State fiscal policies and transitory income fluctuations.
Hines, James R., Jr.
ABSTRACT State and local expenditure and tax revenue respond less
to the business cycle than do federal spending and revenue, thereby
reducing the countercyclicality of total government expenditure and
revenue. This paper considers forces responsible for the cyclical
pattern of state expenditure and revenue. Annual fluctuations in state
personal income are associated with small changes in state spending and
significant changes in tax receipts; receipt of federal grants is
associated with greater state spending. Tax collections, and to a lesser
degree expenditure, of larger states are more closely associated with
annual income fluctuations than are the tax collections and expenditure
of smaller states. These state size differences may proxy for other
state characteristics, such as the extent to which a state faces
interstate competition for mobile businesses and individuals, and the
quality of state government. The spending and tax revenue of states with
less mobile populations closely track income fluctuations, as does
spending in states where convictions of public officials for federal
corruption crimes are more common. In small states, and in states with
more mobile populations and better corruption records, government
expenditure and revenue appear to rise and fall less with income, and in
that respect more closely resemble the federal government.
**********
In the United States, fiscal policy is the province not only of the
federal government but also of 50 state governments and their local
governments. Collectively, these nonfederal governments accounted for
roughly 40 percent of total government expenditure in 2007, and an even
larger fraction of total government revenue. Since states and
municipalities operate independently of the federal government and of
each other, it follows that a significant portion of U.S. fiscal policy
is uncoordinated. As a result, any deliberate countercyclical fiscal
policy conducted by the federal government has the potential to be
affected, possibly even undermined, by the actions of state and local
governments.
The federalist structure of U.S. fiscal policy would be of little
macroeconomic consequence if states and localities behaved just like
smaller versions of the federal government, but there is little reason
to expect them to, and considerable evidence that they do not. States
and localities differ from the federal government and from each other in
the economic functions they serve, their average level and composition
of expenditure, the extent to which they are subject to national
economic shocks, and their beliefs about the role of government in
managing the economy. Perhaps most important, state and local
decisionmaking is driven by interests and constraints that differ from
those of the federal government: states have incentives to pursue
policies that benefit their own stakeholders, and they are subject to
competition from other states.
The aggregate evidence for the postwar period suggests that during
economic downturns, state and local tax collections decline less than do
federal tax collections, and state government spending expands less than
does federal spending. This general pattern has been mirrored in the
recent U.S. experience. For example, in 2007 the federal government
collected $1.638 trillion in taxes other than social insurance
contributions, a figure that declined substantially in the recent
recession, to $1.142 trillion in 2009. State and local tax collections
also declined, but much more gradually: from $1.314 trillion in 2007 to
$1.267 trillion in 2009. Federal government spending of $3.458 trillion
in 2009 was $558 billion greater than in 2007, whereas state and local
spending of $2.026 trillion in 2009 exceeded 2007 spending by just $115
billion. (1) Changes in federal expenditure and in state and local
expenditure reflect policy changes as well as the changing incomes of
the population. In addition, federal grants-in-aid to state and local
governments rose by $104 billion between 2007 and 2009: removing this
component leaves self-financed aggregate state expenditure only $11
billion larger in nominal terms, and smaller in real terms, during 2009
than in 2007. It is clear that state and local governments reacted to
the recession that began at the end of 2007 very differently than did
the federal government.
There are reasons to expect states not to expand their spending
during recessions. Almost all state governments have annual balanced
budget requirements, with only limited exceptions for bad economic
times, so the declining tax revenue that characterizes most economic
downturns is likely to be accompanied by reduced spending. These
balanced budget requirements certainly do not prevent states from
financing additional spending with greater borrowing, as evidenced by
the fact that many states do borrow, but the requirements raise the
political and administrative costs of running deficits, since they
commonly require that state legislatures enact extraordinary measures to
undertake the necessary borrowing. States with balanced budget
requirements can maintain or expand spending across the cycle by running
budget surpluses during good economic times, which can then be drawn
down during bad economic times, but in practice it has proved difficult
for many states to conduct their fiscal affairs that way. Finally, the
adoption of balanced budget requirements reflects a shared expectation
among states that they will not make extensive use of debt finance to
combat adverse economic developments.
A second reason why state governments might fail to increase
expenditure during economic downturns is that the tax revenue reductions
that accompany economic contractions generally discourage spending even
in the absence of a requirement that budgets be balanced. It has been
widely documented that the expenditure of subnational governments is
significantly affected by cash windfalls, whether positive or negative;
this is known as the "flypaper effect": money tends to stick
where it hits. The flypaper effect was first observed in the context of
intergovernmental grants, typically those from the federal government to
state governments, which appeared to have a much greater impact on state
spending than can be easily reconciled with the actual income and
substitution effects of the grants.
There is considerable controversy over the origins of the flypaper
effect, its magnitude, and whether it reflects political dynamics or
more fundamental aspects of the way that individuals make decisions. (2)
But whatever its sources, the flypaper effect describes an empirical
regularity: money in hand and available is more likely to be spent than
is money that is not quite as readily available, even though it is fully
obtainable. Although the flypaper effect is often interpreted narrowly
to apply to specific categories of revenue and expenditure, a more
general view (see, for example, Hines and Thaler 1995) is that it
reflects a feature of human nature that applies broadly to many revenue
sources and expenditure categories, and results from the difficulty that
people have in not spending resources that are available to them. Thus,
for example, Olivier Blanchard, Florencio Lopezde-Silanes, and Andrei
Shleifer (1994) report that firms obtaining cash windfalls from
victorious lawsuits significantly increased their expenditure on
acquisitions of other firms; and in the context of state reactions to
economic fluctuations, Douglas Holtz-Eakin, Whitney Newey, and Harvey
Rosen (1989) find that local revenue shocks are associated with
subsequent spending increases by municipal governments. Any natural
tendency of state governments to spend money that they have on hand
reduces the likelihood that states will pursue countercyclical fiscal
policies. (3)
Against these considerations, however, must be set the potential
benefits of greater spending, including the avoidance of draconian
budget cuts, during economic downturns. To the extent that tax cuts and
greater government spending during recessions facilitate the employment
of underutilized resources within a state, they have the potential to
promote the welfare of state residents? Apart from any desire to put
underutilized resources to work, states that can avoid damaging, cash
flow-driven budget cuts during bad economic times have the potential to
improve their own business conditions and property values. This may
matter more for smaller (that is, less populous) states than for larger
states: since they tend to face the most elastic populations of
businesses and individuals, they may have the strongest incentives to
avoid boom-and-bust fiscal cycles that, to the extent that they are
inefficient or reduce welfare, could result in declining populations,
incomes, and property values.
States differ in the extent to which their levels of expenditure
and taxation vary over the business cycle. Changes in the expenditure
and tax revenue of larger states are more positively correlated with
state income fluctuations than are those of smaller states. This fiscal
pattern of large versus small states may reflect the pressures that all
governments face to spend money during the good times when it is
available, and reduce expenditure at other times; the somewhat greater
immobility of population and business activity in large states arguably
makes it more feasible for their governments to follow such spending
patterns than it is for smaller jurisdictions. Expenditure and tax
revenue in states with less mobile populations (where mobility is
measured by the fraction of adults born in a state who subsequently
reside there), and in states whose public officials are frequently
convicted of federal corruption crimes, exhibit correlations with income
fluctuations similar to those of large states. This suggests that state
size might proxy in these regressions for other attributes of larger
states, such as relatively more captive populations of individuals and
businesses, and perhaps lower quality of government.
Section I of this paper uses quarterly national income accounts
data (available for states only as an aggregate) from 1947 to 2010 to
identify the extent to which state and local tax revenue and spending
change during national income fluctuations. This evidence suggests that
the 2007-09 experience is unusual only in its magnitude: real federal
tax collections per capita tend to decline in years in which the economy
performs poorly, whereas state and local tax collections are much more
stable; furthermore, federal government spending tends to rise more
rapidly during bad economic times than it does during good times, a
pattern that is much less evident among state and local governments.
Section II of the paper uses Census of Governments data to identify
the determinants of state tax and spending patterns, distinguishing
between large and small states. States exhibit marked propensities to
spend out of federal grant dollars, and larger states display tax
revenue and expenditure patterns that are more closely associated with
income fluctuations than are those of smaller states. Section III
concludes.
I. Aggregate Patterns of Federal and State Taxes and Spending
Expenditure by the federal government accounts for about 20 percent
of aggregate national expenditure in virtually every year since World
War II. Figure 1 depicts ratios of real federal spending and of real
state and local spending to real U.S. GDP, as reported in quarterly
national income account data from 1947Q 1 to 2010Q 1. (5) It is
noteworthy that federal spending as a fraction of GDP appears to rise
with the onset of recessions, particularly since the 1960s. Doubtless
recessions have this effect for multiple reasons, including not only
that the federal government pursues a deliberately countercyclical
fiscal policy, but also that a sluggish responsiveness of federal
expenditure to changes in income means that the ratio rises when GDP
fails to grow as rapidly as anticipated. Ratios of state and local
expenditure to GDP appear to exhibit a similar pattern of rising in
recessions, although from visual inspection it is not clear whether they
do so to the same degree as federal expenditure. Certainly state and
local spending during the 2007-09 recession did not rise as sharply as
federal spending. (6) But since the latest recession is just one of
several postwar recessions, it is useful to consider how state and local
spending and federal spending have responded in others.
[FIGURE 1 OMITTED]
Table 1 presents averages of quarterly growth rates of real federal
government and aggregate state and local government tax revenue and
spending per capita. (7) These growth rates are calculated as first
differences of the logarithms of seasonally adjusted quarterly values.
Table entries are cell means; thus, for example, the first column
indicates that, in quarters in which the output gap declined, the log of
real federal income tax collections per capita rose by an average of
0.0177. Strictly speaking, this corresponds to a growth rate of 1.79
percent, but as a convenient approximation, these log differences are
commonly interpreted as percentage growth rates. The output gap is the
difference between the economy's potential GDP and actual GDP,
divided by potential GDP, as reported by the Congressional Budget
Office.
The evidence in table 1 suggests that the cyclical patterns of U.S.
federal tax and spending policies differ from those of state and local
governments. In quarters during which the economy is expanding and the
output gap is narrowing, the mean growth rate of real federal tax
revenue per capita is 1.77 percent, whereas in quarters during which the
output gap is widening, the mean growth rate of this revenue measure is
-1.16 percent. State and local tax collections, by contrast, show a more
muted difference, growing at a quarterly average rate of 1.13 percent
when the output gap narrows and 0.27 percent when it widens. The last
three columns of table 1 show that during recession quarters (as
identified by the Business Cycle Dating Committee of the National Bureau
of Economic Research), the mean growth rate of real federal tax revenue
per capita is -2.91 percent, whereas at other times the mean growth rate
is 1.08 percent; state and local tax revenue grew by an average of -0.02
percent during recession quarters, and 0.93 percent at other times. As
the table indicates, these differences between federal and state and
local tax revenue growth are statistically as well as quantitatively
significant.
Federal spending tends to rise more rapidly when the economy is
performing poorly than when it is performing well; state and local
spending exhibits a similar pattern, but to a much smaller degree. Table
1 presents figures for federal nondefense expenditure, a portion of the
federal budget that is less subject than total expenditure to exogenous
shocks and more amenable to deliberate adjustment in response to
changing macroeconomic conditions. Real federal nondefense spending per
capita rose by a mean of 1.49 percent during quarters in which potential
GDP exceeded actual, and at other times fell by an average of 0.01
percent. State and local government real expenditure per capita rose by
an average of 0.60 percent during quarters in which potential GDP
exceeded the actual; it also rose, by an average of 0.57 percent, at
other times. As indicated in table 1, the relationship between federal
expenditure growth and the output gap differs statistically from that
between the growth of state and local spending and the output gap. The
differences between federal and state and local spending growth during
rising and falling output gaps, and during NBER-defined recessions, are
not significant but point in the same direction: federal spending
generally expands more than does state and local spending during
difficult economic times.
In interpreting the evidence in table 1, it is important to bear in
mind that these patterns reflect not only the degree to which the state
of the economy may influence government tax and expenditure choices, but
also the impact of taxes and government expenditure on economic
performance. The federal government is larger, in terms of both revenue
and spending, than the 50 state governments combined, and it is more
likely to coordinate fiscal and monetary policy; as a result, federal
actions have greater potential to influence the course of the economy. A
simple interpretation of the means in table 1 might take the output gap
and the cyclical status of the economy to be unaffected by government
tax and spending changes, which is inaccurate if federal economic
management is effective, and particularly if taxes and spending are
coordinated with other government policies. To the extent that federal
tax reductions and spending increases reduce the severity of economic
downturns, simple interpretations of the statistics in table 1
understate the extent to which the federal government actively manages
taxation and spending in response to underlying economic conditions. The
same argument applies to state and local governments, although their
smaller aggregate size gives it somewhat less significance. Furthermore,
more, one way in which the federal government actively manages aggregate
demand during downturns is by using federal grant dollars and other
inducements to encourage states to spend money; thus, some federal
spending is effectively channeled through the states and appears as
state expenditure even though it is largely federally determined.
Differences between federal fiscal policy and the fiscal policies
of state and local governments reflect many considerations, including
the federal government's explicit mandate to manage the economy
over the business cycle, and the ability and desire of state governments
to react to changing economic circumstances. Since states are
heterogeneous, it is potentially instructive to compare the fiscal
policies of states with differing characteristics, in order to identify
factors that may contribute to the determination of overall state
policies.
II. Understanding State Fiscal Policy
In the absence of countervailing policy action, tax revenue
generally increases as incomes rise, creating opportunities for states
to fund greater expenditure. A close connection between income changes
and expenditure changes could, of course, constitute efficient state
fiscal policy, although it also is characteristic of the flypaper effect
as applied to state finances. To the extent that the flypaper effect is
behavioral, fiscal policies driven by flypaper considerations are
potentially costly from an efficiency standpoint. Small jurisdictions
generally have the most elastic tax bases (Bucovetsky 1991, Kanbur and
Keen 1993, Wilson and Wildasin 2004) and therefore face significant
costs of adopting suboptimal policies that may chase away their business
and individual populations. The very fact that adopting inefficient
policies is so expensive has the potential to discipline politics in
these smaller states to avoid some of the behavioral inefficiencies that
could persist among their larger neighbors whose tax bases are more
secure.
In evaluating state fiscal policy, then, it is instructive to
distinguish between smaller and larger states in analyzing the impact of
income fluctuations on taxes and expenditure. For that purpose it is
necessary to use data from the Census of Governments. These data are
available from 1951 to 2007 for the lower 48 states and the District of
Columbia; coverage of Alaska and Hawaii is more limited, both in years
and in data items, and so these states are omitted from the analysis.
The data include information on state government expenditure, state
personal incomes, populations, receipts of grants from the federal
government, and collections of major sources of revenue. (8) It is worth
emphasizing that the present analysis does not use data on city, county,
or other local taxes and spending, since the purpose is to analyze the
determinants of taxes and spending at the state level. Personal income
and state tax collections are converted to real 2005 dollars using the
GDP deflator, and state government expenditure is converted to real 2005
dollars using the Bureau of Economic Analysis price index for state and
local government consumption expenditure and gross investment; all
variables are measured in per capita terms.
The state regressions include interactions between balanced budget
requirements and changes in state incomes. Every state except Vermont
has a formal requirement that its budget be balanced on an annual basis.
However, these requirements vary considerably in their ability to
constrain state legislatures, a fact that has permitted analysts to
evaluate their impact by comparing the experiences of states with strict
balanced budget laws with those that have more leeway to run deficits.
Earlier research (for example, Poterba 1994, Alt and Lowry 1994) finds
that states with more binding balanced budget requirements react to
fiscal crises (variously defined) differently than do states with less
binding requirements, although the reactions appear to be idiosyncratic
in that they are influenced by local politics.
Poterba (1994, 1997), Von Hagen (1991), and numerous other studies
use the characterization of state budget stringency from the now-defunct
Advisory Commission on Intergovernmental Relations as their source of
variation in state budget rules. In the simplest of these
classifications, roughly half of the states are classified as having
"strict" rules, with requirements that entail, if necessary,
mid-year budget adjustments and actions by state executive agencies that
make it politically costly (although by no means impossible) to run
deficits. The regressions reported in tables 2, 3, and 4 include this
variable interacted with changes in real state income per capita.
Columns 2-1 and 2-2 of table 2 report regressions in which the
dependent variable is the first difference of the log of real state
expenditure. The regressions include a complete set of year and state
dummy variables. State population is entered as the difference between a
state's share of the U.S. population and the sample mean of 0.0204;
all other independent variables are entered as first differences in
time. The estimated coefficients on the differences in log real income
per capita (-0.0159), and its one-year lag (0.0691), in the regression
in column 2-1, together imply that a 1 percent increase in income
(measured as a deviation from the national average) in a state with
population equal to the sample mean and no strict balanced budget
requirement is accompanied by a 0.0532 percent rise in state spending.
This is a very small effect, and statistically indistinguishable from
zero, suggesting that short-term income changes have little average
effect on state spending. The small size of the association may reflect
a combination of deliberate countercyclical policy and sluggishness in
adjusting spending to genuine income changes, together with any
measurement error in state personal income. The positive coefficients on
the interactions of the balanced budget requirement with the change in
income in column 2-1 are consistent with expenditure by states with
strict budgetary rules tracking state income changes more closely than
does expenditure by other states. However, the small magnitudes and
statistical insignificance of the coefficients suggest that balanced
budget rules do not have powerful average effects over the sample
period.
The coefficient on the interaction between (de-meaned) state
population and the one-year-lagged change in income in column 2-1
(4.714) indicates that government spending by larger states is more
positively associated with once-lagged income fluctuations than is
government spending by smaller states. The negative coefficient on
contemporaneous income changes in the same column (-2.261) implies,
however, that this association is mitigated by effects in the first
year; the sum of these coefficients is 2.453, with a standard error of
2.689. The regression reported in column 2-2 adds as explanatory
variables the first difference of the ratio of grants received from the
federal government to state personal income, the interaction of this
variable with de-meaned state population, and lags of these two
variables. The effect on spending of population interacted with lagged
income changes continues to be positive, although the sum of this
coefficient and the unlagged corresponding term is 4.038, with a
standard error of 2.938 that makes it insignificant.
Intergovernmental grants have the expected positive effect in the
regression reported in column 2-2, the coefficients together implying
that a state with mean population increases its expenditure by 5.468
percent as grants increase by 1 percent of state income. One way to
interpret the impact of intergovernmental grants is to specify the
change in state spending as a function of the change in state personal
income and grant receipts:
(1) [DELTA]ln[S.sub.t] = [beta][DELTA]ln ([y.sub.t] +
[gamma][g.sub.t]),
in which [S.sub.t] is state expenditure in year t, [y.sub.t] is
state personal income in year t, [g.sub.t] is government grant receipts
in year t, and [DELTA] is the first difference operator. The parameter
[gamma], reflects the fact that government grants may not affect
spending to the same extent that personal income does; values of [gamma]
> 1 correspond to larger effects of government grants. Equation 1 can
be rewritten as
(2) [DELTA] ln[S.sub.t] = [beta][DELTA]ln ([y.sub.t] +
[beta][DELTA]ln (1+[gamma] [g.sub.t]/[y.sub.t]).
Using a first-order Taylor approximation to the second term on the
right side of equation 2, evaluated at the mean value of
[g.sub.t]/[y.sub.t], denoted (g/y), produces
(3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
Column 2-2 presents estimated coefficients from a state spending
regression that includes [DELTA] [g.sub.t]/[y.sub.t] as an explanatory
variable. Equation 3 implies that the ratio of the sum of grant effects
(5.468) for a state with mean population to the sum of income effects
for a state without a strict balanced budget requirement (0.22), which
is 24.9, should equal [gamma]/[1 + [gamma](g/y)], or (given the very
small size of government grants relative to income) approximately
[gamma]. The estimates therefore imply that a dollar of grant receipt
has about the same impact on state government spending as $25 in
additional state personal income per capita, which, if taken literally,
implies that the marginal propensity to spend out of grant income is in
the neighborhood of 80 percent. (9) This clearly reflects not only the
significant effects that grants have on spending levels, but also that
grant receipts are influenced by state spending levels, sometimes on a
one-for-one basis. It is noteworthy, however, that the 98.88 coefficient
on the interaction of state population and federal grants suggests that
whatever the process is that determines the association of state
spending and federal grants, this process appears to operate more
strongly for large states than for small states. This is consistent with
a greater willingness of large states to spend out of transitory income,
although it may also reflect aspects of the grant allocation process
that somehow reward expenditure by large states more than expenditure by
small states.
Columns 2-3 and 2-4 of table 2 present estimated coefficients from
regressions in which the dependent variable is the first difference of
the log of state tax revenue, and the independent variables are the same
as those used in the regressions reported in columns 2-1 and 2-2. State
tax revenue rises with income: the sum of the first two coefficients in
column 2-3 indicates that for a state with mean population and no strict
balanced budget requirement, income growth by 1 percent is associated
with 0.667 percent tax revenue growth. The significant positive
coefficients on the interactions of population and the change in state
personal income (current and lagged) imply that rising income is more
strongly associated with increased tax collections in larger than in
smaller states. This is consistent with the less statistically powerful
evidence in columns 2-1 and 2-2 that expenditure of larger states
exhibits stronger income associations than does expenditure of smaller
states. The coefficients imply that increasing the size of an average
state by 1 percent of the U.S. population increases the estimated tax
revenue change associated with a 1 percent income fluctuation by 22
percent (from 0.667 percent to 0.816 percent). Including federal grants
as explanatory variables in the regression reported in column 2-4 has
little effect on these estimates, and the coefficients on grants are
much smaller in magnitude than are the corresponding coefficients in
column 2-2, notably failing to indicate that the spending effects of
grants are crowded out through tax reductions.
It is possible to estimate state-specific income sensitivities of
government expenditure and tax collections using a modified version of
the specification reported in columns 2-2 and 2-4 that substitutes for
the terms [DELTA] income, [[DELTA] income.sub.-1], population x [DELTA]
income, and population x [[DELTA] income.sub.-1], a complete set of
state dummy variables interacted with the sum of the change in log state
personal income and its lag. The estimated coefficients on these
interactions, plotted against state population in the top panel of
figure 2, capture the extent to which the expenditures of different
states respond to income changes. California, Massachusetts, Virginia,
and Wisconsin exhibit the greatest sensitivity of state government
spending to state personal income, and Delaware and Kansas the least.
The bottom panel of figure 2 reveals that California, Michigan,
Pennsylvania, and Virginia exhibit the greatest sensitivity of tax
collections to state personal income, and Delaware and Nebraska are the
least sensitive. The figure shows considerable variation across states
for both revenue and spending sensitivity but suggests a general
positive relationship between both and state population, with no obvious
outliers responsible for the estimation results. As a robustness check,
the regressions reported in columns 2-2 and 2-4 were rerun dropping
potentially important single states, including California, but the
impact on the size and significance of the regression estimates was
small.
[FIGURE 2 OMITTED]
Further evidence comes from distinguishing state expenditure by
function and state tax revenue by source. The first five columns of
table 3 report estimated coefficients from regressions in which the
dependent variables are log changes in real expenditure per capita on
education, health and hospitals, public welfare, highways and roads, and
unemployment insurance. These are the five largest state government
spending categories; education and public welfare alone accounted for
more than half of total state spending in 2006. In these regressions
there is a mild association of spending and intergovernmental grants,
particularly in the categories of health and hospitals, public welfare,
and highways, for which the federal government makes available
substantial matching grants. However, the grants variable used in these
regressions is total grants received by states; the grants are not
distinguished by category, in an effort to avoid the most obvious
sources of endogeneity.
The coefficients reported in these first five columns of table 3
imply that interactions of state population shares and income changes
have sizable positive effects on the growth rates of education and
health and hospital spending, and large but statistically insignificant
effects on the growth rate of public welfare spending. In the case of
education, the sum of the coefficients on current and lagged income
changes interacted with population share is 12.863 (with a standard
error of 3.846), which implies that increasing the size of an average
state by 1 percent of the U.S. population is associated with roughly
one-third greater income sensitivity of education spending. In the case
of health and hospitals, the sum of the population and income
interaction coefficients is 12.331 (with a standard error of 5.987),
which, together with the income coefficients, implies that increasing
the size of an average state by 1 percent of the U.S. population is
associated with a 26 percent greater income sensitivity of health and
hospital expenditure. The sums of the estimated coefficients on the
population-income interaction are negative (but insignificant) in the
case of highway spending and unemployment insurance, the latter of which
expands during economic downturns. In the interpretation that larger
states display greater spending responsiveness to transitory income
fluctuations, it might be expected that changes in unemployment
insurance expenditure, which are negatively associated with income
changes, should react with greater magnitude in large states.
The last four columns of table 3 report coefficients from
regressions in which the dependent variables are log changes in real
individual income tax collections per capita, corporate tax collections,
sales tax collections, and property tax collections. Not all states use
all of these taxes--a fact reflected in the varying sample sizes. Among
these taxes, corporate and sales tax collections display significant
associations with income fluctuations, although the point estimate of
income effects on property taxes is large, with a very large standard
error. Only in the individual income tax regression is there a
significant effect of the interaction between population size and
current plus lagged income changes (the sum of current and lagged
coefficients is 20.51, with an associated standard error of 8.00);
corporate income tax revenue also has a large estimated effect (23.76)
but with a large standard error (15.70). As a practical matter, the
differing income sensitivities associated with different state taxes
make it feasible for state governments to select revenue sources that
tailor their revenue streams, within limits, to desired tax collections
as incomes fluctuate.
The associations of state size with expenditure and tax
responsiveness to changes in state incomes (and changes in federal grant
receipts) raise the possibility that other variables correlated with
state size may be responsible for this pattern. To the extent that state
size matters because larger states have less mobile populations, direct
measures of U.S. population mobility might be used in place of state
size; and if larger states tend to have more dysfunctional governments,
then measures of state government corruption might also be used.
D'Vera Cohn and Rich Morin (2008) report the fractions of adult
Americans born in each state who continued to live in their state of
birth during 2005-07; their figures are reproduced as appendix table 1.
(The figures should be interpreted to mean that, for example, 69 percent
of current American adults who were born in California continued to live
there during 2005-07.) Edward Glaeser and Raven Saks (2006) report
average annual convictions of state public officials for federal
corruption-related crimes per 100,000 state residents between 1976 and
2002; these data do not include figures for the District of Columbia.
Table 4 presents regressions with the same specifications as those
in table 2 but adding as independent variables interactions of measures
of population mobility and state corruption with income fluctuations.
(Mobility and corruption are time-invariant measures for each state and
therefore not included as independent variables.) The results suggest
that the results reported above for state size may partially reflect the
impact of population mobility; the evidence is weaker in the case of
state government performance. In the regression reported in column 4-1,
the estimated effect of the interaction of state mobility and income
fluctuations is large and statistically significant (the sum of the
coefficients is 2.89, with a standard error of 0.51), indicating that
expenditure by states with less mobile populations is more closely
associated with income fluctuations than is expenditure by states with
more mobile populations. The interaction of corruption and income
changes is positive and significant for contemporaneous income changes,
but not for the sum of current and lagged changes; the sum of the
coefficients on contemporaneous and lagged interactions of population
and income changes is also insignificant. Similar results appear with
the introduction of the grants variables in the regression reported in
column 4-2 and for the tax regressions reported in columns 4-3 and 4-4.
Measured immobility as reported in appendix table 1 has a
correlation of 0.53 with state population shares, which in part explains
why the population mobility effects in the regressions reported in table
4 look similar to those reported for state size in table 2. Since
population mobility is potentially influenced by state fiscal policies,
it may be problematic to treat these variables as strictly
exogenous--but the same is true of state population. In both cases there
is reason to expect these variables to be related to the underlying
mobility of the population, but it is very difficult to distinguish the
effect of state size per se from the effects of any other variables that
are strongly correlated with state size.
III. Conclusion
Tax revenue of state and local governments is more robust in
economic downturns than is federal tax revenue, whereas state and local
government spending grows at a slower rate in downturns than does
federal spending. It is tempting to attribute much of this pattern to
choices made long ago about the respective tax bases of federal and
subfederal governments. States and localities rely to a much greater
extent than does the federal government on property taxes and
expenditure-type taxes such as sales and excise taxes, whereas the
federal government depends more on income and payroll taxes. State and
local revenue might as a result be more stable over the business cycle,
and although this does not explain differences in expenditure patterns,
it offers a plausible explanation of patterns of tax receipts. One
difficulty with this interpretation is that property taxes, which are
used almost exclusively by state and local governments, do not appear to
be less sensitive to income fluctuations than are income taxes; a second
is that local, state, and federal governments are entitled to change
their funding models at any time if they are dissatisfied with the
revenue streams they produce.
It is perhaps surprising that larger states do not have spending
and tax policies that more closely resemble those of the federal
government, since they are closer in size to the nation as a whole than
are smaller states, and their governments may be closer in character to
the federal government. Also, as Edward Gramlich (1997) and others note,
states--particularly large states--may be able to internalize a large
share of the benefits of stimulating their economies when there are
underutilized resources. To the extent that larger states have less
mobile populations, they are less subject to the point raised by
Blanchard and Lawrence Katz (1992) that interstate worker mobility
quickly mitigates adverse state economic outcomes (however, see Rowthorn
and Glyn 2006 for a contrary interpretation of the evidence). The logic
of this argument implies that states with more mobile populations might
get smaller returns from trying to increase demand for local factors.
Yet the evidence suggests otherwise and therefore may reflect either
additional economic considerations or a failure of government
optimization.
To the extent that there are national externalities associated with
state and local fiscal policies, the federal government has instruments
at its disposal that can influence patterns of state and local taxes and
spending. The most powerful of these is the provision of grants to state
governments. There is ample evidence that state spending is influenced
by federal grants, so a carefully tailored countercyclical grant policy
has the potential to encourage states, acting on their own behalf, to
behave in a manner that effectively incorporates the interests of other
states in stimulating the economy when resources are otherwise
underutilized.
APPENDIX
Table A1. Interstate Population Mobility, 2005-07
Percent of adults
born in the state
State now living there
Alabama 63.7
Alaska 28.2
Arizona 61.5
Arkansas 54.5
California 69.0
Colorado 54.7
Connecticut 57.1
Delaware 54.0
District of Columbia 13.0
Florida 66.0
Georgia 69.6
Hawaii 57.3
Idaho 48.6
Illinois 59.0
Indiana 62.8
Iowa 54.0
Kansas 50.2
Kentucky 62.6
Louisiana 64.4
Maine 55.3
Maryland 61.1
Massachusetts 58.7
Michigan 67.5
Minnesota 66.3
Mississippi 54.9
Missouri 61.9
Percent of adults
born in the state
State now living there
Montana 47.1
Nebraska 50.2
Nevada 48.7
New Hampshire 52.8
New Jersey 55.6
New Mexico 53.5
New York 55.5
North Carolina 71.4
North Dakota 40.4
Ohio 65.1
Oklahoma 55.6
Oregon 59.2
Pennsylvania 63.8
Rhode Island 53.9
South Carolina 66.0
South Dakota 43.4
Tennessee 66.7
Texas 75.8
Utah 66.1
Vermont 52.5
Virginia 61.9
Washington 64.3
West Virginia 48.9
Wisconsin 68.6
Wyoming 35.7
Source: Cohn and Morin (2008).
ACKNOWLEDGMENTS I thank Molly Saunders-Scott and Daniel Schaffa for
outstanding research assistance, and Kathryn Dominguez, Bill Gale, Lutz
Kilian, Brian Knight, the editors, and participants at the Brookings
Panel conference for extremely helpful comments on earlier drafts.
The author reports no relevant potential conflicts of interest.
References
Aizenman, Joshua, and Gurnain Kaur Pasricha. 2010. "On the
Ease of Overstating the Fiscal Stimulus in the US, 2008-9.'"
Working Paper no. 15784. Cambridge, Mass.: National Bureau of Economic
Research (February).
Alt, James E., and Robert C. Lowry. 1994. "'Divided
Government, Fiscal Institutions, and Budget Deficits: Evidence from the
States." American Political Science Review 88, no. 4:811-28.
Auerbach, Alan J., William G. Gale, and Benjamin H. Harris. 2010.
"Activist Fiscal Policy." Journal of Economic Perspectives 24,
no. 4: 141-64.
Auerbach, Alan J., and Yuriy Gorodnichenko. 2010. "Measuring
the Output Responses to Fiscal Policy." Working Paper no. 16311.
Cambridge, Mass.: National Bureau of Economic Research (August).
Bahl, Roy. 1984. Financing State and Local Government in the 1980s.
Oxford University Press.
Baicker, Katherine. 2005. "Extensive or Intensive Generosity?
The Price and Income Effects of Federal Grants." Review of
Economics and Statistics 87, no. 2: 371-84.
Barro, Robert J., and Charles J. Redlick. 2009. "Macroeconomic
Effects from Government Purchases and Taxes." Working Paper no.
15369. Cambridge, Mass.: National Bureau of Economic Research
(September).
Blanchard, Olivier Jean, and Lawrence F. Katz. 1992. "Regional
Evolutions." BPEA, no. 1 : 1-61.
Blanchard, Olivier, and Roberto Perotti. 2002. "'An
Empirical Characterization of the Dynamic Effects of Changes in
Government Spending and Taxes on Output." Quarterly Journal of
Economics 117, no. 4:1329-68.
Blanchard, Olivier Jean, Florencio Lopez-de-Silanes, and Andrei
Shleifer. 1994. "What Do Firms Do with Cash Windfalls'?"
Journal of Financial Economics 36, no. 3: 337-60.
Bucovetsky, S. 1991. "Asymmetric Tax Competition."
Journal of Urban Economics 30, no. 2: 167-81.
Cogan, John F., Tobias Cwik, John B. Taylor. and Volker Wieland.
2010. "New Keynesian versus Old Keynesian Government Spending
Multipliers." Journal of Economic Dynamics and Control 34, no.
3:281-295.
Cohn, D'Vera, and Rich Morin. 2008. "American Mobility:
Who Moves? Who Stays Put'? Where's Home?" Social and
Demographic Trends Report. Washington: Pew Research Center (December).
Dahlberg, Matz, Eva Mork, Jorn Rattso, and Hanna Agren. 2008.
"Using a Discontinuous Grant Rule to Identify the Effect of Grants
on Local Taxes and Spending." Journal of Public Economics 92, no.
12: 2320-35.
Feldstein, Martin S., and Gilbert E. Metcalf. 1987. "'The
Effect of Federal Tax Deductibility on State and Local Taxes and
Spending." Journal of Political Economy 95, no. 4:710-36.
Glaeser, Edward L., and Raven E. Saks. 2006. "Corruption in
America." Journal of Public Economics 90, no. 6-7: 1053-72.
Gordon, Nora. 2004. "Do Federal Grants Boost School Spending?
Evidence from Title I." Journal of Public Economics 88, no.
9-10:1771-92.
Gramlich, Edward M. 1997. "Subnational Fiscal Policy." In
Financing Federal Systems: The Selected Essays of Edward M. Gramlich.
Cheltenham, U.K.: Edward Elgar.
Hall, Robert E. 2009. "By How Much Does GDP Rise if the
Government Buys More Output?" BPEA, no. 2: 183-231.
Hines, James R., Jr., and Richard H. Thaler. 1995. "The
Flypaper Effect." Journal of Economic Perspectives 9, no. 4:
217-26.
Holtz-Eakin, Douglas, Whitney Newey, and Harvey S. Rosen. 1989.
"The Revenues-Expenditures Nexus: Evidence from Local Government
Data." International Economic Review 30, no. 2:415-29.
Kanbur, Ravi, and Michael Keen. 1993. "Jeux sans Frontieres:
Tax Competition and Tax Coordination When Countries Differ in
Size." American Economic Review 83, no. 4: 877-92.
Knight, Brian. 2002. "Endogenous Federal Grants and Crowd-Out
of State Government Spending: Theory and Evidence from the Federal
Highway Aid Program." American Economic Review 92, no. 1 : 71-92.
Lutz, Byron. 2010. "Taxation with Representation:
Intergovernmental Grants in a Plebiscite." Review of Economics and
Statistics 92, no. 2:316-32.
Poterba, James M. 1994. "State Responses to Fiscal Crises: The
Effects of Budgetary Institutions and Politics." Journal of
Political Economy 102, no. 4: 799-821.
--. 1997. "Do Budget Rules Work?" In Fiscal Policy:
Lessons from Economic Research, edited by Alan J. Auerbach. MIT Press.
Rowthorn, Robert, and Andrew J. Glyn. 2006. "Convergence and
Stability in U.S. Employment Rates." Contributions to
Macroeconomics 6, no. 1: article 4.
www.bepress.com/bejm/contributions/vol6/iss 1/art4.
Romer, Christina D., and David H. Romer. 2010. "The
Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure
of Fiscal Shocks." American Economic Review 100 no. 3: 763-801.
Strumpf, Koleman S. 1998. "A Predictive Index for the Flypaper
Effect." Journal of Public Economics 69, no. 3: 389-412.
Von Hagen, Jargen. 1991. "A Note on the Empirical
Effectiveness of Formal Fiscal Restraints." Journal of Public
Economics 44, no. 2: 199-210.
Wilson, John Douglas, and David E. Wildasin. 2004. "Capital
Tax Competition: Bane or Boon?" Journal of Public Economics 88, no.
6: 1065-91.
Woodford, Michael. 2010. "Simple Analytics of the Government
Expenditure Multiplier." Working Paper no. 15714. Cambridge, Mass.:
National Bureau of Economic Research (January).
JAMES R. HINES JR.
University of Michigan
(1.) The state and local totals are the sum of state, county, city,
and other local taxes and spending. Federal and aggregate state and
local government tax and expenditure data, both at annual and quarterly
frequencies, are drawn from the U.S. national income and product
accounts.
(2.) See, for example, Hines and Thaler (1995), Strumpf (1998), and
Baicker (2005). Knight (2002), Gordon (2004), and Lutz (2010) offer
evidence and interpretations that the magnitude of the flypaper effect
may be overstated by earlier studies that fail to control properly for
the endogeneity of grant receipts to spending levels, although even the
data presented by Lutz (2010) appear to be consistent with significant
flypaper effects.
(3.) An additional reason why states might reduce expenditure
during recessions is that state spending becomes more expensive, on an
after-federal-tax basis, to the degree that fewer taxpayers itemize
deductions or that those who do itemize face lower marginal tax rates.
The ability of taxpayers who itemize on their federal income tax returns
to take deductions for state and local income and property tax payments
generally reduces the alter-tax cost of state spending, and thereby
encourages state governments to spend more than they would were state
taxes nondeductible (Feldstein and Metcalf 1987). Given that only
roughly one-third of American taxpayers itemize their deductions,
however, and that the fraction itemizing does not appear to fall
systematically during recessions (the numbers of taxpayers itemizing
deductions rose during the 1980, 1991-92, and 2001 recessions, according
to IRS data: "SOl Bulletin Historical Table 7,"
www.irs.gov/taxstats/article/0,,id=175812,00.html), the effect of tax
deductibility on state expenditure is unlikely to account for
significant expenditure reactions to economic downturns.
(4.) There is enormous controversy over the extent to which
countercyclical fiscal policy influences the employment of resources
during economic downturns. See, for example, Auerbach, Gale, and Harris
(2010), Auerbach and Gorodnichenko (2010), Barro and Redlick (2009),
Blanchard and Perotti (2002), Cogan and others (2010), Hall (2009),
Romer and Romer (2010), and Woodford (2010). Gramlich (1997) identifies
circumstances in which state governments might benefit from expanding
the employment of underutilized resources through expansionary fiscal
policies, and Bahl (1984) reviews older evidence of the association of
state tax and spending policies and business cycle fluctuations.
(5.) GDP, federal taxes, and state and local taxes are deflated
using the GDP deflator. Government consumption and investment
expenditures are separately deflated using the corresponding deflators,
and all other categories of government expenditure (transfer payments,
interest payments, and others) are deflated using the GDP deflator.
(6.) Aizenman and Pasricha (2010) compare recent levels of certain
(nontransfer) state and federal spending categories with levels
predicted from a simple time-series model and argue that discretionary
state fiscal policy is by this measure so contractionary that it roughly
offsets the recent federal expansion. This conclusion appears to depend
critically on the model used to predict discretionary spending; but
almost any measure shows that state spending failed to expand during the
2007-09 recession, which suggests that states did not pursue active
countercyclical policies.
(7.) Variables are converted to per capita terms using quarterly
interpolations of annual population data from the U.S. Census.
(8.) Data on detailed categories of state government expenditure
are not available for 2007. Annual state populations are available from
the Census.
(9.) The 0.220 estimated income elasticity in column 2-2, together
with the 0.1475 mean ratio of state spending to state personal income,
implies that, evaluated at the mean, the marginal propensity to spend by
state government out of a dollar of personal income is (0.22)(0.15) =
0.033. Then (25)(0.033) = 0.825.
Table 1. Growth in Real Federal and Aggregate State and Local Spending
and Tax Revenue per Capita under Differing Economic Conditions
Average quarterly growth rate (a)
Expenditure or Output gap
revenue measure Narrowing (b) Widening Difference
Expenditure
Federal nondefense 0.61 1.07 -0.45
(0.46) (0.36) (0.59)
State and local 0.58 0.57 0.01
(0.09) (0.10) (0.13)
Difference 0.03 0.49 -0.46
(0.45) (0.38) (0.59)
Tax revenue
Federal 1.77 -1.16 2.93
(0.38) (0.38) (0.53)
State and local 1.13 0.27 0.86
(0.08) (0.12) (0.14)
Difference 0.64 -1.43 2.07
(0.36) (0.36) (0.51)
No. of observations 125 119
Average quarterly growth rate (a)
Expenditure or Output gap
revenue measure Negative (c) Positive Difference
Expenditure
Federal nondefense -0.01 1.49 1.50
(0.52) (0.33) (0.61)
State and local 0.57 0.60 -0.03
(0.11) (0.09) (0.14)
Difference -0.58 0.89 -1.47
(0.52) (0.32) (0.61)
Tax revenue
Federal 1.01 -0.22 1.22
(0.33) (0.43) (0.54)
State and local 0.79 0.67 0.11
(0.11) (0.11) (0.15)
Difference 0.22 -0.89 1.11
(0.30) (0.41) (0.51)
No. of observations 106 139
Average quarterly growth rate (a)
Expenditure or Recession quarter? (d)
revenue measure No Yes Difference
Expenditure
Federal nondefense 0.77 1.74 -0.98
(0.37) (0.62) (0.72)
State and local 0.52 0.91 -0.39
(0.07) (0.19) (0.20)
Difference 0.25 0.84 -0.59
(0.36) (0.66) (0.75)
Tax revenue
Federal 1.08 -2.91 3.99
(0.28) (0.64) (0.70)
State and local 0.93 -0.02 0.95
(0.07) (0.21) (0.22)
Difference 0.15 -2.89 3.04
(0.27) (0.62) (0.68)
No. of observations 200 52
Source: Author's calculations.
(a.) Reported values are simple averages of quarterly growth rates of
real spending or real tax revenue per capita from 1949Q1 to 2010Q1,
calculated as approximations based on first differences in the
logarithms of seasonally adjusted values as reported in the national
income and product accounts. Standard errors are in parentheses.
(b.) Quarters during which the output gap (difference between potential
and actual GDP, divided by potential GDP, as defined by the
Congressional Budget Office) was declining.
(c.) Quarters during which the output gap was negative (actual GDP
exceeded potential GDP).
(d.) Recession quarters are those identified by the Business Cycle
Dating Committee of the National Bureau of Economic Research as
recession months. Data are from 1947Q1 through 2010Q I.
Table 2. Regressions Explaining Annual Changes in State Spending and
Tax Revenue
Dependent variable (a)
[DELTA] log spending
Independent variable (b) 2-1 2-2
[DELTA] income (c) -0.0159 0.119 **
(0.0695) (0.0588)
[DELTA] [income.sub.-1] 0.0691 0.101 *
(0.0598) (0.0593)
Balanced budget req. (d) x [DELTA] income 0.00715 -0.0160
(0.0798) (0.0709)
Balanced budget req. x [DELTA] 0.0186 0.00722
[income.sub.-1] (0.0988) (0.0935)
Population (c) x [DELTA] income -2.261 -0.740
(1.953) (1.751)
Population x [DELTA] [income.sub.-1] 4.714 ** 4.778 **
(2.066) (2.237)
Population -1.374 1.170
(3.788) (3.194)
[Population.sub.-1] 1.675 -0.941
(3.768) (3.133)
A grants-income ratio (f) 4.862 ***
(0.809)
A grants-income [ratio.sub.-1] 0.606 *
(0.311)
Population x [DELTA] grants-income ratio 98.88 **
(39.82)
Population x [DELTA] grants-income -2.500
[ratio.sub.-1] (9.095)
[R.sup.2] 0.279 0.348
Dependent variable (a)
[DELTA] log tax revenue
Independent variable (b) 2-3 2-4
[DELTA] income (c) 0.368 *** 0.367 ***
(0.0836) (0.0836)
[DELTA] [income.sub.-1] 0.299 *** 0.304 ***
(0.0551) (0.0591)
Balanced budget req. (d) x [DELTA] income -0.0644 -0.0618
(0.107) (0.107)
Balanced budget req. x [DELTA] 0.0796 0.0769
[income.sub.-1] (0.0723) (0.0726)
Population (c) x [DELTA] income 9.975 *** 10.30 ***
(2.636) (2.757)
Population x [DELTA] [income.sub.-1] 4.881 *** 4.847 **
(1.795) (1.888)
Population -8.162 ** -8.534 **
(3.998) (4.140)
[Population.sub.-1] 8.235 ** 8.600 **
(3.965) (4.114)
A grants-income ratio (f) 0.100
(0.447)
A grants-income [ratio.sub.-1] 0.00611
(0.361)
Population x [DELTA] grants-income ratio 17.53
(20.58)
Population x [DELTA] grants-income -14.55
[ratio.sub.-1] (12.44)
[R.sup.2] 0.270 0.270
Source: Author's regressions.
(a.) Dependent variables are annual log changes in real state government
spending or tax revenue per capita. All regressions include state and
year fixed effects. Sample size is 2.695 in all regressions. Standard
errors clustered at the state level are in parentheses.
(b.) Variables subscripted "-1" are one-year lags.
(c.) Annual change in the log of real state personal income per capita.
(d.) Dummy variable equal to 1 for states with a strict annual balanced
budget requirement, and zero otherwise.
(e.) Difference between the state's population as a fraction of total
U.S. population and the sample mean of 0.0204.
(f.) Annual change in the ratio of federal grants to state personal
income.
Table 3. Regressions Explaining Annual Changes in State Spending by
Category and Tax Revenue by Source
Dependent variable (a)
[DELTA] log spending
Health and
Independent variable (b) Education hospitals
[DELTA] income 0.258 *** 0.308 ***
(0.0885) (0.107)
[DELTA] [income.sub.-1] 0.127 0.164
(0.0939) (0.115)
Balanced budget req. x [DELTA] income -0.0285 0.179
(0.0970) (0.111)
Balanced budget req. x [DELTA] 0.0755 -0.115
[income.sub.-1] (0.114) (0.129)
Population x [DELTA] income 7.186 ** 7.470
(3.286) (5.132)
Population x [DELTA] [income.sub.-1] 5.677 * 4.861 *
(3.230) (2.876)
Population 2.646 0.122
(4.691) (7.133)
[Population.sub.-1] -2.574 -0.277
(4.644) (6.968)
[DELTA] grants-income ratio 1.512 ** 3.028 **
(0.681) (1.210)
[DELTA] grants-income [ratio.sub.-1] 0.413 0.466
(0.509) (0.793)
Population x [DELTA] grants-income ratio 27.84 46.09
(41.81) (38.94)
Population x [DELTA] grants-income -29.74 * -6.594
[ratio.sub.-1] (17.15) (33.60)
No. of observations 2,646 2,646
[R.sup.2] 0.214 0.087
Dependent variable (a)
[DELTA] log spending
Public
Independent variable (b) welfare Highways
[DELTA] income 0.257 *** 0.247 **
(0.0780) (0.114)
[DELTA] [income.sub.-1] -0.0738 0.347 *
(0.107) (0.201)
Balanced budget req. x [DELTA] income -0.0197 0.188
(0.0952) (0.167)
Balanced budget req. x [DELTA] 0.0726 -0.0621
[income.sub.-1] (0_115) (0.218)
Population x [DELTA] income 7.551 *** -10.88 **
(2.731) (4.921)
Population x [DELTA] [income.sub.-1] 0.559 1.953
(4.417) (4.954)
Population -8.374 1.458
(8.949) (9.406)
[Population.sub.-1] 9.196 -0.493
(9.021) (9.210)
[DELTA] grants-income ratio 8.379 *** 11.16 **"'
(1.103) (1.560)
[DELTA] grants-income [ratio.sub.-1] 1.713 *** 0.952
(0_541) (1.289)
Population x [DELTA] grants-income ratio 213.0 *** -82.00
(49.60) (73.35)
Population x [DELTA] grants-income -18.58 -11.51
[ratio.sub.-1] (20.48) (32.84)
No. of observations 2.646 2,646
[R.sup.2] 0.312 0.209
Dependent variable (a)
[DELTA] log A log tax
spending revenue (c)
Unemployment Individual
Independent variable (b) insurance income tax
[DELTA] income -1.038 * 0.414 **
(0.563) (0.178)
[DELTA] [income.sub.-1] -1.294 *** -0.0244
(0.301) (0.187)
Balanced budget req. x [DELTA] income -0.725 -0.232
(0.732) (0.236)
Balanced budget req. x [DELTA] 0.0607 0.523 *
[income.sub.-1] (0.322) (0.302)
Population x [DELTA] income -24.46 14.68 *
(21.16) (8.654)
Population x [DELTA] [income.sub.-1] -23.90 ** 5.829
(10.38) (6.085)
Population 29.49 * -27.64
(16.16) (18.03)
[Population.sub.-1] -28.97 * 28.21
(16.08) (18.24)
[DELTA] grants-income ratio 5.000 *"' -1.135
(2.286) (1.443)
[DELTA] grants-income [ratio.sub.-1] 0.186 -0.536
(1.047) (1.658)
Population x [DELTA] grants-income ratio 151.5 ** 1.785
(59.79) (55.76)
Population x [DELTA] grants-income -145.0 ** -6.432
[ratio.sub.-1] (71.77) (46.23)
No. of observations 2,646 2,191
[R.sup.2] 0.705 0.132
Dependent variable (a)
A log tax revenue (c)
Corporate Sales
Independent variable (b) income tax tar
[DELTA] income 0.616 ** 0.107
(0.253) (0.0995)
[DELTA] [income.sub.-1] 0.538 *** 0.257 **
(0.191) (0.0857)
Balanced budget req. x [DELTA] income -0.0862 0.0109
(0.303) (0.112)
Balanced budget req. x [DELTA] -0.328 0.0717
[income.sub.-1] (0.283) (0.0962)
Population x [DELTA] income 17.79 * 4.279
(10.29) (2.951)
Population x [DELTA] [income.sub.-1] 5.974 2.851
(8.307) (2.389)
Population 31.07 -14.54 **
(20.55) (7.149)
[Population.sub.-1] -31.13 14.74 **
(20.76) (7.142)
[DELTA] grants-income ratio -1.591 -0.212
(1.627) (0.536)
[DELTA] grants-income [ratio.sub.-1] -2.398 -0.155
(1.575) (0.494)
Population x [DELTA] grants-income ratio 66.85 0.430
(65.95) (26.74)
Population x [DELTA] grants-income 7.299 -22.06
[ratio.sub.-1] (71.28) (22.49)
No. of observations 2,249 2,695
[R.sup.2] 0.216 0.204
Dependent variable (a)
A log tax revenue (c)
Property
Independent variable (b) tax
[DELTA] income 0.708
(0.539)
[DELTA] [income.sub.-1] 0.0425
(0.632)
Balanced budget req. x [DELTA] income -0.499
(0.716)
Balanced budget req. x [DELTA] 0.425
[income.sub.-1] (0.781)
Population x [DELTA] income 0.361
(17.16)
Population x [DELTA] [income.sub.-1] 10.23
(23.43)
Population -9.934
(74.12)
[Population.sub.-1] 6.595
(71.09)
[DELTA] grants-income ratio 6.357 *
(3.268)
[DELTA] grants-income [ratio.sub.-1] 6.740 *
(3.612)
Population x [DELTA] grants-income ratio 296.8
(186.0)
Population x [DELTA] grants-income 51.11
[ratio.sub.-1] (149.2)
No. of observations 2,345
[R.sup.2] 0.027
Source: Author's regressions.
(a.) All regression include state and year fixed effects. Standard
errors clustered at the state level are in parentheses.
(b.) See table 2 for definitions.
(c.) States not collecting tax revenue from the indicated source are
omitted from the sample for that regression.
Table 4. Regressions Investigating the Effect of Population Mobility
and Corruption on State Spending and Tax Revenue
Dependent variable (a)
[DELTA] log spending
Independent variable 4-1 4-2
[DELTA] income -0.00499 0.145 ***
(0.0607) (0.0592)
[DELTA] [income.sub.-1] 0.165 * 0.165 *
(0.0843) (0.0849)
Balanced budget req. x [DELTA] income -0.0142 -0.0372
(0.0689) (0.0638)
Balanced budget req. x -0.0528 -0.0454
[DELTA] [income.sub.-1] (0.0939) (0.0922)
Population x [DELTA] income -5.297 -3.840
(3.245) (2.932)
Population x [DELTA] [income.sub.-1] -1.157 0.307
(3.134) (3.145)
Immobility' x [DELTA] income 1.269 ** 0.920 **
(0.520) (0.453)
Immobility x [DELTA] [income.sub.-1] 1.617 *** 0.894
(0.580) (0.608)
Corruption (c) x [DELTA] income 0.616 ** 0.592 **
(0.245) (0.224)
Corruption x [DELTA] [income.sub.-1] -0.158 -0.251
(0.334) (0.357)
Population -1.956 1.478
(4.000) (3.304)
[Population.sub.-1] 2.437 -1.068
(4.019) (3.250)
[DELTA] grants-income ratio 5.703 ***
(0.572)
[DELTA] grants-income [ratio.sub.-1] 0.743 **
(0.290)
Population x [DELTA] grants- 56.63 **
income ratio (28.26)
Population x [DELTA] grants- -3.067
income [ratio.sub.-1] (10.09)
[R.sup.2] 0.304 0.372
Dependent variable (a)
[DELTA] log tax revenue
Independent variable 4-3 4-4
[DELTA] income 0.480 *** -0.482 ***
(0.0666) (0.0712)
[DELTA] [income.sub.-1] 0.305 0.306 ***
(0.0646) (0.0690)
Balanced budget req. x [DELTA] income -0.133 * -0.132 *
(0.0761) (0.0764)
Balanced budget req. x 0.0619 0.0600
[DELTA] [income.sub.-1] (0.0735) (0.0742
Population x [DELTA] income 3.236 3.354
(2.350) (2.420)
Population x [DELTA] [income.sub.-1] 4.170 * 4.115
(2.368) (2.490)
Immobility' x [DELTA] income 1.983 *** 2.016 ***
(0.385) (0.386)
Immobility x [DELTA] [income.sub.-1] 0.363 0.341
(0.418) (0.450)
Corruption (c) x [DELTA] income -0.320 -0.327
(0.284) (0.285)
Corruption x [DELTA] [income.sub.-1] -0.0226 -0.0159
(0.281) (0.282)
Population -7.909 * -8.298 *
(4.060) (4.205)
[Population.sub.-1] 8.147 ** 8.537 **
(4.036) (4.188)
[DELTA] grants-income ratio 0.121
(0.528)
[DELTA] grants-income [ratio.sub.-1] -0.0774
(0.431)
Population x [DELTA] grants- 15.45
income ratio (22.60)
Population x [DELTA] grants- -17.70
income [ratio.sub.-1] (13.80)
[R.sup.2] 0.278 0.278
Source: Authors' regressions.
(a.) Dependent variables are annual log changes in real state
government spending or tax revenue per capita. All regressions include
state and year fixed effects. Sample size is 2,640 in all regressions.
Standard errors clustered at the state level are in parentheses. See
table 2 for definitions of variables not defined below.
(b.) Difference between the fraction of adults born in the state who
lived there in 2005-07 and the sample mean of 0.5741.
(c.) Difference between the average annual number of federal
corruption convictions of state government officials per 100,000 state
residents during 197(-2002 and the sample mean of 0.2709.
Comments and Discussion
COMMENT BY
WILLIAM G. GALE The Great Recession and the associated fiscal
policy responses in many countries have renewed research interest in the
effects of activist fiscal stabilization and stimulus policies
(Auerbach, Gale, and Harris 2010). This paper by James Hines addresses
an important but often underanalyzed component of that issue, the role
of subnational governments.
In the United States, state and federal governments can have
important influences on overall fiscal stimulus for at least three
reasons. First, these governments' spending and taxes, which
equaled 14 percent and 9 percent of GDP, respectively, in 2009, are
sizable relative to those of the federal government. Second, almost all
states have balanced budget rules. When revenue falls during a
recession, states must either draw down their rainy-day funds, raise
taxes, or cut spending; the latter two options are likely to act as
procyclical policies that could exacerbate the downturn. James Poterba
(1994), for example, finds strong evidence that states contract spending
and raise taxes when faced with a negative fiscal shock. He also finds
that states with stricter budget rules (for example, those that apply to
the enacted budget, rather than just the proposed budget) respond to
unexpected deficits by reducing spending much more than other states.
Third, one federal stimulus option--besides raising government
purchases, raising transfers to individuals or businesses, or cutting
taxes-is to provide transfers to the states, which could ease their
balanced budget constraints and reduce the need for contractionary state
responses. Although the argument that transfers to states are
stimulative is plausible, there is surprisingly little evidence on their
countercyclical effects.
Edward Gramlich (1978, 1979) and Robert Reischauer (1978) evaluate
the effects of three federal grant programs undertaken in response to
the 1973-75 recession. One program offered countercyclical revenue to
the states in the form of block grants, another paid the salaries of
state and local government workers, and a third contributed funding for
capital improvements. The general finding was that states'
short-run response to this federal aid was primarily to bolster their
rainy-day funds; increases in outlays and reductions in taxes were
modest in the short run. The long-run response--particularly in the form
of decreased income tax revenue--was substantial but materialized after
the recession had ended. It is unclear how relevant these findings are
to the current economic downturn, however, given the dated nature of the
evidence, differences between the 1975 economy and today's, and
differences between the states' economic situations then and now:
in the current downturn, states have been hurt both by the recession and
by the housing crisis, which heightened the need for state transfers to
local governments due to reduced municipal property tax revenue.
Concerns that state and local government responses to their budget
difficulties might undermine federal stimulus efforts are also
highlighted by historical considerations. (2) The classic paper by Cary
Brown (1956) shows that states did not conduct countercyclical policy
during the Great Depression. Subfederal governments still accounted for
the majority of government spending during that era, and in the
aggregate they tightened their budgets between 1931 and 1933 and
provided no net stimulus between 1933 and 1942.
Likewise, Kenneth Kuttner and Adam Posen (2001) show that during
Japan's "lost decade," fiscal efforts at the national
level were inconsistent, smaller than commonly thought, and undercut by
a variety of factors. Among these was that Japan's announced public
spending initiatives turned out to be significantly larger than what
ended up being implemented, because many of the announced federal
programs required partial local Hines's paper begins with a
demonstration that state activities in the recent downturn through 2009
provided virtually no net stimulus in the aggregate, after accounting
for federal grants. Other researchers have reported similar findings for
the recent downturn (for example, Aizenman and Pasricha 2010).
With this finding as background, the paper then investigates how
state policy has varied over the business cycle in the past, and reaches
several major conclusions. First, state-level fiscal policy is less
countercyclical than federal policy. This result is as expected, and
although the results do not show states running actively procyclical
policies as suggested by the simple balanced budget rule analysis above,
they still imply that the impact of federal countercyclical policies is
smaller than it would be if states acted exactly like the federal
government.
Second, large states tend to have more procyclical policies than
smaller states. This result may seem backward at first: one might expect
larger states to act more like the federal government (which can be
thought of as an aggregation of all states) and therefore to behave more
countercyclically than small states. However, Hines explains that
because smaller states face more elastic populations of taxpayers and
businesses, they need to be more responsive to local needs during
downturns, and hence need to run more countercyclical policies. In
particular, state spending on education and health, and revenue
collected through the income tax, are more sensitive to income
fluctuations in large states than in small states.
Third, most of the cyclical variation at the state level occurs
through changes in revenue rather than in spending. Aggregate state
spending is not particularly sensitive to income fluctuations, but taxes
are.
The fourth main result is that states have a high propensity to
spend federal grants, and this propensity is higher in large states than
in small. This is important given the key issue of whether federal
grants to states are stimulative. However, the extent to which grant
levels are endogenous with respect to state spending levels is difficult
to ascertain, which muddies the interpretation.
Policymakers were seeking answers on these issues as they were
constructing the American Recovery and Reinvestment Act of 2009, the
main federal stimulus package in the recent recession. If and when
another stimulus package is debated, whether and in what form to extend
aid to the states will continue to be a critical question. At the same
time, several other issues would be worth exploring further. The first
is the sensitivity of Hines's overall results to his use of the
whole 1951-2007 time period. The nation experienced only two recessions
in the 20 years preceding 2007, and the nature of the aggregate fiscal
response to the business cycle changed over that period (Auerbach, Gale,
and Harris 2010). It would be interesting to know how robust the results
are to a divided sample period.
Second, if the focus is on understanding state responses during
economic downturns, it would be interesting to know whether the response
to rises versus falls in income is asymmetric because of inertia or
political constraints. It is certainly plausible that the response of
states to a recession is not the exact opposite of their response to an
expansion.
Another issue of interest is how, precisely, states run
countercyclical policies, given their balanced budget rules* Do they
exploit built-in features of the rules (for example, a requirement that
the budget be balanced, but only on an ex ante basis)? Do they exploit
the fact that such rules typically apply only to the operating budget,
so that debt-financed capital expenditures, one classic type of
countercyclical policy, are still possible? Are they running budget
surpluses on average, building up reserves, and then draining their
rainy-day funds during hard times? Do federal grants in fact enhance
their ability to run countercyclical policies? Hines's paper
provides a foundation from which future research could explore all of
these issues.
REFERENCES FOR THE GALE COMMENT
Aizenman, Joshua, and Gurnain Kaur Pasricha. 2010. "On the
Ease of Overstating the Fiscal Stimulus in the US, 2008-9." Working
Paper no. 15784. Cambridge, Mass.: National Bureau of Economic Research.
Auerbach, Alan J., and William G. Gale. 2009. "Activist Fiscal
Policy to Stabilize Economic Activity." In Financial Stability and
Macroeconomic Policy. Federal Reserve Bank of Kansas City.
Auerbach, Alan J., William G. Gale, and Benjamin H. Harris. 2010.
"'Activist Fiscal Policy." Journal of Economic
Perspectives 24, no. 4:141-64.
Brown, E. Cary. 1956. "Fiscal Policy in the 'Thirties: A
Reappraisal." American Economic Review 46, no. 5: 857-79.
Congressional Budget Office* 2010. "Estimated Impact of the
American Recovery and Reinvestment Act on Employment and Economic Output
from July 2010 through September 2010." Washington (November).
Gramlich, Edward M. 1978. "State and Local Budgets the Day
after It Rained: Why Is the Surplus So High?" BPEA, no. 1: 191-216.
--. 1979. "Stimulating the Macro Economy through State and
Local Governments." American Economic Review 69, no. 2:180-85.
Kuttner, Kenneth N., and Adam S. Posen. 2001. "The Great
Recession: Lessons for Macroeconomic Policy from Japan." BPEA, no.
2: 93-185.
Poterba, James M. 1994. "State Responses to Fiscal Crises: The
Effects of Budgetary Institutions and Politics." Journal of
Political Economy 102, no. 4: 799-821.
Reischauer, Robert D. 1978. "Federal Countercyclical
Policy--the State and Local Role." Prepared for the Seventy-first
Annual Conference on Taxation of the National Tax Association-Tax
Institute of America, Philadelphia, Pa., November 13.
(1.) In an analysis of provisions included in the 2009 federal
stimulus package, the Congressional Budget Office (2010) estimates
output multipliers between 1.0 and 2.5 for transfers to state
governments for infrastructure spending, and between 0.7 and 1.8 for
transfers for other purposes.
(2.) The following two paragraphs are based on Auerbach and Gale
(2010). government funding that did not materialize. Also, coordination
issues with local government limited effective planning and
implementation.
COMMENT BY
BRIAN KNIGHT In this paper, James Hines investigates differences
between state and federal fiscal responses to business cycle
fluctuations and delves into cross-state differences in fiscal policy
over the business cycle.
This is obviously an important topic given both the current
economic climate and the fiscal challenges facing both the federal
government and state and local governments. Given this importance, I was
surprised to learn that Hines's was the first systematic comparison
of federal and state-local fiscal policy and thus was delighted to see
him tackle this issue. I will first summarize his key findings before
turning to some comments on the analysis and a discussion of the
results.
The paper's analysis is divided into two parts. The first is a
national analysis of quarterly national income and product accounts data
covering 1947-2010. Hines finds that federal spending and taxes exhibit
the expected countercyclical pattern, with spending rising and taxes
falling during recessions identified by the National Bureau of Economic
Research (NBER). This result is robust to using the output gap, as
defined by the Congressional Budget Office, as an alternative measure of
recessions. Hines's corresponding subnational measures of revenue
and spending are based on national aggregate data and combine the state
and local sectors into one. Using this measure, Hines finds some
evidence of a countercyclical pattern in state and local fiscal policy,
but one that is smaller in magnitude than the results for the federal
government. The state and local results are also less robust than the
corresponding federal patterns to measures based on the output gap.
Taken together, the key finding from this first analysis is that federal
fiscal policy is more countercyclical than state and local fiscal
policy.
The second part of the paper is based on annual state-level Census
data covering the years 1951-2007. Thus, unlike the aggregate analysis,
this analysis excludes data from the most recent recession. Also, this
analysis incorporates spending and taxes from state governments alone
and thus excludes data on fiscal policy at the local level. The key
findings here can be summarized as follows. First, fiscal policy does
not differ between states with and states without strict balanced budget
rules.
Second, there is a strong flypaper effect: the marginal propensity
of states to spend from federal grants is close to 1, and about 25 times
the marginal propensity to spend from other sources of income. Third,
state policy is more countercyclical in states with small populations
than in more populous states. Hines interprets this finding as
reflecting the fact that large states face less competitive pressure,
since voters in these states are less likely to vote with their feet.
That is, this pattern of procyclical policy in large states may reflect
the suboptimal propensity of politicians to spend, rather than save,
budget surpluses during economic booms, and pressures to save any
surpluses may be less relevant in large states. As corroborating
evidence for this "political failure" interpretation, Hines
shows that, conditional on population, fiscal policy is more procyclical
in states with less mobile populations, where voting with one's
feet is arguably more costly, and in states with high levels of
corruption, where these political failures are presumably more salient.
I found the results from the first analysis to be quite convincing
in terms of documenting that federal fiscal policy tends to be more
countercyclical than fiscal policy at the state and local level. This is
certainly consistent with normative principles, as outlined by Wallace
Oates (1999), who argues that, in most cases, stabilization policy is
best carried out at the central level. This result follows from the fact
that state and local governments have open economies and thus neither
the means nor the incentives to counteract economic downturns with
expansionary fiscal policy. Although this is the conventional wisdom in
the federalism literature, Edward Gramlich (1987), as discussed in more
detail below, puts forward a case for a more active macroeconomic
stabilization policy at the state and local level.
Given Hines's finding of a difference in countercyclicality
between the federal government and state and local governments, a
natural question is what the root causes of this difference might be.
One possible answer is that nearly all states have balanced budget
rules, which, by limiting borrowing during recessions, may constrain
their ability to conduct countercyclical fiscal policy. One weakness of
this interpretation, however, is that it is seemingly inconsistent with
another of Hines's findings, cited above, that state fiscal policy
is independent of the presence of strict balanced budget rules at the
state level. A second possibility derives from the fact that federal
taxation is largely based on income, whether individual or corporate,
whereas state and local tax bases are more based on sales and property.
To the extent that income is more volatile than either consumption or
land and housing values, then, holding tax rates fixed, federal tax
revenue will automatically exhibit more volatility over the business
cycle than do state and local tax revenues. Hines provides some evidence
in favor of this hypothesis in table 3, but the standard errors in these
regressions are quite large, making it difficult to reject the
hypothesis that revenues from income, sales, and property tax systems at
the state level exhibit similar patterns over the business cycle.
A third possibility, mentioned above, is that because states are
less able than the federal government to prevent the benefits of their
countercyclical policy from spilling over to other jurisdictions, they
have neither the means nor the incentive to counteract a weak economy.
But this idea is inconsistent with Hines's finding, also described
above, that fiscal policy is more countercyclical in less populous
states, which presumably have more open economies and thus less control
over economic activity within their borders than do larger states.
A final possibility, not explored in the paper, is that federal tax
systems may be more progressive than state tax systems on average, and
that this greater progressivity builds greater countercyclicality into
federal fiscal policy. All else equal, reductions in individual income
under a progressive system tend to move taxpayers into tax brackets with
lower tax rates and hence lower tax payments.
As mentioned above, Hines also finds strong evidence of a flypaper
effect. In interpreting this result, however, some lessons from the
existing literature on the flypaper effect are worth noting. First, many
federal grant programs have matching provisions. For example, the
federal share of Medicaid spending varies across states from a minimum
of 50 percent, consistent with a dollar-for-dollar federal match of
state spending, to around 75 percent, consistent with a 3-to-I federal
match. These substantial matching provisions introduce significant
substitution effects, and consequently one should not expect the
marginal propensity to consume from grants to equal the marginal
propensity to consume from income. That is, the standard theoretical
prediction of equivalence between grant income and private income is
based upon a model with purely lump-sum federal grants. Indeed, Robert
Moffitt (1984) investigates this issue in the context of the Aid to
Families with Dependent Children program (the predecessor of
today's Temporary Assistance for Needy Families) and shows that the
flypaper effect can be explained entirely by the program's matching
provisions. Second, several recent papers, such as Brian Knight (2002),
Nora Gordon (2004), and Byron Lutz (2010), argue that the flypaper
effects documented in the literature might be explained by endogeneity
problems. That is, if federal funds tend to flow to jurisdictions with
strong preferences for public spending, then the observed flypaper
effects may reflect the presence of this third factor. Moreover, results
using exogenous variation in grant receipts tend to exhibit
significantly weaker flypaper effects.
I found Hines's results on the differences in fiscal policy
between small and large states to be quite interesting, and the idea
that political failures of fiscal policy are more salient in large
states to be compelling. On the other hand, when one combines this
finding with the finding that federal fiscal policy is more
countercyclical, one is left with the puzzling result that it is fiscal
policy in small states rather than in large states that more closely
resembles federal fiscal policy. If differences between small states and
large states reflect the differing ability to vote with one's feet,
why is the inability to save surpluses during booms not even more of an
issue at the federal level, where the population is even less mobile?
This issue of differences in fiscal policy between small and large
states certainly deserves further exploration.
Regarding this comparison, a potential statistical issue that is
not discussed in the paper involves measurement error: any error
associated with measuring economic activity may be more pronounced in
small states than in large states. This might explain why state taxes
and spending more closely track state income in large states than in
small states. Although this differential measurement error seems
unlikely to entirely explain the differences in fiscal policy that Hines
documents, it may lead to their being overstated.
A related fiscal institution not explored in the paper is the
rainy-day fund, which, in the context of the author's
"political failure" story, can be interpreted as a mechanism
through which state governments can commit to saving during economic
booms for use during economic contractions. Indeed, many of these funds
have provisions designed to address political failure. For example, some
states require deposits into the rainy-day fund during favorable
economic times, and many allow withdrawals only during downturns.
Consistent with the idea that states are unable to commit to saving
during booms, Knight and Arik Levinson (1999) document that
contributions to these rainy-day funds tend to increase overall state
savings on a dollar-for-dollar basis and, moreover, that funds with
stricter rules regarding contributions and withdrawals tend to have
higher overall savings than funds with more lax rules.
A final issue that this paper raises involves regional business
cycles. The first part of the paper is framed around the national
business cycle and the associated differences in responses to national
shocks by the federal government and state and local governments. Yet
because some shocks are industry-specific, many recessions have
important regional components or might even be limited to one region.
Consideration of such spatial variation in business cycles gives rise to
a number of interesting issues that are relevant to the paper's
results. First, states and localities may be in a better position to
respond to local economic shocks than the federal government: state
officials may have better information about local economic conditions,
or, perhaps for political reasons, the federal government may be unable
to respond to highly localized economic shocks. Whatever the reason, the
better ability of states and localities to respond to local shocks
suggests an important role for state and local governments in
macroeconomic stabilization policy (Gramlich 1987).
Second, any federal response to localized economic downturns raises
moral hazard considerations. The expectation of increased federal
funding for states and localities in fiscal distress may provide
incentives for them to run excessive deficits and engage in other risky
fiscal practices. To the extent that the federal government cannot
commit to not bailing out state and local governments, strict balanced
budget rules may be a necessary ingredient of fiscal policy at the state
and local level.
Third, federal automatic stabilizers provide a type of insurance
against regional contractions: relatively vibrant regions
cross-subsidize regions suffering from economic downturns, with the
expectation that these roles will be reversed when the economic
conditions are reversed. In a careful examination of this issue,
Pierfederico Asdrubali, Bent Sorensen, and Oved Yosha (1996) document
that about 13 percent of shocks to gross state product are smoothed by
federal fiscal policy, with federal taxes, transfers, and grants all
providing significant contributions.
In summary, I found this to be a very interesting paper that makes
a significant contribution to the literature on stabilization policy and
federalism. I hope that this paper will help to stimulate research in
this area in the coming years.
REFERENCES FOR THE KNIGHT COMMENT
Asdrubali, Pierfederico, Bent E. Sorensen, and Oved Yosha. 1996.
"Channels of Interstate Risk Sharing: United States
1963-1990." Quarterly Journal of Economics 111, no. 4 (November):
1081-1110.
Gordon, Nora. 2004. "Do Federal Grants Boost School Spending?
Evidence from Title I." Journal of Public Economics 88, no. 9-10
(August): 1771-92.
Gramlich, Edward M. 1987. "Subnational Fiscal Policy."
Perspectives on Local Public Finance and Public Policy 3: 3-27.
Knight, Brian. 2002. "Endogenous Federal Grants and Crowd-Out
of State Government Spending: Theory and Evidence from the Federal
Highway Aid Program." American Economic Review 92, no. 1 (March):
71-92.
Knight, Brian, and Arik Levinson. 1999. "Rainy Day Funds and
State Government Savings." National Tax Journal 52, no 3
(September): 459-72.
Lutz, Byron. 2010. "Taxation with Representation:
Intergovernmental Grants in a Plebiscite Democracy." Review of
Economics and Statistics 92, no. 2 (May): 316-32.
Moffitt, Robert. 1984. "The Effects of Grants-in-aid on State
and Local Expenditures: The Case of AFDC." Journal of Public
Economics 23, no. 3 (April) 279-305.
Oates, Wallace E. 1999. "An Essay on Fiscal Federalism."
Journal of Economic Literature 37, no. 3 (September): 1120-49.
GENERAL DISCUSSION Alan Auerbach noted the lack of a distinction in
the paper between automatic and discretionary responses. Presumably some
state programs, like some federal programs, are automatically cyclical
in their expenditure. It would be interesting to know whether the
variation found by Hines relates to differences in these types of
programs across states. Are states behaving differently because some but
not others are deliberately responding to recessions, or is everything
on automatic pilot, and do the differences across states have to do with
different compositions of expenditure across these program types? Henry
Aaron agreed that the distinction is important, and he suggested that
the hypothesis that discretionary tax policy is almost unambiguously
procyclical would soon be tested in an effort by the Center on Budget
and Policy Priorities to monitor virtually every proposed state tax
increase.
Jonathan Parker argued that state income might not be the right
dependent variable. A state that cuts back a lot on spending may see a
much bigger rebound in income than one that cuts back little or not at
all. Also, very high income households account for a large amount of tax
revenue in some states. In New York, Wall Street bonuses alone accounted
for about $30 billion of state income in 2007, so that when bonuses fell
by more than half in 2008, it blew a multibillion-dollar hole in the
state budget. This raises the question of whether the recent high
cyclicality of high incomes is responsible. Parker noted that that would
be the case only for states and localities that rely on the income tax:
a state that derives most of its revenue from a progressive income tax
is going to have higher revenue cyclicality than one that relies on a
flat consumption tax.
Benjamin Friedman was struck by the differential effect that Hines
found between larger and smaller states. Because of spillover effects,
one might have thought a priori that countercyclical spending policy,
and especially discretionary spending, would be more effective in larger
than in smaller states. But then the differential would be in the
opposite direction from Hines's finding, leading to the question:
is the expected effect offset by other factors?
Steven Davis wanted to see more evidence supporting the theory that
smaller states are more efficient in their spending. For example,
Hines's argument suggests that smaller states would have more
efficient court systems. Is this in fact the case? If the theory is
correct, smaller states should perform better on average than larger
states on other attributes, too, even the choice of tax base and the
quality of public services.
Laurence Ball argued that it might be beneficial to consider a more
continuous measure of the business cycle, such as unemployment, rather
than a binary indicator like NBER recession dating. He suspected that
the results could be quite different. Justin Wolfers added that NBER
recessions are defined as times when the economy is getting worse, but
active fiscal policy is typically used both in the downswing and on the
rebound. Henry Aaron mentioned what he saw as a more fundamental point.
At least three of the recessions covered in the paper were caused by
cutbacks in defense outlays. This reverses the causation altogether and
undermines the relevance of the comparison between state and federal
countercyclicality.
Wolfers also noted that when the federal government wants to
respond to economic weakness, it can respond forcefully. Much of the
time, however, the federal government does not respond at all, perhaps
because of politics or gridlock. Aggregating across all states, then,
the response of the states may be much more reliable than that of the
federal government. Politics has not yet entered into the discussion,
but it certainly plays a role.
Robert Hall pointed out that before the early 1960s it was illegal
for state and local employees to unionize. Public sector unionization
has become important only in recent decades, so that today a
governor's job is to figure out how to make up the difference after
powerful unions have taken one large share of the revenue needed to
balance the budget, and stingy taxpayers have withheld another. In such
an environment it is hard to see how a state could possibly act
countercyclically. Following up on Parker's comment, Hall noted
that another big change is the greater dependence of many states on a
progressive income tax, which gathers abundant revenue in good times. He
also agreed with Ball that focusing on recessions was not the right way
of looking at the issue. The need for stimulus is after the end of a
recession. A better analysis would be one that compared bad times with
good times.
Gary Burtless argued that the discussion of countercyclical
stimulus was incomplete without a consideration of the revenue side. As
he understood it, the single biggest element of automatic
countercyclical stimulus is the loss of income tax revenue from the
corporate and personal income tax.
Burtless was also interested in the treatment of unemployment
insurance and its allocation between state and federal responsibility.
The UI program is essentially mandated by the federal government, but
states set the benefit levels and the tax rates on employers to pay for
the benefits. An unambiguously discretionary part of policy is the
decision whether to provide extensions to UI during downtums. That is
entirely a federal initiative, mandated and funded by the federal
government, although the state governments write the checks. If such
extensions are a large part of what the paper is counting as state
stimulus spending, it may not be an accurate characterization.
Refet Gurkaynak was curious whether the difference that Hines
observed between small and large states would be found among small and
large member states of the European Union. The available history is not
as long, but it would be informative to see whether the smaller European
states have behaved differently from the larger ones during the last
recession and its aftermath.
Donald Kohn noted that the depth of the recent recession and the
length and severity of the output gap likely caused many states to run
through their rainy-day funds, even if they had built them up as
planned. He also wondered whether the dependence of most states on real
estate taxes made a difference in this cycle. Because this recession
started and was deepest in the real estate sector, the resulting steep
loss in property taxes might have cut into state revenue deeper and more
quickly than in other recessions. To the extent that states were
constrained by a balanced budget, this would have forced them to cut
back spending more aggressively than otherwise.
Robert Gordon observed that state and local real spending as a
share of potential GDP dropped sharply between 1931 and 1933 and
continued to drop throughout the 1930s. Roughly half of the federal
stimulus in those years was thus offset by tightening at the state and
local level, even before including the tax side. The nation's
economic outlook was dismal in the second quarter of 1940, but then, in
a little over a year, the total government share of spending doubled
with the surge in wartime spending. In the recent period, the state and
local share of potential GDP has drifted downward steadily since 2003,
but the federal government has come nowhere close to offsetting this
decline. In fact, it is hard to see any increase in federal government
spending on goods and services through the second quarter of 2010. The
spending has all been in transfer payments, and most of that in
unemployment insurance.
