Compensation Contract for Trade Liberalization

We define a compensation contract as the inter-branch consensus between the executive and legislative branches with regard to the depth of trade liberalization and whom and how much to compensate in order to secure the successful ratification of a trade agreement.

Our theory is based on four well-established premises about actors, incentives, and the institutional environment of the politics of trade agreements. First, in both parliamentary and presidential democracies, the executive branch negotiates trade agreements with a foreign government and the legislative branch ratifies them and appropriates the budget to compensate losers. The executive branch alone cannot enact trade liberalization without securing a majority support among legislators on the floor (Milner and Rosendorff Reference Milner1997; Naoi Reference Naoi2015).Footnote ¹ Second, we assume that in both systems, the executive is relatively more supportive of free trade than the median legislator due to its representation of broader domestic constituencies, i.e., the median voter (Karol Reference Karol2000; Milner Reference Milner1988; Thies and Porche Reference Thies and Porche2007). The executive branch’s free trade bias may be even more pronounced in presidential systems, where a president runs for a nationwide district.Footnote ² We further assume that when the executive branch must choose between tariffs and subsidies to secure ratification, the executive prefers subsidy over tariff because subsidy is associated with fewer deadweight losses than tariffs and because imposing tariffs will prompt a trade partner to retaliate by reducing market access for export industries (Tullock Reference Tullock, Alt and Shepsle1990). Subsidy, on the other hand, invokes less retaliation (Gilligan Reference Gilligan1997). Third, we assume that a trade agreement generates distributional effects geographically, i.e., some electoral districts, defined by geographical units, benefit from the agreement while others lose. Finally, we assume that legislators are office-seeking and that, for analytical simplicity, individual legislator’s decision to support ratification is purely determined by the net income effects of a trade agreement and compensation to their districts. This implies that legislators representing import-competing districts would prefer tariffs over subsidies when the probability of legislating compensation is low and would be indifferent between tariffs and subsidies when the probability is high. Legislators representing export-oriented districts and non-trading districts support ratification for the income benefits to industries and consumers, regardless of compensation.Footnote ³

Building on these premises, we argue that the executive seeks to achieve two interrelated goals. One is to negotiate an agreement that can be ratified by the legislature. Failure to ratify an agreement signals incompetence and a lack of leadership (Mansfield, Milner, and Rosendorff Reference Mansfield, Milner and Peter Rosendorff2000; Milner Reference Milner1997; Naoi Reference Naoi2015). The second goal is the electoral survival of co-partisans. The executive wants its co-partisan legislators to survive the next election; to stay in power by reducing adverse income effects of trade liberalization in districts.

Consequently, the key to successful ratification of a trade agreement is to identify representatives who might vote against trade liberalization on the floor and preempt Nay votes with the promise or actual delivery of compensation (Naoi Reference Naoi2015). A compensation contract is critical as targeted income compensation to losers—in the forms of subsidies, loans, and grants—can address both ratification and electoral survival problems by reducing the adverse income effects of a trade liberalization as long as these policy instruments can be distributed geographically.

Commitment and Information Asymmetry Problems in Compensation Contract

A primary challenge in negotiating a compensation contract is that the bargaining between the executive and legislative branches is supralegal, regardless of differences in constitutional structures (Weingast and Marshall Reference Weingast and Marshall1988). In fact, without legal means to bind the contract, neither the executive nor legislative branch could credibly commit to enforcing a compensation contract. The executive branch might promise compensation to legislators, but renege on them once a trade agreement is ratified (see Weingast and Marshall Reference Weingast and Marshall1988, on logrolling). Likewise, legislators may not comply with their initial commitments due to heterogeneous district interests or the lack of sufficient side-payments at the time of voting. Note that the timing of negotiating trade agreements and a legislature ratifying the agreement is inherently sequential, which worsens the commitment and enforcement problems (Drazen Reference Drazen2000).

Another challenge in negotiating and enforcing a compensation contract is the information asymmetry problem between the executive and legislative branches. In the context of trade liberalization, the contract between the executive and legislative branches requires that the executive branch has relatively accurate information about which districts or industries gain and lose from a trade agreement and the appropriate price for compensating these legislators (Naoi Reference Naoi2015). Yet, there exists information asymmetry between the two branches due to their inherently misaligned interests. On the one hand, legislators have incentives to maximize the amount of compensation they can extract from the executive branch by exaggerating the potential damage. On the other hand, the executive has incentives to buy off support with the lowest compensatory offer possible. Indeed, incomplete information is one of the key explanations for ratification failure in the literature (Milner and Rosendorff Reference Milner and Peter Rosendorff1996).

Below, we demonstrate that the geographic concentration of industries mitigates the commitment and information asymmetry problems in parliamentary systems, while geographic diffusion mitigates these problems in presidential systems.

Geographic Concentration and Compensation Contract

The differences between parliamentary and presidential systems in who negotiates a compensation contract (party leaders versus a legislature) and where it is enforced (within a party or in the legislature) leads to distinct geographic profiles of industries that are likely to be liberalized and compensated. In parliamentary systems where party leaders enforce a compensation contract within a party, the geographic concentration of industries facilitates the formation and enforcement of a compensation contract with a smaller number of concerned legislators. Accordingly, we predict that parliamentary systems are more likely to liberalize and compensate geographically concentrated products and protect geographically diffused products with tariffs. This prediction challenges the canonical logic of collective action which predicts that geographically concentrated industries win government protection by overcoming the free-rider problem (Olson Reference Olson1965). Our theory suggests that, from the perspective of party leaders, the geographic concentration facilitates liberalization because of its lower negotiation and information costs in striking the compensation deal.

On the other hand, in presidential systems with non-overlapping appointments between the executive and legislative branches, a compensation contract is more likely to be negotiated and enforced on the legislative floor during the ratification phase. This is because the separation of power with non-concurrent appointments between the executive and legislative branches entails severe commitment and information asymmetry problems in striking a compensation contract. Importantly, what binds the contract in presidential systems is a majoritarian threshold. A majoritarian threshold on the legislative floor means that numbers become power: geographically diffused industries, with a larger number of concerned legislators, yield more power to pass the compensation bills. Accordingly, in presidential systems, governments are more likely to liberalize and compensate geographically diffused industries.

In sum, the constitutional differences between parliamentary and presidential systems shape where interbranch contracts are negotiated and enforced: within a party for parliamentary systems and in a legislature for presidential systems. The geographic concentration of industries facilitates the formation and enforcement of a contract within a party in parliamentary systems, while geographic diffusion facilitates contract formation in the legislature with a majority threshold in presidential systems. We thus predict heterogeneous effects of constitutional differences across varying levels of the geographic concentration of industries on trade liberalization.

We consider trade liberalization as the result of successful policy substitution from the executive’s point of view to secure ratification, rather than as the results of how party or legislators maximize electoral survival. The logic of our hypothesis is consistent with McGillivray (Reference McGillivray1997)’s argument that diffused industry allows legislators to form a majority coalition with the lowest costs and hence wins trade protection in countries with weak party discipline (the U.S. is her case). However, our prediction is the opposite of hers because her model only considers a single policy instrument (tariff), while we consider geographic diffusion to facilitate substitution between tariff and compensation in presidential systems.Footnote ⁷

Moreover, our theory differs from the party discipline argument in two important ways (McGillivray Reference McGillivray1997; Nielson Reference Nielson2003). First, we consider commitment and information asymmetry problems that arise from the negotiations between the two branches, while the party discipline argument tends to focus on party behavior in the legislative branch alone. Second, the commitment and information asymmetry problems that our theory highlights are rooted in the constitutional structures rather than originating from the internal party dynamics. Accordingly, we argue that while strengthening party discipline may solve commitment problems to some degree, it does not solve the information asymmetry problems that plague compensation deals.Footnote ⁸

Finally, our theory is subject to two important scope conditions. First, our theory requires that legislative elections have districts comprised of geographic units (Catalinac and Motolinia Reference Catalinac and Motolinia2021; Fiva, Halse, and Smith Reference Fiva, Halse and Smith2021). Therefore, our theory does not apply to a country with a nation-wide closed-list proportional representation district, where voters choose parties not individual candidates. Second, another important scope condition is that the government can target either crop or geographic area for compensation.Footnote ⁹ Our theory of exchange between ratification vote and compensation does not hold if a government only has universalistic or formula-based compensation programs.

Data and Measurement

Despite the theoretical significance of the geographic concentration of industries in explaining economic policy outcomes, assessing its impact on trade liberalization poses two significant empirical challenges. First, various political and economic factors shape the geographic concentration of industries, which may confound our inference regarding the effect of constitutional structure on trade policies. Second, our empirical test requires that we link product-level tariff reduction data with product-level geographic concentration data. While product-level data on trade liberalization is publicly available, it is difficult to obtain geographic concentration data at the product-level that is comparable across countries.

Geographic Concentration of Crop Production with Remote-Sensed Data

To overcome these challenges, we focus on agricultural trade liberalization. Agricultural trade provides an ideal case to test our theory. First, the geographic distribution of agricultural production is relatively exogenous to politics and correlates more with terrain, soil, and climate conditions. In this regard, we follow Nunn and Qian (Reference Nunn and Qian2011) which shows that crop production indeed depends heavily on time-invariant geoclimatic conditions and is less likely to be influenced by political institutions or constitutional design. Second, compensatory policy tools for the agricultural sector tend to be product-specific across democracies (Anderson, Martin, and Valenzuela Reference Anderson, Martin and Valenzuela2006), allowing us to analyze parliamentary and presidential systems in a relatively comparable manner. This uniformity in the design of agricultural compensation programs is in contrast to a diversity of compensation programs for manufacturing sectors that range from a universalistic welfare program (e.g., Northern European welfare states that are not industry-specific), to formulaic compensation specifically designed for trade losers (e.g., Trade Adjustment Assistance [TAA] in the U.S.), and to targeted subsidies for firms (e.g., Japan, see Estévez-Abe Reference Estévez-Abe2008).

For a measure of the geographic concentration of agricultural production, we leverage novel data that measures geographic distributions of agricultural production for 145 crops around the world using remote-sensed satellite images. In particular, we use the crop yield data constructed by Monfreda, Ramankutty, and Foley (Reference Monfreda, Ramankutty and Foley2008), which combines satellite image data of land-use from Ramankutty et al. (Reference Ramankutty, Evan, Monfreda and Foley2008) with national-, state-, or county-level census data for each country. The census data of crop production supplements the satellite image data to improve the accuracy of estimated yields. We obtain this crop yield data that combines remote-sensed data and census information from EarthStat.Footnote ¹⁰ For each country and for each crop, we extract the production volumes for each grid cell from the obtained raster format data overlaid over the map of each country. Unlike census data where data information is collected at county- or state-level, EarthStat data are not bound by administrative borders. Therefore, we can compare the concentration of crop production at the exact same unit across countries and across crops. The resulting dataset offers yields of 145 individual crops around the year 2000 at 5 minutes × 5 minutes spatial resolution in latitude by longitude (approximately 10 km × 10 km).Footnote ¹¹ Figure 1 illustrates the granularity and the comparability of our remote-sensed data across various crops, visualizing the geographic distribution of sugarcane and rice production in Japan. Notice that rice (right) is produced in most parts of Japan whereas sugarcane (left) production is concentrated in two prefectures in the south.

Figure 1. Sugarcane and Rice Production per Grid (Approximately 10 km × 10 km)

Note: The figure shows sugarcane (left) and rice (right) production for Japan. It is visually clear that sugarcane production is more concentrated than rice production in Japan.

To quantify the geographic concentration of the crop yield from this land cover data, we then calculate the HHI. Specifically, we compute HHI for each crop, k, and for each country, i, to obtain our measure of the geographic concentration of agricultural production given by the following formula:

(1) $$ \begin{array}{rl}{Z}_{ik}:={\displaystyle \sum_{l=1}^{L_i}}\left(\right.\frac{P_{ikl}}{P_{ik}}{\left)\right.}^2,& \end{array} $$

where $ {P}_{ikl} $ is the production of crop k at grid cell l in country i, $ {P}_{ik} $ is the total production of crop k in country i, and $ {L}_i $ is the total number of grid cells in country i. The more (less) concentrated the crop production is, the larger (smaller) the HHI for the product becomes. For instance, $ \mathrm{HHI} $ will be unity if a crop is produced in one grid (i.e., concentrated), while $ \mathrm{HHI} $ approaches zero as the production is equally distributed in every single grid (i.e., dispersed). In terms of our example in Figure 1, the HHI value of sugarcane production ( $ \approx 0.0148 $ ) is much larger than that of rice ( $ \approx 0.0007 $ ). Given the skewed distribution of the geographic concentration measure, we log-transform the value for our empirical analysis.

HHI is appropriate for our theoretical framework because it is a standardized and distance-free metric to capture the geographic concentration of crop production. In particular, our argument considers the number of concerned electoral districts to be theoretically more relevant for the formation and enforcement of a contract than the distance among them. In this regard, HHI is preferred to other measures of the geographic concentration proposed in the literature such as the one used in Busch and Reinhardt (Reference Busch and Reinhardt2000), which takes into account the spatial proximity.Footnote ¹² To illustrate the advantage of our distance-free measure, consider tomatoes in the U.S., which are primarily produced in California and Florida. Although these two states are geographically apart, tomatoes are a concentrated crop in our theoretical framework. HHI is consistent with our theoretical framework because it measures overall concentration regardless of the geographic proximity of the area of crop production.

Outcome Measure of Trade Liberalization

Once the matches between crop names and HS product codes are established, we develop the outcome measure of trade liberalization that varies across each crop within each PTA. Our universe of trade liberalization includes a total of 155 PTAs signed by two parties that have become in force between 1995 and 2016, where at least one of the two signatories is democracy. The PTA information comes from the Regional Trade Agreement Database of World Trade Organization (WTO).Footnote ¹³ We follow the convention in the literature and consider a country democracy when the Polity IV score is equal to or greater than 6. Our sample consists of 38 democracies that signed at least one PTA during this time period, of which 25 are presidential systems and 14 are parliamentary systems. Two countries in the dataset switched their constitutional structures once in the given time frame.Footnote ¹⁴ We remove the European Union member countries from our sample because it is a supra-national organization with a unified trade policy.Footnote ¹⁵ This, combined with the heterogeneous geographic dispersion across crops, allows us to systemically evaluate the effect of constitutional structure across democracies.

To accurately measure crop-level episodes of trade liberalization in each trade agreement, we calculate the difference between preferential rate and the Most-Favored Nation (MFN) tariff rate for each of the matched HS6 product categories. Specifically, we classify crops to have experienced trade liberalization when a preferential tariff rate is lower than the pre-existing MFN tariff rate. If PTA tariff rate for a given crop is identical to pre-existing MFN rate (non-zero) or higher, we consider the product to remain protected. Thus, for each of PTA’s importing country and crop, we construct an indicator variable that takes the value of 1 if the PTA tariff rate of a given crop is lower than the MFN tariff rate in post-PTA period, while zero otherwise. Formally, our outcome measure of trade liberalization $ {Y}_{pik}\in \{0,1\} $ for PTA p, importing country i, and crop k is computed as follows:

(2) $$ \begin{array}{rl}{Y}_{pik}=\unicode{x1D7D9}\{{\tau}_{pik}^{\mathrm{PTA}}<{\tau}_{pik}^{\mathrm{MFN}}\},& \end{array} $$

where $ {\tau}_{pik}^{\mathrm{PTA}} $ and $ {\tau}_{pik}^{\mathrm{MFN}} $ are the average preferential and MFN rates at the HS6-level, respectively. Both PTA and MFN tariff rates data are obtained from the World Integrated Tariff Solutions (WITS).Footnote ¹⁶ Note that using the data available through the WITS from a particular year might be limited as countries sometimes fail to report preferential tariffs even years after the agreements are in force (Barari and Kim Reference Barari and Kim2020), while some products receive delayed tariff reduction over time due to tariff phase-out negotiated among trade partners (Baier and Bergstrand Reference Baier and Bergstrand2007). Since our goal is to accurately capture the episode of trade liberalization, we verify and correct any inaccuracies in the timing reported by each country. We then compute the moving average of the tariff rates over varying time window lengths. For the ensuing analysis, we adopt a conservative approach, using a broad time window that spans ten years both before and after each trade agreement. Moreover, when a crop aligns with multiple HS6 categories, as illustrated above, we average the tariff rates across all the corresponding HS6 categories to minimize potential measurement errors.

Empirical Findings

Our objective is to estimate the effect of crop concentration and constitutional systems on tariff reduction. We focus on the difference between PTA and MFN tariff rates in the post-PTA period, while accounting for any crop-specific heterogeneity across trade agreements. To achieve this, we formulate a correlated random effect (CRE) model as described in Equation 3. This model permits arbitrary correlations between the crop-specific intercepts and predictors, while avoiding the incidental parameter problem present in non-linear logistic regression with crop fixed effects. In this regard, this model can be considered as a logistic regression with crop fixed effects (Wooldridge Reference Wooldridge2019).Footnote ¹⁷ Formally, for each PTA p, importing country i, and for each crop k,

(3) $$ {Y}_{pik}={\displaystyle \begin{array}{l}{\mathrm{logit}}^{-1}(\alpha +{\beta}_1{Z}_{ik}+{\beta}_2{D}_{pi}+{\beta}_3{Z}_{ik}\times {D}_{pi}\\ {}+\hskip2px {\Delta}^{\top }{X}_{pik}+{\Lambda}^{\top }{\overline{M}}_k),\end{array}} $$

where $ {Y}_{pik} $ is defined in Equation 2, $ {Z}_{ik} $ is the geographic concentration of crop k in country i defined in Equation 1, and $ {D}_{pi} $ is the binary indicator of whether country i has a parliamentary system (= 1) or a presidential system (= 0).Footnote ¹⁸ We rely on the DPI to measure constitutional structures (Scartascini, Cruz, and Keefer Reference Scartascini, Cruz and Keefer2021). We include a set of various potential confounders $ {X}_{pik} $ that vary across PTAs, countries, and/or crops. This includes GDP per capita, population, agricultural population, production volume of each crop, Polity score, land size, and crop-level and total trade volume with the world for both PTA parties, the proportion of non-agricultural products liberalized in each PTA, crop-level and total trade volume with PTA partners and with all countries, average tariff rates, indicator variables for joint WTO and regional trade agreements membership, shared border with PTA partners, constitutional institutions of PTA partners, MFN tariff rates in pre-PTA period, an indicator variable if pre-PTA MFN tariff rate is equal to zero, and crop-level elasticity and differentiation. Table A2 in the Supplementary Material summarizes sources of variables used in this analysis. To account for crop-specific heterogeneity, we include the vector that consists of the mean of each variable for each crop across countries and PTA, $ {\overline{M}}_k $ , such that

(4) $$ \begin{array}{rl}& {\overline{M}}_k:={\left[\begin{array}{l}\qquad {\overline{Z}}_k,\overline{D},{\overline{Z}}_k\times \overline{D},{\overline{X}}_k\qquad \end{array}\right]}^{\top }.\end{array} $$

Our quantity of interest (QOI) is the difference in predicted probabilities of liberalization between parliamentary and presidential systems at different levels of geographic concentration. To account for the varying baseline tendencies toward liberalization between these two regime types, we subtract the average levels of trade liberalization across crops specific to each institutional type. That is:

(5) $$ {\mathrm{QOI}}_z:={\displaystyle \begin{array}{l}\left({\pi}_{\mathrm{Parliamentary},\ z}\hskip0.35em -\hskip0.35em {\overline{\pi}}_{\mathrm{Parliamentary}}\right)\\ {}-\hskip2px \left({\pi}_{\mathrm{Presidential},\ z}\hskip0.35em -\hskip0.35em {\overline{\pi}}_{\mathrm{Presidential}}\right),\end{array}} $$

where $ {\pi}_{c,z} $ is the predicted probability of liberalization for a country with constitutional institution $ c\in \{\mathrm{Parliamentary},\mathrm{Presidential}\} $ and geographic concentration level of z, and $ {\overline{\pi}}_c $ is the mean predicted probability of liberalization for institution c across all crops. The predicted probabilities $ {\pi}_{c,z} $ for each political institution c is computed as follows:

$$ \left\{\begin{array}{l}{\pi}_{\mathrm{Parliamentary},\ z}={\mathrm{logit}}^{-1}\left(\alpha +{\beta}_1z+{\beta}_2+{\beta}_3z+{\Delta}^{\top }x+{\Lambda}^{\top}\overline{m}\right)\\ {}{\pi}_{\mathrm{Presidential},\ z}={\mathrm{logit}}^{-1}\left(\alpha +{\beta}_1z+{\Delta}^{\top }x+{\Lambda}^{\top}\overline{m}\right)\end{array}\right. $$

where x ( $ \overline{m} $ ) indicates the median of (demeaned) covariates included in the model.

To draw a robust statistical inference about QOI, we rely on a Bayesian framework to estimate the parameters of our model. In doing so, we specify the weakly informative prior distribution as follows:

(6) $$ \begin{array}{rl}\beta, \Delta, \Lambda & \hskip-0.5em \sim \hskip0.5em \mathcal{N}(0,{(2.5/{s}_v)}^2)\end{array} $$

(7) $$ \begin{array}{rl}\alpha & \hskip-0.5em \sim \hskip0.5em \mathcal{N}(0,2.{5}^2),\end{array} $$

where $ {s}_v $ stands for the standard deviation of the predictor, v. This prior specification helps stabilize the posterior inference (Gelman, Hill, and Vehtari Reference Gelman, Hill and Vehtari2020) and the default choice for rstanarm package.Footnote ¹⁹ Sampling is conducted with a Markov Chain Monte Carlo (MCMC) method. We run 4 MCMC chains with 1,000 burn-in periods and 1,000 posterior draws each. We then simulate the predicted probabilities of liberalization for each system over different levels of the geographic concentration in our data, while holding other variables to their median values. The uncertainty estimates of QOI come from the posterior distributions of parameters. We verify the convergence with the Gelman–Rubin statistic and confirm the statistic is lower than 1.1 for all the estimated parameters.

Figure 2 shows the estimated difference in the predicted probabilities of liberalization for each political system as a function of the geographic concentration as defined in Equation 5. The positive (negative) values indicate that the predicted probability of liberalization is higher for parliamentary (presidential) systems. The figure reveals that parliamentary systems are more likely to liberalize geographically concentrated products, whereas presidential systems are more likely to liberalize geographically diffused products, consistent with our theoretical expectations. Additionally, the 95% credible intervals do not include zero for both highly concentrated and highly diffused crops, supporting our predictions. Note that the rug plots indicate an even distribution of the moderator variable (log of HHI) across both political systems, ensuring sufficient overlap in observations across a wide range of values.Footnote ²⁰

Figure 2. The Effect of Geographic Concentration on the Differences in Predicted Probabilities of Liberalization Between Parliamentary and Presidential Systems

Note: The horizontal axis represents the level of the geographic concentration, while the vertical axis displays the difference in predicted probabilities between parliamentary and presidential systems, adjusted by their respective average levels of trade liberalization. The positive (negative) difference indicates that the predicted probability of liberalization is higher for parliamentary (presidential) systems. The colored rug plots on the x-axis indicate the distributions of parliamentary (red) and presidential (blue) systems at the different levels of the geographic concentrations.

The estimated effects hold substantive significance. At the 90th percentile of the geographic concentration (i.e., concentrated product), we observe that parliamentary systems are 3.0 percentage points more likely, and presidential systems are 2.2 percentage points less likely, to liberalize crops with the given concentration level, relative to their respective mean predicted probabilities of liberalization. Conversely, at the 10th percentile of the geographic concentration (i.e., diffused product), parliamentary systems exhibit a 2.6 percentage point lower likelihood of liberalization, while presidential systems show a 1.8 percentage point higher likelihood for crops with the given concentration level, compared to their respective mean predicted probabilities of liberalization. Taken together, parliamentary system countries are 5.6 percentage points more likely and presidential system countries are 4.1 percentage points less likely to liberalize concentrated crops compared to diffused crops. This suggests that our findings are not necessarily driven by one system alone, as observed in Section A.3 of the Supplementary Material.Footnote ²¹ Section A.3 of the Supplementary Material shows additional analyses that demonstrate that our results are robust when we include covariates such as party discipline and district magnitude, use alternative prior specifications, exclude Japan or the U.S., and account for potential selection mechanisms arising from pre-PTA subsidy levels.Footnote ²²

Comparative Case Study: Sugar Industry in Japan and the U.S.

The sugarcane industry is an ideal case study because the geographic concentration of the crop in many economies is primarily determined by climate and soil suitability for growth. This makes its concentration largely exogenous to domestic politics. Despite the constitutional differences, sugar production in Japan and the U.S. is geographically concentrated, resulting in limited legislative voting power concerning compensation legislation. The sugar industry in both countries is import-competing and has consistently demanded government protection, which allow us to focus on institutional differences.

As shown in Figure 1, sugarcane production in Japan (left panel) is primarily concentrated in Kagoshima and Okinawa prefectures.Footnote ²⁴ In the U.S., sugarcane and sugar beets are predominantly produced in the states of Florida and Louisiana. The left panel of Figure 3 illustrates this concentration, especially when compared to maize (corn) on the right panel, one of the most widely dispersed crops.Footnote ²⁵

Figure 3. Sugarcane and Maize Production per Grid (Approximately 10 km × 10 km) in the U.S.

Note: The figures show sugarcane and maize production for the U.S. It is visually clear that sugarcane production is more concentrated than the production of maize.

Our theory posits that Japan would liberalize tariffs for geographically concentrated industries, such as sugar, and compensate them with subsidies and party leaders would enforce this contract within a party. Conversely, we expect that the U.S. would protect geographically concentrated industries like sugar with tariffs due to high uncertainty regarding legislating compensation for sugar.

Evidence from Elite Interviews in Japan and the U.S.

This section further provides qualitative evidence from fifteen elite interviews we conducted with top policymakers in Japan and the U.S. We validate our theoretical mechanisms on (1) how constitutional rules shape who negotiates and enforces the contract (party leaders or the legislature) and where the contract is formed and enforced (within a party or in a legislature) and (2) why geographical concentration facilitates dealmaking in parliamentary regimes but geographic diffusion helps dealmaking in presidential systems. Table 1 shows the list of interviewees and Section C of the Supplementary Material details how we chose interviewees and Section D of the Supplementary Material describes how we adhered to the ethical and human subject principles set forth by the APSA.

Table 1. List of Interviewees

Note: Interviews are conducted between September 2018 and June 2021. Positions described in the table above reflect retired, past, or current positions at the time of interviews. We do not specify the position status among the above three to further protect the anonymity of our subjects. Interview 1 is a group interview where interviewees are multiple officials from the ministry. Interviews 3 and 4 and 8 and 9 are different individuals holding the same title.

Parliamentary Systems: Fused Role of Legislators and Enforcement within a Party

The fused roles of party leaders in trade negotiation and domestic compensation became evident during these interviews. In 2013, when the Japanese government commenced TPP negotiations, the Abe Cabinet established the “Headquarter for Countermeasures for TPP,” comprising a “Negotiation Team” and a “Domestic Adjustment Team.” The Negotiation Team led Japan’s TPP negotiations with other governments, while the Domestic Adjustment Team negotiated compensation with domestic industries. The Abe administration appointed senior LDP legislators to both teams, who traveled together for the TPP negotiations. While the Negotiation Team attended international meetings, the Domestic Adjustment Team met with agricultural interest groups that stayed at nearby hotels, highlighting the fused role of party leaders in negotiating a trade agreement and compensation (Nishikawa Reference Nishikawa2017; Interviews 3, 5, 7).

The government’s ability to appoint concurrent legislators to the executive branch helped the Abe administration address the commitment problem. After the Domestic Adjustment Team made informal promises about compensation during trade negotiations, Prime Minister Abe quickly promoted the head of the team (Koya Nishikawa) to the Minister of Agriculture, Forestry and Fisheries (MAFF) upon their return. MAFF oversees budget allocation and regulation for farmers and thus, the Minister can follow through with the informal promises. The next head of the Domestic Adjustment Team (Hiroshi Moriyama) followed the exact same promotion path (Interview 4).

The fused role of party leaders is also evident from the simultaneous process of negotiating trade agreements and domestic compensation revealed in our interviews. Interest groups and legislators engage with the executive branch from the early stages of trade negotiations (Naoi and Urata Reference Naoi and Urata2013), allowing party leaders in both branches to collaboratively design trade liberalization and domestic compensation strategies (Interviews 1, 5, 6, 8, 9). To further quote our interviews with the bureaucrats (Interviews 8, 9):

1. Author(s): “What is the sequence of trade negotiation and negotiation for compensation?”

2. Bureaucrats: “Completely parallel. We present (a policy mix, added by the authors) in one package; here is the proposed tariff reduction and here is the amount of compensation.”

In sum, the evidence from the interviews lends support to our claim that party leaders negotiate both a trade agreement and compensation and that a contract is enforced within a party. Party leaders’ ability to design tariff reduction and compensation simultaneously during trade negotiations alleviates the commitment problems, compared to the sequential process in presidential systems, as we describe in the next section.

The geographic concentration of industries streamlines information gathering and reduces the costs of negotiating compensation packages, as it involves a smaller number of legislators. Our hypothesis is further corroborated by interviews with two Japanese bureaucrats who participated in the negotiation of a bilateral trade agreement (Interviews 8, 9).

1. Bureaucrats: “It is only Hokkaido (where potential trade losers are located, added by the authors), so there will be no problem if we can tame Hokkaido.”

2. Author(s): “How does it help to have only Hokkaido being affected?”

3. Bureaucrats: “A fewer legislators pressure us. Compensation also requires a smaller budget.”

Note that the bureaucrats mentioned the smaller number of legislators (districts) as the primary reason for the ease of “taming” opposition to tariff reduction, rather than the collective action capacity of geographically concentrated industries, which further supports our theory.

By contrast, the geographic diffusion of crops deters the executive branch to even consider trade liberalization. Rice is a good example. When we asked a government official who was responsible for budgeting agricultural compensation, whether they think about the feasibility of tariff reduction or the feasibility of compensation first, s/he responded that the feasibility of compensation was considered first and added that this is precisely why rice cannot be liberalized (Interview 11).

Party leaders serving a dual role of negotiating a trade agreement and the budget also help mitigate information asymmetry problems. The geographic concentration of industries (i.e., a smaller number of concerned legislators) further enables party leaders to invest more time in acquiring accurate information about how much liberalization that incumbents can tolerate and the optimal price for compensation to necessary to stay in the office. In our interview with a LDP legislator who participated in Japan’s negotiation of the Trans-Pacific Partnership Agreement, s/he described the process of vetting legislator demands for compensation (Interview 6):

Furthermore, the Domestic Adjustment Team was led by senior LDP legislators such as Koya Nishikawa and Hiroshi Moriyama, who represented agricultural constituencies and had a history of opposing agricultural liberalization. These appointments of legislators with agricultural expertise to the Domestic Adjustment Team were deliberate efforts to address the information asymmetry problem. Their close relationship with the agricultural sector assisted the executive branch in verifying the legitimacy of legislator demands for compensation (Nishikawa Reference Nishikawa2017).

Presidential Systems: Separation of Power and Enforcement on the Legislative Floor

Our American interviewees include two former American trade policy advisors, a former USTR official who led the negotiation of trade agreements, another USTR official who led Congressional affairs, and a member of the Council of Economic Advisers.Footnote ²⁸ The separation of power in presidential systems means that there is no concurrent legislator appointment at the USTR, the agency that is tasked to negotiate a trade agreement.Footnote ²⁹ Legislators do not just lack a formal role in negotiating a trade agreement. A former USTR official observed much lower legislator participation and initiative to acquire information during trade negotiations compared to the Japanese counterparts. Significantly fewer U.S. legislators, for instance, travel to trade negotiation sites compared to the Japanese counterparts. When USTR officials seek to receive feedback from Congressional members during the active negotiation phase, Congressional members are generally disinterested and not available to talk (Interview 12). There is “a mad rush” to meet with USTR officials, however, after the trade negotiation is complete and before an agreement hits the legislative floor for ratification (see Office of the United States Trade Representative 2015; U.S. Government Accountability Office 2007). As a former USTR official put it: “USTR does not have money to compensate,” so legislators do not invest time to influence compensation (or the content of trade agreements) until a negotiated agreement is about to reach the legislative floor for ratification (Interview 12). This sequential nature of negotiating a trade agreement and then compensation is in contrast to the simultaneous process reported by the Japanese interviewees. The difference is a direct consequence of the constitutional structure that defines a separate and independent role of the executive and legislative branches.

The geographic diffusion of industries facilitates mobilizing votes for ratification and compensation because a majority threshold binds a compensation contract on a legislative floor. An illustrative example is the Obama administration’s promise to increase funding for the beef industry to mobilize majority support for ratification of the Korea-U.S. Trade Agreement (KORUS). Cattle production is geographically diffused in the U.S., where the top 10 states constitute only 57% of total beef production. Despite the fact that beef was projected to be one of the largest beneficiaries of KORUS with improved access to the Korean market, beef still received USDA funding to “compensate” for receiving not enough market access with KORUS. This funding, in conjunction with the appointment of Montana senator Max Baucus as the Chair of the Senate Committee on Finance where reciprocal trade agreement legislation is reviewed (Interview 12), illustrates how mobilizing support from diffused industries is key to the successful ratification of trade agreements in presidential systems.

Non-overlapping appointments exacerbate the information asymmetry problem that the executive faces in vetting truthful information from legislators. When we asked how the USTR or executive branch would predict which legislators would cast a Nay vote, three of our interviewees said it was the past record of roll-call votes on trade bills—publicly available information—that the USTR found the most informative (Interviews 13, 14, 15). One of the interviewees said past roll-call votes, especially regarding Trade Promotion Authority, helped them predict (Interview 12).Footnote ³⁰ The USTR’s prediction of roll-call votes is more uncertain with the entry of new Congressional members who lack past record of roll-call votes (Interview 12). This uncertainty has become an issue for KORUS ratification vote prediction, which led to the Obama administration renegotiating the agreement and adding extra deals such as beef funding to mobilize ratification support from Congress. The USTR rarely look into legislators’ district characteristics or record of receiving political donations (Interview 14). This information gathering process differs from the detailed and personal process described by party leaders in Japan (Interview 6).

In sum, our elite interviews have verified that party leaders design trade liberalization and compensation simultaneously and enforce them within a party in parliamentary systems, while a legislature ties the fate of ratification with compensation with a majority threshold in presidential systems. Moreover, the geographic concentration facilitates a compensation contract to form and be enforced within a party in parliamentary systems, while geographic diffusion helps legislators build a majority coalition in the legislature in presidential systems.