The Effect of Spending Levels on the Process of Distributive Politics Barry S. Rundquist Department of Political Science (M/C 276) Universit y of Illinois at C hicago Chicago, IL 60607 [email protected]and Thomas M. Carsey Department of Political Science Florida State University Tallahassee, FL 32306 [email protected]Prepared for delivery at the 2002 Annual Meeting of the American Political Science Association, Boston, MA, August 29 - September 1, 2002. Copyright by the American Political Science Association.
Transcript
The Effect of Spending Levels on the Process of Distributive Politics
Barry S. RundquistDepartment of Political Science (M/C 276)
University of Illinois at ChicagoChicago, IL 60607
Prepared for delivery at the 2002 Annual Meeting of the American Political Science Association,Boston, MA, August 29 - September 1, 2002. Copyright by the American Political ScienceAssociation.
1 This research was supported in part by a grant from the National Science Foundation(SES#98-9370). We also want to thank Lisa Schmit, who worked very hard to compile most ofthe data used here.
1
The Effect of Spending Levels on the Process of Distributive Politics
Introduction
Distributive theory argues that the re-election goals of members of Congress (MCs)
combine with the institutional rules of legislative decision making to produce the resulting
geographic distribution of federal expenditures.1 Specifically, scholars have found that
committee representation, MC partisanship and ideology, constituency need and demand (see
Rundquist and Carsey 2002 for evidence and a review), and even the efforts of organized interest
groups (Lowry and Potoski 2002) may all play a part in this process. Following work by Lee
(1998, 2000) and Lee and Oppenheimer (1999), we recently argued that the difference in
representation between the House and Senate should also be incorporated into the set of
institutional rules that may structure the distribution of policy benefits (Carsey and Rundquist
2002).
In this paper, we seek to define several new fronts for the continuing study of how the
institutional structure of policy making in Washington relates to the geographic distribution of
policy benefits and provide some exploratory analysis and findings. Our initial question centers
on how the overall level of spending in a policy area might affect the process of distributive
politics. As part of this, we intend to look specifically at the small state advantage stemming
from Senate representation. Our second consideration deals with whether the Republican take-
over of Congress after the 1994 election fundamentally altered the process of distributive
politics. In this regard, we intend to look specifically at the small state advantage stemming from
2
Senate representation. Our second consideration deals with whether the Republican take-over of
Congress after the 1994 election fundamentally altered the process of distributive politics.
Throughout this study we conceive of policy benefits as the year-to-year changes in per capita
federal expenditures received by counties. We conduct our analysis using federal spending data
in five substantive policy areas: agriculture, crime, defense, health, and transportation. Our
analysis covers the period from 1983 through 1996. Again, the analysis is exploratory, so the
results are necessarily tentative.
Levels of Spending and Distributive Politics
Most studies of distributive politics take the level of overall programmatic spending as
given and focus on the factors that affect the distribution of those funds (e.g. Carsey and
Rundquist 2002, Levitt and Snyder 1995). Others consider the number of programs rather than
dollar amounts, but again take the overall size of the government’s effort in a policy area as given
(Bickers and Stein 1995). To the degree that the overall level of spending is considered in
distributive models, it takes the form of an assumption or conclusion that distributive politics is
merely a synonym for pork barrel politics, which by definition involves wasteful or unnecessary
spending and therefore higher levels of programmatic spending that would occur in the absence
of distributive politics. In other words, evidence of distributive politics is assumed to lead to
higher than optimal levels of spending in a policy area.
There are at least two conceptual/theoretical problems with this view of distributive
politics. First, distributive studies are plagued by the problem of defining “optimal,” thereby
making empirical tests of this assumption difficult at best (see Ferejohn 1974, and Weingast,
3
Shepsle, and Johnson 1981 for relevant discussions). We have begun to address the question of
the efficient and/or effective distribution of policy benefits in military procurement (Rundquist
and Carsey 2002), but our own work focuses on where the benefits are directed and not on the
overall level of spending. That is, we consider the efficiency of military procurement in terms of
whether funds are directed to places with the capacity for producing military goods and services
and not in terms of the overall level of procurement contracting. The result is that a fundamental
assumption of traditional distributive theory remains insufficiently addressed.
The second issue concerning the distributive politics/level of spending relationship is a
failure to consider that the existence of, or nature of, distributive politics in a policy area may
depend in part on the overall level of spending in that area. Again, we have offered speculation
that the amount of money being spent for military procurement may affect the distributive
politics of this policy area (2002), and we have produced some preliminary work for other policy
areas (Rundquist and Carsey 2000), but a great deal more work conceptual and empirical work is
necessary. We can think of several ways in which the overall level of spending in a policy area
might affect the presence or nature of distributive politics operating within that policy area. But
first, we need to clarify what we mean by distributive politics.
Distributive politics is a process of allocating policy benefits that predicts a particular
geographic pattern to those benefits. The process is driven by the re-election concerns of MCs
and is shaped by the institutions that structure Congressional policy making. This definition
differs from earlier conceptions of distributive politics that tended to define the concept in terms
of a type of policy rather than a policy-making process. (e.g. Lowi 1964). Distributive theory
predicts that whatever rules and institutions structure policy making will result in some
4
geographic constituencies receiving disproportionately larger shares of policy benefits while
others receive less. Following Lowi’s (1964) assertion that committees structure distributive
policy making while parties play a more prominent role in regulatory and redistributive policy
making, early studies of distributive politics focused on whether committee representation was
the fundamentally important institutional factor determining the distribution of policy benefits
(e.g. Rundquist and Ferejohn 1975; Arnold 1979). Others have argued that political parties
provide the fundamental structure to policy making (e.g. Aldrich 1995; Cox and McCubbins
1993), including what Lowi would call distributive policies. More recently, scholars have argued
that thinking about distributive politics and policy making as structured by either committees or
parties may be misleading, and that what might be more prominent is a hybrid theory involving
both institutional structures (e.g. Shepsle and Weingast 1995, Rundquist and Carsey 2002,
Carsey and Rundquist 1999a).
Returning to our question of principle interest, one way in which the party versus
committee argument might be tied back to Lowi’s (1964) original argument may be to consider
the overall level of spending within a policy area. Lowi assumed a committee-dominated process
because the kinds of policies he was describing were expected to produce little or no conflict
within Congress. In other words, Lowi implicitly assumed that what he thought of as distributive
policies were not worth fighting for at a partisan or ideological level. We have speculated that
one reason why we uncovered evidence of a hybrid model of party and committee effects in the
distribution of military procurement benefits was because the amount of money involved is worth
fighting for (Rundquist and Carsey 2002). More specifically, the total amount of policy benefits
available in a policy area might influence whether the distribution of those benefits is structured
5
by committee representation or by majority versus minority party status (Rundquist and Carsey
2000).
Another way in which the level of programmatic spending might be related to the nature
of distributive politics centers on whether the process of distributive politics is essentially
proactive or reactive. Early work on focusing on committee representation devoted considerable
attention the recruitment and benefit hypotheses (e.g. Arnold 1979, Rundquist and Ferejohn
1975, Rhode and Shepsle 1973). The recruitment hypothesis states that MCs from places that
receive substantial benefits from a particular policy area will seek out committee assignments
that allow them to protect those benefits. In contrast, the benefit hypothesis states that MCs who
serve on committees that oversee spending in a particular policy area are able to direct a
disproportionate share of those benefits back to their geographic constituencies. Thus, the
recruitment hypothesis implies that distributive politics is essentially a reactive process – MCs
try to protect existing benefits but do not proactively direct new or increased benefits back to
their districts. The benefit hypothesis implies the opposite – that distributive politics results from
the proactive efforts of MCs. Sorting out this particular puzzle requires a multi-equation model,
and our own work suggests evidence that both processes are at work (Carsey and Rundquist
1999a, 1999b, Rundquist and Carsey 2002).
However, this stream of research points to another possible forum for untangling the
proactive vs. reactive nature of distributive politics at it relates to the overall level of spending.
Specifically, periods of programmatic expansion and budgetary growth create the circumstances
for MCs to proactively pursue the capture of new benefits. If committees or parties are the
institutional tools used by MCs in a proactive way in distributive politics, they should play a
6
more prominent role in determining the geographic distribution of policy benefits when budgets
are growing. In contrast, if distributive politics is essentially the politics of using party and/or
committee representation to protect current policy benefits, then the effect of committee and
party representation should be most evident in periods of overall budgetary decline. When the
budget is shrinking, it should shrink less in those places represented by MCs in a position to
protect current benefits.
The Small-State Advantage
Lee (1998, 2000) and Lee and Oppenheimer (1999) argue that the difference in how seats
are apportioned in the U.S. House and the U.S. Senate creates a small-state advantage in the
distribution of national policy benefits, at least in per-capita terms. This stems directly from the
fact that every state – whether small and large in population – is equally represented in the
Senate. This creates a disproportionate level of representation per capita in Congress overall for
small states which produces higher levels of benefits per capita for small-state residents. This
small-state bias leads to a preference among small-state senators for allocation formulas and
decision rules based on equal distribution across states as opposed to population-based formulas
(Lee and Oppenheimer 1999). This may be further enhanced by the tendency for Senators from
small states to be more attractive coalition partners when it comes to assembling majority support
for policy expenditures. Compared to senators from larger states, small state senators can be
induced to join the majority coalition for a cheaper aggregate price in programatic benefits (Lee
2000).
Lee and Oppenheimer (1999) argue that the small-state effect should be evident in
7
formula-based expenditures that are not subject to annual re-authorization or short-term
Congressional or bureaucratic manipulation. This suggests that inter-chamber differences in
representation operate at times and within policy areas that are precisely the opposite of those
times and policy areas where distributive politics is traditionally expected to operate. In a recent
paper, we find evidence supporting the small-state advantage hypothesis, but we find it operating
across policy areas and regardless of whether expenditures are formula based or not (Carsey and
Rundquist 2002). We also find evidence that committee representation interacts with the small-
state advantage, though the form that interaction takes varies across policy areas. In that paper,
we conclude that the small-state advantage implied by the equal representation states receive in
the Senate should be integrated into distributive theory rather than offered as a contrast to it.
Linking back to our general theory of distributive politics, one question implied by this
conclusion is whether the small-state advantage works principally in a proactive or reactive
manner. As with the recruitment versus benefit hypotheses regarding committee representation,
small states may reap rewards from their Senate representation primarily from Senators reactively
protecting the current benefits their states receive. Or small states may benefit from their
senators proactively producing new benefits. In other words, changes in the overall level of
expenditures in a policy area might provide a setting in which to evaluate the proactive versus the
reactive hypothesis regarding both intra- and inter-institutional structures in congressional policy
making.
Majority Party Status
As mentioned above, scholars have increasingly turned to political parties as the
8
institutions providing structure to distributive politics. Aldrich, for example, argues that the
collective action problems associated with distributive policy making create the incentives for
MCs to construct and maintain parties in Congress (1995). Levitt and Snyder (1995) uncover
evidence of a majority party effect on the geographic distribution of federal outlays. Others point
more to a hybrid model of parties and committees shaping distributive politics (Rundquist and
Carsey 2002, Shepsle and Weingast 1995). The basic argument is that committee members
themselves do not constitute a sufficiently large block to guarantee passage of their preferred mix
of policy benefits. Universal logrolling is an option (e.g. Weingast 1979), but greater gains from
exchange can be captured if a less than universal majority coalition can be established and
maintained. Parties server this function.
The presumption then is that members of the majority party in Congress should benefit
overall from the distribution of policy benefits, and that within any one policy area, it will be
majority party members represented on the committees with jurisdiction over that policy area that
benefit. We provide evidence of this for military procurement (Carsey and Rundquist 1999a,
Rundquist and Carsey 2002). Preliminary evidence for other policy areas, as shown below, is
more mixed.
Most studies of distributive politics, however, employ data from a period in which the
House of Representatives was controlled exclusively by the Democratic party. Some discussion
is devoted to the GOP control of the Senate in the early 1980s (e.g. Carsey and Rundquist
1999a), though most distributive studies focus exclusively on the House. Thus, most of what we
think we know about the importance of party control of Congress is based on evidence gathered
when it was always the same party in control. History has provided us with a better means of
2 Of course, numerous studies exist that examine the role of party control of stategovernments. See Winters 1976; Erikson et al. 1989; Dye 1984; ????
9
testing party control theories, however, since the GOP took control of both chambers as a result
of the 1994 elections.2 Bickers and Stein (2000) offer a first look at this by comparing
expenditure patterns under the 103rd and 104th Congresses. They conclude, “Republican control
of Congress does not appear to have significantly altered the politics of domestic spending. But .
. . Republican control has influenced the content of domestic policy.” (p. 1084). By this they
mean that the 104th Congress under GOP control increased the amount of spending on contingent
liability programs while reducing spending in most entitlement programs except those that award
benefits directly to individuals. However, political factors like partisanship, freshman status,
incumbency, or electoral vulnerability did not appear to shape the patterns of spending in the
104th Congress any differently than they did in the 103rd. Instead, business cycle factors and
district characteristics were the biggest determinants of where funds were allocated differently
before and after the GOP takeover. This suggests that the geographic distribution of
expenditures would only be altered to the degree that beneficiaries of contingent liabilities are
clustered in different areas from beneficiaries of non-individual entitlement programs. This
contrasts with a recent Associated Press report that the GOP takeover after 1994 has produced a
“seismic shift” in federal spending, moving tens of billions of dollars from Democratic to GOP
districts. . .” (Pace 2002). In the analysis that follows, we examine the impact of the GOP
takeover on the geographic distribution of policy benefits and whether the factors that influence
that distribution changed. Like Bickers and Stein, our analysis only contains data for the first
two years after the GOP takeover. This necessarily limits our findings, but it does provide some
10
insight into the short-term impact of a relatively surprising change in majority party control.
Data and Methods
Our analysis covers federal spending in the areas of agriculture, crime, defense, health,
and transportation policy from 1983 through 1996. As in several of our recent studies, the unit of
analysis here is the U.S. county. Heitshusen (1991), Anton et al. (1980), Crump and Archer
(1993), and Carsey and Rundquist (2002) have demonstrated the utility of county-level analysis,
with Anton et al. (1980) noting that, at the county level, “Federal program outlays are closely
associated with “need” in programs designed to address those needs (p. 78).” Counties are useful
because they are components of both most congressional districts and all states, and because they
allow for more narrow specifications of local constituency characteristics than do either
congressional districts or states. County borders also remain stable, meaning that all of the
problems introduced in district-level analyses by the decennial redistricting can be avoided.
Our analysis examines per capita spending in five policy areas: agriculture, crime,
defense procurement, health, and transportation. Data on programmatic expenditures are from
the Census Bureau’s Consolidated Federal Funds Report (CFFR). Data on the various measures
of constituency characteristics included as substantive control variables come from a variety of
sources. Those for agriculture, defense, and transportation are taken from the Regional
Economic Information Systems (REIS) data. Data for crime were taken from the U.S.
Department of Justice Federal Bureau of Investigation’s “Uniform Crime Reporting Program
Data, County Level.” Data on health care came from the “Area Resource File: 1940-94"
produced by the U.S. Department of Health and Human Services’ Health Resources and Services
11
Administration, Bureau of Health Professions. Member ideology is measured using annual
Conservative Coalition scores. The specific constituency characteristic measures are reported in
the note to Table 1. Representation on a relevant committee is measured as a simple dummy
variable for most counties, coded 1 if the county is represented and 0 otherwise. However, there
are several large urban counties that encompass multiple congressional districts. In these
instances, the committee representation variables were computed as proportions. So, if a county
included 10 House districts, and five of those representatives served on a relevant committee, the
committee representation variable was coded as .5. The specific committees in question are
listed in the note to Table 1. State population is measured in millions of persons. Each
dependent variable is measured in dollars per capita, adjusted for inflation.
In our analysis, we employ a statistical model that represents a modification of a portion
of a model of distributive politics that we have employed in several studies elsewhere (see
Rundquist and Carsey 2002 for the most complete treatment). Specifically, we model the level
of funding received in a particular county per capita as a function of: previous levels of spending,
whether the county was represented on a House and/or Senate committee with substantive
jurisdiction over the policy area in question, the partisan make-up of the House and Senate
delegations that represent the county in Congress, and the ideological orientations of those same
delegations. This model further separates committee representation into representation by either
fo the two parties. We also control for constituency characteristics related to the policy problem
under consideration (e.g., Adler and Lapinski 1997). Finally, we include a measure of state
population to capture the potential small-state advantage of equal Senate representation.
Because the model includes a lagged value of the dependent variable, the other variables in the
3 These models are identical to those presented in Table 1 in Carsey and Rundquist(2002). The results differ slightly due to a previous failure to include the full set of year dummyvariables.
12
model are really predicting change in the level of spending received in a county from one year to
the next.
Findings
To begin, we provide the results for our basic model for each of the five policy areas
pooling the data over the entire 1983 to 1996 period. Without interpreting every coefficient, the
results presented inTtable 13 show some support for House committee representation effects on
the distribution of per capita benefits in Agriculture, Defense, and Health for counties
represented by House committee Democrats and for Agriculture, Crime, and Defense for
counties represented by House committee Republicans. Looking at Agriculture, for example,
being represented on the House Agriculture committee by either a Democrat or a Republican
translates into an increase in county-level per capita agricultural expenditures of between $59.50
and $61.40.
Looking at Senate committee representation, we see roughly the same pattern of partial
support for an effect in some policy areas, but not others. Again, the effect is strong and
statistically significant for committee representation by either a Democrat or a Republican on the
Senate Agriculture committee, though here we see the effect is more than three times larger for
Democrats than it is for Republicans. Oddly, we find statistically significant negative effects for
Senate committee representation by a Republican for health care expenditures.
Turning to the effect of state population, Table 1 reveals that counties located within
4 This analysis is similar to that presented in Chapter 8 of Rundquist and Carsey (2002).
13
relatively smaller states received significantly larger relative changes in per capita expenditures
in four out of five of our policy areas compared to counties located in more populous states. The
one policy area where this does not hold is crime spending. For example, the coefficient in Table
1 operating on state population in the agriculture spending model can be interpreted as indicating
that an increase in state population of one million leading to a decrease in per capita agriculture
spending in a county of $2.79 per capita. These negative and significant coefficients support
Lee’s (1998, 2000) and Lee and Oppenheimer’s (1999) claim of a small state advantage in the
distribution of federal expenditures.
In sum, the results in Table 1 illustrate the average relationships between spending in
these five policy areas and each of the independent variables over the entire 1983-1996 period.
As such, they provide a backdrop to our subsequent exploration of how these results change over
time in response to, a) increases or decreases in the overall level of spending in each policy area,
and, b) the GOP takeover after 1994.
Because this analysis is exploratory, we wanted to proceed with the most flexible method
of estimation possible. Thus, we ran a series of 2-year rolling average models for each policy
area in order to see whether there are any changes in the magnitudes or signs of the coefficients
operating on each independent variable.4 So, for example, we re-ran the model for agriculture
spending presented in the first column of Table 1 thirteen times; once for 1983-84, then for 1984-
85, again for 1985-86 and so forth up through 1995-96. The 1983-84 regressions include roughly
3,000 cases each – one observation for each county – while the subsequent models include
5 The 1983-84 model contains fewer observations because of the use of lagged values forthe independent variables in the model and the fact that we do not have data for 1982.
14
roughly 6,000 cases.5 These sample sizes are sufficiently large to produce efficient estimates of
the parameters in the model for each two-year period.
We conducted this analysis for all five policy areas, resulting in a total of sixty-five
regression models with a total of more than 850 coefficients. This obviously presents too much
information to report in tabular form (imagine thirteen times the information presented in Table
1!). We also know that random chance alone will produce dozens of statistically significant
coefficients from these analyses. So, rather than focus on specific coefficient estimates or tests
of statistical significance, we chose instead to plot the values of each coefficient estimate as it
changes over time. We then compare these plots to plots of total level of programmatic
spending over time to look for relationships between increasing or decreasing levels of spending
and the factors that influence the geographic distribution of policy benefits. These graphs also
allow us to readily examine the impact of the GOP takeover after the 1994 elections.
We chose this method examining the impact of increasing or decreasing budgets over
time over other methods because of its flexibility. Our approach imposes very little in the way of
constraints on the analysis. In contrast, we could have chosen to include in our baseline models a
direct measure of the level of spending in each policy area and then a complete set of
multiplicative interaction terms between total spending and the other variables in the model.
This approach, while attractive for its simplicity and the substantially fewer number of parameter
estimates it would produce, is also more restrictive. Specifically, it imposes a linear structure on
whether or not, say, the effect of committee representation on changes in the distribution of per
6 All spending measures are adjusted for inflation. Agriculture, Defense, Health, andTransportation spending is reported in billions of dollars. Crime spending is reported in millionsof dollars.
15
capita benefits changes when the level of spending changes. It also imposes the assumption that
increases and decreases in spending have the opposite impact on how factors like the small state
advantage operates, but that the magnitude (absolute value) of these effects is the same. Without
any exiting theory or empirical evidence in the literature to guide us, we felt that the least
restrictive approach was the most reasonable for an exploratory analysis.
Figures 1 through 5 present several graphs based on our rolling average analysis. The
graph in the upper left-hand side of each figure plots the total annual spending for each policy
area.6 This provides the basis for comparison to the other graphs in each figure as they show
whether the overall level of programmatic spending is increasing or decreasing over time. So,
for example, Figure 1 shows that agricultural spending stood at about $45.6 billion in 1983, that
it dropped dramatically in 1984 and 1985 to below $25 billion, then increased sharply again in
1986 to about $34 billion. Agricultural spending crept up again in 1987 before declining steadily
to 1990. From 1990 to 1996 we see a gradual, though erratic, pattern of increases in agricultural
spending.
The remaining graphs included in figures 1 through 5 plot the magnitude of the
coefficient estimates from our distributive politics models as estimated from our rolling average
analyses. These results are plotted against time using the most recent year as a reference point.
Thus, the analysis for 1983-84 plots the coefficient estimates on the Y-axis against 1984 on the
X-axis. The 1984-85 analysis plots the coefficient estimates on the Y-axis against 1985 on the
X-axis, and so forth. To facilitate comparison, we plotted standardized coefficient estimates.
16
Again, to illustrate, the second graph in figure 1 plots the standardized coefficient estimates
operating on the lagged value of agriculture spending produced from our rolling average analysis
of agriculture spending. That standardized coefficient was estimated as .93 for 1983-84 (labeled
on the figure as 84). It declined in magnitude in 1984-85 and again in 1985-86. It then returned
to values above .90 until dropping dramatically in 1995 and again in 1996.
These plots allow us to explore the effect of changing levels of spending, as well as the
impact of the GOP takeover after 1994, on the nature of distributive politics. We address the
question of changing levels by comparing the graph for overall spending with the graphs plotting
the standardized regression coefficients. We examine the impact of the post-1994 takeover by
looking at the plots for the coefficient estimates before and after 1994.
Rather than examine each specific graph contained in figures 1 through 5, we devote
some attention to general patterns we observe across the graphs. The first conclusion we reach is
that few clear patterns linking the level of spending to the impact that factors like committee
representation, delegation partisanship, or delegation ideology to changes in the amount of per
capita benefits received by a county are apparent. What we are looking for are trends over time
in the graphs that plot the standardized coefficients that either parallel or mirror the trends in
overall programmatic spending that are shown.
For example, looking at Figure 1, we see that the precipitous decline in total agriculture
spending that occurred in 1984 and 85 does seem to be mirrored by a steep increase in the effect
of all four committee representation variables. This suggests that MCs serving on agricultural
committees were better able to protect the level of benefits residents of counties they represented
received during the rapid decline in federal agricultural spending than were MCs not on
17
agricultural committees. Thus, committee representation appears to have can served in a reactive
capacity when overall agricultural spending declined. Looking at later years for agricultural
spending, the patterns are less clear. Subsequent increases and decreases in spending do not
appear to produce a systematic effect on all of the committee representation variables. One
finding of note is the increased importance of representation on the Senate agriculture committee
by a Republican following the 1994 election. This coincides with an increase in agricultural
spending, suggesting that Senate committee Republicans were able to proactively exploit
committee representation to capture new agricultural benefits for their constituents.
Looking at Figures 2 through 5, trends in overall policy spending are harder to link to
trends in the impact of committee representation. Figure 2 shows steady increases in crime
spending took place from 1983 through 1996. However, the influence of committee
representation on the distribution of these funds was erratic. Figure 3 shows a steady decline in
defense procurement from 1985 forward. That is roughly paralleled by a steadily increasing
effect of House defense committee representation by a Democrat. This suggests a reactive
element to defense committee representation. However, Figure 3 also reveals a steep jump in the
impact of Senate defense committee representation by a Republican after 1994. Thus, while
Republican Senate committee members seemed to capture gains from an increasingly large
agricultural spending “pie,” their Republican counterparts on Senate defense committees were
successful in getting benefits even as the overall level of defense procurement spending
continued to decline.
Figure 4 shows the results for health policy. Unlike defense, there is a steady increase in
overall health policy spending for most of this time period. This growth in total spending is
18
accompanied by a steady increase in the impact of House committee representation by a
Democrat. Thus, whereas House committee Democrats appeared to dig in their heals to protect
their share of a shrinking defense procurement budget, House Democrats on health policy
committees appear to have proactively pursued larger shares of a growing health care budget.
The behavior of the other committee representation variables in the area of health policy are
more erratic.
Finally, Figure 5 shows an up and down pattern of spending for transportation that is
similar in this respect to agriculture spending. In the House, committee representation by
Democrats appears to be most influential during periods of relatively high transportation
spending. Thus, committee Democrats capture a disproportionate share of benefits for the
counties they represent when the overall level of expenditures is higher. The other transportatoin
committee representation variables reveal no readily discernable pattern.
Thus far, we have fairly mixed evidence of the impact of changing levels of spending in a
policy area on the nature of the distributive politics that shapes the geographic distribution of
benefits. In agricultural, defense, and health policy, we see evidence of both reactive benefit
protection and proactive benefit targeting, particularly by House committee Democrats. We also
see that Senate committee Republicans were able to capture benefits for counties they
represented after the 1994 takeover in both agriculture and defense procurement policy – two
traditionally “distributive” policy areas. However, in many instances we see no clear pattens
regarding committee representation, leading us to be cautious about our interpretations thus far.
We see even less in the way of consistent patterns regarding the direct effects of the
partisan and ideological make-up of a county’s House or Senate delegation. However, we don’t
19
find this to be too surprising. Our previous work on defense spending led to the conclusion that
the effects of party in any one policy area might be constrained to only those majority party
members serving on committees with jurisdiction over that policy. When aggregated to overall
domestic spending, this would translate into a direct effect of partisanship (e.g. Levitt and Synder
1995), but it is not surprising to find an erratic pattern within specific policy areas. Less attention
has been given to ideology in the distributive politics literature, but we suspect a similarly erratic
process at work.
Turning to the effect of state population on the distribution of per capita expenditures
over time, Figures 1 through 5 present some interesting results. The most obvious pattern is the
sharp movement to a strongly negative coefficient in 1995 in all but transportation spending. In
fact, for each of these four other policy areas, models that account for this substantial one-time
dip in the magnitude of the coefficient, the small state advantage effect for the remainder of the
period drops to statistical insignificance. It is only in transportation spending that we find a
consistently negative coefficient estimate operating on state population, thereby indicating a
small-state advantage, though even here the magnitude of the effect varies. However, the
overwhelming pattern is one in which the bulk of the small-state advantage reported in Table 1
stems from a one-time advantage following the GOP takeover.
The meaning of this finding is unclear at this point, though we offer the following
speculation. We suspect that GOP takeover represents a substantial disruption in what had been
the normal course of policy making in Congress for much fo the previous forty years when
Democrats control the House and typically the Senate. Added to this is the element of surprise
that the GOP capture of the House in 1994 represents. Scholars, pundits, and politicians alike
20
expected GOP gains in the 1994 midterm elections, but few seriously believed the GOP would
become the majority. We suspect that as Congress adjusted to the transfer of power, institutional
structures that remained unaffected – like the small-state advantage in representation due to the
Senate – more prominently structured policy making. Party leadership changed, committee
leadership and committee composition changed, but small states were still small states. In each
of the four affected policy areas, the impact of state population returned quickly in 1996 to what
it had been prior to the GOP takeover, suggesting that the disruption caused by the surprise GOP
victory was short-lived.
What we do not see in Figures 1 through 5 is evidence that the small state advantage is
more likely to appear in times of budgetary increases or decreases. The patterns in each graph of
the standardized coefficients’ operation on state population seem completely unrelated to any
pattern in the graphs of total programmatic spending.
The findings related to the small-state advantage reveal at least one important effect of
the GOP takeover on the process of distributive politics and it runs counter to the conclusion
reached by Bickers and Stein (2000). But there is another pattern that emerges in Figures 1
through 5 corresponding to the GOP takeover. In every policy area except transportation, the
GOP takeover seems correlated with a substantial dip in the effect of the previous year’s level of
per capita benefits on the current year’s level of benefits. Specifically, the level of benefits
received by a county in 1995, after the GOP takeover, is less strongly correlated with the pre-
existing level of benefits received in 1994, before the takeover, than it has been in any previous
period included in the study. The effect continues to decline in agriculture spending in 1996,
though it reverts back to previous levels in crime, defense, and health policy. What this means is
21
that, for the first year after the GOP takeover, the old adage of funds being distributed this year in
pretty much the same way that the were distributed last year does not hold particularly well. This
does not seem to be directly related to committee representation or partisanship, but rather
suggests a blanket shift in the geographic distribution of funds after the GOP took over. This
finding is consistent with the assertion by Bickers and Stein (2000) that a fundamental shift in the
types of policy expenditures, and thus the types of constituents who would benefit, occurred after
the 1994 election. Our results are compatible with the notion that a wholesale shift took place in
the kinds of constituencies that benefitted from the distribution of federal spending after the GOP
took over as compared to those who benefitted before, and that this shift was not targeted only to
those places with newfound majority party representation on the relevant congressional
committees. That the effect of the lagged value of spending snaps back up to previously high
levels in crime, defense, and health policy in 1996 suggests that whatever across-the-board shift
that took place after 1994 was quickly institutionalized.
Conclusions and Discussion
These exploratory findings point scholars in a new direction for thinking about
distributive politics. First, we see evidence that the previously found effect of majority party
representation on the relevant policy committees holds across several policy areas, though not all
of them. In some, the effects are weaker. In others, the effects of committee representation exist
regardless of party. However, in each policy area we find evidence of a distributive politics of
policy making at work. Similarly, in four of five policy areas, we find evidence of the small state
advantage that is assumed to stem from the nature of representation in the Senate.
22
Our evidence on whether the nature of distributive politics differs in times of budgetary
growth versus budgetary decline is mixed. In some policy areas, there is just no pattern at all. In
others, we find some evidence of both a proactive and a reactive effect of committee
representation across both the House and the Senate, though the effects are far from uniform and
consistent across policy areas. Our previous work examining the reciprocal relationship between
committee representation and the distribution of defense benefits found evidence for both the
recruitment and benefit hypothesese (e.g. Carsey and Rundquist 1999b). Thus, we are not
surprised in this setting to find evidence of both a proactive and a reactive effects of committee
representation.
The small state advantage does not appear to be sensitive to changing levels of
programmatic spending. It does, however, seem to respond to a change in overall partisan
control of Congress. We speculate that the small state advantage emerged most clearly
immediately after the GOP takeover because the small state effects of Senate apportionment is
the only institutional structure that provided the same incentives to all MCs both before and after
the 1994 elections. The lack of an effect of state population on the distribution of policy benefits
during other time periods, however, brings into question our understanding of the prominence of
the small-state advantage on a consistent basis. Only in transportation policy does there appear
to be a fairly consistent small-state advantage.
Finally, the GOP takeover appears to have resulted in a short-term fundamental disruption
in pattern of benefit distribution in four of five policy areas. The effect does not appear to be
driven my any particular institutional structure, perhaps with the exception of the small-state
advantage. Rather, the effect of the GOP takeover appears to be one of broad-scale change in
23
which places receive funds.
Further research requires a number of things. First, we must satisfy ourselves that we
have sufficiently controlled for constituency characteristics. Second, we must extend the data set
in order to assert with more confidence the short- and longer-term impact of the GOP takeover in
1994. Third, we might consider different measures of the level of policy spending. Perhaps the
total budget doesn’t matter as much as the total amount received in the nearby area (say, the
state). If the state continues to receive funds at a steady pace, the overall national growth or
decline in a policy area’s budget may not create the proactive or reactive pressures on MCs that
we considered here. It may also be that increases or decreases in explicitly discretionary funds
might be more important, or that increases or decreases in funding within a policy area should be
measured relative to the entire federal budget. Finally, policy specific studies like ours always
run the risk of missing the potential interaction that may take place across policy areas. The
sensitivity of MCs to spending decreases in one policy area may be a function of whether or not
they can readily substitute those benefits with expenditures from other policy areas.
24
References
Adler, Scott E. and John Lapinski. 1997. “Demand-side Theory and Congressional CommitteeComposition: A Constituency Characteristics Approach.” American Journal of Political Science. 41:895-918.
Aldrich, John H. 1995. Why Parties? The Origin and Transformation of Party Politics inAmerica. Chicago: University of Chicago Press.
Anton, Thomas J., Jerry P. Hawley, and Kevin L. Kramer. 1980. Moving Money: An EmpiricalAnalysis of Federal Expenditure Patterns. Cambridge, MA: Oelgeschlager, Gunn, and Hain.
Arnold, R. Douglas. 1979. Congress and the Bureaucracy. New Haven: Yale University Press.
Bickers, Kenneth N. and Robert M. Stein. 2000. “The Congressional Pork Barrel in aRepublican Era.” Journal of Politics. 62(4): 1070-86.
Carsey, Thomas M. and Barry S. Rundquist. 1999a. “Party and Committee in CongressionalPolicy Making: Evidence from the Domestic Distribution of Defense Expenditures.” Journal ofPolitics. 61: 1156-69.
Carsey, Thomas M. and Barry S. Rundquist. 1999b. “The Reciprocal Relationship between StateDefense Interest and Committee Representation in Congress.” Public Choice. 99:455-63.
Carsey, Thomas M. and Barry S. Rundquist. 2002. “Inter- and Intra-Chamber Differences in theDistribution of Policy Benefits.” Paper presented a the Annual Meeting of the Midwest PoliticalScience Association, Chicago, IL.
Cox, Gary W. and Mathew D. McCubbins. 1993. Legislative Leviathan. Berkeley: Universityof California Press.
Crump, Jeffrey and Clark Archer. 1993. "Spatial and Temporal Variability in the Geography ofAmerican Defense Outlays". Political Geography. Vol. 12, No.1, pp 38-63.
Dye, Thomas R. 1984. “Party and Policy in the States.” Journal of Politics, 46: 1097-1116.
Erikson, Robert S., Gerald C. Wright, Jr., John P. McIver. 1989. “Political Parties, PublicOpinion, and State Policy in the United States.” American Political Science Review, 83: 729-750.
Ferejohn, John A. 1974. Pork Barrel Politics: Rivers and Harbors Legislation, 1947-1968. Stanford: Stanford University Press.
25
Heitshusen, Valerie. 1991. “Do Committee Members Get a Bigger Piece of the Pie? TheDistribution of Program Expenditures to Members of Congress in the 1980s.” A paper presentedat the Annual Meeting of the American Political Science Association, Chicago, IL.
Lee, Frances E. 1998. “Representation and Public Policy: The Consequences of SenateApportionment for the Geographic Distribution of Federal Funds.” Journal of Politics. 60:34-62.
Lee, Frances E. 2000. “Senate Representation and Coalition Building in Distributive Politics.” American Political Science Review. 94: 59-72.
Lee, Frances E. and Bruce I. Oppenheimer. 1999. Sizing Up the Senate: The UnequalConsequences of Equal Representation. Chicago: The University of Chicago Press.
Levitt, Steven D. and James M. Snyder. 1995. “Political Parties and Distribution of FederalOutlays.” American Journal of Political Science. 39:958-80.
Lowi, Theodore. 1964. “American Business, Public Policy, Case Studies, and Political Theory.” World Politics. 16: 677-715.
Lowry, Robert C. and Matthew Potoski. 2002 “Interest Groups, Citizens and Politicians:Explaining State Success in Distributive Policy.” Paper presented at the Second Annual StatePolitics and Policy Conference, Milwaukee, WI. May 24-25.
Pace, David. 2002. “GOP Shifted Billions to Districts.” Associated Press, August 8th, 2002. Available at http://wire.ap.org
Rohde, David and Kenneth Shepsle. 1973. “Democratic Committee Assignments in the Houseof Representatives: Strategic Aspects of a Social Choice Process.” American Political ScienceReview. 67: 889-905.
Rundquist, Barry S. and Thomas M. Carsey. 2000. “Distributive Politics and TargetingBenefits: Is it Really Worth Fighting For?” Paper presented at the Annual Midwest PoliticalScience Association Meeting, Chicago, IL.
Rundquist, Barry S. and Thomas M. Carsey. 2002. Congress and Defense Spending: TheDistributive Politics of Military Procurement. Norman: University of Oklahoma Press.
Rundquist, Barry and John Ferejohn. 1975. “Observations on a Distributive Theory of PolicyMaking: Two American Expenditure Programs Compared.” In Craig Liske, William Loehr, andJohn McCamant, eds., Comparative Public Policy: Issues, Theories, and Methods, 87-108. NewYork: Wiley.
26
Shepsle, Kenneth A. and Barry R. Weingast. 1995. “Positive Theories of CongressionalInstitutions.” in Kenneth Shepsle and Barry Weingast, eds., Positive Theories of CongressionalInstitutions, 5-36. Ann Arbor: University of Michigan Press.
Stein, Robert and Kenneth Bickers. 1995. Perpetuating the Pork Barrel: Policy Subsystems andAmerican Democracy. New York: Cambridge University Press.
Weingast, Barry R. 1979. “A Rational Choice Perspective on Congressional Norms.” AmericanJournal of Political Science. 23:245-62.
Weingast, Barry R. Kenneth A. Shepsle, and Christopher Johnson. 1981. “The PoliticalEconomy of Benefits and Costs: A Neo-classical Approach to Distributive Politics.” Journal ofPolitical Economy. 96: 132-63.
Winters, Richard. 1976. “Party Control and Policy Change.” American Journal of PoliticalScience. 20: 597-636.
27
Table 1: Baseline models of factors that influence county-level per capita expenditures in five policy areas
Note: Cell entries are unstandardized regression coefficients, two-tailed significance levels in parentheses. Models alsoinclude year dummy variables. The relevant committees and constituency characteristic variables are, in order: a House Agriculture Committee, Senate Agriculture Committee, per capita earning from agriculture. b House Judiciary Committee, Senate Judicial Committee, per capita offenses, per capita police employment. c House Armed Services Committee, Senate Defense Committee, economic capacity in Gun Belt states, per capita income. d House Commerce Committee, House Ways and Means Committee, Senate Labor Committee, Senate FinanceCommittee, doctors per capita, hospital beds per capita. e House Public Works Committee, Senate Banking Committee, Senate Public Works Committee, per capita income fromhighway construction.
