EXPLORING ALTERNATIVE MEASURES OF NET RENTS TO FARMLAND
THROUGH THE ECONOMETRIC CAPITALIZATION
FORMULA FOR FARMLAND PRICE
by
ZsiZsi Tiziana Rachman
A thesis submitted in partial fulfillment
of the requirements for the degree
of
Master of Science
in
Applied Economics
MONTANA STATE UNIVERSITY
Bozeman, Montana
March 1988
ii
APPROVAL
of a thesis submitted by
ZsiZsi Tiziana Rachman
This thesis has been read by each member of the thesis
committee and has been found to be satisfactory regarding
content, English usage,
format, citations, bibliographic
style, and consistency, and is ready for submission to the
College of Graduate Studies.
Date
Co-chairperson, Graduate Committee
Date
co-chairperson, Graduate Committee
Approved for the Major Department
Date
Head, Major Department
Approved for the College of Graduate Studies
Date
Graduate Dean
iii
STATEMENT OF PERMISSION TO USE
In
presenting
this
thesis in partial fulfillment of
the requirements for a
master's
University,
that
I
agree
degree
the
Library
available to borrowers under rules of
quotations from
perrr:ission,
this thesis
provided
that
at
Montana State
shall
make
the Library.
it
Brief
are allowable without special
accurate·
acknowledgement
of
source is made.
Permission
for
extensive
reproduction of this thesis
professor
when, in
or,
in
his
the opinion
material is
may
absence,
of either,
quotation
be
granted
from
by
or
my major
by the Dean of Libraries
the proposed
for scholarly purposes.
use of the
Any copying or use of
the material in this thesis for financial gain shall not be
allowed without my written permission.
Signature
Date
iv
ACKNOWLEDGEMENTS
I
would
like
committee members,
My thanks
to
express
Dr. Bruce
my
appreciation
Beattie and
my
Dr. John Marsh.
to Dr. Myles Watts for his interest and guidance
above and beyond the call of duty.
and gratitude
goes to
Special acknowledgement
my co-chairmen,
Dr. Jeffrey LaFrance for their guidance
for their
to
infinite patience
Dr. Oscar Burt and
and, most
of all,
throughout the course of this
thesis.
Thank you friends.
Finally, my deepest love and gratitude
Alhambra
and
Daria
Rachman,
along with sister Vivienne
to my parents,
for their monetary support,
and brother
for your unconditional love and support.
Rendy.
Thank you
v
TABLE OF CONTENTS
APPROVAL . . . . . . . . . . . . .
~
P":<fe
................................... ll
STATEMENT OF PERMISSION TO USE ••••••••••••••••••••••••• i i i
ACKNOWLEDGEMENTS •••••••••••••••••••••••••••••••••••••.•• i v
TABLE OF CONTENTS ••••••••••••••••••••••••••.•.••.•••••••. v
LIST OF TABLES •••••••••••••••••••••••••••••.•••••.•••••• vi
ABSTRACT ••••••••••••••.•••••••••••••••••..•••••••••••. v i i i
CHAPTER
1.
INTRODUCTION ••••••••••••••••••••••••••••••••••• 1
Statement of the Problem •••••••••••••••••••• 2
Methode logy ••••••• ·• •••••••••••••••••••••••.• 3
2.
LITERATURE REVIEW ••••••••••••••..•••••••••••.•. 6
3•
MODEL DEVELOPMENT ••••••••••••••••••••••••••••• 2 0
Distributed Lags ••••••••••••••••••••••••••• 20
Farmland Price Model ••••••••••••••••••••••• 25
Data Compilation ••••••••••••••••••••••••••• 33
4.
EMPIRICAL RESULTS ••••••••••••••••••••••.•••••• 3 6
USDA Accounting Data Measure •.••••••••.•••• 36
Cash Rent Measure ••••••••.•••.•••••.••..•.• 46
Gross Revenue Measure ••••••••••••••••••••.• 57
5.
SUMMARY AND CONCLUSIONS •••••••••••••••.••••.•• 8 5
REFERENCES CITED ••••••••••••••••••••••.•••.•••••••••.... 91
APPENDIX ••••••••••••••••••••••••••••••.•••••.•••.••••••• 96
Original Data Set ••••••••••••••••.•••••••••.••.•.•• 97
vi
LIST OF TABLES
Table
Page
1
Regression results for final U.S. land price
model using USDA accounting data (sample period
1942-83) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2
Regression results for land prices regressed
on cash rents for Illinois (sample period
1961-83) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3
Regression results for cash rents regressed
on land prices for Illinois (sample period
1961-83) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ...... 56
4
Initial regression results for Illinois land
prices using gross revenue from corn and
soybeans (sample period 1961-83) . . . . . . . . . . . . . . . . . . . 61
5
Regression results for final Illinois land
price model using gross revenue from corn and
soybeans (sample period 1960-83) . . . . . . . . . . . . . . . . . . . 66
6
Regression results for final land price model
for Iowa, Indiana, and Ohio using gross revenue
from corn and soybeans (sample period 1960-83) ..... 69
7
Distributed lag land price elasticities (Y(T)) ..... 73
8
Cumulative distributed lag land price
elasticities (e(T)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
9
Post-sample forecasts of final estimated
land price model (classic disturbance) for
Illinois (Burt's land price data) .................. 76
10
·Post-sample forecasts of final estimated
land price model (classic disturbance) for
Illinois (land price index data) ................... 77
11
Post-sample forecasts of final estimated
land price model (classic disturbance) for
I ow a • . . • . • • . . . . . . . • • . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 8
vii
12
Post-sample forecasts of final estimated
land price model (classic disturbance) for
Indiana . ........................ ·· . . . . . . . . . . . . . . . . . . . ~19
13
Post-sample forecasts of final estimated
land price model (AR(1) disturbance) for
Indiana . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 0
14
Post-sample forecasts of final estimated
land price model (classic disturbance) for
Ohio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
15
Post-sample forecasts of final estimated
land price model (AR(1) disturbance) for Ohio ...... 82
16
Sample data used for exploring USDA accounting
data measure ..................... • ................. 9 7
17
Sample data for Illinois used for exploring
cash rent and gross revenue measures ............... 99
18
Sample data for Iowa .............................. lOl
19
Sample da·ta for Indiana ........................... 102
20
Sample data for Ohio .............................. 103
viii
ABSTRACT
Farmland prices began to diverge from farm income
trends during the mid-1950's.
Traditional capitalization
theory is the accepted mechanism with which to value
farmland.
The central idea to this theory is that land
values must derive from net rents to land. The divergence
between the time path of farm income and land prices has
focused
economists'
attention
on
the
validity of
traditional capitalization theory in explaining farmland
prices, and the appropriateness of net farm income as a
measure of net returns imputed to farmland.
This study explores three alternative measures of
returns imputed to farmland.
The measures are the United
States Department of Agriculture (USDA) accounting data,
cash rents, and gross revenue from corn and soybeans
production.
Each of
these measures
is fitted to
essentially
a
second order non-stochastic difference
equation framework; an economic capitalization formula for
farmland price that is developed in a different study and
is tested empirically using Illinois crop-share rent data
as the measure of net rents to farmland.
Problems are encountered concerning two of the three
measures explored; data problems for the USDA accounting
data measure and a joint dependency problem with farmland
prices for the cash rent measure. Encouraging results are
encountered when gross revenue from corn and soybeans is
explored. The regression results for this measure are
similar to the results of the study that uses cropshare
rents. An implicit capitalization rate cannot be computed
from the estimated land price model~ But it appears that
the estimated model can
be
used
for conditionally
forecasting short to intermediate time periods.
Additional research on exploring other alternative
measures or methods of imputing a net return to farmland is
suggested, but the need for a good set of farm accounts
data seems to be more pressing.
1
CHAPTER 1·
INTRODUCTION
In
1960,
Scofield
farmland prices were
trends.
(Doll,
sharply
et.al.,
diverging
income,
thus
coining
an
the
accompanying
term
Historically, farmland prices are
farm
income,
thus
farm income
must
Divergence between
rise
paradox."
quite closely
linked to
.Ricardo's
derive
from
classical
net
rent to
income-price trends occurred in
the mid-1950's, and has widened more after that.
income has
in farm
"land-price
supporting
argument that land values
land.
from
He concluded that it was paradoxical for farmland
prices to increase without
net
1983) noted that
Real farm
consistently declined since 1973, with farmland
prices continuing to increase,
then
dropping
slightly in
1981 and the subsequent two years.
Rising
farmland
prices
are
elation; landowners have benefitted
their wealth
position from
viewed
from
the rise.
by
the
Some
some
with
increase in
view it with
concern; those that borrow to buy land have to pay a larger
downpayment,
and
service
the
decreasing funds from farm income.
makes it
rest
of
the
This cash
debt
with
flow problem
difficult for new farmers to get started, for the
tenant farmer to become
an owner-operator,
or for smaller
2
farms to finance additional land.
discuss the
cash
flow
prqblem
temporal disassociation
Robison and Blake (1980)
which
emanates
from the
of generated returns and the usual
nominally amortized flat debt payment schedule.
may be
confused with
Cash flow
the returns resulting in a perceived
discrepancy between returns and farmland prices.
Many believe that land prices have reached unrealistic
levels before
are
the drop
overvalued
conclusion is
in
in the early 1980's, and that they
relation
to
reached through
farm
the use
capitalization formula, and the use
of
earnings.
This
of the traditional
farm
income
as a
measure of returns to farmland.
Statement of the Problem
The "unrealistic"
economists' attention
capitalization
This, in
each
on the
theory
farmland prices.
farm income
levels of
as·a
measure of
to
to a
the
levels of
traditional
imputing
inappropriate
returns
to
and
used
to
explain
appropriateness of net
returns imputed to farmland.
come
new research studies,
up with new approaches to
land.
Some hypothesize
capitalization
simplistic and have modified
income
mechanism
host of
have
explain the high price
that
a
applicability of traditional
Some question the
turn, leads
purporting
as
land prices have focused
it,
have
farmland,
some
formula
consider
suggested
while
is
too
net farm
other ways for
others
offer
new
3
imputing
returns
theories that
farmland,
to
replace
while
traditional
others
offer
new
capitalization theory.
Very few of these new models are tested empirically.
The purpose
that
models
of this
the
traditional
behavior
of
capitalization
is formulated
by Burt
using
data
annual
results.
study is to look at one approach
theory
correctly to
will solely
determine
Burt's land
within a
This model
is tested empirically
for Illinois, and produces encouraging
are imputed
on the
farmland
study
is
price model.
idea that
farmland, then
capitalization
objective of this
prices
framework.
(1986), then
The model is based
traditional
farmland
price
net rent to land
stated
as
Thus
formula.
to
if returns
further
in the
the
main
test empirically
Since the measure of net rent to
land used in Burt's study is
available only
an
rent that may be derived from
alternative
measure
of
for Illinois,
more commonly available data bases is needed.
Methodology
This study explores
estimating
net
returns
several
to
alternative
farmland.
farmland prices through use of the
The measure
of net
returns to
series published by Scott (1983)
methods for
Burt (1986) models
capitalization formula.
farmland that is used is a
and
is
defined
as "net
returns in dollars to landowners on high quality crop-share
grain
farms
in
Illinois.''
This
crop-share
rent data
4
originates
from
farm
accounting data associated with the
farm records program at the University of Illinois, and the
uniform application
of accounting
in collecting the data
type of
throughout
in
Burt's
period
is
easily
crop-share rent
analysis,
alternative measure of net
that
the
makes this
data unique to the state of Illinois (Burt, 1986).
Due to the uniqueness of the
used
procedures that is used
there
is
returns to
collected
in
any
data that is
a
need
for
an
farmland using data
state.
It will give
Burt's model greater plausibility if the results reached in
the
Illinois
study
can
be
reproduced with a measure of
rents that uses a more common data base.
Three alternative measures of net returns
are
explored
in
this
study.
Department of Agriculture (USDA)
to
impute
a
residual
return
First,
to farmland
United
accounting data
to farmland.
States
are used
Second, cash
rents are used as a proxy for net returns to farmland.
The
third alternative measure is gross revenue for the dominant
cash crops for the state.
These measures are then fitted to essentially the same
dynamic regression
The
results
model that is developed by Burt (1986).
of. estimation
are
compared
to
regression
results arrived at when crop-share rent data was used, thus
using crop-share rent as
is first
the reference
measure.
Analysis
done for Illinois, because of the opportunity for
direct comparison.
If encouraging results emanate from one
5
extended from
Corn Belt
Illinois to
region
and
other neighboring states in the
comparisons
made.
The estimated
equations in this last part of the study are then tested by
evaluating
their
predictive
relative
performance
via
prediction errors.
Comparisons
of
using alternative
estimation
precise
used.
net
results
performance,
and
results
rent
are
with
statistical
from the estimated equations
measures
made
no
with
results
when the
generous
aspirations
as
It is hoped that the same
would prevail
with .Burt's (1986)
criteria of
about
getting as
when crop-share data are
general dynamic structure
alternate measures
of net returns
are used, and that such a model would do reasonably well in
·conditionally forecasting within short t·o intermediate time
periods.
This
thesis
summarizes
major
is
organized
works
agricultural land prices.
price
model
section
and
data
on
theory
plus
some
thoughts
fulfillment of
in Chapter 5.
Chapter 2
follows.
are releva'nt and pertain to
that
Chapter 3 describes the farmland
of
the
compilation.
empirical results.
as
estimation
Chapter
used plus a
4
delineates
A summary of the findings of the study
on
their
the objectives
relevance
towards
the
of this study are presented
6
CHAPTER 2
LITERATURE REVIEW
Fisher, in his Theory of Interest
"the value
value as
that
of any
property, or
a source
expected
of income
income"
(p.
(1967), stated that
rights to
and is
12).
wealth, is its
found by discounting
This is the basis for
traditional capitalization theory.
This approach
roots in basic intertemporal choice analysis.
starts out
now
as
by assuming
opposed
future.
So
to
later,
people
preference.
that human
say
premium
one time period into the
basically
have
a
positive
that
r,
in
reflects
a period
to the
this
case,
is
the
extra
worth
into the
future.
denote
the
value
of
future, or
"goods"
The
a special price; a
of
"goods" now
its time value.
Let Vo denote the value of "goods" now (time period
V1
time
This positive time preference translates into
rate
instead of
The analysis
nature prefers "goods"
"goods" being worth more now relative
interest
has its
0) and
one time period later.
Then,
v 0 (1+r)
=v1
reflects this extra worth of Vo
(2.1)
over V1
which equals rv 0 •
Another way to look at equation (2.1) is:
vo
=
( 2. 2)
(1+r)
7
which transforms
future into
an
immediately.
the value
equivalent
of a
dollar one per:j..od in the
value
for
a
dollar receiVed
This is called discounting a future value to
the present, or said
another way,
the present
value of a
future worth.
According
Fisher's
to
concept
of
value,
the
traditional capitalization formula is represented as
00
Po=
I
t=1
( 2. 3 )
Rt/(1 + i 1 )(1 + i 2 ) ... (1 +it)
where
Po= price of farmland at time zero (current price);
Rt
= returns
it
occurring at the end of time period t;
discount rate at time t.
If returns and discount
perpetuity, i.e.,
•
•
• I
Rt
=
rates
are
Rand
it=
assumed
i for
constant into
all t
= 1,
2,
oo, the equation reduces to
00
Po
=
I
R/(1 + i)t
( 2. 4)
t=1
with a closed form solution of
Po
=
R
( 2. 5)
i
Equation
is what
land.
(2.~)
with current returns and discount rates
is commonly
Use
of
(2.3)
used to evaluate the current value of
requires
earnings and discount rates.
knowledge
of
all future
8
When
diverge
land
in
commonly
price
the
used
farm
and
evaluate
time people concluded that
relation to
equation
mid-1950's,
to
farm income.
income
farmland
land
was
was
value.
being
still
During this
overvalued in
However, equation (2.5)
assumption, the long-run equilibrium capitalization
formula under restrictive assumptions.
should
(2. 5)
began to
When (2.5) was used, land prices
were lower than the actual prices.
is, by
trends
be
used
to
farmland
associated
interest
rates,
estimate
with
and
the
Thus
this equation
equilibrium
constant
rents
value
and
of
constant
not short-term farmland prices where
economic variables are subject to.continuing perturbations.
Research in the 1960's and the early 1970's focused on
the apparent land price
net
farm
income
divergence.
did
not
fully
It was
claimed that
explain rising farmland
prices, and researchers sought other variables
Reynolds
and
Timmons
were factors other than
contributing
to
the
that might.
(1969) hypothesized that there
net
farm
change
in
income,
that
could be
farmland prices and thus
causing the divergence between the time paths of income and
prices.
Net farm income, though, was still believed to be
the major determinant of
considered
were
advancements,
pressure
from
land prices.
government
farm
farm
enlargement,
increasing
The
other factors
programs, technological
transfers
population,
of
farmland,
and capital gains.
These factors were modeled through a recursive cobweb model
9
(Ezekiel, 1938),
was
where ori
determined
ownership
by
the
transferred
exogenously
in
the
income, government
of return
current
plus
quantity
other
market
including
stock and
the current
of
variables
determined
net farm
capital gains, rate
farm enlargement.
On the
quantity of farmland transferred
was a function of expected capital gains, ratio
nonfarm earnings,
farmland
expected
payments, expected
on common
supply side,
the demand side, farmland price
a measure
of technology,
of farm to
ratio of farm
mortgage debt to equity, and change in the number of farms.
The model
was estimated
the years 1933-1965.
included
as
an
using two stage least squares for
A cross-sectional
alternative
analysis was also
approach
to
.time
series
analysis, although the authors acknowledged the coefficient
estimates are not directly comparable.
A similar
land prices
equation
land
in
and
by Klinefelter (1973), in which
were
Illinois
model,
prices
Reynolds'
study done
factors
involved
that
were
Timmons'
modeled
very
and expected
equation was
reduced only
average farm size, number of voluntary
and expected
capital gains
those
used in
expected
net
farm enlargements,
capital gains).
multicollinearity between variables in
the final
to
(inflation,
rents, government programs, technology,
farmland transfers,
a single
hypothesized to affect
similar
study
through
Due to
the proposed model,
to include net rent,
farmland transfers,
as explanatory variables.
Net
10
rent, in this study, was defined as the remaining amount of
the landlord's share of
(including
inventory
subtracted
out
gross farm
change)
(Reiss,
landowners based
after
1969).
the earning
a
three-year
moving
all of his costs are
It
was
potential of
average of recent past trends.
was
income from production
Thus the
average
assumed
that
farmland on an
net rent variable
of
net
rents for the
previous three years.
Pope
et
al,
credibility of
(1979)
the above
evaluated
two models
the
in explaining recent
data and tested their predictive ability,
and
Cochrane's
(1966)
simultaneous
Tweeten and Martin's (1966)
of
the
farmland
equation
When
these
estimated to include recent data, there
reversals and
lack of
et al,
1979) had
changes and statistical
was chosen
model
and
recursive model
models
were re-
were numerous sign
statistical significance of the re-
estimated equation coefficients.
model (Pope
along with Herdt
five equation
market.
structural
The modified Klinefelter
the least problem with sign
insignificance,
so
this equation
as the econometric model to be tested against a
Box-Jenkins time
series equation
relative predictive
performance.
of land
prices only for
Results were that a time
series model provided as good or better short run forecasts
than Klinefelter's modified econometric model.
Melichar
(1979)
provided
price paradox phenomenon when
some insight into the land
he
proposed
that comparing
11
the USDA
index of farm real estate value with operator net
farm income, the common measures
returns
to
land,
was
apples with oranges.
of
erroneous.
land
prices
It was like
and net
com~aring·
First, an aggregated return component
is compared to the unit price of a single asset (land) with
net farm income imputed to real
the other
estate alone
productive assets.
Second, operators' net farm
income, as traditionally computed
appropriate
measure
productive assets.
of
exclusive of
by the
return
USDA, is
not an
to land, or of returns to
To obtain a valid measure of returns to
farm production assets, Melichar suggested that net rent to
non-operator landlords
and interest
paid on
farm debt be
added to operators' net farm income.
Also, the part of net
income
is
imputed
productive
to
farm
only a return to
management
and
farm
dwellings
productive assets,
labor;
to
farm
faster rate
up
Melichar
but also
a return to
Melichar then compared
the
parallel
a
value.
productive
than
relatively
makes
of
path of this residual return to productive assets
with the time path of its
return
part
so these last two return
components must be subtracted out.
the time
a
This adjusted component is not
assets.
operator
not
major
concluded
value
assets
of
that
the
the
1950's the
had been growing at a
these
assets,
but with
And since farm real estate
trends.
portion
Since
of
farm
upward
productive
pressure
assets,
caused by
12
returns to
productive assets
on the value of these assets
should also be felt by farm real estate values.
In the second part
parallel
of
relationships
with the key
(returns).
his
found
assumption
of
paper,
in
Melichar examined
the asset-pricing model
constant
growth
in earnings
Through this analysis, he showed that the total
rate of return to productive assets is made up of
of
current
return
capital gains
to
on these
steady-state rate
rate.
productive assets plus the rate of
assets.
of capital
of current returns,
assumed
Melichar then
for
the
past
estate in
to
a
constant geometric
and
twenty-five
the
growth
years,
rate of
on the whole,
Because of the dominant position of
historical
data
of
capital gains
estate assets to growth rates of current returns
to productive
gains
be
the makeup of farm productive assets,
Melichar also compared
from real
showed that the
tested this last result empirically,
supported this result.
farm real
He also
gains equals the growth rate
and concluded that capital gains
returns
the rate
were,
assets, and
in
a
sense,
found that
fully
real estate capital
explained by the growth
exhibited by current returns to productive assets.
In summation, Melichar's analysis showed that as asset
earnings are expected to grow, then these expected earnings
will translate more into real capital gains rather
current returns.
He concluded that
than to
13
according to asset-pricing theory, a far~
economy characterized by rapid growth in the real
current return to assets will tend to experience
large annual real capital gains and a low rate of
current returns to assets--which corresponds to
actual experience in most years since the mid
1950's. (p. 1085)
Asset
returns
if
properly
explain asset values,
at
measured
least
would seem to fully
within
the
framework in
which Melichar chose to make his comparisons.
But there
Melichar's
were those that were not fully convinced by
explanation.
between general
"There
price inflation
land that deserves particular
(1980).
is
a
fundamental link
and the relative price of
attention," argued Feldstein
He explained that the combination of inflation and
the tax laws will raise the return to
the price
land, and eventually
of land will have to adjust itself upward during
inflationary times.
Feldstein incorporated the land market
·into his model from a speculative viewpoint.
derive
price
equations
capital (business
for
Feldstein then
results of
the effects
the
price
comparative
and
for
reproducible
capital) in an explicit portfolio-choice
framework.
and
land
He went on to
of
looked
at
of inflation
reproducible
on the price of land
capital.
analysis
statics
comparative statics
pointed
inappropriateness of assuming that the effect
Results
of
to
the
of inflation
on land prices is neutral (Feldstein, 1980).
The
study
farmland market
by
from
Castle
an
and
Hoch (1982) discussed the
investor's
point
of
view.
A
14
prospective
investor
starts
about future land prices.
current
price
then
out
by forming expectations
If expected price exceeds actual
he
will
want
to buy, but if actual
current price exceeds expected price then
sell or
at least
be discouraged
Hoch hypothesized that this
two components.
the earnings
earnings
value of
is made
into account
the "pure"
for
assets
measures the
"all the forces which
part
of
of infinite life.
effect that
cause real
second
component
inflation on land prices.
complicated description
farm
real
estate
inflation.
is determined by
measured
to change
A subsidiary
the
effect
of
Castle and Hoch offered a fairly
of
their
prices
capital gains
This latter
This
include another
estate prices
expectations
based
capitalization formula; two components
return were
the effect of
the general price level" (p. 8).
the
up of
return obtained from
concept was extended by Castle and Hoch to
relative to
Castle and
This is the familiar capitalized value
concept
component that
want to
The first component in land price, labeled
real estate assets.
of
from buying.
expected price
component, takes
the capitalized
he will
on
the
model for
traditional
in addition
to net
and farm real estate debt under
component
emanates
from
the lag
between interest rates and inflation rates so that existing
long-term farm debt has value because
to be
of
the
less than market interest rates.
three
components
were
the loan
rates tend
Equations for each
developed
and
tested
15
empirically.
Predicted
values
for
components were added up to produce
land.
each
of
the value
a predicted
value for
Comparisons between actual and predicted land prices
were made.
Castle and Hoch
the empirical
concluded that
testing show
net return to real
that the
estate assets
the results of
capitalized value of
used in
farm production
explains only about one half of real estate values.
Shalit and Schmitz (1982) offered a "the other side of
the coin" approach
paper focused
through
to
farmland
production,
speculative "motives"
as did
paper first derived an
for
these individual
Their
did
than
not
consider
Castle and Hoch (1982).
land
The
derived demand
through lifetime utility
profit
maximation.
Then from
derived demands, the authors developed an
aggregate derived
of
and
individual farmer's
agricultural
maximization rather
authors note
analysis.
on the derived demand for farmland generated
agricultural
function
market
demand
model
for
farmland,
which the
was just a more general and dynamic extension
traditional
capitalization
theory.
Lastly,
aggregate model was empirically tested for the
1950-78 period.
The net return
used in
u.s.
the
for the
their model
is a
series compiled by Hottel and Evans (1979), which "consists
of
residual
income
deducting
imputed
dwellings
from
the
to
real
returns
estate
to
operator's
equity
labor,
total
(Shalit and Schmitz, 1982, p. 717).
obtained by
management,
net
farm
and
income"
16
The analysis resulted in Shalit.and Schmitz concluding
that "the price of farmland is
determined not
only by the
profit it generates (agricultural income and capital gains)
but also by the debt it can carry" (p. 718), and also that,
"the
expansion
and
contraction
of
credit
importantly
affects the pace at which land prices increase or decrease"
(p. 718).
conventional
Melichar's
growth
in
land
returns to
prices
land has
solely
a plausible
I
I
1986).
Feldstein's
papers
a
Alston mentioned that both
tested
with
results
However, rapid
with rapid
to model
the effect of
Alston (1986) combined the two
generated
into
growth in
Feldstein was able to formulate
land prices.
hypotheses
explaining
expected
been associated
theoretical mechanism
inflation on
competing
from
of
been widely accepted.
growth in land prices has also
inflation (Alston,
hypothesis
simple
from
Melichar's
model of land prices.
hypotheses had
that
and
been previously
reject inflation as a cause of
land price movements and favor Melichar's more conventional
explanation.
A
price
equation
returns and rate of
for
land
inflation was
which
incorporated net
initially derived using
the traditional capitalization approach.
Then a regression
equation that incorporated expected inflation as a separate
independent variable
was derived.
eight midwestern states, and
Analysis
cash rent
was done for
data were
used to
17
measure
the
regression
net
return
equation
variable.
for
the
Estimation
eight
states
of
the
gave
a
statistically significant negative net effect for inflation
(contrary
to
Feldstein's
hypothesis)
on
land
prices,
although empirically small.
There seemed
taken by researchers
farmland
major differences in the approach
to be
explaining
in
prices.
addition
In
the
recent
trend of
to this, there is yet no
agreement on the exact definition of net returns imputed to
farmland.
Alston (1986)
and Castle
and Hoch (1982) used
USDA accounting rent data
(plus some
adjustments) for the
return variable
in their respective models.
time series plots of cash
state)
indicate
that
rents
there
seemed to lag land prices.
in
their
adjustment
process that
needed
to
must
account
production costs.
factor
in
All of
this could
years
because
experienced
for
changed
prices (by
when cash rents
are somewhat sticky
of
the renegotiation
whenever
a
commodity
change is
prices
and
Recently past land prices could become a
determining
amount, thus making
were
farmland
Cash rents
path
be
and
Comparison of
cash
the
new
rent
renegotiated
a
cash
rent
non-exogenous variable.
add up to the possibility of cash rents
being jointly determined with
land
prices;
the exogenous
variable(s) that effect land prices may also determine cash
rents as well.
18
Some agreed that operator's net farm income is a valid
starting point from which
farmland
(Melichar,
to impute
Phipps,
Melichar's adjustments to
a residual
Shalit
operator's
net
return to
and
Schmitz).
farm
income to
produce a measure of return to total farm productive assets
was fairly straightf9rward.
part of
But to impute a
return to one
total farm productive assets, namely farmland, can
become much more difficult and much more arbitrary.
For example, Phipps (1984) used a
the
one
Melichar
farmland.
to
developed
estimate a net return to
First, Phipps subtracted out "an implicit return
non-land
durable
productive
operator's net farm income.
of
to
depreciation
on
assets"
He used the
non-land
durable
opportunity cost of investment as the
to
the
non-land
opportunity cost
85%
of
the
subtracted
dwelling
similar approach to
component.
of a
non-farm
returns
from
imputed
opera.tor' s
427)
from
USDA's definition
assets
measur~
plus
the
of the return
He also subtracted out the
farm operator's
wage
(p.
rate.
to
net
time, evaluated at
Shalit
and
management,
Schmitz
labor,
and
farm income to arrive at a
return to farm real estate (1982).
Furthermore,
capitalization
recent
theory
as
literature
the
springboard
develop a more involved model for
models
usually
involved
component, making them
traditional
to launch and
farmland prices.
variables
arduous
used
to
These
other that the return
follow
and comprehend
19
(Castle and Hoch, Shalit and Schmitz).
There is a need for
a much simpler model for farmland prices, because according
to traditional
capitalization theory,
returns to an asset
should fully justify its value.
Burt
(1986)
modeled
regression framework
farmland
that is
prices
tied back
traditional capitalization formula.
in
a dynamic
eventually to the
The model
was tested
empirically using annual Illinois data, producing empirical
results that
devoted to
were encouraging.
a more
The
in-depth discussion
following chapter is
of Burt's farmland
price model, as the model is central to this study.
20
CHAPTER 3
MODEL DEVELOPMENT
This
chapter
distributed lag
of
the
price
presents
a
theory before
econometric
developed
short
Burt
on
summarizing the formulation
capitalization
by
discussion
formula
(1986).
for farmland
Lastly,
a
section
discussing sources of data compilation is included.
Distributed Lags
Oftentimes an economic phenomenon is best described by
a dynamic
another,
system consisting
where
more
than
measure the full effect a
comparison,
a
static
this study, for
one
has
consists
one
the
that affect one
period is needed to
the effect(s}
(during
example,
time
variable
system
affect one another where
contemporaneously
of variables
on
another.
of variables that
are fully realized
time period}.
time
In
Relating to
adjustment
between a
change in rents and its effect on land prices may not occur
instantaneous~y
market.
because
A dynamic
lag structure.
depending on
of
system is
uncertainties
the
land
modeled using a distributed
This structure could be finite
whether one
in
variable vlill
or infinite
have the "ripple"
21
effect on another variable over a finite or infinite length
of time.
A finite distributed lag model would have the form
~t
Yt =a+ ooXt + o1 xt_ 1 + ... + okXt-k +
where Xt is an
lag
effect
exogenous variable
on
Yt
~t
and
that has
(3.1)
a distributed
is an unobservable error term.
Equation (3.1) indicates that the order of lag coefficients
higher
than
ok
are
assumed
independent variable X does
to
not
be
zero,
so
that the
affect
Y
beyond
k time
periods.
Examples
of
finite
distributed
arithmetic lag, inverted V-lag,
A basic
disadvantage to
the
distributed
Since there is
knowledge
little
of
the
structures are
and Almon
polynomial lag.
using the finite lag structure is
in deciding what lag k should be, thus
when
lag
lag
a priori specifying
effect will be fully realized.
theoretical
industry
basis
and/or specific
to identify the length of the
distributed lag period, specifying k becomes
further problem
with the
finite distributed
due to the fact that economic time series
to change.
the lagged
This
arbitrary.
A
lag model is
are usually slow
often results in multicollinearity among
independent
problems (Theil, 1978).
variables,
.leading
to estimation
Precision in the estimation of the
distributed lag coefficients is lost due to the increase in
the standard
errors of the coefficient estimates.
problem that may arise
is the
loss of
Another
degrees of freedom
22
due
to
the
number
of
parameters
freedom equals total number
of
total number
to be
of parameters
estimated (degrees of
sample
data
points minus
estimated including the
intercept) .
Assuming·
necessarily
an
lag
infinite
alleviate
having
structure
to
include
geometric
the
Jorgensen's (1966) rational lag.
assumes a
lag structure
structure.
not
arbitrarily assign k in
equation (3.1) as it arbitrarily sets k = +oo.
structures
will
lag,
A
that takes
Infinite lag
Pascal,
geometric
and
lag model
on a geometric series
This lag structure is specified as
Yt =a+ ~Xt + ~AXt-1 + ~A 2 Xt-2 + ... + Pt
(3.2)
where o' s are assigned weights equal to oj=~>),
j=O, .1 1 2 1
•
•
•
I
A lies in the interval of 0 < A < 1.
and
only three parameters (a,
the infinite
~
1 and A)
lag structure.
unestimable due
to an
are
Notice that
needed to describe
The equation in this form is
infinite series
required on
X.
A
Koyck transformation (Theil, 1978) on (3.2) will produce
Yt = a(1-A) +
which
is
~Xt
estimable
because
The difference equation
path of
adjustment of
of the finite lag structure.
parameter
A determines
the dependent
change in the independent variable X.
indicate a
(3. 3)
+ AYt-1 + Pt - APt-1
slower time
variable Y
Higher
the time
due to a
values of A
rate towards the adjustment, while
lower values of A indicate a faster time rate of adjustment
in Y.
Note
also that
due to Koyck's transformation, the
23
error term
in (3.2) has an additional moving average error
(MA) component in (3.3).
A lag
structure with
geometrically declining weights
may not
be appropriate for describing the dynamics of land
prices.
The distributed
as having
a "peak"
lag effect
could be hypothesized
(inverted V-lag
structure) or several
"peaks" at certain lag periods similar
polynomial function
type of
would take.
distributed
lag
Yt
= W(L)Xt
with the
+
path that a
The general form of this
structure
developed by Jorgensen (1966).
to the
is
the
rational lag
This model is specified as
~t
(3.4)
rational generating function W(L) defined as W(L)
= B(L)/A(L), where Lis the lag operator
such that
Ljxt =
Xt-j· Jorgensen defines B(L) and A(L) as
B(L)
A(L)
where
= ~0 + ~lL + ~2L2 + .• ; + ~mLm
= 1 - AlL~ A2L2- ... - AnLn
is
usually
normalized
identification purposes. B(L) is
lag on
the independent
an
variable X,
to
mth
equal
one
for
order polynomial
while A(L)
is an nth
order polynomial lag structure on the dependent variable Y.
A(L) also
determines the
order of the difference equation
and correlation structure of
the error
term.
Multiplying
equation (3.4) through by A(L) yields
( 3 • 5)
24
Equation (3.5)
is an nth order difference equation with an
nth order mov1ng average
disturbance term
if l-It
is white
noise.
The roots of a difference equation determine the shape
of the distributed lag pattern through time.
lag allows
for real
and/or complex
the polynomial A(L) is
difference equation,
at
least
for example,
The rational
roots if the order of
2.
In
a second-order
real roots
can imply a
unimodal lag structure that dampens off
after reaching the
"peak" lag,
for an oscillatory
while complex
roots allow
pattern in the distributed lag structure.
Jorgensen (1966) showed that
can approximate
any arbitrary
the
rational
lag structure
lag model
which has an
asymptote of zero.
A Pascal distributed lag
the rational
lag in
positive and equal.
order zero
that the
With
while A(L)
the
model is
a special
case of
roots are constrained to be
Pascal
model,
B(L)
is of
is of order r with equal roots such
that the equation looks like
(3. 6)
The. geometric
when r=l,
lag model
thus making
the rational lag model.
is a
special case
of the Pascal
the geometric lag a special case of
25
Farmland Price Model
(1986)
Burt
quantity
of
determines
begins
farmland
price"
by
pointing
out
that
"with
fixed, the demand equation entirely
(p.
11).
With
the
assumptions
of
competition among buyers (potential and realized alike) and
a world of certainty,
use
of
the
farmland price
traditional
is explained through
capitalization
formula.
So the
basic model for farmland prices would be
00
Po=
I
t=l
Rt/[(1 + r 1 )(1 + r 2 ) ... (1 + rt)]
(3 •7 )
where
P0 =price of land in year 0;
Rt = net returns to land in year t;
rt = real discount rate in year t.
Returns and prices are thought of
of the
year.
as occurring
at the end
If discount rates are assumed constant, then
(3.7) reduces to
00
Po =
I
Rt/(1 + r)t.
( 3-. 8)
t=l
Letting net
constant into
returns
and
the
rate
of
capitalization be
perpetuity produces the long run equilibrium
·land price equation
( 3 • 9)
where
26
= equilibrium price of land;
R* = equilibrium {fixed) annual net returns to land;
a = 1/r = reciprocal of the real cap~talization rate.
P*
The dynamic
regression equation
for farmland
developed is constrained to .have an
price to be
equilibrium structure
of {3.9), with a an unknown parameter.
The
dynamic
regression
equation is specified with a
multiplicative distributed lag on
likely
that
land
changes rather
Random shocks
it is more
market participants consider percentage
than
absolute
in the
changes
in
this variable.
economy emanating from discrepancies
of the constant capitalization
and measurement
rents, since.
rate implicit
in the model
errors on land prices would also impact in
a proportional, rather than in an additive way.
assumptions, the
dynamic regression
From these
equation for farmland
prices is specified as
( 3. 10)
... ) llt
where llt is a random
unknown parameters;
geneous of degree one
disturbance
and (3.10)
(~o
+
~1
term;
~0,
is considered
+ •.•
= 1).
~1,
... are
to be homo-
Equation (3.10)
after a natural logarithmic transformation becomes
00
log Pt
where log
= log
a +
.E
~jlog
Rt-j + log llt
{3.11)
J=O
llt is assumed to follow an autoregressive-moving
average {ARMA) process of unknown order and E(log llt)
= 0.
27
The form of equation (3.11) cannot be estimated when a
finite
data
set
is
used
because
of
its
infinite lag
structure (thus an infinite number of unknown parameters to
be
estimated),
technique
so
an
chosen
approximation
a
is
second-order
approximation of the general
lag in
was proposed
Burt
by Jorgensen.
a
lag as
"
second
order
is
parsimonious and
specification
needed.
The
rational
lag
equation (3.11) which
describes this rational
flexible form
in which a
is sufficiently flexible" (p.
13), which allows for relatively few
unknown parameters to
be estimated.
The number of potentially estimable unknown
parameters
crucial
is
when· working
with
annual
data.
Therefore, (3.11) is approximated by
log Pt
where
L
= log
is
a
a.
+
ho + YlL) log Rt
( 1 - AlL - A2L2)
lag
operator
such
+ log J..lt
that
Multiplying both sides of ( 3.12) by ( 1 - AlL
(1 - AlL ~ A2 L2)log Pt
=
LjXt
-
(3.12)
=
Xt-j·
A2L2) yields
(1 - .AlL - A2L2)log a.
+ Yolog Rt + Yllog Rt-1 + (1 - AlL - A2L2)log J..lt (3.13)
or equivalently
log Pt
=
(1 - Al - A2)log a. + Yolog Rt
+ Yllog
Rt-1 +
AllogPt-1 + A2log Pt-2 + log J..lt - A1log J..lt-1 A2log J..lt-2
Grouping terms produces
(3.14)
28
=
log Pt
(1 - A1 - Ai)log a + YQlog Rt + y1log
Al(log Pt-1 - log ~t-1> +
log
A2(log Pt-2
- log
~t
Equation
Rt~1
+
~t- 2 >
+
(3.15)
(3.15)
can
be
transformed
into a second-order
difference equation in expected values,
log Pt
= o + y0 log
Rt + Y1log Rt-1
A2E(1og Pt_ 2 ) + log
Pt-1) +
~t
where E(log Pt-j>
=
- A2)log
Equation
a.
+ A1 E(log
(3.16)
log Pt-j - log
~t-j
and
is
a
(3.16)
=
o
(1 - A1
second-order
nonstochastic difference equation (Burt, 1980).
If at
land prices
P*)
=
equilibrium, log
will approach
E(log Pt)
= E(log
Rt
=
log Rt-1
a steady-state
Pt-1>
= E(log
=
log R*, then
such that E(log
Pt-2>·
Transposing
the disturbance term in (3.16), we can write
. E(log Pt)
=o+
YQlog Rt + Y1log Rt-1
+ A1 E(log Pt-1)
(3.17)
+ A2 E(log Pt-2)
For the equilibrium state, like terms can be grouped to get
E(log P*)
= log
a+
(y 0 + y 1 )log R*
( 3. 18)
(1 - A1 - A2)
The homogeneity
constraint,
~0
+
~1
+ ...
= 1,
which was
imposed initially can be combined with (3.18) to produce
( 3.19)
Notice that if (3.19) is true, then (3.18) is the result of
a
natural
logarithmic
transformation
of
(3.9).
noted that E(log P*) is interpreted as log P*.
It was
29
The approach taken in
model is
specifying
them out.
These
sources of
into two components.
(input
price,
affect expectations
other
and not
try to separate
dynamic behavior are grouped
Sources that
aff·ect net
returns to
commodity price, and technology) will
affect expectations of
The
farmland price
to assume that the sources of dynamic behavior in
farmland prices are "confounded,"
land
the
net
returns,
of what
component
and
land prices
consists
of
will eventually
are going to be.
sources
of
dynamic
adjustments in land price itself.
An
example
Shalit-Schmitz
of
the
model.
uncertainty about
latter
When
component
there
occurs in the
is
considerable
future rents, lending institutions would
practice capital rationing and "focus more
on market value
of collateral, rather than on prospective net rents" (Burt,
1986, p. 12), thus producing a cyclical adjustment path for
farmland
prices.
Researchers frequently try to separate
out these sources of
formulate
hypotheses
specification of these
adjustments"
dynamic behavior
on
the
"price
components
are
usually using time series data.
evaluate the
estimation results
forms
in land
that they take.
expectations"
then
prices and
tested
The
and "dynamic
empirically,
When the criterion used to
are variable coefficients
with significant t-ratios and the "right" sign, time series
data usually would not contain enough information to reject
30
the
~
priori specified mechanisms in the price of farmland.
Burt concludes that
if statistical estimation of the general model
yields precise estimates of unknown parameters,
then an attempt can be made to justify one or
more plausible structures for the formation of
expectations about
rents and capital gains.
(p. 12)
In
specifying
capitalization is
the
the
model,
assumed constant
argue against this
assumption
implicit
rate
over time.
on
the
point
of
One could
that
it is
Tanzi (1980) believed that the real rate was
unrealistic.
associated with the position of the economy in the business
cycle
or
that
it
depended
on
the
inflation
rate
as
hypothesized by Feldstein (1980).
The
classical theory of
interest, originated by Fisher, came to the conclusion that
the
equilibrium
constant
over
real
time,
rate
of
because
interest
in
mainly depends on intertemporal
with
in
changes
produced results
rate of
The rational
(1973)
assumption of a constant
support for
to support
productivity for
short
run
participants are
study
(1984)
an almost constant
the past few decades.
discussed by Sargent
evidence
that
capitalization
this assumption
term investment
long run this rate
Darby's
expectations hypothesis
implies
fairly
consumer preferences along
productivity.
that seem
change in
the
remains
supports
rate.
the
Intuitive
would be that due to the long
characteristics
of
the
farmland market,
more likely to use longer term real rates
31
of interest
to capitalize
farmland values.
In summation,
Burt states that
the empirical question
is
whether farmland
investors take- account of these year-to-year
movements in their decisions or think of a longer
run equilibrium.
(pp. 12-13)
The traditional
capitalization approach
model developed here does not explicitly
the
influence
of
tax
rates
explains that there are
delaying payment
"
of capital
and thus the
take into account
on
farmland
many
devices
prices.
Burt
available for
gains taxes, even to the next
generation ... " (p. 13) and that
changes in tax rates, especially effective rates
on
farmers
and
owners of farmland, occur
infrequently and in an evolutionary way as new
loopholes are discovered and then lost with new
I.R.S. rulings and legislation. (p. 13)
Burt then points
weakly · with
out
rents,
that
the
because
"would-be
of
taxes correlating
independent variables
associated with tax rates over time can be
compounded into
the disturbance term if the form of the regression equation
is such that these variables enter in an
additive way with
the disturbance" (p.13).
Net rent
data used
crop-share
~ent
data
constructed
are
to estimate
data mentioned previously.
from
all agricultural land in
this
index
the model is Scott's
into
a
grainland in Illinois.
the USDA land value index for
Illinois.
land
The land price
value
Burt
series
Estimation of
(1986) adjusts
for high quality
the model
using the
32
Scott data not only resulted in statistically credible land
price
equations,
constraint
near 4%.
but
near
also
one
and
A study by
an
estimated
homogeneity
an implicit capitalization rate
Watts and
Johnson (1985) empirically
estimated this rate at 4. 5.%.
When inferior
measures of
net returns
are used, one
would expect estimation results not quite as encouraging as
above.
The
land
price
reliable rent data may
equations
have some
estimated
from less
statistical credibility,
but it might be asking too much from the data to produce an
estimated homogeneity constraint coefficient
It
would
not
be
surprising if an "unrealistic" implicit
capitalization rate is estimated using this
One method
equal to one.
inferior data.
to be explored in this study is a per acre
gross return measure (price x yield) for the
crops of
the state.
One
dominant cash
could posit that the "true" net
rent measure is nearly proportional to the gross measure so
that in
separated
constant
the logge·d
from
model, the proportionality constant is
the
term
of
gross
the
constant term has this
measure
regression
a
in the
Because the
added (log
of the
this term cannot be interpreted
as an implicit capitalization rate.
in
combined
equation.
extra component
proportionality constant),
reflected
and
potentially
This phenomenon may be
"unrealistic" capitalization
rate resulting from estimation using an inferior measure of
net rents to farmland.
33
Dynamic regression equations
estimated
using
Townsend,
and
estimation
.
the
program
LaFrance
of
for
DYNEREG
(1986).
distributed
farmland
lag
price are
developed by Burt,
This
program
models
is
for
and/or regression
.
models with time series error term.
language,
the
computational
squares, specifically
Written
in FORTRAN 77
algorithm is nonlinear least
Marquardt's
compromise
(Draper and
Smith, 1981).
Data Compilation
Almost
all
of
the
data compiled for this study are
collected and published by the USDA.
The components
used
to
calculate
Returns
to Total
Agricultural Asset,
plus data for Total Agricultural Asset
Value, Agricultural
Real Estate
are collected
Value, and
Land in Farms
from Economic Indicators of the Farm Sector:
National Financial Summary, 1984.
Land
Value index series
for the U.S. are collected from various issues of Farm Real
Estate Market Development:
changed its
Outlook
name to
Outlook
and
Situation (which
Agricultural Land Values ·and Markets:
in 1985).
For deflating purposes, the Personal
Consumption Expenditure Index (PCEI) is used; this index is
supplied by Burt but can be found in
The Economic
Report of
the yearly
the President.
above are for the years 1942-84,
and are
issues of
All data collected
used when (USDA)
34
accounting data
is explored
as an
alternative measure of
net rents to land.
Data sources cited
and/or gross
used in
revenues are
this part
plus unity,
land
the
price
are
used
when
being explored.
of the
are supplied
yearly issues of
Illinois
below
study, PCEI
by Burt
cash rents
The deflators
and inflation rate
but can be found in the
Economic
Report
data
supplied by Burt, as it is
is
of
the President.
constructed from Scott's (1983) data by Burt for use in his
1986 research
paper.
Cropshare rent
data originate from
Scott's 1983 paper, while cash rents are compiled by Alston
for use
in his
doctorate dissertation
originate from his PhD.
values
(Illinois,
thesis.
Iowa,
research, and thus
Statewide index
Indiana,
of land
and Ohio) are compiled
from various issues of Farm Real Estate Market Developments
publications (cited
in the
previous paragraph).
real estate tax are collected from
various issues
of Farm
of
prices paid by
farmers (for production of all commodities)
originate from
Real
Estate
Taxes,
while
the
index
Per acre
various issues of Agricultural Statistics.
Direct
government
payments
Indicators of the Farm
Sheet Statistics,
Sector:
1984.
originate
from Economic
State Income
and Balance
Components used to compute Gross
Income from
Farming
Expense are
collected from Economic Indicators of the Farm
Sector:
State Income
(excluding
and Balance
dwelling)
and Production
Sheet Statistics, 1984.
35
Finally,
data
grain) and
for
the
value of production for corn (for
for soybeans
Field Crops
originate from
(discontinued in
1980 onward).
various issues of
1979) and
Crop Values (from
Data for total acres harvested for corn (for
grain) and for soybeans are compiled from various issues of
Crop Production (Annual Summary).
After data
measures of
are
net rent
compiled
and
to farmland
the
three alternative
are computed, then each
alternative net rent measure is explored by fitting them to
a farmland price equation similar to the dynamic regression
equation
Estimation
for
land
results
price
for
developed
the
by
alternative
Burt
measures
explored are presented in the following chapter.
(1986).
being
36
CHAPTER 4
EMPIRICAL RESULTS
This chapter presents and
discusses empirical results
for the three alternative measures of net rents to farmland
being exploreQ in this
divided into
use of USDA
farmland,
study.
three parts.
accounting
the
impute
section
viable measure of net rents,
devoted
to
exploring
chapter is
The first section explores the
to
second
Therefore, the
residual
explores
while
gross
a
the
revenues
return to
cash
rents as a
third
section is
also
as
a viable
measure of net rents to farmland.
USDA Accounting Data Measure
This method uses
returns
to
agricultural
assets to
indirectly model
the behavior of farmland prices.
to
Assets
Agricultural
Production
Expenses
are
from
Excluding Dwelling (Watts and
computed
Gross
by
Income
Johnson, 1985).
Returns
subtracting
from
Farming
Production
Expenses are defined as total production expense (excluding
operator households) minus both
net rent
to all landlords
and interest payments (both non-real estate and real estate
excluding operator households), and with costs
labor added.
of operator
Gross Income from Farming Excluding Dwelling
is defined as gross income from farming minus
gross rental
37
value of
data
dwellings (including operator households).
are
collected
agricultural real
total
agricultural
estate values,
asset
Other
values,.
land in farms, and index
of land values.
All value and return data
land
deflated
are
values)
since
only deflated by PCEI,
The index
the calendar year.
current land
the
PCEI
into
constant 1982
by
per acre
it
is
data and
are heavily
per acre
used
inventory adjustments
estimates of
The rest
to
of the data
weighted towards the
This is partly due
the harvest
method
a
recent information that could very
latter part of the year.
accounting
already
Farmers should base their
well go back into the previous year.
collected after
measure by the
of land values are collected early in
values on
are accounting
index of
The index of land values series is
farms series.
measure.
than
into a
dollars, then is deflated
land in
(other
to data being
and mainly due to the accrual
collect
the
data
makes its
at the end of the year (Burt, 1986).
The accounting data are adjusted by deflating them
year's inflation
to make
them more
for one
commensurate with the
land value index data by deflating them with
the following
year's PCEI measure.
A direct measure of returns to farmland is unavailable
with the accounting data.
. and the
Therefore, the land
price model
dynamic regression equation for farmland prices is
modified to account for this.
Let
38
A = total agricultural assets value
0 = agricultural non-real estate assets Value
P = agricultural real estate value
TR = return to total agricultural assets
In equilibrium, the farmland price model is defined as
1
P = -(TR - rO)
(4. 1 )
r
where (TR
- rO)
equals the
returns to real estate (farm-
land) with the assumption that the
real estate
rate of
return to non-
assets is equal to that of real estate assets.
Note that (4.1) can
be rewritten
with the
term 0
on the
right hand side,
P
+
1
= A = --r
0
indicating
an
TR
(4•2)
equilibrium
capitalization
equation
for
determining the value of total agricultural assets.
Following a similar logical
to
arrive
at
equation
progression as
(3.19),
the
that used
modified
dynamic
regression equation for farmland is represented as
(4•3)
with
~t
are
unknown
that a. =
as a multiplicative random disturbance;
1/r.
parameters,
Taking
~0
the
+
~1
~0 ,
~1 ,
...
+ ... = 1, and recall
natural
logarithm
of (4.3)
results in
00
log Pt =log a.+
.~
J=O
~jlog
(TRt-j) +log
~t
(4. 4)
39
Approximation of
(4.4) by a second order rational log then
produces
o+
log Pt =
YQlog (TRt - rOt) +
y1 log
(TRt_ 1 - r0t_ 1 )
+ AlE(log Pt-1) + A2E(log Pt-2) + log ~t
which is similar to (3.16), with
= (1 -- A1- A2)log a and
~t·
E(log Pt-j> =log Pt-j - log
This model
o
(4.5)
is explored by searching over a few values
of r and getting a rough approximation to the least squares
solution; i.e., an r that produces an estimated equation of
(4.5) with the smailest residual sum of squares.
specification
of
r
results
farmland as Rt-j = (TRt-j
having a
in computation of returns to
rOt-j>,
similar structure
A priori
j=1,2, ...
as (3.16).
and (4.5)
But there is not a
strong a priori .basis for imposing the constraint
rate of
return on
that the
real estate assets be equal to the rate
of return on non-real estate assets.
It is not
return on
clear
that
non-real estate
assuming
a
constant
assets is as easily justifiable
as imposing this constraint on real
estate assets.
two classes of assets do not have much in common.
(real estate) is
(durable
asset),
aggregation
of
fixed
in
while
very
having a fairly
lived.
supply
non-real
heterogeneous
livestock, inventories,
elastic
rate of
and
of
estate
These
Farmland
infinite life
assets
capital
is
an
(machinery,
financial assets, etc.) as well as
supply
function
and
are short-
40
Assuming an
additive distributed lag structure on net
returns (and error structure) enables r to be
estimated as
a
the
free
parameter.
An
additive
regression equation for farmland
form
price·
of
dynamic
approximated
by a
second order rational lag is represented as
Pt
= a1(TR1
A2E(Pt-2) +
with
~t
- ~Ot) + a2(TRt-1 - ~Ot-1) + A1E(Pt-1) +
~t
a white
(4.6)
noise error
notation for the parameter r.
assumption
enables
the
Ot can
TRt- 1 =
a more consistent
The additive distributed lag
than in
logged form)
so that
be separated out, thus estimating its
coefficient as a free parameter.
TRt =
~
dynamic regression equation to be
specified in levels (rather
the variable
term and
At the equilibrium state,
TR*, Ot = Ot_ 1 = 0*, Pt = Pt- 1 = Pt- 2
= P*;
so that (4.6) becomes
E(P*) (1 - A1
A2)
= a1(TR*
-
~0*)
+ a2(TR* -
( a1 + a2)
(4. 7)
As shown, specifying the dynamic regression
in levels
and without
constraint.
~0*)
equation (4.6)
an intercept forces the homogeneity
Equation (4.7)
interprets
the
reciprocal of
the capitalization rate as
( a1 + a2)
1
( 4. 8)
=
r
(1 - A1 - A2)
An easily estimated form of (4.6) is
41
Pt = a1TRt + a2TRt-1 + Y10t + Y20t_1 + A1E(Pt-1) +
A2E(Pt-2) + ~t
a1~
where Y1 =
the implicit
+ a2)
is
(4.9)
and Y2 =
a2~·
With the estimation of (4.9),
rate r
capitali~ation
estimated,
and
=
(1
the hypothesis that the rate of
return to farmland is equal to the rate of return
assets
=
(~
r)
can
be
checked.
of other
In the event that the
distributed lag structure on TRt-1 is negligible so that a 2
= Y2 = 0, equation (4.9)' reduces to
(4.10)
and
the
problem
of
estimating
redundant parameters (a 2
and Y2) is eliminated so that the least squares estimate of
r becomes
~
=y1/a1.
Equation
(U.S.)
using
(4.9)
a
is
estimated
sample
period
for
of
the United States
1942-1984,
estimated equation turns out to be unstable.
taken is to shorten the sample period to
but
the
The next step
1951-1972 for the
'
purpose of trying to locate a more "stable" period in which
to re-estimate ( 4. 9); but it still
estimated
equation.
The
results in
deflating
method
an un·stable
used on the
variables is redefined so as to possibly make the variables
more
commensurate
in
terms
of time with respect to each
other, but instability still
persists.
index
place
of
the
estimation of (4.9) produces
an
unstable
is
used
encouraging note
in
is that
When
accounting
land value
measure,
equation.
One
in all of the trial estimations,
42
the estimated coefficients for
variable
always
has
the
opposite
coefficients
for
qualitatively
corroborating
respect to the
returns
signs
non..:real
the
on
estate value
sign of the estimated
total
to
the
assets
thus
structure of (4.6) (with
non-real
estate
and
return to
assets variables).
The
poor
results
from
estimating
(4.9)
with
the
underlying assumption of an additive distributed lag on net
returns prompt
a return
to estimation of (4.6) based on a
multiplicative distributed lag
structure
on
net returns.
Returning to the originally proposed model means that a way
must be found to
returns
with
structure
(4.6)
one ·variable
so
instead
as
of
to
reflect net
two.
A priori
determining r (rate of return to non-real estate assets) as
distinct from
a
=
1/r would
be a way to fulfill this and
alleviate the necessity to modify the DYNREG·program needed
to estimate (4.6) in its original form.
An r
of 3.5%
is looked
at so that net returns could
be calculated for each year, and
can be
estimated with
1950-52, 1974-75
are
such that
equation (4.6)
the structure of (3.16).
dummied
out
such
that
The years
the year's
effect of farm returns is excluded from the distributed lag
structure and that
realized.
its
effect
contemporaneously.
on
land
The
prices
years
is fully
1950-52
are
associated with the Korean War, while the years 1974-75 are
associated
with
years
following the quadrupling of crude
43
oil prices and Russia buying up the surplus stocks of grain
in the
U.S.
During these
extremely high,
farm income.
years,
u.s.
grain prices were
thus feeding into extremely high levels of
The
extremely
high
levels
of
income are
considered aberrations due to a one-time violent shock from
outside the system.
because at
The sample period is
cut off
at 1983
the time the data was being collected, the 1984
levels were preliminary measures subject to revisions.
comparison
purposes,
equations
are
For
estimated using both
the accounting and index data for the land price variable.
Estimation of (4.6) produced
first
differenced
measures, although
The
first
model
reproduced in
models
Table 1.
series.
An
is
suggest a
accounting
and index
estimated
with
and
constraint, with results
1 and
land
2 are estimated
price
variable while
and 4 are estimated using the land price index
implicit
estimated because
model
are
Equations
the
that
equations are unstable.
the homogeneity
using accounting data for
equations 3
both
both estimated
differenced
without imposing
for
results
capitalization
rate
can
not
be
the intercept is subtracted out when the
differenced.
A
second
order
non-stochastic
difference model (in first differences) with a second order
lag on the rent variable is also estimated.
But
only the
estimated coefficients from the first order lag on rent and
land price are significant,
thus
only
regression results
from estimation of the first order model are presented.
44
Table 1:
Regression results
model using USDA
period 1942-83)a
for final u.s. land price
accounting
data (sample
----------------------------------------------------------2
Equation No. :
1
3
4
Rentt
.0395
(.0163)
.0425b
.0485
( .0329)
.0915b
E(Pt-1)
.8801
(.2686)
.9575
(.0157)
.6138
(.3647)
.. 9085
(.0279)
AR(1)
error
Linear
homogeneous
.8189
(.0885)
no
.8487
(.1542)
yes
.1399
(.1528)
no
.2731
(.1543)
yes
Adj. R2
.4681
.4793
-.0472
-.0632
Std. Error
est.
.0321
.0318
.0608
.0639
Error Sum
sq.
.03498
.03540
.12561
.14294
Durbin
Watson
Degrees of
Freedom
1. 78
1. 79
2.01
2.07
34
35
34
35
===========================================================
arn first differences and with the years 1950~51 and 197475 dummied out
bstandard error not computed
45
One important observation from the results
is that
returns have
no explanatory
of land values are used (equations
that the
results from
respect to a measure
return to
of the
index
of
4).
This means
land prices.
It also demonstrates
accounting
data
are
not a
of per acre land prices if we assume that
land
values
correlation observed
to the
3 and
relationship between residual
that land prices derived from
the
value when the index
equations 1 and 2 are spurious with
farmland and
reliable measure
in Table 1
is
reliable.
Some
of the
in equations 1 and 2 are probably due
ineffective
removal
of
"other
assets"
from the
"total assets" measure by use of (TRt- rOt)·
The next step is to disaggregate to the state level to
see if
the
analysis
encouraging results
national level.
level
does
level.
would
have
similar,
or
compared to the results arrived at the
But accounting data compiled for
not
guarantee
a
reproduction
excluding
dwelling
can
calculate
production
be
found,
expenses
are
included households
households were needed).
Net
but
rent
from farming
data
only
the U.S.
for the state
Data needed to calculate gross income
measures that
even more
needed
compiled
to
for
(measures that excluded
to
all
landlords and
operator labor measures are not collected also.
The "missing"
question the
not
"missing"
data at the state level can lead one to
validity of
at
the
the same
national
type of
level.
data that was
Aside from that,
46
there are
questions as
used to
collect the
applied
somewhat
to the
data, and
uniformly
Measurement of economic
type of accounting methods
also if
these methods are
across
depreciation
the
for
sample
area.
non-real estate
assets is nearly impossible, causing the accounting data to
be only a crude
measure at
probability
aggregation
and it
is
of
relatively
best.
Add to
these the high
problems inherent in the data
easy
to
see
how
poor estimation
results could be produced.
Cash Rent Measure
The second
alternative measure
proxy for net returns to farmland.
the
amount
of
rent
(paid
uses cash
A
cash rent
rents as a
lease has
to the landlord) specified as
either a fixed amount per acre or a fixed lump
sum.
Under
a cash lease,_ the landlord furnishes the land and buildings
while the tenant receives all of the income
typically
pays
insurance,
and
all
major
expenses
repair
except
costs
generated, and
property
to
taxes,
buildings
and
improvements (Kay, 1981).
In comparison,
landlord
is
produced with
( Kay , 19 81 ) .
midwest and
to
a crop-share
receive
a
lease specifies that the
certain
share
of
the crops
the proceeds from the sale becoming the rent
This
type of
other areas
lease is
where cash
Many crop-share leases have
more popular
in the
grain farms dominate.
the landlord
pay part
of the
47
variable
costs
in
the
same
proportion
production to be re'cei ved (as rent) .
measure
is
crop-share
rent
used
Since
in
analysis begins by using Illinois data.
leases are
the reference
Burt
(1986),
the
Although cash rent
few in number in Illinois, this will not affect
its predictive
ability
closely
of
those
if
cash
crop-share
rent
a measure
of net
adjustments follow
rent and imputed returns to
land in owner-operator situations.
as good
as the share of
If
rents to
cash rents provide
farmland as the crop-
share rent data.in Burt (1986), this can be a start towards
a more
common (easily
collectible) measure of net returns
to farmland.
Cash rents tend to be inflexible in their adjustments,
and thus slow to change because landlord and tenant have to
renegotiate the
contract
whenever
reflect new economic conditions.
takes time and effort
parties
so
that
leases
are
prices
lease
can
is
renegotiation
a
cost
for both
commodities, resource
But by
recent
affect
process;
made to
induce either (both) landlord
renegotiated,
possibly
into
of
or (and) tenant to · renegotiate.
rent
is
negotiated less frequently.
prices
technology will
change
The renegotiation process
translating
Pressure from changes in
prices, and
a
the
especially
the time
changes in farmland
decisions
on
a cash
the
made
part
in· the
of
the
landlord wanting a higher rent because land has become more
"valuable."
So
there
is
a
joint
dependency
between
48
farmland prices and cash rents in that changes in the level
of one affect the level of the other and vice versa.
Regressing
coefficient
land
price
estimates
on
on
cash
cash
prices.
the
as a
result of
The method
problem
method
of
involves
correlated with
of
are ·correlated
with the
its joint dependency with land
instrumental
biased
will produce
rents that are biased and
inconsistent, because cash rents
error term
rents
variables alleviates
and inconsistent estimates.
choosing
a
variable
that
is
This
highly
the endogenous "independent" variable, and
at the same time uncorrelated, in the limit, with the error
term.
For this
analysis,a special
instrumental variables
squares regression.
function
of
is
used,
case of the method of
namely
two
stage least
This method specifies cash rents as a
the . exogenous
variables
in
the
system.
Regressing cash rents· on these exogenous variables produces
predicted values of cash
the
error
term
in
rents that
the
land
values are highly correlated
cash rent,
are uncorrelated with
price equation.
with the
And these
original values for
thus fulfilling the two conditions specified by
the method of instrumental variables.
The
suitable
variable.
first
stage
equation
for
Crop-share
exogenous variable
of
the
the
rent
with which
is also a net rental payment
analysis
instrument
is
is
producing
a
for the cash rent
probably
be
the "best"
to explain cash rents as it
for the
leasing of farmland.
49
These
measures
two
although
crop-share
signals
price.
follow
since
it
rent
is
similar
is
a
more
adjustment
sensitive
function
rents
adjustments, the
·this plus
indicates
that,
former seem
its strength
market
of commodity yield and
Comparison of time series plots of
crop-share
to
paths,
to lag
cash rents with
with
respect
the latter.
to
All of
in explaining land prices indicate
crop-share rents to best explain cash rents.
Regression of cash rents on crop-share
second order
rents within a
non-stochastic difference equation similar to
(3.16) results in crop-share
rent
variation
The deflation process that is
in
cash
rents.
explaining
used on the variables is similar to the
with
the
cash
rent
from
a
survey
by Burt,
that
asks farmers
rents are in the surrounding area for that year.
Since crop-share rents seem to explain
variations
one used
of the
data originating from Alston (1986).
This data is collected
what cash
96%
in
cash
a major
portion of
rents, regressing crop-share rents on
potential exogenous variables would be a logical first step
in trying
to locate
exogenous variables
that affect cash
rents.
Both crop-share rents and cash rents
to the
landlord by
the tenant
farmland, so the factors
(which is
rents.
which affect
production) would
These factors
for use
are sums paid
of the landlord's
usage of
this land
affect determination of these
are imputed
through two components,
50
return and cost of production.
The two rents are basically
the net amount of returns minus the cost of production.
Several proxies for the return and cost components are
ekplored.
and
The proxies
Soybeans
dwelling) for
or
are either Gross Revenue from Corn
expense
from
Recall
that the
measures
are
section.
the dominant
in Illinois
measure of
cash crops
farmland
going
to
accounting
Corn and
crop
production expense
not collected at the
(excluding
gross income and
from
explored in the previous
Illinois
Farming
variable, and Production Expense
the return
for .the cost variable.
production
Income
Gross
data
soybeans are
with a· major part of
production.
Since a
that excludes households is
state level,
a method
that prorates
the components of production expense (including households)
with
the
ratio
of
each
production
expense
component
excluding households
with the same component that includes
households
national
(at
components up
the
construct
also constructed
adding
all
the final
Gross
Farming
from
this
measure.
Ope.rator
through a prorating method, and
added on to produce
Income
and
(except interest expense and net rent to all
landlords) is used to
labor is
level)
Production Expense measure.
and
Production
Expense
are
deflated to a per acre measure by the land in farms measure
described in the previous section.
and soybeans is
computed
as
a
Gross revenue from corn
weighted
number of acres harvested for each.
average
of the
51
The
equation
regression
for
crop-share
rent
is
formulated as
SRt
=n
+
~GRt
+ oCt +
(4.11)
~t
with SR representing crop-share
of the
rent, GR
two measures
of gross
Different
combinations
costs~
proxies,
and
also
the
soybean production
than the other
variation in
returns, and C representing
of
return
explain crop-share rent.
representing one
the
cost
and return
proxies alone are used to
Gross
Revenues
from
corn and
by itself performs statistically better
combinations
as
crop-share rent,
it
explains
78%
of the
and has the lowest standard
error of the estimate and sum of squared error.
Cash rents are then
return proxies.
In
level are
an alternative
the hope
paths.
proxies are explored
explain cash
in
index and follow
order
to
find
a
better than
set
variables in
cash rent.
First
t-1
noise
and
error
t-2
term)
equations explored.
that can
the others and
a~d
the equation
second order
non-stochastic difference equations with lags of
or
the cost
Combinations of cost and return
rents relatively
instrument for
proxy for
this national
thus produce a set of exogenous
of the
cost and
that production costs at the state
proportional to
similar adjustment
the above
addition, the index of prices paid by
farmers is included as
variable with
regressed on
t and t-1
on the explanatory variables (with white
are
the
basic
structures
for
the
52
Regression results
because
the
are
structures
counterintuitive
and
varied
and
that
are
implausible
on
a
not encouraging,
estimated
priori
are
grounds.
Detailed results of these regression runs are not reported,
but an attempt is made
to
summarize
these
results
in a
rather descriptive way.
The proxies
for cost
of production do not do well as
a group regardless of their combination with gross returns.
Estimates of the coefficients for this variable either have
the wrong sign (should be negative), or have the right sign
and not
be statistically significant.
return variable perform only
income from
The proxies for the
slightly better.
When gross
farming is used in conjunction with production
expense, the estimated coefficients for the return variable
are
statistically
significant.
Also
some
equations
estimated with this equation are unstable, which made these
accounting measures
somewhat suspect.
Gross revenues from
corn and soybeans production produces coefficient estimates
with
the
right
sign
(positive),
but
which
are
not
statistically significant.
For further information concerning the appropriateness
of
cash
rents
as
an
alternative
farmland prices are
then
rents.
Cash rents
are also
see
there
if
between the
is
regressed
evidence
two variables.
measure of net .rents,
directly
regressed on
to
support
on
to cash
land prices to
joint dependency
lnitial regression results for
53
the two
These
models indicate
equations
homogeneity
acre
for
are
a need for differencing the data.
reestimated
constraint.
high
adjusted with
without
the
Scott's (1983) "··· dollars per
quality
grainland
in
Illinois
and
the USDA land price index" (p. 797) that was
reconstructed by Burt (1986) are used
The cash
imposing
rent data
as land
value data.
is from Alston's study (1986) found in
the same publication as that of the land price index and is
measured
the
same
way
(survey), therefore no adjustment
(for inflation) is needed.
Regression results from land prices regressed
rents are presented in Table 2.
for the difference parameters
may indicate
that lagged
able
to
are relatively
If cash rents are
explain
prices, then these lagged
The estimated coefficients
small.
This
values of cash rents do not have
much explanatory capacity.
variable
on cash
the
values
an exogenous
dynamic
behavior of land
should
have substantial
explanatory capacity.
Burt's
rents
produced
estimated difference equation parameters that are
far more
significant
results
than
the
that lagged values of
explaining
land
using
rent
4
as
prices
the
variable estimates indicating
rent have
equation 1 (or 2) as the
equation
crop-share
than
a more
important role in
current year rents.
unrestricted (general)
restricted
model,
the
Taking
model and
F-statistic
computed is not statistically significant at the 25%
54
Table 2:
Regression Results for Land Prices Regressed on
Cash Rents for Illinois (sample period 1961-83)a
===========================================================
Equation No. :
1
2
3
4
.9243
(.2423)
.9149
(.2325)
.9210
(.2259)
.9447
·(.2302)
Rentt_ 1
-.0597
(.7779)
.0045
(.6244)
.3178
(.2282)
E(Price>t- 1
.40171
(.8108)
.3023.
(.5706)
E(Price>t- 2
-.1076
(.2862)
AR(1) error
.4654
(.1846)
.4547
( .1857)
.4480
( .1864)
.4632
( .1848)
.6855
(.1518)
Adj. R2b
.5507
.5824
.5959
.5816
.2242
St. error est.
.0556
.0546
.0536
.0548
.0690
Error sum sq.
.05558
.05661
.05750
.06307
.10009
Durbin Watson
Degrees of
freedom
5
.1792
(.2931)
2.06
2.07
2.04
2.14
2.10
18
19
20
21
21
==========================================================
arn first differences
bExclusive of autoregressive (AR) disturbance
55
level.
Looking
estimated
at
equation
coefficient
for
3,
the
the
rent
t-ratio
for
variable lagged one
period is not significant at conventional levels of
10%.
Only
contemporaneous
significance with
rents
t-ratios that
the
show
5% and
statistical
are relatively consistent
in all of the estimated equations (around 4.0).
These results
suggest that
contemporaneous rents as the
good
or
better
job
a static
model with only
explanatory variable
of explaining the variations in land
prices than when dynamics are incorporated.
inclusion
of
a
regression
does
does as
distributed
not
lag
affect
the
Also.note that
structure
into
autoregressive
the
error
remained fairly constant and statistically
estimates which
significant throughout.
Regression results of
cash
rents
prices are presented in Table 3.
the
regressed
on land
They are somewhat similar
to the results
from
similar F-test
is conducted between equations 1 (or 2) and
4, with the computed
values
of
land
F-statistics
price
explaining cash rents.
2 and
from Table
previous
are
regression
indicating
even
less
than
land
A
that lagged
significant
in
Comparison of equation 3 from Table
3 indicates
that cash
rents lagged qne
period are relatively more significant (in
prices)
model.
prices
(in
explaining
explaining land
cash
Comparison of equation 5 for the two tables suggests
rents).
56
Table 3:
Regression Results for Cash Rents Regressed on
Land Prices for Illinois (sample period 196183)a:
----------------------------------------------------------2
Equation No. :
3
1
.4723
( .1250)
4
.4696
( .1222)
5
.4752
.5011
( .1191) ( .0975)
Pricet_ 1
-.1334
( .3844)
-.1077
( .3835)
E(Rent>t-1
.3677
(. 7133)
.3414
( .6956)
E(Rent>t-2
.0249
( .2399)
AR(1) error
.2064
(.2040)
.2060
(.2040)
.1970
(.2044)
.2082
(.2040)
.1634
(.2057)
Adj. R2b
.4947
.5257
.5594
.5816
.2388
St. error est.
.0400
.0390
.0383
.0375
.0501
Error sum sq.
.028B5
.02886
.02935
.02959
.05268
Durbin Watson
Degrees of
freedom
.0507
( .1247)
.3453
( .1328)
2.02
2.02
2.02
2.03
2.04
18
19
20
21
21
==========================================================
arn first differences
bExclusive of AR disturbance
57
otherwise.
But
the
magnitude
fairly trivial from a
of these differences are
statistical point
of view,
and are
also of little practical importance.
Regression results
any dynamics
rents
from Tables
involved
and
land
in
the
prices
are
2 and 3 indicate that
relationship
extremely
between cash
weak,
and
that
basically only contemporaneous levels are needed to explain
the variation
indication
of one
of
a
variable by
phenomenon
spurious regression."·
What
exogenous variable (or
set
values of
the other.
kn.own
as
"third
is happening
of
This is an
variable
is that a third
variables),
namely lagged
crop-share rents, is (are) separately explaining
both cash rents and farmland
prices.
mirage
dependency
effect
variables
of
that
determined by
joint
are,
in
the third
This
reality,
between
being
variable.
is not
the
two
simultaneously
These results lead to
the conclusion that cash rent, as a measure
to farmland,
produces the
of net returns
useful in explaining the movements of
farmland prices.
Gross Revenue Measure
The
farmland
third
alternative
explored
in
production of dominant
rents
proxies
on
the
this
study
crops.
different
briefly
measure
mentioned
of
is
net
in
and
the
to
gross revenue from
Regression
cost
returns
return
of crop-share
to production
preceding
section
58
suggests
use
the
of
this
that gross
regressions are
production (in
measure.
Results
revenue from
Illinois) does
of
the
corn and soybean
a relatively
better job of
explaining Illinois crop-share rents than when cost proxies
are included,
or when
Dwelling (Watts,
Gross Income from Farming Excluding
Johnson,
proxies are used.
1985)
a
likely
rents.
to
So
candidate
as
A major portion of
crop
production,
and
soybeans are dominant.
reflect
a
major
with
the cost
Crop-share rents explain farmland prices
rather well (Burt, 1986).
is
together
this gross
an
alternative to crop-share
Illinois farmland
among
these
Gross revenues
portion
revenue measure
of
the
is delegated
crops,
corn
from the
return
and
two crops
to farmland in
Illinois, albeit a gross rather than a net measure.
Gross revenue
value
from corn
and soybean
of
production
measure
soybeans.
This value
of production
the season
average price
for
series 'is
grain and of
defined as
crop's production estimates
Each value of production estimate is deflated to a
per acre measure by the
crop.
corn
received by farmers for the crop
($/bu.) that is applied to the
(bu.).
of
production is a
Land
value
total
data
acreage
harvested
is a series constructed by Burt
(1986) from Scott's "··· dollars per acre for
grain land
in Illinois
p. 797).
high quality
for comparable and paired sales in
1961, 1967, and 1981 and adjusted with the
index" (1983,
for the
USDA land price
The land price index is published
59
as a
series for
March 1 before 1976, February 1 for 1976-
81, and April 1 for 1982 and onward.
on surveys
that ask
The indices are based
farmers to estimate the average value
of farmland in the surrounding area.
Land value indices are used
comparison
with
dollar
value
these surveys occur early
that farmers
Gross revenue
go
is heavily
the end of the
year,
As
land
in the
value
data for
price levels.
year, it
corn and
usual,
back
Since
is conceivable
into
the
previous year.
weighted by revenues received at
during
harvest
time.
Thus gross
soybeans should be deflated for one
year's inflation to be
data.
land
base their estimates of land prices on recent
information that could
revenue from
as
commensurate
both
land
with
the
land value
value and revenue data are
deflated to a base year dollar measure (1982) by the PCEI.
Initially, land prices are regressed on. gross revenue
and
a
cost
variable.
A proxy for the cost variable is
Production Expenses (Watts and
the first
is
a
section of
cost
imputed
Johnson,
this chapter.
to
all
1985)
Although this measure
agricultural
imputed to farmland should follow
a
defined in
assets,
costs
similar time path since
farmland is a major component of total agricultural assets.
But
regression
results
suggest
because the estimated coefficients
reflect a
that
this
for
the
is
not
so,
cost variable
total distributed lag effect on land prices that
are positive
when
it
theoretically
should
be negative.
60
These coefficients
are also not
statistical~y
significant,
so the cost variable is dropped from the model.
Land
prices
alone, within
are
then
a structure
regressed
estimated
is
lagged
with
at
and
constraint imposed.
(3.8)
t-1
except
and
without
t-2.
the
that
the rent
The equations are
long-run
homogeneity
regression results are presented
The
One notices
in equations 1-4, Table 4.
on
coefficients
estimated
gross revenues
equivalent to (3.16) and also a
structure that is similar to
variable
on
instantly that the
difference
the
equation
parameters (A1, A2) of equations 1 and 2 are very
Burt's estimated
one
of
coefficients (equation 1, Table 5).
Burt's
estimated
comparison purposes
his
estimated
equations
the
equations
as all
of the
is
reproduced
regression results for
estimated
with the homogeneity constraint,
estimated
difference
An F
similar.
equation
When
parameters
the null
(h=1) constraint.
model.
two
become
The
hypothesis,
Computed
hypothesis of restricting the
model (h=1) are 68.47 and 37.75 both with 1 and
null
the
test is conducted to test .the statistical
F-statistics under
freedom.
for
are
importance of the homogeneity
of
Only
equations
are
unstable.
close to
19 degrees
F-statistics indicate rejection of the
thus
favoring
non-restriction
of
the
61
Table 4:
Initial Regression Results for
Prices Using by Gross Revenue
Soybeans (Sample Period 1960-83)
Illinois Land
from Corn and
===========================================================
Equation No. :
Intercept
1
2
3
4
-.0831
(.0392)
-.5473
(.0410)
.2068
(.0193)
.0512
(.0292)
.2137
(.0304)
.1426
(.0519)
-.0175
(.0263)
Rentt-1
.2130
( .0359)
-.0418
( .0475)
Rentt_ 2
-.4329
(.0628)
E(Price)t_ 1
1. 6214
( .0440)
1. 6808
( .0502)
1. 8149
(.0326)
1.9430
(.0203)
E(Price)t_ 2
-.7638
(.0405)
-.8090
( .0421)
-.9628
(.0262)
-.9872
(.0172)
no
no
yes
yes
Linear
homogeneous
.9908
.9910
.9637
.9752
St. error est.
.0339
.0338
.0708
.0569
Error sum sq.
.02179
.02167
.10032
.06473
Durbin Watson
Degrees of
freedom
1. 41
1. 31
.43
.70
19
19
20
20
==========================================================
astandard error not computed
62
Table 4:
Continued
===========================================================
Equation No. :
Intercept
5
6
7
8
-.6371
(.0507)
-.5411
( .0720)
-.0785
(.0381)
-.6260
( .0416)
.2367
( .0307)
.1895
(.0089)
.2037
{.0091)
-.0097
( .0269)
Rentt_ 1
.2169
(.0369)
-.0575
( .0482)
Rentt-2
E(Price)t-l
1.6266
(.0431)
1. 6900
(.0475)
1. 6400
(.0345)
1.6367
(.0338)
E(Price)t_ 2
-.7629
(.0398)
-.8114
(.0402)
-.7786
(.0342)
-.7710
(.0335)
no
no
no
no
Linear
homogeneous
.9920
.9925
.9912
.9924
St. error est.
.0345
.0337
.0333
.0337
Error sum sq.
~02260
.02160
.02227
.02274
Durbin Watson
Degrees of
freedom
1. 45
1. 40
1. 34
1.42
19
19
20
20
==========================================================
63
The
indices
models
two
are
results
with
then estimated using land value
presented
in
equations
5
and 6.
Overall, the regression results are not much different from
the results of equations 1 and 2.
land
price
estimated
1.37, 1.34, 1.52,
that
forcing
from
and
the
equations
1.48,
homogeneity
in the
corn and
do a
soybeans may
two
of
model.
imposes
an
Gross revenue from
longer
term.
The long-term
the true values of land price for that
Consistency in
kinds
constraint
This suggests
better job of forecasting for
the short-term than for the
period.
1, 2, 5, and 6 ate
respectively.
implausible.constraint
forecasts overshoot
Long-run elasticities of
land
the regression
results using the
value data indicate the usefulness of
gross revenue in explaining high quality grain land as well
as other farmland.
Attempts to
fine tune the land price model results in
a model similar to (3.16) but
at
lag
t-1.
variables at
The
lag
t
with only
estimated
and
lag
one rent variable
coefficients
t-2
are
for the rent
not statistically
significant at conventional levels of significance, so both
variables are dropped from
the final
model.
Results for
this now model are presented in equations 7 and 8, Table 4,
with equation
8
using
Long-run elasticity
and 8
are 1.38
the
statewide
land
value index.
coefficients estimated for equations 7
and 1.52
respectively, further indicating
64
that the data suggest using a model without the homogeneity
constraint.
Equations 1-8 are also estimated with
inflation
adjustment
on
but
regression
results
only
presented in
with the
table 4.
adjustment
estimate than
the rent data discussed earlier,
with
to
land
the
adjustment
are
This is because equations estimated
have
lower
standard
equations estimated
Since the inflation adjustment
fit
and without the
prices,
errors
of the
without the adjustment.
on rents
produces a better
this adjustment is incorporated and
assumed from here onward.'
There
is
a
problem
with
using
the
gross revenue
measure because it is a return to the factors of production
with farmland being only one component
of it.
Therefore,
gross revenue is adjusted in this study to re'flect a return
to farmland alone.
Most
landlord
crop-share
receiving
translate to
The
rent
leases
50%
one-half of
data
is
then
of
in
the
Illinois
crops
produqed,
revenue received
adjusted
original revenue estimates.
Real
provide
the
which
for the crops.
to include only 50% of
estate
tax
is
a cost
borne by the landlord, and is subtracted out.
What is left
is a more precise estimate of returns imputed
to land, but
is
still
a
landlord are
gross
measure
left out
since other costs paid by the
(primarily the
fertilizer and chemicals).
landlord's share of
65
The
adjusted
rent
estimation process
to
data
that
is
run
through
described
a similar
previously.
This
results in
a model with only one rent variable at lag t-1.
Regression
results
presented in
using land
Table 5,
the
little, but
the
7 and
estimated
model
are
and 3, with equation 3
Comparing
these two equations
8 (Table 4) indicate that almost all
intercept)
that the
revenues are used.
final
equations 2
value indices.
with equations
(except
for
estimated
fit is
coefficients
changed
improved when adjusted gross
This is shown by
the reduction
in the
standard errors of the point estimates.
An
implicit
capitalization
from the final farmland
rate
cannot be computed
price equations
estimated in this
study, as is done in Burt's (1986) study.
Costs imputed to
farmland and proportional to gross revenue
the
rent
confounded
variable;
into
the
transformation is
estimated intercept
are imbedded in
this proportional cost amount becomes
intercept
term
when
the logarithm
done on the land price equation.
coefficient
does
not
have
So the
the same
meaning for the land price model estimated in this study as
that estimated in Burt's study.
The study here digresses a little in order
the potential
to explore
for direct government payments to strengthen
the explanatory value of final estimated land price models.
Direct government payments are deflated to a per-acre
66
Table 5.:
Regression Results for Final Illinois Land Price
Model Using Gross Revenue from Corn and Soybeans
(Sample Period 1960-83)
----------------------------------------------------------1
2
Equation No. :
3
Intercept
.4091
(.0268)
.0780
(.0318)
-.4688
(.0335)
Rentt.
.0708
(.0164)
Rentt_ 1
.0563
(.0238)
.1905
(.0086)
.2050
(.0089)
E(Price>t-1
1. 6317
(.0325)
1.6105
(.0329)
1.6062
(.0322)
E(Price)t_ 2
-.7588
(.0255)
-.7515
( .0324)
-.7431
(.0317)
yes
no
no
Linear
homogeneous
Adj. R2
.9952
.9917
.9929
St. error est.
.0241
.0321
.0325
Error sum sq.
.01049
.02064
.02108
Durbin Watson
Degrees of
freedom
2.59
1. 45
1. 54
18
20
20
----------------------------------------------------------
67
measure
by
the
total
number of acres harvested for both
corn (for grain) and soybeans.
Equations
framework
are
similar
estimated
to
(3.16)
Government
price data.
within
a
distributed
lag
using the two types of land
payments
treated
as
a similar
exogenous variable produces a smaller standard error of the
estimate for its estimated land price equations compared to
when these
estimated
payments are summed up with gross revenue.
coefficients
relatively
low
for
government
t-ratios
and
are
significant at conventional levels
10%).
payments
not
The
have
statistically
of significance
(5% or
This result is caused by the fact that data used for
government payments
sector,
but
also
dairy sectors.
variable
is
contribute
are not
include
It
also
a
gross
much
in
only for
the crop production
payments for the livestock and
indicates
that
since
the rent
measure, government payments do not
the
way
of
explaining
land
price
movements as it is such a small part of gross returns.
Given the
encouraging results in using gross revenues
from corn and soybean
prices,
the
production to
analysis
is
then
explain Illinois land
broadened
to include the
surrounding states in the corn belt region, namely Indiana,
Iowa,
and
Ohio.
It
is
thought that these states have
similar farmland characteristics (a major portion
used for
crop production, and a large portion leased with crop-share
leases dominating)
as
in
Illinois.
Only
the adjusted
68
statewide index
the
of land
analysis
as
constructed only
values are
the
dollar
no
results
the
between
analysis
for
fashion as
value
for Illinois.
that there is basically
using
these
was done
used in this stage of
land
prices
Previous results indicate
difference
in
the regression
two land value measures.
three
are
states
progresses
for Illinois.
The
in similar
Final model regression
results for the three states are presented in Table 6.
Analysis with Iowa, Indiana,
the
same
data,
estimated
namely
a
equation
second
and Ohio
structure
order
data result in
as with Illinois
non-stochastic
difference
equation with one rent variable at lag t-1.
Comparisons
adjusted and
revenue
the
(corn
equations,
between
other
and
estimated
one
with
with
unadjusted gross
soybeans production), indicate similar
regression results for all three states.
using adjusted
estimated
Only
the results
revenue data (50% of gross revenue and real
estate taxes deducted) are presented.
The low Durbin-Watson
statistic
from
final equation
estimates for Indiana and Ohio (equations 2 and 4) indicate
the presence of positively
equation was
reestimated for
error structure with the
and
5
in
autocorrelated errors.
Table
6.
the two states with an AR(l)
results presented
The
So the
t-ratios
for
in equations 3
the
estimated
coefficients of the AR(l) error term for the two states are
69
Table 6:
Regression Results for Final Land Price Model
for Iowa, Indiana, and
Ohio, using Gross
Revenue from Corn and Soybeans (Sample Period
1960-83)
----------------------------------------------------------Indiana
Iowa
State:
Equation No. :
Ohio
1
2
3
4
5
Intercept
-.2866
(.0336)
-.2113
(.0387)
-.1796
(.0592)
-.2079
(.0652)
-.1054
(.1015)
Rentt-1
.1802
(.0084)
.1832
(.0099)
.1788
(.0147)
.2163
(.0206)
.1869
(.0261)
E(Price)t-1
1.7380
(.0348)
1.6468
(.0381)
1.6819
(.0497)
1.5557
(.0702)
1.6655
(.0769)
E(Price>t-2
-.8514
(.0355)
-.7788
(.0384)
-.8167
(.0493)
-.7161
(.0666)
-.8203
(.0733)
no
no
no
no
no
Linear
homogeneous
.587
(.1689)
AR error
.748
(.1385)
.9924
.9863
.9836a
.9755
.9679a
St. error est.
.0383
.0437
.0390
.0560
.0443
Srror sum sq.
.02928
.03826
.02737
.06276
.03537
Durbin Watson
Degrees of
freedom
1.55
.93
1.57
.72
1.71
20
20
18
20
18
---------------------------------------------------------aExclusive of AR disturbance
70
3.48
and
5.4,
indicating
that
they
are
statistically
significant at conventional levels of significance.
Comparison between
regression results for these three
states with that of Illinois (equation 3, Table 4) produces
some
results.
interesting
the difference
coefficients for
similar,
but
Not
that
of
the
equation parameters quite
rent
constant throughout the analysis
The estimated
only · are the estimated
variable remains fairly
within
the
four states.
coefficients for the intercept have the same
negative sign, and are
fairly close.
long-run
for
elasticities
coefficients between 1.20
the
and
Computation
three
1.60,
states
of the
results in
indicating
that the
homogeneity constraint assumption is violated.
Gross
revenue
from
production of principal. crops is
explored as an alternative to revenue from corn and soybean
production in
the analysis
for Ohio.
It is thought that
corn and soybeans are not as dominant in
the
other
three
states.
better when revenues from
But
corn and
this state
as in
the fit is statistically
soybean production are
used.
Additional analysis is also done on two plains states:
Kansas and
these two
production.
crops
is
North Dakota.
Wheat
is the
dominant crop in
states, so gross revenue is estimated from wheat
Gross
also
revenue
explored
from
for
production
comparison
of principal
purposes.
The
71
regression results
for the
two states are not encouraging
and thus not presented.
The
practice
of
summer
fallowing
is
important in
Kansas and North Dakota, so that summer fallow acres should
have
been
included
deflating gross
does not
acres.
with
the
harvested
acreage
revenue to a per acre basis.
publish
an
annual
se.ries
for
before
But the USDA
summer fallowed
Exclusion of the summer fallow acreage distorts the
per acre
amount
regression
of
results
returns
for
to
Kansas
farmland
such
that the
and North Dakota are much
different from the results for the cornbelt states.
Distributed lag land price
respect
to
gross
returns
response elasticities with
are
then
computed
estimated models for Illinois (equation 2,
Indiana (equation
and for
These
for
the
Table 5), Iowa,
3, Table 6), Ohio (equation 5, Table 6),
Burt's estimated
elasticities
a.re
equation (equation
partial
derivatives
1, Table 5).
defined
in
natural logarithms as
Y(T)
= aYt-T/BXt = BYt/BXt-T
where X and Y are
logarithms
prices, respectively.
land prices
similarly
from a
the
period(s) into
response
defined as
of
(4.12)
gross
returns
and land
Y(T) measures the effect on current
change in
effect
of
the future.
elasticities
are
rents T
current
period(s) back, or
rents
on land price T
Cumulative (intermediate-run)
also
computed,
and these are
72
T
e(T)
Y(j)
l:
( 4. 13)
j=O
where e(T)
is the - total cumulative effect on current land
price from a change in rents T periods back.
The computed
elasticities are reproduced in Tables 7 and 8 only for T=O,
1, 2, . . . . ,
20,
with
T=20
reflecting
the
long-run (or
equilibrium) response.
The computed elasticities in Table 7 indicate that the
distributed lag response of
estimated
Burt's
land prices
models
using
gross
estimated
model
using
distributed lag
response (using
to rents
revenues
are
cropshare
for the
similar
rents.
to
The
gross revenues) follows a
similar dampened cyclical path indicative of a second order
difference equation
estimated model.
with complex
roots as
Turning points in the time
that of Burt's
paths of Y(T)
occur at similat lag periods, except for Iowa, which is one
period
off.
One
elasticities for
difference
This is
that
the
computed
estimated models using gross revenues are
relatively larger than the
used.
is
ones when
cropshare rents were
reflected in the intermediate and long-run
cumulative elasticities e(T).
The estimated relationship between
gross
revenue
is
further
Final estimated regression
tested
equations
farmland price and
via prediction errors.
for
Illinois, Iowa,
Indiana, and Ohio (equations 2 and 3, Table 5, and all
73
Table 7:
Distributed lag land price elasticities (Y(T))
----------------------------------------------------------Gross Revenue (Corn and Soybeans)
Crop-share
T
Rent
Illinois
Illinois
Iowa
Indiana
Ohio
0
.0708
.0000
.0000
:oooo
.0000
1
.1718
.1905
.1802
.1788
.1869
2
.2266
.3068
.3132
.3007
.3113
3
.2394
.3509
.3909
.3598
.3651
4
.2187
.3346
.4127
.3595
.3528
5
.1752
.2752
.3845
.3108
.2880
6
.1199
.1917
.3169
.2291
.1903
7
.0627
.1020
.2234
.1316
.0807
8
.0113
.0201
.1184
.0341
-.0217
9
-.0291
-.0442
.0157
-.0500
-.1023
10
-.0561
-.0863
-.0736
-.1120
-.1526
11
-.0694
-.1058
-.1413
-.1476
-.1703
12
-.0707
-.1055
-.1827
-.1567
-.1584
13
-.0627
-.0904
-.1976
-.1430
-.1241
14
-.0487
-.0663
-.1876
-.1126
-.0768
15
-.0318
-.0389
-.1589
-.0725
-.0261
16
-.0150
-.0128
-.1147
-.0301
.1955
17
-.0003
.0087
-.0649
.0087
. 0540
18
.0108
.0235
-.0152
.0392
.0738
19
.0179
.0314
.0289
.0588
.0787
20
.0211
.0329
.0632
.0669
.0705
200
.174E-12
.107E-12
-.307E-7
-.355E-9
-.991E-9
===========================================================
74
Table 8:
Cumulative
distributed
elasticities (e(T))
lag
land
price
----------------------------------------------------------Gross Revenue (Corn and Soybeans)
Crop-share
T
0
1
2
3
4
5
6
7.
8
9
10
11
12
13
14
15
16
17
18
19
20
Rent
Illinois
Illinois
.0708
.2426
.4693
.7087
.9274
1.1026
1.2225
1. 2852
.0000
.1905
.4973
.8482
1.1183
1. 4581
1. 6498
1. 7517
1. 2965
1.2672
1.2113
1.1419
1.0712
1.0085
.9599
.9281
.9130
.9127
.9236
.9415
1.0000
1. 7719
1.7277
1. 6413
1.5355
1. 4300
1. 3396
1.2732
1.2344
1.2216
1. 2303
1.2538
1. 2852
1.3511
Iowa
.0000
.1802
.4934
.8843
.1297
1.1682
1.9984
2.2218
2.3402
2.3539
2.2823
2.1410
1.9581
1.7605
1.5729
1. 4150
1. 3002
1. 2353
1. 2202
1. 2491
1. 5891
Indiana
.0000
.1788
.4795
.8393
1.1988
1.5096
1.7387
1.8703
1. 9044
1. 8544
1. 7423
1. 5948
1.4381
1.2951
1.1825
1.1099
1. 0799
1.0886
1.1277
1.1865
1.3264
Ohio
.0000
.1869
.4982
.8633
1.2161
1. 5041
1.6944
1.7752
1. 7535
. 1. 6512
1. 4985
1.3283
1.1699
1. 0457
.9689
.9428
.9624
1. 0163
1. 0902
1. H?89
1.2074
===========================================================
75
equations
in
post-sample
Table
6)
are
forecasting
re-estimated in a sequential
format
(see
Burt,
Townsend and
LaFrance, 1986).
For each
year at a
equation, the
time,
then
forecasts computed
the
sample period
equation
re-estimated and
up until the end of the original sample
period (in this case it is 1983).
reduction of
is
is reduced one
The step-by-step yearly
the sample period is done for up to 13 years,
thus back until 1970, as
is
also
done
in
Burt's (1986}
study.
These post-sample forecasts are presented in Tables 915.
Although sample period reductions
13 years,
only post-sample
ahead are presented
prediction errors
value
of
the
in
the
for up to
forecasts for up to five years
tables.
Asterisks indicate
that are greater than twice the absolute
standard
approximate 95%
are done
error,
thus
lying
confidence interval.
that these "predictions are
not
ex
outside
an
Burt (1986) mentions
ante
in
a realistic
sense because known rents are used, but the primary purpose
is to detect specification error" (p. 16).
Glancing at
estimated
similarity
land
the
price
arises.
post-sample
equations,
This
point
forecasts
for
one
major
is
that
the six
point
of
asterisked
prediction errors
follow two
main diagonal paths, one for
the year 1975 and
the
for
other
the
year
indicates that 1975 and 1976 are outlier years.
1976.
As
This
76
Table 9:
Post sample forecasts of final estimated land
price model equations (classic disturbance) for
Illinois (Burt's land price data)
===========================================================
Number of Years Beyond the Sample
Sample
Period
2
1
3
4
5
----------------------------~------------------------------
1960-82
-.052
(.047)
1960-81
-.048
(.049)
-.090
( .061)
1960-80
.025
(.050)
-.028
( .065)
-.064
( .080)
1960-79
.001
(.052)
.026
(.067)
-.027
( .085)
-.063
(.102)
1960-78
.029
(.055)
.024
( .070)
.052
( .087)
.001
(.103)
-.036
(.116)
1960-77
.007
( .060)
.035
(.079)
.031
(.095)
.059
(.110)
.007
(.122)
1960-76
.148*
(.048)
.145*
(.062)
.190*
(.074)
.179*
(.082)
.186*
(.088)
1960-75
-.016
( .053)
.131
(.074)
.126
(.093)
.171
(.104)
.162
(.106)
1960-74
-.115*
(.035)
-.135*
(.049)
-.021
( .065)
-.037
( . 077)
.021
( .083)
1960-73
-.071*
(.033)
-.165*
(.039)
-.183*
(.048)
-.059
( .058)
-.061
(.065)
1960-72
-.050
(.029)
-.113*
(.039)
-.196*
(.041)
-.198*
(.048)
-.060
(.055)
1960-71
.055*
( .023)
-.006*
(.030)
-.091*
(.036)
-.179*
(.033)
-.205*
( .035)
1960-70
.001
(.027)
.054
(.035)
-.007
(.043)
-.072
(.045)
-.180*
(.036)
RMSE
Mean
.064
-.007
.096
-.010
.112
-.015
.122
-.020
.127
-.018
===========================================================
77
Table·10:
Post sample forecasts of final estimated land
price model (classic disturbance) for Illinois
(land price index data)
===========================================================
Number of Years Beyond the Sample
Sample
Period
1
2
3
4
5
1960-82
-.055
(.048)
1960-81
-.052
( .049)
-.096
(.061)
1960-80
.025
( .051)
-.031
(.065)
-.070
(.081)
1960-79
.001
( .052)
.026
(.068)
-.031
(.086)
-.069
(.103)
1960-78
.031
(.055)
.025
(.070)
.054
(.087)
-.001
(.103)
-.040
(.117)
1960-77
.008
( .060)
.038
( .078)
.033
(.095)
.062
(.109)
1960-76
.156*
( . 046)
.153*
(.059)
.202*
(.071)
.191*
( . 07 8)
.007
(.122)
.199*
(.084)
1960-75
-.011
( .051)
.144*
( .070)
.140
(.088)
.188
(.098)
.179
(.100)
1960-74
-.108*
(.035)
-.065
( .034)
-.122*
( .050)
.003
(.065)
-.012
(.077)
.047
(.082)
-.155*
( .040)
-.169*
(.050)
-.035
( .060)
-.038
(.067)
1960-72
-.048
( .030)
-.106*
( .041)
-.184*
(.043)
-.183*
( .050)
-.037
(.058)
1960-71
.058*
(.024)
.000
(.031)
-.061
(.037)
-.166*
( .034)
-.189*
(.037)
1960-70
.001
( .028)
.059
( .036)
.001
(.045)
-.060
( .047)
-.165*
( .038)
RMSE
Mean
.064
-.005
.096
-.005
.111
-.007
.121
-.009
.125
-.004
1960-73
===========================================================
78
Table 11:
Post sample forecasts of final estimated land
price model (classic disturbance) for Iowa
----------------------------------------------------------Number of Years Beyond the Sample
Sample
Period
1
2
3
4
5
1960-82
-.097
(.058)
1960-81
-.082
( .059)
-.175*
(.098)
1960-80
.021
(.063)
-.061
(.087)
-.146
(.116)
1960-79
.078
( .060)
.099
(.083)
.046
(.112)
-.011
(.143)
1960-78
.022
(.060)
.096
(.081)
.122
(.107)
.073
( .136)
.017
(.164)
1960-77
-.045
(.066)
-.015
(.088)
.056
(.112)
.084
(.136)
.042
(.162)
1960-76
.141*
(.058)
.086
( .078)
.122
(.094)
.135
( .109)
.164
(.121)
1960-75
-.036
(.063)
.036
(.120)
. 076·
(.132)
.141
(.135)
1960-74
-.136*
(.033)
.098
(.093)
..;..168*
(.041)
-.066
(.053)
-.130*
(.063)
-.066
(.068)
1960-73
-.056
(.034)
-.175*
(.039)
-.200*
(.045)
-.085
(.052)
-.136*
(.060)
1960-72
-.004
(.038)
-.060
(.050)
-.178*
(.052)
-.202*
(.051)
-.086
(.056)
1960-71
.080*
(.033)
.081
(.048)
.034
(.058)
-.103
'(.053)
-.157*
(.040)
1960-70
-.006
( .041)
.072
(.064)
.069
(.087)
.022
(.097)
-.113
(.083)
RMSE
Mean
.076
-.009
.110
-.010
.113
-.010
.112
-.010
.114
-.022
===========================================================
79
Table 12:
Post sample forecasts of final estimated land
price model (classic disturbance) for Indiana
====================~======================================
Number of Years Beyond the Sample
Sample
Period
1
2
1960-82
-.130*
(.057)
1960-81
-.141*
(.049)
-.228*
(.061)
1960-80
-.014
(.050)
-.151*
(.062)
1960-79
.008
(.053)
1960-78
3
4
5
-.009
(.064)
-.240*
(.078)
-.145
(.079)
-.233*
( .095)
-.042
(.057)
-.024
( .072)
-.045
( . 086)
-.184
(.103)
-.271
( .117)
1960-77
-.077
(.061)
-.113
(.083)
-.110
( .105)
-.137
(.122)
-.275
(.137)
1960-76
.016
(.064)
-.061
(.092)
-.092
(.121)
-.086
(.145)
-.113
(.161)
1960-75
-.096
(.058)
-.086
(.084)
-.192
(.113)
-.244
( .141)
-.243
(.160)
1960-74
-.128*
( .049)
-.218*
(.066)
1960-73
-.140*
(.038)
-.273*
(.051)
-.234*
(. 086)
-.380*
(.062)
-.391*
(.119)
-.493*
(.081)
1960-72
.028
(.035)
-.238*
(.069)
1960-71
.053
(.033)
-.109*
(.056)
.084
(.050)
-.348*
(.105)
-.399*
(.073)
-.348*
( .076)
-.036
(.072)
-.167
(.082)
-.292*
(.080)
1960-70
-.085*
(.032)
.170*
(.055)
.249*
(.082)
.162
(.104)
(.104)
.088
-.024
.150
-.085
.204
-.133
.251
-.198
.306
-.270
RMSE
Mean
-.371*
(.082)
.016
===========================================================
80
Table 13:
Post sample forecasts of final estimated land
price model (AR(1) disturbance) for Indiana
===========================================================
Number of Years Beyond the Sample
Sample
Period
1
2
3
4
5
1960-82
-.070
( .055)
1960-81
-.135*
(.046)
-.161*
(.062)
1960-80
-.016
( .047)
-.146*
( .058)
-.174*
( .078)
1960-79
.024
(.049)
-.001
(.058)
-.127
( .071)
-.150
( .088)
1960-78
-.017
(.053)
.012
(
~063)
-.014
( . 07 4)
-.141
( .089)
-.167
(.106)
1960-77
-.084
( .055)
-.082
(.073)
-.062
(.090)
-.091
(.101)
-.215
(.115)
1960-76
.064
( .053)
-.030
( .070)
-.010
(.083)
.016
(.093)
-.017
(.098)
1960-75
-.030
(.059)
.030
( .082)
-.068
(.107)
-.056
(.127)
-.031
(.139)
1960-74
-.150*
( .055)
-.259*
( .082)
-.293*
(.113)
-.410*
(.133)
-.467*
(.162)
1960-73
-.149*
(.039)-
-.278*
( .066)
-.380*
(.102)
-.394*
(.133)
-.485*
( .142)
1960-72
.025
(.037)
-.118
(.060)
-.259*
(.082)
-.378*
(.109)
-.408*
(.136)
1960~71
.091*
( .034)
.160*
(.056)
.073
(.081)
-.129*
(.092)
-.346*
( .095)
1960-70
-.081*
(.031)
.206*
(.059)
.324*
(.094)
.273*
(.125)
.080
(.120)
RMSE
Mean
.086
-.028
.153
-.056
.205
-.090
.247
-.146
.302
-.228
===========================================================
81
Table 14:
Post sample forecasts of final estimated land
price model (classic disturbance) for Ohio
===========================================================
Number of Years Beyond the Sample
Sample
Period
1
2
3
4
5
1960-82
-.150
(.076)
1960-81
-.236*
(.058)
-.327*
( . 07 4)
1960-80
-.162
(.044)
-.365*
(.060)
-.503*
(.081)
1960-79
-.049
(.046)
-.200*
( .058)
-.419*
(.083)
-.573*
( .112)
1960-78
-.010
(.050)
.057
( .065)
-.209*
(.082)
-.432*
(.115)
-.588*
(.148)
1960-77
-.044
(.058)
-.055
( . 083)
-.113
(.110)
-.272*
( .134)
-.506*
(.174)
1960-76
.108*
(.044)
.075
( .059)
.095
(.073)
.057
(.087)
-.093
( .100)
1960-75
.009
(.054)
.118*
(.091)
.088
(.095)
.110
(.113)
.073
(.128)
1960-74
-.176*
(.031)
-.239*
(.049)
-.231*
(.072)
-.338*
( .093)
-.356*
(.108)
1960-73
-.073*
(.024)
-.261*
(.035)
-.342*
(.048)
-.344*
(.061)
-.444*
(.071)
1960-72
.007
(.022)
-.066
(.036)
-.253*
(.047)
-.334*
( .055)
-.337*
(.066)
1960-71
.022
(.022)
.028
(.033)
-.042
(.047)
-.234*
( .053)
-.325*
(.056)
1960-70
.009
(.029)
.033
(.044)
.042
(.064)
-.026
( . 07 8)
-.222*
( . 07 6)
RMSE
Mean
.110
-.057
.191
-.110
.260
-.172
.317
-.170
.366
-.311
===========================================================
82
Table 15:
Post sample forecasts of final estimated land
price model (AR(l) disturbance) for Ohio
----------------------------------------------------------Number of Years Beyond the Sample
Sample
Period
1
2
3
4
5
1960-82
.014
(.064)
1960-81
-.178*
(.058)
-.191*
(.081)
1960-80
-.153*
(.045)
-.324*
( .070)
-.417*
( .113)
1960-79
-.047
(.047)
-.189*
( .061)
-.378*
(.096)
-.488*
(.148)
1960-78
-.003
(.051)
-.049
( .066)
-.192*
(.083)
-:-.383*
(.126)
-.495*
(.178)
1960-77
-.083
(.059)
-.080
(.088)
-.145
( .118)
-.291
(.149)
-.479*
1960-76
.105*
(.039)
.025
(.055)
.059
(.062)
.011
( . 07 5)
-.123
( .085)
1960-75
.054
( .037)
.140*
(.042)
.054
(.062)
.094
( . 066)
.047
( . 07 9)
1960-74
-.149*
(.034)
-.156*
( . 068)
-.120
(.095)
-.226
-.214
(.138)
-.082*
( . 026)
-.301*
( .044)
-.447*
(.088)
-.476*
(.113)
-.574*
.006
(.025)
-.075
( .041)
-.293*
(.058)
-.441*
( .096)
-.472*
1960-71
.021
(.025)
.028
(.039)
-.048
(.057)
-.265*
( .068)
-.413*
(.098)
1960-70
.015
(.035)
.041
(.057)
.056
(.084)
-.017
(.107)
-.243*
(.106)
RMSE
Mean
.091
-.037
.166
-.094
.250
-.170
.320
-.248
.383
-.330
1960-73
1960-72
( .113)
( • 206)
(.120)
(.119)
===========================================================
83
previously mentioned, this is due to a combination
increases in
oil prices
and Russia
stocks, creating an aberration
form
of
extremely
high
u.s.
in
grain
buying out
prices
of high
u.s.
grain
agriculture in the
translating
into
extremely high farm income for several years after 1974.
These two outlier years,
period,
throw
off
if
forecasts
shown especially in the
included
in
the sample
for later periods.
two sets
This is
of post-sample forecasts
for Ohio, where there is a group of prediction errors lying
outside the approximate
years
beyond
1975
95%
and
confidence
1976.
interval
for the
Burt's (1986) post-sample
fo·recasts also produce the 1975-76 outliers.
Another point to be noted
sample
forecasts
increases as the
is
that
number
sample increases.
these
sets
of post-
root mean squared error (RMSE)
of
This
from
years
forecasted
beyond the
indicates that the estimated land
price equations do better with short term rather
than long
term forecasting as was hypothesized earlier in the chapter
of this study.
arrived at
long-run
greater
This is also consistent with the conclusion
previously in
elasticities
than
one,
this section.
of
these
these
Because computed
estimated
equations
will
equations are
not
do
well
forecasting for the long run.
This concludes the presentation and discussion
empirical
results
for
this
study.
A
summary
of the
of the
84
empirical results along
with
conclusions
and suggestions
for further study are presented in the following chapter.
85
CHAPTER 5
SUMMARY AND CONCLUSIONS
Three
alternative
measures
of
net
returns imputed to
farmland were explored in this study.
The
three measures
were returns estimated from aggregate accounting data, cash
rents, and gross revenue from production
(in this
case, corn
measure was
framework
model.
for grain
estimated within
similar
that
to
farml~nd
prices
measure with
(Burt,
which
and soybeans).
a second
of
the
1986)
Burt's
was
(1986) land price
rents in explaining
used
performance
Each rent
order rational lag
of crop-share
The performance
of dominant. crops
as
of
a reference
each alternative
returns measure was compared against.
Returns
did
not
to· farmland
do
very
Regression results
computed from USDA accounting data
well
in
explaining
suggested that
returns and land prices were of
farmland
prices.
any correlation between
a spurious
nature.
Since
net returns imputed to farmland were computed indirectly by
removal of returns to
to
total
assets,
it
non-real estate
assets from returns
was thought that this procedure was
ineffective in computing the
true measure
of net
rent to
u.s.
data and
land.
Analysis was
initially done
for aggregate
was to proceed down to individual state data.
If
86
analysis at
the state level (initial point being Illinois)
produced similar
level, one
regression
could conclude
results
that the
computing net rent to land was
the
data
that
was
as
needed
at
indirect procedure of
not useful.
for
the national
the
Unfortunately
state level was not
published.
Cash rents were the
rents
explored.
jointly
second
It
dependent
was
with
procedure
was
done
framework,
specifically
alternative
thought
land
that
prices,
within
an
two
stage
measure
cash
so
of net
rents are
the estimation
instrumental
least
variable
squares.
But
regression results using this approach were implausible.
Intuitively, cash
rents would
variable for farmland prices
by
the
renegotiation
through
whenever
conditions)
usually
contracts
are
up
to
for
due to
process
change
needed,
is
for
a
not be a good explanatory
three
that
(in
and
their rigidity caused
both parties must go
response
to
new market
that cash rent contracts are
years
while
year-to-year.
crop-share
rent
Farmland prices would
respond to new economic conditions a
lot faster
than cash
rents.
Cash
land
rents
prices,
current
and
were
and
lagged
then
also
cash
both models suggested that
to
a
"third"
variable
regressed
land
on current and lagged
prices
rents.
were
regressed
on
Regression results from
both variables
were responding
(or set of variables), namely the
87
lagged
that
values
the
of
crop-share
correlations
prices are
rents.
between
So it was concluded
cash
rents
and farmland
spurious, and that each are ultimately respond-
ing to changes in this "third" variable mentioned above.
The last alternative measure
from production.
(so that
explored was
gross revenue
Since analysis was initiated in Illinois
direct comparison
could be
made with crop-share
rents), the per acre value of production (yield x price) of
corn for grain and
returns
to
soybeans
were
farmland.
A
used
major
as
a
part
measure of
of
Illinois
agricultural land is used for crop production with corn and
soybeans being its dominant crops.
The
regression
results
measure of net rent.
difference
indicated some promise for this
The
equation
estimated coefficients
parameters
estimated using crop-share rents
coefficients
for
the
those in Burt's study.
from
corn
and
rent
This would mean
soybean
rents,
but
very close to those
(Burt, 1986).
variable
Estimated
were different from
that gross revenue
production gave a distributed lag
land price response pattern very
share
were
for the
general
close
conclusions
to
that
about
of croppredictive
performance could not be made.
The estimated intercept could not be used to
estim~ted
capitalization
rate
variable used a gross rather than
returns to
farmland.
compute the
of farmland, as the return
a net
Landlord costs
return measure of
imputed to farmland
88
that are proportional to gross returns were confounded with
the intercept term, and
thus prevented
the computation of
an implicit capitalization rate.
Analysis was
then broadened to include other surrounding
cornbelt states of Ohio, Iowa, and Indiana.
results
also
states.
Estimated coefficients for the difference equation
parameters
proved
for
the
to
be
three
encouraging
The regression
states
for these three
.were very close to the
Illinois analysis (using either gross revenue or crop-share
rents).
The
only
difference
structure on the estimated
was
that
equations for
an AR(l) error
both Indiana and
Ohio came out statistically significant. Regression results
for the four states proved to be consistent, with estimated
coefficients
for
varying much
across the
indicated that
the
computed
each
variable (including intercept) not
states.
the homogeneity
long-run
Regression results also
constraint was violated as
elasticities
were
considerably
greater than one.
Distributed
lag
response
paths
between Burt's estimated model
final
estimated
compared.
The
distributed lag
run
models
were all
gross
for
lag
greater than
somewhat
rents and
were then
similar
Computed cumulative long-
response
models
revenues
indicated
response paths.
distributed
elasticities)
using crop-share
using
comparison
of land price to rents
estimated
one, suggesting
(long-run
response
using gross revenues
that the estimated
89
equations
for
farmland
measure would only
be
price
useful
using
for
the
gross
revenue
conditional short-term
forecasting of land prices.
Finally,
a
set
of
post-sample forecasts were done for
these final estimated farmland
basically
indicated
that
price
the
forecast farmland prices better
equations.
Results
estimated
equations would
within
short-term time
a
framework as was alluded to in the previous paragraph.
The results
agricultural
properly
areas,
came
returns
measure
study indicate
gross
computed)
measure of
revenue
of this
to
revenues
close
to
land.
seemed
to
that in homogeneous
from
production (if
approximating
Analysis
have
order difference equation (with
with
verified
a good
the gross
Burt's second
complex roots)
land price
model.
If
a
gross
measure
of
revenue
produced near similar
regression results as when crop-share rents (a net measure)
were used, it would not be unrealistic to expect that a net
measure of
revenues imputed
to farmland
will improve the
fit produced by use of the gross measure.
Further
encompass
research
finding
farmland from
a
on
farmland
way
to
prices
subtract
would
costs
either
imputed to
gross revenues or to explore new alternative
measures of returns to farmland that is a net measure.
It seems that we
finding a
may
proper way
be
back
to impute
to
the
old
problem of
a net return to farmland.
90
But the need for a
good
set
of
farm
accounts
data for
deriving a net return measure seems to be more pressing.
91
REFERENCES CITED
92
Alston, Julian M. "An Analysis of Growth of
Prices:
1963-1982."
American
Agricultural Economics 68 (1986):1-9.
Barry, Peter
Estate. 11
u.s.
Farmland
Journal
of
J.
"Capital Asset Pricing and Farm Real
American Journal of Agricultural Economics
62 (1980):547-543.
Burt, Oscar P. "Econometric Modeling of the Capitalization
Formula for Farmland Prices."
American Journal of
Agricultural Economics 68 (1986):10-25.
Burt,
Oscar.
"Estimation· of Distributed Lags as
Nonstoi::hastic Difference Equations."
Staff Paper No.
80-1,
Department
of
Agricultural Economics and
Economics, Montana State University, Bozeman, Montana,
1980.
Burt,
Oscar,
S.
Townsend,
and
J.
T.
LaFrance.
"Instruction Manual for DYNEREG:
a Nonlinear Least
Squares Algorithm for Distributed Lag Models and/or
Regression Models with Time Series Error Terms."
Staff Paper No. 86-4,. Department of Agricultural
Economics and Economics, Montana State University,
Bozeman, Montana, August, 1986.
Castle, Emery N. and Irving Hoch. 11 Farm Real Estate Price
Components,
1920-1978. 11
American
Journal
of
Agricultural Economics 64 (1982):8-18.
Darby, Michael R. 11 The u.s. Productivity Slowdown: A Case
of Statistical Myopia."
American Economic Review 74
(1984):301-322.
Doll, John P. and Richard Widdows. "Capital Gains Versus
Current Income in the Farming Sector:
Comment."
American
Journal
of
Agricultural
Economics 63
(1981):729-733.
Doll,
John P., Richard Widdows, and
Paul D. Velde.
Research REview.
The Value of Agricultural Land in
the United
States:
A
Report
on Research."
Agricultural Economics Research 35 (1983):39-44.
11
Ezekiel, Mordecai.
"The Cobweb Theorem. 11
Journal of Economics (1938):255-280.
Quarterly
Feldstein, Martin. "Inflation, Portfolio Choice, and the
Prices of Land and Corporate Stock." American Journal
of Agricultural Economics 62 (1980):910-916.
93
Fisher, Irving.
The Theory of Interest: As Determined by
Impatience to Spend Income and Opportunity to Invest
It. (reprint) New York: A. M. Kelley, 1967.
Harvey, A.
York:
c.
The Econometric Analysis of Time Series.
John Wiley and Sons, 1981.
New
Herdt, Robert W., and Willard W. Cochrane.
"Farmland
Prices and Technological Advance." Journal of Farm
Economics 48 (1966):243-263.
Hirshleifer, Jack. Price Theory and Applications. 3rd ed.
Englewood Cliffs, New Jersey: Prentice-Hall, Inc.,
1984.
Hottel, J. B., and c. D. Evans.
"Returns to Equity Capital
in the U.S. Farm Production Sector." Balance Sheet of
~t~h~e~~F~a~rm~i~n~g~;~S~e~c~t~o~r~,~~l~9~7~9~~S~u~p~p_l~e~m~e~n~t.
Washington,
D.C.: USDA ESCS Agr. Info. Bull~ no. 430, 1979.
Jorgensen, Dale N.
"Rational Distributed Lag Functions."
Econometricia 32 ( 1966): 135-148,.
Kay, Donald.
Farm Management:
Planning, Control, and
Implementation. New York: McGraw-Hill, Inc., 1981.
Klinefelter, Danny
in Illinois."
33.
A.
"Factors Affecting Farmland Values
Illinois Agr. Econ. Res. 13 (1973):27-
Kmenta,
Jan.
Elements
of Econometrics.
Macmillian Publishing Co., Inc., 1986.
New
York:
Melichar, Emanuel. "Capital Gains versus Current Income in
the Farming Sector." American Journal of Agricultural
Economics 61 (1979):1085-1092.
Phipps, Tim T.
"Land
American
Journal
(1984):422-429.
Prices and Farm-Based Returns."
of
Agricultural
Economics 66
Plaxico, James S. "Implications of Divergences in Sources
of Resurns in Agriculture."
American Journal of
Agricultural Economics 68 (1979):1098-1102.
Pope, Rulon D., Randall A. Kramer, Richard D. Green, and B.
Delworth Gardner.
"An Evaluation of Econometric
Models of U.S. Farmland Prices."
Western Journal of
Agricultural Economics 4 (1979):107-119.
94
Reinsel, Robert D. and Edward I. Reinsel. "The Economics
of Asset Values and Current Income in Farming."
American
Journal
of
Agricultural
Economics 61
(1979):1093-1097.
Reiss, Franklin J.
"Landlord
and
Tenant Shares."
University of Illinois Department of Agriculture AERR
no. 102, 1969.
Reynolds, John E. and John F. Timmons. "Factors Affecting
Farmland Values in the United States." Iowa Agr. Exp.
Sta. Res. Bull.· No. 566, 1969.
Robison, Lindon J., and John R. Brake.
"Inflation, Cash
Flows, and Growth:
Some Implications for the Farm
Firm."
Southern Journal of Agricultural Economics
(1980):131-137 .
"Rational Expectations, the Real Rate
. Sargent, Thomas J.
of Interest, and the Natural Rate of Unemployment."
Brookings Paper on Economic Activity (1973):429-480.
Scott, John
T., Jr.
"Factors Affecting Land Price
Decline." American Journal of Agricultural Economics
65 (1983):796-800.
Shalit, Haim, and Andrew Schmitz. "Farmland Accumulation
and Prices."
American
Journal of· Agricultural
Economics 64 (1982):710-719.
Tanzi,
Vito.
"Inflationary
Expectations,
Activity, Taxes
and Interest
Rates."
Economic Review 70 (1980):12-21.
Theil, Henri. Principles of Econometrics.
Wiley and Sons, 1971.
Economic
American
New York:
John
Tweeten, L. G., and J. E. Martin.
"A Methodology for
Predicting U.S. Farm Real Estate Price Variation."
Journal of Farm Economics 48 (1966):378-393.
U.S. Congress, Joint Economic Committee.
of the President.
Washington:
Printing Office (various issues).
Economic Reports
U.S. Government
U.S. Department of Agriculture.
Agricultural Statistics.
Washington:
u.s. Government Printing Office (various
issues).
95
.s.
Department of Agriculture,
Crop Reporting Board
(Statistical Reporting Service).
Crop Production
(Annual Summary):
Acreage,
Yield, Production.
Washington:
U.S. Government Printing Office (various
issues).
Crop Values:
Season Average Prices Received
by Farmers and Value of Production. Washington: u.s.
Government Printing Off~ce (various issues).
Field Crops: Production, Disposition, Value.
Washington:
u.s. Government Printing Office (various
issues) .
.s.
Department of Agriculture, Economic Research Service.
Economic Indicators of the Farm Sector: National
Financial Summary, 1984. Washington: U.S. Government
Printing Office, 1985.
Economic Indicators of the Farm Sector: State
Income
and
Balance
Sheet
Statistics,
1984.
Washington: u.s. Government Printing Office, 1985.
Farm Real Estate Market Development: Outlook
and Situation. Washington: U.S. Government Printing
Office (various issues).
Farm Real Estate Taxes. Washington:
Government Printing Office (various issues).
u.s.
ttts, Myles, and James Johnson.
The Relationship of
Inflation and Agricultural Income, Asset Values, and
Firm Financial
Analysis."
Staff
Paper 85-6,
Department of Agricultural Economics and Economics,
Montana State University, Bozeman, Montana, 1985.
96
APPENDIX
Original Data Set ·
97
Table 16:
Sample data used for exploring USDA accounting
data measure
===========================================================
Year
Deflator
(PCEI)
Deflator
(PCEI)
Ret. to Tot.
Agr. Assets
(million$)
Agr. Asset
Value
(million$)
---------------------------------------~------------------
1940
1941
1942
1943
1.944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
··1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
.141
.152
.168
.184
.194
.202
.220
.243
.257
.256
.262
.278
.284
.290
.291
.295
.301
.310
.316
.323
.329
.333
.339
.344
.350
.356
.367
.376
.393
.410
.429
.449
.467
.496
.548
.592
.626
.667
.716
.782
.866
.946
1. 000
1. 039
.152
.168
.184
.194
.202
.220
.243
.257
.256
.262
.278
.284
.290
.291
.295
.301
.310
.316
.323
.329
.333
.339
.344
.350
.356
.367
.376
.393
.410
.429
.449
.467
.496
.548
.592
.626
.667
.716
.782
.866
.946
1. 000
1. 039
1. 082
1733
3262
5428
5701
3989
4248
6831
7450
9493
5494
7133
8423
7378
5380
5325
4519
4763
5273
7625
5114
6249
7577
8210
8307
7504
10922
12291
10320
10443
12472
12834
13315
19136
36259
28631
27511
22472
22401
31834
38733
28979
39869
37352
27576
42900
49600
59100
67800
74900
81800
91700
100800
106400
105800
122400
136000
133800
130400
134100
137800
146300
154400
170200
172900
174700
182600
190300
197900
205500
221400
234100
246100
259300
270500
280200
303100
341400
418900
442300
510100
590400
656700
783700
918100
1003200
1005200
977800
956500
98
Table 16:
Continued
===========================================================
Year
Agr. Real
Estate Value
(million$)
Land in
Farms
(million ac. )
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
31700
35200
41000
45900
52100
58400
62600
65100
65500
75900
83600
85100
84100
87500
92400
99900
105600
114200
120100
121800
127500
133000
140900
149300
160000
169100
179000
187900
194200
201300
216400
241800
297100
327000
381100
453500
507700
600700
704200
779200
780200
745600
736100
1093
1109
1125
1142
1145
1148
1152
1155
1202
1203
1204
1205
1206
1201
1197
1191
1184
1181
1174
1166
1157
1149
1144
1137
1130
1121
1113
1106
1102
1096
1092
1087
1084
1059
1054
1047
1044
1042
1038
1034
1027
1024
1019
Index of
Land Values
----------------------------------------------------------28900
1077
.0750
.0750
.0821
.0893
.1000
.1107
.1250
.1392
.1535
.1571
.1535
.1749
.1964
.1964
.1892
.2035
.2035
.2178
.2321
.2535
.2428
.2642
.27~5
.2749
.2927
.3070
.3356
.3570
.3820
.4034
.4200
.4300
.5355
.5300
.6600
.7500
.8600
] . 0000
1. 0900
1.2500
1.4500
1.5800
1.5700
1.4600
99
Table 17:
Sample data for Illinois used for exploring cash
rent and gross revenue measures
===========================================================
Year
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
Land
Price
(Burt's)
($/acre)
511
551
551
535
534
580
607
646
726
767
797
833
823
825
882
981
1296
1548
1906
2555
2807
3162
3411
3605
3303
2966
Statewide
Index of
Land Val.
($/acre)
18.6
20.1
20.1
19.5
20.2
21.2
22 .·2
23.7
26.7
28.3
29.4
30.8
30.4
30.5
32.7
36.6
49.0
59.1
73.5
100.0
110.4
125.1
135.5
143.6
131.0
117.0
Real
Estate
Tax
($/acre)
3.79
3.91
4.03
4.16
4.30
4.44
4.69
4.83
5.29
5.69
6.60
7.03
7.07
7.83
8.30
8.82
9.10
9.34
10.15
10.96
11.75
12.61
13.02
14.08
13.73
13.55
Gross Rev.
from corn
& Soybeans
($/acre)
Direct
Gov't
Payments
($/acre)
135.05
123.28
125.30
149.57
159.13
167.84
156.95
186.63
175.33
180.43
176.02
195.37
191. 28
217.51
329.12
455.88
406.42
480.92
461.63
447.83
470.14
580.90
564.85
557.55
575.35
491.77
3.11
1. 31
1. 20
7.74
7.89
7.23
8.78
8.85
7.43
5.81
9.61
11.89
9.95
10.27
14.56
7.48
.53
2.04
.51
1. 61
5.04
1. 60
1.73
2.39
5.72
33.06
===========================================================
100
Table 17:
Continued
===========================================================
Year
Cropshare
rent
($/acre)
Cash
rent
($/acre)
Product ion
expense
($/acre)
Gross Inc.
from farm
ex. dwell.
($/acre)
Index of
pr pd by
farmers
($/acre)
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
17
17
21
23
26
29
27
30
33
29
34
30
33
34
48
85
107
80
103
89
95
110
108
93
90
102
20.04
19.44
20.53
20.75
20.64
22.16
22.87
24.33
27.79
29.69
33.44
34.47
35.56
36.71
36.57
39.38
48.72
56.47
68.04
81.00
85.00
92.00
99.00
105.80
112.80
111.40
122.90
117.15
107.58
114.75
121.27
116.10
111.73
103.19
114.28
114.51
120.04
125.52
126.11
123.08
118.96
115.34
155.65
163.81
188.64
200.96
196.75
206.66
234.26
248.61
242.30
236.15
165.53
139.42
138.02
161.21
171.50
167.62
162.02
159.49
184.61
163.43
171.85
184.16
180.91
172.23
216.13
275.45
316.89
278.86
327.83
291.76
330.56
341.22
389.84
374.06
372.56
371.07
.8700
.9300
.9200
.9300
.9400
.9500
.9400
1.0000
1.0000
1.0000
1.0000
1. 0400
1.0800
1.1300
1. 2100
1. 4600
1. 6600
1. 8200
1.9352
1.9950
2.1546
2.4938
2.7531
2.9526
2.9925
3.0524
-----------------------------------------------------------
===========================================================
101
Table 18:
Sample data for Iowa
===========================================================
Year
Deflator
Inflation
Rate Plus
Unity
Gross
Revenue
($/acre)
Statewide
Real
Index of
Estate
J. . and Price
Tax
($/acre)
($/acre)
----------------------------------------------------------1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
.692
'. 706
.719
.726
.737
.748
.759
.772
.794
.814
.846
.884
.925
.965
1.000
1. 057
1.164
1. 253
1. 317
1.393
1.491
1.625
1.790
1.945
2.060
2.136
1. 0206
1.0202
1. 0184
1.0097
1.0152
1. 0149
1. 0147
1. 0171
1.0285
1. 0252
1.0393
1.0449
1.0464
1. 0432
1.0363
1. 0570
1.1012
1.0765
1.0511
1. 0577
1.0704
1.0899
1.1015
1. 0866
1. 0591
1.0369
124.85
117.12
115.91
146.52
147.24
159.18
160.14
160.52
183.49
158.13
177.59
186.66
199.15
205.86
362.04
468.16
415.68
398.06
405.10
381.30
502.28
538.72
637.05
541. 77
545.34
541.82
66
71
73
69
72
73
76
79
89
100
105
111
114
114
122
141
189
234
294
397
413
475
553
597
553
481
2.65
2.86
3.06
3.23
3.37
3.54
3.71
3.80
4.24
4.17
4.63
5.53
5.87
5.89
5.61
5.55
5.67
6.40
7.57
8.02
8.39
8.96
9.85
10.32
8.63
8.84
===========================================================
102
Table 19:
Sample data for Indiana
===========================================================
Year
Deflator
Inflation
Rate Plus
Unity
Gross
Revenue
($/acre)
Statewide
Real
Index of
Estate
Land Price
Tax
($/acre)
($/acre)
-----------------------------------------------------------
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
.692
.706
.719
.726
.737
.748
.759
.772
.794
.814
.846
.884
.925
.965
1.000
1.057
1.164
1. 253
1.317
1. 393
1.491
1.625
1.790
1.945
2.060
2.136
1.0206
1. 0202
1. 0184
1.0097
1.0152
1. 0149
1. 0147
1. 0171
1.0285
1.0252
1.0393
1.0449
1.0464
1.0432
1. 0363
1.0570
1.1012
1.0765
1. 0511
1. 0577
1. 0704
1. 0899
1.1015
1.0866
1.0591
1. 0369
126.73
117.25
124.28
138.48
151.34
162.11
142.68
174.72
167.96
141.94
166.05
184.83
191.71
200.02
287.61
436.20
391.38
409.87
458.80
407.77
459.65
513.12
594.00
469.87
538.82
487.66
64
67
69
66
68
71
76
80
92
100
106
106
104
109
113
131
161
200
244
321
361
415
481
517
449
391
2.09
2.33
2.42
2.48
3.01
3.06
3.20
3.41
3.75
4.17
4.51
4.97
5.43
5. 93.
5.90
6.06
5.01
5.03
5.09
.5 .17
5.34
6.66
7.43
7.80
8.05
8.53
===========================================================
103
Table 20:
Sample data for Ohio
·===========================================================
Year
Deflator
Inflation
Rate Plus
Unity
Gross
Revenue
($/acre)
Statewide
Real
Index of
Estate
Land Price
Tax
($/acre)
($/acre)
-----------------------------------------------------------
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
.692
.706
.719
.726
.737
.748
.759
.772
.794
.814
.846
.884
.925
.965
1.000
1.057
1.164
1.253
1.317
1. 393
1. 491
1.625
1.790
1. 945
2.060
2.136
1.0206
1.0202
1. 0184
1.0097
1. 0152
1.0149
1.0147
1. 0171
1. 0285
1.0252
1.0393
1. 0449
1. 0464
1. 0432
1.0363
1.0570
1.1012
1.0765
1.0511
1.0577
1.0704
1.0899
1.1015
1. 0866
1. 0591
1. 0369
127.96
117.85
119.37
139.22
138.05
143.93
132.65
150.03
184.50
134.47
163.08
168.74
188.38
190.40
258.92
348.94
388.71
397.50
446.97
417.66
459.81
524.87
648.30
414.21
514.17
515.52
69
72
73
72
75
77
83
86
93
100
106
110
115
120
127
147
184
208
252
331
373
448
513
526
450
398
1.92
2.05
2.21
2.32
2.44
2.58
2.75
2.91
2.98
3.08
3.32
3.68
4.31
4.10
4.68
4.96
5.42
5.79
6.49
7.31
7.78
8.27
8.35
8.33
8.35
8.51
===========================================================