Title I: Characteristics of Tests Will Influence Expenses;	 
Information Sharing May Help States Realize Efficiencies	 
(08-MAY-03, GAO-03-389).					 
                                                                 
The No Child Left Behind Act of 2001 (NCLBA) reauthorized the $10
billion Title I program, which seeks to improve the educational  
achievement of 12.5 million students at risk. In passing the	 
legislation, Congress increased the frequency with which states  
are to measure student achievement in mathematics and reading and
added science as another subject. Congress also authorized	 
funding to support state efforts to develop and implement tests  
for this purpose. Congress mandated that GAO study the costs of  
implementing the required tests. This report describes		 
characteristics of states' Title I tests, provides estimates of  
what states may spend to implement the required tests, and	 
identifies factors that explain variation in expenses.		 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-03-389 					        
    ACCNO:   A06842						        
  TITLE:     Title I: Characteristics of Tests Will Influence	      
Expenses; Information Sharing May Help States Realize		 
Efficiencies							 
     DATE:   05/08/2003 
  SUBJECT:   Data collection					 
	     Education program evaluation			 
	     Educational research				 
	     Educational standards				 
	     Educational testing				 
	     Strategic planning 				 
	     Academic achievement				 
	     Education or training costs			 
	     Colorado						 
	     Delaware						 
	     Maine						 
	     Massachusetts					 
	     North Carolina					 
	     Texas						 
	     Virginia						 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO 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.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
******************************************************************
GAO-03-389

Report to Congressional Requesters

United States General Accounting Office

GAO

May 2003 TITLE I Characteristics of Tests Will Influence Expenses;
Information Sharing May Help States Realize Efficiencies

GAO- 03- 389

The majority of states administer statewide tests and customize questions
to measure student learning against their state standards. These states
differ along other characteristics, however, including the types of
questions on their tests and how they are scored, the extent to which
actual test questions are released to the public following the tests, and
the number of new tests they need to develop to comply with the NCLBA. GAO
provides three estimates of total expenditures between fiscal year

2002 and 2008, based on different assumptions about the types of test
questions states may choose to implement and how they are scored. The
method by which tests are scored largely explains the differences in GAO*s
estimates.

If all states use tests with multiple- choice questions, which are machine
scored, GAO estimates that the total state expenditures will be about $1.9
billion. If all states use tests with a mixture of multiple- choice
questions and a limited number of open- ended questions that require
students to write their response, such as an essay, which are hand scored,
GAO estimates spending to be about $5.3 billion. GAO estimates that
spending will be at about $3.9 billion, if states keep the mix of question
types states reported to GAO. In general, hand scoring is more expensive
and time and labor intensive than machine scoring. Benchmark funding for
assessments as specified in NCLBA will cover a larger percentage of
estimated expenditures for tests comprised of multiple- choice questions
and a smaller percentage of estimated expenditures for tests comprised of
a mixture of multiple- choice and open- ended questions. Several states
are exploring ways to reduce assessment expenses, but information on their
experiences is not broadly shared among states.

Dollars in billions 0 2

4 6

Multiple- choice Current questiontype

Multiple- choice Estimates

Benchmark appropriations

and open- ended

Source: GAO analysis. 2.7

1.9 3.9

5.3

The No Child Left Behind Act of 2001 (NCLBA) reauthorized the $10 billion
Title I program, which seeks to improve the educational achievement of 12.
5 million

students at risk. In passing the legislation, Congress increased the
frequency with which states are to measure student achievement in
mathematics and reading and

added science as another subject. Congress also authorized funding to
support state efforts to develop and implement tests for this purpose.

Congress mandated that GAO study the costs of implementing the required
tests. This report describes characteristics of states*

Title I tests, provides estimates of what states may spend to implement
the required tests, and identifies factors that explain variation in
expenses.

Given that significant expenses may be associated with testing, GAO is
recommending that Education facilitate the sharing of information on
states* experiences in attempting to reduce expenses. Education agreed
with GAO*s

recommendation but raised concerns about GAO*s methodology for estimating
expenditures.

www. gao. gov/ cgi- bin/ getrpt? GAO- 03- 389. To view the full report,
including the scope and methodology, click on the link above. For more
information, contact Marnie S. Shaul at (202) 512- 7215 or shaulm@ gao.
gov. Highlights of GAO- 03- 389, a report to

Congressional Requesters

May 2003

TITLE I

Characteristics of Tests Will Influence Expenses; Information Sharing May
Help States Realize Efficiencies

Page i GAO- 03- 389 Title I Letter 1 Results in Brief 3 Background 4
States Generally Report Administering Statewide Assessments

Developed to Measure Their State Standards, but Differ Along Other
Characteristics 7 Estimates of Spending Driven Largely by Scoring
Expenditures 14 Conclusions 22 Recommendation 22 Agency Comments 22
Appendix I Objectives, Scope, and Methodology 24

Appendix II Accountability and Assessment Requirements under the 1994 and
2001 Reauthorizations of Title I 31

Appendix III Number of Tests States Reported They Need to Develop or
Augment to Comply with NCLBA (as of March 2003) 33

Appendix IV Estimates of Assessment Expenditures NCLBA Required, but Not
in Place at the Time of Our Survey, FY 2002- 08 35

Appendix V State Development and Nondevelopment Estimates 36

Appendix VI Fiscal Years 2002- 08 Estimated Expenditures for Each Question
Type 38 Contents

Page ii GAO- 03- 389 Title I Appendix VII Comments from the Department of
Education 39

Appendix VIII GAO Contacts and Staff Acknowledgments 42 GAO Contacts 42
Staff Acknowledgments 42 Tables

Table 1: Number of Assessments and Subject Areas Required by the 1994 and
2001 ESEA Reauthorizations 5 Table 2: Assessment Minimum Amounts under
NCLBA 6 Table 3: The Number of Tests States Reported Needing to Develop

or Augment Varies 13 Table 4: Estimated Expenditures by States for Title I
Assessments, Fiscal Years 2002- 08 15 Table 5: Estimated Total
Expenditures for Test Development Are Lower Than for Test Administration,
Scoring, and Reporting 17 Table 6: Total Estimated Expenditures by States
for Title I

Assessments, Fiscal Years 2002- 08 19 Table 7: States Selected for Study
25 Table 8: Examples of Assessment Expenditures 26 Table 9: Average Annual
Expenditures for the 7 States (adjusted to 2003 dollars) 27 Table 10:
Estimated Expenditures to Implement Title I Assessments in a Given Year 30
Table 11: Estimates of Expenditures for the Assessments Required by NCLBA
That Were Not in Place at the Time of Our Survey, Fiscal Years 2002- 08 35
Table 12: Estimates by State, Development, and Nondevelopment

Expenditures 36 Table 13: Estimated Expenditures for Each Question Type,
Fiscal Years 2002- 08 38 Figures

Figure 1: The Majority of States Report They Currently Use Statewide Tests
and Plan to Continue to Do So 8

Page iii GAO- 03- 389 Title I

Figure 2: The Majority of States Reported That They Currently Use and Plan
to Develop New Tests That Are Customized to Measure Their State*s
Standards 9 Figure 3: The Majority of States Reported They Use a
Combination

of Multiple- choice and Open- ended Questions on Their Tests, but Many
States Are Uncertain about Question Type on Future Tests 11 Figure 4:
States Split in Decision to Release Test Questions to the

Public Following Tests 12 Figure 5: Estimated Scoring Expenditures Per
Assessment Taken for Selected States, Fiscal Year 2002 16 Figure 6:
Various Factors Are Likely to Affect What States Spend on Title I
Assessments 18 Figure 7: Total Expenditures Likely to Be Lower in First
Few Years and Benchmark Funding in NCLBA Estimated to Cover Most of
Expenditures in First Few Years 21 Abbreviations

ESEA Elementary and Secondary Education Act LEA local educational agency
NASBE National Association of State Boards of Education

NCLBA No Child Left Behind Act This is a work of the U. S. Government and
is not subject to copyright protection in the United States. It may be
reproduced and distributed in its entirety without further permission from
GAO. It may contain copyrighted graphics, images or other materials.
Permission from the copyright holder may be necessary should you wish to
reproduce copyrighted materials separately from GAO*s product.

Page 1 GAO- 03- 389 Title I

May 8, 2003 The Honorable Judd Gregg Chairman, Committee on Health,
Education,

Labor, and Pensions United States Senate

The Honorable Edward M. Kennedy Ranking Minority Member, Committee on

Health, Education, Labor, and Pensions United States Senate

The Honorable John A. Boehner Chairman, Committee on Education

and the Workforce House of Represenatives The Honorable George Miller

Ranking Minority Member, Committee on Education and the Workforce House of
Representatives

Title I, the largest source of federal funding for primary and secondary
education, provided states $10.3 billion in fiscal year 2002 to improve
the educational achievement of 12.5 million students at risk. In passing
the No Child Left Behind Act of 2001 (NCLBA), Congress increased funding
for Title I and placed additional requirements on states and schools for
improving student performance. To provide an additional basis for making
judgments about student progress, NCLBA increased the frequency with which
states are to assess students in mathematics and reading and added science
as another subject. Under NCLBA, states can choose to administer
statewide, local, or a combination of state and local assessments, but
these assessments must measure states* content standards for learning. If
a state fails to fulfill NCLBA requirements, the Department of Education
(Education) can withhold federal funds designated for state administration
until the requirements have been fulfilled. To support states in
developing and implementing their assessments, Congress authorized
specific funding to be allocated to the states between fiscal year 2002
and 2007.

United States General Accounting Office Washington, DC 20548

Page 2 GAO- 03- 389 Title I

NCLBA requires that states test all students annually in grades 3 through
8 in mathematics and reading or language arts and at least once in one of
the high school grades by the 2005- 06 school year. It also requires that
states test students in science at least once in elementary, middle, and
high school by 2007- 08. Some states have already developed assessments in
many of the required subjects and grades.

In the conference report accompanying passage of the NCLBA, Congress
mandated that we do a study of the anticipated aggregate cost to states,
between fiscal year 2002 and 2008, for developing and administering the
mathematics, reading or language arts, and science assessments required
under section 1111( b) of the act. As agreed with your offices, this
report (1) describes characteristics of states* Title I assessments and
(2) provides estimates of what states may spend to implement the required
assessments between fiscal year 2002 and 2008 and identifies factors that

explain variation in expenses. 1 To determine the characteristics of
states* Title I assessments, we collected information through a survey
sent to the 50 states, the District of Columbia, and Puerto Rico; all 52
responded to our survey. We also reviewed published studies detailing the
characteristics of states* assessments. To estimate projected expenditures
all states are expected to incur, we reviewed 7 states* expenditures* all
of which had implemented the 6 assessments required by the 1994 Elementary
and Secondary Education Act (ESEA) reauthorization and were testing
students in many of the additional subjects and grades required by NCLBA.
The 7 states were Colorado, Delaware, Maine, Massachusetts, North
Carolina, Texas, and Virginia. To estimate projected expenditure ranges
for all states, we used expenditures from these 7 states coupled with key
information gathered through a survey completed by each state*s assessment
director. We estimated projected state expenditures for test development,
administration, scoring, and reporting results for both assessments that
states need and assessments that states currently have in place. Our
methodology for estimating expenditures was reviewed by several internal
and external experts and their suggestions have been incorporated as
appropriate. Education officials were also briefed on our methodology and
raised no substantial concerns. As agreed with your offices, we did not

1 NCLBA authorizes funding through fiscal year 2007 for assessments.
However, consistent with the mandate for this study, we examined
expenditures between fiscal years 2002 through 2008, enabling us to more
fully capture expenditures associated with the science assessments, which
are required to be administered in school year 2007- 08.

Page 3 GAO- 03- 389 Title I

determine expenditures for alternate assessments for students with
disabilities nor expenditures for English language proficiency testing. In
addition, we did not determine the expenditures local school districts may
incur with respect to these assessments. To determine what factors account
for variation in projected expenditures, we reviewed the 7 states*
expenditures, noting the test characteristics that were associated with
specific types and levels of expenditure. We supplemented our examination
of state expenditures with interviews of test publishers and contractors
and state assessment officials in these states regarding the factors that
account for price and expenditure variation. The expenditure data that we
received were not audited. Actual expenditures may vary from projected
amounts, particularly when events or circumstances are different from
those assumed. All estimates are reported in nominal dollars unless
otherwise noted.

We conducted our work in accordance with generally accepted government
auditing standards between April 2002 and March 2003. (See app. I for more
details about our scope and methodology.)

The majority of states share two characteristics* they administer
statewide assessments rather than individual local assessments and use
customized questions to measure the content taught in the state schools
rather than questions from commercially available tests. However, states
differ in many other respects. For example, some states use assessments

that include multiple- choice questions and other states include a mixture
of multiple- choice questions and a limited number of questions that
require students to write their response, such as an essay. Many states
that use questions that require students to write their response believe
that such questions enable them to more effectively measure certain
skills, such as writing. However, others believe that multiple- choice
questions also allow them to assess such skills. In addition, some states
make actual test questions available to the public after testing but
differ with respect to the percentage of test questions they publicly
release and consequently, the number of questions they will need to
replace. States also vary in the number of new tests they reported needing
to develop to comply with the NCLBA, which ranged from 0 to 17.

We provide three estimates*$ 1.9, $3.9, and $5.3 billion* of total
spending by states between fiscal year 2002 and 2008, with the method by
which assessments are scored largely explaining the differences in our
estimates. These estimates are based on expenditures associated with new
assessments as well as existing assessments. The $1.9 billion estimate is
Results in Brief

Page 4 GAO- 03- 389 Title I

based on the assumption that all states will use multiple- choice
questions, which are machine scored. The $3.9 billion estimate is based on
the assumption that all states keep the mix of question types* whether
multiple- choice or a combination of multiple- choice and open- ended*
states reported to us. The $5.3 billion estimate is based on the
assumption that all states will use a combination of multiple- choice
questions and questions that require students to write their response,
such as an essay, which are hand scored. Several states are exploring ways
to reduce assessment expenses. This information could be beneficial to
others, however, it is currently not being broadly shared. Given that
significant expenses may be associated with testing, we are recommending
that Education facilitate the sharing of information on states*
experiences as they attempt to reduce expenses. Education agreed with our
recommendation, but raised concerns about our methodology for estimating
expenditures.

Enacted as part of President Johnson*s War on Poverty, the original Title
I program was created in 1965, but the 1994 and most recently, the 2001
reauthorization of ESEA, mandated fundamental changes to Title I. The 1994
ESEA reauthorization required states to develop state standards and
assessments to ensure that students served by Title I were held to the
same standards of achievement as other students. Some states

had already implemented assessments prior to 1994, but they tended to be
norm referenced* a student*s performance was compared to the performance
of all students nationally. The 1994 ESEA reauthorization required
assessments that were criterion referenced* students* performance was to
be judged against the state standards for what children should know and be
able to do. 2 In passing the NCLBA, Congress built on the 1994
requirements by, among other things, increasing the number of grades and
subject areas in which states were required to assess students, as shown
in table 1. NCLBA requires annual testing of students in third through
eighth grades, in mathematics and reading or language arts. It also
requires mathematics and reading or language arts testing in one of the
high school grades (10- 12). States must also assess

2 A norm referenced test evaluates an individual*s performance in relation
to the performance of a large sample of others, usually selected to
represent all students nationally in the same grade or age range.
Criterion referenced tests are assessments that measure the mastery of
specific skills or subject content and focus on the performance of an
individual as measured against a standard or criterion rather than the
performance of others taking the test. Background

Page 5 GAO- 03- 389 Title I

students in science at least once in elementary (3- 5), middle (6- 9), and
high school (10- 12). NCLBA gives the states until the 2005- 06 school
year to administer the additional mathematics and reading or language arts
assessments and until the 2007- 08 school year to administer the science
assessments (see app. II for a summary of Title I assessment
requirements).

Table 1: Number of Assessments and Subject Areas Required by the 1994 and
2001 ESEA Reauthorizations

Number of required assessments Subject 1994 ESEA

reauthorization 2001 ESEA reauthorization

Reading or language arts 3 7 Mathematics 3 7 Science 0 3

Total 6 17

Source: P. L. No. 103- 382 (1994) and P. L. No. 107- 110 (2001).

Unlike the 1994 ESEA reauthorization, NCLBA does not generally permit
Education to allow states additional time to implement these assessments
beyond the stated time frames. 3 Under the 1994 ESEA reauthorization,
Congress allowed states to phase in the 1994 ESEA assessment requirements
over time, giving states until the beginning of the 2000- 01 school year
to fully implement them with the possibility of limited time extensions.
In April 2002, we reported that the majority of states were not in
compliance with the Title I accountability and assessment provisions
required by the 1994 law. 4 Every state applying for Title I funds must
agree to implement the changes described in the 2001 act, including those
related to the additional

assessments. In addition to the regular Title I state grant, NCLBA
authorizes additional funding to states for these assessments between
fiscal year 2002 and 2007. 5 These funds are to be allocated each year to

3 The Secretary of Education may provide states 1 additional year if the
state demonstrates that exceptional or uncontrollable circumstances, such
as a natural disaster or precipitous and unforeseen decline in the
financial resources of the state prevented full implementation of the
academic assessments by the deadlines.

4 U. S. General Accounting Office, Title I: Education Needs to Monitor
States* Scoring of Assessments, GAO- 02- 393 (Washington, D. C.: Apr. 1,
2002). 5 According to Education, there are also other sources of funding
in NCLBA that states may draw upon for assessment related expenses.

Page 6 GAO- 03- 389 Title I

states, with each state receiving $3 million, regardless of its size, plus
an amount authorized based on its share of the nation*s school age
population. States must use the funds to pay the cost of developing the
additional state standards and assessments. If a state has already
developed the required standards and assessments, it may use these funds
to, among other things, develop challenging state academic content and
student academic achievement standards in subject areas other than those
required under Title I and to ensure the validity and reliability of state
assessments. NCLBA authorized $490 million for fiscal year 2002 for state
assessments and such funds as may be necessary through fiscal year 2007.
However, if in any year Congress appropriates less than the amounts shown
in table 2, states may defer or suspend testing; however, states are still
required to develop the assessments. In fiscal year 2002, states received
$387 million for assessments.

Table 2: Assessment Minimum Amounts under NCLBA Fiscal year Appropriation
benchmark

2002 $370,000,000 2003 380,000,000 2004 390,000,000 2005 400,000,000 2006
400,000,000 2007 400,000,000

Total $2.34 billion

Source: P. L. No. 107- 110 (2001).

Other organizations have provided cost estimates of implementing the
required assessments. The National Association of State Boards of
Education (NASBE) estimated that states would spend between $2.7 to $7
billion to implement the required assessments. AccountabilityWorks
estimated that states would spend about $2.1 billion. 6 States can choose
to use statewide assessments, local assessments, or

both to comply with NCLBA. States can also choose to develop their own
test questions or augment commercially available tests with questions so

6 NASBE and AccountabilityWorks made different assumptions regarding what
costs would vary with the number of students tested and which would be
invariant costs. For example, NASBE assumed that development costs would
vary by the number of students taking the

test and AccountabilityWorks did not. Additionally, AccountabilityWorks
reports having verified its assumptions with officials from two states,
while the authors of the NASBE study do not report having verified their
assumption with state officials.

Page 7 GAO- 03- 389 Title I

that they measure what students are actually taught in school. However,
NCLBA does not permit states to use commercially available tests that have
not been augmented.

NCLBA provides Education a varied role with respect to these assessments.
Education is responsible for determining whether or not states*
assessments comply with Title I requirements. States submit evidence to
Education showing that their systems for assessing students and holding
schools accountable meet Title I requirements, and Education contracts
with individuals who have expertise in assessments and Title I to review
this evidence. The experts provide Education with a report on the

status of each state regarding the degree to which a state*s system for
assessing students meets the requirements and, therefore, warrants
approval. Under NCLBA, Education can withhold federal funds provided for
state administration until Education determines that the state has
fulfilled those requirements. 7 Education*s role also includes reporting
to Congress on states* progress in developing and implementing academic
assessments, and providing states, at the state*s request, with technical
assistance in meeting the academic assessment requirements. It also
includes disseminating information to states on best practices.

The majority of states report using statewide assessments developed to
measure student learning against the content they are taught in the
states* schools, but their assessments differ in many other ways. For
example,

some states use assessments that include multiple- choice questions, while
others include a mixture of multiple- choice questions and questions that
require students to write their answer by composing an essay or showing
how they calculated a math answer. In addition, some states make actual
test questions available to the public but differ with respect to the
percentage of test questions they publicly release. Nearly all states
provide accommodations for students with disabilities and some states
report offering their assessments in languages other than English. States
also vary in the number of new tests they will need to develop to comply
with the NCLBA.

7 This amount is generally 1 percent of the amount that states receive
under Title I or $400,000, whichever is greater. States Generally

Report Administering Statewide Assessments Developed to Measure Their
State Standards, but Differ Along Other Characteristics

Page 8 GAO- 03- 389 Title I

Forty- six states currently administer statewide tests to students and 44
plan to continue using statewide tests for future tests NCLBA requires
them to add. 8 (See fig. 1.) Only 4 states* Idaho, Kansas, Pennsylvania,
and Nebraska* currently use a combination of state and local assessments
and only Iowa currently uses all local assessments.

Figure 1: The Majority of States Report They Currently Use Statewide Tests
and Plan to Continue to Do So Note: Percentages do not add to 100 because
of rounding.

The majority of states (31) report that all of the tests they currently
use consist of questions customized, that is, developed specifically to
assess student progress against their state*s standards for learning for
every grade and subject tested. (See fig. 2.) Many of the remaining states
are using different types of tests for different grades and subjects. For
example, some states are using customized tests for some grades and
subjects and

8 The District of Columbia and Puerto Rico are included in our state
totals. The Majority of States Use

Statewide Tests That They Report Are Written to Their State Standards

90% (46)

2% (1) 8%

(4)

2% (1) Current Future

85% (44)

2%( 1) 10% 4% (2)

Statewide Local Combination Don't know/ missing Source: GAO survey.

(5)

Page 9 GAO- 03- 389 Title I

commercially available tests for other grades and subjects. Seven states
reported using only commercially available tests in all the grades and
subjects they tested.

In the future, the majority of states (33) report that all of their tests
will consist of customized questions for every subject and grade.
Moreover, those states that currently use commercially available tests
report plans to replace these tests with customized tests or augment
commercially available tests with additional questions to measure what
students are taught in schools, as required by NCLBA.

Figure 2: The Majority of States Reported That They Currently Use and Plan
to Develop New Tests That Are Customized to Measure Their State*s
Standards

Note: Percentages do not add to 100 due to rounding. In the current
period, *other* includes states that reported using commercially available
tests for all grades and subjects tested that had not been augmented with
additional questions to measure state standards. These states reported
plans to augment these tests with additional questions or replace them
with customized tests.

Customized test only Other Don't know/ missing Source: GAO survey.

Current Future

Type of test 60% (31) 37% (19)

4%

(2)

63% (33) 19% (10) 17% (9)

Page 10 GAO- 03- 389 Title I

In developing their assessments, nearly all states (50) reported providing
specific accommodations for students with disabilities. 9 These often
include Braille, large print, and audiotape versions of their assessments
for visually impaired students, as well as additional time and oral
administration.

About a quarter of the states (12) report offering these assessments in
languages other than English, typically Spanish. Both small and larger
states scattered across the United States offer assessments in languages
besides English. For example, states such as Wyoming and Delaware and
large states such as Texas and New York offer Spanish language versions of
their assessments. New York and Minnesota offer their assessments in as
many as four other languages besides English. 10 While a quarter of the
states currently translate or offer assessments in languages other than
English, additional states may provide other accommodations for students
with limited English proficiency, such as additional time to take the
test, use of bilingual dictionaries, or versions of the test that limit
use of idiomatic expressions.

Thirty- six states report they currently use a combination of
multiplechoice and a limited number of open- ended questions for at least
some of the assessments they give their students. (See fig. 3.) For
example, in Florida, third grade students* math skills are assessed using
multiplechoice questions, while fifth grade students* math skills are
assessed using a combination of multiple- choice and open- ended
questions. Twelve states reported having tests that consist entirely of
multiple- choice questions. For example, all of Georgia*s and Virginia*s
tests are multiple- choice. Almost half of the states reported that they
had not made a decision about

the ratio of multiple- choice to open- ended questions on future tests. Of
the states that had made a decision, most reported plans to develop
assessments using the same types of questions they currently use.

9 Two states reported that they did not provide accommodations for
students with disabilities at the state level, however, accommodations may
have been provided at the local school level.

10 New York offers its assessments in Spanish, Korean, Haitian Creole, and
Russian and Minnesota offers its mathematics assessments in Spanish,
Hmong, Somali, and Vietnamese. States Vary in Approach to

Specific Accommodations States Are Using Different Types of Questions to
Assess Students

Page 11 GAO- 03- 389 Title I

Figure 3: The Majority of States Reported They Use a Combination of
Multiplechoice and Open- ended Questions on Their Tests, but Many States
Are Uncertain about Question Type on Future Tests

States choose to use a mixture of question types on their tests for
varying reasons. For example, some officials believe that open- ended
questions, requiring both short and long student responses, more
effectively measure certain skills such as writing or math computation
than multiple- choice questions. Further, they believe that different
question types will render a more complete measure of student knowledge
and skills. In addition, state laws sometimes require test designers to
use more than one type of question. In Maine, for example, state law
requires that all state and local assessments employ multiple measures of
student performance.

69% (36)

23% (12) Current Future

Mix of multiple- choice and written response Multiple- choice Don't know
Missing Source: GAO survey.

35% (18)

13%

(7) Question type

8%

(4)

48% (25)

Don't know 4%

(2)

Page 12 GAO- 03- 389 Title I

Slightly over half of the states currently release actual test questions
to the public, but differ in the percent of questions they release. (See
fig. 4.) Texas, Massachusetts, Maine, and Ohio release their entire tests
to the public following the tests, allowing parents and other interested
parties to see every question their children were asked. Other states,
such as New Jersey and Michigan release only a portion of their tests.
Moreover, even those states that do not release questions to the general
public may release

a portion of the questions to teachers, as does North Carolina, so that
they can better understand areas where students are having the most
difficulty, and improve instructions. States that release questions must
typically replace them with new questions.

Figure 4: States Split in Decision to Release Test Questions to the Public
Following Tests

Often, states periodically replenish their tests with new questions to
improve test security. For example, states like Florida, Kentucky,
Maryland, and South Carolina that do not release test questions, replenish
or replace questions periodically.

In addition to replenishing test items, many states use more than one
version for each of their tests and do so for various reasons. For
example, States Split as to Whether

They Make Actual Test Questions Available to the Public Following Tests

54% (28) 44% (23)

2% (1) Do release Do not release Don't know/ missing Source: GAO survey.

Page 13 GAO- 03- 389 Title I

Virginia gives a different version of its test to students who may have
been absent. Some states use multiple test versions of their high school
tests to allow those students who do not pass it to take it multiple
times. Still other states, such as Massachusetts and Maine, use multiple
versions to enable the field testing of future test questions.

States differ in the number of additional tests they reported they need to
meet NCLBA requirements, with some having all of the tests needed while
others will need to develop new tests or augment commercially available
tests with additional questions to fulfill the new requirements for a
total of 17 tests. (See table 3.) Appendix III has information on the
number of tests each state needs to develop or augment to comply with
NCLBA.

The majority of states (32) report they will need to develop or augment 9
or fewer tests and the rest (20) will need to develop or augment 10 or
more tests. Eight states* Alabama, New Mexico, Montana, South Dakota,
Idaho, West Virginia, Wisconsin, and the District of Columbia report that
they need to develop or augment all 17 tests. Maryland is also replacing a
large number of its tests (15); although its assessments were certified as
compliant with the 1994 law, the tests did not provide scores for
individual students. Although Education waived the requirement that
Maryland*s

tests provide student level data, Maryland is in the process of replacing
them so that it can provide such data, enabling parents to know how well
their children are performing on state tests.

Table 3: The Number of Tests States Reported Needing to Develop or Augment
Varies

Range in number of test states need to comply with NCLBA Number

of states

None 5 1* 3 4 4- 6 6 7- 9 17 10- 12 10 13 or more 10 Source: GAO survey.

Most states reported plans to immediately begin developing the tests,
which according to many of the assessment directors we spoke with,
typically take 2 to 3 years to develop. For example, most states reported
that by 2003 they will have developed or will begin developing the reading
and mathematics tests that must be administered by the 2005- 06 school
States Vary in the Number

of Additional Tests They Reported They Need to Develop or Augment

Page 14 GAO- 03- 389 Title I

year. Similarly, most states reported that by 2005 they will have
developed or will begin developing the science tests that must be
administered by the 2007- 08 school year.

To help them develop these tests, most states report using one or more
outside contractors to help manage testing programs. Nearly all states
report that developing, administering, scoring, and reporting will be a
collaborative effort involving contractors and state and local education
agencies. However, while states report that contractors and state

education agencies will share the primary role in developing, scoring, and
reporting new assessments, local education agencies will have the primary
role in administering the assessments.

We provide three estimates*$ 1.9, $3. 9, and $5.3 billion* of total state
spending between fiscal years 2002 and 2008 for test development,
administration, scoring, and test reporting. These figures include
estimated expenses for assessments states will need to add as well as
continuing expenditures associated with assessments they currently have in
place. The method of scoring largely explains the differences in the
estimates. However, various other factors, such as the extent to which
states release assessment questions to the public after testing and
therefore need to replace them, also affect expenditures. Between states,
however, the number of students assessed will largely explain variation in
expenditures. Moreover, because expenditures for test development are

small in relation to test administration, scoring, and reporting
(nondevelopment expenditures), we estimate that state expenditures may be
lower in the first few years when states are developing their assessments
and higher in subsequent years as states begin to administer and score
them and report the results.

We estimate that states may spend $1.9, $3.9, or $5.3 billion on Title I
assessments between fiscal years 2002 through 2008, with scoring
expenditures largely accounting for differences in our estimates. Table 4
shows total state expenditures for the 17 tests required by Title I. In
appendix IV, we also provide separate estimates for expenses associated
with the subset of the 17 assessments that states reported they did not
have in place at the time of our survey but are newly required by NCLBA.
Estimates of Spending

Driven Largely by Scoring Expenditures

Different Estimates Primarily Reflect Differences in How Assessments Are
Scored

Page 15 GAO- 03- 389 Title I

Table 4: Estimated Expenditures by States for Title I Assessments, Fiscal
Years 2002- 08

Question type Estimate Questions and scoring methods used

Multiple- choice $1.9 billion Estimate assumes that all states use
machinescored multiple- choice questions. Current question type $3.9
billion Estimate assumes that states use the mix of

question types reported in our survey. Multiple- choice and open- ended
$5.3 billion

Estimate assumes that all states use both machine- scored multiple- choice
questions and some hand scored open- ended questions. Source: GAO
projections based on state assessment plans and characteristics and
expenditure data gathered from 7 states.

The $1.9 billion estimate assumes that all states will use multiple-
choice questions on their assessments. Multiple- choice questions can be
scored by scanning machines, making them relatively inexpensive to score.
For instance, North Carolina, which uses multiple- choice questions on all
of its assessments and machine scores them, spends approximately $0.60 to
score each assessment.

The $3.9 billion estimate assumes that states will implement assessments
with questions like the ones they currently use or plan to use based on
state education agency officials* responses to our survey. However, 25
states reported that they had not made final decisions about question type
for future assessments. Thus, the types of questions states ultimately use
may be different from the assessments they currently use or plan to use.

Finally, the $5.3 billion estimate assumes that all states will implement
assessments with both multiple- choice and open- ended questions. Answers
to open- ended questions, where students write out their responses, are
typically read and scored by people rather than by machines, making them
much more expensive to score than answers to multiple- choice questions.
We found that states using open- ended questions had much higher scoring
expenditures per student than states using multiple- choice questions, as
evidenced in the states we visited, as shown in figure 5. 11 For example,
Massachusetts, which uses many openended questions on its Title I
assessments, spends about $7.00 to score each assessment. Scoring
students* answers to open- ended questions in Massachusetts involves
selecting and training people to read

11 In Texas and Colorado, we were unable to separate scoring expenditures
from other types of expenditures.

Page 16 GAO- 03- 389 Title I

and score the answers, assigning other people to supervise the readers,
and providing a facility where the scoring can take place. In cases where
graduation decisions depend in part on a student*s score on the
assessment, the state requires that two or three individuals read and
score the student*s answer. By using more than one reader to score
answers, officials ensure consistency between scorers and are able to
resolve disagreements about how well the student performed.

Figure 5: Estimated Scoring Expenditures Per Assessment Taken for Selected
States, Fiscal Year 2002

We estimate that, for most states, much of the expense associated with
assessments will be related to test scoring, administration, and
reporting, not test development, which includes such expenses as question
development and field testing. 12 (See table 5.) In Colorado, for example,

12 This may not be true for smaller states because they may have fewer
assessments to administer, score, and report.

Estimated scoring expenditures per assessment taken 0 2

4 6

8 Delaware

Maine North Carolina

Virginia Massachusetts

Source: GAO analysis of expenditure data provided by state education
agencies. Open- ended and multiple- choice

Primarily multiple- choice

Page 17 GAO- 03- 389 Title I

test administration, scoring, and reporting expenditures comprise 89
percent of the total expenditures, while test development expenditures
comprised only 11 percent. (See app. V for our estimates of development
and nondevelopment expenditures by state.)

Table 5: Estimated Total Expenditures for Test Development Are Lower Than
for Test Administration, Scoring, and Reporting

In millions Multiple- choice Current question type Multiple- choice and
open- ended

Development $668 $706 $724 Administration, scoring, and reporting 1,233
3,237 4,590

Total $1,901 $3,944 $5,313

Source: GAO projections based on state assessment plans and
characteristics and expenditure data gathered from 7 states.

While the scoring method explains a great deal of the variation in
expenditures among states, other factors are likely to affect
expenditures. These factors include the number of different test versions
used, the extent to which the state releases assessment questions to the
public after testing, fees for using copyrighted material, and factors
unique to the state. (See fig. 6.) For example, states that use multiple
test versions will have higher expenditures than those that have one.
Massachusetts used 24 different test versions for many of its assessments
and spent approximately $200,000 to develop each assessment. Texas used
only 1 version for its assessments and spent approximately $60,000 per
assessment. In addition, states that release test items to the public or
require rapid reporting of student test scores are likely to have higher
expenditures than states that do not because they need to replace these
items with new ones to protect the integrity of the tests and assign
additional staff to more rapidly score the assessments by the specified
time frame. States that customize their assessments may have higher
expenditures than states that augment commercially available tests.
Moreover, factors unique to the state may affect expenditures. Maine,
which had one of the lowest assessment development expenses of all of the
states we visited (about $22,000 per assessment), has a contract with a

nonprofit testing company. Between states, the number of students tested
generally explains much of the variation in expenditures, particularly
when question types are similar. States with large numbers of students
tested will generally have higher expenditures than states with fewer
students. Various Factors are Likely

to Affect Expenditures for Title I Assessments

Page 18 GAO- 03- 389 Title I

Figure 6: Various Factors Are Likely to Affect What States Spend on Title
I Assessments Using the benchmark funding levels specified in NCLBA, we
estimate that

these amounts would cover varying portions of estimated expenditures. (See
table 6.) In general, these benchmark amounts would cover a larger
percentage of the estimated expenditures for states that choose to use
multiple- choice tests. To illustrate, we estimated that Alabama would
spend $30 million if it continued to use primarily multiple- choice
questions, but $73 million if the state used assessments with both
multiplechoice and open- ended questions. The specified amount would cover
151 percent of Alabama*s estimated expenditures if it chose to use all
multiple- choice questions, but 62 percent if the state chose to use both
multiple- choice and open- ended questions. Benchmark Amounts in

NCLBA Will Cover Varying Portions of States* Estimated Expenditures and
Amount Covered Will Vary Primarily by Type of Test Questions States Use

Use of open- ended questions Number of students taking assessments
Customizing assessments to align with state standards Extent of public
release of questions Number of different versions of the assessments
Faster turnaround time for scoring Factors unique to the state

Likely affect on estimated Factor expenditures

Source: State education agency official interviews.

Page 19 GAO- 03- 389 Title I

Table 6: Total Estimated Expenditures by States for Title I Assessments,
Fiscal Years 2002- 08 Estimates (in millions) Appropriation benchmark as
percent of

estimated expenses Multiplechoice Current

question type

Multiplechoice and open- ended

Appropriation benchmark (in millions) a Multiplechoice Current

question type

Multiple- choice and openended

Alabama $30 $30 $73 $46 151% 151% 62% Alaska 17 25 28 26 154 106 93
Arizona 39 108 108 51 132 47 47 Arkansas 23 42 53 37 158 88 70 California
178 235 632 219 123 93 35 Colorado 32 87 87 46 145 53 53 Connecticut 28 68
68 41 147 59 59 Delaware 14 24 24 26 183 106 106 District of Columbia 13
13 17 24 184 184 144 Florida 83 211 281 102 123 48 36 Georgia 54 54 174 67
124 124 39 Hawaii 17 31 31 28 162 91 91 Idaho 18 23 30 30 167 131 98
Illinois 65 164 211 92 141 56 44 Indiana 40 113 113 56 140 49 49 Iowa 24
62 62 38 158 62 62 Kansas 23 36 51 38 164 106 73 Kentucky 28 62 71 43 155
70 61 Louisiana 31 81 81 49 158 60 60 Maine 18 33 33 29 159 86 86 Maryland
35 91 91 51 146 56 56 Massachusetts 38 109 109 55 144 50 50 Michigan 57
177 177 80 140 45 45 Minnesota 34 91 91 51 149 56 56 Mississippi 25 63 63
39 154 61 61 Missouri 36 99 99 54 150 54 54 Montana 18 28 29 27 149 97 94
Nebraska 18 34 34 32 177 93 93 Nevada 21 26 45 33 152 125 72 New Hampshire
17 32 32 29 168 92 92 New Jersey 43 127 127 67 153 53 53 New Mexico 21 39
41 33 155 84 81 New York 83 276 276 121 146 44 44 North Carolina 49 49 152
65 132 132 43 North Dakota 16 23 23 26 162 109 109 Ohio 55 171 171 86 158
50 50 Oklahoma 27 37 66 42 156 114 63

Page 20 GAO- 03- 389 Title I

Estimates (in millions) Appropriation benchmark as percent of estimated
expenses

Multiplechoice Current

question type

Multiplechoice and open- ended

Appropriation benchmark (in millions) a Multiplechoice Current

question type

Multiple- choice and openended

Oregon 28 28 70 40 145 145 57 Pennsylvania 58 162 181 87 150 54 48 Puerto
Rico 28 28 70 47 167 167 67 Rhode Island 17 28 28 27 161 98 98 South
Carolina 31 82 85 43 139 53 51 South Dakota 18 18 27 26 145 145 97
Tennessee 33 33 85 52 158 158 61 Texas 126 232 441 147 116 63 33 Utah 24
44 61 37 154 84 60 Vermont 16 25 25 25 155 102 102 Virginia 43 60 129 59
136 99 46 Washington 41 118 118 55 135 47 47 West Virginia 23 23 43 31 135
135 72 Wisconsin 29 66 72 53 180 80 73 Wyoming 15 21 21 25 171 119 119

Total $1,901 $3,944 $5,313 $2,733 144% 69% 51%

Source: GAO analysis. a Figures in these columns are based largely on
benchmark funding levels in NCLBA. If Congress appropriates less than the
benchmark amounts, states may defer test administration. For fiscal years
2002 and 2003, however, we used the actual appropriation. In addition,
because we were mandated to estimate spending for fiscal year 2008, for
purposes of this analysis, we assumed a fiscal year 2008 benchmark of $400
million, the same amount as for fiscal years 2005, 2006, and 2007. It
should be noted, however, that Congress has not authorized funding past
fiscal year 2007, when Title I would be reauthorized. Benchmarks by state
were calculated based on the formula in NCLBA for allocating assessment
funds to the states.

Estimated expenditures are likely to be lower in the first few years when
tests are being developed and increase in later years when greater numbers
of tests are administered, scored, and reported. As a result, the
benchmark funding amounts in NCLBA would cover a larger percentage of
estimated expenditures in the first few years. Under some circumstances,
the funding benchmarks in NCLBA exceed estimated state expenditures. For
example, as shown in figure 7, the fiscal year 2002 allocation would more
than cover all of the estimated expenses if all states were to use
multiple- choice questions or continue with the types of questions they
currently use. If all states were to choose to use a mixture of
multiplechoice and open- ended questions, the most expensive option,
fiscal year 2002 funding would cover 84 percent of states* total
expenditures. We estimate a similar pattern for fiscal year 2003. (See
app. VI for fiscal year 2002 through 2008 estimated expenditures for each
question type.) In fiscal year 2007 and 2008, benchmark funding would
continue to cover all of the estimated expenditures if all states were to
use all multiple Total Expenditures Likely

to Be Lower in the First Few Years, Increasing Over Time as States Begin
to Administer, Score, and Report Additional Assessments

Page 21 GAO- 03- 389 Title I

choice questions, about two- thirds of estimated expenditures if all
states continued using their current mix of questions, and a little over
50 percent of estimated expenditures if all states were to use a mixture
of question types, the most expensive option.

Figure 7: Total Expenditures Likely to Be Lower in First Few Years and
Benchmark Funding in NCLBA Estimated to Cover Most of Expenditures in
First Few Years

Some states are exploring ways to control expenses related to assessments
and their experiences may provide useful information to other states about
the value of various methods for controlling expenditures. Recently,
several states, in conjunction with testing industry representatives, met
to discuss ways of reducing test expenditures. For example, the group
discussed a range of possible options for reducing expenditures, including
computer- administered tests; commercially available tests that can be
customized to states standards by adding additional questions;
computerized scoring of written responses, and Opportunities May Exist to

Share Information on Efforts to Reduce Testing Expenditures

Dollars in millions 0 200

400 600

800 1000

2002 2003 2004 2005 2006 Fiscal year

Benchmark appropriations Multiple- choice Current question type Multiple-
choice and open- ended

2007 2008

Source: GAO analysis.

Page 22 GAO- 03- 389 Title I

computer scanning of students* written responses. Information about
individual states experiences as they attempt to reduce expenses could
benefit other states. However, such information is currently not
systematically shared.

The 1994 and 2001 ESEA reauthorizations raised student assessments to a
new level of importance. These assessments are intended to help ensure
that all students are meeting state standards. Congress has authorized
funding to assist states in developing and implementing these assessments.
We estimate that federal funding benchmarks in NCLBA will cover a larger
percentage of expenses in the first few years when states are developing
their assessments, with the covered percentage decreasing as states begin

to administer, score, and report the full complement of assessments.
Moreover, the choices states make about how they will assess students will
influence expenditures. Some states are investigating ways to reduce the
expenses, but currently information on states* experiences in attempting
to reduce expenses is not broadly shared. We believe states could benefit
from information sharing.

Given the large federal investment in testing and the potential for
reducing test expenditures, we recommend that Education use its existing
mechanisms to facilitate the sharing of information on states* experiences
as they attempt to reduce expenses.

The Department of Education provided written comments on a draft of this
report, which we have summarized below and incorporated in the report as
appropriate. (See app. VII for agency comments.) Education agreed with our
recommendation, stating that it looks forward to continuing and enhancing
its efforts to facilitate information sharing that might help states
contain expenses. However, Education raised concerns about our
methodology, noted the availability of additional federal

resources under ESEA that might support states* assessment efforts, and
pointed out that not all state assessment costs are generated by NCLBA.

With regard to our estimates, we have confidence that our methodology is
reasonable and provides results that fairly represent potential
expenditures based on the best available information. Education*s comments
focus on the uncertainties that are inherent in estimation of any kind*
the necessity of assumptions, the possibility of events or trends not
readily predicted, and other potential sources of error that are
acknowledged in the report* without proposing an alternative methodology.
Because of the uncertainty, we produced three estimates Conclusions

Recommendation Agency Comments

Page 23 GAO- 03- 389 Title I

instead of one. In developing our approach, we solicited comments from
experts in the area and incorporated their suggestions as appropriate. We
also discussed our estimation procedures with Education staff, who raised
no significant concerns. Second, Education cites various other sources of
funds that states might use to finance assessments. While other sources
may be available, we focused primarily on the amounts specifically
authorized for assessments in order to facilitate their comparison to
estimated expenses and because they are the minimum amounts that Congress
must appropriate to ensure that states continue to develop as well as
implement the required assessments.

We are sending copies of this report to the Secretary of Education,
relevant congressional committees, and other interested parties. Please
contact me on (202) 512- 7215 or Betty Ward- Zukerman on (202) 512- 2732
if you or your staff have any questions about this report. In addition,
the

report will be available at no charge on GAO*s Web site at http:// www.
gao. gov. Other GAO contacts and staff acknowledgments are listed in
appendix VIII.

Marnie S. Shaul, Director Education, Workforce and Income Security Issues

Appendix I: Objectives, Scope, and Methodology

Page 24 GAO- 03- 389 Title I

The objectives of this study were to provide information on the basic
characteristics of Title I assessments, and to estimate what states would
likely spend on Title I assessments between fiscal year 2002 and 2008, and
identify factors that explain variation in estimated expenditures. To

address the first objective, we collected information from a survey sent
to the 50 states, the District of Columbia, and Puerto Rico, and reviewed
documentation from state education agencies and from published studies
detailing the characteristics of states* assessments. To address the
second objective, we collected detailed assessment expenditure information
from 7 states, interviewed officials at state education agencies,
discussed cost factors with assessment contractors, and estimated
assessment expenditures under three different scenarios. The methods we
used to address the objectives were reviewed by several external
reviewers, and we incorporated their comments as appropriate. This
appendix discusses the scope of the study, the survey, and the methods we
used to estimate assessment expenditures.

We surveyed all 50 states, the District of Columbia, and Puerto Rico, all
of which responded to our survey. We asked them to provide information
about their Title I assessments, including the characteristics of current
and planned assessments, the number and types of new tests they needed to
develop to satisfy No Child Left Behind Act (NCLBA) requirements, when
they planned to begin developing the new assessments, the types of
questions on their assessments, and their use of contractors. We also
reviewed documentation from several states about their assessment programs
and published studies detailing the characteristics of states*
assessments. This study estimates likely expenditures on Title I
assessments by states

between fiscal year 2002 and 2008, and identifies factors that may explain
variation in the estimates. It does not estimate expenditures for
alternate assessments for students with disabilitiess for English language
proficiency testing, or expenditures incurred by school districts. 1
Instead, we estimated expenses states are expected to incur based on
expenditure data obtained for this purpose from 7 states combined with
data on these and other states* assessment plans and characteristics
obtained through a

1 The study also does not estimate the opportunity costs of assessments.
Appendix I: Objectives, Scope, and

Methodology Providing Information on the Basic Characteristics of Title I
Assessments

Estimating Assessment Expenditures and Explaining Variation in the
Estimates

Appendix I: Objectives, Scope, and Methodology

Page 25 GAO- 03- 389 Title I

survey. 2 In the 7 states, we requested information and documentation on
expenditures in a standard set of areas, met with state officials to
discuss the information and asked that they review our subsequent analysis
of information regarding their state. The expenditure data that we
received from the 7 states were not audited. Moreover, actual expenditures
may

vary from projected amounts, particularly when events or circumstances are
different from those assumed, such as changes in the competitiveness of
the market for student assessment or changes in assessment technology. We
selected 7 states that had assessments in place in many of the grades

and subjects required by the NCLBA from the 17 states with assessment
systems that had been certified by Education as in compliance with
requirements of the Improving America*s Schools Act of 1994 when we began
our work. We included states with varying student enrollments, including 2
states with relatively small numbers of students. The states we selected
were Colorado, Delaware, Maine, Massachusetts, North Carolina, Texas and
Virginia. (See table 7 for information about the selected states.) Table
7: States Selected for Study

Number of assessments State Date approved

by Education Number of students Reading

(out of 7) Math (out of 7) Science

(out of 3) Total

Colorado July 2001 724,508 7 5 1 13 Delaware December 2000 114,676 7 7 3
17 Maine February 2002 207,037 3 3 3 9 Massachusetts January 2001 975,150
5 4 3 12 North Carolina June 2001 1,293,638 7 7 0 14 Texas March 2001
4,059,619 7 7 2 16 Virginia January 2001 1,144,915 4 4 3 11 Source: U. S.
Department of Education, National Center for Education Statistics, and
state education agencies. We collected detailed assessment expenditure
information from officials

in the 7 states. We obtained actual expenditures on contracts and state
assessment office budget expenditures for fiscal year 2002 for all 7
states

2 Because our expenditure data were limited to 7 states, our estimates may
be biased. For example, if the 7 states we selected had higher average
development expenditures per ongoing assessment than the average state,
then our estimate of development expenditures would be biased upwards.
Selection of 7 States Collection of Expenditure

Information from 7 States

Appendix I: Objectives, Scope, and Methodology

Page 26 GAO- 03- 389 Title I

and for previous years in 4 states. 3 In site visits to the 7 states, we
interviewed state education agency officials who explained various
elements of their contracts with assessment publishing firms and the
budget for the state*s assessment office. To the extent possible, we
collected expenditure data, distinguishing expenditures for assessment
development from expenditures for assessment administration, scoring, and
reporting, because expenditures vary differently between these two
expenditure categories. Assessment development expenditures vary with the
number of assessments while administration, scoring, and reporting
expenditures vary with the number of students taking the assessments. (See
table 8 for examples of expenditures.)

Table 8: Examples of Assessment Expenditures Type of expenditure Example
of expenditure

Development Question writing Question review (e. g., for bias)
Administration Printing and delivering assessment

booklets Scoring Scanning completed booklets into scoring machines

Reporting Producing individual score reports Source: State education
agencies. Using annual assessment expenditures for all 7 states, the
number of

assessments developed and implemented, and the number of students who took
the assessments, we calculated average expenditures for ongoing
development (assessments past their second year of development) and
average expenditures for administration, scoring, and reporting for each
state. (See table 9.)

3 We were unable to obtain information on personnel expenditures from 5 of
the 7 states, and so we did not include personnel expenditures in our
analysis. In the 2 states in which we obtained personnel expenditures,
such expenditures were a relatively small part of the assessment budget.
Calculation of Averages for Development and for

Administration, Scoring, and Reporting

Appendix I: Objectives, Scope, and Methodology

Page 27 GAO- 03- 389 Title I

Table 9: Average Annual Expenditures for the 7 States (adjusted to 2003
dollars) State

Average development expenditures (per ongoing assessment)

Average expenditures for administration, scoring, and reporting (per
assessment taken)

Both multiple- choice and open- ended

questions Multiplechoice

questions

Colorado $72,889 $10.35

Appendix I: Objectives, Scope, and Methodology

Page 28 GAO- 03- 389 Title I

mean of the average annual expenditures for Colorado, Delaware, Maine, and
Massachusetts, states that use both types of questions.

To estimate development expenditures, we obtained information from each
state regarding the number of assessments it needed to develop, the year
in which it planned to begin development of each new assessment, and the
number of assessments it already had. For each assessment the state
indicated it needed to develop, we estimated initial development
expenditures beginning in the year the state said it would begin
development and also for the following year because interviews with
officials revealed that developing an entirely new assessment takes
approximately 2 to 3 years. For the 7 states that provided data, we were
typically not able to separate expenditures for new test development from
expenditures for ongoing test development. Where such data were

available, we determined that development expenses for new assessments
were approximately three times the expense of development expenses for
ongoing assessments, and we used that approximation in our estimates. For
each state each year, we multiplied the number of tests in initial

development by three times the average ongoing development expenditure to
reflect that initial development of assessments is more expensive than
ongoing development. 4 We multiplied the number of ongoing tests by the
average ongoing development expenditure. The sum of these two products
provides a development expenditure for each state in each year and
provides a total development estimate. We calculated three estimates as
follows:

 using the expenditure information from states that use multiple- choice
questions, we produced a lower estimate;  using the information from the
state survey on the types of tests they

planned to develop (some indicated both open- ended/ multiple- choice
tests and some multiple- choice), we produced a middle estimate; 5 and 
using the expenditure information from the states that use open- ended and

multiple- choice questions, we produced the higher estimate. 4 We found
estimates were not sensitive to changes in assumptions regarding
development costs, partly because they proved to be a generally small
portion of overall expenses. 5 For states that reported that they did not
know the kinds of question they would use on future tests, we assumed that
future test would be the same as they currently use. Where

data were missing, we assumed that states would use assessments with both
multiplechoice and open- ended questions, potentially biasing our
estimates upward. Estimating Development

Expenditures

Appendix I: Objectives, Scope, and Methodology

Page 29 GAO- 03- 389 Title I

To produce an estimate for administration, scoring, and reporting, we used
three variables: the average number of students in a grade; the number of
administered assessments; and the average administration, scoring, and
reporting expenditure per assessment taken. We calculated the average

number of students in a grade in each year using data from the National
Center for Education Statistics* Common Core of Data for 2000- 01 and
their Projection of Education Statistics to 2011. We obtained data on the
number of administered assessments from our state education agency

survey. Data on average expenditures come from the states in which we
collected detailed expenditure information.

For each state in each year, we multiplied the average number of students
in a grade by the number of administered assessments and by the
appropriate average assessment expenditure. Summing over states and years
provided a total estimate for administration, scoring, and reporting. As
above, we performed these calculations, using the expenditure information
from multiple- choice states to produce the lower estimate, using the
information from the state survey and expenditure information from both
combination and multiple- choice states to produce a middle estimate, and
using the expenditure information from the combination states to produce
the higher estimate. We also estimated what states are likely to spend on
the assessments that states did not have in place at the time of our
survey, but are required by NCLBA, using the same basic methodology. Table
10 provides an overview of our approach to estimating states* likely
expenditures on Title I assessments. Estimating Administration, Scoring,
and Reporting

Expenditures

Appendix I: Objectives, Scope, and Methodology

Page 30 GAO- 03- 389 Title I

Table 10: Estimated Expenditures to Implement Title I Assessments in a
Given Year

A Total estimated development expenditure for ongoing assessments

= Number of ongoing assessments  * Average development expenditure for
each ongoing assessment

B Total estimated development expenditure for new assessments = Number of
new assessments  * Three times the average development expenditure for
each ongoing assessment

C Total estimated expenditures for administration, scoring, and reporting
(ongoing and new assessments)

= Average number of students in each grade  * Average administration,
scoring, and reporting

expenditure for each assessment taken, times the number of assessments
administered, for each ongoing and new assessment

A + B + C = States* estimated expenditures to implement Title I
assessments

Source: GAO analysis. We conducted our work in accordance with generally
accepted government auditing standards between April 2002 and March 2003.

Appendix II: Accountability and Assessment Requirements under the 1994 and
2001 Reauthorizations of Title I

Page 31 GAO- 03- 389 Title I

Requirements for 1994 Requirements for 2001 Developing standards for
content and performance Develop challenging standards for what students
should know in

mathematics and reading or language arts. In addition, for each of these
standards, states should develop performance standards representing three
levels: partially proficient, proficient, and advanced. The standards must
be the same for all children. If the state does not have standards for all
children, it must develop standards for Title I children that incorporate
the same skills, knowledge, and performance expected of other children.

In addition, develop standards for science content by 2005- 06. The same
standards must be used for all children.

Implementing and administering assessments Develop and implement
assessments aligned with the content and performance standards in at least
mathematics and reading or language arts.

Add assessments aligned with the content and performance standards in
science by the 2007- 08 school year. These science assessments must be
administered at some time in each of the following grade ranges: grades 3
through 5, 6

through 9, and 10 through 12. Use the same assessment system to measure
Title I students as the state uses to measure the performance of all other
students. In the absence of a state system, a system that meets Title I
requirements must be developed for use in all Title I schools.

Use the same assessment system to measure Title I students as the state
uses to measure the performance of all other students. If the state
provides evidence to the Secretary that it lacks authority to adopt a
statewide system, it may meet the Title I requirement by adopting an
assessment system on a statewide basis and limiting its applicability to
Title I students or by ensuring that the Title I local educational agency
(LEA) adopts standards and aligned assessments. Include in the assessment
system multiple measures of student performance, including measures that
assess higher order thinking skills and understanding.

Unchanged Administer assessments for mathematics and reading in each of
the following grade spans: grades 3 through 5, 6 through 9, and 10 through
12.

Administer reading and mathematics tests annually in grades 3 through 8,
starting in the 2005- 06 school year (in addition to the assessments
previously required sometime within grades 10 through 12). States do not
have to administer mathematics and reading or language arts tests annually
in grades 3 through 8 if Congress does not provide specified amounts of
funds to do so, but states have to continue to work on the development of
the standards and assessments for those grades. Have students in grades 4
and 8 take the National Assessment of Educational Progress examinations in
reading and mathematics every other year beginning in 2002- 03, as long as
the federal government pays for it.

Assess students with either or both criterion referenced assessments and
assessments that yield national norms. However, if the state uses only
assessments referenced against national norms at a particular grade, those
assessments must be augmented with additional items as necessary to
accurately measure the depth and breath of the state*s academic contents
standards.

Unchanged Assess students with statewide, local, or a combination of state
and local assessments. However, states that use all local or a combination
of state and local assessments, must ensure, among other things, such
assessments are aligned with the state*s academic content standards, are
equivalent to one another, and

Unchanged Appendix II: Accountability and Assessment Requirements under
the 1994 and 2001

Reauthorizations of Title I

Appendix II: Accountability and Assessment Requirements under the 1994 and
2001 Reauthorizations of Title I

Page 32 GAO- 03- 389 Title I

Requirements for 1994 Requirements for 2001

enable aggregation to determine whether the state has made adequate yearly
progress Implement controls to ensure the quality of the data collected
from the assessments. Unchanged Including students with limited English
proficiency and with disabilities in assessments Assess students with
disabilities and limited English proficiency

according to standards for all other students. Provide reasonable
adaptations and accommodations for students with disabilities or limited
English proficiency to include testing in the language and form most
likely to yield accurate and reliable information on what they know and
can do.

By 2002- 03 annually assess the language proficiency of students with
limited English proficiency. Students who have attended a U. S. school for
3 consecutive years must be tested in English unless an individual
assessment by the district shows testing in a native language will be more
reliable.

Reporting data Report assessment results according to the following: by
state, local educational agency (LEA), school, gender, major racial and
ethnic groups, English proficiency, migrant status, disability, and
economic disadvantage.

Unchanged. LEAs must produce for each Title I school a performance profile
with disaggregated results and must publicize and disseminate these to
teachers, parents, students, and the community. LEAs must also provide
individual student reports, including test scores and other information on
the attainment of student performance standards.

Provide annual information on the test performance of individual students
and other indicators included in the state accountability system by 2002-
03. Make this annual information available to parents and the public and
include data on teacher qualifications. Compare high- and low- poverty
schools with respect to the percentage of classes taught by teachers who
are *highly qualified,* as defined in the law, and conduct similar
analyses for subgroups listed in previous law.

Measuring improvement Use performance standards to establish a benchmark
for improvement referred to as *adequate yearly progress.* All LEAs and
schools must meet the state*s adequate yearly progress standard, for
example, having 90 percent of their students performing at the proficient
level in mathematics. LEAs and schools must show continuous progress
toward meeting the adequate yearly progress

standard. The state defines the level of progress a school or LEA must
show. Schools that do not make the required advancement toward the
adequate yearly progress standard can face consequences, such as the
replacement of the existing staff.

In addition to showing gains in the academic achievement of the overall
school population, schools and districts must show that the following
subcategories of students have made gains in their academic achievement:
pupils who are economically disadvantaged, have limited English
proficiency, are disabled, or belong to a major racial or ethnic group. To
demonstrate gains among these subcategories of students, school districts
measure their progress against the state*s definition of adequate yearly
progress. States have 12 years for all students to perform at the
proficient level.

Consequences for not meeting the adequate yearly progress standard LEAs
are required to identify for improvement any schools that fail to make
adequate yearly progress for 2 consecutive years and provide technical
assistance to help failing schools develop and implement required
improvement plans. After a school has failed to meet the adequate yearly
progress standard for 3 consecutive years, LEAs must take corrective
action to improve the school.

New requirements are more specific as to what actions an LEA must take to
improve failing schools. Actions are defined for each year the school
continues to fail leading up to the 5th year of failure when a school may
be restructured by changing to a charter school, replacing school staff,
or state takeover of the school administration. The new law also provides
that LEAs offer options to children in failing schools. Depending on the
number of years a school has been designated for

improvement, these options may include going to another public school with
transportation paid by the LEA or using Title I funds to pay for
supplemental help. Source: P. L. No. 103- 382 (1994) and Pub. L No. 107-
110 (2001).

Appendix III: Number of Tests States Reported They Need to Develop or
Augment to Comply with NCLBA (as of March 2003) Page 33 GAO- 03- 389 Title
I

State Number of tests needed

Alabama 17 Alaska 9 Arizona 9 Arkansas 9 California 5 Colorado 4
Connecticut 8 Delaware 0 District of Columbia 17 Florida 0 Georgia 0
Hawaii 9 Idaho 17 Illinois 6 Indiana 9 Iowa 0 Kansas 11 Kentucky 8
Louisiana 8 Maine 8 Maryland 15 Massachusetts 6 Michigan 8 Minnesota 11
Mississippi 3 Missouri 8 Montana 17 Nebraska 11 Nevada 11 New Hampshire 9
New Jersey 10 New Mexico 17 New York 8 North Carolina 3 North Dakota 11
Ohio 8 Oklahoma 8 Oregon 6 Pennsylvania 11 Puerto Rico 10 Rhode Island 11
South Carolina 3

Appendix III: Number of Tests States Reported They Need to Develop or
Augment to Comply with NCLBA (as of March 2003)

Appendix III: Number of Tests States Reported They Need to Develop or
Augment to Comply with NCLBA (as of March 2003) Page 34 GAO- 03- 389 Title
I

State Number of tests needed

South Dakota 17 Tennessee 15 Texas 1 Utah 0 Vermont 9 Virginia 6
Washington 8 West Virginia 17 Wisconsin 17 Wyoming 11 Source: GAO survey.

Appendix IV: Estimates of Assessment Expenditures NCLBA Required, but Not
in Place at the Time of Our Survey, FY 2002- 08 Page 35 GAO- 03- 389 Title
I

Table 11 provides estimates of assessment expenditures states may incur
for grades and subjects they reported they would need to add to meet the
additional assessment requirements under NCLBA. These estimates do not
include any expenditures for continuing development or administration of
assessments in grades and subjects already included in states* reported
assessment program, unless states indicated plans to replace its existing
assessments. Estimates reflect total expenditures between fiscal year 2002
and 2008, and are based on the assumptions we made regarding question
types.

Table 11: Estimates of Expenditures for the Assessments Required by NCLBA
That Were Not in Place at the Time of Our Survey, Fiscal Years 2002- 08
Dollars in billions Question type Estimate Questions and scoring methods
used

Multiple- choice $0.8 Estimate assumes that all states use machine- scored
multiple- choice questions.

Current question type $1.6 Estimate assumes that states use the mix of

question types they reported in our survey. Multiple- choice and open-
ended

$2.0 Estimate assumes that all states use both

machine scored multiple- choice questions and some hand scored open- ended
questions. Source: GAO. Note: Projections based on state assessment plans
and characteristics and expenditure data

gathered from 7 states.

Appendix IV: Estimates of Assessment Expenditures NCLBA Required, but Not
in Place at the Time of Our Survey, FY 2002- 08

Appendix V: State Development and Nondevelopment Estimates

Page 36 GAO- 03- 389 Title I

Table 12 provides test development and nondevelopment expenditures by
state between fiscal year 2002- 08. Test development estimates reflect
expenditures associated with both new and existing tests. Nondevelopment
expenditures reflect expenditures associated with administration, scoring,
and reporting of results for both new and existing assessments. Table 12:
Estimates by State, Development, and Nondevelopment Expenditures

Dollars in millions Multiple- choice and open- ended Current question type
Multiple- choice Development Nondevelopment Development Nondevelopment
Development Nondevelopment

Alabama $16 $57 $15 $15 $15 $15 Alaska 15 14 15 10 13 4 Arizona 15 93 15
93 14 25 Arkansas 14 39 14 28 13 10 California 13 619 12 223 12 166
Colorado 13 74 13 74 12 20 Connecticut 14 54 14 54 13 15 Delaware 12 13 12
13 11 3 District of Columbia 13 4 12 1 12 1 Florida 12 269 12 200 11 72
Georgia 12 162 11 44 11 44 Hawaii 14 17 14 17 13 5 Idaho 15 16 14 9 14 4
Illinois 13 198 13 151 12 53 Indiana 14 99 14 99 13 27 Iowa 12 50 12 50 11
14 Kansas 14 37 13 23 13 10 Kentucky 14 58 14 48 13 16 Louisiana 14 67 14
67 13 18 Maine 14 19 14 19 13 5 Maryland 16 75 16 75 15 20 Massachusetts
13 96 13 96 12 26 Michigan 14 163 14 163 13 44 Minnesota 15 76 15 76 14 20
Mississippi 13 51 13 51 12 14 Missouri 14 85 14 85 13 23 Montana 16 13 16
12 15 3 Nebraska 13 21 13 21 12 6 Nevada 14 31 13 13 13 8 New Hampshire 13
18 13 18 12 5 New Jersey 14 113 14 113 13 30

Appendix V: State Development and Nondevelopment Estimates

Appendix V: State Development and Nondevelopment Estimates

Page 37 GAO- 03- 389 Title I

Dollars in millions Multiple- choice and open- ended Current question type
Multiple- choice Development Nondevelopment Development Nondevelopment
Development Nondevelopment

New Mexico 16 25 16 24 15 7 New York 14 262 14 262 13 70 North Carolina 13
139 12 37 12 37 North Dakota 14 9 14 9 13 2 Ohio 13 158 13 158 12 42
Oklahoma 14 53 13 24 13 14 Oregon 14 57 13 15 13 15 Pennsylvania 15 166 15
147 14 45 Puerto Rico 14 56 13 15 13 15 Rhode Island 14 13 14 13 13 4
South Carolina 13 73 13 70 12 19 South Dakota 17 10 15 3 15 3 Tennessee 15
70 14 19 14 19 Texas 12 429 11 221 11 115 Utah 12 50 12 33 11 13 Vermont
15 10 15 10 14 3 Virginia 13 116 12 48 12 31 Washington 14 104 14 104 13
28 West Virginia 17 26 16 7 16 7 Wisconsin 15 57 15 51 14 15 Wyoming 14 7
14 7 13 2

Total $724 $4,590 $706 $3,237 $668 $1,233

Source: GAO estimates based on state assessment plans and characteristics
and expenditure data gathered from 7 states.

Appendix VI: Fiscal Years 2002- 08 Estimated Expenditures for Each
Question Type

Page 38 GAO- 03- 389 Title I

Table 13 provides estimates for each question type and the benchmark
appropriations by fiscal years from 2002 through 2008. Each estimate
reflects assumptions about the type of questions on the assessments. For
example, the multiple- choice estimate assumes that all states will use
assessments with only multiple- choice questions. These estimates also
assume that states implement the assessment plans reported to us. The
benchmark appropriation is based on actual appropriations in 2002 and 2003
and on the benchmark funding level in NCLBA for 2004- 07. We assumed a
benchmark of $400 million in 2008, the same as in 2005, 2006, and 2007.

Table 13: Estimated Expenditures for Each Question Type, Fiscal Years
2002- 08 Fiscal year (in millions)

Question type 2002 2003 2004 2005 2006 2007 2008 Total

Multiple- choice $165 237 288 291 293 308 318 $1,901

Current question type $324 442 572 615 633 665 692 $3,944

Multiplechoice and open- ended $445 586 761 824 855 903 941 $5,313

Benchmark appropriation $366 376 390 400 400 400 400 $2,733

Source: GAO estimates based on state assessment plans and characteristics
and expenditure data gathered from 7 states. Note: Fiscal years 2002
through 2008 sums may not equal the total because of rounding.

Appendix VI: Fiscal Years 2002- 08 Estimated Expenditures for Each
Question Type

Appendix VII: Comments from the Department of Education

Page 39 GAO- 03- 389 Title I

Appendix VII: Comments from the Department of Education

Appendix VII: Comments from the Department of Education

Page 40 GAO- 03- 389 Title I

Appendix VII: Comments from the Department of Education

Page 41 GAO- 03- 389 Title I

Appendix VIII: GAO Contacts and Staff Acknowledgments

Page 42 GAO- 03- 389 Title I

Sherri Doughty (202) 512- 7273 Jason Palmer (202) 512- 3825

In addition to those named above, Lindsay Bach, Cindy Decker, and Patrick
DiBattista made important contributions to this report. Theresa Mechem
provided assistance with graphics. Appendix VIII: GAO Contacts and Staff

Acknowledgments GAO Contacts Staff Acknowledgments

(130126)

The General Accounting Office, the audit, evaluation and investigative arm
of Congress, exists to support Congress in meeting its constitutional
responsibilities and to help improve the performance and accountability of
the federal government for the American people. GAO examines the use of
public funds; evaluates federal programs and policies; and provides
analyses, recommendations, and other assistance to help Congress make
informed oversight, policy, and funding decisions. GAO*s commitment to
good government is reflected in its core values of accountability,
integrity, and reliability.

The fastest and easiest way to obtain copies of GAO documents at no cost
is through the Internet. GAO*s Web site (www. gao. gov) contains abstracts
and fulltext files of current reports and testimony and an expanding
archive of older products. The Web site features a search engine to help
you locate documents using key words and phrases. You can print these
documents in their entirety, including charts and other graphics.

Each day, GAO issues a list of newly released reports, testimony, and
correspondence. GAO posts this list, known as *Today*s Reports,* on its
Web site daily. The list contains links to the full- text document files.
To have GAO e- mail

this list to you every afternoon, go to www. gao. gov and select
*Subscribe to daily E- mail alert for newly released products* under the
GAO Reports heading.

The first copy of each printed report is free. Additional copies are $2
each. A check or money order should be made out to the Superintendent of
Documents. GAO also accepts VISA and Mastercard. Orders for 100 or more
copies mailed to a single address are discounted 25 percent. Orders should
be sent to: U. S. General Accounting Office 441 G Street NW, Room LM
Washington, D. C. 20548 To order by Phone: Voice: (202) 512- 6000

TDD: (202) 512- 2537 Fax: (202) 512- 6061

Contact: Web site: www. gao. gov/ fraudnet/ fraudnet. htm E- mail:
fraudnet@ gao. gov Automated answering system: (800) 424- 5454 or (202)
512- 7470 Jeff Nelligan, managing director, NelliganJ@ gao. gov (202) 512-
4800

U. S. General Accounting Office, 441 G Street NW, Room 7149 Washington, D.
C. 20548 GAO*s Mission Obtaining Copies of

GAO Reports and Testimony

Order by Mail or Phone To Report Fraud, Waste, and Abuse in Federal
Programs Public Affairs
*** End of document. ***