Protectionist Executives

ABSTRACT

The trade policy literature long presumed that legislatures favor relatively more protectionist policies than executives do, but more recent studies have found little evidence to support the idea. This article clarifies these mixed results with a simple formal model. It finds that legislative protectionism requires very specific circumstances involving the combination of universalistic legislative norms and a particular economic geography, with trade policy interests large relative to legislative districts but small compared to countries. Empirical evidence on this latter point suggests that the necessary spatial patterning is likely to have diminished in many countries since the 1930s, when the Smoot-Hawley Tariff and the Reciprocal Trade Agreements Act focused attention on institutional differences. The assumption of legislative protectionism based on theories and measures of that era thus may not hold today, and widespread presumptions about institutional preferences may accordingly be uncomfortably fragile.

KEYWORDS:

• Geography
• institutions
• trade barriers

Acknowledgments

An early draft of this paper was presented at the Midwest Political Science Association meeting. I thank Jeffry Frieden, Ken Shepsle, Amy Widestrom, Gerald Schneider, Maya Hadar, the anonymous reviewers, and especially Mark Nieman for helpful comments.

Notes

¹ Similarly, Rogowski (1987:204) argues that “In small districts, localistic ‘parish-pump’ and ‘pork-barrel’ orientations are likelier to dominate.”

² The executive’s larger constituency forces protectionists to band together in larger coalitions, raising the costs of collective action, but protection’s concentrated benefits can cover even a large group’s costs of organization (Grossman and Helpman 1994).

³ These constituencies can, but need not, be thought of as geographical entities, as in Westminster-style systems.

⁴ Voters may not be self-interested in their trade policy preferences, even if the balance of the evidence suggests they are (Fordham and Kleinberg 2012), and in any case many are essentially unconcerned with trade and international-economic issues (Guisinger 2009). Any such voters would be irrelevant to the model. However, politicians trying to use trade policy to capture votes are apt to focus, just as the model does, on constituents who do care about trade and who are likely to want protection for themselves.

⁵ Except where noted, the model produces identical results if the status quo protects every sector.

⁶ Unless otherwise noted, I assume uncertainty is large enough to guarantee that –ω < π_i – π* < ω— that is, that economic shocks can be large enough to affect vote choices.

⁷ Formally, + – > – for all N > 1. If the economy has exactly one industry, the voter is equally likely to vote for the incumbent whether or not that industry receives protection, since the costs to protecting the entire economy are large enough to balance the benefits of receiving protection.

⁸ The agenda setter here is indifferent over the possible sets of coalition partners and industries that receive protection, but institutional structures or lobbying behavior could determine which industries and representatives obtain protection.

⁹ If the agenda setter maximized her expected number of votes instead of expected probability of reelection, she might choose to protect a large supermajority of her own industries. This would raise the protectionism of the legislature, but not by very much (for large legislatures).

¹⁰ A similar result might apply if there were sector-specific economic shocks rather than a single economy-wide one.

¹¹ If the status quo is for economy-wide protection instead of economy-wide free trade, the expected level of protection is , higher than any equilibrium outcome for k > 1.

¹² Values of (that is, n) likely reflect institutional differences, with generally higher values in first-past-the-post electoral systems and lower values in proportional systems. This model assumes single-member districts, however.

¹³ The representative strictly prefers any such proposal to the status quo unless it protects all industries in all constituencies.

¹⁴ This potentially implies increased diversification at the national level, too, but the change is proportionally less dramatic when applied to larger, national populations.

¹⁵ Population figures are unadjusted for the breadth of the franchise: Nonvoter members of society can, after all, join and influence interest groups. Extensions of voting rights over time mean eligible voters per legislator increase even more dramatically.

¹⁶ There are exceptions. Germany’s constitution keeps the Bundestag roughly proportional in size to the voting population, just as the pre-war Reichstag was. Notably, however, Bundestag constituencies are about 30% larger than were the Reichstag’s.

¹⁷ The RTAA might still have been influential for other reasons—say, by increasing the ability to negotiate reciprocal concessions or reducing the number of veto players.

¹⁸ Even without explicit attempts to match industries to districts, natural configurations of economic geography may lead to concentration (Chen and Rodden 2013).

¹⁹ Moomaw (1998) shows two-digit industrial codes’ effectiveness for measuring agglomeration economies to be comparable to that of more-detailed classification systems.

²⁰ The SIC periodically changes to account for structural changes in the economy. The most substantial change came with the 1997 revision that formally replaced the SIC with the North American Industrial Classification System (NAICS), which significantly reduced the number of two-digit codes.

²¹ Taking the logarithm of the concentration ratios to alleviate skew increases the t-statistic on the difference of means to 9.12.