Since the 1970s, globalization has advanced broadly. Some four billion
people, roughly two-thirds of the world's population, have joined the
world economy during the past twenty-five years as part of the increasing
integration of ever more countries into the world trading system. Declining
trade barriers have contributed significantly to this expansion of world
trade. Countries across the globe—ones as diverse as the Philippines,
Zambia, Mexico, Poland, Chile, Bangladesh, Ghana, Korea, and
Morocco—have all chosen to liberalize unilaterally their trade
policies. In the developing countries, this “rush to free trade”
gathered momentum in the mid-1980s.1
As the
International Monetary Fund (IMF) pointed out in 1992,
“[s]ince the mid-1980s, there has been a marked shift
in the orientation of the trade and industrial policies of most
developing countries away from a heavy reliance on direct intervention
and inward-looking industrial policies toward less controlled and more
export-oriented trade regimes.” More recent studies show that
this change in policy has occurred in most regions and continued
throughout the 1990s.
2 Countries have
chosen to integrate their economies into a global one by dismantling
protectionist barriers.
This substantial change in trade policy is surprising. Over the
years, many scholars have emphasized the durability of the status quo
in economic policy.3
In trade
policy the status quo bias is viewed as even more significant, as the
benefits of protectionism are highly concentrated while its costs are
diffuse.
4 A status quo bias exists if
trade liberalization is welfare enhancing but policy change is not made
because the political costs of liberalization outweigh the economic
benefits to be gained from it.
Vested interests in
protectionist sectors can be tenacious pressure groups with preferred
access to policymakers. Any change in the protectionist status quo is
thus unexpected. For the less-developed countries (LDCs) to abandon
their protectionist, import-substituting industrialization (ISI)
strategy was surprising, given existing models of the political economy
of trade policy.
5 Explanations for this change have emphasized three factors. Some
scholars claim that economic crises have forced countries to reform and
liberalize. Others have claimed that external pressures from the United
States, Western countries in general, or international institutions,
such as the World Trade Organization (WTO), IMF, and World Bank, are
responsible. Finally, the spread of neoliberal policy ideas is often
credited with bringing economic liberalization. We examine these claims
with respect to trade policy and introduce another factor that we think
is important and underappreciated.
Beginning nearly a decade before this move to free trade was a global
movement toward democracy. In 1975, there were approximately thirty
democracies in the world; by 1992 there were about eighty-nine, which
was roughly half the total number of independent countries in the
world.6
By 2002, this figure had risen to more
than 120 countries.
7 Are these two trends in
economic and political reform related? Are democracies more likely to
initiate trade liberalizing reforms than autocracies?
Many different claims about how political institutions affect trade
policy exist, but systematic theory and data are lacking. Rodrik argues
that any change in a political regime is likely to induce trade
reforms. “Historically sharp changes in trade policy have almost
always been preceded (or accompanied) by changes in the political
regime. Not all political transformations result in trade reform, but
sharp changes in trade policy are typically the result of such
transformations.”8
The conventional wisdom,
however, has been that democracy is not propitious for economic reform,
and indeed that nondemocratic countries should be more likely to
liberalize, as Chile in the 1970s suggests. As Geddes summarizes,
“until recently, it was widely accepted that democracies,
especially fragile, uninstitutionalized new democracies have difficulty
carrying out economic liberalization because its costs make it
unpopular and hence politically suicidal to elected officials.
Consequently, it was argued, authoritarian governments should be more
capable of initiating and sustaining major economic reforms.”
9 Other scholars have argued that regime type makes little difference
for economic reform.10
Finally, Haggard and Kaufmann argue that
differences within regime types (that is, among democracies and among
autocracies) are likely to have more effect than is regime type
itself.
11 A few systematic studies of the impact
of democracy on trade liberalization exist.
12 But only recently have some
claimed that democracy might make economic reform more likely.
13 In sum, the systematic investigation of the
impact of political institutions on the decision to liberalize trade
merits greater attention.
In the political economy literature, the question is approached from
the opposite end: what determines trade policy? A vast literature
explores this issue. We point out several contributions that lay the
groundwork for our argument, focusing on the Heckscher-Ohlin and
Stopler-Samuelson theorems.14
The main
alternative to this political economy model is the Ricardo-Viner, or
specific factors, model. It postulates that sectors of the economy, not
factors of production, gain or lose from trade. The more mobile agents
are, the more Heckscher-Ohlin models apply. See, for example, Hiscox 2002.
Mayer explores the dependence
of tariff rates on the distribution of factor ownership, the costs of
voting, and the degree of factor mobility and industry diversification
in the economy.
15 His median voter model using
the Heckscher-Ohlin framework (two factors and two sectors) has
produced interesting findings about the political economy of trade. In
a series of papers, for instance, Dutt and Mitra have shown that in
such a framework an increase in inequality raises trade barriers in
capital-rich countries and lowers them in capital-scarce ones, and that
left-wing governments adopt more protectionist policies in capital-rich
countries but more free trade policies in labor-rich economies than do
right-wing governments.
16 These cross-sectional findings are evidence that Heckscher-Ohlin
models may describe the politics of trade policy well. Others using
them have also been able to explain important political outcomes.17
But these
models have not been used often to explain change in policy over time,
such as in the recent move to free trade.
18 Using a different model of trade policy, Grossman and Helpman 1994 analyze how interest
groups bid for protection with their campaign support. These works show
how changes in political participation, especially among interest
groups, affect trade policy outcomes. These theories seem less capable
of explaining how developing countries around the world were suddenly
able to liberalize their trade policies.
This trade
liberalization process is especially anomalous in light of the
well-known tendency of countries to resist reform.
19 However, many
agree that an economy in crisis should facilitate reform.
20 Few systematic tests of these hypotheses
exist, and we will control for these factors in our empirical
analysis.
The three main arguments explaining economic reform in the LDCs focus
on external pressures, political leaders and their ideas, and economic
crisis. Many scholars have argued that a variety of pressures external
to the LDCs have forced them to change their policies and join the
global economy. Some arguments focus on U.S. hegemony associated with
the end of the Cold War, others on the role of private investors and
countries' desire for foreign investment, and others on pressures
from international institutions, such as the IMF, World Bank, and
WTO.21
Some scholars attribute economic reform to changes either in
political leaders or in the ideas that leaders hold about economic
development.22
For them, economic failure
(both absolute and relative) prompted leaders to give up on ISI and
move toward more market-friendly economic policies; reform came from
new leaders with different ideas or from old leaders with new ideas.
Other scholars emphasize economic crisis as the spur to reform.23
Crises
underscore the failure of old policies and create an environment in
which radical, new policies can be tried. They also increase
countries' reliance on external benefactors, such as private
investors or the IMF. The crisis argument is related to external
pressures, as well as the claims about new ideas and wars of attrition.
In order to show that regime type also plays a role, one must try to
control for these factors.
This article is divided into four sections. In the next section, we
present some facts about trade policy and democratization in the LDCs
during the past thirty years. In the following section, we outline how
democracy may be related to trade liberalization. In the third section,
we present econometric evidence covering more than 100 LDCs from
1970–99. The final section concludes by arguing that regime type
and changes in it affect trade policy, even when controlling for many
other factors. Since the late 1970s, democracies and democratization
have led to lower levels of trade barriers, ceteris paribus.
Data on Trade Liberalization and Democracy
In the 1960s and 1970s, many LDCs had trade regimes marked by
extensive, overlapping, and often prohibitive trade restrictions; ISI
was the policy regime of choice. Groups that gained from these policies
tended to be powerful supporters of the political leaders, and changing
trade policies, it was believed, would inflict severe costs on the
regime's main backers. Many scholars agree that in large parts of
Latin America, Africa, and Asia, the groups that gained from ISI were
urban owners of industry (that is, capitalists) and urban,
higher-skilled and often unionized workers (more capital-endowed
workers); the losers tended to be less-skilled, poorer workers,
generally rural ones.24
Even the debt crisis of the early 1980s was unable to shake this
coalition from its economically inefficient policies. Many scholars
report that protection increased in the wake of the debt crisis,
especially in Latin America. As Haggard and Webb point out, “if
the interest group configuration does not change, it is unclear how the
status quo (protection) could ever be transcended.”25
The status quo bias against liberalization is
thus a key issue; something must change so that this bias can be
overcome.
Beginning in the mid-1980s, the status quo was overturned. As the
African Development Bank notes, “a series of reform and
liberalization efforts undertaken by developing countries in the past
decade and a half represents an effective shift in development strategy
from an inward-oriented, import-substituting framework … to an
outward-oriented export-promoting framework.”26
The data we have collected on the LDCs demonstrates a
massive change from 1970–99 for five different measures of trade
policy. First, a dramatic decline of about 60 percent occurred in the
average tariff level for about forty LDCs from 1982–99; statutory
tariff rates fell from an average of about 30 percent in the early
1980s to around 12 percent in 1999.
27 World Bank 2000b compiles this
data from UN Conference on Trade and Development TRAINS and the
WTO's integrated database.
Tariff duties collected as a
percent of imports have also fallen for a large number of LDCs. This
data, available for about eighty-five to ninety countries during the
1973–97 period, shows that duties peaked in 1973 at 21 percent of
imports and then fell almost 53 percent, ending at about 10 percent of
the value of imports.
28 This is consistent with the statutory tariff
rate data; by the late 1990s, both series show that tariff rates had
dropped in the LDCs to around 10 percent.
More evidence of trade liberalization comes from nontariff barriers
(NTBs). While coverage is very limited (on average, thirty LDCs are
sampled every few years), the data show that beginning in 1984 at
around 38 percent, average NTB coverage ratios declined to around 17
percent in 1998, or about a 55 percent decrease.29
UN Conference on Trade and Development 1994 and 2000.
Falling tariffs were not compensated for by rising NTBs.
Another measure of trade policy is the dichotomous categorization of
countries into open and closed trade regimes, as constructed by Sachs
and Warner and updated by Horn and Wacziarg.30
Sachs and Warner 1995 define an
economy as closed if any one of the following is true: NTBs cover 40
percent or more of trade; average tariff rates are 40 percent or more;
the black market exchange rate depreciated by 20 percent or more
relative to the official exchange rate during the 1970s or 1980s; a
socialist economy existed as defined by Kornai
1992; or there was a state monopoly on exports. Updated by Horn Welch and Wacziarg 2003.
According to
this data, which covers ninety LDCs per year from 1970–99, the
percentage of LDCs scored as open rose from 15 percent in the early
1970s to 64 percent in 1999, with the biggest changes beginning in the
mid-1980s.
Finally, using an outcome measure, trade dependence (exports plus
imports as a percent of gross domestic product [GDP]),
liberalization has been associated with real changes in the developing
economies' exposure to international competition. From 1970, trade
as a percentage of the domestic economy rose from an average of 55 percent
to nearly 85 percent in 1999, or a 55 percent increase.31
All five
measures of trade policy underscore the same message: since the mid-1980s,
countries across the globe have decided to dramatically reduce their trade
barriers and move toward freer trade.
Our central question, then, is what disturbed the historical
equilibrium involving interest groups and political leaders around
protectionism? Why did political leaders in many developing countries
choose to lower their trade barriers? As noted before, standard
political economy models predict that vested interest groups with
concentrated gains from protection will strongly resist any such
reduction in barriers; furthermore, the gains from freer trade will be
diffuse and thus present collective action problems for groups desiring
lower levels of protection. For some reason, however, political leaders
decided to alter this status quo by lowering average trade barriers
significantly.32
We have data on average
barriers. These aggregate rates across many sectors of the economy.
Within this overall average decline, it is probable that some interest
groups succeeded in maintaining protection but many sectors failed to
do so.
We argue that the movement toward democracy in developing countries
can provide part of the answer. Democratization opened up new avenues
of support for freer trade. Leaders recognized that previously
disenfranchised groups had become part of the voting public; as we show
later, these new groups benefited more from trade liberalization than
continued high protectionism. Needing the support of these new groups
in a democratic setting, leaders saw that their ability to use trade
barriers as a strategy for garnering political support had declined.
Democratic political competition meant that leaders were likely to
liberalize trade to appeal to these new groups to ensure their
political survival. We agree that “democracies possess greater
capacities for promoting change and breaking free of unholy interest
group coalitions than generally thought.”33
Democratic leaders in a number of developing countries chose trade
liberalization as a means of gaining broader political support. In no
country were trade barriers reduced to zero, and thus we are not
arguing that protectionist interest groups no longer mattered. But in
democratizing countries they mattered less than before, and thus
leaders could liberalize more than previously.
Evidence of democratization among the LDCs is plentiful. Beginning
before the move to free trade, a global movement toward democracy
erupted. To measure regime type, we use the 21-point Polity index
constructed by Gurr et al., Jaggers and Gurr, and Marshall and Jaggers,
ranging from −10 for a highly autocratic state to 10 for a highly
democratic one.34
Using data from Polity IV, the average regime
score for about 110 LDCs fell from −3.4 in 1970 to a nadir of
−4.71 in 1977, and then rose to a high of 1.8 in 1999.
35 For more discussion, see the section on
empirical analysis below.
Similarly, the dichotomous regime
classification from Alvarez et al. shows how the number of democracies
has increased over time.
36 Starting from
about 16 percent in the early 1970s, the percentage of democracies
falls to a low of 14 percent among 110 LDCs in 1977, but then rises to
49 percent in 1999. Similar to the Polity data, this series shows that
the percentage of democratic countries was falling in the 1970s but
began a vigorous rise after 1977. Importantly, both figures
show that the process of democratization among the LDCs began in the
late 1970s, almost a decade before widespread trade liberalization got
underway. The wave of democratization preceded that of trade
liberalization.
A number of cases illustrate our claim. For instance, in 1982 Bolivia
began a democratic transition. Shortly after this, in 1985, the new
government launched an economic reform program called the “New
Economic Policy,” which eliminated all quantitative restrictions
(QRs) and lowered tariffs.37
Argentina shows a similar picture. There the
democratic transition began in 1983 and substantial trade reform
followed in 1988; it included tariff reductions and the elimination of
import licensing.
38 In other parts of the world this pattern has been repeated. For
instance, in the Philippines the first parliamentary elections occurred
in 1984 and were followed by presidential elections in 1986, after the
ouster of the dictator Ferdinand Marcos. After this political change
(1986–88), the new leaders led by Corazón Aquino began
changing economic policy, replacing QRs with tariffs and then reducing
tariffs.39
As the
Wall Street Journal reported in 1986,
economic decline felled Ferdinand Marcos and unless Mrs.
Aquino can turn the Philippine economy around, she will find it difficult
to lay the foundations for enduring democracy…. What really will
matter will be how much economic reform Mrs. Aquino can push through the
Philippine political system…. Economic recovery still depends,
however, on whether the Philippines can throw off decades of
import-substitution and cronyism, and start selling its wares in the
world marketplace…. For months, Mrs. Aquino has been gamely chipping
away at monumental domestic trade barriers, which routinely include
effective protection rates of 80 percent and 90 percent. Predictably, this
liberalization is drawing sharp protests from entrenched, protected
industries…. Mrs. Aquino is not giving up on this grueling
liberalization process.40
Wall Street
Journal, 18 September 1986, 1.
South Korea also shows a similar pattern. A democratic transition
occurred during 1987–88, and the new government followed this
with an economic reform program beginning in 1989 that lowered trade
barriers significantly by 1992.41
Bangladesh is another interesting case. Between 1986 and
1992, the country underwent a democratic transition; beginning in 1987,
President Hussain Muhammad Ershad's government introduced a slow
liberalization process that only became substantial after the early
1990s when the political situation stabilized. Tariffs fell from
roughly 90 percent in 1990 to 20 percent by 1996.
42 The democratization
of many East Central European countries, such as Poland and Hungary, was
followed by economic reform, including massive trade liberalization.
43 Recently, democratization in Africa has lent an impetus to trade
liberalization.44
Zambia is a
telling example. In 1991, a democratic transition occurred when
Frederick Chiluba and his Movement for Multiparty Democracy party (MMD)
overwhelmed Kenneth Kaunda in the presidential election. Chiluba's
MMD campaigned for a radical change in economic policy, and this was
one factor in his victory over Kaunda, who had run the country since
its independence. In 1992, Chiluba then launched a massive economic
reform program that featured trade liberalization; tariffs were lowered
from 30 percent in the late 1980s to 13 percent by 1996.
45
These cases, among others, suggest that changes in regime type may have
an influence on changes in trade policy. Trade liberalization followed a
change in regime type—in particular, democratization. Why did
leaders choose to liberalize trade in the face of vested protectionist
interests?
Trade Liberalization and Democratization:
An Argument
How might democracy contribute to trade liberalization?
Democratization means a movement toward majority rule with universal
suffrage in contested elections. We call the group of actors who
participate in the selection of political leaders “the
selectorate,” following the terminology of Bueno de Mesquita et
al.46
In a democracy, the
selectorate is the part of the population that is eligible to vote. In a
nondemocracy, the selectorate is that subset of the population upon whose
political support leaders could potentially rely to remain in office. The
winning coalition is the minimal set of individuals in the selectorate
whose support an incumbent needs to remain in office. In a democracy, this
may be a simple majority of voters, whereas in an autocracy it may be
more complex. Increasing the selectorate tends to imply an increase in
the size of the winning coalition; in a majority voting situation, this
means a change in the median voter.
Democratization is a process involving an expansion of the
selectorate. Democracies choose political leaders through popular
elections, while autocrats maintain their position with the backing of
small groups, such as the military elite, large landowners, or heavy
industrialists. Democratization implies an expansion of the selectorate
and the winning coalition, which changes the optimal policies that
leaders will choose. This expansion implies a change in the composition
of the selectorate, and hence a change in its preferred policies. To
retain office, leaders must adjust their policies to be responsive to
the preferences of the expanded selectorate.
With democratization, leaders can build new coalitions in favor of
different policies because they can appeal for support to newly
enfranchised groups. As Bienen and Herbst point out, “political
liberalization may change constituencies and therefore promote economic
reform…. Democratization would alter processes of economic
decision making and the nature of economic policies…. A regime
that has to take account of voters who have not previously had
influence may well shift the allocation and distribution of
resources.”47
Democratization is important for trade policy: the optimal level of
protectionism for political leaders is a declining function of the size
of the winning coalition. Even if the same political leaders remain in
office, an expansion of the winning coalition reduces the amount of
protection that is optimal for them. We expect that, as democratization
occurs, political leaders will alter their trade policies—with or
without public urging—to gain the support of this larger
selectorate.
Our argument follows the logic of the Heckscher-Ohlin and
Stopler-Samuelson theorems as they have been advanced by Mayer and by
Yang.48
See Mayer
1984; and Yang 1995. An alternative
would be a specific factors model in which sectors of the economy
(import-competing versus export-oriented firms), rather than factors of
production such as capital and labor, form the major coalitions. It is
much debated as to which of these provides a better account of the
political economy of LDCs.
In a Heckscher-Ohlin world, the
Stopler-Samuelson theorem shows how individuals benefit or lose from
changes in trade policy given their endowments of capital and labor.
Mayer and Yang each connect this to politics by showing how political
leaders respond to voters' preferences vis-à-vis trade
policy. Developing countries by definition possess relatively less
capital than labor.
49 Almost all
developing countries trade mostly with developed ones. More than 80
percent of LDC trade is with the OECD countries. Markusen and Wiggle 1990. Of total world trade,
only about 6 percent is between LDCs. Hence it makes sense to consider
all LDCs relatively well endowed with labor given their primary trading
partners, the OECD countries.
Because the vast majority of
their trade is with rich developed countries, their import-competing
sectors tend to be capital-intensive, and therefore protectionism will
benefit those individuals well endowed with the relatively scarce
factor, capital. Liberalizing trade policy in the LDCs results in a
gain in income for, and a reduction in the prices of imported goods
bought by, those well endowed with the relatively abundant
factor—that is, labor— in these economies.
In developing countries, workers and the poor tend to gain from trade
liberalization through increases in their income and reductions in the
prices they must pay (especially for import-competing goods). As an
economist notes, “the protection of capital-intensive industries
affected [Latin America's] ability to create
employment…. In developing countries more-open trade regimes
result in higher employment and in a more even distribution of income
than protectionist regimes…. Export-able industries tend to be
significantly more labor intensive than import-competing sectors
… and the removal of external sector distortions tends to
strengthen the process of employment creation in most developing
countries.”50
Democratization will
thus enfranchise a new group of voters with preferences for lower
levels of protectionism.
As noted above, in nondemocratic countries those eligible to
determine their countries' leaders are part of a very restricted
selectorate. In many autocracies, either the voting that takes place
does not affect the choice of leadership, or voting does not occur at
all. However, even dictators must earn the support of some
“majority” of the populace that has the right to determine
the leadership. In many autocracies, this selectorate consists of the
richest individuals, and hence those who own the most capital. Those
with “voting” rights in autocracies thus own above-average
levels of capital, and benefit the most from high levels of protection.
In Latin America, for example, “the protective system generated
large benefits to local industrialists … and urban workers.
This, of course, was achieved at the cost of depressing the earnings
and incomes of rural workers.”51
The Philippines under the autocrat
Marcos was another glaring example of this.
Democratization changes which groups political leaders must garner
support from; political competition within democracies induces leaders
to appeal to new coalitions of voters and hence offer new policies to
win their support. As a country democratizes, the selectorate grows and
the preferences of the enlarged selectorate will differ from before. As
democracy advances to include those who own less and less capital (that
is, workers and the rural poor), the median voter's capital
ownership will decline, and his or her most preferred tariff rate
decreases as well.52
Hence political leaders in a
democracy can appeal to these groups who are less well endowed with
capital and whose interests may be better served by less protectionism.
Leaders in this setting have new sources of support and ones that
prefer lower trade barriers.
As Weyland suggests, “democratization reduces the political
clout of the vested interests that benefited the most from the old
development model, such as protectionist business sectors and the
military. At the same time, it enhances the role of the electorate,
including the large mass of poor people who received meager benefits
under the old development model … Democratization weakens
politically some prime beneficiaries of state interventionism….
[It] creates new cleavages that diminish the political power
of business associations, which were once dominated by protectionist
sectors.”53
This seems to have been
the case in countries such as Bolivia, the Philippines, Bangladesh, and
Zambia, where democratization preceded, and helped leaders initiate,
trade liberalization.
In sum, in developing countries where autocratic governments depend
on support from a small selectorate and thus are not responsive to the
overall population, the governments can employ extensive protectionism.
Democratization, however, may break down the old coalition supporting
protectionism, and can thus lead to change in the status quo. As the
selectorate grows, leaders may find it in their political interests to
modify their trade policies. As the democratic selectorate in
developing countries become less well endowed with capital, high levels
of trade barriers no longer compensate them for their loss of income
from, and the higher costs imposed by, protectionism. Thus as the
political regime becomes more democratic and leaders build new
coalitions of support, political competition may induce leaders to make
their trade policies less restrictive. Leaders may reason that by
lowering barriers, they can increase the incomes of workers and garner
more of their support in future elections. Protectionist interest
groups may remain powerful in these democracies, but they will be less
influential than before because leaders now have new sources of support
they can turn to.
Is it realistic to assume that workers and the poor gain from trade
liberalization? Do they not lose from it and thus oppose it? It is
important to separate trade liberalization from the other economic
reforms. Different economic reforms have different distributional
consequences. Many reforms, such as privatization, pension reform, and
increasing labor market flexibility, may have significant negative
effects (at least in the short term) for workers, and are often
bitterly opposed by them.54
In contrast, the
distributional consequences of trade liberalization may not hurt
workers or the rural poor. In the same countries where opposition to
other reforms has been high, trade liberalization has often been
supported—or at least not opposed—by workers and their
organizations, as Murillo shows in Mexico, Argentina, and
Venezuela.
55 Weyland notes that public support for market
reform, including trade liberalization, was actually strong in a large
number of Latin American countries, and Fishlow shows that these
reforms have not been visibly opposed most of the time.
56 Baker provides systematic evidence that trade
liberalization was and has remained very popular in Latin America. As
he notes, “aggregate support for free trade is much higher than
support for privatization, indicating that positive orientations toward
free trade are not simply a vague expression of support for economic
liberalization. Instead, many citizens favorably single out free trade
from the list of Latin America's recent market reforms.”
57 Baker 2003, 428.
Baker has a different explanation for this continued support than we do
for the initial liberalization.
In Africa, as van de Walle
points out, trade policy reform—although limited—has not
been opposed by social groups either.
58 Other political scientists point out that voters may support
governments even when times turn bad because they think bad times now
are a signal of good times to come. As Stokes shows for a wide variety
of countries, “Our most startling result is that in every country
people sometimes reacted to economic deterioration by supporting the
government and its economic program more strongly.”59
Polls in Mexico show that voters strongly favored trade
liberalization via the North American Free Trade Agreement (NAFTA) and
thought it would mean more jobs and higher wages for them.
60 Similarly, research on the transition economies in East
Central Europe shows that in most of these countries, public support
for trade liberalization was strong and never wavered.
61 Political leaders do not seem to have lost
support by pursuing trade liberalization; publics often are willing to
keep supporting regimes that liberalize trade even if the economy slows
in the short term, in the hope of future gains.
62 As for whether trade liberalization actually benefits workers in
LDCs, the data are mixed. Some studies show that increased openness
leads to faster economic growth, which benefits workers;63
others
cast doubt on claims relating trade policy to growth.
64 At the
microeconomic level, scholars have shown, surprisingly, that trade
liberalization may not have important positive or negative effects for
firms or workers.
65
Theoretically, Heckscher-Ohlin (and Stopler-Samuelson) models suggest
that workers should gain from trade liberalization in LDCs; and
empirically the extant data do not reject this claim.
Empirical Analysis
According to our argument, democracies should be oriented more toward
free trade than nondemocracies, and an increase in the degree of
democracy should induce a move to liberalize trade. Many theories
describe the relationship between political and economic reforms, but
little systematic empirical work exists on trade liberalization. Part
of the reason is that both political and trade regimes are difficult to
measure. We test our argument while also trying to control for the
leading contending propositions, relating to economic crises, changes
in leaders and ideas, and external pressures. Our data set is a
time-series cross-section (TSCS) one, containing 179 developing
countries, territories, and dependencies from 1970–99.66
In the analysis, we have only about 100
countries represented; most small countries, territories, and
dependencies are missing data.
Our central hypothesis is that
more democratic countries should have fewer trade barriers, ceteris
paribus; an increase in democracy should prompt a reduction in trade
barriers.
Our central independent variable is the type of political regime in
place in a country at time t. The political regime variable
comes from Polity III and Polity IV, which collected data on the
political characteristics of 177 countries between 1800 and 1999.67
To measure each state's regime type, we
employ the widely used index constructed by Gurr et al. and Jaggers and
Gurr.
68 This index combines data on five factors that
capture the institutional differences between democracies and
autocracies: (1) the competitiveness of the process for selecting a
country's chief executive, (2) the openness of this process, (3)
the extent to which institutional constraints limit a chief
executive's decision-making authority, (4) the competitiveness of
political participation within a country, and (5) the degree to which
binding rules govern political participation within it. Each of these
five measures is directly related to our emphasis on political
competition and the size of the selectorate.
Following Gurr et al. and Jaggers and Gurr, these data are used to
create an 11-point index of each state's democratic
characteristics (democ) and an 11-point index of its
autocratic characteristics (autoc).69
The difference between these indices,
regime = democ −
autoc, yields a summary
measure of regime type that takes on values ranging from −10 for
a highly autocratic state to 10 for a highly democratic one. This
measure captures the variation both within democracies and among
autocracies. For instance, not all autocracies are the same. Some
autocracies have a more expansive selectorate than others. Mexico, for
example, is usually seen as more democratic than Saudi Arabia or China;
Polity catches these differences. For instance, in 1980, Mexico scores
a −3, while Saudi Arabia gets −10 and China, −7.
Polity's scoring of autocracies correlates highly with a
categorization of autocracies created by Geddes; countries she codes as
single-party regimes, then military ones and then personalistic ones,
represent a declining scale of democracy which correlates at roughly
the 0.6 level with
regime.
70 Similarly,
differences among relatively democratic countries can be discerned.
Summary statistics for
regime, as well as our other variables,
are listed in
Table 1. We use a lagged
version of
regime (from one to three periods) in the
regressions to mitigate endogeneity issues.
To increase the robustness of our analysis, we use two other measures
of regime type. First, from Geddes's data on autocracies we
construct a variable ranging from 1 to 8, where 8 is most democratic.
Because more than two-thirds of the regimes in Polity are autocracies,
it is interesting to see if their differences matter. Polity does this
one way, and Geddes does it another. Geddes argues that certain
autocracies are longer-lived because of their internal characteristics;
she sometimes equates these differences with the nature and extent of
the groups that support the autocrat.71
We interpret this as arguing that
autocracies vary according to the size of their selectorate and winning
coalition. She claims that single-party systems have the broadest
selectorates, followed by military-run governments and then personalist
regimes.
In personalist regimes, one individual dominates the military,
state apparatus, and the ruling party if there is one. Because so much
power is concentrated in the hands of one individual in personalist regimes,
he generally controls the coalition-building agenda…. In contrast to
single-party regimes, the leader's faction in a personalist regime may
actually increase benefits to itself by excluding the rival faction from
participation. Where the main benefits of participation in the government
come from access to rents and illicit profit opportunities, benefits to
individual members of the ruling group may be higher if they need not be
shared too widely.72
We code as the most autocratic regimes those with personalist
elements, then those with military involvement, and least of all those
with a single party. We call this variable dictator, which is
supplemented with data from Przeworski et al. to add a code for
democracy.73
Hence all
countries not classified as autocracies by either Geddes or Alvarez et
al. (for which Alvarez et al. have data) are coded as democratic.
74 dictator is coded as 1 =
personalist regimes; 2 = mixed regimes with some personalist element; 3
= personalist mixed with military; 4 = personalist mixed with single
party; 5 = military; 6 = military mixed with single party; 7 = single
party; and 8 = democracy. All countries coded as autocracies by Alvarez et al. 1996, but missing in Geddes, 1999, are coded as mixed (= 2).
Note that this measure is blunter than Polity because it does not
discern among democracies. We use both a lagged version of the
dictator variable and dummy variables for each category.
Our third measure of regime type is the dichotomous categorization
created by Alvarez et al. and Przeworski et al.75
Their
measure codes a regime as democratic if and only if high political
offices are chosen through fair and free contested elections where
alternation of leaders occurs. This measure, which is quite blunt since
it assumes that the democratic transition occurs completely in one
year, has been criticized.
76 It makes no
distinction between types of autocracies or levels of democracy. But it
is highly correlated with the other two measures (r = 0.80 with
regime and r = 0.70 with
dictator). We use this
variable
dem in lagged form.
Our central dependent variable measures a country's trade policy
in year t. We want to predict the extent of protectionism, or
conversely openness, of the trade regime. This is notoriously difficult
to measure.77
Pritchett, for instance,
finds little correlation among different measures of openness in the
literature.
78 We therefore follow
Edwards, and use a variety of measures.
79 However, we face a difficult
challenge in finding time-series as well as cross-sectional data, as we
are interested in how openness changes over time. We use two
alternative ways of measuring trade policy.
Our first measure is a country's (unweighted) average
statutory tariff rate (tariff).80
This is the most
appropriate measure for our model, which predicts a decline in tariff rates
in response to the shift toward democracy. But it is poorly measured.
Various countries were sampled several times in the 1980s and then
almost yearly from 1992 to 1999, giving 907 total observations.
Our second measure of trade liberalization is a dichotomous
classification of trade regimes into open and closed ones.81
Sachs and Warner code a country as
closed (
sw open = 0) if any one of the following is true: NTBs
cover 40 percent or more of trade; average tariff rates are 40 percent
or more; the black market exchange rate depreciated by 20 percent or
more relative to the official exchange rate during the 1970s or 1980s;
a socialist economy existed as defined by Kornai; or there was a state
monopoly on exports.
82 This measure is very useful
because it considers many forms of protectionism; it is much broader
than tariff rates and thus more comprehensive. Others have used it, and
it seems highly correlated with more precise data on trade
liberalization episodes.
83 It is correlated
with statutory tariff rates at −0.49. Our regressions using this
variable are logistic since it is dichotomous. Because trade policy is
hard to measure and comes in a variety of forms, using both measures
gives us a broader picture of how trade policy is changing over time
and across countries. If both show that democracy is related to trade
liberalization, then confidence in our claims should be enhanced.
Changes in trade policy may also arise because of factors other than
changes in the political regime, and we need to control for these.
Conventional wisdom as well as scholarly work suggests that three sets
of factors should be included. First, we need to control for several
economic variables.84
It is
often argued that small countries tend to be more open than large ones.
85 We thus measure a
country's size by its population, using the lagged value of the
natural log of population as our control (
lnpop). A
country's level of economic development is also likely to affect
its trade policy; more-developed countries tend to have smaller trade
barriers.
86 Hence we add the lagged value of per capita real
GDP as a control (
gdp pc).
The second set of control variables relates to both internal and
external political factors that might affect trade policy. First among
these is economic crisis. As noted above, the war of attrition models
of economic reform often point out that the greater the distortions
caused by the policy, the more likely reform is; economic crisis is one
way to measure these distortions. Tornell, among others, claims that
countries are likely to liberalize their trade regimes after an
economic crisis because a crisis generates conflict among the
powerful.87
He defines a country as being in crisis if either
its inflation rate is skyrocketing or its real income is plummeting. We
use a similar definition. Crises are either occurring or not; they are
not long-continuing events by definition, and they are extreme events,
not yearly changes in economic variables.
One important difficulty with this variable is defining what
constitutes a crisis. Different economic problems may be more important
in different countries, and different levels of those problems may
trigger different evaluations of whether a crisis exists. We use two
different notions of economic crisis, both of which stress that crises
are unusual and extreme shocks. One notion from Tornell deems a crisis
to exist if one of two conditions holds: either the country's
inflation rate is 40 percent or more and it increases by 25 percent or
more from the year before, or per capita GDP falls by 15 percent or
more from the previous year (ec crisis = 1). Our second form
of crisis involves the balance of payments. Here a crisis exists
(bp crisis = 1) if a country's level of international
reserves falls to less than the equivalent of three months' worth
of imports. This second notion of crisis relates to a country's
debt and capital flight problems.88
Although balance-of-payments crises tend to most affect
countries with fixed exchange rates, most of the countries here had
some form of fixed exchange rate or managed rate for the period. More
than 60 percent of our total observations are for countries with fixed
or managed rates. Even though countries are increasingly using more
flexible rates, by the 1990s only 20 percent of our observations were
for country-years with pure floating rates, and more than 60 percent
had some form of fixed or managed exchange rates.
Interestingly, there is practically no correlation (r = 0.012, not
significant at .10) between these two forms of crisis, yet both are
cited as reasons for economic reform. We include each of these in
lagged form.
A second factor seen as responsible for trade liberalization in the
LDCs is external pressure—from the various international
financial institutions (IFIs) that provide funds to LDCs (such as the
IMF or World Bank), or from the world's hegemon, the United
States. In particular, the claim is that around periods of economic
crisis, LDCs are especially vulnerable to external pressure, and that
in exchange for loans or aid, countries have been forced to liberalize
their trade regimes—so-called conditionality. The counterfactual
is that LDCs would never have made these changes without overwhelming
external pressure. To control for these external forces, we include a
variable indicating whether the country has just signed an IMF
agreement to help bail it out of a crisis. Such a signing should
represent a period of high external pressure as the country is claiming
an inability to fund its own needs; IMF loans are intended to provide
help for countries experiencing severe balance-of-payments or reserves
crises. imf created by Przeworski and Vreeland is equal to 1
if an IMF agreement has been signed in that year; it is lagged in the
models.89
Przeworski
and Vreeland 2000; they also have another variable called
under, which shows whether a country is subject to an IMF
agreement that year. We use both, although we think imf is a
superior measure of the magnitude of external pressure. In addition, we
also look at the amount of total foreign aid that a country receives as
a percent of its central government budget (aid). This
measures the dependence of the government on foreign sources of
(nonprivate) capital.
More external pressure of any sort
should be related to lower trade barriers.
90 We include several variables designed to capture the impact of other
external pressures. A measure of U.S. hegemony captures the widespread
claim that American power is responsible for economic reform. Given our
interest in trade, hegemony (us heg) is measured as the sum of
U.S. exports and imports as a percent of world trade.91
If Hegemonic Stability Theory
is correct, greater U.S. influence should induce trade
liberalization.
92 We also examine the impact of the GATT/WTO. Joining the
GATT/WTO should induce countries to lower their trade barriers.
Recent work by Rose, however, suggests that it might have no impact;
moreover, given the exceptions that LDCs were allowed in the GATT
regime, it may be that GATT membership had a negative impact on
them.93
Recent research suggests that the Generalized System of Preferences
(GSP) in the GATT may also have induced developing country members to
maintain higher trade barriers than otherwise.
94 We include a lagged variable (
gatt)
indicating whether a country is in the GATT/WTO (=1) or not.
Finally, some scholars argue that competitive pressures among states
may drive policy changes around the world.95
In order to control for this, we create a
variable that indicates the average tariff level for all LDCs in that
year (
av tariff) and the average level of openness (according
to Sachs and Warner) for all countries in that year (
av open).
We use the lagged version of these to test for strategic competition
among LDCs.
A third factor involves the ideas that leaders possess about the best
policies available to them. Many claim that the turnabout in trade
policy was caused by a change in the ideas that leaders held about the
policies that would best promote economic development. Whether these
new ideas resulted from policy failure or external pressure is debated.
But the claim is that in their search for better development
strategies, leaders decided that an open trade regime was preferred to
the ISI one. It is difficult to find measures of such ideational
change, and the measures we use are not ideal. One measure we employ is
the number of years a government has been in office (office).
A new government might indicate a change in leadership and hence a
change in ideas. Others have examined whether governments in their
first year of office are more likely to reform; we considered this
(first = 1 if office < 2; 0 otherwise), but it was
never significant in the regressions.96
Finding measures of how policy ideas have changed throughout the
world is difficult. Quinn has developed an indicator of changing global
ideas about economic policy.97
He uses a measure of the
degree to which the top five advanced industrial countries have opened
their capital markets (
five open) to suggest how, globally,
changes in ideas about the ideal set of foreign economic policies are
evolving. Increases in this measure indicate that anticapitalist
sentiment is waning worldwide. This variable captures both changes in
ideas about optimal policies and the potential contagion of those ideas
from powerful developed states to the LDCs. This measure has many
problems, and if it is not related to our outcomes it can hardly be
seen as a fair test of the ideas claims. We include it in some models
to try to control for all of the main alternative explanations for
trade liberalization in the LDCs.
The basic equation estimating the relationship between democracy and
trade policy is:
We also include us heg, gatt, and five
open in some equations to check for robustness. TSCS data have
numerous problems that violate the standard assumptions necessary for
ordinary least squares (OLS) to be unbiased and efficient. We try to
correct for these in the standard ways. We use panel-corrected standard
errors to mitigate problems caused by various forms of
heteroskedasticity, as recommended by Beck and Katz.98
We
include country fixed effects and a time trend or decade fixed effects
to deal with problems of omitted variable bias. The use of a time trend
allows us to address concerns about whether the relationship between
democracy and trade policy is solely related to their both trending in
one direction over time.
The use of country fixed effects is particularly interesting in this
model.99
In all the regressions, a
Hausman test rejects the suitability of using random effects, not
surprisingly.
The fixed effects—or
“within”—estimator exploits the time-series component
of the data around the country averages. The within estimator examines
variation over time and thus addresses the question of the impact on
trade policy of a change in regime type within a country over time. We
address problems of serial correlation by using an AR1 correction. In
the logistic model, we estimate a natural spline function with three
knots; we use the count variable and three splines generated by this
procedure to handle temporal dependence, as recommended by Beck, Katz,
and Tucker.
100Regime Type
In almost all of the regressions, regime type is correctly signed and
significant. The regressions on tariff rates are the most direct test
of our argument. As Tables 2 and 3 show, more democratic regimes tend to have lower
tariff rates. Setting all the other variables at their means in
equation (3) in Table 3, a one standard
deviation increase in democracy from its mean leads to a 12.2 percent
decrease in tariff rates. Tariff rates drop from about 19 percent to
about 16.5 percent. Moving from an absolute autocracy (−10) to a
perfect democracy (10) induces a 31 percent decline in tariffs. Rates
fall from close to 22 percent to about 15 percent. These results are
robust to a wide variety of controls.
Table 3 also shows that using other
measures of regime type does not undermine our result. The
dictator variable is negative and quite significant,
indicating again that systems with larger selectorates tend to have
lower trade barriers. An increase in dictator by one standard
deviation from its mean leads to roughly a 17 percent reduction of
tariff rates, from about 16 percent to roughly 13 percent. The
dichotomous dem variable is also negative, but it is not
significant at conventional levels.
When we include the regime variable from Polity with dummy
variables for the three major types of autocracies, the type of
autocracy matters, even when controlling for overall regime type. More
democratic countries still tend to have lower barriers. But compared to
personalistic regimes (which is the excluded category), single-party
ones (sgl pty) are much less prone to protectionism. This
finding also supports our argument because selectorates are larger in
single-party autocracies than in personalistic ones. This suggests that
variations within both democracies and autocracies may help explain the
choice of trade policy.101
We also
turned Polity's regime variable into a dichotomous
variable with countries scoring below 6 as autocracies (= 0) and those
at or above 6 as democracies (= 1). Using equation (3) in Table 3, we found that once again, the regime
coefficient is negative and very significant; more democratic countries
have lower trade barriers.
To assess the robustness of our results, we address concerns about
multicolinearity. (The use of panel-corrected standard errors helps
assuage problems with various types of heteroskedasticity; and the
inclusion of country fixed effects, decade fixed effects, and a time
trend should address concerns about omitted variable biases.) In Table 4, we lag the regime variable by two and then
three periods. This did not affect our results greatly. The regime
variable either by itself or jointly with all of the lags was correctly
signed and significant for all the tariff regressions. We also lagged
the main alternatives to our argument: the crisis variables (ec
crisis and bp crisis) and an external pressure variable
(imf) for two and three periods for both dependent variables.
They were never significant and did not affect the regime
variable. This suggests that multicolinearity (at least among the main
alternative hypotheses) is not driving these results. Moreover, it
suggests that the effect of regime on trade policy is fairly
long lasting.102
We also dropped the
outliers from equation (3) in Table 3 and
reran the regressions, which did not change the regime
variable's sign or significance. This entailed dropping
observations whose residuals were more than 2, 3, or 5 standard
deviations from the mean.
Table 5 shows that increasing democracy
also increases the likelihood of a country opening its trade regime. A
one-unit increase in democracy in equation (1) raises the probability
of a change to openness by 0.523. Equivalently, holding all other
variables constant, each additional unit of increase in democracy
multiplies the log odds of openness occurring by 1.69. Using
alternative measures of regime type yields similar results. As above,
we created a dichotomous version of Polity's regime
score, with countries at or above 6 scoring as democracy. This variable
is also positive and significantly related to trade liberalization.
When we lag the regime variable, it remains positively related to trade
liberalization, but it is only significant when considered jointly with
all three lags. Entering the lags of the crisis and external pressure
variables does not change the results materially either.
Sachs-Warner trade liberalization
As Table 6 shows, the dictator
variable is also positive and statistically significantly related to
trade policy liberalization. The same is true for the dichotomous
dem variable. More-democratic countries are more likely to
liberalize their trade policies. As before, we also get interesting
results when using both the Polity measure and the dummy variables for
autocratic regime types. The regime measures are all jointly
significant (p > .05) with regime still having a
positive relationship. The broad Sachs-Warner measure of trade policy
yields similar results to the narrower tariff measure. For the two
different measures of trade policy, movement toward democracy is
positively associated with more openness, even when we account for many
other influences.
Sachs-Warner trade liberalization
To further test the robustness of our results, we also address
concerns about endogeneity. To explore whether trade policy itself
promotes regime change, we regressed all the independent variables in
equation (1) of Table 3—including the
two trade policy measures lagged (each independently) and excluding the
average level of tariffs in the world—on our measure of regime
type. None of the variables, except GDP per capita, the time trend, and
the log of population, was near statistical significance, implying
again that colinearity is not a major problem. Most importantly, the
trade policy measures never reached conventional levels of
significance, suggesting that they are not causing regime change.
In addition, several tests examined endogeneity problems with the
regime variable. First, following Wooldridge, we took the residuals
from two fixed effects regressions of our independent variables on
regime and included them in regressions identical to those in
equation (3) in Table 3 and equation (1) in
Table 5.103
We then checked whether the coefficients on the
residuals were significant; they were not at the .10 level. This test
indicates that endogeneity is not acute.
Another way to deal with endogeneity is to instrument for the
variable in question. Finding useful instruments for regime type is not
a trivial matter. In our case, they must be measures that predict
regime type well and are not related to either trade policy or the
errors. We used two instruments for regime: the average age of the
party system in a country year (ptyage variable from Beck et
al.) and the level of secondary school completion among the population
over fifteen years (sschool variable from Barro and Lee).104
We expect both variables to be positively related to democracy. Using
these two instruments, we estimate the impact of regime type on
statutory tariff rates in
Table 7. The regime
variable remains negative and quite significant.
105 A test devised by Davidson and
MacKinnon 1993, which is similar to the (Durbin-Wu-) Hausman
test, reveals that the null hypothesis that an OLS version of the same
equation would be consistent can be rejected at the .01 level,
suggesting that the instrumental variables are useful. A Sargan-like
test for overidentification does not allow rejection of the null
hypothesis that the instruments are useful and uncorrelated with the
error term at the .05 level.
These results give some
confidence that even correcting for possible endogeneity, regime type
still affects trade policy.
Instrumental variable regression: Tariff rates
The Political Control Variables
Our first set of control variables explored the impact of economic
crisis on a country's decision to liberalize. We included two
distinct variables, ec crisis and bp crisis, to
capture the pressures from such crises. These variables do not seem to
matter significantly, either independently or jointly. Countries may
respond differently to crises, sometimes raising trade barriers and
other times lowering them. These results are not unexpected. Many
claims about the role of crises come from case studies that often
select on the dependent variable; that is, they explore cases of
economic reform and often find that crises (of varying types) existed
as well. But they rarely look at the cases where countries experienced
crises and did not reform.106
A simple
cross-tabulation of our data shows that in 3 percent of the cases,
liberalization according to Sachs and Warner occurred just after a bout
of economic crisis; in 3 percent of cases, such crises occurred with no
trade liberalization; and in 30 percent of them liberalization occurred
with no crisis beforehand. For balance-of-payments crises, 40 percent
of the cases had crises but no trade policy reform, compared to 16
percent with crises preceding the reforms.
As noted earlier,
the debt crisis of the early 1980s did not lead to trade
liberalization, but to greater closure. The effects of crises may be
highly contingent on the environment. Sorting through the myriad types
of economic crises and their various political effects is an important
area for future research.
The second set of external political factors involves external
pressures. The main way we measured this was to look at whether
countries had signed an agreement with the IMF in the previous period
(we explored up to three lags). This variable (imf) was never
significant. Increases in external pressure via the IMF have little
measurable effect on trade policy, and when they do (in the regressions
on Sachs-Warner openness), their effect is to increase the closure of
the economy, not to open it. We also tried using whether a country was
under any IMF agreement (in one to three previous periods), and this
was never significant either. These results may not be that surprising
given the findings of Przeworski and Vreeland and Vreeland, among
others, which show that countries do not receive IMF loans when they
are most vulnerable and in crisis, and that conditionality rarely seems
to have much impact on policy.107
We also employed a variable
measuring the amount of foreign aid a country received relative to its
government budget. While this measure has many fewer observations, it
was not significant either. We concur with Weyland, who argues that
“economic-structural arguments alone cannot provide a complete
explanation for the enactment of market reform….
‘Markets’ and ‘leverage’ did not determine
governmental decisions; political leaders retained a margin of
choice.”
108 Other measures of external influence show greater impact. Whether a
country is in the GATT/WTO has a significant influence on both
tariff rates and on Sachs-Warner openness. But the relationship is
surprising. Being a member keeps one's tariff rates higher than
otherwise and lowers the probability that a country opens its trading
regime. This finding is consistent with recent research by Rose, which
shows that being a member of the GATT/WTO does not increase a
country's trade flows, nor does it lead them to reduce their trade
barriers.109
Many countries, it seems, choose to reduce
their barriers before they enter the GATT or even if they have no plans
for entering it. Furthermore, countries that are already in the
international organization are much less likely to become more open,
ceteris paribus. In part, this results from the myriad exceptions in
GATT rules (some tightened by the WTO) to reducing barriers. Among
these exceptions, the GSP scheme was very important for the LDCs; it
allowed them to maintain their barriers while still gaining access to
developed countries' markets. Moreover, it is very consistent with
Ozden and Reinhardt, who show that the GSP system in the GATT
encouraged many LDCs to maintain higher trade barriers than
otherwise.
110 In part, this finding may reflect the fact
that many of these countries decided to lower their trade barriers
unilaterally before joining the GATT/WTO. Once in the organization,
further liberalization might be quite slow given the glacial pace of
recent multilateral trade negotiations.
Our final set of political controls looked at the change in ideas
around the globe. Our measures are weak indicators for this factor;
hence, our results should be interpreted cautiously. We examined
governments' tenure in office to see whether new governments were
more likely to undertake trade policy reform. This variable was only
significant in the case of tariff rates, and here the findings were
surprising. The longer governments had been in office, the lower were
their tariff rates, ceteris paribus. Trade policy reform may require
political stability and a government with a firm hold on power.111
A variable for whether the government was in its
first year or not was never significant, suggesting that in general,
new governments either did not have new ideas about trade policy or
could not implement them.
112 Our measure
of the global spread of procapitalist ideas, five open, has
weak and inconsistent results. This measure is a poor substitute for
more direct evidence about the global spread of ideas, but few
alternatives exist.
Existing data do not allow us to conclude
much about the role of ideas in the change in trade policy. We have
tried to control for them simply to give greater credibility to our
claim about democracy. But our research should not be interpreted as
concluding that the spread of new ideas about trade policy did not
matter.
We looked at several other, less obvious factors that might affect
economic reform. We included a variable for a government's
relative political capacity.113
This measures
a government's ability to extract resources from its society. One
might expect a more capable government to need to use trade taxes less,
and hence be more likely to liberalize. It was not significant in
regressions on either dependent variable—equation (3) in
Table 3 and equation (1) in
Table 5—and did not affect the sign or
significance of the
regime variable. We also included a
variable indicating whether a country was involved in a war that year
or previously.
114 war was coded
from the latest Correlates of War data set and includes all three types
of war.
War is expected to make protectionism more likely and
to reduce the chances of trade liberalization. Its effect on tariff
rates was positive but not near conventional levels of significance.
Its effect on liberalization was negative, as expected, and significant
in some cases. It did not alter the sign or significance of the
regime variable, however. In addition, we employed a variable
intended to show the similarity between the foreign policy interests of
an LDC and the United States by measuring the overlap in UN voting.
115 Such similarity should be associated with a more
open economy and hence with lower trade barriers. This variable had no
statistically significant impact on Sachs-Warner liberalization, but it
sometimes had a modestly negative impact on tariff rates. It did not
materially affect the
regime variable. Finally, we added data
on inequality.
116 These data are
few and of low quality (some are imputed), so results should be
regarded with low confidence. But for neither dependent variable was
inequality near conventional levels of significance, nor did it affect
the sign or significance of the
regime variable.
The Economic Factors
We included several obvious economic controls for trade policy. Our
variable measuring country size, logpop, is always
significant, but not as expected. Big countries tend to have higher
tariff rates, as often suggested, but they tend to be more likely to
liberalize, ceteris paribus. The level of economic development (gdp
pc) seems to matter at times. But among the LDCs, more-developed
countries have more restrictive trade regimes, ceteris paribus.
Finally, we looked at a country's flows of direct foreign
investment as a percentage of gross national product (GNP), which is
represented by variable fdi.117
For
the
tariff variable, these flows were always positive and
significant, indicating that foreign investment flows earlier may have
built up a constituency for continued protection of the host market.
We looked at a number of other economic factors that might have some
relationship to trade policy. First, we controlled for countries with
heavy dependence on oil and fuel exports. Unfortunately, given the use
of country fixed effects, and given that the oil producers change
little over time, we were unable to address this point. Second, we
included a measure of a country's exchange-rate regime, measured
along a continuum of fixed to floating.118
This variable had no statistically
significant relationship to tariff rates, and a mildly negative impact
on trade liberalization à la Sachs-Warner; its inclusion did
not affect the sign or significance of
regime. Finally, we also
looked at (lagged) yearly changes in GNP per capita, GDP, exchange rates,
and inflation.
119 These variables had
no statistically significant relationship to tariff rates, sometimes a
positive impact on the Sachs-Warner measure of trade liberalization, and
no effect on the sign or significance of
regime.
In sum, these results support our claim that democratization is one
force that leads to a more open trade regime. This finding is robust to
three different measures of regime type; it also withstands a variety
of robustness checks. Moreover, this influence was never negligible,
even when controlling for many alternative explanations. Conventional
wisdom about economic reform depending on crises and external pressures
is not supported by this study, while the impact of regime change
appears more important than previously thought.
Conclusion
Why countries that long pursued protectionism should suddenly
liberalize their trade regimes is an important and underexplored
question. While economists have long preached the benefits of free
trade, developing countries have only recently begun to heed their
advice. Indeed, much of the extant literature argues that economic
reforms, such as trade liberalization, rarely occur. However, many
developing countries began liberalizing trade in the mid-1980s, and the
move to free trade since has been remarkable.
Our argument is that a change in the political regime toward more
democracy should be followed by a move to liberalize trade. Autocratic
political leaders in LDCs can cater to the capital-rich segment of the
population because the “selectorate” that picks them is
limited. Trade barriers are then imposed on capital-intensive imports
so that wealth is redistributed from those who are not part of the
selectorate to those who are. Democratization, which implies an
increase in the selectorate's size, changes the calculations of
political leaders about the optimal level of trade barriers; it induces
the adoption of trade policies that better promote the welfare of
consumers/voters at large, which implies trade liberalization in
this context. While protectionist interest groups remain important in
developing democracies, other groups preferring lower trade barriers
become more important for political leaders because they are now part
of the selectorate upon which leaders depend for their political
survival.
We think that future research should try to disaggregate regime type
further. Our data suggest that autocracies may vary in the likelihood
of choosing economic reform, with single-party and military-controlled
systems being more likely than personalistic ones. The two former types
of regimes rely on a broader selectorate and are not as able to use
protectionism to garner political support. The likelihood of reform may
also depend on the type of democratic institutions in place. Examining
the impact of different political institutions on trade policy is an
unexplored area of great potential interest.
We view democratization as exogenous. However, trade policy could
exert an impact on political regimes. Although we lag all of our
independent variables (from one to three periods) and include tests for
endogeneity, this could be a much longer-term effect. Most models
predicting regime type, however, do not include trade policy or even
the extent of openness of the economy as a predictor.120
Moreover, we found no evidence
of such an impact in our data; trade policy did not predict democracy.
Even after instrumenting for democracy, regime type still played an
important role in explaining the move to free trade. Democracies choose
lower levels of trade barriers, even when holding many other factors
constant.
Our results cast some doubt on the leading alternative theories of
trade policy reform. Although much discussion exists of the role of
economic crises, external pressures, and the role of ideas on economic
reform, little systematic research has been done. To the extent that
our measures adequately control for these factors, they did not seem to
play a consistent role in explaining trade policy. Neither crises, nor
international pressures, nor new leaders seem to account very well for
the move to free trade. We concur with Jenkins, who points out that
“the existence of a crisis is no guarantee that a government will
respond, and more importantly, that it will be successful in convincing
interest groups that ‘something must be done.’”121
Moreover, international institutions that were supposed to
foster trade liberalization, such as the GATT/WTO and the IMF, do
not appear to be playing that role. As Rose has argued, the GATT does
not seem to promote a more liberal trade policy for most countries; and
as Ozden and Reinhardt show, the GSP in the WTO has slowed down
liberalization in LDCs.
122 Our research
certainly does not rule out any of these factors, especially the spread
of neoliberal ideas. Our measures of the rise of new ideas are very
crude, and better research into this topic requires more and different
data.
Our argument does not explain all cases or all pressures for
liberalization. No single variable can possibly account for the
dramatic change in economic policies in the LDCs during the past twenty
years. Changes in domestic political institutions, however, have been
an underappreciated factor. Hence we highlight their role.
Additionally, we cannot explain all countries. India, for instance,
remains a puzzle; long a democracy, the government has only recently
chosen to lower trade barriers. Although a large number of cases seem
to fit our claim, as discussed in the first section of this article, no
single variable can possibly account for this move to free trade in all
countries.
In general, more-democratic countries are more willing to open their
markets to the international economy, even when holding many other
factors constant. Democratization thus may have promoted the
globalization of the past two decades. As we show vis-à-vis
trade, and as Quinn shows relative to capital controls, democracies in
the late twentieth century may have been more likely to join the global
economy by eliminating the barriers protecting their markets.123
Democratization may have fostered the increasing
globalization of the past two decades. Whether the new democracies in
the developing world will survive and thrive in a globalized world is a
separate issue that should command future research.
