Electoral Rules,Programmatic Competition,and Redistribution: Evidence From Interwar France

Electoral Rules and Programmatic Policy

Several canonical contributors to the study of the political economy of electoral rules have argued that proportional representation (PR) encourages politicians to supply non-targeted policies that benefit voters irrespective of their geographic position, while single-member district systems (SMD) encourage the opposite (e.g., Persson & Tabellini, 2003). The logic is that the re-election prospects of politicians under SMD are tied to the welfare of voters in certain districts, allowing them to disregard the welfare of other voters. This encourages the provision of targeted goods (pork) to particular districts financed by taxes on voters living elsewhere. Under PR, politicians’ re-election prospects depend on the welfare of voters in larger districts (at the limit, comprising the entire country), meaning that targeted spending loses its allure relative to policies that benefit broad swaths of the electorate irrespective of geography – what we call programmatic policies. While different in the specifics, Lizzeri and Persico (2001) and Milesi-Ferretti et al. (2002) offer arguments in the same vein. ³

These arguments easily generalize to other forms of “personalistic” behavior beyond the provision of pork barrel spending. Persson and Tabellini’s (2003) argument implies that politicians whose fortunes are tied to the welfare of a small geographic area are likely to expend more effort on constituency service and local hobnobbing than are those whose fortunes are tied to broader districts. These personalistic activities can be understood as geographically targeted non-material benefits. Extending the logic of Persson and Tabellini (2003), the payoff to programmatic policies relative to the payoff from personalistic appeals should increase under proportional representation.

The possible limitation of arguments like that of Persson and Tabellini (2003) is that they may grant insufficient attention to institutional diversity among PR systems. Their results may only hold for a subset of PR countries, depending on differences in electoral rules that they consider to be of marginal significance. ⁴ In a canonical article, Carey and Shugart (1995) argue that the value of personal reputation should increase with district magnitude when the electoral rule allows for competition between candidates in the same party. Under open-list PR, voters have the opportunity to select which candidates within a given party list they favor, pitting co-partisans against one another within a given multi-member district and increasing candidates' incentives to distinguish themselves from their co-partisans through non-programmatic appeals to their voters. On this basis, we might expect politicians in open-list PR to devote more energy to personalistic appeals and non-programmatic forms of distribution than under SMD, since they face the additional challenge of distinguishing themselves from their co-partisans in the same district. ⁵ Given that politicians are constrained in the total amount of effort they can supply, this implies that politicians under open-list PR should supply less effort pushing for programmatic policy than politicians under SMD.

A countervailing consideration is that politicians may face greater difficulty claiming credit for targeted policies like local public goods in the multi-member districts characteristic of PR systems (Ashworth & Bueno de Mesquita, 2006; Lancaster, 1986; Tromborg & Schwindt-Bayer, 2022). Voters may not know whether a given local public good is attributable to the efforts of one legislator or another within their multi-member district. For this reason, Lancaster places open-list PR between SMD and closed-list PR in terms of its effect on geographic targeting (also Rickard, 2018).

Carey and Shugart (1995, 430–431) grant Lancaster’s (1986) point that credit-claiming for pork barrel projects may become more difficult as district magnitude increases. They therefore propose that pork-barreling may be lower in open-list PR systems than under SMD. But even if pork-barreling declines as district magnitude increases, this does not necessarily mean that legislators will shift their efforts to programmatic competition. After all, the provision of pork is only one way of cultivating a personal reputation among voters. Carey and Shugart argue that the cultivation of personal reputation may take a different form under open-list PR than under SMD: rather than providing pork, legislators may engage in other non-programmatic activities to distinguish themselves from co-partisans in the eyes of their voters. These authors’ conclusion remains that the total effort directed to personal reputation-building will be higher under OLPR than under SMD, since under SMD politicians are not compelled to distinguish themselves from their co-partisans within a given district. If we assume that all effort devoted to personal reputation-building comes at the expense of effort devoted to programmatic policy, this means that legislators face weaker incentives to advocate for programmatic policy under OLPR than under SMD.

We can thus sketch out two competing positions. The first is a version of the classic view articulated by Persson and Tabellini, which holds that PR – irrespective of list type – generates fewer incentives for non-programmatic appeals and correspondingly stronger incentives for programmatic policy than does SMD. Following Lancaster (1986), this view can be amended to incorporate a difference between open- and closed-list PR. Open-list PR may offer stronger incentives for personalism and non-programmatic targeting than closed-list PR, but weaker incentives than SMD, given the aforementioned credit-claiming difficulties. This interpretation of the ordering of electoral systems is consistent with recent comparative work by Rickard (2018), who characterizes open-list PR as a middle ground between SMD and closed-list PR. This ordering of electoral systems is reflected in ordering A of Figure 1 below. The second possibility is one that follows Carey and Shugart (1995) in predicting that open-list PR will generate stronger incentives for personalistic appeals than SMD does because the former forces candidates to compete against co-partisans within the same district. This ordering of electoral systems is reflected in ordering B of Figure 1.OPEN IN VIEWER

The existing evidence does not allow us to adjudicate decisively between these alternative views. Crisp et al. (2004) use cross-national evidence from Latin America to show that open-list PR encourages personalism, a finding that is consistent with ordering B above. Similarly, Chang and Golden (2007) present evidence indicating that non-programmatic appeals (corruption) increase with district magnitude under open-list PR. Edwards and Thames (2007) show cross-nationally that public goods spending decreases with district magnitude in open-list systems, a finding that supports ordering B. Shugart et al. (2005) show that candidates in open-list PR are more likely to have personal attributes that may endear them to local constituents than candidates under other electoral rules, consistent with ordering B. André and Depauw (2013) show that legislators’ time in-district and their propensity to advocate for local development projects increases with district magnitude under OLPR but decreases under CLPR, consistent with ordering B above.

Tromborg and Schwindt-Bayer (2022), on the other hand, present cross-national evidence indicating that geographic targeting declines with district magnitude irrespective of list type, which supports possibility A above. Selb and Lutz (2015) also challenge Carey and Shugart’s hypothesis with fine-grained evidence from Switzerland. Rickard (2018) shows that there are fewer geographically targeted subsidies under PR than under SMD, and fewer geographically targeted subsidies under CLPR than under OLPR, though she does not directly compare OLPR and SMD. Menendez (2016) shows that geographically targeted trade compensation schemes are less common under PR, but does not compare open- and closed-list cases. Gagliarducci et al. (2011) show that Italian legislators elected via SMD engage in more targeted spending than those elected via CLPR, but their study does not include a comparison to OLPR. Cheibub and Gisela (2020) offer evidence from Brazil indicating that the extent of intra-party competition under open-list PR is exaggerated, lending support to ordering A above.

A smaller number of studies leverage within-country variation to study the effects of changes in electoral rules on electoral promises or spending patterns. For the most part, these studies do not throw direct light on the question at hand in this paper. Pachón and Shugart (2010) examine the change from single non-transferable vote (SNTV) to PR in Colombia, while Catalinac (2016) studies the switch from SNTV to a mixed-member majoritarian system in Japan. Catalinac’s finding that intraparty competition encourages legislative particularism and pork-barrel spending is broadly consistent with ordering B above. On the other hand, Funk and Gathmann (2013) present evidence that switching from SMD to (mostly) OLPR in Swiss cantons decreased geographically targeted spending on roads, consistent with ordering A, though their paper lacks the measures of legislator-level behavior that would be necessary to validate the hypothesized mechanisms. Studying within-country change in electoral rules (along with fine-grained measures of legislator behavior) is the most promising way of assessing the distributive effects of electoral institutions (as argued, e.g., in Rickard, 2017). Nevertheless, work using this kind of research design remains underdeveloped. This is the avenue we pursue in this paper.

Electoral System Change in France

In what follows, we study the relationship between electoral rules and legislator behavior in France during the Third Republic, which was founded in 1875 and survived until the country’s defeat by Germany in 1940. Possessing universal male suffrage and a democratically accountable executive, the Third Republic compared favorably to its peer countries in Europe. The French case is relevant because it offers a rare instance of a shift from a single-member majoritarian system to open-list PR and back again. Between 1889 and 1914, French legislators were elected to its lower chamber, the Chamber of Deputies, from single-member districts (the scrutin d’arrondissement) using a two-round rule: if no candidate received a majority in the first round, all candidates could proceed to a second round in which the winner was determined by plurality. Voices in favor of proportional representation had existed since before the founding of the Third Republic (e.g., Rosanvallon, 1998, pp. 153–164), but pressure for electoral reform intensified after the turn of the century. Legislation to this effect passed the lower chamber in 1909 and 1912 but was stopped through procedural maneuvering and subsequently stymied in the upper chamber (Ahmed, 2013, pp. 158–159). A renewed effort following the end of the First World War led to the adoption of PR in July 1919 for use in the elections of November 1919.

In the public debate over electoral reform both before and after the war, it was commonly argued that the single-member scrutin d’arrondissement favored non-programmatic and personality-driven electoral competition, and that switching to proportional representation would incentivize programmatic competition (e.g., Lachapelle, 1911; Hoffmann, 2008, p. 200). Indeed, the single-member scrutin d’arrondissement was criticized as a “scrutin de gladiateurs” on the basis of its alleged tendency to yield electoral conflict based on personalities rather than on policy programs. ⁶ Others were more cautious about proportional representation, emphasizing that its effects would depend on the precise characteristics of the electoral rule in place. Among these voices was the mathematician Henri Poincaré. In a preface to a 1911 book on electoral reform, Poincaré argued that if France were to adopt proportional representation, it must avoid a system that would allow voters to select their preferred candidates from party lists: such an electoral rule would “permit a range of dubious maneuvers” and “embed in voters’ minds the notion that questions of persons must prevail over questions of principle” (Lachapelle, 1911, xi). In effect, Poincaré endorsed Carey and Shugart (1995) avant la lettre: whatever the shortcomings of France’s SMD system, he predicted that electoral personalism would only increase under proportional representation if lists were open. Similar concerns were aired in the parliamentary committee on electoral reform (Commission du Suffrage Universel), where it was argued that open-list PR would promote “vigorous but perhaps unscrupulous competition between candidates on the same list” (Archives Nationales, (n.d), C 14826).

As is often the case when it comes to the reform of electoral systems (e.g., Ahmed, 2013; Boix, 1999; Leemann & Mares, 2014; Walter & Emmenegger, 2019), parties’ attitudes toward reform were shaped in part by their expectations regarding its effects on their own electoral prospects. The adoption of proportional representation was favored by Socialists and by the parties of the right, both of whom thought that PR would improve their electoral prospects (Monier, 1999, p. 67). The adoption of PR was opposed by the centrist Radical Party, which was the dominant party after 1905 (Parti Républicain Radical et Radical Socialiste, 1910). The two-round SMD system that existed before the adoption of PR favored parties that could form local coalitions to win second-round elections. The centrist and ideologically flexible Radicals were able to form coalitions with Socialists or Rightist candidates (as local circumstances dictated), placing them in a strong position to form winning coalitions. Accordingly, they supported the majoritarian electoral system. ⁷

The PR system adopted in July 1919, known as representation proportionelle avec prime a la majorité, was a compromise solution in several important respects (e.g., Guieu, 2015, p. 196). First, the multi-member electoral districts were formed at the level of the pre-existing departments (France’s primary sub-national administrative unit) and the number of seats assigned to each department was equal to the number of seats within each department under the single-member system. This led to moderately sized districts with an average magnitude of 6.5 and an average voting population of 123,000 (Kreuzer, 2001, pp. 55–59). ⁸ Second, the new electoral rule allowed for an element of majoritarianism in that any candidates who received a vote from an absolute majority of voters in a given department would automatically receive a seat (Journal officiel 1919). Finally, the PR system was open-list: it allowed for voters to select their preferred candidates within a given party list and to split their votes between candidates on different lists (panachage) (e.g., Stuart, 1920). ⁹ In Appendix Q, we discuss the workings of France’s SMD system and its open-list system of proportional representation in detail and score them on the basis of Carey and Shugart’s (1995) criteria of ‘Ballot,’ ‘Pool,’ and ‘Votes.’ This yields the conclusion that France’s system of open-list PR created distinctively strong incentives for candidates to cultivate a personal reputation and was thus conducive to personalistic rather than programmatic electoral competition. ¹⁰

This open-list PR system was used in two elections, 1919 and 1924, before being abandoned prior to the 1928 elections. There is wide scholarly agreement that the proportional system was abandoned because it was seen as providing favorable of electoral opportunities for the French Communist Party, which had split from the Socialists in 1920. This made the PR system unpalatable to its former proponents: the Socialists were dissatisfied with an electoral rule that allowed for Communist inroads in parliament at their expense, while the Right sought to thwart the expansion of a party committed to a break with capitalism. Their response was therefore to return to the two-round SMD system that had previously been in place. This institutional change achieved its immediate goal of squeezing out the Communists, whose seat share dropped from 28 to 14 (Tiersky, 1974, p. 48). Its other possible effects are to be investigated below.

During the war, France had been governed by a broad national coalition (l’Union sacrée) and then by a center-right coalition. The 1919 election under PR yielded a large conservative majority in the Chamber of Deputies, as can be seen in Appendix Table F2 (Bernard & Dubief, 1988, p. 89; Guieu, 2015, p. 202). This election was a major defeat for the Socialists, contrary to their expectations about the effect of PR on their electoral performance. These fortunes reversed, however, in the 1924 election under PR, yielding a center-left majority (a Radical government supported by the Socialists, the Cartel des gauches) that “kindled great hopes in the French left” for redistributive reform (Bernard & Dubief, 1988, p. 163; also Cagé & Piketty, 2023, pp. 496–497, 509).

The Cartel was, however, unsuccessful in its ambitions and collapsed in the summer of 1926, leading to a mixed government under Raymond Poincaré, who pursued an austerity program to stabilize the franc (see Bernard & Dubief, 1988, pp. 98–99, 162–164; Keiger, 1997, pp. 320–324; Valance, 2017, pp. 411–414; Eichengreen, 1996, pp. 177–183). The 1928 election after the return to SMD reaffirmed conservative control of government. Poincaré, still in charge, announced shortly after the 1928 election that his government would subordinate all reform projects to the overriding aim of “fiscal prudence” (Journal Officiel 1928a, Dèbats Chambre, 7 June 1928). As we show in Appendix F, the governments during the 1928–1932 legislative period were considerably more conservative on average than those of the 1924–1928 legislative period. Accordingly, there is little prima facie reason to expect a surge in programmatic redistribution following the return to SMD in 1928, since this period was marked by conservative rule; the only time that the left had a chance to pursue its programmatic objectives in government was during the period under proportional representation.

Data and Empirical Strategy

In what follows, we examine how electoral system may have changed politicians’ incentives to make programmatic policy commitments, both as incumbents and as candidates. To measure the behavior of incumbents, we construct a dataset that measures the number of bills proposed by each legislator over a range of policy areas during four legislative periods between 1914 and 1932. Our data set is therefore in a panel format with a distinct observation for each deputy during each legislative period during which he sat in the Chamber. During the period 1914–1932, 889 deputies sat for a single legislative period, 347 deputies sat for two periods, 172 deputies sat for three periods, and 80 deputies sat for all four periods. In two of the legislative periods in our dataset (1914–1919 and 1928–1932), politicians were elected in a majoritarian electoral system, while in the other two periods (1919–1924 and 1924–1928) they were elected through the open-list proportional rule.

The source for the construction of our variables measuring deputies’ bill proposals are publications of the French Chamber of Deputies, which we consulted at the Archives of the National Assembly in Paris (Annales, various years). These publications record legislative proposals written and submitted by each deputy to the Chamber. We take advantage of the fact that deputies in the French parliament had an unlimited right of legislative initiative: any deputy could put forward a bill on any topic at any time (Larcher, 1896, p. 150; Gooch, 1928, pp. 150–151; Zeldin, 1993, p. 585). The deputy would request that the president of the chamber grant him the opportunity to introduce the bill and, having received permission, would read out the title of the bill and sometimes a brief summary of its contents. The president of the chamber could, at most, require the deputy to wait until after the debate had ended or until the next day to introduce the bill to the chamber (Larcher, 1896, p. 159). Once the deputy had introduced the bill to the chamber, the bill would be assigned to the committee of initiative, which would conduct a preliminary assessment of its merits before passing it along to subject-specific committees (Gooch, 1928, p. 154).

The unlimited right of legislative initiative possessed by every member of the Chamber of Deputies was unusual in comparative perspective and quite controversial in the Third Republic. Some believed that France’s deputies had “excessively fertile minds” and exercised legislative initiative with “regrettable abundance” (Larcher, 1896, pp. 213, 231). Any bill proposed by a deputy was printed in the chamber’s official journal. Since it was common for newspapers in the legislator’s home district or department to reproduce the bill in full, proposing bills in parliament was a formidable source of “electoral propaganda” (227–228). Deputies would likely discuss their recent bill proposals during their trips to their home districts, which in many cases occurred on a weekly basis (Siegfried, 1930, p. 214), and in their apparently voluminous correspondence with constituents (de Jouvenel, 1914, pp. 22–24). Economic liberals like the lawyer Émile Larcher feared that the unlimited right of legislative initiative encouraged undue regulatory intervention in the market and undesirably high levels of public spending: as Larcher wrote, a deputy might propose a law granting a public pension to every elderly citizen in order to win popular support (Larcher, 1896, pp. 228–229). This would be an example of a deputy using his right of legislative initiative to advocate for programmatic redistribution. Deputies would also, of course, use the right of legislative initiative to push bills that would channel non-programmatic benefits to their constituents (Siegfried, 1930, pp. 212–215; Larcher, 1896, p. 229).

We draw on the publications of the Chamber of Deputies to code information about the content of deputies’ legislative proposals. We record the number of programmatic bill proposals in three core substantive domains: social policy (social insurance, welfare spending, etc.), labor market regulation, and tax policy. We also measure the number of bills proposed by each legislator that entailed the targeted use of public funds for projects in their home district (pork barrel spending). Our measure of incumbent behavior thus directly captures legislator effort on behalf of programmatic policy provision, on one hand, and particularistic spending, on the other. Detailed information on the construction of each of these variables is in Appendix A and descriptive statistics in Appendix C. Although it is possible that our focus on these policy domains overlooks important programmatic activity in other areas, we believe that the policy areas we study capture most of the legislative activities that are relevant for testing theories about the relationship between electoral systems and programmatic redistribution, such as Persson and Tabellini (2003). Figure 2 reports the mean number of bills proposed per legislator in each of these four policy areas for the four parliamentary periods we study. This shows a very considerable increase in bill proposals in each of the three programmatic areas we study during the 1928–1932 period relative to the three earlier periods. The stability in the mean number of programmatic bills proposed in the 1919–1924 and 1924–1928 periods relative to the 1914–1919 period is striking, since one would expect a large increase in parliamentary advocacy for programmatic reform in the immediate aftermath of the First World War (e.g., Smith, 2018). In other words, given received views about the effect of war on welfare state development, it is unsurprising that we observe an increase in programmatic policymaking in the interwar period: what is surprising is that this increase does not come until the 1928–1932 period, a decade ‘late.’ This shift in legislator behavior is to be further investigated below.OPEN IN VIEWER

To measure politicians’ behavior as candidates, we code the programmatic content of the manifestos issued by candidates for the election of 1924 (under PR) and 1928 (under SMD). The source for this information is a publication of the manifestos (professions de foi) of all elected deputies (JORF, 1925; JORF, 1930). These manifestos were short documents (typically one to two pages in length) addressed from the candidate to the voters in his constituency. They typically included some biographical information, including the candidate’s past service to his constituency, along with political material ranging from denunciations of the incumbent government to warnings about the perils of socialism. Candidates usually made some statements about the groups in society that they stood for, but only some candidates made concrete policy commitments. Our sources only include the manifestos of winning candidates. This yields 542 observations of 271 legislators who won both elections. Using multiple researchers, we code the manifestos for programmatic policy content by examining whether the manifesto includes an endorsement of the following social, labor market, and fiscal policies: a commitment to provide public unemployment insurance, to establish or increase public pensions, to institute paid vacations, to institute a minimum wage, or to raise taxes on high incomes or wealth. ¹¹ We develop a binary variable that indicates the presence or absence of a policy commitment in each of these domains for each candidate in each election, as well as a variable that reports the average support across these five policy areas. Information on the construction of each of these variables can be found in Appendix B and descriptive statistics in Appendix C.

Empirical Strategy

We first analyze the relationship between the electoral system and politicians’ behavior as legislators in the Chamber of Deputies. The complete specification takes the following form:

L i , t = β 1 P R t + X i , t ψ + Z d i , t ϕ + α i + ϵ i , t

(1)

L i , t represents the bills proposed by legislator i in legislative period t and PR _t is a variable that takes a 1 for the periods under PR and a 0 for the periods under SMD. X_i,t is a vector of variables capturing time-varying legislator characteristics measured in time t, Z d i , t is a vector of variables measuring time-varying department-level characteristics measured for district d of legislator i at time t, α _i represents individual fixed effects (capturing time-invariant legislator characteristics), and β₁, ψ, and ϕ are parameters to be estimated. We estimate this model using ordinary least squares. The results using a Poisson model can be found in Appendix O.

In an alternative specification, we use a continuous measure of district magnitude rather than a binary measure of the electoral rule in place. This takes the following form:

L i , t = β 1 D M d i , t + β 2 D M d i , t 2 + X i , t ϕ + Z d i , t ψ + α i + ϵ i , t

(2)

D M d i , t is a measure of the district magnitude of district d of legislator i at time t. The variable DM naturally takes a 1 for every politician during the legislative periods under SMD (1914–1919 and 1928–1932), but varies with the magnitude of the district during the periods under open-list PR. The squared term D M d i , t 2 allows us to examine whether the relationship between district magnitude and bill proposals varies as magnitude increases. This will tell us whether the relationship between open-list proportional representation and programmatic policy provision becomes weaker or stronger in higher-magnitude districts. This model is estimated using ordinary least squares. Appendix O contains the results of a Poisson model.

In what follows, we first report the results of simplified ‘baseline’ versions of these models that omit time-varying legislator and department characteristics (X_i,t and Z d i , t ). We then report the results of the full model with time-varying controls. In the full model, we control for whether a given deputy was also a mayor during a given legislative period (under the so-called cumul des mandats, politicians could hold multiple positions; combining the role of deputy with the role of mayor was common). Becoming or ceasing to be a mayor could plausibly affect a politicians’ incentives to engage in programmatic policy-making. We also control for deputies’ party label, using variables to indicate Radical, Socialist, or Communist party membership or membership in one of the parties of the Republican Right. The reference group is politicians without a party affiliation. Finally, we control for a deputy’s status as a party leader, based on the list of parliamentary party group secretaries and presidents in the Feuilletons of the Chamber of Deputies (Journal Officiel de la République Française, 1919, Journal Officiel de la République Francaise b. Chambre des Députés, 1928b, Journal Officiel de la République Francaise. Chambre des Députés, 1915, Journal Officiel de la République Francaise. Chambre des Députés, 1924).

These specifications also include a set of department-level control variables ( Z d i , t ) . The department was the relevant electoral district during the elections under PR. We derive a measure of department-level income per capita for each legislative period from Cagé and Piketty’s (2023) annual department-level estimates, which report department income per capita as a share of the national average. This allows us to address the possibility that deputies from lower-income or economically declining areas may be more supportive of programmatic public policies that redistribute to disadvantaged groups. To control for the possibility that a legislator’s support for programmatic policy during any given legislative period may depend on the strength of the electoral threat from the radical left within his department, we control for the share of seats held by Communist deputies in each department for each legislative period. This is derived from our own data set. We also control for population density with data from the French census (Statistique Agricole de la France, 1935). Finally, to address the possibility that a department’s exposure to war-related destruction and predation may affect its residents’ policy demands and, by extension, the behavior of its representatives, we include a dummy variable for departments that were partially or wholly under German occupation during the period 1914–1918, which includes those that were on the front lines (i.e., partially occupied). This is based on Nivet’s (2011) list of occupied departments. Descriptive statistics for all of these variables are available in Appendix C.

Candidate Behavior

We then proceed to assess the relationship between the electoral system in place and politicians’ behavior as candidates. Our data on candidates’ manifesto commitments covers the 1924 and 1928 elections, conducted under open-list PR and SMD, respectively. We rely on a similar model to the one used to analyze incumbent behavior, which was discussed above. The model takes the following form:

C i , t = β 1 S M D t + X i , t ϕ + Z d i , t ψ + α i + ϵ i , t

(3)

where C_i,t indicates the presence or absence of a platform commitment in a given policy area, SMD _t indicates the electoral rule in place (taking a 0 for the 1924 election and a 1 for the 1928 election), X_i,t indicates the same set of time-varying individual-level controls discussed above, Z d i , t indicates the same set of department-level controls discussed above, and α _i indicates individual fixed effects. As in the models of incumbent behavior discussed above, this model assesses the relationship between electoral system change and within-person change in politicians’ behavior. The model is estimated using ordinary least squares.

Results

The results of our analyses are consistent with the view that the adoption of open-list proportional representation induced a decline in programmatic policy-making as politicians re-allocated effort toward non-programmatic intra-party competition. We begin by presenting the results concerning legislative behavior and then turn to the analysis of candidates’ manifesto commitments. ¹²

Legislator Behavior

The coefficient plot in Figure 3 below displays the estimated correlation between the presence of proportional representation and the number of bills proposed by French legislators in the Chamber of Deputies. The estimates depicted with diamonds are from an analysis using the baseline model (no time-varying controls) and the full data set covering four legislative periods (1914–1919, 1919–1924, 1924–1928, and 1929–1932), two of which were elected under the proportional rule (1919–1924 and 1924–1928). The estimates depicted with triangles are from an analysis using the model with time-varying controls covering the same period. The estimates depicted with circles are from an analysis that uses the baseline model and is restricted to the three postwar periods (1919 onward) to address concerns that the results are driven by wartime dynamics. The estimates depicted with squares are from an analysis using the model with time-varying controls covering the three postwar periods. The models estimated to generate these coefficients include legislator fixed effects, meaning that the estimates capture within-person variation in behavior across different electoral rules. The tabular version of the results reported in Figure 3 can be found in Appendix D.OPEN IN VIEWER

Figure 3.

The relationship between open-list proportional representation and the number of bills proposed by each deputy in France’s Chamber of Deputies in the listed policy areas per legislative period

The coefficient plot in Figure 3 shows that proportional representation is associated with significantly fewer bill proposals per legislator in the areas of social policy, labor market regulation, and taxation. The baseline results from the analysis of all four legislative periods (diamonds in the plot) show that legislators proposed, on average, 0.25, 0.33, and 0.08 fewer bills in each of these policy areas under PR than under SMD (amounting to roughly a sixth, a third, and a 10th of a standard deviation). ¹³ When we restrict the analysis to the postwar legislative periods, we obtain similar and somewhat stronger results. The models with time-varying controls generate similar results. This supports the notion that legislators reallocated their efforts away from programmatic policy-making and toward other activities under open-list proportional representation. The results of the Poisson model, reported in Appendix O, are similar to those presented here.

The plot also shows that there was no statistically significant difference in the quantity of pork bills proposed by legislators under open-list PR and SMD. Taken together, the evidence indicates that the adoption of PR was followed by a decline in legislator efforts on behalf of programmatic policy but did not lead to a discernible spike in pork-barreling. This is consistent with the possibility that open-list PR induces legislators to devote greater effort to non-programmatic activities other than pork-barreling, as suggested by Carey and Shugart (1995). In general, the results presented in Figure 3 contradict the hypothesis proposed by Persson and Tabellini (2003), which was that the adoption of proportional representation should be followed by an increase in programmatic redistribution. We arrive at similar results when analyzing the behavior of legislators in each of France’s main party groups separately, indicating that our findings are driven by behavioral changes among legislators from across the political spectrum (Appendix M).

To further investigate the robustness of these findings, we widen the time frame of our analysis by incorporating observations from three earlier periods in which legislators were elected under SMD. These were the 1902–1906, 1906–1910, and 1910–1914 parliaments. Doing so increases the number of observations by close to 75% and allows us to track long-serving legislators over a greater span of their career in the Chamber of Deputies before the introduction of PR. The results, reported in Appendix I, are consistent with the findings reported in Figure 3, albeit at attenuated levels of significance in some models. In general, this analysis strengthens our confidence in the validity of the results on the consequences of PR adoption reported in Figure 3.

We now turn to our alternative specification (from equation (2) above), which uses a continuous measure of district magnitude rather than a binary measure for proportional representation. As discussed in the previous section, the district magnitude variable takes a 1 for all politicians during the legislative periods under SMD but varies with the magnitude of the multi-member district under PR (see Appendix H for the full distribution of departments by district magnitude). The results in Table 1 show that an increase in district magnitude is associated with a significant reduction in the number of bills proposed in each of the three programmatic domains studied here. As in the previous specifications, we see no significant relationship between district magnitude and pork-barreling. The coefficient on the squared term suggests that the negative relationship between district magnitude and programmatic bill proposals may be attenuated at high levels of district magnitude. Table 1 reports the results of a model without the full set of time-varying controls. Results of the model with the full set of controls, found in Appendix E, look very similar to those reported below. The results here include all four legislative periods (1914–1919, 1919–1924, 1924–1928, and 1928–1932), but the results are similar if the analysis is restricted to the postwar legislative periods (1919 onward). These estimates are available in Appendix E. The results with the Poisson model look similar and can be found in Appendix O.

OPEN IN VIEWER

Table 1.

The Relationship Between District Magnitude and the Number of Bills Proposed by Each Deputy in France’s Chamber of Deputies in the Listed Policy Areas per Legislative Period

	Dependent variable:
	Social policy bills	Labor market bills	Taxation bills	Pork
	(1)	(2)	(3)	(4)
District magnitude	−0.036**	−0.050***	−0.013+	−0.003
	(0.013)	(0.008)	(0.007)	(0.007)
District magnitude2	0.001*	0.001***	0.0003*	0.00005
	(0.0002)	(0.0001)	(0.0001)	(0.0001)
Legislator fixed effects	Yes	Yes	Yes	Yes
Period covered	1914-1932	1914-1932	1914-1932	1914-1932
Observations	2,419	2,419	2,419	2,419
R2	0.007	0.049	0.006	0.0002

⁺p < 0.10; *p < 0.05; **p < 0.01; ***p < 0.001.

Note. Table 1 reports results from the analysis covering all four legislative periods we study (1914–1932). Standard errors are heteroskedasticity-robust and clustered by legislator. The model estimated is model 2 with time-varying controls omitted. Appendix E contains estimates with the full set of time-varying controls as well as estimates for the 1919–1932 period. See Appendix A for the construction of the measure of bill proposals and Appendix C for descriptive statistics.

The analyses reported in this section assess the relationship between electoral rules and legislator behavior using individual fixed effects. This addresses the question of whether the introduction of a new electoral rule induced legislators to behave differently than they had under the previous electoral rule. Naturally, this restricts the analysis to legislators who served in parliament under both electoral rules. To assess the overall consequences of electoral rule change for legislator behavior, it is also important to consider the effect that a change in electoral rules may have on the composition of the legislature, i.e., on the kinds of candidates who get elected. In addition to encouraging incumbent legislators to push for more programmatic policies than they had previously, the return to an SMD system for the 1928 election may have led to the replacement of some legislators with new, more vigorous advocates of programmatic policy expansion.

To assess the extent to which the behavior of legislators re-elected under SMD in 1928 differed from those who were newly elected in 1928, we first estimate a interaction model without legislator fixed effects. This allows us to study whether the ‘effect’ of SMD adoption in 1928 on programmatic policymaking was stronger among freshman legislators in the 1928–1932 parliament (for whom we have only one observation). The results, reported in Appendix J, show that the freshman legislators elected in 1928 were generally no more or less likely than legislators re-elected in 1928 to propose new bills in the domains of social policy, labor market regulation, and taxation. In other words, we observe that both new and experienced legislators proposed more programmatic policy bills in the 1928–1932 parliament than they or their counterparts had in earlier parliaments under OLPR. The same point can be shown by estimating a model that compares the behavior of re-elected legislators and freshman legislators in the 1928–1932 parliament. As shown in Appendix K, members of these two groups were generally not very different from one another in terms of the number of programmatic bills proposed during the 1928–1932 parliament, though freshman legislators were somewhat more likely to propose labor market bills and somewhat less likely to propose pork barrel legislation than were their more experienced counterparts. In general, these findings suggest that the return to SMD in France may have influenced programmatic policymaking both by changing the behavior of legislators who were already serving under the previous electoral rule and by facilitating the election of candidates committed to programmatic reform.

Alternative Explanations

A key limitation of a single-country study of national-level electoral system change like ours is that all units are ‘treated’ at the same time. We take several steps to ameliorate the inferential concerns that arise from this design. First, we conduct a series of placebo tests to examine whether we observe similar shifts in programmatic activity between legislative periods in which there is no change in electoral rules, going back to 1902. The results of these tests, reported in Appendix R, indicate that the behavioral changes documented above are indeed distinctive to the periods in which there was a change in electoral rules. Second, we develop a comparative case study between France and the United Kingdom in which we examine variation in the level of programmatic activity during the interwar period (as proxied by fiscal outcomes). This comparison shows that the United Kingdom experienced a surge in taxation and government expenditure after the First World War in a manner consistent with established theories of the relationship between war and programmatic reform (e.g., Obinger & Schmitt, 2020; Scheve & Stasavage, 2016). France, in contrast, remained at its prewar levels of taxation and government expenditure for most of the 1920s, despite much public discussion of the need for programmatic redistribution to compensate for wartime sacrifices (on which, Smith, 2018). Tax yields and public spending in France remained at their prewar levels until after 1928, when levels of taxation and public spending rapidly increased and converged with those of the UK. These findings, presented in Appendix S, are consistent with the idea that France’s adoption of open-list PR in 1919 stood in the way of a ‘normal’ path of postwar programmatic development (represented by the United Kingdom) by discouraging politicians from advocating for programmatic reform. The return to SMD for the 1928 elections may have allowed France to converge with the United Kingdom.

Still, it remains possible that changes in the behavior of French legislators during and after the 1928 elections were driven by factors other than those we have stressed above. First of all, it could be the case that the return to SMD influenced the behavior of legislators in a different way than the one conjectured above. Second, it could be the case that other changes within France around 1928 led to increase in programmatic policymaking and campaigning for reasons unrelated to the change in electoral system. Third, it could be that the changes in the electoral system we study were downstream from other changes in the political environment that affected levels of programmatic activity. Given the limits of our design, we now lay out possible rival explanations of the observed patterns and assess their plausibility in light of the evidence, as recommended by Stokes (2014).

Conclusion and Implications

What drives politicians in democracies to make programmatic policy commitments? In this paper, we have examined the relationship between electoral rules and programmatic policy development. A prominent body of research, especially by economists, suggests that the adoption of proportional representation should generate stronger incentives for politicians to compete on the basis of programmatic policies that benefit voters irrespective of their place of residence. In particular, proportional representation is thought to spur the introduction of redistributive social programs, yielding larger welfare states (Persson & Tabellini, 2003). A competing line of scholarship by political scientists has challenged this conclusion, arguing that proportional representation systems differ radically from one another in terms of the electoral incentives they generate for programmatic policy-making. According to the latter view, the stylized contrast between single-member and proportional electoral systems obscures more than it clarifies. Some kinds of PR systems – in particular, open-list PR – may actually generate weaker incentives for programmatic appeals than do single-member systems. This is because open-list PR may push politicians to dissipate their energies in non-programmatic intra-party competition (Carey & Shugart, 1995).

Progress on this question has been hindered by the fact that electoral systems rarely change, generating few opportunities for the kind of within-country comparisons that are the most plausible way of studying how electoral rules shape politicians’ behavior. Moreover, scholars are usually unable to directly measure the individual-level behaviors predicted by existing theories. This paper attempts to overcome these limitations by leveraging the adoption and subsequent abandonment of open-list proportional representation in Third Republic France. Using original legislator-level data hand-coded from historical sources, we have examined how open-list PR affected politicians’ behavior as both incumbents and candidates. Our results show that under proportional representation, French politicians were less likely to propose programmatic social policies and less likely to campaign on programmatic policy promises than they themselves were under the single-member electoral rule.

The evidence presented in this paper indicates that open-list proportional representation generates weaker incentives for programmatic policy development than does a single-member rule. This does not mean, of course, that SMD is the ideal system if the goal is to maximize programmatic electoral competition. Closed-list PR may still outperform single-member majoritarianism in terms of generating incentives for legislators to invest in programmatic redistribution. The point of this paper is to throw light on the relative standing of open-list PR and SMD with regard to candidates’ incentives to form programmatic linkages with voters.

Due to several important empirical limitations, this paper cannot decisively resolve the scholarly disagreement about the effects of different electoral systems on politicians’ incentives to supply programmatic policy. First of all, the changes in electoral rules that we study in this paper may have been endogenous to other changes in the political environment in interwar France that could have also exerted a direct influence on politicians’ behavior. Second, we lack a control group of French legislators who were unaffected by the changes in electoral rules, preventing us from cleanly separating the effect of rule changes from other features of the political environment that may have changed at the same time. Third, our analysis focuses on winning electoral candidates and legislators in parliament, introducing potential bias from the omission of losing candidates who faced the same electoral rule but perhaps behaved in systematically different ways. Fourth, our study is restricted to a single historical case with many distinctive features: among other things, France operated under an unusual form of SMD involving a second-round runoff and experimented with an unusual form of open-list PR that granted a considerable premium to lists that won a majority. These French peculiarities would impose a constraint on the generalizability of our findings even if the first three problems listed above were resolved.

Even with these limitations in mind, this paper’s findings indicate that scholars should retire the stylized distinction between proportional and single-member systems as a heuristic for predictions about legislators’ incentives to push for programmatic redistribution. This means that theories connecting proportional representation to welfare state development must be rethought with more attention to institutional variation among proportional regimes. When it comes to electoral rules and the incentives they generate, the devil is in the details.

Supplemental Material