Full text of Working Paper (Federal Reserve Bank of Cleveland) : Dedicated Taxes and Rent Capture by Public Employees, Working Paper 89-04

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Working Paper 8904
DEDICATED TAXES AND RENT CAPTURE
BY PUBLIC EMPLOYEES

by Brian A. Cromwell

Brian A. Cromwell is an economist at the Federal
Reserve Bank of Cleveland. The author thanks Paul
Bauer, Erica Groshen, Randall Eberts, William
Wheaton, and James Poterba for useful discussion
and suggestions. Seminar participants at the
Federal Reserve Bank of Cleveland and Case Western
Reserve University provided useful comments.
Invaluable help with the data was received from
Ralph Day and from Dottie Nichols and William Lyons
of the Transportation Systems Center. Financial
support from the National Graduate Fellowship Program
and the MIT Center for Transportation Studies is gratefully
acknowledged.
Working papers of the Federal Reserve Bank of
Cleveland are preliminary materials circulated to
stimulate discussion and critical comment. The views
stated herein are those of the author and not necessarily
those of the Federal Reserve Bank of Cleveland or of
the Board of Governors of the Federal Reserve System.

April 1989

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

1
Introduction
This paper tests for evidence of rent capture by public-sector
employees when a dedicated tax for a local public service is enacted. The use
of special districts for the provision of specific public services has grown
increasingly common in recent years. These districts provide services such as
water, sewer, electric, or transit and typically generate some revenue in
exchange for provision. When user fees do not cover costs, however, public
funding from federal, state, and local governments makes up the difference.
This research reports that alternative funding mechanisms and budgeting
practices appear to provide different opportunities for public-sector unions
to obtain higher wage rates and payrolls.

Two distinct types of local public funding are considered: general
revenues and dedicated taxes. Districts receiving funding from general
revenues typically compete with other districts, departments, or agencies
through a budget process for a limited pool of funds raised by traditional
taxation methods. In contrast, agencies that receive earmarked revenues from
a dedicated tax are assured of funding without the need to justify their
budget or level of senrice. Their funding falls outside the traditional local
budget process, which is characterized by the comparison of costs and benefits
of competing uses of public funds. This lack of competition potentially
1
yields increased political and budgetary autonomy.

Proponents of earmarked revenues hold that this type of funding
permits long-term planning and more efficient operation.

Managers who are not

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

2
concerned with competing for funds on an annual basis can take a longer-term
view with respect to capital budgets, planned expansions, and other
operations.

Nevertheless, the permanent nature of the funding and the lack

of checks and reviews through a normal budget process also creates
opportunities for rent capture by public-sector employees and unions through
higher wages, fringe benefits, and staffing levels. These higher costs
potentially offset any efficiency gains from earmarking and argue for the use
of traditional budget and financing methods.

This paper tests whether the enactment of a dedicated tax leads to
higher payroll and wages for public employees. Data from-the Section 15
reporting system administered by the Urban Mass Transportation Administration
(UMTA) provide payroll and wage levels for a homogeneous type of employee--bus
operators--for 165 public transit systems over the period 1982-1985. By 1985,
64 of these systems used local dedicated revenue sources to support their
operations. I independently collected information on when these taxes were
enacted and the historical circumstances leading to their enactment. Eight
were enacted during the sample period, while the other 56 were enacted during
the 15 years prior to 1982.

In the sample,'average hourly wage rates for public-sector
operators rise from a pre-tax level of $7.97 to more than $10.00 in the
two-year period following a tax enactment, and subsequently remain well above
the average of $8.59 per hour for unionized systems with no dedicated taxes.
In further analysis, pooled time-series, cross-section regressions provide a
systematic look at the level and time path of payroll and wages following a

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

3

tax enactment. The level of dedicated revenues and the time since tax
enactment are used as independent explanatory variables, with further controls
made for local private-sector wage rates, system size, unionization, and
demographic characteristics.

A potentially serious objection to the analysis is that the

existence of a dedicated tax may be endogenous to the local wage process. The
history of these taxes in the local mass-transit industry, however, is
discussed in detail and strongly suggests that the principal determinant for
the existence of dedicated taxes for transit are state-level policies for
transit funding. A more plausible channel for bias is the presence of
unobserved fixed effects that are correlated with both public-sector wages and
dedicated taxes, such as union strength, the voting power of public employees,
tastes for good or bad government, or the local political' environment. The
empirical analysis thus explicitly tests and controls for fixed effects.

The econometric analysis uses two procedures. The first is a
generalized least squares (GLS) procedure that controls for cross-section
heteroscedasticity and correlation of errors across time. The GLS results
suggest that there are significant and permanent increases of 8 to 10 percent
in payroll and wages following the enactment of a tax. These results,
however, are potentially biased due to unobserved fixed effects that are
correlated with both wages and the presence of a dedicated tax. To control
for fixed effects, I use the standard "fixed-effects" or "within-groups"
estimator, differencing the cross-section observations from their individual
means and running ordinary least squares (OLS) on the transformed data.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Hausman specification tests strongly reject the null hypothesis of no
correlated fixed effects. The fixed-effect results, however, suggest an even
larger impact on payrolls and wages than do the GLS results. Payroll and
wages rise a statistically significant 20 percent in the six years following
the tax enactment, and this increase remains stable over time.

Simulations of the extent to which the payroll increase represents
the capture of additional revenues resulting from the dedicated tax suggest
that immediately following the enactment of a dedicated tax, transit systems
expand significantly in size, revenues, and payroll. But while the higher
size and payroll are stable over time, the additional revenues gradually
diminish to the point that no significant difference exists after lfc years.
Holding system size constant, the results show that total revenues are
significantly higher in the five-year period following the tax enactment, but
decrease steadily from an initial upward shift in the first year. Payroll,
however, rises 20 percent in the five years following tax enactment, then
drifts upward to a 30-percent higher level by year 10. Thus, the payroll
share of revenues initially drops with the surge of new revenues. But as
payroll increases, the payroll share then rises from 27 percent at year one to
almost 40 percent by year 15 of the tax. In sum, the results suggest that.
increases in labor costs eventually absorb all additional net revenues that
result from enactment of a dedicated tax, and that these gains are maintained
in spite of a gradual falloff in revenues over time.

The paper is organized as follows. Section I reviews previous work
on ability-to-pay and rent capture and discusses how a dedicated funding

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

source increases the potential for rent capture. Section I1 provides evidence
on trends in the use of special districts and dedicated taxes nationwide.
Section I11 discusses the data, including survey results on the use of
dedicated taxes in the local mass-transit industry. Section IV presents the
econometric evidence on changes in payroll and wage rates following the
enactment of a dedicated tax. Section V provides additional econometric and
simulation results on changes in size, revenues, and payroll share following
tax enactment. Finally, section VI summarizes the conclusions and suggests
areas for future research.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

6

I. Abilitv-to-Pav and Rent Ca~ture

Abilitv-to-Pay
The labor and industrial relations literature includes numerous
studies on the effect of "ability-to-pay" on wage levels.

Early studies

hypothesized that, other things being equal, the managers of firms in lesscompetitivefiighly profitable industries paid higher wages in order to lower
labor turnover rates, enhance their public image (Weiss, 1966), improve worker
morale (Slichter, 1950), or assure a queue of available workers who can be
hired to meet increased product demand (Ross and Wachter, 1973).

Such high

wages in less-competitive industries, however, can also be interpreted as a
capture of economic rents by labor. Furthermore, workers' gains from
unionization are potentially higher in less-competitive industries because of
the presence of rents to be captured and because of lower threat of entry by
nonunion firms.

Empirical studies of labor's ability to share in any excess return
due to product-market power employ cross-industry comparisons and have
produced mixed results.

Early analyses using industry concentration as a

measure of market power include Rapping (1967), Masters (1969), Haworth and
Rasmussen (1971), and Ashenfelter and Johnson (1972).

These studies find no

statistically significant relationship between market concentration and wages.
In contrast, Pugel (1980) uses profits as a measure of concentration and finds
that labor receives 7 to 14 percent of the total excess return. More recent
studies by Clark (1984), Ruback and Zimmerman (1984), and Salinger (1984) also
find significant evidence of union rent-sharing using cross-section data on

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

profits of firms and lines of business. Finally, several studies suggest that
union workers capture economic rents created by industrywide regulations.3
Rose (1987), for example, finds significant declines in union wages resulting
from deregulation in the trucking industry. In sum, a substantial body of
empirical evidence suggests that private-sector unions are able to capture
economic rents created through monopoly power and regulation.

Public-Sector Rent Capture
The analysis of rent capture also has a strong tradition in the
public-choice literature. Niskanen (1971, 1975) posits that a bureaucracy
maximizes the level of service it provides (and hence the size of its budget)
subject to its production constraints and to the resources allocated by its
political superiors.

Since an agency negotiates with political leaders over a

total budget as opposed to incremental units of service, and since the agency
is often the sole provider of the service, it can use its monopoly power to
establish a level of service greater than that desired by voters. While the
service-maximizing model implies that bureaucrats minimize production costs
per unit of service, models by Migue and Belanger (1974) and Orzechowski
(1977) explicitly recognize that bureaucrats desire higher wages, fringe
benefits, and staff levels and may use their monopoly powers to obtain them.

Public-sector unions may share in bureaucratic rents in the same
way that private-sector unions share economic rents. Empirical studies,
however, generally show that public-sector unions have a more moderate effect
on wages than do their private-sector counterparts. Freeman's (1986)
literature review suggests a public-union wage premium on the order of 5

1

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

8
percent. Gyourko and Tracy (1988) control for endogenous choice of both
government and union status and find a 4 percent public-union wage
differential, versus 14 percent in the private sector. Although the ability
of public unions to raise wages appears limited, Gyourko and Tracy also find
(in a forthcoming article) lower property values in cities that pay their
public-sector workers relatively high wages. This suggests that public
workers' success in raising wages can be interpreted as rent capture from a
community.

Earmarking and Rent Ca~ture
The size of the overall governmental budget in Niskanen-type models
is larger than socially optimal because budgetary procedures allow bureaus to
act as price-discriminating revenue maximizers. However, the ability to use
this market power is constrained by competition from other bureaus and by the
preferences of legislative committees. Other constraints on local government
spending, pointed out by Courant, et al. (1979), include voters' direct
referenda on tax collections, and potential mobility to jurisdictions offering
alternative expenditure-taxation packages.

The earmarking of revenues directly affects most of these
constraints on public employee market power. First, the lack of alternative
uses for earmarked revenues weakens the negotiating position of a political
authority in relation to the designated receiving agency. Second, many
dedicated taxes are virtually permanent because periodic voter reapproval is

4

often not required and because the costs of repeal are high.

Permanent,

exclusive access to funding should lower the variance of funding and therefore

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

raise the expected present value of dedicated tax revenues as compared to
general revenues.

Finally, since most dedicated taxes accrue to regionwide

special districts, voting with one's feet entails a relatively high-cost move:
that is, to another metropolitan area, not just to a neighboring municipality.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

10
11.

Trends in S~ecialDistricts and Dedicated Taxes

Special districts are the fastest-growing form of local government
in the United States and represent an increasingly important method of
providing specific local public services, such as water, sewers, airports,
parks, and mass transit.

As shown in table 1, the number of special districts

nationwide doubled from 14,405 in 1957 to 28,719 in 1982. They now account for
35 percent of all governmental units--including states, counties,
municipalities, and school districts. Revenues in real terms rose nearly 500
percent in the same period, totaling $31 billion in 1982. Although fees
collected from users comprise the largest share of revenues (64 percent in
1982), local dedicated taxes are historically the largest component of state
and local assistance and are surpassed only by federal assistance in the late

While the share of revenues accounted for by local dedicated taxes
nationwide is not large overall, those districts that have special tax
authority depend heavily on it for revenue. As shown in table 2, in 1977--the
last year in which dedicated tax authority status was reported--districts with
tax authority were similar in budget size to districts without tax authority,
but received 25 percent of their total revenues from dedicated sources. While
this was down from the 33 and 30 percent shares in the 1967 and 1972 census
years, respectively, it represents the dominant source of public assistance in
these districts. There is also general evidence of rent capture in districts

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

11
with tax authority. Wages and salaries accounted for 30.9 percent of revenues
in these districts as opposed to a 25.9 percent share in districts without tax
authority.

While tax authority status by district was not reported in 1982, a
rank ordering of special districts by states in terms of the statewide use of
local dedicated taxes reveals a similar pattern. As shown in table 3, for
special districts in states with the highest use of these taxes, quintile 5,
dedicated revenues accounted for 25.1 percent of revenues, versus 0.3 percent
for the lowest quintile. Wages and salaries represent a much higher portion
of revenues in states where the use of dedicated taxes is widespread,
accounting for 30.2 percent in the highest quintile versus an average of 20.5
percent in the lowest three quintiles, again suggesting that the use of these
taxes presents opportunities for rent capture.

In summary, special districts are an increasingly common form of
government, and dedicated taxes are the largest source of public funding for
these districts from the state and local sector. In states in which these
taxes are predominantly used, revenues from dedicated taxes exceed federal
assistance as well. Wages and salaries in districts with dedicated tax
authority on average constitute a higher share of revenues compared to
districts without tax authority. This evidence of rent capture is seen in
government censuses from 1967 through 1982.

There are several explanations other than rent capture, however,
for the differences in wage shares. The heterogeneous output mix across

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

special districts could produce the observed correlation if districts using
dedicated taxes tend to produce output that requires labor-intensive
technology. Or, districts with dedicated taxes may employ higher-skilled and
higher-quality workers. Finally, districts that use dedicated taxes could be
in higher-wage areas. To account for these factors, the analysis will control
for prevailing wage rates and examine wages in districts producing a
homogeneous output and employing homogeneous employees--bus drivers in local
mass-transit systems.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

111. Dedicated Taxes and Mass Transit

The local mass-transit industry is the focus of the empirical
analysis for several reasons. First, the production processes involved are
relatively homogeneous and their inputs (labor hours and vehicle miles) are
measurable, facilitating comparisons of cost efficiency across transit
providers. Second, the employees of these systems are also relatively
homogeneous. In particular, it is assumed that the human capital of bus
drivers is similar across transit systems. Finally, transit districts receive
revenues from a wide variety of sources: fares, federal operating assistance,
state and federal capital grants, local general revenues, and local dedicated
taxes. This heterogeneity permits control for variations in operating
conditions and measurement of the impact of revenue sources and institutional
settings on wage rates. In particular, I look for evidence of wage changes
that are systematically related to the enactment of a local dedicated tax.

Data
The data source for this work is the Section 15 Reporting System
administered by UMTA.

Section 15 of the Urban Mass Transportation Act (UMT

Act) establishes a uniform accounting system for public mass-transportation
finances and operations. All applicants and direct beneficiaries of federal
assistance under Section 9 of the UMT Act are subject to this system and are
required to file annual reports with UMTA.

Section 15 data for fiscal year (FY) 1979 through FY 1985 are
available for some 435 transit systems and include detailed information on

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

14
revenue sources, expenses, employees, and hours and miles of service provided.
These data provide an extremely detailed view of a cross-section of local
government entities that perform similar activities. The revenue data are
broken down into revenues from both transit operations and public subsidies,
including information on federal, state, and local contributions for
operations and capital procurement. State and local revenues are broken down
into those from dedicated taxes versus general revenues.

The expense data include information on payroll, fringe benefits,
materials, and services for the areas of administration, operations, and
maintenance. Data on labor hours for types of employees are provided as well.
Using the expense and employee data, average hourly pay rates can be
constructed for the different types of employees. Operating statistics
include data on passengers, vehicle miles, and vehicle hours. The detailed
data on employee hours, payroll, fringe benefits, and local revenue sources
are of particular interest for this work.

Payroll and employee hour data were obtained for a homogeneous
type of employee--bus operators--for 165 public transit systems from
1982-1985. The sample was limited to systems that operated only bus service
(as opposed to subway, commuter rail, ferryboat, etc.), that operated at least
five vehicles, that did not contract out senrice, and that provided complete
information for all years of the survey.5 These payroll and wage data were
combined with revenue data to look for evidence of rent capture from dedicated
taxes.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Survev Results on Dedicated Taxes
By the end of the sample period, 64 of the systems reported having
local dedicated revenue sources to support their operations. The UMTA data
report the exact dollar value of monies spent from dedicated sources, but
provide little institutional information on the type of tax used. For this
research, I have supplemented the UMTA data through a telephone survey of the
managers or staff of transit systems reporting dedicated revenues. This
information--the type of tax, year the tax was enacted, and historical
circumstances surrounding the tax enactment--is summarized in table 4.

Property taxes were the most common type of tax observed (28 of 64
systems), followed by gasoline excise (17), sales (14), and payroll taxes (5)
California, which accounts for 15 of the observed gasoline taxes, enacted a
gasoline tax in 1972 that is administered by the state but returns money to
the local level for transit based on the money collected in that area.

For

the purposes of this study and for the UMTA statistics, these funds are
assumed to be local dedicated revenues.6

Eight of the dedicated taxes were

enacted during the sample period and the other 56 during the 15-year period
prior to 1982.

A majority were enacted between 1967 and 1973, when federal

capital assistance was being provided for the creation of publicly owned and
operated transit systems. Only five of the taxes observed required periodic
voter approval, and none of the taxes was repealed during the sample period.

A potentially serious objection to the analysis is that the presence

of a dedicated tax is the result of high public-sector wages. However, the

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

16
process by which most of these taxes were enacted renders this possibility
tenuous at best.

The historical circumstances surrounding the enactment of these
taxes in the local mass-transit industry strongly suggest that funding
policies at the state level were the major determinant of the existence of
dedicated taxes for transit. In the late 1960s and early 1970s, the federal
government provided capital assistance for the establishment of public transit
systems subject to the provision of matching funds and operating assistance by
state and local governments. States with established subsidy programs for
transportation met the federal requirements from existing funding sources at
the state level and from local general revenues. Many states without existing
funding sources, however, encouraged the formation of special districts with
taxing authority to meet the federal requirements. Thus, most of the taxes
observed in this period were established when the district was created.

Transit systems in Pennsylvania--whose state government plays an
activist role in mass transit--are barred by state law from enacting dedicated
taxes.

In contrast, neighboring Ohio provides no state funding for mass

transit but allows voter referenda to grant new local transit systems the
authority to enact taxes. .Of the sample of eight transit systems in Ohio,
seven had dedicated taxes.

No local dedicated taxes exist in New York and

Wisconsin, both of which have extremely generous state programs for mass
transit.

Rural-dominated siate legislatures in Texas, Kansas, Kentucky,

Missouri, Oregon, and Washington, however, provide authority for the use of
local dedicated taxes because of their aversion to providing financial support

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

17
for inner-city bus service. (There are reportedly also anti-union and racial
motivations to some of these decisions, since most transit systems are
unionized and a disproportionate number of riders are black.)

Taxes enacted after 1974 typically were instituted after the transit
system had been publicly owned and briefly supported by local general
revenues. Circumstances surrounding these taxes are varied, but apparently
the drain on local general revenues for operating assistance was larger than
originally expected and new funding mechanisms were required. This was
especially true in areas desiring to expand their systems. Groups supporting
the expansion of transit and dedicated taxes in general included downtown
business interests; transit-dependent populations such as the poor,
handicapped, and elderly; and transit unions.

Finally, small, city-run systems in the South typically did not use
dedicated taxes. The service areas of these systems were historically covered
by the bounds of the city government, eliminating the need for a regional
agency or specihl district. States reporting such service were North
Carolina, Georgia, Alabama, Tennessee, and Mississippi.

Preliminarv Em~iricalEvidence
Sample means (and standard deviations) for system size, revenues,
and vehicle operator wage rates in 1985 are reported in table 5. The systems
with dedicated taxes are, on average, three times as large as other systems,
in terms of both miles of service delivered and number of vehicle operators.
Revenues, however, are four times as large on average, suggesting that systems

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

18
with dedicated taxes have larger budgets relative to service provided. A
breakdown of revenue composition implies that the use of dedicated taxes
precludes the use of local general-revenue sources. Among the dedicated-tax
systems, earmarked revenues provide 40.4 percent of total funding, while local
general revenues account for 1.9 percent. Systems without dedicated taxes,
however, receive 23.7 percent of their funding from local general revenues and
only 0.1 percent from earmarked sources.8 Thus, these two groups of transit
systems exhibit extreme differences both in local funding mechanisms and in
composition of revenues, suggesting that comparison of wages and payroll
between the two groups is a useful natural experiment.

Average wages paid to vehicle operators were 16 percent higher in
systems using dedicated taxes versus those without. This finding provides
only prima facie evidence of rent capture, because several other explanations
exist for this differential.

First, dedicated-tax systems may have a higher

rate of unionization than nondedicated-tax systems. Second, operators in
dedicated-tax systems could be higher-quality, more productive workers.
Third, dedicated taxes may be more common in high-wage areas.

Finally, there

could be several unobservable fixed effects correlated with both high
public-sector wages and the use of dedicated taxes: for example, strong
public-sector unions, tastes for good or bad government, and political
structure.

These concerns are addressed in turn below. In the case of union
status, I found that 81 percent of the nondedicated-tax systems were unionized

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

19
versus 97 percent of the dedicated-tax systems. Union status can be
controlled for through the standard use of a dummy variable. No information
is available, however, on the human capital of the public vehicle operators.

I used the number of collisions and fatalities involving transit vehicles as
measures of driver quality, but they had no significant explanatory value and
do not affect the estimates presented here. I am left to assume that the
quality of bus drivers is homogeneous across transit systems.

To measure regional variation in private-sector wage rates, I ran a
standard human-capital wage regression on the universe of private-sector
workers in the Current Population Survey, controlling for industry,
occupation, and 94 distinct geographic areas interacted with union status.
The average human-capital measures for motor vehicle operators were then used
with these results to project a union and nonunion wage rate for vehicle

+

operators in each geographic area. This procedure was repeated for each of
the four years in the sample period. These estimated private-sector wages
were then used as independent explanatory variables for vehicle operator wages
in public transit systems.

Sample means for the private-sector wages and other demographic
variables are reported in.table6. Wages average 4.6 percent higher for
nonunion, private vehicle operators in areas using dedicated taxes versus
those without, explaining part of the differential observed in public wages
The areas using dedicated taxes also have higher populations, density,
incomes, and population growth. Areas without dedicated taxes have higher
poverty rates and a higher percentage of black population, reflecting the use

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

20
of city financing in the South.

No significant difference is observed in

transit demand in these areas as measured by the percentage of people who
drive to work versus take mass transit.

In general, -the sample means suggest that wages are higher among
transit systems that use dedicated tax rates.

Vehicle operators in the

dedicated tax systems receive, on average, a 12.3 percent, or $1.07 per hour,
premium above the private-sector wage for nonunion drivers. Operators in
systems without dedicated taxes receive a premium of only 1.2 percent, or
$0.10 per hour. These statistics, of course, do not systematically control
for demographic variables, unionization, economies of scale, and other
observable variables influencing wage rates. Finally, as already discussed, a
series of possible unobservable fixed effects associated with dedicated taxes
and public-sector wages are also not controlled. The empirical analysis of
the next section takes into account both these observable independent
variables and unobservable fixed effects in order to test the robustness of
this evidence of rent capture.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

21

IV. Payroll and Wane Chan~esafter Tax Enactment

While the sample period in this study is only four years, the
dedicated taxes were enacted over a 17-year period. I exploit this variation
in the age of taxes in order to make inferences regarding the time path of
payroll, wages, and revenue following tax enactment. To this end, I construct
a variable called YEAR that equals the number of years that have passed since
a tax was first enacted. The values of YEAR for a system with a tax enacted in
1980, for example, are 2, 3, 4, and 5 for the 1982-1985 sample period years,
respectively.

Table 7 reports the sample means (and standard deviations) for
payroll, public operator wage, private nonunion operator wage, and system
size. The means for systems without dedicated taxes are broken down by union
status. For dedicated-tax systems, the means are broken down by the values of

YEAR observed during the four-year sample period. There are 404 time-series
cross-section observations for the nondedicated-tax systems, and 256 such
observations for the dedicated-tax systems. Values are reported in 1985
dollars. Caution is needed in interpreting these results, as individual
dedicated-tax systems appear in four different year groups, introducing
correlation among the sample values.

The sample means do not suggest that dedicated taxes are enacted in
predominantly high-wage areas or in systems that already have high wages. The
average public wage in the pre-tax observations is $7.97 per hour versus an
average $8.59 per hour for the unionized nondedicated-tax systems. Compared

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

22

with the private-sector wages, these wages represent $0.20 and $0.55 premiums,
respectively, suggesting that unions in the systems initially without
dedicated taxes were not doing as well as their peers. In the year following
tax enactment, however, the means suggest that public-sector wages jump to
$9.17 per hour, a $1.13 premium, and peak at $10.55 per hour by year 4, a
$2.34 premium. While the wages and premiums vary in the ensuing years, they
remain above the average for the unionized nondedicated average and are above
$9.00 in all years but 7 and those over 16.

Examination of payroll per mile

of service shows a 34-percent jump following the enactment of the dedicated
tax.

In general, these sample means provide further evidence that
enactment of a dedicated tax leads to rent capture by the vehicle operators
Whether measured in payroll or wages, labor costs rise substantially in the
years immediately following the enactment of the tax and remain above the
labor costs existing in systems without dedicated taxes over time. These
statistics, of course, do not systematically control for demographic
variables, unionization, economies of scale, private-sector wages, and other
observable variables influencing wage rates.

The econometric analysis uses pooled cross-section regressions with
vehicle operator payroll and wage rates as dependent variables. Wage rates
are a commonly used measure of union success in delivering rents to workers,
but are probably not the best measure in this study for two reasons. First,
payroll measures reflect both wage rates and a union's ability to expand
membership. Thus, comparing payroll changes with revenue changes stemming

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

23
from a new tax provides a more complete measure of rent capture.

Second, the

wage rates used below were constructed from the payroll data and data on
employee labor hours. While the payroll data come from balance sheets and are
reportedly accurate, labor hours are more poorly measured, potentially
introducing error into the wage measures. For these reasons, the discussion
will focus principally on the payroll estimates, although the wage regressions
are also reported and deliver qualitatively similar results.

The basic equation for estimation, shown in equation (I), uses
the log of public operators' payroll (wage) as the dependent variable as
follows :

Time-variant explanatory variables include the log of miles of service
provided (LSIZE), the log of the private-sector nonunion wage (LPRIWAGE),
sample-year dummies (DUM83, DUM84, DUM85).

and

The impact of a dedicated tax is

measured by the dollar amount of dedicated revenues collected per mile
(DEDREV), DEDREV-squared, and a length j vector of YEARj dummies, j=1,. . . ,17,

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

24
measuring time since tax enactment. YEAR5 equals one, for example, when
YEAR=5. The YEARj and DEDREV variables are set to zero for systems with no
dedicated taxes.

Finally, the specification includes a vector of six

variables measuring union status and area demographics, which were not
observed to change over the sample period. (In the case of the demographic
variables, this was due to data limitations.) The variables collected for
each system area were union status, log of population, log of population
density, percent change in population between 1980 and 1984, percent of black
9
population, log of per capita income, and the poverty rate.

Two procedures are used.

The first is a generalized least squares

(GLS) procedure that controls for cross-section heteroscedasticity and
time-wise first-order correlation of errors as discussed in Kmenta (1986).

In

the absence of correlated fixed effects, this procedure is efficient. If fixed
effects are present and correlated with the independent variables, however,
the GLS estimates are inconsistent. The second procedure used is the standard
fixed-effects (FE) or "within-group" estimator discussed in Hausman and Taylor
(1981).

The cross-section observations are differenced from their individual

means, and ordinary least squares (OLS) is run on the transformed data. All
time-invariant variables are eliminated. In the absence of correlated fixed
effects, the FE estimator is consistent but inefficient. If fixed effects are
present, however, the FE estimator still yields consistent results. Hausman
specification tests are then conducted to test the null hypothesis of no
correlated fixed effects.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Generalized Least Sauares Results
Results for the GLS payroll regressions are reported in table 8a.
In equation (I), which omits the dedicated tax variables, LSIZE has an
estimated coefficient (standard error) of 0.981 (0.014), suggesting that
payroll has a scale elasticity of nearly one.

LPRIWAGE has an estimated

coefficient of 0.455 (0.086), showing that nearly half of the variation in
private-sector wages is reflected in the public-sector payroll wage. When the
demographic variables are omitted (not shown here), the private wage
coefficient is 0.578 (0.091).

While my prior expectation was a relation

closer to one, restricting the coefficient to one did not substantially affect
the results that follow. Standard errors increased, but the estimated
10
coefficients changed little and were still significant.

The estimated coefficient for the union dummy is 0.325 (0.28),
suggesting that unions raise payrolls by 38 percent. This is a large effect
compared to most estimates of private and public union wage premiums. It
reflects in part the use of low-wage, part-time operators by nonunion systems
and, to the extent that the size and private wage variables are not perfect
controls, the small size of these systems and their concentration in low-wage
areas of the South. It also implies, however, that transit unions are in
general successful in capturing rents for their members. Finally, the time
dummies show an upward trend in wage rates and, among the demographic
variables, population and density have positive and significant effects on
payroll, while percent black, poverty, and income have negative effects.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Equation (2) in table 8a measures the impact of the dedicated tax on
payroll by entering the level of dedicated revenues per mile, DEDREV, and
DEDREV-squared into the specification. The estimated coefficient for DEDREV
is 0.0640 and is statistically significant at the 99-percent confidence level
with a t-statistic of 4.64. The estimated coefficient (standard error) for
DEDREV-squared is -0.0023 (0.0026).

Evaluated at the average value of DEDREV

for dedicated-tax systems in 1985 ($1.33 per mile), these estimates suggest
that use of a dedicated tax increases operator payrolls by 8.4 percent.

Equation (3) of the table omits the level variables but includes the
YEAR dummies in order to explore the time path of payroll following tax
enactment. Payroll is significantly higher in YEAR3, YEAR4, and YEARS, with
estimated coefficients of 0.082, 0.100, and 0.084, respectively. The
estimated effect diminishes in YEAR6 and YEAR7 (where there are few
observations), but returns to the 0.07 to 0.10 range for the following years
and remains statistically significant through YEAR15. These estimates suggest
that payroll rises 7 to 10 percent in the years immediately following a tax
enactment and that these gains are relatively stable over time.

Equation (4) of table 8a controls for both the level of revenues and
the time path. The estimated coefficients for the level variables change
little from equation (2), and the estimate for DEDREV remains statistically
significant with a t-statistic of 2.90. The YEAR dummies, however, no longer
reveal a significant change in payroll over time. This result changes in the
fixed-effects (FE) estimates, which I will now discuss.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

27
Fixed-Effects Estimation Results
Results from the FE estimation are reported in table 8b, with
equations (5) through (8) corresponding to equations (1)
GLS estimates.

through (4) of the

Equation (5) again omits all dedicated-tax variables. The

estimated coefficient (standard error) of the size variable is 0.711 (0.026).
This is below the GLS estimates and suggests that economies of scale with
respect to labor costs exist within transit systems. The private-sector wage
variables and the time dummies, however, change little from the GLS estimates,
and the time-invariant demographic and union variables, of course, are
omitted.

The FE results for the dedicated-tax variables remain
statistically significant and increase in magnitude from the GLS estimates.
The estimated coefficient on DEDREV in equation (6), which omits the year
dummies, is 0.0947 and has a t-statistic of 4.53.

This suggests that the

average dedicated tax of $1.33 per mile raises payroll by 12.1 percent. When
the YEAR dummies are included instead of the level variables, however,
allowing for payroll growth over time, the estimated increase is much larger
As reported in equation (7), payroll grows by 22 percent in the four-year
period following tax enactment, remains 20 to 23 percent higher through year

7, then increases to a 33 .to 39 percent higher level in years 9 through 15,
before falling off in years 16 and beyond.

Equation (8) includes both the DEDREV variables and YEAR dummies.
The estimated coefficient for DEDREV is 0.0607, which is little changed from

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

28
the GLS estimates, and has a t-statistic of 2.28. The YEAR dummies, however,
in contrast to the GLS results, are statistically significant and reveal
growth in payroll in the early years of the dedicated tax. For the average
dedicated tax, the results suggest that payroll grows by 21 percent during the
first four years following tax enactment, then reaches a 30- to 34-percent
higher level in years 10 through 15.

Note that the FE estimates of the long-term impact of a tax rely on
observed payroll changes occurring within individual systems during the
four-year sample period.

Because taxes were enacted at different times, an

estimate of the cumulative impact of a tax lasting J years is feasible. The
FE estimate for a tax lasting J years links together payroll changes observed
in individual systems with taxes of age 1 through J.

For higher values of J ,

therefore, the standard error of the total estimated change grows. This
increasing imprecision is seen in the steady growth of standard errors of the
FE estimates of the YEARj coefficients for higher values of YEAR. Indeed, for

J > 15, the estimated payroll change is not statistically significant,
although the point estimate is still large. Greater confidence can thus be
placed in the projections for the early years of a dedicated tax, particularly
those immediately following tax enactment.

S~ecificationTests
In the absence of correlated fixed effects, the GLS estimates are
efficient, while the FE estimates are inefficient but consistent. If
correlated fixed effects are present, however, the GLS estimates are
inconsistent, while the FE estimates remain consistent.

Hausman (1978) shows

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

that a specification test can be constructed by comparing the two estimators.
Under the null hypothesis of no correlated fixed effects, the difference in
the estimated P's,

PFE - BGLS,should

be small relative to the difference

in the covariance matrix of the estimates, VFE - VGLS. The Hausman
test-statistic M, shown in equation (2) below, is distributed

under

x2
q

H,, where q is the number of potentially biased coefficients tested.

Hausman tests of the GLS specifications (1) through (4)--FE
specifications (5) through (8)--reject

the null hypothesis of no correlated

fixed effects at the 99.5-percent confidence level. The test statistics
(critical values) are 125.2 (16.8), 131.0 (20.3), 49.8 (44.2), and 156.4
(46.8), respectively. The GLS results reveal a similar pattern of the effect
of dedicated taxes, but of lower magnitude. Unobserved fixed effects appear
to bias downward the estimated change in payroll that follows a tax enactment.

Wage Regression Results .
Using wage rates as a dependent variable instead of payroll yields
qualitatively similar results. The GLS regressions, reported in table 9a,

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

30
show that including the DEDREV variables alone (equation (10))

results in an

estimated coefficient (standard error) of 0.0598 (0.0119) for DEDREV and
-0.0074 (0.0026) for DEDREV-squared, representing 6.8 percent higher wages for
the average dedicated tax. The significant negative coefficient on
DEDREV-squared suggests a decreasing ability to raise wages for higher revenue
levels. Including the YEARj dummies alone (equation (11))

indicates a

statistically significant 8.4 percent rise in wages from an initial level of
$8.23 to $8.92 by year 2 of the tax. Wages remain in the $8.85 to $9.23 range
between years 4 and 11, then decline tp $8.69 in year 14. The higher levels
are statistically significant throughout this period. Including both the
DEDREV and YEARj variables (equation (12)) indicates that wages rise
significantly with the level of dedicated revenues and also have significantly
higher levels in years 4, 6, and 8 through 11, suggesting an upward time path
after tax enactment.

The FE estimates for wages are reported in table 9b. Contrary to
the GLS and payroll results, a statistically significant increase stemming
from dedicated revenue levels (equations (14) and (16))

is no longer

indicated. The YEARj dummies (equations (15) and (16)), however, show a
statistically significant rise in wages following tax enactment. The increase
is on the same order of magnitude and follows a similar pattern to the payroll

FE estimates. As shown in figure 1, wages rise from an initial base of $8.23
to $10.00 by year 6 and remain at this level through year 12 before declining
slightly. The higher wages are statistically significant in year 2 and in
years 4 through 13. As with the payroll FE results, the 95-percent confidence
interval surrounding the estimate grows with higher values of YEAR. In spite

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

of the increasing imprecision of the FE estimates, the simulated wage path
follows closely the wage rates observed in the sample means, raising
confidence in the robustness of the results.

Hausman specification tests reject the null hypothesis of no
correlated fixed effects at the 99.5-percent confidence level for GLS
specification(9) and at the 95-percent confidence level for specifications
(10) and (12)--corresponding

to equations (13), (14), and (16) of the FE

results. The test statistics (critical values) are 24.6 (16.8), 17.9 (14.1),
and 39.2 (37.7), respectively. GLS specification 11, however, is not
rejected.

The test statistic is 1.2, versus a 95-percent critical value of

33.9. (The test statistic of this specification in the payroll regression
showed a similar drop, but still rejected.) Given the overwhelming rejection
of the GLS specification in all of the payroll regressions and in three out of
four wage regressions, I use FE estimates in the next section to simulate the
impact of a dedicated tax on system size, revenues, and the payroll share of
revenues.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

V. Simulation Results

To explore the implications of a dedicated tax for system size and
revenues, I ran FE regressions using LSIZE and the log of revenues, LREV, as
dependent variables, and the YEARj and other time-varying variables as
independent variables. These results are reported in table 10 in equations
(17) and (18), respectively. For comparative purposes and for the convenience
of the reader, I also report in equation (19) the FE payroll estimate from
equation (8).

-

Finally, I use the payroll share of revenues (measured as LPAY

LREV) as a dependent variable to directly estimate changes in the labor

claim on revenues. These results form the basis for the simulation exercises
that follow.

As with the other FE estimates, the standard errors grow for

higher values of YEAR; therefore, the focus of the discussion will be
principally on the years immediately following the enactment of a dedicated
tax.

Equation (17) indicates that systems that enact a dedicated tax
undergo a significant and large expansion in the six-year period following a
tax enactment. As shown in figure 2, the simulation suggests that the average
system grows 40 percent by year 3 and reaches a 45- to 52-percent higher level
by year 6, at which point it stabilizes. Such a large expansion is consistent
with the reports I received from transit managers who just recently enacted
taxes. The taxes are often presented to the voters as an opportunity to
increase service significantly. This large expansion in size naturally
corresponds with higher revenues and payroll. I therefore conduct simulations
allowing both for system expansion and for keeping the size of a system

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

constant. In simulations of expansions, I use the point estimate of the size
results from equation (17).

As in the previous section, the base values for

the simulations are the mean (log) values for systems without dedicated taxes

The change in revenues that follows the enactment of a dedicated
tax is estimated in equation (18) in table 10.

Total revenues from all

sources, as opposed to only dedicated tax revenues, is used as the dependent
variable for several reasons. While I know the exact amount of total revenues
spent and the percentage of those revenues from local dedicated taxes, the
total amount collected from a dedicated tax is not reported. Some dedicated
tax revenues may be stored in trust funds for use in expansions or in future
years, especially in the early years of the tax. Furthermore, while I know
the tax rates in most cases, I have no information on the size of the tax
base, preventing direct measurement of its change or erosion over time.
Finally, enactment of a dedicated tax usually coincides with a substitution
away from other revenue sources--principally local general revenues, though
sometimes fares are reduced as well.

Because labor is concerned with its

share of the total pie, examining labor's share in the change of total
revenues is a natural measurement of rent capture.

Coinciding with the expansion in size, revenue estimates show a
dramatic rise in the years immediately following tax enactment. (To simulate
the change in revenues, I sum the increase resulting from the YEARj dummies
'

with the increase implied by the assumed expansion path and the estimated

LSIZE coefficient.) As shown in figure 3, revenues jump a statistically

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

34
significant 40 percent during the first year of the tax, and remain at a 35to 45-percent higher level through year 14 before declining in the last few
years.

A similar projection for payroll, from the estimates in equation (19),

is shown in figure 4.

Payroll rises 50 percent by year 4, then increases to a

60- to 80-percent higher level in years 8 through 14. The increase is
statistically significant throughout the projection period. To compare the
payroll and revenue increases, figure 5 charts the dollar amount of changes
occurring in both on the same scale. Following tax enactment, payroll takes
up a steadily increasing percentage of the additional revenues, absorbing 27
percent by year 3, 47 percent by year 9 , and all additional revenues by the
end of the projection period.

To net out the direct effects on revenues and payroll resulting
from system expansion, or possible economies of scale, I also project revenues
and payroll while holding system size constant. Figure 6 shows that the
dedicated tax results in significantly higher revenues of 17 to 29 percent in
years 1 through 4. In the years immediately following a tax increase, the
system thus has much higher funding relative to the level of service provided
prior to the dedicated tax. This increase can thus be viewed as a measure of
"excess" revenues, which are a potential target for capture by labor unions.

A similar projection for payroll, shown in figure

7, suggests a

steady and permanent increase following tax enaction. Payroll 'rises to a 2 8 to 39-percent higher level by year 10, and the higher level is statistically
significant throughout the projection period. As shown in figure 8, the
change in payroll appears to absorb all excess revenues by year

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

6.

Payroll remains at this higher level in spite of the disappearance of

excess revenues by year 14.

Note that the point estimate of payroll edges

down once excess revenues disappear. In general, these results show that all
additional revenues are absorbed in payroll by year 6 of the tax, and that
these gains are permanent in spite of the gradual disappearance of additional
revenues resulting from the tax.

Finally, equation (20) in table 10 reports the direct impact of
a tax on the payroll share of revenues.

As shown in figure 9, payroll share

takes an initial drop from 2 8 . 7 percent to 2 5 percent, corresponding with the
initial surge in revenues. By year 6 , however, payroll share rises to 33
percent, a statistically significant higher level, and steadily increases to
38 percent by year 15.

Enactment of a dedicated tax thus appears to result in

a steady increase in the labor share of revenues. It should be noted that the
payroll share estimates fall well within the range observed in the raw data.

In sum, these results strongly suggest that the use of a
dedicated tax system provides significant opportunities for rent capture by
public-sector unions. Controlling for private-sector wages and system size,
statistically significant growth is seen in wages, payroll, and the labor
share of revenues. Within six years, the simulations suggest that higher
payroll absorbs all additional revenues resulting from a dedicated tax. The
payroll share of revenues continues to increase throughout the sample period,
in spite of a fall in revenues.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

VI.

Conclusion

This paper provides a direct test of whether the enactment of a
dedicated tax leads to rent capture by public employees. In a natural
experiment provided by the wide variation in funding arrangements for local
mass-transit systems, the empirical analysis reveals a systematic link between
changes in the wages and payroll for a homogeneous type of employee--bus
operators--and enactment of a dedicated tax. The results are robust across
several specifications, and the simulation values of wages, payroll, revenues,
and payroll shares from a hypothetical tax are well within the observed range
of values.

Sample means show that hourly wage rates for public-sector
operators rise from a pre-tax level of $7.97 to more than $10.00 in the
two-year period following a tax enactment, and over time remain substantially
above the average of $8.59 per hour for unionized systems with no dedicated
taxes. Pooled time-series cross-section regressions are used to control for
local private-sector wage rates, system size, unionization, and demographic
characteristics. A generalized least squares (GLS) procedure is initially
used and suggests that significant wage and payroll gains of 8 to 10 percent
follow a tax enactment.

Although it is possible that the existence of a dedicated tax
results from high local wages, history strongly suggests that the principal
determinant for the existence of dedicated taxes for transit are state-level
policies. The more plausible channel for bias is the presence of unobserved

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

37

fixed effects that are correlated with both public-sector wages and the
existence of dedicated taxes. This potential bias is explicitly tested and
controlled for with the standard fixed-effects (FE) estimator, which reveals a
statistically significant increase in labor costs of greater magnitude than
the GLS estimates. Payroll and wages rise 20 percent in the six years
following the tax enactment, and this increase remains stable over time.
Hausman specification tests strongly reject the null hypothesis of no
correlated fixed effects.

The FE results are used in simulations that measure the extent to
which the payroll increases represent the capture of additional revenues
resulting from a dedicated tax. The results suggest that in the period
immediately following the enactment of a dedicated tax, transit systems expand
significantly in size, revenues, and payroll. Holding system size constant,
the results show that payroll absorbs all additional revenues by year 6 of the
tax. The payroll share of total revenues takes an initial drop with the surge
of new revenues, but as payroll and wages steadily increase, the payroll share
rises from 27 percent at year 1 to almost 40 percent at the end of the 17-year
projection period.

The results,suggest that enactment of a dedicated tax leads to
significant rent capture by public-sector unions in the local mass-transit
industry. They support the argument that traditional budgeting methods that
weigh the costs and benefits of competing uses of funds act as a check on
public employee power. Thus, efficiencies that result from earmarked funding
may be offset by increased labor costs.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

These results should be treated with some caution, however, for
the following reasons.

First, the simulation of payroll and revenues for a

hypothetical tax lasting up to 17 years was estimated from only a four-year
sample period. Future work will entail collecting wage and payroll data prior
to 1982 to see if the wage gains observed in the sample period also occurred
in earlier years. Second, while the assumption that bus drivers are
homogeneous seems reasonable, I have no direct measures of human capital with
which to test this assumption. Changes in the quality and composition of
employees, however, are reported to occur in the administrative area following
a tax enactment. Systems with dedicated taxes appear to hire highly paid
managers and additional planning, marketing, and public relations personnel.
(The general manager of Bi-State Transit in St. Louis is reportedly the
highest-paid public employee in Missouri, with the exception of the governor.)
One possible approach in future work will be to examine whether changes in the
composition of administrative staff result in higher-quality service or
cost-effectiveness.

In general, the evidence of rent capture appears to be robust for
the local mass-transit industry and confirms differences in wages and salary
share of revenues observed in the aggregate data on the use of dedicated'taxes
in special districts. While the local mass-transit industry provides a clean
experiment due to its homogeneous output, inputs, and production processes,
examination of districts producing other types of outputs would provide
additional confirmation. Also of interest would be examination of dedicated

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

39

taxes at the state and federal levels.

Examination of the impact of

earmarking for education at the state level is a possible avenue for future
research.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Footnotes
1. For an extreme example of political empire-building resulting from a
dedicated revenue source, see the Caro (1984) account of Robert Moses and the
toll revenues of the Tri-borough Bridge Commission in New York.

2. For a general discussion of earmarking, see Gold, et. al. (1987)
3. For for a general discussion of economic rents and regulation, see Joskow
and Rose (forthcoming).
Studies of airline deregulation include Card (1986).
4. None of the dedicated taxes observed in this study was repealed during the
four-year sample period.
5. Thirteen systems that provided data that were obviously "wrong" or
inconsistent over time were also omitted. For example, one system reported
having no expenses for fuel or tires. Others reported having no operating
employees. Still others reported large swings in the size of operations and
revenues. In general, the payroll, revenue, and operating data such as
mileage are of high quality, while the data on employee hours and passengers
is less so.
6 . Omitting California from the sample results in larger standard errors but
does not substantively change the results.

7. Wisconsin subsidizes operating expenses at a 36-percent matching rate.
8. Certain systems without dedicated taxes receive trace amounts of revenues
earmarked from local sources such as parking meter fees, license fees, and
other local charges.
9. Demographic variables were collected from the U.S. Census Bureau (1986).
10. Other specifications tested included those entering crashes, fatalities,
and the private-sector union wage. Coefficients on these variables were
insignificant and did not affect the results.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

References
Ashenfelter, O., and G. Johnson, "Unionism, Relative Wages, and Labor Quality
in U.S. Manufacturing Industries," International Economic Review, 13 (October
1972), pp. 488-507.
Buse, A., "Goodness of Fit in Generalized Least Squares Estimation," American
Statistician, 27 (1973), pp. 106-108.
Card, D., "The Impact of Deregulation on the Employment and Wages of Airline
Mechanics," Industrial and Labor Relations Review, 39 (July 1986), pp.
527-538.
Caro, R., The Power Broker, (1984)
Clark, K., "Unionization and Firm Performance: The Impact on Profits, Growth,
and Productivity," American Economic Review, 74 (December 1984), pp. 893-919.
Courant, P., E. Gramlich, and D. Rubinfeld, "Public Employee Market Power and
the Level of Government Spending," American Economic Review, 69 (December
1979), pp. 806-817.
Freeman, R., "Unionism Comes to the Public Sector," Journal of Economic
Literature, 24 (March 1986), pp. 41-86.
Gold, S., B. Erickson, and M. Kissell, Earmarking State Taxes, National
Conference of State Legislatures, Denver (1987).
Gyourko, J., and J. Tracy, "On the Political Economy of Land Value
Capitalization and Local Public Sector Rent-Seeking in a Tiebout Model,"
Journal of Urban Economics, (forthcoming).

, "An Analysis of Public and Private Sector Wages Allowing for
Endogenous Choices of Both Government and Union Status," Journal of Labor
Economics,(April 1988).
Hausman, J., "Specification Tests in Econometrics," Econometrica, 46 (November
1978), pp. 1251-1271.

, and W. Taylor, "Panel Data and Unobservable Individual Effects,"
Econometrica, 49 (November 1981), pp. 1377-1398.
Haworth, C., and D. Rasmussen, "Human Capital and Interindustry Wages in
Manufacturing," Review of Economics and Statistics, 53 (November 1971), pp.
376-380.
Joskow, P., and N. Rose, "The Effects of Economic Regulation," in R.
Schmalensee and R. Willig (Eds.), Handbook of Industrial Organization,
North-Holland Publishing Co., (forthcoming).

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Kmenta, J., Elements of Econometrics. Second Edition, Macmillan, New York,
(1986).
Masters, S., "Wages and Plant Size: An Interindustry Analysis," Review of
Economics and Statistics, 51 (August 1969), pp. 341-345.
Migue, J.L., and G. Belanger, "Toward a General Theory of Managerial
Discretion," Public Choice, 17 (1974), pp. 27-44.
~iskanen,W. , "Bureaucrats and Politicians," Journal of Law and Economics, 18
(December 1975), pp. 617-643.

, Bureaucracv and Re~resentativeGovernment, Aldine Publishers,
Chicago (1971).
Orzechowski, W., in T. Borcherding (Ed.), Budgets and Bureaucrats: The Sources
of Government Growth, Duke University Press (1977), pp. 229-259.
Pugel, T., "Profitability, Concentration, and the Interindustry Variation in
Wages," Review of Economics and Statistics, 62 (May 1980), pp. 248-253.
Rapping, L., "Monopoly Rents, Wage Rates, and Union Wage Effectiveness,"
Ouarterlv Review of Economics and Business, 7 (Spring 1967), pp. 31-47.
Rose, N., "Labor Rent Sharing and Regulation: Evidence from the Trucking
Industry," Journal of Political Economv, 95 (December 1987), pp. 1146-1178.
Ross, S., and M. Wachter, "Wage Determination, Inflation, and Industrial
Structure," American Economic Review, 63 (September 1973), pp. 675-692.
Ruback, R., and M. Zimmerman, "Unionization and Profitability: Evidence from
the Capital Market," Journal of Political Economy, 92 (December 1984), pp.
1134-1157.
Salinger, M.. "Tobin's q , Unionization and the Concentration-Profits
Relationship," Rand Journal of Economics, 15 (Summer 1984), pp. 159-170
Schmenner, R., "The Determination of Municipal Employee Wages," Review of
Economics and Statistics, 55 (February 1973), pp. 83-90.
Slichter, S., "Notes on the'Structure of Wages," Review of Economics and
Statistics, 32 (February 1950), pp. 80-91.
U.S. Census Bureau, State and Metro~olitanArea Data Book, U.S. Government
Printing Office, Washington, D.C.,(1986).
U.S. Census Bureau, Census of Governments, U.S. Government Printing Office,
Washington, D.C.,(various issues).

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Urban Mass Transportation Administration, U.S. Department of Transportation.
National Urban Mass Transportation Statistics. Section 15 Annual Report, U.S.
Government Printing Office, Washington, D.C., (various issues).
Weiss, L., "Concentration and Labor Earnings," American Economic Review, 56
(March 1966), pp. 96-117.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 1
Trends in Special Districts:
Number of Units and Revenue Composition

--

Number of units
Percent of total
governmental units

14,405

21,264

14.1

26.2

-

25,987

28,719

32.5

34.9

\

Revenue~
*
(1982 $ , millions)
6.280

11.838

21.795

30.961

Federal aid

315

763

3,723

4,405

State aid

140

475

1,274

1,810

Local general
revenues

119

747

1,555

2,057

Local dedicated
taxes

1,213

1,841

2,597

2,846

O w n source
revenues

4,493

8,013

12,644

19,843

Total

Revenues
(percent)

*

Total

-

100

-

100
-

100
-

Federal aid

5.0

6.4

17.1

14.2

State aid

2.2

4.0

5.8

5.8

Local general
revenues

1.9

6.3

7.1

6.6

Local dedicated
taxes

19.3

15.5

11.9

9.2

Own source
revenues

71.5

67.7

58 .O

64.1

100
-

Calculated with GNP deflator for state and local purchases.

Sources: U.S. Census Bureau, Census of Governments, and author's calculations.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 2
The Use of Dedicated Taxe,s:
Special D i s t r i c t Revenues and Wages
Census
Year
Total revenues
(1982 $ , m i l l i o n s )

*

Districts
With Dedicated
Tax Authority

Districts
Without Dedicated
Tax Authority

1967

5,371

6,466

1972

6,837

8,527

Dedicated t a x
revenues
(percent)

Wage and s a l a r y
share of revenues
(percent)

*

1967
1972

Calculated with GNP d e f l a t o r f o r s t a t e and l o c a l purchases

Sources: U . S . Census Bureau, Census of Governments, and a u t h o r ' s c a l c u l a t i o n s .

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 3
Composition of 1982
Special District Revenues (Percent):
States Rank-Ordered by Use of Local Dedicated Taxes

Quintile

Total

National
averape

100

* Unweighted

Federal State
Aid
Aid

14.2

5.8

---Local---Gen.
Ded .
Rev.
Taxes

6.6

9.2

Source

Wage and
Salary Share

64.1

25.2

Own

averages, Hawaii excluded. Quintile 3 has 9 states.

Sources: U.S. Census Bureau, 1982 Census of Governments, and author's
calculations.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 4
Dedicated Taxes for Local Mass-Transit Systems
Year
Enacted State

System Name

Type
of Tax

Ft. Lauderdale-Browrd Cnty TA
Waterloo-Black Hawk Cnty TA
Austin TA
Fort Worth CITRAN

gasoline
property
sales
sales

Gainesville-RTS

gasoline

Periodic
Renewal ?

Hillborough Area RTA
sales
Parkrsbrg-Mid-Ohio Valley TA property
property
Huntington-Tri-State TA

no
no
yes (3 year)

Lansing-Capital Area TA
Columbus-Central Ohio TA
Spokane TA

property
sales
sales

no
yes (5 year)
no

1980

WA

Tacoma-Pierce Cnty TS

sales

no

1978

KY
TX
TX
WV

Ft. Wright-TA No. Kentucky
San Antonio-VIA Metro Tr.
Houston-MTA
Wheeling-Ohio Valley RTA

payroll
sales
sales
property

no
no
no
no

IA

Sioux City TS

property

no

S Daytona-E Volusia TA
-Indianapolis PTC

sales
sales

Denver-RTD
Louisville-TA River City

sales
payroll

Pinellas Suncoast TA
City of Dubuque-Keyline TS
Gary PTC
Topeka MTA
Ypsilanti-Ann Arbor TA
St. Louis-Bi-State TA
Cincinnati-SORTA
Portland-Tri-County MTD
Eugene-Lane County MTD
Charlestn-Kanawha Vly RTA

property
property
property
property
property
sales
payroll
payroll
payroll
property

no
no
no
no
no
no
no
no
no
yes (5 year)

Sacramento RTD
Los Angeles-SCRTD
Monterey-Salinas TA
Montebello Muni Bus

gasoline
gasoline
gasoline
gasoline

no
no
no
no

1976

CA
CA
CA
CA

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 4 (cont.)
Dedicated Taxes for Local Mass-Transit Systems
Year
Enacted

state

System Name

Type
of Tax

CA
CA
CA
CA
CA
CA
CA
CA
CA
CA
CA
NE
OH

Santa Monica Muni Bus
San Diego TS
Alameda-Contra Costa TD
Oxnard-S Coast Area Transit
Gardena-Municipal Bus
Santa Barbara MTD
Fresno TS
Stockton MTD
Bakrsfld-Golden Empire TD
Riverside TA
N San Diego TS
TA of Omaha
Akron-Metropolitan RTA

gasoline
gasoline
gasoline
gasoline
gasoline
gasoline
gasoline
gasoline
gasoline
gasoline
gasoline
property
property

IL
IL
IL
IN
MO
OH
OH

Champaign-Urbana MTD
Rock Island County MTD
Greater Peoria Mass TD
Greater Lafayette PTC
Kansas City Area TA
Canton RTA
Youngstown-Western Res. TS

property
property
property
property
sales
property
property

MN
MN
OH
UT

Minneapolis MTC
Duluth TA
Toledo RTA
Salt Lake City-Utah TA

property
property
property
sales

1968

IL
IN

Springfield MTD
Fort Wayne PTC

property
property

1967

IA
IN
KS

Cedar Rapids Bus Dept.
South Bend PTC
Wichita MTA

property
property
property

1972
(cont . )

Source: Telephone survey by.author.

Periodic
Renewal ?

no
no
no
no
no
yes (5 year)
yes (10 year)

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 5
Transit System Summary Statistics:
Size, Revenues, Wages
Variable

Systems With
Dedicated Taxes

Systems Without
Dedicated Taxes

Number of
observations
System size
(miles, 000)
Vehicle
operators
Revenues
( $ millions)
Revenue composition
(percent)
Federal aid
State aid
Local general
revenues
Local dedicated
taxes

40.4
(16.3)

Own source
(fares)
Unionized

63/65

82/101

Wages
( $ per hour)

*

1985 sample means (standard deviations).

Sources: UMTA (1985) and author's calculations.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 6
*
Private-Sector Wages and Demographic Variables
I

Variable Name

Areas With
Systems Using
Dedicated Taxes

Areas With
Systems Not Using
Dedicated Taxes

Private operator
nonunion wage
Private operator
union wage
Population
(000

Density
(pop. per sq. mi.)
Pop. growth
( % change, 1980-84)

531.8
(567.3)
14.4
(74.9)

Income
( $ per capita)
Poverty
( % of pop.)
Black
( % of POP.)
Drive to work
( % of POP.)
Bus to work
( % of pop.)

*

1985 sample means (standard deviations).

Sources : U.S . Census ~ureau'
(1986) and author 's calculations.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 7
Trends A f t e r Tax E n a c t m e ~ t :
P a y r o l l , Wages, and Size
Vehicle Operator
Wages ( $ per h r . )
Number,gf
Obs .
Svstems without
dedicated taxes
Nonunionized

Unionized

Systems with
dedicated taxes

-

Pre t a x
Years a f t e r
t a x enactment

Payroll
( $ per mi.)

---------------Public

Private

System
Size
(000 m i . )

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 7 (cont.)
Trends after Tax Enactmen$:
Payroll, Wages, and Size

Years after
tax enactment

Number,gf
Obs .

Vehicle Operator
Wages ( $ per hr.) System
Payroll
---------------Size
( $ per mi.) Public
Private (000 mi.)

YEAR-15

YEAR= 16

YEAR > 16

* Means (standard deviations). All $ figures are in 1985 values, calculated
using the Consumer Price Index.
** Time-series/cross-sectional observations: 165 transit systems were observed
over a 4-year (1982-85) period for a total of 660 observations.
Source: Author's calculations.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 8a
*
Generalized Least Squares Payroll Regressions
Dependent Variable: Log of Operators' Payroll
Independent
Variables

LSIZE
(log miles)
LPRIWAGE
(log private wage)
DEDREV
( $ per mile)

Years after
tax enactment

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 8a (cont.)
*
Generalized Least Squares Payroll Regressions
Dependent Variable: Log of Operators' Payroll
Independent
Variables

YEAR17
(-1 if YEAR > 16)

Transit union
(-1 if yes)
Population
(log)
Density
(log)
% change in
population, 80

Black
( % of pop)

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 8a (cont.)
*
Generalized Least Squares Payroll Regressions
Dependent Variable: Log of Operators' Payroll
Independent
Variables
Income
(per capita, log)
Poverty
( % of pop.)

(1)
-0.198
(0.047)
-0.00008
(0.00004)

(2
-0.162
(0.047)
-0.00009
(0.00004)

(3)

(4)

-0.208
(0.051)

-0.181
(0.049)

-0.00009 0.00009
(0.00004) (0.00004)

Constant

Buse R-squared
Mean dep. var.
Log likelihood
Estimated rho

0.983
7.219
627.07
0.823

0.976
7.219
636.70
0.825

0.980
7.219
634.10
0.817

0.974
2.093
680.666
0.767

* Estimated coefficients (standard errors). GLS prodedure used on
cross-sectionally heteroscedastic and time-wise autoregressive model discussed
in Kmenta (1986). Cross-sections restricted to have same autoregressive
parameter. Buse R-Squared defined in Buse (1973).
Source: Author's calculations.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 8b
*
F i x e d - E f f e c t s P a y r o l l Regressions
Dependent V a r i a b l e : Log o f O p e r a t o r s ' P a y r o l l
Independent
Variable

(5)

(6

(7)

(8)

0.435
(0.141)

0.436
(0.137)

0.367
(0.142)

0.362
(0.141)

LSIZE
(log miles)
LPRIWAGE
( l o g p r i v a t e wage)
DEDREV
($ p e r mile)

Years a f t e r
t a x enactment

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 8b (cont.)
*
Fixed-Effects Payroll Regressions
Dependent Variable: Log of Operators' Payroll
Independent
Variable

YEAR17
(-1 if YEAR > 16)

-----

-----

0.210
(0.111)

0.068
(0.125)

................................................................
R-squared
Mean dep. var.
Log likelihood
Hausman test stat.

*

0.673
7.219
806.96
125.2

0.690
7.219
824.51
131.0

Estimated coefficients (standard errors).

Source: Author's calculations.

0.693
7.219
828.07
49.8

0.699
7.219
834.78
156.4

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 9a
*
Generalized Least Squares Wage Regressions
Dependent Variable: Log of Operators' Wage
Independent
Variables
LSIZE
(log miles)
LPRIWAGE
(log private wage)
DEDREV
( $ per mile)

Years after
tax enactment

0.237
(0.069)

0.134
(0.070)

0.126
(0.078)

0.117
(0.078)

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 9a (cont.)
*
Generalized Least Squares Wage Regressions
Dependent Variable: Log of Operators' Wage
Independent
Variables

YEAR17
(=1 if YEAR > 16)

Transit union
(-1 if yes)
Population
(log)
Density
(log)
% change in
population, 80-84

Black
( % of pop)

(9)

(10)

(11)

(12)

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 9a (cont.)
*
Generalized Least Squares Wage Regressions
Dependent Variable: Log of Operators' Wage
Independent
Variables
Income
(per capita, log)

(9

(10)

(11)

(12)

0.007
(0.037)

0.004
(0.037)

0.003
(0.039)

0.010
(0.039)

Poverty
( % of pop)
Constant

..................................................................
Buse Risquared
Mean dep. var.
Log likelihood
Estimated rho

0.664
2.090
717.43
0.814

0.677
2.090
723.21
0.808

0.682
2.090
722.59
0.804

0.680
2.090
724.58
0.804

* Estimated coefficients (standard errors). GLS procedure used on
cross-sectionally heteroscedastic and time-wise autoregressive model discussed
in Kmenta (1986). Cross-sections restricted to have same autoregressive
parameter. Buse R-squared defined in Buse (1973).
Source: Author's calculations.

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 9b
*
F i x e d - E f f e c t s Wage R e g r e s s i o n s
Dependent V a r i a b l e : Log o f O p e r a t o r s ' Wage
Independent
Variables

LSIZE
(log miles)

0.028
(0.026)

0.023
(0.027)

0.013
(0.027)

0.013
(0.027)

LPRIWAGE
( l o g p r i v a t e wage)

0.203
(0.142)

0.201
(0.141)

0.165
(0.145)

0.165
(0.145)

DEDREV
($ per mile)

Years a f t e r
t a x enactment

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 9b (cont.) *
Fixed-Effects Wage Regressions
Dependent Variable: Log of Operators' Wage
Independent
Variables

YEAR17
(=1 if YEAR > 16)

-----

-----

0.092
(0.113)

0.080
(0.129)

.................................................................
R- squared
Mean dep. var.
Log likelihood
Hausman test stat.

*

0.470
2.090
803.60
24.6

0.472
7.219
805.31
17.9

Estimated coefficients (standard errors).

Source: Author's calculations.

0.488
7.219
815.41
1.2

0.489
7.219
815.59
39.2

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 1 0
Size, Revenue, Payroll, and Paygoll Share
Fixed-Effects Regressions

(17)

Dependent Variable
(18)
- (19)

Independent
Variable

Log
Size

Log
Revenue

LSIZE
(log miles)

---

(20)

.............................................

LPRIWAGE
(log private wage)
Years after
tax enactment

YEAR?
(-1 if YEAR-7)

0.664
(0.032)

Payroll Share
Log
(Log Payroll Payroll
Log Revenue)
0.682
(0.027)

0.018
(0.026)

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Table 10 (cont.)
Size, Revenue, Payroll, and Paxroll Share
Fixed-Effects Regressions
(17)

Dependent Variable
(18)
(19)

Log
Size

Log
Revenue

(20)

.............................................
Independent
Variable

YEAR17
(=1 if YEAR > 16)

DUM85

Payroll Share
Log
(Log Payroll Payroll Log Revenue)

0.362
(0.165)

-0.013
(0.055)

0.150
(0.044)

0.081
(0.037)

-0.069
(0.036)

0.070
7.465
421.08

0.631
8.460
710.34

0.693
7.219
828.07

- 1.240

-----------------------------------------------------------------------R- squared
Mean dep. var.
Log-likelihood

*

Estimated coefficients (standard errors).

Source: author's calculations.

0.107

845.47

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Figure 1

Waae Rates
$ per hour
15.00

0.00
0

2

4

6

8

10

12

14

Years since enactment of dedicated tax
Source: Author's Calculations

16

18

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Figure 2

System Size
Vehicle miles (000)

3,000

2

4

6

8

10

12

14

Years since enactment of dedicated tax
Source: Author's Calculations

16

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Figure 3

Revenues

2

4

6

8

10

12

14

Years since enactment of dedicated tax
Source: Author's Calculations

16

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Figure 4

Payroll
$ (000)

3,000 -

9 5 % contldonco Intorval

.:-. . ........ ................. -.

.. -.....

2,000 -

,

.. ........................

.. - -.-

0

0

1

I

I

I

I

I

1

I

2

4

6

8

10

12

14

16

Years since enactment of dedicated tax
Source: Author's Calculations

18

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Change in ~ e i e n u e svs. Payroll

$ (000)
-

rovonuoo

I

\

/

-

,/-.
.---

/ - - - - - -

/
F

\
%

payroll

\ \ N f

/

/
. C - \ - - -

0
/

/
/

/
/

/
/

/

C

I

I

i

I

I

I

I

1

2

4

6

8

10

12

14

16

Years since enactment of dedicated tax
Source: Author's Calculations

-

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Figure 6

Revenues
(size constant)

0

0

2

4

6

8

10

12

14

Years since enactment of dedicated tax
Source: Author's Calculations

16

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

(size constant)
3,000

2,000
@ S X oontldanoa I n t a r v a l

:.. ....*-..........-...
............. --..

..........
.

astlmata

................... - - - - . .- -_ .
........ -.-._..-----.._..
.... ..- --.

-

0

0

l

I

2

4

I

6'

-.-..

-

-C

I

I

I

I

I

8

10

12

14

16

Years since enactment of dedicated tax
Source: Author's Calculations

A-

18

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Figure 8

Change in Revenue vs. Payroll
(size constant)
-

A

revanuem

payroll
\
\
\

..

\

0
/

0

/
/
/
/
1

0

2

4

6

8

10

12

14

Years since enactment of dedicated tax

16

http://clevelandfed.org/research/workpaper/index.cfm
Best available copy

Figure 9

Payroll Share
(size constant)

percent

100

95% eontldoneo Interval

. - -..-... - -

e m - . . . . . . . . . .

.-

._-.-.--

/
. - . . . . - - I

-u-&-A-A-z,,

*--.-.

..9--------

oatlmato

_.....-.-.-....-..._.--------- ---

,
,
,,
,
,
,

.. -.

i

I

1

I

I

I

1

I

Years since enactment of dedicated tax
Source: Author's Calculations

1