude oil overthe period 1963}1997.

In a policy context, the results of the foregoing analysisshow that the price of energy can be used to promote theuse of conservation tillage. By increasing the real price ofenergy via, for example, a tax, the extent to which conser-vation tillage is used will expand. On the negative side,however, is the fact that a fairly substantial increase mustbe made in order for any signi"cant change to occur inthe use of conservation tillage. Whether other available

12Because they provide few insights, complete numerical results arenot presented. They are available upon request.

policy options are more suited to increase the use ofconservation tillage is beyond the scope of the presentstudy (Uri, 1998).

References

Alchien, A., 1959. Costs and output. The Allocation of EconomicResources. Stanford University Press, Stanford.

Batte, M., 1993. Technology and its impact on American agriculture.Size, Structure, and the Changing Face of American Agriculture.Westview Press, Inc., Boulder, CO.

Batte, M., Forster, L., Bacon, K., 1993. Performance of AlternativeTillage Systems on Ohio Farms. Department of Agricultural Econ-omics and Rural Sociology, The Ohio State University, Columbus,OH.

Belsley, D., Kuh, E., Welsch, R., 1980. Regression Diagnostics. Wiley,New York.

Box, G.E.P., Jenkins, G.M., 1971. Time Series Analysis: Forecastingand Control. Holden Day, Inc., San Francisco.

Brown, R.L., Durbin, J., Evans, J., 1975. Techniques for testing theconstancy of regression relationships over time. Journal of theRoyal Statistical Society, 37, 149}163.

Carter, M., Kunelius, H., 1990. Adapting conservation tillage in cool,humid regions. Journal of Soil and Water Conservation, 45, 454}456.

Conservation Technology Information Center. (1996) National CropResidue Management Survey. West Lafayette, IN, annual.

Council of Economic Advisors. 1979. Economic Report of the Presi-dent. U.S. Government Printing O$ce, Washington.

Energy Information Administration, Monthly Energy Review, 1997.DOE/EIA-0035 (95/01), US Department of Energy, Washington,December.

Frye, W., 1984. Energy requirements in no-tillage. In: Phillips, R.E.,Phillips, S.H. (Eds.), No-Tillage Agriculture: Principles and Practi-ces. Van Nostrand Reinhold Company, New York, NY.

Frye, W., 1995. Energy use in conservation tillage. Farming for a BetterEnvironment. Soil and Water Conservation Society, Ankeny, IA.

Geweke, J., 1984. Inference and causality in economic time seriesmodels. In: Griliches, Z., Intriligator, M., (Eds.), Handbook ofEconometrics, North-Holland Publishing Company, Amsterdam.

Geweke, J., Meese, R., Dent, W., 1983. Comparing alternative tests ofcausality in temporal systems: Analytic results and experimentalevidence. Journal of Econometrics, 21, 161}194.

Granger, C., 1969. Investigating causal relations by econometric modelsand cross-spectral methods. Econometrica, 37, 424}438.

Granger, C., Newbold, P., 1977. Spurious regressions in econometrics.Journal of Econometrics, 2, 111}20.

Gray, R., Taylor, J., Brown, W., 1996. Economic factors contributing tothe adoption of reduced tillage technologies in Central Sas-katchewan. Journal of Plant Science, 37, 7}17.

Gri$th, D., Mannering, J., Richey, C., 1977. Energy requirements andareas of adaptation for eight tillage-planting systems for corn. InLockertz, W. (Ed.), Agriculture and Energy. Academic Press, NewYork, NY.

Griliches, Z., 1957. Hybrid corn: an exploration in the economics oftechnical change. Econometrica, 25, 501}522.

Harvey, A., 1981. The Econometric Analysis of Time Series. PhillipAllen, Ltd., Oxford.

Judge, G., Gri$ths, W., Hill, R.C., Lutkepohl, H., Lee, T.C., 1985. TheTheory and Practice of Econometrics, 2nd edn. Wiley, New York.

Kislev, Y., Schori-Barach, N., 1973. The process of an innovation cycle.American Journal of Agricultural Economics, 55, 28}37.

Ljung, G., Box, G., 1978. On a measure of lack of "t in time seriesmodels. Biometrika, 66, 297}304.

Mikesell, C., Williams, J., Long, J., 1988. Evaluation of net returndistributions from alternative tillage systems for grain sorghum and

N.D. Uri / Journal of Energy Policy 27 (1999) 299}306 305

Page 8: Energy and the use of conservation tillage in US agriculture

soybean rotations. North Central Journal of Agricultural Econ-omics, 10, 255}271.

Mills, T., 1990. Time Series Techniques for Economists, CambridgeUniversity Press, Cambridge.

Nowak, P., 1992. Why farmers adopt production technology. Journalof Soil and Water Conservation, 47, 14}16.

Pagoulatos, A., Debertin, D., Sjarkowi, F., 1989. Soil erosion, intertem-poral pro"t, and the soil conservation decision. Southern Journal ofAgricultural Economics, 21, 55}62.

Pierce, D.A., 1975. Forecasting in dynamic models with stochasticregressors. Journal of Econometrics, 3, 349}374.

Pierce, D.A., Haugh, L.D., 1977. Causality in temporal systems. Journalof Econometrics, 5, 265}293.

Pierce, D., Haugh, L., 1979. The characterization of instantaneouscausality. Journal of Econometrics, 7, 257}259.

Priestley, M.B., 1981. Specral Analysis and Time Series. AcademicPress, London.

Sawa, T., 1978. Information criteria for discriminating among alterna-tive models. Econometrica, 46, 1273}1291.

Schertz, D., 1988. Conservation tillage: an analysis of acreage projec-tions in the United States. Journal of Soil and Water Conservation,33, 256}258.

Torgerson, D., Duncan, J., Dargan, A., 1987. Energy and US Agricul-ture, US Department of Agriculture, Economic Research Service,Washington, DC.

US Congress, Public Law No. 96}294, 94 Stat. 611, 16 USC 590h(1980).

Uri, N., 1997. Conservation tillage and input use. Environmental Geol-ogy, 29, 188}200.

Uri, N., 1998. The role of public policy in the use of conservation tillagein US Agriculture. International Journal of Energy, Environmentand Economics, forthcoming.

Uri, N., Day, K., 1992. Energy e$ciency, technological change and thedieselization of American agriculture in the United States. Trans-portation Planning and Technology, 16, 221}231.

Waterson, M., 1984. Economic Theory of Industry. Cambridge Univer-sity Press, Cambridge.

Westra, J., Olson, K., 1997. Farmers' Decision Processes and theAdoption of Conservation Tillage. Department of Applied Econ-omics, University of Minnesota, Minneapolis.

Williams, J., Llewelyn, R., Goss, L., Long, J., 1989. Analysis of NetReturns to Conservation Tillage from Corn and Soybeans in North-east Kansas, Kansas Agricultural Experiment Station Bulletin,Kansas State University, Manhattan.

306 N.D. Uri / Journal of Energy Policy 27 (1999) 299}306