Wheat Pricing: Information on Transition to New Tests for Protein (Letter
Report, 12/08/94, GAO/RCED-95-28).
--------------------------- Indexing Terms -----------------------------
REPORTNUM: RCED-95-28
TITLE: Wheat Pricing: Information on Transition to New Tests for
Protein
DATE: 12/08/94
SUBJECT: Grain inspection
Standards evaluation
Test equipment
Grain and grain products
Food supply
Commodity sales
Technology transfer
Testing
Price adjustments
**************************************************************************
* This file contains an ASCII representation of the text of a GAO *
* report. Delineations within the text indicating chapter titles, *
* headings, and bullets are preserved. Major divisions and subdivisions *
* of the text, such as Chapters, Sections, and Appendixes, are *
* identified by double and single lines. The numbers on the right end *
* of these lines indicate the position of each of the subsections in the *
* document outline. These numbers do NOT correspond with the page *
* numbers of the printed product. *
* *
* No attempt has been made to display graphic images, although figure *
* captions are reproduced. Tables are included, but may not resemble *
* those in the printed version. *
* *
* A printed copy of this report may be obtained from the GAO Document *
* Distribution Facility by calling (202) 512-6000, by faxing your *
* request to (301) 258-4066, or by writing to P.O. Box 6015, *
* Gaithersburg, MD 20884-6015. We are unable to accept electronic orders *
* for printed documents at this time. *
**************************************************************************
Cover
================================================================ COVER
Report to Congressional Requesters
December 1994
WHEAT PRICING - INFORMATION ON
TRANSITION TO NEW TESTS FOR
PROTEIN
GAO/RCED-95-28
Wheat Pricing
(150910)
Abbreviations
=============================================================== ABBREV
EGIS - Export Grain Inspection System
FGIS - Federal Grain Inspection Service
GAO - General Accounting Office
HRS - hard red spring
HRW - hard red winter
NIRR - Near Infrared Reflectance
NIRT - Near Infrared Transmittance
USDA - U.S. Department of Agriculture
Letter
=============================================================== LETTER
B-258389
December 8, 1994
Congressional Requesters
Protein levels in wheat are an important factor in determining hard
red spring (HRS) wheat prices, particularly for HRS wheat grown in
Minnesota, Montana, North Dakota, and South Dakota. Higher protein
commands higher prices in the market. Therefore, the accuracy and
reliability of protein testing is of primary importance to these
areas and to those who buy and sell high-protein wheat. In 1993,
concerns were raised that a new technology for estimating the protein
levels of wheat--the Near Infrared Transmittance (NIRT)
technology--was producing estimates that were lower than those
provided by an older technology. This new technology was introduced
by the Federal Grain Inspection Service (FGIS)--an agency in the U.S.
Department of Agriculture (USDA) that provides official inspections
of grain. Inspections by laboratories other than those supervised by
FGIS are known as unofficial inspections. While official inspections
must meet FGIS' standards and are used for both domestic and export
sales, they are generally required for export sales. In contrast,
unofficial inspections are not subject to FGIS' standards.
Unofficial inspectors can range from "in-house" graders at grain
elevators and processing plants to third-party inspection agencies.
Because of the above concerns, you asked us to (1) describe the
pricing situation for wheat in 1993, (2) evaluate FGIS' introduction
of the NIRT technology, (3) analyze the economic impact of the NIRT
technology on segments of the industry, and (4) describe recent
efforts to standardize unofficial protein testing of wheat.
RESULTS IN BRIEF
------------------------------------------------------------ Letter :1
Prices for wheat with high amounts of protein were at record levels
in 1993. During the previous 5 years, these prices ranged from $3.04
per bushel to $5.18 per bushel in the Minneapolis market--a major
market for HRS wheat. In comparison, in 1993, prices ranged from
$4.58 to $7.19 per bushel for similar wheat in the same market.
High-protein wheat received record prices in 1993 largely because
supplies were low at the beginning of the year and unusually severe
weather conditions during the 1993 growing season further decreased
supplies.
FGIS took reasonable steps in introducing the NIRT technology for
official protein testing. However, difficulties commonly associated
with the transition to a new technology, as well as the unusual 1993
crop conditions that were not reflected in the initial calibration
for protein levels, led to concerns about FGIS' actions and eroded
confidence in the NIRT technology.
Our analysis showed that because the NIRT technology generally
provided lower protein readings on some damaged high-protein wheat,
the market reacted by raising the premiums for high-protein wheat.
The resulting higher premiums generally offset any losses that would
have occurred from the lower NIRT readings. We estimate that because
of the NIRT technology, price premiums rose 50 cents per bushel more,
on average, than they would have risen if the NIRT technology had not
been used. However, this overall conclusion does not discount the
possibility that some individual farmers and grain elevator operators
incurred losses.
The National Conference on Weights and Measures, in conjunction with
FGIS, has proposed standards for unofficial protein testing. These
standards would help promote greater uniformity in commercial grain
inspection. Even though the standards are not yet enforceable, they
can be used by manufacturers as guidelines in the design of
protein-testing equipment. The adoption of these standards would
help to make unofficial protein readings more consistent.
BACKGROUND
------------------------------------------------------------ Letter :2
Components of wheat quality include damage, protein levels, and the
test weight of wheat. HRS wheat, one of the six basic classes of
wheat, maintains the highest protein content--usually 13 percent or
above. This wheat, planted in the spring, grows primarily in the
North Central states of Minnesota, Montana, North Dakota, and South
Dakota. In 1993, USDA estimated that HRS wheat production in these
states represented approximately 90 percent of the HRS wheat grown in
the United States.
Federal or official protein testing is managed by FGIS, which (1)
approves equipment for the official inspection of grain, (2) operates
testing laboratories, (3) authorizes qualified state agencies and
private laboratories to inspect grain, and (4) provides federal
oversight of official grain inspection. In fiscal year 1993, FGIS
and other entities authorized to perform official protein inspections
conducted over 660,000 wheat protein inspections. FGIS performed
approximately 10 percent of these inspections.
For official protein testing, FGIS uses near-infrared spectroscopy,
which provides results more quickly than other technologies. FGIS
has two types of this technology: the Near Infrared Reflectance
(NIRR) technology and the NIRT technology. The NIRR technology
estimates protein by measuring the amount of infrared light that is
reflected from a portion of ground grain, while the NIRT technology
estimates protein by measuring the amount of infrared light that
passes through whole kernels of grain. Since NIRT instruments
analyze whole grain samples, they reduce operator errors, decrease
analysis time, and provide more consistent results. After using the
NIRR technology since 1978, FGIS began using the NIRT technology for
official protein testing in May 1993.
To calibrate the NIRR and NIRT technologies, FGIS uses a chemical
process, referred to as the reference technology, that measures,
rather than estimates, protein levels in wheat samples. This process
is costly and time-consuming, making it too impractical for routine
testing.
PRICES FOR HIGH-PROTEIN WHEAT
WERE AT RECORD LEVELS IN 1993
------------------------------------------------------------ Letter :3
The unusual protein-pricing situation in 1993 occurred in two phases
and was the result of short supplies of high-quality, high-protein
wheat. Initially, the short supply was due to quality problems
resulting from high moisture levels in the 1992 crop, which reduced
stocks of high-quality, high-protein wheat at the end of 1992. The
prices for this wheat, which had averaged $3.36 per bushel in
calendar year 1991, increased to an average of $4.63 per bushel in
1992. These short supplies continued into 1993, helping to keep
premium prices at a high level. The 1993 winter wheat harvest had
low levels of protein and/or other quality problems that further
decreased supplies. In addition, the 1993 HRS wheat harvest did not
replenish the stocks of high-quality wheat. Heavy moisture, combined
with cold temperatures, delayed the harvest and produced lower yields
and uneven quality. The overall quality of the crop varied in the
four North Central states, depending on rainfall levels. Farmers,
grain elevator operators, millers, and other representatives of the
grain industry told us that the 1993 crop was one of the worst they
had seen for quality.
Moreover, the shortage of high-protein wheat was perceived to be even
greater than was actually the case because the NIRT technology
underestimated the protein levels in some damaged high-protein wheat,
according to FGIS. Consequently, the price premium for high-protein
wheat increased further. For example, as figure 1 shows, the price
for 15-percent HRS wheat went from $4.58 per bushel in April 1993 to
$6.60 in October and to $7.19 by November.
Figure 1: Variations in
Protein Prices for HRS Wheat
for the Minneapolis Market,
October 1988-January 1994
(See figure in printed
edition.)
Source: USDA's Economic Research Service.
FGIS' PROCEDURES WERE GENERALLY
REASONABLE, BUT CERTAIN
CONDITIONS ERODED CONFIDENCE IN
TEST RESULTS
------------------------------------------------------------ Letter :4
FGIS took a systematic approach to evaluating the NIRT technology and
followed established procedures for implementing it. However,
several conditions during implementation, such as weather-related
damage to the grain, made the transition more difficult than
anticipated and led to concerns about FGIS' thoroughness in deciding
to implement the NIRT technology. (See app. I for the chronology of
the evaluation and implementation of the NIRT technology.)
FGIS TOOK A SYSTEMATIC
APPROACH TO INTRODUCING THE
NIRT TECHNOLOGY
---------------------------------------------------------- Letter :4.1
FGIS followed reasonable procedures in evaluating and introducing the
NIRT technology into the official protein testing system in 1993.
For its initial assessment of the technology, FGIS followed the broad
criteria cited in its Type Evaluation Handbook to determine whether
further evaluation was warranted. According to FGIS' documentation,
the NIRT technology met these broad criteria. This handbook also
provides performance standards for any testing technology and
prescribes procedures for evaluating and approving the technology.
After this phase, FGIS conducted more rigorous evaluations.
First, FGIS weighed the advantages and disadvantages of switching to
the NIRT technology for the agency's testing. In doing so, FGIS
determined that the NIRT technology's advantages outweighed the
benefits of continuing to use the NIRR technology. The NIRT
technology lowers labor costs because it requires fewer operating
steps and reduces the potential for errors by operators. As a
result, the NIRT technology's results on the same sample are more
consistent from instrument to instrument and, except in certain cases
when grain quality problems are unusual, should be more
representative of the actual protein levels than the NIRR
technology's results.
However, agency officials recognized that upgrading the equipment at
all of FGIS' official testing stations would be costly and would
cause some market disruptions. The disruptions would occur
principally because, for a period of time, official testing stations
could be using either the NIRR or the NIRT equipment, thus producing
different estimates.
Second, FGIS evaluated different approaches to introducing the new
technology into the market. After considering various implementation
alternatives, FGIS decided to replace NIRR equipment with NIRT
equipment at FGIS-operated official testing stations when the
implementation of the NIRT technology began. FGIS' goal was to
minimize the time during which both types of instruments were used
for official testing. In turn, this would minimize the exposure of
the wheat market system to different official testing results at
different locations. However, FGIS did not require other official
inspection sites to replace NIRR equipment with the NIRT equipment
but strongly encouraged them to do so as soon as funds became
available. By August 1994, all of the official locations that were
testing HRS wheat had switched to the NIRT technology.
Third, FGIS completed a pilot study at 16 official inspection
stations in 1992 to compare the results of testing conducted with
NIRR and NIRT instruments. For the 5-week study, both instruments at
each location analyzed HRS wheat market samples for protein levels.
The pilot study showed that in the field, the NIRT technology's
results were more consistent and, on average, closer to the reference
technology results than were the NIRR technology's.
FGIS planned to recalibrate the NIRT equipment periodically, as it
had done for the NIRR equipment. FGIS periodically recalibrates the
testing equipment with new wheat samples because different wheat
characteristics may be present in each new crop year (June 1 to May 1
of the following year). These characteristics need to be reflected
in the calibration data for continued accurate protein readings.
FGIS periodically discussed the development of the NIRT technology
with the agency's Advisory Committee. Members of the Committee
include farmers, university faculty, grain elevator managers,
representatives of mills and grain companies, and others involved in
the grain industry. FGIS kept the Committee apprised of the pilot
testing results. At its June 1992 meeting, the Committee recommended
that FGIS "continue moving toward NIRT technology as the standard
protein testing method, and continue to explore means of phasing in
the technology to minimize the impact on the industry."
FGIS also sought the advice of grain industry representatives in
determining the best time to implement the switch to the NIRT
technology. To do this, FGIS held a telephone conference with
industry officials from major grain companies and milling and trade
associations in February 1993. These officials generally agreed that
in order to minimize any impacts on the market, FGIS should introduce
the NIRT technology in the beginning of May, as it did, when there is
generally a low amount of wheat in the marketing system.
SEVERAL CIRCUMSTANCES
IMPEDED A SMOOTH TRANSITION
TO THE NIRT TECHNOLOGY
---------------------------------------------------------- Letter :4.2
As FGIS had expected, not all official non-FGIS inspection sites
began using the NIRT technology on the official implementation date.
Consequently, official results were provided on the basis of testing
conducted using two different technologies that did not necessarily
produce equivalent results. Those in the wheat market that purchased
wheat on the basis of one technology's test results and sold it on
the basis of another's saw, in some cases, differences in protein
results. These differences created disparities between expected and
actual prices.
However, these anticipated difficulties were made much worse by the
unanticipated damage to the HRS wheat crop in 1993. When FGIS
originally calibrated the NIRT equipment in 1991, it used market
samples from 1991 and prior years that did not represent some of the
unusual quality problems found in the 1993 crop. As a result, the
NIRT equipment was not producing accurate readings on some samples of
1993 HRS wheat. In January 1994, FGIS updated the original
calibration. This new calibration is based on samples from 1988-93
crop years representing a range of growing conditions and protein
levels. These samples also reflected the quality of the 1993 crop.
CONCERNS ERODED CONFIDENCE
IN FGIS AND THE NIRT
TECHNOLOGY
---------------------------------------------------------- Letter :4.3
Many local farm and grain elevator organizations and others in the
industry we spoke to raised concerns about FGIS' decision-making
process and the procedures the agency used to introduce the NIRT
technology. Although FGIS informed its Advisory Committee and other
grain industry representatives about its progress toward implementing
the NIRT technology, many farmers and grain elevator managers we
spoke with were not fully aware of FGIS' decisions. For these
individuals, an FGIS notice announcing the effective dates of the
changes may have been the only information they received on the NIRT
testing technology. Knowing little about FGIS' testing and
calibration efforts, many of these individuals had little confidence
in the validity of the NIRT technology or in FGIS' decision to use
the technology for protein testing. Table 1 shows the market's
concerns about FGIS' actions and GAO's findings related to those
actions.
Table 1
Wheat Market's Concerns About the
Introduction of the NIRT Technology
Market's concern GAO's findings
--------------------------------------- ---------------------------------------
The NIRT technology could not predict When FGIS initially calibrated the NIRT
protein well on damaged wheat because technology in 1991, FGIS used samples
FGIS calibrated the NIRT technology by from the wheat that the industry sent
using only selected samples of premium to official laboratories in 1991 and
quality wheat rather than by using previous years. While representative of
wheat representing actual quality that 5-year period, the samples did not
conditions. represent the high damage that occurred
after 1991.
FGIS required official laboratories to To avoid confusion over test results at
remove NIRR instruments from their the same site, FGIS asked only that
premises as soon as they switched to official laboratories not use both
the NIRT technology. This was probably instruments for testing.
done to prohibit comparative testing.
FGIS' pilot study results were based The pilot study evaluated the
only on premium quality samples that performance of the NIRT technology on
were not representative of the quality 1992 samples of wheat traded in the
of the wheat traded in the market. market.
The results of the NIRT technology FGIS' testing results indicate that the
might be more consistent than the NIRR NIRT technology's results in the field
technology's results, but the results are more consistent and more accurate
of the NIRT technology are not as than the NIRR technology's results.
accurate as the NIRR technology's This is because of the NIRT
results. technology's operational advantages.
The NIRT technology always In 1993, the NIRT technology
underestimates protein levels. underestimated protein in high-damage,
high-protein wheat--the kind that was
primarily harvested in North Dakota
that year. The 1994 recalibration of
the NIRT technology has taken this
condition into account.
FGIS made an implementation mistake by FGIS determined that it was not cost-
not running the NIRT and NIRR effective to operate the two
technologies in tandem for a period of technologies at all official points.
time before proceeding with full FGIS ran the two technologies side by
implementation of the NIRT technology. side in its pilot study at 16 official
locations and found that the NIRT
technology's results were closer to the
reference technology's results.
FGIS made an implementation mistake by FGIS switched its locations to the NIRT
not requiring all official inspection technology at the same time. The agency
points to switch at the same time. strongly recommended that all other
official inspection locations switch to
the NIRT technology, but taking into
account budget constraints at these
sites, it did not require the switch.
As of August 1994, all official
inspection locations testing HRS wheat
were using the NIRT technology.
FGIS' 1994 recalibration of the NIRT According to FGIS officials, FGIS had
technology was an attempt to correct planned the NIRT recalibration as an
the agency's original faulty update of the 1991 calibration rather
calibration. than as a correction. The recalibration
would have taken place in the absence
of a particular problem. However,
because of the unusual characteristics
of the 1993 crop, FGIS expedited its
NIRT recalibration.
--------------------------------------------------------------------------------
ECONOMIC LOSSES IN 1993 WERE
OFFSET BY HIGH PREMIUMS
------------------------------------------------------------ Letter :5
Although farmers and operators of grain elevators in the wheat market
were concerned about losses resulting from lower protein readings by
the NIRT technology, our economic analysis suggests that such losses
were offset, overall, by the increase in price premiums that resulted
from the lower NIRT readings. By underestimating protein levels in
heavily damaged HRS wheat, the NIRT technology indicated an apparent
additional shortage of high-protein wheat. Farmers and grain
elevator operators generally benefited from this perception. The
artificial shortage resulting from the lower readings created price
premiums that were higher than they would have been if the protein
levels had been measured accurately. (See app. II for a more
detailed discussion of our economic analysis.)
For example, our analysis shows that wheat exported through the
Pacific Northwest had price premiums that were, on average, 53 cents
higher per bushel because of the NIRT technology's original
calibration. In addition, the NIRT technology's readings did not
generally drop farmers and others out of the high-protein categories
altogether. While our economic analysis shows that losses from lower
protein readings were offset by higher premiums in the aggregate,
some farmers we spoke with believed that they had incurred severe
financial losses.
EFFORTS TO STANDARDIZE
UNOFFICIAL PROTEIN TESTING ARE
UNDER WAY
------------------------------------------------------------ Letter :6
Unofficial protein testing is generally more common than official
protein testing for U.S. wheat traded in the domestic market. For
example, it is common practice for elevator managers to use in-house
NIRR or NIRT equipment to measure protein levels in wheat before they
offer farmers a price for their wheat. However, without federal
standards, this unofficial protein testing varies from site to site
in terms of equipment operation, maintenance, and calibration. As a
result, determinations of protein levels can vary extensively within
the commercial sector and between this sector and the official
protein testing.
To promote greater uniformity in commercial grain inspections, in
1990, the Congress amended the United States Grain Standards Act,
authorizing FGIS to work with the Department of Commerce's National
Institute of Standards and Technology and the National Conference on
Weights and Measures to standardize unofficial grain inspections.\1
The Institute sponsors the Conference, a standards-writing
organization whose members include weights and measures officials
from states and local communities, federal government officials,
manufacturers, trade representatives, and consumer organizations.
The 1990 act authorized FGIS, along with the Conference and the
Institute, to develop a program to evaluate equipment for the
unofficial testing system. This program is to include identifying
inspection instruments that require standardization, establishing
performance criteria, developing a national testing program for
instruments used on commercial inspections, and developing standard
reference materials or other means necessary for the calibration or
testing of approved instruments.
In January 1994, the Conference introduced its proposal to
standardize measurement practices for wheat protein in the commercial
sector and its proposal to have the unofficial system use only
equipment that meets the Conference's standards by 2005. If enforced
by states, these standards could ensure consistent operation of the
unofficial equipment and minimize operator error.
In July 1994, members of the Conference voted to approve the proposed
standards. However, they deferred a decision on when these standards
could be enforced by the states. Manufacturers can start using these
standards as guidelines for designing protein-testing instruments.
In addition, FGIS planned to start testing instruments for adherence
to the standards in the fall of 1994.
--------------------
\1 Grain Quality Incentives Act of 1990, P.L. 101-624, title XX,
section 2009, Nov. 28, 1990. Prior to this legislation, FGIS'
activities were restricted to the official inspection system. FGIS'
authority to work cooperatively on standardized commercial grain
inspections was subsequently expanded to include "other appropriate
governmental, scientific, or technical organizations." United States
Grain Standards Act Amendments of 1993, P.L. 103-56, section 11,
Nov. 24, 1993.
AGENCY AND INDUSTRY COMMENTS
AND OUR RESPONSE
------------------------------------------------------------ Letter :7
We discussed a draft of this report with FGIS' Acting Director,
Quality Assurance and Research Division and with the Chief of the
Quality Control and Testing Branch, Quality Assurance and Research
Division. These officials agreed with our results. However, they
suggested minor technical revisions to our draft. Where appropriate,
we incorporated these revisions into the report.
We also discussed the draft with the Vice President and Director of
Planning and Evaluation, U.S. Wheat Associates. He also agreed with
our results and did not suggest any changes.
SCOPE AND METHODOLOGY
------------------------------------------------------------ Letter :8
To address our objectives, we interviewed officials at FGIS and the
Agricultural Stabilization and Conservation Service in USDA and the
Congressional Research Service. We spoke with representatives of
national farm, milling, and trade organizations as well as local
organizations in Minnesota, Montana, North Dakota, and South Dakota.
We also talked to representatives from FGIS-authorized laboratories,
state departments of agriculture, and academia. In addition, we
spoke to officials of the Canadian Grain Commission. We also
obtained and analyzed documentation and data from these agencies and
organizations. (See app. III for a list of organizations
contacted.) Finally, we spoke to a number of individual farmers and
elevator operators.
To evaluate the economic impacts of the new technology on industry,
we used an economic model that describes spring wheat prices as a
function of the various wheat characteristics such as protein, test
weight, total damage, and moisture content. Our economic analysis
encompassed the states of Minnesota, Montana, North Dakota, and South
Dakota.
We conducted our review between November 1993 and July 1994.
---------------------------------------------------------- Letter :8.1
As arranged with your offices, unless you publicly announce its
contents earlier, we plan no further distribution of this report
until 7 days from the date of this letter. At that time, we will
send copies to the Secretary of Agriculture. We will also make
copies available to others on request.
Please contact me at (202) 512-5138 if you or your staff have any
questions about this report. Major contributors to this report are
listed in appendix IV.
John W. Harman
Director, Food
and Agriculture Issues
LIST OF REQUESTERS
---------------------------------------------------------- Letter :8.2
The Honorable Conrad Burns
The Honorable Kent Conrad
The Honorable Byron L. Dorgan
The Honorable Larry Pressler
United States Senate
The Honorable Tim Johnson
The Honorable David Minge
The Honorable Collin C. Peterson
The Honorable Earl Pomeroy
House of Representatives
CHRONOLOGY OF THE EVALUATION AND
IMPLEMENTATION OF THE NIRT
TECHNOLOGY
=========================================================== Appendix I
LATE 1980S
------------------------------------------------------- Appendix I:0.1
The U.S. Department of Agriculture's (USDA) Federal Grain Inspection
Service (FGIS) began to use the Near Infrared Transmittance (NIRT)
technology for determining protein levels in soybeans. Because it
found the technology successful, FGIS decided to investigate the NIRT
technology's usefulness for determining wheat protein levels.
JUNE 1990
------------------------------------------------------- Appendix I:0.2
FGIS' testing showed that the NIRT technology met established
requirements for estimating wheat protein levels.
JULY 1991
------------------------------------------------------- Appendix I:0.3
FGIS completed the NIRT calibration for hard red spring (HRS) wheat,
which was necessary to begin implementing the NIRT technology.
SEPTEMBER 1991
------------------------------------------------------- Appendix I:0.4
FGIS presented four alternatives for implementing the NIRT technology
to the agency's Advisory Committee: (1) not implementing the NIRT
technology for wheat protein, (2) using the NIRT technology
concurrently with the NIRR technology, (3) phasing out the NIRR
technology as part of the NIRT technology's implementation, and (4)
converting all official inspection sites from the NIRR technology to
the NIRT technology at the same time. FGIS discussed with the
Advisory Committee (whose members included farmers) the inevitable
differences between the two technologies.
OCTOBER 1991
------------------------------------------------------- Appendix I:0.5
Between June of 1990 and October of 1991, FGIS assessed the impact of
including the NIRT technology in the official system. It determined
that there was little difference in accuracy between the NIRT
technology and the reference technology but determined that there
were differences in results between the NIRR and NIRT technologies.
NOVEMBER 1991
------------------------------------------------------- Appendix I:0.6
FGIS decided to implement the NIRT technology for official protein
testing. It planned to verify the performance of the first several
units before allowing their official use. FGIS would proceed to
purchase and phase in units as soon as possible by January 1, 1994.
MARCH 1992
------------------------------------------------------- Appendix I:0.7
FGIS announced a pilot study to compare the NIRT technology's
performance with that of the NIRR technology using market samples at
16 FGIS-operated locations. The pilot study started in the week
ending March 14, 1992, and included 5 weeks of data for each location
in the study.
JUNE 1992
------------------------------------------------------- Appendix I:0.8
Pilot study results showed that the expected accuracy was better for
the NIRT technology than for the NIRR technology when used under
field conditions as opposed to controlled laboratory conditions.
FGIS reported on the status of the NIRT technology to the FGIS
Advisory Committee. FGIS officials reported that they found the NIRT
technology to be closer to the reference technology than the NIRR
technology and that the use of the NIRT technology would improve the
accuracy of the entire system. In addition, FGIS found that field
NIRR technologies, on average, estimated higher protein levels than
did the NIRT technology, implying that in some instances, using the
NIRT technology would yield lower protein measurements. The Advisory
Committee recommended that FGIS continue moving toward the NIRT
technology as the standard protein-testing method and continue to
explore means of phasing in the technology to minimize the impact on
the industry.
FGIS constructed an NIRT technology implementation schedule.
JULY 1992
------------------------------------------------------- Appendix I:0.9
FGIS announced in the Federal Register that the agency would
introduce the NIRT technology as an "additional" type of technology
in the agency's national program to inspect wheat protein levels.
Because FGIS found that the calibration for hard red winter (HRW)
wheat did not accurately estimate protein levels, FGIS suspended its
announced implementation of the NIRT technology for official protein
testing for HRS wheat, soft white wheat, and HRW wheat, pending
additional verification testing.
AUGUST 1992
------------------------------------------------------ Appendix I:0.10
FGIS planned to develop updated NIRT calibrations for HRS wheat in
the fall and winter of 1993.
JANUARY 1993
------------------------------------------------------ Appendix I:0.11
FGIS announced that it had completed testing for soft white wheat and
HRS wheat and was prepared to begin using the NIRT instruments for
these classes on February 22, 1993.
FEBRUARY 1993
------------------------------------------------------ Appendix I:0.12
FGIS participated in a conference call with industry representatives.
The discussion centered around (1) the best time for implementing the
NIRT technology and (2) industry's concerns about introducing two
different official methods for determining wheat protein that do not
provide identical results. FGIS said it would address any
measurement discrepancies on a case-by-case basis.
After reviewing industry's comments, FGIS decided not to implement
the NIRT program for soft white and HRS wheat on February 22, 1993.
FGIS rescheduled its implementation of the NIRT program for all
classes of wheat to May 2, 1993. As suggested during the call, this
date coincides with the start of the marketing year for wheat.
Although spring wheat is typically harvested later, several
participants indicated that stocks should be sufficiently low by May
2 so the market impact of the new testing program would be reduced.
Furthermore, establishing one implementation date for all classes of
wheat would minimize the time needed to check and maintain two
separate instruments.
FGIS started the first of two training programs for Protein
Coordinators (FGIS employees located around the country who are
responsible for direct technical oversight of official wheat protein
testing) so that they would be prepared to train NIRT operators in
the field.
MARCH 1993
------------------------------------------------------ Appendix I:0.13
FGIS announced the use of the NIRT technology for official wheat
protein testing, effective May 2, 1993. All FGIS locations providing
wheat protein testing were being equipped with the NIRT technology.
FGIS planned to use the NIRT technology for official testing at
export locations and for federal appeals--the highest level of
official retests. Non-FGIS official sites without NIRT instruments
could continue to provide wheat protein testing with their NIRR
equipment.
MAY 1993
------------------------------------------------------ Appendix I:0.14
FGIS began using the NIRT technology for official wheat protein
inspections. All FGIS-operated stations began to use the NIRT
technology only.
Within 1 to 2 weeks after the switch to the NIRT technology, FGIS
learned of differences between the NIRR technology's and NIRT
technology's results for wheat protein in HRS wheat. These
differences were found in both previous stocks and Canadian wheat
imports. For example, the Montana Department of Agriculture started
receiving calls from export elevator operators in Portland, Oregon,
expressing concern that the NIRT technology was predicting lower
protein levels than the NIRR technology did.
FGIS started a series of studies to identify why the differences
occurred and whether corrective measures were needed. Those tests,
according to FGIS, showed that the NIRT technology was performing
within the expectations of protein levels for normal HRS wheat. The
differences between the NIRR technology and the NIRT technology were,
in many cases, equally due to the error in the NIRR technology as
well as the NIRT technology. Canadian feed wheat, which was not
covered in the NIRT calibration, consistently fell outside the
expected statistical tolerances for accuracy.
On May 21, FGIS scheduled NIRT recalibrations for HRS wheat, soft
white wheat, and soybeans. Resources were scheduled from May 1993
through September 1993 to work on the development of a new HRS wheat
protein calibration.
On May 27, FGIS conducted a teleconference with 37 grain industry
representatives--from Minnesota, North and South Dakota, Montana,
Idaho, Oregon, and Washington State--to address industry concerns
about the NIRT technology's results. At this meeting, FGIS agreed to
monitor the situation in order to identify any problems with the NIRT
technology's results.
OCTOBER 1993
------------------------------------------------------ Appendix I:0.15
According to FGIS, one of the first indications of problems with the
readings from the NIRT technology on the 1993 HRS wheat crop came
from the Grand Forks Grain Inspection Department, Inc. To address
this concern, FGIS obtained 10 HRS wheat samples from that company
for laboratory analysis.
NOVEMBER 1993
------------------------------------------------------ Appendix I:0.16
FGIS developed a preliminary new HRS wheat recalibration that
included a small number of 1993 crop samples that were obtained early
in the harvest and represented typical production for the areas in
which they were harvested. Had there been no quality problem with
the 1993 crop of HRS wheat, FGIS would probably have implemented this
recalibration in May 1994 as a routine update.
According to FGIS' analysis of samples from several locations chosen
to reasonably represent the HRS wheat market, the current calibration
for HRS wheat showed lower protein levels than it should have on a
significant portion of 1993 crop samples, particularly those samples
with quality problems that affected their appearance.
NOVEMBER-DECEMBER 1993
------------------------------------------------------ Appendix I:0.17
Between November 16 and December 7, 1993, FGIS developed and
evaluated a new NIRT calibration for HRS wheat that incorporated 1993
crop samples.
DECEMBER 1993
------------------------------------------------------ Appendix I:0.18
FGIS determined that the new calibration would significantly improve
the accuracy of protein estimates for the 1993 HRS wheat. A final
evaluation using a group of samples with a protein range chosen from
samples coming from all major HRS wheat-marketing areas indicated
that the new NIRT calibration should transfer well to field
instruments, maintain accuracy on normal HRS wheat, and improve the
accuracy of protein determinations for samples that were damaged and
therefore difficult for NIRT technology to predict. The calibration
was based on samples from the 1988-93 crop years. FGIS believed that
the revised calibration would particularly improve official protein
results for the 1993 crop that was getting lower estimates from the
NIRT technology. The updated calibration was expected to increase
the protein levels of affected samples by 0.1 to 0.5 percent, with
the larger increases occurring above the 13.5-percent protein level.
These changes can make a significant difference in prices for farmers
selling wheat during a shortage of high-protein, high-quality wheat.
While the recalibration would not completely eliminate the tendency
to give lower protein readings in lower-quality samples, it would
improve the overall accuracy of the calibrations.
FGIS sent instructions for the upcoming new calibration for HRS wheat
and its implementation to all locations testing HRS wheat using the
NIRT technology. FGIS would begin implementation of the new
calibration for HRS wheat on January 10, 1994, but because of
expected market impacts, the new calibration would not be used until
January 24, 1994.
FGIS expected that there would be differences between the new and old
calibrations on some types of samples. Therefore, FGIS delayed the
implementation of the new calibration following the agency's
announcement to have all instruments set with the new calibration
before the date set for implementation. This was intended to
minimize the possibility that some entities would gain an unfair
market advantage from prior knowledge or earlier use of the
calibration.
JANUARY 1994
------------------------------------------------------ Appendix I:0.19
FGIS held a teleconference on January 5 with all official
protein-testing locations to review the processes that FGIS would use
to convert to the new calibration to make the change as smooth as
possible.
FGIS announced that on January 24, it would update the official NIRT
protein calibration for HRS wheat.
On January 24, all NIRT instruments began to use the new calibration
for HRS wheat.
AUGUST 1994
------------------------------------------------------ Appendix I:0.20
All official locations testing HRS wheat had switched to the NIRT
technology.
ECONOMIC IMPACT OF WHEAT PROTEIN
TESTING
========================================================== Appendix II
This appendix discusses our estimation of the economic impact of low
protein readings by FGIS' NIRT technology during the 1993 crop year.
Overall, we found that the economic impact of lower protein readings
from the NIRT technology was generally balanced by increased protein
premiums\1 that resulted from these lower readings. The lower
protein readings exacerbated the market's response to a scarcity of
high-protein wheat during a year in which high-protein wheat was
already in short domestic and international supply because of
weather- and disease-related problems. We estimate that initial low
readings by the NIRT technology inflated protein premiums by
approximately 53 cents per bushel for HRS wheat sold between the 13-
and 15-percent protein levels in the Pacific Northwest export market.
For the Great Lakes export market, although data are limited, we
estimate that premiums were approximately 82 cents higher.
First, we provide an explanation of our choice of the hedonic
regression framework for this analysis. We then furnish some
intuition on the theoretical framework for the hedonic model. Next,
we discuss the empirical model and estimation techniques that we used
to obtain the effects of low protein readings. A description of our
data and data sources follows. For both the Pacific Northwest and
the Great Lakes markets, we then provide and examine the regression
results. Last, using these regression results, we calculate the
economic impact of low protein readings on both of these markets.
--------------------
\1 We define protein premiums in this appendix as the difference
between an average protein percentage price and a base price. Using
13 percent as the base price, the premium for 14-percent protein
would be the average 14-percent market price over the average
13-percent market price. The additional premium at the 15- percent
level would be the difference between the average market price for
15-percent protein and 14-percent protein wheat.
ANALYSIS USED
-------------------------------------------------------- Appendix II:1
We used a hedonic regression model, incorporating monthly price and
quality data, to estimate the effect of artificially low protein
measurements on protein premiums for HRS wheat. Protein premiums
were estimated to be a function of different wheat
characteristics--protein, damage, moisture--as well as a variable
representing the supply of protein and a variable representing low
NIRT readings in that period of time during which the NIRT technology
was in effect but before the machines were recalibrated in January of
1994.
The hedonic approach was valuable in this analysis, since it is a
"characteristics" regression approach, and characteristics such as
protein and damage can be of vital importance in determining the
price that particular wheat shipments/crops can command in the
market. It also helped us to incorporate the fact that wheat with
different levels of protein, while similar, is a quite different
product. For example, a characteristic of one wheat protein level,
such as the moisture in 13 percent-protein wheat, may have an effect
on the demand and price of 14-percent protein wheat. In addition,
the fact that blending is so prevalent in this industry makes the
characteristics of one type of wheat an important part of another's
demand. By controlling for these important wheat characteristics,
the coefficient on the NIRT calibration variable helped us to isolate
the separate effect on protein premiums of the low protein readings.
We used monthly price and quality data at export markets from 1985 to
1993. Since they are official testing sites, export markets were the
only points where all observations were definitely tested using the
NIRT technology from May to December 1993. In addition, we only
estimated effects on wheat protein premiums from the Pacific
Northwest and the Great Lakes export markets because these were the
only markets that had a complete record of prices by average protein
level.
HEDONIC THEORETICAL MODEL
-------------------------------------------------------- Appendix II:2
In a hedonic model, the individual coefficients of the regression
variables represent the implicit, or nonmarket, price of each
characteristic found in that product. Following the hedonic work in
the economics literature (Ladd and Martin\2 , Wilson\3 , Espinosa and
Goodwin\4 , and others), processors demand a differentiated
agricultural product such as wheat because of the particular
characteristics it possesses. These characteristics can be stated as
input arguments into the processor's production function. For
example, the amount of flour demanded for making bread dough and
rolls is highly influenced by a characteristic such as protein
content. Maximization of the processor's profit function, which
includes this production function, yields a first-order condition
that can be interpreted as the hedonic price function:
where,
ri = the price of input i
xj.h = the total quantity of characteristic j that enters
into the production of product h
vih = the quantity of the i\th input used in the
production of the h\th product
�xj.h/�vih = the marginal yield of characteristic j in
production of output h from input i
�fh/�xj.h = the marginal physical product from one unit
of characteristic j used in the production
of the h\th product
ph(�fh/�xj.h) = the value of the marginal product of
the j\th characteristic used in the production of h
The last term, therefore, can be interpreted as a marginal implicit
price (or imputed price) paid for the j\th product characteristic
used in the production of output h. The last equation can be
simplified by assuming that j = ph (�fh /�xj.h) and xjih =
�xj.h /�vih are both constant\5 and is therefore:
where j is the marginal implicit value of the characteristic
j and xjih is the quantity of characteristic j contained in each unit
of input i that goes into the production of h. Thus, by regressing
input prices on input characteristics, as measured by xjih, we can
determine the effect that physical characteristics have on the prices
paid for inputs and measure the marginal implicit values (or hedonic
prices) of these characteristics.
In our model, we assume that ri, the price variable, is a protein
premium. Protein premiums are prices that millers and others must
pay in the marketplace to obtain wheat of desired quality. We use
protein premiums instead of the total market price of wheat because
we want to estimate the effects of the low protein readings on
premiums directly. Variations in protein premiums, or price
differentials, are more likely to be captured by variations in
characteristics than are variations in the base price, which
fluctuate in response to broader market conditions. Similarly, we
did not want variations in the base wheat price to overwhelm the
variations in premiums. As with total price, premiums are affected
by the levels of characteristics of the different protein
percentages.
--------------------
\2 George W. Ladd and Marvin B. Martin, "Prices and Demands for
Input Characteristics," American Journal of Agricultural Economics,
Vol. 58 (1976), pp. 21-30.
\3 William W. Wison, "Differentiation and Implicit Prices in Export
Wheat Markets," Western Journal of Agricultural Economics, Vol. 14,
No. 1 (1989), pp. 67-77.
\4 Juan A. Espinosa and Barry K. Goodwin, "Hedonic Price Estimation
for Kansas Wheat Characteristics," Western Journal of Agricultural
Economics, Vol. 16, No. 1 (1991), pp. 72-85.
\5 This assumption means that each additional unit of input
contributes the same amount of a given characteristic and that the
marginal implicit price for the characteristic is constant.
EMPIRICAL MODEL AND ESTIMATION
PROCEDURE
-------------------------------------------------------- Appendix II:3
We specified the empirical relationship between protein premiums and
implicit prices of HRS wheat characteristics by the following linear
sum:
where, PRit is the average protein premium for HRS wheat (dollars per
bushel) from the i\th protein percentage in time t, c is the constant
term, the j's represent marginal implicit prices for the j =
1, ..., m hard red wheat characteristics, as measured by the xitj's,
and is the error term.
We used conventional measures of wheat quality from the agricultural
economics literature in our estimation of wheat price premiums.
These measures included the percentage of protein, test weight,
percentage of moisture, and percentage of total defects. With hard
wheat, processors use protein to predict the quantity of wheat
gluten, which makes protein a desirable quality for bread-making.
Producers are paid for protein content based on quarters or fifths of
a percentage point. Therefore, higher protein levels should have a
positive influence on protein premiums. We also used test weight,
which measures the density of wheat kernels and is an important
indicator of flour yield. Therefore, we expect a positive
relationship between test weight and premiums. We expect to see a
negative relationship between moisture content and premiums, since a
high moisture content means a possibility of moisture damage in the
storage and handling of the wheat. Total defects are the sum of
three factors: damaged kernels, shrunken and broken kernels, and
foreign material. We expect total defects to have a negative effect
on premiums. Damage was an especially important variable to include,
since the NIRT technology initially displayed low protein readings in
wheat with high damage.
Because of blending and other demand considerations, we also realized
the impact that a certain protein level's characteristics may have on
another category's premium. In order to capture these effects, we
included in some specifications the influence of characteristics
outside of that particular protein premium category. For example, in
the 15-percent protein specification for the Pacific Northwest
market, we include the moisture level for both 13- and 14-percent
protein wheat. Here, we assume that as the moisture level increases
for the 13- and 14-percent protein categories, there will be a
greater demand for higher quality 15- percent protein wheat for
blending purposes. Similarly, we assume that as the moisture level
increases for 13-percent protein wheat, there will be an economic
incentive to blend with 14-percent protein wheat. We do not,
however, include moisture in 15-percent protein wheat in the
14-percent premium specification. This is because we assume that a
lower-protein wheat will not be blended with a higher-protein wheat
with too much moisture.
We also had to deal with the complication of the importance of the
supply effects for the 1992 and 1993 crop years because of the
natural shortage of high-protein wheat on the market. In general,
economists have found that the relative supplies of higher-protein
wheat to lower-protein wheat or the dispersion of protein in wheat is
an important factor in determining premiums. For instance, during
times in which higher-protein wheat is in relatively short supply,
premiums for higher-protein wheat are likely to be high. We
incorporated these supply effects within the hedonic framework,
starting with the following expression where the marginal implicit
value for an additional unit of wheat protein is a function of the
supply of protein of that level:
where prot is the marginal implicit value of an additional
unit of wheat protein and Qprot is the ratio of the quantity of wheat
at a certain protein level to the total domestic supply of wheat. We
expect there to be a negative relationship between this implicit
value for protein, prot, and Qprot, the ratio of that
particular supply of protein to total domestic supply. That is, the
sign on 1 is expected to be negative because as the supply
of a particular protein category increases (decreases) relative to
the total supply of protein, the relative value placed on that
category of protein decreases (increases). Therefore, substituting
expression 4 into expression 3 gives us:
The explanatory variable of interest in this analysis, however, is
not a wheat characteristic per se, but is a dummy variable that
represents the separate effect of low protein readings on premiums,
holding constant the observed wheat characteristics and conditions of
protein supply. This dummy variable represents that time period
between the introduction of the NIRT technology and the recalibration
of the NIRT technology. The NIRT machines with the initial, lower
calibration were installed in early May of 1993 and were recalibrated
upward in January of 1994. Specifically, to incorporate the price
effect of the lower protein readings, we included a dummy variable
which takes on a value of 1 for May to December 1993, and zero
otherwise. During this time period, the market was reacting to the
effects of these lower protein readings. Since this is a highly
competitive market, wheat premiums are very sensitive to the
distribution of high-protein wheat. Thus, by switching this variable
on, we could obtain a value for a protein premium differential
representing the effects of low protein readings, holding constant
other factors. We expect this variable to be positively related to
protein premiums. First, the market could have been reacting to
information about low protein readings by the NIRT technology.
Additionally, this variable could be positively related to premiums,
since our supply ratio variable, Qprot, was based on yearly data and
may not be sensitive to the monthly variation introduced by the
calibration problem. Consequently, the effect of reduced supply
shows up in this dummy variable. (See discussion in "Data and Data
Sources" section.)
In addition to these other supply factors, spring wheat premiums are
also affected by the supply of HRW wheat, since this class
substitutes for spring wheat in certain categories. In particular,
higher-protein, 13-percent HRW wheat substitutes on the margin for
lower-protein, 14-percent spring wheat. Therefore, in order to
capture this competitive supply effect, we included the monthly
percentage of HRW wheat to total spring and winter wheat at the
13-percent protein level.
Finally, in order to account for seasonality, we also included a
dummy variable, which we labelled 1 during preharvest months, when
supplies are smaller, and zero during postharvest months, when
supplies and stocks are larger.
PACIFIC NORTHWEST MARKET
ESTIMATION
------------------------------------------------------ Appendix II:3.1
Using this hedonic empirical model, we estimated protein premiums at
the 14- and 15-percent protein levels for the Pacific Northwest
export market. First, we corrected for autocorrelation using a
single equation autoregressive procedure--the Yule-Walker method.\6
First-order autocorrelation is often prevalent in time-series
analysis such as this. Using the transformed variables from this
procedure, we estimated the equations jointly by the Seemingly
Unrelated Regression model.\7 This estimation technique provides a
gain in efficiency by using information on explanatory variables that
are included in the system but excluded from the ith equation. Using
78 observations, we jointly estimated equations 7 and 8:
(See figure in printed
edition.)
where:
PR14 = Protein premium for 14-percent protein wheat calculated as the
difference between the average monthly price for 14-percent protein
and 13-percent protein for HRS wheat.
PR15 = Protein premium for 15-percent protein calculated as the
difference between the average monthly price for 15-percent protein
and 14-percent protein for HRS wheat.
C = Constant
PROT13 = Average monthly protein percent at the 13-percent level
(between 13.0 percent and 13.99 percent) for the Portland market.
PROT14 = Average monthly protein percent at the 14-percent level
(between 14.0 and 14.99) for the Portland market.
PROT15 = Average monthly protein percent at the 15-percent level
(between 15.0 percent and 15.99 percent) for the Portland market.
WTPR14 = Ratio of 14-percent protein HRS wheat on the domestic market
to the total quantity of 13-, 14-, and 15-percent HRS wheat times the
average 14-percent protein level (PROT14).
WTPR15 = Ratio of 15-percent protein HRS wheat on the domestic market
to the total quantity of 13-, 14-, and 15-percent HRS wheat times the
average 15-percent protein level (PROT15).
TD13 = Average monthly total defects at the 13-percent protein level;
total damage includes total damaged kernels, shrunken and broken
kernels, and foreign material.
TD14 = Average monthly total defects at the 14-percent protein level.
TD15 = Average monthly total defects at the 15-percent protein level.
MST13 = Average monthly moisture content at the 13-percent protein
level.
MST14 = Average monthly moisture content at the 14-percent protein
level.
MST15 = Average monthly moisture content at the 15-percent protein
level.
CAL = Dichotomous variable to represent low protein reading; 1 for
May 1993 to December 1993 and zero otherwise.
SEAS = Dichotomous variable representing seasonality; 1 during the
preharvest months and zero for postharvest months.
TW13 = Average monthly test weight in pounds of grain per bushel at
the 13-percent protein level.
TW14 = Average monthly test weight in pounds of grain per bushel at
the 14-percent protein level.
TW15 = Average monthly test weight in pounds of grain per bushel at
the 15-percent protein level.
HRW13= Monthly proportion of 13-percent HRW wheat of the total amount
of 13-percent protein wheat.
--------------------
\6 This is described in A.R. Gallant and J.J. Goebel, "Nonlinear
Regression With Autoregressive Errors," Journal of the American
Statistical Association, Vol. 71 (1976).
\7 A. Zellner, "An Efficient Method of Estimating Seemingly
Unrelated Regressions and Tests for Aggregation Bias," Journal of the
American Statistical Association, Vol. 57 (1962), pp. 348-368.
GREAT LAKES MARKET
ESTIMATION PROCEDURE
------------------------------------------------------ Appendix II:3.2
For the Great Lakes export market, there was an insufficient number
of observations at the 15-percent protein level to warrant using the
SUR method. Therefore, we estimated premiums for the 14- and
15-percent protein levels by the single equation autoregressive
method above. Using this procedure, we were able to use 47
observations at the 14-percent protein level and 39 observations at
the 15-percent protein level. However, since these numbers were
still quite small, we reduced the total number of explanatory
variables in both equations. The Great Lakes equations we estimated
for 14- and 15-percent protein premiums were:
(See figure in printed
edition.)
DATA AND DATA SOURCES
-------------------------------------------------------- Appendix II:4
We obtained all quality data from the FGIS Export Grain Inspection
System (EGIS) data files for each month and each protein percentage.
These data files keep records of each shipment or shipload sent to
each export market in the United States for all of the grains. Using
this database for 1985 through 1993, we calculated monthly average
quality characteristics--such as average defects and moisture levels
for our explanatory variables--for 13- to 13.9-percent, 14- to
14.9-percent, and 15- to 15.9-percent protein levels. For some
months, we had missing observations for certain percentage categories
of HRS wheat. Because of this, for the Pacific Northwest market, our
number of observations were reduced from 99 to 78 including 2 missing
observations in the key May to December 1993 period.
We then weighted these quality characteristics by the size of
shipload or shipload bushels to obtain a more representative estimate
of the average monthly quality characteristics. In addition, we
included HRW wheat in the 13-percent category, since quality
characteristics (such as sprout damage) of HRW wheat in this category
have an effect on the demand for HRS wheat at the 13- and 14-percent
levels. Including HRW wheat at the 13-percent level also gave us a
greater number of observations.
We obtained quantity data on the production of HRS wheat from USDA's
February 1994 Situation and Outlook Report as well as from the U.S.
Wheat Associates' U.S. Wheat: 1993 Crop Quality Report. We
obtained the percentage of U.S. production of different protein
categories such as 13-percent, 14-percent, or 15-percent protein
wheat from various issues of North Dakota State University's Regional
Quality Report for HRS wheat for the years 1985 to 1993. In order to
calculate our supply ratio variable, Qprot, we used yearly quantity
data. We then divided this yearly data into months and applied a
monthly smoothing process. To these monthly total quantity figures,
we multiplied the percentage of the different protein categories such
as 13, 14, and 15 percent that we obtained from North Dakota State
University. Although this was somewhat of a limited measure in that
it may not capture month-to-month fluctuations, it does represent
broader supply conditions related to HRS wheat.
To calculate price premium data, we used monthly price data for 13-,
14-, and 15-percent HRS wheat obtained from USDA's Livestock and
Grain Market News from the Pacific Northwest arket and the
Minneapolis market from 1985 until 1993. To find premiums, we
subtracted the average 14-percent price from the 15-percent price to
obtain the average monthly 15-percent price premium. Similarly, we
calculated the average 14-percent differential by subtracting the
average 13-percent price from the average 14-percent price. In order
to express these price data in constant dollars, we adjusted the
price series using the Bureau of Labor Statistics' Producer Price
Index for crude materials.
Data used in order to calculate economic loss included the quantities
sold into the export market. This was taken again from the EGIS
database and organized at each protein level. Protein premium data
used for these calculations were taken from Livestock and Grain
Market News as well as other USDA publications.
RESULTS OF REGRESSION MODELS
-------------------------------------------------------- Appendix II:5
For both the Pacific Northwest and the Great Lakes market, the
variable used to reflect low protein readings, CAL, had a positive
effect on protein premiums and was highly statistically significant.
This result was unaffected by the type of model used or the inclusion
of different variables. However, greater confidence should be placed
on the results for the Pacific Northwest market because of the larger
number of observations at both the 14- and 15-percent protein levels.
PACIFIC NORTHWEST REGRESSION
RESULTS
------------------------------------------------------ Appendix II:5.1
Regression results for the Pacific Northwest market revealed that
CAL, the variable used to approximate the effect of low protein
readings, was significant and positively related to both 14- and
15-percent protein premiums. For the price difference between 14
percent and 13 percent or the 14-percent premium, PRE14, the CAL
variable indicated that low readings increased premiums by
approximately 31 cents. This result was consistent throughout the
trial of many model specifications and inclusion of various
combinations of explanatory variables.
Other significant variables included 14-percent protein percentage;
total defects, 14 percent; moisture, 13 percent; test weight, 13
percent; and percentage of HRW wheat, 13 percent. Interestingly,
14-percent protein was significantly and negatively related to
premiums. This result may represent the distribution of protein
across protein levels at the export market. We expect that
higher-protein wheat should command higher premiums than lower-
protein wheat in a relative sense. However, if the entire wheat
protein distribution was skewed because of a supply shortage of
high-protein wheat, then a lower average percentage of a certain
protein level would be associated with a higher premium.\8 Thus, we
interpret the negative relationship between average protein percent
and protein premiums as the result of this variable capturing this
supply effect. Nevertheless, inclusion or exclusion of the protein
percentage variable, PROT, did not affect the robustness of the
estimated coefficient of CAL, our variable of interest.
Total defects at the 14-percent level had the expected negative sign
and decreased premiums by 16 cents for every 1 percentage point
increase in damage. Moisture at the 13-percent protein level was
significant and positively related to 14-percent premiums. We
interpret this relationship to mean that the greater the moisture for
13-percent wheat, the greater the demand for 14-percent wheat. As
expected, the 13-percent HRW wheat revealed a competitive effect and
was negatively related to 14-percent HRS wheat premiums.
For the 15-percent premium percentage equation, statistically
significant variables include protein percentage, PRO14; average
moisture at the 13-percent and 14-percent level, MST13 and MST14;
test weight at the 14-percent level, TW14; and the seasonality
variable, SEAS; and CAL. The CAL variable indicated that premiums
were increased by 22 cents during those months in which there were
low test readings.
Table II.1
Regression Results From the Pacific
Northwest Export Market
Coefficien
t/
Independent Coefficient/ Independent significan
variable significance variable ce
-------------------- -------------- -------------------- ----------
CONSTANT -1.623 CONSTANT -1.792
WTPR14 -0.011 WTPR15 -0.011
PROT13 -0.196 PROT14 -
0.442\***
PROT14 -0.289\* PROT15 -0.003
TD13 0.045 TD14 -0.008
TD14 -0.155\** TD15 0.067
MST13 0.094\** MST13 0.091\**
MST14 0.045 MST14 0.108\**
N/A N/A MST15 -0.053
TW13 0.118\*** TW14 0.104\*
TW14 0.006 TW15 0.004
CAL 0.313\*** CAL 0.216\**
HRW13 -0.314\*** N/A N/A
SEAS 0.012 SEAS 0.081\*
System Weighted
R\2 = 0.82
130 degrees of freedom
----------------------------------------------------------------------
----------------------------------------------------------------------
Note: PRE14 = 14-percent protein premium. PRE15 = 15-percent
protein premium. N/A = not applicable.
\*** Denotes significance at the 1-percent level or greater.
\** Denotes significance at the 5-percent level.
\* Denotes significance at the 10-percent level.
--------------------
\8 Although we attempted to control for supply effects at each
protein level with the variable WTPR, our control did not perform
well.
GREAT LAKES REGRESSION
RESULTS
------------------------------------------------------ Appendix II:5.2
Regression results for the Great Lakes market also showed the CAL
variable to be highly statistically significant and positive at both
the 14-percent and 15-percent protein premium levels. Interestingly,
for the 14-percent protein premium level, the effect of low protein
readings was to increase premiums by approximately 26 cents, which
was less than that for the Pacific Northwest market. However, at the
15-percent protein level, the CAL variable indicated that protein
premiums would be approximately 56 cents higher because of low
protein readings. Overall, adding the 14- and 15-percent effects of
this variable together, premiums increased by 82 cents per bushel.
The only other significant variable was the supply variable at the
15-percent level, WTPR15. However, less confidence should be placed
in these results than in the Pacific Northwest results because of the
lower number of observations in each model, the lower R\2 s, and the
fact that there were few significant explanatory variables.
Table II.2
Regression Results From the Great Lakes
Export Market
Coefficien
t/
Coefficient/ significan
Independent variable significance Independent variable ce
-------------------- -------------- -------------------- ----------
CONSTANT 3.051 Constant -6.686
WTPR14 0.019 WTPR15 -
0.068\***
PROT14 -0.200 PROT15 0.359
TD14 -0.027 TD15 0.098
TW14 -0.024 TW15 -0.022
MST14 0.110 MST15 0.248
SEAS -0.001 SEAS 0.002
CAL 0.256\*** CAL 0.560\*
R\2 = 0.39 Degrees R\2 = 0.54 Degrees
of freedom = 38 of freedom = 30
----------------------------------------------------------------------
Note: PRE14 = 14-percent protein premium. PRE15 = 15-percent
protein premium. This table has no coefficients at the 5-percent
significance level (\** ).
\*** Denotes significance at the 1-percent level or greater.
\* Denotes significance at the 10-percent level.
ECONOMIC IMPACT DETERMINATION
-------------------------------------------------------- Appendix II:6
To determine the economic impact of low protein readings on the
producer, we compared the effect of premium increases due to the
market's reaction to these lower readings with the effect of losses
from lower protein levels of wheat. Overall, we found that a loss in
premium income from lower protein readings was offset by gains from
increases in protein premiums owing to a perceived decrease in wheat
protein in the marketplace.
Our estimates of economic impacts from gains and losses are based on
"average" price impacts over the affected period. Given the data
available to us, we were able to estimate the increase in premiums in
this period, holding everything else constant. We obtained the gains
or the price effects from the low protein readings from the CAL
variable in the regression results (see tables II.1 and II.2). For
example, in the 14-percent premium equation of the Pacific Northwest
market, the coefficient on the CAL variable translated into an
increase in premiums of approximately 31 cents. In the 15-percent
premium category, this gain was approximately 22 cents. We then
compared these gains from the market with losses from being forced
into the next lower protein category.
To calculate loss, we assumed that producers were forced into the
next lower protein category. This was based on a study by USDA's
Economic Research Service that reported, on average, that all
readings for HRS wheat were 0.16 percentage points lower under the
new testing devise.\9
For the Pacific Northwest market, to obtain premiums at each
percentage, we calculated average price differences at each protein
percentage from Grain and Livestock Market News and divided the price
difference into quarters. Since protein scales for premiums are
graduated into quarters or fifths of a percentage point, an
adjustment downward does not necessarily mean a move into the next
whole protein percentage. In the Pacific Northwest market, premiums
are based on quarters of a percentage, while the Great Lakes market
follows the Minneapolis market, which is based on fifths of a
percentage. For the Great Lakes market, we obtained average monthly
premiums at 13-, 14-, and 15-percent protein levels from the Economic
Research Service's study mentioned above. The assumption that all
producers were bumped down into the next protein category was
conservative, since some experts noted that only high-damaged,
high-protein wheat was affected by the new protein-testing
technology.
Holding everything else constant, the net effect of these two
estimates--gain from the market and loss from lower protein
readings--provides an estimate of net economic impact for the two
export markets--the Pacific Northwest and the Great Lakes. Using
data from the EGIS data file, we were able to calculate how much 13-,
14-, and 15-percent HRS wheat was exported from the Pacific Northwest
and the Great Lakes markets. From these calculations, we found that
there was a modest gain in the Pacific Northwest export market and a
small loss in the Great Lakes export market. However, these net gain
or loss estimates for each market are minor compared with the total
value of wheat traded in these export markets. Therefore, on
balance, losses from the lower protein readings were generally offset
by gains in the marketplace from a perceived short supply of
high-protein wheat. While individual farmers and others may have
experienced significant losses or gains, the aggregate economic
effects seemed to be minimal.
Table II.3
Economic Impact of Low Protein Readings
on the Pacific Northwest Market
Quantities
(bushels Premium
in change Totals
Total gain or loss thousands) (cents) (dollars)
------------------------------------ ---------- -------- ----------
Gain
----------------------------------------------------------------------
13 N/A No gain N/A
14 75,845\a 0.31\b $23,511,95
0
15 1,520\a 0.22\b 334,400
Total gain 23,846,350
Loss
13 16,377\a 0.20\c 3,275,400
14 75,845\a 0.20\c 15,169,000
15 1,520\a 0.14\c 212,800
Total loss 18,657,200
Net gain or (loss) $5,189,150
----------------------------------------------------------------------
Note: N/A = not applicable.
\a These figures are compiled at different protein percentages for
the Pacific Northwest market and are taken from the EGIS data files.
\b These figures are estimates of gains in premium (CAL variable)
from the market in cents taken from the above model of the Pacific
Northwest Market.
\c These figures represent average Pacific Northwest premium
categories calculated by taking price differences between protein
levels and dividing these into quarters; they are taken from
Livestock and Grain Market News.
Table II.4
Economic Impact of Low Protein Readings
on the Great Lakes Market
Quantity Premiu
(bushels m
in change
thousand (cents Totals
Total gain or loss s) ) (dollars)
---------------------------------------- -------- ------ ----------
Gain
13 N/A No N/A
gain
14 14,915\a 0.26\b $3,877,900
15 581\a 0.56\b 325,360
Total gain 4,203,260
Loss
13 13,900\a 0.17\c 2,363,000
14 14,915\a 0.15\c 2,237,250
15 581\a 0.23\c 133,630
Total loss 4,733,880
Net gain or (loss) $(530,620)
----------------------------------------------------------------------
Note: N/A = not applicable.
\a These figures are compiled at different protein percentages for
the Great Lakes Market and are taken from the EGIS data files.
\b These figures are estimates of gains in premiums (CAL variable)
from the market in cents taken from the above model of the Great
Lakes Market.
\c These figures represent average monthly Minneapolis protein
premium categories divided into fifths; they are taken from Economic
Effects of Updating Protein Calibration for Hard Red Spring Wheat,
USDA, Economic Research Service, Staff Report No. AGES9417, p. 11.
--------------------
\9 William Lin, Economic Effects of Updating Protein Calibration for
Hard Red Spring Wheat, USDA, Economic Research Service, Commodity
Economics Division, Staff Report No. AGES9417 (June 1994).
ORGANIZATIONS CONTACTED
========================================================= Appendix III
MINNESOTA
----------------------------------------------------- Appendix III:0.1
Minnesota Association of Wheat Growers
Minnesota Department of Agriculture
Minnesota Farmers Elevator Association
MONTANA
----------------------------------------------------- Appendix III:0.2
Montana Department of Agriculture
Montana Farmers Union
Montana Grain Growers Association
Montana Wheat and Barley Committee
NORTH DAKOTA
----------------------------------------------------- Appendix III:0.3
North Dakota Farm Bureau Federation
North Dakota Farmers Union
North Dakota Grain Dealers Association
North Dakota Grain Growers Association
North Dakota State University
North Dakota Wheat Commission
SOUTH DAKOTA
----------------------------------------------------- Appendix III:0.4
South Dakota Farm Bureau Federation
South Dakota Farmers Union
South Dakota Grain and Feed Association
South Dakota Grain Growers Association
South Dakota Wheat Commission
NATIONAL ASSOCIATIONS
----------------------------------------------------- Appendix III:0.5
Millers National Federation
National Grain and Feed Association
National Grain Trade Council
U.S. Wheat Associates
MAJOR CONTRIBUTORS TO THIS REPORT
========================================================== Appendix IV
RESOURCES, COMMUNITY, AND ECONOMIC
DEVELOPMENT DIVISION, WASHINGTON,
D.C.
Thomas E. Slomba, Assistant Director
Patrick J. Kalk, Assignment Manager
Barbara J. El Osta, Economist
Carol Herrnstadt Shulman, Communications Analyst
CHICAGO REGIONAL OFFICE
Pauline Seretakis Lichtenfeld, Evaluator-in-Charge
Jacqueline M. Garza, Staff Evaluator