Explaining the Internationalist Orientation in American Foreign Economic Policy: Theories of Legislative Politics in Trade and Aid Policy Helen V. Milner B. C. Forbes Professor of Politics and International Affairs, Princeton University [email protected]Dustin H. Tingley 1 PhD Student, Politics Department, Princeton University [email protected]Robertson Hall, Princeton University, Princeton NJ, 08544 Abstract Since 1945, American foreign economic policy has been oriented toward engagement with the international system. Given the changes in world politics and economics as well as American domestic politics over the past twenty years, scholars have wondered whether American foreign economic policy might change. What groups have supported this internationalist policy since the late 1970s, and have they changed over this period? We examine legislative voting in the US House of Representatives from the 96 th to the 108 th Congress (1979-2004) on trade and aid. Beyond description, we use theory to develop testable hypotheses about sources of support for and opposition to aid and trade. We show that a bipartisan group has persisted supporting trade and aid; Stolper-Samuelson models best explain this. But supporters in the two areas differ in important ways. Most interestingly, labor and liberal Democrats remain part of the group supporting aid, while they no longer do on trade. 1 Corresponding author. The authors would like to thank Raymond Hicks for assistance.
41
Embed
Explaining the Internationalist Orientation in …pcglobal/conferences/domestic08/papers/Trading... · Explaining the Internationalist Orientation in American Foreign Economic Policy:
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Explaining the Internationalist Orientation in American Foreign Economic Policy: Theories of Legislative Politics in Trade and Aid Policy
Helen V. Milner
B. C. Forbes Professor of Politics and International Affairs, Princeton University [email protected]
Dustin H. Tingley1 PhD Student, Politics Department, Princeton University
Abstract Since 1945, American foreign economic policy has been oriented toward engagement with the international system. Given the changes in world politics and economics as well as American domestic politics over the past twenty years, scholars have wondered whether American foreign economic policy might change. What groups have supported this internationalist policy since the late 1970s, and have they changed over this period? We examine legislative voting in the US House of Representatives from the 96th to the 108th Congress (1979-2004) on trade and aid. Beyond description, we use theory to develop testable hypotheses about sources of support for and opposition to aid and trade. We show that a bipartisan group has persisted supporting trade and aid; Stolper-Samuelson models best explain this. But supporters in the two areas differ in important ways. Most interestingly, labor and liberal Democrats remain part of the group supporting aid, while they no longer do on trade.
1 Corresponding author. The authors would like to thank Raymond Hicks for assistance.
Great powers in world politics make different decisions about what strategies to pursue,
and these decisions have critical influences on how the international system operates. Since 1945,
the central element of American foreign economic policy has been its active engagement with the
international system. This internationalist orientation meant that American trade policy favored
lowering barriers to trade in goods and services, and that the US pursued a relatively generous
program of giving foreign assistance to other countries. Prior to 1945, US policy was much more
isolationist, relying upon protection of the domestic market and eschewing foreign aid. This post-
World War II internationalist policy, it is claimed, was built on two pillars: a bipartisan coalition
in Congress and presidential dominance of foreign policy (e.g., (Johnson, 2006)). Given the
changes in world politics and economics as well as in American domestic politics over the past
twenty-five years, scholars have wondered whether the direction of American foreign policy
might change dramatically (Ikenberry, 2005). Would this internationalist policy be replaced by a
more isolationist one?2 This question reflects concerns about the durability of the bipartisan
group that has supported this internationalist policy, as well as presidential dominance of it. This
study is one of the first to inquire systematically into the nature of the groups that have supported
and opposed foreign aid and trade liberalization in the US over the past twenty five years.
Further, we provide a theoretical explanation for these patterns and empirical tests of our
hile
y
rs—
rnationalism? Finally, is presidential dominance of
the poli
have
propositions.
Our central questions are four. First, what groups have supported an internationalist
policy in aid and trade since the late 1970s? Is bipartisanship still evident? Second, are the groups
that support international engagement via trade the same as those that support foreign aid? W
it is often assumed that groups supporting internationalism are similar across issues, are the
supporters of aid and trade the same? Third, have the groups supporting aid and trade polic
changed over this quarter of a century? American foreign economic policy after 1945 was
designed for a world driven by the Cold War and by a serious North-South divide. Both of these
organizing features of world politics have disappeared, and new political and economic facto
such as globalization- have come to shape American politics and the world economy. What
consequences have these changes had for inte
cy agenda in trade and aid apparent?
To address these questions, we examine legislative voting in the US House of
Representatives from the 96th to the 108th Congress (1979-2004) on trade and aid issues. We show
that for each area a bipartisan group has been necessary for continuing engagement, but that the
groups in aid differs from those in trade in substantial ways. Further, the supporters in trade
2 A second dimension of US foreign policy concerns unilateralism versus multilateralism.
3
’ districts
and thei
s of
wed with
sts
ic
but
nners
e
one of t
e trade
e
more for
aid than trade, while the president has more influence in trade than aid policy. Bringing
evolved over time, whereas those in aid have been more stable. Our theory claims that the
distinct support groups in aid and trade depend upon the characteristics of legislators
r ideologies, rather than their party affiliations or presidential preferences.
In general, the groups supporting aid and trade differ, but both policies require bipartisan
support. Democrats are more supportive of foreign aid than are Republicans, while Republicans
tend to be more supportive of trade liberalization. But fractures within the parties are large and
salient; party identification only explains a small part of legislators’ votes. Many discussion
American foreign policy focus on parties and presidents, but we think this misses a central
element of the politics surrounding foreign policy (Destler, 1995; Lancaster, 2007). A central
core of support for international engagement in trade and aid cuts across the parties and lies in the
regions that gains economically from trade and aid, i.e., those regions that are well endo
human and physical capital. Groups possessing such endowments are the winners from
internationalism and its biggest supporters. On the other hand, groups whose economic intere
seem to imply they should oppose internationalism are divided. Organized labor is a staunch
opponent of trade liberalization, as expected, but is supportive of foreign aid giving. Labor’s
division on the twin policies of internationalism is reflected in the division of the Democrat
Party. Cleavages across class lines thus dominate trade politics; capital and labor and their
respective supporters line up as opponents in this area. Political ideology also plays a role,
mainly in the aid area where legislators representing conservative voters oppose aid. The
cleavages around aid thus cut across class and party lines; supporters include economic wi
from aid plus its ideological proponents among liberal voters and labor. The power of the
presidency, in contrast, is apparent only in the trade arena. In particular, Democratic presidents
have used their influence to push Democratic legislators to support trade policies they endorse.
No party in the US today wholeheartedly supports international engagement; hence, presidents
must craft bipartisan legislative support, which limits their dominance of the agenda. Further, aid
and trade seem to represent substitutes in foreign policy terms, as majorities in the two parties se
hem, but not the other, as advancing US national interests (Palmer and Morgan, 2006).
Our project is one of the first to study these great power preferences and to compar
and aid policy. Beyond describing these groups, we theorize about the sources of different
groups’ preferences for aid and trade openness, showing that the economic characteristics of
domestic groups matter, that organized interest group pressures are salient, and that ideology is
sometimes important. We use standard theories from the political economy literature to compar
the power of alternative explanations for these groups. We find strong support for (Heckscher-
Ohlin based) Stolper-Samuelson theories of trade and aid. We find that ideology matters
4
systematic theory and empirical tests to our understanding of American foreign economic policy
is an important step in advancing research in this area.
II. Theories of American Foreign Economic Policy
Most theories that address questions about the nature of foreign policy focus on one of
three sets of factors. Some tend to look at a country’s place in the international system and
predict policy given that relative global position (see for example (Elman, 1995; Haas, 2007)).
This approach is less helpful since it seems to suggest that unanimous support for policy should
be evident at home since all actors in the country share the country’s same global position. In
terms of American foreign economic policy, no such unanimity in the executive branch,
legislature or public has been apparent since 1945. Others focus most attention on the executive
branch and on the preferences and beliefs of the executive (whether prime minister or president
and cabinet)(Canes-Wrone et al., 2006; Howell and Pevehouse, 2007). For this group, presidents
have much more influence over foreign policy than they do domestic policy, and more influence
than the public or Congress. Others pay more attention to domestic economic and social factors;
for instance, there exists a large literature in trade policy that emphasizes the economic
characteristics of political actors’ constituencies (e.g., (Hiscox, 1999; Ladewig, 2006)). Our
approach uses the last two set of theories to explore the nature of the groups supporting and
opposing aid and trade policy in the US.
Unlike most other studies of foreign policy, we focus our attention on the legislature and
in particular the House. We think that legislators in the House closely reflect the interests, ideas,
and concerns of their constituents. They have the shortest (re)election periods, the smallest
constituencies, and higher turnover rates (Collier and Munger, 1994). Further, some claim that
the House after the mid 1970s took over the dominant position on most international policy issues
from the Senate (Johnson, 2006). Hence they may well provide the best mapping of the groups
that support and oppose trade and aid policy.
The President proposes most foreign policy initiatives, such as trade liberalization
agreements, and presents budgets to Congress for foreign aid giving. Majorities in Congress must
approve his proposals before they can be implemented. Why would legislators support an
internationalist policy of generous aid and trade liberalization? A first hypothesis that must be
entertained, our null hypothesis, is that legislators simply vote idiosyncratically on these issues.
Both issue areas may seem arcane and distant from the domestic concerns of their constituents;
they may also be seen as unimportant, having few, if any, consequences for their constituents. In
this view, no predictable group of legislators should appear in support of aid and trade. No set of
5
factors should be able to systematically explain legislators’ votes on trade and aid policy, since
they (and their constituents) either do not have preferences or do not know them in these areas.
To the extent that trade and aid are marginal to the US economy, this perspective gains
plausibility. How important are trade and aid to the US? American trade dependence has grown
much since the 1970s. By 2000, roughly 25% of the US economy was exposed to trade; that is,
exports and imports accounted for a quarter of GDP. Imports since 2000 have been well over $1
trillion per year, and exports from the US close to $1 trillion, combined they are equivalent to the
size of the entire US government budget (WTO, 2006, table II:4). Economic aid has been a less
salient part of the US economy, but not insignificant.3 It has accounted for less than 1% of US
GDP since the Marshall Plan ended. Nevertheless, the US is the largest absolute donor in the
world since its GDP is so large. The amount of aid is also similar in many years to the amount of
direct government spending on agricultural supports, an area that has received considerable
scholarly attention (GPO, 2004; Hansen, 1991; OECD/DAC, 2007; Poole and Daniels, 1985).
The fact that much foreign aid is tied to domestic suppliers suggests that legislators and domestic
groups see aid as valuable; an estimated 72% of US aid in 2006 remains tied
(CenterForGlobalDevelopment, 2007). Since aid is a much smaller part of the economy, one
might expect that economic and social factors would explain it less well.
A milder version of the null hypothesis is that legislators simply follow the president’s
lead. Presidents propose foreign policy and legislators vote in favor of it if they come from the
president’s party and against it if they are from the opposition party (Howell and Pevehouse,
2007). In this version, legislators (and their constituents) take cues from the President and
following party loyalty vote in accord with the president. Legislators thus have no preferences
regarding trade and aid and/or no incentives to find them out. This argument suggests no
predictable group of legislators supporting aid and trade policy should exist; rather, support will
change as the president’s party and his interests change. We also test this presidential dominance
hypothesis.
In addition to the null hypothesis and the presidential dominance ones, we explore two
other sets of theories that claim to explain legislators’ votes. Below we examine the specific
hypotheses that are derived from such political economy models and ideological arguments.
III. Theory and Hypotheses about Economic Interests.
3 Note that non-military aid is no longer primarily directed toward Israel and Egypt. In 1998, a deal was struck to end all non-military aid to Israel and to halve all non-military aid to Egypt in ten years. In the FY08 budget proposed, Israel no longer gets any non-military aid and Egypt gets less than half of what it got in the 90s—less than $400 million.
6
Two sets of factors may systematically explain legislators’ votes on aid and trade: their
economic interests and their ideological predispositions. Here we explore how the economic
interests of legislators’ constituents may affect their voting. Legislators desire most of all to
remain in office. This office seeking motivation leads them to pay attention to their constituents.
Voters may not know much about policy nor pay much attention to issues, but they tend to vote
legislative incumbents out of office when bad outcomes arise. Such punishment of incumbents
and the fear of it motivate legislators to develop preferences about all sorts of policies that could
affect outcomes in their districts.
Avoiding such bad outcomes (and sometimes being rewarded for good ones) means that
legislators may vote according to the distributional consequences that policies are expected to
have for their constituents. These distributional consequences in turn depend on the economic
characteristics of their districts. Trade and aid, according to various theories, have distributional
consequences, and different districts because of their different economic compositions will
experience the costs and benefits of aid and trade flows differently. The two most prominent
theories of such distributional consequences are Stolper-Samuelson and Ricardo-Viner models.
These lead us to anticipate that the district level economic consequences that legislators expect to
follow from aid and trade policies shape their preferences about trade and aid policy. We focus on
two basic elements that affect their political support for foreign economic policy: 1) the economic
composition of their districts and 2) the activities of interest groups like labor unions and business
associations.
These competing theoretical models, the Stolper-Samuelson (SS) and Ricardo-Viner
(RV) theorems, make different predictions about who the winners and losers will be from
international economic integration (Rogowski, 1989). Many have shown that these theorems
accurately predict congressional voting patterns on US trade policy, but controversy has arisen
over which of these two models best predicts trade policy (e.g., (Bailey, 2001; Baldwin and
McGee, 2000; Beaulieu, 2002a; Fleck, 2002; Ladewig, 2006). We extend these well-known
results to foreign aid because international foreign aid patterns are closely related to other
international economic flows ((Alesina and Dollar, 2000; Brakman and Marrewijk, 1998; Husain,
1993); for a longer discussion of why aid may be similar to trade see (Milner and Tingley, 2007)).
In this paper we concentrate on the Stolper-Samuelson model since it has received substantial
support in the literature. We also conducted an extensive series of tests of RV, and generally
found much less support for this theory. We discuss some of this research below, but concentrate
on SS in order to focus the paper.
The Stolper-Samuelson theorem (SS), which is based on the Heckscher-Ohlin model of
international trade, assumes that factors can easily move across industries; it predicts that the
7
distributional consequences of international economic policies will vary by factors of production.
That is, those who own more capital and those who own relatively more labor will differ in their
preferences over these policies since they will differentially gain (and lose) from them. This
model sees the primary cleavage as one between capital and labor. In advanced industrial
countries, which are abundant in capital--especially human capital, the relatively scarce factor,
labor—especially unskilled labor—will lose from policies that open the economy to the world
and its poorer economies.
In addition, the Stolper-Samuelson framework predicts that encompassing interest groups
which represent capital or labor should be active players, lobbying and pressuring legislators to
adopt their favored position. These groups should be internally united on their preferences toward
trade; thus, it predicts that PAC support to legislators by capital and labor groups should be an
important component of legislators’ decisions (Hiscox, 2002). Labor groups should contribute to
legislators who vote against bills to liberalize trade, while capital groups should actively support
such bills. We are not arguing that PAC contributions are necessarily causal; instead we want to
see which groups support and oppose aid and trade. This model then generates a series of
predictions that are distinct from those of our null hypothesis.
In contrast, the Ricardo-Viner theorem (RV), or specific factors model, sees the major
groups in the economy as those organized around different industries, and not factors of
production. Since factors of production are not mobile but specific to an industry, groups like
capital and labor that work in the same industry face the same distributional consequences. If an
industry gains (loses) from foreign trade or aid, all factors of production (i.e., capital and labor) in
that industry should support (oppose) policies that promote trade or aid. The Ricardo-Viner model
focuses on the district level characteristics associated with trade flows: in particular, export
oriented versus import competing industries. In addition, the specific factors model predicts that
interest groups representing capital owners across the economy or labor generally should be
unable to form a unified position on trade; they will be divided in their policy preferences by their
sectoral characteristics (Hiscox, 2002, pgs. 38-40). The specific factors model then sets forth
predictions for these two sets of influences on legislators, ones which differ from the Stolper-
Samuelson model and our null hypothesis.
How do these models apply to foreign aid? Following models proposed by a variety of
economists, we point out the distributive consequences of aid for donor countries (e.g., (Bhagwati
et al., 1983, 1984; Brakman and Marrewijk, 1998; Jones, 1984; Kemp, 1995; Mayer and
Raimondos-Moller, 1999)). There are two effects associated with aid as an international transfer:
a direct income effect (the transfer of purchasing power usually financed by taxes) and a change
in the terms of trade. An individual’s preferences for foreign aid in the donor country depend on
8
their factor ownership and on these terms of trade effects. Because aid affects the international
terms of trade, it in turn changes the distribution of income among factor owners in the donor
country. Their preferences then depend on how the terms of trade effect interacts with their
endowments.
As Mayer and Raimondos-Moeller (2003) note, a necessary condition for foreign aid to
increase the welfare of a person in the donor is that the transfer raises their income. An increase
in a person’s income will occur if the recipient country’s propensity to consume exceeds the
donor’s for the good which uses relatively intensively the factor that the person owns relatively
more of than the average person. For example, a transfer will increase a person’s income if the
recipient country has a higher propensity to consume the capital-intensive good than the donor
does and the person’s capital ownership ratio exceeds that of the average person in the donor.
Since poor recipient countries have a higher marginal propensity to consume certain goods, such
as food and capital intensive imports, than rich donor countries, a transfer would raise the world
prices of these goods. Then individuals in the donor country whose factors of production are
intensively used in the production of these goods have incentives to favor foreign aid, as the
Stolper-Samuelson theorem would anticipate. Since exports from rich countries to poor tend to be
capital intensive, this means that owners of capital in the donor benefit from aid through the terms
of trade effects.
Thus the main conclusion of the distributional models of aid, using the Stolper-
Samuelson framework, is that owners of capital in donor countries tend to gain from aid and thus
are more likely to support giving aid. On the other hand, owners of relatively unskilled labor in
the donor are likely to lose from aid and thus should oppose it. Likewise interest groups
representing capital should support aid, while those representing unskilled labor should oppose it.
The conclusions of the Ricardo-Viner model are that export-oriented districts should favor aid
and trade, while import-competing ones should oppose them both. As with trade, interest groups
representing capital and labor should be too divided in their preferences to make significant PAC
contributions to one side or the other.
These two theorems generate different predictions about the underlying bases of support
for and opposition to trade and aid policies. If the Stolper-Samuelson models are correct, we
should see class cleavages over aid and trade. A number of studies have asked which of these
two theorems provides the better explanation for trade (e.g., (Beaulieu, 2002a; Ladewig, 2006)).
Consistent with these scholars, we find much less support in trade voting for specific factors
models compared to Stolper-Samuelson ones in the time period we consider, and no support in
foreign aid voting. Thus in our empirical analysis, we focus on Stolper-Samuelson effects, and
9
briefly discuss the relative performance of Ricardo-Viner models in the appendix and in a
separate paper (Tingley, 2008).
IV. Ideology and Its Influence.
A long debate has occurred over the relative role of ideology and interests in legislative
voting (e.g., (Kalt and Zupan, 1993)). We think that it is important to try to distinguish these two
factors, but that they both are likely to matter for legislators. Ideology is harder to define than
economic interests. Here we identify ideology with a set of beliefs about the proper role of
government in the economy and society, especially relating to government attempts to
redistribute income. As noted above, we look at the beliefs of legislators’ constituencies and
assume that legislators think that their constituents (or the majority of them) hold these beliefs
and thus act on their behalf.
In American politics, a liberal-conservative ideological spectrum is often used to describe
political beliefs (e.g., McCarty, Poole and Rosenthal, 2006). This traditional left to right
ideological scale may help explain views toward foreign aid and trade. The liberal-conservative
political spectrum often identifies liberals as holding strong beliefs in the importance of
government intervention in the economy, especially to deal with redistribution of wealth to the
poor (Bobbio, 1996; McCarty et al., 2006). The conservative position on this spectrum is
associated with beliefs about the importance of individual effort and the value of the market as
means of wealth generation and distribution; government intervention is often seen as inefficient
and ineffective, as well as morally undesirable. Given these beliefs, one would expect individuals
holding liberal values to favor aid and to be opposed to free trade; on the other hand, those
holding conservative values should favor free trade and oppose aid as a form of government
intervention to redistribute wealth globally. We would expect districts (or legislators) with more
liberal views to support aid and oppose free trade, while more conservative ones should support
trade and oppose aid. Ideology defined on this liberal-conservative scale would split
internationalist groups over the value of aid versus trade. A president valuing both might have to
construct an ideological coalition of opposites to garner support for both elements of
internationalism.
In sum, our hypotheses are:
1. Economic Interests: The greater the endowment of human (or physical) capital in a district, the
higher the probability that the legislator votes in favor of trade liberalization and foreign aid.
2. Economic Interests: The larger the contributions and activities of interest groups representing
capital (labor), the higher (lower) the probability that the legislator votes in favor of trade
liberalization and foreign aid.
10
3. Ideology: The more conservative the members of a district are, the greater the probability that
the legislator votes in favor of trade liberalization but the less likely the legislator votes in favor
of foreign aid.
4. Presidential Dominance: The probability that a legislators votes in favor of an aid or trade bill
rises when the president both supports that policy and is from the same party as the legislator.
V. Empirical Evaluation of the Hypotheses
Research Design
Our analysis focuses on legislative voting in the US House of Representatives from the 96th-108th
Congress. In trade and aid the president needs the assent of Congress to implement his policies. In
trade, the Constitution gives the Congress explicit control over most trade policy, largely since it
is a matter of taxes. For the president to lower trade barriers or sign agreements with other
countries requires congressional assent. Presidents thus must bargain with Congress for trade
negotiating authority, which allows the President to lower US barriers by certain amounts in
exchange for foreign concessions; the President must then bring the international agreement back
to Congress for majority votes in both houses with no amendments. In addition, legislators can
introduce trade legislation of their own, usually out of (sub) committees, since they have
constitutional authority over national trade policy. We examine all three types of bills below:
presidential authorization to negotiate in trade, final passage of trade agreements, and individual
bills to regulate trade policy.
In foreign aid, the president also needs congressional approval since this involves taxing
and spending. Congress must agree to his proposals to appropriate and then allocate funds for
foreign aid each year. Unlike in trade, aid spending authority is usually part of a much larger
foreign operations bill, which contains spending authority for all forms of international activity in
the US government. Committees amend the president’s proposals and we generally examine
these amendments, rather than the annual omnibus foreign operations bills. Hence, although the
exact processes differ, the president requires congressional majorities to implement his foreign
policies in both of these areas.
We seek to tap legislators’ general sentiments toward trade and foreign aid. Our
dependent variable is the probability that a legislators votes in favor of trade and foreign aid. To
identify the proper population of votes relating to aid and trade policy, we utilized the Voteworld
program, various publications by the Congressional Quarterly, and the Congressional Record.
With this population of votes (itself a subset of all House votes), we selected a sample of votes
that met certain a priori criteria to identify the votes that tapped legislator positions on aid and
trade.
11
We include aid votes that satisfied the following criteria. First, we identified the universe
of amendments related to foreign aid that received roll call votes in the House between 1979-
2004 (96th-108th Congress). Second, we selected a subset that had the clear legislative
consequence of increasing or decreasing foreign aid funding, such that they could be
unambiguously coded. This requires that we exclude procedural, conference report, and final
passage votes; and it means that most of our votes involve amendments to final passage bills.
While interested in separately analyzing the universe of final passage votes, our reading of the
Congressional Record is that these votes concern a wide range of foreign operations issues, many
of which are orthogonal to foreign aid. A separate paper considers these final passage votes
(Milner and Tingley, 2007). Third, we do not consider amendments that involved specific
countries or programs that primarily support specific countries (like the Economic Support Fund).
Such votes usually tap geopolitical and country specific issues not related to general sentiments
toward aid. Fourth, we excluded amendments that dealt with ‘hot-button’ issues, like abortion and
AIDS, because these votes also address a very different set of political issues. Fifth, we exclude
votes on two specific, and important, types of foreign aid: food aid (PL480: Food for Peace
program) and export promotion (Import Bank or the Overseas Private Investment Corporation).
These programs represent an important means of international engagement. They also tap very
particular and narrow interests that more general aid votes might not. For example, in separate
analysis of food aid votes, we see that (as expected) district agricultural production has a strong
effect on support food aid. We do not see this relationship with the more general votes we include
in our sample. Thus combining a range of different types of votes could bias inferences. For
export promotion programs we expect that if economic interests influence voting in our current
sample, then they would be especially salient for votes on these programs. We look forward to
giving all of these types of aid legislation a more detailed treatment than space allows here. The
resulting set of foreign aid votes taps salient, yet still general, preferences for and against aid.
We pursue a similar strategy with regard to trade votes. We include trade votes that 1)
had clear consequences for US trade policy (e.g., were not procedural votes or ‘sense of
Congress’ votes), 2) did not deal with individual products unless those products dealt with major
US industries (e.g., steel, automobiles, textiles, sugar), and 3) had been used by previous scholars
in roll call vote analysis and that further review indicated that the votes were sufficiently trade
oriented.4 Most of our trade votes are final passage votes, as preceding floor votes on a particular
trade bill—if they happened—tended to be procedural. A thorough description of each bill is
provided in our on-line appendix.
4 Some votes used by other scholars involve issues that, upon detailed inspection of the Congressional Record, were rather tangential to actual trade policy and thus we do not include them.
12
We chose our sample of votes from the 96th-108th Congress primarily because this period
covers a broad span of American political history that has witnessed many changes domestically
and internationally. This sample gives us variation on the party of the President, divided
government, and the state of the US economy. Our votes should indicate general preferences
about foreign aid or trade. Our votes are similar enough to get at this general preference, but not
so similar that they merely inflate our number of observations without adding new information.
Our dependent variables then are two: a legislator’s vote on aid bills and his/her vote on trade
bills. Votes were recoded so that a 1 equaled support for aid and trade liberalization, while a 0
indicated opposition. We estimate the probability that a legislator votes in favor of aid or trade
given a series of characteristics about his or her district.
Empirical Analysis
We begin our discussion of the empirical results with a series of graphs that illustrate
interesting differences between voting on aid and trade. This gives us a general understanding of
trade and aid voting in the US during the years in our sample prior to moving into a more detailed
multivariate analysis. Table 1 charts the average DW-Nominate scores of Republicans and
Democrats by their position on aid and trade. DW-Nominate scores are a commonly used
measure of legislator ideology, usually thought of as arrayed along the liberal-conservative
(Left/Right) dimension with a higher number being more conservative. The scores are derived
from extracting a common dimension from voting over all bills within a Congressional session.5
For each vote, we took the average DW-Nominate score of legislators in each party by whether
they supported or opposed foreign aid or trade liberalization. The horizontal axis arrays the votes
along Congressional sessions, while the vertical axis shows the calculated average DW-Nominate
score for each group, with 95% confidence intervals surrounding each calculated average.
Two interesting patterns emerge. First, aid supporters when arrayed along the ideological
dimension are an ideologically ‘non-connected’ group. Support for foreign aid comes from liberal
Republicans and liberal Democrats; opponents are conservative Republicans and conservative
Democrats. This pattern is quite stable throughout our sample of aid votes (1979-2004). Support
for trade policy, on the other hand, has shifted around in this ideological space over time. Prior to
the 1990’s support for free trade came from conservative Democrats and conservative
Republicans. While conservative Democrats tended to maintain their support for free trade, more
conservative Republicans started to oppose free trade—a change which could be titled the “Ross
Perot effect.” Increasingly from 1992 onward, the ideologically connected group of liberal
Republicans and conservative Democrats formed the core of support for freer trade. Thus,
5 Unfamiliar readers may wish to consult http://voteview.com/dwnomin.htm, or (Poole and Rosenthal, 2006)
preliminary evidence suggests that while the groups supporting foreign aid (liberal Democrats
and liberal Republicans) have persisted, the groups supporting free trade have been less stable.
The second observation is that differences within parties tend to be much greater for aid
than they are for trade. Ideological differences also seem to split Democrats much more than
Republicans for aid votes, while this is less true for trade votes. Table 1 suggests that ideology
will be a better predictor of legislative voting on aid than on trade.
<Insert Table 1 here>
Table 1 also suggests an important question in thinking about internationalist policy: does
it require bipartisan support? Do Democrats need Republicans to support aid and do Republicans
need Democrats to support free trade? Table 2 investigates this by graphing, for each vote, half of
all votes cast and the number of pro-aid or pro-trade votes coming from the Democratic or
Republican parties (respectively). This clearly shows that in both aid and trade, support is needed
from members of the opposite party. Internationalism is a bipartisan policy.
<Insert Table 2 here>
Another way to compare how legislators within parties vote on trade and aid is to match
up all trade and aid votes within a congressional session, and calculate simple cross-tabulations.
E.g., one might cross-tabulate the NAFTA final passage vote and the amendment to cut aid funds
to the World Bank in the 103rd Congress. Unlike many other studies we include many votes in our
analysis, and thus presenting the results from all such vote pairs in cross-tabulated form would be
unwieldy. Instead we graph a measure of ordinal association, the gamma score (which is
equivalent to the Yule’s-Q that Broz reports (2005) in his work on similar issue areas), between
trade and aid votes for each party for all combinations of votes in our sample that are in the same
session. The left hand axis for the top two graphs in table 3 is the gamma score, and the right
hand side is the z-score obtained by dividing the gamma score by its asymptotic standard error.
The organization of our cross-tabulations that produce these scores is such that the top row is an
anti-aid vote, the bottom row is a pro-aid vote, the left column is anti-trade and right column pro-
trade. Given this organization, a positive gamma implies that the population tends to lie along the
diagonal from top left to bottom right. A negative gamma implies the population tends to lie on
the bottom left to top right diagonal (Liebetrau, 1983, pg. 75). I.e., legislators tend to be in the
anti-trade/anti-aid and pro-trade/pro-aid cells when the gamma score is positive and in the off
14
diagonal cells when the gamma is negative. When the associated z-score is larger in magnitude,
then the relationship (either on or off diagonal) is stronger.
<Insert Table 3 here>
The top left graph of table 3 shows Democrats in the early 1980’s were significantly
more likely to lie on the off diagonal: either pro-aid/anti-trade or anti-aid/pro-trade.6 Inspection of
the actual cross-tabulations suggests that during this period if a Democrat was pro-aid, they were
much more likely to be anti-trade. This block of legislators drives the negative gamma score we
see for this period. This association fades away for Democrats in the 1990’s and later, after which
a slight increase in the average number of pro-trade Democrats occurs on several trade votes. For
Democrats in the 1990’s and up, the gamma score was insignificant for most vote cross-
tabulations. The off diagonal relationship in the early period gets lessened as more Democrats
support trade in the later period. Republicans, on the other hand, did not have a significant
measure of association between the two policy areas in the 1980’s. But in the 1990’s and later
Republicans who were pro-aid tended to also be pro-trade. Of course, most Republicans were
anti-aid and pro-trade. But as more Republicans became pro-aid/pro-trade during this period the
gamma score turns positive and is significant in this later period. This observation couples nicely
with table 1, as in the later periods more liberal Republicans began to favor both aid and trade
liberalization, whereas more conservative Republicans have increasingly opposed
internationalism in both issue areas. Taken together, this means that the saliency of party in
staking out pro or anti-internationalist positions has changed over time. Democrats in the earlier
period were better able to keep a pro-aid anti-trade position, but some Democrats began to take
more pro-trade positions in the 1990’s. This possibly suggests that changes in executive branch
leadership, from the Reagan/Bush I era to the Clinton era may be driving these changes. Clinton’s
pressure on Democrats to vote for NAFTA may be part of this. More curious is the relatively high
percentage of Democrats voting in favor of the trade liberalization under George W. Bush.
Republicans on the other hand had a less structured set of policy positions in the early period, but
have become increasingly divided along pro and anti internationalism lines.
Multivariate Analysis
The above analyses provide an overview of the groups supporting internationalism, but
treat party and legislator ideology as the only things we need to know. The next section tests
6 Non-ordinal measures of association, such as similar chi-square tests, reveal a similar story but do not give the directional inferences that the gamma measure does. Tau-b tests results are basically identical to the gamma measures.
15
predictions deduced from the political economy and ideology models using multivariate
regression analysis. This allows us to estimate how district level variables affect the voting of
legislators, and then to make comparisons across issue areas. In this section we turn to a panel
data setting that analyzes voting over our entire sample.
Our dependent variable is dichotomous, with a one indicating a vote in favor of
internationalism. Our data are collected in a panel format with the legislator-vote as the unit of
analysis. We estimate a series of probit models using our set of votes for each issue area. Our
model specification includes vote fixed effects to control for any unmodeled heterogeneity across
votes and differences in the yeas-nays margin across votes. Here we present results from a
marginal effects specification (‘population averaged’) (Liang and Zeger, 1993; Neuhaus et al.,
1991), which uses a GEE estimator with a probit link and an exchangeable within group
correlation structure. Thus, slope coefficients indicate the influence on a population of legislators,
not individual legislators per se; put differently, they tell us the average impact of a variable on an
average legislator’s probability of voting in favor of aid or trade. Heterogeneity across legislators
is accounted for by calculating robust standard errors. Our results do not change if we use a
random effects specification, which relaxes the assumption that the intercept for particular
legislators is identical across votes. Because of the relatively small number of observations per
legislator, we do not use legislator fixed effects. The panel specification means that we are
combining votes within and across Congressional sessions, which allows us to compactly analyze
our data. In our robustness checks below we show that running separate probit regressions for
each vote yields identical conclusion.
Independent Variables
Before presenting our results, we briefly describe our independent variables; a more
thorough discussion is in appendix 1. We test predictions made by Stolper-Samuelson models by
measuring capital endowments at the district level; Stolper-Samuelson models propose that the
greater the amount of capital (human capital or physical) used in the district (relative to unskilled
labor), the more likely is a vote in favor of aid and trade. Following other scholars (Beaulieu,
2002a, b; Broz, 2005; Broz and Hawes, 2006), we measure this by the percentage of people
working in high skill jobs (% HighSkill).7 We expect this measure of human capital to be
positively related to support for aid and freer trade.
7 Others have used similar measures, such as the percentage college educated (Bailey, 2001). We operationalized RV type dynamics by measuring the percentage of people in a district identified as being in net export oriented industries relative to those in import competing at the SIC four digit level (Baldwin and McGee, 2000). We constructed this measure for every Congressional session in our sample. We also constructed several measures used by other authors for all the years in our sample (Broz, 2005; Ladewig, 2006). We then conducted a series of model fit exercises using both Bayesian Information Criterion (BIC)
16
We also constructed measures of relative physical capital endowments by aggregating the
amounts of fixed capital and labor in manufacturing industries following the procedure used by
Ladewig (2006). The lnCapitalEstab variable measures the (log) of capital stock in all
manufacturing sectors, and the lnLaborManuf variable measures the (log) of manufacturing
employment, at the district level for all Congressional sessions in our sample.8 We expect the
capital measure to be positively related to support for internationalism, and the labor measure to
be negatively related. Stolper-Samuelson models also suggest that districts abundant in
agriculture should favor both aid and trade. We measure this by calculating a district’s total value
for livestock and crop production (MktValAgProd).
PAC contributions from corporate sources (CorpPAC%), labor groups (LabPAC % PAC),
and money-center banks (BankPAC %) are operationalized as a percentage of total PAC
contributions. All of these campaign contributions are implemented as the percentage of total
PAC contributions (per the advice in (Roscoe and Jenkins, 2005, pg. 60)) and taken from the
electoral cycle preceding the Congress where we observe an actual vote.9 Stolper-Samuelson
models predict that corporate PAC’s, including money-center banks, will support aid and free
trade, while labor groups will be in opposition. We also added a variable for unionization
(%union) in order to see if districts that had higher union density were less favorable to aid and
trade, as Stolper-Samuelson models would predict.
Ideologically based theories suggest that legislators from more conservative districts will
support free-trade but oppose foreign aid, while legislators from more liberal will have opposite
preferences. Following scholars in American politics, we measure district ideology as the
percentage of the two party vote for the President that goes to the Republican candidate,
PrezVoteRepub% variable. In the US two party system, we argue that an individual’s political
ideology is well indicated by his/her vote for one of the two major parties. The district average of
Republican votes should indicate how conservative the district is likely to be.
If the Presidential dominance hypothesis is right, then what should matter are Presidential
party dynamics. Thus we created a Presidential support variable using data from David Rohde, and Bayes’ factors. Our results generally extend Ladewig’s findings beyond the 102nd Congress where his analysis left off: SS fits Congressional voting significantly better than RV. However, RV fit the data better on several votes in this post 102nd Congress period, including the important NAFTA and GATT votes in the 103rd Congress. In order to focus this paper on comparisons between aid and trade we briefly discuss these results in an online appendix and a more thorough discussion in a separate research note (Tingley, 2008). 8 We used GIS software to translate county level data to district level data whereas Ladewig manually made estimates from district maps. Nevertheless, our variables correlate at a very high level for the Congressional sessions where our analyses overlap (r>.9). 9 Readers should not that we are not using PAC contributions to argue that PAC money buys votes. Instead, we use PAC contributions as a way to measure the affinity of different groups in society to different types of policies. Ultimately we would love to use an appropriate instrument that would help us get around the various problems that plague the analysis of PAC contributions, and we invite readers to make concrete suggestions to this end.
17
coded as a 1 if the President was of the legislator’s same party and the President supported aid or
trade liberalization, and 0 otherwise (Rohde, 2004). If the Presidential dominance hypothesis is
right, this variable PrezSupport should be positive for both aid and trade. We also explored more
nuanced versions of this variable, but our substantive results did not change and thus we retain
our single dummy variable measurement.10 The nature of the data prevents us from investigating
a variety of other theoretical conceptions about the interactions between the President and
Congress. For example, we are unable to incorporate various forms of strategic interaction
between the President and Congress. This could arise if the President were only to issue positions
if a vote was expected to be close. Thus, while we think it is important to measure the potential
influence of the President (something that previous roll call voting analyses have not done),we
note that future work will be needed to better establish the relationship between the two branches
on these and other international economic policy areas. We think that if the president is powerful,
he should at least be able to pull his own party members to support his favored policies.
In addition to testing our hypotheses, we include a variety of common control variables
so that our results do not suffer from omitted variable bias. We include regional control variables
because other foreign policy studies have concluded that regional variations are important.
Trubowitz, for instance, argues that “Once the champion of Cold War internationalism, the
Northeast had become the defender of neo-isolationism by the 1980s”, while the South and West
had turned into the supporters of international engagement (1998, pg 232). His argument stresses
that regions within the US have distinctive economic stakes in American foreign policy;
isolationist versus internationalist policies affect the regions quite differently and hence make
their representatives differentially supportive of such policies. Many previous analyses of roll
call voting on foreign policy issues do not include these measures. By estimating models with
these controls, we are more confident that our inferences are not biased by regional differences
that we do not otherwise measure.
Finally, we also included a number of demographic variables at the district and state level
(when district level variables were not available). A more complete discussion of these variables
is in appendix 1.
10 Instead of the dichotomous PrezSupport variable we considered the full range of dichotomous variables produced by 1) the President’s position (support aid/trade, oppose aid/trade, no position) and 2) whether legislator was of the same party as the President. Unfortunately, including these variables and vote fixed-effects generate problems in the statistical analysis because particular dummy variables started to make vote fixed effects drop. For example, take a vote where the President did not take a position. Then the dummy variable representing the President taking a pro position and the legislator being a member of the President’s party would be collinear with the dummy variable for that vote, because all legislators would have a “0” for that vote. Hence, the vote fixed effect drops, and the interpretation of the other dummy variables becomes quite problematic because the omitted categories become combinations of vote fixed effects and the omitted Presidential category.
We present six different multivariate models in table 5 for aid and trade votes. We wish
to avoid both criticisms of omitted variables and ‘garbage can’ regressions (Achen, 2002), and
thus we move from relatively simple regressions to more complicated ones. Our first regression
contains our core political economy variable (%HighSkill) and PAC variables. The second two
sets of regressions add party, demographic control variables, and regional control variables. Next
we drop party and add our district level ideology variable. The final set of regressions includes
uses alternative district level measures of Stolper-Samuelson.11 We first discuss the direction and
significance of our variables for each issue area, and then move to comparing models and
estimating the substantive impact of our variables. We estimate models separately explaining
support for aid and trade in order to isolate the effect of our variables on each issue area. While
we believe that there could be a relationship between legislator’s preferences about the two issue
areas, this relationship is neither uniform nor constant, as shown in table 3, and thus we do not
believe that our estimation strategy is biased in any systematic way by not including relationships
across the two issue areas.
Our results strongly suggest that the null hypothesis of idiosyncratic votes by legislators
can be rejected. Further, the presidential dominance hypothesis receives no support in the aid
case and only limited support in the trade one. Our economic interest and ideological variables
11 We estimated a number of other models to check the robustness of our results. We discuss some of these robustness checks below, but note that our results are consistent across models.
19
predict votes well and suggest that a core group of legislators supporting an internationalist policy
exists.
The Stolper-Samuelson model does quite well in both aid and trade votes. Our measure
of capital endowments (%HighSkill) is highly significant and positive for both aid and trade
votes, indicating that legislators from districts with high levels of human capital are more likely to
vote in favor of foreign aid and free-trade. The results for trade are consistent with those found in
the literature. The strong positive relationship for aid is as predicted by the Stolper-Samuelson
model, but is perhaps more surprising. The same variable used by others to study trade
preferences and support for financial bailouts has the same effect for aid voting. This result
indicates that a core group of supporters exists for an internationalist policy that is composed of
similar groups on trade and aid. Thus our results provide an important extension of previous
findings in trade policy to that of foreign aid voting.
One criticism of the %HighSkill variable is that it might capture other effects beyond
economic factors, and thus the measure might be problematic despite the fact that a range of
published studies use this measure. For example, those districts with higher skill levels might also
be more cosmopolitan and hence open to international engagement. Unfortunately, criticisms of
this type often stop without suggesting ways to control for these other influences. One suggestion
is to include the percentage of adults that have a college degree; however, this correlates with the
skill variable at a high level, ranging from .65 in the 1980’s to .97 in the 2000’s. This multi-
collinearity is a statistical nuisance (akin to having a small n or low variance in IV’s), but over-
determination between Stopler-Samuelson models and cosmopolitanism is a theoretical problem.
The extent of this theoretical over-determination is not clear. Those who think that the college
education variable taps cosmopolitanism assume a tight relationship between education and
cosmopolitanism, though we are unaware of why this necessarily must be the case. Similarly it is
not clear why more education always leads to greater acceptance of free market ideas, as higher
level education need not be in neo-liberal or classical economics.
As noted above, we collected other measures of capital and labor endowments, following
the procedure laid out by Ladewig (2006). We obtain largely the same results for these variables
as we do for our high skill variable. The lnCapitalEstab variable is positive for aid and trade,
whereas the lnCapitalEstab is negative for both aid and trade. These variables are significant
in our simpler models, but lose their significance once we condition on many other variables.
Taken together with our results for our skill measure, we find relatively strong support for our
models of economic interests in both aid and trade. Thus, the same empirical relationship
identified in trade seems to apply to aid.
20
Our next set of results focuses on the contributions of interest groups representing capital
and labor. We argue that the Stolper-Samuelson model can help explain the direction of these
contributions. This model suggests that capital and labor groups will be united in their
preferences that their contributions will correlate with legislative voting records. We expected
similarities across trade and aid voting in the role of capital and labor PACs. But the data do not
show this. The role of organized labor in particular follows a different pattern than expected:
labor PAC contributions (LabPAC%) tend to go to legislators that supported foreign aid and
opposed free trade, even when we control for party. While labor’s opposition to trade is
understandable in terms of our model, the positive relationship to aid is more puzzling. This
positive relationship between labor and foreign aid has been documented elsewhere using
archival material from the AFL-CIO (Milner and Tingley, 2007). These results pose an
interesting contrast across the issue areas. While organized labor was pro-trade before the 1970s,
changing economic conditions eroded its support for free trade. By the time our sample starts in
the late 1970’s, organized labor has become increasingly hostile towards trade liberalization.
Conversely, organized labor, which supported and benefited greatly from early aid programs like
the Marshall Plan, continued to support foreign aid programs.
Another interesting difference between aid and trade policy is that while money-center
bank PAC contributions (BankPAC%) tend to make up a higher percentage of total PAC
contributions for legislators voting in favor of foreign aid, this relationship is not significant for
trade votes. This extends the results of several studies (Broz, 2005; Broz and Hawes, 2006) to
foreign aid, and illustrates how large banks appear to have preferences for international
engagement beyond their support for the financial bailouts and IMF programs that Broz and
Hawes have considered.12 Instead these banks also support legislators that support foreign aid
programs more generally. In a separate paper we consider the historical bases of this support for
aid more directly. Their impact is negligible, however, for trade. Money center banks do not
appear to have salient and unified interests with respect to trade policy, or at least these interests
are not discernible from the data we consider. Thus a key difference between aid and trade policy
appears to be that large, internationally exposed banks help maintain support for aid but have less
influence or salient preferences over trade policy.
12 We do not consider here the difference-in-differences approach that Broz also uses. This procedure took the set of legislators that changed their vote from one financial bail out vote to a later vote. This change was regressed on changes in bank PAC contributions. We also conducted these analyses for our PAC variables. Depending on the model specification our results were in the expected direction and significant or insignificant. This was further complicated by not having a large sample of “switchers”, making us less confident in the statistical technique. Finally, this procedure explicitly samples on the dependent variable, and hence we have good reason to doubt the inferences that come out of it.
21
Contributions from corporate PAC’s (CorpPAC%) had a surprising negative relationship
in several models for trade votes. While negative significant in some of the trade models, this
variable was insignificant for the set of aid votes. Our Corporate PAC result is somewhat
surprising for trade.13 Labor’s opposition to trade and bank PACs support for aid confirm our
hypotheses about PAC’s, but labor’s support for aid and corporate PACs sometimes significant
opposition to trade do not. Taken together, the organized interest groups provide mixed support
for the economic interest hypotheses.14
District agricultural production (MktValAgProd) is not significantly related to the set of
foreign aid votes in our sample,15 but highly agricultural districts tend to be more free trade
oriented. As we control for region, we do not believe this is picking up regional effects. This
result is consistent with Stolper-Samuelson predictions on trade, which view the US’s relative
abundance in land favoring trade liberalization.
One important criticism of models 1 and 2 in table 5 might be that our economic
measures do not pick up on the ideological orientation of congressional districts, and thus suffer
omitted variable bias. It is possible, if not likely, that many of our economic and demographic
variables correlate with aggregate district ideological orientation. Thus including district level
measures of ideology may decrease the impact of our economic variables. Nevertheless, we re-
estimate our models to include district ideological variables. Our liberal-conservative measure,
the percentage of the district’s two party vote that is for the Republican Presidential candidate
(PrezVoteRepub%), is negative and significant for aid, while positive and significant for trade. As
predicted by theories of political ideology, more conservative districts favor trade and oppose aid,
largely because of their beliefs about the role of the government in the economy.
It is also common for scholars to include a legislator-specific ideological variable in
regression analysis of voting, though the appropriateness of this is debated. While we are more
interested in evaluating the influence of district level economic variables, our results largely do
not change if we include legislator’s DW-Nominate scores.16 Interested readers may consult these
13 This measure of capital PAC correlates negatively with free-trade support in the early 1980’s, but then trends positively. We note that both our capital and labor PAC measures aggregate across organizations, possibly obscuring more fine-grained dynamics. We obtain similar results if we restrict our attention to contributions by the AFL-CIO. 14 We also constructed a variable that measured the percentage of PAC contributions from the top 50 exporters, recreating and extending a measure first used by Nollen and Quinn (1994) This variable was positive but insignificant for both our trade and aid panels, though the variable was much closer to attaining conventional significance levels for the trade panel. 15 Not surprisingly, when we look at the set of votes on food aid legislation (PL 480: Food for Peace), legislators from districts with high agricultural production significantly favor foreign aid. We discuss these results in a separate paper. 16 The only change is that our LaborPAC% variable remains positive, but loses its statistical significance. Our union measure remains positive and significant. Labor’s tight support of foreign aid is documented elsewhere, and thus our substantive conclusions remain unchanged (Milner and Tingley, 2007).
22
statistical models in our on-line appendix. However, we take a moment to comment on the results
from using legislator-specific ideological variables instead of district level measures of liberal-
conservatism. As we saw in table 1, more conservative legislators are more likely to oppose
foreign aid and support free trade, while more liberal legislators do the reverse. As the critical
reader may have observed, DW-Nominate scores do a better job sorting out within party voting
behavior over foreign aid than for foreign trade. Not surprisingly, when we estimate the marginal
effect of a legislator’s DW-Nominate scores, the magnitude of the effect is almost four times as
large for aid as for trade. 17 We observe a similar difference between aid and trade using the
district level measure of ideology (PrezVoteRepub%) (see below). Controlling for a range of
other factors, ideology appears to matter much more for foreign aid than it does for trade.
Political party (Party) is highly significant for both our aid and trade votes. Yet while
Republicans are more likely to vote in favor of free trade, they are less likely to vote in favor of
foreign aid. While Democrats are more likely to vote against free trade, they generally favor
foreign aid. Thus party, not surprisingly, follows a patter similar to what we saw with the liberal-
conservative ideology measure. Clearly including both party and ideology would be inappropriate
as they are highly correlated. Nevertheless, table 1 shows that ideology can have a significant
impact within parties, especially for aid. We return to these within party differences below.
Models 3 and 4 include regional variables. Compared to legislators from the Northeast,
southern legislators were less likely to support foreign aid. Legislators from the West (West) and
Midwest (Midwest) are more likely to vote in favor of free trade compared to Northeastern
legislators. This supports only part of Trubowitz’s claims; the West has become a solid bulwark
of support for trade policy. These regional controls do not, however, change the results for our
other variables, and thus regional differences do not seem to powerfully predict aid and trade
votes.
We also consider several district-level demographic variables that may be related to
support for aid and free trade. Legislators from districts with large foreign-born populations
(%ForBorn) are likely to vote in favor of foreign aid in several of our models. A district’s foreign
born population had a less salient effect on trade votes. While positive, this variable was not
consistently significant. Consistent with evidence that the African-American lobby supports
foreign aid, legislators from districts with large African-American populations (%Black) are more
likely to favor foreign aid. No consistent relationship holds for trade votes. Our measure of
unemployment level (Unemploy%) and change in (state level) unemployment were almost always
negative but never significant. While the economic stress unemployment causes was a frequently
17 Marginal effects results calculated using the mfx2 command in STATA show that the larger aid marginal effect does not have a confidence interval that overlaps with the small trade marginal effect. See on-line appendix and discussion below.
23
cited source of opposition to both trade and aid, this does not materialize in our data. There is no
consistent relationship between (the log of) median income (LogMdnIncm) and either trade or aid
voting. Some have used income as a measure of altruism, suggesting that richer districts should
provide more support for aid (and perhaps trade) (Krueger, 1996); we do not find this result.18
The preceding discussion focused on district and legislator preferences as predicted by
political economy models and those emphasizing ideology. As discussed in our literature review,
an alternative (though complimentary) way to explain legislative behavior is to more formally
consider the relationship between the Executive and Legislative branch. The President clearly
plays a key role in shaping both aid and trade policy, and thus the President may be able to also
influence legislative voting by taking particular positions on the bills.
Our multivariate results show that the effect of having a President of the same party issue
a statement in support of aid or free trade, versus not taking a position, is positive and significant
for trade, but negative and insignificant for aid (PrezSupport).19 There is also some evidence that
the impact of the President taking a position is less than that of some of our other economic
variables.20 This contrasts with Kesselman’s (1961) findings that emphasized the role of the
President in influencing aid role call voting, and supports the skepticism about this result
articulated by Asher and Weisberg (1978). To see why there is a difference between aid and
trade, we can look at this relationship using simple cross-tabulations. In our sample of 10 high
saliency aid votes, the President took a pro-aid position on 6 votes, and no position on the 4
others. In our sample of 20 high saliency trade votes, the President took a pro-trade position on 16
votes, and no position on 4 others. Consistent with previous work, the President—regardless of
his party—tends to have a strong preference for both foreign aid and free trade. The question that
interests us, however, is whether the President’s position taking has an influence on legislative
voting.
Table 6 suggests that the influence of the President on aid voting is very different from
his influence in trade policy. Whereas the President’s decision to take a position on trade
18 This measure does have a positive and significant coefficient when we use Ladewig’s measures of capital and labor concentration, but only for trade voting. 19 As discussed previously we also measured the influence of the President in several other ways, but generally came to the same set of conclusions. One exception was if we simply took the President’s position (either pro-internationalism (1) or no position (0) in our sample). This variable was negative and significant for both aid and trade. This suggests that the influence of the President in trade policy may operate chiefly through members of the President’s party. We note that we do not model potential sources of strategic interaction between the President and the legislature, and hence we treat the set of cases where the President issues a position as no different from cases where the President does not issue a position. This is clearly inappropriate. However, given our interest is between comparing aid and trade, we see little reason why the strategic dynamics would be different across the issue areas. And hence the difference in impact that we identify remains interesting. 20 The marginal effect of our Presidential variable is several orders of magnitude smaller than the marginal impact of several of our economic variables. For dichotomous variables like Presidential support, the marginal effect is calculated as a discrete change in predicted probability.
24
legislation results in more legislators voting in favor of trade, this is not the case for aid votes. In
fact, legislators who are of the same party as the President are less likely to vote in favor of the
President’s position. We suspect that the influence of the President on trade policy is driven by
the shift of some Democrats towards a pro-trade position in the 1990’s when Clinton was
president, as shown in table 3.
<Insert Table 6 here>
To summarize, there are interesting similarities and differences between legislative
voting on foreign aid and trade policy. The similarities suggest that there are underlying forces
that keep together—or prevent the breakdown of—a broad internationalist policy orientation.
Predictions from the Stolper-Samuelson model help explain the similarities in these groups. Our
measure of district skill levels, which indicates human capital endowments relative to unskilled
labor, correlates positively with both foreign aid and free trade support. The population of
legislators coming from districts with relatively high levels of capital endowments is more likely
to support both foreign aid and free trade. This can readily be seen by looking at “defectors”
within each party: i.e., Democrats voting against aid or for trade, and Republicans voting in favor
of aid or against trade. Table 7 gives the mean levels of our skill variable, with 95% confidence
intervals, for defectors and party line voters over all votes in our regression analyses. Republicans
supporting aid tend to come from districts with much higher human capital, as do Democrats
supporting trade, relative to members of their parties voting the standard party position.
<Insert Table 7 here>
Differences in support for the two issue areas suggest ways that support for
internationalism is weaker—or could become weaker—in one area as opposed to the other.
Organized labor’s opposition to free trade, and support for foreign aid, is perhaps the most
interesting difference. Elsewhere we have documented labor’s relationship to foreign aid more
closely by using records from the AFL-CIO memorial archives. While it is clear that umbrella
organizations like the AFL-CIO, which have a high profile overseas, are generally supportive of
foreign aid, it is less clear what other non-affiliated union organizations prefer (if they have
salient preferences at all). Labor’s opposition to free trade has been extensively documented
elsewhere. Money-center banks appear to be strong supporters of foreign aid, most likely because
of the complementary role of aid for private finance. They do not appear to play a role in keeping
together support for, or against, free trade.
25
There are several potential explanations for these similarities and differences. Our results
suggest that distributional consequences matter for both aid and trade (as reflected by our human
capital measure), but that the effect of ideology for trade votes is smaller. The substantive
effect—the change in probability of supporting aid or trade—of increasing the skill variable by a
single standard deviation while holding other variables at their mean values is very similar across
the two issue areas: 9.4% for aid and 8.5% for trade (using model 4 from table 5). The
substantive effect of increasing the district ideology variable by a single standard deviation (using
model 4 from table 5) is dissimilar across the two issue areas: -8% for aid and 4% for trade. Table
8 lists these substantive effects for several of our variables. Thus the substantive impact of
ideology is approximately twice as large for aid as for trade. Interestingly, the substantive impact
of our skill measure is also larger than for our ideology measure for both issue areas, though
making more thorough comparisons would require more statistical detail than space permits here.
<Insert table 8 here>
The fact that ideology has a larger substantive effect in aid than in trade also holds using
marginal effects calculations. One explanation of this is that aid policy is an explicitly re-
distributional program, and hence taps ideological motivations more centrally. The effect of our
measure of relative capital (to unskilled labor) endowments is similar across the issue areas,
though slightly higher in aid than in trade. The substantive effects of the skill variable are similar
in magnitude to those seen in studies in other issue areas (e.g., (Broz, 2005)). The substantive
effects of using Ladewig’s measures of capital and labor endowments show a slightly larger
effect in trade than in aid. Thus the effect of these district economic variables is similar across the
two issue areas. Economic interests generate a clear set of preferences regarding aid and trade;
class cleavages strongly mark the trade area but are more diluted in the aid area where they are
cross cut by ideological factors.
VI. Robustness
Our models correctly predict a very high percentage of actual votes by legislators:
between 71%-74% for trade and 74%- 76% for aid. The proportional error reduction for both
issue areas is between 29-34%.21 The accuracy of our model does not appear to be driven by our
estimation strategy or other potential misspecifications. As mentioned, our results do not change
if we use a random effects specification.
21 To do this we used the PRE post-estimation routine in STATA.
26
We find similar results if we look at our votes individually. Analyzing votes separately
and displaying results in an economical way is difficult when there are many votes under
consideration. We adopt two approaches that 1) allow us to compare the effects of our variables
across aid and trade and 2) compactly display meaningful results from the analysis of many votes.
Our first procedure calculates a probit model for each vote separately, and then takes the
marginal effect for each of the variables at its mean. The marginal effect is the slope of the line
tangent to the probability link function, here calculated at the mean of the variable. We then
calculate the 95% confidence intervals of the marginal effects. Tables 9 and 10 display these
results for two of our key variables: %HighSkill and PrezVoteRepub%. On the left hand side of
these graphs are the results for aid votes and on the right hand side are results for trade votes.
There is a relatively consistent relationship between voting and these variables, with a significant
or near significant, marginal effect in a majority of votes. For trade the effect of our district
ideology level was more mixed than with aid. In aid there was a consistent and strong negative
relationship. Here again we see how the effect of district ideology is much more salient for aid
than for trade. These tables also demonstrate how our pooling strategy does not drive our results.
<table 9 here>
<table 10 here>
These tables do not however calculate substantive effects that many political scientists
find useful. Our on-line appendix explains and presents substantive effects from looking at each
vote separately using a Bayesian strategy with non-informative priors. By using Bayesian probit
models, we generated the posterior distribution of a variable’s slope coefficient for every vote in
our sample. We then use a procedure to calculate substantive effects of variables and see if there
are changes over time.22 We use this procedure for sheer convenience in reporting the substantive
effects of a variable for many votes within one graph and in a way that also picks up our
estimation uncertainty. Analogous results would obtain using a straight MLE approach. We find
similar substantive effects to those calculated using the pooled model in table 5. We find little
change over time; that is, the significance and impact of our key variables does not systematically
22 Specifically, we used the MCMCprobit package in the MCMCpack library to obtain posterior distributions on our slope coefficients for every vote. Then we fixed all variables at their session mean and (and party at Republican if party variable included), drew the ‘jth’ observation from each of these posterior distributions of coefficients, and calculated the predicted probability of voting in favor of aid or trade. Then, we increased each of the variables by their session standard deviation one at a time and calculated the change in probability, again for all j observations from each of the coefficient posterior distributions. Finally, we subtract the jth observation when all variables are held at the mean from the jth observation when a variable was increased by a standard deviation. This leaves us with a distribution of substantive effects that incorporates our uncertainty in the estimated coefficients without having to rely on delta method approximations.
27
change over time if we estimate the results one vote at a time. Our use of a panel estimator does
not drive our results, and again our inferences would be substantively similar were we to look at
individual votes.
Another possibility is that our results might be biased by including legislators with
heterogeneous electoral incentives. Legislators with particularly secure electoral prospects might
respond to district level factors differently compared to legislators elected by very close margins.
To address this, we reran our model 6 from table 5 and excluded legislators winning by more than
a certain percentage. Our results, presented in table 11, change little and, if anything, become
stronger for several of our key variables.
<Insert Table 11 here>
We also estimated models separately by party. The skill variable is positive and
significant for both issue areas and both parties. The Ladewig measures of economic interests are
correctly signed and significant for both issue areas for Democrats, but lose significance in the
model with many district level control variables. The effect of our district level ideology variable
for aid is more significant among Republicans, whereas for trade this variable was more
significant for Democrats. The effect of the percent of foreign born variable was more significant
for Democrats than Republicans in both aid and trade, with a positive coefficient in both issue
areas. The effect of our labor and bank PAC variables remained strong and significant for both
parties and issue areas. Overall, then our results are robust when separating our sample by party,
and we provide these results in our appendix.
Religiously based ideological preferences may also relate to support for international
policy (Busby, 2007). We collected religious membership data through the Association of
Religious Data Archives at the county level, and then transformed to the district level using
geographic concordance software. Regression models including the percentage of Jewish,
Evangelical, Protestant, and Catholic adherents by district had the same results as our main
models. The Evangelical measure was negative and significant for aid and the Jewish measure
was positive and significant for trade. Including these additional controls did not change our
results. Finally, we estimated our models including votes that we classified as being ‘low
saliency’. These votes satisfied most of our selection criteria, but were not as salient in tapping
preferences for aid or trade policy. Our results barely change by including these additional votes
and we report these results in our appendix. Overall, our results are quite robust.
VII. Conclusion and future research
Trade and foreign aid are two key dimensions of American foreign economic policy, and
as such play important roles in engaging the US with the rest of the world. In this paper we have
28
taken a first step toward theoretically identifying the political groups that support American
internationalism. We show that in both areas legislators do not vote idiosyncratically or simply on
the basis of the president’s position. Nor is partisanship the best predictor of these preferences.
Legislators appear to understand the economic effects of trade and aid policies on their districts
and on their political support within them, and vote accordingly. The distributional consequences
of aid and trade as reflected in the economic characteristics of their districts seem to affect their
preferences. Their votes are also shaped by the ideological preferences that their constituents
hold. Our theories do a better job of explaining support for aid and trade than do either random
voting, party affiliation, or presidential domination models. In particular, Stolper-Samuelson
models of trade and aid preferences receive important corroboration here. The strong support for
international engagement through aid and trade policy by districts with large percentages of high
skill workers is the most striking common source of support for the two policies; given that skill
levels have been rising across the US since the beginning of our sample this seems evidence for
future support for internationalism. This economic model then helps explain commonalities in
support for internationalism that would not be revealed by other models.
We show that bipartisan support for aid and trade remain essential. Presidents must build
cross party support; they cannot expect majorities in either party to favor aid or trade. But unlike
studies which assume that the groups supporting (or opposing) internationalism are similar across
issue areas, we demonstrate that their nature differs between aid and trade. In aid this group
includes liberal Republicans and liberal Democrats who come from districts that have relatively
high levels of capital endowments (especially human capital); in addition, organized labor
supports this as does the international banking community. The surprise here, compared to trade
policy, is organized labor’s continuing support for aid, and with it liberal Democrats’. Labor and
liberal Democrats support for aid is unexpected given the Stolper-Samuelson model. Opposition
to aid is centered in districts with conservative Republicans and conservative Democrats
especially from areas with high levels of unskilled labor. Again, the economic interests model we
use does not predict Republicans’ distaste for aid. Districts that have greater support for
conservative ideology also tend to oppose aid.
In trade, economic interests play a stronger role and they pit capital against labor more,
but still this splits the parties. For trade openness, the supporting groups are conservative
Democrats and liberal Republicans who come from districts with high concentrations of human
capital, high levels of agricultural production, and locations in the Midwest and West. Opponents
of trade include liberal Democrats, primarily those from regions with low skilled populations,
labor organizations and their PACs, and, increasingly, conservative Republicans from low skill
labor districts. Support from presidents of a legislator’s party also helps to turn legislators away
29
from opposing freer trade. While party affiliation or presidential power are important elements
for explaining a legislator’s preferences vis-à-vis trade or aid, our economic and ideological
theories provide a more complete understanding of these groups’ preferences.
The groups supporting trade have also changed over time, while those supporting aid
have been quite stable. Conservative Republicans have started to oppose trade, and in their
opposition to trade and aid have become a major anti-internationalist bloc. Moderate
Republicans, while a shrinking proportion of Congress overall, remain the bulwark of support for
internationalism, as they have since World War II. Besides these changes, the groups supporting
and opposing trade seem to have been fairly stable over this twenty-five year period (1979-2004).
Examining our data by individual votes yields a very similar story to our pooled analyses,
showing little systematic change. This stability is remarkable given the momentous changes in
world and American politics during this period. The end of the Cold War, globalization, the war
on terror (although just starting as we end), the polarization of American politics, and the
increasing Republican control of Congress all mark this period and could have dramatically
changed these patterns. Our data suggest that they did not.
The groups supporting foreign aid and free trade have important, and interesting,
similarities and differences. While economic interests appear to weigh on legislators in both issue
areas (something that is consistent with what legislators, Treasury Secretaries, etc. say in the
historical records we consulted (e.g., debates, testimony)), ideology appears to play a stronger
role for foreign aid than for trade. Given the weaker distributional consequences of aid, this result
might not be surprising. Those who believe the government’s role should be minimalist do not
support foreign aid, while those with a more expansive view of the government’s role in the
economy are more supportive.
One potential implication of our analysis is that support for foreign aid would be
significantly compromised if labor began to oppose aid, just as it did with trade beginning in the
early 1970’s. While supporters can change, labor continues to play an important role in support
for foreign aid. But this support is not guaranteed. For example, in the early 1990’s several labor
groups expressed frustration with aid money going to fund export-processing zones in developing
countries. This suggests that while scholars have become increasingly interested in the influence
of Evangelical Christian groups on foreign aid (Busby, 2007), scholars—and pro-aid
politicians—should not overlook underlying support from powerful labor groups like the AFL-
CIO.
Another implication of our comparison of voting on aid and trade is that the interaction
between the executive and legislative branches can differ depending on the issue area. Thus,
scholars who make broad claims about the role of the President versus Congress in foreign policy
30
(Canes-Wrone et al., 2006; Howell and Pevehouse, 2007) may miss important heterogeneity
across issue areas. Foreign policy comprises a number of interconnected policy areas. The politics
of international engagement are not monolithic across these areas. While the President
undoubtedly plays a crucial role in the conduct of foreign policy, the Congress matters in
important ways for matters beyond military issues. Presidential dominance of the policy agenda is
not the norm in aid, though there does appear to be some influence in trade policy.
Further work remains to be done to understand support for internationalism in the US.
We have taken the most systematic steps so far towards analyzing the domestic politics around
two important areas of international engagement. Theories of political economy and ideology can
help us to explain support for internationalism in the US. They add considerably beyond party
affiliation or presidential domination to our understanding of the political dynamics of foreign
policy making in the US. These models enable us to better understand the domestic sources of
preferences for internationalism in the world’s biggest great power.
______________________________________________
Appendix 1: Variable Descriptions
Votes
All roll call votes were obtained using the Voteworld program
(http://ucdata.berkeley.edu:7101/new_web/VoteWorld/voteworld/) or (www.voteview.com) . A
vote in favor of foreign aid or free-trade was coded as a 1, a vote in opposition coded as a 0.
Money-center Bank PAC Percent
The percentage of Political Action Committee campaign contributions made to a candidate by a
money- center bank in the previous election cycle. Money center banks are classified by the
Mean DW-Nominate by party; Confidence intervals in dashesAid Support and Ideology
-.70
.7D
W-N
OM
Mea
n
97 98 98 99 99 100
100
100
102
102
103
103
105
105
105
105
106
106
106
107
108
108
108
108
Congressional Session
Dem Supp Dem OppRep Supp Rep Opp
Mean DW-Nominate by party; Confidence intervals in dashesFree Trade Support and Ideology
Table 2
100
150
200
Vot
es
96 96 96 97 98 102
103
104
104
108
Congressional Session
Half All Votes Dem Pro-Aid
Aid VotesDo Dems Need Republicans for Aid Support?
5010
015
020
025
0V
otes
97 98 98 99 99 100
100
100
102
102
103
103
105
105
105
105
106
106
106
107
108
108
108
108
Congressional Session
Half All Votes Rep Pro-Trade
Trade VotesDo Repubs Need Dems for Trade Support?
33
Table 3
-6-4
-20
2G
amm
aZ-S
core
-.6-.4
-.20
.2.4
Gam
ma
97 98 102 103 108 108Session
Gamma Score
Gamma/AsympStanError
Gamma Measure of AssociationLegislator Votes on Individual Trade and Aid Bills: Dems
-20
24
6G
amm
aZ-S
core
-.20
.2.4
.6.8
Gam
ma
97 98 102 103 108 108Session
Gamma Score
Gamma/AsympStanError
Gamma Measure of AssociationLegislator Votes on Individual Trade and Aid Bills: Repubs
0.2
.4.6
.81
% P
ro-A
id
96 96 96 97 98 102
103
104
104
108
Congressional Session
Republicans
Democrats
% Pro-Aid, Within Party
0.2
.4.6
.81
% P
ro-T
rade
97 98 98 99 99 100
100
100
102
102
103
103
105
105
105
105
106
106
106
107
108
108
108
108
Congressional Session
Republicans
Democrats
% Pro-Trade, Within Party
<Table 5 Below> [For Formatting Purposes]
Table 6: Presidential Support Aid No Pres Position Pres Supports
% voting anti-aid
% voting pro-aid
% voting anti-aid
% voting pro-aid
Democrats with Dem President 81 19 64 36Democrats with Rep President 83 17 55 45Republicans with Rep President 51 49 19 81Republicans with Dem President 24 76 32 68
Trade No Pres Position Pres Supports
% voting anti-trade
% voting pro-trade
% voting anti-trade
% voting pro-trade
Democrats with Dem President 19 81 56 44Democrats with Rep President 9 91 28 72Republicans with Rep President 62 38 79 21Republicans with Dem President 65 35 63 37
34
Table 7
Mean High Skill % (95% conf. Intervals) Democrat Aid Supporter .223 (.220, .226) Democrat Aid Opponent .192 (.190, .195) Republican Aid Supporter .265 (.260, .271) Republican Aid Opponent .228 (.225, .231) Democrat Trade Supporter .261 (.257, .265) Democrat Trade Opponent .228 (.226, .231) Republican Trade Supporter .276 (.273, .278) Republican Trade Opponent .240 (.237, .244)
Table 8
Variable Aid Trade Movement From Baseline Model Skill Increase by 1 SD 9.4% 8.5% BankPAC Increase by 1 SD 4.1% 0.8% LaborPAC Increase by 1 SD 21.8% -19.0% DistrictIdeology Increase by 1 SD -8.7% 4.1%
Move from Pro-trade position to no position PrezSupport 1.0% -26.2%
Estimation uses model 4 from table 5 with region set at Northeast
Table 9
-2 0 2 4 6MargEffect
aid
-2 0 2 4 6MargEffect
trade
Variables in model: HighSkill PrezVoteRepub ForBorn UnemplChangelnMdnIncm Union% Unemploy% %BlackCongressional Sessions run lower (earlier) to higher (later)
HighSkill
35
Table 10
-2 0 2 4 6MargEffect
aid
-2 0 2 4 6MargEffect
trade
Variables in model: HighSkill PrezVoteRepub ForBorn UnemplChangelnMdnIncm Union% Unemploy% %BlackCongressional Sessions run lower (earlier) to higher (later)
Bibliography Achen, C. H., 2002, Toward a New Political Methodology: Microfoundations and ART. Annual Review of Political Science 5, 423-450. Adler, S. E., Congressional District Data File, [congressional term]. University of Colorado, Boulder, CO. http://socsci.colorado.edu/~esadler/research.htm. Alesina, A. and D. Dollar, 2000, Who Gives Foreign Aid to Whom and Why? . Journal of Economic Growth 5, 33-63. Asher, H. B. and H. Weisberg, 1978, Voting Change in Congress: Some Dynamic Perspectives on an Evolutionary Process. American Journal of Political Science 22, 391-425. Bailey, M., 2001, Quiet Influence: The Representation of Diffuse Interests on Trade Policy, 1983-1994. Legislative Studies Quarterly February, 45-80. Baldwin, R. E. and C. S. McGee, 2000, Is trade policy for sale? Congressional voting on recent trade bills. Public Choice 105, 79-101. Beaulieu, E., 2002a, Factor or Industry Cleavages in Trade Policy? An Empirical Analysis of the Stopler-Samuelson Theorem. Economics and Politics 14, 99-131. Beaulieu, E., 2002b, The Stolper-Samuelson Theorem Faces Congress. Review of International Economics 10, 343-360. Bhagwati, J., R. Brecher and T. Hatta, 1983, The generalized theory of transfers and welfare: bilateral transfers in a multilateral world. American Economic Review 73, 57-76. Bhagwati, J., R. Brecher and T. Hatta, 1984, The paradoxes of immiserizing growth and donor-enriching ’recipient-immiserizing’ transfers: a tale of two literatures. Weltwirtschaftliches Archiv 110, 228-243. Bobbio, N., 1996, Left and right : the significance of a political distinction ( Polity Press, Cambridge, UK) Pages. Brakman, S. and C. v. Marrewijk, 1998, The economics of international transfers (Cambridge University Press, Cambridge, U.K. ; New York) Pages. Broz, J. L., 2005, Congressional Politics of International Financial Rescues. American Journal of Political Science 49, 479-496. Broz, J. L. and M. B. Hawes, 2006, Congressional Politics of Financing the International Monetary Fund. International Organization 60, 367-399. Bureau_of_Labor_Statistics, 2005, http://www.bls.gov/news.release/union2.t05.htm. accessed_07/29/06. Busby, J. W., 2007, Bono Made Jesse Helms Cry: International Norm Take-up and the Jubilee 2000 Campaign for Debt Relief. International Studies Quarterly 51, 247-275. Canes-Wrone, B., W. G. Howell and D. E. Lewis, 2006, Toward a Broader Understanding of Presidential Power: A Re-Evaluation of the Two Presidencies Thesis. working paper. CenterForGlobalDevelopment, 2007, http://www.cgdev.org/section/initiatives/_active/cdi/_country/united_states/. Collier, K. and M. Munger, 1994, A Comparison of Incumbent Security in the House and Senate. Public Choice 78, 145-154. Destler, I., 1995, American Trade Politics (Institute for International Economics) Pages. Elman, M. F., 1995, The foreign policies of small states: Challenging neorealism in its own backyard. British Journal of Political Science 25, 171.
Fleck, R. K., 2002, Democratic Opposition to the Fair Labor Standards Act of 1938. The Journal of Economic History 62, 25-53. GPO, 2004, http://www.gpoaccess.gov/usbudget/fy05/sheets/b97.xls accessed 02/28/07. Haas, M. L., 2007, The United States and the End of the Cold War: Reactions to Shifts in Soviet Power, Policies, or Domestic Politics? International Organization 61, 145. Hansen, J. M., 1991, Gaining access : Congress and the farm lobby, 1919-1981 (University of Chicago Press, Chicago) Pages. Hiscox, M., 1999, The Magic Bullet? The RTAA, Institutional Reform and Trade Liberalization. International Organization 53, 669-698. Hiscox, M., 2002, Commerce, Coalitions, and Factor Mobility: Evidence from Congressional Votes on Trade Legislation. American Political Science Review 96. Howell, W. G. and J. C. Pevehouse, 2007, While Dangers Gather: Congressional Checks on Presidential War Powers Pages. Husain, I., 1993, Trade, Aid, and Investment in Sub-Saharan Africa. World Bank Policy Research Working Paper 1214. Ikenberry, G. J., 2005, Creating America’s World:The Sources of Postwar Liberal Internationalism. working paper. Johnson, R. D., 2006, Congress and the Cold War (Cambridge University Press, New York) Pages. Jones, R., 1984, The Transfer Problem in a Three Agent Setting. Canadian Journal of Economics 17, 1-14. Kalt, J. P. and M. A. Zupan, 1993, Capture and Ideology in the Economic Theory of Politics. Public choice theory. Volume 2: The characteristics of political equilibrium Elgar Reference Collection series. International Library of Critical Writings in Economics, vol. 24, 238-259. Kemp, M. C., 1995, The Gains from Trade and the Gains from Aid: Essays in International Trade Theory (Routledge, New York) Pages. Kesselman, M., 1961, Presidential leadership in congress on foreign policy. Midwest Journal o f Political Science 5, 284-289. Krueger, A., 1996, Observations on International Labor Standards and Trade. Annual World Bank Conference on Development Economics April. Ladewig, J. W., 2006, Domestic Influences on International Trade Policy: Factor Mobility in the United States, 1963 to 1992. International Organization 60, 69-103. Lancaster, C., 2007, Foreign Aid: Diplomacy, Development, Domestic Politics (University of Chicago Press, Chicago) Pages. Liang, K.-Y. and S. L. Zeger, 1993, REGRESSION ANALYSIS FOR CORRELATED DATA. Annu. Rev. Pub. Health 14, 43-58. Liebetrau, A. M., 1983, Measures of Association (Sage Publications, Beverly Hills) Pages. Mayer, W. and P. Raimondos-Moller, 1999, The Politics of Foreign Aid. University of Cincinati; Copenhagen Business School. Mayer, W. and P. Raimondos-Møller, 2003, The Politics of Foreign Aid: A Median Voter Perspective. Review of Development Economics 7, 165-178. McCarty, N., K. Poole and H. Rosenthal, 2006, Polarized America: The Dance of Ideology and Unequal Riches (MIT Press, Cambridge) Pages.
Milner, H. and D. Tingley, 2007, The Domestic Politics of Foreign Aid: American Legislators and the Politics of Donor Countries, (APSA 2006) http://www.wws.princeton.edu/hmilner/Research.htm. Neuhaus, J. M., J. D. Kalbfleisch and W. W. Hauck, 1991, A Comparison of Cluster-Specific and Population-Averaged Approaches for Analyzing Correlated Binary Data. International Statistical Review 59, 25-35. Nollen, S. D. and D. P. Quinn, 1994, Free Trade, Fair Trade, Strategic Trade, and Protectionism in the U.S. Congress. International Organization 48, 491-525. OECD/DAC, 2007, Table 2a: Disbursements. http://www.oecd.org/dataoecd/50/17/5037721.htm. Palmer, G. and T. C. Morgan, 2006, A Theory of Foreign Policy (Princeton University Press, Princeton, N.J.) Pages. Poole, K. T. and R. S. Daniels, 1985, Ideology, Party, and Voting in the U.S. Congress, 1959-1980. The American Political Science Review 97, 373-399. Poole, K. T. and H. L. Rosenthal, 2006, Ideology and Congress (Transaction Publishers, Somerset, NJ) Pages. Rogowski, R., 1989, Commerce and coalitions: how trade affects domestic political alignments (Princeton University Press, Princeton, NJ) Pages. Rohde, D. W., 2004, Roll Call Voting Data for the United States House of Representatives, 1953-2004. Compiled by the Political Institutions and Public Choice Program, Michigan State University. Roscoe, D. D. and S. Jenkins, 2005, A Meta-Analysis of Campaign Contributions' Impact on Roll Call Voting. Social Science Quarterly 86, 52-69. Tingley, D., 2008, Factoral Versus Sectoral Models of Trade Voting: A Research Update. working paper. Trubowitz, P., 1998, Defining the National Interest: Conflict and Change in American Foreign Policy (University of Chicago Press, Chicago) Pages. WTO, 2006, International Trade Statistics. Geneva.