Show Summary Details

Page of

Printed from Oxford Research Encyclopedias, Economics and Finance. Under the terms of the licence agreement, an individual user may print out a single article for personal use (for details see Privacy Policy and Legal Notice).

date: 26 January 2022

Political Economy of Protectionfree

Political Economy of Protectionfree

  • Xenia MatschkeXenia MatschkeEconomics, University of Trier; CESifo Group Munich


The political economy of protection is a field within economics, but it has significant overlap with its sister discipline, political science. For a political economy of protection, one needs at a minimum two types of economic agents: political decision makers who provide protection, and economic agents who are protected or even actively seek protection. The typical political economy scenario leads to an economic outcome that is not Pareto-optimal: From a general welfare perspective, the political interaction is not desirable. An important task of political economy research is to explain why and how political interaction takes place. For the first part of the question, it appears clear that if protection is actively sought, the protection seeker intends to benefit from his activities. However, if the policymakers were truly interested in Pareto optimality and welfare maximization, they would refuse to protect. Hence a crucial assumption in the political economy literature is that the politicians’ objective function differs from the general welfare function. For the second part of the question, theoretical political economy models consider either the election campaign phase when politicians are eager to win a majority of votes (preelection models) or the phase when the politicians have been elected and may benefit from the spoils associated with holding office (postelection models). Whereas in the election phase, politicians have an incentive to cater to the interests of that part of the electorate that is considered pivotal for the election outcome, in the postelection phase they may be open to, for example, special interest group (SIG) influences from which they derive utility.

A first wave of theoretical political economy models originates from the 1980s. Building on these early advances, more elaborate models have been proposed. The most prominent one is the Grossman–Helpman protection for sale (PfS) model. It delivers a postelection general equilibrium framework of trade policy determination. In this common agency model, industry interest groups act as principals and offer the government a menu of contracts of campaign contributions in exchange for trade policy. The PfS model predicts that industries that lobby for protection will obtain trade protection in equilibrium, whereas nonlobbying industries will face import subsidies. Numerous papers have evaluated the PfS model empirically and found that the implied weight on contributions in the governmental welfare function and the implied share of the population represented by lobbies are both very high. Remedies for this surprising result exist, but it has also been argued that the found empirical regularities may be spurious.

At the beginning of the 21st century, the majority of political economy literature is still theoretical, but better data availability increasingly offers the opportunity to empirically test theoretical results. A number of challenges remain for the political economy literature, however. In particular, more work is required to better understand policymaker interests. Moreover, an incorporation of political economy aspects into the new trade theory models that allow for intra-industry trade and firm diversity appears to be a promising avenue for future research.


  • History of Economic Thought
  • International Economics
  • Public Economics and Policy

Political Economy and Political Economy Models in Modern Economics

Political economy, which merges ideas from economics and political science, is an interdisciplinary field in which standard mathematically founded economic theory, where economic agents optimize under constraints, is combined with a very active role for politicians and governments that maximize an objective function that is typically not the social welfare function. Moreover, nonpoliticians influence the political decision process, often competing against each other in the process. Accordingly, Weingast and Wittman (2006) defined political economy as “the methodology of economics applied to the analysis of political behavior and institutions” (p. 3).

Classification of Political Economy Models

Today, two general types of political economy models are commonly used: pre- and postelection models.1 They come in many facets, and only the simple baseline models that have been extensively used in the analysis of protection are discussed here.

(Pre)Election Models

(Pre)election models consider the election phase in democracies. Politicians need to maximize their vote share in order to maximize their chance of winning elections, and they need to win the election in order to be able to choose policies. In the two early baseline models in this class of models, it is assumed either that the politicians have their own preferences over the available policy choices or that they are indifferent about the proposed policies but commit to a certain policy to attract a majority of votes.

In the direct democracy model described by Black (1948, 1958), the decision-finding process within a committee is investigated. Any committee member has preferences over a one-dimensional policy issue and can bring forward a motion, suggesting a policy that is best from their perspective. In an election, that candidate wins whose policy platform is preferred to the competitors’ platforms by a majority of voters.

In opportunistic politician models (Downs, 1957), in contrast, the politicians do not have preferences over policies but only care about vote shares.2 The sole objective of opportunistic, office-seeking politicians consists of maximizing their vote share to win the election, and they choose their policy platform accordingly.

It turns out, however, that the outcome of either type of (pre)election model is similar in that the wishes of the median voter will determine the winning policy platform. The median voter theorem, first described in detail by Black (1948), holds when voter preferences are single peaked and the election is over one policy issue.3 It says that the median voter’s preferred policy is a Condorcet winner, that is, the policy outcome at the center of the preferred policy distribution will win against any other alternative policy in a pairwise majority vote. How exactly the other voters’ preferred policies look is immaterial for this result.

The median voter theorem has been criticized on multiple fronts. In the Downsian representative democracy framework, it only holds in the two-party case. With as few as three parties, all choosing the median voter’s preferred policy will no longer constitute a Nash equilibrium. Moreover, the median voter theorem falls apart when dropping the assumption that the vote is over a one-dimensional policy issue. It is also difficult to explain why rational voters, in view of the costs of voting, would vote if the voting outcome were already predetermined. Because of its analytical simplicity and its clear predictions, however, the median voter scenario is still widely used, although its relevance for real-world policy choices appears questionable, and more elaborate and realistic models, such as the probabilistic voting model (Enelow & Hinich, 1982; Hinich et al., 1973; Lindbeck & Weibull, 1987), have been developed.

Postelection Models

In postelection models, politicians have been elected to office and use their position for their own interests, typically not bound by any earlier political promises. In this situation, the politicians are susceptible to the influence of powerful special interest groups (SIGs), also called lobby groups or lobbies, who can offer rewards, for example monetary contributions, in exchange for political favors. “The potential uses of public resources and powers to improve the economic status” of lobby groups create a demand for regulation, just as the “characteristics of the political process” explain its supply (Stigler, 1971, p. 3). Various reasons exist why lobby contributions could increase the welfare of politicians: They could be used as income for additional consumption expenditure, but they could also improve reelection chances because, for example, contributions can be used to finance campaigning. This latter aspect provides a natural link between pre- and postelection models.

Forming a lobby group is not a trivial task. Its success depends on the size of the benefits to be gained from acting collectively, on the one hand, and the costs of organization and of precluding free riding, on the other hand (Olson, 1965). The extent to which a lobby can influence the political process is dependent on the politicians’ openness to special interests and the extent to which other lobby groups compete for influence. From the lobby’s perspective, the existence of competing groups reduces its welfare. From the society’s perspective, lobby competition is a double-edged sword. It reduces the actual policy distortion as the competing lobby influences partly offset each other (Becker, 1983). But it is also true (Krueger, 1974; Tullock, 1980) that rent-seeking lobby competition causes an additional social cost because lobbies waste resources to capture regulation rents (rent dissipation).

Protection in a Political Economy Framework

What Is Protection?

Protection is the act of keeping harm away from somebody or something (Latin protegere: to shield, cover, or protect). As such, it carries a positive and intrinsically nonaggressive connotation in standard language. This contrasts with the way protection appears in economic models, where it is often actively sought by powerful lobby groups who exploit political structures to have protection established to their advantage, causing a deadweight loss for society and thus a Pareto-inferior outcome.

Agents Involved

At a minimum, a political economy model of protection contains an economic agent, for example a firm or an industry that demands protection, and a policymaker who supplies protection. In accordance with this view, Rodrik (1995, p. 1458) stated that any political economy model consists of four elements: individual preferences over policy choices, a description of how preferences are channeled to create a political demand for protection, policymaker preferences, and an institutional setting describing the political supply side of protection. The explicit modeling of the policymakers’ side of the model, what goals they pursue and how other economic agents can influence them, is at the core of political economy models of protection, sometimes also called endogenous models of protection because they focus the process on how protection arises in equilibrium.

The idea that individuals act to further their self-interest is a key assumption in economic models. It thus appears natural to assume the same about politicians, as pointed out by Buchanan and Tullock (1962) in their landmark book The Calculus of Consent. Individual interests take center stage in their theory of public choice. Because politicians have their own objectives that may not coincide with those of society, constitutional restrictions need to be placed on the use of the political process, to prevent its abuse for exploitative purposes (Buchanan & Tullock, 1962, p. 13). A second important feature in Buchanan and Tullock’s book is the idea of the political process as a kind of exchange when voting is over multidimensional issues (Van den Hauwe, 1999). Logrolling, that is, exchanging votes or favors, as a method of equalizing the costs and benefits of a decision opens the possibility for a broad consensus and can improve welfare compared to the outcome of a majority vote.

The welfare functions of politicians and other economic agents are commonly assumed to differ. Politicians in a political economy model either maximize their vote share in the preelection period, or they maximize some weighted welfare function in which spoils from holding political office are included in the objective function. Some older political economy models also assume specific influence functions, with economic agent activities as input and policies as output.

Trade Protection

Protection typically results in overall welfare losses but benefits population subgroups by redistributing income toward them. However, not every redistribution can be considered protection in a strict sense, as protection implies that an economic agent is safeguarded from a harmful outer influence. A direct subsidy to domestic producers redistributes income but is not, strictly speaking, protection, in contrast to an import tariff directed against the external threat of imports from abroad.4 The literature on political economy of protection reflects this fact in that most of it deals with the political economy of trade protection, notably in the form of import tariffs.

Trade taxes, such as import tariffs, are a relatively costly means of redistributing income to domestic producers: Production subsidies as direct measure to boost producer surplus can create the same income for producers as the more indirect import tariff; but the latter also depresses consumer surplus. Rodrik (1986) pointed out, however, that this seeming disadvantage of trade protection can also be an advantage, because the costliness of redistribution in a model with an endogenous degree of income distribution implies that the equilibrium redistribution will be lower. Moreover, more efficient income transfers may intensify the competition between lobbies for protection, such that any advantages from more direct transfers may be offset by the additional cost of lobbying (Grossman & Helpman, 1994). Consequently, lobbies would prefer a more inefficient way to transfer income, to reduce the degree of competition.

Indirect measures may also be more opaque than direct measures. The principle of optimal obfuscation suggested by Magee et al. (1989) stated that the negative effects of trade protection may be less obvious than the effects of more direct redistributive measures. But whether the tax rate is slightly higher to pay for direct production subsidies or whether the consumer prices are slightly higher because of import protection probably will both go largely unnoticed in practice, and in a complete information framework, optimal obfuscation is impossible.

The impact of the positive connotation of “protection” should not be underestimated, either. Protecting the weak, in particular against unfair external pressure, is viewed positively by society. Baldwin (1985, 1989) argued that any theory trying to explain the emergence of protectionism should not solely build on the narrow self-interests of politicians but also include broad social concerns as explanation. Status quo protection appears to be a powerful trade policy goal, and already early on, economists have tried to model this status quo bias: Corden (1974) formulated a conservative social welfare function where group welfare losses enter with a higher weight than group welfare gains. In the General Agreement on Tariffs and Trade–World Trade Organization (GATT-WTO) framework, antidumping duties, countervailing duties, and safeguard tariffs are all special tariffs designed to protect industries in distress. However, whereas anti-dumping and countervailing duties allegedly protect from unfair competition (protection against outward “aggression” in the form of unfair foreign practices such as dumping of imports or payment of subsidies to foreign competitors (Art. VI GATT 1994), safeguard tariffs (Art. XIX GATT plus Safeguard Agreement) are used as protection against sudden import surges. A count of measures for the years 1996 to 2014 reveals that the combined sum of antidumping and countervailing duty applications to 6-digit HS sectors clearly exceeds the number of safeguard tariff applications. For the United States, the contrast is especially stark (Kuenzel, 2020).

Political Economy of Trade Protection Models

The political economy of trade protection models can be classified into models predating Grossman and Helpman’s influential protection for sale model, those directly building on it and those later models that depart from its basic assumptions.

Early Theoretical Models in the 1980s

The early postelection models often employ ad hoc governmental welfare functions that take lobby contributions or effort as arguments and generate trade policy outcomes as functional value. These policy or influence functions are not derived from first principles of optimization and are thus subject to a black box criticism (Ethier, 2010). A selection of postelection models, based on the survey article by Rodrik (1995), provides a flavor of these early models.

Findlay and Wellisz (1982) presented a model with two industry lobby groups i = 1,2, where group 1 is in favor of tariff protection and group 2 opposes it. These groups exert effort Lito influence the tariff. The government chooses the tariff according to a tariff formation function t(L1,L2), which is increasing in L1 and decreasing in L2. The resulting tariff is the outcome of a Nash equilibrium in effort choices where neither SIG can improve its welfare unilaterally by changing its lobby activity. This tariff formation approach has been criticized as lacking micro-foundations: The tariff formation function is simply assumed and not derived from optimization, and it remains unclear what the government’s objectives are.

The political support function model is due to Hillman (1989, chap. 2). The government maximizes a political support function gπPπPPP, where P denotes the domestic price and Pthe world market price. The political support function is increasing in its first argument and decreasing in the second, where πPπP denotes the change in the protected industry’s profit and PP the specific tariff, written as a wedge between domestic and world market price. The political support thus positively depends on the industry gains from protection, and negatively on the welfare losses incurred by the general population. The government maximizes this political support function by choice of the tariff, but the function itself is not micro-founded.

Another, more comprehensive political economy of protection model is the campaign contributions approach by Magee et al. (1989). This model considers two parties, one associated with capital interests and the other with labor interests. The parties accept campaign contributions from their respective lobby group. Contributions increase election chances, but caving in to lobby demands reduces election chances. In equilibrium, a party will choose its policy in such a way that at the margin, an infinitesimal increase in protection leads to a zero net effect on the election probability as marginal costs and benefits cancel each other. The Magee et al. model has been severely criticized in a book review by Austen-Smith (1991). One major criticism regards the highly restrictive strategy spaces for the political parties (Austen-Smith, 1991, p. 74). The protectionist (pro-labor) party can only choose an import tariff, and the pro-export (pro-capital) party can only choose an export subsidy. Austen-Smith (1991, pp. 77ff., p. 88) also contended that the Magee et al. model informally introduces a probabilistic voting framework without any apparent rational choice foundation. Compared to other models in the 1980s, however, it is appealing because it combines the pre- and post-election perspectives.

(Pre)Election Models of Trade Protection

Mayer (1984) adapted the Downs model to investigate trade protection choice. Each voter owns capital and labor, but in different proportions. A voter’s favorite trade policy depends on their factor mix. Moving from autarky to free trade, the relative price of that good will increase whose production uses the domestically abundant factor intensively. This price change diminishes the demand for the domestically scarce factor, reducing its price, and raises the demand for the abundant factor, increasing its price (Stolper–Samuelson Theorem). In the Mayer model, the equilibrium tariff choice will reflect the relative factor endowments of the median voter who is typically a worker, assuming that capital ownership is relatively concentrated. Hence, in a capital-abundant country, an import tariff on the labor-intensive import good will result. In a labor-abundant country, the equilibrium policy would be an import subsidy for the capital-intensive import good.

Because, in reality, trade policies are typically trade restricting, this latter theoretical result is clearly at odds with empirical observation. However, Dutt and Mitra (2002) empirically tested another prediction of the Mayer model, namely, that holding the relative factor endowments fixed, an increase in inequality affects the equilibrium trade policy in labor- and capital-abundant countries differently. In a cross-country sample, they showed that an increase in inequality raises trade barriers in capital-abundant countries, whereas in labor-abundant countries, an increase in inequality is linked to a decrease in trade protection. In a related article, Dutt and Mitra (2005) found that left- and right-leaning governments pick different trade policies depending on their country’s factor abundance. Left-leaning governments adopt more protectionist trade policies in capital-abundant countries but choose a pro-trade policy in labor-abundant countries. Both studies thus provide some empirical support for the original Mayer (1984) trade protection model.

A dynamic median voter model in an overlapping generations model with equally sized generations and infinite time horizon was proposed by Blanchard and Willmann (2011). In their model, heterogeneous agents born with different ability levels, drawn from a continuous distribution, live for two periods. Agents are born unskilled and can immediately produce an unskilled good, the country’s import good, in the first period. Also in the first period, an agent may spend a fixed fraction of time on education to become skilled and can then produce a skilled good, the country’s export good, in the second period. Higher ability agents will, if educated, produce a greater quantity of the skilled good than lower ability agents of the same educational level. The time cost of education and worker productivity in the unskilled sector are independent of ability level. Thus, only agents above a certain threshold ability level will choose education.

Each period, the agents vote on whether to impose a low or a high import tariff. Any tariff reform becomes effective immediately. Because old workers are sectorally immobile, they unequivocally either favor the low (skilled workers) or the high (unskilled workers) tariff. The median voter will be a young voter, whose identity, as described by their ability level, will depend on the status quo (tariff before the vote) and the realized (tariff after the vote) contemporary trade regime through the older generation’s skill composition. The higher the share of unskilled old workers and thus the vested interest in a high tariff, the lower will be the median voter’s ability level, because fewer young voters are needed for a protectionist majority. In a political steady state, the skill composition of the population under the steady state tariff is constant. The steady-state trade policy can be always protectionist, always liberal, or a one-time policy transition can occur, with the equilibrium tariff at first high and then low. Whereas each generation would be better off in the liberal steady state, young voters, for a given future tariff, would benefit from protection today because they are unskilled, explaining the existence of a protectionist steady state: The median voter may choose to remain unskilled because a protectionist policy is expected in the future. By the same logic, however, it is also possible to transition from a protectionist to a liberal steady state. Young voters, expecting a future liberal tariff regime, may vote for a low tariff today because they plan to become educated and reap the educational rewards tomorrow. Such a move from a protectionist to a liberal steady state can be induced by paying educational subsidies, by trading partners’ trade liberalization, or by suggesting a bold rather than a timid trade policy reform.5 Tax-paid transfer payments to workers who lose from trade liberalization may not be helpful, however, because perpetual transfers lower the incentives to acquire education and thus inadvertently reduce the future support for trade liberalization.

Grossman and Helpman (2005) presented a model that combines a preelection scenario where party leaders choose trade policy platforms before the election with a subsequent policy game where the elected legislators, who represent different districts, form a coalition to implement a trade policy. The model has a very simple structure with two political parties, three geographic districts, three nonnumeraire goods, and a continuum of voters. Capital is industry and district specific. The industries are asymmetrically spread across the districts; hence the trade policy interests differ across districts. The political game consists of three stages. First, the parties choose their trade policy platforms to maximize their election probabilities. Under this assumption, they will both announce the same policy platform in equilibrium, as in the Downs model. In the second stage, elections are held. For a party to win, it must garner a majority of votes in at least two of the three districts. Once the election is over, the legislators of the victorious party implement a trade policy. In principle, they would want to choose the trade policy that maximizes the joint welfare of their districts. However, although the party cannot force its legislators to choose the promised trade policy, it can fine them for deviations from the announced policy platform. The corresponding party discipline parameter δ‎ lies between 0 (no party discipline, legislators choose their preferred trade policy) and infinity (complete party discipline, legislators choose the promised trade policy).

The equilibrium of the game typically exhibits a protectionist bias both with respect to the party platform (“policy rhetoric”) and the actual trade policy (“policy reality”). The parties promise free trade in the limit when δ‎ approaches infinity. Otherwise, for δ‎ > 0, free trade is only promised if either industry output is not increasing in its domestic price or if the industries are evenly distributed across districts. If the policy platform is free trade, this policy will be enacted by the legislators. Otherwise, the chosen trade policies, though typically not identical to the promised policy platform, also exhibit a protectionist bias. As an empirically testable hypothesis, Grossman and Helpman (2005) posited that among countries with majoritarian systems, the chosen trade policies in countries with stronger party discipline will be less protectionist. Moreover, they also conjectured that majoritarian systems may exhibit a protectionist bias compared to proportional systems, assuming that legislators in proportional systems are more likely to maximize national welfare. Empirical evidence for such a majoritarian bias across countries and over time was presented in Evans (2009) for tariffs and Rickard (2012) for subsidies. The overall evidence is far from conclusive, however (Rickard, 2015).

The preelection models of trade protection have been criticized because trade policy rarely constitutes a major issue in election campaigns, and the politicians’ trade policy stance thus appears unlikely to decide elections. Events in the 21st century, like Brexit, Trump’s trade wars, and popular campaigns against free trade agreements, however, show that trade policy can be turned into a general interest topic, in particular if a shift in trade flows affects the employment and income of large segments of the population (Autor et al., 2013; Feigenbaum & Hall, 2015). In the year 2000 U.S. congressional elections, the extension of permanent normal trade relations to China led to higher voter turnout and a higher Democratic vote share in counties more affected by this change (Che et al., 2016). A similar effect was recorded by Margalit (2011) for U.S. presidential elections: From his first election in 2000 to his reelection in 2004, George W. Bush’s vote share dropped more in counties that faced increased foreign competition, as measured by the number of job losses claimed in applications for Trade Adjustment Assistance, a U.S. national program meant to help workers laid off because of increased foreign competition. Autor et al. (2020) found that from 2000 to 2016, U.S. electoral districts facing higher trade exposure were more likely to elect a Republican representative and in general became more polarized. Majority white districts turned toward conservative Republican candidates, whereas majority minority districts chose more liberal Democratic candidates. Blanchard et al. (2019) and Fajgelbaum et al. (2020) both investigated the effects of former U.S. president Donald Trump’s trade war initiated in 2018. Fajgelbaum et al. (2020) were able to show that the newly introduced U.S. import tariffs favored sectors concentrated in politically competitive counties, a result in line with the median voter theorem, whereas the counties most negatively affected by foreign retaliatory tariffs were counties with a high Republican vote share. According to Blanchard et al. (2019), the retaliatory tariffs cost the Republican party seats in the 2018 midterm election. Early-21st-century political economy research thus provides ample evidence that trade and trade policy may feed back into the general populace’s political preferences.

Protection for Sale Model

In the postelection protection for sale (PfS) model (Grossman & Helpman, 1994), special interest groups pressure the government to provide trade protection in exchange for financial benefits.


PfS is a micro-founded general equilibrium (GE) model, but it is set up in such a way that it actually behaves like a partial equilibrium model, where each market can be considered separately. Domestic welfare is then calculated as the sum of producer surplus, consumer surplus, and governmental revenue in the nonnumeraire industries.

The model assumes a small (price-taking) country with n+1 industries, of which n nonnumeraire industries are of interest, whereas the numeraire industry helps simplify the analysis. The country owns fixed endowments of labor and capital, and neither international migration nor investment (foreign or domestic) are considered.

Labor is completely mobile across industries, but capital is sector specific, a key assumption of the specific factors trade model (also known as the Ricardo–Viner model, which complements the Heckscher–Ohlin model for the short to medium run time horizon). Markets are perfectly competitive; that is, firms and consumers act as price takers.

Whereas all workers earn a wage of 1 per labor unit, capital owners of any nonnumeraire industry receive the difference between revenue and wage payment; that is, they are residual claimants.

Some of the industries have formed SIGs to influence the government’s trade policy choice. A lobby group, if it could choose the trade policy vector itself, would choose an import tariff for its own product and an import subsidy for other products, because it consumes the latter. However, it would not choose extreme policies, for example a prohibitive import tariff, because it also receives part of the tariff revenue and pays part of the import subsidy.6

The government is susceptible to interest group pressure because it does not maximize domestic welfare, but rather a weighted sum WGof domestic welfare W and financial contributions C from lobby groups:7

(1) W G = aW + C = a W C + 1 + a C .

As the weight a on domestic welfare W approaches infinity, domestic welfare net of contributions and contributions receive equal weight.

The interaction between SIGs and government is set up as a common agency model where several principals (lobbies) try to influence a single agent (government) in its policy choice (Bernheim & Whinston, 1986). To this purpose, each lobby prepares a menu for the government, that is, a list with all possible tariff vectors and the contributions the lobby promises in exchange for a particular tariff vector. Once the government has chosen the tariff vector, its policy choice is implemented, and the promised contributions are collected.

The equilibrium ad valorem tariff rate τifor any nonnumeraire industry i in the resulting subgame-perfect Nash equilibrium can be derived as

(2) τ i 1 + τ i = I i Θ a + Θ F i M i e i ,

where Mi denotes imports, Fidomestic output, and eithe absolute price elasticity of import demand of good i. Iiis a lobby indicator variable, and Θ‎ denotes the total population share represented by SIGs.

By (2), an import tariff for lobbying industries results, whereas industries that do not lobby face import subsidies. The absolute trade policy measure is increasing in the inverse import penetration ratio Fi/Mi and decreasing in the absolute price elasticity of import demand ei.8

The question of which industries will form lobbies is left open by the original PfS model. Mitra (1999) endogenized the lobby formation and found that, mirroring in particular the predictions in Olson (1965) about the role of concentration, industries with more capital, more capital concentration, and more inelastic demand are more likely to form an SIG.

The PfS model assumes a small country that takes the world market prices as given and can freely choose the tariffs. Both assumptions are problematic. Most countries are big in an economic sense in at least some markets. Moreover, most countries are WTO members and thus bound in their tariff decisions by the GATT principles. Grossman and Helpman (1995) developed a framework where two governments that value domestic welfare and financial contributions interact, either in a trade war or in trade negotiations (“trade talks”). The resulting equilibrium tariff in a trade war reflects PfS political support, but also terms-of-trade motives: As in the classical Johnson (1954) model, a smaller foreign export supply elasticity increases the equilibrium tariff. In trade talks, however, the terms-of-trade motive disappears from the equilibrium tariff equation. Instead, the difference between the domestic and the foreign tariff is determined by the difference in the industry lobby strength at home and abroad. A strong lobby at home and a weak one abroad results in a relatively high domestic tariff compared to the foreign tariff.

Trade agreements can not only solve the terms-of-trade externality problem, where the unilateral desire to manipulate the terms of trade via an optimal tariff results in an inefficient trade war, but may also help governments commit against trade protection demands from lobbies. In Maggi and Rodriguez-Clare (2007), capital is mobile across sectors, contrary to the original PfS model. In this scenario, capital will want to move to sectors where protection is expected, creating a welfare-reducing distortion. Signing a free trade agreement provides the government with a way to commit to free trade vis-à-vis domestic lobbies and thus prevent this distortion. From the model, it also follows that free trade agreements should be deeper for sectors with a higher degree of capital mobility.

First Empirical Evidence

The PfS model spurred a new interest in the political economy of protection and stimulated research that empirically tested its main predictions. An empirical political economy literature existed, but typically consisted of reduced form regressions that lacked solid theoretical foundations. The PfS model offered the opportunity to take a theory-driven political economy look at real-world protection data.

The first two articles in this vein were written by Goldberg and Maggi (GM; 1999) and Gawande and Bandyopadhyay (GB; 2000). Both use basically the same 1983 cross-sectional U.S. data set. The baseline structural estimation equation is derived directly from (2) as

(3) τ i 1 + τ i = β 0 + β 1 I i F i M i e i + β 2 F i M i e i + ε i

for a cross-section of U.S. standard industrial classification (SIC) 3-digit manufacturing industries i.9 In lieu of ad valorem tariff rates, both GM and GB use cross-sectional U.S. nontariff barrier (NTB) data, more precisely NTB coverage ratios, to calculate the dependent variable, because as a WTO member, the U.S. is bound by GATT-WTO regulations and cannot freely adjust tariff rates. Although NTB coverage ratios are, at the tariff line level, simple (0,1) measures, they may be an appropriate measure for the severity of trade restrictions at the industry level, as Trefler (1993) found a high correlation between average tariff rates and NTB coverage ratios for the United States. Import and output levels are readily available, and even import demand elasticity estimates exist.10

Determining which industries actively lobby poses a major problem, however. Both GM and GB use political action committee (PAC) contributions to assign values of 0 or 1 to the lobby indicator variable Ii. In principle, industries with active lobbies are easily identifiable as those that pay strictly positive contributions. But in reality, at the 3-digit SIC level of aggregation, all industries contribute strictly positive amounts. GM and GB argue that this does not mean that all industries lobby for trade policy. Rather, PACs may pay money for different reasons, and it is thus necessary to sort the industries into those that lobby for trade protection and those that pay contributions for different reasons. The sorting procedures in GM and GB differ, however: GM simply used a cutoff value for PAC contributions. If the PAC contributions for industry i lie above the cutoff level, Ii=1is assigned. Robustness results for different cutoff levels are presented. GB used a more elaborate procedure. They ran auxiliary regressions where the PAC contributions divided by value-added were regressed on the product of SIC 2-digit dummies and import penetration ratios vis-à-vis several main U.S. trading partners. If the predicted value of the dependent variable was positive, the PAC contributions were assumed sensitive to import pressure; PAC spending of the 2-digit SIC industry was trade-induced; and thus trade policy lobbying takes place. Both GM and GB also acknowledged that the explanatory variables are most likely endogenous because of reverse causality, and they instrumented for them with several variables borrowed from Trefler (1993).

Although the GB and GM approaches slightly differ, their empirical findings are very similar. The PfS model is empirically confirmed: In (3), the β1estimate is positive and the β2estimate negative, and both are statistically different from 0. More importantly, however, the structural parameters, recovered from the regression parameters as Θ=β2β1 and a=1+β2β1, conform to theory. The estimated lobby population share Θ‎ is strictly positive, but below 1, and the weight a on social welfare in the governmental welfare function is also strictly positive.

Despite this prima facie empirical evidence in favor of the PfS model, the structural estimates raise some questions. The estimated lobby population share Θ‎ is very close to 1, suggesting that almost the entire population is represented by lobbies, and the estimated social welfare weight a is very high (e.g., more than 3,000 in GB). Because the weight on political contributions is normalized to 1 in the PfS model, the a estimate suggests that the government is de facto putting equal weight on the domestic welfare net of contributions and the contributions themselves.

Further Empirical Evidence and Extensions

The existing empirical PfS literature is extensive, and any overview of its scope and problems must be necessarily selective.

“Protection for sale” may not only explain unilateral trade policies but also bilateral or multilateral trade liberalization. Baldwin and Magee (2000) investigated how labor and business contributions changed the voting behavior of U.S. legislators in 1993–1994, when the North American Free Trade Agreement (NAFTA), the GATT Uruguay Round agreement, and the granting of Most Favored Nation (MFN) status to China were up for ratification in Congress. They found evidence that both the NAFTA and GATT decisions were influenced by contributions. In particular, without pro-trade business contributions, NAFTA would not have passed Congress.

Gawande et al. (2006) considered the role of foreign lobbies in the determination of U.S. trade policy. Whereas foreign nationals are not allowed to financially contribute to U.S. politicians directly, they can commission U.S. citizens, who then act on their behalf, provided these people register as “foreign agents” with the U.S. Department of Justice and pay the contributions out of their own pocket. Gawande et al. (2006) found that such lobbying activity reduces U.S. trade barriers.

Matschke and Sherlund (2006) investigated the role that organized labor and the degree of labor mobility play for U.S. trade policy. In the baseline PfS model, the protection rents go to capital owners. But workers may receive part of these rents via collective bargaining. Then, compared to the baseline model, theory predicts equilibrium protection will be lower when capital owners lobby, but trade unions do not, whereas equilibrium protection will be higher when trade unions lobby, but capital owners do not. Matschke and Sherlund (2006) found empirical support for these predictions in the original 1983 U.S. protection data used by GM and GB. Moreover, the parameter estimates for the lobby population share and the domestic welfare parameter are lower and thus more realistic than in the baseline model.

Gawande et al. (2012) similarly found lower and more plausible structural parameter estimates by augmenting the baseline PfS model by lobby competition between upstream and downstream producers. Whereas downstream producers are interested in higher tariffs for their own products, they prefer a lower protection level for their inputs and will lobby against the interests of their upstream suppliers. In the original PfS model, in contrast, all goods are final goods. A competition between industry SIGs only arises because capital owners as consumers consume goods from all industries and thus have an interest in import subsidies for those goods that they do not produce, but consume.

Matschke (2008) addressed the question of why, typically, import protection and not export promotion is observed. She argued that because tariffs generate revenue, whereas income taxation to pay subsidies entails additional costs, a natural bias toward the use of trade taxes exists. The PfS framework with U.S. tariffs as trade policy variable fits the data well if revenue considerations are included, but poorly otherwise. Moreover, the model can be used to obtain a precise estimate for the cost of raising taxes.

In Bombardini (2008), individual firms rather than monolithic industry lobbies choose the level of political contributions. In the presence of a lobbying fixed cost, only the largest firms in a sector will contribute, and the share of industry output produced by contributing firms can thus serve as a continuous measure of political activity at the industry level. As industries with a higher firm size dispersion (keeping the mean firm size constant) contain more firms that are politically active, higher firm-size dispersion should be positively correlated with the protection level. Bombardini (2008) indeed found empirical support for this prediction, using an augmented version of the original GB (2000) data set.

Saha (2019) dropped the assumption that the government weighs all contributions equally and assumed instead that the effectiveness of lobbying varies by industry. Contributions from more effective industries receive a higher weight in the governmental welfare function. For the most effective industries, trade protection and inverse import penetration ratio will be positively correlated, whereas the opposite is true for the least effective industries. More geographically concentrated industries producing more highly differentiated goods are found to be more effective in lobbying, as proxied by firm membership in trade associations. The modified PfS specification was tested and confirmed for Indian trade policy data.

Evidently, assigning a value of either 0 or 1 for the lobby indicator, as is standard in most of the empirical PfS literature, introduces a major element of data arbitrariness. Imai et al. (2009) showed that a simpler model than PfS, namely, one where organized industries receive higher protection in response to import surges, would yield similar estimates as PfS itself, which casts doubt on whether the empirical tests really support PfS. Imai et al. (2013) went one step further and claimed that the empirical evidence, at least with regard to the original U.S. 1983 data that triggered the empirical PfS literature wave, is actually at odds with the PfS predictions. They proposed a new test for the PfS model that does not require any assumptions about whether an industry is politically organized, building on one of the central predictions of PfS, namely, that for given controls, and in particular given inverse import penetration ratio, organized industries receive higher protection than unorganized ones. Moreover, for these industries, a higher inverse import penetration ratio should be positively correlated with higher protection. However, when sorting industries into quantiles based on the protection level, a graphical analysis reveals that for the highest quantiles, the observed correlation is opposite to this prediction. This result is confirmed in a formal econometric quantile regression. Imai et al. (2013) concluded that the PfS model needs to be theoretically further developed to yield a result that matches the empirical relationship between import penetration ratio and protection.

Ederington and Minier (2008) similarly voiced concern about the validity of empirical PfS tests, pointing out two key problems: First, all industries lobby, evidenced by strictly positive PAC contributions. Distinguishing between industries that have formed a trade-related SIG and those that have not by claiming that part of the PAC contributions constitute lobbying for non-trade-related issues is at odds with the PfS assumption that lobbying exclusively targets trade policy and that other redistributive policy instruments are not available. Secondly, the fact that import subsidization is extremely rare contradicts the prediction that nonorganized industries will face import subsidies in equilibrium. Introducing a constant and an error term in the estimation equation to account for extraneous, trade-policy influencing factors similarly leads to model misspecification. Ederington and Minier (2008) concluded that “researchers must more rigorously account for the presence of domestic policy support and extraneous political factors in any empirical work” (p. 503) related to the PfS model.

The extremely high estimates of the structural model parameters, namely, welfare weight and lobby population share, have haunted the empirical PfS literature from its start and may be due to model misspecification. However, not all empirical PfS papers focus on the derivation of such estimates. Using a slightly extended PfS framework, Cadot et al. (2004) found that the PfS model matches observed differences between trade policies in industrialized and developing countries remarkably well. To obtain this result, they assumed that all industries lobby for trade policy and that lobbying competition exists as final producers oppose trade protection for their upstream suppliers. Moreover, the assumption of constant wage is dropped, as the numeraire sector no longer uses only labor in production. In this setting, the following stylized empirical facts can be explained:


The protection rate increases with the degree of processing because the industries further down the supply chain face less lobby competition.


Developing countries on average impose higher trade restrictions than industrialized countries. This is true because developing countries possess fewer inter-industry linkages, so lobby competition is less intense.


Industrialized countries protect agriculture more than manufacturing, whereas for developing countries, the opposite holds. This result follows if the agricultural labor share in value added, which reduces the lobbying incentive for capital owners, is low in industrialized countries and high in developing countries.


The PfS model has been the fountainhead for a large empirical literature on the political economy of trade protection. However, the evidence found in support of this model is mixed.

Because there does not exist a clear-cut answer on how to distinguish lobbying from nonlobbying industries, the lobby indicator construction remains arbitrary, leading to possible model misspecification.

The structural estimates of the model parameters vary considerably and typically appear very high, both with regard to the lobby population share and the welfare weight.

Trade policy is often measured by nontariff barrier coverage ratios, not by tariffs for which the theory was developed. Moreover, the considered trade policy measures are often quite persistent and may thus have been determined earlier than the political economy variables used to explain them, casting doubt on any causal claims.

The inclusion of tariff revenue in the lobby objective function is a key element for an interior solution but appears at odds with real world perception about its relevance. Moreover, the empirical relationship between import penetration ratio and protection rate should depend on whether an industry lobbies, but whether it really does is still an open question.

The PfS model predicts import subsidies that are rarely observed, and it also does not help explain why trade policy is heavily biased in favor of import protection, whereas export promotion is rarely observed. Already Rodrik (1995, p. 1477) stated that on this latter puzzle, “we get very little help from the literature.”

Although the PfS model still dominates the political economy of trade policy literature, a need for theoretical and empirical advances beyond PfS clearly exists.

Beyond Protection for Sale

The PfS model is said to have solid micro foundations. This is true with regard to the principal agent framework where one common agent, the government, interacts with several contract-setting principals, the lobbies. But questions remain. Why does the government maximize a weighted sum of domestic welfare and SIG contributions? How does it use these contributions? Is this quid pro quo approach, where money is paid in exchange for trade policy concessions—something that noneconomists would typically call corruption or bribery—really what takes place in practice? The literature has started to to move beyond PfS, and one major step is to consider other forms of lobbying that do not build on a quid-pro-quo approach.

Lobbying Expenditures and Informational Lobbying

It has long been noted (by Tullock [1972] for the United States) that the amount of money in politics appears surprisingly small, a puzzle reconfirmed by Ansolabehere et al. (2003), in particular if solely political action committee (PAC) contributions are considered.11 Low political spending at nonbinding levels and large SIG gains from government intervention (and much higher general welfare costs) led Ansolabehere et al. (2003) to conclude that campaign contributions to politicians should not be interpreted as investment in political outcomes, but rather as consumption expenditure, similarly to charity donations.

Kang (2016) structurally estimated a full game-theoretic model of rent-seeking lobbying activity in the energy sector and found a positive but very small lobby impact on the probability of policy enactment, about 0.05 percentage points. Yet, the estimated average return of contributions exceeds 130%. Kang’s findings thus cannot solve the Tullock puzzle, and she mentioned market frictions in the political influence market as possible explanation.

Bombardini and Trebbi (2011) pointed out, however, that the astronomical rates of return for political spending reported by Ansolabehere et al. (2003) and others can be reduced considerably if one does not view contributions as the only channel for which interest groups can influence policy choices. They developed a model where lobbies exert influence via campaign contributions and offering votes. The model yields an inverted U relationship between contributions and votes; that is, sectors with a medium employment size (as proxy for votes) give the most contributions. Smaller and larger sectors pay less.

The quid pro quo approach to lobbying, that is, the view that contributions are paid in exchange for policy favors, is something that economists may consider a natural way of modeling the lobbying process, but the fact that lobbying expenditure exceeds campaign contributions by far is a point in case that the quid pro quo view may fall short of explaining the lobbying process appropriately.

Political scientists as well as some economists have suggested instead that lobbies exert influence on politicians by providing information (e.g., Austen-Smith, 1992, 1993, 1995; Austen-Smith & Wright, 1994; Lohmann, 1995; Potters & Van Winden, 1990, 1992; Schlozman & Tierney, 1986) and that business campaign contributions buy access (for empirical evidence, see Herndon, 1982; Langbein, 1986) to politicians to provide information, rather than directly buying policies. It is thus information transmission, not so much money transmission, that connects SIGs with politicians, notwithstanding that obtaining access to legislators typically comes at a cost (Austen-Smith, 1995; Grossman & Helpman, 2001).

Grossman and Helpman (2001, chap. 4 for cheap talk and chap. 5 for costly lobbying) built on this and related work (in particular, Krishna & Morgan, 2001). One of their findings in the cheap talk scenario where information transmission is costless is that the more aligned the preferences of interest group and policymaker are, the more precise the information provided by the SIG can be. A lack of credibility due to conflicting interests, in contrast, leads to coarser information transmission. In cases where lobbying from different sides occurs, this can actually raise the credibility of a lobby, whereas information coming from more extreme SIGs of the same side will be deemed incredible and will thus be ignored. With lobbying costs, the mere act of lobbying may serve as a signal and convey information about the state of the world. When the SIG can actually choose the lobbying effort (endogenous lobbying), it can use higher lobbying to signal a higher urgency of its case. An interesting situation arises when the policymaker charges the lobbies a fee for access. In this case, only lobbies whose interests are quite aligned with the policymaker’s and those that have very different interests will have an incentive to pay the access fee.

Bennedsen and Feldmann (2002) formulated a public good provision model where informational lobbying takes place within a multimember legislature; that is, the assumption of a monolithic policymaker is dropped, similarly to Grossman and Helpman (2005). The SIG is interested in a high public good provision. The incentives for lobbying then depend on the legislative structure. Bennedsen and Feldmann (2002) found that in a congressional system where no fixed governing coalition exists, lobbies fulfill a vital role in helping to identify legislators (districts) with high public good preference, who can then be included in an enacting coalition, whereas other legislators (districts) with low public good demand will be excluded. This model thus combines an analysis of the legislative structure with informational lobbying. Although the model does not explicitly address trade policy, it could be adjusted for this purpose.

Ludema et al. (2018) used an informational lobbying model to explain tariff suspensions. Upstream and downstream firms lobby for and against tariff suspensions via costless messages and costly lobby expenditures. Higher lobby expenditures by proponents of a tariff suspension ceteris paribus raise the suspension probability. In the absence of spending by opponents, verbal opposition itself reduces the suspension probability, and the same is true for more spending by opponents, although the estimated effect is smaller. Ludema et al. (2018) tested these predictions for the United States by estimating a linear probability model with the outcome of voting on tariff suspension bills, which pertain to product level tariffs, as dependent variable. This approach circumvents a problem present in most of the empirical trade protection literature: Tariffs and nontariff barriers, the dependent variable of choice in such models, are typically quite persistent over time and thus unlikely to be caused by the current political economy variables found on the right-hand side of the estimation equation. On the other hand, the firm-level trade policy lobbying data used by Ludema et al. (2018) still leave open the question of whether lobbying was really with regard to a specific tariff suspension bill.

Tovar (2011) combined an informational lobbying model with the PfS model. In contrast to the baseline PfS version, the factor ownership is so highly concentrated that an SIG does not consider consumer surplus or tariff revenue shares. For this reason, any industry SIG i will only lobby for trade policy pertaining to its own product. As in the PfS model, the industry i lobby presents a contribution schedule to the government depending on the chosen tariff for its product. But this schedule is also contingent on the productivity parameter μi, which the SIG does not know yet. To convey the information about the productivity level once this information becomes available, the lobby group needs to spend money L(μi), an endogenous lobbying expenditure. From the size of the lobbying expenditure, the government can infer the productivity parameter. Higher productivity will lead to more output, a higher lobbying expenditure, and a higher tariff in equilibrium. The equilibrium tariff equation resembles the one of the pure PfS model, namely,

(4) τ i 1 + τ i = I i a F i M i e i I i 1 + a a L i ' M i e i

where the notation is as in (2) and Li'denotes the positive derivative of lobbying cost with respect to the tariff. Ceteris paribus, a steeper lobbying cost curve implies a lower equilibrium tariff.12 The model is estimated at the 3-digit ISIC industry level as a double-censored Tobit model, using NTB tariff equivalents to calculate the dependent variable.

These last two examples show that the informational lobbying literature has found its way into the empirical political economy of trade protection literature, but they also document that these attempts typically constitute augmentations of PfS.

Firm-Level Lobbying and New Trade Theory

The political economy of trade protection literature discussed so far is grounded in the traditional neoclassical trade theory. The Stolper–Samuelson theorem of Heckscher–Ohlin theory predicts that freer trade hurts owners of the domestically scarce factor in the long run. The Ricardo–Viner (specific factors) theory, which complements Heckscher–Ohlin theory for the short to medium run, stipulates that because of short-run capital immobility, freer trade hurts import-competing industries.

Voting models based on Heckscher–Ohlin theory thus predict that trade protection arises if the median voter owns high relative endowments of the domestically scarce production factor. Lobbying models, in contrast, typically use a Ricardo–Viner setup and find that lobbying by import-competing industries leads to trade protection.

However, since the late 1970s and early 1980s, a new trade theory has emerged.13 It is firm centered (each firm is a price-setting monopolist for a particular variety of a differentiated good, monopolistically competing against firms that offer similar but distinct varieties) and can explain intra-industry trade flows, which are the predominant trade form, in particular between industrialized countries. The improved availability of firm-level data has enabled the empirical validation of its predictions.

The new trade theory is still rarely applied to trade policy, and in particular the political economy of trade protection. Part of the reason is that trade barriers are, on average, low today. Moreover, the move away from traditional trade models has blurred the lines between those who benefit and those who suffer from trade liberalization: In the new trade models, profits are driven to zero by free entry and exit, and the wage rate is often the normalized price. Gains from freer trade exist but only accrue at the consumer level, whereas producer and factor owner welfare is not affected.

It was a political scientist who first pointed out that the emergence of new trade models did not necessarily imply the obsolescence of the political economy of trade protection. Gilligan (1997) argued that in a world where monopolistic firms are the main actors, trade policy lobbying switches from being a public to being a private good. Consequently, for a single firm with well-specified interests, the free rider problem of lobby organization vanishes. The firm can lobby itself to pursue its interests without having to fear that other firms can free-ride on its lobbying. Bombardini and Trebbi (2012) indeed found empirical evidence that firms in differentiated-goods sectors with higher concentration are more likely to lobby individually.

Another political scientist, In Song Kim from the Massachusetts Institute of Technology (MIT), built on this insight and provided evidence that shifting from an industry to a firm-level focus on lobbying is a very fruitful avenue for future research. To this purpose, Kim (2017) assembled a comprehensive data set on firm lobby activities with regard to trade policy that links lobbying reports, available from the Senate Office of Public Records under the Lobbying Disclosure Act of 1995, and firm information from Compustat and Orbis. Kim complemented these data with information about the content of lobbying from the respective congressional bills. Information about the firms’ product range and industry classification are linked to trade barrier and trade volume information and the level of product differentiation.

Kim (2017, pp. 2–3) showed that 70% of the current variance in MFN U.S. tariff rates can be attributed to the variation within industries. With the conclusion of the GATT Uruguay Round in the mid-1990s, the resulting increase in the number of tariff lines, by 1,303 from the year 1993 to 1996, was accompanied by a steep increase in the ad valorem U.S. tariff rate variance. This increase, however, was relatively moderate with regard to the between-industry variance, but astronomical for the within-industry variance.

In the theory section of his article, Kim (2017) combined a simple new trade model with four firms (two domestic and two foreign, of which one domestic and one foreign firm are productive enough to also sell to the respective foreign market) with a PfS lobbying model. The domestic government is lobbied by the two domestic firms and the foreign exporter. The following hypotheses are derived:


“Productive firms are more likely to lobby for trade liberalization when they compete in an industry with differentiated products than when they compete in an industry with substitutable products.”


“Differentiated products will have lower tariffs than substitutable products, on average.” (Kim, 2017, p. 9)

Kim empirically tested his hypotheses for a panel of all publicly traded firms in manufacturing and agriculture between 1999 and 2014, 4,030 firms from 535 North American Industry Classification System (NAICS) 6-digit industries in all. In a logistic regression with a lobby indicator as dependent variable, he found a big and statistically significant positive impact of the interaction between the degree of product differentiation and productivity on lobbying activity. The separate effect of productivity is U-shaped; that is, the least and the most productive firms are ceteris paribus the most active lobbyists. Secondly, the variation of tariff policies as a function of substitutability at the Harmonized System HS-8 product level is investigated in a random coefficients model. Lower substitutability is associated with lower applied MFN tariff rates.

Kim and Osgood (2019) provided additional descriptive evidence on the importance of understanding firm lobbying behavior. Lobbying firms have higher revenue, and among these firms, revenue is higher for those that lobby for trade-related issues than for those that lobby for non-trade-related issues. When comparing firms that support free trade with those that do not, the finding is even more pronounced: Free trade supporters earn considerably higher revenue. Moreover, within any given manufacturing industry, lobbying activity is quite heterogeneous.


The political economy of protection is a large and active research area at the boundary of economics and political science. This article has focused on the political economy of trade protection, following the approach by Hillman, whose 1989 book shares the title with this article. Since 1989, however, the political economy of (trade) protection has progressed considerably. The Grossman–Helpman (1994) protection for sale (PfS) model has provided a game-theoretic micro foundation for the political support function proposed in Hillman (1989) and has become the standard framework for political economy models that investigate the influence of special interest groups (SIGs) on policymaking. But voting models have also progressed further: Dynamic versions of the originally static Mayer (1984) median voter model have been developed, and the decision-finding processes within legislatures have been explored.

With regard to the driving forces behind the political economy of trade protection, however, the existing literature has its limitations, and further work is required.


The policymaker preferences are still not well understood. Regarding the policymaker as one monolithic entity constitutes a problem if trade policy is decided within a legislature where a majority needs to be secured for every policy issue. Although some literature on this issue exists (e.g., Bennedsen & Feldmann, 2002; Grossman & Helpman, 2005), more research on the political decision process is needed. But even considering the policymaker as one entity with a well-defined objective function, it is not clear whether the widely used PfS governmental welfare function, consisting of a weighted average of social welfare and campaign contributions, is really appropriate. PfS theory assumes that SIGs influence the government’s decision via campaign contributions, but the size of these contributions is small and often not linked to trade policy issues. The view that SIGs influence policymakers not via monetary contributions, but by providing information, has gained influence but leaves open questions, in particular why the information provision would lead policymakers to rig the policy in favor of the informants. In the PfS model, for example, information is perfect, and with this information and without SIG contributions, the government would choose free trade for all industries. Moreover, the fact that most empirical trade policy papers claiming to test for informational lobbying use the PfS model as a backbone appears troubling.


The economic interests behind protection, be it the interests of voters or SIGs, are derived from the traditional neoclassical trade theories, namely, the Heckscher–Ohlin and the Ricardo–Viner models. According to the Heckscher–Ohlin theory, owners of the domestically scarce factor will gain from trade protection. According to the Ricardo–Viner theory, capital owners of import-competing industries demand protection because they cannot move their capital out of these shrinking sectors in the short to medium run. Cleavages in political interests thus appear along factor or industry lines. The new trade theory, however, considers monopolistically competitive firms of possibly differing productivity that produce distinct varieties of a certain good. Especially the more productive firms may have an interest in reciprocally open trade to secure access to foreign markets while also not worrying about competition from abroad, as the specificity of their products provides some protection.

Advances focusing on firm lobbying activity are coming out of political science, for example Kim (2017) and related work. Newly available micro-level U.S. lobbying data merged with product-level data allow for a careful empirical investigation of such lobbying. Much of the empirical literature on the political economy of trade protection still relies on macro-level data at the industry or even country level. Because trade policy, as many other policies, is often very persistent at the aggregate level, current political activity probably does not have much effect on it. Using micro-level data instead appears as the right way to proceed in the empirical political economy literature. Such micro level data on the interaction between politicians and interest groups is not available for most countries, however; hence the econometric analysis of the interplay between political interactions, on the one hand, and policy choices, on the other hand, will be regionally limited for the moment. Without available cross-section, time series, or panel data to conduct an econometric analysis, case studies may instead prove informative. Schattschneider (1935), often cited as the first book on the political economy of trade protection, is just that: a case study how the notorious Smoot–Hawley Tariff Act of 1930 came to pass Congress.

The benevolent and omniscient planner of classical economic theory is a useful yardstick for comparison. To understand actual policy outcomes, however, an analysis of the political economy, that is, the interaction between policymakers and economic agents pursuing their own interests within a given political framework that structures this interaction, is crucial. Such theoretical and empirical analysis not only helps in understanding political outcomes but can also provide guidance on how to adjust the regulatory framework in case unsatisfactory outcomes are observed.

Further Reading



  • 1. The distinction between pre- and post-election period is a theoretical construct, however, because in a democracy, after the election is always before the election, especially in the United States, where elections to Congress take place every two years, and contributions help finance the candidates’ electoral campaigns.

  • 2. The model is very closely related to Hotelling’s (1929) spatial duopoly model of location choice, where two firms sell a homogeneous good and simultaneously choose their position in a line market.

  • 3. When individual preferences are single peaked, the individual has a preferred choice (bliss point), and any deviations from this bliss point reduce their utility, with more distant choices reducing utility more. Assuming single-peakedness of preferences is important to rule out Condorcet voting cycles, where the winner of an election depends on the order in which motions enter the voting process.

  • 4. This view is in contrast to that of Rickard (2012), who considers subsidies a means of protection.

  • 5. The latter point mirrors results in Krishna and Mitra (2008).

  • 6. This revenue-sharing assumption is common in the literature but raises a cognitive-dissonance issue, as a real interest in trade revenue is rarely present in industrialized countries (Ethier, 2010).

  • 7. Because the country is small in the world market, a domestic welfare–maximizing government would choose free trade.

  • 8. This definition of import penetration is somewhat nonstandard because import penetration is usually defined as imports divided by domestic consumption. Here, domestic consumption is replaced by domestic production.

  • 9. GB also incorporate intermediate goods in the estimation equation.

  • 10. The fact that the price elasticity of import demand eis a g. enerated (estimated) regressor means that a statistical correction is needed in the estimation, as GB point out.

  • 11. The “political spending” becomes considerably higher, however, if total lobbying spending and the contributions by Super PACs, a new PAC form created in 2010, are taken into account (see Bombardini & Trebbi, 2019). For the U.S. electoral cycle of 2018, PAC contributions amounted to about 570 million USD, whereas lobby expenditures equaled about 5.9 billion USD, according to the Center for Responsive Politics.

  • 12. The lobby expenditure derivative is estimated from an auxiliary regression of lobbying expenditures on protection.

  • 13. Krugman (1979) is the seminal paper for this literature.