Information about http://www.stanford.edu/~wacziarg/downloads/pork.pdf
Quarterly Journal of Political Science, 2007, 2: 95106 Research…
Tags: central prediction, compromise version, economics department, empirical results, graduate school of business, gsb stanford, hauk, highway bill, journal of political science, malapportionment, pork barrel spending, small states, st columbia, stanford ca, stanford university graduate, stanford university graduate school, transportation equity act, university graduate school, university of south carolina, usd 286,Pages: 12
Language: english
Created: Fri Mar 16 11:50:49 2007
Display cached document
Page 1
Page 2
Page 3
Page 4
Page 5
Page 6
Page 7
Page 8
Page 9
Page 10
Page 11
Page 12
Quarterly Journal of Political Science, 2007, 2: 95106
Research Note
Small States, Big Pork
William R. Hauk, Jr.1 and Romain Wacziarg2
1
Economics Department, Moore School of Business, University of South Carolina, 1705
College St, Columbia, SC 29208 USA; hauk@moore.sc.edu; tel: (803) 777 6044.
2
Stanford University, Graduate School of Business, 518 Memorial Way, Stanford CA
94305, wacziarg@gsb.stanford.edu, tel: (650) 723 6069.
ABSTRACT
Using data on authorizations from the 2005 Highway Bill, we show that the leg-
islative allocation of pork barrel spending by U.S. state (measured by the value of
transportation earmarks per capita) greatly favors smaller states. We exploit the dif-
ference between two versions of the bill: the version that was passed by the House
and the compromise version passed in conference committee. Our empirical results
provide strong evidence in favor of theories of legislative malapportionment.
On 10 August 2005 President Bush signed a USD 286.4 billion transportation bill. The
bill, a reauthorization of the Transportation Equity Act for the 21st Century (TEA-
21), provides a renewable six-year plan to fund the nation's highway system. In total,
USD 24.2 billion, or 8.4% of the bill's total spending, were earmarked for 6373 specific
highway construction and improvement projects throughout the 50 states, the District of
Columbia and U.S. territories. These earmarks, often referred to as pork barrel spending,
constitute an ideal setting to analyze the incidence of small state overrepresentation
in the U.S. Congress. In this paper we test the hypothesis that small states receive
a disproportionate share of pork barrel spending, a central prediction of models of
legislative malapportionment.
A major innovation of our approach, compared to existing work, is to analyze different
versions of the bill as it progressed through the legislative process. This allows us to dis-
tinguish the effect of legislative malapportionment from the effect of increasing returns
to the provision of local public goods. Both effects would lead small states to receive more
spending per capita. However, if legislative malapportionment is at work, the small state
Supplementary electronic data for this article is available at
MS submitted 20 October 2005; final version received 4 October 2006
ISSN 1554-0626; DOI 10.561/100.00005048
© 2007 W.R. Hauk and R. Wacziarg
96 Hauk, Wacziarg
effect should be more pronounced in the Senate version, since small states are overrep-
resented to a much greater extent in the Senate relative to the House of Representatives.1
Consistent with this insight, we find that the effect of a state's population on per-capita
transportation pork expenditures is large, negative and significant in the final version of
the bill, after passage through the Senate and the conference committee. In contrast, the
effect is small and statistically insignificant in the version passed by the House of Repre-
sentatives, where the bill originated. These results are unchanged after holding constant
a number of presumptive determinants of earmarked spending. Thus, this paper offers
compelling empirical evidence supportive of models of legislative malapportionment, in
the context of a single, well-defined legislative initiative.
THEORY AND EVIDENCE ON LEGISLATIVE MALAPPORTIONMENT
The idea that the overrepresentation of small states in the U.S. Senate should lead to the
disproportionate allocation of public spending, trade protection, and other geographi-
cally targetable governmental activities to smaller states is neither new nor surprising.
Formal models of this process, however, are relatively rare, as most models of legisla-
tive policy-making assume that constituencies of equal sizes are equally represented.2
Departing from this assumption, Ansolabehere, Snyder, and Ting (2003) present a bar-
gaining model in a bicameral legislature with malapportionment. With a supermajority
rule in the malapportioned chamber (such as in the U.S. Senate), they predict that
transfers across states disproportionately benefit smaller states. In a different legislative
bargaining model, Knight (2004) notes that the advantage small states enjoy through
over-representation in the Senate gives them stronger bargaining power in the appropri-
ations process. Since their constituents will be paying a smaller share of tax revenues,
small states are more interested in expanding government spending. Finally, Hauk (2005)
presents a model explaining international trade protection in the context of a lobbying
model with a malapportioned legislature. Again, the benefits of trade protection in this
model are geographically targetable because the concentration of different industries
differs across legislative districts.
On the empirical side, a number of studies identify a correlation between inequalities
in representation and inequalities in the receipt of government benefits. In a study of
federal government spending by state, Atlas, Gilligan, Hendershott, and Zupan (1995)
observe that less-populated states tend to receive a greater share of federal funds than
their share of the population. They argue that this discrepancy is the consequence of U.S.
1
Due to single-member House delegations from seven states, House representation is also charac-
terized by some malapportionment, but obviously to a much smaller extent than the Senate, where
every state receives two Senators irrespective of population.
2
While the theoretical literature on malapportionment and pork is not very developed, this issue
relates to the broader literature on the determinants of Congressional spending.(e.g. Levitt and
Snyder 1997). The question of whether governments target spending to maximize vote share in
swing states, for instance, is addressed theoretically by Dixit and Londregan (1996) and empirically
by Dahlberg and Johansson (2002), among others.
Small States, Big Pork 97
Senate malapportionment. A simple finding that smaller states spend more per capita
is not, however, definitive evidence in favor of models of legislative malapportionment.
Such an effect could stem from the fact that public goods, by their very nature as non-
rival goods, involve increasing returns to scale: whether a bridge serves 50 people, as
an Alaskan bridge funded in the recent highway bill does, or several million, like the
Brooklyn Bridge, their costs remain commensurate, or at least less than proportional to
the population served.3
In a more direct test of legislative malapportionment, Ansolabehere, Gerber, and
Snyder (2002) use state and county-level data on government spending, noting that in
the aftermath of several Supreme Court decisions on apportionment that forced more
equal representation in state legislatures, spending and revenue transfers tended to flow
more equally across constituencies. Knight (2004) analyzes state-level data on earmarked
projects across several years and projects and finds a small-state bias in earmarks in
bills that were initiated by the Senate. Using similar data from a taxpayer watchdog
group, Herron and Shotts (2003) provide evidence that the U.S. Senate provides more
projects considered to be "pork" to smaller states.Their results are robust to the inclusion
of a number of covariates capturing different state characteristics. Hauk (2005) shows
that industry-specific international trade protection is biased towards industries that are
disproportionately concentrated in small states, due to malapportionment.
This paper marks an advance on the existing empirical literature by considering a
pork-laden highway bill and analyzing data on the small-state bias in earmarked projects
for both the House and conference committee versions of the bill. We are therefore able
to tell apart theories of legislative malapportionment from other theories predicting a
small state bias, such as those based on increasing returns to the provision of public
goods.
EARMARKS DATA FROM THE 2005 HIGHWAY BILL
The Transportation Bill was passed as a six-year authorization to fund highway projects
in 1991. Reauthorized in 1997, the Transportation Equity Act for the 21st Century
(TEA-21) expired in 2003.4 In April 2004 the U.S. House of Representatives passed a
reauthorization of TEA-21 as H.R. 3550, the "Transportation Equity Act: A Legacy
for Users (TEA-LU)". The bill was introduced by Don Young, Republican of Alaska
and Chairman of the House Transportation and Infrastructure Committee.5 TEA-LU
3
See Alesina and Wacziarg (1998) for a detailed discussion of this point, and accompanying cross-
country evidence. They note: "To the extent that public goods are of a non-rival nature, increasing
returns stem from the fact that, while the required level of provision is independent of population
size (or grows less than proportionately to it in the case of partial non-rivalry), the cost of public
goods can be spread over a larger pool of taxpayers in larger countries".
4
For a historical discussion of pork barrel politics in the context of transportation legislation up to
1991, see Evans (2004), chapter 4.
5
Don Young's name will be immortalized with the renaming of Knik Arm Bridge in Anchorage to
"Don Young's Way". Among the most egregious pork barrel projects in the 2005 Highway Bill is
98 Hauk, Wacziarg
contained about USD 11 billion in earmarked projects, but in July 2004 the bill died
in conference committee over disagreements between the House and Senate conferees
over the extent of total spending and over formulas to divide the non-earmarked funding
among states. Don Young reintroduced the bill in the House as H.R. 3 in February 2005,
and the House passed it on 10 March 2005 by a vote of 4179.
H.R. 3 was more successful that its predecessor. The Senate adopted a version of
the bill in May 2005, by a vote of 8911. The bill then made its way to conference
committee, where conferees from the House and Senate agreed on a single version in
late July. While we have limited information on how the conferees were selected in this
specific case, we note that all three members of Alaska's congressional delegation were
conferees. As we will show below, Alaska happens to have a relatively small population
and also received a lot of earmark dollars per capita. The "Safe, Accountable, Flexible,
and Efficient Transportation Equity Act of 2005" (SAFETEA) became law on 10 August
2005. By the time it became law, the total amount allocated to earmarked projects was
more than doubled, to USD 24.2 billion.6
In this paper we use the earmarks data compiled by the taxpayer watchdog group Tax-
payers for Common Sense (TCS). TCS has been monitoring earmarks in the successive
versions of the bill, starting with H.R. 3550. The group painstakingly compiled a com-
prehensive database of earmarks in three successive version of the bill: the 2004 House
version (H.R. 3550), the 2005 House version (H.R. 3) and the final version adopted
by Congress. These earmarks have been aggregated at the state level, i.e. we observe
aggregate spending per state, as well as the number of earmarks per state.7
the Gravina Bridge, connecting an island of 50 inhabitants to mainland Alaska. According to the
watchdog group Taxpayers for Common Sense, "this bridge will be nearly as long as the Golden
Gate Bridge and taller than the Brooklyn Bridge" (http://www.taxpayer.net/Transportation/
gravinabridge.htm).
6
An interesting issue arising in the 2005 Transportation Bill is the possible role of a veto threat from
the President. The White House initially indicated that a bill providing more than USD 256 billion
would be vetoed. It then increased this limit to USD 283 billion, and ended up signing the USD 286
billion bill. This suggests the veto threat was not very credible, particularly since this White House
was not known for its willingness to veto spending bills. Moreover, the veto threat pertained only to
the overall amount of spending in the bill, not to the earmarked amount or its allocation across states.
Third, the bill passed with such overwhelming margins in the House and Senate that a veto-override
might have been a realistic possibility. Thus, the presidential veto threat played virtually no role in
the cross-state allocation of earmarks. For a discussion of the effect of the presidential veto on pork
spending and its cross-jurisdictional allocation in a more general context, see McCarty (2000).
7
The data are available at http://www.taxpayer.net/Transportation/safetealu/states.htm (we used
the version posted on 12 August 2005). We do not observe earmarks from the Senate version of the
bill, as there were none in that version. Therefore, all the Senate earmarks were added in conference.
One possible explanation for the fact that the Senate version did not include any earmarks is that
Senators are reluctant to go on the record as favoring specific projects in their home states, since that
might draw the ire of constituents in areas not receiving any earmarks. Instead, introducing Senate-
favored earmarks in conference committee makes it hard for voters to tell whether they originated
from House members (with narrow constituencies) or Senators (with statewide constituencies).
Empirical tests of this hypothesis could exploit the fact that there is variation in the degree of
homogeneity of House districts: in districts that include both urban and rural interests, for instance,
House members should be less willing to go on record as favoring earmarks benefiting a specific
Small States, Big Pork 99
Our data have several advantages over those previously used. First, transportation
earmarks are the quintessential example of a geographically targetable form of public
expenditure.8 Past studies have often used broader categories of spending (such as total
public spending per state) that are less easily targetable by members of Congress (for
instance, entitlements are harder to target geographically). Second, we observe all ear-
marks included in the bill, i.e. we do not select a specific category of earmarks in our
analysis, limiting the potential for subjective judgement calls as to what constitutes pork
barrel spending. Third, we observe several versions of the same bill, so we can separately
analyze the determinants of state per capita spending in the different versions. Past stud-
ies, in contrast, did not exploit these differences.9 Thus, the 2005 Transportation Bill
provides an ideal setting in which to test theories of legislative malapportionment.
EMPIRICAL RESULTS
Our econometric specification is the following:
log Gi = 0 + 1 log Si + 2 Xi + i (1)
where Gi is earmarked spending per capita in USD at the state level, Si is state population
from the 2000 Census, and Xi is a vector of control variables, further detailed below.
We use various series for Gi , corresponding to the three observed versions of the bill.
Equation 1 is estimated using ordinary least squares with robust standard errors, to
correct for the possible incidence of heteroskedasticity in i . In Equation 1 the proper
interpretation of 1 is the percentage increase in per capita pork barrel spending resulting
from a 1% increase in state population.
Table 1 presents some summary statistics for the main variables in our analysis. Unsur-
prisingly, Alaska is the biggest recipient of per capita earmarks no matter which version
of the bill is considered (the final version earmarks USD 1508.704 of spending per capita
for Alaskan projects). In the final version, Arizona is the smallest recipient, with only
USD 23.177 per capita. Table 2 presents correlations for the main variables. Per capita
earmarks are highly correlated across the three versions of the bill: the correlation is
99.6% between the 2004 and 2005 House versions, but lower at 87.8% between the
House and conference versions in 2005. Moreover, the correlation between earmarks
per capita and state population is negative but small for the House version (-13%), but
segment of their electorate, relative to Representatives from more homogeneous districts. Such a
test is beyond the scope of this short paper.
8
We are not claiming that they are perfectly targetable. For instance, a bridge built in New York state
could benefit residents of New Jersey traveling to New York, and some transportation dollars could
be spent on out-of-state suppliers and contractors. We are claiming that transportation expenditures
are more easily targeted geographically than most other types of public expenditures.
9
Moreover, since we are relying on differences in the bill between different chambers, we can rule
out Electoral College effects: since Electoral College votes are the sum of House and Senate mem-
bership, most existing studies cannot differentiate between the effects of legislative representation
and Electoral College weight.
100 Hauk, Wacziarg
Table 1. Summary statistics for the main variables (51 observations)
Variable Mean Std. Dev. Min Max
Earmarks per capita, 52.252 120.964 0.000 (SD) 889.309 (AK)
House version 2004
Earmarks per capita, 62.785 148.018 14.226 (DE) 1087.754 (AK)
House version 2005
Earmarks per capita, 158.805 219.904 23.177 (AZ) 1508.704 (AK)
Conference version
2005
Population, 2000 5811.969 6559.095 509.294 (WY) 36,132.147 (CA)
census, thousands
State personal income 32,340.350 5207.890 24,650.000 (MS) 51,803.000 (DC)
per capita, 2005
Table 2. Correlations among the main variables (51 observations)
Earmarks per
Earmarks per Earmarks per capita, Population,
capita, House capita, House Conference 2000 census,
version 2004 version 2005 version 2005 thousands
Earmarks per 0.996 1.000
capita, House
version 2005
Earmarks per 0.876 0.878 1.000
capita,
Conference
version 2005
Population, -0.121 -0.130 -0.318 1.000
2000 census,
thousands
State personal 0.100 0.109 0.039 0.101
income per
capita, 2005
* Denotes statistical significance at the 5% level.
larger in magnitude (-31.8%) when considering the conference version. This is prelim-
inary evidence in favor of a legislative malapportionment effect in the allocation of pork
barrel spending across states.
Table 3 presents results from estimating Equation 1 using earmarks data from the
House versions of the bill. Column 1 presents the simplest specification, without con-
Small States, Big Pork 101
trols. Our main coefficient of interest, on the log of population, is negative but statistically
insignificant. The subsequent columns add control variables sequentially. We first add
controls representing a state's transportation spending requirement, in column 2. On the
one hand, states that already have a lot of infrastructure might need less incremental
spending. On the other hand, maintenance and upkeep expenditures might imply a posi-
tive effect of installed infrastructure.To control for the density of installed infrastructure,
we include the total mileage of open highways divided by the state's surface area.10 The
coefficient on this variable comes out positive and highly significant. Geographically
large states may also require more transportation spending per capita. This hypothesis
is borne out, as the log of land area bears a positive and significant sign a 1% increase
in state area is associated with a 0.328% increase in per capita spending.
Next, we control for the log of a state's per capita personal income, obtained from
the U.S. Bureau of Economic Analysis (column 3). The coefficient on this variable turns
out to be statistically indistinguishable from zero. We also hypothesize that states with
a large Republican congressional representation might be at an advantage in the allo-
cation of pork barrel spending, since the Republican Party controlled both houses of
Congress as well as the Presidency. We worry that if Republican representatives tend to
be elected predominantly from smaller states, our estimated size effect may capture in
part a majority party effect. Thus, we control for the share of Republicans in the state's
House delegation, as well as the number of Republican Senators in a state's Senatorial
delegation (column 4). Contrary to our expectations, the proportion of Republicans in
the House representation of a state is negatively related to state per capita earmarks (as
expected, the number of Republican Senators is unrelated to spending in the version
of the bill initiated by the House). Thus, Republican House representation is associated
with pork barrel spending restraint when it comes to the 2005 Highway Bill, although the
effect is small in magnitude: a one standard deviation increase in the share of Republicans
in the state House delegation is associated with a 0.132 point increase in the log of pork
per capita, which represents only 3.709% of the mean of this variable. While one should
not over-interpret such a small estimated effect, it is interesting to note that majority
status does not seem to confer a special ability (or desire) to direct pork disproportion-
ately to the majority party's constituents.11 In column 5 we add the proportion of a state's
House delegation that is on the House Transportation Committee, hypothesizing that
strong representation of a state on this committee might be positively associated with
10
The data on highway length per state in 2003 are from the U.S. Department of Transportation. Our
results are not sensitive to including variables based on alternative measures of installed roadway
infrastructure by unit of land area.
11
A similar observation, referring to pre-1991 earmarks in various pork-laden bills, appears in Evans
(2004). For instance, on p. 226, she states "In nearly every case (. . .), projects were distributed in
a bipartisan manner. . . . Leaders give to members of their own party to hold their loyalty through
various challenges on the floor. They also give to members of the other party, who might be won
by pork barrel benefits away from support for their own party's position if it is in opposition to
the bill". The fact that the 2005 Highway Bill was passed with such overwhelming margins is an
indication that party affiliation may not have mattered much for the allocation of highway funds,
including earmarked funds. Why the majority party does not seem to receive a disproportionate
share of earmarks, however, remains unknown, and beyond the scope of this paper.
102
Table 3. Determinants of earmarks per capita in the House versions of the Transportation Bill (dependent variable: total
value of earmarks per capita, in USD, per state)
(1) (2) (3) (4) (5) (6)
2005 2005 2005 2005 2005 2004
House House House House House House
version version version version version version
Log of population, 2005 -0.112 -0.160 -0.182 -0.194 -0.097 -0.101
(0.119) (0.132) (0.145) (0.149) (0.058) (0.067)
Open highway miles/land area in km2 6.844 6.172 6.181 0.480 2.745
(2.124) (1.580) (1.423) (1.578) (1.528)
Log of land area, km2 0.328 0.355 0.404 0.162 0.232
(0.174) (0.188) (0.178) (0.065) (0.076)
Log state personal income per capita, 2004 0.839 0.846 0.236 -0.270
(0.975) (1.143) (0.456) (0.530)
Percent Republicans in House delegation -0.556 -0.427 -0.115
(109th Congress) (0.250) (0.196) (0.184)
Number of Republican Senators 0.002 -0.032 -0.112
(109th Congress) (0.099) (0.065) (0.063)
Share of state representatives on House 2.149 1.793
Transportation Committee (0.596) (0.615)
Constant 5.410 2.096 -6.563 -6.715 0.695 4.957
(1.867) (1.072) (10.212) (11.919) (4.607) (5.363)
Observations 51 51 51 51 51 50*
Adjusted R-squared 0.02 0.26 0.27 0.31 0.71 0.64
Robust standard errors in parentheses.
* South Dakota was dropped when taking logarithms because that state had zero earmarks per capita in the 2004 House version.
Hauk, Wacziarg
Small States, Big Pork 103
that state's earmarks per capita. This is indeed the case, as the estimated coefficient is
positive and highly significant statistically. Moreover, this variable adds greatly to the
explanatory power of our model, as adding it to the regression raises the adjusted R 2
from 0.31 to 0.71.
As a final check, we considered earmarked spending per capita in the unsuccessful 2004
version of the Transportation Bill as a dependent variable (column 6). Unsurprisingly,
given the high correlation between this variable and its 2005 counterpart, the results are
not greatly affected. No matter which specification is considered, the coefficient on the
log of population remains negative but insignificant and small in magnitude.
We now turn to the determinants of earmarked spending per capita in the final version
of the bill the version that came out of conference committee and became law. The
bill has now been the subject of a compromise with the Senate, where representation is
malapportioned. Table 4 presents the results. We proceed as before, including controls
sequentially. The estimated coefficient on the log of population is again negative, but it
is now highly significant statistically and it is about six times larger than in the House
version, irrespective of the specification: the malapportionment effect has appeared. Our
estimates suggest that a 1% increase in a state's population is associated with a 0.595
percentage point decrease in pork barrel spending per capita, in the baseline specification
of column 5. This is an economically large effect. The effect is estimated with great
precision, as the t-statistic always exceeds 7 across specifications. Moreover, the log of
population alone can explain 63% of the variation in the dependent variable (column 1),
whereas this number was essentially zero in Table 3.
As expected, the effect of the number of Republican senators in each state's delegation
becomes larger in magnitude than in Table 3, and is now significant at the 5% level
(column 5). Thus, a larger Republican senatorial delegation seems associated with less
pork barrel spending, although the effect of the share of Republicans in the House
delegation becomes statistically insignificant. The effect of the installed highway density
also becomes insignificant in Table 4. Coefficients on other control variables are in line
with those found using the House version (a positive effect of the log of land area and
an effect of log personal income per capita that is statistically indistinguishable from
zero). In column 5 we include two variables reflecting the composition of the conference
committee: the share of a state's representatives that are on the conference committee, and
the number of a state's Senators on the conference committee. Both variables come out
statistically significant, and their inclusion raises the explanatory power of the regression.
However, they only marginally impact the coefficient on the log of population, which
falls from 0.655 to 0.595. Thus, we find only weak evidence that small states derive
their ability to attract disproportionate earmarks through strong representation on the
conference committee, the Alaskan example notwithstanding.12
12
Further, including in our regressions the share of a state's representation on the House transporta-
tion committee and the number of Senators on the Senate Transportation Committee did not change
the results. The first variable, which is highly correlated with conference committee representation,
was positively related to earmarks per capita, while the second variable was insignificant.
Table 4. Determinants of earmarks per capita in the Conference version of the 2005 Transportation Bill (dependent variable: 104
total value of earmarks per capita, in USD, per state)
(1) (2) (3) (4) (5) (6) (7)
2005 2005 2005 2005 2005 Excluding Excluding
Conference Conference Conference Conference Conference Alaska Arizona
version version version version version
Log of population, 2005 -0.585 -0.628 -0.644 -0.655 -0.595 -0.564 -0.590
(0.081) (0.084) (0.088) (0.086) (0.044) (0.048) (0.044)
Open highway miles/land 1.935 1.444 1.883 1.540 1.915 1.833
area in km2 (1.434) (1.350) (1.294) (1.208) (1.210) (1.181)
Log of land area, km2 0.164 0.184 0.245 0.229 0.197 0.238
(0.106) (0.110) (0.115) (0.063) (0.063) (0.064)
Log state personal income 0.613 0.284 0.259 -0.110 0.177
per capita, 2004 (0.612) (0.667) (0.385) (0.456) (0.382)
Percent Republicans in House -0.329 -0.183 -0.230 -0.177
delegation (109th Congress) (0.206) (0.165) (0.157) (0.164)
Number of Republican -0.146 -0.133 -0.154 -0.120
Senators (109th Congress) (0.085) (0.063) (0.061) (0.060)
Share of state in Conference 0.507 0.241 0.440
Committee (0.224) (0.261) (0.221)
Number of Senators on 0.416 0.369 0.395
Conference Committee (0.080) (0.087) (0.078)
Constant 13.513 12.191 5.863 9.059 8.207 12.012 8.882
(1.248) (1.409) (6.345) (6.835) (3.870) (4.604) (3.873)
Observations 51 51 51 51 51 50 50
Adjusted R-squared 0.63 0.65 0.65 0.69 0.83 0.79 0.86
Hauk, Wacziarg
Robust standard errors in parentheses.
Small States, Big Pork 105
As a final robustness check on our findings, we carried out an outlier analysis. Plot-
ting the residuals from the regression in column 4 of Table 4, two apparent outliers
are revealed: Alaska's earmarked spending is vastly underpredicted by the model and
Arizona's is somewhat overpredicted. The Alaska case is particularly interesting because
the entire Alaskan congressional delegation was part of the conference committee, and
Alaska is a small state in terms of its population. Excluding each of these states from
the sample, however, hardly changes the results at all (columns 6 and 7 of Table 4).
In fact, excluding Arizona slightly raises the coefficient on log population. Excluding
Alaska reduces it slightly, although the precision of the estimate is raised. At any rate, the
Alaskan example is a basket case of the effects of legislative malapportionment.
CONCLUSION
In this paper we presented a simple empirical test of theories of legislative malapportion-
ment. Using data from the 2005 Highway Bill, we showed that earmarked spending on
transportation infrastructure projects is disproportionately allocated to smaller states.
However, this effect only appears after the bill made its way through the Senate and
Conference Committee, and is not apparent in the House version of the legislation. This
provides direct evidence that legislative malapportionment is primarily responsible for
the small state bias in pork barrel spending.
Future research should seek to generalize our findings to other types of geographi-
cally targetable expenditures and policies. Using data on different versions of the same
legislation as it makes its way in the legislative process is a promising way to identify the
effects of institutional rules on public policy.
ACKNOWLEDGMENTS
We thank Keith Ashdown, Keith Krehbiel, Ken Shotts, Erich Zimmerman, the editors
and two anonymous referees for useful comments. Taxpayers for Common Sense gener-
ously made public their earmarks data. The dataset used in this paper is available upon
request, and all errors are ours.
REFERENCES
Alesina, Alberto, and Romain Wacziarg. 1998. "Openness, Country Size and Government", Journal of
Public Economics 69 (September): 30521.
Ansolabehere, Stephen, Alan Gerber, and James Snyder. 2002. "Equal Votes, Equal Money: Court
Ordered Redistricting and the Distribution of Public Expenditures in the American States",
American Political Science Review 96 (December): 76777.
Ansolabehere, Stephen, James Snyder, and Michael Ting. 2003. "Bargaining in Bicameral Legislatures:
When and Why Does Malapportionment Matter?", American Political Science Review 97 (August):
47181.
106 Hauk, Wacziarg
Atlas, Cary M., Thomas Gilligan, Robert Hendershott, and Mark Zupan. 1995. "Slicing the Federal
Government Net Spending Pie", American Economic Review 85 (September): 62429.
Dahlberg, Matz, and Eva Johansson. 2002. "On the Vote Purchasing Behavior of Incumbent Govern-
ments", American Political Science Review 96 (March): 2740.
Dixit, Avinash, and John Londregan. 1996. "The Determinants of Success of Special Interests in
Redistributive Politics", Journal of Politics 58 (November): 113255.
Evans, Diana. 2004. Greasing the Wheels: Using Pork Barrel Projects To Build Majority Coalitions in
Congress. Cambridge: Cambridge University Press.
Hauk, William R. 2005. "Protection with Many Sellers: An Application to Legislatures with Malap-
portionment", Working Paper, Stanford University.
Herron, Michael C. and Kenneth W. Shotts. 2003. "Winners, Losers and the Congressional Pork
Barrel", Working Paper, Northwestern University.
Knight, Brian. 2004. "Legislative Representation, Bargaining Power, and the Distribution of Federal
Funds: Evidence from the U.S. Senate", NBER Working Paper No. 10385 (March).
Levitt, Steven, and James Snyder. 1997. "The Impact of Federal Spending on House Election Out-
comes", Journal of Political Economy 105 (February): 3053.
McCarty, Nolan. 2000. "Presidential Pork: Executive Veto Power and Distributive Politics", American
Political Science Review 94 (March): 117129.
Taxpayers for Common Sense. 2005. "Database of Earmarks in Conference Agreement to the Trans-
portation Bill", http://www.taxpayer.net/Transportation/safetealu/states.htm (August 12, 2005).