Federal-Aid Highways: Trends, Effect on State Spending, and	 
Options for Future Program Design (31-AUG-04, GAO-04-802).	 
                                                                 
In 2004, both houses of Congress approved separate legislation to
reauthorize the federal-aid highway program to help meet the	 
Nation's surface transportation needs, enhance mobility, and	 
promote economic growth. Both bills also recognized that the	 
Nation faces significant transportation challenges in the future,
and each established a National Commission to assess future	 
revenue sources for the Highway Trust Fund and to consider the	 
roles of the various levels of government and the private sector 
in meeting future surface transportation financing needs. This	 
report (1) updates information on trends in federal, state, and  
local capital investment in highways; (2) assesses the influence 
that federal-aid highway grants have had on state and local	 
highway spending; (3) discusses the implications of these trends 
for the federal-aid highway program; and (4) discusses options	 
for the federal-aid highway program.				 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-04-802 					        
    ACCNO:   A11979						        
  TITLE:     Federal-Aid Highways: Trends, Effect on State Spending,  
and Options for Future Program Design				 
     DATE:   08/31/2004 
  SUBJECT:   Federal aid for highways				 
	     Federal funds					 
	     Federal grants					 
	     Funds management					 
	     Grant monitoring					 
	     Highway research					 
	     Intergovernmental relations			 
	     Public roads or highways				 
	     Highway Trust Fund 				 
	     National Highway System				 

******************************************************************
** 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-04-802

United States Government Accountability Office

  GAO	Report to the Ranking Minority Member Subcommittee on Transportation and
     Infrastructure, Committee on Environment and Public Works, U.S. Senate

August 2004

FEDERAL-AID HIGHWAYS

    Trends, Effect on State Spending, and Options for Future Program Design

                                       a

GAO-04-802

Highlights of GAO-04-802, a report to the Ranking Minority Member,
Subcommittee on Transportation and Infrastructure, Committee on
Environment and Public Works, U.S. Senate

In 2004, both houses of Congress approved separate legislation to
reauthorize the federal-aid highway program to help meet the Nation's
surface transportation needs, enhance mobility, and promote economic
growth. Both bills also recognized that the Nation faces significant
transportation challenges in the future, and each established a National
Commission to assess future revenue sources for the Highway Trust Fund and
to consider the roles of the various levels of government and the private
sector in meeting future surface transportation financing needs.

This report (1) updates information on trends in federal, state, and local
capital investment in highways; (2) assesses the influence that
federal-aid highway grants have had on state and local highway spending;
(3) discusses the implications of these trends for the federal-aid highway
program; and (4) discusses options for the federal-aid highway program.

Congress may wish to consider expanding the mandate of the proposed
National Commission to consider options to redesign the federal-aid
highway program in light of these issues.

DOT officials commented on a draft of this report and said that the report
raised important issues that merit further study.

www.gao.gov/cgi-bin/getrpt?GAO-04-802.

To view the full product, including the scope and methodology, click on
the link above. For more information, contact JayEtta Hecker at (202)
512-2834 or [email protected].

August 2004

FEDERAL-AID HIGHWAYS

Trends, Effect on State Spending, and Options for Future Program Design

The Nation's investment in its highway system has doubled in the last 20
years, as state and local investment outstripped federal investment-both
in terms of the amount of and growth in spending. In 2002, states and
localities contributed 54 percent of the Nation's capital investment in
highways, while federal funds accounted for 46 percent. However, as state
and local governments faced fiscal pressures and an economic downturn,
their investment from 1998 through 2002 decreased by 4 percent in real
terms, while the federal investment increased by 40 percent in real terms.

Evidence suggests that increased federal highway grants influence states
and localities to substitute federal funds for funds they otherwise would
have spent on highways. Our model, which expanded on other recent models,
estimated that states used roughly half of the increases in federal
highway grants since 1982 to substitute for state and local highway
funding, and that the rate of substitution increased during the 1990s.
Therefore, while state and local highway spending increased over time, it
did not increase as much as it would have had states not withdrawn some of
their own highway funds. These results are consistent with our earlier
work and with other evidence. For example, the federal-aid highway program
creates the opportunity for substitution because states typically spend
substantially more than the amount required to meet federal matching
requirements-usually 20 percent. Thus, states can reduce their own highway
spending and still obtain increased federal funds.

These trends imply that substitution may be limiting the effectiveness of
strategies Congress has put into place to meet the federal-aid highway
program's goals. For example, one strategy has been to significantly
increase the federal investment and ensure that funds collected for
highways are used for that purpose. However, federal increases have not
translated into commensurate increases in the nation's overall investment
in highways, in part because while Congress can dedicate federal funds for
highways, it cannot prevent state highway funds from being used for other
purposes.

GAO identified several options for the future design and structure of the
federal-aid highway program that could be considered in light of these
issues. For example, increasing the required state match, rewarding states
that increase their spending, or requiring states to maintain levels of
investment over time could all help reduce substitution. On the other
hand, the ability of states to meet a variety of needs and fiscal
pressures might be better accomplished by providing states with funds
through a more flexible federal program-this could also reduce
administrative expenses associated with the federal-aid highway program.
While some of these options are mutually exclusive, others could be
enacted in concert with each other. The commission separately approved by
both houses of Congress in 2004 may be an appropriate vehicle to examine
these options.

Contents

  Letter

Results in Brief
Background
States and Localities Invest More in Highways Than the Federal

Government; However, Recent Federal Investment Has Outpaced State and
Local Investment

Evidence Suggests Federal Highway Grants Have Increasingly Been Used to
Substitute for Rather Than Supplement Spending from States' Own Resources

Substitution May Be Limiting the Effectiveness of Strategies to Accomplish
the Federal-Aid Highway Program's Overall Goals We Identified Several
Options for the Design and Structure of the

Federal-Aid Highway Program Conclusions Matter for Congressional
Consideration Agency Comments and Our Evaluation

1 3 7

16

21

32

40 46 47 47

Appendixes

Appendix I: Appendix II:

Appendix III:

Appendix IV: Appendix V: Objectives, Scope, and Methodology

Description of Grant Substitution Model, Statistical Methods, and Results

Summary of Previous Studies Description of GAO's Statistical Model
Statistical Results

State Characteristics Associated with States' Level of Effort to Fund
Highways from State Resources

Program Options Designed to Reduce Substitution

GAO Contacts and Staff Acknowledgments

GAO Contacts
Staff Acknowledgments

50

53 53 62 72

87

89

94 94 94

Tables     Table 1: Federal-Aid Highway Program Grant Programs and      
                                      Formulas                              9 
              Table 2: Range of Estimates of Highway Substitution Rates by 
                                                                      Time 
                   Period Based on a 95 Percent Confidence Level           24 
                      Table 3: Options to Reduce Substitution              41 
                  Table 4: Summary of Fiscal Substitution Studies          56 

                                    Contents

Table 5:	Highway Grant Substitution Rates Reported in Fiscal Substitution
Studies 60

Table 6:	Variables Considered in the Second Stage State Highway
Expenditure Equation 63

Table 7:	Variables Used to Explain the Distribution of Federal Highway
Grants 67

Table 8: Descriptive Statistics 70

Table 9:	Instrumental Variables Estimator of Federal Grants per Capita 73

Table 10: Instrumental Variables Estimates of State Highway Spending
Model, Without Correcting for Autocorrelation 75

Table 11: Instrumental Variables Estimates of State Highway Spending
Model, Correcting for Autocorrelation 77

Table 12: Summary Results of the Statistical Testing of the Variable
Coefficients 79

Table 13: State Highway Spending Model with Statistically Insignificant
Variables Removed 81

Table 14: State Highway Spending Model with Substitution Rates by Time
Period 83

Table 15: Statistical Tests for the Endogeneity of Federal Grants and
State Highway Spending 85

Table 16: Stepwise Regression Analysis of the Fixed Effects 86

Figures	Figure 1: Figure 2: Figure 3: Figure 4: Figure 5: Figure 6:

Figure 7:

Figure 8: Illustrative Effects of $1 Increase in Federal Highway Grant 15
Federal and State and Local Highway Capital Expenditures, 1982 through
2002 (2001 dollars) 17 Amount of Yearly Capital Expenditures, 1998 through
2002 (2001 dollars) 19 Federal and State and Local Highway Capital
Investment, 1991 through 2002 (2001 dollars) 20 Rates of Fiscal
Substitution into Nonhighway Uses by Time Period 23 Summary of Federal
Grant Substitution Rates Reported in Various Studies Using Data from
Various Time Periods 28 State and Local Highway Spending for Capital
Projects as a Percent of Total (Federal Plus State and Local) Capital
Spending (1997 through 2000) 31 Federal Obligations by System in Fiscal
Year 2001 37

Contents

Figure 9:	DOT Performance Measures for Condition and System Performance 39

Abbreviations

DOT Department of Transportation
BEA Bureau of Economic Analysis
FHWA Federal Highway Administration
GPRA Government Performance and Results Act
ISTEA Intermodal Surface Transportation Efficiency Act
MOE Maintenance of Effort
NHTSA National Highway Traffic Safety Administration
TEA-21 Transportation Equity Act for the 21st Century

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. However, because this
work may contain copyrighted images or other material, permission from the
copyright holder may be necessary if you wish to reproduce this material
separately.

A

United States Government Accountability Office Washington, D.C. 20548

August 31, 2004

The Honorable Harry Reid
Ranking Minority Member
Subcommittee on Transportation and Infrastructure
Committee on Environment and Public Works
United States Senate

Dear Senator Reid:

In 2004, both houses of Congress approved separate legislation to
reauthorize the federal-aid highway program to help meet the Nation's
surface transportation needs, enhance mobility, and promote economic
growth. Each bill also recognized that the Nation faces significant
transportation challenges in the future. Many transportation experts have
noted that eventually, the introduction of more fuel-efficient vehicles
and
clean fuels may undermine the sustainability of financing the Nation's
surface transportation program through motor fuel taxes. As such, both
bills established a National Commission to assess future revenue sources
for the Highway Trust Fund and to consider the roles of the various levels
of government and the private sector in meeting future surface
transportation financing needs. In the longer term, broader fiscal
challenges face the Nation, including federal and state budget deficits
and
the fiscal crisis looming as the baby boomer generation retires, causing
mandatory commitments to Social Security and Medicare to consume a
greater share of the Nation's resources, squeezing funding available for
all
domestic discretionary programs. These challenges require the Nation to
think critically about existing government programs and commitments.

In light of these issues, you asked us to provide information on past
trends
in the federal, state, and local capital investment in highways, and on
how
federal-aid highway program grants influence the level of state and local
highway spending. We responded to the first part of your request in June
2003.1 This report (1) updates information on trends in federal, state,
and
local capital investment in highways; (2) assesses the influence that
federal-aid highway grants have had on state and local highway spending;
(3) discusses the implications of these trends for the federal-aid highway
program; and (4) discusses options for the structure and design of the

1GAO, Trends in Federal and State Capital Investment in Highways,
GAO-03-744R (Washington, D.C.: June 18, 2003).

federal-aid highway program that could be considered in light of these
issues. In addition, this report identifies characteristics associated
with differences among states' levels of effort for highway investment
(see app. III).

To respond to your request, we reviewed data from the Department of
Transportation's (DOT) Federal Highway Administration (FHWA), the Bureau
of the Census, and other sources for the period from 1982 through 2002. We
determined that the data were sufficiently reliable for the purposes of
our analyses. We also reviewed and synthesized the research literature on
the influence that federal highway grants have had on the level of state
and local highway spending. Our literature review revealed a number of
studies that used statistical models developed by the studies' authors to
estimate the influence of federal funding on state spending choices. These
models examined different time periods, employed different statistical
methods, and considered different social, demographic, economic, and
political factors that may affect state highway spending decisions. None
of the models used in the studies we reviewed included the most recent
data now available on highway funding, and none examined whether the
effect of federal grants on state spending changed over the time period
covered in the study. Therefore, based on the models used in the earlier
studies, we developed a statistical model of state highway spending
outcomes to estimate the fiscal effects of federal highway funding on
state highway spending choices. This model included the most recent data
available and examined whether the effect of federal grants on state
spending changed over the time period covered by our data. The purpose of
the statistical model was to isolate the effect that federal grants have
on state highway spending by controlling for other factors that also
affect state spending decisions. The model therefore takes into account
changing state economic conditions, the size and intensity of highway
usage, and other factors that may be associated with states' willingness
to support highway spending. This statistical model was reviewed by
experts in DOT and peer reviewed by three authors of the earlier studies
on the fiscal effects of federal highway grants. These experts and authors
generally agreed with our methods, and we made revisions based on their
comments as appropriate. A more detailed description of the literature and
the statistical model is contained in appendix II. We conducted our work
from August 2003 through July 2004 in accordance with generally accepted
government auditing standards.

Results in Brief	The Nation's capital investment in its highway system has
doubled in the last 20 years, and during that time period as a whole,
state and local investment in highways outstripped federal investment in
highways-both in terms of the amount of and growth in spending. Between
1982 and 2002, state and local capital investment in highways increased
150 percent, from $14.1 billion to $35.7 billion in real terms, whereas
the federal investment increased 98 percent, from $15.5 billion to $30.7
billion in real terms.2 For every year after 1986, states and localities
invested more in the Nation's highways than did the federal government.
Most recently, in 2002, states and localities contributed 54 percent of
the Nation's capital investment in highways, spending $35.7 billion, while
the federal government contributed 46 percent, or $30.7 billion. However,
since the early 1990s, state and local investment in highways has
increased at a slower rate than federal investment in highways. From 1991,
when the Intermodal Surface Transportation Efficiency Act (ISTEA) was
enacted, through 2002, state and local investment increased 23 percent,
from $29.0 to $35.7 billion in real terms. During that same time period,
federal investment increased 47 percent, from $20.9 to $30.7 billion in
real terms. In the period following the enactment of the Transportation
Equity Act for the 21st Century (TEA-21), from 1998 through 2002, during
which state and local governments faced fiscal pressures and an economic
downturn, the trend intensified, with state and local investment
decreasing by 4 percent-from $37.0 to $35.7 billion in real terms-and
federal investment increasing by 40 percent- from $21.9 to $30.7 billion
in real terms.

The preponderance of evidence suggests that federal-aid highway grants
have influenced state and local governments to substitute federal funds
for state and local funds that otherwise would have been spent on
highways. Therefore, according to our model-which refined and expanded on
other recent models and controlled for the effects of other factors-and
according to other studies, when federal highway grants increased, total
highway spending did not increase as much as it would have had states not
withdrawn some of their own highway-related funds. Specifically, our model
examined how federal highway spending affected state spending, and it
estimated that state and local governments have used roughly half of the
increases in federal highway grants since 1982 to substitute for funding
they would otherwise have spent from their own resources. In addition, our

2Dollar amounts in this report are adjusted to 2001 dollars, matching
adjustments made in the earlier related report GAO-03-744R.

model estimated that the rate of grant substitution increased
significantly over the past two decades, rising from about 18 cents on the
dollar during the early 1980s to roughly 60 cents on the dollar during the
1990s, when ISTEA and TEA-21 were in effect.3 Three previous studies of
this issue all found that substitution occurred, although their estimates
of levels of substitution varied, probably due to differences in time
periods studied, definitions of substitution, and statistical methods
employed. There are a number of reasons why substitution may occur. Our
earlier work found that in general, the federal grant system as a whole
does not encourage states to use federal dollars as a supplement rather
than a substitute for their own spending.4 Specifically, the structure of
the federal-aid highway program creates an opportunity for substitution
because states typically spend substantially more in state and local funds
than is required to meet current federal matching requirements. As a
consequence, when federal funding increases, states are able to reduce
their own highway spending and yet obtain the increased federal funds. If
states substitute some of the increase in federal funds for their own
funds, then total highway spending may increase, but not by as much as it
would have had substitution not occurred.

These trends imply that substitution may be limiting the effectiveness of
the strategies Congress has put into place to meet the federal-aid highway
program's overall goals. Congress and DOT have at various times enumerated
goals for the federal-aid highway program, including, among other things,
enhancing safety, promoting economic growth, enhancing mobility,
supporting interstate and international commerce, and meeting national
security needs. To meet these goals, Congress has put in place strategies
that include significantly increasing the federal investment in the
highway system-particularly since 1991-and ensuring that funds collected
by the federal government for highways are used for that purpose. However,
due to probable substitution, the sizable increases in

3While these estimates represent our most likely estimates of the
substitution that occurred, they are only estimates. The uncertainty
surrounding our estimates can be expressed in terms of a level of
confidence that a given range of values encompasses the actual
substitution rate. This is discussed later in this report. For example,
the estimate of an 18 percent substitution rate for the early 1980s is not
statistically different from a finding of no substitution. Regarding our
estimate of 60 cents on the dollar during the 1990s, the actual
substitution rate, with a 95 percent level of confidence, may be as high
as 96 percent or as low as 21 percent.

4GAO, Federal Grants: Design Improvements Could Help Federal Resources Go
Further (GAO/AIMD-97-7), December 18, 1996.

dedicated federal funding that Congress has provided for highways since
1991 have not translated into commensurate increases in the Nation's
overall investment in its highway system. In part this is because, while
Congress can dedicate federal funds for highways, it cannot prevent state
highway funds from being used for other purposes. Furthermore, Congress
has sought to meet the goals of the program through a strategy of
emphasizing states' priorities and decision-making. Specifically, Congress
has incorporated return-to-origin features into the highway program and
returned to each state more of the fuel and other taxes collected in that
state, and has given states wide latitude in deciding how to use and
administer federal grants to best meet their transportation needs.
However, substitution may be limiting the effectiveness of this strategy.
Although the federal-aid highway program has a considerable regulatory
component that requires states to follow and enact certain laws as a
condition of receiving federal funds, from a funding standpoint, the
program's return-to-origin features and flexibility, combined with
substitution and the use of state and local highway funds for other
purposes, means that the federal-aid highway program is to some extent
functioning as a cash transfer, general purpose grant program. This raises
broader questions about the effectiveness of the federal investment in
highways in accomplishing the program's goals and outcomes. While under
the Government Performance and Results Accountability Act (GPRA), DOT has
established performance measures and outcomes for the federal-aid highway
program to enhance mobility and economic growth, the program's current
structure does not link funding with performance or the accomplishment of
these goals and outcomes.

We identified several options for the design and structure of the
federal-aid highway program that could be considered in light of these
issues. These options include program designs that have been used for
other federal programs and which could reduce substitution. For example,
increasing the required state match on federal highway projects, rewarding
states that increase their highway spending effort, or requiring states to
maintain levels of highway investment over time to receive federal funds
could all reduce substitution. On the other hand, the ability of states to
meet a variety of needs and fiscal pressures might be better accomplished
by providing states with funds through a more flexible federal program.
Adopting such an option could be seen as recognizing substitution as an
appropriate response on the part of states to increasing fiscal challenges
and competing demands. It could also reduce the level of administrative
involvement needed and thereby reduce administrative expenses associated
with the federal-aid highway program. Finally, policy makers may wish to
consider the design of the federal-aid highway program in the

broader context of aligning the program with program-related goals,
possibly taking into account performance measures and results. While some
of these options are mutually exclusive, others could be enacted in
concert with each other. For instance, requiring states to maintain levels
of highway investment over time or other options to limit substitution
could be combined with an effort to align funding with the accomplishment
of performance measures. Similarly, aligning funding with the
accomplishment of performance measures could also be carried out in
conjunction with creating a more flexible federal program.

The proposed National Commission to assess future revenue sources to
support the Highway Trust Fund may be an appropriate vehicle through which
to examine these options. This commission is to consider how the program
is financed and the roles of the federal and state governments and other
stakeholders in financing it; the appropriate program structure and
mechanisms for delivering that funding are important components of making
these decisions. Therefore, in light of the issues raised in this report
and the fiscal challenges the Nation faces in the 21st Century, Congress
may wish to consider expanding the mandate of the National Commission to
assess possible changes to the federal-aid highway program to maximize the
effectiveness of federal funding and promote national goals and
strategies. Consideration could be given to the program's design,
structure, and funding formulas; the roles of the various levels of
government; and the inclusion of greater performance and outcome oriented
features.

We provided a draft of this report to DOT for review and obtained comments
from departmental officials, including FHWA's Director of Legislation and
Strategic Planning. These officials said that our analysis raised
interesting and important issues regarding state funding flexibility and
the federal-aid highway program that merit further study. We agree with
DOT's characterization of the importance of the issues raised in this
report, and we continue to believe that Congress has the opportunity to
maximize the effectiveness of federal funding and promote national goals
and strategies by expanding the proposed mandate of the National
Commission. DOT also provided some technical comments, which we
incorporated where appropriate.

Background	Federal funding for highways is provided to the states mostly
through a series of formula grant programs collectively known as the
federal-aid highway program.5 Periodically Congress enacts multiyear
legislation that authorizes the Nation's surface transportation programs,
including highway, transit, highway safety, and motor carrier programs.
This legislation authorizes the federal-aid highway program and the
individual grant programs that comprise it, and it sets overall funding
for it and other surface transportation programs. In 1991, for example,
Congress enacted ISTEA, which authorized $121 billion for highways for the
6-year period from fiscal years 1992 through 1997, and in 1998 Congress
enacted TEA-21, which authorized $171 billion for the federal-aid highway
program from fiscal years 1998 through 2003. In 2004, the House and Senate
each approved separate legislation to reauthorize the federal-aid highway
program, the House authorizing $226.3 billion and the Senate authorizing
$256.4 billion for fiscal years 2004 through 2009. These authorizations
provide multiyear "contract authority" that gives the states notice
several years in advance of the size of the federal-aid program and the
approximate amount of federal funding they may expect to receive.

Funding for the federal-aid highway program is provided through the
Highway Trust Fund. Established by the Highway Revenue Act of 1956, the
Highway Trust Fund is a dedicated source of revenues generated by highway
user fees such as taxes on motor fuels, tires, and trucks. TEA-21
established two additional mechanisms to support the dedication of highway
user fees to highways. First, the act established guaranteed funding for
certain highway, transit, and highway safety programs, including the
federal-aid highway program, by protecting them with "firewalls" from
competing for funding with other domestic discretionary programs through
the congressional budget process. Second, the act provided that the
highway program funding authorizations would be adjusted to reflect
changes in estimates of Highway Trust Fund revenue, ensuring that funding
available for the federal-aid highway program reflected the revenue taken
in by the Highway Trust Fund. Both the Senate and the House have each
approved separate legislation to extend the collection of fuel taxes to
the Highway Trust Fund, the Senate through 2009 and the House through
2011. Amid concerns that the introduction of more

5The federal-aid highway program also includes discretionary grants and
research and development programs. While grants are provided to states,
localities may also sponsor federal-aid projects and can receive some
federal funds, primarily through their state.

fuel-efficient vehicles and clean fuels may undermine the sustainability
of financing the Highway Trust Fund through fuel taxes in the future, both
houses also included provisions to create a National Commission to examine
future revenue sources to support the Highway Trust Fund and to consider,
among other things, the roles of the various levels of government and the
private sector in meeting future surface transportation financing needs.

Once Congress authorizes funding, FHWA makes federal funding available to
the states annually at the start of each fiscal year through
apportionments based on formulas specified in law for each of the several
formula grant programs that make up the federal-aid highway program.
Ninety-two percent of the funds apportioned to the states in fiscal year
2003 were apportioned by formula. The remaining highway program funds were
distributed through allocations to states with qualifying projects. The
highway programs with apportionments based on formulas are shown in table
1.

        Table 1: Federal-Aid Highway Program Grant Programs and Formulas

FY 2003 funding Program Purpose (in billions)a Grant formula Minimum
apportionment

Interstate Maintenance Resurfacing, restoring,

Program	rehabilitating, and reconstructing most routes on the Interstate
Highway System.

$4.2	Interstate System lane miles (33 1/3%)

Vehicle miles traveled on the Interstate System (33 1/3%)

Annual contributions to the Highway Account of the Highway Trust Fund
attributable to commercial vehicles (33 1/3%)  1/2 percent of Interstate
Maintenance and National Highway System apportionments combined National
Highway System Improvements to rural

Program	and urban routes that are part of the National Highway System
(including the Interstate System) and designated connections to major
intermodal terminals.

$5.1	Lane miles on principal arterial routes, excluding the Interstate
System (25%)

Vehicle miles traveled on principal arterial routes, excluding the
Interstate System (35%)

Diesel fuel used on highways (30%)

Total lane miles on principal arterial highways divided by the State's
total population (10%)  1/2 percent of Interstate Maintenance and National
Highway System apportionments combined Surface Transportation Program
Projects on any federalaid highway, bridge projects on any public road,
transit capital projects, intracity and intercity bus terminals and
facilities, and other uses.

$5.9	Total lane miles of federalaid highways (25%)

Total vehicle miles traveled on federal-aid highways (40%)

Estimated tax payments attributable to highway users paid into the Highway
Account of the Highway Trust Fund (35%)

                                  1/2 percent

    Highway Bridge      Replacing or     $3.6 Relative share of   1/4 percent 
                                                    total         (10 percent 
Replacement and     rehabilitating         cost to repair or    maximum)   
                         deficient                 replace       
    Rehabilitation  highway bridges and       deficient highway  
       Program                                bridges            
                     seismic retrofits              (100%)       
                            for                                  
                     bridges on public                           
                           roads.                                

(Continued From Previous Page)

FY 2003 funding Program Purpose (in billions)a Grant formula Minimum
apportionment

Congestion          Projects which reduce $1.4 Weighted        1/2 percent 
Mitigation and                                 population in   
Air Quality         transportation             non- attainment 
Improvement         related                    and             
Program             emissions in air           maintenance     
                       quality                    areas           
                       nonattainment and          (100%)          
                       maintenance areas for                      
                       ozone, carbon                              
                       monoxide, and                              
                       particulate matter.                        

Minimum Guarantee Program

Funding to states based on equity considerations including specific shares
of overall program funds and minimum return on contributions to the
highway account of the Highway Trust Fund. A portion of the funds are
distributed among core highway programs while remaining funds are eligible
under the same rules as the Surface Transportation Program.

$6.4	90.5 percent of the percentage share of contributions to the Highway
Account of the Highway Trust Fund from motor fuel and other taxes
collected in that state based on latest available data

                                N/A Other b $0.5

Source: FHWA.

aReflects amounts apportioned by formula before the distribution of
Minimum Guarantee Program funding among the core programs.

bIncludes funds for the Appalachian Development Highway System and
Recreational Trails Programs.

As we reported in 1995, the federal funding formula derives from a
complicated set of calculations and is a complex process in which the
underlying data and factors are ultimately not meaningful because they are
overridden by other provisions that yield a predetermined outcome. 6 One
reason is the presence of "equity provisions" that ensure that states
receive set amounts based on historic funding levels and other
considerations. These equity provisions were strengthened after our 1995
report. For example, as table 1 shows, TEA-21's Minimum Guarantee Program
ensures that each state's share of apportionments from nearly all
federal-aid highway funds is not less than 90.5 percent of that state's
percentage share

6GAO, Highway Funding: Alternatives for Distributing Federal Funds
(GAO/RCED-96-6) Nov. 28, 1995.

of contributions to the Highway Account of the Highway Trust Fund.7 Funds
from this program accounted for nearly a quarter of all highway funding in
fiscal year 2003. Under separate legislation approved by both the House
and the Senate, each state's share of apportionments could rise to 95
percent by 2009.8 Furthermore, as table 1 shows, states receive minimum
apportionments regardless of the formula for several grant programs.

States have broad flexibility to transfer funds between the various grant
programs. For example, states may transfer up to 50 percent of their
Interstate Maintenance and National Highway System Program funds to other
programs, including the Surface Transportation Program, which, as table 1
shows, has broad eligibility rules. In addition, ISTEA and TEA-21 provided
the states broad authority to transfer federal-aid highway funds to
transit projects and vice versa. Between fiscal years 1992 and 2002, 47
states and the District of Columbia transferred about $8.8 billion from
federal-aid highway funds to transit programs to fund rail line
improvements, motor vehicle purchases, new or improved passenger
facilities, and other projects. During that same time, about $40 million
was transferred from FTA to FHWA for highway projects.

Once FHWA apportions funds to the states, funds are available to be
obligated by the states for construction, reconstruction, and improvement
of highways and bridges on eligible federal-aid highway routes and for
other purposes authorized in law. About 1 million of the Nation's 4
million miles of roads are eligible for federal aid; however, these roads
accounted for 85 percent of the vehicle miles traveled on the Nation's
roadways in 2001. The roads that are generally ineligible are functionally
classified as local roads or minor collectors. Around 161,000 miles of
federally eligible roadways are on the National Highway System, of which
around 47,000 belong to the Interstate Highway System. With few
exceptions, federal funds for highways must be matched by funds from other
sources-usually state and local governments. The matching requirement on
most projects is

7According to FHWA, although never legally described and named as such,
the portion of the Highway Trust Fund that is not specifically credited by
law to the Mass Transit Account of the Highway Trust Fund has come to be
called the "Highway Account" and receives all Trust Fund receipts not
specifically designated for the Mass Transit Account.

8The Senate approved the 95 percent amount while the House bill contained
a "reopener" provision that would delay fiscal year 2006 funding for most
federal-aid highway programs from October 2005 until August 2006 if
Congress has not enacted legislation by September 30, 2005, raising each
states' guaranteed rates of return to 95 percent, effective in fiscal year
2009.

80 percent federal and 20 percent state or local funding. In addition to
matching federal funds, states and localities spend funds to finance
highway capital projects and to maintain existing roadways.

The federal-aid highway program is administered by FHWA, whose
responsibilities include reviewing periodic transportation improvement
plans prepared by state and local governments, approving projects for
federal aid, apportioning grant funding to the states, providing technical
support, and overseeing federally funded projects. In fiscal year 2004,
FHWA received $334 million to provide these services, with an authorized
staff level of 2,931 positions. FHWA personnel are located in Washington,
D.C., and in 52 field offices located in each state, the District of
Columbia, and Puerto Rico, as well as a regional "resource center" with
four offices across the country that provide specialized technical
assistance to the field offices and the states.

The federal-aid highway program has a considerable regulatory component.
As a condition of receiving federal aid, states agree to apply and enforce
certain federal laws on federally aided projects, such as the
environmental assessment provisions in the National Environmental Policy
Act, the Americans With Disabilities Act, the nondiscrimination
protections found in the Civil Rights Act of 1964, and others. In
addition, states are required to establish goals and to award a set
percentage of contracts (the national goal is 10 percent) on federally
aided projects to small businesses owned and controlled by socially and
economically disadvantaged individuals, including minority and women-owned
businesses. Furthermore, in accepting federal-aid highway funds, states
must enact certain laws to improve highway safety or face penalties in the
form of either withholdings or transfers in their federal grants.9 In
addition to these penalties, states may apply for and receive highway
safety incentive grants through programs administered outside the
federal-aid highway program by the National Highway Traffic Safety
Administration (NHTSA). For example, states in which the use of seat belts
exceeds the national average or improves over time are eligible for
incentive grants based on NHTSA's

9Under TEA-21, states are subject to withholdings or transfers in their
federal grants if they fail to enact laws that (1) prohibit open alcoholic
beverage containers in the passenger area of a motor vehicle, (2)
establish minimum penalties for repeat drunk-driving offenders, and (3)
establish laws making it illegal for people to drive with the specified
level of alcohol in their blood of .08 blood alcohol concentration-the
level at which a person's blood contains 2/25th of 1 percent alcohol.

calculation of the annual savings to the federal government in medical
costs that resulted from the increased use.

In general, there are three possible ways that federal grant funding can
influence state spending for a program, as illustrated in figure 1. First,
increased federal funding may stimulate, or leverage, additional spending
from state resources. For example, a state may have to increase its own
spending in order to meet federal matching requirements and obtain federal
funds, thus increasing the overall level of spending by more than the
amount of the federal grant.10 As the federal-aid highway program in most
cases requires that states must contribute 20 percent of the total cost of
a project in order to receive federal matching funds of 80 percent of the
total cost, the suggestion is that every $1.00 increase in federal funds
would go towards a total spending increase of $1.25 ($1.00 is 80 percent
of $1.25), $0.25 of which would be funded with state and local government
funds ($0.25 is 20 percent of $1.25). The result of a stimulative effect
of federal grant funding is illustrated in the first panel of figure 1, in
which an additional $1.00 of federal aid increases spending from state
resources by 25 cents, increasing the overall level of highway spending by
$1.25. Alternatively, increased federal funding may supplement state
spending by adding to what states would otherwise have spent, increasing
the overall level of spending by the amount of the federal grant, as
illustrated in the second panel of figure 1. To the extent that states
maintain their own spending when they receive additional federal funding,
either because federal policy requires that they do so or because they do
so voluntarily, then the additional federal aid supplements state
spending. Finally, states may use increased federal funding to substitute
for, or replace, what they would otherwise have spent from state
resources, so that the overall level of spending increases by less than
the amount of the federal grant. This substitution of federal funds for
state funds is illustrated in the third panel of figure 1, in which an
additional $1.00 in federal funding results in only a 50 cent increase to
total spending because in response to the influx of

10With matching requirements, states must contribute their own funds in
order to receive federal matching funds. Economic theory suggests that
grants requiring matching, by lowering the effective price of aided
programs relative to other state spending priorities, encourage states to
spend more of their own funds. Matching grants typically contain either a
single rate (e.g., 50 percent) or a range of rates (e.g., 50 percent to 80
percent) at which the federal government will match state spending on an
aided program.

federal funds, the state withdraws 50 cents of its own spending on the
program and uses these funds for other purposes.11

11Although the fiscal effect of grants has been described in the text only
in terms of an increase in federal grant funding, stimulation and
substitution may also occur when federal funding is declining. If in
response to a decline in federal aid, for example, states increase
spending from state resources to compensate for the loss in federal
funding, this too represents grant substitution, the substitution of state
funds for federal funding.

     Figure 1: Illustrative Effects of $1 Increase in Federal Highway Grant

                                  Source: GAO.

States and Localities Invest More in Highways Than the Federal Government;
However, Recent Federal Investment Has Outpaced State and Local Investment

The Nation's capital investment in its highway system has doubled in the
last 20 years, and during that time period as a whole, state and local
investment in highways outstripped federal investment in highways-both in
terms of the amount of and growth in spending. Between 1982 and 2002,
state and local capital investment in highways increased 150 percent, from
$14.1 billion to $35.7 billion in real terms, whereas the federal
investment increased 98 percent, from $15.5 billion to $30.7 billion in
real terms.12 For every year after 1986, states and localities invested
more in the Nation's highways than did the federal government. (See fig.
2.) Most recently, in 2002, states and localities contributed 54 percent
of the Nation's capital investment in highways, spending $35.7 billion,
while the federal government contributed 46 percent or $30.7 billion in
real terms.

12To determine trends in real terms, we adjusted the data to 2001-year
dollars to coincide with the data in our related report, GAO-03-744R,
which presented data from 1982 through 2001. We converted these data using
the Bureau of Economic Analysis' (BEA) Price Indexes for Gross Government
Fixed Investment-Highways and Streets.

Figure 2: Federal and State and Local Highway Capital Expenditures, 1982
through 2002 (2001 dollars) Dollars in billions (2001)

45 40 35 30 25 20 15 10 5 0

1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996
1997 1998 1999 2000 2001 2002 Year

Federal
State and local

Source: GAO analysis of FHWA data.

In addition to the billions of dollars states and localities invest in
capital highway projects to expand highway capacity or rehabilitate
existing highways, states and localities spend additional funds
maintaining and policing their roadways. For example, in 2001, states and
localities spent about 27 percent of their total capital and maintenance
funding on maintenance activities, including fixing potholes, sealing
cracks in bridge decks, and fixing highway lighting.

Although states and localities still spend more on highway capital
investment than the federal government, recently, state and local highway
investment has increased at a slower pace than federal highway investment.
In addition, state and local investment has decreased in real terms three
times since 1996: between 1996 and 1997, between 1999 and 2000, and
between 2001 and 2002. Last year, we reported that since TEA-21 was
passed, from 1998 through 2001, federal investment increased faster

than state and local investment.13 In real terms, federal investment
increased 29 percent, while state and local investment increased 2
percent.14 This trend of federal investment increasing more quickly than
state and local investment continued in 2002. From 2001 through 2002,
federal investment increased 8.5 percent, while state and local investment
decreased 5 percent in real terms. Thus, from 1998 through 2002, federal
investment increased 40 percent, while state and local investment
decreased by 4 percent. Figure 3 shows the annual federal and state and
local capital expenditures on highways during these years.

13The percent change from 1998 through 2001 is computed by comparing the
investment in these 2 years. The calculation does not describe the
variations in the intervening years.

14As we reported, federal investment did not follow this pattern from 1997
to 1998, despite the large increase in funding authorized by TEA-21. When
comparing the change in funding from 1997 through 2001, federal investment
increased 23 percent while state and local investment increased 16
percent. This lower level of increase in federal expenditures was likely
due to the midyear passage of TEA-21 in June 1998 and the amount of time
it takes states to spend capital project funds.

Figure 3: Amount of Yearly Capital Expenditures, 1998 through 2002 (2001
dollars)

Dollars in billions (2001)

45

40 $38.2 $37.6

35

30

25

20

15

10

5

0 1998 1999 2000 2001 2002 Year

Federal State and local Source: GAO analysis of FHWA data.

The general trend of federal investment in highways increasing at a faster
pace than state and local investment in highways holds over a longer
period of time as well, including the period following the passage of
ISTEA in 1991. Although there was some variation on a year-by-year basis,
from 1991, when ISTEA was enacted, through 2002, state and local
investment increased 23 percent, from $29.0 to $35.7 billion in real
terms. During that same time period, federal investment increased 47
percent, from $20.9 to $30.7 billion in real terms, as shown in figure 4.

Figure 4: Federal and State and Local Highway Capital Investment, 1991
through 2002 (2001 dollars)

Dollars in billions (2001)

          1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

Year

State and local Federal

                       Source: GAO analysis of FHWA data.

Although the reasons for this change in spending patterns by level of
government are unclear, tough economic times, with a majority of states
needing to reduce spending to avoid budget deficits, along with large
increases in federal funds for highways may have influenced these spending
patterns. For example, a recent survey of states by the National
Conference of State Legislatures found that even after the economy began
growing after the March 2001 national recession, 36 states still have
budget shortfalls with a cumulative gap of about $25.7 billion.15

15National Conference of State Legislatures, State Budget Update: February
2003.

Evidence Suggests Federal Highway Grants Have Increasingly Been Used to
Substitute for Rather Than Supplement Spending from States' Own Resources

The preponderance of evidence suggests that increases in federal-aid
highway grants influence state and local governments to substitute federal
funds for funding they would have otherwise spent on highway projects from
their own resources.16 We built on earlier studies to develop a model that
analyzed data from 1982 through 2000 to examine whether and to what extent
states have substituted increases in federal highway funds for state
highway funds. Our preferred model analyzes data from 1983 through 2000
because of the statistical techniques we used.17 Our analysis suggests
that significant substitution has occurred and that the rate of grant
substitution increased significantly over the past two decades, rising
from 18 percent in the early 1980s to about 60 percent during the
1990s-the periods that ISTEA and TEA-21 were in effect. Three previous
studies of this issue also found that substitution existed, although their
estimates of levels of substitution varied.18 The structure of the federal
grant system as a whole may encourage substitution. Specifically, the
structure of the federal-aid highway program creates an opportunity for
substitution because states typically spend substantially more in state
and local funds than is required to meet current federal matching
requirements. As a consequence, when federal funding increases, states are
able to reduce their own highway spending and yet obtain the increased
federal funds. If states substitute some of the increase in federal funds
for their own funds, then total highway spending may increase, but not by
as much as it would have had substitution not occurred.

16Alternatively, our results suggest that during periods of declining
federal aid, states may replace some of the decline in federal funding
with additional funding from state resources.

17See appendix II for a description of the various statistical models we
considered and the rationale for our selection of a preferred model.

18Shama Gamkhar, "The Role of Federal Budget and Trust Fund Institutions
in Measuring the Effect of Federal Highway Grants on State and Local
Government Highway Expenditure," Public Budgeting and Finance, Spring
2003; Brian Knight, "Endogenous Federal Grants and Crowd-out of State
Government Spending: Theory and Evidence from the Federal Highway Aid
Program," The American Economic Review, Vol. 92 No. 1, March 2002, pp.
71-92; and Harry Meyers, "Displacement Effects of Federal Highway Grants,"
National Tax Journal, Vol. XL, No. 2, June 1987, pp. 221-235.

Our Statistical Model Suggests Federal Highway Funds Have Increasingly
Been Substituted for State Funds That Were Shifted to Nonhighway Uses

Our statistical model, which we developed from previous models, estimates
that states have used a significant portion of increases in federal
highway funding to substitute for state and local funding for highways,
and that the rate of substitution increased during the 1990s. According to
our preferred model, for the entire period from 1983 through 2000, state
governments used roughly half of the increases in federal highway grants
to substitute for funding they would have otherwise spent from their own
resources on highways.19 When our model examined four separate time
periods from 1983 through 2000 that corresponded to the four authorization
periods for the federal-aid highway program, the results suggest that the
rate of grant substitution increased in the 1990s, during the periods in
which ISTEA and TEA-21 were in effect, in comparison to the early 1980s.20
Specifically, our model suggests that states substituted approximately 18
cents (not statistically significant) of every dollar increase in federal
aid from 1983 to 1986 for funds they would have spent on highways from
their own resources. Our model suggests that the substitution rate rose to
approximately 36 cents of every dollar increase in federal aid for the
period from 1987 to 1991, and that the substitution rates then rose again
to approximately 60 cents for every dollar increase in federal aid for the
two periods examined in the 1990s: 1992 through 1997 and 1998 through
2000. (See fig. 5.)

19Our model and all the studies we examined used grant expenditures
recorded by states as the measure of federal grants. Grant expenditures
are recorded when the federal government reimburses states for eligible
project expenses. One study, described later in the report, also used an
alternative measure.

20Because grant allotments remain available for expenditures for up to 4
years, some of the grant expenditures for a given time period includes
grant obligations from prior periods.

Figure 5: Rates of Fiscal Substitution into Nonhighway Uses by Time Period

Percentage

                                     59 58

1983-1986

                         1987-1991 1992-1997 1998-2000

Time period

Source: GAO.

The rates of grant substitution for the time periods reported in figure 5
are derived from our statistical model of state spending choices and are
subject to some uncertainty. While these estimates represent our most
likely estimates of the rate at which states substituted federal funds for
state and local funds, the actual substitution may be larger or smaller
than these estimates. The uncertainty surrounding our estimates can be
expressed in terms of a level of confidence that a given range of values
encompasses the actual substitution rate. The range of values surrounding
each of our estimates is shown in table 2 at a 95 percent level of
confidence. The size of each interval provides a sense of the uncertainty
associated with our estimates. The intervals associated with the two time
periods during the 1980s contain possible values of zero, meaning that we
cannot be 95 percent confident that substitution occurred during these
periods. In contrast, the range of estimates for both time periods in the
1990s does not encompass zero; therefore, they are statistically different
from zero, which means that our results imply at least a 95 percent level
of confidence that substitution occurred. Our most likely estimates for
the two periods we looked at in the 1990s are in both cases just under 60
percent, and we can

be 95 percent confident that the actual substitution rate was between 21
percent and 97 percent.

Table 2: Range of Estimates of Highway Substitution Rates by Time Period
Based on a 95 Percent Confidence Levela

                            Point estimate   Low estimate       High estimate 
            Time period          (percent)        (percent)         (percent) 
              1983-1986                 18              -21 
              1987-1991                 36               -2 
              1992-1997                 59               22 
              1998-2000                 58               21 

Source: GAO

aPositive values represent grant substitution and negative values indicate
grant stimulation.

These results are roughly consistent with previous studies that, when
taken together, also seem to suggest increasing substitution rates over
time. We made four primary enhancements to the models used in previous
studies in developing our model. First, we used more recent data on
highway expenditures than were available for previous studies. Second, we
used a conservative definition of substitution. Our model defined
substitution as occurring only when, in response to increased federal
highway funds, state and local funds were moved out of highway-related
projects altogether. We did not consider it substitution if in response to
increased federal highway funds, state and local funds were moved from
highway projects that were eligible for federal aid to highway projects
that were not eligible for federal aid. Third, our model is structured to
examine substitution rates over time, rather than being limited to one
estimate covering all the years included in our study. Finally, compared
to previous studies, we employed a more comprehensive collection of
factors related to state spending decisions.

Combined, we believe these enhancements increase the ability of our model
to provide a conservative and more reliable estimate of the extent to
which states substitute federal highway aid for spending that would
otherwise have come from state and local resources. However, all estimates
that are based on statistical models, particularly of complex processes
such as the determination of states' budget choices, are subject to
uncertainty. This uncertainty can derive from both choices about what
factors to include in a model and the inherent impreciseness in estimating
relationships between one factor-in this case federal highway grants-

and another, state and local highway spending. While we have attempted to
take many factors affecting state spending decisions into account, there
may be other factors that are not subject to precise measurement, such as
the influence of citizen and interest groups on states' funding decisions,
that could not be included in our analysis. As a result of the uncertainty
in both the data and the statistical formulation of our model, the
precision of our estimate, or any other estimate, is limited and our
estimate should be considered one point in a range within which the actual
extent of substitution falls, and one piece of a body of evidence on the
existence of substitution. (See app. II for additional details on our
statistical model.)

In commenting on a draft of this report, DOT officials said that to the
extent substitution occurred and increased during the 1990s, it was likely
due to a number of factors, including changes in states' revenues and
priorities. While our analysis specifically took changing economic
conditions into account when assessing state spending choices, determining
specific causes is beyond the scope of our statistical model. For example,
states faced rising demands for health care and education during the 1980s
and early 1990s that they may have funded, in part, by reducing their own
levels of highway funding effort when federal highway funding increased.
Accordingly, our model establishes an association between substitution and
increases in federal highway grants; it does not identify the specific
causes responsible for these rising rates.

Earlier Studies Found That Federal Grants Reduced States' Highway
Spending, Although Substitution Estimates Varied

Three other studies, including two published in the past 3 years, have
reported that states substituted additional federal highway spending for
state spending. These studies reported a wide range of estimates for the
percentage of federal funds that has been used as a substitute for state
and local funds, from zero to nearly 100 percent. The wide range of
estimates is the result of different time periods examined, different
definitions of substitution, and differences in the statistical methods
employed.21

A study by Brian Knight, which, of the three studies, included the most
recent data, found that from 1983 through 1997, roughly 90 percent of

21Issues related to the differing statistical methods employed in previous
studies are discussed in appendix II.

increased federal aid was substituted for state highway spending.22 Knight
used a different definition of substitution than we used in our study.
Knight defined substitution as occurring when, in response to increased
federal highway funds, state funds were moved out of highway-related
projects. He did not take into account local spending on highways, which
might possibly have mitigated the reduction in state funds.

Another study, by Shama Gamkhar,23 analyzed data from 1976 through 1990
using two different measures of federal grants. Gamkhar reported an
average substitution rate of 63 percent when measuring federal grants
through grant expenditures (the same measure of federal grants used by the
other studies, including our model) and an average substitution rate of 22
percent when measuring federal grants through grant obligations.24 Gamkhar
defined substitution the same way our model did, as when, in response to
increased federal highway funds, state and local funds were moved out of
highway-related projects altogether.

22Knight, op. cit.

23Gamkhar, op. cit.

24The grant distribution process first allots federal funding to states.
States then obligate these funds for eligible highway projects, and,
finally, the federal government reimburses states at the time obligated
balances are actually spent. Thus, obligations are the second step in the
federal grant making process, and grant expenditures are the final step of
the process.

A study by Harry G. Meyers examined data from 1976 through 1982, and
modeled substitution based on two different definitions of substitution.25
Using a definition of substitution similar to the definition employed in
our model, the study found no evidence of substitution during this period.
Meyers also modeled the substitution rate based on a different definition
of substitution, defining substitution as occurring when state funds were
moved out of federal-aid highway projects, even if those funds were used
for highway projects that were ineligible for federal aid. Using this
definition of substitution, the study found a substitution rate of 63
percent. The findings of these studies and GAO's results are summarized in
figure 6. In this figure, we placed next to our finding the findings of
the three models that used the same measure of federal grants and the same
or a similar definition of substitution that we did, organizing these
chronologically.26

25Meyers, op. cit.

26Gamkhar and Meyers's findings on the second and third bars of the figure
used the same measure of federal grants and similar definitions of
substitution. Knight's study used the same measure of federal grants that
we did and a definition of substitution that was closer to our definition
than Meyers's second analysis, and so we placed his finding as the fourth
bar on the figure. Gamkhar and Meyers's alternative ways of modeling
(shown in the fifth and sixth bars of the figure) used considerably
different measures of federal grants (Gamkhar) and substitution (Meyers)
than we did in our model.

Figure 6: Summary of Federal Grant Substitution Rates Reported in Various
Studies Using Data from Various Time Periods

Percentage

                                       91

                                       a

                                   1976-1990

                                       c

                                Knight1983-1996

                                       d

                                   1976-1990

a

                                       eb

OGA

1983-2000

                                       ss

ery

Me1976-1982

ery

Me1976-1982

Source: GAO, Meyers, op. cit., Gamkhar, op. cit., and Knight, op. cit.

Alternative approaches employed to measure grant substitution:

aSubstitution defined as the reduction in state and local government
spending on all highway-related projects; federal grants measured as grant
expenditures.

bSubstitution defined as the increase in state and local government
nonhighway spending; federal grants measured as grant expenditures.

cSubstitution defined as the reduction in state (but not local) government
spending on all highway	related projects; federal grants measured as grant
expenditures.

dSubstitution defined as the reduction in state and local government
spending on all highway-related projects; federal grants measured as grant
obligations.

eSubstitution defined as the reduction in state and local government
spending on federal-aid eligible highway projects; federal grants measured
as grant expenditures.

As can be seen from this figure, generally, those studies with the same or
similar definitions of substitution as our model also suggest that
substitution rates may have increased over time. Specifically, Meyers
reported no evidence of substitution into nonhighway spending from 1976
through 1982; Gamkhar, based on data through 1990, reported higher rates
of substitution; and Knight, based on data through 1997, reported even
higher rates of substitution, although using a somewhat different
definition of substitution. Our model also found evidence of such a trend.

Structure of the Federal Grant System in General May Encourage
Substitution

In 1996, we reported that the federal grant system as a whole does not
encourage states to use federal dollars to supplement their own spending
but rather results in states using federal grants to substitute for their
own spending.27 In summarizing research over the past 30 years for a wide
variety of federal grant programs, we reported that each additional dollar
of federal grant funding substitutes for between 11 and 74 cents of
funding states otherwise would have spent. On balance, we found that for
every dollar of additional federal aid, states have withdrawn about 60
cents of their own funding.

Our 1996 study found that federal grant programs produced a variety of
fiscal effects, in part depending on the grant program's structure. For
example, grants are considered "open-ended" when there is no limit on
federal matching, and "closed-ended" when total federal matching funds are
capped. The influence of federal matching is essentially the same for both
types of grants until a state obtains the maximum federal contribution for
a closed-ended grant. After this point, closed-ended grants no longer
provide additional matching funds in response to additional state
spending. This lack of additional federal matching funds reduces the
incentive for states to increase their own spending on aided activities.
As a result, we found that open-ended grant programs, for example, Foster
Care, Adoption Assistance and Medicaid, generally stimulated additional
spending from state resources because the more states spent of their own
resources, the more federal resources they would obtain.28 In contrast,
closed-ended matching grant programs, such as the federal-aid highway
program, which place a limit on the total amount of federal funds that
states can receive through meeting matching requirements, as well as
programs that do not require states to contribute matching funds to
receive federal funds, were associated with higher rates of grant
substitution and stimulated less additional spending on the aided
activity.

27GAO/AIMD-97-7.

28The median estimate, from the studies reviewed, was that each additional
dollar of federal matching aid leverages an additional $0.38 in state
spending.

Structure of Federal-Aid Highway Program Creates an Opportunity for
Substitution

The federal-aid highway program is particularly susceptible to
substitution because in general the current matching requirement for
states is not high enough to require states to maintain or increase their
spending in order to receive increases in federal funds. In most cases,
the federal-aid highway program requires that the federal contribution be
no more than 80 percent of the total cost of the project, while the
state's matching contribution be at least 20 percent. If the federal
highway program worked to stimulate state spending, this might suggest
that every $1.00 increase in federal funds would result in a total
spending increase of $1.25 ($1.00 is 80 percent of $1.25), $0.25 of which
would be funded with state and local government funds ($0.25 is 20 percent
of $1.25). However, because in most cases state funding already exceeds
the required state matching contribution, often by large amounts, states
are not required to increase or even maintain their level of funding for
projects in order to receive increases in federal funds.

Several studies have demonstrated that state highway spending
substantially exceeds federal matching requirements. The earliest study we
reviewed found that, during the 1960s, 38 percent of aggregate state
capital spending for noninterstate federal-aid highways was in excess of
federal matching requirements.29 This study found that for the large
majority of states, state spending on federal-aid highway system projects
exceeded federal matching requirements by more than 10 percent. Another
study found that in 1982, state spending on federal-aid highway system
projects exceeded the required federal match by more than 19 percent.30
Other studies that have analyzed the fiscal effects of federal highway aid
have also reported that state spending typically exceeds federal matching
requirements.31

In general, states continue to spend more than their required match on
federal-aid highway projects. In 2000, the most recent year for which data
are available for federal-aid highways, states accounted for approximately
49 percent of all federal-aid-eligible highway capital spending, which is
over twice the required 20 percent match on most federal-aid highway

29Edward Miller, "The Economics of Matching Grants: The ABC Highway
Program," National Tax Journal, Vol. XXVII, No. 2 pp. 221-229, June 1974.

30Meyers, op. cit., considered his estimate of the over match by states
conservative due to data limitations.

31Gamkhar, op. cit., and Knight, op. cit.

projects.32 Figure 7 shows the variation among states in their highway
capital spending as a percent of total (federal plus state and local)
highway capital spending during the period from 1997 through 2000.
Although these data include spending on nonfederal-aid-eligible highways
and therefore can not be used to determine precisely to what extent states
are exceeding federal matching requirements, they show that in the
majority of states, state and local spending counts for over half of total
capital highway spending.

Figure 7: State and Local Highway Spending for Capital Projects as a
Percent of Total (Federal Plus State and Local) Capital Spending (1997
through 2000)

Number of states

                                       19

20% orless

                          21-35% 36-50% 51-65% 66-80%

ve

abo81% andPercentage of state and local capital spending

Source: GAO analysis of FHWA data.

32States and localities invest in capital projects on both their
federal-aid-eligible highways and roads where federal aid is not eligible
to be used, such as roads functionally classified as local. The amount of
funding spent on only federal-aid-eligible roads is periodically estimated
by FHWA. This estimate is used for the national number. However, this
information is not available for state-by-state analysis. Thus, figure 7
includes state and local spending on roads that are not eligible for
federal aid, overstating the amount of state "match" on federal-aid
eligible roads.

Substitution May Be Limiting the Effectiveness of Strategies to Accomplish
the Federal-Aid Highway Program's Overall Goals

The trends in funding and probable substitution described in this report
imply that substitution may be limiting the effectiveness of strategies
Congress has put into place to help the federal-aid highway program
accomplish its overall goals. Congress and DOT have at various times
enumerated goals for the federal-aid highway program, and, to meet these
goals, Congress has put in place a number of strategies, including
increasing its investment in highways and giving states wide latitude in
deciding how to use and administer federal grants to best meet their
transportation needs. However, because of substitution, the sizable
increases Congress provided in federal funding for highways have not
translated into commensurate increases in the Nation's overall spending in
its highway system. In part, this is because, while Congress can dedicate
federal funds to highways, it cannot prevent state highway funds from
being used for other purposes. Congress has also sought to meet the goals
of the program through a strategy of emphasizing states' priorities and
decision-making. However, substitution may be limiting the effectiveness
of this strategy. Although the federal-aid highway program has a
considerable regulatory component, from a funding standpoint, the program
is to some extent functioning as a cash transfer, general purpose grant
program. This raises broader questions about the effectiveness of the
federal investment in highways in accomplishing the program's goals and
outcomes, for although DOT has created performance measures and outcomes
under GPRA, currently there is no link between the achievement of these
measures and outcomes and federal funding provided to the states.

Congress and DOT Have Set Out Goals for the Federal-Aid Highway Program

Congress and DOT have at various times enumerated goals for the federalaid
highway program to, among other things, enhance safe and reliable travel,
promote economic growth, enhance mobility, support interstate and
international commerce, and meet national security needs. According to
DOT's 2003-08 Strategic Plan, the department's mission is enumerated in 49
U.S.C. 101, which states that "the national objectives of general welfare,
economic growth and stability, and the security of the United States
require the development of transportation policies and programs that
contribute to providing fast, safe, efficient, and convenient
transportation...". In establishing the Interstate Highway System,
Congress, in the Federal-Aid Highway Act of 1956, stated that the
Interstate system was to serve principal metropolitan areas and industrial
centers, support the national defense, and connect with routes of
continental importance in Canada and Mexico. Current law defines the
primary focus of the federal-aid highway program as completion and
expansion of the National Highway System, of

which the Interstate is a part, to provide interconnected routes that
serve, among other things, major population centers, international border
crossings, commercial ports, airports, and major travel destinations.

Congress continued to set out these goals in reauthorization legislation
that the Senate and House each passed in 2004. For example, the
legislation approved by the Senate states that:

"...among the foremost needs that the surface transportation system must
meet to provide for a strong and vigorous national economy are safe,
efficient, and reliable (i) national and interregional personal mobility
(including personal mobility in rural and urban areas) and reduced
congestion; (ii) flow of interstate and international commerce and freight
transportation; and (iii) travel movements essential for national
security."

To meet the program's goals, Congress has set out a number of strategies,
including increasing investment in highways and providing states
flexibility to best meet their transportation needs. Furthermore, under
Congress' direction, DOT has established strategic goals and performance
measures and outcomes for the federal-aid highway program to enhance
mobility and economic growth. Among these goals are to reduce the growth
of congestion on the Nation's highways and improve the condition of the
National Highway System.

One Strategy to Meet Goals Has Been to Increase Investment and Ensure
Federal Highway Funds Go to Highway Program

Since the Federal-Aid Highway Act was enacted in 1956, every time Congress
has reauthorized the highway program it has expanded either the size or
scope, or both, of the federal-aid highway program.33 Since 1991, Congress
has provided significant increases in federal spending on highways.
ISTEA's authorization of $121 billion for highways for the 6-year period
from fiscal years 1992 through 1997 was a 73 percent increase over the $70
billion authorized in the prior 6-year bill, and TEA-21's authorization of
$171 billion for the federal-aid highway program from fiscal years 1998
through 2003 represented an increase of 41 percent over ISTEA's
authorization level. In 2004, the House and Senate each approved separate
legislation to reauthorize the federal-aid highway program, increases of
32 percent and 50 percent over TEA-21, respectively.34 Despite these
increases, numerous congressional transportation leaders stated that

33See CRS report 98-221: ISTEA Reauthorization: Highway and Transit
Legislative Proposals in the 105th Congress, 2nd Session.

34Increases are shown in nominal dollars.

these increases were not enough, and that further spending was required to
meet the country's needs.

Congress has also included features in the design of the federal-aid
highway program to attempt to ensure that funds collected by the federal
government for highways are used for that purpose. Prior to 1956, federal
fuel and motor vehicle taxes were directed to the General Fund of the U.S.
Treasury, and there was no relationship between the receipts from these
taxes and federal funding for highways. Amid concerns that federal taxes
on motor fuel were being used for nontransportation purposes, Congress
established the Highway Trust Fund in 1956 and specifically provided that
revenues from most highway user taxes would be used to finance the greatly
expanded highway program enacted by the Federal-Aid Highway Act of 1956.
Despite having a dedicated source of funding, highways competed for
federal funding with other forms of domestic discretionary spending
through the appropriations process over the years. As a result, Congress
often appropriated less money than was authorized, even though sufficient
funds were being collected in the Highway Trust Fund to support the
authorized levels. So Congress took further action in TEA-21, establishing
guaranteed spending levels for highway programs that protected highway
programs from having to compete for funding through the congressional
budget and appropriations process. It also established "Revenue Aligned
Budget Authority," directly linking highway revenues collected into the
Highway Trust Fund with the apportionments provided annually to the states
for their highway programs.

Despite congressional efforts to increase the federal investment in the
highway system and to ensure that funds collected by the federal
government for highways are used for that purpose, due to probable
substitution, the sizable increases in dedicated federal funding that
Congress has provided for highways have not translated into commensurate
increases in the Nation's overall investment in its highway system.
Moreover, the effectiveness of Congress' strategy to dedicate federal
funds to highways is limited because Congress has no similar ability to
prevent state and local highway funds, where most of the investment
occurs, from being used for other purposes. Therefore, while Congress can
ensure that certain federal moneys are dedicated to highways and given to
the states for that purpose, it cannot ensure that state and local highway
funds are not used for other purposes. When substitution occurs, some
dedicated federal highway funds replace state highway funds, and those
state highway funds are then used for other purposes.

Another Strategy Has Been to Emphasize Importance of States' Priorities
and Decision-making

Congress has also sought to meet the goals of the program by emphasizing
the importance of states' priorities and decision-making regarding how to
meet their most pressing transportation needs. One way it has done so is
by incorporating return-to-origin features into the program--returning to
the states more of the money collected in fuel taxes. TEA-21's Minimum
Guarantee provisions ensure that each state receives back from most
highway programs 90.5 percent of the total estimated percentage share of
contributions to the Highway Account of the Highway Trust Fund from motor
fuel and other taxes collected in that state. Under separate legislation
passed by both the House and Senate in 2004, this amount could rise to 95
percent by 2009.35

In addition, Congress has given the states broad flexibility in the use of
its federal aid grant funds by providing states significant discretion to
use these funds flexibly across highway, bridge, transit, and other
transportation projects. States have, if they choose, broad flexibility in
the use of slightly more than half of their federal-aid highway funds. For
example, the Surface Transportation grant program has broad eligibility
rules, and states can use those funds for highways, bridges, transit
capital projects, bus terminals, and many other uses. States may use some
of their Minimum Guarantee Program grant funds under the same rules;36 in
fiscal year 2003, the funds apportioned under these two programs accounted
for one third of all federal aid highway funds apportioned nationwide. For
eight states that receive higher levels of Minimum Guarantee grant funds,
these two programs account for more than 40 percent of their funding, and
in one of these eight states, for just over 50 percent. While other
federal-aid highway grant funds have more limited uses, states have the
authority to transfer funds from these limited programs to more flexible
programs and uses. For example, states may transfer up to 50 percent of
their National Highway System and Interstate Maintenance program funds to
the Surface Transportation Program or certain other grant programs, and,
in the case of

35Specifically, the Senate bill provides that each state would achieve a
95 percent return on payments to the highway account of the Highway Trust
Fund by 2009. While the House bill does not contain this provision, it
would delay fiscal year 2006 funding for most federal-aid highway programs
from October 2005 until August 2006, if Congress has not enacted
legislation by September 30, 2005, raising each state's guaranteed rate of
return to 95 percent, effective in fiscal year 2009.

36Specifically, those funds that are not distributed to the core highway
programs. See table 1.

the National Highway System program, 100 percent under certain conditions.

Furthermore, states have broad flexibility in deciding which projects to
pick and how to implement them. The projects for which states use federal
funding must be for construction, reconstruction, and improvement on
eligible federal-aid highway routes. Nevertheless, federal law (23 U.S.C.
S:145) provides that the authorization or appropriation of federal funds
"shall in no way infringe on the sovereign rights of the States to
determine which projects shall be federally financed." Moreover, FHWA's
role in overseeing the design and construction of most projects is
limited. Specifically, only high cost construction or reconstruction
projects on the Interstate Highway System are always subject to "full"
oversight in which FHWA prescribes design and construction standards,
approves design plans and estimates, approves contract awards, inspects
construction progress, and renders final acceptance when projects are
completed. For projects that are not located on the National Highway
System, states are required to assume oversight responsibility for the
design and construction of projects unless a state determines that it is
not appropriate for it to do so.37 As figure 8 shows, in 2002, about $1
out of every $5 obligated for federal-aid projects occurred on the
Interstate system, while projects off the National Highway System
accounted for about 57 percent, nearly 3 times as much.

37FHWA approves state transportation plans, environmental assessments, and
property acquisition for all federally financed highway projects. On
projects that are not located on the Interstate system but are part of the
National Highway System, states may assume responsibility for overseeing
the design and construction of projects unless either the state or FHWA
determines that this responsibility is not appropriate. While FHWA and
each state enter into an agreement documenting the types of projects for
which the state will assume these oversight responsibilities, FHWA does
not maintain information centrally on how many states have opted for
federal versus state oversight in cases where discretion is permitted.

          Figure 8: Federal Obligations by System in Fiscal Year 2001

                              Interstate projects

Projects on the National Highway System (noninterstate)

Projects off the National Highway System

Source: GAO analysis of FHWA data.

Substitution may be limiting the effectiveness of Congress' strategy of
emphasizing the role of states' priorities and decision-making regarding
how to meet their most pressing transportation needs. The program does
have a substantial regulatory component that requires states to enact and
follow certain laws as a condition of receiving federal funds; for
example, states are required to enact drunk-driving laws, such as .08
blood alcohol laws, and to contract with disadvantaged business
enterprises. However, from a funding standpoint, the federal-aid highway
program's return-toorigin features and flexibility, combined with
substitution and the use of state and local highway funds for other
purposes, means that the program is, to some extent, functioning as a cash
transfer, general purpose grant program. This raises broader questions
about the effectiveness of the federal investment in highways in
accomplishing the program's goals and outcomes.

Broader Questions Exist Our findings on substitution lead to broader
questions about whether the about the Program's Goals federal-aid highway
program is effective in meeting its goals. As required and Outcomes by the
Government Performance and Results Act (GPRA), DOT has

articulated goals for the department's programs, including the federal-aid
highway program, to achieve by establishing measurable performance goals,
measures, and outcomes. One of the purposes of GPRA is to provide

decisionmakers a means of allocating resources to achieve desired results.
Linking resources and results will become even more important than it is
today in the years ahead, as the Nation faces a fiscal crisis in which
mandatory commitments to Social Security and Medicare will consume a
greater share of the Nation's resources, squeezing the funding available
for discretionary programs, potentially including highways. These
challenges require the Nation to think critically about all existing
government programs and commitments.

Among its performance goals, DOT has articulated goals for mobility and
economic growth, including to improve the condition of the transportation
system, reduce travel times, and increase access to and reliability of the
transportation system. Two major performance measures related to the
federal-aid highway program are to (1) improve the percentage of travel on
the National Highway System meeting pavement performance standards for
acceptable ride and (2) slow the growth of congestion--in particular, to
limit the annual growth of urban area travel time under congested
conditions to one-fifth of 1 percent below the growth that has been
projected. These goals are shown in figure 9.

Figure 9: DOT Performance Measures for Condition and System Performance

NHS Pavement Condition Travel in Urban Areas in Congested Conditions

Percentage miles with acceptable ride quality

                              Percentage of travel

100

                                      31.0

95

90

                                      28.5

85

80 26.0

     0                            0                                      
1996   1998   2000  2002 2004 1996 1997 1998 1999 2000 2001 2002 2003 2004 
          Trend                         Trend                            
         Target                        Target                            

Unconstrainted growth

Source: DOT's Fiscal Year 2004 Performance Plan.

Although DOT has articulated performance measures, the federal-aid highway
program does not have the mechanisms to link funding levels with the
accomplishment of specific performance-related goals and outcomes. In
contrast, NHTSA has some incentive grant programs that link funding to
particular outcomes, such as increasing the use of seat belts within
states. As we have reported, although a variety of tools are available to
measure the costs and benefits of transportation projects, they often do
not drive investment decisions, and many political and other factors
influence project selections.38 For example, the law in one state requires
that most highway funds, including federal funds, be distributed equally
across all the state's congressional districts. Consequently, there is
currently no way to measure how funding provided to the states is being
used to accomplish particular performance-related results such as reducing
congestion or improving conditions.

38GAO, Surface Transportation: Many Factors Affect Investment Decisions,
GAO-04-744 (Washington, D.C.: June 30, 2004).

We Identified Several Options for the Design and Structure of the
Federal-Aid Highway Program

We identified several options for the design and structure of the
federal-aid highway program that could be considered in light of the
issues raised by our findings. On the one hand, there are options that
have been used in other federal programs that could limit substitution.
Another option to consider may be to simplify the program towards a more
flexible approach. Another option would be to consider whether a different
program structure and different financing mechanisms could be used to
target funding and more closely align resources with desired results.

Reduce Substitution	To increase the extent to which federal-aid highway
program funds are used to supplement state highway funds rather than
substitute for them, several options exist to re-design the program to
limit substitution. These include:

o 	Revising federal matching requirements to increase the percentage of
projects' costs that must be paid for with state and local funds.

o 	Instituting the use of funding formulas that reward states that
increase state and local highway funding by increasing their federal
funding, while reducing the federal funding of those states that do not.

o 	Adding a requirement that states maintain their own level of highway
spending effort over time in order to receive additional federal funds.

All three options are designed to reduce or eliminate substitution. The
first two options are designed to stimulate additional state spending on
highways, while the third option is designed so that increased federal
funding will supplement state spending rather than replace it. These
objectives may not be perfectly achieved because models of substitution,
like any models, produce estimates that are subject to uncertainty. As
such, there is no way to objectively determine with certainty what states
would have spent in the absence of increased federal funding. Table 3
summarizes the options, along with possible approaches that could be taken
in implementing them.

Table 3: Options to Reduce Substitution

Option Approaches

Revise matching Revise state match to a higher percentage than current 20
requirements percent of total funding for project

Keep state match at 20 percent but count only state spending in excess of
base time period for match

Link federal funding to Provide federal funds to states proportionally,
based on their
states' highway funding effort compared to average effort of all states
effort

Provide federal funds to states proportionally, based on each state's own
effort relative to an initial base time period

Institute a maintenance Require states to maintain existing levels of
state spending in of effort provision order to receive federal funds

Source: GAO.

Each of these options and approaches would be likely to have somewhat
different effects and would require careful consideration of various
factors. Some possible effects are summarized below; see appendix IV for
additional discussion of these options.

o 	The likely effect of revising the matching requirement would depend on
the magnitude of the change. For example, if the requirement was changed
so that states generally had to provide 60 percent of the total funding
for eligible projects, states currently spending less than 60 percent of
total highway funds for eligible projects would have an incentive to
increase their spending in order to obtain the maximum federal match,
while those spending more than 60 percent would not have an incentive to
increase their spending. A few states with a low state/federal spending
ratio might have to more than double their current spending in order to
receive additional federal funds. Setting the required match at 40 percent
would give fewer states an incentive to increase their spending and would
generally require less of an increase in spending from those states with
low state/federal spending ratios. An advantage of continuing to set the
state match at 20 percent but counting only state spending in excess of
what each state spent during a base time period towards the match is that
it would stimulate state spending in all states to a similar degree.

o 	Using funding formulas that link federal funds to states' highway
funding effort could also be achieved through various approaches.39 For
example, providing federal funds to states proportionally based on their
effort in comparison to the average effort of all states would put states
in competition with each other, rewarding states whose funding effort is
already high and penalizing states whose funding effort is currently low.
On the other hand, providing federal funds to states proportionally based
on each state's own effort during an initial base time period would put
each state in competition with the funding effort it made in the base
period, rewarding states whose spending grew more quickly in comparison to
their spending during the base period and penalizing states whose spending
stayed the same or dropped when compared to their spending during the base
period.40 Such provisions could be designed so they could be suspended in
a recession or severe economic downturn in order to prevent states from
having to make disproportionate reductions in other state services to
maintain highway funding.

o 	Instituting a maintenance of effort provision would require each state
to continue to spend what it spent in a defined base period, plus
inflation, in order to obtain increased federal funds. Therefore, it would
not stimulate state spending, but it would attempt to ensure that states
used federal funds to supplement rather than replace state and local
funds. In previous work, we concluded that, to be effective, maintenance
of effort provisions need to define a minimum level of state spending
effort that can be objectively quantified and updated to keep pace with
inflation in program costs so that the maintenance of effort provision
ensures a continued level of activity when measured in inflation adjusted
dollars.41 This could be achieved by defining a state's base spending
level as the

39Defining a state's highway spending effort would need to avoid unfairly
penalizing lowincome states, which may not have the resources to compete
with the highway spending of wealthier states. Therefore, each state's
highway funding effort could be defined as the state's highway spending
compared to some measure of the state's taxing capacity. The most
comprehensive measure of states' taxing capacity that is available
annually is Total Taxable Resources (TTR), which is produced annually by
the Department of the Treasury.

40Defining a state's spending during a base time period should, to the
extent possible, be established by measuring spending levels that are
typical rather than unusually high or low.

41GAO, Proposed Changes in Federal Matching and Maintenance of Effort
Requirements for State and Local Governments, GAO/GGD-81-7 (Washington,
D.C.: Dec. 23, 1980); and Block Grants: Issues in Designing Accountability
Provisions, GAO/AIMD-95-226 (Washington, D.C.: Sept. 1, 1995).

amount spent per year during a recent historical period and then adjusting
that base spending level for inflation.42

Increase Flexibility in States' Use of Funds and Reduce Administrative
Expenses

Another potential option would be to build on trends giving states greater
flexibilities and discretion with their federal-aid highway program funds.
In contrast to changes in program designs that would limit substitution,
adopting such an option could be seen as recognizing substitution as an
appropriate response on the part of states to increasing fiscal challenges
and competing demands. Adopting such an option could also be seen as
recognizing that the ability of states to meet a variety of needs and
fiscal pressures might be better accomplished by providing states with
federal funding for highways through a more flexible federal program.

Such an option would also recognize the changing nature of FHWA's role and
the federal-aid highway program. Currently, FHWA reviews and approves
transportation plans and environmental reviews, and-on some
projects-designs, plans, specifications, estimates, and contract awards.
FHWA also has duties related to the program's considerable regulatory
component. To carry out these responsibilities, FHWA has among the largest
field office structure in DOT, and a larger field structure than many
other federal agencies. FHWA has personnel in over 50 field offices,
including one office in each state, and has had a field office in each
state since 1944. However, the federal-aid highway program has changed
considerably in 60 years. In 2004, the program's return-to-origin features
and flexibility, combined with substitution and the use of state and local
highway funds for other purposes, means that from a funding standpoint,
the federal-aid highway program is, to some extent, functioning as a cash
transfer, general purpose grant program. Devolving funding
responsibilities to the states in a manner consistent with that function
would build on the flexibilities already present and obviate much of the
need for FHWA's extensive field organization, allowing it to be greatly
reduced in scope. This could produce budgetary savings of some portion of
FHWA's $334 million annual budget.

42One drawback of a maintenance of effort provision is that basing it on a
historical spending period could result in a base spending period that
represents an unusually high spending level for some states, effectively
locking them into continued high spending in future years. This could be
ameliorated however by establishing waivers for states that are able to
demonstrate that spending in the base period chosen is unusually high, to
allow a more "typical" spending level for purposes of the maintenance of
effort provision.

Adopting such an option would involve weighing numerous factors, including
FHWA's role and value. But devolving funding responsibilities to the
states would not require abandoning the program's regulatory component.
Some federal laws and requirements in place originated outside the
transportation program and would doubtless remain in force, such as civil
rights compliance. Others that are currently part of the transportation
program could also remain in effect. Depending on priorities, these could
continue to be overseen by FHWA directly or a process could be established
through which states certify their compliance with the requirements, as is
done in other programs. In this manner, it would be possible to enforce
these laws and requirements without an extensive field structure, as other
federal agencies and programs do.

Devolving authority to the states could also take the form of devolving
not only the federal programs, but the revenue sources that support it.
Considerable federal effort goes into collecting and accounting for motor
fuel taxes and other highway user fees. One argument for maintaining a
federal fuel tax is that this tax may be a useful public policy to prevent
tax competition between states to avoid the disinvestment in the highway
system that could potentially result. Such a "turnback" provision was
considered in the form of an amendment to TEA-21 in the House of
Representatives in 1998, but it did not pass.

Devolving federal responsibilities to the states is not dissimilar to the
Surface Transportation System Performance Pilot Program that was proposed
in the administration's reauthorization proposal, but which was not
included in either the House or Senate version of the bill. Up to five
states could have participated in the program, which would have allowed a
state to assume some or all of FHWA's authorities and responsibilities
under most federal law or regulations.43 Once approved to participate, a
state would have had to identify annually what goals it wanted to achieve
with its federal funds and what performance measures it would use to gauge
success. A state would also have had to agree to a maintenance of effort
requirement that it maintain its total combined state and federal

43Under the proposed pilot program, the federal government would not have
devolved its responsibilities (1) to review states and local governments'
transportation plans, (2) to oversee "major" projects costing over $1
billion, (3) under Title VI of the Civil Rights Act of 1964, or (4) under
any laws relating to federally recognized tribes. In addition, the
proposal specified that nothing in it would be interpreted to relieve any
project from the requirements of the National Environmental Policy Act,
nor would it preclude DOT from issuing rulemaking actions as needed.

highway program expenditures at the level of at least the average level of
the three previous years. A state's participation in the pilot program
would have been terminated if that state did not achieve the agreed
performance for two consecutive years.

Link Federal-Aid Highway Funding with Program Goals

Another option could be to consider whether a different program structure
and different financing mechanisms could be used to target funding and
more closely align resources with desired results. Restructuring the
program in this way could take several forms. For example, the program
could be reoriented to function more like a competitive discretionary
grant program, in which program sponsors justify projects seeking federal
aid based on an assessment of their potential benefits. This is not
dissimilar to the program used by DOT to fund large transit capital
projects.44 The program could also be revised to include the use of
incentive grant programs similar to those that NHTSA has to link funding
to particular outcomes, such as increasing the use of seat belts within
states.

Adopting such an option would require asking the following questions:

o 	What policy goals have been established by Congress for the performance
of the federal-aid highway program, what outcomes and results have been
articulated in DOT's strategic plans to fulfill those goals, and are they
the right goals and outcomes?

o 	What is the appropriate role of each level of government? Would the
roles need to be redefined in order to align federal spending more closely
with a greater performance and outcome orientation? In particular, what
refocusing of federal involvement (e.g., interstate commerce, homeland
security, national defense) would need to occur?

o 	How could the design of the federal-aid highway program's grants and
funding mechanisms best support accomplishment of agreed-upon

44Under the Federal Transit Administration's New Starts Program, local
transit agencies compete for project funds based on specific financial and
project justification criteria. FTA assesses the technical merits of a
major transit project proposal and its finance plan and then notifies
Congress that it intends to commit, subject to appropriations, New Starts
funding to certain projects through full funding grant agreements. The
agreement establishes the terms and conditions for federal participation
in the project, including the maximum amount of federal funds-which by law
must be no more than 80 percent of the estimated net cost of the project,
but in practice is often less than that percentage.

performance goals and outcomes? What funding incentives are needed to
introduce a greater performance and outcome orientation?

o 	What type of departmental administrative structure for the federal-aid
highway program would best ensure that the performance goals established
by Congress and articulated in DOT's strategic plans and outcomes are
measured and accomplished?

o 	Can a greater performance and outcome orientation to the federal-aid
highway program be reconciled with congressional and state legislative
policies and preferences toward providing at least some transportation
funding in the form of specific project earmarks?

Conclusions	Addressing the issues raised in this report would require
weighing competing and sometimes conflicting options and strategies. If,
for example, reducing the level of grant substitution is an important
concern, then design changes in the current program, including adopting
features that have been used in other federal programs, may be warranted.
If, on the other hand, preserving states' flexibility, including their
ability to meet a variety of needs and fiscal pressures is a higher
priority, then design changes in the direction of a different, more
flexible program may be warranted. While some options are mutually
exclusive, others could be enacted in concert. For instance, an option to
limit substitution could be combined with efforts to align resources with
desired results, and returning program authorities and resources to the
states could be accompanied by adding performance measures.

Beyond these options, our work raises broader and more fundamental issues
given the challenges the Nation faces in the 21st Century. The fact that
both the federal and state governments face budget deficits totaling
hundreds of billions of dollars and a growing fiscal crisis requires
policymakers to think critically about existing government programs and
commitments and make tough choices in setting priorities and linking
resources to results to ensure that every federal dollar is wisely and
effectively spent. The opportunity to better align the federal-aid highway
program with performance goals and outcomes comes at a time when both
houses of Congress have already approved separate legislation to create a
National Commission to examine future revenue sources to support the
Highway Trust Fund and to consider the roles of the various levels of
government and the private sector in meeting future surface transportation
financing needs. The proposed commission is to consider how the program

is financed and the roles of the federal and state governments and other
stakeholders in financing it; the appropriate program structure and
mechanisms for delivering that funding are important components of making
these decisions. Thus, this commission may be an appropriate vehicle
through which to examine these options for the future structure and design
of the federal-aid highway program.

Matter for In light of the issues raised in this report and the fiscal
challenges the

Nation faces in the 21st Century, Congress may wish to consider
expandingCongressional the proposed mandate of the National Commission to
assess possible Consideration changes to the federal-aid highway program
to maximize the effectiveness

of federal funding and promote national goals and strategies.
Consideration could be given to the program's design, structure, and
funding formulas; the roles of the various levels of government; and the
inclusion of greater performance and outcome-oriented features.

Agency Comments and Our Evaluation

We provided DOT a draft of this report for review and obtained comments
from departmental officials, including FHWA's Director of Legislation and
Strategic Planning. These officials said that our analysis raised
interesting and important issues regarding state funding flexibility and
the federal-aid highway program that merit further study. DOT officials
also stated that while they recognize that federal-aid highway grants can
influence state and local governments to substitute federal funds for
state and local funds that otherwise might have been spent on highways,
they believe that this substitution is likely due to numerous factors.
Specifically, the officials said that to the extent substitution occurred
and increased during the 1990s, it was also likely due to changes in
states' revenues and priorities. DOT officials also emphasized that
regardless of changes in the availability of state funds for highway
programs, the overall federal share of capital spending on highways
declined during the period we studied, from over 55 percent in the early
1980s to around 45 percent today. DOT officials also emphasized that there
is no evidence that the substitution discussed in our report resulted in
the diversion of federal-aid highway funds apportioned to the states. They
further stated that substitution may reflect appropriate resource
allocations by states and that preserving states' flexibility has been a
priority of the federal-aid highway program and is a goal of DOT's
reauthorization proposal. Finally, regarding options for changes in the
design of the federal-aid highway program, officials emphasized that FHWA
adds considerable value to the federal-aid highway program by providing

program oversight and sharing its expertise with states to ensure states
uniformly address key areas of national concern including safety and
environmental protection.

We agree with DOT's characterization of the importance of the issues
raised in this report, including the effect that federal-aid highway
grants have on state spending decisions and states' funding flexibility.
We also agree with DOT officials that many factors influence state
budgetary decisions, including changing state budget priorities and the
availability of state revenues. It was for this reason that we used a
statistical model that specifically took changing economic conditions and
revenues into account in order to better isolate the effect of federal
grants on state spending choices. We believe that our model has reasonably
distinguished between the effects of changing economic conditions and
revenues and the effect of federal grants, and, consistent with earlier
models and studies, we found the relationship between federal grants and
state spending, indicating substitution, to be statistically significant,
particularly during the 1990s. However, determining specific causes of
substitution is beyond the scope of our statistical model. For example,
while states faced rising demands for health care and education during the
1980s and 1990s that could have resulted in states reducing their highway
spending when federal highway funding increased, our model does not
identify the specific causes responsible for rising substitution rates.
Although DOT officials said that the overall federal share of capital
spending on highways declined during the period we studied, these relative
shares do not affect our findings on substitution since substitution can
occur when the federal share of funding is either rising or falling; if
substitution occurs when state funding is rising it simply means that
state spending increased less than the increase that might have occurred
had there been no substitution. While DOT officials stated that there is
no evidence that substitution resulted in the diversion of federal-aid
highway funds, there are important differences between diversion and
substitution. In the context in which DOT officials raised it, diversion
is the transfer of federal funds for purposes other than those authorized
by law, while substitution, as we have reported it, is the transfer of
state funds that would have otherwise been spent on highways. States can
both use federal funds for the purposes authorized by law and at the same
time substitute federal funds for state funds. Thus, while we agree that
there is no evidence that substitution resulted in the diversion of
federal-aid highway funds, we do not believe our report suggests the
existence of such evidence.

Finally, we agree with DOT officials that states' flexibility and FHWA's
role are important factors in the federal-aid highway program; however, we
believe that options for changing the design, structure, and funding
mechanisms of the federal-aid highway program should be considered in
light of substitution and the issues raised in this report, and that a
variety of factors, including but not limited to these two, should be
weighed when considering such changes. While the department took no
position on the matter for congressional consideration to expand the
mandate of the proposed National Commission, officials did state that they
believe these issues merit further study. We continue to believe that
Congress has the opportunity to maximize the effectiveness of federal
funding and promote national goals and strategies by expanding the
proposed mandate of the National Commission to consider these issues.

We are sending copies of this report to the Honorable Norman Mineta,
Secretary of Transportation. We will also make copies available to others
upon request. In addition, the report will be available at no charge on
the GAO Web site at http://www.gao.gov.

If you have any questions about this report, please contact me at
[email protected], or (202) 512-2834, or contact Jerry Fastrup at
[email protected] or (202) 512-7211, or Steve Cohen at [email protected] or
(202) 512-4864. GAO contacts and acknowledgments are listed in appendix V.

Sincerely yours,

JayEtta Z. Hecker Director, Physical Infrastructure

Appendix I

                       Objectives, Scope, and Methodology

In light of the increasing federal-aid highway program funding and
concerns over future federal revenues for highways, you asked us to
provide information on past trends in the federal, state, and local
capital investment in highways, and how federal-aid highway program grants
influence the level of state and local highway spending. We responded to
the first part of your request in June 2003. This report (1) updates
information on trends in federal, state, and local capital investment in
highways; (2) assesses the influence that federal-aid highway grants have
had on state and local highway spending; (3) discusses the implications of
these issues on the federal-aid highway program; and (4) discusses options
for the federal-aid highway program that could be considered in light of
these issues. In addition, this report identifies characteristics
associated with differences among states' levels of effort for highways
(see app. III).

To update information on federal, state, and local capital investment in
highways, we obtained 2002 (the most recent year available) expenditure
data from the Federal Highway Administration. We converted these
expenditure data to 2001-year dollars to coincide with the data in our
previous report,1 which presented data from 1982 through 2001.

To assess the influence that federal-aid highway grants have had on all
state and local highway spending, we reviewed and synthesized the research
literature on this issue. Our literature review revealed a number of
studies that used statistical models to estimate the influence of federal
funding on state spending. These models examined different time periods,
employed different statistical methods, and considered different potential
social, demographic, economic, and political factors that may affect state
highway spending decisions. None of the models used in the studies we
reviewed included the most recent data now available on highway funding,
and none examined whether the effect of federal grants on state spending
changed during the time period covered in the study. Therefore, based on
the models used in the earlier studies, we developed our own statistical
model of state highway capital and maintenance outcomes to estimate the
fiscal effects of federal highway funding on state highway spending. The
purpose of our statistical model was to isolate the effect of federal
grants on highway spending in states by controlling for other factors that
affect state spending decisions. Our model therefore considered a wide
range of potential factors such as economic conditions and the size of a
state's highway system, that may affect state spending choices. In
addition, our

1GAO, Trends in Federal and State Highway Investment (GAO-03-744R).

Appendix I
Objectives, Scope, and Methodology

model included the most recent data available and examined whether the
effect of federal grants on state spending changed during the time
period.2 A more detailed description of the literature and our statistical
model is contained in appendix II. Finally, our model was reviewed by
experts in the Department of Transportation (DOT) and peer reviewed by
three authors of the earlier studies on the fiscal effects of federal
highway grants. These experts and authors generally agreed with our
methods, and we made revisions based on their comments as appropriate.

To address the implications of the effect of federal highway grants on
state and local highway spending and options raised by these implications,
we reviewed pertinent legislation and congressional actions affecting the
federal-aid highway program, including goals, funding trends, program
features, and financing mechanisms. We reviewed the Government Performance
and Results Act and DOT's strategic and performance plans and reports for
2003 and 2004. We then evaluated how our model results and other analysis
on the existence of substitution affect the design and performance of the
federal-aid highway program.

Finally, to identify state characteristics associated with their effort to
fund highways from state resources, we defined a state's level of effort
broadly to include both a state's and its local governments' spending for
highway maintenance and capital construction relative to the personal
income of state residents.3 We determined a multivariate analysis is
required so that other factors, in addition to the state characteristic
under consideration, can be taken into account and held constant. (See
app. III for results.)

To perform this multivariate analysis, we utilized the same statistical
model of state highway spending used to analyze the fiscal effect of
federal highway grants (see app. II). The variables expected to affect
state highway spending fall into four broad categories: (1) fiscal
capacity, (2) the cost of transportation services to the representative
voter/consumer (tax price),

2We adjusted for the changing cost of highway services over time by
deflating expenditures using the chain-price deflator for state and local
government streets and highways published by the Bureau of Economic
Analysis.

3Our previous report defined level of effort more narrowly as highway
capital spending as a percentage of gross state product. We have adopted a
broader definition for the current analysis that is consistent with the
scope of this report's focus on highway spending broadly defined to
include capital as well as maintenance spending. Also consistent with this
report, we have used the personal income of state residents rather than
gross state product to reflect states' fiscal capacities.

Appendix I
Objectives, Scope, and Methodology

(3) federal grants, and (4) indicators of state preferences for highway
spending. The specific variables we considered are listed in
table 6.

We conducted our work from August 2003 through July 2004 in accordance
with generally accepted government auditing standards.

Appendix II

Description of Grant Substitution Model, Statistical Methods, and Results

This appendix presents a thorough description of the statistical analysis
that we conducted to estimate the extent to which states substitute
federal highway grants for funds that would have been spent on highways
from their own resources. The first section summarizes the literature on
this topic because we built upon models from previous studies in
developing our model. The next section describes the model that we
developed. The final section describes the statistical tests that we used
and presents the results of those tests.

Summary of Previous Studies

We reviewed a number of studies on substitution and relied most heavily on
the models used in three of them in developing our statistical model.1 The
three studies are similar in that each draws upon economic models that
explain states' highway spending in terms of the demand for mobility that
flows from the construction and maintenance of a highway network. Within
the context of these models, the potential for grant substitution arises
in the response of state highway spending to changes in federal grant
funding. However, these models differ in key details, such as the
statistical methods used to estimate the extent of substitution, the
definition of state highway expenditures, and the control variables used
in the model. They also differ in their estimation of substitution rates.

Conceptual Framework	The models in each of the key studies are built upon
the premise that the political process responds to the preferences of
voters/consumers for highway transportation services. As a result, the
models characterize the demand for and supply of highway spending as
depending on four types of factors:

1The three studies that we relied on most heavily were: Harry G. Meyers,
op. cit; Brian Knight, op. cit; and Shama Gamkhar, op. cit. Other studies
that we reviewed include: Shama Gamkhar, "Is the Response of State and
Local Highway Spending Symmetric to Increases and Decreases in Federal
Highway Grants?" Public Finance Review, Vol. 28 No. 1, January 2000 pp.
3-25: Janet G. Stotsky, "State Fiscal Responses to Federal Government
Grants," Growth and Change, Summer 1991, pp 17-31; Roger D. Congleton and
Randall W. Bennett, "On the Political Economy of State Highway
Expenditures: Some Evidence of the Relative Performance of Alternative
Choice Models," Public Choice, 84 (1995), pp. 1-24; Rajeev K. Goel and
Michael A. Nelson, "Use or Abuse of Highway Tax Revenues?: An Economic
Analysis of Highway Spending," unpublished draft, April 2001; Edward
Miller, "The Economics of Matching Grants: The ABC Highway Program,"
National Tax Journal, XXVII, no. 2, pp. 221-229; Herman B. Leonard, By
Choice or By Chance: Tracking the Values in Massachusetts' Public Spending
(Pioneer Institute for Public Policy Research, 1992).

Appendix II Description of Grant Substitution Model, Statistical Methods,
and Results

1.	Fiscal capacity (FC), which is the ability of states to fund services
using within-state resources;

2.	The tax price (TP) faced by the typical voter/consumer of highway
services, which can be thought of as the cost of an additional unit of
mobility;

3.	Intergovernmental grant funding (G), including both grants intended for
highways and grants for other public services; and

4.	Differences in voter/consumer preferences (P) for highway
transportation services.

This relationship can be summarized in the following relationship:

State Highway Expenditure = f(FC,TP, G, P)

In these models, greater tax paying capacity is expected to result in a
higher demand for mobility that in turn increases the demand for a larger
highway network. Similarly, more grant funding (both for highways as well
as for other public services) increases the resources available to states
and is expected to increase total highway spending. Differences in
political culture are also expected to result in different preferences for
transportation services relative to other public services, such as health
and education. Finally, if the typical voter/consumer faces a higher unit
cost of transportation services, also called the tax price of highway
services, the demand for transportation services is likely to be lower.

The tax price of highway services is, in turn, dependent upon several
factors:

1.	A higher cost of inputs (labor, building materials, supplies, etc.)
used to build and maintain highways results in more expensive
transportation services. A higher unit cost of mobility is expected to
reduce the demand for transportation services, but will increase highway
spending as long as the demand for transportation services is price
inelastic.

2.	Economies and/or diseconomies of scale may also affect the unit cost of
mobility. A required minimum facility size may result in more lane miles
per resident in smaller states, which may result in a higher unit cost for
the typical voter/consumer. Similarly, very low lane miles per

Appendix II Description of Grant Substitution Model, Statistical Methods,
and Results

resident may be associated with more intensive usage, which may also
result in a higher unit cost as well. Thus, unit cost may be U-shaped.

3.	A greater number of voter/consumers with whom the cost of highway
services may be shared is expected to reduce unit cost to the typical
state voter, increasing the demand for transportation services. This will
result in higher total highway spending and lower spending per voter so
long as demand is price inelastic.

4.	More highway users may lead to greater deterioration in the quality of
highways and greater congestion, raising the unit cost of transportation
services to the typical voter/consumer, reducing the demand for highway
services. The effect on spending is expected to be positive if demand is
price inelastic. In addition to cost considerations, more users could also
be thought of as reflecting a stronger preference for highway services
relative to other goods and services.

5.	Matching grants on the marginal dollar of highway spending reduce the
unit cost of services to the typical voter/consumer. To the extent that
matching requirements apply to additional state spending the typical
voter/consumer pays a smaller share of additional spending, lowering the
cost of additional spending to the typical voter/consumer and raising the
demand for highway services.2

In table 4, we summarize three studies that are representative of the
variety of models that have been considered in the literature and upon
which we base our analysis.

2Closed-ended matching programs that limit the availability of federal
matching funds at the margin of spending do not lower the cost of
additional highway spending and hence the tax price faced by the typical
voter/consumer.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

                Table 4: Summary of Fiscal Substitution Studies

        Time  State highway    Fiscal                                      
Study  period  expenditures   capacity Tax price     Grant     Preferences Other     
                                       variables   variables               variables 
                                                                               State 
       1983-                  Personal     1.       Highway    1. Governor     fixed 
Knight        State (but not           Population    grant                   effects 
        1997      local)      Income   2. Drivers expenditures  Democrat   
                              per      per capita per                      
                               capita      3.       capitaa    2. Percent  
                government             Registered                          
               spending for             vehicles                 Democrats 
                                          per                           in 
              highway-related            capita                State House 
              projects                                         3. Percent  
              (capital                                                     
                    and                                          Democrats 
                                                                        in 
               maintenance,                                          State 
                                                                    Senate 
              real per                                                     
              capita)                                                      

Gamkhar 1976-  State & local  Personal 1. Effective  1.   Highway    1. State      
                                                                        fixed      
        1990                  income      nonhighway                     effects   
                government    per                          grant        
               spending for    capita     match rate    expenditures 2. Time fixed 
              highway-related          2. Registered    per capitaa      effects   
              projects                     Vehicles  2.   Highway    3.  Percent   
              (capital                       per                        
                    and                     capita         grant        unemployed 
               maintenance,            3.  Vehicle      obligations  4. Debt as a  
                                            miles                       
              real per                     traveled      per capita     percent of 
              capita)                        per                        
                                            capita   3.  Other fed        income   
                                       4.  Percent                      
                                            light        grants per     
                                           vehicles        capita       
                                       5.  Percent                      
                                            metro                       
                                          population                    
                                       6. Population                    
                                           density                      

Meyers 1976- State capital  Personal   1. Effective     1. Highway       
                                             highway          
          1982  spending on    income per    match ratea      grant         
                federal-aid    capita     2. Effective         expenditures 
                highways (net                nonhighway       per capita    
                of                                            
                interstate                   match rate    2. Other federal 
                highways, real            3. Registered       grants per    
                per capita)                  Vehicles per     capita        
                                             capita           
                                          4. Vehicle miles    
                                             traveled per     
                                             capita           
                                          5. Lane miles       
                                             per              
                                             capita           

Source: GAO analysis.

                      aTreated as an endogenous variable.

Statistical Methods	The three studies employ a variety of statistical
methods in estimating the substitution effect of federal highway grants.
All use simultaneous equations estimators, but they treat different
variables as endogenous. Knight and Gamkhar treat federal grant
expenditures and state own-source

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

highway expenditures as jointly determined and therefore use an
instrumental variable estimator for their per capita federal grant
variable to remove the endogenous component associated with this
variable.3 In contrast, Meyers does not treat per capita federal highway
grants as an endogenous variable and may have a biased estimate of the
substitution rate. He does, however, treat the effective matching rate
associated with highway grants (i.e., the ratio of highway grants to total
highway spending) as endogenous and uses an instrumental variable
procedure to correct for potential bias in that variable.4

Both Gamkhar and Meyers find autocorrelation in their error terms and,
therefore, make an adjustment for autocorrelation. The Knight study does
not correct for autocorrelation. Finally, both Knight and Gamkhar use a
fixed effect estimating procedure to control for unique circumstances
across states that are not captured by the other control variables
included in their models. Neither study reports the significance of fixed
effects in their model. In addition, Gamkhar also includes time dummy
variables to capture systematic effects over time that the other control
variables do not capture. Knight does not include a time adjustment in his
model. Meyers includes neither a fixed effects nor time adjustment.

Differing Definitions of State Highway Expenditures

Each of the three studies define state highway spending differently, which
has important implications regarding how grant substitution is measured
and influences the interpretation of the studies' results.5 The earliest
study, by Meyers, includes state capital spending only for projects
eligible under the federal-aid highway program, excluding spending for
interstate highways. Measuring the dependent variable in this way means
the

3If state expenditures and federal grants were jointly determined,
ordinary least squares (OLS) methods would provide biased estimates of the
substitution rate. Gamkhar does, however, use OLS methods in a model in
which federal grants are measured using grant obligations, arguing that
obligations are known prior to states' expenditure decisions and therefore
treated obligations data as exogenous.

4Neither Knight nor Gamkhar includes a price effect associated with
federal highway grant funding arguing that such an effect is not present
because states spend substantially more than is required to satisfy the
matching requirements associated with federal funding.

5Gamkhar uses the implicit price deflator for government purchases, and
Meyers uses the implicit price deflator for state and local government
purchases to adjust for changes in the purchasing power of a dollar over
time. Knight does not identify the deflator he uses. None of the studies
adjusts for cross-state differences in input costs. All three studies
express state highway spending and federal grants on a per capita basis.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

highway grant coefficient measures only the response of state capital
spending on federal-aid highway projects to changes in federal funding. As
a consequence, Meyers counts increased state or local spending for
maintenance on federal-aid highway projects, or increased state or local
capital and maintenance on nonfederal-aid highway spending, as grant
substitution in the same way as increased spending for other state
services such as education and health or increased state taxpayer relief
would be counted as substitution.

In contrast, Knight defines state highway spending more broadly to include
all highway spending by state governments, whether for federal-aid
highways or for other state highway projects. However, Knight does not
include local spending on highways in his definition of state highway
spending. As a consequence, increased state maintenance spending on
federal-aid highway projects or spending on state government highway
projects that are not part of the federal-aid system is not considered
grant substitution in his study, even though such spending is not eligible
for federal assistance. However, increased highway spending by local
governments is considered to be grant substitution in the same way that
increased state or local spending for other state services and increased
tax relief are considered substitution. Finally, the Gamkhar study defines
highway spending to include both capital and maintenance spending by both
state and local governments. This study, therefore, counts only increased
state or local spending for nonhighway purposes, including increased tax
relief, as representing grant substitution.

The Estimated Effect of Federal Highway Grants on State Highway Spending

All three studies use federal grants expenditures to measure federal
grants received by states. This variable is statistically significant in
all studies. In addition to grant expenditures, Gamkhar also considers
grant obligations as an alternative measure. Since obligated funds are
available for expenditure for several years, she included this variable
with lagged values.6

The reported estimates of substitution rates associated with federal
highway grants vary across the three studies. These differences are, in
part, due to differences in the time periods studied, the definitions of
state

6Gamkhar treats grant obligation data as a predetermined variable and
therefore uses OLS methods to estimate the grant coefficient. However, if
state spending and federal grants are jointly determined this may yield
biased estimates of the substitution rate.

Appendix II Description of Grant Substitution Model, Statistical Methods,
and Results

highway spending, and the statistical methods employed. Among the
highlights of the studies were the following:

o 	Knight's study reports a grant substitution rate of over 90 percent for
the period from 1983 to 1997. Knight defines substitution as the reduction
in state (but not local) government spending on all highway-related
projects.

o 	Gamkhar reports a substitution rate of 63 percent for the period 1976
through 1990. Gamkhar defines substitution as the reduction in state and
local government spending on all highway-related projects; Gamkhar
measured federal grants using grant expenditures. When grants were
measured using obligations rather than actual grant expenditures, a lower
substitution rate of 22 percent is reported.7

o 	Meyers also reports a 63 percent substitution rate for the period 1976
through 1982. Meyers defines the substitution rate as the reduction in
state and local government spending on federal-aid eligible highway
projects net of spending on the Interstate Highway Systems; federal grants
are measured using grant expenditures. However, when he defined
substitution as the increase in state and local government nonhighway
spending, he reports no substitution.

7The estimate based on grant obligations data may be downwardly biased if,
as argued by Knight, federal grants and state expenditures are jointly
determined (see Knight op. cit, p. 77).

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

  Table 5 summarizes the definitions used and findings of these three studies.

Table 5: Highway Grant Substitution Rates Reported in Fiscal Substitution
                                    Studies

Definitions of state highway Study expenditures Substitution rate

               Knight           State (but not local) 91 percent 
                                           government 
                       spending for highway-related   
                           projects (capital and      
                       maintenance, real per capita   
                                 dollars)             

            Gamkhar  State and local government             63 percent (grant 
                                                                expenditures) 
                    spending for highway-related 22 percent (grant            
                                                 obligations)                 
                           projects (capital and 
                    maintenance real per capita  
                                        dollars) 

Meyers	State capital spending on federal-63 percent aid highways (net of
interstate highways, real per capita)

State and local government 0 percent
nonhighway spending, (real per
capita)a

Source: GAO analysis.

aMeyers's formal test for substitution into nonhighway spending is to test
whether federal highway grants are systematically related to state
nonhighway spending. He finds no statistical evidence of such a
relationship.

Controls for Other Factors To isolate the effect of federal highway grants
on state highway spending, Associated with State these studies include
additional variables in their models to control for Spending Choices other
factors also related to state spending. Some of the control variables

               are similar across the studies, but others differ.

Fiscal Capacity 	All three studies use per capita personal income to
represent states' funding capacity, and in each study the variable is
found to be statistically significant.

Tax Price 	All three studies include a wide variety of variables that are
intended to capture various components of the tax price faced by the
typical voter/consumer.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

Input prices	All three studies measure financial variables in real dollars
by adjusting for price level differences over time but otherwise do not
explicitly include an input cost adjustment as a tax price proxy, except
to the extent that the fixed effects procedure employed by Knight and
Gamkhar capture these differences.8

Highway System Size 	Only Meyers uses an indicator of highway system size:
lane miles on federal-aid highways. While this variable has the expected
positive sign it is statistically insignificant. However, a quadratic term
to capture a possible U-shaped functional form was not used.

Highway usage 	All studies use the number of registered vehicles as a
measure of highway usage. In addition, Knight uses the number of drivers,
whereas Meyers includes vehicle miles traveled. Gamkhar includes several
additional proxies for highway use that are not included in the other
studies: the percentage of light motor vehicles, population density, and
percentage of population living in metropolitan areas. However, none of
these factors was statistically significant. In general, only one of the
use variables is statistically significant in each study and no one
measure is statistically significant across studies. In several instances
the coefficient has a negative sign, although a positive relationship
between highway usage and state spending would be expected.

Highway matching rates	Although highway grants require state matching,
Knight and Gamkhar do not include highway matching rates as part of their
models because they found that states' highway spending exceeds the
amounts required for their federal grant allotments and, therefore, have
only an income effect but no price effect. Meyers, in contrast, does
include the effective matching rate (highway grants as a percentage of
highway expenditures) and reports a price elasticity of one.

Nonhighway matching rates	Other grants may also have a price effect
because programs such as Medicaid, Foster Care, and Adoption Assistance
are all open-ended matching grants. Including the effective matching rate
associated with other grant spending (i.e., other grants as a percent of

8However, to the extent that cross-state differences in input costs is
relatively stable over time, Knight and Gamkhar may have accounted for
these differences through the fixed effects estimating procedure.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

nonhighway spending) captures the potential price effect of other grants.9
The sign on the effective matching rate is expected to be negative because
higher demand for other state services would reduce the demand for highway
spending. These variables are statistically significant in both studies.
The Knight study does not consider the tax price effect of nonhighway
grant funding.

Cost sharing 	Only Knight, by including population in his model, includes
a factor that could be interpreted as reflecting the cost-reducing effect
of having more taxpayers sharing the cost of highway services. Neither
Gamkhar nor Meyers includes such a factor.

Other Grant Funding 	In addition to the tax price effect of nonhighway
grant funding, the studies may also have income effects. Both Meyers and
Gamkhar include other nonhighway grants per capita in their models to
capture the income effect of these grants.10 The income effect is expected
to have a negative effect on own-source spending as some of these grants
may be substituted into highway spending and supplant funding from state
resources. The Knight study does not consider either price or income
effects associated with nonhighway grant funding.

Political Culture/Preferences 	Only Knight includes variables that are
intended to reflect differences in state preferences for highway spending
that may be associated with the political party of the state governor and
the partisan representation in the state legislature. He finds the party
of the state governor to be statistically significant at the 10 percent
level, while the other political variables are not statistically
significant.

Description of GAO's Statistical Model

Consistent with previous studies, we model state spending choices as being
conditioned on states' fiscal capacities, the tax price faced by state
voters, federal grant funding for highways and for other state services,
and preferences of state voters for highway spending. Because both theory
and the results of previous studies suggest that federal grants and state

9The effective matching rate is measured by expressing other grant funding
as a percentage of total nonhighway spending.

10As noted above, both studies also include these grants as a share of
total nonhighway spending to capture any price effect these grants may
have.

Appendix II Description of Grant Substitution Model, Statistical Methods,
and Results

spending decisions are jointly determined, we use an instrumental
variables (IV) approach to estimate the fiscal effect of federal grants.

To capture other factors that may be systematically associated with
differences in state spending choices, we estimate the model using a fixed
effects estimating procedure. The fixed effects procedure is intended to
capture factors such as topographical differences and weather conditions
across states that do not change over time and to capture other unmeasured
factors with large cross-state variation that exhibit relatively little
change over time.11 In addition, we include a time trend to capture trend
changes in state spending that may not be captured by the other variables
included in our model.

The specific variables considered for our model are listed in table 6.12

Table 6: Variables Considered in the Second Stage State Highway
Expenditure Equation

Variable name

                        Dependent variable (per capita)

  Real state and local government spending for highway capital and maintenance

                          Fiscal capacity (per capita)

Real personal income

Real income squared

Tax price

              Vehicle miles traveled per capita Drivers per capita

Registered motor vehicles per capita

                   Effective match rate of nonhighway grants

Population

                             Lane miles per capita

11Both Knight and Gamkhar adopt this strategy.

12Consistent with previous studies, we expressed state highway spending on
a per capita basis and adjusted for the changing cost of highway services
over time by deflating expenditures using the chain-price deflator for
state and local government streets and highways published by the Bureau of
Economic Analysis as part of the National Income Accounts.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

(Continued From Previous Page)

                                 Variable name

                         Squared lane miles per capita

Federal grants (per capita)

Real highway grantsa

Real other federal grants

State preferencesb

Governor democratic

           Percentage of state House represented by Democratic party

           Percentage of state Senate represented by Democratic party

                                Other variables

         Utah Olympics (=1 for 1997-2000) Time trend Time trend squared

                               Inverse time trend

State fixed effects

Source: GAO analysis.

aPredicted values of federal highway grants.

bWith the exception of some independents, office holders are either
Democratic or Republican. Therefore, the choice of using the percentage
Democrats or Republicans is arbitrary and has no effect on the statistical
results except to change the sign of the regression coefficient.

Testing the Stability of the Substitution Rate

With each time period, various rules and regulations change that may
affect the ability of states to substitute federal grants for state
spending. Given the range of estimates over different time periods
reported in past research, we also want to test whether the rate of grant
substitution, if found, systematically differ across the time periods
included in our data. To see if the substitution rate differs over time,
we introduce dummy variables for each of the time periods covered in our
study into our model.13 We then multiply these dummy variables by the
grants variables and included these interaction variables in the model. If
statistically significant, these variables would provide evidence that
substitution has varied from one time period to another.

13Our grant expenditures variable reflects, in part, obligations made in
prior years. As a consequence, the substitution rate for a particular time
period will, in part, be conditioned on grant obligations made in prior
time periods.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

Definition of State Highway Spending

The estimated effect of federal highway grants on state highway spending
is measured by the regression coefficient associated with federal highway
grants. As a consequence, the interpretation of that coefficient is
directly affected by how the dependent variable, state expenditures, is
defined. If we defined state highway spending narrowly as only capital
expenditures on federal-aid highway projects, the federal grants
coefficient in our model would be interpreted as the response of state
capital spending to changes in federal highway aid. This approach, taken
by Meyers,14 represents a definition of state spending that is consistent
with the requirements of the federal-aid highway program, which restricts
federal grants to authorized uses, such as capital investment on eligible
federal-aid highway routes. Under this approach, grant funds that are used
for purposes that are not eligible for federal aid, would represent grant
substitution in the same way that increased spending for health and
education and for state tax relief would represent grant substitution.
Some policymakers may not view this as substitution, perhaps arguing that
state transportation officials are better positioned to determine the best
use of available funding for highway-related projects. Our analysis uses
this broader definition of state highway spending. Thus, our measure of
grant substitution considers only state grant funds that are effectively
used for nonhighway purposes as substitution.15

We adopt this approach for two reasons. First, we want to be conservative
in our definition of grant substitution. A broader definition of state
highway spending that includes state and local spending on highway
projects not eligible for federal funding would yield a lower estimate for
the substitution rate because some types of adjustments would not be
treated as grant substitution. Second, an estimate of grant substitution
that is based only on state (but not local) government spending would be
affected by cross-state differences in the extent to which highway
spending is centralized at the state level. Since there are large
differences across states in the extent to which highway spending is
centralized, we include local as well as state government spending so that
our measure of highway spending would be comparable across states.

14Meyers, op. cit.

15Knight, op. cit, takes an intermediate approach by using only state
government capital and maintenance spending, excluding highway spending by
local governments.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

Definition of Federal Highway Grants and Specification of the Federal
Grants Equation

Because federal highway grants are provided on a reimbursement basis, we
obtained from FHWA federal highway grant expenditures that are
contemporaneous with states' reported own-source highway spending. As with
state spending, we express federal grants in real per capita dollars,
using the BEA chain-price index for state and local government streets and
roads. Because federal grant expenditures, by definition, represent
formula grant allotments from current and prior years, any lagged response
in state spending to federal highway grant funds is already included in
our grants variable. We therefore do not include lagged values of federal
highway grants in our model.

Knight provides an economic argument explaining that state highway
spending and federal grant funding are jointly determined because elected
officials reflect the preferences of state voter/consumers both in state
legislatures and in Congress. His study tests for and finds confirming
evidence for his theoretical argument.16 Based on these findings, we also
employ an IV estimator that provides a consistent estimate of the federal
grant coefficient to measure the fiscal effect of federal grants. Using
this approach, we estimate a first stage instrumental variable equation
that models federal highway funding in terms of exogenous variables that
are expected to influence the distribution of federal grants. The
instrumental variables include the exogenous variables from the state
expenditure equation (e.g., fiscal capacity, the individual components of
tax price, and preferences) and variables that are highly correlated with
federal grants but uncorrelated with state highway spending (e.g.,
variables included in federal grant formulas and those that may affect the
distribution of discretionary grants). Predicted values of federal grants,
derived from the instrumental variables (highway grants) equation, are
then used in lieu of actual grant values to correct for the bias in
ordinary least squares (OLS) estimates of the federal grants coefficient
in the state expenditure equation.

The excluded exogenous variables we consider include state contributions
to the highway trust fund and variables that are intended to reflect the
influence of state representatives on the distribution of federal highway
grants: tenure in Congress, state representation on transportation
committees, and state representation in the majority party. The exogenous
variables we consider are summarized in table 7.

16Knight, op. cit.

Appendix II Description of Grant Substitution Model, Statistical Methods,
and Results

 Table 7: Variables Used to Explain the Distribution of Federal Highway Grants

Instrumental Variables Equation:Federal Highway Grants

Exogenous variables from spending equation

Real personal income per capita

Real personal income per capita squared

Real nonhighway federal grants per capita

Effective nonhighway matching rate

Lane miles per capita - 1-yr. lag

Lane miles per capita squared - 1 yr. lag

Vehicle miles traveled per capita

               Drivers per capita Registered vehicles per capita

Population

                     Governor democratic (1=Dem. 0 = other)

Percent Democrats in state house

Percent Democrats in state senate

Utah Olympics (=1 for 1997 -2000)

Time trend

Time trend squared

Inverse time trend

State fixed effects dummy variables Excluded exogenous variables
Percentage of state representatives in majority party (in year grants were
authorized) Percentage of state representatives on House transportation
authorization committee Average tenure of state representatives in House
(in year grants were authorized) Percentage of state senators in majority
party (in the year grants authorized) Percentage of state representatives
on Senate transportation authorization committee Average tenure of state
Senators (in year grants were authorized) Real federal highway trust fund
receipts per capita Source: GAO analysis.

Consistent with the state highway spending equation, we include real per
capita income, real nonhighway grant funding, registered vehicles,
licensed drivers, and vehicle miles traveled-including 1-and 2-year lagged
values for each of these variables-and use a fixed effects estimating
procedure. Fixed effects are intended to capture factors that have
substantial variation across states with little variation over time.
Examples would be factors such as state land area-a factor that has been
part of highway funding

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

formulas and that does not change over time-and constraints that are
applied to funding formulas, such as the  1/2-of 1 percent minimum state
grant that is included in highway funding formulas (see table 1).

State Funding Capacity 	Consistent with previous studies, we use real per
capita personal income to measure states' taxing capacities. Unlike
previous studies, we also include the squared value of per capita income
to capture the possibility that demand for highways does not increase in
proportion to increases in income, perhaps signifying that as basic
transportation needs are met, increases in income are increasingly
allocated to other uses such as health and education. Personal income is
published by the BEA in the Department of Commerce. We include 1- and
2-year lagged values of real per capita income in the model to allow for
lagged responses to changes in income and also to reflect cyclical changes
affecting the level of state revenues.

Tax Price	The tax price faced by state voters/consumers is reflected in a
number of variables included in the model. Highway usage is reflected by
vehicle miles traveled on state highways, and by registered vehicles and
licensed drivers in the state, as reported by FHWA. We include 1-and
2-year lagged values in each of these variables to allow for lagged
responses in spending to changes in highway usage.

Consistent with prior studies, we do not include the matching rate on
highway grants because states spend more than the required federal match,
and therefore, states pay 100 percent of the cost of funding additional
highway projects, and because highway matching rates vary little both over
time and across states. However, we do include the effective match rate on
other grant funding to capture the price effect of other grant funding.
Medicaid, Foster Care, and Adoption Assistance, for example, are openended
matching programs with price effects that may encourage states to spend
less on highways in order to provide matching funds for these and possibly
other matching programs. We include 1-and 2-year lagged values to capture
these effects. Using data from the Census Bureau, we measure the effective
matching rate for nonhighway spending by deducting states' federal highway
grants from their total federal grants and expressing the net amount
(nonhighway grants) as a proportion of each state's nonhighway spending,
also calculated by deducting highway spending from total spending.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

Although previous studies do not include the size of the highway network
to be maintained, we expect the per capita cost of maintaining an existing
highway network to be higher in states with more miles of road per capita.
Therefore, we include this variable in our model along with its squared
value to test for evidence of per capita costs varying with the scale of
the road network-that is, economies or diseconomies of scale. We obtained
data on total lane miles of state highways from FHWA.

Other Federal Grants	In addition to federal highway grants, states receive
federal grants for a variety of other purposes, including health,
education, and welfare. While it is possible that state highway funds may
be substituted into spending for other state services, it is also possible
that some state funds that would have otherwise been used for other
purposes may be redirected into highways. For this reason, we also include
other federal grant funding in our model to capture the income effect of
these grants and their potential substitution into highway spending. While
some of this aid is provided on a reimbursement basis (Medicaid, for
example) other grants can remain eligible for expenditure in subsequent
years. For this reason, we include 1and 2-year lagged values of other
federal grants to capture these potential effects. Other federal grants
are also expressed in real per capita dollars.

State Preferences	The political culture of states may affect both the
overall level of spending on public services, as well as spending
priorities for different types of services, such as highways, versus
education and health care. Differences in political culture and spending
priorities may be relatively stable over time, in which case the fixed
effects adjustment may adequately control for cross-state differences in
these spending preferences. Nonetheless, in addition to including fixed
effects, we have also included variables that may be associated with
differences in political culture. For this purpose, we have included dummy
variables that are equal to one if the state governor is Democratic and
zero otherwise, and the percentage of the state Senate and state House
that is represented by the Democratic Party. With the exception of some
independents, office holders are either Democratic or Republican.
Therefore, the choice of using the percentage Democrats or Republicans is
arbitrary and has no effect on the statistical results except to change
the sign of the regression coefficient. We obtained these data from the
Elections section of the Census Bureau's Statistical Abstract.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

Other Variables	To capture trend changes in state spending that cannot be
captured by the other variables included in our model, while allowing for
a possible curvilinear trend, we have also included time, time squared,
and the inverse of time. Finally, we include a dummy variable for the
state of Utah that was equal to 1 during the years 1997 through 2000 and
zero otherwise to account for the unusually large increase in highway
spending in that state just prior to the 2002 Winter Olympics.

The means and standard deviations for the variables included in our
statistical model are shown in table 8.

Table 8: Descriptive Statistics

Standard Variables Units Mean deviation

        Real State Highway Spending           Dollars per person        
                         Per Capita                               $203    $80 
        Real Federal Highway Grants           Dollars per person        
                         Per Capita                                 $94   $59 
        Real Non-Hwy Federal Grants           Dollars per person        
                         Per Capita                               $671   $235 
          Federal Nonhighway Grants                        Ratio        
              percent of Nonhighway                                     
                       Expenditures                                 16%    4% 
              Road Miles Per Person         Lane Miles per 1,000        
                                                      population     53 
                                         Registered vehicles per        
            Registered Vehicles per                        1,000        
                             Person                   population    786 
         Vehicle Miles Traveled per       1,000 miles per person        
                             Person                                   9 
        Licensed Drivers per Person   Licensed drivers per 1,000        
                                                      population    679 
             Real Per Capita Income           Dollars per person 21,272 4,223 
        Percent Democratic in State                      Percent        
                              House                                 57%   17% 
        Percent Democratic in State                      Percent        
                             Senate                                 58%   18% 
                 Governor Democrat,               Not applicable        
          (1=Democrat; 0=otherwise)                                 52%   50% 

Source: GAO analysis.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

Statistical Methods 	Because we use time series and cross-section data to
estimate the model, we expect autocorrelation to bias the estimates of the
standard errors associated with variables in our model. To reduce the
problem of heteroscedasticity, we normalize variables by expressing them
on a per capita basis (except for those already expressed in ratio or
percentage terms). We conducted statistical tests to determine if our data
are affected by autocorrelation and found statistical evidence of its
presence. Therefore, we estimate all our models using a correction for
autocorrelation. As noted above, we use a fixed effects procedure that
allows for a separate constant term associated with each state to
represent differences in state funding that are unique to each state and
independent of the other variables included in the model.

Additional Analysis of Fixed Effects

The fixed effects coefficients of our model represent state differences in
highway spending, after controlling for the other explanatory variables in
our model. They are intended to capture the effect of variables that have
comparatively little variation over time but are systematically associated
with differences in spending across states. To identify those state
characteristics that are systematically related to the fixed effects
associated with state highway expenditures, we perform an additional
stepwise regression analysis that regresses the following explanatory
variables on our estimated fixed effects, using the following:

o  Heating degree days,

o  State land area,

o  Lane miles per capita,

o  Population,

o  Vehicles per capita,

o  Drivers per capita,

o  Federal land area,

o  Percentage of Democrats in state House,

o  Percentage of Democrats in state Senate,

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

o  Governor Democratic,

o  Federal nonhighway grants per person, and

o 	Ratio of federal nonhighway grants per person to state nonhighway
spending

We use the mean value of 21 observations from 1980 to 2000 per state to
represent each variable in explaining our estimated fixed effects.

Statistical Results	We report the statistical results explaining federal
highway grants in terms of exogenous instrumental variables in table 9.
Based on the R2 statistics, the fixed effects adjustment accounts for 85
percent of the variation in federal grants funding and the additional
exogenous variables added to the model increases the R2 by 5 percent to 90
percent. Variables that are statistically significant at the 5 percent
level appear in bold in the table. In addition to the fixed effects
coefficients, per capita income, highway lane miles of roads, and state
contributions to the Highway Trust Fund are strongly associated with the
distribution of federal highway funding.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

Table 9: Instrumental Variables Estimator of Federal Grants per Capita Model R2

                            Constant term only 0.000

                         State fixed effects only 0.854

                            X -variables only 0.739

                           X and group effects 0.895

            Variables           Coefficients Standard error Probability value 
    Personal income, real per          0.015          0.007             0.035 
              capita                                        
    Personal income, real per          0.008          0.009             0.375 
          capita, (t-1)                                     
    Personal income, real per         -0.021          0.006             0.000 
          capita, (t-2)                                     
     Personal income squared,    -0.3121E-06     0.1500E-06             0.037 
         real per capita                                    
     Personal income squared,    -0.1150E-06     0.2005E-06             0.566 
      real per capita, (t-1)                                
     Personal income squared,     0.4831E-06     0.1288E-06             0.000 
      real per capita, (t-2)                                
    Nonhighway federal grant,          0.035          0.027             0.187 
         real per capita                                    
    Nonhighway federal grant,          0.133          0.032             0.000 
      real per capita, (t-1)                                
    Nonhighway federal grant,         -0.051          0.026             0.048 
      real per capita, (t-2)                                
    Effective nonhighway match      -201.622        129.481             0.119 
              ratea                                         
    Effective nonhighway match      -388.050        148.409             0.009 
           ratea (t-1)                                      
    Effective nonhighway match       139.497        118.758             0.240 
           ratea (t-2)                                      
    Vehicle miles traveled per        -7.331          3.204             0.022 
              capita                                        
    Vehicle miles traveled per        -0.852          3.721             0.819 
          capita, (t-1)                                     
    Vehicle miles traveled per         2.688          2.637             0.308 
          capita, (t-2)                                     
     Registered vehicles per           0.048          0.024             0.041 
              capita                                        
     Registered vehicles per          -0.013          0.028             0.643 
          capita, (t-1)                                     
     Registered vehicles per          -0.007          0.024             0.785 
           capita,(t-2)                                     
        Drivers per capita            -0.022          0.034             0.514 
     Drivers per capita,(t-1)         -0.018          0.038             0.634 
    Drivers per capita, (t-2)          0.041          0.033             0.221 
       Population, in 1,000            0.002          0.002             0.138 
Road miles per capita, (t-1)        1.231          0.313             0.000 
      Road miles squared per          -0.004          0.001             0.002 
          capita, (t-1)                                     
        Governor Democrat,            -3.836          1.737             0.027 
    (1=Democrat; 0=otherwise)                               
Percent Democratic in State         2.758         14.498             0.849 
              House                                         
Percent Democratic in State        32.801         11.695             0.005 
              Senate                                        
    Utah Olympics = 1 for Utah       -12.757         12.055             0.290 
    in 1997-2000, 0 otherwise                               
            Time trend                -3.123          2.627             0.234 

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

                         (Continued From Previous Page)

            Variables           Coefficients Standard error Probability value 
        Inverse time trend            66.101         33.632             0.049 
        Time trend squared            -0.009          0.089             0.923 
      Highway trust fund per           0.294          0.088             0.001 
              capita                                        
      Highway trust fund per           0.330          0.089             0.000 
          capita, (t-1)                                     
      Highway trust fund per           0.225          0.088             0.010 
          capita, (t-2)                                     
Percent of the State's House                             
        delegation on the                                   
           authorizing                                      
            committee                  8.233          5.761             0.153 
      Percent of the State's                                
     Senate delegation on the                               
           authorizing                                      
            committee                -14.760          4.515             0.001 
      Percent of the State's                                
     Senate delegation in the                               
             majority                                       
              party                    1.649          2.223             0.458 
Percent of the State's House                                               
    delegation in the majority         5.940          3.863             0.124
              party                                         
Tenure for each State's U.S.        0.148          0.283             0.601 
         House Delegation                                   
Tenure for each State's U.S.        0.063          0.202             0.756 
        Senate Delegation                                   

Source: GAO analysis.

aRatio of federal nonhighway grants to state and local nonhighway
expenditures.

Displacement Effect of Federal Highway Grants

We report the results for the second stage expenditure equation without a
correction for autocorrelation in table 10. Again, regression results for
variables that are statistically significant at the 5-percent level appear
in bold in the table. The model explains 78 percent of the variation in
state own-source highway spending and fixed effects alone account for 69
percent of the variation. The estimated substitution rate associated with
federal grants is 84 percent17 and is statistically significant. That is,
other things being equal, a dollar increase in federal highway grants is
associated with an 84-cent reduction in highway spending from state
own-source revenues. Alternatively, the coefficient also implies that
states replace 84 cents of each dollar decline in federal funding.18 These
results are similar to the findings reported by Knight, who reported a
substitution rate of 91 percent, higher than the substitution rates
reported by Gamkhar.

17The substitution rate is the coefficient of the federal grants variable
after removing the negative sign.

18In a related paper Gamkhar tests whether the substitution rate is
symmetrical during periods of rising and falling federal aid and found no
statistically significant difference; see Shama Gamkhar, op. cit.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

Table 10: Instrumental Variables Estimates of State Highway Spending
Model, Without Correcting for Autocorrelation

                              Model statistics R2

                            Constant term only 0.000

                         State fixed effects only 0.685

                            X -variables only 0.445

                           X and group effects 0.783

               Variables              Coefficients Standard error Probability 
     Predicted FHWA payments / per         -0.8412          0.233       0.000 
                 capita                                           
Personal income, real per capita,        0.0160          0.013       0.221 
                  (t)                                             
Personal income, real per capita,        0.0196          0.017       0.249 
                 (t-1)                                            
Personal income, real per capita,        0.0110          0.011       0.322 
                 (t-2)                                            
Personal income squared, real per   -0.6259E-07   2.860E-07          0.827 
              capita, (t)                                         
Personal income squared, real per   -0.0416E-07   3.857E-07          0.280 
             capita, (t-1)                                        
Personal income squared, real per   -0.9145E-07   2.572E-07          0.722 
             capita, (t-2)                                        
    Nonhighway federal grants, real         0.0719          0.051       0.159 
            per capita, (t)                                       
    Nonhighway federal grants, real         0.0789          0.070       0.261 
           per capita, (t-1)                                      
    Nonhighway federal grants, real        -0.0483          0.051       0.341 
           per capita, (t-2)                                      
    Effective nonhighway grant match     -509.8810    250.807           0.042 
               ratea, (t)                                         
    Effective nonhighway grant match     -394.9280    300.760           0.189 
              ratea, (t-1)                                        
    Effective nonhighway grant match     -102.9480    230.545           0.655 
              ratea, (t-2)                                        
Vehicle miles traveled per capita,      -6.4858          6.106       0.288 
                  (t)                                             
Vehicle miles traveled per capita,       7.8859          7.135       0.269 
                 (t-1)                                            
       Vehicle miles traveled per           2.0446          5.113       0.689 
              capita,(t-2)                                        
    Registered vehicles per capita,         0.0444          0.046       0.338 
                  (t)                                             
    Registered vehicles per capita,         0.0210          0.054       0.695 
                 (t-1)                                            
    Registered vehicles per capita,        -0.0683          0.046       0.135 
                 (t-2)                                            
        Drivers per capita, (t)            -0.1290          0.065       0.046 
       Drivers per capita, (t-1)           -0.0498          0.072       0.491 
       Drivers per capita, (t-2)            0.0777          0.064       0.227 
    Governor Democrat, (1=Democrat;         1.2802          3.400       0.707 
              0=otherwise)                                        
      % Democratic in State House          -4.5781         27.136       0.866 
      % Democratic in State Senate         46.8598         23.539       0.047 
Utah Olympics = 1 for 1997-2000 in     154.8240         22.803       0.000 
           Utah, 0 otherwise                                      
               Time trend                 -32.6570          4.883       0.000 
           Inverse time trend             -95.4868         54.237       0.078 
           Time trend squared               1.0301          0.153       0.000 

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

                         (Continued From Previous Page)

               Variables              Coefficients Standard error Probability 
          Population, in 1,000              0.0033          0.003       0.287 
      Road miles per capita, (t-1)          1.1212          0.677       0.098 
     Road Miles squared per capita,        -0.0013          0.003       0.630 
                 (t-1)                                            
      Autocorrelation coefficient           0.5298                

Source: GAO analysis.

aRatio of federal nonhighway grants to state and local nonhighway
expenditures.

However, the model also indicates the presence of autocorrelation (r=0.53,
shown in the last row of table 10). As a consequence the standard error
for the grants coefficient is biased downward, which raises the prospect
that the grants coefficient may not be statistically significant. We
therefore reestimated the model adjusting for autocorrelation using two
methods: Cochrane-Orcutt and Newey-West.19 The results are reported in
table 11.20 The Cochrane-Orcutt procedure produces a feasible
generalized-least squares estimate of the grants coefficient and its
standard error. With this procedure, the point estimate of the
substitution rate drops from 84 to 39 percent and is statistically
insignificant (shown in the second column of tables 10 and 11). The
Newey-West correction for autocorrelation does not involve re-estimating
the grants coefficient, so the estimated substitution rate remains at 84
percent. The coefficient continues to be statistically significant after
correcting for the bias in its standard error.

19We used two software packages to estimate the model: Limdep version 7
and Stata version

8. Limdep uses the Cochrane-Orcutt method to correct for autocorrelation,
which results in a revised generalized-least-squares estimate of the model
coefficients and their standard errors, while Stata uses the Newey-West
method, which only corrects the estimates of the standard errors. Each
method yields a different but equally valid estimate of the substitution
rate. Because the Cochrane-Orcutt method does not include the first
observation for each state, these estimates are based on observations from
1983 through 2000. In contrast, the Newey-West method includes the first
observation.

20We also adjusted for heteroscedasticity using White's methods. The
effect of this adjustment was minor and therefore we did not report it.

Appendix II Description of Grant Substitution Model, Statistical Methods,
and Results

  Table 11: Instrumental Variables Estimates of State Highway Spending Model,
    Correcting for Autocorrelation Autocorrelation corrected model estimates

             Model statistics                    R2                        R2 
            Constant term only                 0.000        
         State fixed effects only              0.481        
             X -variables only                 0.233        
           X and group effectsa                0.558                    0.783 

                                         Cochrane-Orcutt Newey-West estimates 
                                               estimates 
            Variables           Coefficients Probability         Coefficients 
                                                                  Probability 
    FHWA payments / per capita             -0.3859 0.222         -0.841 0.002 
    Personal income, real per               0.0028 0.824          0.016 0.226 
           capita, (t)                                   
    Personal income, real per               0.0177 0.157          0.020 0.211 
          capita,, (t-1)                                 
    Personal income, real per               0.0058 0.567          0.011 0.309 
          capita, (t-2)                                  
     Personal income squared,            1.083E-07 0.683    -0.0626E-07 0.827 
       real per capita, (t)                              
     Personal income squared,           -3.273E-07 0.231     -4.160E-07 0.237 
      real per capita, (t-1)                             
     Personal income squared,          -0.6321E-07 0.785    -0.9140E-07 0.716 
      real per capita, (t-2)                             
    Nonhighway federal grants,              0.1065 0.012          0.072 0.140 
       real per capita, (t)                              
    Nonhighway federal grants,              0.0312 0.595          0.079 0.214 
      real per capita, (t-1)                             
    Nonhighway federal grants,             -0.0715 0.118         -0.048 0.330 
      real per capita, (t-2)                             
    Effective nonhighway grant           -624.9567 0.004       -509.881 0.034 
         match rateb, (t)                                
    Effective nonhighway grant           -311.9012 0.176       -394.928 0.124 
        match rateb, (t-1)                               
    Effective nonhighway grant            119.4099 0.553       -102.948 0.646 
        match rateb, (t-2)                               
    Vehicle miles traveled per              2.1109 0.689         -6.486 0.268 
           capita, (t)                                   
    Vehicle miles traveled per              3.6408 0.461          7.886 0.159 
          capita, (t-1)                                  
    Vehicle miles traveled per             -0.2714 0.953          2.045 0.657 
           capita,(t-2)                                  
     Registered vehicles per                0.0180 0.640          0.044 0.319 
           capita, (t)                                   
     Registered vehicles per                0.0336 0.348          0.021 0.633 
          capita, (t-1)                                  
     Registered vehicles per               -0.0342 0.353         -0.068 0.118 
          capita, (t-2)                                  
     Drivers per capita, (t)               -0.0826 0.107         -0.129 0.036 
    Drivers per capita, (t-1)              -0.0555 0.277         -0.050 0.407 
    Drivers per capita, (t-2)               0.0381 0.470          0.078 0.206 
        Governor Democrat,                  3.4241 0.387          1.280 0.735 
    (1=Democrat; 0=otherwise)                            
% Democratic in State House             -9.4411 0.753         -4.578 0.878 
% Democratic in State Senate           -13.8934 0.613         46.860 0.073 
      Utah Olympics, = 1 for                                                  
       1997-2000 in Utah, 0               142.8233 0.000        154.824 0.000
            otherwise                                    
            Time trend                    -17.7499 0.199        -32.657 0.000 

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

                         (Continued From Previous Page)

                   Autocorrelation corrected model estimates

                                       a
                                     The R
                                       2
       with the correction for autocorrelation is not comparable to the R
                                       2
without the correction because the dependent variable is different between
             the two models due to the autocorrelation adjustment.
                                                                                                                                        Road                   Road                                        
                                                                Inverse                   Time                                          miles                 Miles                                        
 Cochrane-Orcutt Newey-West Variables Coefficients Coefficients  time   110.4277 -95.487  trend  0.5692 1.030 Population, 0.0018 0.003   per   -0.3525 1.121 squared, 0.0034 -0.001                        
       estimates  estimates            Probability  Probability  trend     0.895   0.234 squared  0.089 0.000  in 1,000    0.741 0.361 capita,   0.711 0.147   per     0.364  0.674                        
                                                                                                                                        (t-1)                capita,                Autocorrelation        
                                                                                                                                                              (t-1)                   coefficient   0.5297 

Source: GAO analysis.

bRatio of federal non-highway grants to state and local nonhighway
expenditures.

The Effect of Removing Statistically Insignificant Variables

The full model includes over 30 variables when all the lags are included
and many of these variables are statistically insignificant. To simplify
the model, we performed F-tests for the statistical significance of
variables and removed variables with a statistical significance level
below 10 percent. We tested variables that were included with 1-and 2-year
lags as a group and removed them as a group if found insignificant. We
summarize the results of these tests in table 12. The primary result is
that neither the highway usage variables nor the variables intended to
capture state differences in preferences are statistically significant.
The only variables that are systematically associated with differences in
state highway spending are the variables reflecting financial resources
that could be used to fund highways.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

Table 12: Summary Results of the Statistical Testing of the Variable
Coefficients

                                         Statistical results 
                   Variables tested              F Statistic      Probability 
               Personal income, all                   3.6169           0.0000 
                             Linear                   2.5524           0.0545 
                            Squared                   1.1269           0.3373 
             Nonhighway grants, all                   3.5888           0.0016 
                            Amounts                   2.3880           0.0677 
                              Ratio                   3.5081           0.0150 
                     Use variablesa                   0.8952           0.5447 
               Political preference                   0.3476           0.7909 
                         Time trend                   4.3000           0.0050 
                         Lane miles                   1.3333           0.2642 

Source: GAO analysis.

aVehicle miles traveled per capita, registered vehicles per capita, and
licensed drivers per capita.

The result of removing statistically insignificant variables is shown in
table

13. With the Cochrane-Orcutt method for autocorrelation correction, the
grant substitution coefficient is 0.50 and with the Newey-West correction
the coefficient is 0.58; both estimates are statistically significant at
the 1 percent significance level. Thus, the difference in estimated
substitution rates under the two methods narrowed with the simplified
model. To be conservative in our findings regarding grant substitution, we
are using the lower estimate of 0.50, based on the Cochrane-Orcutt method,
as our preferred estimate. The 95 percent confidence interval ranges from
12 to 88 percent, which includes Gamkhar's estimate of 63 percent but not
Knight's higher estimate of 91 percent. Because the Cochrane-Orcutt method
does not include the first observation for each state, these estimates are
based on observations from 1983 through 2000.

Analysis of Remaining The full model includes per capita income squared to
test for nonlinear

Explanatory Variables	effects of income on state spending. However, the
squared term is statistically insignificant. We conclude that state
spending is proportional to income, which implies that both high-and
low-income states respond to changes in income in roughly the same
proportion, once other factors affecting state spending choices are taken
into account. The lag structure on per capita income indicates that the
largest increase occurs in the first

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

year, but prior year changes in income also affect state expenditures (see
table 13).

The effect of nonhighway grants enters into the model in two ways: the
absolute size of other grant funding, measured in per capita terms,
representing the income effect of other-grant funding; and the ratio of
the nonhighway grants to state nonhighway spending, representing the tax
price effect of other-grant funding. The net income effect of other grants
is small but positive. The coefficients on the nonhighway grant variables
sum to a small positive effect with a statistically significant positive
effect in the current year and a statistically significant negative effect
in year 2. This result is contrary to expectations in that the net effect
would be expected to result in some of the funding from other federal
grants to be used as a substitute for states' own highway spending.

In contrast, the tax price effect of other grants is strongly negative
indicating that matching requirements associated with other federal
programs, such as Medicaid, result in states spending less of their own
resources on highways. For every dollar spent by a state, the federal
government reimburses the state for a percentage of the cost, reducing the
tax price of these services to the state. The lower price for other public
services raises the demand for those services and reduces the demand for
highways, suggesting that highways and other public services are
substitute goods.

We enter the time trend variable into the model in linear, quadratic, and
inverse form to provide a flexible functional form. The inverse term was
statistically insignificant and we dropped it from the model. The
coefficients on the linear and quadratic term indicate a negative trend
for most of the years in state highway spending when other factors
affecting state spending are taken into account.

Varying Substitution Rates As we noted in our summary of previous studies,
Meyers reports no

Over Time	evidence of substitution into nonhighway spending during the
1976 to 1982 time period. Gamkhar, based on data from 1976 through 1990,
reports higher rates of substitution, and Knight's study, based on data
from 1983 through 1997, reports even higher rates of substitution. We
therefore tested for evidence of increasing substitution rates using the
Cochran-Orcutt method, which, as discussed earlier, uses the estimation
period 1983 to 2000. The results are shown in table 14.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

Table 13: State Highway Spending Model with Statistically Insignificant
Variables Removed

                              Model statistics R2

                            Constant term only 0.000

                         State fixed effects only 0.426

                            X - variables only 0.078

                           X and group effects 0.479

Autocorrelation Adjustment Method

                           Cochrane-Orcut Newey-West

                                                  95 Percent                       
                             Standard Probability confidence              Standard Probability 
Variables    Coefficients   errors      values   interval Coefficients   errors      values 
 FHWA highway                                                                      
 grants, real         -0.501    0.194       0.010     -0.881       -0.580    0.174       0.001 
  per capita                                          -0.121                       
Personal                                                                        
 income, real                                                                      
per capita, (t)        0.006    0.002       0.005      0.002        0.009    0.003       0.000 
                                                       0.010                       
Personal                                                                        
 income, real                                                                      
  per capita,          0.005    0.002       0.033      0.001        0.006    0.003       0.041 
     (t-1)                                             0.009                       
Personal                                                                        
 income, real                                                                      
  per capita,          0.002    0.002       0.280     -0.002        0.001    0.002       0.652 
     (t-2)                                             0.006                       
  Nonhighway                                                                       
    federal                                                                        
 grants, real                                                                      
      per                                                                          
  capita, (t)          0.115    0.040       0.004      0.037        0.064    0.046       0.168 
                                                       0.193                       
  Nonhighway                                                                       
    federal                                                                        
 grants, real                                                                      
      per                                                                          
 capita, (t-1)         0.044    0.049       0.370     -0.052        0.037    0.058       0.522 
                                                       0.140                       
  Nonhighway                                                                       
    federal                                                                        
 grants, real                                                                      
      per                                                                          
 capita, (t-2)        -0.081    0.041       0.045     -0.161       -0.036    0.047       0.445 
                                                      -0.001                       
Effective                                                                       
  nonhighway                                                                       
  grant match       -630.218  204.465       0.002 -1030.969      -475.541 231.228        0.040 
  ratea, (t)                                      -229.467                         
Effective                                                                       
  nonhighway                                                                       
  grant match       -310.799  208.831       0.137 -720.108       -238.395 242.269        0.325 
  rate, (t-1)                                     98.510                           
Effective                                                                       
  nonhighway                                                                       
  grant match        214.391  186.720       0.251 -151.580       -172.916 212.623        0.416 
  rate, (t-2)                                     580.362                          
Utah Olympics,                                                                     
    = 1 for                                                                        
 1997-2000, 0                                                                      
otherwise         140.309   27.778       0.000     85.864      165.540   25.423       0.000 
                                                     194.754                       
  Time trend         -17.804    5.107       0.000  -27.814        -15.166    3.150       0.000 
                                                    -7.794                         
  Time trend           0.570    0.162       0.000      0.252        0.461    0.098       0.000 
    squared                                            0.888                       
Autocorrelation                                                                    
  coefficient          0.584                                                       

                             Source: GAO analysis.

Appendix II Description of Grant Substitution Model, Statistical Methods,
and Results

aRatio of federal nonhighway grants to state and local nonhighway
expenditures

To test whether the substitution rate has increased over the period of our
sample data, we divided our sample into four time estimation periods,
corresponding with the authorization periods for the federal-aid highway

21

program.

o  1983 to 1986,

o  1987 to 1990,

o  1991 to 1997, and

o  1998 to 2000.

Allowing the substitution rate to vary over time improves the explanatory
power of the model, increasing the R2 of our preferred model from 48
percent to 57 percent. The first period from 1983 to 1986 shows a
substitution rate of 18 percent that is not significantly different from
zero. The estimated substitution rate increases to 36 percent in the 1987
to 1990 period and is significant at the 10 percent level. The
substitution rate rises to just under 60 percent during the two periods of
the 1990s and is statistically significant at the 1 percent level. As
shown in Table 14, these results are roughly consistent with previous
studies that, when taken together, seem to suggest increasing matching
rates over time.

21As stated earlier, grant expenditures represent, in part, obligation
authority provided in prior years.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

 Table 14: State Highway Spending Model with Substitution Rates by Time Period
                                    Model R2

                            Constant term only 0.000

                         State fixed effects only 0.494

                            X - variables only 0.145

                           X and group effects 0.565

                                          Standard Probability     95 Percent 
                                                                   confidence 
           Variables         Coefficients   errors      values       interval 
     FHWA Grant, real per          -0.178    0.198       0.370   -0.566 0.211 
     capita for 1983-1986.                                     
     FHWA Grant, real per          -0.360    0.195       0.065   -0.742 0.022 
     capita for 1987-1990.                                     
     FHWA Grant, real per          -0.592    0.190       0.002   -0.965 -0.22 
     capita for 1991-1997.                                     
     FHWA Grant, real per          -0.581    0.188       0.002   -0.95 -0.213 
     capita for 1998-2000.                                     
Personal income, real per        0.007    0.002       0.004    0.002 0.011 
          capita, (t)                                          
Personal income, real per        0.000    0.002       0.857   -0.004 0.005 
         capita, (t-1)                                         
Personal income, real per        0.004    0.002       0.067        0 0.007 
         capita, (t-2)                                         
    Non-Hwy federal grants,         0.147    0.041       0.000    0.068 0.227 
     real per capita, (t)                                      
    Non-Hwy federal grants          0.046    0.049       0.346    -0.05 0.142 
    real per capita, (t-1)                                     
    Non-Hwy federal grants         -0.084    0.041       0.039  -0.163 -0.004 
    real per capita, (t-2)                                     
       Effective federal                                            -1156.819 
    nonhighway-grant match       -756.328 204.332        0.000       -355.837 
          ratea, (t)                                           

Effective federal nonhighway-grant -243.191 208.038 0.242 -650.945 164.563 
           match ratea, (t-1)                                         
Effective federal nonhighway-grant  241.691 188.098 0.199 -126.981 610.363 
           match ratea, (t-2)                                         
Utah Olympics, = 1 for 1997-2000,   144.022 26.282  0.000    92.51 195.534 
              0 otherwise                                             
               Time Trend               -5.423  5.193  0.296 -15.602    4.756 
           Time Trend Squared            0.187  0.168  0.265   -0.142   0.516 
      Autocorrelation coefficient      0.51599                        

Source: GAO analysis.

aRatio of federal nonhighway grants to state and local nonhighway
expenditures.

When the substitution rate is allowed to vary over time, the time trend
coefficients become statistically insignificant. This lack of significance
suggests that there is no negative time trend in state spending once the
increasing substitution rate associated with different time periods is
taken into account.

We use an IV estimator because we assume federal grants and state spending
are jointly determined. To test the reliability and validity of the IV
estimator we ran three additional statistical tests: (1) a weak
instruments

Appendix II Description of Grant Substitution Model, Statistical Methods,
and Results

test, (2) a test for exogeneity of excluded exogenous instruments, and (3)
a test for endogeneity of federal grants.

The weak instruments test is intended to verify that the excluded
exogenous instrumental variables included in the grants equation are
correlated with federal grants. If they are not, the IV estimator provides
no advantage to a simple (and more efficient) OLS estimator. To test the
significance of the excluded exogenous variables, we calculated the
partial R2 associated with the excluded exogenous instruments and found
the instruments to be statistically significant at the 1 percent level.

To test for the exogeneity of excluded exogenous instruments, we conducted
a Hausman over-identifying restrictions test.22 This test compares the
estimated federal grant coefficients for each time period using the full
set of excluded exogenous variables with coefficients derived from using a
subset of instruments composed of predetermined variables that can safely
be assumed to be exogenous. A finding that the set of grant coefficients
from the two models are not statistically different from one another lends
support for the hypothesis that the full set of excluded exogenous
instruments are independent of the error term in the second stage
expenditure equation. For this test, we used a subset of excluded
exogenous variables. Differences between the grant coefficients for each
time period using all instruments, and the coefficients using the subset
of exogenous instruments, were not statistically significant and are
quantitatively very similar to one another. Thus, we found no evidence
that our excluded exogenous instruments were correlated with the error
term of the expenditure equation.

Finally, we conducted a Hausman test for the endogeneity of the federal
grant variable.23 This test consists of comparing the IV estimate of the
grant coefficient for each time period with the corresponding grant
coefficient based on the OLS estimate. If the differences were not
statistically significant there would be little justification for using
the IV estimator. This test yielded statistically significant differences
between the two sets of estimates, lending support for the assumption that
federal grants and state

22Jeffrey M. Woolridge, Econometric analysis of Cross-section and Panel
Data, The MIT Press, Cambridge, Massachusetts, 2002.

23William H. Greene, Econometric Analysis, 5th Edition, Prentice Hall, New
Jersey, 2003, pp. 80-81.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

spending are jointly determined. The results of each of the three tests
are summarized in table 15.

Table 15: Statistical Tests for the Endogeneity of Federal Grants and
State Highway Spending

 Weak instrument test Partial R2 Probability value 0.10 0.000 Grant coefficient

Hausman over
identifying
restrictions test All instruments

Subset of instruments Probability value

                   '83-`86            -0.178      -0.187                0.531 
                   '87-`90            -0.360      -0.366      
                   '91-`97            -0.592      -0.590      
                   '98-`00            -0.581       .581       

                               Grant coefficient

                       Hausman                         
              endogeneity test    2SLS             OLS      Probability value 
                       '83-`86   -0.178         -0.047                  .0001 
                       '87-`90   -0.360         -0.160 
                       '91-`97   -0.592         -0.340 
                       '98-`00   -0.582         -0.343 

                             Source: GAO analysis.

Additional Tests for Varying Substitution Rates

We also estimated alternative models that allow the substitution rate to
vary according to state size (measured by population), per capita income,
and state per capita spending on mass transit to test for a varying
substitution rate related to these factors. The results of these models
were negative. Overall we found no evidence that substitution rates
systematically differ by either population size or the level of mass
transit spending. We did obtain higher estimates of substitution rates in
states with higher per capita income (56-66 percent in high income states
compared to just over 30 percent in lower income states), but these
differences were not statistically different from the average substitution
rate of 50 percent found for the period from 1983 to 2000.

 Appendix II Description of Grant Substitution Model, Statistical Methods, and
                                    Results

State Characteristics Associated with Fixed Effects

In the models reported above, state fixed effects account for most of the
variation in state highway spending. Based on our preferred model (the
model in table 12 using the Cochrane-Orcutt autocorrelation correction
method), differences in state spending associated with these fixed effects
can be as much as $400 per capita. However, these fixed effects are
difficult to interpret since they represent all factors that are
systematically related to cross-sectional differences in state spending
not included in the model (e.g., geography, weather, and other variables
that have substantial crosssectional variation).

To determine if the differences in state spending measured by the fixed
effects of our model are systematically associated with particular state
characteristics, we performed a step-wise regression using the fixed
effects from our preferred model as the dependent variable. Of the 12
variables we considered, 3 are statistically significant: per capita
highway lane miles, per capita income, and heating degree days (see table
16). In the first step, lane miles account for 51 percent of the
cross-state variation in our fixed effects, the second step equation added
per capita income, and increases the explained variation to 68 percent.
The third step equation adds heating degree-days, raising the variation
explained to 77 percent. The remaining variables are statistically
insignificant and provide little additional explanatory power.

Table 16: Stepwise Regression Analysis of the Fixed Effects

                                                            Probability  
          Step                       Variables Coefficient     value       R2 
             1   Average lane miles per capita      0.58322    0.002      51% 
             2  Average real income per capita     -0.01778    0.000      68% 
             3     Average heating degree days      0.01324    0.004      77% 

Source: GAO analysis.

The positive coefficient on lane miles per capita may reflect a higher per
capita cost of maintaining a larger highway network. The negative
coefficient on per capita income suggests that, other things being equal,
states with high average real incomes per capita spend less for highways
than states with low average real incomes per capita. Finally, the
positive coefficient on heating-degree days indicates that states in
colder climates spend more on highways than states in warmer climates, all
other things remaining equal.

Appendix III

State Characteristics Associated with States' Level of Effort to Fund
Highways from State Resources

Based on our model of state highway spending, we found a number of factors
that are systematically related to state highway spending and, in turn, a
state's level of effort to fund highway from state resources.1 Perhaps
most importantly, more federal highway aid is associated with less state
effort to fund highways from state resources once other factors related to
state spending are taken into account. Our conservative estimate of grant
substitution suggests that about half the increase in federal highway
grants is used to reduce states' level of highway spending effort.

Increases in federal grant funding for nonhighway purposes, such as
health, education, and welfare, are also associated with reduced effort on
the part of states to fund highways. Based on our model of state highway
spending, we found that states with a higher percentage of their
nonhighway spending funded by federal grants reduced their effort to fund
highways, presumably, to provide matching funds for programs like
Medicaid, which is an open-ended matching program.

In addition to federal grants, we found two cost factors that are
systematically related to states' levels of highway spending effort, other
things being equal. States with large highway networks, as measured by the
number of highway lane miles, systematically spend more per capita.
Presumably, a larger road network is more expensive to maintain and states
must therefore devote a larger share of their funding capacity to
maintaining their highway network. In addition, we found that colder than
average temperatures, as measured by heating degree days, are associated
with higher state spending, suggesting that colder weather creates more
wear and tear on the highways and hence the need for states to make a
greater spending effort to maintain their highway network, other things
being equal.

Finally, we found that high per capita income states make less effort than
states with lower incomes. This result is, perhaps, not surprising since
the same effective tax rate, (level of effort), generates more revenues in
high

1Since level of effort is defined as state spending relative to funding
capacity, factors that are shown to be directly related to state spending
are also directly related to a state's level of effort. In an earlier
report, GAO, Trends in Federal and State Capital Investment in Highways,
(GAO-03-744R, June 18, 2003), we reported considerable variation among
states in the level of highway funding effort that persisted over the
entire period of our study from 1982 through 2000. In addition, we
reported wide movement in the states' relative levels of effort over time.
That is, some states making a low level of effort in the 1980s ranked
above average in the 1990s and vice versa.

Appendix III
State Characteristics Associated with States'
Level of Effort to Fund Highways from State
Resources

income states than in states with lower incomes. Thus, the same level of
highway spending can be funded with less effort in high-income states and
low-income states compensate by undertaking a greater effort to fund
highways from state resources.2

2The relationship between income and level of effort is complex. We found
that low-income states make a greater level of effort to fund highway from
state resources compared to higher income states. At the same time, there
was no evidence of a difference in the spending response of high- and
low-income states to a change in income, that is, the squared term from
the substitution model was not statistically significant. In our model of
state highway spending, we found that the fixed effect coefficients were
negatively related to state per capita income, indicating that high-income
states make less effort than states with lower income. In addition to the
fixed effects, we also included per capita income and per capita income
squared in our model to test whether the spending response to changes in
income differed between high-and low-income states. As reported in
appendix II we found no evidence of a differential response.

Appendix IV

Program Options Designed to Reduce Substitution

One program option that could be designed to reduce substitution would be
to modify the matching requirement to leverage additional state highway
spending. While the use of matching requirements as an economic tool is
designed to leverage additional spending, the federal-aid highway
program's current matching requirements, which typically call for 20
percent state funding and 80 percent federal funding of eligible projects,
permit substitution because most states' highway funding is already higher
than 20 percent of their total highway funds. The matching requirement,
therefore, does not provide states with an incentive to increase or even
maintain their level of funding in order to receive additional federal
funds. Instead, states are free to substitute federal funds for funds they
would have spent from their own resources and to use their own funds in
other ways.

For the matching requirement to leverage additional state spending, the
states' matching portion would have to be set high enough so that states
would not receive additional federal funds without spending beyond what
they would have otherwise spent without additional federal assistance.
This objective cannot be perfectly achieved because models of
substitution, like any models, produce estimates that are subject to
uncertainty, and there is no way to objectively determine with certainty
what states would have spent in the absence of increased federal funding.
However, the likelihood that increased federal funding will leverage
additional state highway spending can be achieved in several ways.

Increase State Matching Requirements

The most direct approach would be to change the current 80 percent
federal/20 percent state match ratio to a matching ratio closer to the 45
percent federal/55 percent state division of funding in fiscal year 2002.
This would likely mean that some states (those whose spending is less than
60 percent of combined federal and state spending) would be required to
increase their highway spending in order to qualify for any increased
federal funding, while other states whose spending is already over 60
percent of combined federal and state spending would not have to increase
or maintain their spending in order to receive increased federal funds.

Increasing the required state match from 20 percent to 60 percent might
require a few states, whose state highway funding levels are currently a
comparatively small proportion of their total highway spending, to more
than double their current level of highway spending to avoid losing
federal funds. If increases of this magnitude were deemed too extreme, a
more moderate increase in the state match could be established. For
example,

          Appendix IV Program Options Designed to Reduce Substitution

raising the state matching share to 40 percent instead of 60 percent would
require smaller funding increases in states whose state and local spending
is currently a smaller proportion of the total highway spending, but it
would also reduce the number of states that would be required to increase
their level of funding in response to increased federal funding.

Another drawback of simply increasing state matching requirements is that
even substantial increases in the requirements (raising the required state
match from 20 percent to 60 percent) would not be likely to leverage
additional state spending in all states. An alternative that would
increase the likelihood of leveraging additional state spending in all
states would be to continue with the 80 percent federal/20 percent state
matching ratio but stipulate that only state spending in excess of what
the state had spent for highways in an appropriate base time period be
counted against its federal matching requirement. This approach has the
advantage of maintaining the current 20 percent state matching rate, yet
provides a leveraging incentive in all states rather than in only those
states with below average spending. However, it might have the effect of
making it easier for those states that were not spending much in the base
time period to increase their spending and receive increased federal funds
than it would be for those states whose spending was already high.

Modify Funding Formulas to Reward State Highway Funding Effort

Another approach that would reduce substitution by creating an incentive
for states to increase their own highway spending would be to directly
link the level of federal highway aid to each state's level of highway
funding effort. This link could be achieved by setting aside a fixed
percentage of formula grant funding to be distributed in accordance with
states' highway funding efforts. As stated in the text, to avoid
penalizing low income states, each state's highway funding effort could be
defined as the state's highway spending compared to some measure of the
state's taxing capacity. There are a variety of indicators that could
serve as a measure of states' funding capacity. The most comprehensive
that is available annually is Total Taxable Resources (TTR), which is
produced annually by the Department of the Treasury and used to distribute
substance and mental health block grants. Less comprehensive measures
would include Gross State Product (GSP) and Personal Income (PI), both
published annually by the Department of Commerce.

This approach could be implemented in a variety of ways. One approach
would be to compare each state's funding effort to the average effort of
all states. If, for example, $100 per capita were set aside and
distributed in this

Appendix IV Program Options Designed to Reduce Substitution

way, states whose highway spending efforts were above the average spending
effort would receive funding proportionally above the $100 per capita
average and states whose effort was below the average spending effort
would receive funding proportionally below the $100 per capita average.

Initially, those states with an above-average highway funding effort would
be rewarded with higher per capita funding, and those states with a
belowaverage highway funding effort would be penalized with lower per
capita funding. In following years, each state's highway spending effort
would continue to be compared to the average state highway spending
effort, so that states whose funding effort rose relative to the national
average would automatically be rewarded with higher per capita funding,
while states whose effort fell relative to the national average would
automatically be penalized. Distributing the set aside in this fashion
would, in effect, put all states in competition with one another,
automatically rewarding states whose effort rose compared to the national
average and penalizing states whose effort fell compared to the national
average.

The approach just described would reward those states whose funding effort
is currently high and penalize those whose effort is currently low.
However, this approach could be modified to avoid rewarding or penalizing
states based on their current level of effort. Instead, the linking of
federal funds to state effort could be based only on future changes in
each state's level of highway funding effort. In this approach, each
state's highway funding effort would be compared to its own effort during
an initial time period, such as the year (or an appropriate average of
years) prior to initiation of the set aside. For example, all states could
be awarded the same per capita grant amount in the first year of the
set-aside program. Then, in future years, each state's funding effort
would be compared to its own funding effort in the first year of the set
aside program and adjusted accordingly. Each state whose funding effort
increased compared to the initial base year would receive an increase in
federal funding proportionate to the increase in its own spending. Such an
approach would, in effect, put

          Appendix IV Program Options Designed to Reduce Substitution

each state in competition with the effort it made in the base period.1 If
both approaches to rewarding state highway funding effort were deemed
desirable, a combination of the two approaches could be employed. The
strength of the incentive would depend on the amount of total formula
funding distributed through the set aside program; the greater the amount
of funding distributed in this manner, the larger the financial
consequences to states of changing their level of highway funding effort.

Introduce a State Maintenance of Effort Requirement

If, instead of seeking to stimulate additional state spending on highways,
the goal of federal policy makers is for federal grants to supplement
state spending on highways, then instituting a maintenance-of-effort (MOE)
provision may be a more appropriate approach. MOE provisions require
states to maintain existing levels of state spending on an aided program
as a condition of receiving federal funds. As a tool, MOE requirements are
designed not to stimulate additional state spending but to guard against
grant substitution so that increased federal spending will supplement
rather than replace states' own spending.

As with matching requirements, this objective cannot be perfectly achieved
because models of substitution, like any models, produce estimates that
are subject to uncertainty and there is no way to objectively determine
with certainty what states would have spent in the absence of increased
federal funding. However, the likelihood that increased federal funding
will not be used as a substitute for state spending can be strengthened if
MOE requirements are designed appropriately. In previous work, we
concluded that, to be effective, MOE provisions should define a minimum
level of state spending effort that can be objectively quantified based on
reasonably current expenditures on the aided activity. 2 Adjusting the MOE
requirement for inflation in program costs would ensure the minimum
spending level is maintained when measured in inflation adjusted dollars.

1To work in the way described here, funding for the set-aside program, as
described in the text, would have to be adjusted annually to reflect the
overall change in states' funding effort. If states' overall level of
state effort remained unchanged from year-to-year, funding for the
set-aside program would not have to change in order to ensure that states
whose effort increased were rewarded and states whose effort declined were
penalized. If, however, the overall average state effort increased,
funding for the set-aside program would have to increase proportionally to
ensure that all states with increased effort were rewarded and those with
declining effort were penalized.

2GAO, Proposed Changes in Federal Matching and Maintenance of Effort
Requirements for State and Local Governments, GAO/GGD-81-7, December 23,
1980.

Appendix IV Program Options Designed to Reduce Substitution

This could be achieved by defining a state's base spending level as the
amount spent per year during a recent historical period and then adjusting
that base spending level for inflation.

One drawback of an MOE provision is that basing it on historical spending
period could result in a base spending period for the MOE provision that
represents an unusually high spending level for some states, effectively
locking them into continued high spending in future years. This could be
ameliorated however by establishing waivers for states that are able to
demonstrate that spending in the base period chosen is unusually high, to
allow a more "typical" spending level for purposes of the MOE provision.

Developing an indicator of state highway spending effort to link federal
funding to state spending or establishing a state's base spending level to
design an MOE requirement would require careful consideration. Among other
issues, in defining these indicators, consideration would have to be given
to whether to measure:

o 	Capital expenditures for highways or capital plus maintenance spending;

o  Expenditures on all state roads or for federal-aid roads only;

o 	State government expenditures only, or spending by state and local
governments;

o 	Total expenditures, or expenditures normalized on a per capita, per
lane mile, or other basis.

In addition, an indicator of state funding effort or a state's base
funding level for an MOE provision should, to the extent possible, be
established by measuring spending levels that are typical rather than
unusually high or low. Highway capital expenditures in a state can
increase or decrease dramatically from year to year and may be unusually
high or low for variety of reasons (e.g., Utah's unusually high spending
during preparations for the 2002 Winter Olympics, or a state particularly
hard hit by recession that drops spending below its usual effort). To some
extent, such factors can be taken into account by defining a state's
funding effort or base level of spending for an MOE provision using
multi-year averages so that such unique circumstances are averaged out.

Appendix V

                     GAO Contacts and Staff Acknowledgments

GAO Contacts	JayEtta Z. Hecker (202) 512-2834 Steve Cohen (202) 512-4864
Jerry Fastrup (202) 512-7211

Staff Acknowledgments

(544074)

In addition to those named above, Jay Cherlow, Catherine Colwell, Gregory
Dybalski, Edda Emmanuelli-Perez, Scott Farrow, Donald Kittler, Alex
Lawrence, Sara Ann Moessbauer, Robert Parker, Paul Posner, Teresa Renner,
Stacey Thompson, and Alwynne Wilbur made key contributions to this report.

GAO's Mission	The Government Accountability 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.

Obtaining Copies of The fastest and easiest way to obtain copies of GAO
documents at no cost

is through GAO's Web site (www.gao.gov). Each weekday, GAO postsGAO
Reports and newly released reports, testimony, and correspondence on its
Web site. To Testimony have GAO e-mail you a list of newly posted products
every afternoon, go to

www.gao.gov and select "Subscribe to Updates."

Order by Mail or Phone	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. Government Accountability 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

To Report Fraud, Contact:
Waste, and Abuse in Web site: www.gao.gov/fraudnet/fraudnet.htm

E-mail: [email protected] Programs Automated answering system: (800)
424-5454 or (202) 512-7470

Congressional	Gloria Jarmon, Managing Director, [email protected] (202)
512-4400 U.S. Government Accountability Office, 441 G Street NW, Room 7125

Relations Washington, D.C. 20548

Public Affairs	Jeff Nelligan, Managing Director, [email protected] (202)
512-4800 U.S. Government Accountability Office, 441 G Street NW, Room 7149
Washington, D.C. 20548

                               Presorted Standard
                              Postage & Fees Paid
                                      GAO
                                Permit No. GI00

United States
Government Accountability Office
Washington, D.C. 20548-0001

Official Business
Penalty for Private Use $300

Address Service Requested
*** End of document. ***