Electoral institutions define essential rules of the game in democratic politics. An enormous literature in political science has examined how they affect elections, policy, and political representation.Footnote 1 Thus, scholars of international political economy argue that electoral institutions have important effects on trade policies and other types of foreign policy. For instance, they can shape trade barriers, subsidies, and immigration policy.Footnote 2 They also influence the extent to which governments compensate trade losers and pursue income redistribution,Footnote 3 which may moderate the backlash against globalization.Footnote 4
Given that electoral institutions are themselves the result of political choice, how can the politics of electoral system choice be explained? Studies on how electoral institutions shape trade policy imply that such institutions are endogenous to domestic conflict over international trade. Yet theories emphasizing economic explanations of electoral institutions are controversial, and the evidence is limited. A prolific literature has examined the politics surrounding the choice of electoral systems. It has paid special attention to domestic explanations of why several European countries adopted proportional representation (PR) in the late nineteenth century and the first half of the twentieth century.Footnote 5 We reconsider the role of international trade as an economic explanation of the politics of electoral system choice during the first globalization (from the 1870s to 1914). We argue that political conflict over trade policy shaped the political struggle over electoral system reform in important ways that prior work has overlooked.
The idea that international trade affects countries’ choice of electoral system is not new. Early contributions put forth the idea that there is a “natural affinity” between free trade and proportional electoral systems.Footnote 6 However, recent research on the origins of PR has paid less attention to trade, and credible empirical tests are scant. By shifting from cross-national analyses using small samples to evaluating legislative votes and elite positions on electoral reform, recent studies have made tremendous progress in demonstrating the importance of factors related to inter-partyFootnote 7 and intra-party competition.Footnote 8 But they have largely failed to address the possibility that trade shapes conflict over electoral rules.
At the core of the trade-based theory is the notion that economic interests have preferences regarding electoral institutions that are partly based on expectations about whether these institutions affect trade policy. Trade was the principal—and hotly contested—instrument of economic policy at a time when European countries were first debating the introduction of PR.Footnote 9 Thus, interests that are hurt by rising tariffs enacted under the status quo majoritarian electoral system should have economic motives to support the introduction of PR. Those benefiting from protection, by contrast, have reasons to oppose changing the electoral system. This theory highlights that among ordinary people and organized interests, who do not compete for office but mobilize (or are mobilized) in elections, divisions over trade policy can affect divisions over electoral reform. This matters because electoral reforms do not take place in a social vacuum. Some reform proposals are subject to a referendum, as in our main empirical case, (historical) Switzerland, as well as recent reform debates in Italy, New Zealand, and the United Kingdom. Electoral reform can also become an issue in electionsFootnote 10 or mass protests during crises.Footnote 11
Testing the trade perspective on electoral reform has proven challenging. As already noted by Ronald Rogowski,Footnote 12 political accounts of the introduction of PR tend to focus on elections and parliamentary elites, making it easy to overlook the potential relevance of the trade cleavage.Footnote 13 The fact that several reforms took place in times of political crisis may further obscure the role of longer-run economic factors. Using cross-national data makes it difficult to rule out alternative explanations.Footnote 14 Hence, we follow much of the recent scholarship on electoral reform in turning to within-country studies using more fine-grained data. We use district-level data on popular votes during the first era of globalization to address two important challenges. The first is that data on mass support for electoral reform are generally not available until the second half of the twentieth century. The second challenge is disentangling the effects of trade interests versus partisan politics.
To address these challenges, our main analysis uses canton-level data from Switzerland: popular votes on trade policy, the adoption of PR, and other issues. The focus on within-canton variation and the availability of detailed data from the population and firm census enable us to control for a range of alternative explanations, such as the role of issue bundling, inequality, and skill-based economic institutions.
Our first main finding is that support for the omnibus tariff law of 1903, which increased protection for many goods, is tightly linked to opposition to the introduction of PR in two subsequent referendums (1910 and 1918). This supports the theory’s central proposal that preferences related to trade policy significantly shape preferences regarding electoral institutions.
Our second main finding is based on a panel analysis, which enables us to account for district and period fixed effects, canton-specific trends, and flexible time-varying controls. The general argument is agnostic as to why people want protection. Historical accounts and the factor endowments model of trade suggest that relatively scarce agriculture was a central pillar of support for protectionism, with protectionist farmers pitted against workers and urban consumers benefiting from free trade. According to the theory, the agricultural cleavage should also explain disagreements on electoral reform. We establish that changes in the share of the agricultural population are approximately one-to-one related to changes in support for PR.
In a secondary analysis, we turn to Germany and find some evidence that the theory applies more broadly. In the absence of mass data, we use legislative votes. We find that support for the general tariff of 1902 in the Reichstag predicts opposition to the (partial) introduction of PR in 1918.
Altogether, we bring trade back into the debate on the politics of electoral system choice. Our analysis addresses the critique that economic explanations lack direct evidence of a link between economic interests and preferences on electoral rules.Footnote 15 We demonstrate that conflict over trade shaped the mass politics of electoral reform during the first globalization, challenging the view that “economic explanations cannot account for the origins of PR.”Footnote 16 By underscoring that the rules of the game can be endogenous to conflicts over trade, we show they are also relevant to the literature on how electoral rules affect trade policy.Footnote 17
Some studies of electoral reform that do not explicitly focus on trade note its salience.Footnote 18 And one within-country study, also on Switzerland but focused on legislative votes, uses a proxy for industries’ positions on the tariff as a control for trade interests.Footnote 19 While it finds no industry effects, this is not inconsistent with our findings. We focus on how support for tariffs, measured directly from popular votes, shapes mass support for PR. Our evidence underlines the importance of the agricultural cleavage over trade. Our findings demonstrate that mass preferences over trade, not just strategies of parliamentary elites, can play a role in driving electoral reform.
Theoretical Considerations
Here, we develop empirical implications about how conflict over trade policy may shape the politics of electoral reform and explain how they relate to important theoretical debates in international and comparative political economy. In an article published almost forty years ago in this journal, Ronald Rogowski argued that there is a natural affinity between trade and the choice of PR.Footnote 20 He theorizes that a welfare-maximizing state will choose its electoral institutions based on the extent of its dependence on trade. He argues that electoral institutions influence domestic conflicts over countries’ trade policy. Even if free trade generates aggregate gains, some sectors, regions, and firms will want protection.Footnote 21 Because collective-action problems and distributive politics influence political competition, there is no guarantee that political equilibrium achieves the first-best trade policy.Footnote 22
According to Rogowski’s trade theory of electoral institutions, electoral rules shape whether democratic policy makers are sufficiently isolated and autonomous from protectionist pressures and have incentives to pursue stable policies. Thus, beliefs about the trade effects of electoral institutions shape the choice of the rules themselves. Subsequent scholarship has analyzed game-theoretically and empirically how electoral rules impact trade policy. While some theories suggest that certain electoral institutions are generally more likely to produce protectionist policies than others,Footnote 23 an equally common line of thought is that the interaction of interests and electoral institutions shapes trade policy.Footnote 24 The latter implies that electoral institutions do not have uniform effects across contexts.Footnote 25
Trade Interests and Electoral Reform
We build on these analytical foundations to analyze how distributive conflict over trade shapes the domestic politics of electoral system reform. Where trade is a salient policy issue, as it was during the first wave of democratization, we argue that the calculus of electoral system choice is shaped by whether economic interests—free traders versus protectionists—can expect better parliamentary representation (and thus more influence over trade policy) under PR compared to alternative systems. That is, citizens who want to lower tariffs should prefer the electoral institution that they believe benefits the political party representing their trade interests. We deliberately do not test the thesis that “the more an economically advanced state relies on external trade, the more it will be drawn to the use of PR, a parliamentary system, and large districts.”Footnote 26 We focus on an equally important (and causally prior) issue.
The core of the institutional logic is agnostic as to why people are for or against trade. Foundational international political economy models maintain that who wants protection is a function of factor endowments and industry of employment.Footnote 27 The factor endowments model of international trade posits that protection benefits (and trade harms) owners of factors with which a given country is scarcely endowed (relative to the rest of the world), as well as producers who use the scarce factor intensively. By contrast, protection harms (and trade benefits) owners of factors with which the country is abundantly endowed, and producers that use them intensively. In sectoral models, policy preferences stem from an individual’s industry of employment: those who work in exporting industries (industries facing strong import competition) will want free trade (protection).Footnote 28
Standard accounts of trade politics in Europe during the first globalization emphasize the relevance of factoral cleavages.Footnote 29 In many labor-abundant and land-scarce European economies at the turn of the twentieth century, conservative and farmers’ parties representing agricultural interests strongly supported protectionism; the emerging left parties aggregated worker demands to remove protectionist barriers on agricultural products.Footnote 30 Industrial workers also consumed agricultural products, and thus had additional reasons to want freer trade as food prices rose. Of course, it is unlikely that “any particular approach monopolizes the truth on these issues.”Footnote 31 Multiple mechanisms can be at play.
Which electoral system should those favoring free trade support? If the status quo majoritarian system is biased against a group’s trade policy preferences—as it was in several Western European countries at the time—PR is the preferred alternative. In PR systems, group size is key for parliamentary representation. A proportional electoral formula applied to multi-member districts tends to generate a stronger relationship between votes and seats, which tightens as district magnitude increases.Footnote 32 In majoritarian systems, by contrast, strategic voting and electoral geography more easily undermine the link between group size and political representation.Footnote 33 Partisan control over the drawing of the district map intensifies unequal parliamentary representation in majoritarian systems through gerrymandering or malapportionment.Footnote 34 Thus, two groups of similar size can have dramatically different parliamentary representation. Evidence from this period indicates that introducing PR can boost the representation of broad-based interests that were underrepresented under the old electoral system.Footnote 35
The literature also emphasizes that PR fosters inter-party bargaining. If no party controls a majority of seats, trade policy is determined in post-election negotiations between parties.Footnote 36 In Western Europe at the turn of the twentieth century, no economic group on its own (such as farmers, manual workers, or employees in the tertiary sector) could plausibly win a majority under a PR system.Footnote 37 Thus, parties representing worker and consumer interests might negotiate with representatives of the middle classes in commerce and industry, or seek a compromise with farmers. Ultimately, supporting the introduction of PR will be preferable to the majoritarian status quo if incumbents use it to lock in the protectionist status quo.
In short, people who do not obtain their preferred trade policy have instrumental reasons to support an electoral reform that gives them more political weight. We test the following hypothesis in a majoritarian electoral system where free trade supporters are underrepresented: Greater political support for free trade policies among voters increases support for introducing PR.
We now briefly discuss how our argument relates to three countervailing mechanisms invoked in other accounts. First, trade policy preferences need not reflect well-defined interests; and even when they do, they may not affect support for electoral reforms. An important body of survey-based work on contemporary trade preferences argues that voters do not have self-interested preferences.Footnote 38 Our reading of the evidence is that both material and nonmaterial motivations affect preference formation.Footnote 39 Trade was a politically salient issue during our period of study. Studies emphasize that farmers and workers, as well as the parties representing them, cared a lot about tariffs.Footnote 40
Second, trade preferences may not matter because of the uncertainty associated with electoral reform, especially before the advent of modern opinion polls.Footnote 41 We assume that expectations are formed using the available information. For example, historical debates reveal attention to reforms in other countries and at the subnational level, and newspapers published statistical simulations of results under alternative rules.
Third, some models of electoral reform assume that politics is one-dimensional, and thus that all issues can be summarized along a left–right ideological dimension.Footnote 42 This suggests that trade policy will be closely bundled with other issues, such as social insurance or the state’s role in education, making it difficult to identify a systematic relationship between trade policy preferences and support for electoral reform when controlling for other salient issues. Our empirical strategy addresses this possibility.
Contesting Tariffs and Electoral Institutions: Switzerland in the First Globalization
We test the empirical implications of our theory in the first globalization, during a period of rising protectionism. Given the theoretical focus on economic interests rather than parliamentary elites, a hard test ought to consider mass behavior outside the magnificent parliamentary buildings often erected at this time throughout Europe.Footnote 43 Switzerland is an especially suitable test case: like its European neighbors, it experienced intense conflict over both trade and electoral institutions from the 1880s until the end of World War I. The fact that it held popular votes on tariffs and electoral reform enables us to measure mobilized mass support without distortions of the electoral system and without strategic voting.
We draw on district-level data on popular votes and district characteristics from the decennial population census and Switzerland’s first firm census in 1905. At the time, it had twenty-five cantons (states) with considerable policy autonomy. Cantons were further divided into 186 administrative regions that we call districts (district in French and Bezirk in German). Importantly, the borders of these administrative districts were stable and difficult to change—unlike electoral districts, which were subject to gerrymandering. The comprehensive data set compiled by Wolf Linder and colleagues contains historical district-level data on popular votes and population census data.Footnote 44 We also coded data from the 1905 firm census to measure relevant variables concerning the structure of agriculture, industry, and skill formation.
The Protectionist Turn and the Struggle over PR
Important changes in the international political economy at the end of the nineteenth century paved the way for a protectionist turn in Europe, which in turn fueled conflict over electoral institutions. The decline in transportation costs thanks to steamships and the expansion of railroad networks exposed domestic producers to competition in commodities such as grain and textiles.Footnote 45 The subsequent rise of agricultural tariffs in Germany and other European countries from the early 1880s reduced the market for Swiss producers and, along with the depression of 1870, drove up support for protectionism in the agricultural sector.Footnote 46
In a context of rising protectionism, the federal government, dominated by the liberal Free Democratic Party (Freisinn), proposed a sweeping new tariff law increasing tariffs on “a large part”Footnote 47 of imported goods.Footnote 48 The proposal increased tariffs on 48 percent of all 1,113 included items, left 39 percent unchanged, and included mixed changes or reductions in other categories.Footnote 49 It entailed a strong increase in agricultural protection.Footnote 50
The Free Democratic Party was the dominant party from the 1880s until the adoption of PR and represented a mix of agricultural, commercial, and industrial interests. The tariff was supported by the Catholic-Conservative Party and bitterly opposed by the younger Social Democratic Party. The Social Democratic Party opposed the new tariff in the name of labor and consumer interests, similar to labor’s position in most of Western Europe.Footnote 51 Beyond possible income effects, workers benefited from free trade through lower consumer prices on food and industrial products.
After both chambers of parliament passed the law, opponents tried to stop it by calling for a popular vote—the referendum on the general tariff—in 1903. It is not hard to understand why. Switzerland’s economy was relatively abundant in labor and capital and scarce in land, a situation conducive to conflict between farmers demanding more protection and workers and (some) producers opposing it.Footnote 52 The tariff raised protection for high-value-added agricultural products in which Swiss farmers had specialized, and raised consumer prices on food. While there were some large estates, family farming was common.Footnote 53 The passage of the tariff bill was seen as a “great victory” for agricultural interests.Footnote 54 As a result, by 1913 tariffs in Switzerland stood at about 15 percent for foodstuffs and at 9 percent for industrial manufactured goods—considerably higher than in free-trade Britain and just a bit lower than in Germany.Footnote 55
Using regression analysis, we confirm the relevance of agriculture for the protectionist tariff (see Appendix B.1 in the online supplement). There is a strong positive correlation between the share of the agricultural workforce in a district and popular support for protection. The relationship is robust to accounting for canton fixed effects as well as a host of district-level controls (supplemental Table B.1). These results are consistent with accounts that emphasize factoral cleavages over the trade issue during most of the nineteenth and early twentieth centuries.Footnote 56 At the mass level (where capitalists are a small minority), we find that the main cleavage was between scarce agriculture and abundant labor. Of course, no theoretical model is a complete description of the world. In additional analysis, we engage with implications of sectoral models (supplemental Table B.2).
Conflict over trade spilled over into conflict over electoral reform. Contemporaries linked trade to the introduction of PR, arguing that the proposed electoral reform would affect tariff policy. For example, a week before the popular vote on PR, a trade-union newspaper ran a “Plea to Swiss Citizens” on its front page, calling on workers to vote in favor of introducing PR. It meticulously linked unequal political representation in the majoritarian system to unequal substantive representation on economic issues, prominently the tariff:
The Chambers, deceiving the people, have imposed on us a customs tariff that has made life considerably more expensive…. We have a great interest in having as many representatives of the working class party as possible in the National Council…. The more representatives of the working class there are in the Assembly, the more these laws will be impregnated with our spirit, will be in conformity with our interests and needs; the better they will be applied. The Swiss Socialist Party, the only party that defends our interests without calculation and without restriction, despite its … voters and although the working class is the most important class in Switzerland, has only 7 representatives in the National Council, out of 195. (Le Prolétaire, 15 October 1910, our translation. For the original text and more examples, see Appendix C.)
As in several other countries at the time (and until 1919), Switzerland elected its lower house (the National Council of the Federal Assembly) using a majoritarian electoral system with multi-member districts and runoff elections, with the last runoff decided by plurality rule.Footnote 57 The Free Democratic Party gerrymandered its parliamentary majority in the face of rapid industrialization and the rise of the working class and the political left, and opposed the introduction of PR.Footnote 58 The Social Democratic Party was thus severely underrepresented in the lower chamber of parliament.
Trade and Popular Support for PR
We begin by testing the link between mass support for protection and mass support for PR to explore the core mechanism of the trade account: conflict over trade policy induces conflict over electoral reform. It took Switzerland three attempts to adopt PR from below in a referendum (1900, 1910, and 1918). We focus on the two popular votes that occurred after the political debate and the vote on the general tariff in 1902–1903, which crystallized preferences on trade policy and mobilized interests around the issue to a much greater degree.Footnote 59
The year 1910 saw an attempt to introduce PR through a referendum over the opposition of the bicameral legislature dominated by the Free Democratic Party.Footnote 60 Under the proposal, each canton would constitute an electoral district, effectively doubling the average district’s magnitude (number of seats); the legislature would also lose its power to gerrymander electoral districts. Lawmakers rejected the proposal, but following popular initiative, it was put to a referendum. The Social Democrats campaigned in favor of PR; the Free Democratic Party campaigned against it.Footnote 61 The outcome was close: with 47.5 percent in favor, the majoritarian system remained in place. But PR supporters learned that a future victory would require only a small gain in support. They eventually won a majority in the 1918 referendum.
Cross-Sectional Analysis
To systematically test our main implication, we use data from sequential popular votes on the protectionist general tariff and the introduction of PR. Our baseline model takes the form:
$$V_i^{{\rm{PR}}} = {\beta _1}V_i^{\rm{T}} + f\left( {{X_i}} \right){\beta _2} + {\delta _c} + {\varepsilon _i}$$
where
$$V_i^{{\rm{PR}}}$$
is the vote share in favor of introducing PR in district
$$i$$
in 1910 or 1918,
$$V_i^{\rm{T}}$$
is the prior support for protectionism in the 1903 vote on the general tariff,
$${X_i}$$
is a matrix of district-level controls and
$${\delta _c}$$
denotes canton fixed effects. The within-country research design holds country-level variables constant to account for the possibility that small countries may be inclined to support a particular combination of trade policy and electoral institutions due to critical junctures or other unobserved factors.Footnote
62
Using district-level data allows us to use canton fixed effects, which help partially account for several explanations offered in the literature (such as differences in institutions, party systems, socialist threat, and gerrymandering across cantons).Footnote
63
As we have mentioned, one-dimensional models of the politics of electoral reform introduce the possibility that support for protectionism is perfectly bundled with other issues on a left–right dimension.Footnote 64 To address this possibility, we control for district-level votes on other issues. We extract the first-dimension factor score from all other popular votes between 1897 and 1902, excluding protectionism and PR. In supplemental analyses, we show that the results are robust to using an alternative measure based on party endorsements (Table B.4) and including all other votes individually (Table B.5).
An alternative economic explanation emphasizes (land) inequality,Footnote 65 which has also been linked to elites’ position on the tariff.Footnote 66 We use three variables to tap into the heterogeneity of agricultural interests. We coded the district-level tables based on data from the first Swiss firm census (known as the census of economic establishments) in 1905 and calculated the Gini coefficient of land inequality. To capture additional differences in the structure of agriculture, we also include the average farm size and the share of agricultural land used for meadows and pasture.
Another economic explanation focuses on how economic institutions enhance skill formation and thereby foster an inter-class coalition that prefers PR as a commitment device.Footnote 67 While these economic institutions mostly vary at higher levels of aggregation, which our within-country research design and canton fixed effects account for, we nonetheless include a district-level proxy for skill formation to capture any remaining variation. We calculate the share of apprentices attending a vocational school. We also include data on the composition of the industrial sector from the firm census (ten subgroups), which capture potential industry-level variation in skills as well as industry-varying demand for protectionism.Footnote 68 Since geographic areas with diverse languages and religions may vary in their support for PR, we include district-level controls from the 1900 population census, including the (logged) population size, share of German speakers, share of Protestants, and language and religious fractionalization.Footnote 69
An additional model relaxes the assumption that the control variables enter the specification in a linear way. To enable arbitrary interactions between all district-level controls (including quadratic terms) with sparse data, we use double-selection lasso linear regression (DSLR).Footnote 70 This and other elements of our empirical strategy are designed to rule out alternative explanations of the effects of theoretical interest, not to conduct a horse race between variables.Footnote 71
Table 1 displays the main estimation results. All specifications related to the 1910 vote to introduce PR (models 1–4) display a negative relationship between district-level support for protectionism and subsequent support for PR that is statistically significant at the 5 percent level. The specification in model 1 establishes that, accounting for canton fixed effects, a one point increase in the protectionist vote is associated with a 0.5-point decline in the vote for PR. The estimated coefficient hardly changes when adding the district-level control for votes on other issues (model 2). This indicates that issue bundling is not driving the relationship between trade and PR. We can thus distinguish trade from other issues that were salient at the time. Entering all controls (model 3) and allowing for flexible functional forms (model 4) somewhat reduces the trade coefficient, but it remains substantively and statistically significant. In substantive terms, the estimate of –0.32 from the most flexible statistical specification (model 4) suggests that increasing support for protectionism by ten percentage points is associated with a three percentage-point decline, on average, in support for PR, plus or minus one point.
Table 1. Protectionism and the vote to introduce PR
Notes: Coefficient estimates from district-level ordinary least squares (OLS) regressions (columns 1–3, 5–7) and DSLR (columns 4, 8), which permits interactions between all district controls as well as quadratic terms (it selects 26–28 out of 255 possible control terms). The dependent variable is the vote share in favor of introducing PR in the 1910 (columns 1–4) or 1918 (columns 5–8) referendum. Robust (heteroskedastic-consistent) standard errors in parentheses.
1 Log of population, German speakers (share), Protestants (share), language fractionalization, religious fractionalization (from 1900 census).
2 Land inequality (Gini), pastureland (share), average farm size (calculated from the 1905 firm census).
3 Composition of industrial employment (10 subsectors, textiles is baseline), share of women (calculated from the 1905 firm census).
4 Share of apprentices attending vocational school (calculated from the 1905 firm census).
5 First-dimension factor score from other popular votes 1897–1902 (excluding tariff and PR).
*
$$p \lt 0.05$$
(two-tailed tests).
The estimates for the 1918 vote on the introduction of PR, shown in models 5 to 8 of Table 1, yield the same picture. The coefficient on the trade variable is of comparable magnitude and statistically significant in three of the four specifications. While not significant in the saturated OLS model, in the more general DSLR model the effect of protectionist support is statistically significant and substantively identical to the one in the 1910 vote.
To mitigate omitted-variable bias, DSLR optimally selects control variables based on their relevance for both the treatment (vote on trade) variable and the outcome (votes on PR), enabling interactions and higher-order terms (see Appendix B.2). Given the large number of potential confounders relative to the sample size, naive OLS can limit our ability to draw valid inferences due to overfitting and low precision.Footnote 72 Model 7 includes forty-five controls but cannot accommodate the large number of possible interactions, which may lead to (attenuation) bias. DSLR mitigates these issues by regularizing the variable-selection process, effectively balancing the trade-off between bias and variance.
These results also indicate that disagreements over trade policy before the war had an enduring impact on mass support for electoral reform. They suggest that considering international trade complements explanations focused on political events. While Switzerland did not fight in the war, it mass mobilized its army and was exposed to the war’s economic and political effects, notably inflation and mass strikes. Some studies argue that this experience softened some incumbent elites’ willingness to fight PR and perhaps made some voters more accommodating of electoral reform.Footnote 73 We do not dispute this. In contrast to the trade account, however, the war on its own does not explain why some people supported the reform while others opposed it, or why agriculture was a persistent cleavage.
There are no individual-level voting data for the population of eligible voters in our study period. Thus, evaluating individual- or group-level predictors of mass voting behavior requires using district-level data, which entails some degree of ecological inference. This remains true for studies of election results, due to the secret ballot.Footnote 74 While we cannot entirely rule out ecological fallacies, they are unlikely to drive our main conclusions, for four reasons.
First, vote choice in referendums is simpler than voting in multi-party elections in Europe, which has been a focus of controversial debates.Footnote 75 Second, our analysis accounts for state fixed effects as well as a large set of district-level confounders, under flexible functional form assumptions. This mitigates the concern that the observed relationship masks omitted individual-level variables. Third, one may conceive of ecological inference as a measurement problem in which the key explanatory variable is measured with noise, which leads to attenuation bias. One standard approach to such an errors-in-variables problem is to use an instrumental variable. While valid instruments are hard to find, in an additional analysis we build on standard (factoral) trade theory and instrument the protectionist vote with the share of the agricultural workforce (Table B.6 and Figure B.2). Fourth, where data are available, we replicate our main analysis at the municipal level (Table B.7).
Panel Analysis
Using panel data, we take an additional step to rule out confounding based on unobserved district heterogeneity. The panel analysis leverages the fact that after the passage of the general tariff there were two popular votes on the introduction of PR, in 1910 and 1918. The trade perspective implies that a decline (growth) in the size of the agricultural sector between 1910 and 1918 should generate growth (decline) in support for PR during the same period, beyond nation-wide or canton-specific swings in support for PR.
The panel analysis does not estimate the same quantity of interest as the cross-sectional analysis, which focused on the link between mass support for trade policy and support for electoral reform.Footnote 76 It tests an additional implication of the trade perspective based on the fact that agriculture was a pillar of support for protectionism at this time (see Appendix B.1).
The baseline two-way fixed effects panel model we estimate is
$$V_{it}^{{\rm{PR}}} = {\alpha _i}{ + \lambda_t} + {\theta _1}{A_{it}} + f\left( {{X_{it}}} \right){\theta _2} + {\varepsilon _{it}}$$
where
$$V_{it}^{{\rm{PR}}}$$
is the vote share in favor of introducing PR in district
$$i$$
in period
$$t$$
(1910 or 1918) and the main time-varying explanatory variable of interest is the district’s share of the agricultural workforce,
$${A_{it}}$$
. The equation includes fixed effects for districts (
$${\alpha _i}$$
), a common time effect (
$${\lambda _t}$$
) and other time-varying district-level characteristics from the closest population census (
$${X_{it}}$$
). This specification allows for sorting into agriculture based on fixed unobservable characteristics; the period effect captures events such as World War I.
The baseline specification assumes that there are no unobserved changes within districts that explain changes in support for PR (net of time-varying controls and time trends). In an extended specification, we relax this assumption and add canton-specific time trends to account for heterogeneity stemming from factors such as canton-level changes in electoral disproportionality and socialist threat,Footnote 77 canton-level electoral reforms,Footnote 78 and varying exposure to mobilization during the war. We also allow districts to have varying trends based on the rate of change in the agricultural sector between 1900 and 1910. As before, we use DSLR regression, which always includes two-way fixed effects for district and time and permits interactions between all time-variant district-level controls.
Table 2 provides the estimates from the panel analysis. All models include canton and period fixed effects. We sequentially include canton-specific period effects and time-varying district-level controls from the 1910 and 1920 population census. The most flexible specification (model 5) allows interactions between all district controls. Across specifications, a one percentage-point decline in the agricultural workforce corresponds to an approximately one percentage-point increase in support for PR, with a standard error of about one-third of a point. There is a tight co-movement between agricultural change and increasing support for PR—beyond national and cantonal trends, and net of changes in observed district characteristics. These results bolster the idea that economic interests strongly linked to trade policy shape mass support for electoral reform.
Table 2. Panel evidence on the link between agriculture and the PR vote
Notes: Coefficient estimates from two-period panel regressions with district and year fixed effects estimated using OLS (columns 1–4) and DSLR (column 5), which permits interactions between all district controls. The dependent variable is the vote share in favor of introducing PR in each year. Standard errors in parentheses clustered at the district level.
1 Interaction of time and district-level change in agricultural workforce, 1900–1910.
2 From closest population census (1910, 1920): log of population, German speakers (share), Protestants (share), language and religious fractionalization.
*
$$p \lt 0.05$$
(two-tailed tests).
A potential alternative explanation for these results is that people in more agricultural districts simply prefer the status quo on these issues (after supporting a significant change in the status quo policy in the 1903 tariff vote). However, looking at other popular votes reveals that agriculture was not generally against change. For example, an additional analysis shows a positive and significant relationship between the share of the agricultural workforce and support for policy change on other issues, such as regulation of food products or military reform (supplemental Figure B.3).Footnote 79
Trade and PR in Imperial Germany
To illustrate the broader scope of the theory, we examine the case of Germany, which features prominently in the largely separate literatures on electoral reformFootnote 80 and the political economy of tariffs.Footnote 81 After unification in 1871, Germany’s lower house of parliament (the Reichstag) was elected under universal male suffrage in 397 single-member districts. In the hybrid political regime, the government appointed by the emperor required parliamentary majorities in the Reichstag to make policies considered key to Germany’s staying economically and politically competitive, such as budgets and tariffs. Tariff policy and electoral reform were two central issues on the political agenda.
Germany’s majoritarian electoral system was severely malapportioned. Despite rapid industrialization and population growth in cities and industrial areas, the map and magnitude of electoral districts were not changed between 1871 and 1918. Regardless of their population, each district elected one representative, which led to the overrepresentation of agricultural districts. As in Switzerland, agriculture tended to be protectionist. The 1879 tariff and subsequent tariff increases benefited large East Elbian grain producers as well as middle-class farmers west of the Elbe specializing in livestock farming, especially pork, to serve the increased demand from rapidly industrializing areas.Footnote 82 By 1890, agricultural interests organized by the powerful Agricultural League had “enough votes in the Reichstag and Prussian Landtag to block any liberalization of the tariff policy.”Footnote 83 Agricultural interests—and the Conservative Party that represented them—opposed electoral reform because they benefited from the status quo. The Social Democratic Party, which represented workers’ and urban consumers’ interests, demanded the abolition of tariffs on agricultural products and the introduction of PR (for example, in its 1891 program).
Since no data are available for popular votes, we cannot directly assess the trade–PR linkage at the mass level. As a second-best strategy, we use roll-call votes in the Reichstag. Leveraging the stable district boundaries, we explore whether legislative support for the protectionist tariff of 1902 predicts legislative opposition to the electoral reform of 1918 within districts. The general tariff law of 1902 signaled a “resumption of high agricultural tariffs.”Footnote 84 In 1917 and 1918, the national parliament debated and passed a partial electoral reform that introduced PR in large (that is, populous) districts that suffered from malapportionment.Footnote 85
The parliamentary debate on PR illustrates the potential relevance of trade. For example, in July 1918 Kuno Graf von Westarp, an aristocratic deputy from Prussia and member of the Conservative Party, justified his party’s rejection of the electoral reform because it was expected to weaken the representation of rural interests, which demanded protection, at the expense of urban consumers and workers:
In our opinion, there is no compelling reason to make such a far-reaching change to the current electoral law ... The main advantage of this law will accrue to the Social Democrats, but above all it will result in the metropolitan industrial consumer circles being strengthened in the Reichstag ... The rural districts will then see that this law also prevents the necessary protection of agriculture.Footnote 86
Table 3 reports the results of a regression analysis using roll-call votes on the protectionist tariff of 1902 and the electoral reform of 1918 in the Reichstag.Footnote 87 The dependent variable is a dummy capturing a vote in favor of introducing PR in large districts.Footnote 88 The explanatory variable of theoretical interest is a dummy variable capturing whether the deputy representing the same district voted for the tariff increase of 1902.Footnote 89 The theory implies that we should observe a negative relationship between voting for the tariff and the probability of voting against the electoral reform.
Table 3. Legislative votes on tariffs and electoral reform in the German Reichstag
Notes: Coefficient estimates from OLS regressions at the electoral district level. The dependent variable is a vote in favor of introducing PR in large districts in 1918. Robust (heteroskedastic-consistent) standard errors in parentheses.
1 Log of population, land inequality (Gini), skill ratio, religious fractionalization.
2 Region fixed effects for East Prussia, West Prussia, Bavaria, Saxony, Württemberg, Baden, Hesse, Elsaß-Lothringen, other.
*
$$p \lt 0.05$$
(two-tailed tests).
The estimation results are consistent with this hypothesis. In the specification without control variables (model 1), a protectionist vote is associated with approximately thirty-five percentage points less support for electoral reform. Adding exogenous district characteristics such as population size, land inequality, and skills (model 2), and fixed effects for regions (model 3), somewhat reduces the estimated coefficient, but it remains substantively and statistically significant.
The results from Germany suggest that trade is relevant for understanding political conflict over electoral reform beyond the Swiss case. Extending the analysis to other cases is beyond the scope of the article. We leverage the advantages of within-country designs to rule out alternative explanations, but we acknowledge that this comes at the expense of providing a comprehensive cross-national picture.
Conclusion
Does trade affect the politics of electoral system choice? We demonstrate that conflict over international trade can be a deeper cause of the historical struggle over the adoption of PR. Spurred by Rogowski’s seminal study, we provide evidence that conflict over trade policy shaped conflict over electoral institutions during the first globalization. While most prior work in this area has focused on domestic explanations, we reconsider the role of international trade. Our results help make sense of how two key political debates, which are usually the subject of separate literatures, are related. These findings matter because historical reforms often locked in electoral institutions that arguably shaped political and economic outcomes for decades, if not more. Thus, the results shed new light on the longer, “seemingly improbable”Footnote 90 causal chains connecting the political economy of trade with institutional choice.
Data Availability Statement
Replication files for this research note may be found at <https://doi.org/10.7910/DVN/Q1VSUZ>.
Supplementary Material
Supplementary material for this research note is available at <https://doi.org/10.1017/S0020818325000104>.
Acknowledgments
A previous version of this paper was circulated under the title “Trade Origins of Proportional Representation” and won APSA’s Fiona McGillivray Award for best political economy paper in 2023. We thank Amel Ahmed, Pablo Beramendi, Patrick Emmenegger, Jeffry A. Frieden, Miriam Golden, Erica Owen, Evgeny Postnikov, Pedro Riera, Ron Rogowski, Ken Scheve, Theo Serlin, André Walter, and the anonymous reviewers, as well as participants at APSA 2022, EPSA 2022, EUI Comparative Politics Seminar, IC3JM 2024, IPES 2021, University of Zurich Historical Political Economy Workshop, and the Virtual Historical Political Economy seminar, for comments on previous drafts. Jean Durand and Katja Sonntag provided excellent research assistance.
Funding
We are also grateful for financial support from the Deutsche Forschungsgemeinschaft (German Research Foundation) under Germany’s Excellence Strategy – EXC-2035/1 – 390681379.
Authors
Michael Becher is Professor of Political Economy in the School of Politics, Economics, and Global Affairs at IE University. He can be reached at michael.becher@ie.edu.
Irene Menéndez González is Assistant Professor of International Political Economy in the School of Politics, Economics, and Global Affairs at IE University. She can be reached at irene.menendez@ie.edu.
