Journal of Public Economics 90 (2006) 1115 – 1132

[Link]/locate/econbase

The dynamics of municipal fiscal adjustment

Thiess Buettner a, David E. Wildasin b,*
a
ifo and Munich University, Poschingerstrasse 5, D-81679 Munich, Germany
b
Martin School of Public Policy, University of Kentucky, Lexington, KY 40506-0027, USA
Received 20 December 2002; received in revised form 8 September 2005; accepted 8 September 2005
Available online 27 October 2005

Abstract

The dynamic fiscal policy adjustment of local jurisdictions is investigated empirically using a panel of
more than 1000 U.S. municipalities over a quarter of a century. Distinguishing own-source revenue, grants,
expenditures, and debt service, the analysis is carried out using a vector error-correction model which takes
account of the intertemporal budget constraint. The results indicate that a large part of the adjustment in
response to fiscal imbalances takes place by offsetting changes in future expenditures. In addition, the
results show that fiscal imbalances are financed to a significant extent by subsequent changes in grants.
Decomposition of the sample according to average city population reveals that the basic pattern of fiscal
adjustment is robust, although intergovernmental grants play a much more pronounced role in maintaining
budget balance for large cities.
D 2005 Elsevier B.V. All rights reserved.

JEL classification: H70; H72; H77

Keywords: Municipal fiscal policy; Dynamic fiscal policy

1. Introduction

Our goal in this paper is to shed new light on the dynamics of local government
policymaking, with specific reference to the fiscal policies of municipal governments in the
United States. The U.S. federal system, of which local governments are an important part, is a
durable institutional structure that decentralizes significant portions of public sector decision-
making authority to subnational governments. Governments at all levels within this structure
operate under a variety of constraints, and these constraints create the incentives that, in part,

* Corresponding author. Tel.: +1 859 257 2456; fax: +1 859 323 1937.
E-mail address: dew@[Link] (D.E. Wildasin).

0047-2727/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/[Link].2005.09.002
1116 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132

elicit the observed behavior of policymakers. The fiscal policies of subnational governments are
important in themselves, but they also illuminate the nature of the constraints which these
governments face and thus the institutional structure of the public sector itself.
Local governments in the U.S. are numerous, diverse, and economically important.1 The
services performed by these governments, their financing, and their relationships with the
national and state governments have evolved over time in a complex process involving the
interplay of all branches of government (executive, legislative, and judicial) at all levels of
government (federal, state, and local), all against the background of ongoing demographic,
technological, economic, and social change and widely varying local circumstances.
Occasionally—but rarely, in U.S. experience—local fiscal policies result in crises in which a
government’s financial obligations to creditors, vendors, and employees cannot be met from
existing revenues. In such situations, local authorities, as units of government subordinate to
states, are often subjected to special oversight mechanisms even as the state government assists
the locality with additional funding to meet its most pressing contractual and public-service
delivery obligations. The financial crises of New York City in the 1970s and of Philadelphia,
Orange County, California, and Washington DC in the 1990s provide well-known examples of
local fiscal policies gone awry, and smaller localities also encounter fiscal distress from time to
time.2 These events, though noteworthy, are nonetheless exceptions to the rule. Somehow,
despite (or perhaps in part because of) the conflicting demands imposed upon them by taxpayers,
interest groups, creditors, vendors, state governments, and others, local policymakers face an
incentive structure that, in equilibrium, results in behavior that for the most part preserves the
financial integrity of local governments. Whether the fiscal policies chosen by local governments
are economically desirable according to normative criteria (efficiency, equity) is a separate and
very important question. Leaving this question aside, one can observe that the institutional
structure of American federalism has created a system of local governments that pass a basic
survival test while permitting a relatively high degree of local fiscal autonomy—a fundamental
requirement for any fiscally decentralized public sector.
Since fiscal viability cannot be taken for granted at all times and places, it is a matter of some
importance to understand better how these governments manage the financial stresses to which
they are inevitably exposed. Our analysis is intended to shed new light, from an empirical
viewpoint, on the dynamics of municipal fiscal adjustment. Quantitatively speaking, how do
municipal governments adjust their revenues, expenditures, and debt policies over time? What role
do transfers from higher-level governments play in their fiscal dynamics? U.S. municipal

1
The Bureau of the Census publishes a quinquennial Census of Governments. As of the 1997 census, there were over
3000 countries, almost 20,000 municipalities, almost 35,000 special districts, almost 14,000 school districts, and almost
17,000 townships, making almost 90,000 units of local government in total. Total public expenditure by all localities
amounted to $837 billion, of which municipalities—the focus of the present analysis—accounted for $275B, school
districts $257B, counties $198B, special districts $89B, and townships $28B. Total local government spending in 1997
amounted to some 10.1% of GDP. In 1995 local government spending amounted to 26.9% of all public expenditures in
the US.
2
See GAO (1995) for series of case studies of localities in fiscal distress, some findings of which are summarized in
Holloway (1996a,b). The infrequency of formal municipal bankruptcy proceedings in the U.S. is quite remarkable. Since
the passage of the Municipal Bankruptcy Act in 1937, there have been fewer than 500 bankruptcy filings. Municipal
bankruptcies are a minuscule fraction of all bankruptcies. Of the total of 1.2 million bankruptcy proceedings commenced
in the 12 months preceding Sept. 30, 2000, 6 were filed under Chapter 9 (the portion of the bankruptcy code governing
municipalities). A large portion of Chapter 9 filings that do occur are accounted for by small special districts (such as
water or sewer districts), and bankruptcies by municipalities proper are therefore even more rare. See Administrative
Office of the US Courts (2000a,b, Table F-2) and National Bankruptcy Review Commission (1997).
 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132 1117

governments present an exceptionally attractive subject for the systematic empirical analysis of
decentralized fiscal policies. We have assembled a balanced panel of annual fiscal data for more
than 1000 municipalities for over a quarter-century. These data have been collected using
consistent definitions and, though they are of course subject to imperfections, they likely represent
the best available collection of fiscal data on such a large number of governmental units.3
Our analysis utilizes methods that have been exploited previously in macroeconomic analyses
devoted to the study of intertemporal government budget constraints for national governments
(e.g., Wilcox, 1989; Trehan and Walsh, 1991). Previous studies have generally focused on
testing for the stationarity or bsustainabilityQ of fiscal policies; by contrast, we pay closer
attention to the adjustment process that maintains this balance. Suppose, for example, that
municipal revenue declines, such that the local deficit is increased. Long-run budget balance
requires some offsetting adjustment to this increased deficit. What form does this adjustment
take, and when does it occur? Do municipalities with low revenue tend to reduce spending in
order to restore balance? Do they simply run bigger deficits for a period of time, delaying
adjustments in taxes and spending? Do lower revenues trigger additional transfers from higher-
level governments, enabling municipalities to maintain spending without having to raise local
taxes? Or is lower revenue in one period simply offset in subsequent periods, reducing the need
for further adjustment? Any of these types of fiscal adjustment, or some combination of all of
them, is conceivable, and the same can be said about possible paths of fiscal adjustment in
response to fiscal imbalances due to innovations in grants, spending, or debt service.
Given the complexity of the political and market constraints under which municipal
authorities operate, it is difficult to justify strong prior expectations about which particular form
of adjustment must dominate, and our goal here is to examine the dynamics of fiscal adjustment
with a minimum of prior structure. We do this using a vector error-correction approach, outlined
in Section 2. Section 3 describes the data and estimation approach, and provides specification
tests as well as further background information about the interpretation of the basic model.
Section 4 presents the results of the empirical analysis, using the entire sample. These results
indicate that municipalities respond differently to fiscal imbalances originating from different
sources. As an illustration, we find that higher municipal public spending is typically followed
by offsetting expenditure changes in subsequent periods, with relatively modest adjustments in
other fiscal variables; by contrast, imbalances associated with municipal revenue tend to persist
and to be followed by substantial changes in municipal spending.
For reasons of institutional structure and political economy, the incentives for fiscal
adjustment may differ significantly across municipalities, in particular, between large and small
cities. Because our sample contains a large cross-section of municipalities, it is possible to
analyze sub-samples with differences in population size separately, as is done in Section 5. We
find, in fact, that the basic pattern of fiscal adjustment is robust across cities of different sizes,
but intergovernmental grants play a much more pronounced role in maintaining budget balance
for large cities.
Section 6 summarizes the main findings and discusses some of the many directions for
interesting future research that they suggest.
3
Numerous studies have examined fiscal policymaking at the level of stage governments; see, for example, Poterba
(1994), Bohn and Inman (1996), and McCarty and Schmidt (1997). At the municipal level, data on large municipalities
are more readily available and have a principal subject of previous analyses; see, e.g., Inman (1989). An exception is
Holtz-Eakin et al. (1991) who use a sample of 171 municipalities drawn randomly from the Census of Governments.
Large municipalities are clearly of great importance because they account for a large fraction of total municipal fiscal
activity, but, as we shall see below, their behavior differs in significant ways from that of smaller cities.
1118 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132

2. A framework for analysis of fiscal adjustment

In order to examine the process of budgetary adjustment with a minimum of prior restrictions,
we analyze the evolution of fiscal flows like revenues, expenditures, and debt service, as well as
their interrelationship over time, by means of a vector autoregression. Although fiscal
adjustments might in principle take place at any future date, a now-standard approach in
macroeconomic analysis is to fit a time-series model that captures the most significant
interrelationships between the variables with a limited number of lags. If municipalities, on
average, pursue fiscal policies consistent with intertemporal budget balance, the components of
their budgets will display a cointegrating relationship, and, hence, the deficit will be stationary
(e.g., Trehan and Walsh, 1988). In order to model the dynamic adjustment to current changes in
fiscal imbalances, one can exploit this stochastic implication of the intertemporal budget
constraint and employ a vector error-correction framework, relating the change of expenditures,
revenues, and debt service to the lagged deficit. Bohn (1991), for example, conducts such an
analysis of fiscal policy at the level of the US Federal government, and we utilize a similar
approach at the city level. Unlike macro models applied to national governments, however, it is
necessary to recognize that local governments obtain substantial amounts of revenue not only
from own-sources like taxes, but from higher levels of government; in our sample, about 28% of
municipal revenue, on average, is obtained from intergovernmental transfers. Furthermore, as
already pointed out in the Introduction, those transfers may be crucial in restoring the balance of
the budget. We therefore explicitly decompose the revenue side of the budget into own-source
and intergovernmental revenue.
Formally, the empirical analysis focuses on a four-dimensional vector of budgetary
components Y t = ( G t , DSt , R t , Z t )V, where G t is bprimaryQ government expenditure, DSt denotes
current debt service expenditures, R t is own-source revenue, and Z t is intergovernmental
revenue. The current deficit D t is defined as

Dt ubVYt ¼ Gt þ DSt  Rt  Zt ; with bV ¼ ð1; 1;  1;  1Þ: ð1Þ

Following the literature, the empirical model assumes that the current deficit is stationary, and
describes the changes of the elements of the vector Y t as a function of lagged changes of Y t as
well as of its lagged level, i.e. the lagged deficit

Að LÞDY t ¼ cbVY t1 þ ut ; ð2Þ

where D is the difference operator and A (L) is a polynomial in the lag operator. The lagged
deficit term captures the error-correction property of the system, implying that deficits or
surpluses lead to budgetary adjustments reflected in DY t ; the parameter vector c describes the
magnitude (and sign) of the impact of the previous year’s deficit on the current changes in the
budget components and thus the speed of the error-correction process. This approach can be
utilized only if the deficit is stationary, which must be verified empirically. It does not, however,
impose any a priori restrictions on the direction or magnitude of the adjustments of individual
budget components; instead, by estimating these adjustments empirically, the analysis yields
insights about how each of the components of the fiscal policy vector Y t reacts, over time, to
innovations in itself or in one of the other components.
As in a vector autoregressive system, dynamic adjustments to deficit shocks can be described
by impulse–response functions. More specifically, the system can be used to trace the fiscal
adjustment to temporary imbalances, i.e. to surpluses or deficits, which cannot be traced back
 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132 1119

statistically to previous changes in the budget components.4 Graphs of impulse–response

functions offer visual displays showing the time-paths of budgetary adjustment for fiscal
variables (see Buettner and Wildasin, 2003). To summarize the fiscal adjustment process in a
more compact and quantitative fashion, one can compute the present value of the response of
each variable with respect to shocks in every other variable.
Following Bohn (1991) we can state the intertemporal budget constraint in terms of
innovations as (see Buettner and Wildasin (2003) for details)
c t ð DRÞ þ PV
PV c t ðDZ Þ  PV
c t ðDGÞ ¼ Ĝ c t  R̂
G t þ DS R t  Ẑ
Z t; ð3Þ
c t ð DX Þ
where X̂t denotes the innovation in a variable X t , i.e., the change in its expected value. PV
is the expected present value of all changes in this variable in the future. Accordingly, the
innovations in the budgetary components on the right-hand side should evoke an offsetting linear
combination of innovations in the present value of the responses.
More specifically, based on an estimate of system (2), we can compute the present value of
the projected changes of each budgetary component in response to innovations in itself and in
any other budgetary component. Let k( Y[i], Y[ j]) denote the present value of the impulse–
response function of variable Y[ j] given a unit innovation in variable Y[i]. Following Eq. (3) a
unit innovation in each of the budgetary components triggers offsetting responses of the
components of the primary surplus such that
kðY ½i; RÞ þ kðY ½i; Z Þ  kðY ½i; GÞ ¼ b½i; ð4Þ
where b[i] is the i-th component of b defined in (1).
Since Eq. (4) follows from the definition of the intertemporal budget constraint, one might
think of it as an exact relationship. However, aside from possible inconsistencies in the data,
this is not necessarily the case, empirically. Whereas the underlying intertemporal budget
constraint assumes a given interest rate to discount future budgetary flows, the interest rate is
generally not known with certainty, and it may also vary over time. In addition, as discussed
further below, the data display significant variation in the size of municipalities, which
requires scaling fiscal variables in per-capita terms. As a consequence, the appropriate
discount rate is a function of both the interest rate as well as the rate of population growth
and, hence, differs from the interest rate (see Buettner and Wildasin (2003) for details).
Finally, it is possible that intertemporal budget balance, in practice, is achieved over much
longer time periods than the roughly 25-year horizon of our data, and hence could not be
detected in a model with a lag structure of only a few years. For all of these reasons, the
intertemporal budget constraint, applied to the available data, does not hold as an accounting
identity within the context of our empirical model. Despite these qualifications, we shall see
that the adjustment pattern found in the present study follows the predictions of Eqs. (3) and
(4) rather closely.

3. Data and model specification

The empirical analysis employs annual data for individual municipalities from all over the
U.S. obtained from the quinquennial Census of Governments (COG) and the accompanying

4
Note that these temporary imbalances are not bstructuralQ shocks in the terminology of standard VAR analysis,
because a deviation in any one budget component may be accompanied by an immediate adjustment in other budget
components.
1120 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132

Table 1
Definition of fiscal variables
Variable Components (Bureau of Census categories)
(i) Own revenue (R t ) Total taxes, total general charges, total miscellaneous general revenue excluding
interest revenue
(ii) Grants (Z t ) Intergovernmental revenue from federal government and from state governments
(iii) General expenditure ( G t ) Total general expenditure including intergovernmental expenditure net of local
intergovernmental revenue, and excluding interest on general debt
(iv) Debt service (DSt ) Total interest on general debt net of interest revenue
(v) General deficit (D t ) (iii) + (iv)  (i)  (ii)

Annual Survey of Government Finances (ASOGF). In order to trace budgetary adjustments

across time the analysis focuses on a subsample of all cities annually reported in the COG/
ASOGF, yielding a balanced panel for 1270 cities over 26 years from 1972 to 1997 for a total
of 33,020 city-year observations.5 The dataset comprises four fiscal variables, which are
constructed from Bureau of Census fiscal classifications as shown in Table 1.
There are two revenue variables, own-source revenue and intergovernmental revenue
(bgrantsQ) obtained from higher-level governments. There are also two variables on the
expenditure side, general expenditure and net debt-service expenditures. In addition to fiscal
transfers from higher-level governments, small amounts of municipal expenditures and
revenues are payments to or receipts from other local governments; the net amount of these
payments is included as part of general expenditure. Many municipalities hold significant
interest-bearing financial assets, but, since asset values are not always reported in the data, it is
not possible to determine net indebtedness. It is therefore preferable to utilize the flow of net
debt service.
Table 2 presents some descriptive statistics with fiscal variables scaled in terms of
population size. The mean of real per-capita expenditures is $756. There is, however, strong
variation in expenditures, with a standard deviation of $534. The debt service, defined net of
interest earnings, shows a mean around zero. The figures for expenditure correspond to the
mean values on the revenue side, i.e. own revenue of $553 and intergovernmental revenues of
$213. The mean of the residual difference between the first four components (denoted as
general government deficit) is at minus $8, indicating that on average the cities run a small
surplus. However, there is marked variation in the sample between a per capita deficit of as
much as $2464 and a surplus of $1771. This variation in budget outcomes is also reflected in
differences in the debt service, where some cities show high spending whereas others actually
report positive net interest earnings. The fiscal variables show modest mean values of annual
growth in expenditures and revenues, and again substantial variation between rather large
extremes.
The bottom of Table 2 reports statistics for population and income showing that the average
population size is around 75,000. Population size ranges from below 1000 to almost 8 million

5
Although the data are available in digital form, their preparation for analysis is non-trivial, particularly because they
are not coded uniformly across years. The final data have been checked for consistency with state-level aggregates
reported in Census publications. The Census Bureau makes occasional revisions in these data without, however, updating
the publicly available data. Since the revisions by the Census Bureau are not reported, preliminary regressions have been
run to detect influential observations. If a further inspection of these observations revealed apparent inconsistencies with
previous and subsequent observations, the corresponding city was completely removed from the dataset. As result, 76
cities were removed from the 1346 cities in the basic balanced sample.
 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132 1121

Table 2
Descriptive statistics
Mean Std. Dev. Min. Max. Median
Fiscal variables Levels per capita, 1972–1997
Own revenue 0.553 0.399 0.004 7.501 0.445
General expenditure 0.756 0.534 0.005 7.826 0.596
Grants 0.213 0.233 0.000 3.150 0.136
Debt service (net) 0.002 0.051 0.530 1.104  0.000
General deficit 0.008 0.164 1.771 2.464  0.018

Annual change per capita, 1973–1997

Own revenue 0.014 0.102 1.441 1.486 0.009
General expenditure 0.017 0.191 2.348 2.485 0.011
Grants 0.004 0.104 1.412 1.610 0.000
Debt service (net) 0.001 0.032 0.527 0.463  0.001

Other variables 1972–1997

Income in $10,000 per capita 1.920 0.516 0.722 6.737 1.838
Population (in 1000) 74.77 267.1 0.671 7922 31.38
Statistics for pooled observations for 1270 cities in 1996 dollars (deflated with common US GDP deflator). Fiscal
variables in $1000 per-capita.

(New York City) indicating strong variation in the dataset.6 Thus, to model the fiscal adjustment
process, fiscal variables should be scaled with the size of the considered jurisdiction; since
income data are only available at the county level the analysis utilizes per-capita figures.7
However, variations in absolute population size are used in Section 5 to decompose the sample
in order to determine whether there are important differences in the budgetary adjustment pattern
in large and small cities.
Estimation of the system outlined in Section 2 is basically carried out by means of regressions
of the annual changes in each of the budgetary components, i.e. in own revenue, grants, general
expenditures, and in debt service, on the deficit in the previous year and on lagged values of the
changes in each of the budget components, such that the basic set of estimation equation is:
X
p Xp X
p
DGi;t ¼  1 Dt1 þ a10 þ a11;k DGi;tk þ a12;k DDSi;tk þ a13;k DRi;tk
k¼1 k¼1 k¼1

X
p
þ a14;k DZi;tk þ u1i;t
k¼1

X
p X
p X
p
DDSi;t ¼  2 Dt1 þ a20 þ a21;k DGi;tk þ a22;k DDSi;tk þ a23;k DRi;tk
k¼1 k¼1 k¼1

X
p
þ a24;k DZi;tk þ u2i;t
k¼1

6
The population data reported in the COG/ASOGF public use files do not correspond strictly to the year of the fiscal
data. In addition, they are generally not updated on an annual basis. Therefore, the population data have been smoothed
by a moving average using a cubic trend polynomial (Kendall and Stuart, 1976:381f).
7
The income figures report per-capita income for the corresponding county or county area as reported by the Bureau of
Economic Analysis.
1122 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132

X
p X
p X
p
DRi;t ¼  3 Dt1 þ a30 þ a31;k DGi;tk þ a32;k DDSi;tk þ a33;k DRi;tk
k¼1 k¼1 k¼1

X
p
þ a34;k DZi;tk þ u3i;t
k¼1

X
p X
p X
p
DZi;t ¼  4 Dt1 þ a40 þ a41;k DGi;tk þ a42;k DDSi;tk þ a43;k DRi;tk
k¼1 k¼1 k¼1

X
p
þ a44;k DZi;tk þ u4i;t :
k¼1

Note that the basic system is formulated in first differences to take account of possible non-
stationarity of the individual budgetary components.8
Each equation in this system relates the current change in one of the key fiscal variables to
changes in the previous values of all fiscal variables and to contemporaneous shocks. These
shocks, which trigger (possibly very complex) chains of future fiscal adjustments, can arise from
several sources. For example, a positive shock to G might arise from a change in voters’ tastes or
incomes, or a change in demographic structure, that raises the demand for public spending.
Alternatively, positive expenditure shocks could result from mandates imposed by state or federal
mandates (including those imposed by the judicial system). Technological changes (e.g.,
information systems, civil engineering technology) could result in negative expenditure shocks
(due to cost savings) or positive ones (due to the costs of system upgrades or of newly feasible
policies). Similarly, fluctuations in interest rates or financial management shocks (in a celebrated
case, Orange County lost money in the derivatives markets) could cause unpredictable fluctuations
in debt service (resulting in a shock in the second equation); changes in business investment,
housing prices, or employment could produce positive or negative revenue shocks, as could tax-
limitation requirements (the third equation); and new programmatic initiatives by higher-level
governments, fluctuations in state or federal revenues that are shared with municipalities, and
court-ordered fiscal assistance could cause shocks to intergovernmental revenues (the fourth
equation). These sources of fiscal disturbance, as well as many others, require offsetting fiscal
adjustments over time if municipal finances are to maintain long-run balance (which they do, as we
confirm empirically). Our empirical estimates, presented below, show how municipalities make
these adjustments in practice, from the many feasible adjustment paths permitted by the general 4-
equation system. We also investigate some of the possible sources of fiscal disturbances—a
subject, which, however, warrants more attention than we are able to devote to it here.
Estimation of the VECM (2) requires specification of the lag length. Given the limited overall
time dimension of the dataset (26 years), we begin with a lag of 4 years in the differenced data,
subsequently testing for possible reductions in the number of lags. As shown in Table 3, a
reduction of the lag length is always rejected. This suggests employing a model with four lags.9

8
Unit root tests have been carried out using a statistic suggested by Im et al. (2002). As expected, non-stationarity of
the levels cannot be rejected, but stationarity is not rejected for the deficit and for the first differences of all of the four
budgetary components (details presented in Buettner and Wildasin (2003)).
9
Estimates of models with 5 and 6 lags (available upon request) did not show major differences in the adjustment
pattern.
 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132 1123

Table 3
Specification tests
Lag length 2 3 4
Indiv. eff (v 2 (5076)) 4019 4186 4437
Lag order reduction (v 2 (16)) 1518 711.0 571.7
Likelihood-ratio statistics on cross-equation restrictions.

Typically, panel data studies allow for individual effects capturing differences in the
characteristics of individual units.10 The following analysis deals essentially with first
differences of fiscal flow variables, and, in this respect, will not be affected by cross-sectional
differences in local characteristics. The fiscal deficit variable, however, is entered in levels. The
presence of individual effects would imply that the jurisdictions converge to different (per-
capita) deficit levels. Comparing estimation with and without individual effects it turns out that
joint tests reject the presence of individual effects, regardless of lag lengths (see Table 3).11 This
indicates that cities are commonly converging toward the same level of deficit. As no indication
of individual effects is found, it is appropriate to estimate individual equations of the system (2)
separately with OLS; in this case, joint estimation does not improve efficiency as the set of
regressors is the same across equations.12

4. Empirical results

Since the system is a four-dimensional vector error-correction model, estimation produces a

large number of parameters. Central parameters are the coefficients of the error-correction term
in the individual equations. As shown in Table 4, the results clearly confirm convergence toward
the intertemporal budget constraint, since a higher deficit shows a positive impact on own
revenue and on grants received, whereas a higher deficit shows a negative impact on
expenditures. The positive impact on debt service is consistent with the fact that the deficit
results in a rise in debt levels and thus creates higher debt service in the subsequent period.
Given a constant rate of interest, and in the absence of population growth, the coefficient of the
deficit in the debt service equation should reflect the real interest rate.13
One way to trace the estimated adjustment pattern resulting from the complete model is to
compute impulse–response functions showing how the necessary adjustment actually takes

10
The literature on dynamic panel data has emphasized bias of standard panel data approaches in the presence of lagged
endogenous variables and suggests the use of instrumental techniques (e.g., Holtz-Eakin et al., 1991). With the rather
long time period available in our sample, the Nickell (1981) bias should not be a significant problem, and it is neglected
in the tests for the presence of individual effects.
11
Testing is carried out using individual fixed effects for all equations since Hausman tests rejected the use of a random
effects model for the own-source revenue and expenditure equations.
12
Avery (1977) has emphasized that in the presence of individual error components, estimation of individual equations
separately is not efficient and proposed simultaneous estimation techniques (see, also, Baltagi, 1995: 106 pp).
13
With a constant rate of interest r, and denoting the rate of change in population with n, the change in debt service per-
capita can be expressed as
   
DS t n DBt n Bt1
D ¼ 1 r  r :
Pt 1 þ n Pt1 1 þ n Pt1
As population growth shows an average rate of 0.98% in the sample, the impact of a change in debt per capita on the debt
service is less than proportional to the real interest rate (even with a constant interest rate this relationship is only an
approximate one, because of the difficulty of measuring government assets and liabilities).
1124 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132

Table 4
Estimates for the error-correction term
Equation g (Std. err.)
Own revenue .098 (.013)
Gen. expend. .297 (.018)
Debt service .013 (.003)
Vert. grants .069 (.009)
Heteroscedasticity robust standard errors in parentheses.

place given an initial fiscal disturbance. The proper interpretation is that the impulse
responses trace the adjustment to a fiscal imbalance hypothetically arising from changes in
the individual budget components. Later, we discuss the possible sources of these
imbalances.
A convenient way to summarize the impulse responses is to calculate the total response to
temporary imbalances in present value terms, as outlined in Section 2.14 In order to calculate
the present values, we fix the discount rate at 3%.15 The columns of Table 5 show the long-
run response given unit innovations in per-capita values of the fiscal variables, expressed in
present value terms. The table also displays standard errors obtained by sampling from the
normal joint distribution of the VECM estimates and computing the corresponding
distribution in the impulse–response functions as suggested by Sims (1987) and Hamilton
(1994, 337).16
It is instructive to consider the findings reported in Table 5 from two different perspectives.
Reading down the columns of the table shows how innovations in any one fiscal variable affect
the subsequent adjustments of itself and the other variables. Reading across the row for any one
fiscal variable shows how responsive it is to changes in its own value or in that of other fiscal
variables.
Consider first the own revenue column. This column shows how a $1 increase or
decrease in revenues in one period affects the subsequent evolution of expenditures,
intergovernmental transfers, debt service, and revenues themselves, all expressed in present
value terms. To illustrate, suppose that revenues increase by $1 resulting in a current deficit
reduction (or current surplus increase) of $1. What kind of subsequent fiscal adjustment can
we predict on basis of our statistical model? The revenue column shows that an innovation
to own revenue by $1 is followed by an increase in future expenditures by 51 cents and to
reductions of own revenue and grants of 35 and 9 cents, respectively, all measured in present
value terms.
Since part of the adjustment to a change in each of the fiscal variables takes the form of an
offsetting change in its own future value, it is also instructive to assess the response to a
permanent $1 increase in each variable. Dividing by the permanent component of the innovation
in own revenue (1  .348), it turns out that 78 cents of a permanent increase in own revenue by

14
Detailed estimates of the VAR system are available of the Journal of Public Economics website. The precise time-
path of adjustment can be illustrated graphically by plotting impulses-response functions (Buettner and Wildasin (2003)).
15
Probably due to the fact that most of the adjustment takes place in the periods, the qualitative results are not sensitive
to the actual value of the discount rate.
16
Sims (1987) argues that a possible deficiency of this approach is that it ignores the randomness of the estimated
covariance matrix of the errors. However, in the current context, this estimate is obtained from a large cross-section as in
seemingly unrelated regression analysis. Note that the sampling is carried out using a heteroscedasticity consistent
estimate of the variance–covariance matrix of the VECM.
 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132 1125

Table 5
Present value responses
Response Innovation to
Own revenue Gen. expend. Grants Debt service
Own revenue  .348 (.026) .162 (.019)  .144 (.023) .145 (.037)
Gen. expend. .508 (.027) .716 (.020) .338 (.027) .370 (.037)
Grants  .086 (.012) .082 (.010)  .473 (.017) .049 (.016)
Debt service  .005 (.005) .019 (.004)  .015 (.004) .387 (.014)

Response to permanent increase

Own revenue .571 (.040)  .273 (.044) .236 (.059)
Gen. expend. .780 (.021) .641 (.043) .604 (.063)
Grants  .131 (.019) .287 (.033) .079 (.026)
Debt service  .008 (.008) .068 (.014)  .028 (.008)
Standard errors in parentheses obtained by sampling from the normal joint distribution of the VECM estimates based on a
heteroscedasticity consistent estimate of the variance–covariance matrix.

$1 is translated into higher spending (cf. the bottom panel of Table 5), whereas
intergovernmental grants are reduced by 13 cents.
Following the predictions of Eq. (4), the innovations in each of the budgetary components
will be fully balanced by the present value of changes in own revenue, grants, and
expenditures, which make up the primary surplus. For example, summing across the first three
rows in the first column, an additional dollar of own revenue is estimated to result in an
offsetting change of 94 cents in the primary surplus. Computing the present value of
adjustments in the primary surplus to innovations in expenditures and grants yields similar
figures of $0.959 and $0.955, respectively. For innovations in debt service, the sum of the
present value of changes in expenditures, revenues, and grants is much lower ($0.564). But,
future changes in the debt service play a major role in balancing the budget, indicating strong
temporal fluctuations in the debt service. With regard to permanent increases in the debt
service by $1, the present value response of the primary surplus amounts to $0.919. Given that
the intertemporal budget constraint holds only approximately in empirical data, as the true
interest rate, its time path, and the amount of non-interest bearing assets of the municipalities
are not known, these figures are indicative of reasonable properties of the empirical model of
municipal fiscal adjustment.
But the results also indicate that jurisdictions do not respond solely with the components of
their primary surplus to innovations in budgetary components. A small but statistically
significant fraction of additional grants, 2.8 cents out of a permanent increase of $1, is used to
lower the debt burden. In addition, an increase in expenditures is followed by an increase in debt
service by 6.8 cents per dollar of additional permanent expenditures.
Generally, the results show that innovations in the components of the budget tend to be partly
offset by future changes in the same component. This is particularly true for expenditures, where
more than two thirds of a change are balanced with an offsetting change in the present value of
future expenditures. Considering permanent innovations in budgetary components, Table 5
displays a key role of expenditures for fiscal adjustment, where we find that three quarters of
each dollar in additional own revenue and almost two thirds of each dollar in additional grants
show up in the form of added spending. Changes in debt service also have much larger effects on
expenditures than on own revenue. Nevertheless, smaller but still significant parts of the
adjustment are obtained by changes in own revenue and grants, in the sense that lower revenues
1126 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132

Table 6
Significance of conditioning variables
Conditioning variables Equations
Own reven. Gen. expend. Grants Debt serv.
Period-specific effects 399 (20)* 346 (20)* 581 (20)* 970 (20)*
Predicted change in tax revenuea 292 (1)* 45.3 (1)* 1.06 (1) 0.05 (1)
Predicted change in expenditurea 211 (1)* 396 (1)* 123 (1)* 1.18 (1)
Predicted change in fed. grantsa 0.38 (1) 14.9 (1)* 117 (1)* 0.82 (1)
Change in employmenta 0.01 (1) 0.34 (1) 1.68 (1) 0.61 (1)
Change in incomea 8.56 (1)* 2.49 (1) 4.54 (1)* 3.69 (1)
Likelihood-ratio statistics for restricting the respective set of conditioning variables to zero. aPeriod-specific effects
included as further conditioning variables. Significance at the 5% level is marked with a star, degrees of freedom in
parentheses.

and higher expenses are balanced with significant future increases in own revenue as well as
grants.17
So much for the interpretation of the columns of Table 5. Next, reading across the rows in
Table 5, one can see that each fiscal variable adjusts in the expected direction to innovations in
the others, but by varying degrees. Own revenue, for example, adjusts more strongly to an
innovation in expenditures than to an innovation in grants or debt service. The third row,
showing the response of grants, is of particular interest. Many theoretical and empirical studies
highlight the role of intergovernmental transfers as instruments through which higher level
governments can influence the behavior of recipient governments. However, as recent
discussions of soft-budget constraints have emphasized it may also be the case that recipient
governments can induce higher fiscal transfers from donor governments through their own
policy actions. The third row of the table shows that fiscal transfers from higher-level
governments do indeed respond quite significantly to innovations in municipal own revenues
and expenditures, but not very strongly to debt service burdens.
The results in Table 5 provide a picture of a coherent fiscal adjustment mechanism, showing
how cities adapt, in accordance with the constraint of long-run fiscal balances, to unpredicted
changes in key fiscal variables. Our analysis has left open, however, what the sources of these
unpredicted changes actually are. Indeed, each such change might itself be the result of
underlying exogenous shocks to local economic conditions, macroeconomic conditions, or any
number of other factors. In order to clarify the sources of shocks to municipal fiscal variables, we
include some additional conditioning variables in our basic model. Table 6 reports likelihood-
ratio statistics, which provide a natural way to summarize the gain in the predictive power from
the inclusion of additional variables.18 In the first row of Table 6 we report statistics for the
inclusion period-specific effects. The effects would include macroeconomic conditions such as

17
It may be of interest to relate the response to an innovation in grants to the discussion of the bflypaperQ effect,
according to which the public sector has a high propensity to spend out of grants is (for overviews see Gramlich, 1977,
and Hines and Thaler, 1995). The results in Table 5 indicate that the response in spending to a permanent increase in
grants by one dollar amounts to 64 cents, which generally is in accordance with the results in the literature (see Hines and
Thaler, 1995). This results differs, however, from Holtz-Eakin et al. (1991) who use a panel VAR (with only large
municipalities) to estimate the relationship between the levels of fiscal variables, and do not find a positive effect of an
innovation in grants on spending.
18
More precisely, the likelihood-ratio statistics indicate whether implicit restrictions in the basic, unconditional model
can be rejected on statistical grounds.
 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132 1127

GDP growth, unemployment, and financial market conditions such as interest rates.19 We see
that period-specific effects do partly account for the unpredicted changes in fiscal variables to
which municipalities must adjust; this is especially true for the debt-service equation, probably
indicating the importance of fluctuations in interest rates as determinants of municipal borrowing
costs.
As we have seen, municipal own source and intergovernmental revenues trigger significant
adjustments in municipal finances. There may be also common factors that affect the evolution
of revenues for all municipalities nationwide and that are thus exogenous to individual cities.
These nationwide trends will affect individual municipalities differently, depending on their
individual revenue structures and, thus, act as exogenous fiscal shocks. To examine the
relationship with the fiscal imbalances empirically, we use detailed data on the budget structure
in order to compute averages of annual nationwide trends in the components of tax revenue and
federal grants, weighted by the specific share in tax revenue or federal grants received by the
individual municipality. Inclusion of these variables enables us to determine the extent to which
nationwide trends in these sub-categories of the budget account for contemporaneous
innovations in municipal budget.20
As shown in Table 6, the national trends in municipal tax revenues have a large impact on the
own-source revenues of individual cities; additionally, they have a significant though more
modest impact on expenditures. It is notable that it has no significant effect on grants or
municipal debt service. Similarly, the national trends in federal grants have a large impact on
municipal grant revenues, but no strong effects on other municipal fiscal variables. These results
indicate that national trends in tax revenues and federal grants are reasonable proximate
determinants of innovations in the revenues of the individual cities. As we have seen in Table 5
these innovations have large and persistent effects on all components of municipal budgets;
common national trends in individual revenue components thus emerge as important shocks,
exogenous to individual cities, to which municipal finances must adjust.
One might also expect that local fiscal imbalances, including innovations in local revenues,
depend importantly on local economic and demographic conditions (in addition to other possible
idiosyncratic factors). To investigate this possibility, we use county-level data on the change of
employment and personal income, both in per-capita terms.21 As shown in the last two rows of
Table 6, these variables have only weak predictive power; in fact, changes in employment
demonstrate no significant effect at all. For changes in income we find some limited though
significant effects on municipal revenues.

19
Note that we do not use period-specific effects in the basic equation, as this would imply to model only adjustments to
idiosyncratic innovations, although the intertemporal budget constraint requires adjustments to all innovations. Moreover,
the inclusion of period-specific effects would tend to limit the comparability between the results for different subsamples
as carried out below.
20
If DX i,t denoted the annual change in one of the budgetary components, its prediction can be obtained as a weighted
average of the national trends in each of the sub-categories of the COG/ASOGF data making up the budget components
where the weights are the shares of each sub-category. Formally
Xn
d
DX i;t ¼ a ji;t1 DX jt ;
P

j¼1

where a ji,t1 is the share of sub-category j in the respective budget component in municipality i and DX̄ tj is the national
per-capita change in component j. The predicted change in tax revenue is an average of national trends in each of the 18
different types of taxes reported in the data. The predicted change in federal grants distinguishes 12 sub-groups.
21
The data has been taken from the Bureau of Economic Analysis.
1128 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132

To summarize, the analysis shows that imbalances in municipal revenue, whether own-source or
intergovernmental, trigger significant adjustments in all components of municipal budgets, and
that these imbalances are importantly influenced by common national trends, exogenous to
individual cities. Somewhat surprisingly, shocks to local economic variables have only small
impacts on municipal fiscal imbalances.

5. Fiscal adjustment and city size

Our analysis so far has imposed the assumption that all U.S. municipalities follow a common
fiscal adjustment process, and, indeed, the empirical findings of Section 4 lend support to this
view. The institutional environments within which different cities choose their fiscal policies are
not necessarily all the same, however. For example, large cities, which have been the focus of
most previous studies of municipal finances, may face quite different administrative and political
constraints than smaller ones. On the one hand, they may be able, effectively, to lobby higher-
level governments for fiscal assistance or other special treatment. On the other hand, all states
distinguish cities into size classes for purposes of state statutes and administrative regulations
concerning fiscal policies, personnel management and staffing rules, financial accounting and
management procedures, and other controls.22 Since municipalities of different sizes may thus
operate in quite different, fiscal, political, and regulatory regimes, it is natural to wonder to what
extent our estimates of fiscal adjustment parameters are robust with regard to city size.
Results from estimating the system separately for large and small cities are reported in Table
7. (Results for cities in the 25th to 75th percentiles by population size are omitted to save space,
but are generally in accordance with those reported for the entire sample.) With regard to the
innovations on the revenue side the results are generally similar to those above. An innovation in
own revenue is offset to a slightly greater extent by future revenue reductions for small cities.
Moreover, the response to an innovation in grants differs by city size: in small cities, own
revenues fall more, and expenditures rise less, as compared to large cities. Generally speaking,
however, the results concerning innovations in own-source and intergovernmental revenue
confirm the findings for the entire sample.
The situation is different when it comes to innovations arising on the expenditure side. As
reported in the second column, small cities raise own revenue more in response to an innovation
in expenditures than do large jurisdictions. For example, small jurisdictions respond to a
permanent expenditure increase of $1 with an increase in own revenue of about 67 cents,
whereas large jurisdictions raise their own revenue by only 38 cents. By contrast, large cities rely
much more on transfers from higher-level governments to finance permanent increases in
expenditures. For small municipalities, only about 18 cents out of a dollar, or 18%, of a
permanent increase in spending is financed by increased grants, whereas the corresponding

22
Cities of the bfirst classQ, for example, are those with the large populations, the next size category defines the cities of
the bsecond classQ, and so forth. The preamble to the Pennslyvania Intergovernmental Cooperative Authority Act for
Cities of the First Class, a law which set up elaborate financing (a bbailoutQ) and fiscal control mechanisms for
Philadelphia during its fiscal crisis in the 1990s, exemplifies the potential importance of city size. bIt is hereby declared to
be a public policy of the Commonwealth . . . to foster the fiscal integrity of cities of the first class to assure that cities
provide for the health, safety, and welfare of their citizens; pay principal and interest owed on their debt obligations when
due; meet financial obligations to their employees, vendors, and suppliers; and provide for proper financial planning
procedures and budgeting practices.Q bCities of the first classQ in Pennsylvania are those with populations in excess of 1.5
million. Philadelphia is the only city in this class; the second largest city in the state is Pittsburgh, with a population of
less than .4 million.
 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132 1129

Table 7
Decomposition with respect to city size
Response Innovation to
Own revenue Gen. expend. Vert. grants Debt service
Small cities (bottom quartile)
Own revenue .420 (.047) .204 (.040) .188 (.049) .306 (.082)
Gen. expend. .443 (.049) .696 (.039) .262 (.051) .319 (.084)
Vert. grants .075 (.023) .056 (.018) .502 (.029) .018 (.034)
Debt service .002 (.008) .015 (.006) .012 (.007) .337 (.027)

Response to permanent increase

Own revenue .673 (.070) .378 (.097) .462 (.117)
Gen. expend. .765 (.044) .525 (.094) .482 (.129)
Vert. grants .130 (.040) .184 (.059) .027 (.051)
Debt service .004 (.014) .050 (.020) .025 (.014)

Large cities (top quartile)

Own revenue .320 (.062) .115 (.031) .132 (.039) .058 (.079)
Gen. expend. .511 (.069) .696 (.039) .404 (.057) .298 (.084)
Vert. grants .112 (.026) .148 (.023) .424 (.040) .180 (.037)
Debt service .014 (.010) .029 (.007) .033 (.008) .408 (.025)

Response to permanent increase

Own revenue .380 (.075) .230 (.071) .098 (.132)
Gen. expend. .752 (.048) .702 (.069) .503 (.144)
Vert. grants .165 (.044) .487 (.060) .304 (.063)
Debt service .021 (.015) .096 (.026) .057 (.015)
Sample decomposition based on the quartiles of the long-run distribution of population. Small cities have populations
between 1 and 15 thousand, while large cities have populations between 63 thousand and 7.4 millions. Standard errors in
parentheses obtained by sampling from the normal joint distribution of the VECM estimates based on a
heteroscedasticity-consistent estimate of the variance–covariance matrix.

figure for large municipalities is almost 49%. It is interesting to compare these adjustment
responses to the average shares of expenditures financed by grants. Larger jurisdictions, on
average, depend more on grants: grants as a share of expenditures are 31% for large cities and
only 26% for small cities. For large cities, then, innovations in municipal expenditures are even
more substantially financed by increased intergovernmental transfers than the already higher
average share of grants in total expenditures suggests. Conversely, for smaller cities, innovations
in municipal expenditures are even less substantially financed through transfers than is indicated
by the relatively low share of grants in total expenditures. Expressed somewhat differently, the
bmarginalQ response of grants to innovations in expenditures magnifies the existing baverageQ
differential importance of grants in the finances of cities in different size categories.
The response of fiscal transfers to an innovation in debt service is particularly noteworthy. We
find that small cities tend to respond by raising own revenue, whereas large cities do not. Instead,
innovations in debt service give rise to large increases in grants for large cities, where almost a
third of the innovation is offset by transfers from higher-level governments. For other cities this
response is negligible. Evidently, grants play an unusually important role in the fiscal adjustment
process for large cities.
Taken together, these results suggest that our principal findings from Section 4 are robust with
respect to city size. Table 5’s summary statistics about the fiscal adjustment of municipalities in
response to fiscal shocks on the revenue side are essentially confirmed. On the expenditure side
1130 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132

we do see, however, some important differences. Expenditure and debt service innovations for
large cities tend to be much more offset by intergovernmental transfers, whereas for small cities
the main adjustment occurs through changes in own revenues.
These differences in the dynamics of fiscal adjustment for large and small cities cannot
conclusively demonstrate that large cities operate under soft budget constraints. They do,
however, demonstrate that the fiscal adjustment process for large cities, especially in its
intergovernmental dimensions, differs from that for small cities, suggesting that the institutional
structure of intergovernmental fiscal relations also vary by city size. One implication of this
finding is that empirical analysis of large cities, useful though it may be because of their
quantitative importance, may lead to findings that are unrepresentative of municipalities in
general.

6. Conclusion

The preceding analysis has investigated empirically the dynamic fiscal policy adjustment of
U.S. municipalities using a vector error-correction model which takes account of government
intertemporal budget constraints. The results point to an important role of expenditures in
maintaining intertemporal budget balance, especially in response to shocks in own-source and
intergovernmental revenues. We find that an additional dollar of own-source revenue gives rise
to 78 cents of additional expenditures, expressed in present value terms. In response to an
additional dollar of grant revenue, expenditures rise by 64 cents, in present value. However,
revenue-side adjustments are significant, too; it is particularly interesting to note that grants from
higher-level governments are quite sensitive to municipal fiscal imbalances. A sample
decomposition shows that these patterns are generally robust with respect to city size. However,
it is noteworthy that small cities tend to rely more on own-source revenue whereas additional
revenue from grants play a much larger role in restoring budget balance for large cities—much
larger than the budget share of grants in the finances of these cities would suggest. Thus,
intergovernmental transfers seem to bcushionQ the process of fiscal adjustment for municipalities
generally. Since municipalities use bexternalQ funds to balance their budgets, it is possible that
this apparent softening of budget constraints distorts local policy decisions. If so, our results
suggest that this effect may be particularly relevant for larger cities.
The statistical methods that we have used do not test for the importance of any particular
institutional, economic, or other determinants of municipal fiscal adjustment, but rather shed light
on the empirically-relevant contours of the underlying institutions as revealed in the dynamic
adjustment process. Our findings have identified important empirical relationships, not
previously discerned, the explanation of which presents new challenges and opportunities for
future research. For example, why should expenditures and revenues play different roles in the
dynamic adjustment process for municipalities? Since expenditure levels seem to be heavily
influenced by changes in revenues, what does this imply about the impact of state-level regulatory
constraints on the types of revenue instruments available to municipalities, or limitations on their
utilization? Would these types of policies or institutional constraints have more significant
impacts on municipal spending than, say, budget oversight or review agencies which, at first
glance, might appear to be more directly related to expenditure policy? Why should
intergovernmental transfers interact with municipal finances as they do? Why should fiscal
adjustment for large cities differ from that of small cities? As indicated in some of the preceding
discussion, there are many interesting hypotheses that might be examined in an attempt to explain
these and other empirical results revealed in our analysis. An enhanced view of the empirical
 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132 1131

landscape should be of value in discriminating among competing theoretical models of local

government policymaking in a federal structure, and, ultimately, in understanding better the
complex institutional structures, interacting with underlying economic, demographic, and
technological fundamentals that produce the observed dynamic fiscal adjustment process.

Acknowledgements

We are indebted to the editor and referees for helpful comments on an earlier version of this
paper, and to R. Graycarek for research assistance. Much of this work was undertaken while the
first author was a postdoctoral scholar at the University of Kentucky, whose support is gratefully
acknowledged.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at
doi:10.1016/[Link].2005.09.002.

References

Administrative Office of the US Courts, 2000a. Bankruptcy Basics, Revised second edition Administrative Office of the
US Courts, Bankruptcy Judges Division, Washington.
Administrative Office of the US Courts, 2000b. Annual Report of the Director. Administrative Office of the US Courts,
Washington.
Avery, R.B., 1977. Error components and seemingly unrelated regressions. Econometrica 45, 199 – 209.
Baltagi, B.H., 1995. Econometric Analysis of Panel Data. Wiley, New York. et al.
Bohn, H., 1991. Budget balance through revenue or spending adjustment. Journal of Monetary Economics 27, 333 – 359.
Bohn, H., Inman, R.P., 1996. Balanced-budget rules and public deficits: evidence from U.S. States. Carnegie-Rochester
Conference Series on Public Policy 45, 13 – 76.
Buettner, T., Wildasin, D.E., 2003. The Dynamics of Municipal Fiscal Adjustment. CESifo working paper.
General Accounting Office, 1995. Background Studies on Boston, Cleveland, New York. unpublished.
Gramlich, E., 1977. Intergovernmental grants: a review of the empirical literature. In: Oates, W. (Ed.), The Political
Economy of Fiscal Federalism. Heath, Lexington.
Hamilton, J.D., 1994. Time Series Analysis. Princeton University Press, Princeton.
Hines, J., Thaler, R., 1995. The flypaper effect. Journal of Economic Perspectives 9, 217 – 226.
Holloway, G.M., 1996a. District Government: Information on its Fiscal Condition and the Authority’s First Year of
Operations. GAO report T-AIMD-96-126.
Holloway, G.M., 1996b. District Government: Information on its Fiscal Condition. GAO report T-AIMD-96-133.
Holtz-Eakin, D., Newey, W., Rosen, H., 1991. The revenue–expenditure nexus: evidence from local government data.
International Economic Review 30, 415 – 429.
Im, K.S., Pesaran, M.H., Shin, Y., 2002. Testing for unit roots in heterogeneous panels. Journal of Econometrics 115,
53 – 74.
Inman, R.P., 1989. The local decision to tax. Regional Science and Urban Economics 19, 455 – 491.
Kendall, M., Stuart, A., 1976. The Advanced Theory of Statistics: III. Hafner, New York.
McCarty, T.A., Schmidt, S.J., 1997. A vector autoregression analysis of state–government expenditure. American
Economic Review 87 (2), 278 – 282.
National Bankruptcy Review Commission, 1997. Bankruptcy: The Next Twenty Years. National Bankruptcy Review
Commission, Washington.
Nickell, S., 1981. Biases in dynamic models with fixed effects. Econometrica 49, 1417 – 1426.
Poterba, J.M., 1994. State responses to fiscal crisis: the effects of budgetary institutions and policy. Journal of Political
Economy 102, 799 – 821.
Sims, C.A., 1987. Comment on D. E. Runkle, vector autoregression and reality. Journal of Business and Economic
Statistics 5, 443 – 449.
1132 T. Buettner, D.E. Wildasin / Journal of Public Economics 90 (2006) 1115–1132

Trehan, B., Walsh, C.E., 1988. Common trends, the government’s budget constraint, and revenue smoothing. Journal of
Economic Dynamics and Control 12, 425 – 444.
Trehan, B., Walsh, C.E., 1991. Testing intertemporal budget constraints: theory and applications to U.S. Federal budget
and current account deficits. Journal of Money, Credit, and Banking 23, 206 – 223.
Wilcox, V., 1989. The sustainability of government deficits: implications of the present value borrowing constraint.
Journal of Money, Credit, and Banking 21, 291 – 306.