Sweden’s Peculiar Adoption of Proportional Representation: The Overlooked Effects of Time and History

Abstract

Sweden’s adoption of proportional representation (PR) is interesting because it involved static structural and institutional factors, well captured by variance-based left-threat thesis, and four temporal factors—sequencing, timing, historical change, and duration—that historical case studies highlight. We integrate these two sets of factors. We fuse the more static, temporally homogeneous world created by the left-threat thesis, that is well suited to explain cross-sectional variations, with the more dynamic, temporally heterogenous world presumed by the case studies that is attuned to temporal processes. It illustrates how comparative historical analysis (CHA) can translate temporal anomalies into generalizable temporal mechanisms and how nested analysis, together with causal graphs, provide helpful tools for updating theories. We ultimately employ an abductive approach that evaluates evidence not just for its inferential leverage of confirming theories but also for its inductive potential to generate new, more test-worthy hypotheses.

Over the last three decades, the literature on the origins of proportional representation (PR) has grown in numbers and methodological diversity. It has focused on the same 1880–1920 period, studied the identical two dozen cases, tested a wide range of theories, and involved a multi-method dialogue. It also was marked by genuine knowledge cumulation as it converged around the so-called left-threat thesis, first articulated by Stein Rokkan (1968) and subsequently formalized by Carles Boix (1999) and retested by Marcus Kreuzer (2010), and Lucas Leeman, and Isabella Mares (2014).

Sweden’s PR adoption in 1907 poses a challenge for the literature on the origins of PR. Historical case studies of Sweden confirm the left-threat thesis’s key structural and institutional elements (Carstairs 1980; Eckelberry 1964; Lewin 1988; Rokkan 1968; Rustow 1971; Verney 1957). They discuss Conservatives’ growing concern that Sweden’s ascendant working-class party and widening franchise would electorally strengthen Socialists and thereby threaten the right’s economic and political privileges. Those case studies further show how Conservatives turned to PR in anticipation that the new electoral system might offer them minority protection against the looming parliamentary majorities of Socialists. But they also identify temporal complexities that the left-threat thesis does not address. The Swedish case thus confirms most empirical implications of the left-threat thesis while simultaneously foregrounding new, not yet theorized temporal factors. We show how unexplained factors can serve as inductive insights to update the left-threat thesis, rather than becoming disconfirming evidence to reject the theory. It shifts the focus from the inferential leverage of evidence for testing theories to their less explored abductive potential to update theories.

This situation, that the case studies confirm the left-threat thesis while also identifying important gaps, results from the fact that PR adoption in Sweden was entangled with other constitutional reforms rather than being a discrete event as in many other countries. The historical case studies therefore analyze Sweden’s PR adoption in the context of its broader democratization that led them to identify four temporal factors that remain unaddressed in the left-threat thesis. First, Sweden’s late democratization and parliamentarization entangled PR with broader constitutional struggles, thus changing the sequencing of key causal factors laid out in the left-threat thesis. Second, its early adoption of PR relative to other European democracies changed the timing and reducing the ability of parties to learn from other PR adopters. Third, Sweden’s PR adoption extended over a long time-range that required actors to cope with historical change. One such change involved the shifting legitimacy between 1880 and 1920 of the very institutions over which parties bargained (Andersson 1998; Blais, Dobrzynska, and Indridason 2005; Carstairs 1980; Teorell 2017). Parties had to factor those changes in public attitudes into their constitutional calculus. Fourth, the long period it took to eventually adopt PR added duration as a temporal factor. The PR adoption was a multi-event process that involved staggered reform bills in 1907 and 1918 as well as four failed reform bills preceding them. In sum, sequencing, timing, historical change, and duration are key for understanding Sweden’s PR adoption because they increased the uncertainty under which actors had to reform the existing institutions.

The left-threat thesis does not incorporate these temporal elements as it focuses on static institutional and structural factors best suited to explain cross-sectional variations. What, then, is the process for updating theories in response to newly discovered empirical anomalies? And what are the challenges if those empirical anomalies involve temporal dynamics that differ from the static assumptions of an existing theory? The nested analysis proposed by Evan Lieberman (2005), together with comparative historical analysis (CHA) (Kreuzer 2023), both emphasize completing the causal inference techniques used to test with the theorizing necessary to generate test-worthy hypotheses. Together, these two approaches offer a methodological toolkit suitable for handling Sweden’s unexplained empirical anomalies.

Such CHA-inspired nested analysis makes four assertions. First, theories involve abstractions that simplify a phenomenon being studied (Tavory and Timmermans 2014; Swedberg 2014; Kreuzer 2024). Such simplification is necessary to generate testable hypotheses, but these simplifications come at the cost of backgrounding causal factors that could potentially confound test results (Gerring 2012; Lieberson 1985; Pearl and Mackenzie 2018). Valid causal inferences therefore require, what Judea Pearl calls, de-confounding (Kreuzer 2024; Pearl and Mackenzie 2018). De-confounding involves the active search for causal factors that a given theory backgrounds. Second, CHA and nested analysis conceptualize causal inferences as an iterative process that should be mindful at the outset about confounders (Kreuzer 2023, 116–34, 164–80). Large N analysis (LNA) and small N analysis (SNA) must consider prior research findings before deciding whether to proceed with model-testing, when prior research is well established and the likelihood of confounders is small, or model-building strategies, when prior research is limited, and hence subject-overlooked confounders (Lieberman 2005, 2016; more generally, see Abbott 2004; Yom 2015). Third, LNA model testing should be complemented with SNA model-building. SNA plays a particularly central role in de-confounding because it pays greater attention to contextual and temporal complexities than LNA. SNA assists in updating theories and generating new, more test-worthy hypotheses that should guide subsequent LNA replications (Lieberman 2005). Fourth, confounders should be treated as inductive insights that can update existing theories rather than disconfirming evidence that invalidates established theories (Kreuzer 2023, 118–30, 164–80).

We use Sweden for a SNA to both evaluate and update the left-threat thesis. We employ a CHA framework for SNA to better address the temporal nature of the confounders encountered in Sweden’s PR adoption. CHA belongs to a larger repertoire of qualitative methods that pays close attention not just to context but also to qualitative changes of those contexts over time and the temporal dynamics related to sequencing, timing, historical change, and duration (Adams, Clemens, and Orloff 2005; Kreuzer 2023; Mahoney and Rueschemeyer 2003; Mahoney and Thelen 2015; Pierson 2004). CHA is part of a larger tradition of historical social science that involves historical sociology, historical institutionalism, economic sociology, economic history, and other historically informed modes of social inquiry.

We begin by subjecting Sweden to the standard process-tracing model-testing SNA (Beach and Pedersen 2019; Bennett and Checkel 2014). This model-testing broadly confirmed the left-threat thesis while also identifying untheorized temporal complexities. The second part shifts to model-building SNA to explore the temporal confounders more extensively: sequencing, timing, historical change, and duration. This exploration and related theorizing are necessary for translating these confounders, which originally were simply inchoate empirical anomalies, into theoretically grounded temporal mechanisms that operated at the same level of abstraction as the left-threat thesis. The third part engages in a second SNA to empirically demonstrate the relevance of the temporal mechanisms. It engages in another theorizing step by demonstrating how these temporal mechanisms are compatible with the original left-threat thesis. Demonstrating this compatibility is a necessary step for updating a theory. We conclude with conjectures about the possible relevance of the four temporal mechanisms beyond the Swedish case. Overall, we demonstrate how evidence serves both to test theories as well as to update them; it ultimately questions the positivist orthodoxy that evidence used to generate theories should never be employed to also verify it (Yom 2015).

Nested Analysis: Large N Analysis and Model-Testing Small N Analysis

Nested analysis starts by embedding new research in the findings of prior scholarship. This requires well-developed theories and available data sets to baseline new research against LNA of the existing theories (Lieberman 2005, 438). The proportional representation literature is advanced enough so that we can skip the opening LNA. The literature includes several LNA and model-test SNA studies that warrant the plausibility of the left-threat thesis. Online appendix 1 retraces the provenance of the left-threat thesis over four research cycles. We begin our analysis with the thesis that emerged from the most recent, fourth research cycle and specify its empirically testable implications. We then use Sweden as an additional case for model-testing SNA.

Theoretical Benchmark: Empirical Implications of Left-Threat Thesis

Figure 1 provides a causal arrow diagram of the key empirical implications made by the most up-do-date version of the left-threat thesis. ¹ It breaks the theory into three components: the outcome, empirically tested causal factors, and theorized, but so far untested, background factors.

Figure 1 Left-threat thesis

Note: The figure differentiates two sets of factors generating the left threat: empirically tested and untested background factors.

The left-threat thesis stipulates a singular, snapshot-like outcome: choosing proportional representation over an existing majoritarian system. It situates PR during the late nineteenth-century industrialization and the resulting conflict between old constitutional monarchies and emerging democracies. Conservatives and Liberals sought to obtain minority protections in return for enfranchising largely propertyless workers to contain the perceived electoral threat that those workers posed for private property (Przeworski 2018, 28–47; Ziblatt 2017, 24–53). The empirical studies assume these macro-historical factors to be uniform across all cases and thus making it unnecessary to empirically verify these factors even though they are part of the larger theoretical argument. The studies focus instead on less macro and less historical factors in the middle box of figure 1 that are temporally more proximate to the outcome.

PR became interesting for Conservatives and Liberals because it could replace earlier, increasingly anachronistic electoral safeguards—outright franchise restriction, plural voting, indirect elections, public voting, mal-apportionment, or gerrymandering—or legislative safeguards like bicameralism, or limiting parliamentary sovereignty. PR provided minority protection by solving Conservatives and Liberals’ coordination problem that arose under the existing majoritarian systems. The entry of Socialists as a new contestant confronted Conservatives and Liberals with the risk of splitting their votes and thereby allowing the more disciplined Socialists to win parliamentary majorities with simple electoral pluralities. Conservatives and Liberals could hypothetically have addressed this risk by cooperating and strategically withdrawing candidates across districts, thus converting the new three-way back into the old two-way races. The complexity of such cooperation depended on whether the existing system was first-past-the-post (FPTP), double ballot, or a more complicated blocked vote system (these systems are described in Carstairs 1980, 9–14). It also was conditional on whether Conservatives and Liberals were divided over clerical issues or had managed to build disciplined national party organizations (Ahmed 2013; Boix 1999; Calvo 2009; Kreuzer 2001; Schröder and Manow 2014). PR became attractive because it did not require electoral coordination. Furthermore, Amel Ahmed (2013) and Nina Barzachka (2014) extended the left-threat thesis beyond the electoral arena. They pointed out that the left’s protests and strikes reinforced electoral threat or served as substitutes in instances where the threat itself was modest. They also recognized that socialist parties varied in their ideological radicalism and thus suggested that their platforms needed to be analyzed in addition to their electoral strength.

Model-Testing Small N Analysis

The model-testing SNA confirms all key elements of the left-threat thesis and identifies a few test anomalies. (Refer to online appendix 2 for quotes from political historians directly supporting the left-threat thesis.) It juxtaposes the thin description offered by a theory against the thick description provided by historical sources (Beach and Pedersen 2019; Bennett and Checkel 2014). We resort to process tracing since the unavailability of district level voting data and legislative rollcalls make it impossible to test the left-threat thesis with more standard statistical research designs.

Figure 2 offers a synopsis of our test results on which the rest of the current section expands. It updates figure 1—which identified the empirical implications of the left-threat thesis—in light of the Swedish model-testing SNA findings. It sorts those findings into three categories: causal factors in bold indicate confirming evidence, factors in strikethrough identify disconfirming evidence, and factors in italics point to evidence unpredicted by the original theory. This unpredicted evidence was deemed causally relevant by historical case studies but was unrelated to sequencing, timing, historical change, and duration discussed in subsequent sections. Figure 2 underscores the validity of the left threat for explaining Sweden’s adoption of proportional representation. It fully confirms its three empirical implications (ideological radicalism, disproportionality of a non-PR system, and social protest), fully disconfirms one implication (fragmentation of the right), and partially confirms four implications (Conservatives championed PR but not Liberals; the left posed a threat but also were allied with Liberals in so-called Lib-Lab coalitions pioneered in the UK by the Liberal and Labour Party; the franchise was expanded but not fully; parliament had legislative powers but was not fully sovereign). It also discovered unpredicted factors (role of temperance and free church movements). We next elaborate on these findings.

Figure 2 Revised left-threat thesis

Notes: Figure 2 updates the original left-threat thesis considering the Swedish test results. Causal factors in bold are confirmed, factors in strikethrough disconfirmed, and in italics and boxes point to new inductive insights.

To understand these findings more readily, let us offer a very brief historical contextualization. We elaborate on this historical context and particularly its temporal complexities more fully in the following two sections. Sweden’s PR adoption was entangled with three other political reforms: the widening of the franchise for the two legislative chambers, re-balancing their respective powers, and parliament wrestling from the King his prerogatives to appoint the PM and dissolve parliament. These reforms pivoted around the 1907 Lindman Law which significantly expanded the franchise for the lower house and introduced PR for both houses. The 1907 Law was preceded by four unsuccessful reform bills, so-called near misses, going back to 1896. And it was followed by the 1918 constitutional reforms that fully expanded the male and female franchise and shifted full authority of the appointment and dismissal of government from the King to parliament. With this historical synopsis, we discuss the findings of our model-testing SNA.

Electoral threats: Sweden introduced proportional representation in 1907 when the conservative Prime Minister Arvid Lindman passed the namesake Lindman Law with the express purpose of protecting the right against the growing strength of the Social Democrats after their founding in 1889. The law itself was not fully implemented until 1911 because the Swedish constitution required that the original 1907 law be reaffirmed after the next national election held in 1909 (Verney 1957; Eckelberry 1964).

As table 1 shows, the left electoral threat was modest with Socialists winning only 9.5% of the vote in 1905, the last election before the PR adoption. Sweden’s very restrictive suffrage lessened the electoral threat for the lower house because it enfranchised only 24% of the male population in 1896 when in the rest of Europe on average 64.6% of men were enfranchised (Flora 1983, 141–42). ² Still, the share of Swedes who could vote for the lower house increased steadily “as economic growth helped more and more industrial workers to fulfill the economic requirements to vote” (Lewin 1988, 70; Rustow 1971, 19). Sweden also experienced “one of the most rapid periods of economic growth of any industrial country.” Its increase of per capita national product of 26.2% was more than twice that of the UK and only slightly lower than the United States (Rustow 1971, 16–17). These incremental increases in the electorate resulting from voters economic requirements to vote were particularly pronounced in urban working-class districts (Eckelberry 1964, 20; Jusko 2017, 130–39).

Table 1 Vote share of Swedish parties in lower house

Source: Mackie and Rose 1991, 405–6

The electoral threat was even lower in the upper house, which was indirectly selected from rural and urban councils. The franchise for those local councils technically was wider but the plural voting based on income and property ownership restricted it so much that its electorate amounted to only one tenth of the already highly restricted lower house electorate (Bengtsson 2019, 136–39). Not surprisingly, Socialists had no elected members in the Conservative dominated upper house. Overall, the restrictive franchises kept the immediate electoral threat moderate. But the massive and rapidly growing opposition to those restrictions meant that those protections would not endure for much longer and thus significantly increased the electoral threat in the near future.

The restrictive franchises for both houses and continued royal interference in forming governments converted the left-threat into a Lib-Lab threat. Bengtsson observed that “the lack of democracy in Sweden would through its effect on social coalitions foster a thorough democratization. The exclusion of such large part of the population from formal national politics meant that the petty bourgeoisie, lower middle class and working class united in suffrage reform movements” (2019, 141). This Lib-Lab coalition also highlights that Conservatives were alone in advocating for PR and that the right’s fragmentation, as stipulated by the left-threat thesis, was not a factor in Sweden’s PR adoption. Scholars disagree on how moderate or radical the Swedish socialists were and on the impact of the1903 Socialist-organized General Strike (Ahmed 2010; Berman 1998; Marks, Mbaye, and Kim 2009; Rustow 1955, 42; Verney 1957, 193–201).

Social protest: Well-organized, large scale protest movements—rooted in unions, temperance groups, free churches, and the National Suffrage Association (NSA)—underscored that the moderate electoral left threat would not last long. These movements sought to expand the franchise and reform parliament so that they could pursue their social reforms more effectively (Schrad 2021). They also supported Liberal and Socialist candidates in return for having their own members elected on liberal or socialist tickets (Lundkvist 1975, 182–83).

As figure 3 shows, the free churches—the Mission Covenant Church, Baptist Union, and United Methodist Church—were the oldest of the social movements and began to organize in the 1880s. By 1890, they had around 100,000 members and a decade later 130,000 members (Lundkvist 1977, 68). These three groups challenged the religious monopoly of the Lutheran state church, pressed for more religious freedom, and engaged in short-lived efforts to organize workers in Christian Labor Unions (Anderson 2009, 215–18; Sandell 2001, 674). After 1890, the free churches also began to work with the temperance movement to restrict the sale of liquor (Lundkvist 1977, 171; Sandell 2001, 675). The temperance movement emerged to address Sweden’s alcoholism that was a central social problem in the late nineteenth century (Sandell 2001, 675). It comprised five different groups who had 228,000 members by 1900 and 355,000 by 1910, making it the largest social movement in Sweden (Johansson 1947, 37; Edquist 2001, 300). The temperance movement pushed for the adoption of so-called local vetoes, that would give citizens the right to hold local referenda on whether to ban the sale of alcohol in their municipalities. These local vetoes were quite radical because they would have given all citizens the right to vote even as their electoral franchises remained restricted. They also would have deprived local governments of their liquor taxes—their most important revenue source. Conservatives consequently opposed the local vetoes because they were afraid that the loss of liquor revenue would lead to higher income or property taxes (Lundkvist 1977, 231). It turns out that Conservatives were threatened not just by revolutionaries but also by teetotalers.

Figure 3 Swedish social movements and popular mobilization

Notes: Figure 3 displays the membership of Sweden’s three largest social movements and contextualizes them with important events in its democratization. No data was available for the NSA, the fourth major movement.

Source: Lundkvist and Andrae 1998

The free churches and temperance movements worked together with the National Suffrage Association (NSA) founded in 1890 by Liberals and Socialists. The NSA engaged in educational work and took the extraordinary step in 1893 and 1896 to organize parallel elections with universal suffrage to elect, what they called, two People’s Parliaments that discussed strategies and passed petitions to introduce universal suffrage (Franzén 1985, 147; Lundkvist 1977, 47). The People’s Parliaments also discussed the Socialists’ demands to organize a general strike on behalf of franchise demands. Liberals opposed this strategy as being too radical. This did not prevent Socialists from organizing a general strike on their own in 1903. Labor Unions were the last of the four movements to form in 1890. They worked closely with the Social Democrats to push for political and social reforms (Sandell 2001, 676). Overall, these social movements reinforced the left threat by advocating for the same reforms as the Socialists, providing additional mobilizational capacity and joining the Lib-Lab coalition.

Electoral systems: The left-threat thesis stipulates that the disproportionality of the existing, majoritarian systems complicated the electoral coordination of Conservatives and Liberals and thus favored Socialists. Given the Lib-Lab coalition, the coordination problems under Sweden’s FPTP system did not play a significant role. However, two features of Sweden’s electoral system contributed to majority bonuses that potentially amplified the left threat. The five largest urban districts employed a rare, so-called blocked voting system also found in Belgium (Mackie and Rose 1991, 400). It was an artifact of a clause in the electoral law that required that each seat represent a fixed number of voters. The original law did not anticipate a rapid urbanization that would change the population sizes of districts. It also did not make any accommodations for re-districting (Verney 1957, 109). As a result, politicians complied with the law by adding seats to the growing urban districts. These seats would be aggregated into lists and the list winning the largest vote share would win all seats. Blocked voting—combining multiple member lists with winner-take all plurality rule—created super-majoritarian districts precisely where Socialists grew fastest. Furthermore, the composition of electoral lists in the urban districts was very complex and gave an advantage to better organized parties like the Socialists. The electoral system required “no formal parties, no registration of candidates before the election; every eligible man was a possible candidate, and voters could write any names they wanted on the ballots, although parties and other political organization recommended certain lists and printed ballot were used by most voters” (Janson 2018, 73; Proportional Representation Society 1908, 24).

Overall, the Swedish case strongly supports the left-threat thesis’s structural and institutional claims. The rapid industrialization and demands for democratization created a left threat both through elections and social protest (Vogler 2022). Conservatives recognized that the existing FPTP system increased the probability of manufacturing legislative majorities for Socialists even though Socialists might win only a plurality of the votes. They also realized that PR lowered the probability of such over-representations by guaranteeing the proportional translation of votes into seats. Sweden’s Lib-Lab coalition contradicts the left-threat thesis prediction that right’s fragmentation was a key factor. Overall, the SNA suggests that core element of the left-threat thesis hold.

But what are we to make of the fact that the historical case studies we used to test the existing left-threat thesis also revealed empirical anomalies that are not accounted for by the same thesis. That is, the narratives advanced by the case studies reference elements that are crucial for explaining Sweden’s PR adoption, but these same elements neither contradict the left-threat thesis nor does the thesis account for them. What are the inferential implications for such empirical anomalies? Should they be read as falsifying a theory, as is commonly done, even though they do not directly contract a theory’s implications? CHA treats untheorized empirical anomalies as the raw material for updating theories rather than as disconfirming evidence used to reject them (Swedberg 2014; Yom 2015). And nested analysis employs model-building SNA to update a theory. ³ The introduction already listed sequencing, timing, historical change, and duration as the proposed theoretical updates and in the process skips over the inductive process that translated the initially inchoate empirical anomalies into these four temporal mechanisms. The next section, therefore, illustrates this translation process that is central to theorizing.

Model-Building SNA: Translating Historical Complexities into Temporal Mechanisms

The historical case studies highlight temporal complexities in Sweden’s adoption of proportional representation that the existing left-threat thesis overlooks. More specifically, Sweden’s PR adoption was entangled with other constitutional reforms rather than being a discrete choice. Moreover, this choice extended over a 25-year period, rather than taking place during a single legislative session. Sweden’s PR adoption involved a layered, concurrently unfolding set of processes, rather than a single easily dateable snapshot-like event. These temporal complexities of the Swedish case generated at the first reading of the secondary sources inchoate empirical anomalies that were difficult to bring into conversation with the left-threat thesis. To establish such a dialogue required drawing on the CHA toolkit and translating the anomalies into four temporal mechanisms. This translation process is rarely explicated, even though it is a crucial element of theorizing (Kreuzer 2024; for more details on this translation, see 2023, 73–81, 116–33).

Chronologies provide a crucial step for explicating from temporal complexities first patterns, then link those patterns to temporal mechanisms, and ultimately use those mechanisms to update the original theory. Chronologies impose an initial, strictly calendric order on the temporal complexities we encounter in historical narratives (Kreuzer 2023, 73–94). They reduce the noise by selecting from the totality of historical occurrences or facts the subset of potentially relevant events (Abbott 1983; on the difference between occurrences and events see Abrams 1982, 191; Falleti and Mahoney 2015, 212–14). Chronicling has a long tradition in historical analysis (Rosenberg and Grafton 2010) where it is employed to generate old-town chronicles or to construct genealogical trees. Historians, in turn, go beyond such chronicles by seeking to explain the connection among events, come up with periodization schemes, and assess the significance of historical changes (Appleby, Hunt, and Jacob 2011, 257). Social scientists often lack the historians’ detailed contextual knowledge and find historical narratives sufficiently overwhelming that chronicling helps them reduce historical noise.

Table 2 provides a chronology of the seven key events discussed in the historical case studies. This chronology is not exhaustive and leaves out smaller reforms like amendments or private member bills that were less relevant. The seven events listed in the chronology included three fully implemented laws and four bills that were introduced to parliament but failed to pass and become laws. We will refer to these bills as near misses because they sought to make constitutional changes like the 1907 and 1918 laws but were unsuccessful. Near misses frequently are overlooked because they involve defeats and fail to bring about change. ⁴ These near misses are helpful to understand the generative process that ultimately produced the change (Riker 1990; Tulis and Mellow 2018). We also make the chronology more analytical by adding six event categories: lower house PR, upper house PR, lower house franchise, upper house franchise, bicameralism, and parliamentarization.

Table 2 Event chronology

Notes: This chronology lists seven key events in Sweden’s political development. It differentiates between bills that failed to win parliamentary majorities and laws that were enacted.

* Involved double ballot electoral system instead PR.

# Changes in norms rather than formal constitutional amendments.

✓ Minor and ✓✓ major reform proposals.

⨯ Existing institutions were not considered for reform and remained unchanged.

For the substantive details refer to online appendix 3.

The chronology ultimately underscores the entanglement of Sweden’s PR adoption with reforms about the franchise, bicameralism, and parliamentary sovereignty. This chronology thus provides a strong prima facia case that Sweden’s PR adoption involved temporal dynamics that mattered but are unaccounted for in the left-threat thesis. CHA has long been attentive to such temporal complexities and provides the concept for translating the complexities into discrete causal mechanisms. Adam Przeworski and Henry Teune (1970), proper names into variables, so CHA translates dates into four recurring temporal mechanisms: sequencing, timing, historical change, and duration. CHA uses sequencing to differentiate the temporal order of causal factors. Sequencing involves a before/simultaneous/after differentiation to figure out whether event x₁ happened before, concurrently with, or after event x₂. It explores to what extent variation in sequencing produces different causal effects. It departs from linear assumptions of causality in which the temporal order of causal factors does not matter (Abbott 1988; Falleti and Mahoney 2015; Pierson 2000). Timing compares the starting dates of single but qualitatively identical events across multiple cases to make ordinal distinctions between early versus late timings. It explores to what extent later events learned from earlier ones (Kreuzer 2023, 54–59; Pierson 2000; Pink 2018). Historical change acknowledges that we live in a changing world and that preferences, norms, concepts, resources, or pretty much any causal factors are moving objects. And the movement of those objects through time transforms them qualitatively and irreversibly (Bloch 1953; Carr 1961; Conrad 2017; Kreuzer 2023, 36–52; Maza 2017; Sewell 1996). The casual effects consequently have a half-life and might change as their attributes change and either weaken or strengthen the original effects of causal factors. Finally, duration measured the time that elapses between the starting and end point of a process and enables a short/long distinction (Aminzade 1992; Bartolini 1993; Cohen 2018, 1–21; Grzymala-Busse 2011; Kreuzer 2023, 54–59).

These four temporal mechanisms make it possible to translate the temporal complexities shown in the chronology of events (i.e., table 2) into more general causal factors that are necessary for SNA model-building. Table 3 contrasts the original left-threat thesis with the new temporal mechanisms derived from the Swedish case and illustrates initial results of the model-building process that is necessary for ultimately updating of the left-threat thesis.

Table 3 Sweden’s temporal factors

Notes: Political historians operate under assumptions of temporal heterogeneity and thus attentive to temporal complexities invisible if operating under assumptions of temporal homogeneity. Table 3 identifies the four key temporal factors ignored by the left-threat thesis. Refer to the glossary in the online appendix for elaborations of these four temporal factors.)

The first column points out the left threat’s inattention to causal mechanisms. The second column points out the left threat’s inattention to temporal mechanisms because it treats PR adoption as a single-shot event and a single process. The third column foregrounds the temporal mechanisms at work in Sweden that are not accounted for by the left-threat thesis. These temporal mechanisms do not invalidate the left-threat thesis because they involve heretofore untheorized causal factors that should be treated as inductive insights for potentially updating the theory. Such updating, however, requires first that we illustrate the relevance of the four temporal mechanisms for explaining the Swedish case. We hypothesize that the four temporal factors mattered because they profoundly affected the uncertainty under which Swedish actors had to make institutional choices. The left-threat thesis itself makes uncertainty over the Socialists’ electoral threat and the ability of PR to provide effective minority protection the key element of its theoretical framework. The thesis therefore follows the larger literature that also places uncertainty at the center of its analysis (refer to online appendix 4 for a synopsis of this literature). The next section illustrates how sequencing, timing, and historical change increased this uncertainty. The section on managing uncertainty discusses how the fourth factor, the long duration of the multi-event outcome, explains how actors managed to make decisions despite the high uncertainty that they faced.

Small N Analysis Model-Testing: Temporal Mechanism

The previous section demonstrated that the four temporal mechanisms—sequencing, timing, historical change, and duration—provide plausible new inductive insights to update the left-threat thesis. The model-building of the second section now needs to be complemented by empirically demonstrating the relevance of those four mechanisms for explaining the Swedish case. We are in effect proceeding with a second SNA model-testing to verify the causal relevance of newly identified the temporal mechanisms.

Reverse and Simultaneous Sequencing: Complexifying the Choice Set

CHA pays attention to sequencing because the temporal order in which causal factors unfold have important causal effects. Paul Pierson uses a cooking analogy to demonstrate the causal effects of sequencing. He points out that the sequence with which ingredients are cooked profoundly affects the tastiness of the final meal (Pierson 2004, 1–2). The complex sequencing of Sweden’s franchise expansion, parliamentarization, and proportional representation adoption became causally significant because it increased the uncertainty actors faced. We refer to this sequence as non-linear because it deviates from the parliamentarization, franchise-expansion, proportional representation adoption sequence stipulated by the left-threat thesis that is presumed to be fixed and uniform across all cases (refer to online appendix 4 for a brief review of the literature on uncertainty and institutional design).

Table 4 illustrates the link between sequencing and uncertainty by demonstrating how Sweden’s non-linear sequence complexified actors’ choice sets. The left-threat thesis’s linear sequence meant that actors faced a simple choice set of whether to adopt PR. Table 4 lists this choice set on the far right. It contrasts this choice set with nine others that result from Sweden’s non-linear sequence. And table 4 lumps these nine options into five broader categories. One, mostly hypothetical choice sets involved ignoring all reform options and doubling down on the constitutional status quo and defending the 1866 constitutions. Such intransigeance risked more radical pro-democracy protest and thus was politically too costly. Parties therefore had to choose among the middle eight options—a complex undertaking. And the contrast between their complexity with the simple choice set stipulated by the left-threat thesis underscores the uncertainty that Swedish parties faced.

Table 4 Sweden’s hypothetical choice sets

Legend: ⨯: reform rejected.

✓ = reform accepted.

Notes: The column on the far left lists the five constitutional choices over which Swedish parties negotiated simultaneously. The subsequent columns lump them into five broad categories: status quo, single PR, double PR, FPTP, and full PR reform. The column on the far right lists the far less complex choice set stipulated by the left-threat thesis.

The middle eight choice sets required parties to consider complex trade-offs. Conservatives had to figure out how many of the existing, increasingly anachronistic minority protections (see the status quo column) they had to give up to win the more durable minority protection offered by PR. Liberals and Socialists, in turn, had to ascertain how many democratic concessions to demand, and how quickly, as well as how many concessions on minority protections they were willing to make. Parties had to evaluate constitutional trade-offs that were far more complex than doing a simple cost-benefit calculus between PR and whatever majoritarian system was going to be replaced. And this complexity created enormous uncertainty under which parties had to define their preferences and to determine the trade-offs required to secure parliamentary majorities. For example, the restrictive franchise excluded workers and complicated Conservatives’ task to accurately assess the left’s eventual electoral threat under universal suffrage. This task was further complicated by the Socialists running candidates on joint lists with Liberals so that workers ended up voting for Liberal candidates and bourgeois voters for Socialist candidates. This made it impossible to ascertain how many votes socialists would have received had they run on their own lists.

Early Timing: Limited Opportunity to Learn

The chronology also highlights the early timing of Sweden’s proportional representation adoption that contributed to uncertainty by creating an informational deficit about the effects of PR. The 1896 Annerstedt Bill introduced the first PR proposal and made Sweden one of the earliest European country to consider PR. This limited its opportunity to learn from other countries about PR’s political consequences and confronted its politicians with an informational deficit. In the 1890s, PR constituted a new and untried voting technology that required understanding how technicalities like district magnitude, electoral formula, nature of lists, or vote pooling would affect the translation of parties’ votes into seats. PR also had important secondary effects in how parties campaigned, built their organizational infrastructures, and structured their internal decision making (Kreuzer 2001). Being an earlier PR adopter, Sweden was unable to learn about the workings of PR from other countries and thereby lessening the uncertainty this new voting technology posed. Lewin, for example, reports that the Lindman government was aware in 1907 “that the shortcomings of majority elections should be quite imperceptible to those who are the rulers at a given point in time” (1988, 73). Furthermore, Sweden’s unusual block voting system in urban districts added another complication. These districts involved lists and thus blurred the standard distinction between FPTP and PR systems; refer to online appendix 5, which provides additional evidence about parties’ limited understanding of PR.

Historical Change: Dealing with Moving and Hence Unpredictable Objects

The chronology of figure 2 makes clear that Sweden’s proportional representation adoption extended over a 25-year period that was marked by rapid socio-economic changes and democratization as well as World War I, followed by revolutionary uprisings. These were tumultuous years that made for an unpredictable political environment. A key element of this unpredictability was the shifting public acceptance of the minority protections that Conservatives championed. This changing acceptance draws attention to the role of public opinion, a factor the left-threat thesis ignores altogether. In doing so, the left-threat thesis misses out on an element whose changing nature makes it far more historical than the comparatively static structural and institutional factors. We use the term historical in the sense historians employ it (Bloch 1953; Kreuzer 2023, 36–52). Historians associate historical with events that did not just occur in the past but also events connected in a process in which the past becomes qualitatively different from the present. The protest activities of Sweden social movements constituted the events that shifted public opinion about the legitimacy of various minority protections. The event-driven and irreversible nature of this shift that makes the change in public opinion just as historical as the slower moving changes in socio-economic structures and institutions (Hirschman 1991; Przeworski 2018, 13–36). Conservatives sought minority protection not just through PR but also through preserving franchise restrictions, bicameralism, and restricting parliamentary sovereignty—all elements of the increasingly antiquated 1866 Constitution that the social movements challenged vocally. The changing public attitudes became a source of uncertainty as it affected the long-term durability of the protections that various long-term, anti-majoritarian safeguards might provide.

Table 5 lists the key minority protections and describes the overall trend in their public acceptance between roughly 1880 and 1910 to ascertain the long-term certainty these safeguards would provide Conservatives. It ordinally ranks the options from least to most certain in providing long-term protection.

Table 5 Public acceptance of minority protections

Notes: Conservatives had to choose among minority protections that differed in their public acceptance and hence the certainty of their long-term minority protection for Conservatives. Online appendix 6 elaborates on the coding.

Sources: Carstairs 1980; Lewin 1988; Verney 1957

The first three options in table 5, property or tax requirements for the franchise or allocating voters multiple votes based on their income, wealth, or family status, provided very limited long-term protections. These voting restrictions were widely seen as totally anachronistic and had mobilized massive social movements that sought to abolish them. By contrast, PR is at the bottom of the table because it was new and untarnished by partisan electoral engineering that was associated with various franchise restrictions. It also reflected egalitarian norms that resonated with popular demands for more democracy (Blais, Dobrzynska, and Indridason 2005; Carstairs 1980, 101). PR thus was highly certain to provide long-term minority protections (Carstairs 1980, 102). Finally, the middle three minority protections—bicameralism, indirect elections, and limiting parliamentary sovereignty—provided modest long-term minority protections because their public acceptance was less clear cut than that of PR or franchise restrictions. Conservatives and Liberals supported the indirectly elected upper house, modelled after the U.S. Senate, because it would counter-balance the directly elected lower house (Lewin 1988, 83–106). Liberals, however, wanted to curtail the powers of the upper house to make Sweden’s bicameralism more symmetrical. Socialists supported such reforms and even advocated for a unicameral legislature. They accepted bicameralism after the franchise expansion to the upper house increased their seat share. Conservatives thus faced moderate certainty about the upper house’s long-term minority protection.

The 1866 Constitution separated executive and legislative powers by placing the hereditary king in charge of selecting the prime minister, his cabinet, dissolving parliament, and even setting the national agenda. This division of power was meant to place the king above the emerging party politics to represent more effectively the national interest (Lewin 1988, 88; Verney 1957, 96–97). The idea of the king as supra-party and unifying figure proved uncontentious as long as Conservatives won parliamentary majorities which allowed the king to appoint governments from members of his own political milieu (Verney 1957, 119, 134–35). This changed in 1906 when the Liberal Party won its first parliamentary majority and the king invited the Liberal Party leader Karl Staaff to form the government (Carstairs 1980, 100; Verney 1957, 120–25). From this point onward, limiting parliamentary sovereignty became more contentious. The king, for example, opposed Staaff’s request to dissolve parliament and call new elections that Staaff intended to use as a referendum on his failed 1906 bill. The king re-asserted his royal prerogatives and appointed Lindman as Staaff’s successor instead. He continued during the next decade to insert himself in parliamentary affairs by claiming that wartime national defense and growing prominence of political parties after the 1907 franchise expansion required an executive above party politics more than ever. The king accepted full parliamentary sovereignty only in 1918 when Liberals and Socialists won increasingly larger parliamentary majorities, demands among Socialists to abolish the monarchy grew louder, and monarchies collapsed across Europe in 1918 (Carstairs 1980, 101; Ericsson and Nyzell 2017, 337–40; Lewin 1988, 102–4). The political legitimacy of limiting parliamentary sovereignty thus was very much in flux between 1900–1920, creating a moderate and short-term certainty to offer minority protection for Conservatives.

Managing Uncertainty: Staggered Reforms and Near Misses

To recap, the first section confirmed the link between structural and institutional factors of the left-threat thesis. The next section translated the inchoate temporal complexities, observed in the historical case studies, into four theoretically grounded temporal mechanisms with the help of CHA. The third section showed that sequencing, timing, and historical change interjected considerable uncertainty in actors’ decision-making, making proportional representation adoption highly unlikely. This section demonstrates how duration, in the form of Sweden’s staggered reform and near misses prior to 1907, created a prolonged outcome that reduced uncertainty and allowed parties to adopt PR after all. It shows how the staggering of the 1907 and 1918 reforms lowered uncertainty by allowing actors to reduce the numbers of decision to be reached and bought them time to better assess the left threat. The 1907 reforms also had a pre-history in the form of four near misses that involved parliamentary debates, negotiations, and ultimately failures. These near misses had an important educative effect that reduced uncertainty sufficiently for actors to pass the 1907 Law.

Staggered Reforms: Buying Time to Reach Compromises

The adoption of proportional representation was inescapably linked to reaching a larger constitutional compromise. The staggered reforms lowered uncertainty by de-complexifying the larger constitutional bargaining because they reduced the number of issues over which parties had to negotiate and deferred controversial issues into the future. The 1907 and 1918 constitutional debates considered jointly six reforms of which the 1907 law addressed four and the 1918 law addressed two. The controversial demand to adopt a unicameral parliament remained unresolved until 1970. The number of reforms tell only half the story how the staggering reduced uncertainty.

To get a fuller story requires looking at the specific reforms undertaken and how their mix bought actors time to figure out the trade-offs among different constitutional compromises. For example, if the Lindman Law was indeed the second-best solution for all involved, the actors had to figure out what opportunities the updated constitutional order offered them to pursue their first-order goals in the future. The 1907 Lindman Law reduced uncertainty by providing Conservatives with enough minority protection that they would accept the current and potential future democratic reforms. It simultaneously gave Liberals and Socialists enough new political leverage so that they could follow up on the democratization demands to which Conservatives successfully objected in 1907.

Let us explore in more detail how the 1907 reforms reduced uncertainty first for Conservatives, then for Liberals, and finally for Socialists.

We already discussed how Sweden’s restrictive franchise was a key source of uncertainty for Conservatives because it lacked legitimacy and prevented directly assessing the electoral left threat. Conservatives could make indirect inferences using union and Socialist party membership figures or strike activities as proxy indicators for the Socialists’ full electoral potential. But such inferences remained uncertain and required Conservatives to speculate about how much the Socialist electoral strength would grow depending on which tax, property, age, or other franchise restriction they removed. The growth potential of the Socialists 9.5% vote share in 1905 was difficult to gauge on the eve of the first franchise reforms (Mackie and Rose 1991, 405–6). The 1911 election—the first held under the expanded franchise—provided a more definite answer as Socialists increased their vote share by 19% and won 28.5% of the votes.

For Conservatives, the left threat resulting from the Socialists’ electoral surge in 1911 was limited because of the simultaneously introduced PR worked as advertised. It kept Conservatives’ seats share within roughly 1% of their vote share and provided minority protection in the sense of averting a far more disproportional vote seat translation for Socialists than was anticipated under a majoritarian system (Mackie and Rose 1991, 405–6). Furthermore, Conservatives could rely on the other minority protections. The remaining suffrage requirements continued to disenfranchise 20% of the electorate, the indirectly and plutocratically elected upper house assured them a veto point, and the king still retained the right to appoint the PM and cabinet (Eckelberry 1964, 102). As Leif Lewin observed, the Lindman Law “prevented the creation of a Parliament totally dominated by the Left. The age of upper-class rule was past. But the party that represented the upper class had been given a chance to survive” (1988, 82). One could add that the law not just increased Conservatives’ chances to survive but also provided them with time to copy the British Tories and build a national party organization that could compete effectively for votes (Carstairs 1980, 104; Eckelberry 1964, 59–61; Ziblatt 2017). The British example showed that not all minority protections had to involve constitutional, anti-majoritarian set asides.

Most Liberals and Socialists voted against the Lindman Law. They objected to some of the surviving minority protections (Carstairs 1980, 102) but also recognized that the expanded franchise increased their political leverage and would allow them to revisit the concession to which they acceded in 1907. Socialists and Liberals particularly realized that reforming the plutocratic upper house had to wait, given that its minority protection assuaged Conservatives’ concerns over widening the franchise (Eckelberry 1964, 88–94). By 1917, Liberals and Socialists won a significant parliamentary majority and set the stage for a second reform push.

Full parliamentarization was accomplished in 1917, when the leader of the Liberal Party, Nils Edén, obtained a promise from the Crown that it would no longer appoint prime ministers who did not have majorities in both chambers. The Crown also agreed to defer to parliament in appointing cabinet members or in calling early elections (Verney 1957, 115, 204). Establishing full parliamentary sovereignty thus was accomplished through changing constitutional norms, rather than through formal constitutional amendments. The king realized that he had to accede to full parliamentary sovereignty to preempt growing anti-monarchist sentiments.

Next, Edén, together with his socialist coalition partner, turned to the remaining franchise restrictions. In early 1918 his government tabled a bill to introduce female suffrage, remove remaining franchise restrictions for the lower house, and, most importantly, abolish the far more restrictive franchise requirements for the Council and City elections, whose members also served as the electors for the indirectly elected upper house. The upper house was unwilling to be so extensively democratized and rejected the bill (Sarlvik 2002, 241). However, by late 1918 Conservatives acquiesced to an almost identical bill that the government reintroduced as changed political circumstances weakened Conservatives’ opposition. Since the failed spring bill, the war had ended, the Socialists increasingly demanded a unicameral republic, and above all, the revolutionary threats in Germany, Russia, Finland, Hungary, and the Baltics dramatically increased the left threat. The bill passed this time, completing Sweden’s democratization. However, the left threat reduced Conservatives’ opposition to fuller democratization rather than, as it had in 1907, leading them to act preemptively to champion reforms to mitigate the left threat (Nielsson 2019, 140–42; Verney 1957, 202–5). Swedish Conservatives thus ended up in the same place as their European counterparts—full democratization with PR as the only minority protection—but it got there via a different path (Verney 1957, 204–8).

In sum, the staggered reforms reduced uncertainty by simplifying the choice set and buying Conservatives time to assess the left threat and the effectiveness of different minority protections. In turn, it gave Liberals and Socialists time to revisit and overturn the anti-democratic concessions they made in 1907 to pass the Lindman Law. In buying actors time, the staggered reforms made up for the time that they lost because Sweden’s late democratization compressed multiple constitutional reforms into a short, single, and uncertain historical moment. Those same reforms were settled more incrementally and over a longer period in other countries that democratized earlier.

Near Misses: Learning from Mistakes

So far, our account, that the staggered reforms reduced uncertainty, reads history backwards (Møller 2021). Actors in 1907 had no advance knowledge that reforms would be staggered as we do with the benefit of historical hindsight. We therefore cannot attribute all the uncertainty reductions to the staggered reforms. Reading history forward points to the four near misses between 1895 and 1906 as another factor for reducing uncertainty. These near misses reduced uncertainty because they involved deliberations, formation of coalitions, and ultimately failures that generated new information about the trade-offs of different constitutional compromises and their likelihood of obtaining parliamentary majorities. ⁵

Conservatives used the 1896 and 1902/03 bills as trial balloons to explore the possible trade-off between maximizing minority protection while minimizing democratic concessions. These early reform bills failed because they overshot the mark. The 1896 Annerstedt Bill promised to marginally lower the income barrier for the lower house franchise while retaining the existing property restrictions. It also proposed only a partial PR for the urban districts of the lower house. These districts operated under the super-majoritarian block vote system and thus exaggerated Liberals and Socialists electoral strength. The bill offered too little and was readily defeated.

The 1902/1903 Otter and Boström Bills were slightly less partisan versions of the Annerstedt Bill. They proposed PR for the entire lower house but wanted to curtail Liberal and Socialist strength in cities by creating hybrid urban/rural districts. Conservatives were willing to drop the income and property requirements for the lower house franchise but insisted that voters pay taxes, serve in the military, and that married and older voters receive extra votes (Eckelberry 1964, 49–50; Verney 1957, 138–42). Liberals and Socialists were lukewarm in their response to the Boström bill. Interestingly enough, they also backed off their initial support for PR, recognizing that they could use their opposition to extract more significant democratic reforms from Conservatives (Eckelberry 1964, 47–52; Lewin 1988, 185).

The 1905 election gave Liberals a legislative majority and led to the appointment of Karl Staaff as the first Liberal prime minister. His reform followed up on his party’s longstanding commitment to expand the male franchise for the lower house, while leaving the franchise for the upper house unchanged because Liberals realized that the upper house would never pass any reforms that would directly diminish its power. The proposed electoral system was designed to put Conservatives on the defensive. Staaff proposed PR for rural and urban councils that indirectly elected the upper house and a double ballot for the lower house. He believed that, as the centrist party, Liberals could draw disproportionate support in the run-off elections from both sides of the political spectrum (Eckelberry 1964, 78–80; Verney 1957, 145–48). The bill failed in both houses, but it served as a wake-up call for Conservatives to get their treasured lower house PR before Liberals and Socialist gave the double ballot system another try. With either camp having failed to pass their most preferred reforms, the 1907 Lindman Bill split enough of the differences to finally win a parliamentary majority.

The prior analysis demonstrated that the near misses were connected and that the failure in one produced new information, learning, and changes in subsequent reform bills. In doing so, they reduced uncertainty sufficiently for the 1907 Lindeman Bill to finally win the required bicameral majority to become law. Conservatives came to realize how many democratic concessions were necessary to get PR. Liberals and Socialists shifted from PR to double ballot to extract more democratic concessions from Conservatives. They also realized that the 1907 reforms gave them sufficient electoral growth opportunities that they could challenge the concessions that they had made to Conservatives in the near future (Lewin 1988, 84–85; Timelin 1928, 205–8).

Updating the Left-Threat Thesis

The four temporal mechanisms clearly played a central role in Sweden’s proportional representation adoption together with the factors identified by the original left-threat thesis. Social science methodology offers no clear guidelines for updating theories. Specifically, it offers no guidelines for when new inductive insights update a theory as opposed to become the basis for an altogether new theory. Scholars are often prone to reject a theory only to present a slightly modified version of the very same theory as a new one. The pressure for product differentiation thus contributes to unproductive theory swapping that was evident in the early research cycles on the origins of PR (Kreuzer 2023, 68–72). Such theory swapping is unfortunate because it ignores the importance of generating strong tests that juxtapose theories making as many unique predictions as possible (Beach and Pedersen 2019, 87, 129–30; Bennett and Checkel 2014, 16–18; Kreuzer 2023, 164–80). Hypothetically, a theory could be infinitely updated as it always is possible to extend the length of the causal pathway and identify ever more causal interaction effect. Such infinite updating, though, would mean that at some point a theory morphs into a causal narrative. One way to make theory updating more transparent is by using of causal graphs (on the use of causal graphs, see Kreuzer 2024; Pearl and Mackenzie 2018). Figure 4 visualizes an updated left-threat thesis that combines the static, institutional, and structural elements of the original theory with the newly identified temporal mechanisms. It demonstrates that the temporal mechanisms constitute second order causal factors that refine the causal effects of the original factors by turning them from static ones into dynamic ones. As such, the temporal mechanisms on their own have little explanatory power and would not warrant a new theory. Instead, their contribution is more modest by foregrounding the temporal dynamics of an existing, empirically solid theory. We presume that bringing a CHA perspective to other static theories could foreground similar temporal dynamics that currently remain behind ahistorical ceteris paribus assumptions. The theoretical potential of history remains greatly underutilized (Kreuzer 2024).

Figure 4 Left-threat thesis with temporal mechanism

Notes: Figure 4 incorporates in the original temporally heterogeneous left-threat thesis (dark blue box) with the temporal elements observed in the Swedish case: sequencing, timing, historical change, long outcome, and uncertainty.

Do Temporal Mechanisms Make Sweden Exceptional?

Figure 4 represents a modified the left-threat thesis to explain the Swedish case. To elevate these modifications to an update of the theory requires the probability that they are generalizable beyond the Swedish case. Nested analysis would suggest pivoting back to LNA and testing the updated theory across the original population of cases. However, the inclusion of the four temporal mechanism makes such a LNA strategy difficult to execute. Sequencing, timing, and historical change are difficult to measure and involve causal complexities that are not readily compatible with LNA (Abbott 1988; Hall 2003; Lieberson 1985; Page 2018). We therefore conclude our analysis with a more informal conjecture about the relevance of temporal mechanisms beyond the Swedish case. ⁶

Table 6 reports data on three of the four temporal mechanisms. First, we assess the outcome duration of fourteen cases by exploring whether they also included near misses, staggered reforms, and even reversals. Second, we evaluate sequencing by looking to the extent cases deviated from the linear and fixed order of parliamentarization, founding of socialist parties, franchise expansion, socialist electoral success, and PR adoption. Third, we look at the timing of PR adoption. It was impossible to evaluate historical change. Table 6 also lumps the countries into four groups based on their degree of temporal complexity. This taxonomy provides a rough estimate of the extent cases share Sweden’s temporal complexity or resemble the temporally less complex scenarios laid out by the original left-threat thesis.

Table 6 Temporal heterogeneity of cases

Notes: This sequencing and timing require comparing the dates of political identical events across multiple cases. Table 5 provides dates for the establishment of parliamentary sovereignty, three stages in franchise expansions, and PR near misses, actual reforms, further reforms, and reversals. Refer to online appendix 7 for the sources.

Sweden clearly is the temporally most heterogeneous of the fourteen cases without, however, being exceptional. France and Greece are almost as heterogenous as Sweden, but their temporal heterogeneity is linked to the long duration resulting from their multiple outcomes and reversals. These two countries defy claims that PR systems rarely ever are reversed because they facilitated multi-party systems whose parties are highly unlikely to revert to majoritarian systems threatening the survival of smaller parties (Colomer 2005). Both countries have histories that were long marked by political instability, frequent regime changes, and repeated partisan electoral engineering resulting in near misses and reversals. France and Greece also expanded the male franchise well before the first PR adoption, thus adding timing and duration as additional temporal factors.

Austria, Germany, and Italy established PR after their monarchies collapsed which required drafting entirely new constitutions, like in Sweden, thus entangling PR with other constitutional reforms. In Germany, the franchise preceded parliamentarization and occurred almost half a century before PR adoption. Ireland and Finland also established PR in the context of obtaining independence, having to establish an entirely new political order. These five cases had short outcomes, but the sequencing of their causal factors was reversed just as in Sweden. This leaves Belgium, Denmark, the Netherlands, and the UK as the only homogeneous cases. They followed the linear causal sequence laid out by the left-threat thesis: parliamentarization, founding of socialist parties, male franchise expansion, and emerging electoral left threat. Their outcomes also involved short, discrete events. But we also can identify minor temporal complexities. Belgium, Switzerland, and the UK had failed PR reforms. Switzerland experienced a significant time lag between the franchise expansion and PR adoption and its PR reforms were introduced through popular initiatives rather than the legislative route.

Overall, these comparisons underscore that Sweden’s PR adoption was far less peculiar when compared to other cases. This finding underscores the key premise of nested analysis that the confirmation of a theory is not just a question of empirical fit but of updating of theories used to analyze the cases. Updating the left-threat thesis turns Sweden from being an outlier to a more typical case that provides useful new theoretical insights.